path: root/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c

/**************************************************************************
 * 
 * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
 * All Rights Reserved.
 * 
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, sub license, and/or sell copies of the Software, and to
 * permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 * 
 * The above copyright notice and this permission notice (including the
 * next paragraph) shall be included in all copies or substantial portions
 * of the Software.
 * 
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 * 
 **************************************************************************/

 /*
  * Authors:
  *   Keith Whitwell <keith@tungstengraphics.com>
  */

#include "util/u_memory.h"
#include "draw/draw_context.h"
#include "draw/draw_private.h"
#include "draw/draw_vbuf.h"
#include "draw/draw_vertex.h"
#include "draw/draw_pt.h"
#include "translate/translate.h"
#include "translate/translate_cache.h"

/* The simplest 'middle end' in the new vertex code.  
 * 
 * The responsibilities of a middle end are to:
 *  - perform vertex fetch using
 *       - draw vertex element/buffer state
 *       - a list of fetch indices we received as an input
 *  - run the vertex shader
 *  - cliptest, 
 *  - clip coord calculation 
 *  - viewport transformation
 *  - if necessary, run the primitive pipeline, passing it:
 *       - a linear array of vertex_header vertices constructed here
 *       - a set of draw indices we received as an input
 *  - otherwise, drive the hw backend,
 *       - allocate space for hardware format vertices
 *       - translate the vertex-shader output vertices to hw format
 *       - calling the backend draw functions.
 *
 * For convenience, we provide a helper function to drive the hardware
 * backend given similar inputs to those required to run the pipeline.
 *
 * In the case of passthrough mode, many of these actions are disabled
 * or noops, so we end up doing:
 *
 *  - perform vertex fetch
 *  - drive the hw backend
 *
 * IE, basically just vertex fetch to post-vs-format vertices,
 * followed by a call to the backend helper function.
 */


struct fetch_emit_middle_end {
   struct draw_pt_middle_end base;
   struct draw_context *draw;
   
   struct translate *translate;
   const struct vertex_info *vinfo;

   /* Cache point size somewhere it's address won't change:
    */
   float point_size;

   struct translate_cache *cache;
};




static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
                                unsigned prim,
				unsigned opt,
                                unsigned *max_vertices )
{
   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
   struct draw_context *draw = feme->draw;
   const struct vertex_info *vinfo;
   unsigned i, dst_offset;
   boolean ok;
   struct translate_key key;


   ok = draw->render->set_primitive( draw->render, 
                                     prim );
   if (!ok) {
      assert(0);
      return;
   }
   
   /* Must do this after set_primitive() above:
    */
   vinfo = feme->vinfo = draw->render->get_vertex_info(draw->render);
   
   

   /* Transform from API vertices to HW vertices, skipping the
    * pipeline_vertex intermediate step.
    */
   dst_offset = 0;
   memset(&key, 0, sizeof(key));

   for (i = 0; i < vinfo->num_attribs; i++) {
      const struct pipe_vertex_element *src = &draw->pt.vertex_element[vinfo->src_index[i]];

      unsigned emit_sz = 0;
      unsigned input_format = src->src_format;
      unsigned input_buffer = src->vertex_buffer_index;
      unsigned input_offset = src->src_offset;
      unsigned output_format;

      switch (vinfo->emit[i]) {
      case EMIT_4F:
	 output_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
	 emit_sz = 4 * sizeof(float);
         break;
      case EMIT_3F:
	 output_format = PIPE_FORMAT_R32G32B32_FLOAT;
	 emit_sz = 3 * sizeof(float);
         break;
      case EMIT_2F:
	 output_format = PIPE_FORMAT_R32G32_FLOAT;
	 emit_sz = 2 * sizeof(float);
         break;
      case EMIT_1F:
	 output_format = PIPE_FORMAT_R32_FLOAT;
	 emit_sz = 1 * sizeof(float);
         break;
      case EMIT_1F_PSIZE:
	 input_format = PIPE_FORMAT_R32_FLOAT;
	 input_buffer = draw->pt.nr_vertex_buffers;
	 input_offset = 0;
	 output_format = PIPE_FORMAT_R32_FLOAT;
	 emit_sz = 1 * sizeof(float);
         break;
      default:
         assert(0);
	 output_format = PIPE_FORMAT_NONE;
	 emit_sz = 0;
	 continue;
      }

      key.element[i].input_format = input_format;
      key.element[i].input_buffer = input_buffer;
      key.element[i].input_offset = input_offset;
      key.element[i].output_format = output_format;
      key.element[i].output_offset = dst_offset;
      
      dst_offset += emit_sz;
   }

   key.nr_elements = vinfo->num_attribs;
   key.output_stride = vinfo->size * 4;

   /* Don't bother with caching at this stage:
    */
   if (!feme->translate ||
       translate_key_compare(&feme->translate->key, &key) != 0) 
   {
      translate_key_sanitize(&key);
      feme->translate = translate_cache_find(feme->cache,
                                             &key);


      feme->translate->set_buffer(feme->translate, 
				  draw->pt.nr_vertex_buffers, 
				  &feme->point_size,
				  0);
   }
   
   feme->point_size = draw->rasterizer->point_size;

   for (i = 0; i < draw->pt.nr_vertex_buffers; i++) {
      feme->translate->set_buffer(feme->translate, 
                                  i, 
                                  ((char *)draw->pt.user.vbuffer[i] + 
                                   draw->pt.vertex_buffer[i].buffer_offset),
                                  draw->pt.vertex_buffer[i].pitch );
   }

   *max_vertices = (draw->render->max_vertex_buffer_bytes / 
                    (vinfo->size * 4));

   /* Return an even number of verts.
    * This prevents "parity" errors when splitting long triangle strips which
    * can lead to front/back culling mix-ups.
    * Every other triangle in a strip has an alternate front/back orientation
    * so splitting at an odd position can cause the orientation of subsequent
    * triangles to get reversed.
    */
   *max_vertices = *max_vertices & ~1;
}





static void fetch_emit_run( struct draw_pt_middle_end *middle,
                            const unsigned *fetch_elts,
                            unsigned fetch_count,
                            const ushort *draw_elts,
                            unsigned draw_count )
{
   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
   struct draw_context *draw = feme->draw;
   void *hw_verts;
   
   /* XXX: need to flush to get prim_vbuf.c to release its allocation?? 
    */
   draw_do_flush( draw, DRAW_FLUSH_BACKEND );

   hw_verts = draw->render->allocate_vertices( draw->render,
                                               (ushort)feme->translate->key.output_stride,
                                               (ushort)fetch_count );
   if (!hw_verts) {
      assert(0);
      return;
   }
         
					
   /* Single routine to fetch vertices and emit HW verts.
    */
   feme->translate->run_elts( feme->translate, 
			      fetch_elts,
			      fetch_count,
			      hw_verts );

   if (0) {
      unsigned i;
      for (i = 0; i < fetch_count; i++) {
         debug_printf("\n\nvertex %d:\n", i);
         draw_dump_emitted_vertex( feme->vinfo, 
                                   (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
      }
   }

   /* XXX: Draw arrays path to avoid re-emitting index list again and
    * again.
    */
   draw->render->draw( draw->render, 
                       draw_elts, 
                       draw_count );

   /* Done -- that was easy, wasn't it: 
    */
   draw->render->release_vertices( draw->render, 
                                   hw_verts, 
                                   feme->translate->key.output_stride, 
                                   fetch_count );

}


static void fetch_emit_run_linear( struct draw_pt_middle_end *middle,
                                   unsigned start,
                                   unsigned count )
{
   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
   struct draw_context *draw = feme->draw;
   void *hw_verts;

   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
    */
   draw_do_flush( draw, DRAW_FLUSH_BACKEND );

   hw_verts = draw->render->allocate_vertices( draw->render,
                                               (ushort)feme->translate->key.output_stride,
                                               (ushort)count );
   if (!hw_verts) {
      assert(0);
      return;
   }

   /* Single routine to fetch vertices and emit HW verts.
    */
   feme->translate->run( feme->translate,
                         start,
                         count,
                         hw_verts );

   if (0) {
      unsigned i;
      for (i = 0; i < count; i++) {
         debug_printf("\n\nvertex %d:\n", i);
         draw_dump_emitted_vertex( feme->vinfo,
                                   (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i );
      }
   }

   /* XXX: Draw arrays path to avoid re-emitting index list again and
    * again.
    */
   draw->render->draw_arrays( draw->render,
                              0, /*start*/
                              count );

   /* Done -- that was easy, wasn't it:
    */
   draw->render->release_vertices( draw->render,
                                   hw_verts,
                                   feme->translate->key.output_stride,
                                   count );

}


static void fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle,
                                        unsigned start,
                                        unsigned count,
                                        const ushort *draw_elts,
                                        unsigned draw_count )
{
   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;
   struct draw_context *draw = feme->draw;
   void *hw_verts;

   /* XXX: need to flush to get prim_vbuf.c to release its allocation??
    */
   draw_do_flush( draw, DRAW_FLUSH_BACKEND );

   hw_verts = draw->render->allocate_vertices( draw->render,
                                               (ushort)feme->translate->key.output_stride,
                                               (ushort)count );
   if (!hw_verts) {
      assert(0);
      return;
   }

   /* Single routine to fetch vertices and emit HW verts.
    */
   feme->translate->run( feme->translate,
                         start,
                         count,
                         hw_verts );

   /* XXX: Draw arrays path to avoid re-emitting index list again and
    * again.
    */
   draw->render->draw( draw->render, 
                       draw_elts, 
                       draw_count );

   /* Done -- that was easy, wasn't it:
    */
   draw->render->release_vertices( draw->render,
                                   hw_verts,
                                   feme->translate->key.output_stride,
                                   count );

}




static void fetch_emit_finish( struct draw_pt_middle_end *middle )
{
   /* nothing to do */
}

static void fetch_emit_destroy( struct draw_pt_middle_end *middle )
{
   struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle;

   if (feme->cache)
      translate_cache_destroy(feme->cache);

   FREE(middle);
}


struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw )
{
   struct fetch_emit_middle_end *fetch_emit = CALLOC_STRUCT( fetch_emit_middle_end );
   if (fetch_emit == NULL)
      return NULL;

   fetch_emit->cache = translate_cache_create();
   if (!fetch_emit->cache) {
      FREE(fetch_emit);
      return NULL;
   }

   fetch_emit->base.prepare    = fetch_emit_prepare;
   fetch_emit->base.run        = fetch_emit_run;
   fetch_emit->base.run_linear = fetch_emit_run_linear;
   fetch_emit->base.run_linear_elts = fetch_emit_run_linear_elts;
   fetch_emit->base.finish     = fetch_emit_finish;
   fetch_emit->base.destroy    = fetch_emit_destroy;

   fetch_emit->draw = draw;
     
   return &fetch_emit->base;
}