From b44b70dc114ddcfb4d359759928df4d7b9efaafb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 14 Nov 2009 19:46:45 -0800 Subject: mesa/st: don't calculate unused vs input semantic tags --- src/mesa/state_tracker/st_program.c | 71 ++----------------------------------- 1 file changed, 2 insertions(+), 69 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index a9be80ce8f..b6c9e7463a 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -69,8 +69,6 @@ st_translate_vertex_program(struct st_context *st, GLuint attr, i; GLuint num_generic = 0; - ubyte vs_input_semantic_name[PIPE_MAX_SHADER_INPUTS]; - ubyte vs_input_semantic_index[PIPE_MAX_SHADER_INPUTS]; uint vs_num_inputs = 0; ubyte vs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; @@ -100,71 +98,6 @@ st_translate_vertex_program(struct st_context *st, stvp->input_to_index[attr] = slot; stvp->index_to_input[slot] = attr; - switch (attr) { - case VERT_ATTRIB_POS: - vs_input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - vs_input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_WEIGHT: - /* fall-through */ - case VERT_ATTRIB_NORMAL: - /* just label as a generic */ - vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs_input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_COLOR0: - vs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs_input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_COLOR1: - vs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs_input_semantic_index[slot] = 1; - break; - case VERT_ATTRIB_FOG: - vs_input_semantic_name[slot] = TGSI_SEMANTIC_FOG; - vs_input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_POINT_SIZE: - vs_input_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; - vs_input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_TEX0: - case VERT_ATTRIB_TEX1: - case VERT_ATTRIB_TEX2: - case VERT_ATTRIB_TEX3: - case VERT_ATTRIB_TEX4: - case VERT_ATTRIB_TEX5: - case VERT_ATTRIB_TEX6: - case VERT_ATTRIB_TEX7: - assert(slot < Elements(vs_input_semantic_name)); - vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs_input_semantic_index[slot] = num_generic++; - break; - case VERT_ATTRIB_GENERIC0: - case VERT_ATTRIB_GENERIC1: - case VERT_ATTRIB_GENERIC2: - case VERT_ATTRIB_GENERIC3: - case VERT_ATTRIB_GENERIC4: - case VERT_ATTRIB_GENERIC5: - case VERT_ATTRIB_GENERIC6: - case VERT_ATTRIB_GENERIC7: - case VERT_ATTRIB_GENERIC8: - case VERT_ATTRIB_GENERIC9: - case VERT_ATTRIB_GENERIC10: - case VERT_ATTRIB_GENERIC11: - case VERT_ATTRIB_GENERIC12: - case VERT_ATTRIB_GENERIC13: - case VERT_ATTRIB_GENERIC14: - case VERT_ATTRIB_GENERIC15: - assert(attr < VERT_ATTRIB_MAX); - assert(slot < Elements(vs_input_semantic_name)); - vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs_input_semantic_index[slot] = num_generic++; - break; - default: - assert(0); - } - input_flags[slot] = stvp->Base.Base.InputFlags[attr]; } } @@ -330,8 +263,8 @@ st_translate_vertex_program(struct st_context *st, /* inputs */ vs_num_inputs, stvp->input_to_index, - vs_input_semantic_name, - vs_input_semantic_index, + NULL, /* input semantic name */ + NULL, /* input semantic index */ NULL, input_flags, /* outputs */ -- cgit v1.2.3 From 0ffdd1c1e897c448cbc359aea3e3a3ba098bbfe5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 14 Nov 2009 19:48:42 -0800 Subject: mesa/st: don't calculate unused input_flags data --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 1 - src/mesa/state_tracker/st_mesa_to_tgsi.h | 1 - src/mesa/state_tracker/st_program.c | 16 ++-------------- 3 files changed, 2 insertions(+), 16 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 3d6c215819..81f02d01e9 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -704,7 +704,6 @@ st_translate_mesa_program( const ubyte inputSemanticName[], const ubyte inputSemanticIndex[], const GLuint interpMode[], - const GLbitfield inputFlags[], GLuint numOutputs, const GLuint outputMapping[], const ubyte outputSemanticName[], diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index c0d1ff59e1..384831762d 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -49,7 +49,6 @@ st_translate_mesa_program( const ubyte inputSemanticName[], const ubyte inputSemanticIndex[], const GLuint interpMode[], - const GLbitfield inputFlags[], GLuint numOutputs, const GLuint outputMapping[], const ubyte outputSemanticName[], diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index b6c9e7463a..fd1b213eb8 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -75,11 +75,9 @@ st_translate_vertex_program(struct st_context *st, ubyte vs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; uint vs_num_outputs = 0; - GLbitfield input_flags[MAX_PROGRAM_INPUTS]; GLbitfield output_flags[MAX_PROGRAM_OUTPUTS]; // memset(&vs, 0, sizeof(vs)); - memset(input_flags, 0, sizeof(input_flags)); memset(output_flags, 0, sizeof(output_flags)); if (stvp->Base.IsPositionInvariant) @@ -91,14 +89,10 @@ st_translate_vertex_program(struct st_context *st, */ for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { if (stvp->Base.Base.InputsRead & (1 << attr)) { - const GLuint slot = vs_num_inputs; + stvp->input_to_index[attr] = vs_num_inputs; + stvp->index_to_input[vs_num_inputs] = attr; vs_num_inputs++; - - stvp->input_to_index[attr] = slot; - stvp->index_to_input[slot] = attr; - - input_flags[slot] = stvp->Base.Base.InputFlags[attr]; } } @@ -266,7 +260,6 @@ st_translate_vertex_program(struct st_context *st, NULL, /* input semantic name */ NULL, /* input semantic index */ NULL, - input_flags, /* outputs */ vs_num_outputs, outputMapping, @@ -316,11 +309,9 @@ st_translate_fragment_program(struct st_context *st, ubyte fs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; uint fs_num_outputs = 0; - GLbitfield input_flags[MAX_PROGRAM_INPUTS]; GLbitfield output_flags[MAX_PROGRAM_OUTPUTS]; // memset(&fs, 0, sizeof(fs)); - memset(input_flags, 0, sizeof(input_flags)); memset(output_flags, 0, sizeof(output_flags)); /* which vertex output goes to the first fragment input: */ @@ -392,8 +383,6 @@ st_translate_fragment_program(struct st_context *st, stfp->input_semantic_index[slot] = num_generic++; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; } - - input_flags[slot] = stfp->Base.Base.InputFlags[attr]; } } @@ -451,7 +440,6 @@ st_translate_fragment_program(struct st_context *st, stfp->input_semantic_name, stfp->input_semantic_index, interpMode, - input_flags, /* outputs */ fs_num_outputs, outputMapping, -- cgit v1.2.3 From 9953fe4cb48b02a0d75735b88173f0ed170a77f2 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 14 Nov 2009 19:50:45 -0800 Subject: mesa/st: don't calculate unused output_flags data either --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 3 +-- src/mesa/state_tracker/st_mesa_to_tgsi.h | 3 +-- src/mesa/state_tracker/st_program.c | 22 ++-------------------- 3 files changed, 4 insertions(+), 24 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 81f02d01e9..d84832f539 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -707,8 +707,7 @@ st_translate_mesa_program( GLuint numOutputs, const GLuint outputMapping[], const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - const GLbitfield outputFlags[] ) + const ubyte outputSemanticIndex[] ) { struct st_translate translate, *t; struct ureg_program *ureg; diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index 384831762d..dc0362fe79 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -52,8 +52,7 @@ st_translate_mesa_program( GLuint numOutputs, const GLuint outputMapping[], const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - const GLbitfield outputFlags[] ); + const ubyte outputSemanticIndex[] ); void st_free_tokens(const struct tgsi_token *tokens); diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index fd1b213eb8..5450b59a2d 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -75,11 +75,6 @@ st_translate_vertex_program(struct st_context *st, ubyte vs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; uint vs_num_outputs = 0; - GLbitfield output_flags[MAX_PROGRAM_OUTPUTS]; - -// memset(&vs, 0, sizeof(vs)); - memset(output_flags, 0, sizeof(output_flags)); - if (stvp->Base.IsPositionInvariant) _mesa_insert_mvp_code(st->ctx, &stvp->Base); @@ -115,7 +110,6 @@ st_translate_vertex_program(struct st_context *st, assert(i < Elements(vs_output_semantic_name)); vs_output_semantic_name[i] = TGSI_SEMANTIC_GENERIC; vs_output_semantic_index[i] = 0; - output_flags[i] = 0x0; } num_generic = 0; @@ -201,9 +195,6 @@ st_translate_vertex_program(struct st_context *st, vs_output_semantic_index[slot] = num_generic++; } } - - assert(slot < Elements(output_flags)); - output_flags[slot] = stvp->Base.Base.OutputFlags[attr]; } } @@ -264,8 +255,7 @@ st_translate_vertex_program(struct st_context *st, vs_num_outputs, outputMapping, vs_output_semantic_name, - vs_output_semantic_index, - output_flags ); + vs_output_semantic_index ); stvp->num_inputs = vs_num_inputs; stvp->driver_shader = pipe->create_vs_state(pipe, &stvp->state); @@ -309,11 +299,6 @@ st_translate_fragment_program(struct st_context *st, ubyte fs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; uint fs_num_outputs = 0; - GLbitfield output_flags[MAX_PROGRAM_OUTPUTS]; - -// memset(&fs, 0, sizeof(fs)); - memset(output_flags, 0, sizeof(output_flags)); - /* which vertex output goes to the first fragment input: */ if (inputsRead & FRAG_BIT_WPOS) vslot = 0; @@ -420,8 +405,6 @@ st_translate_fragment_program(struct st_context *st, break; } - output_flags[fs_num_outputs] = stfp->Base.Base.OutputFlags[attr]; - fs_num_outputs++; } } @@ -444,8 +427,7 @@ st_translate_fragment_program(struct st_context *st, fs_num_outputs, outputMapping, fs_output_semantic_name, - fs_output_semantic_index, - output_flags ); + fs_output_semantic_index ); stfp->driver_shader = pipe->create_fs_state(pipe, &stfp->state); -- cgit v1.2.3 From 07fafc7c9346aa260829603bf3188596481e9e62 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 15 Nov 2009 11:15:25 -0800 Subject: mesa/st: refactor vertex and fragment shader translation Translate vertex shaders independently of fragment shaders. Previously tried to make fragment shader semantic indexes always start at zero and exclude holes. This was unnecessary but meant that vertex shader translation had to be adjusted to take this into account. Now use a fixed scheme for labelling special FS input semantics (color, etc), and another fixed scheme for the generics. With this, vertex shaders can be translated independently of the bound fragment shader, assuming mesa has done its own job and ensured that the vertex shader provides at least the inputs the fragment shader is looking for. The state-tracker didn't attempt to do anything about this previously, so it shouldn't be needed now. --- src/mesa/state_tracker/st_atom.c | 3 +- src/mesa/state_tracker/st_atom.h | 3 +- src/mesa/state_tracker/st_atom_shader.c | 326 ++++++++---------------------- src/mesa/state_tracker/st_cb_bitmap.c | 5 - src/mesa/state_tracker/st_cb_drawpixels.c | 8 +- src/mesa/state_tracker/st_cb_program.c | 43 +--- src/mesa/state_tracker/st_context.h | 2 + src/mesa/state_tracker/st_debug.c | 3 +- src/mesa/state_tracker/st_draw.c | 2 +- src/mesa/state_tracker/st_draw_feedback.c | 8 +- src/mesa/state_tracker/st_program.c | 267 +++++++++++------------- src/mesa/state_tracker/st_program.h | 75 +++++-- 12 files changed, 276 insertions(+), 469 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index ca15ce1b47..dfce955fd9 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -46,7 +46,8 @@ static const struct st_tracked_state *atoms[] = &st_update_clip, &st_finalize_textures, - &st_update_shader, + &st_update_fp, + &st_update_vp, &st_update_rasterizer, &st_update_polygon_stipple, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index c7cffd85c8..f34b49203b 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -47,7 +47,8 @@ void st_validate_state( struct st_context *st ); extern const struct st_tracked_state st_update_framebuffer; extern const struct st_tracked_state st_update_clip; extern const struct st_tracked_state st_update_depth_stencil_alpha; -extern const struct st_tracked_state st_update_shader; +extern const struct st_tracked_state st_update_fp; +extern const struct st_tracked_state st_update_vp; extern const struct st_tracked_state st_update_rasterizer; extern const struct st_tracked_state st_update_polygon_stipple; extern const struct st_tracked_state st_update_viewport; diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index ee649be885..09baff875b 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -56,82 +56,18 @@ #include "st_mesa_to_tgsi.h" -/** - * This represents a vertex program, especially translated to match - * the inputs of a particular fragment shader. - */ -struct translated_vertex_program -{ - struct st_vertex_program *master; - - /** The fragment shader "signature" this vertex shader is meant for: */ - GLbitfield frag_inputs; - /** Compared against master vertex program's serialNo: */ - GLuint serialNo; - /** Maps VERT_RESULT_x to slot */ - GLuint output_to_slot[VERT_RESULT_MAX]; - ubyte output_to_semantic_name[VERT_RESULT_MAX]; - ubyte output_to_semantic_index[VERT_RESULT_MAX]; - - /** Pointer to the translated vertex program */ - struct st_vertex_program *vp; - - struct translated_vertex_program *next; /**< next in linked list */ -}; - - -/** - * Given a vertex program output attribute, return the corresponding - * fragment program input attribute. - * \return -1 for vertex outputs that have no corresponding fragment input +/* + * Translate fragment program if needed. */ -static GLint -vp_out_to_fp_in(GLuint vertResult) -{ - if (vertResult >= VERT_RESULT_TEX0 && - vertResult < VERT_RESULT_TEX0 + MAX_TEXTURE_COORD_UNITS) - return FRAG_ATTRIB_TEX0 + (vertResult - VERT_RESULT_TEX0); - - if (vertResult >= VERT_RESULT_VAR0 && - vertResult < VERT_RESULT_VAR0 + MAX_VARYING) - return FRAG_ATTRIB_VAR0 + (vertResult - VERT_RESULT_VAR0); - - switch (vertResult) { - case VERT_RESULT_HPOS: - return FRAG_ATTRIB_WPOS; - case VERT_RESULT_COL0: - return FRAG_ATTRIB_COL0; - case VERT_RESULT_COL1: - return FRAG_ATTRIB_COL1; - case VERT_RESULT_FOGC: - return FRAG_ATTRIB_FOGC; - default: - /* Back-face colors, edge flags, etc */ - return -1; - } -} - - -/** - * Find a translated vertex program that corresponds to stvp and - * has outputs matched to stfp's inputs. - * This performs vertex and fragment translation (to TGSI) when needed. - */ -static struct translated_vertex_program * -find_translated_vp(struct st_context *st, - struct st_vertex_program *stvp, - struct st_fragment_program *stfp) +static void +translate_fp(struct st_context *st, + struct st_fragment_program *stfp) { - static const GLuint UNUSED = ~0; - struct translated_vertex_program *xvp; const GLbitfield fragInputsRead = stfp->Base.Base.InputsRead; - /* - * Translate fragment program if needed. - */ if (!stfp->state.tokens) { GLuint inAttr, numIn = 0; @@ -141,7 +77,7 @@ find_translated_vp(struct st_context *st, numIn++; } else { - stfp->input_to_slot[inAttr] = UNUSED; + stfp->input_to_slot[inAttr] = -1; } } @@ -151,170 +87,63 @@ find_translated_vp(struct st_context *st, st_translate_fragment_program(st, stfp, stfp->input_to_slot); } +} - /* See if we've got a translated vertex program whose outputs match - * the fragment program's inputs. - * XXX This could be a hash lookup, using InputsRead as the key. - */ - for (xvp = stfp->vertex_programs; xvp; xvp = xvp->next) { - if (xvp->master == stvp && xvp->frag_inputs == fragInputsRead) { - break; - } - } - /* No? Allocate translated vp object now */ - if (!xvp) { - xvp = ST_CALLOC_STRUCT(translated_vertex_program); - xvp->frag_inputs = fragInputsRead; - xvp->master = stvp; +/** + * Find a translated vertex program that corresponds to stvp and + * has outputs matched to stfp's inputs. + * This performs vertex and fragment translation (to TGSI) when needed. + */ +static struct st_vp_varient * +find_translated_vp(struct st_context *st, + struct st_vertex_program *stvp ) +{ + struct st_vp_varient *vpv; + struct st_vp_varient_key key; - xvp->next = stfp->vertex_programs; - stfp->vertex_programs = xvp; - } + /* Nothing in our key yet. This will change: + */ + memset(&key, 0, sizeof key); + key.dummy = 0; - /* See if we need to translate vertex program to TGSI form */ - if (xvp->serialNo != stvp->serialNo) { - GLuint outAttr; - const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten; - GLuint numVpOuts = 0; - GLboolean emitPntSize = GL_FALSE, emitBFC0 = GL_FALSE, emitBFC1 = GL_FALSE; - GLbitfield usedGenerics = 0x0; - GLbitfield usedOutputSlots = 0x0; - - /* Compute mapping of vertex program outputs to slots, which depends - * on the fragment program's input->slot mapping. + /* Do we need to throw away old translations after a change in the + * GL program string? + */ + if (stvp->serialNo != stvp->lastSerialNo) { + /* These may have changed if the program string changed. */ - for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { - /* set defaults: */ - xvp->output_to_slot[outAttr] = UNUSED; - xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_COUNT; - xvp->output_to_semantic_index[outAttr] = 99; - - if (outAttr == VERT_RESULT_HPOS) { - /* always put xformed position into slot zero */ - GLuint slot = 0; - xvp->output_to_slot[VERT_RESULT_HPOS] = slot; - xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_POSITION; - xvp->output_to_semantic_index[outAttr] = 0; - numVpOuts++; - usedOutputSlots |= (1 << slot); - } - else if (outputsWritten & (1 << outAttr)) { - /* see if the frag prog wants this vert output */ - GLint fpInAttrib = vp_out_to_fp_in(outAttr); - if (fpInAttrib >= 0) { - GLuint fpInSlot = stfp->input_to_slot[fpInAttrib]; - if (fpInSlot != ~0) { - /* match this vp output to the fp input */ - GLuint vpOutSlot = stfp->input_map[fpInSlot]; - xvp->output_to_slot[outAttr] = vpOutSlot; - xvp->output_to_semantic_name[outAttr] = stfp->input_semantic_name[fpInSlot]; - xvp->output_to_semantic_index[outAttr] = stfp->input_semantic_index[fpInSlot]; - numVpOuts++; - usedOutputSlots |= (1 << vpOutSlot); - } - else { -#if 0 /*debug*/ - printf("VP output %d not used by FP\n", outAttr); -#endif - } - } - else if (outAttr == VERT_RESULT_PSIZ) - emitPntSize = GL_TRUE; - else if (outAttr == VERT_RESULT_BFC0) - emitBFC0 = GL_TRUE; - else if (outAttr == VERT_RESULT_BFC1) - emitBFC1 = GL_TRUE; - } -#if 0 /*debug*/ - printf("assign vp output_to_slot[%d] = %d\n", outAttr, - xvp->output_to_slot[outAttr]); -#endif - } - - /* must do these last */ - if (emitPntSize) { - GLuint slot = numVpOuts++; - xvp->output_to_slot[VERT_RESULT_PSIZ] = slot; - xvp->output_to_semantic_name[VERT_RESULT_PSIZ] = TGSI_SEMANTIC_PSIZE; - xvp->output_to_semantic_index[VERT_RESULT_PSIZ] = 0; - usedOutputSlots |= (1 << slot); - } - if (emitBFC0) { - GLuint slot = numVpOuts++; - xvp->output_to_slot[VERT_RESULT_BFC0] = slot; - xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR; - xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 0; - usedOutputSlots |= (1 << slot); - } - if (emitBFC1) { - GLuint slot = numVpOuts++; - xvp->output_to_slot[VERT_RESULT_BFC1] = slot; - xvp->output_to_semantic_name[VERT_RESULT_BFC1] = TGSI_SEMANTIC_COLOR; - xvp->output_to_semantic_index[VERT_RESULT_BFC1] = 1; - usedOutputSlots |= (1 << slot); - } - - /* build usedGenerics mask */ - usedGenerics = 0x0; - for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { - if (xvp->output_to_semantic_name[outAttr] == TGSI_SEMANTIC_GENERIC) { - usedGenerics |= (1 << xvp->output_to_semantic_index[outAttr]); - } - } + st_prepare_vertex_program( st, stvp ); - /* For each vertex program output that doesn't match up to a fragment - * program input, map the vertex program output to a free slot and - * free generic attribute. + /* We are now up-to-date: */ - for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { - if (outputsWritten & (1 << outAttr)) { - if (xvp->output_to_slot[outAttr] == UNUSED) { - GLint freeGeneric = _mesa_ffs(~usedGenerics) - 1; - GLint freeSlot = _mesa_ffs(~usedOutputSlots) - 1; - usedGenerics |= (1 << freeGeneric); - usedOutputSlots |= (1 << freeSlot); - xvp->output_to_slot[outAttr] = freeSlot; - xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_GENERIC; - xvp->output_to_semantic_index[outAttr] = freeGeneric; - } - } - -#if 0 /*debug*/ - printf("vp output_to_slot[%d] = %d\n", outAttr, - xvp->output_to_slot[outAttr]); -#endif + stvp->lastSerialNo = stvp->serialNo; + } + + /* See if we've got a translated vertex program whose outputs match + * the fragment program's inputs. + */ + for (vpv = stvp->varients; vpv; vpv = vpv->next) { + if (memcmp(&vpv->key, &key, sizeof key) == 0) { + break; } + } - assert(stvp->Base.Base.NumInstructions > 1); - - st_translate_vertex_program(st, stvp, xvp->output_to_slot, - xvp->output_to_semantic_name, - xvp->output_to_semantic_index); - - xvp->vp = stvp; - - /* translated VP is up to date now */ - xvp->serialNo = stvp->serialNo; + /* No? Perform new translation here. */ + if (!vpv) { + vpv = st_translate_vertex_program(st, stvp, &key); + if (!vpv) + return NULL; + + vpv->next = stvp->varients; + stvp->varients = vpv; } - return xvp; + return vpv; } -void -st_free_translated_vertex_programs(struct st_context *st, - struct translated_vertex_program *xvp) -{ - struct translated_vertex_program *next; - - while (xvp) { - next = xvp->next; - _mesa_free(xvp); - xvp = next; - } -} static void * @@ -328,32 +157,19 @@ get_passthrough_fs(struct st_context *st) return st->passthrough_fs; } - static void -update_linkage( struct st_context *st ) +update_fp( struct st_context *st ) { - struct st_vertex_program *stvp; struct st_fragment_program *stfp; - struct translated_vertex_program *xvp; - - /* find active shader and params -- Should be covered by - * ST_NEW_VERTEX_PROGRAM - */ - assert(st->ctx->VertexProgram._Current); - stvp = st_vertex_program(st->ctx->VertexProgram._Current); - assert(stvp->Base.Base.Target == GL_VERTEX_PROGRAM_ARB); assert(st->ctx->FragmentProgram._Current); stfp = st_fragment_program(st->ctx->FragmentProgram._Current); assert(stfp->Base.Base.Target == GL_FRAGMENT_PROGRAM_ARB); - xvp = find_translated_vp(st, stvp, stfp); + translate_fp(st, stfp); - st_reference_vertprog(st, &st->vp, stvp); st_reference_fragprog(st, &st->fp, stfp); - cso_set_vertex_shader_handle(st->cso_context, stvp->driver_shader); - if (st->missing_textures) { /* use a pass-through frag shader that uses no textures */ void *fs = get_passthrough_fs(st); @@ -362,16 +178,48 @@ update_linkage( struct st_context *st ) else { cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader); } +} + +const struct st_tracked_state st_update_fp = { + "st_update_fp", /* name */ + { /* dirty */ + 0, /* mesa */ + ST_NEW_FRAGMENT_PROGRAM /* st */ + }, + update_fp /* update */ +}; + + + + +static void +update_vp( struct st_context *st ) +{ + struct st_vertex_program *stvp; + + /* find active shader and params -- Should be covered by + * ST_NEW_VERTEX_PROGRAM + */ + assert(st->ctx->VertexProgram._Current); + stvp = st_vertex_program(st->ctx->VertexProgram._Current); + assert(stvp->Base.Base.Target == GL_VERTEX_PROGRAM_ARB); + + st->vp_varient = find_translated_vp(st, stvp); + + st_reference_vertprog(st, &st->vp, stvp); + + cso_set_vertex_shader_handle(st->cso_context, + st->vp_varient->driver_shader); - st->vertex_result_to_slot = xvp->output_to_slot; + st->vertex_result_to_slot = stvp->result_to_output; } -const struct st_tracked_state st_update_shader = { - "st_update_shader", /* name */ +const struct st_tracked_state st_update_vp = { + "st_update_vp", /* name */ { /* dirty */ 0, /* mesa */ - ST_NEW_VERTEX_PROGRAM | ST_NEW_FRAGMENT_PROGRAM /* st */ + ST_NEW_VERTEX_PROGRAM /* st */ }, - update_linkage /* update */ + update_vp /* update */ }; diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index a22fa68299..226d1ab14c 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -169,11 +169,6 @@ make_bitmap_fragment_program(GLcontext *ctx, GLuint samplerIndex) stfp = (struct st_fragment_program *) p; stfp->Base.UsesKill = GL_TRUE; - /* No need to send this incomplete program down to hardware: - * - * st_translate_fragment_program(ctx->st, stfp, NULL); - */ - return stfp; } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 13e454b7a7..60394731f7 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -927,7 +927,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; struct st_renderbuffer *rbRead; - struct st_vertex_program *stvp; + void *driver_vp; struct st_fragment_program *stfp; struct pipe_texture *pt; GLfloat *color; @@ -976,14 +976,14 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, rbRead = st_get_color_read_renderbuffer(ctx); color = NULL; stfp = combined_drawpix_fragment_program(ctx); - stvp = st_make_passthrough_vertex_shader(st, GL_FALSE); + driver_vp = st_make_passthrough_vertex_shader(st, GL_FALSE); } else { assert(type == GL_DEPTH); rbRead = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer); color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; stfp = make_fragment_shader_z(st); - stvp = st_make_passthrough_vertex_shader(st, GL_TRUE); + driver_vp = st_make_passthrough_vertex_shader(st, GL_TRUE); } srcFormat = rbRead->texture->format; @@ -1116,7 +1116,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, pt, - stvp->driver_shader, + driver_vp, stfp->driver_shader, color, GL_TRUE); diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index b2d5c39a3a..8c276f8128 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -138,24 +138,7 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; - - if (stvp->driver_shader) { - cso_delete_vertex_shader(st->cso_context, stvp->driver_shader); - stvp->driver_shader = NULL; - } - - if (stvp->draw_shader) { -#if FEATURE_feedback || FEATURE_drawpix - /* this would only have been allocated for the RasterPos path */ - draw_delete_vertex_shader(st->draw, stvp->draw_shader); - stvp->draw_shader = NULL; -#endif - } - - if (stvp->state.tokens) { - st_free_tokens(stvp->state.tokens); - stvp->state.tokens = NULL; - } + st_vp_release_varients( st, stvp ); } break; case GL_FRAGMENT_PROGRAM_ARB: @@ -177,8 +160,6 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) _mesa_reference_program(ctx, &prg, NULL); stfp->bitmap_program = NULL; } - - st_free_translated_vertex_programs(st, stfp->vertex_programs); } break; default: @@ -219,8 +200,6 @@ static void st_program_string_notify( GLcontext *ctx, stfp->state.tokens = NULL; } - stfp->param_state = stfp->Base.Base.Parameters->StateFlags; - if (st->fp == stfp) st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; } @@ -229,25 +208,7 @@ static void st_program_string_notify( GLcontext *ctx, stvp->serialNo++; - if (stvp->driver_shader) { - cso_delete_vertex_shader(st->cso_context, stvp->driver_shader); - stvp->driver_shader = NULL; - } - - if (stvp->draw_shader) { -#if FEATURE_feedback || FEATURE_drawpix - /* this would only have been allocated for the RasterPos path */ - draw_delete_vertex_shader(st->draw, stvp->draw_shader); - stvp->draw_shader = NULL; -#endif - } - - if (stvp->state.tokens) { - st_free_tokens(stvp->state.tokens); - stvp->state.tokens = NULL; - } - - stvp->param_state = stvp->Base.Base.Parameters->StateFlags; + st_vp_release_varients( st, stvp ); if (st->vp == stvp) st->dirty.st |= ST_NEW_VERTEX_PROGRAM; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 18adb35e87..b760728658 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -127,6 +127,8 @@ struct st_context struct st_vertex_program *vp; /**< Currently bound vertex program */ struct st_fragment_program *fp; /**< Currently bound fragment program */ + struct st_vp_varient *vp_varient; + struct gl_texture_object *default_texture; struct { diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c index 3009cde9d5..6e699ca552 100644 --- a/src/mesa/state_tracker/st_debug.c +++ b/src/mesa/state_tracker/st_debug.c @@ -86,7 +86,8 @@ st_print_current(void) } #endif - tgsi_dump( st->vp->state.tokens, 0 ); + if (st->vp->varients) + tgsi_dump( st->vp->varients[0].state.tokens, 0 ); if (st->vp->Base.Base.Parameters) _mesa_print_parameter_list(st->vp->Base.Base.Parameters); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index c76bff9181..337c21a6c4 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -563,7 +563,7 @@ st_draw_vbo(GLcontext *ctx, /* must get these after state validation! */ vp = ctx->st->vp; - vs = &ctx->st->vp->state; + vs = &ctx->st->vp_varient->state; #if 0 if (MESA_VERBOSE & VERBOSE_GLSL) { diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c index b2d682ef64..d793f820bc 100644 --- a/src/mesa/state_tracker/st_draw_feedback.c +++ b/src/mesa/state_tracker/st_draw_feedback.c @@ -120,10 +120,10 @@ st_feedback_draw_vbo(GLcontext *ctx, /* must get these after state validation! */ vp = ctx->st->vp; - vs = &st->vp->state; + vs = &st->vp_varient->state; - if (!st->vp->draw_shader) { - st->vp->draw_shader = draw_create_vertex_shader(draw, vs); + if (!st->vp_varient->draw_shader) { + st->vp_varient->draw_shader = draw_create_vertex_shader(draw, vs); } /* @@ -136,7 +136,7 @@ st_feedback_draw_vbo(GLcontext *ctx, draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); draw_set_rasterizer_state(draw, &st->state.rasterizer); - draw_bind_vertex_shader(draw, st->vp->draw_shader); + draw_bind_vertex_shader(draw, st->vp_varient->draw_shader); set_feedback_vertex_format(ctx); /* loop over TGSI shader inputs to determine vertex buffer diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 5450b59a2d..6a2b99cf1b 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -49,6 +49,36 @@ #include "st_mesa_to_tgsi.h" #include "cso_cache/cso_context.h" + /* Clean out any old compilations: + */ +void +st_vp_release_varients( struct st_context *st, + struct st_vertex_program *stvp ) +{ + struct st_vp_varient *vpv; + + for (vpv = stvp->varients; vpv; ) { + struct st_vp_varient *next = vpv->next; + + if (vpv->driver_shader) + cso_delete_vertex_shader(st->cso_context, vpv->driver_shader); + + if (vpv->draw_shader) + draw_delete_vertex_shader( st->draw, vpv->draw_shader ); + + if (vpv->state.tokens) + st_free_tokens(vpv->state.tokens); + + FREE( vpv ); + + vpv = next; + } + + stvp->varients = NULL; +} + + + /** * Translate a Mesa vertex shader into a TGSI shader. @@ -58,22 +88,13 @@ * \return pointer to cached pipe_shader object. */ void -st_translate_vertex_program(struct st_context *st, - struct st_vertex_program *stvp, - const GLuint outputMapping[], - const ubyte *outputSemanticName, - const ubyte *outputSemanticIndex) +st_prepare_vertex_program(struct st_context *st, + struct st_vertex_program *stvp) { - struct pipe_context *pipe = st->pipe; - GLuint defaultOutputMapping[VERT_RESULT_MAX]; - GLuint attr, i; - GLuint num_generic = 0; - - uint vs_num_inputs = 0; + GLuint attr; - ubyte vs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; - ubyte vs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; - uint vs_num_outputs = 0; + stvp->num_inputs = 0; + stvp->num_outputs = 0; if (stvp->Base.IsPositionInvariant) _mesa_insert_mvp_code(st->ctx, &stvp->Base); @@ -84,92 +105,56 @@ st_translate_vertex_program(struct st_context *st, */ for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { if (stvp->Base.Base.InputsRead & (1 << attr)) { - stvp->input_to_index[attr] = vs_num_inputs; - stvp->index_to_input[vs_num_inputs] = attr; - - vs_num_inputs++; + stvp->input_to_index[attr] = stvp->num_inputs; + stvp->index_to_input[stvp->num_inputs] = attr; + stvp->num_inputs++; } } -#if 0 - if (outputMapping && outputSemanticName) { - printf("VERT_RESULT written out_slot semantic_name semantic_index\n"); - for (attr = 0; attr < VERT_RESULT_MAX; attr++) { - printf(" %-2d %c %3d %2d %2d\n", - attr, - ((stvp->Base.Base.OutputsWritten & (1 << attr)) ? 'Y' : ' '), - outputMapping[attr], - outputSemanticName[attr], - outputSemanticIndex[attr]); - } - } -#endif - - /* initialize output semantics to defaults */ - for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) { - assert(i < Elements(vs_output_semantic_name)); - vs_output_semantic_name[i] = TGSI_SEMANTIC_GENERIC; - vs_output_semantic_index[i] = 0; - } - - num_generic = 0; - /* - * Determine number of outputs, the (default) output register - * mapping and the semantic information for each output. + /* Compute mapping of vertex program outputs to slots. */ for (attr = 0; attr < VERT_RESULT_MAX; attr++) { - if (stvp->Base.Base.OutputsWritten & (1 << attr)) { - GLuint slot; - - /* XXX - * Pass in the fragment program's input's semantic info. - * Use the generic semantic indexes from there, instead of - * guessing below. - */ - - if (outputMapping) { - slot = outputMapping[attr]; - assert(slot != ~0); - } - else { - slot = vs_num_outputs; - vs_num_outputs++; - defaultOutputMapping[attr] = slot; - } + if ((stvp->Base.Base.OutputsWritten & (1 << attr)) == 0) { + stvp->result_to_output[attr] = ~0; + } + else { + unsigned slot = stvp->num_outputs++; + + stvp->result_to_output[attr] = slot; switch (attr) { case VERT_RESULT_HPOS: - assert(slot == 0); - vs_output_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - vs_output_semantic_index[slot] = 0; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + stvp->output_semantic_index[slot] = 0; break; case VERT_RESULT_COL0: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs_output_semantic_index[slot] = 0; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + stvp->output_semantic_index[slot] = 0; break; case VERT_RESULT_COL1: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs_output_semantic_index[slot] = 1; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + stvp->output_semantic_index[slot] = 1; break; case VERT_RESULT_BFC0: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; - vs_output_semantic_index[slot] = 0; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; + stvp->output_semantic_index[slot] = 0; break; case VERT_RESULT_BFC1: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; - vs_output_semantic_index[slot] = 1; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; + stvp->output_semantic_index[slot] = 1; break; case VERT_RESULT_FOGC: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_FOG; - vs_output_semantic_index[slot] = 0; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_FOG; + stvp->output_semantic_index[slot] = 0; break; case VERT_RESULT_PSIZ: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; - vs_output_semantic_index[slot] = 0; + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; + stvp->output_semantic_index[slot] = 0; break; case VERT_RESULT_EDGE: assert(0); break; + case VERT_RESULT_TEX0: case VERT_RESULT_TEX1: case VERT_RESULT_TEX2: @@ -178,87 +163,50 @@ st_translate_vertex_program(struct st_context *st, case VERT_RESULT_TEX5: case VERT_RESULT_TEX6: case VERT_RESULT_TEX7: - /* fall-through */ + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + stvp->output_semantic_index[slot] = attr - VERT_RESULT_TEX0; + break; + case VERT_RESULT_VAR0: - /* fall-through */ default: - assert(slot < Elements(vs_output_semantic_name)); - if (outputSemanticName) { - /* use provided semantic into */ - assert(outputSemanticName[attr] != TGSI_SEMANTIC_COUNT); - vs_output_semantic_name[slot] = outputSemanticName[attr]; - vs_output_semantic_index[slot] = outputSemanticIndex[attr]; - } - else { - /* use default semantic info */ - vs_output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs_output_semantic_index[slot] = num_generic++; - } + assert(attr < VERT_RESULT_MAX); + stvp->output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + stvp->output_semantic_index[slot] = (FRAG_ATTRIB_VAR0 - + FRAG_ATTRIB_TEX0 + + attr - + VERT_RESULT_VAR0); + break; } } } +} - if (outputMapping) { - /* find max output slot referenced to compute vs_num_outputs */ - GLuint maxSlot = 0; - for (attr = 0; attr < VERT_RESULT_MAX; attr++) { - if (outputMapping[attr] != ~0 && outputMapping[attr] > maxSlot) - maxSlot = outputMapping[attr]; - } - vs_num_outputs = maxSlot + 1; - } - else { - outputMapping = defaultOutputMapping; - } -#if 0 /* debug */ - { - GLuint i; - printf("outputMapping? %d\n", outputMapping ? 1 : 0); - if (outputMapping) { - printf("attr -> slot\n"); - for (i = 0; i < 16; i++) { - printf(" %2d %3d\n", i, outputMapping[i]); - } - } - printf("slot sem_name sem_index\n"); - for (i = 0; i < vs_num_outputs; i++) { - printf(" %2d %d %d\n", - i, - vs_output_semantic_name[i], - vs_output_semantic_index[i]); - } - } -#endif - - /* free old shader state, if any */ - if (stvp->state.tokens) { - st_free_tokens(stvp->state.tokens); - stvp->state.tokens = NULL; - } - if (stvp->driver_shader) { - cso_delete_vertex_shader(st->cso_context, stvp->driver_shader); - stvp->driver_shader = NULL; - } +struct st_vp_varient * +st_translate_vertex_program(struct st_context *st, + struct st_vertex_program *stvp, + const struct st_vp_varient_key *key) +{ + struct st_vp_varient *vpv = CALLOC_STRUCT(st_vp_varient); + struct pipe_context *pipe = st->pipe; - stvp->state.tokens = + vpv->state.tokens = st_translate_mesa_program(st->ctx, TGSI_PROCESSOR_VERTEX, &stvp->Base.Base, /* inputs */ - vs_num_inputs, + stvp->num_inputs, stvp->input_to_index, NULL, /* input semantic name */ NULL, /* input semantic index */ NULL, /* outputs */ - vs_num_outputs, - outputMapping, - vs_output_semantic_name, - vs_output_semantic_index ); + stvp->num_outputs, + stvp->result_to_output, + stvp->output_semantic_name, + stvp->output_semantic_index ); - stvp->num_inputs = vs_num_inputs; - stvp->driver_shader = pipe->create_vs_state(pipe, &stvp->state); + vpv->driver_shader = pipe->create_vs_state(pipe, &vpv->state); if ((ST_DEBUG & DEBUG_TGSI) && (ST_DEBUG & DEBUG_MESA)) { _mesa_print_program(&stvp->Base.Base); @@ -266,9 +214,11 @@ st_translate_vertex_program(struct st_context *st, } if (ST_DEBUG & DEBUG_TGSI) { - tgsi_dump( stvp->state.tokens, 0 ); + tgsi_dump( vpv->state.tokens, 0 ); debug_printf("\n"); } + + return vpv; } @@ -291,7 +241,6 @@ st_translate_fragment_program(struct st_context *st, GLuint attr; const GLbitfield inputsRead = stfp->Base.Base.InputsRead; GLuint vslot = 0; - GLuint num_generic = 0; uint fs_num_inputs = 0; @@ -341,14 +290,25 @@ st_translate_fragment_program(struct st_context *st, break; case FRAG_ATTRIB_FACE: stfp->input_semantic_name[slot] = TGSI_SEMANTIC_FACE; - stfp->input_semantic_index[slot] = num_generic++; + stfp->input_semantic_index[slot] = 0; interpMode[slot] = TGSI_INTERPOLATE_CONSTANT; break; - case FRAG_ATTRIB_PNTC: - stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - stfp->input_semantic_index[slot] = num_generic++; - interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; - break; + + /* In most cases, there is nothing special about these + * inputs, so adopt a convention to use the generic + * semantic name and the mesa FRAG_ATTRIB_ number as the + * index. + * + * All that is required is that the vertex shader labels + * its own outputs similarly, and that the vertex shader + * generates at least every output required by the + * fragment shader plus fixed-function hardware (such as + * BFC). + * + * There is no requirement that semantic indexes start at + * zero or be restricted to a particular range -- nobody + * should be building tables based on semantic index. + */ case FRAG_ATTRIB_TEX0: case FRAG_ATTRIB_TEX1: case FRAG_ATTRIB_TEX2: @@ -357,16 +317,17 @@ st_translate_fragment_program(struct st_context *st, case FRAG_ATTRIB_TEX5: case FRAG_ATTRIB_TEX6: case FRAG_ATTRIB_TEX7: - stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - stfp->input_semantic_index[slot] = num_generic++; - interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; - break; + case FRAG_ATTRIB_PNTC: case FRAG_ATTRIB_VAR0: - /* fall-through */ default: + /* Actually, let's try and zero-base this just for + * readability of the generated TGSI. + */ + assert(attr >= FRAG_ATTRIB_TEX0); + stfp->input_semantic_index[slot] = (attr - FRAG_ATTRIB_TEX0); stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - stfp->input_semantic_index[slot] = num_generic++; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; + break; } } } diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index e2e5eddef2..88aadbd751 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -64,41 +64,70 @@ struct st_fragment_program struct pipe_shader_state state; void *driver_shader; - GLuint param_state; - - /** List of vertex programs which have been translated such that their - * outputs match this fragment program's inputs. - */ - struct translated_vertex_program *vertex_programs; - /** Program prefixed with glBitmap prologue */ struct st_fragment_program *bitmap_program; uint bitmap_sampler; }; + +struct st_vp_varient_key +{ + char dummy; /* currently unused */ +}; + + +/** + * This represents a vertex program, especially translated to match + * the inputs of a particular fragment shader. + */ +struct st_vp_varient +{ + /* Parameters which generated this translated version of a vertex + * shader: + */ + struct st_vp_varient_key key; + + /** TGSI tokens -- why? + */ + struct pipe_shader_state state; + + /** Driver's compiled shader */ + void *driver_shader; + + /** For using our private draw module (glRasterPos) */ + struct draw_vertex_shader *draw_shader; + + /** Next in linked list */ + struct st_vp_varient *next; +}; + + + + /** * Derived from Mesa gl_fragment_program: */ struct st_vertex_program { struct gl_vertex_program Base; /**< The Mesa vertex program */ - GLuint serialNo; + GLuint serialNo, lastSerialNo; /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */ GLuint input_to_index[VERT_ATTRIB_MAX]; /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ GLuint index_to_input[PIPE_MAX_SHADER_INPUTS]; - GLuint num_inputs; - struct pipe_shader_state state; - void *driver_shader; + /** Maps VERT_RESULT_x to slot */ + GLuint result_to_output[VERT_RESULT_MAX]; + ubyte output_semantic_name[VERT_RESULT_MAX]; + ubyte output_semantic_index[VERT_RESULT_MAX]; + GLuint num_outputs; - /** For using our private draw module (glRasterPos) */ - struct draw_vertex_shader *draw_shader; - - GLuint param_state; + /** List of translated varients of this vertex program. + */ + struct st_vp_varient *varients; }; @@ -143,13 +172,21 @@ st_translate_fragment_program(struct st_context *st, const GLuint inputMapping[]); +/* Called after program string change, discard all previous + * compilation results. + */ extern void +st_prepare_vertex_program(struct st_context *st, + struct st_vertex_program *stvp); + +extern struct st_vp_varient * st_translate_vertex_program(struct st_context *st, - struct st_vertex_program *vp, - const GLuint vert_output_to_slot[], - const ubyte *fs_input_semantic_name, - const ubyte *fs_input_semantic_index); + struct st_vertex_program *stvp, + const struct st_vp_varient_key *key); +void +st_vp_release_varients( struct st_context *st, + struct st_vertex_program *stvp ); extern void st_print_shaders(GLcontext *ctx); -- cgit v1.2.3 From 133501bef2933395f14b2ebdfeda84279be93c60 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Nov 2009 16:16:16 -0700 Subject: mesa: fix assorted compiler warnings --- src/mesa/shader/program_parser.h | 2 +- src/mesa/shader/slang/slang_codegen.c | 2 +- src/mesa/state_tracker/st_atom.c | 8 ++++---- src/mesa/state_tracker/st_cb_bitmap.c | 2 +- src/mesa/state_tracker/st_program.c | 4 ++-- src/mesa/vbo/vbo_exec_array.c | 2 +- 6 files changed, 10 insertions(+), 10 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/mesa/shader/program_parser.h b/src/mesa/shader/program_parser.h index bce6041381..c170948f73 100644 --- a/src/mesa/shader/program_parser.h +++ b/src/mesa/shader/program_parser.h @@ -35,7 +35,7 @@ enum asm_type { at_attrib, at_param, at_temp, - at_output, + at_output }; struct asm_symbol { diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 344dfdc680..ee5a50ca82 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -925,7 +925,7 @@ gen_return_with_expression(slang_assemble_ctx *A, slang_operation *oper) slang_operation_copy(rhs, &oper->children[0]); } - ///blockOper->locals->outer_scope = oper->locals->outer_scope; + /*blockOper->locals->outer_scope = oper->locals->outer_scope;*/ /*slang_print_tree(blockOper, 0);*/ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index ca15ce1b47..0e89a624c4 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -137,7 +137,7 @@ void st_validate_state( struct st_context *st ) if (state->st == 0) return; -// _mesa_printf("%s %x/%x\n", __FUNCTION__, state->mesa, state->st); + /*_mesa_printf("%s %x/%x\n", __FUNCTION__, state->mesa, state->st);*/ if (1) { /* Debug version which enforces various sanity checks on the @@ -152,7 +152,7 @@ void st_validate_state( struct st_context *st ) const struct st_tracked_state *atom = atoms[i]; struct st_state_flags generated; -// _mesa_printf("atom %s %x/%x\n", atom->name, atom->dirty.mesa, atom->dirty.st); + /*_mesa_printf("atom %s %x/%x\n", atom->name, atom->dirty.mesa, atom->dirty.st);*/ if (!(atom->dirty.mesa || atom->dirty.st) || !atom->update) { @@ -162,7 +162,7 @@ void st_validate_state( struct st_context *st ) if (check_state(state, &atom->dirty)) { atoms[i]->update( st ); -// _mesa_printf("after: %x\n", atom->dirty.mesa); + /*_mesa_printf("after: %x\n", atom->dirty.mesa);*/ } accumulate_state(&examined, &atom->dirty); @@ -175,7 +175,7 @@ void st_validate_state( struct st_context *st ) assert(!check_state(&examined, &generated)); prev = *state; } -// _mesa_printf("\n"); + /*_mesa_printf("\n");*/ } else { diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index a22fa68299..1960d171bf 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -526,7 +526,7 @@ reset_cache(struct st_context *st) struct pipe_screen *screen = pipe->screen; struct bitmap_cache *cache = st->bitmap.cache; - //memset(cache->buffer, 0xff, sizeof(cache->buffer)); + /*memset(cache->buffer, 0xff, sizeof(cache->buffer));*/ cache->empty = GL_TRUE; cache->xmin = 1000000; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index a9be80ce8f..6d02722c13 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -80,7 +80,7 @@ st_translate_vertex_program(struct st_context *st, GLbitfield input_flags[MAX_PROGRAM_INPUTS]; GLbitfield output_flags[MAX_PROGRAM_OUTPUTS]; -// memset(&vs, 0, sizeof(vs)); + /*memset(&vs, 0, sizeof(vs));*/ memset(input_flags, 0, sizeof(input_flags)); memset(output_flags, 0, sizeof(output_flags)); @@ -386,7 +386,7 @@ st_translate_fragment_program(struct st_context *st, GLbitfield input_flags[MAX_PROGRAM_INPUTS]; GLbitfield output_flags[MAX_PROGRAM_OUTPUTS]; -// memset(&fs, 0, sizeof(fs)); + /*memset(&fs, 0, sizeof(fs));*/ memset(input_flags, 0, sizeof(input_flags)); memset(output_flags, 0, sizeof(output_flags)); diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c index fd70b57b72..6de8f059b7 100644 --- a/src/mesa/vbo/vbo_exec_array.c +++ b/src/mesa/vbo/vbo_exec_array.c @@ -146,7 +146,7 @@ check_array_data(GLcontext *ctx, struct gl_client_array *array, array->Ptr, array->BufferObj->Name); f[k] = 1.0; /* XXX replace the bad value! */ } - //assert(!IS_INF_OR_NAN(f[k])); + /*assert(!IS_INF_OR_NAN(f[k]));*/ } } break; -- cgit v1.2.3 From 5606dfb572bf4b89b4882265924705bacc8c182b Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 17 Nov 2009 16:10:24 -0800 Subject: Merge branch 'outputswritten64' Add a GLbitfield64 type and several macros to operate on 64-bit fields. The OutputsWritten field of gl_program is changed to use that type. This results in a fair amount of fallout in drivers that use programs. No changes are strictly necessary at this point as all bits used are below the 32-bit boundary. Fairly soon several bits will be added for clip distances written by a vertex shader. This will cause several bits used for varyings to be pushed above the 32-bit boundary. This will affect any drivers that support GLSL. At this point, only the i965 driver has been modified to support this eventuality. I did this as a "squash" merge. There were several places through the outputswritten64 branch where things were broken. I foresee this causing difficulties later for bisecting. The history is still available in the branch. Conflicts: src/mesa/drivers/dri/i965/brw_wm.h --- src/mesa/drivers/dri/i965/brw_clip.c | 2 +- src/mesa/drivers/dri/i965/brw_clip.h | 2 +- src/mesa/drivers/dri/i965/brw_context.h | 6 +----- src/mesa/drivers/dri/i965/brw_gs.h | 2 +- src/mesa/drivers/dri/i965/brw_sf.c | 6 +++--- src/mesa/drivers/dri/i965/brw_sf.h | 2 +- src/mesa/drivers/dri/i965/brw_sf_emit.c | 18 +++++++++--------- src/mesa/drivers/dri/i965/brw_util.c | 2 +- src/mesa/drivers/dri/i965/brw_util.h | 2 +- src/mesa/drivers/dri/i965/brw_vs.c | 2 +- src/mesa/drivers/dri/i965/brw_vs_emit.c | 8 ++++---- src/mesa/drivers/dri/i965/brw_wm.c | 4 ++-- src/mesa/drivers/dri/i965/brw_wm.h | 2 +- src/mesa/drivers/dri/i965/brw_wm_fp.c | 2 +- src/mesa/drivers/dri/i965/brw_wm_glsl.c | 2 +- src/mesa/drivers/dri/i965/brw_wm_pass2.c | 4 ++-- src/mesa/drivers/dri/i965/brw_wm_state.c | 2 +- src/mesa/drivers/dri/r200/r200_tcl.c | 15 ++++++++------- src/mesa/main/config.h | 2 +- src/mesa/main/context.c | 8 +++++--- src/mesa/main/ffvertex_prog.c | 2 +- src/mesa/main/mtypes.h | 27 ++++++++++++++++++++++++++- src/mesa/main/texenvprogram.c | 2 +- src/mesa/shader/prog_print.c | 6 +++--- src/mesa/shader/program_parse.tab.c | 2 +- src/mesa/shader/program_parse.y | 2 +- src/mesa/shader/programopt.c | 8 ++++---- src/mesa/shader/slang/slang_link.c | 25 +++++++++++++++---------- src/mesa/state_tracker/st_atom_shader.c | 2 +- src/mesa/state_tracker/st_program.c | 2 +- src/mesa/swrast/s_fragprog.c | 12 ++++++------ src/mesa/tnl/t_context.c | 2 +- src/mesa/tnl/t_vb_program.c | 8 ++++---- 33 files changed, 111 insertions(+), 82 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c index f45dcf8282..dbd10a5297 100644 --- a/src/mesa/drivers/dri/i965/brw_clip.c +++ b/src/mesa/drivers/dri/i965/brw_clip.c @@ -78,7 +78,7 @@ static void compile_clip_prog( struct brw_context *brw, delta = REG_SIZE; for (i = 0; i < VERT_RESULT_MAX; i++) - if (c.key.attrs & (1<= VERT_RESULT_TEX0 && i <= VERT_RESULT_TEX7) { @@ -147,7 +147,7 @@ static void upload_sf_prog(struct brw_context *brw) * edgeflag testing here, it is already done in the clip * program. */ - if (key.attrs & (1<key.attrs & (1<key.attrs & BITFIELD64_BIT(attr)) ? 1 : 0; } /*********************************************************************** @@ -122,8 +122,8 @@ static void do_twoside_color( struct brw_sf_compile *c ) * Flat shading */ -#define VERT_RESULT_COLOR_BITS ((1<nr_setup_regs - 1); - GLuint persp_mask; - GLuint linear_mask; + GLbitfield64 persp_mask; + GLbitfield64 linear_mask; if (c->key.do_flat_shading || c->key.linear_color) persp_mask = c->key.attrs & ~(FRAG_BIT_WPOS | @@ -331,10 +331,10 @@ static GLboolean calculate_masks( struct brw_sf_compile *c, *pc_linear = 0; *pc = 0xf; - if (persp_mask & (1 << c->idx_to_attr[reg*2])) + if (persp_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2])) *pc_persp = 0xf; - if (linear_mask & (1 << c->idx_to_attr[reg*2])) + if (linear_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2])) *pc_linear = 0xf; /* Maybe only processs one attribute on the final round: @@ -342,10 +342,10 @@ static GLboolean calculate_masks( struct brw_sf_compile *c, if (reg*2+1 < c->nr_setup_attrs) { *pc |= 0xf0; - if (persp_mask & (1 << c->idx_to_attr[reg*2+1])) + if (persp_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2+1])) *pc_persp |= 0xf0; - if (linear_mask & (1 << c->idx_to_attr[reg*2+1])) + if (linear_mask & BITFIELD64_BIT(c->idx_to_attr[reg*2+1])) *pc_linear |= 0xf0; } diff --git a/src/mesa/drivers/dri/i965/brw_util.c b/src/mesa/drivers/dri/i965/brw_util.c index ce21aa4869..bba9249d1b 100644 --- a/src/mesa/drivers/dri/i965/brw_util.c +++ b/src/mesa/drivers/dri/i965/brw_util.c @@ -35,7 +35,7 @@ #include "brw_util.h" #include "brw_defines.h" -GLuint brw_count_bits( GLuint val ) +GLuint brw_count_bits(uint64_t val) { GLuint i; for (i = 0; val ; val >>= 1) diff --git a/src/mesa/drivers/dri/i965/brw_util.h b/src/mesa/drivers/dri/i965/brw_util.h index 33e7cd87e4..04f3175d3e 100644 --- a/src/mesa/drivers/dri/i965/brw_util.h +++ b/src/mesa/drivers/dri/i965/brw_util.h @@ -35,7 +35,7 @@ #include "main/mtypes.h" -extern GLuint brw_count_bits( GLuint val ); +extern GLuint brw_count_bits(uint64_t val); extern GLuint brw_parameter_list_state_flags(struct gl_program_parameter_list *paramList); extern GLuint brw_translate_blend_factor( GLenum factor ); extern GLuint brw_translate_blend_equation( GLenum mode ); diff --git a/src/mesa/drivers/dri/i965/brw_vs.c b/src/mesa/drivers/dri/i965/brw_vs.c index f0c79efbd9..fd055e225e 100644 --- a/src/mesa/drivers/dri/i965/brw_vs.c +++ b/src/mesa/drivers/dri/i965/brw_vs.c @@ -56,7 +56,7 @@ static void do_vs_prog( struct brw_context *brw, c.prog_data.inputs_read = vp->program.Base.InputsRead; if (c.key.copy_edgeflag) { - c.prog_data.outputs_written |= 1<prog_data.outputs_written & (1 << i)) { + if (c->prog_data.outputs_written & BITFIELD64_BIT(i)) { c->nr_outputs++; assert(i < Elements(c->regs[PROGRAM_OUTPUT])); if (i == VERT_RESULT_HPOS) { @@ -1124,7 +1124,7 @@ static void emit_vertex_write( struct brw_vs_compile *c) /* Update the header for point size, user clipping flags, and -ve rhw * workaround. */ - if ((c->prog_data.outputs_written & (1<prog_data.outputs_written & BITFIELD64_BIT(VERT_RESULT_PSIZ)) || c->key.nr_userclip || BRW_IS_965(p->brw)) { struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); @@ -1134,7 +1134,7 @@ static void emit_vertex_write( struct brw_vs_compile *c) brw_set_access_mode(p, BRW_ALIGN_16); - if (c->prog_data.outputs_written & (1<prog_data.outputs_written & BITFIELD64_BIT(VERT_RESULT_PSIZ)) { struct brw_reg psiz = c->regs[PROGRAM_OUTPUT][VERT_RESULT_PSIZ]; brw_MUL(p, brw_writemask(header1, WRITEMASK_W), brw_swizzle1(psiz, 0), brw_imm_f(1<<11)); brw_AND(p, brw_writemask(header1, WRITEMASK_W), header1, brw_imm_ud(0x7ff<<8)); @@ -1224,7 +1224,7 @@ static void emit_vertex_write( struct brw_vs_compile *c) */ GLuint i, mrf = 0; for (i = c->first_overflow_output; i < VERT_RESULT_MAX; i++) { - if (c->prog_data.outputs_written & (1 << i)) { + if (c->prog_data.outputs_written & BITFIELD64_BIT(i)) { /* move from GRF to MRF */ brw_MOV(p, brw_message_reg(4+mrf), c->regs[PROGRAM_OUTPUT][i]); mrf++; diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c index d8971321f3..77e3b2c32a 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.c +++ b/src/mesa/drivers/dri/i965/brw_wm.c @@ -231,7 +231,7 @@ static void brw_wm_populate_key( struct brw_context *brw, ctx->Color.AlphaEnabled) lookup |= IZ_PS_KILL_ALPHATEST_BIT; - if (fp->program.Base.OutputsWritten & (1<program.Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) lookup |= IZ_PS_COMPUTES_DEPTH_BIT; /* _NEW_DEPTH */ @@ -347,7 +347,7 @@ static void brw_wm_populate_key( struct brw_context *brw, key->nr_color_regions = brw->state.nr_color_regions; /* CACHE_NEW_VS_PROG */ - key->vp_outputs_written = brw->vs.prog_data->outputs_written & DO_SETUP_BITS; + key->vp_outputs_written = brw->vs.prog_data->outputs_written; /* The unique fragment program ID */ key->program_string_id = fp->id; diff --git a/src/mesa/drivers/dri/i965/brw_wm.h b/src/mesa/drivers/dri/i965/brw_wm.h index b3c05eb0ad..9dcb6e14bb 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.h +++ b/src/mesa/drivers/dri/i965/brw_wm.h @@ -79,7 +79,7 @@ struct brw_wm_prog_key { GLuint program_string_id:32; GLushort origin_x, origin_y; GLushort drawable_height; - GLuint vp_outputs_written; + GLbitfield64 vp_outputs_written; }; diff --git a/src/mesa/drivers/dri/i965/brw_wm_fp.c b/src/mesa/drivers/dri/i965/brw_wm_fp.c index 1c4f62ba48..7d03179588 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_fp.c +++ b/src/mesa/drivers/dri/i965/brw_wm_fp.c @@ -986,7 +986,7 @@ static void emit_render_target_writes( struct brw_wm_compile *c ) } else { /* if gl_FragData[0] is written, use it, else use gl_FragColor */ - if (c->fp->program.Base.OutputsWritten & (1 << FRAG_RESULT_DATA0)) + if (c->fp->program.Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DATA0)) outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_DATA0); else outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLOR); diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c index 3ab446164c..e8c2cb66ec 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c +++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c @@ -371,7 +371,7 @@ static void prealloc_reg(struct brw_wm_compile *c) for (j = 0; j < 4; j++) set_reg(c, PROGRAM_PAYLOAD, fp_input, j, reg); } - if (c->key.vp_outputs_written & (1 << i)) { + if (c->key.vp_outputs_written & BITFIELD64_BIT(i)) { reg_index += 2; } } diff --git a/src/mesa/drivers/dri/i965/brw_wm_pass2.c b/src/mesa/drivers/dri/i965/brw_wm_pass2.c index 6faea018fb..31303febf0 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_pass2.c +++ b/src/mesa/drivers/dri/i965/brw_wm_pass2.c @@ -82,8 +82,8 @@ static void init_registers( struct brw_wm_compile *c ) for (j = 0; j < c->nr_creg; j++) prealloc_reg(c, &c->creg[j], i++); - for (j = 0; j < FRAG_ATTRIB_MAX; j++) { - if (c->key.vp_outputs_written & (1<key.vp_outputs_written & BITFIELD64_BIT(j)) { int fp_index; if (j >= VERT_RESULT_VAR0) diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c index 361f91292b..f89ed9bce7 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_state.c @@ -106,7 +106,7 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key) /* as far as we can tell */ key->computes_depth = - (fp->Base.OutputsWritten & (1 << FRAG_RESULT_DEPTH)) != 0; + (fp->Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) != 0; /* BRW_NEW_DEPTH_BUFFER * Override for NULL depthbuffer case, required by the Pixel Shader Computed * Depth field. diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c index c702910ef2..e7d48a7f29 100644 --- a/src/mesa/drivers/dri/r200/r200_tcl.c +++ b/src/mesa/drivers/dri/r200/r200_tcl.c @@ -509,25 +509,26 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx, prog to a not enabled output however, so just don't mess with it. We only need to change compsel. */ GLuint out_compsel = 0; - GLuint vp_out = rmesa->curr_vp_hw->mesa_program.Base.OutputsWritten; + const GLbitfield64 vp_out = + rmesa->curr_vp_hw->mesa_program.Base.OutputsWritten; vimap_rev = &rmesa->curr_vp_hw->inputmap_rev[0]; - assert(vp_out & (1 << VERT_RESULT_HPOS)); + assert(vp_out & BITFIELD64_BIT(VERT_RESULT_HPOS)); out_compsel = R200_OUTPUT_XYZW; - if (vp_out & (1 << VERT_RESULT_COL0)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_COL0)) { out_compsel |= R200_OUTPUT_COLOR_0; } - if (vp_out & (1 << VERT_RESULT_COL1)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_COL1)) { out_compsel |= R200_OUTPUT_COLOR_1; } - if (vp_out & (1 << VERT_RESULT_FOGC)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_FOGC)) { out_compsel |= R200_OUTPUT_DISCRETE_FOG; } - if (vp_out & (1 << VERT_RESULT_PSIZ)) { + if (vp_out & BITFIELD64_BIT(VERT_RESULT_PSIZ)) { out_compsel |= R200_OUTPUT_PT_SIZE; } for (i = VERT_RESULT_TEX0; i < VERT_RESULT_TEX6; i++) { - if (vp_out & (1 << i)) { + if (vp_out & BITFIELD64_BIT(i)) { out_compsel |= R200_OUTPUT_TEX_0 << (i - VERT_RESULT_TEX0); } } diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h index 8a09efdb53..c5048970cc 100644 --- a/src/mesa/main/config.h +++ b/src/mesa/main/config.h @@ -205,7 +205,7 @@ #define MAX_VARYING 16 /**< number of float[4] vectors */ #define MAX_SAMPLERS MAX_TEXTURE_IMAGE_UNITS #define MAX_PROGRAM_INPUTS 32 -#define MAX_PROGRAM_OUTPUTS 32 +#define MAX_PROGRAM_OUTPUTS 64 /*@}*/ /** For GL_ARB_vertex_program */ diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 101d3c6b67..b5bf46718f 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -601,9 +601,11 @@ _mesa_init_constants(GLcontext *ctx) ASSERT(MAX_NV_VERTEX_PROGRAM_INPUTS <= VERT_ATTRIB_MAX); ASSERT(MAX_NV_VERTEX_PROGRAM_OUTPUTS <= VERT_RESULT_MAX); - /* check that we don't exceed various 32-bit bitfields */ - ASSERT(VERT_RESULT_MAX <= 32); - ASSERT(FRAG_ATTRIB_MAX <= 32); + /* check that we don't exceed the size of various bitfields */ + ASSERT(VERT_RESULT_MAX <= + (8 * sizeof(ctx->VertexProgram._Current->Base.OutputsWritten))); + ASSERT(FRAG_ATTRIB_MAX <= + (8 * sizeof(ctx->FragmentProgram._Current->Base.InputsRead))); } diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 356476e35a..fe2416d894 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -438,7 +438,7 @@ static struct ureg register_input( struct tnl_program *p, GLuint input ) */ static struct ureg register_output( struct tnl_program *p, GLuint output ) { - p->program->Base.OutputsWritten |= (1<program->Base.OutputsWritten |= BITFIELD64_BIT(output); return make_ureg(PROGRAM_OUTPUT, output); } diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 34c51b5442..881d233ca3 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -78,6 +78,31 @@ #endif +/** + * \name 64-bit extension of GLbitfield. + */ +/*@{*/ +typedef GLuint64 GLbitfield64; + +#define BITFIELD64_ONE 1ULL +#define BITFIELD64_ALLONES ~0ULL + +/** Set a single bit */ +#define BITFIELD64_BIT(b) (BITFIELD64_ONE << (b)) + +/** Set a mask of the least significant \c b bits */ +#define BITFIELD64_MASK(b) (((b) >= 64) ? BITFIELD64_ALLONES : \ + (BITFIELD64_BIT(b) - 1)) + +/** + * Set all bits from l (low bit) to h (high bit), inclusive. + * + * \note \C BITFIELD_64_RANGE(0, 63) return 64 set bits. + */ +#define BITFIELD64_RANGE(l, h) (BITFIELD64_MASK((h) + 1) & ~BITFIELD64_MASK(l)) +/*@}*/ + + /** * \name Some forward type declarations */ @@ -1670,7 +1695,7 @@ struct gl_program struct prog_instruction *Instructions; GLbitfield InputsRead; /**< Bitmask of which input regs are read */ - GLbitfield OutputsWritten; /**< Bitmask of which output regs are written to */ + GLbitfield64 OutputsWritten; /**< Bitmask of which output regs are written */ GLbitfield InputFlags[MAX_PROGRAM_INPUTS]; /**< PROG_PARAM_BIT_x flags */ GLbitfield OutputFlags[MAX_PROGRAM_OUTPUTS]; /**< PROG_PARAM_BIT_x flags */ GLbitfield TexturesUsed[MAX_TEXTURE_UNITS]; /**< TEXTURE_x_BIT bitmask */ diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index d7e77e759e..f439d4addb 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -367,7 +367,7 @@ static GLbitfield get_fp_input_mask( GLcontext *ctx ) else { /* calculate from vp->outputs */ struct gl_vertex_program *vprog; - GLbitfield vp_outputs; + GLbitfield64 vp_outputs; /* Choose GLSL vertex shader over ARB vertex program. Need this * since vertex shader state validation comes after fragment state diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index ba4d39452f..52c102cbaa 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -826,11 +826,11 @@ _mesa_print_program(const struct gl_program *prog) * XXX move to imports.[ch] if useful elsewhere. */ static const char * -binary(GLbitfield val) +binary(GLbitfield64 val) { - static char buf[50]; + static char buf[80]; GLint i, len = 0; - for (i = 31; i >= 0; --i) { + for (i = 63; i >= 0; --i) { if (val & (1 << i)) buf[len++] = '1'; else if (len > 0 || i == 0) diff --git a/src/mesa/shader/program_parse.tab.c b/src/mesa/shader/program_parse.tab.c index b9ef88b64b..d4f8429488 100644 --- a/src/mesa/shader/program_parse.tab.c +++ b/src/mesa/shader/program_parse.tab.c @@ -2622,7 +2622,7 @@ yyreduce: YYERROR; } - state->prog->OutputsWritten |= (1U << (yyval.dst_reg).Index); + state->prog->OutputsWritten |= BITFIELD64_BIT((yyval.dst_reg).Index); } ;} break; diff --git a/src/mesa/shader/program_parse.y b/src/mesa/shader/program_parse.y index d07bf85b36..8ca6f9805b 100644 --- a/src/mesa/shader/program_parse.y +++ b/src/mesa/shader/program_parse.y @@ -643,7 +643,7 @@ maskedDstReg: dstReg optionalMask optionalCcMask YYERROR; } - state->prog->OutputsWritten |= (1U << $$.Index); + state->prog->OutputsWritten |= BITFIELD64_BIT($$.Index); } } ; diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c index 3b8529592d..a0daac1b80 100644 --- a/src/mesa/shader/programopt.c +++ b/src/mesa/shader/programopt.c @@ -109,7 +109,7 @@ _mesa_insert_mvp_dp4_code(GLcontext *ctx, struct gl_vertex_program *vprog) vprog->Base.Instructions = newInst; vprog->Base.NumInstructions = newLen; vprog->Base.InputsRead |= VERT_BIT_POS; - vprog->Base.OutputsWritten |= (1 << VERT_RESULT_HPOS); + vprog->Base.OutputsWritten |= BITFIELD64_BIT(VERT_RESULT_HPOS); } @@ -211,7 +211,7 @@ _mesa_insert_mvp_mad_code(GLcontext *ctx, struct gl_vertex_program *vprog) vprog->Base.Instructions = newInst; vprog->Base.NumInstructions = newLen; vprog->Base.InputsRead |= VERT_BIT_POS; - vprog->Base.OutputsWritten |= (1 << VERT_RESULT_HPOS); + vprog->Base.OutputsWritten |= BITFIELD64_BIT(VERT_RESULT_HPOS); } @@ -613,7 +613,7 @@ _mesa_nop_fragment_program(GLcontext *ctx, struct gl_fragment_program *prog) prog->Base.Instructions = inst; prog->Base.NumInstructions = 2; prog->Base.InputsRead = 1 << inputAttr; - prog->Base.OutputsWritten = 1 << FRAG_RESULT_COLOR; + prog->Base.OutputsWritten = BITFIELD64_BIT(FRAG_RESULT_COLOR); } @@ -657,7 +657,7 @@ _mesa_nop_vertex_program(GLcontext *ctx, struct gl_vertex_program *prog) prog->Base.Instructions = inst; prog->Base.NumInstructions = 2; prog->Base.InputsRead = 1 << inputAttr; - prog->Base.OutputsWritten = 1 << VERT_RESULT_COL0; + prog->Base.OutputsWritten = BITFIELD64_BIT(VERT_RESULT_COL0); /* * Now insert code to do standard modelview/projection transformation. diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index 144c126525..0a2bc49780 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -515,7 +515,7 @@ _slang_update_inputs_outputs(struct gl_program *prog) } if (inst->DstReg.File == PROGRAM_OUTPUT) { - prog->OutputsWritten |= 1 << inst->DstReg.Index; + prog->OutputsWritten |= BITFIELD64_BIT(inst->DstReg.Index); if (inst->DstReg.RelAddr) { /* If the output attribute is indexed with relative addressing * we know that it must be a varying or texcoord such as @@ -528,14 +528,17 @@ _slang_update_inputs_outputs(struct gl_program *prog) if (prog->Target == GL_VERTEX_PROGRAM_ARB) { if (inst->DstReg.Index == VERT_RESULT_TEX0) { /* mark all texcoord outputs as written */ - const GLbitfield mask = - ((1 << MAX_TEXTURE_COORD_UNITS) - 1) << VERT_RESULT_TEX0; + const GLbitfield64 mask = + BITFIELD64_RANGE(VERT_RESULT_TEX0, + (VERT_RESULT_TEX0 + + MAX_TEXTURE_COORD_UNITS - 1)); prog->OutputsWritten |= mask; } else if (inst->DstReg.Index == VERT_RESULT_VAR0) { /* mark all generic varying outputs as written */ - const GLbitfield mask = - ((1 << MAX_VARYING) - 1) << VERT_RESULT_VAR0; + const GLbitfield64 mask = + BITFIELD64_RANGE(VERT_RESULT_VAR0, + (VERT_RESULT_VAR0 + MAX_VARYING - 1)); prog->OutputsWritten |= mask; } } @@ -807,7 +810,8 @@ _slang_link(GLcontext *ctx, if (shProg->VertexProgram) { _slang_update_inputs_outputs(&shProg->VertexProgram->Base); _slang_count_temporaries(&shProg->VertexProgram->Base); - if (!(shProg->VertexProgram->Base.OutputsWritten & (1 << VERT_RESULT_HPOS))) { + if (!(shProg->VertexProgram->Base.OutputsWritten + & BITFIELD64_BIT(VERT_RESULT_HPOS))) { /* the vertex program did not compute a vertex position */ link_error(shProg, "gl_Position was not written by vertex shader\n"); @@ -825,7 +829,7 @@ _slang_link(GLcontext *ctx, if (shProg->FragmentProgram) { const GLbitfield varyingRead = shProg->FragmentProgram->Base.InputsRead >> FRAG_ATTRIB_VAR0; - const GLbitfield varyingWritten = shProg->VertexProgram ? + const GLbitfield64 varyingWritten = shProg->VertexProgram ? shProg->VertexProgram->Base.OutputsWritten >> VERT_RESULT_VAR0 : 0x0; if ((varyingRead & varyingWritten) != varyingRead) { link_error(shProg, @@ -836,9 +840,10 @@ _slang_link(GLcontext *ctx, /* check that gl_FragColor and gl_FragData are not both written to */ if (shProg->FragmentProgram) { - GLbitfield outputsWritten = shProg->FragmentProgram->Base.OutputsWritten; - if ((outputsWritten & ((1 << FRAG_RESULT_COLOR))) && - (outputsWritten >= (1 << FRAG_RESULT_DATA0))) { + const GLbitfield64 outputsWritten = + shProg->FragmentProgram->Base.OutputsWritten; + if ((outputsWritten & BITFIELD64_BIT(FRAG_RESULT_COLOR)) && + (outputsWritten >= BITFIELD64_BIT(FRAG_RESULT_DATA0))) { link_error(shProg, "Fragment program cannot write both gl_FragColor" " and gl_FragData[].\n"); return; diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index ee649be885..6e311e537e 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -176,7 +176,7 @@ find_translated_vp(struct st_context *st, /* See if we need to translate vertex program to TGSI form */ if (xvp->serialNo != stvp->serialNo) { GLuint outAttr; - const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten; + const GLbitfield64 outputsWritten = stvp->Base.Base.OutputsWritten; GLuint numVpOuts = 0; GLboolean emitPntSize = GL_FALSE, emitBFC0 = GL_FALSE, emitBFC1 = GL_FALSE; GLbitfield usedGenerics = 0x0; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 6d02722c13..190b6a5526 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -469,7 +469,7 @@ st_translate_fragment_program(struct st_context *st, */ { uint numColors = 0; - GLbitfield outputsWritten = stfp->Base.Base.OutputsWritten; + GLbitfield64 outputsWritten = stfp->Base.Base.OutputsWritten; /* if z is written, emit that first */ if (outputsWritten & (1 << FRAG_RESULT_DEPTH)) { diff --git a/src/mesa/swrast/s_fragprog.c b/src/mesa/swrast/s_fragprog.c index 77a77f0bcb..a22d34415d 100644 --- a/src/mesa/swrast/s_fragprog.c +++ b/src/mesa/swrast/s_fragprog.c @@ -190,7 +190,7 @@ run_program(GLcontext *ctx, SWspan *span, GLuint start, GLuint end) { SWcontext *swrast = SWRAST_CONTEXT(ctx); const struct gl_fragment_program *program = ctx->FragmentProgram._Current; - const GLbitfield outputsWritten = program->Base.OutputsWritten; + const GLbitfield64 outputsWritten = program->Base.OutputsWritten; struct gl_program_machine *machine = &swrast->FragProgMachine; GLuint i; @@ -201,7 +201,7 @@ run_program(GLcontext *ctx, SWspan *span, GLuint start, GLuint end) if (_mesa_execute_program(ctx, &program->Base, machine)) { /* Store result color */ - if (outputsWritten & (1 << FRAG_RESULT_COLOR)) { + if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_COLOR)) { COPY_4V(span->array->attribs[FRAG_ATTRIB_COL0][i], machine->Outputs[FRAG_RESULT_COLOR]); } @@ -212,7 +212,7 @@ run_program(GLcontext *ctx, SWspan *span, GLuint start, GLuint end) */ GLuint buf; for (buf = 0; buf < ctx->DrawBuffer->_NumColorDrawBuffers; buf++) { - if (outputsWritten & (1 << (FRAG_RESULT_DATA0 + buf))) { + if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_DATA0 + buf)) { COPY_4V(span->array->attribs[FRAG_ATTRIB_COL0 + buf][i], machine->Outputs[FRAG_RESULT_DATA0 + buf]); } @@ -220,7 +220,7 @@ run_program(GLcontext *ctx, SWspan *span, GLuint start, GLuint end) } /* Store result depth/z */ - if (outputsWritten & (1 << FRAG_RESULT_DEPTH)) { + if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) { const GLfloat depth = machine->Outputs[FRAG_RESULT_DEPTH][2]; if (depth <= 0.0) span->array->z[i] = 0; @@ -256,12 +256,12 @@ _swrast_exec_fragment_program( GLcontext *ctx, SWspan *span ) run_program(ctx, span, 0, span->end); - if (program->Base.OutputsWritten & (1 << FRAG_RESULT_COLOR)) { + if (program->Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_COLOR)) { span->interpMask &= ~SPAN_RGBA; span->arrayMask |= SPAN_RGBA; } - if (program->Base.OutputsWritten & (1 << FRAG_RESULT_DEPTH)) { + if (program->Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) { span->interpMask &= ~SPAN_Z; span->arrayMask |= SPAN_Z; } diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c index f2771cde09..db21b4589d 100644 --- a/src/mesa/tnl/t_context.c +++ b/src/mesa/tnl/t_context.c @@ -171,7 +171,7 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state ) if (vp) { GLuint i; for (i = 0; i < MAX_VARYING; i++) { - if (vp->Base.OutputsWritten & (1 << (VERT_RESULT_VAR0 + i))) { + if (vp->Base.OutputsWritten & BITFIELD64_BIT(VERT_RESULT_VAR0 + i)) { RENDERINPUTS_SET(tnl->render_inputs_bitset, _TNL_ATTRIB_GENERIC(i)); } diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c index c10a27614f..e69f7d5766 100644 --- a/src/mesa/tnl/t_vb_program.c +++ b/src/mesa/tnl/t_vb_program.c @@ -329,7 +329,7 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage ) /* make list of outputs to save some time below */ numOutputs = 0; for (i = 0; i < VERT_RESULT_MAX; i++) { - if (program->Base.OutputsWritten & (1 << i)) { + if (program->Base.OutputsWritten & BITFIELD64_BIT(i)) { outputs[numOutputs++] = i; } } @@ -407,14 +407,14 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage ) /* Fixup fog and point size results if needed */ if (program->IsNVProgram) { if (ctx->Fog.Enabled && - (program->Base.OutputsWritten & (1 << VERT_RESULT_FOGC)) == 0) { + (program->Base.OutputsWritten & BITFIELD64_BIT(VERT_RESULT_FOGC)) == 0) { for (i = 0; i < VB->Count; i++) { store->results[VERT_RESULT_FOGC].data[i][0] = 1.0; } } if (ctx->VertexProgram.PointSizeEnabled && - (program->Base.OutputsWritten & (1 << VERT_RESULT_PSIZ)) == 0) { + (program->Base.OutputsWritten & BITFIELD64_BIT(VERT_RESULT_PSIZ)) == 0) { for (i = 0; i < VB->Count; i++) { store->results[VERT_RESULT_PSIZ].data[i][0] = ctx->Point.Size; } @@ -472,7 +472,7 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage ) } for (i = 0; i < ctx->Const.MaxVarying; i++) { - if (program->Base.OutputsWritten & (1 << (VERT_RESULT_VAR0 + i))) { + if (program->Base.OutputsWritten & BITFIELD64_BIT(VERT_RESULT_VAR0 + i)) { /* Note: varying results get put into the generic attributes */ VB->AttribPtr[VERT_ATTRIB_GENERIC0+i] = &store->results[VERT_RESULT_VAR0 + i]; -- cgit v1.2.3 From 8468234bfa98be77cbceecc8e91325c00e4e424b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Dec 2009 08:59:38 -0700 Subject: st/mesa: fix up comment --- src/mesa/state_tracker/st_program.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index d66f45d13e..5c81a033f9 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -49,8 +49,11 @@ #include "st_mesa_to_tgsi.h" #include "cso_cache/cso_context.h" - /* Clean out any old compilations: - */ + + +/** + * Clean out any old compilations: + */ void st_vp_release_varients( struct st_context *st, struct st_vertex_program *stvp ) -- cgit v1.2.3 From 325f045c04121b989bddfca02711e3d03f7f9c3f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Dec 2009 18:39:44 +0000 Subject: st/mesa: move assert on nr vs insns until after pos_invarient expansion It is possible to have a 1-instruction vertex shader before expanding pos_invarient. Not sure what this assert achieved, but at least move it where it is correct. --- src/mesa/state_tracker/st_atom_shader.c | 2 -- src/mesa/state_tracker/st_program.c | 2 ++ 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 6e311e537e..9208f3f446 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -287,8 +287,6 @@ find_translated_vp(struct st_context *st, #endif } - assert(stvp->Base.Base.NumInstructions > 1); - st_translate_vertex_program(st, stvp, xvp->output_to_slot, xvp->output_to_semantic_name, xvp->output_to_semantic_index); diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 190b6a5526..cf19f8f1a8 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -87,6 +87,8 @@ st_translate_vertex_program(struct st_context *st, if (stvp->Base.IsPositionInvariant) _mesa_insert_mvp_code(st->ctx, &stvp->Base); + assert(stvp->Base.Base.NumInstructions > 1); + /* * Determine number of inputs, the mappings between VERT_ATTRIB_x * and TGSI generic input indexes, plus input attrib semantic info. -- cgit v1.2.3 From a08e348a84f57ed5e8bf5888f1ce13934d2ce8fa Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 9 Dec 2009 19:03:10 +0100 Subject: gallium: first steps to treat edgeflags as regular vertex element The idea here is to eliminate the set_edgeflags() call in pipe_context by treating edgeflags as a regular vertex element. Edgeflags provoke special treatment in hardware, which means we need to label them in some way, in this case we'll be passing them through the vertex shader and labelling the vertex shader output with a new TGSI semantic (TGSI_SEMANTIC_EDGEFLAG). --- src/gallium/auxiliary/draw/draw_context.c | 7 --- src/gallium/auxiliary/draw/draw_context.h | 3 - src/gallium/auxiliary/draw/draw_private.h | 2 - src/gallium/auxiliary/draw/draw_pt.c | 4 +- src/gallium/auxiliary/draw/draw_pt.h | 5 -- src/gallium/auxiliary/draw/draw_pt_fetch.c | 32 ++++------- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 3 +- src/gallium/auxiliary/draw/draw_pt_post_vs.c | 53 +++++++++++++++--- src/gallium/include/pipe/p_context.h | 8 --- src/gallium/include/pipe/p_shader_tokens.h | 3 +- src/mesa/state_tracker/st_atom_shader.c | 11 +++- src/mesa/state_tracker/st_draw.c | 64 ---------------------- src/mesa/state_tracker/st_mesa_to_tgsi.c | 13 ++--- src/mesa/state_tracker/st_program.c | 32 ++++++++++- src/mesa/state_tracker/st_program.h | 2 +- 15 files changed, 111 insertions(+), 131 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index a4f1fcddc1..cc5f7f0105 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -366,13 +366,6 @@ void draw_set_render( struct draw_context *draw, draw->render = render; } -void draw_set_edgeflags( struct draw_context *draw, - const unsigned *edgeflag ) -{ - draw->pt.user.edgeflag = edgeflag; -} - - /** diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index d529e4e9a2..465b8f10c6 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -143,9 +143,6 @@ void draw_set_mapped_constant_buffer(struct draw_context *draw, const void *buffer, unsigned size ); -void draw_set_edgeflags( struct draw_context *draw, - const unsigned *edgeflag ); - /*********************************************************************** * draw_prim.c diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 41fcb16a0a..0750e6e379 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -142,8 +142,6 @@ struct draw_context /* user-space vertex data, buffers */ struct { - const unsigned *edgeflag; - /** vertex element/index buffer (ex: glDrawElements) */ const void *elts; /** bytes per index (0, 1, 2 or 4) */ diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 4865a2d854..139ae1fe55 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -318,8 +318,10 @@ draw_arrays(struct draw_context *draw, unsigned prim, boolean draw_pt_get_edgeflag( struct draw_context *draw, unsigned idx ) { - if (draw->pt.user.edgeflag) + if (draw->pt.user.edgeflag) { + float *ef = draw->pt.verted_buffer[idx] return (draw->pt.user.edgeflag[idx/32] & (1 << (idx%32))) != 0; + } else return 1; } diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 7a17a9fb6b..b5c8c82f4a 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -149,11 +149,6 @@ struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw ); struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *draw); -/* More helpers: - */ -boolean draw_pt_get_edgeflag( struct draw_context *draw, - unsigned idx ); - /******************************************************************************* * HW vertex emit: diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 65c3a34c34..cb609f8c41 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -120,7 +120,12 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, fetch->translate = translate_cache_find(fetch->cache, &key); { - static struct vertex_header vh = { 0, 1, 0, UNDEFINED_VERTEX_ID, { .0f, .0f, .0f, .0f } }; + static struct vertex_header vh = { 0, + 1, + 0, + UNDEFINED_VERTEX_ID, + { .0f, .0f, .0f, .0f } }; + fetch->translate->set_buffer(fetch->translate, draw->pt.nr_vertex_buffers, &vh, @@ -128,9 +133,6 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, } } - fetch->need_edgeflags = ((draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) && - draw->pt.user.edgeflag); } @@ -158,16 +160,10 @@ void draw_pt_fetch_run( struct pt_fetch *fetch, count, verts ); - /* Edgeflags are hard to fit into a translate program, populate - * them separately if required. In the setup above they are - * defaulted to one, so only need this if there is reason to change - * that default: + /* Extract edgeflag values from vertex data into the header. */ if (fetch->need_edgeflags) { - for (i = 0; i < count; i++) { - struct vertex_header *vh = (struct vertex_header *)(verts + i * fetch->vertex_size); - vh->edgeflag = draw_pt_get_edgeflag( draw, elts[i] ); - } + extract_edge_flags( fetch, count ); } } @@ -194,16 +190,12 @@ void draw_pt_fetch_run_linear( struct pt_fetch *fetch, count, verts ); - /* Edgeflags are hard to fit into a translate program, populate - * them separately if required. In the setup above they are - * defaulted to one, so only need this if there is reason to change - * that default: + /* Extract edgeflag values from vertex data into the header. XXX: + * this should be done after the vertex shader is run. + * Bypass-vs-and-clip interaction with pipeline??? */ if (fetch->need_edgeflags) { - for (i = 0; i < count; i++) { - struct vertex_header *vh = (struct vertex_header *)(verts + i * fetch->vertex_size); - vh->edgeflag = draw_pt_get_edgeflag( draw, start + i ); - } + extract_edge_flags( fetch, count ); } } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index df6c265b7e..d41436858a 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -86,7 +86,8 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, (boolean)draw->bypass_clipping, (boolean)(draw->identity_viewport || draw->rasterizer->bypass_vs_clip_and_viewport), - (boolean)draw->rasterizer->gl_rasterization_rules ); + (boolean)draw->rasterizer->gl_rasterization_rules, + need_edgeflags ); if (!(opt & PT_PIPELINE)) { diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index 6c1cb48e8b..0745b168de 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -147,6 +147,34 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs, +/* As above plus edgeflags + */ +static boolean +post_vs_cliptest_viewport_gl_edgeflag(struct pt_post_vs *pvs, + struct vertex_header *vertices, + unsigned count, + unsigned stride ) +{ + if (!post_vs_cliptest_viewport_gl( pvs, vertices, count, stride)) + return FALSE; + + /* If present, copy edgeflag VS output into vertex header. + * Otherwise, leave header as is. + */ + if (pvs->draw->vs.edgeflag_output) { + struct vertex_header *out = vertices; + int ef = pvs->draw->vs.edgeflag_output; + + for (j = 0; j < count; j++) { + const float *edgeflag = out->data[ef]; + out->edgeflag = (edgeflag[0] != 1.0f); + } + } +} + + + + /* If bypass_clipping is set, skip cliptest and rhw divide. */ static boolean post_vs_viewport( struct pt_post_vs *pvs, @@ -203,15 +231,26 @@ void draw_pt_post_vs_prepare( struct pt_post_vs *pvs, boolean bypass_viewport, boolean opengl ) { - if (bypass_clipping) { - if (bypass_viewport) - pvs->run = post_vs_none; - else - pvs->run = post_vs_viewport; + if (!need_edgeflags) { + if (bypass_clipping) { + if (bypass_viewport) + pvs->run = post_vs_none; + else + pvs->run = post_vs_viewport; + } + else { + /* if (opengl) */ + pvs->run = post_vs_cliptest_viewport_gl; + } } else { - /* if (opengl) */ - pvs->run = post_vs_cliptest_viewport_gl; + /* If we need to copy edgeflags to the vertex header, it should + * mean we're running the primitive pipeline. Hence the bypass + * flags should be false. + */ + assert(!bypass_clipping); + assert(!bypass_viewport); + pvs->run = post_vs_cliptest_viewport_gl_edgeflag; } } diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index f896001eb1..11bcdc0a24 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -57,14 +57,6 @@ struct pipe_context { void (*destroy)( struct pipe_context * ); - - /* Possible interface for setting edgeflags. These aren't really - * vertex elements, so don't fit there. - */ - void (*set_edgeflags)( struct pipe_context *, - const unsigned *bitfield ); - - /** * VBO drawing (return false on fallbacks (temporary??)) */ diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 588ca5e026..c051d4dae8 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -127,7 +127,8 @@ struct tgsi_declaration_range #define TGSI_SEMANTIC_GENERIC 5 #define TGSI_SEMANTIC_NORMAL 6 #define TGSI_SEMANTIC_FACE 7 -#define TGSI_SEMANTIC_COUNT 8 /**< number of semantic values */ +#define TGSI_SEMANTIC_EDGEFLAG 8 +#define TGSI_SEMANTIC_COUNT 9 /**< number of semantic values */ struct tgsi_declaration_semantic { diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 09baff875b..e209634c90 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -106,7 +106,16 @@ find_translated_vp(struct st_context *st, /* Nothing in our key yet. This will change: */ memset(&key, 0, sizeof key); - key.dummy = 0; + + /* When this is true, we will add an extra input to the vertex + * shader translation (for edgeflags), an extra output with + * edgeflag semantics, and extend the vertex shader to pass through + * the input to the output. We'll need to use similar logic to set + * up the extra vertex_element input for edgeflags. + */ + key.passthrough_edgeflags = (ctx->Polygon.FrontMode != GL_FILL || + ctx->Polygon.BackMode != GL_FILL); + /* Do we need to throw away old translations after a change in the * GL program string? diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index e13ae57a0e..773a3f17cd 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -217,59 +217,7 @@ st_pipe_vertex_format(GLenum type, GLuint size, GLenum format, } -/* - * If edge flags are needed, setup an bitvector of flags and call - * pipe->set_edgeflags(). - * XXX memleak: need to free the returned pointer at some point - */ -static void * -setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count, - const struct gl_client_array *array) -{ - struct pipe_context *pipe = ctx->st->pipe; - - if ((primMode == GL_TRIANGLES || - primMode == GL_QUADS || - primMode == GL_POLYGON) && - (ctx->Polygon.FrontMode != GL_FILL || - ctx->Polygon.BackMode != GL_FILL)) { - /* need edge flags */ - GLint i; - unsigned *vec; - struct st_buffer_object *stobj = st_buffer_object(array->BufferObj); - ubyte *map; - - if (!stobj || stobj->Base.Name == 0) { - /* edge flags are not in a VBO */ - return NULL; - } - - vec = (unsigned *) _mesa_calloc(sizeof(unsigned) * ((count + 31) / 32)); - if (!vec) - return NULL; - - map = pipe_buffer_map(pipe->screen, stobj->buffer, PIPE_BUFFER_USAGE_CPU_READ); - map = ADD_POINTERS(map, array->Ptr); - for (i = 0; i < count; i++) { - if (*((float *) map)) - vec[i/32] |= 1 << (i % 32); - - map += array->StrideB; - } - - pipe_buffer_unmap(pipe->screen, stobj->buffer); - - pipe->set_edgeflags(pipe, vec); - - return vec; - } - else { - /* edge flags not needed */ - pipe->set_edgeflags(pipe, NULL); - return NULL; - } -} /** @@ -671,10 +619,6 @@ st_draw_vbo(GLcontext *ctx, * through to driver & draw module. These interfaces still * need a bit of work... */ - setup_edgeflags(ctx, prims[i].mode, - prims[i].start + indexOffset, prims[i].count, - arrays[VERT_ATTRIB_EDGEFLAG]); - pipe->draw_range_elements(pipe, indexBuf, indexSize, min_index, max_index, @@ -683,10 +627,6 @@ st_draw_vbo(GLcontext *ctx, } else { for (i = 0; i < nr_prims; i++) { - setup_edgeflags(ctx, prims[i].mode, - prims[i].start + indexOffset, prims[i].count, - arrays[VERT_ATTRIB_EDGEFLAG]); - pipe->draw_elements(pipe, indexBuf, indexSize, prims[i].mode, prims[i].start + indexOffset, prims[i].count); @@ -699,10 +639,6 @@ st_draw_vbo(GLcontext *ctx, /* non-indexed */ GLuint i; for (i = 0; i < nr_prims; i++) { - setup_edgeflags(ctx, prims[i].mode, - prims[i].start, prims[i].count, - arrays[VERT_ATTRIB_EDGEFLAG]); - pipe->draw_arrays(pipe, prims[i].mode, prims[i].start, prims[i].count); } } diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 1611d53e2f..9fd670cac2 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -741,6 +741,7 @@ emit_face_var( struct st_translate *t, const struct tgsi_token * st_translate_mesa_program( GLcontext *ctx, + struct ureg_program *ureg; uint procType, const struct gl_program *program, GLuint numInputs, @@ -754,7 +755,6 @@ st_translate_mesa_program( const ubyte outputSemanticIndex[] ) { struct st_translate translate, *t; - struct ureg_program *ureg; const struct tgsi_token *tokens = NULL; unsigned i; @@ -764,11 +764,7 @@ st_translate_mesa_program( t->procType = procType; t->inputMapping = inputMapping; t->outputMapping = outputMapping; - t->ureg = ureg_create( procType ); - if (t->ureg == NULL) - return NULL; - - ureg = t->ureg; + t->ureg = ureg; /*_mesa_print_program(program);*/ @@ -899,8 +895,7 @@ st_translate_mesa_program( t->insn[t->labels[i].branch_target] ); } - tokens = ureg_get_tokens( ureg, NULL ); - ureg_destroy( ureg ); + return PIPE_OK; out: FREE(t->insn); @@ -919,7 +914,7 @@ out: debug_assert(0); } - return tokens; + return PIPE_ERROR_OUT_OF_MEMORY; } diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 5c81a033f9..876d92539e 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -192,10 +192,16 @@ st_translate_vertex_program(struct st_context *st, { struct st_vp_varient *vpv = CALLOC_STRUCT(st_vp_varient); struct pipe_context *pipe = st->pipe; + struct ureg_program *ureg; - vpv->state.tokens = + ureg = ureg_create( TGSI_PROCESSOR_VERTEX ); + if (ureg == NULL) + return NULL; + + error = st_translate_mesa_program(st->ctx, TGSI_PROCESSOR_VERTEX, + ureg, &stvp->Base.Base, /* inputs */ stvp->num_inputs, @@ -209,6 +215,20 @@ st_translate_vertex_program(struct st_context *st, stvp->output_semantic_name, stvp->output_semantic_index ); + if (ret) + goto fail; + + /* Edgeflags will be the last input: + */ + if (key.passthrough_edgeflags) { + ureg_MOV( ureg, + ureg_DECL_output( ureg, TGSI_SEMANTIC_EDGEFLAG, 0 ), + ureg_DECL_next_vs_input(ureg)); + } + + tokens = ureg_get_tokens( ureg, NULL ); + ureg_destroy( ureg ); + vpv->driver_shader = pipe->create_vs_state(pipe, &vpv->state); if ((ST_DEBUG & DEBUG_TGSI) && (ST_DEBUG & DEBUG_MESA)) { @@ -222,6 +242,10 @@ st_translate_vertex_program(struct st_context *st, } return vpv; + +fail: + ureg_destroy( ureg ); + return NULL; } @@ -243,6 +267,7 @@ st_translate_fragment_program(struct st_context *st, GLuint interpMode[16]; /* XXX size? */ GLuint attr; const GLbitfield inputsRead = stfp->Base.Base.InputsRead; + struct ureg_program *ureg; GLuint vslot = 0; uint fs_num_inputs = 0; @@ -377,6 +402,11 @@ st_translate_fragment_program(struct st_context *st, if (!inputMapping) inputMapping = defaultInputMapping; + ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); + if (ureg == NULL) + return NULL; + + stfp->state.tokens = st_translate_mesa_program(st->ctx, TGSI_PROCESSOR_FRAGMENT, diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 88aadbd751..91392785e5 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -73,7 +73,7 @@ struct st_fragment_program struct st_vp_varient_key { - char dummy; /* currently unused */ + boolean passthrough_edgeflags; }; -- cgit v1.2.3 From a0127b6ced257919180ba3a1bf534b68d9c750be Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Mon, 14 Dec 2009 18:36:33 +0100 Subject: gallium: more work for edgeflags changes fixes, cleanups, etc. not working yet --- src/gallium/auxiliary/draw/draw_private.h | 1 + src/gallium/auxiliary/draw/draw_pt.c | 11 ----------- src/gallium/auxiliary/draw/draw_pt.h | 3 ++- src/gallium/auxiliary/draw/draw_pt_fetch.c | 15 +-------------- src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 5 ++--- src/gallium/auxiliary/draw/draw_pt_post_vs.c | 5 ++++- src/gallium/auxiliary/draw/draw_vs.c | 4 ++++ src/gallium/auxiliary/draw/draw_vs.h | 1 + src/gallium/drivers/softpipe/sp_context.c | 2 -- src/gallium/drivers/softpipe/sp_draw_arrays.c | 8 -------- src/gallium/drivers/softpipe/sp_state.h | 4 ---- src/mesa/state_tracker/st_atom_shader.c | 4 ++-- src/mesa/state_tracker/st_mesa_to_tgsi.c | 10 ++++------ src/mesa/state_tracker/st_mesa_to_tgsi.h | 4 +++- src/mesa/state_tracker/st_program.c | 15 ++++++++++----- 15 files changed, 34 insertions(+), 58 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 0750e6e379..3850cede1e 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -182,6 +182,7 @@ struct draw_context struct draw_vertex_shader *vertex_shader; uint num_vs_outputs; /**< convenience, from vertex_shader */ uint position_output; + uint edgeflag_output; /** TGSI program interpreter runtime state */ struct tgsi_exec_machine *machine; diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 139ae1fe55..2801dbafe4 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -314,14 +314,3 @@ draw_arrays(struct draw_context *draw, unsigned prim, /* drawing done here: */ draw_pt_arrays(draw, prim, start, count); } - -boolean draw_pt_get_edgeflag( struct draw_context *draw, - unsigned idx ) -{ - if (draw->pt.user.edgeflag) { - float *ef = draw->pt.verted_buffer[idx] - return (draw->pt.user.edgeflag[idx/32] & (1 << (idx%32))) != 0; - } - else - return 1; -} diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index b5c8c82f4a..20edf7a227 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -212,7 +212,8 @@ boolean draw_pt_post_vs_run( struct pt_post_vs *pvs, void draw_pt_post_vs_prepare( struct pt_post_vs *pvs, boolean bypass_clipping, boolean bypass_viewport, - boolean opengl ); + boolean opengl, + boolean need_edgeflags ); struct pt_post_vs *draw_pt_post_vs_create( struct draw_context *draw ); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index cb609f8c41..305bfef435 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -42,11 +42,11 @@ struct pt_fetch { struct translate *translate; unsigned vertex_size; - boolean need_edgeflags; struct translate_cache *cache; }; + /* Perform the fetch from API vertex elements & vertex buffers, to a * contiguous set of float[4] attributes as required for the * vertex_shader->run_linear() method. @@ -160,11 +160,6 @@ void draw_pt_fetch_run( struct pt_fetch *fetch, count, verts ); - /* Extract edgeflag values from vertex data into the header. - */ - if (fetch->need_edgeflags) { - extract_edge_flags( fetch, count ); - } } @@ -189,14 +184,6 @@ void draw_pt_fetch_run_linear( struct pt_fetch *fetch, start, count, verts ); - - /* Extract edgeflag values from vertex data into the header. XXX: - * this should be done after the vertex shader is run. - * Bypass-vs-and-clip interaction with pipeline??? - */ - if (fetch->need_edgeflags) { - extract_edge_flags( fetch, count ); - } } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index d41436858a..932113783d 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -85,10 +85,9 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, draw_pt_post_vs_prepare( fpme->post_vs, (boolean)draw->bypass_clipping, (boolean)(draw->identity_viewport || - draw->rasterizer->bypass_vs_clip_and_viewport), + draw->rasterizer->bypass_vs_clip_and_viewport), (boolean)draw->rasterizer->gl_rasterization_rules, - need_edgeflags ); - + (draw->vs.edgeflag_output ? true : false) ); if (!(opt & PT_PIPELINE)) { draw_pt_emit_prepare( fpme->emit, diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index 0745b168de..08d77649a3 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -155,6 +155,7 @@ post_vs_cliptest_viewport_gl_edgeflag(struct pt_post_vs *pvs, unsigned count, unsigned stride ) { + unsigned j; if (!post_vs_cliptest_viewport_gl( pvs, vertices, count, stride)) return FALSE; @@ -170,6 +171,7 @@ post_vs_cliptest_viewport_gl_edgeflag(struct pt_post_vs *pvs, out->edgeflag = (edgeflag[0] != 1.0f); } } + return TRUE; } @@ -229,7 +231,8 @@ boolean draw_pt_post_vs_run( struct pt_post_vs *pvs, void draw_pt_post_vs_prepare( struct pt_post_vs *pvs, boolean bypass_clipping, boolean bypass_viewport, - boolean opengl ) + boolean opengl, + boolean need_edgeflags ) { if (!need_edgeflags) { if (bypass_clipping) { diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index 790e89ed82..3553689532 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -101,6 +101,9 @@ draw_create_vertex_shader(struct draw_context *draw, if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_POSITION && vs->info.output_semantic_index[i] == 0) vs->position_output = i; + else if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_EDGEFLAG && + vs->info.output_semantic_index[i] == 0) + vs->edgeflag_output = i; } } @@ -120,6 +123,7 @@ draw_bind_vertex_shader(struct draw_context *draw, draw->vs.vertex_shader = dvs; draw->vs.num_vs_outputs = dvs->info.num_outputs; draw->vs.position_output = dvs->position_output; + draw->vs.edgeflag_output = dvs->edgeflag_output; dvs->prepare( dvs, draw ); } else { diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index 89ae158751..e3b807ebd0 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -107,6 +107,7 @@ struct draw_vertex_shader { struct tgsi_shader_info info; unsigned position_output; + unsigned edgeflag_output; /* Extracted from shader: */ diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index f8bf3e9974..2a33587b5a 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -238,8 +238,6 @@ softpipe_create( struct pipe_screen *screen ) softpipe->pipe.draw_arrays = softpipe_draw_arrays; softpipe->pipe.draw_elements = softpipe_draw_elements; softpipe->pipe.draw_range_elements = softpipe_draw_range_elements; - softpipe->pipe.set_edgeflags = softpipe_set_edgeflags; - softpipe->pipe.clear = softpipe_clear; softpipe->pipe.flush = softpipe_flush; diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index d4045816d0..518ef8806e 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -184,11 +184,3 @@ softpipe_draw_elements(struct pipe_context *pipe, 0, 0xffffffff, mode, start, count ); } - - -void -softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags) -{ - struct softpipe_context *sp = softpipe_context(pipe); - draw_set_edgeflags(sp->draw, edgeflags); -} diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index d488fb8710..26d5c3fbb2 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -189,10 +189,6 @@ softpipe_draw_range_elements(struct pipe_context *pipe, unsigned max_index, unsigned mode, unsigned start, unsigned count); -void -softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags); - - void softpipe_map_transfers(struct softpipe_context *sp); diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index e209634c90..629bf8953e 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -113,8 +113,8 @@ find_translated_vp(struct st_context *st, * the input to the output. We'll need to use similar logic to set * up the extra vertex_element input for edgeflags. */ - key.passthrough_edgeflags = (ctx->Polygon.FrontMode != GL_FILL || - ctx->Polygon.BackMode != GL_FILL); + key.passthrough_edgeflags = (st->ctx->Polygon.FrontMode != GL_FILL || + st->ctx->Polygon.BackMode != GL_FILL); /* Do we need to throw away old translations after a change in the diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 9fd670cac2..72bd17bf41 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -738,11 +738,11 @@ emit_face_var( struct st_translate *t, * * \return array of translated tokens, caller's responsibility to free */ -const struct tgsi_token * +enum pipe_error st_translate_mesa_program( GLcontext *ctx, - struct ureg_program *ureg; uint procType, + struct ureg_program *ureg, const struct gl_program *program, GLuint numInputs, const GLuint inputMapping[], @@ -755,7 +755,6 @@ st_translate_mesa_program( const ubyte outputSemanticIndex[] ) { struct st_translate translate, *t; - const struct tgsi_token *tokens = NULL; unsigned i; t = &translate; @@ -904,16 +903,15 @@ out: if (t->error) { debug_printf("%s: translate error flag set\n", __FUNCTION__); - FREE((void *)tokens); - tokens = NULL; } +/* ??? if (!tokens) { debug_printf("%s: failed to translate Mesa program:\n", __FUNCTION__); _mesa_print_program(program); debug_assert(0); } - +*/ return PIPE_ERROR_OUT_OF_MEMORY; } diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index dc0362fe79..9dae373ede 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -30,6 +30,7 @@ #define ST_MESA_TO_TGSI_H #include "main/mtypes.h" +#include "tgsi/tgsi_ureg.h" #if defined __cplusplus @@ -39,10 +40,11 @@ extern "C" { struct tgsi_token; struct gl_program; -const struct tgsi_token * +enum pipe_error st_translate_mesa_program( GLcontext *ctx, uint procType, + struct ureg_program *ureg, const struct gl_program *program, GLuint numInputs, const GLuint inputMapping[], diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 876d92539e..24f2387429 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -193,6 +193,7 @@ st_translate_vertex_program(struct st_context *st, struct st_vp_varient *vpv = CALLOC_STRUCT(st_vp_varient); struct pipe_context *pipe = st->pipe; struct ureg_program *ureg; + enum pipe_error error; ureg = ureg_create( TGSI_PROCESSOR_VERTEX ); if (ureg == NULL) @@ -215,18 +216,18 @@ st_translate_vertex_program(struct st_context *st, stvp->output_semantic_name, stvp->output_semantic_index ); - if (ret) + if (error) goto fail; /* Edgeflags will be the last input: */ - if (key.passthrough_edgeflags) { + if (key->passthrough_edgeflags) { ureg_MOV( ureg, ureg_DECL_output( ureg, TGSI_SEMANTIC_EDGEFLAG, 0 ), ureg_DECL_next_vs_input(ureg)); } - tokens = ureg_get_tokens( ureg, NULL ); + vpv->state.tokens = ureg_get_tokens( ureg, NULL ); ureg_destroy( ureg ); vpv->driver_shader = pipe->create_vs_state(pipe, &vpv->state); @@ -266,6 +267,7 @@ st_translate_fragment_program(struct st_context *st, GLuint defaultInputMapping[FRAG_ATTRIB_MAX]; GLuint interpMode[16]; /* XXX size? */ GLuint attr; + enum pipe_error error; const GLbitfield inputsRead = stfp->Base.Base.InputsRead; struct ureg_program *ureg; GLuint vslot = 0; @@ -404,12 +406,13 @@ st_translate_fragment_program(struct st_context *st, ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); if (ureg == NULL) - return NULL; + return; - stfp->state.tokens = + error = st_translate_mesa_program(st->ctx, TGSI_PROCESSOR_FRAGMENT, + ureg, &stfp->Base.Base, /* inputs */ fs_num_inputs, @@ -423,6 +426,8 @@ st_translate_fragment_program(struct st_context *st, fs_output_semantic_name, fs_output_semantic_index ); + stfp->state.tokens = ureg_get_tokens( ureg, NULL ); + ureg_destroy( ureg ); stfp->driver_shader = pipe->create_fs_state(pipe, &stfp->state); if ((ST_DEBUG & DEBUG_TGSI) && (ST_DEBUG & DEBUG_MESA)) { -- cgit v1.2.3 From 50caff5675888c0063c73fa64b88129db7aa11dd Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 16 Dec 2009 22:12:16 +0100 Subject: gallium: edgeflags change fixes use correct number of vertex inputs fix not running pipeline in case of edgeflags changes to mesa to tgsi translation still very broken --- src/gallium/auxiliary/draw/draw_pt_post_vs.c | 11 +++++--- src/gallium/auxiliary/tgsi/tgsi_dump.c | 3 ++- src/mesa/state_tracker/st_draw.c | 38 ++++++++++++++++------------ src/mesa/state_tracker/st_program.c | 8 +++++- src/mesa/state_tracker/st_program.h | 3 +++ 5 files changed, 41 insertions(+), 22 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index 08d77649a3..9dfb47837e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -156,8 +156,9 @@ post_vs_cliptest_viewport_gl_edgeflag(struct pt_post_vs *pvs, unsigned stride ) { unsigned j; - if (!post_vs_cliptest_viewport_gl( pvs, vertices, count, stride)) - return FALSE; + boolean needpipe; + + needpipe = post_vs_cliptest_viewport_gl( pvs, vertices, count, stride); /* If present, copy edgeflag VS output into vertex header. * Otherwise, leave header as is. @@ -168,10 +169,12 @@ post_vs_cliptest_viewport_gl_edgeflag(struct pt_post_vs *pvs, for (j = 0; j < count; j++) { const float *edgeflag = out->data[ef]; - out->edgeflag = (edgeflag[0] != 1.0f); + out->edgeflag = !(edgeflag[0] != 1.0f); + needpipe |= !out->edgeflag; + out = (struct vertex_header *)( (char *)out + stride ); } } - return TRUE; + return needpipe; } diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index d09ab92565..d3a5da4b2b 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -120,7 +120,8 @@ static const char *semantic_names[] = "PSIZE", "GENERIC", "NORMAL", - "FACE" + "FACE", + "EDGEFLAG" }; static const char *immediate_type_names[] = diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 773a3f17cd..6b0007a824 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -227,6 +227,7 @@ st_pipe_vertex_format(GLenum type, GLuint size, GLenum format, */ static GLboolean is_interleaved_arrays(const struct st_vertex_program *vp, + const struct st_vp_varient *vpv, const struct gl_client_array **arrays, GLboolean *userSpace) { @@ -236,7 +237,7 @@ is_interleaved_arrays(const struct st_vertex_program *vp, GLuint num_client_arrays = 0; const GLubyte *client_addr = NULL; - for (attr = 0; attr < vp->num_inputs; attr++) { + for (attr = 0; attr < vpv->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; const struct gl_buffer_object *bufObj = arrays[mesaAttr]->BufferObj; const GLsizei stride = arrays[mesaAttr]->StrideB; /* in bytes */ @@ -269,7 +270,7 @@ is_interleaved_arrays(const struct st_vertex_program *vp, } } - *userSpace = (num_client_arrays == vp->num_inputs); + *userSpace = (num_client_arrays == vpv->num_inputs); /* printf("user space: %d (%d %d)\n", (int) *userSpace,num_client_arrays,vp->num_inputs); */ return GL_TRUE; @@ -281,15 +282,16 @@ is_interleaved_arrays(const struct st_vertex_program *vp, */ static void get_arrays_bounds(const struct st_vertex_program *vp, - const struct gl_client_array **arrays, - GLuint max_index, - const GLubyte **low, const GLubyte **high) + const struct st_vp_varient *vpv, + const struct gl_client_array **arrays, + GLuint max_index, + const GLubyte **low, const GLubyte **high) { const GLubyte *low_addr = NULL; const GLubyte *high_addr = NULL; GLuint attr; - for (attr = 0; attr < vp->num_inputs; attr++) { + for (attr = 0; attr < vpv->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; const GLint stride = arrays[mesaAttr]->StrideB; const GLubyte *start = arrays[mesaAttr]->Ptr; @@ -321,6 +323,7 @@ get_arrays_bounds(const struct st_vertex_program *vp, static void setup_interleaved_attribs(GLcontext *ctx, const struct st_vertex_program *vp, + const struct st_vp_varient *vpv, const struct gl_client_array **arrays, GLuint max_index, GLboolean userSpace, @@ -331,7 +334,7 @@ setup_interleaved_attribs(GLcontext *ctx, GLuint attr; const GLubyte *offset0; - for (attr = 0; attr < vp->num_inputs; attr++) { + for (attr = 0; attr < vpv->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; struct st_buffer_object *stobj = st_buffer_object(bufobj); @@ -342,7 +345,7 @@ setup_interleaved_attribs(GLcontext *ctx, if (attr == 0) { const GLubyte *low, *high; - get_arrays_bounds(vp, arrays, max_index, &low, &high); + get_arrays_bounds(vp, vpv, arrays, max_index, &low, &high); /*printf("buffer range: %p %p %d\n", low, high, high-low);*/ offset0 = low; @@ -383,6 +386,7 @@ setup_interleaved_attribs(GLcontext *ctx, static void setup_non_interleaved_attribs(GLcontext *ctx, const struct st_vertex_program *vp, + const struct st_vp_varient *vpv, const struct gl_client_array **arrays, GLuint max_index, GLboolean *userSpace, @@ -392,7 +396,7 @@ setup_non_interleaved_attribs(GLcontext *ctx, struct pipe_context *pipe = ctx->st->pipe; GLuint attr; - for (attr = 0; attr < vp->num_inputs; attr++) { + for (attr = 0; attr < vpv->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; GLsizei stride = arrays[mesaAttr]->StrideB; @@ -503,6 +507,7 @@ st_draw_vbo(GLcontext *ctx, { struct pipe_context *pipe = ctx->st->pipe; const struct st_vertex_program *vp; + const struct st_vp_varient *vpv; const struct pipe_shader_state *vs; struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS]; GLuint attr; @@ -521,7 +526,8 @@ st_draw_vbo(GLcontext *ctx, /* must get these after state validation! */ vp = ctx->st->vp; - vs = &ctx->st->vp_varient->state; + vpv = ctx->st->vp_varient; + vs = &vpv->state; #if 0 if (MESA_VERBOSE & VERBOSE_GLSL) { @@ -534,21 +540,21 @@ st_draw_vbo(GLcontext *ctx, /* * Setup the vbuffer[] and velements[] arrays. */ - if (is_interleaved_arrays(vp, arrays, &userSpace)) { + if (is_interleaved_arrays(vp, vpv, arrays, &userSpace)) { /*printf("Draw interleaved\n");*/ - setup_interleaved_attribs(ctx, vp, arrays, max_index, userSpace, + setup_interleaved_attribs(ctx, vp, vpv, arrays, max_index, userSpace, vbuffer, velements); num_vbuffers = 1; - num_velements = vp->num_inputs; + num_velements = vpv->num_inputs; if (num_velements == 0) num_vbuffers = 0; } else { /*printf("Draw non-interleaved\n");*/ - setup_non_interleaved_attribs(ctx, vp, arrays, max_index, + setup_non_interleaved_attribs(ctx, vp, vpv, arrays, max_index, &userSpace, vbuffer, velements); - num_vbuffers = vp->num_inputs; - num_velements = vp->num_inputs; + num_vbuffers = vpv->num_inputs; + num_velements = vpv->num_inputs; } #if 0 diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 24f2387429..45ab8504ae 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -113,6 +113,9 @@ st_prepare_vertex_program(struct st_context *st, stvp->num_inputs++; } } + /* bit of a hack, presetup potentially unused edgeflag input */ + stvp->input_to_index[VERT_ATTRIB_EDGEFLAG] = stvp->num_inputs; + stvp->index_to_input[stvp->num_inputs] = VERT_ATTRIB_EDGEFLAG; /* Compute mapping of vertex program outputs to slots. */ @@ -199,6 +202,8 @@ st_translate_vertex_program(struct st_context *st, if (ureg == NULL) return NULL; + vpv->num_inputs = stvp->num_inputs; + error = st_translate_mesa_program(st->ctx, TGSI_PROCESSOR_VERTEX, @@ -224,7 +229,8 @@ st_translate_vertex_program(struct st_context *st, if (key->passthrough_edgeflags) { ureg_MOV( ureg, ureg_DECL_output( ureg, TGSI_SEMANTIC_EDGEFLAG, 0 ), - ureg_DECL_next_vs_input(ureg)); + ureg_DECL_vs_input( ureg, vpv->num_inputs )); + vpv->num_inputs++; } vpv->state.tokens = ureg_get_tokens( ureg, NULL ); diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 91392785e5..6b9a9226df 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -100,6 +100,9 @@ struct st_vp_varient /** Next in linked list */ struct st_vp_varient *next; + + /** similar to that in st_vertex_program, but with information about edgeflags too */ + GLuint num_inputs; }; -- cgit v1.2.3 From 890809d71b6ab40c56d341a5dad9f5e6b06469d0 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Fri, 18 Dec 2009 15:54:28 +0100 Subject: gallium: fix mesa to tgsi translation for edgeflags test apps (progs/trivial/tri-edgeflags and similar) seem to work now --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 15 ++++++++++++++- src/mesa/state_tracker/st_mesa_to_tgsi.h | 3 ++- src/mesa/state_tracker/st_program.c | 28 +++++++++++++++------------- 3 files changed, 31 insertions(+), 15 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 72bd17bf41..7b334e21d2 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -718,6 +718,16 @@ emit_face_var( struct st_translate *t, t->inputs[t->inputMapping[FRAG_ATTRIB_FACE]] = ureg_src(face_temp); } +static void +emit_edgeflags( struct st_translate *t, + const struct gl_program *program ) +{ + struct ureg_program *ureg = t->ureg; + struct ureg_dst edge_dst = t->outputs[t->outputMapping[VERT_RESULT_EDGE]]; + struct ureg_src edge_src = t->inputs[t->inputMapping[VERT_ATTRIB_EDGEFLAG]]; + + ureg_MOV( ureg, edge_dst, edge_src ); +} /** * Translate Mesa program to TGSI format. @@ -752,7 +762,8 @@ st_translate_mesa_program( GLuint numOutputs, const GLuint outputMapping[], const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[] ) + const ubyte outputSemanticIndex[], + boolean passthrough_edgeflags ) { struct st_translate translate, *t; unsigned i; @@ -823,6 +834,8 @@ st_translate_mesa_program( outputSemanticName[i], outputSemanticIndex[i] ); } + if (passthrough_edgeflags) + emit_edgeflags( t, program ); } /* Declare address register. diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index 9dae373ede..e3c5bd1d94 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -54,7 +54,8 @@ st_translate_mesa_program( GLuint numOutputs, const GLuint outputMapping[], const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[] ); + const ubyte outputSemanticIndex[], + boolean passthrough_edgeflags ); void st_free_tokens(const struct tgsi_token *tokens); diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 45ab8504ae..fcc04782de 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -185,6 +185,10 @@ st_prepare_vertex_program(struct st_context *st, } } } + /* similar hack to above, presetup potentially unused edgeflag output */ + stvp->result_to_output[VERT_RESULT_EDGE] = stvp->num_outputs; + stvp->output_semantic_name[stvp->num_outputs] = TGSI_SEMANTIC_EDGEFLAG; + stvp->output_semantic_index[stvp->num_outputs] = 0; } @@ -197,12 +201,18 @@ st_translate_vertex_program(struct st_context *st, struct pipe_context *pipe = st->pipe; struct ureg_program *ureg; enum pipe_error error; + unsigned num_outputs; ureg = ureg_create( TGSI_PROCESSOR_VERTEX ); if (ureg == NULL) return NULL; vpv->num_inputs = stvp->num_inputs; + num_outputs = stvp->num_outputs; + if (key->passthrough_edgeflags) { + vpv->num_inputs++; + num_outputs++; + } error = st_translate_mesa_program(st->ctx, @@ -210,29 +220,21 @@ st_translate_vertex_program(struct st_context *st, ureg, &stvp->Base.Base, /* inputs */ - stvp->num_inputs, + vpv->num_inputs, stvp->input_to_index, NULL, /* input semantic name */ NULL, /* input semantic index */ NULL, /* outputs */ - stvp->num_outputs, + num_outputs, stvp->result_to_output, stvp->output_semantic_name, - stvp->output_semantic_index ); + stvp->output_semantic_index, + key->passthrough_edgeflags ); if (error) goto fail; - /* Edgeflags will be the last input: - */ - if (key->passthrough_edgeflags) { - ureg_MOV( ureg, - ureg_DECL_output( ureg, TGSI_SEMANTIC_EDGEFLAG, 0 ), - ureg_DECL_vs_input( ureg, vpv->num_inputs )); - vpv->num_inputs++; - } - vpv->state.tokens = ureg_get_tokens( ureg, NULL ); ureg_destroy( ureg ); @@ -430,7 +432,7 @@ st_translate_fragment_program(struct st_context *st, fs_num_outputs, outputMapping, fs_output_semantic_name, - fs_output_semantic_index ); + fs_output_semantic_index, FALSE ); stfp->state.tokens = ureg_get_tokens( ureg, NULL ); ureg_destroy( ureg ); -- cgit v1.2.3 From 9d938e29f202552a98605906b71f31485af1ef60 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Fri, 18 Dec 2009 16:09:32 +0100 Subject: fix debug output mising from edgeflag changes --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 7 ------- src/mesa/state_tracker/st_program.c | 7 +++++++ 2 files changed, 7 insertions(+), 7 deletions(-) (limited to 'src/mesa/state_tracker/st_program.c') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 7b334e21d2..5c9be46a77 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -918,13 +918,6 @@ out: debug_printf("%s: translate error flag set\n", __FUNCTION__); } -/* ??? - if (!tokens) { - debug_printf("%s: failed to translate Mesa program:\n", __FUNCTION__); - _mesa_print_program(program); - debug_assert(0); - } -*/ return PIPE_ERROR_OUT_OF_MEMORY; } diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index fcc04782de..feaac013c1 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -236,6 +236,9 @@ st_translate_vertex_program(struct st_context *st, goto fail; vpv->state.tokens = ureg_get_tokens( ureg, NULL ); + if (!vpv->state.tokens) + goto fail; + ureg_destroy( ureg ); vpv->driver_shader = pipe->create_vs_state(pipe, &vpv->state); @@ -253,6 +256,10 @@ st_translate_vertex_program(struct st_context *st, return vpv; fail: + debug_printf("%s: failed to translate Mesa program:\n", __FUNCTION__); + _mesa_print_program(&stvp->Base.Base); + debug_assert(0); + ureg_destroy( ureg ); return NULL; } -- cgit v1.2.3