diff options
Diffstat (limited to 'src/vulkan/gen8_pipeline.c')
| -rw-r--r-- | src/vulkan/gen8_pipeline.c | 164 |
1 files changed, 110 insertions, 54 deletions
diff --git a/src/vulkan/gen8_pipeline.c b/src/vulkan/gen8_pipeline.c index 6e2e65d6344..0038bca01b4 100644 --- a/src/vulkan/gen8_pipeline.c +++ b/src/vulkan/gen8_pipeline.c @@ -30,6 +30,7 @@ #include "anv_private.h" #include "gen8_pack.h" +#include "gen9_pack.h" static void emit_vertex_input(struct anv_pipeline *pipeline, @@ -38,9 +39,11 @@ emit_vertex_input(struct anv_pipeline *pipeline, const uint32_t num_dwords = 1 + info->attributeCount * 2; uint32_t *p; + static_assert(ANV_GEN >= 8, "should be compiling this for gen < 8"); + if (info->attributeCount > 0) { p = anv_batch_emitn(&pipeline->batch, num_dwords, - GEN8_3DSTATE_VERTEX_ELEMENTS); + GENX(3DSTATE_VERTEX_ELEMENTS)); } for (uint32_t i = 0; i < info->attributeCount; i++) { @@ -48,7 +51,7 @@ emit_vertex_input(struct anv_pipeline *pipeline, &info->pVertexAttributeDescriptions[i]; const struct anv_format *format = anv_format_for_vk_format(desc->format); - struct GEN8_VERTEX_ELEMENT_STATE element = { + struct GENX(VERTEX_ELEMENT_STATE) element = { .VertexBufferIndex = desc->binding, .Valid = true, .SourceElementFormat = format->surface_format, @@ -59,9 +62,9 @@ emit_vertex_input(struct anv_pipeline *pipeline, .Component2Control = format->num_channels >= 3 ? VFCOMP_STORE_SRC : VFCOMP_STORE_0, .Component3Control = format->num_channels >= 4 ? VFCOMP_STORE_SRC : VFCOMP_STORE_1_FP }; - GEN8_VERTEX_ELEMENT_STATE_pack(NULL, &p[1 + i * 2], &element); + GENX(VERTEX_ELEMENT_STATE_pack)(NULL, &p[1 + i * 2], &element); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VF_INSTANCING, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VF_INSTANCING), .InstancingEnable = pipeline->instancing_enable[desc->binding], .VertexElementIndex = i, /* Vulkan so far doesn't have an instance divisor, so @@ -69,7 +72,7 @@ emit_vertex_input(struct anv_pipeline *pipeline, .InstanceDataStepRate = 1); } - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VF_SGVS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VF_SGVS), .VertexIDEnable = pipeline->vs_prog_data.uses_vertexid, .VertexIDComponentNumber = 2, .VertexIDElementOffset = info->bindingCount, @@ -83,7 +86,7 @@ emit_ia_state(struct anv_pipeline *pipeline, const VkPipelineInputAssemblyStateCreateInfo *info, const struct anv_graphics_pipeline_create_info *extra) { - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VF_TOPOLOGY, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VF_TOPOLOGY), .PrimitiveTopologyType = pipeline->topology); } @@ -110,8 +113,8 @@ emit_rs_state(struct anv_pipeline *pipeline, [VK_FRONT_FACE_CW] = Clockwise }; - struct GEN8_3DSTATE_SF sf = { - GEN8_3DSTATE_SF_header, + struct GENX(3DSTATE_SF) sf = { + GENX(3DSTATE_SF_header), .ViewportTransformEnable = !(extra && extra->disable_viewport), .TriangleStripListProvokingVertexSelect = 0, .LineStripListProvokingVertexSelect = 0, @@ -122,19 +125,25 @@ emit_rs_state(struct anv_pipeline *pipeline, /* FINISHME: VkBool32 rasterizerDiscardEnable; */ - GEN8_3DSTATE_SF_pack(NULL, pipeline->gen8.sf, &sf); + GENX(3DSTATE_SF_pack)(NULL, pipeline->gen8.sf, &sf); - struct GEN8_3DSTATE_RASTER raster = { - GEN8_3DSTATE_RASTER_header, + struct GENX(3DSTATE_RASTER) raster = { + GENX(3DSTATE_RASTER_header), .FrontWinding = vk_to_gen_front_face[info->frontFace], .CullMode = vk_to_gen_cullmode[info->cullMode], .FrontFaceFillMode = vk_to_gen_fillmode[info->fillMode], .BackFaceFillMode = vk_to_gen_fillmode[info->fillMode], .ScissorRectangleEnable = !(extra && extra->disable_scissor), +#if ANV_GEN == 8 .ViewportZClipTestEnable = info->depthClipEnable +#else + /* GEN9+ splits ViewportZClipTestEnable into near and far enable bits */ + .ViewportZFarClipTestEnable = info->depthClipEnable, + .ViewportZNearClipTestEnable = info->depthClipEnable, +#endif }; - GEN8_3DSTATE_RASTER_pack(NULL, pipeline->gen8.raster, &raster); + GENX(3DSTATE_RASTER_pack)(NULL, pipeline->gen8.raster, &raster); } static void @@ -192,11 +201,11 @@ emit_cb_state(struct anv_pipeline *pipeline, [VK_BLEND_OP_MAX] = BLENDFUNCTION_MAX, }; - uint32_t num_dwords = GEN8_BLEND_STATE_length; + uint32_t num_dwords = GENX(BLEND_STATE_length); pipeline->blend_state = anv_state_pool_alloc(&device->dynamic_state_pool, num_dwords * 4, 64); - struct GEN8_BLEND_STATE blend_state = { + struct GENX(BLEND_STATE) blend_state = { .AlphaToCoverageEnable = info->alphaToCoverageEnable, .AlphaToOneEnable = info->alphaToOneEnable, }; @@ -210,7 +219,7 @@ emit_cb_state(struct anv_pipeline *pipeline, blend_state.IndependentAlphaBlendEnable = true; } - blend_state.Entry[i] = (struct GEN8_BLEND_STATE_ENTRY) { + blend_state.Entry[i] = (struct GENX(BLEND_STATE_ENTRY)) { .LogicOpEnable = info->logicOpEnable, .LogicOpFunction = vk_to_gen_logic_op[info->logicOp], .ColorBufferBlendEnable = a->blendEnable, @@ -248,9 +257,9 @@ emit_cb_state(struct anv_pipeline *pipeline, } } - GEN8_BLEND_STATE_pack(NULL, pipeline->blend_state.map, &blend_state); + GENX(BLEND_STATE_pack)(NULL, pipeline->blend_state.map, &blend_state); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_BLEND_STATE_POINTERS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_BLEND_STATE_POINTERS), .BlendStatePointer = pipeline->blend_state.offset, .BlendStatePointerValid = true); } @@ -285,6 +294,7 @@ emit_ds_state(struct anv_pipeline *pipeline, /* We're going to OR this together with the dynamic state. We need * to make sure it's initialized to something useful. */ + /* FIXME: gen9 wm_depth_stencil */ memset(pipeline->gen8.wm_depth_stencil, 0, sizeof(pipeline->gen8.wm_depth_stencil)); return; @@ -292,7 +302,7 @@ emit_ds_state(struct anv_pipeline *pipeline, /* VkBool32 depthBoundsTestEnable; // optional (depth_bounds_test) */ - struct GEN8_3DSTATE_WM_DEPTH_STENCIL wm_depth_stencil = { + struct GENX(3DSTATE_WM_DEPTH_STENCIL) wm_depth_stencil = { .DepthTestEnable = info->depthTestEnable, .DepthBufferWriteEnable = info->depthWriteEnable, .DepthTestFunction = vk_to_gen_compare_op[info->depthCompareOp], @@ -309,11 +319,11 @@ emit_ds_state(struct anv_pipeline *pipeline, .BackfaceStencilTestFunction = vk_to_gen_compare_op[info->back.stencilCompareOp], }; - GEN8_3DSTATE_WM_DEPTH_STENCIL_pack(NULL, pipeline->gen8.wm_depth_stencil, &wm_depth_stencil); + GENX(3DSTATE_WM_DEPTH_STENCIL_pack)(NULL, pipeline->gen8.wm_depth_stencil, &wm_depth_stencil); } VkResult -gen8_graphics_pipeline_create( +genX(graphics_pipeline_create)( VkDevice _device, const VkGraphicsPipelineCreateInfo* pCreateInfo, const struct anv_graphics_pipeline_create_info *extra, @@ -353,34 +363,34 @@ gen8_graphics_pipeline_create( emit_ds_state(pipeline, pCreateInfo->pDepthStencilState); emit_cb_state(pipeline, pCreateInfo->pColorBlendState); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VF_STATISTICS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VF_STATISTICS), .StatisticsEnable = true); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_HS, .Enable = false); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_TE, .TEEnable = false); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_DS, .FunctionEnable = false); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_STREAMOUT, .SOFunctionEnable = false); + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_HS), .Enable = false); + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_TE), .TEEnable = false); + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_DS), .FunctionEnable = false); + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_STREAMOUT), .SOFunctionEnable = false); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_VS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PUSH_CONSTANT_ALLOC_VS), .ConstantBufferOffset = 0, .ConstantBufferSize = 4); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_GS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PUSH_CONSTANT_ALLOC_GS), .ConstantBufferOffset = 4, .ConstantBufferSize = 4); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_PUSH_CONSTANT_ALLOC_PS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PUSH_CONSTANT_ALLOC_PS), .ConstantBufferOffset = 8, .ConstantBufferSize = 4); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_WM_CHROMAKEY, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_WM_CHROMAKEY), .ChromaKeyKillEnable = false); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_AA_LINE_PARAMETERS); + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_AA_LINE_PARAMETERS)); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_CLIP, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_CLIP), .ClipEnable = true, .ViewportXYClipTestEnable = !(extra && extra->disable_viewport), .MinimumPointWidth = 0.125, .MaximumPointWidth = 255.875); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_WM, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_WM), .StatisticsEnable = true, .LineEndCapAntialiasingRegionWidth = _05pixels, .LineAntialiasingRegionWidth = _10pixels, @@ -394,30 +404,30 @@ gen8_graphics_pipeline_create( uint32_t log2_samples = __builtin_ffs(samples) - 1; bool enable_sampling = samples > 1 ? true : false; - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_MULTISAMPLE, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_MULTISAMPLE), .PixelPositionOffsetEnable = enable_sampling, .PixelLocation = CENTER, .NumberofMultisamples = log2_samples); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_SAMPLE_MASK, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_SAMPLE_MASK), .SampleMask = 0xffff); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_URB_VS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_URB_VS), .VSURBStartingAddress = pipeline->urb.vs_start, .VSURBEntryAllocationSize = pipeline->urb.vs_size - 1, .VSNumberofURBEntries = pipeline->urb.nr_vs_entries); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_URB_GS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_URB_GS), .GSURBStartingAddress = pipeline->urb.gs_start, .GSURBEntryAllocationSize = pipeline->urb.gs_size - 1, .GSNumberofURBEntries = pipeline->urb.nr_gs_entries); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_URB_HS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_URB_HS), .HSURBStartingAddress = pipeline->urb.vs_start, .HSURBEntryAllocationSize = 0, .HSNumberofURBEntries = 0); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_URB_DS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_URB_DS), .DSURBStartingAddress = pipeline->urb.vs_start, .DSURBEntryAllocationSize = 0, .DSNumberofURBEntries = 0); @@ -427,9 +437,9 @@ gen8_graphics_pipeline_create( length = (gs_prog_data->base.vue_map.num_slots + 1) / 2 - offset; if (pipeline->gs_vec4 == NO_KERNEL) - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_GS, .Enable = false); + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_GS), .Enable = false); else - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_GS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_GS), .SingleProgramFlow = false, .KernelStartPointer = pipeline->gs_vec4, .VectorMaskEnable = Dmask, @@ -475,14 +485,14 @@ gen8_graphics_pipeline_create( length = (vue_prog_data->vue_map.num_slots + 1) / 2 - offset; if (pipeline->vs_simd8 == NO_KERNEL || (extra && extra->disable_vs)) - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VS), .FunctionEnable = false, /* Even if VS is disabled, SBE still gets the amount of * vertex data to read from this field. */ .VertexURBEntryOutputReadOffset = offset, .VertexURBEntryOutputLength = length); else - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_VS, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_VS), .KernelStartPointer = pipeline->vs_simd8, .SingleVertexDispatch = Multiple, .VectorMaskEnable = Dmask, @@ -525,8 +535,8 @@ gen8_graphics_pipeline_create( else fs_input_map = &gs_prog_data->base.vue_map; - struct GEN8_3DSTATE_SBE_SWIZ swiz = { - GEN8_3DSTATE_SBE_SWIZ_header, + struct GENX(3DSTATE_SBE_SWIZ) swiz = { + GENX(3DSTATE_SBE_SWIZ_header), }; int max_source_attr = 0; @@ -548,20 +558,59 @@ gen8_graphics_pipeline_create( swiz.Attribute[input_index].SourceAttribute = source_attr; } - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_SBE, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_SBE), .AttributeSwizzleEnable = true, .ForceVertexURBEntryReadLength = false, .ForceVertexURBEntryReadOffset = false, .VertexURBEntryReadLength = DIV_ROUND_UP(max_source_attr + 1, 2), .PointSpriteTextureCoordinateOrigin = UPPERLEFT, .NumberofSFOutputAttributes = - wm_prog_data->num_varying_inputs); + wm_prog_data->num_varying_inputs, + +#if ANV_GEN >= 9 + .Attribute0ActiveComponentFormat = ACF_XYZW, + .Attribute1ActiveComponentFormat = ACF_XYZW, + .Attribute2ActiveComponentFormat = ACF_XYZW, + .Attribute3ActiveComponentFormat = ACF_XYZW, + .Attribute4ActiveComponentFormat = ACF_XYZW, + .Attribute5ActiveComponentFormat = ACF_XYZW, + .Attribute6ActiveComponentFormat = ACF_XYZW, + .Attribute7ActiveComponentFormat = ACF_XYZW, + .Attribute8ActiveComponentFormat = ACF_XYZW, + .Attribute9ActiveComponentFormat = ACF_XYZW, + .Attribute10ActiveComponentFormat = ACF_XYZW, + .Attribute11ActiveComponentFormat = ACF_XYZW, + .Attribute12ActiveComponentFormat = ACF_XYZW, + .Attribute13ActiveComponentFormat = ACF_XYZW, + .Attribute14ActiveComponentFormat = ACF_XYZW, + .Attribute15ActiveComponentFormat = ACF_XYZW, + /* wow, much field, very attribute */ + .Attribute16ActiveComponentFormat = ACF_XYZW, + .Attribute17ActiveComponentFormat = ACF_XYZW, + .Attribute18ActiveComponentFormat = ACF_XYZW, + .Attribute19ActiveComponentFormat = ACF_XYZW, + .Attribute20ActiveComponentFormat = ACF_XYZW, + .Attribute21ActiveComponentFormat = ACF_XYZW, + .Attribute22ActiveComponentFormat = ACF_XYZW, + .Attribute23ActiveComponentFormat = ACF_XYZW, + .Attribute24ActiveComponentFormat = ACF_XYZW, + .Attribute25ActiveComponentFormat = ACF_XYZW, + .Attribute26ActiveComponentFormat = ACF_XYZW, + .Attribute27ActiveComponentFormat = ACF_XYZW, + .Attribute28ActiveComponentFormat = ACF_XYZW, + .Attribute29ActiveComponentFormat = ACF_XYZW, + .Attribute28ActiveComponentFormat = ACF_XYZW, + .Attribute29ActiveComponentFormat = ACF_XYZW, + .Attribute30ActiveComponentFormat = ACF_XYZW, +#endif + ); uint32_t *dw = anv_batch_emit_dwords(&pipeline->batch, - GEN8_3DSTATE_SBE_SWIZ_length); - GEN8_3DSTATE_SBE_SWIZ_pack(&pipeline->batch, dw, &swiz); + GENX(3DSTATE_SBE_SWIZ_length)); + GENX(3DSTATE_SBE_SWIZ_pack)(&pipeline->batch, dw, &swiz); - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_PS, + const int num_thread_bias = ANV_GEN == 8 ? 2 : 1; + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PS), .KernelStartPointer0 = pipeline->ps_ksp0, .SingleProgramFlow = false, @@ -571,7 +620,7 @@ gen8_graphics_pipeline_create( .ScratchSpaceBasePointer = pipeline->scratch_start[VK_SHADER_STAGE_FRAGMENT], .PerThreadScratchSpace = ffs(wm_prog_data->base.total_scratch / 2048), - .MaximumNumberofThreadsPerPSD = 64 - 2, + .MaximumNumberofThreadsPerPSD = 64 - num_thread_bias, .PositionXYOffsetSelect = wm_prog_data->uses_pos_offset ? POSOFFSET_SAMPLE: POSOFFSET_NONE, .PushConstantEnable = wm_prog_data->base.nr_params > 0, @@ -587,20 +636,25 @@ gen8_graphics_pipeline_create( .KernelStartPointer2 = pipeline->ps_ksp2); bool per_sample_ps = false; - anv_batch_emit(&pipeline->batch, GEN8_3DSTATE_PS_EXTRA, + anv_batch_emit(&pipeline->batch, GENX(3DSTATE_PS_EXTRA), .PixelShaderValid = true, .PixelShaderKillsPixel = wm_prog_data->uses_kill, .PixelShaderComputedDepthMode = wm_prog_data->computed_depth_mode, .AttributeEnable = wm_prog_data->num_varying_inputs > 0, .oMaskPresenttoRenderTarget = wm_prog_data->uses_omask, - .PixelShaderIsPerSample = per_sample_ps); + .PixelShaderIsPerSample = per_sample_ps, +#if ANV_GEN >= 9 + .PixelShaderPullsBary = wm_prog_data->pulls_bary, + .InputCoverageMaskState = ICMS_NONE +#endif + ); *pPipeline = anv_pipeline_to_handle(pipeline); return VK_SUCCESS; } -VkResult gen8_compute_pipeline_create( +VkResult genX(compute_pipeline_create)( VkDevice _device, const VkComputePipelineCreateInfo* pCreateInfo, VkPipeline* pPipeline) @@ -654,7 +708,7 @@ VkResult gen8_compute_pipeline_create( const struct brw_cs_prog_data *cs_prog_data = &pipeline->cs_prog_data; - anv_batch_emit(&pipeline->batch, GEN8_MEDIA_VFE_STATE, + anv_batch_emit(&pipeline->batch, GENX(MEDIA_VFE_STATE), .ScratchSpaceBasePointer = pipeline->scratch_start[VK_SHADER_STAGE_COMPUTE], .PerThreadScratchSpace = ffs(cs_prog_data->base.total_scratch / 2048), .ScratchSpaceBasePointerHigh = 0, @@ -663,7 +717,9 @@ VkResult gen8_compute_pipeline_create( .MaximumNumberofThreads = device->info.max_cs_threads - 1, .NumberofURBEntries = 2, .ResetGatewayTimer = true, +#if ANV_GEN == 8 .BypassGatewayControl = true, +#endif .URBEntryAllocationSize = 2, .CURBEAllocationSize = 0); |