Fixing some shader stuff

CMakeLists.txt

@@ -77,15 +77,15 @@ target_link_libraries(Engine PUBLIC nlohmann_json::nlohmann_json)
 target_link_libraries(Engine PUBLIC crcpp)
 target_link_libraries(Engine PUBLIC fmt::fmt)
 target_link_libraries(Engine PUBLIC GPUOpen::VulkanMemoryAllocator)
-target_link_libraries(Engine PUBLIC shader-slang)
+target_link_libraries(Engine PUBLIC slang)
 target_link_libraries(Engine PUBLIC spirv-cross-glsl spirv-cross-msl)
 target_link_libraries(Engine PUBLIC glslang::glslang)
 if(APPLE)
+    target_include_directories(Engine PUBLIC /usr/local/include)
+    target_link_directories(Engine PUBLIC /usr/local/lib)
+    target_link_libraries(Engine PUBLIC metalirconverter)
     target_link_libraries(Engine PUBLIC metal)
     SET(CMAKE_OSX_DEPLOYMENT_TARGET 14.3)
-    install(FILES
-      ${EXTERNAL_ROOT}/dxc
-      DESTINATION ${CMAKE_INSTALL_PREFIX}/bin)
 endif()
 if(UNIX)
     target_link_libraries(Engine PUBLIC Threads::Threads)
@@ -121,8 +121,8 @@ else()
 endif()
 if(APPLE)
   #Metal lib throws that internally for some reason
-  target_compile_options(Engine PUBLIC -Wno-gnu-anonymous-struct)
-  target_compile_options(Editor PUBLIC -Wno-gnu-anonymous-struct)
+  target_compile_options(Engine PUBLIC -Wno-gnu-anonymous-struct -Wno-nested-anon-types)
+  target_compile_options(Editor PUBLIC -Wno-gnu-anonymous-struct -Wno-nested-anon-types)
 endif()
 target_compile_options(Engine PUBLIC "$<$<CONFIG:DEBUG>:-DENABLE_VALIDATION>")
 target_compile_options(Engine PUBLIC "$<$<CONFIG:DEBUG>:-DSEELE_DEBUG>")
@@ -137,7 +137,7 @@ if(WIN32)
         DEPENDS Editor)
 elseif(APPLE)
     add_custom_target(dll_copy ALL
-        COMMAND ${CMAKE_COMMAND} -E copy_if_different ${SLANG_ROOT}bin/libslang.dylib $<TARGET_FILE_DIR:Editor>
+        COMMAND ${CMAKE_COMMAND} -E true
         COMMAND_EXPAND_LISTS
         DEPENDS Editor)
 else()
@@ -157,6 +157,7 @@ install(
         Engine
         crcpp
         metal
+        slang
     EXPORT 
         EngineTargets
     FILE_SET HEADERS

cmake/SuperBuild.cmake

@@ -2,36 +2,37 @@ include (ExternalProject)
 
 
 #--------------SLang------------------------------
-add_library(shader-slang-glslang SHARED IMPORTED)
-add_library(shader-slang SHARED IMPORTED)
-if(WIN32)
-    set(SLANG_ROOT ${EXTERNAL_ROOT}/vcpkg/packages/shader-slang_x64-windows/)
-    set_target_properties(shader-slang-glslang PROPERTIES IMPORTED_LOCATION ${SLANG_ROOT}/bin/slang-glslang.dll)
-    set_target_properties(shader-slang-glslang PROPERTIES IMPORTED_IMPLIB ${SLANG_ROOT}/lib/slang.lib)
-    set_target_properties(shader-slang PROPERTIES IMPORTED_LOCATION ${SLANG_ROOT}/bin/slang.dll)
-    set_target_properties(shader-slang PROPERTIES IMPORTED_IMPLIB ${SLANG_ROOT}/lib/slang.lib)
-    target_link_libraries(shader-slang INTERFACE shader-slang-glslang)
-    install(FILES 
-        ${SLANG_ROOT}/bin/slang-glslang.dll
-        ${SLANG_ROOT}/bin/slang.dll
-    DESTINATION ${CMAKE_INSTALL_PREFIX}/bin)
-    install(FILES 
-        ${SLANG_ROOT}/lib/slang.lib
-    DESTINATION ${CMAKE_INSTALL_PREFIX}/lib)
-elseif(APPLE)
-    set(SLANG_ROOT ${EXTERNAL_ROOT}/vcpkg/packages/shader-slang_arm64-osx/)
-    set_target_properties(shader-slang-glslang PROPERTIES IMPORTED_LOCATION ${SLANG_ROOT}/bin/libslang-glslang.dylib)
-    set_target_properties(shader-slang PROPERTIES IMPORTED_LOCATION ${SLANG_ROOT}/bin/libslang.dylib)
-    target_link_libraries(shader-slang INTERFACE shader-slang-glslang)
-    install(FILES
-        ${SLANG_ROOT}/bin/libslang.dylib
-        ${SLANG_ROOT}/bin/libslang-glslang.dylib
-    DESTINATION ${CMAKE_INSTALL_PREFIX}/bin)
-endif()
-target_include_directories(shader-slang INTERFACE 
-    $<BUILD_INTERFACE:${SLANG_ROOT}/include/>
-    $<INSTALL_INTERFACE:include>
-)
+set(SLANG_ROOT ${EXTERNAL_ROOT}/slang)
+#add_library(shader-slang SHARED IMPORTED)
+#if(WIN32)
+#    add_library(shader-slang-glslang SHARED IMPORTED)
+#    set(SLANG_ROOT ${EXTERNAL_ROOT}/slang/bin/windows-x64/release/)
+#    set_target_properties(shader-slang-glslang PROPERTIES IMPORTED_LOCATION ${SLANG_ROOT}/bin/slang-glslang.dll)
+#    set_target_properties(shader-slang-glslang PROPERTIES IMPORTED_IMPLIB ${SLANG_ROOT}/lib/slang.lib)
+#    set_target_properties(shader-slang PROPERTIES IMPORTED_LOCATION ${SLANG_ROOT}/bin/slang.dll)
+#    set_target_properties(shader-slang PROPERTIES IMPORTED_IMPLIB ${SLANG_ROOT}/lib/slang.lib)
+#    target_link_libraries(shader-slang INTERFACE shader-slang-glslang)
+#    install(FILES 
+#        ${SLANG_ROOT}/bin/slang-glslang.dll
+#        ${SLANG_ROOT}/bin/slang.dll
+#    DESTINATION ${CMAKE_INSTALL_PREFIX}/bin)
+#    install(FILES 
+#        ${SLANG_ROOT}/lib/slang.lib
+#    DESTINATION ${CMAKE_INSTALL_PREFIX}/lib)
+#elseif(APPLE)
+#    set(BINARY_ROOT ${SLANG_ROOT}/bin/macosx-aarch64/release/)
+#    set_target_properties(shader-slang PROPERTIES IMPORTED_LOCATION ${BINARY_ROOT}/libslang.dylib)
+#    install(FILES
+#        ${BINARY_ROOT}/libslang.dylib
+#    DESTINATION ${CMAKE_INSTALL_PREFIX}/bin)
+#endif()
+#target_include_directories(shader-slang INTERFACE 
+#    $<BUILD_INTERFACE:${SLANG_ROOT}/>
+#    $<INSTALL_INTERFACE:include>
+#)
+
+add_subdirectory(${SLANG_ROOT})
+target_compile_definitions(compiler-core PUBLIC SLANG_ENABLE_DXIL_SUPPORT=1)
 
 #--------------CRC++------------------------------
 add_library(crcpp INTERFACE)

res/shaders/BasePass.slang

@@ -26,10 +26,10 @@ float4 fragmentMain(in FragmentParameter params) : SV_Target
     {
         result += pLightEnv.directionalLights[i].illuminate(lightingParams, brdf);
     }
-    for(uint i = 0; i < lightCount; ++i)
+    for(uint i = 0; i < pLightEnv.numPointLights; ++i)
     {
-        uint lightIndex = pLightCullingData.lightIndexList[startOffset + i];
-        result += pLightEnv.pointLights[lightIndex].illuminate(lightingParams, brdf);
+        //uint lightIndex = pLightCullingData.lightIndexList[startOffset + i];
+        result += pLightEnv.pointLights[i].illuminate(lightingParams, brdf);
     }
     return float4(params.vertexColor, 1.0f);
 }

res/shaders/Debug.slang

@@ -2,8 +2,8 @@ import Common;
 
 struct DebugVertex
 {
-	float3 position;
-	float3 color;
+	float3 position : POSITION;
+	float3 color : COLOR0;
 };
 
 struct Params

res/shaders/MeshletBasePass.slang

@@ -65,6 +65,12 @@ struct PrimitiveAttributes
 	uint cull: SV_CullPrimitive;
 };
 
+uint unpackPrimitiveIndices(uint index)
+{
+  uint32_t packed = pScene.primitiveIndices[index / 4];
+  return (packed >> (index % 4)) & 0xff;
+}
+
 [numthreads(MESH_GROUP_SIZE, 1, 1)]
 [outputtopology("triangle")]
 [shader("mesh")]
@@ -84,9 +90,9 @@ void meshMain(
 	{
 		uint p = min(i, m.primitiveCount - 1);
 		{
-			uint local_idx0 = pScene.primitiveIndices[m.primitiveOffset + (p * 3) + 0];
-			uint local_idx1 = pScene.primitiveIndices[m.primitiveOffset + (p * 3) + 1];
-			uint local_idx2 = pScene.primitiveIndices[m.primitiveOffset + (p * 3) + 2];
+			uint local_idx0 = unpackPrimitiveIndices(m.primitiveOffset + (p * 3) + 0);
+			uint local_idx1 = unpackPrimitiveIndices(m.primitiveOffset + (p * 3) + 1);
+			uint local_idx2 = unpackPrimitiveIndices(m.primitiveOffset + (p * 3) + 2);
 			indices[p] = uint3(local_idx0, local_idx1, local_idx2);
 		}
 	}

res/shaders/lib/Scene.slang

@@ -14,12 +14,12 @@ struct MeshletDescription
 struct MeshData
 {
 	BoundingSphere bounding;
-    uint32_t numMeshlets;
-    uint32_t meshletOffset;
-    uint32_t firstIndex;
-    uint32_t numIndices;
-    uint32_t indicesOffset;
-    uint32_t pad0[3];
+  uint32_t numMeshlets;
+  uint32_t meshletOffset;
+  uint32_t firstIndex;
+  uint32_t numIndices;
+  uint32_t indicesOffset;
+  uint32_t pad0[3];
 };
 
 static const uint MAX_VERTICES = 64;
@@ -38,7 +38,8 @@ struct Scene
 	StructuredBuffer<InstanceData> instances;
 	StructuredBuffer<MeshData> meshData;
 	StructuredBuffer<MeshletDescription> meshletInfos;
-	StructuredBuffer<uint8_t> primitiveIndices;
+  // uint8_t is not valid in DXIL
+	StructuredBuffer<uint32_t> primitiveIndices;
 	StructuredBuffer<uint32_t> vertexIndices;
 };
 

src/Engine/Graphics/Enums.h

@@ -1596,15 +1596,16 @@ typedef enum SeShaderStageFlagBits
     SE_SHADER_STAGE_COMPUTE_BIT = 0x00000020,
     SE_SHADER_STAGE_ALL_GRAPHICS = 0x0000001F,
     SE_SHADER_STAGE_ALL = 0x7FFFFFFF,
-    SE_SHADER_STAGE_RAYGEN_BIT_NV = 0x00000100,
-    SE_SHADER_STAGE_ANY_HIT_BIT_NV = 0x00000200,
-    SE_SHADER_STAGE_CLOSEST_HIT_BIT_NV = 0x00000400,
-    SE_SHADER_STAGE_MISS_BIT_NV = 0x00000800,
-    SE_SHADER_STAGE_INTERSECTION_BIT_NV = 0x00001000,
-    SE_SHADER_STAGE_CALLABLE_BIT_NV = 0x00002000,
-    SE_SHADER_STAGE_TASK_BIT_NV = 0x00000040,
-    SE_SHADER_STAGE_MESH_BIT_NV = 0x00000080,
-    SE_SHADER_STAGE_FLAG_BITS_MAX_ENUM = 0x7FFFFFFF
+    SE_SHADER_STAGE_RAYGEN_BIT_KHR = 0x00000100,
+    SE_SHADER_STAGE_ANY_HIT_BIT_KHR = 0x00000200,
+    SE_SHADER_STAGE_CLOSEST_HIT_BIT_KHR = 0x00000400,
+    SE_SHADER_STAGE_MISS_BIT_KHR = 0x00000800,
+    SE_SHADER_STAGE_INTERSECTION_BIT_KHR = 0x00001000,
+    SE_SHADER_STAGE_CALLABLE_BIT_KHR = 0x00002000,
+    SE_SHADER_STAGE_TASK_BIT_EXT = 0x00000040,
+    SE_SHADER_STAGE_MESH_BIT_EXT = 0x00000080,
+    SE_SHADER_STAGE_SUBPASS_SHADING_BIT_HUAWEI = 0x00004000,
+    SE_SHADER_STAGE_CLUSTER_CULLING_BIT_HUAWEI = 0x00080000,
 } SeShaderStageFlagBits;
 typedef SeFlags SePipelineVertexInputStateCreateFlags;
 typedef SeFlags SePipelineInputAssemblyStateCreateFlags;

src/Engine/Graphics/Metal/Buffer.mm

@@ -54,7 +54,9 @@ void VertexBuffer::download(Array<uint8>& buffer) {
 void VertexBuffer::executeOwnershipBarrier(Gfx::QueueType newOwner) { currentOwner = newOwner; }
 
 void VertexBuffer::executePipelineBarrier(Gfx::SeAccessFlags srcAccess, Gfx::SePipelineStageFlags srcStage,
-                                          Gfx::SeAccessFlags dstAccess, Gfx::SePipelineStageFlags dstStage) {}
+                                          Gfx::SeAccessFlags dstAccess, Gfx::SePipelineStageFlags dstStage) {
+  
+}
 
 IndexBuffer::IndexBuffer(PGraphics graphics, const IndexBufferCreateInfo& createInfo)
     : Gfx::IndexBuffer(graphics->getFamilyMapping(), createInfo.sourceData.size, createInfo.indexType,

src/Engine/Graphics/Metal/Command.h

@@ -1,6 +1,10 @@
 #pragma once
-#include "Graphics/Command.h"
 #include "Buffer.h"
+#include "Graphics/Command.h"
+#include "Metal/MTLBlitCommandEncoder.hpp"
+#include "Metal/MTLCommandBuffer.hpp"
+#include "Metal/MTLComputeCommandEncoder.hpp"
+#include "Metal/MTLRenderCommandEncoder.hpp"
 #include "RenderPass.h"
 #include "Resources.h"
 
@@ -12,106 +16,111 @@ DECLARE_REF(RenderCommand)
 DECLARE_REF(Graphics)
 DECLARE_REF(IndexBuffer)
 DECLARE_REF(GraphicsPipeline)
-class Command
-{
+class Command {
 public:
-    Command(PGraphics graphics, MTL::CommandBuffer* cmdBuffer);
-    ~Command();
-    void beginRenderPass(PRenderPass renderPass);
-    void endRenderPass();
-    void present(MTL::Drawable* drawable);
-    void end(PEvent signal);
-    void executeCommands(Array<Gfx::ORenderCommand> commands);
-    void executeCommands(Array<Gfx::OComputeCommand> commands);
-    void waitDeviceIdle();
-    void signalEvent(PEvent event);
-    void waitForEvent(PEvent event);
-    MTL::RenderCommandEncoder* createRenderEncoder() { return renderEncoder->renderCommandEncoder(); }
-    MTL::ComputeCommandEncoder* createComputeEncoder() { return cmdBuffer->computeCommandEncoder(); }
-    PEvent getCompletedEvent() const { return completed; }
-    constexpr MTL::CommandBuffer* getHandle() const
+  Command(PGraphics graphics, MTL::CommandBuffer* cmdBuffer);
+  ~Command();
+  void beginRenderPass(PRenderPass renderPass);
+  void endRenderPass();
+  void present(MTL::Drawable* drawable);
+  void end(PEvent signal);
+  void waitDeviceIdle();
+  void signalEvent(PEvent event);
+  void waitForEvent(PEvent event);
+  MTL::RenderCommandEncoder* createRenderEncoder() { return parallelEncoder->renderCommandEncoder(); }
+  MTL::BlitCommandEncoder* getBlitEncoder() {
+    assert(!parallelEncoder);
+    if(blitEncoder == nullptr)
     {
-      return cmdBuffer;
+      blitEncoder = cmdBuffer->blitCommandEncoder();
     }
+    return blitEncoder;
+  }
+  PEvent getCompletedEvent() const { return completed; }
+  constexpr MTL::CommandBuffer* getHandle() const { return cmdBuffer; }
+
 private:
-    PGraphics graphics;
-    OEvent completed;
-    MTL::CommandBuffer* cmdBuffer;
-    MTL::ParallelRenderCommandEncoder* renderEncoder;
+  PGraphics graphics;
+  OEvent completed;
+  MTL::CommandBuffer* cmdBuffer;
+  MTL::ParallelRenderCommandEncoder* parallelEncoder = nullptr;
+  MTL::BlitCommandEncoder* blitEncoder = nullptr;
 };
 DEFINE_REF(Command)
-class RenderCommand : public Gfx::RenderCommand
-{
+class RenderCommand : public Gfx::RenderCommand {
 public:
-    RenderCommand(MTL::RenderCommandEncoder* encoder);
-    virtual ~RenderCommand();
-    void end();
-	virtual void setViewport(Gfx::PViewport viewport) override;
-	virtual void bindPipeline(Gfx::PGraphicsPipeline pipeline) override;
-	virtual void bindDescriptor(Gfx::PDescriptorSet descriptorSet) override;
-	virtual void bindDescriptor(const Array<Gfx::PDescriptorSet>& descriptorSets) override;
-	virtual void bindVertexBuffer(const Array<Gfx::PVertexBuffer>& buffers) override;
-	virtual void bindIndexBuffer(Gfx::PIndexBuffer indexBuffer) override;
-	virtual void pushConstants(Gfx::PPipelineLayout layout, Gfx::SeShaderStageFlags stage, uint32 offset, uint32 size, const void* data) override;
-	virtual void draw(uint32 vertexCount, uint32 instanceCount, int32 firstVertex, uint32 firstInstance) override;
-	virtual void drawIndexed(uint32 indexCount, uint32 instanceCount, int32 firstIndex, uint32 vertexOffset, uint32 firstInstance) override; 
-	virtual void drawMesh(uint32 groupX, uint32 groupY, uint32 groupZ) override;
+  RenderCommand(MTL::RenderCommandEncoder* encode, const std::string& name);
+  virtual ~RenderCommand();
+  void end();
+  virtual void setViewport(Gfx::PViewport viewport) override;
+  virtual void bindPipeline(Gfx::PGraphicsPipeline pipeline) override;
+  virtual void bindDescriptor(Gfx::PDescriptorSet descriptorSet) override;
+  virtual void bindDescriptor(const Array<Gfx::PDescriptorSet>& descriptorSets) override;
+  virtual void bindVertexBuffer(const Array<Gfx::PVertexBuffer>& buffers) override;
+  virtual void bindIndexBuffer(Gfx::PIndexBuffer indexBuffer) override;
+  virtual void pushConstants(Gfx::PPipelineLayout layout, Gfx::SeShaderStageFlags stage, uint32 offset, uint32 size,
+                             const void* data) override;
+  virtual void draw(uint32 vertexCount, uint32 instanceCount, int32 firstVertex, uint32 firstInstance) override;
+  virtual void drawIndexed(uint32 indexCount, uint32 instanceCount, int32 firstIndex, uint32 vertexOffset,
+                           uint32 firstInstance) override;
+  virtual void drawMesh(uint32 groupX, uint32 groupY, uint32 groupZ) override;
+
 private:
-    MTL::RenderCommandEncoder* encoder;
-    PIndexBuffer boundIndexBuffer;
-    PGraphicsPipeline boundPipeline;
+  MTL::RenderCommandEncoder* encoder;
+  PIndexBuffer boundIndexBuffer;
+  PGraphicsPipeline boundPipeline;
+  std::string name;
 };
 DEFINE_REF(RenderCommand)
-class ComputeCommand : public Gfx::ComputeCommand
-{
+class ComputeCommand : public Gfx::ComputeCommand {
 public:
-    ComputeCommand(MTL::ComputeCommandEncoder* encoder);
-    virtual ~ComputeCommand();
-    void end();
-	virtual void bindPipeline(Gfx::PComputePipeline pipeline) override;
-	virtual void bindDescriptor(Gfx::PDescriptorSet set) override;
-	virtual void bindDescriptor(const Array<Gfx::PDescriptorSet>& sets) override;
-	virtual void pushConstants(Gfx::PPipelineLayout layout, Gfx::SeShaderStageFlags stage, uint32 offset, uint32 size, const void* data) override;
-	virtual void dispatch(uint32 threadX, uint32 threadY, uint32 threadZ) override;
+  ComputeCommand(MTL::CommandBuffer* cmdBuffer, const std::string& name);
+  virtual ~ComputeCommand();
+  void end();
+  virtual void bindPipeline(Gfx::PComputePipeline pipeline) override;
+  virtual void bindDescriptor(Gfx::PDescriptorSet set) override;
+  virtual void bindDescriptor(const Array<Gfx::PDescriptorSet>& sets) override;
+  virtual void pushConstants(Gfx::PPipelineLayout layout, Gfx::SeShaderStageFlags stage, uint32 offset, uint32 size,
+                             const void* data) override;
+  virtual void dispatch(uint32 threadX, uint32 threadY, uint32 threadZ) override;
+
 private:
-    MTL::ComputeCommandEncoder* encoder;
+  MTL::CommandBuffer* commandBuffer;
+  MTL::ComputeCommandEncoder* encoder;
+  std::string name;
 };
 DEFINE_REF(ComputeCommand)
-class CommandQueue
-{
+class CommandQueue {
 public:
-    CommandQueue(PGraphics graphics);
-    ~CommandQueue();
-    constexpr MTL::CommandQueue* getHandle()
-    {
-        return queue;
-    }
-    PCommand getCommands() {return activeCommand;} //TODO
-    ORenderCommand getRenderCommand(const std::string& name);
-    OComputeCommand getComputeCommand(const std::string& name);
-    void submitCommands(PEvent signal = nullptr);
+  CommandQueue(PGraphics graphics);
+  ~CommandQueue();
+  constexpr MTL::CommandQueue* getHandle() { return queue; }
+  PCommand getCommands() { return activeCommand; }
+  ORenderCommand getRenderCommand(const std::string& name);
+  OComputeCommand getComputeCommand(const std::string& name);
+  void executeCommands(Array<Gfx::ORenderCommand> commands);
+  void executeCommands(Array<Gfx::OComputeCommand> commands);
+  void submitCommands(PEvent signal = nullptr);
+
 private:
-    PGraphics graphics;
-    MTL::CommandQueue* queue;
-    OCommand activeCommand;
-    Array<OCommand> pendingCommands;
+  PGraphics graphics;
+  MTL::CommandQueue* queue;
+  OCommand activeCommand;
+  Array<OCommand> pendingCommands;
 };
-class IOCommandQueue
-{
+class IOCommandQueue {
 public:
   IOCommandQueue(PGraphics graphics);
   ~IOCommandQueue();
-  constexpr MTL::IOCommandQueue* getHandle()
-  {
-    return queue;
-  }
-  PCommand getCommands() {return activeCommand;} // TODO
+  constexpr MTL::IOCommandQueue* getHandle() { return queue; }
+  PCommand getCommands() { return activeCommand; } // TODO
   void submitCommands(PEvent signal = nullptr);
+
 private:
   PGraphics graphics;
   MTL::IOCommandQueue* queue;
   OCommand activeCommand;
 };
 DEFINE_REF(IOCommandQueue)
-}
-}
+} // namespace Metal
+} // namespace Seele

src/Engine/Graphics/Metal/Command.mm

@@ -1,31 +1,43 @@
 #include "Command.h"
 #include "Buffer.h"
+#include "Containers/Array.h"
 #include "Descriptor.h"
 #include "Enums.h"
+#include "Graphics/Command.h"
 #include "Graphics/Enums.h"
 #include "Graphics/Graphics.h"
-#include "Metal/MTLCommandBuffer.hpp"
+#include "Graphics/Resources.h"
 #include "Pipeline.h"
 #include "Resources.h"
 #include "Window.h"
+#include <Metal/Metal.h>
 
 using namespace Seele;
 using namespace Seele::Metal;
 
 Command::Command(PGraphics graphics, MTL::CommandBuffer* cmdBuffer)
-    : graphics(graphics), completed(new Event(graphics)), cmdBuffer(cmdBuffer), renderEncoder(nullptr) {}
+    : graphics(graphics), completed(new Event(graphics)), cmdBuffer(cmdBuffer) {}
 
-Command::~Command() { cmdBuffer->release(); }
+Command::~Command() {}
 
 void Command::beginRenderPass(PRenderPass renderPass) {
-  renderEncoder = cmdBuffer->parallelRenderCommandEncoder(renderPass->getDescriptor());
+  if (blitEncoder) {
+    blitEncoder->endEncoding();
+    blitEncoder = nullptr;
+  }
+  parallelEncoder = cmdBuffer->parallelRenderCommandEncoder(renderPass->getDescriptor());
 }
 
-void Command::endRenderPass() { renderEncoder->endEncoding(); }
+void Command::endRenderPass() {
+  parallelEncoder->endEncoding();
+  parallelEncoder = nullptr;
+}
 
 void Command::present(MTL::Drawable* drawable) { cmdBuffer->presentDrawable(drawable); }
 
 void Command::end(PEvent signal) {
+  assert(!parallelEncoder);
+  blitEncoder->endEncoding();
   if (signal != nullptr) {
     cmdBuffer->encodeSignalEvent(signal->getHandle(), 1);
   }
@@ -33,29 +45,16 @@ void Command::end(PEvent signal) {
   cmdBuffer->commit();
 }
 
-void Command::executeCommands(Array<Gfx::ORenderCommand> commands) {
-  for (auto& command : commands) {
-    auto cmd = Gfx::PRenderCommand(command).cast<RenderCommand>();
-    cmd->end();
-  }
-}
-
-void Command::executeCommands(Array<Gfx::OComputeCommand> commands) {
-  for (auto& command : commands) {
-    auto cmd = Gfx::PComputeCommand(command).cast<ComputeCommand>();
-    cmd->end();
-  }
-}
-
 void Command::waitDeviceIdle() { cmdBuffer->waitUntilCompleted(); }
 
 void Command::signalEvent(PEvent event) { cmdBuffer->encodeSignalEvent(event->getHandle(), 1); }
 
 void Command::waitForEvent(PEvent event) { cmdBuffer->encodeWait(event->getHandle(), 1); }
 
-RenderCommand::RenderCommand(MTL::RenderCommandEncoder* encoder) : encoder(encoder) {}
+RenderCommand::RenderCommand(MTL::RenderCommandEncoder* encoder, const std::string& name)
+    : encoder(encoder), name(name) {}
 
-RenderCommand::~RenderCommand() { encoder->release(); }
+RenderCommand::~RenderCommand() {}
 
 void RenderCommand::end() { encoder->endEncoding(); }
 
@@ -88,11 +87,10 @@ void RenderCommand::bindDescriptor(const Array<Gfx::PDescriptorSet>& descriptorS
     bindDescriptor(set);
   }
 }
-
 void RenderCommand::bindVertexBuffer(const Array<Gfx::PVertexBuffer>& buffers) {
   uint32 i = 0;
   for (auto buffer : buffers) {
-    encoder->setVertexBuffer(buffer.cast<VertexBuffer>()->getHandle(), 0, i++);
+    encoder->setVertexBuffer(buffer.cast<VertexBuffer>()->getHandle(), 0, METAL_VERTEXBUFFER_OFFSET + i++);
   }
 }
 
@@ -125,11 +123,15 @@ void RenderCommand::drawMesh(uint32 groupX, uint32 groupY, uint32 groupZ) {
   encoder->drawMeshThreadgroups(MTL::Size(groupX, groupY, groupZ), MTL::Size(128, 128, 128), MTL::Size(32, 32, 32));
 }
 
-ComputeCommand::ComputeCommand(MTL::ComputeCommandEncoder* encoder) : encoder(encoder) {}
+ComputeCommand::ComputeCommand(MTL::CommandBuffer* cmdBuffer, const std::string& name)
+    : commandBuffer(cmdBuffer), encoder(cmdBuffer->computeCommandEncoder()), name(name) {}
 
-ComputeCommand::~ComputeCommand() { encoder->release(); }
+ComputeCommand::~ComputeCommand() {}
 
-void ComputeCommand::end() { encoder->endEncoding(); }
+void ComputeCommand::end() {
+  encoder->endEncoding();
+  commandBuffer->commit();
+}
 
 void ComputeCommand::bindPipeline(Gfx::PComputePipeline pipeline) {
   encoder->setComputePipelineState(pipeline.cast<ComputePipeline>()->getHandle());
@@ -159,34 +161,52 @@ void ComputeCommand::dispatch(uint32 threadX, uint32 threadY, uint32 threadZ) {
 
 CommandQueue::CommandQueue(PGraphics graphics) : graphics(graphics) {
   queue = graphics->getDevice()->newCommandQueue();
-  activeCommand = new Command(graphics, queue->commandBuffer());
+  MTL::CommandBufferDescriptor* descriptor = MTL::CommandBufferDescriptor::alloc()->init();
+  descriptor->setErrorOptions(MTL::CommandBufferErrorOptionEncoderExecutionStatus);
+  activeCommand = new Command(graphics, queue->commandBuffer(descriptor));
 }
 
 CommandQueue::~CommandQueue() { queue->release(); }
 
 ORenderCommand CommandQueue::getRenderCommand(const std::string& name) {
-  return new RenderCommand(activeCommand->createRenderEncoder());
+  return new RenderCommand(activeCommand->createRenderEncoder(), name);
 }
 
 OComputeCommand CommandQueue::getComputeCommand(const std::string& name) {
-  return new ComputeCommand(activeCommand->createComputeEncoder());
+  return new ComputeCommand(queue->commandBuffer(), name);
+}
+
+void CommandQueue::executeCommands(Array<Gfx::ORenderCommand> commands) {
+  for (auto& command : commands) {
+    auto metalCmd = Gfx::PRenderCommand(command).cast<RenderCommand>();
+    metalCmd->end();
+  }
+}
+
+void CommandQueue::executeCommands(Array<Gfx::OComputeCommand> commands) {
+  submitCommands();
+  for (auto& command : commands) {
+    auto metalCmd = Gfx::PComputeCommand(command).cast<ComputeCommand>();
+    metalCmd->end();
+  }
 }
 
 void CommandQueue::submitCommands(PEvent signalSemaphore) {
   activeCommand->getHandle()->addCompletedHandler(MTL::CommandBufferHandler([&](MTL::CommandBuffer* cmdBuffer) {
-    for(auto it = pendingCommands.begin(); it != pendingCommands.end(); it++)
-    {
-      if((*it)->getHandle() == cmdBuffer)
-      {
+    for (auto it = pendingCommands.begin(); it != pendingCommands.end(); it++) {
+      if ((*it)->getHandle() == cmdBuffer) {
         pendingCommands.remove(it);
         return;
       }
     }
   }));
   activeCommand->end(signalSemaphore);
+  activeCommand->waitDeviceIdle();
   PEvent prevCmdEvent = activeCommand->getCompletedEvent();
   pendingCommands.add(std::move(activeCommand));
-  activeCommand = new Command(graphics, queue->commandBuffer());
+  MTL::CommandBufferDescriptor* descriptor = MTL::CommandBufferDescriptor::alloc()->init();
+  descriptor->setErrorOptions(MTL::CommandBufferErrorOptionEncoderExecutionStatus);
+  activeCommand = new Command(graphics, queue->commandBuffer(descriptor));
   activeCommand->waitForEvent(prevCmdEvent);
 }
 

src/Engine/Graphics/Metal/Enums.h

@@ -7,6 +7,9 @@ namespace Seele
 {
 namespace Metal
 {
+constexpr uint64 METAL_VERTEXBUFFER_OFFSET = 6;// something about metal vertex fetch
+constexpr uint64 METAL_VERTEXATTRIBUTE_OFFSET = 11;
+
 MTL::PixelFormat cast(Gfx::SeFormat format);
 Gfx::SeFormat cast(MTL::PixelFormat format);
 MTL::LoadAction cast(Gfx::SeAttachmentLoadOp loadOp);

src/Engine/Graphics/Metal/Graphics.mm

@@ -31,6 +31,7 @@ void Graphics::init(GraphicsInitializer)
   queue = new CommandQueue(this);
   ioQueue = new IOCommandQueue(this);
   cache = new PipelineCache(this, "pipelines.metal");
+  meshShadingEnabled = true;
 }
 
 Gfx::OWindow Graphics::createWindow(const WindowCreateInfo &createInfo)
@@ -65,12 +66,12 @@ void Graphics::waitDeviceIdle()
 
 void Graphics::executeCommands(Array<Gfx::ORenderCommand> commands)
 {
-  queue->getCommands()->executeCommands(std::move(commands));  
+  queue->executeCommands(std::move(commands));  
 }
 
 void Graphics::executeCommands(Array<Gfx::OComputeCommand> commands)
 {
-  queue->getCommands()->executeCommands(std::move(commands));
+  queue->executeCommands(std::move(commands));
 }
 
 Gfx::OTexture2D Graphics::createTexture2D(const TextureCreateInfo &createInfo)

src/Engine/Graphics/Metal/PipelineCache.mm

@@ -2,6 +2,7 @@
 #include "Descriptor.h"
 #include "Enums.h"
 #include "Foundation/NSError.hpp"
+#include "Foundation/NSString.hpp"
 #include "Graphics.h"
 #include "Graphics/Descriptor.h"
 #include "Graphics/Enums.h"
@@ -13,6 +14,7 @@
 #include "Metal/MTLVertexDescriptor.hpp"
 #include "Shader.h"
 #include "Texture.h"
+#include <iostream>
 
 using namespace Seele;
 using namespace Seele::Metal;
@@ -26,14 +28,13 @@ PGraphicsPipeline PipelineCache::createPipeline(Gfx::LegacyPipelineCreateInfo cr
 
   MTL::RenderPipelineDescriptor* pipelineDescriptor = MTL::RenderPipelineDescriptor::alloc()->init();
 
-  MTL::VertexDescriptor* vertexDescriptor = MTL::VertexDescriptor::alloc()->init();
-  MTL::VertexAttributeDescriptorArray* attributes = vertexDescriptor->attributes();
-  if(createInfo.vertexInput != nullptr)
-  {
+  MTL::VertexDescriptor* vertexDescriptor = pipelineDescriptor->vertexDescriptor()->init();
+  MTL::VertexAttributeDescriptorArray* attributes = vertexDescriptor->attributes()->init();
+  if (createInfo.vertexInput != nullptr) {
     const auto& vertexInfo = createInfo.vertexInput->getInfo();
     for (size_t attr = 0; attr < vertexInfo.attributes.size(); ++attr) {
-      MTL::VertexAttributeDescriptor* attribute = MTL::VertexAttributeDescriptor::alloc()->init();
-      attribute->setBufferIndex(vertexInfo.attributes[attr].binding);
+      MTL::VertexAttributeDescriptor* attribute = attributes->object(attr + METAL_VERTEXATTRIBUTE_OFFSET)->init();
+      attribute->setBufferIndex(vertexInfo.attributes[attr].binding + METAL_VERTEXBUFFER_OFFSET);
       switch (vertexInfo.attributes[attr].format) {
       case Gfx::SE_FORMAT_R32G32B32_SFLOAT:
         attribute->setFormat(MTL::VertexFormatFloat3);
@@ -42,12 +43,11 @@ PGraphicsPipeline PipelineCache::createPipeline(Gfx::LegacyPipelineCreateInfo cr
         throw std::logic_error("TODO");
       }
       attribute->setOffset(vertexInfo.attributes[attr].offset);
-      attributes->setObject(attribute, attr);
     }
 
-    MTL::VertexBufferLayoutDescriptorArray* bufferLayout = vertexDescriptor->layouts();
+    MTL::VertexBufferLayoutDescriptorArray* bufferLayout = vertexDescriptor->layouts()->init();
     for (size_t binding = 0; binding < vertexInfo.bindings.size(); ++binding) {
-      MTL::VertexBufferLayoutDescriptor* buffer = MTL::VertexBufferLayoutDescriptor::alloc()->init();
+      MTL::VertexBufferLayoutDescriptor* buffer = bufferLayout->object(binding + METAL_VERTEXBUFFER_OFFSET)->init();
       buffer->setStride(vertexInfo.bindings[binding].stride);
       buffer->setStepRate(1);
       switch (vertexInfo.bindings[binding].inputRate) {
@@ -58,7 +58,6 @@ PGraphicsPipeline PipelineCache::createPipeline(Gfx::LegacyPipelineCreateInfo cr
         buffer->setStepFunction(MTL::VertexStepFunctionPerInstance);
         break;
       }
-      bufferLayout->setObject(buffer, binding);
     }
   }
   pipelineDescriptor->setVertexDescriptor(vertexDescriptor);
@@ -118,14 +117,15 @@ PGraphicsPipeline PipelineCache::createPipeline(Gfx::LegacyPipelineCreateInfo cr
     type = MTL::PrimitiveTypeTriangle;
     break;
   }
-
   NS::Error* error;
   graphicsPipelines[hash] =
       new GraphicsPipeline(graphics, type, graphics->getDevice()->newRenderPipelineState(pipelineDescriptor, &error),
                            std::move(createInfo.pipelineLayout));
-  assert(!error);
+  if (error) {
+    std::cout << error->localizedDescription()->cString(NS::ASCIIStringEncoding) << std::endl;
+    assert(false);
+  }
 
-  vertexDescriptor->release();
   pipelineDescriptor->release();
   return graphicsPipelines[hash];
 }
@@ -133,7 +133,7 @@ PGraphicsPipeline PipelineCache::createPipeline(Gfx::LegacyPipelineCreateInfo cr
 PGraphicsPipeline PipelineCache::createPipeline(Gfx::MeshPipelineCreateInfo createInfo) {
   MTL::MeshRenderPipelineDescriptor* pipelineDescriptor = MTL::MeshRenderPipelineDescriptor::alloc()->init();
 
-  pipelineDescriptor->setMeshFunction(createInfo.meshShader.cast<VertexShader>()->getFunction());
+  pipelineDescriptor->setMeshFunction(createInfo.meshShader.cast<MeshShader>()->getFunction());
   if (createInfo.taskShader != nullptr) {
     pipelineDescriptor->setObjectFunction(createInfo.taskShader.cast<TaskShader>()->getFunction());
   }
@@ -154,14 +154,12 @@ PGraphicsPipeline PipelineCache::createPipeline(Gfx::MeshPipelineCreateInfo crea
   if (graphicsPipelines.contains(hash)) {
     return graphicsPipelines[hash];
   }
-
-  graphics->getDevice()->newRenderPipelineState(
-      pipelineDescriptor, MTL::PipelineOptionNone,
-      [&](MTL::RenderPipelineState* state, MTL::RenderPipelineReflection*, NS::Error* error) {
-        assert(!error);
-        graphicsPipelines[hash] =
-            new GraphicsPipeline(graphics, MTL::PrimitiveTypeLine, state, std::move(createInfo.pipelineLayout));
-      });
+  NS::Error* error = nullptr;
+  MTL::AutoreleasedRenderPipelineReflection reflection;
+  graphicsPipelines[hash] = new GraphicsPipeline(
+      graphics, MTL::PrimitiveTypeTriangle,
+      graphics->getDevice()->newRenderPipelineState(pipelineDescriptor, MTL::PipelineOptionNone, &reflection, &error),
+      std::move(createInfo.pipelineLayout));
 
   pipelineDescriptor->release();
   return graphicsPipelines[hash];

src/Engine/Graphics/Metal/Shader.h

@@ -37,8 +37,8 @@ public:
 using VertexShader = ShaderBase<Gfx::VertexShader, Gfx::SE_SHADER_STAGE_VERTEX_BIT>;
 using FragmentShader = ShaderBase<Gfx::FragmentShader, Gfx::SE_SHADER_STAGE_FRAGMENT_BIT>;
 using ComputeShader = ShaderBase<Gfx::ComputeShader, Gfx::SE_SHADER_STAGE_COMPUTE_BIT>;
-using TaskShader = ShaderBase<Gfx::TaskShader, Gfx::SE_SHADER_STAGE_TASK_BIT_NV>;
-using MeshShader = ShaderBase<Gfx::MeshShader, Gfx::SE_SHADER_STAGE_MESH_BIT_NV>;
+using TaskShader = ShaderBase<Gfx::TaskShader, Gfx::SE_SHADER_STAGE_TASK_BIT_EXT>;
+using MeshShader = ShaderBase<Gfx::MeshShader, Gfx::SE_SHADER_STAGE_MESH_BIT_EXT>;
 
 DEFINE_REF(VertexShader)
 DEFINE_REF(FragmentShader)

src/Engine/Graphics/Metal/Shader.mm

@@ -8,9 +8,8 @@
 #include "Metal/MTLLibrary.hpp"
 #include <fstream>
 #include <iostream>
+#include <metal_irconverter/metal_irconverter.h>
 #include <slang.h>
-#include <spirv_cross.hpp>
-#include <spirv_cross/spirv_msl.hpp>
 
 using namespace Seele;
 using namespace Seele::Metal;
@@ -25,33 +24,58 @@ Shader::~Shader() {
 }
 
 void Shader::create(const ShaderCreateInfo& createInfo) {
-  Slang::ComPtr<slang::IBlob> kernelBlob = generateShader(createInfo, SLANG_SPIRV);
-  hash = CRC::Calculate(kernelBlob->getBufferPointer(), kernelBlob->getBufferSize(), CRC::CRC_32());
-  spirv_cross::CompilerMSL comp((const uint32*)kernelBlob->getBufferPointer(), kernelBlob->getBufferSize() / 4);
-  auto options = comp.get_msl_options();
-  options.argument_buffers = true;
-  options.argument_buffers_tier = spirv_cross::CompilerMSL::Options::ArgumentBuffersTier::Tier2;
-  options.set_msl_version(3);
-  comp.set_msl_options(options);
-  std::string metalCode = comp.compile();
-  NS::Error* error = nullptr;
-  MTL::CompileOptions* mtlOptions = MTL::CompileOptions::alloc()->init();
+  Slang::ComPtr<slang::IBlob> kernelBlob = generateShader(createInfo, SLANG_DXIL);
 
-  library = graphics->getDevice()->newLibrary(NS::String::string(metalCode.c_str(), NS::ASCIIStringEncoding),
-                                              mtlOptions, &error);
-  if (error) {
-    std::cout << error->debugDescription() << std::endl;
-    assert(false);
+  hash = CRC::Calculate(kernelBlob->getBufferPointer(), kernelBlob->getBufferSize(), CRC::CRC_32());
+  IRCompiler* pCompiler = IRCompilerCreate();
+  IRCompilerSetEntryPointName(pCompiler, "main");
+
+  IRObject* pDXIL = IRObjectCreateFromDXIL((const uint8*)kernelBlob->getBufferPointer(), kernelBlob->getBufferSize(),
+                                           IRBytecodeOwnershipNone);
+
+  // Compile DXIL to Metal IR:
+  IRError* pError = nullptr;
+  IRObject* pOutIR = IRCompilerAllocCompileAndLink(pCompiler, NULL, pDXIL, &pError);
+
+  if (!pOutIR) {
+    // Inspect pError to determine cause.
+    IRErrorDestroy(pError);
   }
-  function = library->newFunction(NS::String::string("main0", NS::ASCIIStringEncoding));
+  IRShaderStage irStage;
+  switch (stage) {
+  case Gfx::SE_SHADER_STAGE_VERTEX_BIT:
+    irStage = IRShaderStageVertex;
+    break;
+  case Gfx::SE_SHADER_STAGE_FRAGMENT_BIT:
+    irStage = IRShaderStageFragment;
+    break;
+  case Gfx::SE_SHADER_STAGE_COMPUTE_BIT:
+    irStage = IRShaderStageCompute;
+    break;
+  case Gfx::SE_SHADER_STAGE_TASK_BIT_EXT:
+    irStage = IRShaderStageAmplification;
+    break;
+  case Gfx::SE_SHADER_STAGE_MESH_BIT_EXT:
+    irStage = IRShaderStageMesh;
+    break;
+  }
+  // Retrieve Metallib:
+  IRMetalLibBinary* pMetallib = IRMetalLibBinaryCreate();
+  IRObjectGetMetalLibBinary(pOutIR, irStage, pMetallib);
+  dispatch_data_t data = IRMetalLibGetBytecodeData(pMetallib);
+
+  // Store the metallib to custom format or disk, or use to create a MTLLibrary.
+  NS::Error* error;
+  library = graphics->getDevice()->newLibrary(data, &error);
+  function = library->newFunction(NS::String::string("main", NS::ASCIIStringEncoding));
   if(!function)
   {
-    std::ofstream shaderFile("error.metal");
-    shaderFile << metalCode;
-    shaderFile.close();
-    assert(!function);
+    assert(false);
   }
-  mtlOptions->release();
+  IRMetalLibBinaryDestroy(pMetallib);
+  IRObjectDestroy(pDXIL);
+  IRObjectDestroy(pOutIR);
+  IRCompilerDestroy(pCompiler);
 }
 
 uint32 Shader::getShaderHash() const { return hash; }

src/Engine/Graphics/VertexData.h

@@ -99,7 +99,7 @@ protected:
     Map<MeshId, uint64> meshOffsets;
     Map<MeshId, uint64> meshVertexCounts;
     Array<MeshletDescription> meshlets;
-    Array<uint8> primitiveIndices;
+    Array<uint32> primitiveIndices;
     Array<uint32> vertexIndices;
     Array<uint32> indices;
     Gfx::PGraphics graphics;