From 25816bea4153231b7a5a492e9d59cc21fe32ed3e Mon Sep 17 00:00:00 2001 From: Evan Hemsley <2342303+thatcosmonaut@users.noreply.github.com> Date: Mon, 2 Jun 2025 13:39:58 -0700 Subject: [PATCH] GPU: Binding validation and prevent null dereference if expected binding is missing (#13164) --- src/gpu/SDL_gpu.c | 251 +++++++++++++++++++++----- src/gpu/SDL_sysgpu.h | 53 +++++- src/gpu/d3d12/SDL_gpu_d3d12.c | 304 +++++++++++++++++--------------- src/gpu/metal/SDL_gpu_metal.m | 134 +++++++------- src/gpu/vulkan/SDL_gpu_vulkan.c | 183 +++++++++++-------- 5 files changed, 588 insertions(+), 337 deletions(-) diff --git a/src/gpu/SDL_gpu.c b/src/gpu/SDL_gpu.c index a28346dde0..e1ad66c00b 100644 --- a/src/gpu/SDL_gpu.c +++ b/src/gpu/SDL_gpu.c @@ -93,10 +93,10 @@ } \ } -#define CHECK_GRAPHICS_PIPELINE_BOUND \ - if (!((CommandBufferCommonHeader *)RENDERPASS_COMMAND_BUFFER)->graphics_pipeline_bound) { \ - SDL_assert_release(!"Graphics pipeline not bound!"); \ - return; \ +#define CHECK_GRAPHICS_PIPELINE_BOUND \ + if (!((RenderPass *)render_pass)->graphics_pipeline) { \ + SDL_assert_release(!"Graphics pipeline not bound!"); \ + return; \ } #define CHECK_COMPUTEPASS \ @@ -106,7 +106,7 @@ } #define CHECK_COMPUTE_PIPELINE_BOUND \ - if (!((CommandBufferCommonHeader *)COMPUTEPASS_COMMAND_BUFFER)->compute_pipeline_bound) { \ + if (!((ComputePass *)compute_pass)->compute_pipeline) { \ SDL_assert_release(!"Compute pipeline not bound!"); \ return; \ } @@ -174,12 +174,18 @@ #define RENDERPASS_DEVICE \ ((CommandBufferCommonHeader *)RENDERPASS_COMMAND_BUFFER)->device +#define RENDERPASS_BOUND_PIPELINE \ + ((RenderPass *)render_pass)->graphics_pipeline + #define COMPUTEPASS_COMMAND_BUFFER \ ((Pass *)compute_pass)->command_buffer #define COMPUTEPASS_DEVICE \ ((CommandBufferCommonHeader *)COMPUTEPASS_COMMAND_BUFFER)->device +#define COMPUTEPASS_BOUND_PIPELINE \ + ((ComputePass *)compute_pass)->compute_pipeline + #define COPYPASS_COMMAND_BUFFER \ ((Pass *)copy_pass)->command_buffer @@ -511,6 +517,73 @@ void SDL_GPU_BlitCommon( SDL_EndGPURenderPass(render_pass); } +static void SDL_GPU_CheckGraphicsBindings(SDL_GPURenderPass *render_pass) +{ + RenderPass *rp = (RenderPass *)render_pass; + GraphicsPipelineCommonHeader *pipeline = (GraphicsPipelineCommonHeader *)RENDERPASS_BOUND_PIPELINE; + for (Uint32 i = 0; i < pipeline->num_vertex_samplers; i += 1) { + if (!rp->vertex_sampler_bound[i]) { + SDL_assert_release(!"Missing vertex sampler binding!"); + } + } + for (Uint32 i = 0; i < pipeline->num_vertex_storage_textures; i += 1) { + if (!rp->vertex_storage_texture_bound[i]) { + SDL_assert_release(!"Missing vertex storage texture binding!"); + } + } + for (Uint32 i = 0; i < pipeline->num_vertex_storage_buffers; i += 1) { + if (!rp->vertex_storage_buffer_bound[i]) { + SDL_assert_release(!"Missing vertex storage buffer binding!"); + } + } + for (Uint32 i = 0; i < pipeline->num_fragment_samplers; i += 1) { + if (!rp->fragment_sampler_bound[i]) { + SDL_assert_release(!"Missing fragment sampler binding!"); + } + } + for (Uint32 i = 0; i < pipeline->num_fragment_storage_textures; i += 1) { + if (!rp->fragment_storage_texture_bound[i]) { + SDL_assert_release(!"Missing fragment storage texture binding!"); + } + } + for (Uint32 i = 0; i < pipeline->num_fragment_storage_buffers; i += 1) { + if (!rp->fragment_storage_buffer_bound[i]) { + SDL_assert_release(!"Missing fragment storage buffer binding!"); + } + } +} + +static void SDL_GPU_CheckComputeBindings(SDL_GPUComputePass *compute_pass) +{ + ComputePass *cp = (ComputePass *)compute_pass; + ComputePipelineCommonHeader *pipeline = (ComputePipelineCommonHeader *)COMPUTEPASS_BOUND_PIPELINE; + for (Uint32 i = 0; i < pipeline->numSamplers; i += 1) { + if (!cp->sampler_bound[i]) { + SDL_assert_release(!"Missing compute sampler binding!"); + } + } + for (Uint32 i = 0; i < pipeline->numReadonlyStorageTextures; i += 1) { + if (!cp->read_only_storage_texture_bound[i]) { + SDL_assert_release(!"Missing compute readonly storage texture binding!"); + } + } + for (Uint32 i = 0; i < pipeline->numReadonlyStorageBuffers; i += 1) { + if (!cp->read_only_storage_buffer_bound[i]) { + SDL_assert_release(!"Missing compute readonly storage buffer binding!"); + } + } + for (Uint32 i = 0; i < pipeline->numReadWriteStorageTextures; i += 1) { + if (!cp->read_write_storage_texture_bound[i]) { + SDL_assert_release(!"Missing compute read-write storage texture binding!"); + } + } + for (Uint32 i = 0; i < pipeline->numReadWriteStorageBuffers; i += 1) { + if (!cp->read_write_storage_buffer_bound[i]) { + SDL_assert_release(!"Missing compute read-write storage buffer bbinding!"); + } + } +} + // Driver Functions #ifndef SDL_GPU_DISABLED @@ -1489,15 +1562,29 @@ SDL_GPUCommandBuffer *SDL_AcquireGPUCommandBuffer( commandBufferHeader = (CommandBufferCommonHeader *)command_buffer; commandBufferHeader->device = device; commandBufferHeader->render_pass.command_buffer = command_buffer; - commandBufferHeader->render_pass.in_progress = false; - commandBufferHeader->graphics_pipeline_bound = false; commandBufferHeader->compute_pass.command_buffer = command_buffer; - commandBufferHeader->compute_pass.in_progress = false; - commandBufferHeader->compute_pipeline_bound = false; commandBufferHeader->copy_pass.command_buffer = command_buffer; - commandBufferHeader->copy_pass.in_progress = false; - commandBufferHeader->swapchain_texture_acquired = false; - commandBufferHeader->submitted = false; + + if (device->debug_mode) { + commandBufferHeader->render_pass.in_progress = false; + commandBufferHeader->render_pass.graphics_pipeline = NULL; + commandBufferHeader->compute_pass.in_progress = false; + commandBufferHeader->compute_pass.compute_pipeline = NULL; + commandBufferHeader->copy_pass.in_progress = false; + commandBufferHeader->swapchain_texture_acquired = false; + commandBufferHeader->submitted = false; + SDL_zeroa(commandBufferHeader->render_pass.vertex_sampler_bound); + SDL_zeroa(commandBufferHeader->render_pass.vertex_storage_texture_bound); + SDL_zeroa(commandBufferHeader->render_pass.vertex_storage_buffer_bound); + SDL_zeroa(commandBufferHeader->render_pass.fragment_sampler_bound); + SDL_zeroa(commandBufferHeader->render_pass.fragment_storage_texture_bound); + SDL_zeroa(commandBufferHeader->render_pass.fragment_storage_buffer_bound); + SDL_zeroa(commandBufferHeader->compute_pass.sampler_bound); + SDL_zeroa(commandBufferHeader->compute_pass.read_only_storage_texture_bound); + SDL_zeroa(commandBufferHeader->compute_pass.read_only_storage_buffer_bound); + SDL_zeroa(commandBufferHeader->compute_pass.read_write_storage_texture_bound); + SDL_zeroa(commandBufferHeader->compute_pass.read_write_storage_buffer_bound); + } return command_buffer; } @@ -1688,14 +1775,18 @@ SDL_GPURenderPass *SDL_BeginGPURenderPass( depth_stencil_target_info); commandBufferHeader = (CommandBufferCommonHeader *)command_buffer; - commandBufferHeader->render_pass.in_progress = true; - for (Uint32 i = 0; i < num_color_targets; i += 1) { - commandBufferHeader->render_pass.color_targets[i] = color_target_infos[i].texture; - } - commandBufferHeader->render_pass.num_color_targets = num_color_targets; - if (depth_stencil_target_info != NULL) { - commandBufferHeader->render_pass.depth_stencil_target = depth_stencil_target_info->texture; + + if (COMMAND_BUFFER_DEVICE->debug_mode) { + commandBufferHeader->render_pass.in_progress = true; + for (Uint32 i = 0; i < num_color_targets; i += 1) { + commandBufferHeader->render_pass.color_targets[i] = color_target_infos[i].texture; + } + commandBufferHeader->render_pass.num_color_targets = num_color_targets; + if (depth_stencil_target_info != NULL) { + commandBufferHeader->render_pass.depth_stencil_target = depth_stencil_target_info->texture; + } } + return (SDL_GPURenderPass *)&(commandBufferHeader->render_pass); } @@ -1703,8 +1794,6 @@ void SDL_BindGPUGraphicsPipeline( SDL_GPURenderPass *render_pass, SDL_GPUGraphicsPipeline *graphics_pipeline) { - CommandBufferCommonHeader *commandBufferHeader; - if (render_pass == NULL) { SDL_InvalidParamError("render_pass"); return; @@ -1718,8 +1807,10 @@ void SDL_BindGPUGraphicsPipeline( RENDERPASS_COMMAND_BUFFER, graphics_pipeline); - commandBufferHeader = (CommandBufferCommonHeader *)RENDERPASS_COMMAND_BUFFER; - commandBufferHeader->graphics_pipeline_bound = true; + + if (RENDERPASS_DEVICE->debug_mode) { + RENDERPASS_BOUND_PIPELINE = graphics_pipeline; + } } void SDL_SetGPUViewport( @@ -1874,6 +1965,10 @@ void SDL_BindGPUVertexSamplers( { CHECK_SAMPLER_TEXTURES } + + for (Uint32 i = 0; i < num_bindings; i += 1) { + ((RenderPass *)render_pass)->vertex_sampler_bound[first_slot + i] = true; + } } RENDERPASS_DEVICE->BindVertexSamplers( @@ -1901,6 +1996,10 @@ void SDL_BindGPUVertexStorageTextures( if (RENDERPASS_DEVICE->debug_mode) { CHECK_RENDERPASS CHECK_STORAGE_TEXTURES + + for (Uint32 i = 0; i < num_bindings; i += 1) { + ((RenderPass *)render_pass)->vertex_storage_texture_bound[first_slot + i] = true; + } } RENDERPASS_DEVICE->BindVertexStorageTextures( @@ -1927,6 +2026,10 @@ void SDL_BindGPUVertexStorageBuffers( if (RENDERPASS_DEVICE->debug_mode) { CHECK_RENDERPASS + + for (Uint32 i = 0; i < num_bindings; i += 1) { + ((RenderPass *)render_pass)->vertex_storage_buffer_bound[first_slot + i] = true; + } } RENDERPASS_DEVICE->BindVertexStorageBuffers( @@ -1954,10 +2057,13 @@ void SDL_BindGPUFragmentSamplers( if (RENDERPASS_DEVICE->debug_mode) { CHECK_RENDERPASS - if (!((CommandBufferCommonHeader*)RENDERPASS_COMMAND_BUFFER)->ignore_render_pass_texture_validation) - { + if (!((CommandBufferCommonHeader*)RENDERPASS_COMMAND_BUFFER)->ignore_render_pass_texture_validation) { CHECK_SAMPLER_TEXTURES } + + for (Uint32 i = 0; i < num_bindings; i += 1) { + ((RenderPass *)render_pass)->fragment_sampler_bound[first_slot + i] = true; + } } RENDERPASS_DEVICE->BindFragmentSamplers( @@ -1985,6 +2091,10 @@ void SDL_BindGPUFragmentStorageTextures( if (RENDERPASS_DEVICE->debug_mode) { CHECK_RENDERPASS CHECK_STORAGE_TEXTURES + + for (Uint32 i = 0; i < num_bindings; i += 1) { + ((RenderPass *)render_pass)->fragment_storage_texture_bound[first_slot + i] = true; + } } RENDERPASS_DEVICE->BindFragmentStorageTextures( @@ -2011,6 +2121,10 @@ void SDL_BindGPUFragmentStorageBuffers( if (RENDERPASS_DEVICE->debug_mode) { CHECK_RENDERPASS + + for (Uint32 i = 0; i < num_bindings; i += 1) { + ((RenderPass *)render_pass)->fragment_storage_buffer_bound[first_slot + i] = true; + } } RENDERPASS_DEVICE->BindFragmentStorageBuffers( @@ -2036,6 +2150,7 @@ void SDL_DrawGPUIndexedPrimitives( if (RENDERPASS_DEVICE->debug_mode) { CHECK_RENDERPASS CHECK_GRAPHICS_PIPELINE_BOUND + SDL_GPU_CheckGraphicsBindings(render_pass); } RENDERPASS_DEVICE->DrawIndexedPrimitives( @@ -2062,6 +2177,7 @@ void SDL_DrawGPUPrimitives( if (RENDERPASS_DEVICE->debug_mode) { CHECK_RENDERPASS CHECK_GRAPHICS_PIPELINE_BOUND + SDL_GPU_CheckGraphicsBindings(render_pass); } RENDERPASS_DEVICE->DrawPrimitives( @@ -2090,6 +2206,7 @@ void SDL_DrawGPUPrimitivesIndirect( if (RENDERPASS_DEVICE->debug_mode) { CHECK_RENDERPASS CHECK_GRAPHICS_PIPELINE_BOUND + SDL_GPU_CheckGraphicsBindings(render_pass); } RENDERPASS_DEVICE->DrawPrimitivesIndirect( @@ -2117,6 +2234,7 @@ void SDL_DrawGPUIndexedPrimitivesIndirect( if (RENDERPASS_DEVICE->debug_mode) { CHECK_RENDERPASS CHECK_GRAPHICS_PIPELINE_BOUND + SDL_GPU_CheckGraphicsBindings(render_pass); } RENDERPASS_DEVICE->DrawIndexedPrimitivesIndirect( @@ -2130,6 +2248,7 @@ void SDL_EndGPURenderPass( SDL_GPURenderPass *render_pass) { CommandBufferCommonHeader *commandBufferCommonHeader; + commandBufferCommonHeader = (CommandBufferCommonHeader *)RENDERPASS_COMMAND_BUFFER; if (render_pass == NULL) { SDL_InvalidParamError("render_pass"); @@ -2143,15 +2262,22 @@ void SDL_EndGPURenderPass( RENDERPASS_DEVICE->EndRenderPass( RENDERPASS_COMMAND_BUFFER); - commandBufferCommonHeader = (CommandBufferCommonHeader *)RENDERPASS_COMMAND_BUFFER; - commandBufferCommonHeader->render_pass.in_progress = false; - for (Uint32 i = 0; i < MAX_COLOR_TARGET_BINDINGS; i += 1) - { - commandBufferCommonHeader->render_pass.color_targets[i] = NULL; + if (RENDERPASS_DEVICE->debug_mode) { + commandBufferCommonHeader->render_pass.in_progress = false; + for (Uint32 i = 0; i < MAX_COLOR_TARGET_BINDINGS; i += 1) + { + commandBufferCommonHeader->render_pass.color_targets[i] = NULL; + } + commandBufferCommonHeader->render_pass.num_color_targets = 0; + commandBufferCommonHeader->render_pass.depth_stencil_target = NULL; + commandBufferCommonHeader->render_pass.graphics_pipeline = NULL; + SDL_zeroa(commandBufferCommonHeader->render_pass.vertex_sampler_bound); + SDL_zeroa(commandBufferCommonHeader->render_pass.vertex_storage_texture_bound); + SDL_zeroa(commandBufferCommonHeader->render_pass.vertex_storage_buffer_bound); + SDL_zeroa(commandBufferCommonHeader->render_pass.fragment_sampler_bound); + SDL_zeroa(commandBufferCommonHeader->render_pass.fragment_storage_texture_bound); + SDL_zeroa(commandBufferCommonHeader->render_pass.fragment_storage_buffer_bound); } - commandBufferCommonHeader->render_pass.num_color_targets = 0; - commandBufferCommonHeader->render_pass.depth_stencil_target = NULL; - commandBufferCommonHeader->graphics_pipeline_bound = false; } // Compute Pass @@ -2218,7 +2344,19 @@ SDL_GPUComputePass *SDL_BeginGPUComputePass( num_storage_buffer_bindings); commandBufferHeader = (CommandBufferCommonHeader *)command_buffer; - commandBufferHeader->compute_pass.in_progress = true; + + if (COMMAND_BUFFER_DEVICE->debug_mode) { + commandBufferHeader->compute_pass.in_progress = true; + + for (Uint32 i = 0; i < num_storage_texture_bindings; i += 1) { + commandBufferHeader->compute_pass.read_write_storage_texture_bound[i] = true; + } + + for (Uint32 i = 0; i < num_storage_buffer_bindings; i += 1) { + commandBufferHeader->compute_pass.read_write_storage_buffer_bound[i] = true; + } + } + return (SDL_GPUComputePass *)&(commandBufferHeader->compute_pass); } @@ -2226,8 +2364,6 @@ void SDL_BindGPUComputePipeline( SDL_GPUComputePass *compute_pass, SDL_GPUComputePipeline *compute_pipeline) { - CommandBufferCommonHeader *commandBufferHeader; - if (compute_pass == NULL) { SDL_InvalidParamError("compute_pass"); return; @@ -2245,8 +2381,10 @@ void SDL_BindGPUComputePipeline( COMPUTEPASS_COMMAND_BUFFER, compute_pipeline); - commandBufferHeader = (CommandBufferCommonHeader *)COMPUTEPASS_COMMAND_BUFFER; - commandBufferHeader->compute_pipeline_bound = true; + + if (COMPUTEPASS_DEVICE->debug_mode) { + COMPUTEPASS_BOUND_PIPELINE = compute_pipeline; + } } void SDL_BindGPUComputeSamplers( @@ -2266,6 +2404,10 @@ void SDL_BindGPUComputeSamplers( if (COMPUTEPASS_DEVICE->debug_mode) { CHECK_COMPUTEPASS + + for (Uint32 i = 0; i < num_bindings; i += 1) { + ((ComputePass *)compute_pass)->sampler_bound[first_slot + i] = true; + } } COMPUTEPASS_DEVICE->BindComputeSamplers( @@ -2292,6 +2434,10 @@ void SDL_BindGPUComputeStorageTextures( if (COMPUTEPASS_DEVICE->debug_mode) { CHECK_COMPUTEPASS + + for (Uint32 i = 0; i < num_bindings; i += 1) { + ((ComputePass *)compute_pass)->read_only_storage_texture_bound[first_slot + i] = true; + } } COMPUTEPASS_DEVICE->BindComputeStorageTextures( @@ -2318,6 +2464,10 @@ void SDL_BindGPUComputeStorageBuffers( if (COMPUTEPASS_DEVICE->debug_mode) { CHECK_COMPUTEPASS + + for (Uint32 i = 0; i < num_bindings; i += 1) { + ((ComputePass *)compute_pass)->read_only_storage_buffer_bound[first_slot + i] = true; + } } COMPUTEPASS_DEVICE->BindComputeStorageBuffers( @@ -2341,6 +2491,7 @@ void SDL_DispatchGPUCompute( if (COMPUTEPASS_DEVICE->debug_mode) { CHECK_COMPUTEPASS CHECK_COMPUTE_PIPELINE_BOUND + SDL_GPU_CheckComputeBindings(compute_pass); } COMPUTEPASS_DEVICE->DispatchCompute( @@ -2363,6 +2514,7 @@ void SDL_DispatchGPUComputeIndirect( if (COMPUTEPASS_DEVICE->debug_mode) { CHECK_COMPUTEPASS CHECK_COMPUTE_PIPELINE_BOUND + SDL_GPU_CheckComputeBindings(compute_pass); } COMPUTEPASS_DEVICE->DispatchComputeIndirect( @@ -2388,9 +2540,16 @@ void SDL_EndGPUComputePass( COMPUTEPASS_DEVICE->EndComputePass( COMPUTEPASS_COMMAND_BUFFER); - commandBufferCommonHeader = (CommandBufferCommonHeader *)COMPUTEPASS_COMMAND_BUFFER; - commandBufferCommonHeader->compute_pass.in_progress = false; - commandBufferCommonHeader->compute_pipeline_bound = false; + if (COMPUTEPASS_DEVICE->debug_mode) { + commandBufferCommonHeader = (CommandBufferCommonHeader *)COMPUTEPASS_COMMAND_BUFFER; + commandBufferCommonHeader->compute_pass.in_progress = false; + commandBufferCommonHeader->compute_pass.compute_pipeline = false; + SDL_zeroa(commandBufferCommonHeader->compute_pass.sampler_bound); + SDL_zeroa(commandBufferCommonHeader->compute_pass.read_only_storage_texture_bound); + SDL_zeroa(commandBufferCommonHeader->compute_pass.read_only_storage_buffer_bound); + SDL_zeroa(commandBufferCommonHeader->compute_pass.read_write_storage_texture_bound); + SDL_zeroa(commandBufferCommonHeader->compute_pass.read_write_storage_buffer_bound); + } } // TransferBuffer Data @@ -2448,7 +2607,11 @@ SDL_GPUCopyPass *SDL_BeginGPUCopyPass( command_buffer); commandBufferHeader = (CommandBufferCommonHeader *)command_buffer; - commandBufferHeader->copy_pass.in_progress = true; + + if (COMMAND_BUFFER_DEVICE->debug_mode) { + commandBufferHeader->copy_pass.in_progress = true; + } + return (SDL_GPUCopyPass *)&(commandBufferHeader->copy_pass); } @@ -2699,7 +2862,9 @@ void SDL_EndGPUCopyPass( COPYPASS_DEVICE->EndCopyPass( COPYPASS_COMMAND_BUFFER); - ((CommandBufferCommonHeader *)COPYPASS_COMMAND_BUFFER)->copy_pass.in_progress = false; + if (COPYPASS_DEVICE->debug_mode) { + ((CommandBufferCommonHeader *)COPYPASS_COMMAND_BUFFER)->copy_pass.in_progress = false; + } } void SDL_GenerateMipmapsForGPUTexture( diff --git a/src/gpu/SDL_sysgpu.h b/src/gpu/SDL_sysgpu.h index 3c2c786600..21dcfbd54a 100644 --- a/src/gpu/SDL_sysgpu.h +++ b/src/gpu/SDL_sysgpu.h @@ -47,6 +47,20 @@ typedef struct Pass bool in_progress; } Pass; +typedef struct ComputePass +{ + SDL_GPUCommandBuffer *command_buffer; + bool in_progress; + + SDL_GPUComputePipeline *compute_pipeline; + + bool sampler_bound[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + bool read_only_storage_texture_bound[MAX_STORAGE_TEXTURES_PER_STAGE]; + bool read_only_storage_buffer_bound[MAX_STORAGE_BUFFERS_PER_STAGE]; + bool read_write_storage_texture_bound[MAX_COMPUTE_WRITE_TEXTURES]; + bool read_write_storage_buffer_bound[MAX_COMPUTE_WRITE_BUFFERS]; +} ComputePass; + typedef struct RenderPass { SDL_GPUCommandBuffer *command_buffer; @@ -54,15 +68,25 @@ typedef struct RenderPass SDL_GPUTexture *color_targets[MAX_COLOR_TARGET_BINDINGS]; Uint32 num_color_targets; SDL_GPUTexture *depth_stencil_target; + + SDL_GPUGraphicsPipeline *graphics_pipeline; + + bool vertex_sampler_bound[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + bool vertex_storage_texture_bound[MAX_STORAGE_TEXTURES_PER_STAGE]; + bool vertex_storage_buffer_bound[MAX_STORAGE_BUFFERS_PER_STAGE]; + + bool fragment_sampler_bound[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + bool fragment_storage_texture_bound[MAX_STORAGE_TEXTURES_PER_STAGE]; + bool fragment_storage_buffer_bound[MAX_STORAGE_BUFFERS_PER_STAGE]; } RenderPass; typedef struct CommandBufferCommonHeader { SDL_GPUDevice *device; + RenderPass render_pass; - bool graphics_pipeline_bound; - Pass compute_pass; - bool compute_pipeline_bound; + ComputePass compute_pass; + Pass copy_pass; bool swapchain_texture_acquired; bool submitted; @@ -75,6 +99,29 @@ typedef struct TextureCommonHeader SDL_GPUTextureCreateInfo info; } TextureCommonHeader; +typedef struct GraphicsPipelineCommonHeader +{ + Uint32 num_vertex_samplers; + Uint32 num_vertex_storage_textures; + Uint32 num_vertex_storage_buffers; + Uint32 num_vertex_uniform_buffers; + + Uint32 num_fragment_samplers; + Uint32 num_fragment_storage_textures; + Uint32 num_fragment_storage_buffers; + Uint32 num_fragment_uniform_buffers; +} GraphicsPipelineCommonHeader; + +typedef struct ComputePipelineCommonHeader +{ + Uint32 numSamplers; + Uint32 numReadonlyStorageTextures; + Uint32 numReadonlyStorageBuffers; + Uint32 numReadWriteStorageTextures; + Uint32 numReadWriteStorageBuffers; + Uint32 numUniformBuffers; +} ComputePipelineCommonHeader; + typedef struct BlitFragmentUniforms { // texcoord space diff --git a/src/gpu/d3d12/SDL_gpu_d3d12.c b/src/gpu/d3d12/SDL_gpu_d3d12.c index 96a33634df..a92215910c 100644 --- a/src/gpu/d3d12/SDL_gpu_d3d12.c +++ b/src/gpu/d3d12/SDL_gpu_d3d12.c @@ -1015,26 +1015,38 @@ struct D3D12CommandBuffer Uint32 vertexBufferOffsets[MAX_VERTEX_BUFFERS]; Uint32 vertexBufferCount; - D3D12Texture *vertexSamplerTextures[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - D3D12Sampler *vertexSamplers[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - D3D12Texture *vertexStorageTextures[MAX_STORAGE_TEXTURES_PER_STAGE]; - D3D12Buffer *vertexStorageBuffers[MAX_STORAGE_BUFFERS_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE vertexSamplerTextureDescriptorHandles[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE vertexSamplerDescriptorHandles[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE vertexStorageTextureDescriptorHandles[MAX_STORAGE_TEXTURES_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE vertexStorageBufferDescriptorHandles[MAX_STORAGE_BUFFERS_PER_STAGE]; + D3D12UniformBuffer *vertexUniformBuffers[MAX_UNIFORM_BUFFERS_PER_STAGE]; - D3D12Texture *fragmentSamplerTextures[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - D3D12Sampler *fragmentSamplers[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - D3D12Texture *fragmentStorageTextures[MAX_STORAGE_TEXTURES_PER_STAGE]; - D3D12Buffer *fragmentStorageBuffers[MAX_STORAGE_BUFFERS_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE fragmentSamplerTextureDescriptorHandles[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE fragmentSamplerDescriptorHandles[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE fragmentStorageTextureDescriptorHandles[MAX_STORAGE_TEXTURES_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE fragmentStorageBufferDescriptorHandles[MAX_STORAGE_BUFFERS_PER_STAGE]; + D3D12UniformBuffer *fragmentUniformBuffers[MAX_UNIFORM_BUFFERS_PER_STAGE]; - D3D12Texture *computeSamplerTextures[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - D3D12Sampler *computeSamplers[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE computeSamplerTextureDescriptorHandles[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE computeSamplerDescriptorHandles[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE computeReadOnlyStorageTextureDescriptorHandles[MAX_STORAGE_TEXTURES_PER_STAGE]; + D3D12_CPU_DESCRIPTOR_HANDLE computeReadOnlyStorageBufferDescriptorHandles[MAX_STORAGE_BUFFERS_PER_STAGE]; + + // Track these separately because barriers can happen mid compute pass D3D12Texture *computeReadOnlyStorageTextures[MAX_STORAGE_TEXTURES_PER_STAGE]; D3D12Buffer *computeReadOnlyStorageBuffers[MAX_STORAGE_BUFFERS_PER_STAGE]; + + D3D12_CPU_DESCRIPTOR_HANDLE computeReadWriteStorageTextureDescriptorHandles[MAX_COMPUTE_WRITE_TEXTURES]; + D3D12_CPU_DESCRIPTOR_HANDLE computeReadWriteStorageBufferDescriptorHandles[MAX_COMPUTE_WRITE_BUFFERS]; + + // Track these separately because they are bound when the compute pass begins D3D12TextureSubresource *computeReadWriteStorageTextureSubresources[MAX_COMPUTE_WRITE_TEXTURES]; Uint32 computeReadWriteStorageTextureSubresourceCount; D3D12Buffer *computeReadWriteStorageBuffers[MAX_COMPUTE_WRITE_BUFFERS]; Uint32 computeReadWriteStorageBufferCount; + D3D12UniformBuffer *computeUniformBuffers[MAX_UNIFORM_BUFFERS_PER_STAGE]; // Resource tracking @@ -1098,22 +1110,14 @@ typedef struct D3D12GraphicsRootSignature struct D3D12GraphicsPipeline { + GraphicsPipelineCommonHeader header; + ID3D12PipelineState *pipelineState; D3D12GraphicsRootSignature *rootSignature; SDL_GPUPrimitiveType primitiveType; Uint32 vertexStrides[MAX_VERTEX_BUFFERS]; - Uint32 vertexSamplerCount; - Uint32 vertexUniformBufferCount; - Uint32 vertexStorageBufferCount; - Uint32 vertexStorageTextureCount; - - Uint32 fragmentSamplerCount; - Uint32 fragmentUniformBufferCount; - Uint32 fragmentStorageBufferCount; - Uint32 fragmentStorageTextureCount; - SDL_AtomicInt referenceCount; }; @@ -1132,16 +1136,11 @@ typedef struct D3D12ComputeRootSignature struct D3D12ComputePipeline { + ComputePipelineCommonHeader header; + ID3D12PipelineState *pipelineState; D3D12ComputeRootSignature *rootSignature; - Uint32 numSamplers; - Uint32 numReadOnlyStorageTextures; - Uint32 numReadOnlyStorageBuffers; - Uint32 numReadWriteStorageTextures; - Uint32 numReadWriteStorageBuffers; - Uint32 numUniformBuffers; - SDL_AtomicInt referenceCount; }; @@ -2886,12 +2885,12 @@ static SDL_GPUComputePipeline *D3D12_CreateComputePipeline( computePipeline->pipelineState = pipelineState; computePipeline->rootSignature = rootSignature; - computePipeline->numSamplers = createinfo->num_samplers; - computePipeline->numReadOnlyStorageTextures = createinfo->num_readonly_storage_textures; - computePipeline->numReadOnlyStorageBuffers = createinfo->num_readonly_storage_buffers; - computePipeline->numReadWriteStorageTextures = createinfo->num_readwrite_storage_textures; - computePipeline->numReadWriteStorageBuffers = createinfo->num_readwrite_storage_buffers; - computePipeline->numUniformBuffers = createinfo->num_uniform_buffers; + computePipeline->header.numSamplers = createinfo->num_samplers; + computePipeline->header.numReadonlyStorageTextures = createinfo->num_readonly_storage_textures; + computePipeline->header.numReadonlyStorageBuffers = createinfo->num_readonly_storage_buffers; + computePipeline->header.numReadWriteStorageTextures = createinfo->num_readwrite_storage_textures; + computePipeline->header.numReadWriteStorageBuffers = createinfo->num_readwrite_storage_buffers; + computePipeline->header.numUniformBuffers = createinfo->num_uniform_buffers; SDL_SetAtomicInt(&computePipeline->referenceCount, 0); if (renderer->debug_mode && SDL_HasProperty(createinfo->props, SDL_PROP_GPU_COMPUTEPIPELINE_CREATE_NAME_STRING)) { @@ -3172,15 +3171,15 @@ static SDL_GPUGraphicsPipeline *D3D12_CreateGraphicsPipeline( pipeline->primitiveType = createinfo->primitive_type; - pipeline->vertexSamplerCount = vertShader->num_samplers; - pipeline->vertexStorageTextureCount = vertShader->numStorageTextures; - pipeline->vertexStorageBufferCount = vertShader->numStorageBuffers; - pipeline->vertexUniformBufferCount = vertShader->numUniformBuffers; + pipeline->header.num_vertex_samplers = vertShader->num_samplers; + pipeline->header.num_vertex_storage_textures = vertShader->numStorageTextures; + pipeline->header.num_vertex_storage_buffers = vertShader->numStorageBuffers; + pipeline->header.num_vertex_uniform_buffers = vertShader->numUniformBuffers; - pipeline->fragmentSamplerCount = fragShader->num_samplers; - pipeline->fragmentStorageTextureCount = fragShader->numStorageTextures; - pipeline->fragmentStorageBufferCount = fragShader->numStorageBuffers; - pipeline->fragmentUniformBufferCount = fragShader->numUniformBuffers; + pipeline->header.num_fragment_samplers = fragShader->num_samplers; + pipeline->header.num_fragment_storage_textures = fragShader->numStorageTextures; + pipeline->header.num_fragment_storage_buffers = fragShader->numStorageBuffers; + pipeline->header.num_fragment_uniform_buffers = fragShader->numUniformBuffers; SDL_SetAtomicInt(&pipeline->referenceCount, 0); @@ -4637,14 +4636,14 @@ static void D3D12_BindGraphicsPipeline( d3d12CommandBuffer->needFragmentUniformBufferBind[i] = true; } - for (i = 0; i < pipeline->vertexUniformBufferCount; i += 1) { + for (i = 0; i < pipeline->header.num_vertex_uniform_buffers; i += 1) { if (d3d12CommandBuffer->vertexUniformBuffers[i] == NULL) { d3d12CommandBuffer->vertexUniformBuffers[i] = D3D12_INTERNAL_AcquireUniformBufferFromPool( d3d12CommandBuffer); } } - for (i = 0; i < pipeline->fragmentUniformBufferCount; i += 1) { + for (i = 0; i < pipeline->header.num_fragment_uniform_buffers; i += 1) { if (d3d12CommandBuffer->fragmentUniformBuffers[i] == NULL) { d3d12CommandBuffer->fragmentUniformBuffers[i] = D3D12_INTERNAL_AcquireUniformBufferFromPool( d3d12CommandBuffer); @@ -4711,21 +4710,21 @@ static void D3D12_BindVertexSamplers( D3D12TextureContainer *container = (D3D12TextureContainer *)textureSamplerBindings[i].texture; D3D12Sampler *sampler = (D3D12Sampler *)textureSamplerBindings[i].sampler; - if (d3d12CommandBuffer->vertexSamplers[firstSlot + i] != sampler) { + if (d3d12CommandBuffer->vertexSamplerDescriptorHandles[firstSlot + i].ptr != sampler->handle.cpuHandle.ptr) { D3D12_INTERNAL_TrackSampler( d3d12CommandBuffer, sampler); - d3d12CommandBuffer->vertexSamplers[firstSlot + i] = sampler; + d3d12CommandBuffer->vertexSamplerDescriptorHandles[firstSlot + i] = sampler->handle.cpuHandle; d3d12CommandBuffer->needVertexSamplerBind = true; } - if (d3d12CommandBuffer->vertexSamplerTextures[firstSlot + i] != container->activeTexture) { + if (d3d12CommandBuffer->vertexSamplerTextureDescriptorHandles[firstSlot + i].ptr != container->activeTexture->srvHandle.cpuHandle.ptr) { D3D12_INTERNAL_TrackTexture( d3d12CommandBuffer, container->activeTexture); - d3d12CommandBuffer->vertexSamplerTextures[firstSlot + i] = container->activeTexture; + d3d12CommandBuffer->vertexSamplerTextureDescriptorHandles[firstSlot + i] = container->activeTexture->srvHandle.cpuHandle; d3d12CommandBuffer->needVertexSamplerBind = true; } } @@ -4743,10 +4742,10 @@ static void D3D12_BindVertexStorageTextures( D3D12TextureContainer *container = (D3D12TextureContainer *)storageTextures[i]; D3D12Texture *texture = container->activeTexture; - if (d3d12CommandBuffer->vertexStorageTextures[firstSlot + i] != texture) { + if (d3d12CommandBuffer->vertexStorageTextureDescriptorHandles[firstSlot + i].ptr != texture->srvHandle.cpuHandle.ptr) { D3D12_INTERNAL_TrackTexture(d3d12CommandBuffer, texture); - d3d12CommandBuffer->vertexStorageTextures[firstSlot + i] = texture; + d3d12CommandBuffer->vertexStorageTextureDescriptorHandles[firstSlot + i] = texture->srvHandle.cpuHandle; d3d12CommandBuffer->needVertexStorageTextureBind = true; } } @@ -4762,12 +4761,12 @@ static void D3D12_BindVertexStorageBuffers( for (Uint32 i = 0; i < numBindings; i += 1) { D3D12BufferContainer *container = (D3D12BufferContainer *)storageBuffers[i]; - if (d3d12CommandBuffer->vertexStorageBuffers[firstSlot + i] != container->activeBuffer) { + if (d3d12CommandBuffer->vertexStorageBufferDescriptorHandles[firstSlot + i].ptr != container->activeBuffer->srvDescriptor.cpuHandle.ptr) { D3D12_INTERNAL_TrackBuffer( d3d12CommandBuffer, container->activeBuffer); - d3d12CommandBuffer->vertexStorageBuffers[firstSlot + i] = container->activeBuffer; + d3d12CommandBuffer->vertexStorageBufferDescriptorHandles[firstSlot + i] = container->activeBuffer->srvDescriptor.cpuHandle; d3d12CommandBuffer->needVertexStorageBufferBind = true; } } @@ -4785,21 +4784,21 @@ static void D3D12_BindFragmentSamplers( D3D12TextureContainer *container = (D3D12TextureContainer *)textureSamplerBindings[i].texture; D3D12Sampler *sampler = (D3D12Sampler *)textureSamplerBindings[i].sampler; - if (d3d12CommandBuffer->fragmentSamplers[firstSlot + i] != sampler) { + if (d3d12CommandBuffer->fragmentSamplerDescriptorHandles[firstSlot + i].ptr != sampler->handle.cpuHandle.ptr) { D3D12_INTERNAL_TrackSampler( d3d12CommandBuffer, sampler); - d3d12CommandBuffer->fragmentSamplers[firstSlot + i] = sampler; + d3d12CommandBuffer->fragmentSamplerDescriptorHandles[firstSlot + i] = sampler->handle.cpuHandle; d3d12CommandBuffer->needFragmentSamplerBind = true; } - if (d3d12CommandBuffer->fragmentSamplerTextures[firstSlot + i] != container->activeTexture) { + if (d3d12CommandBuffer->fragmentSamplerTextureDescriptorHandles[firstSlot + i].ptr != container->activeTexture->srvHandle.cpuHandle.ptr) { D3D12_INTERNAL_TrackTexture( d3d12CommandBuffer, container->activeTexture); - d3d12CommandBuffer->fragmentSamplerTextures[firstSlot + i] = container->activeTexture; + d3d12CommandBuffer->fragmentSamplerTextureDescriptorHandles[firstSlot + i] = container->activeTexture->srvHandle.cpuHandle; d3d12CommandBuffer->needFragmentSamplerBind = true; } } @@ -4817,10 +4816,10 @@ static void D3D12_BindFragmentStorageTextures( D3D12TextureContainer *container = (D3D12TextureContainer *)storageTextures[i]; D3D12Texture *texture = container->activeTexture; - if (d3d12CommandBuffer->fragmentStorageTextures[firstSlot + i] != texture) { + if (d3d12CommandBuffer->fragmentStorageTextureDescriptorHandles[firstSlot + i].ptr != texture->srvHandle.cpuHandle.ptr) { D3D12_INTERNAL_TrackTexture(d3d12CommandBuffer, texture); - d3d12CommandBuffer->fragmentStorageTextures[firstSlot + i] = texture; + d3d12CommandBuffer->fragmentStorageTextureDescriptorHandles[firstSlot + i] = texture->srvHandle.cpuHandle; d3d12CommandBuffer->needFragmentStorageTextureBind = true; } } @@ -4837,12 +4836,12 @@ static void D3D12_BindFragmentStorageBuffers( for (Uint32 i = 0; i < numBindings; i += 1) { D3D12BufferContainer *container = (D3D12BufferContainer *)storageBuffers[i]; - if (d3d12CommandBuffer->fragmentStorageBuffers[firstSlot + i] != container->activeBuffer) { + if (d3d12CommandBuffer->fragmentStorageBufferDescriptorHandles[firstSlot + i].ptr != container->activeBuffer->srvDescriptor.cpuHandle.ptr) { D3D12_INTERNAL_TrackBuffer( d3d12CommandBuffer, container->activeBuffer); - d3d12CommandBuffer->fragmentStorageBuffers[firstSlot + i] = container->activeBuffer; + d3d12CommandBuffer->fragmentStorageBufferDescriptorHandles[firstSlot + i] = container->activeBuffer->srvDescriptor.cpuHandle; d3d12CommandBuffer->needFragmentStorageBufferBind = true; } } @@ -4923,15 +4922,19 @@ static void D3D12_INTERNAL_WriteGPUDescriptors( gpuBaseDescriptor->ptr = heap->descriptorHeapGPUStart.ptr + (heap->currentDescriptorIndex * heap->descriptorSize); for (Uint32 i = 0; i < resourceHandleCount; i += 1) { - ID3D12Device_CopyDescriptorsSimple( - commandBuffer->renderer->device, - 1, - gpuHeapCpuHandle, - resourceDescriptorHandles[i], - heapType); + // This will crash the driver if it gets a null handle! Cool! + if (resourceDescriptorHandles[i].ptr != 0) + { + ID3D12Device_CopyDescriptorsSimple( + commandBuffer->renderer->device, + 1, + gpuHeapCpuHandle, + resourceDescriptorHandles[i], + heapType); - heap->currentDescriptorIndex += 1; - gpuHeapCpuHandle.ptr += heap->descriptorSize; + heap->currentDescriptorIndex += 1; + gpuHeapCpuHandle.ptr += heap->descriptorSize; + } } } @@ -4966,16 +4969,16 @@ static void D3D12_INTERNAL_BindGraphicsResources( } if (commandBuffer->needVertexSamplerBind) { - if (graphicsPipeline->vertexSamplerCount > 0) { - for (Uint32 i = 0; i < graphicsPipeline->vertexSamplerCount; i += 1) { - cpuHandles[i] = commandBuffer->vertexSamplers[i]->handle.cpuHandle; + if (graphicsPipeline->header.num_vertex_samplers > 0) { + for (Uint32 i = 0; i < graphicsPipeline->header.num_vertex_samplers; i += 1) { + cpuHandles[i] = commandBuffer->vertexSamplerDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER, cpuHandles, - graphicsPipeline->vertexSamplerCount, + graphicsPipeline->header.num_vertex_samplers, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetGraphicsRootDescriptorTable( @@ -4983,15 +4986,15 @@ static void D3D12_INTERNAL_BindGraphicsResources( graphicsPipeline->rootSignature->vertexSamplerRootIndex, gpuDescriptorHandle); - for (Uint32 i = 0; i < graphicsPipeline->vertexSamplerCount; i += 1) { - cpuHandles[i] = commandBuffer->vertexSamplerTextures[i]->srvHandle.cpuHandle; + for (Uint32 i = 0; i < graphicsPipeline->header.num_vertex_samplers; i += 1) { + cpuHandles[i] = commandBuffer->vertexSamplerTextureDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, cpuHandles, - graphicsPipeline->vertexSamplerCount, + graphicsPipeline->header.num_vertex_samplers, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetGraphicsRootDescriptorTable( @@ -5003,16 +5006,16 @@ static void D3D12_INTERNAL_BindGraphicsResources( } if (commandBuffer->needVertexStorageTextureBind) { - if (graphicsPipeline->vertexStorageTextureCount > 0) { - for (Uint32 i = 0; i < graphicsPipeline->vertexStorageTextureCount; i += 1) { - cpuHandles[i] = commandBuffer->vertexStorageTextures[i]->srvHandle.cpuHandle; + if (graphicsPipeline->header.num_vertex_storage_textures > 0) { + for (Uint32 i = 0; i < graphicsPipeline->header.num_vertex_storage_textures; i += 1) { + cpuHandles[i] = commandBuffer->vertexStorageTextureDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, cpuHandles, - graphicsPipeline->vertexStorageTextureCount, + graphicsPipeline->header.num_vertex_storage_textures, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetGraphicsRootDescriptorTable( @@ -5024,16 +5027,16 @@ static void D3D12_INTERNAL_BindGraphicsResources( } if (commandBuffer->needVertexStorageBufferBind) { - if (graphicsPipeline->vertexStorageBufferCount > 0) { - for (Uint32 i = 0; i < graphicsPipeline->vertexStorageBufferCount; i += 1) { - cpuHandles[i] = commandBuffer->vertexStorageBuffers[i]->srvDescriptor.cpuHandle; + if (graphicsPipeline->header.num_vertex_storage_buffers > 0) { + for (Uint32 i = 0; i < graphicsPipeline->header.num_vertex_storage_buffers; i += 1) { + cpuHandles[i] = commandBuffer->vertexStorageBufferDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, cpuHandles, - graphicsPipeline->vertexStorageBufferCount, + graphicsPipeline->header.num_vertex_storage_buffers, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetGraphicsRootDescriptorTable( @@ -5046,7 +5049,7 @@ static void D3D12_INTERNAL_BindGraphicsResources( for (Uint32 i = 0; i < MAX_UNIFORM_BUFFERS_PER_STAGE; i += 1) { if (commandBuffer->needVertexUniformBufferBind[i]) { - if (graphicsPipeline->vertexUniformBufferCount > i) { + if (graphicsPipeline->header.num_vertex_uniform_buffers > i) { ID3D12GraphicsCommandList_SetGraphicsRootConstantBufferView( commandBuffer->graphicsCommandList, graphicsPipeline->rootSignature->vertexUniformBufferRootIndex[i], @@ -5057,16 +5060,16 @@ static void D3D12_INTERNAL_BindGraphicsResources( } if (commandBuffer->needFragmentSamplerBind) { - if (graphicsPipeline->fragmentSamplerCount > 0) { - for (Uint32 i = 0; i < graphicsPipeline->fragmentSamplerCount; i += 1) { - cpuHandles[i] = commandBuffer->fragmentSamplers[i]->handle.cpuHandle; + if (graphicsPipeline->header.num_fragment_samplers > 0) { + for (Uint32 i = 0; i < graphicsPipeline->header.num_fragment_samplers; i += 1) { + cpuHandles[i] = commandBuffer->fragmentSamplerDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER, cpuHandles, - graphicsPipeline->fragmentSamplerCount, + graphicsPipeline->header.num_fragment_samplers, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetGraphicsRootDescriptorTable( @@ -5074,15 +5077,15 @@ static void D3D12_INTERNAL_BindGraphicsResources( graphicsPipeline->rootSignature->fragmentSamplerRootIndex, gpuDescriptorHandle); - for (Uint32 i = 0; i < graphicsPipeline->fragmentSamplerCount; i += 1) { - cpuHandles[i] = commandBuffer->fragmentSamplerTextures[i]->srvHandle.cpuHandle; + for (Uint32 i = 0; i < graphicsPipeline->header.num_fragment_samplers; i += 1) { + cpuHandles[i] = commandBuffer->fragmentSamplerTextureDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, cpuHandles, - graphicsPipeline->fragmentSamplerCount, + graphicsPipeline->header.num_fragment_samplers, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetGraphicsRootDescriptorTable( @@ -5094,16 +5097,16 @@ static void D3D12_INTERNAL_BindGraphicsResources( } if (commandBuffer->needFragmentStorageTextureBind) { - if (graphicsPipeline->fragmentStorageTextureCount > 0) { - for (Uint32 i = 0; i < graphicsPipeline->fragmentStorageTextureCount; i += 1) { - cpuHandles[i] = commandBuffer->fragmentStorageTextures[i]->srvHandle.cpuHandle; + if (graphicsPipeline->header.num_fragment_storage_textures > 0) { + for (Uint32 i = 0; i < graphicsPipeline->header.num_fragment_storage_textures; i += 1) { + cpuHandles[i] = commandBuffer->fragmentStorageTextureDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, cpuHandles, - graphicsPipeline->fragmentStorageTextureCount, + graphicsPipeline->header.num_fragment_storage_textures, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetGraphicsRootDescriptorTable( @@ -5115,16 +5118,16 @@ static void D3D12_INTERNAL_BindGraphicsResources( } if (commandBuffer->needFragmentStorageBufferBind) { - if (graphicsPipeline->fragmentStorageBufferCount > 0) { - for (Uint32 i = 0; i < graphicsPipeline->fragmentStorageBufferCount; i += 1) { - cpuHandles[i] = commandBuffer->fragmentStorageBuffers[i]->srvDescriptor.cpuHandle; + if (graphicsPipeline->header.num_fragment_storage_buffers > 0) { + for (Uint32 i = 0; i < graphicsPipeline->header.num_fragment_storage_buffers; i += 1) { + cpuHandles[i] = commandBuffer->fragmentStorageBufferDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, cpuHandles, - graphicsPipeline->fragmentStorageBufferCount, + graphicsPipeline->header.num_fragment_storage_buffers, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetGraphicsRootDescriptorTable( @@ -5137,7 +5140,7 @@ static void D3D12_INTERNAL_BindGraphicsResources( for (Uint32 i = 0; i < MAX_UNIFORM_BUFFERS_PER_STAGE; i += 1) { if (commandBuffer->needFragmentUniformBufferBind[i]) { - if (graphicsPipeline->fragmentUniformBufferCount > i) { + if (graphicsPipeline->header.num_fragment_uniform_buffers > i) { ID3D12GraphicsCommandList_SetGraphicsRootConstantBufferView( commandBuffer->graphicsCommandList, graphicsPipeline->rootSignature->fragmentUniformBufferRootIndex[i], @@ -5302,15 +5305,15 @@ static void D3D12_EndRenderPass( SDL_zeroa(d3d12CommandBuffer->vertexBufferOffsets); d3d12CommandBuffer->vertexBufferCount = 0; - SDL_zeroa(d3d12CommandBuffer->vertexSamplerTextures); - SDL_zeroa(d3d12CommandBuffer->vertexSamplers); - SDL_zeroa(d3d12CommandBuffer->vertexStorageTextures); - SDL_zeroa(d3d12CommandBuffer->vertexStorageBuffers); + SDL_zeroa(d3d12CommandBuffer->vertexSamplerTextureDescriptorHandles); + SDL_zeroa(d3d12CommandBuffer->vertexSamplerDescriptorHandles); + SDL_zeroa(d3d12CommandBuffer->vertexStorageTextureDescriptorHandles); + SDL_zeroa(d3d12CommandBuffer->vertexStorageBufferDescriptorHandles); - SDL_zeroa(d3d12CommandBuffer->fragmentSamplerTextures); - SDL_zeroa(d3d12CommandBuffer->fragmentSamplers); - SDL_zeroa(d3d12CommandBuffer->fragmentStorageTextures); - SDL_zeroa(d3d12CommandBuffer->fragmentStorageBuffers); + SDL_zeroa(d3d12CommandBuffer->fragmentSamplerTextureDescriptorHandles); + SDL_zeroa(d3d12CommandBuffer->fragmentSamplerDescriptorHandles); + SDL_zeroa(d3d12CommandBuffer->fragmentStorageTextureDescriptorHandles); + SDL_zeroa(d3d12CommandBuffer->fragmentStorageBufferDescriptorHandles); } // Compute Pass @@ -5344,6 +5347,7 @@ static void D3D12_BeginComputePass( D3D12_RESOURCE_STATE_UNORDERED_ACCESS); d3d12CommandBuffer->computeReadWriteStorageTextureSubresources[i] = subresource; + d3d12CommandBuffer->computeReadWriteStorageTextureDescriptorHandles[i] = subresource->uavHandle.cpuHandle; D3D12_INTERNAL_TrackTexture( d3d12CommandBuffer, @@ -5362,6 +5366,7 @@ static void D3D12_BeginComputePass( D3D12_RESOURCE_STATE_UNORDERED_ACCESS); d3d12CommandBuffer->computeReadWriteStorageBuffers[i] = buffer; + d3d12CommandBuffer->computeReadWriteStorageBufferDescriptorHandles[i] = buffer->uavDescriptor.cpuHandle; D3D12_INTERNAL_TrackBuffer( d3d12CommandBuffer, @@ -5403,7 +5408,7 @@ static void D3D12_BindComputePipeline( d3d12CommandBuffer->needComputeUniformBufferBind[i] = true; } - for (Uint32 i = 0; i < pipeline->numUniformBuffers; i += 1) { + for (Uint32 i = 0; i < pipeline->header.numUniformBuffers; i += 1) { if (d3d12CommandBuffer->computeUniformBuffers[i] == NULL) { d3d12CommandBuffer->computeUniformBuffers[i] = D3D12_INTERNAL_AcquireUniformBufferFromPool( d3d12CommandBuffer); @@ -5413,9 +5418,9 @@ static void D3D12_BindComputePipeline( D3D12_INTERNAL_TrackComputePipeline(d3d12CommandBuffer, pipeline); // Bind write-only resources after setting root signature - if (pipeline->numReadWriteStorageTextures > 0) { - for (Uint32 i = 0; i < pipeline->numReadWriteStorageTextures; i += 1) { - cpuHandles[i] = d3d12CommandBuffer->computeReadWriteStorageTextureSubresources[i]->uavHandle.cpuHandle; + if (pipeline->header.numReadWriteStorageTextures > 0) { + for (Uint32 i = 0; i < pipeline->header.numReadWriteStorageTextures; i += 1) { + cpuHandles[i] = d3d12CommandBuffer->computeReadWriteStorageTextureDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( @@ -5431,9 +5436,9 @@ static void D3D12_BindComputePipeline( gpuDescriptorHandle); } - if (pipeline->numReadWriteStorageBuffers > 0) { - for (Uint32 i = 0; i < pipeline->numReadWriteStorageBuffers; i += 1) { - cpuHandles[i] = d3d12CommandBuffer->computeReadWriteStorageBuffers[i]->uavDescriptor.cpuHandle; + if (pipeline->header.numReadWriteStorageBuffers > 0) { + for (Uint32 i = 0; i < pipeline->header.numReadWriteStorageBuffers; i += 1) { + cpuHandles[i] = d3d12CommandBuffer->computeReadWriteStorageBufferDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( @@ -5462,21 +5467,21 @@ static void D3D12_BindComputeSamplers( D3D12TextureContainer *container = (D3D12TextureContainer *)textureSamplerBindings[i].texture; D3D12Sampler *sampler = (D3D12Sampler *)textureSamplerBindings[i].sampler; - if (d3d12CommandBuffer->computeSamplers[firstSlot + i] != sampler) { + if (d3d12CommandBuffer->computeSamplerDescriptorHandles[firstSlot + i].ptr != sampler->handle.cpuHandle.ptr) { D3D12_INTERNAL_TrackSampler( d3d12CommandBuffer, (D3D12Sampler *)textureSamplerBindings[i].sampler); - d3d12CommandBuffer->computeSamplers[firstSlot + i] = (D3D12Sampler *)textureSamplerBindings[i].sampler; + d3d12CommandBuffer->computeSamplerDescriptorHandles[firstSlot + i] = sampler->handle.cpuHandle; d3d12CommandBuffer->needComputeSamplerBind = true; } - if (d3d12CommandBuffer->computeSamplerTextures[firstSlot + i] != container->activeTexture) { + if (d3d12CommandBuffer->computeSamplerTextureDescriptorHandles[firstSlot + i].ptr != container->activeTexture->srvHandle.cpuHandle.ptr) { D3D12_INTERNAL_TrackTexture( d3d12CommandBuffer, container->activeTexture); - d3d12CommandBuffer->computeSamplerTextures[firstSlot + i] = container->activeTexture; + d3d12CommandBuffer->computeSamplerTextureDescriptorHandles[firstSlot + i] = container->activeTexture->srvHandle.cpuHandle; d3d12CommandBuffer->needComputeSamplerBind = true; } } @@ -5513,6 +5518,7 @@ static void D3D12_BindComputeStorageTextures( container->activeTexture); d3d12CommandBuffer->computeReadOnlyStorageTextures[firstSlot + i] = container->activeTexture; + d3d12CommandBuffer->computeReadOnlyStorageTextureDescriptorHandles[firstSlot + i] = container->activeTexture->srvHandle.cpuHandle; d3d12CommandBuffer->needComputeReadOnlyStorageTextureBind = true; } } @@ -5550,6 +5556,7 @@ static void D3D12_BindComputeStorageBuffers( buffer); d3d12CommandBuffer->computeReadOnlyStorageBuffers[firstSlot + i] = buffer; + d3d12CommandBuffer->computeReadOnlyStorageBufferDescriptorHandles[firstSlot + i] = buffer->srvDescriptor.cpuHandle; d3d12CommandBuffer->needComputeReadOnlyStorageBufferBind = true; } } @@ -5585,16 +5592,16 @@ static void D3D12_INTERNAL_BindComputeResources( D3D12_GPU_DESCRIPTOR_HANDLE gpuDescriptorHandle; if (commandBuffer->needComputeSamplerBind) { - if (computePipeline->numSamplers > 0) { - for (Uint32 i = 0; i < computePipeline->numSamplers; i += 1) { - cpuHandles[i] = commandBuffer->computeSamplers[i]->handle.cpuHandle; + if (computePipeline->header.numSamplers > 0) { + for (Uint32 i = 0; i < computePipeline->header.numSamplers; i += 1) { + cpuHandles[i] = commandBuffer->computeSamplerDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER, cpuHandles, - computePipeline->numSamplers, + computePipeline->header.numSamplers, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetComputeRootDescriptorTable( @@ -5602,15 +5609,15 @@ static void D3D12_INTERNAL_BindComputeResources( computePipeline->rootSignature->samplerRootIndex, gpuDescriptorHandle); - for (Uint32 i = 0; i < computePipeline->numSamplers; i += 1) { - cpuHandles[i] = commandBuffer->computeSamplerTextures[i]->srvHandle.cpuHandle; + for (Uint32 i = 0; i < computePipeline->header.numSamplers; i += 1) { + cpuHandles[i] = commandBuffer->computeSamplerTextureDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, cpuHandles, - computePipeline->numSamplers, + computePipeline->header.numSamplers, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetComputeRootDescriptorTable( @@ -5622,16 +5629,16 @@ static void D3D12_INTERNAL_BindComputeResources( } if (commandBuffer->needComputeReadOnlyStorageTextureBind) { - if (computePipeline->numReadOnlyStorageTextures > 0) { - for (Uint32 i = 0; i < computePipeline->numReadOnlyStorageTextures; i += 1) { - cpuHandles[i] = commandBuffer->computeReadOnlyStorageTextures[i]->srvHandle.cpuHandle; + if (computePipeline->header.numReadonlyStorageTextures > 0) { + for (Uint32 i = 0; i < computePipeline->header.numReadonlyStorageTextures; i += 1) { + cpuHandles[i] = commandBuffer->computeReadOnlyStorageTextureDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, cpuHandles, - computePipeline->numReadOnlyStorageTextures, + computePipeline->header.numReadonlyStorageTextures, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetComputeRootDescriptorTable( @@ -5643,16 +5650,16 @@ static void D3D12_INTERNAL_BindComputeResources( } if (commandBuffer->needComputeReadOnlyStorageBufferBind) { - if (computePipeline->numReadOnlyStorageBuffers > 0) { - for (Uint32 i = 0; i < computePipeline->numReadOnlyStorageBuffers; i += 1) { - cpuHandles[i] = commandBuffer->computeReadOnlyStorageBuffers[i]->srvDescriptor.cpuHandle; + if (computePipeline->header.numReadonlyStorageBuffers > 0) { + for (Uint32 i = 0; i < computePipeline->header.numReadonlyStorageBuffers; i += 1) { + cpuHandles[i] = commandBuffer->computeReadOnlyStorageBufferDescriptorHandles[i]; } D3D12_INTERNAL_WriteGPUDescriptors( commandBuffer, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, cpuHandles, - computePipeline->numReadOnlyStorageBuffers, + computePipeline->header.numReadonlyStorageBuffers, &gpuDescriptorHandle); ID3D12GraphicsCommandList_SetComputeRootDescriptorTable( @@ -5665,7 +5672,7 @@ static void D3D12_INTERNAL_BindComputeResources( for (Uint32 i = 0; i < MAX_UNIFORM_BUFFERS_PER_STAGE; i += 1) { if (commandBuffer->needComputeUniformBufferBind[i]) { - if (computePipeline->numUniformBuffers > i) { + if (computePipeline->header.numUniformBuffers > i) { ID3D12GraphicsCommandList_SetComputeRootConstantBufferView( commandBuffer->graphicsCommandList, computePipeline->rootSignature->uniformBufferRootIndex[i], @@ -5764,8 +5771,11 @@ static void D3D12_EndComputePass( } } - SDL_zeroa(d3d12CommandBuffer->computeSamplerTextures); - SDL_zeroa(d3d12CommandBuffer->computeSamplers); + SDL_zeroa(d3d12CommandBuffer->computeSamplerTextureDescriptorHandles); + SDL_zeroa(d3d12CommandBuffer->computeSamplerDescriptorHandles); + + SDL_zeroa(d3d12CommandBuffer->computeReadWriteStorageTextureDescriptorHandles); + SDL_zeroa(d3d12CommandBuffer->computeReadWriteStorageBufferDescriptorHandles); d3d12CommandBuffer->currentComputePipeline = NULL; } @@ -7354,20 +7364,22 @@ static SDL_GPUCommandBuffer *D3D12_AcquireCommandBuffer( SDL_zeroa(commandBuffer->vertexBufferOffsets); commandBuffer->vertexBufferCount = 0; - SDL_zeroa(commandBuffer->vertexSamplerTextures); - SDL_zeroa(commandBuffer->vertexSamplers); - SDL_zeroa(commandBuffer->vertexStorageTextures); - SDL_zeroa(commandBuffer->vertexStorageBuffers); + SDL_zeroa(commandBuffer->vertexSamplerTextureDescriptorHandles); + SDL_zeroa(commandBuffer->vertexSamplerDescriptorHandles); + SDL_zeroa(commandBuffer->vertexStorageTextureDescriptorHandles); + SDL_zeroa(commandBuffer->vertexStorageBufferDescriptorHandles); SDL_zeroa(commandBuffer->vertexUniformBuffers); - SDL_zeroa(commandBuffer->fragmentSamplerTextures); - SDL_zeroa(commandBuffer->fragmentSamplers); - SDL_zeroa(commandBuffer->fragmentStorageTextures); - SDL_zeroa(commandBuffer->fragmentStorageBuffers); + SDL_zeroa(commandBuffer->fragmentSamplerTextureDescriptorHandles); + SDL_zeroa(commandBuffer->fragmentSamplerDescriptorHandles); + SDL_zeroa(commandBuffer->fragmentStorageTextureDescriptorHandles); + SDL_zeroa(commandBuffer->fragmentStorageBufferDescriptorHandles); SDL_zeroa(commandBuffer->fragmentUniformBuffers); - SDL_zeroa(commandBuffer->computeSamplerTextures); - SDL_zeroa(commandBuffer->computeSamplers); + SDL_zeroa(commandBuffer->computeSamplerTextureDescriptorHandles); + SDL_zeroa(commandBuffer->computeSamplerDescriptorHandles); + SDL_zeroa(commandBuffer->computeReadOnlyStorageTextureDescriptorHandles); + SDL_zeroa(commandBuffer->computeReadOnlyStorageBufferDescriptorHandles); SDL_zeroa(commandBuffer->computeReadOnlyStorageTextures); SDL_zeroa(commandBuffer->computeReadOnlyStorageBuffers); SDL_zeroa(commandBuffer->computeReadWriteStorageTextureSubresources); diff --git a/src/gpu/metal/SDL_gpu_metal.m b/src/gpu/metal/SDL_gpu_metal.m index 85e640b60c..7c2aefebb2 100644 --- a/src/gpu/metal/SDL_gpu_metal.m +++ b/src/gpu/metal/SDL_gpu_metal.m @@ -476,33 +476,21 @@ typedef struct MetalShader typedef struct MetalGraphicsPipeline { + GraphicsPipelineCommonHeader header; + id handle; SDL_GPURasterizerState rasterizerState; SDL_GPUPrimitiveType primitiveType; id depth_stencil_state; - - Uint32 vertexSamplerCount; - Uint32 vertexUniformBufferCount; - Uint32 vertexStorageBufferCount; - Uint32 vertexStorageTextureCount; - - Uint32 fragmentSamplerCount; - Uint32 fragmentUniformBufferCount; - Uint32 fragmentStorageBufferCount; - Uint32 fragmentStorageTextureCount; } MetalGraphicsPipeline; typedef struct MetalComputePipeline { + ComputePipelineCommonHeader header; + id handle; - Uint32 numSamplers; - Uint32 numReadonlyStorageTextures; - Uint32 numReadWriteStorageTextures; - Uint32 numReadonlyStorageBuffers; - Uint32 numReadWriteStorageBuffers; - Uint32 numUniformBuffers; Uint32 threadcountX; Uint32 threadcountY; Uint32 threadcountZ; @@ -1059,12 +1047,12 @@ static SDL_GPUComputePipeline *METAL_CreateComputePipeline( pipeline = SDL_calloc(1, sizeof(MetalComputePipeline)); pipeline->handle = handle; - pipeline->numSamplers = createinfo->num_samplers; - pipeline->numReadonlyStorageTextures = createinfo->num_readonly_storage_textures; - pipeline->numReadWriteStorageTextures = createinfo->num_readwrite_storage_textures; - pipeline->numReadonlyStorageBuffers = createinfo->num_readonly_storage_buffers; - pipeline->numReadWriteStorageBuffers = createinfo->num_readwrite_storage_buffers; - pipeline->numUniformBuffers = createinfo->num_uniform_buffers; + pipeline->header.numSamplers = createinfo->num_samplers; + pipeline->header.numReadonlyStorageTextures = createinfo->num_readonly_storage_textures; + pipeline->header.numReadWriteStorageTextures = createinfo->num_readwrite_storage_textures; + pipeline->header.numReadonlyStorageBuffers = createinfo->num_readonly_storage_buffers; + pipeline->header.numReadWriteStorageBuffers = createinfo->num_readwrite_storage_buffers; + pipeline->header.numUniformBuffers = createinfo->num_uniform_buffers; pipeline->threadcountX = createinfo->threadcount_x; pipeline->threadcountY = createinfo->threadcount_y; pipeline->threadcountZ = createinfo->threadcount_z; @@ -1207,14 +1195,14 @@ static SDL_GPUGraphicsPipeline *METAL_CreateGraphicsPipeline( result->depth_stencil_state = depthStencilState; result->rasterizerState = createinfo->rasterizer_state; result->primitiveType = createinfo->primitive_type; - result->vertexSamplerCount = vertexShader->numSamplers; - result->vertexUniformBufferCount = vertexShader->numUniformBuffers; - result->vertexStorageBufferCount = vertexShader->numStorageBuffers; - result->vertexStorageTextureCount = vertexShader->numStorageTextures; - result->fragmentSamplerCount = fragmentShader->numSamplers; - result->fragmentUniformBufferCount = fragmentShader->numUniformBuffers; - result->fragmentStorageBufferCount = fragmentShader->numStorageBuffers; - result->fragmentStorageTextureCount = fragmentShader->numStorageTextures; + result->header.num_vertex_samplers = vertexShader->numSamplers; + result->header.num_vertex_uniform_buffers = vertexShader->numUniformBuffers; + result->header.num_vertex_storage_buffers = vertexShader->numStorageBuffers; + result->header.num_vertex_storage_textures = vertexShader->numStorageTextures; + result->header.num_fragment_samplers = fragmentShader->numSamplers; + result->header.num_fragment_uniform_buffers = fragmentShader->numUniformBuffers; + result->header.num_fragment_storage_buffers = fragmentShader->numStorageBuffers; + result->header.num_fragment_storage_textures = fragmentShader->numStorageTextures; return (SDL_GPUGraphicsPipeline *)result; } } @@ -2412,14 +2400,14 @@ static void METAL_BindGraphicsPipeline( metalCommandBuffer->needFragmentUniformBufferBind[i] = true; } - for (i = 0; i < pipeline->vertexUniformBufferCount; i += 1) { + for (i = 0; i < pipeline->header.num_vertex_uniform_buffers; i += 1) { if (metalCommandBuffer->vertexUniformBuffers[i] == NULL) { metalCommandBuffer->vertexUniformBuffers[i] = METAL_INTERNAL_AcquireUniformBufferFromPool( metalCommandBuffer); } } - for (i = 0; i < pipeline->fragmentUniformBufferCount; i += 1) { + for (i = 0; i < pipeline->header.num_fragment_uniform_buffers; i += 1) { if (metalCommandBuffer->fragmentUniformBuffers[i] == NULL) { metalCommandBuffer->fragmentUniformBuffers[i] = METAL_INTERNAL_AcquireUniformBufferFromPool( metalCommandBuffer); @@ -2650,11 +2638,11 @@ static void METAL_INTERNAL_BindGraphicsResources( // Vertex Samplers+Textures if (commandBuffer->needVertexSamplerBind) { - if (graphicsPipeline->vertexSamplerCount > 0) { + if (graphicsPipeline->header.num_vertex_samplers > 0) { [commandBuffer->renderEncoder setVertexSamplerStates:commandBuffer->vertexSamplers - withRange:NSMakeRange(0, graphicsPipeline->vertexSamplerCount)]; + withRange:NSMakeRange(0, graphicsPipeline->header.num_vertex_samplers)]; [commandBuffer->renderEncoder setVertexTextures:commandBuffer->vertexTextures - withRange:NSMakeRange(0, graphicsPipeline->vertexSamplerCount)]; + withRange:NSMakeRange(0, graphicsPipeline->header.num_vertex_samplers)]; } commandBuffer->needVertexSamplerBind = false; } @@ -2662,10 +2650,10 @@ static void METAL_INTERNAL_BindGraphicsResources( // Vertex Storage Textures if (commandBuffer->needVertexStorageTextureBind) { - if (graphicsPipeline->vertexStorageTextureCount > 0) { + if (graphicsPipeline->header.num_vertex_storage_textures > 0) { [commandBuffer->renderEncoder setVertexTextures:commandBuffer->vertexStorageTextures - withRange:NSMakeRange(graphicsPipeline->vertexSamplerCount, - graphicsPipeline->vertexStorageTextureCount)]; + withRange:NSMakeRange(graphicsPipeline->header.num_vertex_samplers, + graphicsPipeline->header.num_vertex_storage_textures)]; } commandBuffer->needVertexStorageTextureBind = false; } @@ -2673,20 +2661,20 @@ static void METAL_INTERNAL_BindGraphicsResources( // Vertex Storage Buffers if (commandBuffer->needVertexStorageBufferBind) { - if (graphicsPipeline->vertexStorageBufferCount > 0) { + if (graphicsPipeline->header.num_vertex_storage_buffers > 0) { [commandBuffer->renderEncoder setVertexBuffers:commandBuffer->vertexStorageBuffers offsets:offsets - withRange:NSMakeRange(graphicsPipeline->vertexUniformBufferCount, - graphicsPipeline->vertexStorageBufferCount)]; + withRange:NSMakeRange(graphicsPipeline->header.num_vertex_uniform_buffers, + graphicsPipeline->header.num_vertex_storage_buffers)]; } commandBuffer->needVertexStorageBufferBind = false; } // Vertex Uniform Buffers - for (Uint32 i = 0; i < graphicsPipeline->vertexUniformBufferCount; i += 1) { + for (Uint32 i = 0; i < graphicsPipeline->header.num_vertex_uniform_buffers; i += 1) { if (commandBuffer->needVertexUniformBufferBind[i]) { - if (graphicsPipeline->vertexUniformBufferCount > i) { + if (graphicsPipeline->header.num_vertex_uniform_buffers > i) { [commandBuffer->renderEncoder setVertexBuffer:commandBuffer->vertexUniformBuffers[i]->handle offset:commandBuffer->vertexUniformBuffers[i]->drawOffset @@ -2699,11 +2687,11 @@ static void METAL_INTERNAL_BindGraphicsResources( // Fragment Samplers+Textures if (commandBuffer->needFragmentSamplerBind) { - if (graphicsPipeline->fragmentSamplerCount > 0) { + if (graphicsPipeline->header.num_fragment_samplers > 0) { [commandBuffer->renderEncoder setFragmentSamplerStates:commandBuffer->fragmentSamplers - withRange:NSMakeRange(0, graphicsPipeline->fragmentSamplerCount)]; + withRange:NSMakeRange(0, graphicsPipeline->header.num_fragment_samplers)]; [commandBuffer->renderEncoder setFragmentTextures:commandBuffer->fragmentTextures - withRange:NSMakeRange(0, graphicsPipeline->fragmentSamplerCount)]; + withRange:NSMakeRange(0, graphicsPipeline->header.num_fragment_samplers)]; } commandBuffer->needFragmentSamplerBind = false; } @@ -2711,10 +2699,10 @@ static void METAL_INTERNAL_BindGraphicsResources( // Fragment Storage Textures if (commandBuffer->needFragmentStorageTextureBind) { - if (graphicsPipeline->fragmentStorageTextureCount > 0) { + if (graphicsPipeline->header.num_fragment_storage_textures > 0) { [commandBuffer->renderEncoder setFragmentTextures:commandBuffer->fragmentStorageTextures - withRange:NSMakeRange(graphicsPipeline->fragmentSamplerCount, - graphicsPipeline->fragmentStorageTextureCount)]; + withRange:NSMakeRange(graphicsPipeline->header.num_fragment_samplers, + graphicsPipeline->header.num_fragment_storage_textures)]; } commandBuffer->needFragmentStorageTextureBind = false; } @@ -2722,20 +2710,20 @@ static void METAL_INTERNAL_BindGraphicsResources( // Fragment Storage Buffers if (commandBuffer->needFragmentStorageBufferBind) { - if (graphicsPipeline->fragmentStorageBufferCount > 0) { + if (graphicsPipeline->header.num_fragment_storage_buffers > 0) { [commandBuffer->renderEncoder setFragmentBuffers:commandBuffer->fragmentStorageBuffers offsets:offsets - withRange:NSMakeRange(graphicsPipeline->fragmentUniformBufferCount, - graphicsPipeline->fragmentStorageBufferCount)]; + withRange:NSMakeRange(graphicsPipeline->header.num_fragment_uniform_buffers, + graphicsPipeline->header.num_fragment_storage_buffers)]; } commandBuffer->needFragmentStorageBufferBind = false; } // Fragment Uniform Buffers - for (Uint32 i = 0; i < graphicsPipeline->fragmentUniformBufferCount; i += 1) { + for (Uint32 i = 0; i < graphicsPipeline->header.num_fragment_uniform_buffers; i += 1) { if (commandBuffer->needFragmentUniformBufferBind[i]) { - if (graphicsPipeline->fragmentUniformBufferCount > i) { + if (graphicsPipeline->header.num_fragment_uniform_buffers > i) { [commandBuffer->renderEncoder setFragmentBuffer:commandBuffer->fragmentUniformBuffers[i]->handle offset:commandBuffer->fragmentUniformBuffers[i]->drawOffset @@ -2754,38 +2742,38 @@ static void METAL_INTERNAL_BindComputeResources( NSUInteger offsets[MAX_STORAGE_BUFFERS_PER_STAGE] = { 0 }; if (commandBuffer->needComputeSamplerBind) { - if (computePipeline->numSamplers > 0) { + if (computePipeline->header.numSamplers > 0) { [commandBuffer->computeEncoder setTextures:commandBuffer->computeSamplerTextures - withRange:NSMakeRange(0, computePipeline->numSamplers)]; + withRange:NSMakeRange(0, computePipeline->header.numSamplers)]; [commandBuffer->computeEncoder setSamplerStates:commandBuffer->computeSamplers - withRange:NSMakeRange(0, computePipeline->numSamplers)]; + withRange:NSMakeRange(0, computePipeline->header.numSamplers)]; } commandBuffer->needComputeSamplerBind = false; } if (commandBuffer->needComputeReadOnlyStorageTextureBind) { - if (computePipeline->numReadonlyStorageTextures > 0) { + if (computePipeline->header.numReadonlyStorageTextures > 0) { [commandBuffer->computeEncoder setTextures:commandBuffer->computeReadOnlyTextures withRange:NSMakeRange( - computePipeline->numSamplers, - computePipeline->numReadonlyStorageTextures)]; + computePipeline->header.numSamplers, + computePipeline->header.numReadonlyStorageTextures)]; } commandBuffer->needComputeReadOnlyStorageTextureBind = false; } if (commandBuffer->needComputeReadOnlyStorageBufferBind) { - if (computePipeline->numReadonlyStorageBuffers > 0) { + if (computePipeline->header.numReadonlyStorageBuffers > 0) { [commandBuffer->computeEncoder setBuffers:commandBuffer->computeReadOnlyBuffers offsets:offsets - withRange:NSMakeRange(computePipeline->numUniformBuffers, - computePipeline->numReadonlyStorageBuffers)]; + withRange:NSMakeRange(computePipeline->header.numUniformBuffers, + computePipeline->header.numReadonlyStorageBuffers)]; } commandBuffer->needComputeReadOnlyStorageBufferBind = false; } for (Uint32 i = 0; i < MAX_UNIFORM_BUFFERS_PER_STAGE; i += 1) { if (commandBuffer->needComputeUniformBufferBind[i]) { - if (computePipeline->numUniformBuffers > i) { + if (computePipeline->header.numUniformBuffers > i) { [commandBuffer->computeEncoder setBuffer:commandBuffer->computeUniformBuffers[i]->handle offset:commandBuffer->computeUniformBuffers[i]->drawOffset @@ -3133,7 +3121,7 @@ static void METAL_BindComputePipeline( metalCommandBuffer->needComputeUniformBufferBind[i] = true; } - for (Uint32 i = 0; i < pipeline->numUniformBuffers; i += 1) { + for (Uint32 i = 0; i < pipeline->header.numUniformBuffers; i += 1) { if (metalCommandBuffer->computeUniformBuffers[i] == NULL) { metalCommandBuffer->computeUniformBuffers[i] = METAL_INTERNAL_AcquireUniformBufferFromPool( metalCommandBuffer); @@ -3141,22 +3129,22 @@ static void METAL_BindComputePipeline( } // Bind write-only resources - if (pipeline->numReadWriteStorageTextures > 0) { + if (pipeline->header.numReadWriteStorageTextures > 0) { [metalCommandBuffer->computeEncoder setTextures:metalCommandBuffer->computeReadWriteTextures withRange:NSMakeRange( - pipeline->numSamplers + - pipeline->numReadonlyStorageTextures, - pipeline->numReadWriteStorageTextures)]; + pipeline->header.numSamplers + + pipeline->header.numReadonlyStorageTextures, + pipeline->header.numReadWriteStorageTextures)]; } NSUInteger offsets[MAX_COMPUTE_WRITE_BUFFERS] = { 0 }; - if (pipeline->numReadWriteStorageBuffers > 0) { + if (pipeline->header.numReadWriteStorageBuffers > 0) { [metalCommandBuffer->computeEncoder setBuffers:metalCommandBuffer->computeReadWriteBuffers offsets:offsets withRange:NSMakeRange( - pipeline->numUniformBuffers + - pipeline->numReadonlyStorageBuffers, - pipeline->numReadWriteStorageBuffers)]; + pipeline->header.numUniformBuffers + + pipeline->header.numReadonlyStorageBuffers, + pipeline->header.numReadWriteStorageBuffers)]; } } } diff --git a/src/gpu/vulkan/SDL_gpu_vulkan.c b/src/gpu/vulkan/SDL_gpu_vulkan.c index eb7f0b319d..4467ef801b 100644 --- a/src/gpu/vulkan/SDL_gpu_vulkan.c +++ b/src/gpu/vulkan/SDL_gpu_vulkan.c @@ -822,13 +822,13 @@ typedef struct DescriptorSetLayout typedef struct GraphicsPipelineResourceLayoutHashTableKey { Uint32 vertexSamplerCount; - Uint32 vertexStorageBufferCount; Uint32 vertexStorageTextureCount; + Uint32 vertexStorageBufferCount; Uint32 vertexUniformBufferCount; Uint32 fragmentSamplerCount; - Uint32 fragmentStorageBufferCount; Uint32 fragmentStorageTextureCount; + Uint32 fragmentStorageBufferCount; Uint32 fragmentUniformBufferCount; } GraphicsPipelineResourceLayoutHashTableKey; @@ -846,18 +846,20 @@ typedef struct VulkanGraphicsPipelineResourceLayout DescriptorSetLayout *descriptorSetLayouts[4]; Uint32 vertexSamplerCount; - Uint32 vertexStorageBufferCount; Uint32 vertexStorageTextureCount; + Uint32 vertexStorageBufferCount; Uint32 vertexUniformBufferCount; Uint32 fragmentSamplerCount; - Uint32 fragmentStorageBufferCount; Uint32 fragmentStorageTextureCount; + Uint32 fragmentStorageBufferCount; Uint32 fragmentUniformBufferCount; } VulkanGraphicsPipelineResourceLayout; typedef struct VulkanGraphicsPipeline { + GraphicsPipelineCommonHeader header; + VkPipeline pipeline; SDL_GPUPrimitiveType primitiveType; @@ -901,6 +903,8 @@ typedef struct VulkanComputePipelineResourceLayout typedef struct VulkanComputePipeline { + ComputePipelineCommonHeader header; + VkShaderModule shaderModule; VkPipeline pipeline; VulkanComputePipelineResourceLayout *resourceLayout; @@ -1038,25 +1042,33 @@ typedef struct VulkanCommandBuffer Uint32 vertexBufferCount; bool needVertexBufferBind; - VulkanTexture *vertexSamplerTextures[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - VulkanSampler *vertexSamplers[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - VulkanTexture *vertexStorageTextures[MAX_STORAGE_TEXTURES_PER_STAGE]; - VulkanBuffer *vertexStorageBuffers[MAX_STORAGE_BUFFERS_PER_STAGE]; + VkImageView vertexSamplerTextureViewBindings[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + VkSampler vertexSamplerBindings[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + VkImageView vertexStorageTextureViewBindings[MAX_STORAGE_TEXTURES_PER_STAGE]; + VkBuffer vertexStorageBufferBindings[MAX_STORAGE_BUFFERS_PER_STAGE]; - VulkanTexture *fragmentSamplerTextures[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - VulkanSampler *fragmentSamplers[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - VulkanTexture *fragmentStorageTextures[MAX_STORAGE_TEXTURES_PER_STAGE]; - VulkanBuffer *fragmentStorageBuffers[MAX_STORAGE_BUFFERS_PER_STAGE]; + VkImageView fragmentSamplerTextureViewBindings[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + VkSampler fragmentSamplerBindings[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + VkImageView fragmentStorageTextureViewBindings[MAX_STORAGE_TEXTURES_PER_STAGE]; + VkBuffer fragmentStorageBufferBindings[MAX_STORAGE_BUFFERS_PER_STAGE]; + VkImageView computeSamplerTextureViewBindings[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + VkSampler computeSamplerBindings[MAX_TEXTURE_SAMPLERS_PER_STAGE]; + VkImageView readOnlyComputeStorageTextureViewBindings[MAX_STORAGE_TEXTURES_PER_STAGE]; + VkBuffer readOnlyComputeStorageBufferBindings[MAX_STORAGE_BUFFERS_PER_STAGE]; + + // Track these separately because barriers can happen mid compute pass + VulkanTexture *readOnlyComputeStorageTextures[MAX_STORAGE_TEXTURES_PER_STAGE]; + VulkanBuffer *readOnlyComputeStorageBuffers[MAX_STORAGE_BUFFERS_PER_STAGE]; + + VkImageView readWriteComputeStorageTextureViewBindings[MAX_COMPUTE_WRITE_TEXTURES]; + VkBuffer readWriteComputeStorageBufferBindings[MAX_COMPUTE_WRITE_BUFFERS]; + + // Track these separately because they are barriered when the compute pass begins VulkanTextureSubresource *readWriteComputeStorageTextureSubresources[MAX_COMPUTE_WRITE_TEXTURES]; Uint32 readWriteComputeStorageTextureSubresourceCount; VulkanBuffer *readWriteComputeStorageBuffers[MAX_COMPUTE_WRITE_BUFFERS]; - VulkanTexture *computeSamplerTextures[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - VulkanSampler *computeSamplers[MAX_TEXTURE_SAMPLERS_PER_STAGE]; - VulkanTexture *readOnlyComputeStorageTextures[MAX_STORAGE_TEXTURES_PER_STAGE]; - VulkanBuffer *readOnlyComputeStorageBuffers[MAX_STORAGE_BUFFERS_PER_STAGE]; - // Uniform buffers VulkanUniformBuffer *vertexUniformBuffers[MAX_UNIFORM_BUFFERS_PER_STAGE]; @@ -5126,8 +5138,8 @@ static void VULKAN_INTERNAL_BindGraphicsDescriptorSets( currentWriteDescriptorSet->pTexelBufferView = NULL; currentWriteDescriptorSet->pBufferInfo = NULL; - imageInfos[imageInfoCount].sampler = commandBuffer->vertexSamplers[i]->sampler; - imageInfos[imageInfoCount].imageView = commandBuffer->vertexSamplerTextures[i]->fullView; + imageInfos[imageInfoCount].sampler = commandBuffer->vertexSamplerBindings[i]; + imageInfos[imageInfoCount].imageView = commandBuffer->vertexSamplerTextureViewBindings[i]; imageInfos[imageInfoCount].imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; currentWriteDescriptorSet->pImageInfo = &imageInfos[imageInfoCount]; @@ -5150,7 +5162,7 @@ static void VULKAN_INTERNAL_BindGraphicsDescriptorSets( currentWriteDescriptorSet->pBufferInfo = NULL; imageInfos[imageInfoCount].sampler = VK_NULL_HANDLE; - imageInfos[imageInfoCount].imageView = commandBuffer->vertexStorageTextures[i]->fullView; + imageInfos[imageInfoCount].imageView = commandBuffer->vertexStorageTextureViewBindings[i]; imageInfos[imageInfoCount].imageLayout = VK_IMAGE_LAYOUT_GENERAL; currentWriteDescriptorSet->pImageInfo = &imageInfos[imageInfoCount]; @@ -5172,7 +5184,7 @@ static void VULKAN_INTERNAL_BindGraphicsDescriptorSets( currentWriteDescriptorSet->pTexelBufferView = NULL; currentWriteDescriptorSet->pImageInfo = NULL; - bufferInfos[bufferInfoCount].buffer = commandBuffer->vertexStorageBuffers[i]->buffer; + bufferInfos[bufferInfoCount].buffer = commandBuffer->vertexStorageBufferBindings[i]; bufferInfos[bufferInfoCount].offset = 0; bufferInfos[bufferInfoCount].range = VK_WHOLE_SIZE; @@ -5245,8 +5257,8 @@ static void VULKAN_INTERNAL_BindGraphicsDescriptorSets( currentWriteDescriptorSet->pTexelBufferView = NULL; currentWriteDescriptorSet->pBufferInfo = NULL; - imageInfos[imageInfoCount].sampler = commandBuffer->fragmentSamplers[i]->sampler; - imageInfos[imageInfoCount].imageView = commandBuffer->fragmentSamplerTextures[i]->fullView; + imageInfos[imageInfoCount].sampler = commandBuffer->fragmentSamplerBindings[i]; + imageInfos[imageInfoCount].imageView = commandBuffer->fragmentSamplerTextureViewBindings[i]; imageInfos[imageInfoCount].imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; currentWriteDescriptorSet->pImageInfo = &imageInfos[imageInfoCount]; @@ -5269,7 +5281,7 @@ static void VULKAN_INTERNAL_BindGraphicsDescriptorSets( currentWriteDescriptorSet->pBufferInfo = NULL; imageInfos[imageInfoCount].sampler = VK_NULL_HANDLE; - imageInfos[imageInfoCount].imageView = commandBuffer->fragmentStorageTextures[i]->fullView; + imageInfos[imageInfoCount].imageView = commandBuffer->fragmentStorageTextureViewBindings[i]; imageInfos[imageInfoCount].imageLayout = VK_IMAGE_LAYOUT_GENERAL; currentWriteDescriptorSet->pImageInfo = &imageInfos[imageInfoCount]; @@ -5291,7 +5303,7 @@ static void VULKAN_INTERNAL_BindGraphicsDescriptorSets( currentWriteDescriptorSet->pTexelBufferView = NULL; currentWriteDescriptorSet->pImageInfo = NULL; - bufferInfos[bufferInfoCount].buffer = commandBuffer->fragmentStorageBuffers[i]->buffer; + bufferInfos[bufferInfoCount].buffer = commandBuffer->fragmentStorageBufferBindings[i]; bufferInfos[bufferInfoCount].offset = 0; bufferInfos[bufferInfoCount].range = VK_WHOLE_SIZE; @@ -6565,6 +6577,16 @@ static SDL_GPUGraphicsPipeline *VULKAN_CreateGraphicsPipeline( &nameInfo); } + // Put this data in the pipeline we can do validation in gpu.c + graphicsPipeline->header.num_vertex_samplers = graphicsPipeline->resourceLayout->vertexSamplerCount; + graphicsPipeline->header.num_vertex_storage_buffers = graphicsPipeline->resourceLayout->vertexStorageBufferCount; + graphicsPipeline->header.num_vertex_storage_textures = graphicsPipeline->resourceLayout->vertexStorageTextureCount; + graphicsPipeline->header.num_vertex_uniform_buffers = graphicsPipeline->resourceLayout->vertexUniformBufferCount; + graphicsPipeline->header.num_fragment_samplers = graphicsPipeline->resourceLayout->fragmentSamplerCount; + graphicsPipeline->header.num_fragment_storage_buffers = graphicsPipeline->resourceLayout->fragmentStorageBufferCount; + graphicsPipeline->header.num_fragment_storage_textures = graphicsPipeline->resourceLayout->fragmentStorageTextureCount; + graphicsPipeline->header.num_fragment_uniform_buffers = graphicsPipeline->resourceLayout->fragmentUniformBufferCount; + return (SDL_GPUGraphicsPipeline *)graphicsPipeline; } @@ -6659,6 +6681,14 @@ static SDL_GPUComputePipeline *VULKAN_CreateComputePipeline( &nameInfo); } + // Track these here for debug layer + vulkanComputePipeline->header.numSamplers = vulkanComputePipeline->resourceLayout->numSamplers; + vulkanComputePipeline->header.numReadonlyStorageTextures = vulkanComputePipeline->resourceLayout->numReadonlyStorageTextures; + vulkanComputePipeline->header.numReadonlyStorageBuffers = vulkanComputePipeline->resourceLayout->numReadonlyStorageBuffers; + vulkanComputePipeline->header.numReadWriteStorageTextures = vulkanComputePipeline->resourceLayout->numReadWriteStorageTextures; + vulkanComputePipeline->header.numReadWriteStorageBuffers = vulkanComputePipeline->resourceLayout->numReadWriteStorageBuffers; + vulkanComputePipeline->header.numUniformBuffers = vulkanComputePipeline->resourceLayout->numUniformBuffers; + return (SDL_GPUComputePipeline *)vulkanComputePipeline; } @@ -7457,21 +7487,21 @@ static void VULKAN_BindVertexSamplers( VulkanTextureContainer *textureContainer = (VulkanTextureContainer *)textureSamplerBindings[i].texture; VulkanSampler *sampler = (VulkanSampler *)textureSamplerBindings[i].sampler; - if (vulkanCommandBuffer->vertexSamplers[firstSlot + i] != sampler) { + if (vulkanCommandBuffer->vertexSamplerBindings[firstSlot + i] != sampler->sampler) { VULKAN_INTERNAL_TrackSampler( vulkanCommandBuffer, (VulkanSampler *)textureSamplerBindings[i].sampler); - vulkanCommandBuffer->vertexSamplers[firstSlot + i] = (VulkanSampler *)textureSamplerBindings[i].sampler; + vulkanCommandBuffer->vertexSamplerBindings[firstSlot + i] = sampler->sampler; vulkanCommandBuffer->needNewVertexResourceDescriptorSet = true; } - if (vulkanCommandBuffer->vertexSamplerTextures[firstSlot + i] != textureContainer->activeTexture) { + if (vulkanCommandBuffer->vertexSamplerTextureViewBindings[firstSlot + i] != textureContainer->activeTexture->fullView) { VULKAN_INTERNAL_TrackTexture( vulkanCommandBuffer, textureContainer->activeTexture); - vulkanCommandBuffer->vertexSamplerTextures[firstSlot + i] = textureContainer->activeTexture; + vulkanCommandBuffer->vertexSamplerTextureViewBindings[firstSlot + i] = textureContainer->activeTexture->fullView; vulkanCommandBuffer->needNewVertexResourceDescriptorSet = true; } } @@ -7488,12 +7518,12 @@ static void VULKAN_BindVertexStorageTextures( for (Uint32 i = 0; i < numBindings; i += 1) { VulkanTextureContainer *textureContainer = (VulkanTextureContainer *)storageTextures[i]; - if (vulkanCommandBuffer->vertexStorageTextures[firstSlot + i] != textureContainer->activeTexture) { + if (vulkanCommandBuffer->vertexStorageTextureViewBindings[firstSlot + i] != textureContainer->activeTexture->fullView) { VULKAN_INTERNAL_TrackTexture( vulkanCommandBuffer, textureContainer->activeTexture); - vulkanCommandBuffer->vertexStorageTextures[firstSlot + i] = textureContainer->activeTexture; + vulkanCommandBuffer->vertexStorageTextureViewBindings[firstSlot + i] = textureContainer->activeTexture->fullView; vulkanCommandBuffer->needNewVertexResourceDescriptorSet = true; } } @@ -7510,12 +7540,12 @@ static void VULKAN_BindVertexStorageBuffers( for (Uint32 i = 0; i < numBindings; i += 1) { VulkanBufferContainer *bufferContainer = (VulkanBufferContainer *)storageBuffers[i]; - if (vulkanCommandBuffer->vertexStorageBuffers[firstSlot + i] != bufferContainer->activeBuffer) { + if (vulkanCommandBuffer->vertexStorageBufferBindings[firstSlot + i] != bufferContainer->activeBuffer->buffer) { VULKAN_INTERNAL_TrackBuffer( vulkanCommandBuffer, bufferContainer->activeBuffer); - vulkanCommandBuffer->vertexStorageBuffers[firstSlot + i] = bufferContainer->activeBuffer; + vulkanCommandBuffer->vertexStorageBufferBindings[firstSlot + i] = bufferContainer->activeBuffer->buffer; vulkanCommandBuffer->needNewVertexResourceDescriptorSet = true; } } @@ -7533,21 +7563,21 @@ static void VULKAN_BindFragmentSamplers( VulkanTextureContainer *textureContainer = (VulkanTextureContainer *)textureSamplerBindings[i].texture; VulkanSampler *sampler = (VulkanSampler *)textureSamplerBindings[i].sampler; - if (vulkanCommandBuffer->fragmentSamplers[firstSlot + i] != sampler) { + if (vulkanCommandBuffer->fragmentSamplerBindings[firstSlot + i] != sampler->sampler) { VULKAN_INTERNAL_TrackSampler( vulkanCommandBuffer, (VulkanSampler *)textureSamplerBindings[i].sampler); - vulkanCommandBuffer->fragmentSamplers[firstSlot + i] = (VulkanSampler *)textureSamplerBindings[i].sampler; + vulkanCommandBuffer->fragmentSamplerBindings[firstSlot + i] = sampler->sampler; vulkanCommandBuffer->needNewFragmentResourceDescriptorSet = true; } - if (vulkanCommandBuffer->fragmentSamplerTextures[firstSlot + i] != textureContainer->activeTexture) { + if (vulkanCommandBuffer->fragmentSamplerTextureViewBindings[firstSlot + i] != textureContainer->activeTexture->fullView) { VULKAN_INTERNAL_TrackTexture( vulkanCommandBuffer, textureContainer->activeTexture); - vulkanCommandBuffer->fragmentSamplerTextures[firstSlot + i] = textureContainer->activeTexture; + vulkanCommandBuffer->fragmentSamplerTextureViewBindings[firstSlot + i] = textureContainer->activeTexture->fullView; vulkanCommandBuffer->needNewFragmentResourceDescriptorSet = true; } } @@ -7564,12 +7594,12 @@ static void VULKAN_BindFragmentStorageTextures( for (Uint32 i = 0; i < numBindings; i += 1) { VulkanTextureContainer *textureContainer = (VulkanTextureContainer *)storageTextures[i]; - if (vulkanCommandBuffer->fragmentStorageTextures[firstSlot + i] != textureContainer->activeTexture) { + if (vulkanCommandBuffer->fragmentStorageTextureViewBindings[firstSlot + i] != textureContainer->activeTexture->fullView) { VULKAN_INTERNAL_TrackTexture( vulkanCommandBuffer, textureContainer->activeTexture); - vulkanCommandBuffer->fragmentStorageTextures[firstSlot + i] = textureContainer->activeTexture; + vulkanCommandBuffer->fragmentStorageTextureViewBindings[firstSlot + i] = textureContainer->activeTexture->fullView; vulkanCommandBuffer->needNewFragmentResourceDescriptorSet = true; } } @@ -7588,12 +7618,12 @@ static void VULKAN_BindFragmentStorageBuffers( for (i = 0; i < numBindings; i += 1) { bufferContainer = (VulkanBufferContainer *)storageBuffers[i]; - if (vulkanCommandBuffer->fragmentStorageBuffers[firstSlot + i] != bufferContainer->activeBuffer) { + if (vulkanCommandBuffer->fragmentStorageBufferBindings[firstSlot + i] != bufferContainer->activeBuffer->buffer) { VULKAN_INTERNAL_TrackBuffer( vulkanCommandBuffer, bufferContainer->activeBuffer); - vulkanCommandBuffer->fragmentStorageBuffers[firstSlot + i] = bufferContainer->activeBuffer; + vulkanCommandBuffer->fragmentStorageBufferBindings[firstSlot + i] = bufferContainer->activeBuffer->buffer; vulkanCommandBuffer->needNewFragmentResourceDescriptorSet = true; } } @@ -8107,15 +8137,15 @@ static void VULKAN_EndRenderPass( SDL_zeroa(vulkanCommandBuffer->vertexBufferOffsets); vulkanCommandBuffer->vertexBufferCount = 0; - SDL_zeroa(vulkanCommandBuffer->vertexSamplers); - SDL_zeroa(vulkanCommandBuffer->vertexSamplerTextures); - SDL_zeroa(vulkanCommandBuffer->vertexStorageTextures); - SDL_zeroa(vulkanCommandBuffer->vertexStorageBuffers); + SDL_zeroa(vulkanCommandBuffer->vertexSamplerBindings); + SDL_zeroa(vulkanCommandBuffer->vertexSamplerTextureViewBindings); + SDL_zeroa(vulkanCommandBuffer->vertexStorageTextureViewBindings); + SDL_zeroa(vulkanCommandBuffer->vertexStorageBufferBindings); - SDL_zeroa(vulkanCommandBuffer->fragmentSamplers); - SDL_zeroa(vulkanCommandBuffer->fragmentSamplerTextures); - SDL_zeroa(vulkanCommandBuffer->fragmentStorageTextures); - SDL_zeroa(vulkanCommandBuffer->fragmentStorageBuffers); + SDL_zeroa(vulkanCommandBuffer->fragmentSamplerBindings); + SDL_zeroa(vulkanCommandBuffer->fragmentSamplerTextureViewBindings); + SDL_zeroa(vulkanCommandBuffer->fragmentStorageTextureViewBindings); + SDL_zeroa(vulkanCommandBuffer->fragmentStorageBufferBindings); } static void VULKAN_BeginComputePass( @@ -8145,6 +8175,7 @@ static void VULKAN_BeginComputePass( VULKAN_TEXTURE_USAGE_MODE_COMPUTE_STORAGE_READ_WRITE); vulkanCommandBuffer->readWriteComputeStorageTextureSubresources[i] = subresource; + vulkanCommandBuffer->readWriteComputeStorageTextureViewBindings[i] = subresource->computeWriteView; VULKAN_INTERNAL_TrackTexture( vulkanCommandBuffer, @@ -8161,6 +8192,7 @@ static void VULKAN_BeginComputePass( VULKAN_BUFFER_USAGE_MODE_COMPUTE_STORAGE_READ_WRITE); vulkanCommandBuffer->readWriteComputeStorageBuffers[i] = buffer; + vulkanCommandBuffer->readWriteComputeStorageBufferBindings[i] = buffer->buffer; VULKAN_INTERNAL_TrackBuffer( vulkanCommandBuffer, @@ -8212,21 +8244,21 @@ static void VULKAN_BindComputeSamplers( VulkanTextureContainer *textureContainer = (VulkanTextureContainer *)textureSamplerBindings[i].texture; VulkanSampler *sampler = (VulkanSampler *)textureSamplerBindings[i].sampler; - if (vulkanCommandBuffer->computeSamplers[firstSlot + i] != sampler) { + if (vulkanCommandBuffer->computeSamplerBindings[firstSlot + i] != sampler->sampler) { VULKAN_INTERNAL_TrackSampler( vulkanCommandBuffer, sampler); - vulkanCommandBuffer->computeSamplers[firstSlot + i] = sampler; + vulkanCommandBuffer->computeSamplerBindings[firstSlot + i] = sampler->sampler; vulkanCommandBuffer->needNewComputeReadOnlyDescriptorSet = true; } - if (vulkanCommandBuffer->computeSamplerTextures[firstSlot + i] != textureContainer->activeTexture) { + if (vulkanCommandBuffer->computeSamplerTextureViewBindings[firstSlot + i] != textureContainer->activeTexture->fullView) { VULKAN_INTERNAL_TrackTexture( vulkanCommandBuffer, textureContainer->activeTexture); - vulkanCommandBuffer->computeSamplerTextures[firstSlot + i] = textureContainer->activeTexture; + vulkanCommandBuffer->computeSamplerTextureViewBindings[firstSlot + i] = textureContainer->activeTexture->fullView; vulkanCommandBuffer->needNewComputeReadOnlyDescriptorSet = true; } } @@ -8267,6 +8299,7 @@ static void VULKAN_BindComputeStorageTextures( textureContainer->activeTexture); vulkanCommandBuffer->readOnlyComputeStorageTextures[firstSlot + i] = textureContainer->activeTexture; + vulkanCommandBuffer->readOnlyComputeStorageTextureViewBindings[firstSlot + i] = textureContainer->activeTexture->fullView; vulkanCommandBuffer->needNewComputeReadOnlyDescriptorSet = true; } } @@ -8306,6 +8339,7 @@ static void VULKAN_BindComputeStorageBuffers( bufferContainer->activeBuffer); vulkanCommandBuffer->readOnlyComputeStorageBuffers[firstSlot + i] = bufferContainer->activeBuffer; + vulkanCommandBuffer->readOnlyComputeStorageBufferBindings[firstSlot + i] = bufferContainer->activeBuffer->buffer; vulkanCommandBuffer->needNewComputeReadOnlyDescriptorSet = true; } } @@ -8380,8 +8414,8 @@ static void VULKAN_INTERNAL_BindComputeDescriptorSets( currentWriteDescriptorSet->pTexelBufferView = NULL; currentWriteDescriptorSet->pBufferInfo = NULL; - imageInfos[imageInfoCount].sampler = commandBuffer->computeSamplers[i]->sampler; - imageInfos[imageInfoCount].imageView = commandBuffer->computeSamplerTextures[i]->fullView; + imageInfos[imageInfoCount].sampler = commandBuffer->computeSamplerBindings[i]; + imageInfos[imageInfoCount].imageView = commandBuffer->computeSamplerTextureViewBindings[i]; imageInfos[imageInfoCount].imageLayout = VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL; currentWriteDescriptorSet->pImageInfo = &imageInfos[imageInfoCount]; @@ -8404,7 +8438,7 @@ static void VULKAN_INTERNAL_BindComputeDescriptorSets( currentWriteDescriptorSet->pBufferInfo = NULL; imageInfos[imageInfoCount].sampler = VK_NULL_HANDLE; - imageInfos[imageInfoCount].imageView = commandBuffer->readOnlyComputeStorageTextures[i]->fullView; + imageInfos[imageInfoCount].imageView = commandBuffer->readOnlyComputeStorageTextureViewBindings[i]; imageInfos[imageInfoCount].imageLayout = VK_IMAGE_LAYOUT_GENERAL; currentWriteDescriptorSet->pImageInfo = &imageInfos[imageInfoCount]; @@ -8426,7 +8460,7 @@ static void VULKAN_INTERNAL_BindComputeDescriptorSets( currentWriteDescriptorSet->pTexelBufferView = NULL; currentWriteDescriptorSet->pImageInfo = NULL; - bufferInfos[bufferInfoCount].buffer = commandBuffer->readOnlyComputeStorageBuffers[i]->buffer; + bufferInfos[bufferInfoCount].buffer = commandBuffer->readOnlyComputeStorageBufferBindings[i]; bufferInfos[bufferInfoCount].offset = 0; bufferInfos[bufferInfoCount].range = VK_WHOLE_SIZE; @@ -8461,7 +8495,7 @@ static void VULKAN_INTERNAL_BindComputeDescriptorSets( currentWriteDescriptorSet->pBufferInfo = NULL; imageInfos[imageInfoCount].sampler = VK_NULL_HANDLE; - imageInfos[imageInfoCount].imageView = commandBuffer->readWriteComputeStorageTextureSubresources[i]->computeWriteView; + imageInfos[imageInfoCount].imageView = commandBuffer->readWriteComputeStorageTextureViewBindings[i]; imageInfos[imageInfoCount].imageLayout = VK_IMAGE_LAYOUT_GENERAL; currentWriteDescriptorSet->pImageInfo = &imageInfos[imageInfoCount]; @@ -8483,7 +8517,7 @@ static void VULKAN_INTERNAL_BindComputeDescriptorSets( currentWriteDescriptorSet->pTexelBufferView = NULL; currentWriteDescriptorSet->pImageInfo = NULL; - bufferInfos[bufferInfoCount].buffer = commandBuffer->readWriteComputeStorageBuffers[i]->buffer; + bufferInfos[bufferInfoCount].buffer = commandBuffer->readWriteComputeStorageBufferBindings[i]; bufferInfos[bufferInfoCount].offset = 0; bufferInfos[bufferInfoCount].range = VK_WHOLE_SIZE; @@ -8650,9 +8684,12 @@ static void VULKAN_EndComputePass( } } - // we don't need a barrier because sampler state is always the default if sampler bit is set - SDL_zeroa(vulkanCommandBuffer->computeSamplerTextures); - SDL_zeroa(vulkanCommandBuffer->computeSamplers); + // we don't need a barrier for sampler resources because sampler state is always the default if sampler bit is set + SDL_zeroa(vulkanCommandBuffer->computeSamplerTextureViewBindings); + SDL_zeroa(vulkanCommandBuffer->computeSamplerBindings); + + SDL_zeroa(vulkanCommandBuffer->readWriteComputeStorageTextureViewBindings); + SDL_zeroa(vulkanCommandBuffer->readWriteComputeStorageBufferBindings); vulkanCommandBuffer->currentComputePipeline = NULL; @@ -9518,21 +9555,23 @@ static SDL_GPUCommandBuffer *VULKAN_AcquireCommandBuffer( SDL_zeroa(commandBuffer->vertexBufferOffsets); commandBuffer->vertexBufferCount = 0; - SDL_zeroa(commandBuffer->vertexSamplerTextures); - SDL_zeroa(commandBuffer->vertexSamplers); - SDL_zeroa(commandBuffer->vertexStorageTextures); - SDL_zeroa(commandBuffer->vertexStorageBuffers); + SDL_zeroa(commandBuffer->vertexSamplerTextureViewBindings); + SDL_zeroa(commandBuffer->vertexSamplerBindings); + SDL_zeroa(commandBuffer->vertexStorageTextureViewBindings); + SDL_zeroa(commandBuffer->vertexStorageBufferBindings); - SDL_zeroa(commandBuffer->fragmentSamplerTextures); - SDL_zeroa(commandBuffer->fragmentSamplers); - SDL_zeroa(commandBuffer->fragmentStorageTextures); - SDL_zeroa(commandBuffer->fragmentStorageBuffers); + SDL_zeroa(commandBuffer->fragmentSamplerTextureViewBindings); + SDL_zeroa(commandBuffer->fragmentSamplerBindings); + SDL_zeroa(commandBuffer->fragmentStorageTextureViewBindings); + SDL_zeroa(commandBuffer->fragmentStorageBufferBindings); SDL_zeroa(commandBuffer->readWriteComputeStorageTextureSubresources); commandBuffer->readWriteComputeStorageTextureSubresourceCount = 0; SDL_zeroa(commandBuffer->readWriteComputeStorageBuffers); - SDL_zeroa(commandBuffer->computeSamplerTextures); - SDL_zeroa(commandBuffer->computeSamplers); + SDL_zeroa(commandBuffer->computeSamplerTextureViewBindings); + SDL_zeroa(commandBuffer->computeSamplerBindings); + SDL_zeroa(commandBuffer->readOnlyComputeStorageTextureViewBindings); + SDL_zeroa(commandBuffer->readOnlyComputeStorageBufferBindings); SDL_zeroa(commandBuffer->readOnlyComputeStorageTextures); SDL_zeroa(commandBuffer->readOnlyComputeStorageBuffers);