diff options
author | Brian Paul <brianp@vmware.com> | 2010-12-13 17:20:12 -0700 |
---|---|---|
committer | Brian Paul <brianp@vmware.com> | 2010-12-13 17:20:53 -0700 |
commit | 3d203b610045980853d26370ee21fb2ef4aed17e (patch) | |
tree | 4034d5c8eeac3e7d7a70dbad0680de19f794bced /src | |
parent | bb7c2691d25b6aaea2663f85a5b6723edbf56938 (diff) |
Squashed commit of the following (st-mesa-per-context-shaders branch):
commit 4f106f44a32eaddb6cf3fea6ba5ee9787bff609a
Author: Brian Paul <brianp@vmware.com>
Date: Mon Dec 13 14:06:08 2010 -0700
st/mesa: reorganize vertex program translation code
Now it looks like the fragment and geometry program code.
Also remove the serial number fields from programs. It was used to
determine when new translations were needed. Now the variant key is
used for that. And the st_program_string_notify() callback removes all
variants when the program's code is changed.
commit e12d6791c5e4bff60bb2e6c04414b1b4d1325f3e
Author: Brian Paul <brianp@vmware.com>
Date: Mon Dec 13 13:38:12 2010 -0700
st/mesa: implement geometry shader varients
Only needed in order to support per-context gallium shaders.
commit c5751c673644808ab069259a852f24c4c0e92b9d
Author: Brian Paul <brianp@vmware.com>
Date: Sun Dec 12 15:28:57 2010 -0700
st/mesa: restore glDraw/CopyPixels using new fragment program variants
Clean up the logic for fragment programs for glDraw/CopyPixels. We now
generate fragment program variants for glDraw/CopyPixels as needed which
do texture sampling, pixel scale/bias, pixelmap lookups, etc.
commit 7b0bb99bab6547f503a0176b5c0aef1482b02c97
Author: Brian Paul <brianp@vmware.com>
Date: Fri Dec 10 17:03:23 2010 -0700
st/mesa: checkpoint: implement fragment program variants
The fragment programs variants are per-context, as the vertex programs.
NOTE: glDrawPixels is totally broken at this point.
commit 2cc926183f957f8abac18d71276dd5bbd1f27be2
Author: Brian Paul <brianp@vmware.com>
Date: Fri Dec 10 14:59:32 2010 -0700
st/mesa: make vertex shader variants per-context
Gallium shaders are per-context but OpenGL shaders aren't. So we need
to make a different variant for each context.
During context tear-down we need to walk over all shaders/programs and
free all variants for the context being destroyed.
Diffstat (limited to 'src')
-rw-r--r-- | src/mesa/state_tracker/st_atom_shader.c | 143 | ||||
-rw-r--r-- | src/mesa/state_tracker/st_cb_bitmap.c | 88 | ||||
-rw-r--r-- | src/mesa/state_tracker/st_cb_bitmap.h | 8 | ||||
-rw-r--r-- | src/mesa/state_tracker/st_cb_drawpixels.c | 199 | ||||
-rw-r--r-- | src/mesa/state_tracker/st_cb_drawpixels.h | 10 | ||||
-rw-r--r-- | src/mesa/state_tracker/st_cb_program.c | 46 | ||||
-rw-r--r-- | src/mesa/state_tracker/st_context.c | 2 | ||||
-rw-r--r-- | src/mesa/state_tracker/st_context.h | 4 | ||||
-rw-r--r-- | src/mesa/state_tracker/st_program.c | 751 | ||||
-rw-r--r-- | src/mesa/state_tracker/st_program.h | 119 |
10 files changed, 894 insertions, 476 deletions
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 05442ef91b5..f416293a7e4 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -50,99 +50,6 @@ #include "st_program.h" - -/** - * Translate fragment program if needed. - */ -static void -translate_fp(struct st_context *st, - struct st_fragment_program *stfp) -{ - if (!stfp->tgsi.tokens) { - assert(stfp->Base.Base.NumInstructions > 0); - - st_translate_fragment_program(st, stfp); - } -} - -/* - * Translate geometry program if needed. - */ -static void -translate_gp(struct st_context *st, - struct st_geometry_program *stgp) -{ - if (!stgp->tgsi.tokens) { - assert(stgp->Base.Base.NumInstructions > 1); - - st_translate_geometry_program(st, stgp); - } -} - -/** - * 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; - - /* Nothing in our key yet. This will change: - */ - memset(&key, 0, sizeof key); - - /* 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. - * _NEW_POLYGON, ST_NEW_EDGEFLAGS_DATA - */ - key.passthrough_edgeflags = (st->vertdata_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 - * GL program string? - */ - if (stvp->serialNo != stvp->lastSerialNo) { - /* These may have changed if the program string changed. - */ - st_prepare_vertex_program( st, stvp ); - - /* We are now up-to-date: - */ - 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; - } - } - - /* 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 vpv; -} - - /** * Return pointer to a pass-through fragment shader. * This shader is used when a texture is missing/incomplete. @@ -167,12 +74,16 @@ static void update_fp( struct st_context *st ) { struct st_fragment_program *stfp; + struct st_fp_varient_key key; assert(st->ctx->FragmentProgram._Current); stfp = st_fragment_program(st->ctx->FragmentProgram._Current); assert(stfp->Base.Base.Target == GL_FRAGMENT_PROGRAM_ARB); - translate_fp(st, stfp); + memset(&key, 0, sizeof(key)); + key.st = st; + + st->fp_varient = st_get_fp_varient(st, stfp, &key); st_reference_fragprog(st, &st->fp, stfp); @@ -182,7 +93,8 @@ update_fp( struct st_context *st ) cso_set_fragment_shader_handle(st->cso_context, fs); } else { - cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader); + cso_set_fragment_shader_handle(st->cso_context, + st->fp_varient->driver_shader); } } @@ -206,6 +118,7 @@ static void update_vp( struct st_context *st ) { struct st_vertex_program *stvp; + struct st_vp_varient_key key; /* find active shader and params -- Should be covered by * ST_NEW_VERTEX_PROGRAM @@ -214,7 +127,21 @@ update_vp( struct st_context *st ) 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); + memset(&key, 0, sizeof key); + key.st = st; /* variants are per-context */ + + /* 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. + * _NEW_POLYGON, ST_NEW_EDGEFLAGS_DATA + */ + key.passthrough_edgeflags = (st->vertdata_edgeflags && ( + st->ctx->Polygon.FrontMode != GL_FILL || + st->ctx->Polygon.BackMode != GL_FILL)); + + st->vp_varient = st_get_vp_varient(st, stvp, &key); st_reference_vertprog(st, &st->vp, stvp); @@ -231,14 +158,16 @@ const struct st_tracked_state st_update_vp = { _NEW_POLYGON, /* mesa */ ST_NEW_VERTEX_PROGRAM | ST_NEW_EDGEFLAGS_DATA /* st */ }, - update_vp /* update */ + update_vp /* update */ }; + + static void update_gp( struct st_context *st ) { - struct st_geometry_program *stgp; + struct st_gp_varient_key key; if (!st->ctx->GeometryProgram._Current) { cso_set_geometry_shader_handle(st->cso_context, NULL); @@ -248,18 +177,22 @@ update_gp( struct st_context *st ) stgp = st_geometry_program(st->ctx->GeometryProgram._Current); assert(stgp->Base.Base.Target == MESA_GEOMETRY_PROGRAM); - translate_gp(st, stgp); + memset(&key, 0, sizeof(key)); + key.st = st; + + st->gp_varient = st_get_gp_varient(st, stgp, &key); st_reference_geomprog(st, &st->gp, stgp); - cso_set_geometry_shader_handle(st->cso_context, stgp->driver_shader); + cso_set_geometry_shader_handle(st->cso_context, + st->gp_varient->driver_shader); } const struct st_tracked_state st_update_gp = { - "st_update_gp", /* name */ - { /* dirty */ - 0, /* mesa */ - ST_NEW_GEOMETRY_PROGRAM /* st */ + "st_update_gp", /* name */ + { /* dirty */ + 0, /* mesa */ + ST_NEW_GEOMETRY_PROGRAM /* st */ }, - update_gp /* update */ + update_gp /* update */ }; diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index f08697fe23b..320b73f03de 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -185,48 +185,47 @@ find_free_bit(uint bitfield) /** * Combine basic bitmap fragment program with the user-defined program. + * \param st current context + * \param fpIn the incoming fragment program + * \param fpOut the new fragment program which does fragment culling + * \param bitmap_sampler sampler number for the bitmap texture */ -static struct st_fragment_program * -combined_bitmap_fragment_program(struct gl_context *ctx) +void +st_make_bitmap_fragment_program(struct st_context *st, + struct gl_fragment_program *fpIn, + struct gl_fragment_program **fpOut, + GLuint *bitmap_sampler) { - struct st_context *st = st_context(ctx); - struct st_fragment_program *stfp = st->fp; - - if (!stfp->bitmap_program) { - /* - * Generate new program which is the user-defined program prefixed - * with the bitmap sampler/kill instructions. - */ - struct st_fragment_program *bitmap_prog; - uint sampler; + struct st_fragment_program *bitmap_prog; + struct gl_program *newProg; + uint sampler; - sampler = find_free_bit(st->fp->Base.Base.SamplersUsed); - bitmap_prog = make_bitmap_fragment_program(ctx, sampler); - - stfp->bitmap_program = (struct st_fragment_program *) - _mesa_combine_programs(ctx, - &bitmap_prog->Base.Base, &stfp->Base.Base); - stfp->bitmap_program->bitmap_sampler = sampler; + /* + * Generate new program which is the user-defined program prefixed + * with the bitmap sampler/kill instructions. + */ + sampler = find_free_bit(fpIn->Base.SamplersUsed); + bitmap_prog = make_bitmap_fragment_program(st->ctx, sampler); - /* done with this after combining */ - st_reference_fragprog(st, &bitmap_prog, NULL); + newProg = _mesa_combine_programs(st->ctx, + &bitmap_prog->Base.Base, + &fpIn->Base); + /* done with this after combining */ + st_reference_fragprog(st, &bitmap_prog, NULL); #if 0 - { - struct gl_program *p = &stfp->bitmap_program->Base.Base; - printf("Combined bitmap program:\n"); - _mesa_print_program(p); - printf("InputsRead: 0x%x\n", p->InputsRead); - printf("OutputsWritten: 0x%x\n", p->OutputsWritten); - _mesa_print_parameter_list(p->Parameters); - } -#endif - - /* translate to TGSI tokens */ - st_translate_fragment_program(st, stfp->bitmap_program); + { + printf("Combined bitmap program:\n"); + _mesa_print_program(newProg); + printf("InputsRead: 0x%x\n", newProg->InputsRead); + printf("OutputsWritten: 0x%x\n", newProg->OutputsWritten); + _mesa_print_parameter_list(newProg->Parameters); } +#endif - return stfp->bitmap_program; + /* return results */ + *fpOut = (struct gl_fragment_program *) newProg; + *bitmap_sampler = sampler; } @@ -411,11 +410,16 @@ draw_bitmap_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z, struct st_context *st = st_context(ctx); struct pipe_context *pipe = st->pipe; struct cso_context *cso = st->cso_context; - struct st_fragment_program *stfp; + struct st_fp_varient *fpv; + struct st_fp_varient_key key; GLuint maxSize; GLuint offset; - stfp = combined_bitmap_fragment_program(ctx); + memset(&key, 0, sizeof(key)); + key.st = st; + key.bitmap = GL_TRUE; + + fpv = st_get_fp_varient(st, st->fp, &key); /* As an optimization, Mesa's fragment programs will sometimes get the * primary color from a statevar/constant rather than a varying variable. @@ -428,7 +432,7 @@ draw_bitmap_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z, GLfloat colorSave[4]; COPY_4V(colorSave, ctx->Current.Attrib[VERT_ATTRIB_COLOR0]); COPY_4V(ctx->Current.Attrib[VERT_ATTRIB_COLOR0], color); - st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT); + st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT); COPY_4V(ctx->Current.Attrib[VERT_ATTRIB_COLOR0], colorSave); } @@ -454,7 +458,7 @@ draw_bitmap_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z, cso_set_rasterizer(cso, &st->bitmap.rasterizer); /* fragment shader state: TEX lookup program */ - cso_set_fragment_shader_handle(cso, stfp->driver_shader); + cso_set_fragment_shader_handle(cso, fpv->driver_shader); /* vertex shader state: position + texcoord pass-through */ cso_set_vertex_shader_handle(cso, st->bitmap.vs); @@ -462,21 +466,21 @@ draw_bitmap_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z, /* user samplers, plus our bitmap sampler */ { struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS]; - uint num = MAX2(stfp->bitmap_sampler + 1, st->state.num_samplers); + uint num = MAX2(fpv->bitmap_sampler + 1, st->state.num_samplers); uint i; for (i = 0; i < st->state.num_samplers; i++) { samplers[i] = &st->state.samplers[i]; } - samplers[stfp->bitmap_sampler] = &st->bitmap.samplers[sv->texture->target != PIPE_TEXTURE_RECT]; + samplers[fpv->bitmap_sampler] = &st->bitmap.samplers[sv->texture->target != PIPE_TEXTURE_RECT]; cso_set_samplers(cso, num, (const struct pipe_sampler_state **) samplers); } /* user textures, plus the bitmap texture */ { struct pipe_sampler_view *sampler_views[PIPE_MAX_SAMPLERS]; - uint num = MAX2(stfp->bitmap_sampler + 1, st->state.num_textures); + uint num = MAX2(fpv->bitmap_sampler + 1, st->state.num_textures); memcpy(sampler_views, st->state.sampler_views, sizeof(sampler_views)); - sampler_views[stfp->bitmap_sampler] = sv; + sampler_views[fpv->bitmap_sampler] = sv; cso_set_fragment_sampler_views(cso, num, sampler_views); } diff --git a/src/mesa/state_tracker/st_cb_bitmap.h b/src/mesa/state_tracker/st_cb_bitmap.h index d04b2b67795..2bd63b9b741 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.h +++ b/src/mesa/state_tracker/st_cb_bitmap.h @@ -34,6 +34,8 @@ struct dd_function_table; struct st_context; +struct gl_fragment_program; +struct st_fragment_program; #if FEATURE_drawpix @@ -47,6 +49,12 @@ extern void st_destroy_bitmap(struct st_context *st); extern void +st_make_bitmap_fragment_program(struct st_context *st, + struct gl_fragment_program *fpIn, + struct gl_fragment_program **fpOut, + GLuint *bitmap_sampler); + +extern void st_flush_bitmap_cache(struct st_context *st); /* Flush bitmap cache and release vertex buffer. Needed at end of diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c9786024575..323e7e357ec 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -94,88 +94,61 @@ is_passthrough_program(const struct gl_fragment_program *prog) /** * Make fragment shader for glDraw/CopyPixels. This shader is made * by combining the pixel transfer shader with the user-defined shader. - * \return pointer to Gallium driver fragment shader + * \param fpIn the current/incoming fragment program + * \param fpOut returns the combined fragment program */ -static void * -combined_drawpix_fragment_program(struct gl_context *ctx) +void +st_make_drawpix_fragment_program(struct st_context *st, + struct gl_fragment_program *fpIn, + struct gl_fragment_program **fpOut) { - struct st_context *st = st_context(ctx); - struct st_fragment_program *stfp; + struct gl_program *newProg; - if (st->pixel_xfer.program->serialNo == st->pixel_xfer.xfer_prog_sn - && st->fp->serialNo == st->pixel_xfer.user_prog_sn) { - /* the pixel tranfer program has not changed and the user-defined - * program has not changed, so re-use the combined program. - */ - stfp = st->pixel_xfer.combined_prog; + if (is_passthrough_program(fpIn)) { + newProg = (struct gl_program *) _mesa_clone_fragment_program(st->ctx, + &st->pixel_xfer.program->Base); } else { - /* Concatenate the pixel transfer program with the current user- - * defined program. - */ - if (is_passthrough_program(&st->fp->Base)) { - stfp = (struct st_fragment_program *) - _mesa_clone_fragment_program(ctx, &st->pixel_xfer.program->Base); - } - else { #if 0 - printf("Base program:\n"); - _mesa_print_program(&st->fp->Base.Base); - printf("DrawPix program:\n"); - _mesa_print_program(&st->pixel_xfer.program->Base.Base); + /* debug */ + printf("Base program:\n"); + _mesa_print_program(&fpIn->Base); + printf("DrawPix program:\n"); + _mesa_print_program(&st->pixel_xfer.program->Base.Base); #endif - stfp = (struct st_fragment_program *) - _mesa_combine_programs(ctx, - &st->pixel_xfer.program->Base.Base, - &st->fp->Base.Base); - } + newProg = _mesa_combine_programs(st->ctx, + &st->pixel_xfer.program->Base.Base, + &fpIn->Base); + } #if 0 - { - struct gl_program *p = &stfp->Base.Base; - printf("Combined DrawPixels program:\n"); - _mesa_print_program(p); - printf("InputsRead: 0x%x\n", p->InputsRead); - printf("OutputsWritten: 0x%x\n", p->OutputsWritten); - _mesa_print_parameter_list(p->Parameters); - } + /* debug */ + printf("Combined DrawPixels program:\n"); + _mesa_print_program(newProg); + printf("InputsRead: 0x%x\n", newProg->InputsRead); + printf("OutputsWritten: 0x%x\n", newProg->OutputsWritten); + _mesa_print_parameter_list(newProg->Parameters); #endif - /* translate to TGSI tokens */ - st_translate_fragment_program(st, stfp); - - /* save new program, update serial numbers */ - st->pixel_xfer.xfer_prog_sn = st->pixel_xfer.program->serialNo; - st->pixel_xfer.user_prog_sn = st->fp->serialNo; - st->pixel_xfer.combined_prog_sn = stfp->serialNo; - /* can't reference new program directly, already have a reference on it */ - st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL); - st->pixel_xfer.combined_prog = stfp; - } - - /* Ideally we'd have updated the pipe constants during the normal - * st/atom mechanism. But we can't since this is specific to glDrawPixels. - */ - st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT); - - return stfp->driver_shader; + *fpOut = (struct gl_fragment_program *) newProg; } /** - * Create fragment shader that does a TEX() instruction to get a Z and/or + * Create fragment program that does a TEX() instruction to get a Z and/or * stencil value value, then writes to FRAG_RESULT_DEPTH/FRAG_RESULT_STENCIL. * Used for glDrawPixels(GL_DEPTH_COMPONENT / GL_STENCIL_INDEX). * Pass fragment color through as-is. - * \return pointer to the Gallium driver fragment shader + * \return pointer to the gl_fragment program */ -static void * -make_fragment_shader_z_stencil(struct st_context *st, GLboolean write_depth, - GLboolean write_stencil) +struct gl_fragment_program * +st_make_drawpix_z_stencil_program(struct st_context *st, + GLboolean write_depth, + GLboolean write_stencil) { struct gl_context *ctx = st->ctx; struct gl_program *p; - struct st_fragment_program *stp; + struct gl_fragment_program *fp; GLuint ic = 0; const GLuint shaderIndex = write_depth * 2 + write_stencil; @@ -183,7 +156,7 @@ make_fragment_shader_z_stencil(struct st_context *st, GLboolean write_depth, if (st->drawpix.shaders[shaderIndex]) { /* already have the proper shader */ - return st->drawpix.shaders[shaderIndex]->driver_shader; + return st->drawpix.shaders[shaderIndex]; } /* @@ -245,18 +218,15 @@ make_fragment_shader_z_stencil(struct st_context *st, GLboolean write_depth, if (write_stencil) p->SamplersUsed |= 1 << 1; - stp = st_fragment_program((struct gl_fragment_program *) p); + fp = (struct gl_fragment_program *) p; /* save the new shader */ - st->drawpix.shaders[shaderIndex] = stp; + st->drawpix.shaders[shaderIndex] = fp; - st_translate_fragment_program(st, stp); - - return stp->driver_shader; + return fp; } - /** * Create a simple vertex shader that just passes through the * vertex position and texcoord (and optionally, color). @@ -873,6 +843,61 @@ draw_stencil_pixels(struct gl_context *ctx, GLint x, GLint y, /** + * Get fragment program variant for a glDrawPixels or glCopyPixels + * command for RGBA data. + */ +static struct st_fp_varient * +get_color_fp_varient(struct st_context *st) +{ + struct gl_context *ctx = st->ctx; + struct st_fp_varient_key key; + struct st_fp_varient *fpv; + + memset(&key, 0, sizeof(key)); + + key.st = st; + key.drawpixels = 1; + key.scaleAndBias = (ctx->Pixel.RedBias != 0.0 || + ctx->Pixel.RedScale != 1.0 || + ctx->Pixel.GreenBias != 0.0 || + ctx->Pixel.GreenScale != 1.0 || + ctx->Pixel.BlueBias != 0.0 || + ctx->Pixel.BlueScale != 1.0 || + ctx->Pixel.AlphaBias != 0.0 || + ctx->Pixel.AlphaScale != 1.0); + key.pixelMaps = ctx->Pixel.MapColorFlag; + + fpv = st_get_fp_varient(st, st->fp, &key); + + return fpv; +} + + +/** + * Get fragment program variant for a glDrawPixels or glCopyPixels + * command for depth/stencil data. + */ +static struct st_fp_varient * +get_depth_stencil_fp_varient(struct st_context *st, GLboolean write_depth, + GLboolean write_stencil) +{ + struct st_fp_varient_key key; + struct st_fp_varient *fpv; + + memset(&key, 0, sizeof(key)); + + key.st = st; + key.drawpixels = 1; + key.drawpixels_z = write_depth; + key.drawpixels_stencil = write_stencil; + + fpv = st_get_fp_varient(st, st->fp, &key); + + return fpv; +} + + +/** * Called via ctx->Driver.DrawPixels() */ static void @@ -889,6 +914,7 @@ st_DrawPixels(struct gl_context *ctx, GLint x, GLint y, struct pipe_sampler_view *sv[2]; int num_sampler_view = 1; enum pipe_format stencil_format = PIPE_FORMAT_NONE; + struct st_fp_varient *fpv; if (format == GL_DEPTH_STENCIL) write_stencil = write_depth = GL_TRUE; @@ -921,14 +947,25 @@ st_DrawPixels(struct gl_context *ctx, GLint x, GLint y, st_validate_state(st); + /* + * Get vertex/fragment shaders + */ if (write_depth || write_stencil) { - driver_fp = make_fragment_shader_z_stencil(st, write_depth, write_stencil); + fpv = get_depth_stencil_fp_varient(st, write_depth, write_stencil); + + driver_fp = fpv->driver_shader; + driver_vp = make_passthrough_vertex_shader(st, GL_TRUE); + color = ctx->Current.RasterColor; } else { - driver_fp = combined_drawpix_fragment_program(ctx); + fpv = get_color_fp_varient(st); + + driver_fp = fpv->driver_shader; + driver_vp = make_passthrough_vertex_shader(st, GL_FALSE); + color = NULL; if (st->pixel_xfer.pixelmap_enabled) { sv[1] = st->pixel_xfer.pixelmap_sampler_view; @@ -936,6 +973,9 @@ st_DrawPixels(struct gl_context *ctx, GLint x, GLint y, } } + /* update fragment program constants */ + st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT); + /* draw with textured quad */ { struct pipe_resource *pt @@ -1116,6 +1156,7 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy, GLint readX, readY, readW, readH; GLuint sample_count; struct gl_pixelstore_attrib pack = ctx->DefaultPacking; + struct st_fp_varient *fpv; st_validate_state(st); @@ -1125,11 +1166,18 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy, return; } + /* + * Get vertex/fragment shaders + */ if (type == GL_COLOR) { rbRead = st_get_color_read_renderbuffer(ctx); color = NULL; - driver_fp = combined_drawpix_fragment_program(ctx); + + fpv = get_color_fp_varient(st); + driver_fp = fpv->driver_shader; + driver_vp = make_passthrough_vertex_shader(st, GL_FALSE); + if (st->pixel_xfer.pixelmap_enabled) { sv[1] = st->pixel_xfer.pixelmap_sampler_view; num_sampler_view++; @@ -1139,10 +1187,17 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy, assert(type == GL_DEPTH); rbRead = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer); color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; - driver_fp = make_fragment_shader_z_stencil(st, GL_TRUE, GL_FALSE); + + fpv = get_depth_stencil_fp_varient(st, GL_TRUE, GL_FALSE); + driver_fp = fpv->driver_shader; + driver_vp = make_passthrough_vertex_shader(st, GL_TRUE); } + /* update fragment program constants */ + st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT); + + if (rbRead->Base.Wrapped) rbRead = st_renderbuffer(rbRead->Base.Wrapped); @@ -1294,7 +1349,7 @@ st_destroy_drawpix(struct st_context *st) for (i = 0; i < Elements(st->drawpix.shaders); i++) { if (st->drawpix.shaders[i]) - st_reference_fragprog(st, &st->drawpix.shaders[i], NULL); + _mesa_reference_fragprog(st->ctx, &st->drawpix.shaders[i], NULL); } st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h index 575f169e08e..8f73e626240 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.h +++ b/src/mesa/state_tracker/st_cb_drawpixels.h @@ -42,6 +42,16 @@ extern void st_init_drawpixels_functions(struct dd_function_table *functions); extern void st_destroy_drawpix(struct st_context *st); +extern void +st_make_drawpix_fragment_program(struct st_context *st, + struct gl_fragment_program *fpIn, + struct gl_fragment_program **fpOut); + +extern struct gl_fragment_program * +st_make_drawpix_z_stencil_program(struct st_context *st, + GLboolean write_depth, + GLboolean write_stencil); + #else static INLINE void diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 4d83fcc6ccb..3e1b709d25b 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -46,8 +46,6 @@ #include "st_cb_program.h" -static GLuint SerialNo = 1; - /** * Called via ctx->Driver.BindProgram() to bind an ARB vertex or @@ -100,8 +98,6 @@ static struct gl_program *st_new_program( struct gl_context *ctx, case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *prog = ST_CALLOC_STRUCT(st_vertex_program); - prog->serialNo = SerialNo++; - return _mesa_init_vertex_program( ctx, &prog->Base, target, @@ -112,8 +108,6 @@ static struct gl_program *st_new_program( struct gl_context *ctx, case GL_FRAGMENT_PROGRAM_NV: { struct st_fragment_program *prog = ST_CALLOC_STRUCT(st_fragment_program); - prog->serialNo = SerialNo++; - return _mesa_init_fragment_program( ctx, &prog->Base, target, @@ -123,8 +117,6 @@ static struct gl_program *st_new_program( struct gl_context *ctx, case MESA_GEOMETRY_PROGRAM: { struct st_geometry_program *prog = ST_CALLOC_STRUCT(st_geometry_program); - prog->serialNo = SerialNo++; - return _mesa_init_geometry_program( ctx, &prog->Base, target, @@ -152,12 +144,10 @@ st_delete_program(struct gl_context *ctx, struct gl_program *prog) break; case MESA_GEOMETRY_PROGRAM: { - struct st_geometry_program *stgp = (struct st_geometry_program *) prog; + struct st_geometry_program *stgp = + (struct st_geometry_program *) prog; - if (stgp->driver_shader) { - cso_delete_geometry_shader(st->cso_context, stgp->driver_shader); - stgp->driver_shader = NULL; - } + st_gp_release_varients(st, stgp); if (stgp->tgsi.tokens) { st_free_tokens((void *) stgp->tgsi.tokens); @@ -167,23 +157,15 @@ st_delete_program(struct gl_context *ctx, struct gl_program *prog) break; case GL_FRAGMENT_PROGRAM_ARB: { - struct st_fragment_program *stfp = (struct st_fragment_program *) prog; + struct st_fragment_program *stfp = + (struct st_fragment_program *) prog; - if (stfp->driver_shader) { - cso_delete_fragment_shader(st->cso_context, stfp->driver_shader); - stfp->driver_shader = NULL; - } + st_fp_release_varients(st, stfp); if (stfp->tgsi.tokens) { st_free_tokens(stfp->tgsi.tokens); stfp->tgsi.tokens = NULL; } - - if (stfp->bitmap_program) { - struct gl_program *prg = &stfp->bitmap_program->Base.Base; - _mesa_reference_program(ctx, &prg, NULL); - stfp->bitmap_program = NULL; - } } break; default: @@ -212,12 +194,7 @@ static GLboolean st_program_string_notify( struct gl_context *ctx, if (target == GL_FRAGMENT_PROGRAM_ARB) { struct st_fragment_program *stfp = (struct st_fragment_program *) prog; - stfp->serialNo++; - - if (stfp->driver_shader) { - cso_delete_fragment_shader(st->cso_context, stfp->driver_shader); - stfp->driver_shader = NULL; - } + st_fp_release_varients(st, stfp); if (stfp->tgsi.tokens) { st_free_tokens(stfp->tgsi.tokens); @@ -230,12 +207,7 @@ static GLboolean st_program_string_notify( struct gl_context *ctx, else if (target == MESA_GEOMETRY_PROGRAM) { struct st_geometry_program *stgp = (struct st_geometry_program *) prog; - stgp->serialNo++; - - if (stgp->driver_shader) { - cso_delete_geometry_shader(st->cso_context, stgp->driver_shader); - stgp->driver_shader = NULL; - } + st_gp_release_varients(st, stgp); if (stgp->tgsi.tokens) { st_free_tokens((void *) stgp->tgsi.tokens); @@ -248,8 +220,6 @@ static GLboolean st_program_string_notify( struct gl_context *ctx, else if (target == GL_VERTEX_PROGRAM_ARB) { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; - stvp->serialNo++; - st_vp_release_varients( st, stvp ); if (st->vp == stvp) diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 6ec9c699a26..21bb91f47a8 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -252,6 +252,8 @@ void st_destroy_context( struct st_context *st ) _vbo_DestroyContext(st->ctx); + st_destroy_program_variants(st); + _mesa_free_context_data(ctx); st_destroy_context_priv(st); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index d342c0cff1e..26565225e3b 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -131,6 +131,8 @@ struct st_context struct st_geometry_program *gp; /**< Currently bound geometry program */ struct st_vp_varient *vp_varient; + struct st_fp_varient *fp_varient; + struct st_gp_varient *gp_varient; struct gl_texture_object *default_texture; @@ -160,7 +162,7 @@ struct st_context /** for glDraw/CopyPixels */ struct { - struct st_fragment_program *shaders[4]; + struct gl_fragment_program *shaders[4]; void *vert_shaders[2]; /**< ureg shaders */ } drawpix; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index aae2913c202..202f7cb711e 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -32,7 +32,9 @@ #include "main/imports.h" +#include "main/hash.h" #include "main/mtypes.h" +#include "program/prog_parameter.h" #include "program/prog_print.h" #include "program/programopt.h" @@ -44,6 +46,8 @@ #include "tgsi/tgsi_ureg.h" #include "st_debug.h" +#include "st_cb_bitmap.h" +#include "st_cb_drawpixels.h" #include "st_context.h" #include "st_program.h" #include "st_mesa_to_tgsi.h" @@ -52,6 +56,29 @@ /** + * Delete a vertex program varient. Note the caller must unlink + * the varient from the linked list. + */ +static void +delete_vp_varient(struct st_context *st, struct st_vp_varient *vpv) +{ + if (vpv->driver_shader) + cso_delete_vertex_shader(st->cso_context, vpv->driver_shader); + +#if FEATURE_feedback || FEATURE_rastpos + if (vpv->draw_shader) + draw_delete_vertex_shader( st->draw, vpv->draw_shader ); +#endif + + if (vpv->tgsi.tokens) + st_free_tokens(vpv->tgsi.tokens); + + FREE( vpv ); +} + + + +/** * Clean out any old compilations: */ void @@ -62,24 +89,76 @@ st_vp_release_varients( struct st_context *st, for (vpv = stvp->varients; vpv; ) { struct st_vp_varient *next = vpv->next; + delete_vp_varient(st, vpv); + vpv = next; + } - if (vpv->driver_shader) - cso_delete_vertex_shader(st->cso_context, vpv->driver_shader); - -#if FEATURE_feedback || FEATURE_rastpos - if (vpv->draw_shader) - draw_delete_vertex_shader( st->draw, vpv->draw_shader ); -#endif + stvp->varients = NULL; +} + + + +/** + * Delete a fragment program varient. Note the caller must unlink + * the varient from the linked list. + */ +static void +delete_fp_varient(struct st_context *st, struct st_fp_varient *fpv) +{ + if (fpv->driver_shader) + cso_delete_fragment_shader(st->cso_context, fpv->driver_shader); - if (vpv->tgsi.tokens) - st_free_tokens(vpv->tgsi.tokens); + FREE(fpv); +} + + +/** + * Free all varients of a fragment program. + */ +void +st_fp_release_varients(struct st_context *st, struct st_fragment_program *stfp) +{ + struct st_fp_varient *fpv; + + for (fpv = stfp->varients; fpv; ) { + struct st_fp_varient *next = fpv->next; + delete_fp_varient(st, fpv); + fpv = next; + } + + stfp->varients = NULL; +} + + +/** + * Delete a geometry program varient. Note the caller must unlink + * the varient from the linked list. + */ +static void +delete_gp_varient(struct st_context *st, struct st_gp_varient *gpv) +{ + if (gpv->driver_shader) + cso_delete_geometry_shader(st->cso_context, gpv->driver_shader); - FREE( vpv ); + FREE(gpv); +} - vpv = next; + +/** + * Free all varients of a geometry program. + */ +void +st_gp_release_varients(struct st_context *st, struct st_geometry_program *stgp) +{ + struct st_gp_varient *gpv; + + for (gpv = stgp->varients; gpv; ) { + struct st_gp_varient *next = gpv->next; + delete_gp_varient(st, gpv); + gpv = next; } - stvp->varients = NULL; + stgp->varients = NULL; } @@ -92,7 +171,7 @@ st_vp_release_varients( struct st_context *st, * \param tokensOut destination for TGSI tokens * \return pointer to cached pipe_shader object. */ -void +static void st_prepare_vertex_program(struct st_context *st, struct st_vertex_program *stvp) { @@ -196,7 +275,10 @@ st_prepare_vertex_program(struct st_context *st, } -struct st_vp_varient * +/** + * Translate a vertex program to create a new varient. + */ +static struct st_vp_varient * st_translate_vertex_program(struct st_context *st, struct st_vertex_program *stvp, const struct st_vp_varient_key *key) @@ -207,6 +289,8 @@ st_translate_vertex_program(struct st_context *st, enum pipe_error error; unsigned num_outputs; + st_prepare_vertex_program( st, stvp ); + _mesa_remove_output_reads(&stvp->Base.Base, PROGRAM_OUTPUT); _mesa_remove_output_reads(&stvp->Base.Base, PROGRAM_VARYING); @@ -277,201 +361,310 @@ fail: } +/** + * Find/create a vertex program varient. + */ +struct st_vp_varient * +st_get_vp_varient(struct st_context *st, + struct st_vertex_program *stvp, + const struct st_vp_varient_key *key) +{ + struct st_vp_varient *vpv; + + /* Search for existing varient */ + for (vpv = stvp->varients; vpv; vpv = vpv->next) { + if (memcmp(&vpv->key, key, sizeof(*key)) == 0) { + break; + } + } + + if (!vpv) { + /* create now */ + vpv = st_translate_vertex_program(st, stvp, key); + if (vpv) { + /* insert into list */ + vpv->next = stvp->varients; + stvp->varients = vpv; + } + } + + return vpv; +} + /** - * Translate a Mesa fragment shader into a TGSI shader. - * \return pointer to cached pipe_shader object. + * Translate a Mesa fragment shader into a TGSI shader using extra info in + * the key. + * \return new fragment program variant */ -void +static struct st_fp_varient * st_translate_fragment_program(struct st_context *st, - struct st_fragment_program *stfp ) + struct st_fragment_program *stfp, + const struct st_fp_varient_key *key) { struct pipe_context *pipe = st->pipe; - GLuint outputMapping[FRAG_RESULT_MAX]; - GLuint inputMapping[FRAG_ATTRIB_MAX]; - GLuint interpMode[PIPE_MAX_SHADER_INPUTS]; /* XXX size? */ - GLuint attr; - enum pipe_error error; - const GLbitfield inputsRead = stfp->Base.Base.InputsRead; - struct ureg_program *ureg; + struct st_fp_varient *varient = CALLOC_STRUCT(st_fp_varient); - ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; - ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; - uint fs_num_inputs = 0; + if (!varient) + return NULL; - ubyte fs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; - ubyte fs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; - uint fs_num_outputs = 0; + assert(!(key->bitmap && key->drawpixels)); - _mesa_remove_output_reads(&stfp->Base.Base, PROGRAM_OUTPUT); + if (key->bitmap) { + /* glBitmap drawing */ + struct gl_fragment_program *fp; - /* - * Convert Mesa program inputs to TGSI input register semantics. - */ - for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) { - if (inputsRead & (1 << attr)) { - const GLuint slot = fs_num_inputs++; + st_make_bitmap_fragment_program(st, &stfp->Base, + &fp, &varient->bitmap_sampler); - inputMapping[attr] = slot; + varient->parameters = _mesa_clone_parameter_list(fp->Base.Parameters); + stfp = (struct st_fragment_program *) fp; + } + else if (key->drawpixels) { + /* glDrawPixels drawing */ + struct gl_fragment_program *fp; - switch (attr) { - case FRAG_ATTRIB_WPOS: - input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - input_semantic_index[slot] = 0; - interpMode[slot] = TGSI_INTERPOLATE_LINEAR; - break; - case FRAG_ATTRIB_COL0: - input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - input_semantic_index[slot] = 0; - interpMode[slot] = TGSI_INTERPOLATE_LINEAR; - break; - case FRAG_ATTRIB_COL1: - input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - input_semantic_index[slot] = 1; - interpMode[slot] = TGSI_INTERPOLATE_LINEAR; - break; - case FRAG_ATTRIB_FOGC: - input_semantic_name[slot] = TGSI_SEMANTIC_FOG; - input_semantic_index[slot] = 0; - interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; - break; - case FRAG_ATTRIB_FACE: - input_semantic_name[slot] = TGSI_SEMANTIC_FACE; - input_semantic_index[slot] = 0; - interpMode[slot] = TGSI_INTERPOLATE_CONSTANT; - 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_PNTC: - case FRAG_ATTRIB_TEX0: - case FRAG_ATTRIB_TEX1: - case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: - case FRAG_ATTRIB_TEX4: - case FRAG_ATTRIB_TEX5: - case FRAG_ATTRIB_TEX6: - case FRAG_ATTRIB_TEX7: - case FRAG_ATTRIB_VAR0: - default: - /* Actually, let's try and zero-base this just for - * readability of the generated TGSI. - */ - assert(attr >= FRAG_ATTRIB_TEX0); - input_semantic_index[slot] = (attr - FRAG_ATTRIB_TEX0); - input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - if (attr == FRAG_ATTRIB_PNTC) - interpMode[slot] = TGSI_INTERPOLATE_LINEAR; - else - interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; - break; - } + if (key->drawpixels_z || key->drawpixels_stencil) { + fp = st_make_drawpix_z_stencil_program(st, key->drawpixels_z, + key->drawpixels_stencil); } else { - inputMapping[attr] = -1; + /* RGBA */ + st_make_drawpix_fragment_program(st, &stfp->Base, &fp); + varient->parameters = _mesa_clone_parameter_list(fp->Base.Parameters); } + stfp = (struct st_fragment_program *) fp; } - /* - * Semantics and mapping for outputs - */ - { - uint numColors = 0; - GLbitfield64 outputsWritten = stfp->Base.Base.OutputsWritten; - - /* if z is written, emit that first */ - if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) { - fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_POSITION; - fs_output_semantic_index[fs_num_outputs] = 0; - outputMapping[FRAG_RESULT_DEPTH] = fs_num_outputs; - fs_num_outputs++; - outputsWritten &= ~(1 << FRAG_RESULT_DEPTH); - } + if (!stfp->tgsi.tokens) { + /* need to translate Mesa instructions to TGSI now */ + GLuint outputMapping[FRAG_RESULT_MAX]; + GLuint inputMapping[FRAG_ATTRIB_MAX]; + GLuint interpMode[PIPE_MAX_SHADER_INPUTS]; /* XXX size? */ + GLuint attr; + enum pipe_error error; + const GLbitfield inputsRead = stfp->Base.Base.InputsRead; + struct ureg_program *ureg; - if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_STENCIL)) { - fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_STENCIL; - fs_output_semantic_index[fs_num_outputs] = 0; - outputMapping[FRAG_RESULT_STENCIL] = fs_num_outputs; - fs_num_outputs++; - outputsWritten &= ~(1 << FRAG_RESULT_STENCIL); - } + ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; + ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + uint fs_num_inputs = 0; + + ubyte fs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; + ubyte fs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; + uint fs_num_outputs = 0; + + + _mesa_remove_output_reads(&stfp->Base.Base, PROGRAM_OUTPUT); + + /* + * Convert Mesa program inputs to TGSI input register semantics. + */ + for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) { + if (inputsRead & (1 << attr)) { + const GLuint slot = fs_num_inputs++; + + inputMapping[attr] = slot; - /* handle remaning outputs (color) */ - for (attr = 0; attr < FRAG_RESULT_MAX; attr++) { - if (outputsWritten & BITFIELD64_BIT(attr)) { switch (attr) { - case FRAG_RESULT_DEPTH: - case FRAG_RESULT_STENCIL: - /* handled above */ - assert(0); + case FRAG_ATTRIB_WPOS: + input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + input_semantic_index[slot] = 0; + interpMode[slot] = TGSI_INTERPOLATE_LINEAR; + break; + case FRAG_ATTRIB_COL0: + input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + input_semantic_index[slot] = 0; + interpMode[slot] = TGSI_INTERPOLATE_LINEAR; + break; + case FRAG_ATTRIB_COL1: + input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + input_semantic_index[slot] = 1; + interpMode[slot] = TGSI_INTERPOLATE_LINEAR; + break; + case FRAG_ATTRIB_FOGC: + input_semantic_name[slot] = TGSI_SEMANTIC_FOG; + input_semantic_index[slot] = 0; + interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; break; + case FRAG_ATTRIB_FACE: + input_semantic_name[slot] = TGSI_SEMANTIC_FACE; + input_semantic_index[slot] = 0; + interpMode[slot] = TGSI_INTERPOLATE_CONSTANT; + 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_PNTC: + case FRAG_ATTRIB_TEX0: + case FRAG_ATTRIB_TEX1: + case FRAG_ATTRIB_TEX2: + case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX4: + case FRAG_ATTRIB_TEX5: + case FRAG_ATTRIB_TEX6: + case FRAG_ATTRIB_TEX7: + case FRAG_ATTRIB_VAR0: default: - assert(attr == FRAG_RESULT_COLOR || - (FRAG_RESULT_DATA0 <= attr && attr < FRAG_RESULT_MAX)); - fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_COLOR; - fs_output_semantic_index[fs_num_outputs] = numColors; - outputMapping[attr] = fs_num_outputs; - numColors++; + /* Actually, let's try and zero-base this just for + * readability of the generated TGSI. + */ + assert(attr >= FRAG_ATTRIB_TEX0); + input_semantic_index[slot] = (attr - FRAG_ATTRIB_TEX0); + input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + if (attr == FRAG_ATTRIB_PNTC) + interpMode[slot] = TGSI_INTERPOLATE_LINEAR; + else + interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; break; } + } + else { + inputMapping[attr] = -1; + } + } + + /* + * Semantics and mapping for outputs + */ + { + uint numColors = 0; + GLbitfield64 outputsWritten = stfp->Base.Base.OutputsWritten; + + /* if z is written, emit that first */ + if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) { + fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_POSITION; + fs_output_semantic_index[fs_num_outputs] = 0; + outputMapping[FRAG_RESULT_DEPTH] = fs_num_outputs; + fs_num_outputs++; + outputsWritten &= ~(1 << FRAG_RESULT_DEPTH); + } + if (outputsWritten & BITFIELD64_BIT(FRAG_RESULT_STENCIL)) { + fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_STENCIL; + fs_output_semantic_index[fs_num_outputs] = 0; + outputMapping[FRAG_RESULT_STENCIL] = fs_num_outputs; fs_num_outputs++; + outputsWritten &= ~(1 << FRAG_RESULT_STENCIL); + } + + /* handle remaning outputs (color) */ + for (attr = 0; attr < FRAG_RESULT_MAX; attr++) { + if (outputsWritten & BITFIELD64_BIT(attr)) { + switch (attr) { + case FRAG_RESULT_DEPTH: + case FRAG_RESULT_STENCIL: + /* handled above */ + assert(0); + break; + default: + assert(attr == FRAG_RESULT_COLOR || + (FRAG_RESULT_DATA0 <= attr && attr < FRAG_RESULT_MAX)); + fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_COLOR; + fs_output_semantic_index[fs_num_outputs] = numColors; + outputMapping[attr] = fs_num_outputs; + numColors++; + break; + } + + fs_num_outputs++; + } } } - } - ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); - if (ureg == NULL) - return; + ureg = ureg_create( TGSI_PROCESSOR_FRAGMENT ); + if (ureg == NULL) + return NULL; - if (ST_DEBUG & DEBUG_MESA) { - _mesa_print_program(&stfp->Base.Base); - _mesa_print_program_parameters(st->ctx, &stfp->Base.Base); - debug_printf("\n"); - } + if (ST_DEBUG & DEBUG_MESA) { + _mesa_print_program(&stfp->Base.Base); + _mesa_print_program_parameters(st->ctx, &stfp->Base.Base); + debug_printf("\n"); + } - error = - st_translate_mesa_program(st->ctx, - TGSI_PROCESSOR_FRAGMENT, - ureg, - &stfp->Base.Base, - /* inputs */ - fs_num_inputs, - inputMapping, - input_semantic_name, - input_semantic_index, - interpMode, - /* outputs */ - fs_num_outputs, - outputMapping, - fs_output_semantic_name, - fs_output_semantic_index, FALSE ); + error = st_translate_mesa_program(st->ctx, + TGSI_PROCESSOR_FRAGMENT, + ureg, + &stfp->Base.Base, + /* inputs */ + fs_num_inputs, + inputMapping, + input_semantic_name, + input_semantic_index, + interpMode, + /* outputs */ + fs_num_outputs, + outputMapping, + fs_output_semantic_name, + fs_output_semantic_index, FALSE ); + + stfp->tgsi.tokens = ureg_get_tokens( ureg, NULL ); + ureg_destroy( ureg ); + } - stfp->tgsi.tokens = ureg_get_tokens( ureg, NULL ); - ureg_destroy( ureg ); - stfp->driver_shader = pipe->create_fs_state(pipe, &stfp->tgsi); + /* fill in varient */ + varient->driver_shader = pipe->create_fs_state(pipe, &stfp->tgsi); + varient->key = *key; if (ST_DEBUG & DEBUG_TGSI) { tgsi_dump( stfp->tgsi.tokens, 0/*TGSI_DUMP_VERBOSE*/ ); debug_printf("\n"); } + + return varient; } -void + +/** + * Translate fragment program if needed. + */ +struct st_fp_varient * +st_get_fp_varient(struct st_context *st, + struct st_fragment_program *stfp, + const struct st_fp_varient_key *key) +{ + struct st_fp_varient *fpv; + + /* Search for existing varient */ + for (fpv = stfp->varients; fpv; fpv = fpv->next) { + if (memcmp(&fpv->key, key, sizeof(*key)) == 0) { + break; + } + } + + if (!fpv) { + /* create new */ + fpv = st_translate_fragment_program(st, stfp, key); + if (fpv) { + /* insert into list */ + fpv->next = stfp->varients; + stfp->varients = fpv; + } + } + + return fpv; +} + + +/** + * Translate a geometry program to create a new varient. + */ +static struct st_gp_varient * st_translate_geometry_program(struct st_context *st, - struct st_geometry_program *stgp) + struct st_geometry_program *stgp, + const struct st_gp_varient_key *key) { GLuint inputMapping[GEOM_ATTRIB_MAX]; GLuint outputMapping[GEOM_RESULT_MAX]; @@ -494,12 +687,19 @@ st_translate_geometry_program(struct st_context *st, GLuint maxSlot = 0; struct ureg_program *ureg; + struct st_gp_varient *gpv; + + gpv = CALLOC_STRUCT(st_gp_varient); + if (!gpv) + return NULL; + _mesa_remove_output_reads(&stgp->Base.Base, PROGRAM_OUTPUT); _mesa_remove_output_reads(&stgp->Base.Base, PROGRAM_VARYING); ureg = ureg_create( TGSI_PROCESSOR_GEOMETRY ); if (ureg == NULL) { - return; + FREE(gpv); + return NULL; } /* which vertex output goes to the first geometry input */ @@ -529,7 +729,7 @@ st_translate_geometry_program(struct st_context *st, } else ++gs_builtin_inputs; -#if 1 +#if 0 debug_printf("input map at %d = %d\n", slot + gs_array_offset, stgp->input_map[slot + gs_array_offset]); #endif @@ -671,37 +871,35 @@ st_translate_geometry_program(struct st_context *st, st_free_tokens(stgp->tgsi.tokens); stgp->tgsi.tokens = NULL; } - if (stgp->driver_shader) { - cso_delete_geometry_shader(st->cso_context, stgp->driver_shader); - stgp->driver_shader = NULL; - } ureg_property_gs_input_prim(ureg, stgp->Base.InputType); ureg_property_gs_output_prim(ureg, stgp->Base.OutputType); ureg_property_gs_max_vertices(ureg, stgp->Base.VerticesOut); - error = st_translate_mesa_program(st->ctx, - TGSI_PROCESSOR_GEOMETRY, - ureg, - &stgp->Base.Base, - /* inputs */ - gs_num_inputs, - inputMapping, - stgp->input_semantic_name, - stgp->input_semantic_index, - NULL, - /* outputs */ - gs_num_outputs, - outputMapping, - gs_output_semantic_name, - gs_output_semantic_index, - FALSE); - + error = st_translate_mesa_program(st->ctx, + TGSI_PROCESSOR_GEOMETRY, + ureg, + &stgp->Base.Base, + /* inputs */ + gs_num_inputs, + inputMapping, + stgp->input_semantic_name, + stgp->input_semantic_index, + NULL, + /* outputs */ + gs_num_outputs, + outputMapping, + gs_output_semantic_name, + gs_output_semantic_index, + FALSE); stgp->num_inputs = gs_num_inputs; stgp->tgsi.tokens = ureg_get_tokens( ureg, NULL ); ureg_destroy( ureg ); - stgp->driver_shader = pipe->create_gs_state(pipe, &stgp->tgsi); + + /* fill in new varient */ + gpv->driver_shader = pipe->create_gs_state(pipe, &stgp->tgsi); + gpv->key = *key; if ((ST_DEBUG & DEBUG_TGSI) && (ST_DEBUG & DEBUG_MESA)) { _mesa_print_program(&stgp->Base.Base); @@ -712,8 +910,44 @@ st_translate_geometry_program(struct st_context *st, tgsi_dump(stgp->tgsi.tokens, 0); debug_printf("\n"); } + + return gpv; +} + + +/** + * Get/create geometry program variant. + */ +struct st_gp_varient * +st_get_gp_varient(struct st_context *st, + struct st_geometry_program *stgp, + const struct st_gp_varient_key *key) +{ + struct st_gp_varient *gpv; + + /* Search for existing varient */ + for (gpv = stgp->varients; gpv; gpv = gpv->next) { + if (memcmp(&gpv->key, key, sizeof(*key)) == 0) { + break; + } + } + + if (!gpv) { + /* create new */ + gpv = st_translate_geometry_program(st, stgp, key); + if (gpv) { + /* insert into list */ + gpv->next = stgp->varients; + stgp->varients = gpv; + } + } + + return gpv; } + + + /** * Debug- print current shader text */ @@ -759,3 +993,134 @@ st_print_shaders(struct gl_context *ctx) } } } + + +/** + * Vert/Geom/Frag programs have per-context variants. Free all the + * variants attached to the given program which match the given context. + */ +static void +destroy_program_variants(struct st_context *st, struct gl_program *program) +{ + if (!program) + return; + + switch (program->Target) { + case GL_VERTEX_PROGRAM_ARB: + { + struct st_vertex_program *stvp = (struct st_vertex_program *) program; + struct st_vp_varient *vpv, **prev = &stvp->varients; + + for (vpv = stvp->varients; vpv; ) { + struct st_vp_varient *next = vpv->next; + if (vpv->key.st == st) { + /* unlink from list */ + *prev = next; + /* destroy this variant */ + delete_vp_varient(st, vpv); + } + else { + prev = &vpv; + } + vpv = next; + } + } + break; + case GL_FRAGMENT_PROGRAM_ARB: + { + struct st_fragment_program *stfp = + (struct st_fragment_program *) program; + struct st_fp_varient *fpv, **prev = &stfp->varients; + + for (fpv = stfp->varients; fpv; ) { + struct st_fp_varient *next = fpv->next; + if (fpv->key.st == st) { + /* unlink from list */ + *prev = next; + /* destroy this variant */ + delete_fp_varient(st, fpv); + } + else { + prev = &fpv; + } + fpv = next; + } + } + break; + default: + _mesa_problem(NULL, "Unexpected program target in " + "destroy_program_variants_cb()"); + } +} + + +/** + * Callback for _mesa_HashWalk. Free all the shader's program variants + * which match the given context. + */ +static void +destroy_shader_program_variants_cb(GLuint key, void *data, void *userData) +{ + struct st_context *st = (struct st_context *) userData; + struct gl_shader *shader = (struct gl_shader *) data; + + switch (shader->Type) { + case GL_SHADER_PROGRAM_MESA: + { + struct gl_shader_program *shProg = (struct gl_shader_program *) data; + GLuint i; + + for (i = 0; i < shProg->NumShaders; i++) { + destroy_program_variants(st, shProg->Shaders[i]->Program); + } + + destroy_program_variants(st, (struct gl_program *) + shProg->VertexProgram); + destroy_program_variants(st, (struct gl_program *) + shProg->FragmentProgram); + destroy_program_variants(st, (struct gl_program *) + shProg->GeometryProgram); + } + break; + case GL_VERTEX_SHADER: + case GL_FRAGMENT_SHADER: + case GL_GEOMETRY_SHADER: + { + destroy_program_variants(st, shader->Program); + } + break; + default: + assert(0); + } +} + + +/** + * Callback for _mesa_HashWalk. Free all the program variants which match + * the given context. + */ +static void +destroy_program_variants_cb(GLuint key, void *data, void *userData) +{ + struct st_context *st = (struct st_context *) userData; + struct gl_program *program = (struct gl_program *) data; + destroy_program_variants(st, program); +} + + +/** + * Walk over all shaders and programs to delete any variants which + * belong to the given context. + * This is called during context tear-down. + */ +void +st_destroy_program_variants(struct st_context *st) +{ + /* ARB vert/frag program */ + _mesa_HashWalk(st->ctx->Shared->Programs, + destroy_program_variants_cb, st); + + /* GLSL vert/frag/geom shaders */ + _mesa_HashWalk(st->ctx->Shared->ShaderObjects, + destroy_shader_program_variants_cb, st); +} diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 72dbc715fe1..519ee8c0d17 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -40,26 +40,60 @@ #include "st_context.h" +/** Fragment program variant key */ +struct st_fp_varient_key +{ + struct st_context *st; /**< variants are per-context */ + + /** for glBitmap */ + GLuint bitmap:1; /**< glBitmap varient? */ + + /** for glDrawPixels */ + GLuint drawpixels:1; /**< glDrawPixels varient */ + GLuint scaleAndBias:1; /**< glDrawPixels w/ scale and/or bias? */ + GLuint pixelMaps:1; /**< glDrawPixels w/ pixel lookup map? */ + GLuint drawpixels_z:1; /**< glDrawPixels(GL_DEPTH) */ + GLuint drawpixels_stencil:1; /**< glDrawPixels(GL_STENCIL) */ +}; + + +/** + * Variant of a fragment program. + */ +struct st_fp_varient +{ + /** Parameters which generated this version of fragment program */ + struct st_fp_varient_key key; + + /** Driver's compiled shader */ + void *driver_shader; + + /** For glBitmap variants */ + struct gl_program_parameter_list *parameters; + uint bitmap_sampler; + + /** next in linked list */ + struct st_fp_varient *next; +}; + + /** * Derived from Mesa gl_fragment_program: */ struct st_fragment_program { struct gl_fragment_program Base; - GLuint serialNo; struct pipe_shader_state tgsi; - void *driver_shader; - /** Program prefixed with glBitmap prologue */ - struct st_fragment_program *bitmap_program; - uint bitmap_sampler; + struct st_fp_varient *varients; }; struct st_vp_varient_key { + struct st_context *st; /**< variants are per-context */ boolean passthrough_edgeflags; }; @@ -101,7 +135,6 @@ struct st_vp_varient struct st_vertex_program { struct gl_vertex_program Base; /**< The Mesa vertex program */ - GLuint serialNo, lastSerialNo; /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */ GLuint input_to_index[VERT_ATTRIB_MAX]; @@ -120,13 +153,35 @@ struct st_vertex_program struct st_vp_varient *varients; }; + + +struct st_gp_varient_key +{ + struct st_context *st; /**< variants are per-context */ + /* no other fields yet */ +}; + + +/** + * Geometry program variant. + */ +struct st_gp_varient +{ + /* Parameters which generated this translated version of a vertex */ + struct st_gp_varient_key key; + + void *driver_shader; + + struct st_gp_varient *next; +}; + + /** * Derived from Mesa gl_geometry_program: */ struct st_geometry_program { struct gl_geometry_program Base; /**< The Mesa geometry program */ - GLuint serialNo; /** map GP input back to VP output */ GLuint input_map[PIPE_MAX_SHADER_INPUTS]; @@ -145,9 +200,12 @@ struct st_geometry_program ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; struct pipe_shader_state tgsi; - void *driver_shader; + + struct st_gp_varient *varients; }; + + static INLINE struct st_fragment_program * st_fragment_program( struct gl_fragment_program *fp ) { @@ -198,32 +256,43 @@ st_reference_fragprog(struct st_context *st, } -extern void -st_translate_fragment_program(struct st_context *st, - struct st_fragment_program *fp); +extern struct st_vp_varient * +st_get_vp_varient(struct st_context *st, + struct st_vertex_program *stvp, + const struct st_vp_varient_key *key); -extern void -st_translate_geometry_program(struct st_context *st, - struct st_geometry_program *stgp); -/* 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_fp_varient * +st_get_fp_varient(struct st_context *st, + struct st_fragment_program *stfp, + const struct st_fp_varient_key *key); -extern struct st_vp_varient * -st_translate_vertex_program(struct st_context *st, - struct st_vertex_program *stvp, - const struct st_vp_varient_key *key); -void +extern struct st_gp_varient * +st_get_gp_varient(struct st_context *st, + struct st_geometry_program *stgp, + const struct st_gp_varient_key *key); + + + +extern void st_vp_release_varients( struct st_context *st, struct st_vertex_program *stvp ); extern void +st_fp_release_varients( struct st_context *st, + struct st_fragment_program *stfp ); + +extern void +st_gp_release_varients(struct st_context *st, + struct st_geometry_program *stgp); + + +extern void st_print_shaders(struct gl_context *ctx); +extern void +st_destroy_program_variants(struct st_context *st); + #endif |