X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fintel%2Fvulkan%2Fgen8_cmd_buffer.c;h=2999c99261d1b4a34a159725be5f955969d4abdd;hb=7728720f07a1f0931d7d394d5842de3803e9192b;hp=dab1d7411e70d3469bc064e4603a11dc7ad9b4bb;hpb=5879cb0251e7b4593eb4fd01684bd68f0945e3d1;p=mesa.git diff --git a/src/intel/vulkan/gen8_cmd_buffer.c b/src/intel/vulkan/gen8_cmd_buffer.c index dab1d7411e7..2999c99261d 100644 --- a/src/intel/vulkan/gen8_cmd_buffer.c +++ b/src/intel/vulkan/gen8_cmd_buffer.c @@ -33,14 +33,14 @@ #include "genxml/genX_pack.h" #if GEN_GEN == 8 -static void -emit_viewport_state(struct anv_cmd_buffer *cmd_buffer, - uint32_t count, const VkViewport *viewports) +void +gen8_cmd_buffer_emit_viewport(struct anv_cmd_buffer *cmd_buffer) { + uint32_t count = cmd_buffer->state.gfx.dynamic.viewport.count; + const VkViewport *viewports = + cmd_buffer->state.gfx.dynamic.viewport.viewports; struct anv_state sf_clip_state = anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 64, 64); - struct anv_state cc_state = - anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 8, 32); for (uint32_t i = 0; i < count; i++) { const VkViewport *vp = &viewports[i]; @@ -50,189 +50,369 @@ emit_viewport_state(struct anv_cmd_buffer *cmd_buffer, struct GENX(SF_CLIP_VIEWPORT) sf_clip_viewport = { .ViewportMatrixElementm00 = vp->width / 2, .ViewportMatrixElementm11 = vp->height / 2, - .ViewportMatrixElementm22 = 1.0, + .ViewportMatrixElementm22 = vp->maxDepth - vp->minDepth, .ViewportMatrixElementm30 = vp->x + vp->width / 2, .ViewportMatrixElementm31 = vp->y + vp->height / 2, - .ViewportMatrixElementm32 = 0.0, + .ViewportMatrixElementm32 = vp->minDepth, .XMinClipGuardband = -1.0f, .XMaxClipGuardband = 1.0f, .YMinClipGuardband = -1.0f, .YMaxClipGuardband = 1.0f, .XMinViewPort = vp->x, .XMaxViewPort = vp->x + vp->width - 1, - .YMinViewPort = vp->y, - .YMaxViewPort = vp->y + vp->height - 1, - }; - - struct GENX(CC_VIEWPORT) cc_viewport = { - .MinimumDepth = vp->minDepth, - .MaximumDepth = vp->maxDepth + .YMinViewPort = MIN2(vp->y, vp->y + vp->height), + .YMaxViewPort = MAX2(vp->y, vp->y + vp->height) - 1, }; GENX(SF_CLIP_VIEWPORT_pack)(NULL, sf_clip_state.map + i * 64, &sf_clip_viewport); - GENX(CC_VIEWPORT_pack)(NULL, cc_state.map + i * 8, &cc_viewport); } - if (!cmd_buffer->device->info.has_llc) { - anv_state_clflush(sf_clip_state); - anv_state_clflush(cc_state); - } + anv_state_flush(cmd_buffer->device, sf_clip_state); anv_batch_emit(&cmd_buffer->batch, - GENX(3DSTATE_VIEWPORT_STATE_POINTERS_CC), - .CCViewportPointer = cc_state.offset); - anv_batch_emit(&cmd_buffer->batch, - GENX(3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP), - .SFClipViewportPointer = sf_clip_state.offset); + GENX(3DSTATE_VIEWPORT_STATE_POINTERS_SF_CLIP), clip) { + clip.SFClipViewportPointer = sf_clip_state.offset; + } } void -gen8_cmd_buffer_emit_viewport(struct anv_cmd_buffer *cmd_buffer) +gen8_cmd_buffer_emit_depth_viewport(struct anv_cmd_buffer *cmd_buffer, + bool depth_clamp_enable) { - if (cmd_buffer->state.dynamic.viewport.count > 0) { - emit_viewport_state(cmd_buffer, cmd_buffer->state.dynamic.viewport.count, - cmd_buffer->state.dynamic.viewport.viewports); - } else { - /* If viewport count is 0, this is taken to mean "use the default" */ - emit_viewport_state(cmd_buffer, 1, - &(VkViewport) { - .x = 0.0f, - .y = 0.0f, - .width = cmd_buffer->state.framebuffer->width, - .height = cmd_buffer->state.framebuffer->height, - .minDepth = 0.0f, - .maxDepth = 1.0f, - }); + uint32_t count = cmd_buffer->state.gfx.dynamic.viewport.count; + const VkViewport *viewports = + cmd_buffer->state.gfx.dynamic.viewport.viewports; + struct anv_state cc_state = + anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, count * 8, 32); + + for (uint32_t i = 0; i < count; i++) { + const VkViewport *vp = &viewports[i]; + + struct GENX(CC_VIEWPORT) cc_viewport = { + .MinimumDepth = depth_clamp_enable ? vp->minDepth : 0.0f, + .MaximumDepth = depth_clamp_enable ? vp->maxDepth : 1.0f, + }; + + GENX(CC_VIEWPORT_pack)(NULL, cc_state.map + i * 8, &cc_viewport); + } + + anv_state_flush(cmd_buffer->device, cc_state); + + anv_batch_emit(&cmd_buffer->batch, + GENX(3DSTATE_VIEWPORT_STATE_POINTERS_CC), cc) { + cc.CCViewportPointer = cc_state.offset; } } #endif -#define emit_lri(batch, reg, imm) \ - anv_batch_emit(batch, GENX(MI_LOAD_REGISTER_IMM), \ - .RegisterOffset = __anv_reg_num(reg), \ - .DataDWord = imm) - void -genX(cmd_buffer_config_l3)(struct anv_cmd_buffer *cmd_buffer, bool enable_slm) +genX(cmd_buffer_enable_pma_fix)(struct anv_cmd_buffer *cmd_buffer, bool enable) { - /* References for GL state: + if (cmd_buffer->state.pma_fix_enabled == enable) + return; + + cmd_buffer->state.pma_fix_enabled = enable; + + /* According to the Broadwell PIPE_CONTROL documentation, software should + * emit a PIPE_CONTROL with the CS Stall and Depth Cache Flush bits set + * prior to the LRI. If stencil buffer writes are enabled, then a Render + * Cache Flush is also necessary. * - * - commits e307cfa..228d5a3 - * - src/mesa/drivers/dri/i965/gen7_l3_state.c + * The Skylake docs say to use a depth stall rather than a command + * streamer stall. However, the hardware seems to violently disagree. + * A full command streamer stall seems to be needed in both cases. */ + anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) { + pc.DepthCacheFlushEnable = true; + pc.CommandStreamerStallEnable = true; + pc.RenderTargetCacheFlushEnable = true; + } + +#if GEN_GEN == 9 - uint32_t l3cr_slm, l3cr_noslm; - anv_pack_struct(&l3cr_noslm, GENX(L3CNTLREG), - .URBAllocation = 48, - .AllAllocation = 48); - anv_pack_struct(&l3cr_slm, GENX(L3CNTLREG), - .SLMEnable = 1, - .URBAllocation = 16, - .AllAllocation = 48); - const uint32_t l3cr_val = enable_slm ? l3cr_slm : l3cr_noslm; - bool changed = cmd_buffer->state.current_l3_config != l3cr_val; - - if (changed) { - /* According to the hardware docs, the L3 partitioning can only be - * changed while the pipeline is completely drained and the caches are - * flushed, which involves a first PIPE_CONTROL flush which stalls the - * pipeline... - */ - anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), - .DCFlushEnable = true, - .PostSyncOperation = NoWrite, - .CommandStreamerStallEnable = true); - - /* ...followed by a second pipelined PIPE_CONTROL that initiates - * invalidation of the relevant caches. Note that because RO - * invalidation happens at the top of the pipeline (i.e. right away as - * the PIPE_CONTROL command is processed by the CS) we cannot combine it - * with the previous stalling flush as the hardware documentation - * suggests, because that would cause the CS to stall on previous - * rendering *after* RO invalidation and wouldn't prevent the RO caches - * from being polluted by concurrent rendering before the stall - * completes. This intentionally doesn't implement the SKL+ hardware - * workaround suggesting to enable CS stall on PIPE_CONTROLs with the - * texture cache invalidation bit set for GPGPU workloads because the - * previous and subsequent PIPE_CONTROLs already guarantee that there is - * no concurrent GPGPU kernel execution (see SKL HSD 2132585). - */ - anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), - .TextureCacheInvalidationEnable = true, - .ConstantCacheInvalidationEnable = true, - .InstructionCacheInvalidateEnable = true, - .PostSyncOperation = NoWrite); - - /* Now send a third stalling flush to make sure that invalidation is - * complete when the L3 configuration registers are modified. - */ - anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), - .DCFlushEnable = true, - .PostSyncOperation = NoWrite, - .CommandStreamerStallEnable = true); - - emit_lri(&cmd_buffer->batch, GENX(L3CNTLREG), l3cr_val); - cmd_buffer->state.current_l3_config = l3cr_val; + uint32_t cache_mode; + anv_pack_struct(&cache_mode, GENX(CACHE_MODE_0), + .STCPMAOptimizationEnable = enable, + .STCPMAOptimizationEnableMask = true); + anv_batch_emit(&cmd_buffer->batch, GENX(MI_LOAD_REGISTER_IMM), lri) { + lri.RegisterOffset = GENX(CACHE_MODE_0_num); + lri.DataDWord = cache_mode; } -} -static void -__emit_genx_sf_state(struct anv_cmd_buffer *cmd_buffer) -{ - uint32_t sf_dw[GENX(3DSTATE_SF_length)]; - struct GENX(3DSTATE_SF) sf = { - GENX(3DSTATE_SF_header), - .LineWidth = cmd_buffer->state.dynamic.line_width, - }; - GENX(3DSTATE_SF_pack)(NULL, sf_dw, &sf); - /* FIXME: gen9.fs */ - anv_batch_emit_merge(&cmd_buffer->batch, sf_dw, - cmd_buffer->state.pipeline->gen8.sf); +#elif GEN_GEN == 8 + + uint32_t cache_mode; + anv_pack_struct(&cache_mode, GENX(CACHE_MODE_1), + .NPPMAFixEnable = enable, + .NPEarlyZFailsDisable = enable, + .NPPMAFixEnableMask = true, + .NPEarlyZFailsDisableMask = true); + anv_batch_emit(&cmd_buffer->batch, GENX(MI_LOAD_REGISTER_IMM), lri) { + lri.RegisterOffset = GENX(CACHE_MODE_1_num); + lri.DataDWord = cache_mode; + } + +#endif /* GEN_GEN == 8 */ + + /* After the LRI, a PIPE_CONTROL with both the Depth Stall and Depth Cache + * Flush bits is often necessary. We do it regardless because it's easier. + * The render cache flush is also necessary if stencil writes are enabled. + * + * Again, the Skylake docs give a different set of flushes but the BDW + * flushes seem to work just as well. + */ + anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) { + pc.DepthStallEnable = true; + pc.DepthCacheFlushEnable = true; + pc.RenderTargetCacheFlushEnable = true; + } } -#include "genxml/gen9_pack.h" -static void -__emit_gen9_sf_state(struct anv_cmd_buffer *cmd_buffer) +UNUSED static bool +want_depth_pma_fix(struct anv_cmd_buffer *cmd_buffer) { - uint32_t sf_dw[GENX(3DSTATE_SF_length)]; - struct GEN9_3DSTATE_SF sf = { - GEN9_3DSTATE_SF_header, - .LineWidth = cmd_buffer->state.dynamic.line_width, - }; - GEN9_3DSTATE_SF_pack(NULL, sf_dw, &sf); - /* FIXME: gen9.fs */ - anv_batch_emit_merge(&cmd_buffer->batch, sf_dw, - cmd_buffer->state.pipeline->gen8.sf); + assert(GEN_GEN == 8); + + /* From the Broadwell PRM Vol. 2c CACHE_MODE_1::NP_PMA_FIX_ENABLE: + * + * SW must set this bit in order to enable this fix when following + * expression is TRUE. + * + * 3DSTATE_WM::ForceThreadDispatch != 1 && + * !(3DSTATE_RASTER::ForceSampleCount != NUMRASTSAMPLES_0) && + * (3DSTATE_DEPTH_BUFFER::SURFACE_TYPE != NULL) && + * (3DSTATE_DEPTH_BUFFER::HIZ Enable) && + * !(3DSTATE_WM::EDSC_Mode == EDSC_PREPS) && + * (3DSTATE_PS_EXTRA::PixelShaderValid) && + * !(3DSTATE_WM_HZ_OP::DepthBufferClear || + * 3DSTATE_WM_HZ_OP::DepthBufferResolve || + * 3DSTATE_WM_HZ_OP::Hierarchical Depth Buffer Resolve Enable || + * 3DSTATE_WM_HZ_OP::StencilBufferClear) && + * (3DSTATE_WM_DEPTH_STENCIL::DepthTestEnable) && + * (((3DSTATE_PS_EXTRA::PixelShaderKillsPixels || + * 3DSTATE_PS_EXTRA::oMask Present to RenderTarget || + * 3DSTATE_PS_BLEND::AlphaToCoverageEnable || + * 3DSTATE_PS_BLEND::AlphaTestEnable || + * 3DSTATE_WM_CHROMAKEY::ChromaKeyKillEnable) && + * 3DSTATE_WM::ForceKillPix != ForceOff && + * ((3DSTATE_WM_DEPTH_STENCIL::DepthWriteEnable && + * 3DSTATE_DEPTH_BUFFER::DEPTH_WRITE_ENABLE) || + * (3DSTATE_WM_DEPTH_STENCIL::Stencil Buffer Write Enable && + * 3DSTATE_DEPTH_BUFFER::STENCIL_WRITE_ENABLE && + * 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE))) || + * (3DSTATE_PS_EXTRA:: Pixel Shader Computed Depth mode != PSCDEPTH_OFF)) + */ + + /* These are always true: + * 3DSTATE_WM::ForceThreadDispatch != 1 && + * !(3DSTATE_RASTER::ForceSampleCount != NUMRASTSAMPLES_0) + */ + + /* We only enable the PMA fix if we know for certain that HiZ is enabled. + * If we don't know whether HiZ is enabled or not, we disable the PMA fix + * and there is no harm. + * + * (3DSTATE_DEPTH_BUFFER::SURFACE_TYPE != NULL) && + * 3DSTATE_DEPTH_BUFFER::HIZ Enable + */ + if (!cmd_buffer->state.hiz_enabled) + return false; + + /* 3DSTATE_PS_EXTRA::PixelShaderValid */ + struct anv_pipeline *pipeline = cmd_buffer->state.gfx.base.pipeline; + if (!anv_pipeline_has_stage(pipeline, MESA_SHADER_FRAGMENT)) + return false; + + /* !(3DSTATE_WM::EDSC_Mode == EDSC_PREPS) */ + const struct brw_wm_prog_data *wm_prog_data = get_wm_prog_data(pipeline); + if (wm_prog_data->early_fragment_tests) + return false; + + /* We never use anv_pipeline for HiZ ops so this is trivially true: + * !(3DSTATE_WM_HZ_OP::DepthBufferClear || + * 3DSTATE_WM_HZ_OP::DepthBufferResolve || + * 3DSTATE_WM_HZ_OP::Hierarchical Depth Buffer Resolve Enable || + * 3DSTATE_WM_HZ_OP::StencilBufferClear) + */ + + /* 3DSTATE_WM_DEPTH_STENCIL::DepthTestEnable */ + if (!pipeline->depth_test_enable) + return false; + + /* (((3DSTATE_PS_EXTRA::PixelShaderKillsPixels || + * 3DSTATE_PS_EXTRA::oMask Present to RenderTarget || + * 3DSTATE_PS_BLEND::AlphaToCoverageEnable || + * 3DSTATE_PS_BLEND::AlphaTestEnable || + * 3DSTATE_WM_CHROMAKEY::ChromaKeyKillEnable) && + * 3DSTATE_WM::ForceKillPix != ForceOff && + * ((3DSTATE_WM_DEPTH_STENCIL::DepthWriteEnable && + * 3DSTATE_DEPTH_BUFFER::DEPTH_WRITE_ENABLE) || + * (3DSTATE_WM_DEPTH_STENCIL::Stencil Buffer Write Enable && + * 3DSTATE_DEPTH_BUFFER::STENCIL_WRITE_ENABLE && + * 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE))) || + * (3DSTATE_PS_EXTRA:: Pixel Shader Computed Depth mode != PSCDEPTH_OFF)) + */ + return (pipeline->kill_pixel && (pipeline->writes_depth || + pipeline->writes_stencil)) || + wm_prog_data->computed_depth_mode != PSCDEPTH_OFF; } -static void -__emit_sf_state(struct anv_cmd_buffer *cmd_buffer) +UNUSED static bool +want_stencil_pma_fix(struct anv_cmd_buffer *cmd_buffer) { - if (cmd_buffer->device->info.is_cherryview) - __emit_gen9_sf_state(cmd_buffer); - else - __emit_genx_sf_state(cmd_buffer); + if (GEN_GEN > 9) + return false; + assert(GEN_GEN == 9); + + /* From the Skylake PRM Vol. 2c CACHE_MODE_1::STC PMA Optimization Enable: + * + * Clearing this bit will force the STC cache to wait for pending + * retirement of pixels at the HZ-read stage and do the STC-test for + * Non-promoted, R-computed and Computed depth modes instead of + * postponing the STC-test to RCPFE. + * + * STC_TEST_EN = 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE && + * 3DSTATE_WM_DEPTH_STENCIL::StencilTestEnable + * + * STC_WRITE_EN = 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE && + * (3DSTATE_WM_DEPTH_STENCIL::Stencil Buffer Write Enable && + * 3DSTATE_DEPTH_BUFFER::STENCIL_WRITE_ENABLE) + * + * COMP_STC_EN = STC_TEST_EN && + * 3DSTATE_PS_EXTRA::PixelShaderComputesStencil + * + * SW parses the pipeline states to generate the following logical + * signal indicating if PMA FIX can be enabled. + * + * STC_PMA_OPT = + * 3DSTATE_WM::ForceThreadDispatch != 1 && + * !(3DSTATE_RASTER::ForceSampleCount != NUMRASTSAMPLES_0) && + * 3DSTATE_DEPTH_BUFFER::SURFACE_TYPE != NULL && + * 3DSTATE_DEPTH_BUFFER::HIZ Enable && + * !(3DSTATE_WM::EDSC_Mode == 2) && + * 3DSTATE_PS_EXTRA::PixelShaderValid && + * !(3DSTATE_WM_HZ_OP::DepthBufferClear || + * 3DSTATE_WM_HZ_OP::DepthBufferResolve || + * 3DSTATE_WM_HZ_OP::Hierarchical Depth Buffer Resolve Enable || + * 3DSTATE_WM_HZ_OP::StencilBufferClear) && + * (COMP_STC_EN || STC_WRITE_EN) && + * ((3DSTATE_PS_EXTRA::PixelShaderKillsPixels || + * 3DSTATE_WM::ForceKillPix == ON || + * 3DSTATE_PS_EXTRA::oMask Present to RenderTarget || + * 3DSTATE_PS_BLEND::AlphaToCoverageEnable || + * 3DSTATE_PS_BLEND::AlphaTestEnable || + * 3DSTATE_WM_CHROMAKEY::ChromaKeyKillEnable) || + * (3DSTATE_PS_EXTRA::Pixel Shader Computed Depth mode != PSCDEPTH_OFF)) + */ + + /* These are always true: + * 3DSTATE_WM::ForceThreadDispatch != 1 && + * !(3DSTATE_RASTER::ForceSampleCount != NUMRASTSAMPLES_0) + */ + + /* We only enable the PMA fix if we know for certain that HiZ is enabled. + * If we don't know whether HiZ is enabled or not, we disable the PMA fix + * and there is no harm. + * + * (3DSTATE_DEPTH_BUFFER::SURFACE_TYPE != NULL) && + * 3DSTATE_DEPTH_BUFFER::HIZ Enable + */ + if (!cmd_buffer->state.hiz_enabled) + return false; + + /* We can't possibly know if HiZ is enabled without the framebuffer */ + assert(cmd_buffer->state.framebuffer); + + /* HiZ is enabled so we had better have a depth buffer with HiZ */ + const struct anv_image_view *ds_iview = + anv_cmd_buffer_get_depth_stencil_view(cmd_buffer); + assert(ds_iview && ds_iview->image->planes[0].aux_usage == ISL_AUX_USAGE_HIZ); + + /* 3DSTATE_PS_EXTRA::PixelShaderValid */ + struct anv_pipeline *pipeline = cmd_buffer->state.gfx.base.pipeline; + if (!anv_pipeline_has_stage(pipeline, MESA_SHADER_FRAGMENT)) + return false; + + /* !(3DSTATE_WM::EDSC_Mode == 2) */ + const struct brw_wm_prog_data *wm_prog_data = get_wm_prog_data(pipeline); + if (wm_prog_data->early_fragment_tests) + return false; + + /* We never use anv_pipeline for HiZ ops so this is trivially true: + * !(3DSTATE_WM_HZ_OP::DepthBufferClear || + * 3DSTATE_WM_HZ_OP::DepthBufferResolve || + * 3DSTATE_WM_HZ_OP::Hierarchical Depth Buffer Resolve Enable || + * 3DSTATE_WM_HZ_OP::StencilBufferClear) + */ + + /* 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE && + * 3DSTATE_WM_DEPTH_STENCIL::StencilTestEnable + */ + const bool stc_test_en = + (ds_iview->image->aspects & VK_IMAGE_ASPECT_STENCIL_BIT) && + pipeline->stencil_test_enable; + + /* 3DSTATE_STENCIL_BUFFER::STENCIL_BUFFER_ENABLE && + * (3DSTATE_WM_DEPTH_STENCIL::Stencil Buffer Write Enable && + * 3DSTATE_DEPTH_BUFFER::STENCIL_WRITE_ENABLE) + */ + const bool stc_write_en = + (ds_iview->image->aspects & VK_IMAGE_ASPECT_STENCIL_BIT) && + pipeline->writes_stencil; + + /* STC_TEST_EN && 3DSTATE_PS_EXTRA::PixelShaderComputesStencil */ + const bool comp_stc_en = stc_test_en && wm_prog_data->computed_stencil; + + /* COMP_STC_EN || STC_WRITE_EN */ + if (!(comp_stc_en || stc_write_en)) + return false; + + /* (3DSTATE_PS_EXTRA::PixelShaderKillsPixels || + * 3DSTATE_WM::ForceKillPix == ON || + * 3DSTATE_PS_EXTRA::oMask Present to RenderTarget || + * 3DSTATE_PS_BLEND::AlphaToCoverageEnable || + * 3DSTATE_PS_BLEND::AlphaTestEnable || + * 3DSTATE_WM_CHROMAKEY::ChromaKeyKillEnable) || + * (3DSTATE_PS_EXTRA::Pixel Shader Computed Depth mode != PSCDEPTH_OFF) + */ + return pipeline->kill_pixel || + wm_prog_data->computed_depth_mode != PSCDEPTH_OFF; } void genX(cmd_buffer_flush_dynamic_state)(struct anv_cmd_buffer *cmd_buffer) { - struct anv_pipeline *pipeline = cmd_buffer->state.pipeline; + struct anv_pipeline *pipeline = cmd_buffer->state.gfx.base.pipeline; + struct anv_dynamic_state *d = &cmd_buffer->state.gfx.dynamic; - if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE | - ANV_CMD_DIRTY_DYNAMIC_LINE_WIDTH)) { - __emit_sf_state(cmd_buffer); + if (cmd_buffer->state.gfx.dirty & (ANV_CMD_DIRTY_PIPELINE | + ANV_CMD_DIRTY_DYNAMIC_LINE_WIDTH)) { + uint32_t sf_dw[GENX(3DSTATE_SF_length)]; + struct GENX(3DSTATE_SF) sf = { + GENX(3DSTATE_SF_header), + }; +#if GEN_GEN == 8 + if (cmd_buffer->device->info.is_cherryview) { + sf.CHVLineWidth = d->line_width; + } else { + sf.LineWidth = d->line_width; + } +#else + sf.LineWidth = d->line_width, +#endif + GENX(3DSTATE_SF_pack)(NULL, sf_dw, &sf); + anv_batch_emit_merge(&cmd_buffer->batch, sf_dw, pipeline->gen8.sf); } - if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE | - ANV_CMD_DIRTY_DYNAMIC_DEPTH_BIAS)){ + if (cmd_buffer->state.gfx.dirty & (ANV_CMD_DIRTY_PIPELINE | + ANV_CMD_DIRTY_DYNAMIC_DEPTH_BIAS)){ uint32_t raster_dw[GENX(3DSTATE_RASTER_length)]; struct GENX(3DSTATE_RASTER) raster = { GENX(3DSTATE_RASTER_header), - .GlobalDepthOffsetConstant = cmd_buffer->state.dynamic.depth_bias.bias, - .GlobalDepthOffsetScale = cmd_buffer->state.dynamic.depth_bias.slope, - .GlobalDepthOffsetClamp = cmd_buffer->state.dynamic.depth_bias.clamp + .GlobalDepthOffsetConstant = d->depth_bias.bias, + .GlobalDepthOffsetScale = d->depth_bias.slope, + .GlobalDepthOffsetClamp = d->depth_bias.clamp }; GENX(3DSTATE_RASTER_pack)(NULL, raster_dw, &raster); anv_batch_emit_merge(&cmd_buffer->batch, raster_dw, @@ -245,37 +425,35 @@ genX(cmd_buffer_flush_dynamic_state)(struct anv_cmd_buffer *cmd_buffer) * using a big old #if switch here. */ #if GEN_GEN == 8 - if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS | - ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE)) { - struct anv_dynamic_state *d = &cmd_buffer->state.dynamic; + if (cmd_buffer->state.gfx.dirty & (ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS | + ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE)) { struct anv_state cc_state = anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, GENX(COLOR_CALC_STATE_length) * 4, 64); struct GENX(COLOR_CALC_STATE) cc = { - .BlendConstantColorRed = cmd_buffer->state.dynamic.blend_constants[0], - .BlendConstantColorGreen = cmd_buffer->state.dynamic.blend_constants[1], - .BlendConstantColorBlue = cmd_buffer->state.dynamic.blend_constants[2], - .BlendConstantColorAlpha = cmd_buffer->state.dynamic.blend_constants[3], + .BlendConstantColorRed = d->blend_constants[0], + .BlendConstantColorGreen = d->blend_constants[1], + .BlendConstantColorBlue = d->blend_constants[2], + .BlendConstantColorAlpha = d->blend_constants[3], .StencilReferenceValue = d->stencil_reference.front & 0xff, - .BackFaceStencilReferenceValue = d->stencil_reference.back & 0xff, + .BackfaceStencilReferenceValue = d->stencil_reference.back & 0xff, }; GENX(COLOR_CALC_STATE_pack)(NULL, cc_state.map, &cc); - if (!cmd_buffer->device->info.has_llc) - anv_state_clflush(cc_state); + anv_state_flush(cmd_buffer->device, cc_state); - anv_batch_emit(&cmd_buffer->batch, - GENX(3DSTATE_CC_STATE_POINTERS), - .ColorCalcStatePointer = cc_state.offset, - .ColorCalcStatePointerValid = true); + anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CC_STATE_POINTERS), ccp) { + ccp.ColorCalcStatePointer = cc_state.offset; + ccp.ColorCalcStatePointerValid = true; + } } - if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE | - ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK | - ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK)) { + if (cmd_buffer->state.gfx.dirty & (ANV_CMD_DIRTY_PIPELINE | + ANV_CMD_DIRTY_RENDER_TARGETS | + ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK | + ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK)) { uint32_t wm_depth_stencil_dw[GENX(3DSTATE_WM_DEPTH_STENCIL_length)]; - struct anv_dynamic_state *d = &cmd_buffer->state.dynamic; struct GENX(3DSTATE_WM_DEPTH_STENCIL wm_depth_stencil) = { GENX(3DSTATE_WM_DEPTH_STENCIL_header), @@ -285,44 +463,50 @@ genX(cmd_buffer_flush_dynamic_state)(struct anv_cmd_buffer *cmd_buffer) .BackfaceStencilTestMask = d->stencil_compare_mask.back & 0xff, .BackfaceStencilWriteMask = d->stencil_write_mask.back & 0xff, + + .StencilBufferWriteEnable = + (d->stencil_write_mask.front || d->stencil_write_mask.back) && + pipeline->writes_stencil, }; GENX(3DSTATE_WM_DEPTH_STENCIL_pack)(NULL, wm_depth_stencil_dw, &wm_depth_stencil); anv_batch_emit_merge(&cmd_buffer->batch, wm_depth_stencil_dw, pipeline->gen8.wm_depth_stencil); + + genX(cmd_buffer_enable_pma_fix)(cmd_buffer, + want_depth_pma_fix(cmd_buffer)); } #else - if (cmd_buffer->state.dirty & ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS) { + if (cmd_buffer->state.gfx.dirty & ANV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS) { struct anv_state cc_state = anv_cmd_buffer_alloc_dynamic_state(cmd_buffer, - GEN9_COLOR_CALC_STATE_length * 4, + GENX(COLOR_CALC_STATE_length) * 4, 64); - struct GEN9_COLOR_CALC_STATE cc = { - .BlendConstantColorRed = cmd_buffer->state.dynamic.blend_constants[0], - .BlendConstantColorGreen = cmd_buffer->state.dynamic.blend_constants[1], - .BlendConstantColorBlue = cmd_buffer->state.dynamic.blend_constants[2], - .BlendConstantColorAlpha = cmd_buffer->state.dynamic.blend_constants[3], + struct GENX(COLOR_CALC_STATE) cc = { + .BlendConstantColorRed = d->blend_constants[0], + .BlendConstantColorGreen = d->blend_constants[1], + .BlendConstantColorBlue = d->blend_constants[2], + .BlendConstantColorAlpha = d->blend_constants[3], }; - GEN9_COLOR_CALC_STATE_pack(NULL, cc_state.map, &cc); + GENX(COLOR_CALC_STATE_pack)(NULL, cc_state.map, &cc); - if (!cmd_buffer->device->info.has_llc) - anv_state_clflush(cc_state); + anv_state_flush(cmd_buffer->device, cc_state); - anv_batch_emit(&cmd_buffer->batch, - GEN9_3DSTATE_CC_STATE_POINTERS, - .ColorCalcStatePointer = cc_state.offset, - .ColorCalcStatePointerValid = true); + anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_CC_STATE_POINTERS), ccp) { + ccp.ColorCalcStatePointer = cc_state.offset; + ccp.ColorCalcStatePointerValid = true; + } } - if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE | - ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK | - ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK | - ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE)) { - uint32_t dwords[GEN9_3DSTATE_WM_DEPTH_STENCIL_length]; - struct anv_dynamic_state *d = &cmd_buffer->state.dynamic; - struct GEN9_3DSTATE_WM_DEPTH_STENCIL wm_depth_stencil = { - GEN9_3DSTATE_WM_DEPTH_STENCIL_header, + if (cmd_buffer->state.gfx.dirty & (ANV_CMD_DIRTY_PIPELINE | + ANV_CMD_DIRTY_RENDER_TARGETS | + ANV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK | + ANV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK | + ANV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE)) { + uint32_t dwords[GENX(3DSTATE_WM_DEPTH_STENCIL_length)]; + struct GENX(3DSTATE_WM_DEPTH_STENCIL) wm_depth_stencil = { + GENX(3DSTATE_WM_DEPTH_STENCIL_header), .StencilTestMask = d->stencil_compare_mask.front & 0xff, .StencilWriteMask = d->stencil_write_mask.front & 0xff, @@ -332,23 +516,30 @@ genX(cmd_buffer_flush_dynamic_state)(struct anv_cmd_buffer *cmd_buffer) .StencilReferenceValue = d->stencil_reference.front & 0xff, .BackfaceStencilReferenceValue = d->stencil_reference.back & 0xff, + + .StencilBufferWriteEnable = + (d->stencil_write_mask.front || d->stencil_write_mask.back) && + pipeline->writes_stencil, }; - GEN9_3DSTATE_WM_DEPTH_STENCIL_pack(NULL, dwords, &wm_depth_stencil); + GENX(3DSTATE_WM_DEPTH_STENCIL_pack)(NULL, dwords, &wm_depth_stencil); anv_batch_emit_merge(&cmd_buffer->batch, dwords, pipeline->gen9.wm_depth_stencil); + + genX(cmd_buffer_enable_pma_fix)(cmd_buffer, + want_stencil_pma_fix(cmd_buffer)); } #endif - if (cmd_buffer->state.dirty & (ANV_CMD_DIRTY_PIPELINE | - ANV_CMD_DIRTY_INDEX_BUFFER)) { - anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_VF), - .IndexedDrawCutIndexEnable = pipeline->primitive_restart, - .CutIndex = cmd_buffer->state.restart_index, - ); + if (cmd_buffer->state.gfx.dirty & (ANV_CMD_DIRTY_PIPELINE | + ANV_CMD_DIRTY_INDEX_BUFFER)) { + anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_VF), vf) { + vf.IndexedDrawCutIndexEnable = pipeline->primitive_restart; + vf.CutIndex = cmd_buffer->state.restart_index; + } } - cmd_buffer->state.dirty = 0; + cmd_buffer->state.gfx.dirty = 0; } void genX(CmdBindIndexBuffer)( @@ -372,109 +563,35 @@ void genX(CmdBindIndexBuffer)( cmd_buffer->state.restart_index = restart_index_for_type[indexType]; - anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_INDEX_BUFFER), - .IndexFormat = vk_to_gen_index_type[indexType], - .MemoryObjectControlState = GENX(MOCS), - .BufferStartingAddress = { buffer->bo, buffer->offset + offset }, - .BufferSize = buffer->size - offset); - - cmd_buffer->state.dirty |= ANV_CMD_DIRTY_INDEX_BUFFER; -} - -static VkResult -flush_compute_descriptor_set(struct anv_cmd_buffer *cmd_buffer) -{ - struct anv_device *device = cmd_buffer->device; - struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline; - struct anv_state surfaces = { 0, }, samplers = { 0, }; - VkResult result; - - result = anv_cmd_buffer_emit_samplers(cmd_buffer, - MESA_SHADER_COMPUTE, &samplers); - if (result != VK_SUCCESS) - return result; - result = anv_cmd_buffer_emit_binding_table(cmd_buffer, - MESA_SHADER_COMPUTE, &surfaces); - if (result != VK_SUCCESS) - return result; - - struct anv_state push_state = anv_cmd_buffer_cs_push_constants(cmd_buffer); - - const struct brw_cs_prog_data *cs_prog_data = get_cs_prog_data(pipeline); - const struct brw_stage_prog_data *prog_data = &cs_prog_data->base; - - unsigned local_id_dwords = cs_prog_data->local_invocation_id_regs * 8; - unsigned push_constant_data_size = - (prog_data->nr_params + local_id_dwords) * 4; - unsigned reg_aligned_constant_size = ALIGN(push_constant_data_size, 32); - unsigned push_constant_regs = reg_aligned_constant_size / 32; - - if (push_state.alloc_size) { - anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_CURBE_LOAD), - .CURBETotalDataLength = push_state.alloc_size, - .CURBEDataStartAddress = push_state.offset); + anv_batch_emit(&cmd_buffer->batch, GENX(3DSTATE_INDEX_BUFFER), ib) { + ib.IndexFormat = vk_to_gen_index_type[indexType]; + ib.MemoryObjectControlState = GENX(MOCS); + ib.BufferStartingAddress = + (struct anv_address) { buffer->bo, buffer->offset + offset }; + ib.BufferSize = buffer->size - offset; } - assert(prog_data->total_shared <= 64 * 1024); - uint32_t slm_size = 0; - if (prog_data->total_shared > 0) { - /* slm_size is in 4k increments, but must be a power of 2. */ - slm_size = 4 * 1024; - while (slm_size < prog_data->total_shared) - slm_size <<= 1; - slm_size /= 4 * 1024; - } - - struct anv_state state = - anv_state_pool_emit(&device->dynamic_state_pool, - GENX(INTERFACE_DESCRIPTOR_DATA), 64, - .KernelStartPointer = pipeline->cs_simd, - .KernelStartPointerHigh = 0, - .BindingTablePointer = surfaces.offset, - .BindingTableEntryCount = 0, - .SamplerStatePointer = samplers.offset, - .SamplerCount = 0, - .ConstantIndirectURBEntryReadLength = push_constant_regs, - .ConstantURBEntryReadOffset = 0, - .BarrierEnable = cs_prog_data->uses_barrier, - .SharedLocalMemorySize = slm_size, - .NumberofThreadsinGPGPUThreadGroup = - pipeline->cs_thread_width_max); - - uint32_t size = GENX(INTERFACE_DESCRIPTOR_DATA_length) * sizeof(uint32_t); - anv_batch_emit(&cmd_buffer->batch, GENX(MEDIA_INTERFACE_DESCRIPTOR_LOAD), - .InterfaceDescriptorTotalLength = size, - .InterfaceDescriptorDataStartAddress = state.offset); - - return VK_SUCCESS; + cmd_buffer->state.gfx.dirty |= ANV_CMD_DIRTY_INDEX_BUFFER; } -void -genX(cmd_buffer_flush_compute_state)(struct anv_cmd_buffer *cmd_buffer) -{ - struct anv_pipeline *pipeline = cmd_buffer->state.compute_pipeline; - const struct brw_cs_prog_data *cs_prog_data = get_cs_prog_data(pipeline); - VkResult result; - - assert(pipeline->active_stages == VK_SHADER_STAGE_COMPUTE_BIT); - - bool needs_slm = cs_prog_data->base.total_shared > 0; - genX(cmd_buffer_config_l3)(cmd_buffer, needs_slm); - - genX(flush_pipeline_select_gpgpu)(cmd_buffer); - - if (cmd_buffer->state.compute_dirty & ANV_CMD_DIRTY_PIPELINE) - anv_batch_emit_batch(&cmd_buffer->batch, &pipeline->batch); - - if ((cmd_buffer->state.descriptors_dirty & VK_SHADER_STAGE_COMPUTE_BIT) || - (cmd_buffer->state.compute_dirty & ANV_CMD_DIRTY_PIPELINE)) { - result = flush_compute_descriptor_set(cmd_buffer); - assert(result == VK_SUCCESS); - cmd_buffer->state.descriptors_dirty &= ~VK_SHADER_STAGE_COMPUTE_BIT; - } - - cmd_buffer->state.compute_dirty = 0; -} +/* Set of stage bits for which are pipelined, i.e. they get queued by the + * command streamer for later execution. + */ +#define ANV_PIPELINE_STAGE_PIPELINED_BITS \ + (VK_PIPELINE_STAGE_VERTEX_INPUT_BIT | \ + VK_PIPELINE_STAGE_VERTEX_SHADER_BIT | \ + VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT | \ + VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT | \ + VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT | \ + VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | \ + VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT | \ + VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT | \ + VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT | \ + VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | \ + VK_PIPELINE_STAGE_TRANSFER_BIT | \ + VK_PIPELINE_STAGE_BOTTOM_OF_PIPE_BIT | \ + VK_PIPELINE_STAGE_ALL_GRAPHICS_BIT | \ + VK_PIPELINE_STAGE_ALL_COMMANDS_BIT) void genX(CmdSetEvent)( VkCommandBuffer commandBuffer, @@ -484,14 +601,20 @@ void genX(CmdSetEvent)( ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer); ANV_FROM_HANDLE(anv_event, event, _event); - anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), - .DestinationAddressType = DAT_PPGTT, - .PostSyncOperation = WriteImmediateData, - .Address = { - &cmd_buffer->device->dynamic_state_block_pool.bo, - event->state.offset - }, - .ImmediateData = VK_EVENT_SET); + anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) { + if (stageMask & ANV_PIPELINE_STAGE_PIPELINED_BITS) { + pc.StallAtPixelScoreboard = true; + pc.CommandStreamerStallEnable = true; + } + + pc.DestinationAddressType = DAT_PPGTT, + pc.PostSyncOperation = WriteImmediateData, + pc.Address = (struct anv_address) { + &cmd_buffer->device->dynamic_state_pool.block_pool.bo, + event->state.offset + }; + pc.ImmediateData = VK_EVENT_SET; + } } void genX(CmdResetEvent)( @@ -502,14 +625,20 @@ void genX(CmdResetEvent)( ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer); ANV_FROM_HANDLE(anv_event, event, _event); - anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), - .DestinationAddressType = DAT_PPGTT, - .PostSyncOperation = WriteImmediateData, - .Address = { - &cmd_buffer->device->dynamic_state_block_pool.bo, - event->state.offset - }, - .ImmediateData = VK_EVENT_RESET); + anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL), pc) { + if (stageMask & ANV_PIPELINE_STAGE_PIPELINED_BITS) { + pc.StallAtPixelScoreboard = true; + pc.CommandStreamerStallEnable = true; + } + + pc.DestinationAddressType = DAT_PPGTT; + pc.PostSyncOperation = WriteImmediateData; + pc.Address = (struct anv_address) { + &cmd_buffer->device->dynamic_state_pool.block_pool.bo, + event->state.offset + }; + pc.ImmediateData = VK_EVENT_RESET; + } } void genX(CmdWaitEvents)( @@ -529,14 +658,15 @@ void genX(CmdWaitEvents)( for (uint32_t i = 0; i < eventCount; i++) { ANV_FROM_HANDLE(anv_event, event, pEvents[i]); - anv_batch_emit(&cmd_buffer->batch, GENX(MI_SEMAPHORE_WAIT), - .WaitMode = PollingMode, - .CompareOperation = COMPARE_SAD_EQUAL_SDD, - .SemaphoreDataDword = VK_EVENT_SET, - .SemaphoreAddress = { - &cmd_buffer->device->dynamic_state_block_pool.bo, - event->state.offset - }); + anv_batch_emit(&cmd_buffer->batch, GENX(MI_SEMAPHORE_WAIT), sem) { + sem.WaitMode = PollingMode, + sem.CompareOperation = COMPARE_SAD_EQUAL_SDD, + sem.SemaphoreDataDword = VK_EVENT_SET, + sem.SemaphoreAddress = (struct anv_address) { + &cmd_buffer->device->dynamic_state_pool.block_pool.bo, + event->state.offset + }; + } } genX(CmdPipelineBarrier)(commandBuffer, srcStageMask, destStageMask,