X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Fdrivers%2Fdri%2Fi965%2Fbrw_wm.c;h=1943d08ab66cde700b2b2a1b76098ecd30d46fab;hb=1be1114e6b977ef70cc8bba9485adbaace4ecde4;hp=7eccbcb07e18eb0f1bf8dd42bce5afb0f30da097;hpb=2b7bbd89e87c4025cfc5513a078b1e7a10640357;p=mesa.git diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c index 7eccbcb07e1..1943d08ab66 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.c +++ b/src/mesa/drivers/dri/i965/brw_wm.c @@ -1,147 +1,64 @@ /* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - + * Copyright (C) Intel Corp. 2006. All Rights Reserved. + * Intel funded Tungsten Graphics to + * develop this 3D driver. + * + * Permission is hereby granted, free of charge, to any person obtaining + * a copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sublicense, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial + * portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + * LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + * OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + * WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ #include "brw_context.h" #include "brw_wm.h" #include "brw_state.h" +#include "brw_shader.h" #include "main/enums.h" #include "main/formats.h" #include "main/fbobject.h" #include "main/samplerobj.h" +#include "main/framebuffer.h" #include "program/prog_parameter.h" #include "program/program.h" #include "intel_mipmap_tree.h" +#include "intel_image.h" +#include "brw_nir.h" +#include "brw_program.h" -#include "glsl/ralloc.h" +#include "util/ralloc.h" -/** - * Return a bitfield where bit n is set if barycentric interpolation mode n - * (see enum brw_wm_barycentric_interp_mode) is needed by the fragment shader. - */ -static unsigned -brw_compute_barycentric_interp_modes(struct brw_context *brw, - bool shade_model_flat, - bool persample_shading, - const struct gl_fragment_program *fprog) +static void +assign_fs_binding_table_offsets(const struct brw_device_info *devinfo, + const struct gl_shader_program *shader_prog, + const struct gl_program *prog, + const struct brw_wm_prog_key *key, + struct brw_wm_prog_data *prog_data) { - unsigned barycentric_interp_modes = 0; - int attr; + uint32_t next_binding_table_offset = 0; - /* Loop through all fragment shader inputs to figure out what interpolation - * modes are in use, and set the appropriate bits in - * barycentric_interp_modes. + /* If there are no color regions, we still perform an FB write to a null + * renderbuffer, which we place at surface index 0. */ - for (attr = 0; attr < VARYING_SLOT_MAX; ++attr) { - enum glsl_interp_qualifier interp_qualifier = - fprog->InterpQualifier[attr]; - bool is_centroid = (fprog->IsCentroid & BITFIELD64_BIT(attr)) && - !persample_shading; - bool is_sample = (fprog->IsSample & BITFIELD64_BIT(attr)) || - persample_shading; - bool is_gl_Color = attr == VARYING_SLOT_COL0 || attr == VARYING_SLOT_COL1; - - /* Ignore unused inputs. */ - if (!(fprog->Base.InputsRead & BITFIELD64_BIT(attr))) - continue; - - /* Ignore WPOS and FACE, because they don't require interpolation. */ - if (attr == VARYING_SLOT_POS || attr == VARYING_SLOT_FACE) - continue; - - /* Determine the set (or sets) of barycentric coordinates needed to - * interpolate this variable. Note that when - * brw->needs_unlit_centroid_workaround is set, centroid interpolation - * uses PIXEL interpolation for unlit pixels and CENTROID interpolation - * for lit pixels, so we need both sets of barycentric coordinates. - */ - if (interp_qualifier == INTERP_QUALIFIER_NOPERSPECTIVE) { - if (is_centroid) { - barycentric_interp_modes |= - 1 << BRW_WM_NONPERSPECTIVE_CENTROID_BARYCENTRIC; - } else if (is_sample) { - barycentric_interp_modes |= - 1 << BRW_WM_NONPERSPECTIVE_SAMPLE_BARYCENTRIC; - } - if ((!is_centroid && !is_sample) || - brw->needs_unlit_centroid_workaround) { - barycentric_interp_modes |= - 1 << BRW_WM_NONPERSPECTIVE_PIXEL_BARYCENTRIC; - } - } else if (interp_qualifier == INTERP_QUALIFIER_SMOOTH || - (!(shade_model_flat && is_gl_Color) && - interp_qualifier == INTERP_QUALIFIER_NONE)) { - if (is_centroid) { - barycentric_interp_modes |= - 1 << BRW_WM_PERSPECTIVE_CENTROID_BARYCENTRIC; - } else if (is_sample) { - barycentric_interp_modes |= - 1 << BRW_WM_PERSPECTIVE_SAMPLE_BARYCENTRIC; - } - if ((!is_centroid && !is_sample) || - brw->needs_unlit_centroid_workaround) { - barycentric_interp_modes |= - 1 << BRW_WM_PERSPECTIVE_PIXEL_BARYCENTRIC; - } - } - } - - return barycentric_interp_modes; -} + prog_data->binding_table.render_target_start = next_binding_table_offset; + next_binding_table_offset += MAX2(key->nr_color_regions, 1); -bool -brw_wm_prog_data_compare(const void *in_a, const void *in_b) -{ - const struct brw_wm_prog_data *a = in_a; - const struct brw_wm_prog_data *b = in_b; - - /* Compare all the struct (including the base) up to the pointers. */ - if (memcmp(a, b, offsetof(struct brw_wm_prog_data, param))) - return false; - - if (memcmp(a->param, b->param, a->nr_params * sizeof(void *))) - return false; - - if (memcmp(a->pull_param, b->pull_param, a->nr_pull_params * sizeof(void *))) - return false; - - return true; -} - -void -brw_wm_prog_data_free(const void *in_prog_data) -{ - const struct brw_wm_prog_data *prog_data = in_prog_data; - - ralloc_free((void *)prog_data->param); - ralloc_free((void *)prog_data->pull_param); + brw_assign_common_binding_table_offsets(MESA_SHADER_FRAGMENT, devinfo, + shader_prog, prog, &prog_data->base, + next_binding_table_offset); } /** @@ -149,85 +66,122 @@ brw_wm_prog_data_free(const void *in_prog_data) * Depending on the instructions used (i.e. flow control instructions) * we'll use one of two code generators. */ -bool do_wm_prog(struct brw_context *brw, - struct gl_shader_program *prog, - struct brw_fragment_program *fp, - struct brw_wm_prog_key *key) +bool +brw_codegen_wm_prog(struct brw_context *brw, + struct gl_shader_program *prog, + struct brw_fragment_program *fp, + struct brw_wm_prog_key *key) { struct gl_context *ctx = &brw->ctx; - struct brw_wm_compile *c; + void *mem_ctx = ralloc_context(NULL); + struct brw_wm_prog_data prog_data; const GLuint *program; - struct gl_shader *fs = NULL; + struct brw_shader *fs = NULL; GLuint program_size; + bool start_busy = false; + double start_time = 0; if (prog) - fs = prog->_LinkedShaders[MESA_SHADER_FRAGMENT]; + fs = (struct brw_shader *)prog->_LinkedShaders[MESA_SHADER_FRAGMENT]; + + memset(&prog_data, 0, sizeof(prog_data)); + + /* Use ALT floating point mode for ARB programs so that 0^0 == 1. */ + if (!prog) + prog_data.base.use_alt_mode = true; - c = rzalloc(NULL, struct brw_wm_compile); + assign_fs_binding_table_offsets(brw->intelScreen->devinfo, prog, + &fp->program.Base, key, &prog_data); /* Allocate the references to the uniforms that will end up in the * prog_data associated with the compiled program, and which will be freed * by the state cache. */ - int param_count; - if (fs) { - param_count = fs->num_uniform_components; - } else { - param_count = fp->program.Base.Parameters->NumParameters * 4; - } + int param_count = fp->program.Base.nir->num_uniforms; + if (fs) + prog_data.base.nr_image_params = fs->base.NumImages; /* The backend also sometimes adds params for texture size. */ param_count += 2 * ctx->Const.Program[MESA_SHADER_FRAGMENT].MaxTextureImageUnits; - c->prog_data.param = rzalloc_array(NULL, const float *, param_count); - c->prog_data.pull_param = rzalloc_array(NULL, const float *, param_count); + prog_data.base.param = + rzalloc_array(NULL, const gl_constant_value *, param_count); + prog_data.base.pull_param = + rzalloc_array(NULL, const gl_constant_value *, param_count); + prog_data.base.image_param = + rzalloc_array(NULL, struct brw_image_param, + prog_data.base.nr_image_params); + prog_data.base.nr_params = param_count; + + if (prog) { + brw_nir_setup_glsl_uniforms(fp->program.Base.nir, prog, &fp->program.Base, + &prog_data.base, true); + } else { + brw_nir_setup_arb_uniforms(fp->program.Base.nir, &fp->program.Base, + &prog_data.base); + } + + if (unlikely(brw->perf_debug)) { + start_busy = (brw->batch.last_bo && + drm_intel_bo_busy(brw->batch.last_bo)); + start_time = get_time(); + } - memcpy(&c->key, key, sizeof(*key)); + if (unlikely(INTEL_DEBUG & DEBUG_WM)) + brw_dump_ir("fragment", prog, fs ? &fs->base : NULL, &fp->program.Base); - c->prog_data.barycentric_interp_modes = - brw_compute_barycentric_interp_modes(brw, c->key.flat_shade, - c->key.persample_shading, - &fp->program); + int st_index8 = -1, st_index16 = -1; + if (INTEL_DEBUG & DEBUG_SHADER_TIME) { + st_index8 = brw_get_shader_time_index(brw, prog, &fp->program.Base, ST_FS8); + st_index16 = brw_get_shader_time_index(brw, prog, &fp->program.Base, ST_FS16); + } + + char *error_str = NULL; + program = brw_compile_fs(brw->intelScreen->compiler, brw, mem_ctx, + key, &prog_data, fp->program.Base.nir, + &fp->program.Base, st_index8, st_index16, + true, brw->use_rep_send, + &program_size, &error_str); + if (program == NULL) { + if (prog) { + prog->LinkStatus = false; + ralloc_strcat(&prog->InfoLog, error_str); + } - program = brw_wm_fs_emit(brw, c, &fp->program, prog, &program_size); - if (program == NULL) + _mesa_problem(NULL, "Failed to compile fragment shader: %s\n", error_str); + + ralloc_free(mem_ctx); return false; + } - /* Scratch space is used for register spilling */ - if (c->last_scratch) { - perf_debug("Fragment shader triggered register spilling. " - "Try reducing the number of live scalar values to " - "improve performance.\n"); + if (unlikely(brw->perf_debug) && fs) { + if (fs->compiled_once) + brw_wm_debug_recompile(brw, prog, key); + fs->compiled_once = true; - c->prog_data.total_scratch = brw_get_scratch_size(c->last_scratch); + if (start_busy && !drm_intel_bo_busy(brw->batch.last_bo)) { + perf_debug("FS compile took %.03f ms and stalled the GPU\n", + (get_time() - start_time) * 1000); + } + } + if (prog_data.base.total_scratch) { brw_get_scratch_bo(brw, &brw->wm.base.scratch_bo, - c->prog_data.total_scratch * brw->max_wm_threads); + prog_data.base.total_scratch * brw->max_wm_threads); } if (unlikely(INTEL_DEBUG & DEBUG_WM)) fprintf(stderr, "\n"); - brw_upload_cache(&brw->cache, BRW_WM_PROG, - &c->key, sizeof(c->key), + brw_upload_cache(&brw->cache, BRW_CACHE_FS_PROG, + key, sizeof(struct brw_wm_prog_key), program, program_size, - &c->prog_data, sizeof(c->prog_data), + &prog_data, sizeof(prog_data), &brw->wm.base.prog_offset, &brw->wm.prog_data); - ralloc_free(c); + ralloc_free(mem_ctx); return true; } -static bool -key_debug(struct brw_context *brw, const char *name, int a, int b) -{ - if (a != b) { - perf_debug(" %s %d->%d\n", name, a, b); - return true; - } else { - return false; - } -} - bool brw_debug_recompile_sampler_key(struct brw_context *brw, const struct brw_sampler_prog_key_data *old_key, @@ -245,12 +199,29 @@ brw_debug_recompile_sampler_key(struct brw_context *brw, old_key->gl_clamp_mask[1], key->gl_clamp_mask[1]); found |= key_debug(brw, "GL_CLAMP enabled on any texture unit's 3rd coordinate", old_key->gl_clamp_mask[2], key->gl_clamp_mask[2]); - found |= key_debug(brw, "GL_MESA_ycbcr texturing\n", - old_key->yuvtex_mask, key->yuvtex_mask); - found |= key_debug(brw, "GL_MESA_ycbcr UV swapping\n", - old_key->yuvtex_swap_mask, key->yuvtex_swap_mask); found |= key_debug(brw, "gather channel quirk on any texture unit", old_key->gather_channel_quirk_mask, key->gather_channel_quirk_mask); + found |= key_debug(brw, "compressed multisample layout", + old_key->compressed_multisample_layout_mask, + key->compressed_multisample_layout_mask); + found |= key_debug(brw, "16x msaa", + old_key->msaa_16, + key->msaa_16); + + found |= key_debug(brw, "y_uv image bound", + old_key->y_uv_image_mask, + key->y_uv_image_mask); + found |= key_debug(brw, "y_u_v image bound", + old_key->y_u_v_image_mask, + key->y_u_v_image_mask); + found |= key_debug(brw, "yx_xuxv image bound", + old_key->yx_xuxv_image_mask, + key->yx_xuxv_image_mask); + + for (unsigned int i = 0; i < MAX_SAMPLERS; i++) { + found |= key_debug(brw, "textureGather workarounds", + old_key->gen6_gather_wa[i], key->gen6_gather_wa[i]); + } return found; } @@ -268,7 +239,7 @@ brw_wm_debug_recompile(struct brw_context *brw, for (unsigned int i = 0; i < brw->cache.size; i++) { for (c = brw->cache.items[i]; c; c = c->next) { - if (c->cache_id == BRW_WM_PROG) { + if (c->cache_id == BRW_CACHE_FS_PROG) { old_key = c->key; if (old_key->program_string_id == key->program_string_id) @@ -291,18 +262,18 @@ brw_wm_debug_recompile(struct brw_context *brw, old_key->stats_wm, key->stats_wm); found |= key_debug(brw, "flat shading", old_key->flat_shade, key->flat_shade); + found |= key_debug(brw, "per-sample interpolation", + old_key->persample_interp, key->persample_interp); found |= key_debug(brw, "number of color buffers", old_key->nr_color_regions, key->nr_color_regions); found |= key_debug(brw, "MRT alpha test or alpha-to-coverage", old_key->replicate_alpha, key->replicate_alpha); - found |= key_debug(brw, "rendering to FBO", - old_key->render_to_fbo, key->render_to_fbo); found |= key_debug(brw, "fragment color clamping", old_key->clamp_fragment_color, key->clamp_fragment_color); + found |= key_debug(brw, "multisampled FBO", + old_key->multisample_fbo, key->multisample_fbo); found |= key_debug(brw, "line smoothing", old_key->line_aa, key->line_aa); - found |= key_debug(brw, "renderbuffer height", - old_key->drawable_height, key->drawable_height); found |= key_debug(brw, "input slots valid", old_key->input_slots_valid, key->input_slots_valid); found |= key_debug(brw, "mrt alpha test function", @@ -321,13 +292,15 @@ static uint8_t gen6_gather_workaround(GLenum internalformat) { switch (internalformat) { - case GL_R8I: return WA_SIGN | WA_8BIT; - case GL_R8UI: return WA_8BIT; - case GL_R16I: return WA_SIGN | WA_16BIT; - case GL_R16UI: return WA_16BIT; - /* note that even though GL_R32I and GL_R32UI have format overrides - * in the surface state, there is no shader w/a required */ - default: return 0; + case GL_R8I: return WA_SIGN | WA_8BIT; + case GL_R8UI: return WA_8BIT; + case GL_R16I: return WA_SIGN | WA_16BIT; + case GL_R16UI: return WA_16BIT; + default: + /* Note that even though GL_R32I and GL_R32UI have format overrides in + * the surface state, there is no shader w/a required. + */ + return 0; } } @@ -348,7 +321,7 @@ brw_populate_sampler_prog_key_data(struct gl_context *ctx, int unit_id = prog->SamplerUnits[s]; const struct gl_texture_unit *unit = &ctx->Texture.Unit[unit_id]; - if (unit->_ReallyEnabled && unit->_Current->Target != GL_TEXTURE_BUFFER) { + if (unit->_Current && unit->_Current->Target != GL_TEXTURE_BUFFER) { const struct gl_texture_object *t = unit->_Current; const struct gl_texture_image *img = t->Image[0][t->BaseLevel]; struct gl_sampler_object *sampler = _mesa_get_samplerobj(ctx, unit_id); @@ -363,13 +336,8 @@ brw_populate_sampler_prog_key_data(struct gl_context *ctx, if (alpha_depth || (brw->gen < 8 && !brw->is_haswell)) key->swizzles[s] = brw_get_texture_swizzle(ctx, t); - if (img->InternalFormat == GL_YCBCR_MESA) { - key->yuvtex_mask |= 1 << s; - if (img->TexFormat == MESA_FORMAT_YCBCR) - key->yuvtex_swap_mask |= 1 << s; - } - - if (sampler->MinFilter != GL_NEAREST && + if (brw->gen < 8 && + sampler->MinFilter != GL_NEAREST && sampler->MagFilter != GL_NEAREST) { if (sampler->WrapS == GL_CLAMP) key->gl_clamp_mask[0] |= 1 << s; @@ -379,8 +347,9 @@ brw_populate_sampler_prog_key_data(struct gl_context *ctx, key->gl_clamp_mask[2] |= 1 << s; } - /* gather4's channel select for green from RG32F is broken; - * requires a shader w/a on IVB; fixable with just SCS on HSW. */ + /* gather4's channel select for green from RG32F is broken; requires + * a shader w/a on IVB; fixable with just SCS on HSW. + */ if (brw->gen == 7 && !brw->is_haswell && prog->UsesGather) { if (img->InternalFormat == GL_RG32F) key->gather_channel_quirk_mask |= 1 << s; @@ -400,26 +369,71 @@ brw_populate_sampler_prog_key_data(struct gl_context *ctx, struct intel_texture_object *intel_tex = intel_texture_object((struct gl_texture_object *)t); + /* From gen9 onwards some single sampled buffers can also be + * compressed. These don't need ld2dms sampling along with mcs fetch. + */ if (brw->gen >= 7 && - intel_tex->mt->msaa_layout == INTEL_MSAA_LAYOUT_CMS) { + intel_tex->mt->msaa_layout == INTEL_MSAA_LAYOUT_CMS && + intel_tex->mt->num_samples > 1) { key->compressed_multisample_layout_mask |= 1 << s; + + if (intel_tex->mt->num_samples >= 16) { + assert(brw->gen >= 9); + key->msaa_16 |= 1 << s; + } } + + if (t->Target == GL_TEXTURE_EXTERNAL_OES && intel_tex->planar_format) { + switch (intel_tex->planar_format->components) { + case __DRI_IMAGE_COMPONENTS_Y_UV: + key->y_uv_image_mask |= 1 << s; + break; + case __DRI_IMAGE_COMPONENTS_Y_U_V: + key->y_u_v_image_mask |= 1 << s; + break; + case __DRI_IMAGE_COMPONENTS_Y_XUXV: + key->yx_xuxv_image_mask |= 1 << s; + break; + default: + break; + } + } + } } } -static void brw_wm_populate_key( struct brw_context *brw, - struct brw_wm_prog_key *key ) +static bool +brw_wm_state_dirty(const struct brw_context *brw) +{ + return brw_state_dirty(brw, + _NEW_BUFFERS | + _NEW_COLOR | + _NEW_DEPTH | + _NEW_FRAG_CLAMP | + _NEW_HINT | + _NEW_LIGHT | + _NEW_LINE | + _NEW_MULTISAMPLE | + _NEW_POLYGON | + _NEW_STENCIL | + _NEW_TEXTURE, + BRW_NEW_FRAGMENT_PROGRAM | + BRW_NEW_REDUCED_PRIMITIVE | + BRW_NEW_STATS_WM | + BRW_NEW_VUE_MAP_GEOM_OUT); +} + +static void +brw_wm_populate_key(struct brw_context *brw, struct brw_wm_prog_key *key) { struct gl_context *ctx = &brw->ctx; /* BRW_NEW_FRAGMENT_PROGRAM */ const struct brw_fragment_program *fp = - (struct brw_fragment_program *)brw->fragment_program; + (struct brw_fragment_program *) brw->fragment_program; const struct gl_program *prog = (struct gl_program *) brw->fragment_program; GLuint lookup = 0; GLuint line_aa; - bool program_uses_dfdy = fp->program.UsesDFdy; - bool multisample_fbo = ctx->DrawBuffer->Visual.samples > 1; memset(key, 0, sizeof(*key)); @@ -480,13 +494,8 @@ static void brw_wm_populate_key( struct brw_context *brw, key->line_aa = line_aa; /* _NEW_HINT */ - if (brw->disable_derivative_optimization) { - key->high_quality_derivatives = - ctx->Hint.FragmentShaderDerivative != GL_FASTEST; - } else { - key->high_quality_derivatives = - ctx->Hint.FragmentShaderDerivative == GL_NICEST; - } + key->high_quality_derivatives = + ctx->Hint.FragmentShaderDerivative == GL_NICEST; if (brw->gen < 6) key->stats_wm = brw->stats_wm; @@ -501,55 +510,27 @@ static void brw_wm_populate_key( struct brw_context *brw, brw_populate_sampler_prog_key_data(ctx, prog, brw->wm.base.sampler_count, &key->tex); - /* _NEW_BUFFERS */ - /* - * Include the draw buffer origin and height so that we can calculate - * fragment position values relative to the bottom left of the drawable, - * from the incoming screen origin relative position we get as part of our - * payload. - * - * This is only needed for the WM_WPOSXY opcode when the fragment program - * uses the gl_FragCoord input. - * - * We could avoid recompiling by including this as a constant referenced by - * our program, but if we were to do that it would also be nice to handle - * getting that constant updated at batchbuffer submit time (when we - * hold the lock and know where the buffer really is) rather than at emit - * time when we don't hold the lock and are just guessing. We could also - * just avoid using this as key data if the program doesn't use - * fragment.position. - * - * For DRI2 the origin_x/y will always be (0,0) but we still need the - * drawable height in order to invert the Y axis. - */ - if (fp->program.Base.InputsRead & VARYING_BIT_POS) { - key->drawable_height = ctx->DrawBuffer->Height; - } - - if ((fp->program.Base.InputsRead & VARYING_BIT_POS) || program_uses_dfdy) { - key->render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer); - } - /* _NEW_BUFFERS */ key->nr_color_regions = ctx->DrawBuffer->_NumColorDrawBuffers; + /* _NEW_COLOR */ + key->force_dual_color_blend = brw->dual_color_blend_by_location && + (ctx->Color.BlendEnabled & 1) && ctx->Color.Blend[0]._UsesDualSrc; + /* _NEW_MULTISAMPLE, _NEW_COLOR, _NEW_BUFFERS */ key->replicate_alpha = ctx->DrawBuffer->_NumColorDrawBuffers > 1 && (ctx->Multisample.SampleAlphaToCoverage || ctx->Color.AlphaEnabled); /* _NEW_BUFFERS _NEW_MULTISAMPLE */ /* Ignore sample qualifier while computing this flag. */ - key->persample_shading = - _mesa_get_min_invocations_per_fragment(ctx, &fp->program, true) > 1; - - key->compute_pos_offset = - _mesa_get_min_invocations_per_fragment(ctx, &fp->program, false) > 1 && - fp->program.Base.SystemValuesRead & SYSTEM_BIT_SAMPLE_POS; + if (ctx->Multisample.Enabled) { + key->persample_interp = + ctx->Multisample.SampleShading && + (ctx->Multisample.MinSampleShadingValue * + _mesa_geometric_samples(ctx->DrawBuffer) > 1); - key->compute_sample_id = - multisample_fbo && - ctx->Multisample.Enabled && - (fp->program.Base.SystemValuesRead & SYSTEM_BIT_SAMPLE_ID); + key->multisample_fbo = _mesa_geometric_samples(ctx->DrawBuffer) > 1; + } /* BRW_NEW_VUE_MAP_GEOM_OUT */ if (brw->gen < 6 || _mesa_bitcount_64(fp->program.Base.InputsRead & @@ -563,7 +544,8 @@ static void brw_wm_populate_key( struct brw_context *brw, * like GL requires. Fix that by building the alpha test into the * shader, and we'll skip enabling the fixed function alpha test. */ - if (brw->gen < 6 && ctx->DrawBuffer->_NumColorDrawBuffers > 1 && ctx->Color.AlphaEnabled) { + if (brw->gen < 6 && ctx->DrawBuffer->_NumColorDrawBuffers > 1 && + ctx->Color.AlphaEnabled) { key->alpha_test_func = ctx->Color.AlphaFunc; key->alpha_test_ref = ctx->Color.AlphaRef; } @@ -572,47 +554,74 @@ static void brw_wm_populate_key( struct brw_context *brw, key->program_string_id = fp->id; } - -static void +void brw_upload_wm_prog(struct brw_context *brw) { struct gl_context *ctx = &brw->ctx; + struct gl_shader_program *current = ctx->_Shader->_CurrentFragmentProgram; struct brw_wm_prog_key key; struct brw_fragment_program *fp = (struct brw_fragment_program *) brw->fragment_program; + if (!brw_wm_state_dirty(brw)) + return; + brw_wm_populate_key(brw, &key); - if (!brw_search_cache(&brw->cache, BRW_WM_PROG, + if (!brw_search_cache(&brw->cache, BRW_CACHE_FS_PROG, &key, sizeof(key), &brw->wm.base.prog_offset, &brw->wm.prog_data)) { - bool success = do_wm_prog(brw, ctx->Shader._CurrentFragmentProgram, fp, - &key); + bool success = brw_codegen_wm_prog(brw, current, fp, &key); (void) success; assert(success); } brw->wm.base.prog_data = &brw->wm.prog_data->base; } +bool +brw_fs_precompile(struct gl_context *ctx, + struct gl_shader_program *shader_prog, + struct gl_program *prog) +{ + struct brw_context *brw = brw_context(ctx); + struct brw_wm_prog_key key; + + struct gl_fragment_program *fp = (struct gl_fragment_program *) prog; + struct brw_fragment_program *bfp = brw_fragment_program(fp); -const struct brw_tracked_state brw_wm_prog = { - .dirty = { - .mesa = (_NEW_COLOR | - _NEW_DEPTH | - _NEW_STENCIL | - _NEW_POLYGON | - _NEW_LINE | - _NEW_HINT | - _NEW_LIGHT | - _NEW_FRAG_CLAMP | - _NEW_BUFFERS | - _NEW_TEXTURE | - _NEW_MULTISAMPLE), - .brw = (BRW_NEW_FRAGMENT_PROGRAM | - BRW_NEW_REDUCED_PRIMITIVE | - BRW_NEW_VUE_MAP_GEOM_OUT | - BRW_NEW_STATS_WM) - }, - .emit = brw_upload_wm_prog -}; + memset(&key, 0, sizeof(key)); + if (brw->gen < 6) { + if (fp->UsesKill) + key.iz_lookup |= IZ_PS_KILL_ALPHATEST_BIT; + + if (fp->Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH)) + key.iz_lookup |= IZ_PS_COMPUTES_DEPTH_BIT; + + /* Just assume depth testing. */ + key.iz_lookup |= IZ_DEPTH_TEST_ENABLE_BIT; + key.iz_lookup |= IZ_DEPTH_WRITE_ENABLE_BIT; + } + + if (brw->gen < 6 || _mesa_bitcount_64(fp->Base.InputsRead & + BRW_FS_VARYING_INPUT_MASK) > 16) + key.input_slots_valid = fp->Base.InputsRead | VARYING_BIT_POS; + + brw_setup_tex_for_precompile(brw, &key.tex, &fp->Base); + + key.nr_color_regions = _mesa_bitcount_64(fp->Base.OutputsWritten & + ~(BITFIELD64_BIT(FRAG_RESULT_DEPTH) | + BITFIELD64_BIT(FRAG_RESULT_SAMPLE_MASK))); + + key.program_string_id = bfp->id; + + uint32_t old_prog_offset = brw->wm.base.prog_offset; + struct brw_wm_prog_data *old_prog_data = brw->wm.prog_data; + + bool success = brw_codegen_wm_prog(brw, shader_prog, bfp, &key); + + brw->wm.base.prog_offset = old_prog_offset; + brw->wm.prog_data = old_prog_data; + + return success; +}