diff options
Diffstat (limited to 'src/mesa/drivers/dri/r300/r300_state.c')
-rw-r--r-- | src/mesa/drivers/dri/r300/r300_state.c | 554 |
1 files changed, 193 insertions, 361 deletions
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c index 07299f3b36..217a16818e 100644 --- a/src/mesa/drivers/dri/r300/r300_state.c +++ b/src/mesa/drivers/dri/r300/r300_state.c @@ -42,6 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/macros.h" #include "main/context.h" #include "main/dd.h" +#include "main/framebuffer.h" #include "main/simple_list.h" #include "main/api_arrayelt.h" #include "main/texformat.h" @@ -52,21 +53,21 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #include "shader/prog_statevars.h" #include "vbo/vbo.h" #include "tnl/tnl.h" +#include "tnl/t_vp_build.h" -#include "radeon_ioctl.h" -#include "radeon_state.h" #include "r300_context.h" #include "r300_ioctl.h" #include "r300_state.h" #include "r300_reg.h" #include "r300_emit.h" -#include "r300_fragprog.h" #include "r300_tex.h" +#include "r300_fragprog_common.h" +#include "r300_fragprog.h" +#include "r500_fragprog.h" #include "drirenderbuffer.h" extern int future_hw_tcl_on; -extern void _tnl_UpdateFixedFunctionProgram(GLcontext * ctx); static void r300BlendColor(GLcontext * ctx, const GLfloat cf[4]) { @@ -451,18 +452,9 @@ static void r300SetPolygonOffsetState(GLcontext * ctx, GLboolean state) static GLboolean current_fragment_program_writes_depth(GLcontext* ctx) { - r300ContextPtr r300 = R300_CONTEXT(ctx); + struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current; - if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) { - struct r300_fragment_program *fp = (struct r300_fragment_program *) - (char *)ctx->FragmentProgram._Current; - return (fp && fp->WritesDepth); - } else { - struct r500_fragment_program* fp = - (struct r500_fragment_program*)(char*) - ctx->FragmentProgram._Current; - return (fp && fp->writes_depth); - } + return (fp && fp->writes_depth); } static void r300SetEarlyZState(GLcontext * ctx) @@ -535,8 +527,6 @@ static void r300SetAlphaState(GLcontext * ctx) R300_STATECHANGE(r300, at); r300->hw.at.cmd[R300_AT_ALPHA_TEST] = pp_misc; r300->hw.at.cmd[R300_AT_UNKNOWN] = 0; - - r300SetEarlyZState(ctx); } static void r300AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) @@ -584,15 +574,19 @@ static void r300SetDepthState(GLcontext * ctx) r300->hw.zs.cmd[R300_ZS_CNTL_1] |= translate_func(ctx->Depth.Func) << R300_Z_FUNC_SHIFT; } - - r300SetEarlyZState(ctx); } static void r300SetStencilState(GLcontext * ctx, GLboolean state) { r300ContextPtr r300 = R300_CONTEXT(ctx); + GLboolean hw_stencil = GL_FALSE; + if (ctx->DrawBuffer) { + struct radeon_renderbuffer *rrbStencil + = radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL); + hw_stencil = (rrbStencil && rrbStencil->bo); + } - if (r300->state.stencil.hw_stencil) { + if (hw_stencil) { R300_STATECHANGE(r300, zs); if (state) { r300->hw.zs.cmd[R300_ZS_CNTL_0] |= @@ -737,7 +731,12 @@ static void r300ColorMask(GLcontext * ctx, static void r300PointSize(GLcontext * ctx, GLfloat size) { r300ContextPtr r300 = R300_CONTEXT(ctx); - /* same size limits for AA, non-AA points */ + + /* We need to clamp to user defined range here, because + * the HW clamping happens only for per vertex point size. */ + size = CLAMP(size, ctx->Point.MinSize, ctx->Point.MaxSize); + + /* same size limits for AA, non-AA points */ size = CLAMP(size, ctx->Const.MinPointSize, ctx->Const.MaxPointSize); R300_STATECHANGE(r300, ps); @@ -937,15 +936,25 @@ static void r300UpdateWindow(GLcontext * ctx) GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0; GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0; const GLfloat *v = ctx->Viewport._WindowMap.m; + const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF; + const GLboolean render_to_fbo = (ctx->DrawBuffer->Name != 0); + GLfloat y_scale, y_bias; + + if (render_to_fbo) { + y_scale = 1.0; + y_bias = 0; + } else { + y_scale = -1.0; + y_bias = yoffset; + } GLfloat sx = v[MAT_SX]; GLfloat tx = v[MAT_TX] + xoffset + SUBPIXEL_X; - GLfloat sy = -v[MAT_SY]; - GLfloat ty = (-v[MAT_TY]) + yoffset + SUBPIXEL_Y; - GLfloat sz = v[MAT_SZ] * rmesa->state.depth.scale; - GLfloat tz = v[MAT_TZ] * rmesa->state.depth.scale; + GLfloat sy = v[MAT_SY] * y_scale; + GLfloat ty = (v[MAT_TY] * y_scale) + y_bias + SUBPIXEL_Y; + GLfloat sz = v[MAT_SZ] * depthScale; + GLfloat tz = v[MAT_TZ] * depthScale; - R300_FIREVERTICES(rmesa); R300_STATECHANGE(rmesa, vpt); rmesa->hw.vpt.cmd[R300_VPT_XSCALE] = r300PackFloat32(sx); @@ -964,6 +973,8 @@ static void r300Viewport(GLcontext * ctx, GLint x, GLint y, * values, or keep the originals hanging around. */ r300UpdateWindow(ctx); + + radeon_viewport(ctx, x, y, width, height); } static void r300DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval) @@ -996,64 +1007,6 @@ void r300UpdateViewportOffset(GLcontext * ctx) radeonUpdateScissor(ctx); } -/** - * Tell the card where to render (offset, pitch). - * Effected by glDrawBuffer, etc - */ -void r300UpdateDrawBuffer(GLcontext * ctx) -{ - r300ContextPtr rmesa = R300_CONTEXT(ctx); - r300ContextPtr r300 = rmesa; - struct gl_framebuffer *fb = ctx->DrawBuffer; - driRenderbuffer *drb; - - if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) { - /* draw to front */ - drb = - (driRenderbuffer *) fb->Attachment[BUFFER_FRONT_LEFT]. - Renderbuffer; - } else if (fb->_ColorDrawBufferIndexes[0] == BUFFER_BACK_LEFT) { - /* draw to back */ - drb = - (driRenderbuffer *) fb->Attachment[BUFFER_BACK_LEFT]. - Renderbuffer; - } else { - /* drawing to multiple buffers, or none */ - return; - } - - assert(drb); - assert(drb->flippedPitch); - - R300_STATECHANGE(rmesa, cb); - - r300->hw.cb.cmd[R300_CB_OFFSET] = drb->flippedOffset + //r300->radeon.state.color.drawOffset + - r300->radeon.radeonScreen->fbLocation; - r300->hw.cb.cmd[R300_CB_PITCH] = drb->flippedPitch; //r300->radeon.state.color.drawPitch; - - if (r300->radeon.radeonScreen->cpp == 4) - r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_ARGB8888; - else - r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_RGB565; - - if (r300->radeon.sarea->tiling_enabled) - r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_TILE_ENABLE; -#if 0 - R200_STATECHANGE(rmesa, ctx); - - /* Note: we used the (possibly) page-flipped values */ - rmesa->hw.ctx.cmd[CTX_RB3D_COLOROFFSET] - = ((drb->flippedOffset + rmesa->r200Screen->fbLocation) - & R200_COLOROFFSET_MASK); - rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] = drb->flippedPitch; - - if (rmesa->sarea->tiling_enabled) { - rmesa->hw.ctx.cmd[CTX_RB3D_COLORPITCH] |= - R200_COLOR_TILE_ENABLE; - } -#endif -} - static void r300FetchStateParameter(GLcontext * ctx, const gl_state_index state[STATE_LENGTH], @@ -1116,7 +1069,7 @@ void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state) if (!fp) return; - paramList = fp->mesa_program.Base.Parameters; + paramList = fp->Base.Base.Parameters; if (!paramList) return; @@ -1235,9 +1188,8 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings) { r300ContextPtr r300 = R300_CONTEXT(ctx); int i; - struct r300_fragment_program *fp = (struct r300_fragment_program *) - (char *)ctx->FragmentProgram._Current; - struct r300_fragment_program_code *code = &fp->code; + struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current; + struct r300_fragment_program_code *code = &fp->code.r300; R300_STATECHANGE(r300, fpt); @@ -1271,15 +1223,15 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings) } r300->hw.fpt.cmd[R300_FPT_CMD_0] = - cmdpacket0(R300_US_TEX_INST_0, code->tex.length); + cmdpacket0(r300->radeon.radeonScreen, + R300_US_TEX_INST_0, code->tex.length); } static void r500SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings) { int i; - struct r500_fragment_program *fp = (struct r500_fragment_program *) - (char *)ctx->FragmentProgram._Current; - struct r500_fragment_program_code *code = &fp->code; + struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current; + struct r500_fragment_program_code *code = &fp->code.r500; /* find all the texture instructions and relocate the texture units */ for (i = 0; i < code->inst_end + 1; i++) { @@ -1322,7 +1274,7 @@ static GLuint translate_lod_bias(GLfloat bias) static void r300SetupTextures(GLcontext * ctx) { int i, mtu; - struct r300_tex_obj *t; + struct radeon_tex_obj *t; r300ContextPtr r300 = R300_CONTEXT(ctx); int hw_tmu = 0; int last_hw_tmu = -1; /* -1 translates into no setup costs for fields */ @@ -1356,21 +1308,16 @@ static void r300SetupTextures(GLcontext * ctx) /* We cannot let disabled tmu offsets pass DRM */ for (i = 0; i < mtu; i++) { if (ctx->Texture.Unit[i]._ReallyEnabled) { - -#if 0 /* Enables old behaviour */ - hw_tmu = i; -#endif tmu_mappings[i] = hw_tmu; - t = (r300TexObjPtr) r300->state.texture.unit[i].texobj->DriverData; - /* XXX questionable fix for bug 9170: */ + t = radeon_tex_obj(ctx->Texture.Unit[i]._Current); if (!t) continue; - if ((t->format & 0xffffff00) == 0xffffff00) { + if ((t->pp_txformat & 0xffffff00) == 0xffffff00) { WARN_ONCE ("unknown texture format (entry %x) encountered. Help me !\n", - t->format & 0xff); + t->pp_txformat & 0xff); } if (RADEON_DEBUG & DEBUG_STATE) @@ -1381,29 +1328,28 @@ static void r300SetupTextures(GLcontext * ctx) r300->hw.tex.filter.cmd[R300_TEX_VALUE_0 + hw_tmu] = - gen_fixed_filter(t->filter) | (hw_tmu << 28); + gen_fixed_filter(t->pp_txfilter) | (hw_tmu << 28); /* Note: There is a LOD bias per texture unit and a LOD bias * per texture object. We add them here to get the correct behaviour. * (The per-texture object LOD bias was introduced in OpenGL 1.4 * and is not present in the EXT_texture_object extension). */ r300->hw.tex.filter_1.cmd[R300_TEX_VALUE_0 + hw_tmu] = - t->filter_1 | - translate_lod_bias(ctx->Texture.Unit[i].LodBias + t->base.tObj->LodBias); + t->pp_txfilter_1 | + translate_lod_bias(ctx->Texture.Unit[i].LodBias + t->base.LodBias); r300->hw.tex.size.cmd[R300_TEX_VALUE_0 + hw_tmu] = - t->size; + t->pp_txsize; r300->hw.tex.format.cmd[R300_TEX_VALUE_0 + - hw_tmu] = t->format; + hw_tmu] = t->pp_txformat; r300->hw.tex.pitch.cmd[R300_TEX_VALUE_0 + hw_tmu] = - t->pitch_reg; - r300->hw.tex.offset.cmd[R300_TEX_VALUE_0 + - hw_tmu] = t->offset; + t->pp_txpitch; + r300->hw.textures[hw_tmu] = t; - if (t->offset & R300_TXO_MACRO_TILE) { + if (t->tile_bits & R300_TXO_MACRO_TILE) { WARN_ONCE("macro tiling enabled!\n"); } - if (t->offset & R300_TXO_MICRO_TILE) { + if (t->tile_bits & R300_TXO_MICRO_TILE) { WARN_ONCE("micro tiling enabled!\n"); } @@ -1420,37 +1366,36 @@ static void r300SetupTextures(GLcontext * ctx) } r300->hw.tex.filter.cmd[R300_TEX_CMD_0] = - cmdpacket0(R300_TX_FILTER0_0, last_hw_tmu + 1); + cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER0_0, last_hw_tmu + 1); r300->hw.tex.filter_1.cmd[R300_TEX_CMD_0] = - cmdpacket0(R300_TX_FILTER1_0, last_hw_tmu + 1); + cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER1_0, last_hw_tmu + 1); r300->hw.tex.size.cmd[R300_TEX_CMD_0] = - cmdpacket0(R300_TX_SIZE_0, last_hw_tmu + 1); + cmdpacket0(r300->radeon.radeonScreen, R300_TX_SIZE_0, last_hw_tmu + 1); r300->hw.tex.format.cmd[R300_TEX_CMD_0] = - cmdpacket0(R300_TX_FORMAT_0, last_hw_tmu + 1); + cmdpacket0(r300->radeon.radeonScreen, R300_TX_FORMAT_0, last_hw_tmu + 1); r300->hw.tex.pitch.cmd[R300_TEX_CMD_0] = - cmdpacket0(R300_TX_FORMAT2_0, last_hw_tmu + 1); + cmdpacket0(r300->radeon.radeonScreen, R300_TX_FORMAT2_0, last_hw_tmu + 1); r300->hw.tex.offset.cmd[R300_TEX_CMD_0] = - cmdpacket0(R300_TX_OFFSET_0, last_hw_tmu + 1); + cmdpacket0(r300->radeon.radeonScreen, R300_TX_OFFSET_0, last_hw_tmu + 1); r300->hw.tex.chroma_key.cmd[R300_TEX_CMD_0] = - cmdpacket0(R300_TX_CHROMA_KEY_0, last_hw_tmu + 1); + cmdpacket0(r300->radeon.radeonScreen, R300_TX_CHROMA_KEY_0, last_hw_tmu + 1); r300->hw.tex.border_color.cmd[R300_TEX_CMD_0] = - cmdpacket0(R300_TX_BORDER_COLOR_0, last_hw_tmu + 1); + cmdpacket0(r300->radeon.radeonScreen, R300_TX_BORDER_COLOR_0, last_hw_tmu + 1); if (!fp) /* should only happenen once, just after context is created */ return; if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) { - if (fp->mesa_program.UsesKill && last_hw_tmu < 0) { + if (fp->Base.UsesKill && last_hw_tmu < 0) { // The KILL operation requires the first texture unit // to be enabled. r300->hw.txe.cmd[R300_TXE_ENABLE] |= 1; r300->hw.tex.filter.cmd[R300_TEX_VALUE_0] = 0; r300->hw.tex.filter.cmd[R300_TEX_CMD_0] = - cmdpacket0(R300_TX_FILTER0_0, 1); + cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER0_0, 1); } - r300SetupFragmentShaderTextures(ctx, tmu_mappings); - } else - r500SetupFragmentShaderTextures(ctx, tmu_mappings); + } + r300->vtbl.SetupFragmentShaderTextures(ctx, tmu_mappings); if (RADEON_DEBUG & DEBUG_STATE) fprintf(stderr, "TX_ENABLE: %08x last_hw_tmu=%d\n", @@ -1481,7 +1426,7 @@ static void r300SetupRSUnit(GLcontext * ctx) if (hw_tcl_on) OutputsWritten.vp_outputs = CURRENT_VERTEX_SHADER(ctx)->key.OutputsWritten; else - RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->state.render_inputs_bitset); + RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset); if (ctx->FragmentProgram._Current) InputsRead = ctx->FragmentProgram._Current->Base.InputsRead; @@ -1545,6 +1490,7 @@ static void r300SetupRSUnit(GLcontext * ctx) } } + /* We always route 4 texcoord components */ for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (! ( InputsRead & FRAG_BIT_TEX(i) ) ) continue; @@ -1554,36 +1500,30 @@ static void r300SetupRSUnit(GLcontext * ctx) continue; } - int swiz; - - /* with TCL we always seem to route 4 components */ - if (hw_tcl_on) - count = 4; - else - count = VB->AttribPtr[_TNL_ATTRIB_TEX(i)]->size; - - switch(count) { - case 4: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3); break; - case 3: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(R300_RS_SEL_K1); break; - default: - case 1: - case 2: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(R300_RS_SEL_K0) | R300_RS_SEL_Q(R300_RS_SEL_K1); break; - }; - - r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= swiz | R300_RS_TEX_PTR(rs_tex_count); + r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) | R300_RS_TEX_PTR(rs_tex_count); r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg); InputsRead &= ~(FRAG_BIT_TEX0 << i); - rs_tex_count += count; + rs_tex_count += 4; + ++tex_ip; + ++fp_reg; + } + + if (InputsRead & FRAG_BIT_WPOS) { + r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) | R300_RS_TEX_PTR(rs_tex_count); + r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg); + InputsRead &= ~FRAG_BIT_WPOS; + rs_tex_count += 4; ++tex_ip; ++fp_reg; } if (InputsRead & FRAG_BIT_FOGC) { if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_FOGC, _TNL_ATTRIB_FOG)) { - r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) | R300_RS_TEX_PTR(rs_tex_count); + r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_S(0) | R300_RS_SEL_T(R300_RS_SEL_K0) | R300_RS_SEL_R(R300_RS_SEL_K0); + r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_Q(R300_RS_SEL_K1) | R300_RS_TEX_PTR(rs_tex_count); r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg); InputsRead &= ~FRAG_BIT_FOGC; - rs_tex_count += 4; + rs_tex_count += 1; ++tex_ip; ++fp_reg; } else { @@ -1591,16 +1531,6 @@ static void r300SetupRSUnit(GLcontext * ctx) } } - if (InputsRead & FRAG_BIT_WPOS) { - r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) | R300_RS_TEX_PTR(rs_tex_count); - r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg); - InputsRead &= ~FRAG_BIT_WPOS; - rs_tex_count += 4; - ++tex_ip; - ++fp_reg; - } - InputsRead &= ~FRAG_BIT_WPOS; - /* Setup default color if no color or tex was set */ if (rs_tex_count == 0 && col_ip == 0) { r300->hw.rr.cmd[R300_RR_INST_0] = R300_RS_INST_COL_ID(0) | R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(0) | R300_RS_COL_FMT(R300_RS_COL_FMT_0001); @@ -1608,10 +1538,10 @@ static void r300SetupRSUnit(GLcontext * ctx) } high_rr = (col_ip > tex_ip) ? col_ip : tex_ip; - r300->hw.rc.cmd[1] |= (rs_tex_count << R300_IT_COUNT_SHIFT) | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN; + r300->hw.rc.cmd[1] |= (rs_tex_count << R300_IT_COUNT_SHIFT) | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN; r300->hw.rc.cmd[2] |= high_rr - 1; - r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(R300_RS_INST_0, high_rr); + r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_INST_0, high_rr); if (InputsRead) WARN_ONCE("Don't know how to satisfy InputsRead=0x%08x\n", InputsRead); @@ -1632,7 +1562,7 @@ static void r500SetupRSUnit(GLcontext * ctx) if (hw_tcl_on) OutputsWritten.vp_outputs = CURRENT_VERTEX_SHADER(ctx)->key.OutputsWritten; else - RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->state.render_inputs_bitset); + RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset); if (ctx->FragmentProgram._Current) InputsRead = ctx->FragmentProgram._Current->Base.InputsRead; @@ -1696,7 +1626,7 @@ static void r500SetupRSUnit(GLcontext * ctx) } } - + /* We always route 4 texcoord components */ for (i = 0; i < ctx->Const.MaxTextureUnits; i++) { if (! ( InputsRead & FRAG_BIT_TEX(i) ) ) continue; @@ -1706,59 +1636,41 @@ static void r500SetupRSUnit(GLcontext * ctx) continue; } - int swiz = 0; - - /* with TCL we always seem to route 4 components */ - if (hw_tcl_on) - count = 4; - else - count = VB->AttribPtr[_TNL_ATTRIB_TEX(i)]->size; - - if (count == 4) { - swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT; - swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT; - swiz |= (rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT; - swiz |= (rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT; - } else if (count == 3) { - swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT; - swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT; - swiz |= (rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT; - swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT; - } else if (count == 2) { - swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT; - swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT; - swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT; - swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT; - } else if (count == 1) { - swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT; - swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT; - swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT; - swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT; - } else { - swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT; - swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT; - swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT; - swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT; - } + r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) | + ((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) | + ((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) | + ((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT); - r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= swiz; r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg); InputsRead &= ~(FRAG_BIT_TEX0 << i); - rs_tex_count += count; + rs_tex_count += 4; + ++tex_ip; + ++fp_reg; + } + + if (InputsRead & FRAG_BIT_WPOS) { + r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) | + ((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) | + ((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) | + ((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT); + + r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg); + InputsRead &= ~FRAG_BIT_WPOS; + rs_tex_count += 4; ++tex_ip; ++fp_reg; } if (InputsRead & FRAG_BIT_FOGC) { if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_FOGC, _TNL_ATTRIB_FOG)) { - r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) | - ((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) | - ((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) | - ((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT); + r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= (rs_tex_count << R500_RS_IP_TEX_PTR_S_SHIFT) | + (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) | + (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) | + (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT); r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg); InputsRead &= ~FRAG_BIT_FOGC; - rs_tex_count += 4; + rs_tex_count += 1; ++tex_ip; ++fp_reg; } else { @@ -1766,19 +1678,6 @@ static void r500SetupRSUnit(GLcontext * ctx) } } - if (InputsRead & FRAG_BIT_WPOS) { - r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) | - ((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) | - ((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) | - ((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT); - - r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg); - InputsRead &= ~FRAG_BIT_WPOS; - rs_tex_count += 4; - ++tex_ip; - ++fp_reg; - } - /* Setup default color if no color or tex was set */ if (rs_tex_count == 0 && col_ip == 0) { r300->hw.rr.cmd[R300_RR_INST_0] |= R500_RS_INST_COL_ID(0) | R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(0) | R500_RS_COL_FMT(R300_RS_COL_FMT_0001); @@ -1789,7 +1688,7 @@ static void r500SetupRSUnit(GLcontext * ctx) r300->hw.rc.cmd[1] |= (rs_tex_count << R300_IT_COUNT_SHIFT) | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN; r300->hw.rc.cmd[2] |= 0xC0 | (high_rr - 1); - r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(R500_RS_INST_0, high_rr); + r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_INST_0, high_rr); if (InputsRead) WARN_ONCE("Don't know how to satisfy InputsRead=0x%08x\n", InputsRead); @@ -1902,7 +1801,7 @@ static void r300VapCntl(r300ContextPtr rmesa, GLuint input_count, static void r300SetupDefaultVertexProgram(r300ContextPtr rmesa) { - struct r300_vertex_shader_state *prog = &(rmesa->state.vertex_shader); + struct r300_vertex_shader_state *prog = &(rmesa->vertex_shader); GLuint o_reg = 0; GLuint i_reg = 0; int i; @@ -1911,11 +1810,11 @@ static void r300SetupDefaultVertexProgram(r300ContextPtr rmesa) int program_end = 0; for (i = VERT_ATTRIB_POS; i < VERT_ATTRIB_MAX; i++) { - if (rmesa->state.sw_tcl_inputs[i] != -1) { + if (rmesa->swtcl.sw_tcl_inputs[i] != -1) { prog->program.body.i[program_end + 0] = PVS_OP_DST_OPERAND(VE_MULTIPLY, GL_FALSE, GL_FALSE, o_reg++, VSF_FLAG_ALL, PVS_DST_REG_OUT); - prog->program.body.i[program_end + 1] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, VSF_FLAG_NONE); - prog->program.body.i[program_end + 2] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_REG_INPUT, VSF_FLAG_NONE); - prog->program.body.i[program_end + 3] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_REG_INPUT, VSF_FLAG_NONE); + prog->program.body.i[program_end + 1] = PVS_SRC_OPERAND(rmesa->swtcl.sw_tcl_inputs[i], PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, VSF_FLAG_NONE); + prog->program.body.i[program_end + 2] = PVS_SRC_OPERAND(rmesa->swtcl.sw_tcl_inputs[i], PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_REG_INPUT, VSF_FLAG_NONE); + prog->program.body.i[program_end + 3] = PVS_SRC_OPERAND(rmesa->swtcl.sw_tcl_inputs[i], PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_REG_INPUT, VSF_FLAG_NONE); program_end += 4; i_reg++; } @@ -1986,6 +1885,7 @@ static void r300SetupRealVertexProgram(r300ContextPtr rmesa) (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT); } + static void r300SetupVertexProgram(r300ContextPtr rmesa) { GLcontext *ctx = rmesa->radeon.glCtx; @@ -2015,6 +1915,7 @@ static void r300SetupVertexProgram(r300ContextPtr rmesa) */ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state) { + r300ContextPtr rmesa = R300_CONTEXT(ctx); if (RADEON_DEBUG & DEBUG_STATE) fprintf(stderr, "%s( %s = %s )\n", __FUNCTION__, _mesa_lookup_enum_by_nr(cap), @@ -2060,8 +1961,12 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state) case GL_POLYGON_OFFSET_FILL: r300SetPolygonOffsetState(ctx, state); break; + case GL_SCISSOR_TEST: + radeon_firevertices(&rmesa->radeon); + rmesa->radeon.state.scissor.enabled = state; + radeonUpdateScissor( ctx ); + break; default: - radeonEnable(ctx, cap, state); break; } } @@ -2080,7 +1985,7 @@ static void r300ResetHwState(r300ContextPtr r300) if (RADEON_DEBUG & DEBUG_STATE) fprintf(stderr, "%s\n", __FUNCTION__); - r300UpdateWindow(ctx); + radeon_firevertices(&r300->radeon); r300ColorMask(ctx, ctx->Color.ColorMask[RCOMP], @@ -2102,8 +2007,6 @@ static void r300ResetHwState(r300ContextPtr r300) r300UpdateCulling(ctx); - r300UpdateTextureState(ctx); - r300SetBlendState(ctx); r300SetLogicOpState(ctx); @@ -2242,20 +2145,6 @@ static void r300ResetHwState(r300ContextPtr r300) r300BlendColor(ctx, ctx->Color.BlendColor); - /* Again, r300ClearBuffer uses this */ - r300->hw.cb.cmd[R300_CB_OFFSET] = - r300->radeon.state.color.drawOffset + - r300->radeon.radeonScreen->fbLocation; - r300->hw.cb.cmd[R300_CB_PITCH] = r300->radeon.state.color.drawPitch; - - if (r300->radeon.radeonScreen->cpp == 4) - r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_ARGB8888; - else - r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_FORMAT_RGB565; - - if (r300->radeon.sarea->tiling_enabled) - r300->hw.cb.cmd[R300_CB_PITCH] |= R300_COLOR_TILE_ENABLE; - r300->hw.rb3d_dither_ctl.cmd[1] = 0; r300->hw.rb3d_dither_ctl.cmd[2] = 0; r300->hw.rb3d_dither_ctl.cmd[3] = 0; @@ -2271,41 +2160,15 @@ static void r300ResetHwState(r300ContextPtr r300) r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[1] = 0x00000000; r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[2] = 0xffffffff; - r300->hw.zb.cmd[R300_ZB_OFFSET] = - r300->radeon.radeonScreen->depthOffset + - r300->radeon.radeonScreen->fbLocation; - r300->hw.zb.cmd[R300_ZB_PITCH] = r300->radeon.radeonScreen->depthPitch; - - if (r300->radeon.sarea->tiling_enabled) { - /* XXX: Turn off when clearing buffers ? */ - r300->hw.zb.cmd[R300_ZB_PITCH] |= R300_DEPTHMACROTILE_ENABLE; - - if (ctx->Visual.depthBits == 24) - r300->hw.zb.cmd[R300_ZB_PITCH] |= - R300_DEPTHMICROTILE_TILED; - } - r300->hw.zb_depthclearvalue.cmd[1] = 0; - switch (ctx->Visual.depthBits) { - case 16: - r300->hw.zstencil_format.cmd[1] = R300_DEPTHFORMAT_16BIT_INT_Z; - break; - case 24: - r300->hw.zstencil_format.cmd[1] = R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL; - break; - default: - fprintf(stderr, "Error: Unsupported depth %d... exiting\n", ctx->Visual.depthBits); - _mesa_exit(-1); - } - r300->hw.zstencil_format.cmd[2] = R300_ZTOP_DISABLE; r300->hw.zstencil_format.cmd[3] = 0x00000003; r300->hw.zstencil_format.cmd[4] = 0x00000000; r300SetEarlyZState(ctx); - r300->hw.unk4F30.cmd[1] = 0; - r300->hw.unk4F30.cmd[2] = 0; + r300->hw.zb_zmask.cmd[1] = 0; + r300->hw.zb_zmask.cmd[2] = 0; r300->hw.zb_hiz_offset.cmd[1] = 0; @@ -2319,7 +2182,7 @@ static void r300ResetHwState(r300ContextPtr r300) r300->hw.vps.cmd[R300_VPS_ZERO_3] = 0; } - r300->hw.all_dirty = GL_TRUE; + r300->radeon.hw.all_dirty = GL_TRUE; } void r300UpdateShaders(r300ContextPtr rmesa) @@ -2330,8 +2193,8 @@ void r300UpdateShaders(r300ContextPtr rmesa) ctx = rmesa->radeon.glCtx; - if (rmesa->NewGLState && hw_tcl_on) { - rmesa->NewGLState = 0; + if (rmesa->radeon.NewGLState && hw_tcl_on) { + rmesa->radeon.NewGLState = 0; for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) { rmesa->temp_attrib[i] = @@ -2386,24 +2249,18 @@ static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx, } -static void r300SetupPixelShader(r300ContextPtr rmesa) +static GLboolean r300SetupPixelShader(GLcontext *ctx) { - GLcontext *ctx = rmesa->radeon.glCtx; - struct r300_fragment_program *fp = (struct r300_fragment_program *) - (char *)ctx->FragmentProgram._Current; + r300ContextPtr rmesa = R300_CONTEXT(ctx); + struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current; struct r300_fragment_program_code *code; int i, k; - if (!fp) /* should only happenen once, just after context is created */ - return; + /* Program is not native, fallback to software */ + if (fp->error) + return GL_FALSE; - r300TranslateFragmentShader(rmesa, fp); - if (!fp->translated) { - fprintf(stderr, "%s: No valid fragment shader, exiting\n", - __FUNCTION__); - return; - } - code = &fp->code; + code = &fp->code.r300; r300SetupTextures(ctx); @@ -2411,10 +2268,10 @@ static void r300SetupPixelShader(r300ContextPtr rmesa) R300_STATECHANGE(rmesa, fpi[1]); R300_STATECHANGE(rmesa, fpi[2]); R300_STATECHANGE(rmesa, fpi[3]); - rmesa->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_RGB_INST_0, code->alu.length); - rmesa->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_RGB_ADDR_0, code->alu.length); - rmesa->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_ALPHA_INST_0, code->alu.length); - rmesa->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(R300_US_ALU_ALPHA_ADDR_0, code->alu.length); + rmesa->hw.fpi[0].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_RGB_INST_0, code->alu.length); + rmesa->hw.fpi[1].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_RGB_ADDR_0, code->alu.length); + rmesa->hw.fpi[2].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_ALPHA_INST_0, code->alu.length); + rmesa->hw.fpi[3].cmd[R300_FPI_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_US_ALU_ALPHA_ADDR_0, code->alu.length); for (i = 0; i < code->alu.length; i++) { rmesa->hw.fpi[0].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].inst0; rmesa->hw.fpi[1].cmd[R300_FPI_INSTR_0 + i] = code->alu.inst[i].inst1; @@ -2445,15 +2302,17 @@ static void r300SetupPixelShader(r300ContextPtr rmesa) } R300_STATECHANGE(rmesa, fpp); - rmesa->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(R300_PFS_PARAM_0_X, code->const_nr * 4); + rmesa->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_PFS_PARAM_0_X, code->const_nr * 4); for (i = 0; i < code->const_nr; i++) { const GLfloat *constant = get_fragmentprogram_constant(ctx, - &fp->mesa_program.Base, code->constant[i]); + &fp->Base.Base, code->constant[i]); rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat24(constant[0]); rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat24(constant[1]); rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat24(constant[2]); rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 3] = r300PackFloat24(constant[3]); } + + return GL_TRUE; } #define bump_r500fp_count(ptr, new_count) do{\ @@ -2470,27 +2329,21 @@ static void r300SetupPixelShader(r300ContextPtr rmesa) if(_nc>_p->r500fp.count)_p->r500fp.count=_nc;\ } while(0) -static void r500SetupPixelShader(r300ContextPtr rmesa) +static GLboolean r500SetupPixelShader(GLcontext *ctx) { - GLcontext *ctx = rmesa->radeon.glCtx; - struct r500_fragment_program *fp = (struct r500_fragment_program *) - (char *)ctx->FragmentProgram._Current; + r300ContextPtr rmesa = R300_CONTEXT(ctx); + struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current; int i; struct r500_fragment_program_code *code; - if (!fp) /* should only happenen once, just after context is created */ - return; - ((drm_r300_cmd_header_t *) rmesa->hw.r500fp.cmd)->r500fp.count = 0; ((drm_r300_cmd_header_t *) rmesa->hw.r500fp_const.cmd)->r500fp.count = 0; - r500TranslateFragmentShader(rmesa, fp); - if (!fp->translated) { - fprintf(stderr, "%s: No valid fragment shader, exiting\n", - __FUNCTION__); - return; - } - code = &fp->code; + /* Program is not native, fallback to software */ + if (fp->error) + return GL_FALSE; + + code = &fp->code.r500; r300SetupTextures(ctx); @@ -2522,7 +2375,7 @@ static void r500SetupPixelShader(r300ContextPtr rmesa) R300_STATECHANGE(rmesa, r500fp_const); for (i = 0; i < code->const_nr; i++) { const GLfloat *constant = get_fragmentprogram_constant(ctx, - &fp->mesa_program.Base, code->constant[i]); + &fp->Base.Base, code->constant[i]); rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat32(constant[0]); rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat32(constant[1]); rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat32(constant[2]); @@ -2530,6 +2383,7 @@ static void r500SetupPixelShader(r300ContextPtr rmesa) } bump_r500fp_const_count(rmesa->hw.r500fp_const.cmd, code->const_nr * 4); + return GL_TRUE; } void r300UpdateShaderStates(r300ContextPtr rmesa) @@ -2537,7 +2391,10 @@ void r300UpdateShaderStates(r300ContextPtr rmesa) GLcontext *ctx; ctx = rmesa->radeon.glCtx; - r300UpdateTextureState(ctx); + /* should only happenen once, just after context is created */ + if (!ctx->FragmentProgram._Current) + return; + r300SetEarlyZState(ctx); /* w_fmt value is set to get best performance @@ -2561,19 +2418,15 @@ void r300UpdateShaderStates(r300ContextPtr rmesa) rmesa->hw.fg_depth_src.cmd[1] = fgdepthsrc; } - if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) - r500SetupPixelShader(rmesa); - else - r300SetupPixelShader(rmesa); + r300TranslateFragmentShader(ctx, ctx->FragmentProgram._Current); - if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) - r500SetupRSUnit(ctx); - else - r300SetupRSUnit(ctx); + if (!rmesa->vtbl.SetupPixelShader(ctx)) + return; + + rmesa->vtbl.SetupRSUnit(ctx); if ((rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) r300SetupVertexProgram(rmesa); - } /** @@ -2587,15 +2440,18 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state) _swsetup_InvalidateState(ctx, new_state); _vbo_InvalidateState(ctx, new_state); _tnl_InvalidateState(ctx, new_state); - _ae_invalidate_state(ctx, new_state); - if (new_state & _NEW_BUFFERS) { - r300UpdateDrawBuffer(ctx); + if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { + _mesa_update_framebuffer(ctx); + /* this updates the DrawBuffer's Width/Height if it's a FBO */ + _mesa_update_draw_buffer_bounds(ctx); + + R300_STATECHANGE(r300, cb); } r300UpdateStateParameters(ctx, new_state); - r300->NewGLState |= new_state; + r300->radeon.NewGLState |= new_state; } /** @@ -2605,32 +2461,6 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state) */ void r300InitState(r300ContextPtr r300) { - GLcontext *ctx = r300->radeon.glCtx; - GLuint depth_fmt; - - radeonInitState(&r300->radeon); - - switch (ctx->Visual.depthBits) { - case 16: - r300->state.depth.scale = 1.0 / (GLfloat) 0xffff; - depth_fmt = R300_DEPTHFORMAT_16BIT_INT_Z; - break; - case 24: - r300->state.depth.scale = 1.0 / (GLfloat) 0xffffff; - depth_fmt = R300_DEPTHFORMAT_24BIT_INT_Z_8BIT_STENCIL; - break; - default: - fprintf(stderr, "Error: Unsupported depth %d... exiting\n", - ctx->Visual.depthBits); - _mesa_exit(-1); - } - - /* Only have hw stencil when depth buffer is 24 bits deep */ - r300->state.stencil.hw_stencil = (ctx->Visual.stencilBits > 0 && - ctx->Visual.depthBits == 24); - - memset(&(r300->state.texture), 0, sizeof(r300->state.texture)); - r300ResetHwState(r300); } @@ -2641,30 +2471,11 @@ static void r300RenderMode(GLcontext * ctx, GLenum mode) (void)mode; } -void r300UpdateClipPlanes( GLcontext *ctx ) -{ - r300ContextPtr rmesa = R300_CONTEXT(ctx); - GLuint p; - - for (p = 0; p < ctx->Const.MaxClipPlanes; p++) { - if (ctx->Transform.ClipPlanesEnabled & (1 << p)) { - GLint *ip = (GLint *)ctx->Transform._ClipUserPlane[p]; - - R300_STATECHANGE( rmesa, vpucp[p] ); - rmesa->hw.vpucp[p].cmd[R300_VPUCP_X] = ip[0]; - rmesa->hw.vpucp[p].cmd[R300_VPUCP_Y] = ip[1]; - rmesa->hw.vpucp[p].cmd[R300_VPUCP_Z] = ip[2]; - rmesa->hw.vpucp[p].cmd[R300_VPUCP_W] = ip[3]; - } - } -} - /** * Initialize driver's state callback functions */ void r300InitStateFuncs(struct dd_function_table *functions) { - radeonInitStateFuncs(functions); functions->UpdateState = r300InvalidateState; functions->AlphaFunc = r300AlphaFunc; @@ -2700,4 +2511,25 @@ void r300InitStateFuncs(struct dd_function_table *functions) functions->RenderMode = r300RenderMode; functions->ClipPlane = r300ClipPlane; + functions->Scissor = radeonScissor; + + functions->DrawBuffer = radeonDrawBuffer; + functions->ReadBuffer = radeonReadBuffer; +} + +void r300InitShaderFunctions(r300ContextPtr r300) +{ + if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) { + r300->vtbl.SetupRSUnit = r500SetupRSUnit; + r300->vtbl.SetupPixelShader = r500SetupPixelShader; + r300->vtbl.SetupFragmentShaderTextures = r500SetupFragmentShaderTextures; + r300->vtbl.FragmentProgramEmit = r500FragmentProgramEmit; + r300->vtbl.FragmentProgramDump = r500FragmentProgramDump; + } else { + r300->vtbl.SetupRSUnit = r300SetupRSUnit; + r300->vtbl.SetupPixelShader = r300SetupPixelShader; + r300->vtbl.SetupFragmentShaderTextures = r300SetupFragmentShaderTextures; + r300->vtbl.FragmentProgramEmit = r300FragmentProgramEmit; + r300->vtbl.FragmentProgramDump = r300FragmentProgramDump; + } } |