Reimplement Vulkan shader objects. (#1764)

* Reimplement Vulkan shader objects.

This change reimplements Vulkan shader objects in the `gfx` layer so that it is no longer layered on top of the `DescriptorSet` abstraction. Since this is the last implementation that uses `DescriptorSet`, the change also removes all `DescriptorSet` related API from public `gfx` interface.

The Vulkan implementation now passes all test cases, but it still have two issues:
1. The PushConstant setting is not correct, this is because we don't seem to be able to get correct reflection data about the size of push constants for an entry-point.
2. The `shader-toy` example can't run on Vulkan, because it currently sets nullptr to `Texture` bindings, and this change doesn't properly handle setting resource to null in `ShaderObject`s yet. If we can use the `nullDescriptor` feature on vulkan, this implementation will be simple. However we still want to decide whether we want to use a Vulkan 1.2 feature for this.

* Fix up

1 files changed, 15 insertions, 47 deletions

diff --git a/tools/gfx/cuda/render-cuda.cpp b/tools/gfx/cuda/render-cuda.cpp
index dbdc27628..383ccc924 100644
--- a/tools/gfx/cuda/render-cuda.cpp
+++ b/tools/gfx/cuda/render-cuda.cpp
@@ -12,7 +12,6 @@
 #include "slang-com-helper.h"
 #include "../command-writer.h"
 #include "../renderer-shared.h"
-#include "../render-graphics-common.h"
 #include "../slang-context.h"
 
 #   ifdef RENDER_TEST_OPTIX
@@ -1038,14 +1037,6 @@ public:
                 m_writer->bindRootShaderObject(PipelineType::Compute, object);
             }
 
-            virtual SLANG_NO_THROW void SLANG_MCALL setDescriptorSet(
-                IPipelineLayout* layout,
-                UInt index,
-                IDescriptorSet* descriptorSet) override
-            {
-                m_writer->setDescriptorSet(PipelineType::Compute, layout, index, descriptorSet);
-            }
-
             virtual SLANG_NO_THROW void SLANG_MCALL dispatchCompute(int x, int y, int z) override
             {
                 m_writer->dispatchCompute(x, y, z);
@@ -1861,27 +1852,29 @@ public:
         // If this is a specializable program, we just keep a reference to the slang program and
         // don't actually create any kernels. This program will be specialized later when we know
         // the shader object bindings.
-        if (desc.slangProgram && desc.slangProgram->getSpecializationParamCount() != 0)
+        RefPtr<CUDAShaderProgram> cudaProgram = new CUDAShaderProgram();
+        cudaProgram->slangProgram = desc.slangProgram;
+        if (desc.slangProgram->getSpecializationParamCount() != 0)
         {
-            RefPtr<CUDAShaderProgram> cudaProgram = new CUDAShaderProgram();
-            cudaProgram->slangProgram = desc.slangProgram;
             cudaProgram->layout = new CUDAProgramLayout(this, desc.slangProgram->getLayout());
             *outProgram = cudaProgram.detach();
             return SLANG_OK;
         }
 
-        if( desc.kernelCount == 0 )
+        ComPtr<ISlangBlob> kernelCode;
+        ComPtr<ISlangBlob> diagnostics;
+        auto compileResult = desc.slangProgram->getEntryPointCode(
+            (SlangInt)0, 0, kernelCode.writeRef(), diagnostics.writeRef());
+        if (diagnostics)
         {
-            return createProgramFromSlang(this, desc, outProgram);
+            // TODO: report compile error.
         }
-
-        if (desc.kernelCount != 1)
-            return SLANG_E_INVALID_ARG;
-        RefPtr<CUDAShaderProgram> cudaProgram = new CUDAShaderProgram();
-        SLANG_CUDA_RETURN_ON_FAIL(cuModuleLoadData(&cudaProgram->cudaModule, desc.kernels[0].codeBegin));
-        SLANG_CUDA_RETURN_ON_FAIL(
-            cuModuleGetFunction(&cudaProgram->cudaKernel, cudaProgram->cudaModule, desc.kernels[0].entryPointName));
-        cudaProgram->kernelName = desc.kernels[0].entryPointName;
+        SLANG_RETURN_ON_FAIL(compileResult);
+        
+        SLANG_CUDA_RETURN_ON_FAIL(cuModuleLoadData(&cudaProgram->cudaModule, kernelCode->getBufferPointer()));
+        cudaProgram->kernelName = desc.slangProgram->getLayout()->getEntryPointByIndex(0)->getName();
+        SLANG_CUDA_RETURN_ON_FAIL(cuModuleGetFunction(
+            &cudaProgram->cudaKernel, cudaProgram->cudaModule, cudaProgram->kernelName.getBuffer()));
 
         auto slangProgram = desc.slangProgram;
         if( slangProgram )
@@ -1985,31 +1978,6 @@ public:
         return SLANG_E_NOT_AVAILABLE;
     }
 
-    virtual SLANG_NO_THROW Result SLANG_MCALL createDescriptorSetLayout(
-        const IDescriptorSetLayout::Desc& desc, IDescriptorSetLayout** outLayout) override
-    {
-        SLANG_UNUSED(desc);
-        SLANG_UNUSED(outLayout);
-        return SLANG_E_NOT_AVAILABLE;
-    }
-
-    virtual SLANG_NO_THROW Result SLANG_MCALL
-        createPipelineLayout(const IPipelineLayout::Desc& desc, IPipelineLayout** outLayout) override
-    {
-        SLANG_UNUSED(desc);
-        SLANG_UNUSED(outLayout);
-        return SLANG_E_NOT_AVAILABLE;
-    }
-
-    virtual SLANG_NO_THROW Result SLANG_MCALL
-        createDescriptorSet(IDescriptorSetLayout* layout, IDescriptorSet::Flag::Enum flags, IDescriptorSet** outDescriptorSet) override
-    {
-        SLANG_UNUSED(layout);
-        SLANG_UNUSED(flags);
-        SLANG_UNUSED(outDescriptorSet);
-        return SLANG_E_NOT_AVAILABLE;
-    }
-
     virtual SLANG_NO_THROW Result SLANG_MCALL createGraphicsPipelineState(
         const GraphicsPipelineStateDesc& desc, IPipelineState** outState) override
     {