Various fixes to gfx. (#2120)

* Various fixes to gfx. * Fix. * Fixes. * Fix. Co-authored-by: Yong He <yhe@nvidia.com>
author: Yong He <yonghe@outlook.com> 2022-02-09 15:30:38 -0800
committer: GitHub <noreply@github.com> 2022-02-09 15:30:38 -0800
commit: b8982fcf43b86c1e39dcc3dd19bff2821633eda6 (patch)
tree: 0d66dbf46b50e760cce4aee232bd6a020976e6fb
parent: 59f3fdc0a372d19ce4e989514ee3e9ecbcbf234c (diff)
43 files changed, 776 insertions, 227 deletions
diff --git a/build/visual-studio/slang-unit-test-tool/slang-unit-test-tool.vcxproj b/build/visual-studio/slang-unit-test-tool/slang-unit-test-tool.vcxproj
index 74126e92c..fd24865a4 100644
--- a/build/visual-studio/slang-unit-test-tool/slang-unit-test-tool.vcxproj
+++ b/build/visual-studio/slang-unit-test-tool/slang-unit-test-tool.vcxproj
@@ -273,6 +273,7 @@
   <ItemGroup>
     <ClCompile Include="..\..\..\tools\slang-unit-test\unit-offset-container.cpp" />
     <ClCompile Include="..\..\..\tools\slang-unit-test\unit-test-byte-encode.cpp" />
+    <ClCompile Include="..\..\..\tools\slang-unit-test\unit-test-chunked-list.cpp" />
     <ClCompile Include="..\..\..\tools\slang-unit-test\unit-test-command-line-args.cpp" />
     <ClCompile Include="..\..\..\tools\slang-unit-test\unit-test-compression.cpp" />
     <ClCompile Include="..\..\..\tools\slang-unit-test\unit-test-find-type-by-name.cpp" />
diff --git a/build/visual-studio/slang-unit-test-tool/slang-unit-test-tool.vcxproj.filters b/build/visual-studio/slang-unit-test-tool/slang-unit-test-tool.vcxproj.filters
index 9db9bee4c..7e06cf50c 100644
--- a/build/visual-studio/slang-unit-test-tool/slang-unit-test-tool.vcxproj.filters
+++ b/build/visual-studio/slang-unit-test-tool/slang-unit-test-tool.vcxproj.filters
@@ -20,6 +20,9 @@
     <ClCompile Include="..\..\..\tools\slang-unit-test\unit-test-byte-encode.cpp">
       <Filter>Source Files</Filter>
     </ClCompile>
+    <ClCompile Include="..\..\..\tools\slang-unit-test\unit-test-chunked-list.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
     <ClCompile Include="..\..\..\tools\slang-unit-test\unit-test-command-line-args.cpp">
       <Filter>Source Files</Filter>
     </ClCompile>
diff --git a/examples/gpu-printing/main.cpp b/examples/gpu-printing/main.cpp
index 31313830f..d8f680376 100644
--- a/examples/gpu-printing/main.cpp
+++ b/examples/gpu-printing/main.cpp
@@ -73,7 +73,7 @@ ComPtr<gfx::IShaderProgram> loadComputeProgram(slang::IModule* slangModule, char
     gGPUPrinting.loadStrings(linkedProgram->getLayout());
 
     gfx::IShaderProgram::Desc programDesc = {};
-    programDesc.slangProgram = linkedProgram;
+    programDesc.slangGlobalScope = linkedProgram;
 
     auto shaderProgram = gDevice->createProgram(programDesc);
 
@@ -113,7 +113,7 @@ Result execute()
 
     IResourceView::Desc printBufferViewDesc;
     printBufferViewDesc.type = IResourceView::Type::UnorderedAccess;
-    auto printBufferView = gDevice->createBufferView(printBuffer, printBufferViewDesc);
+    auto printBufferView = gDevice->createBufferView(printBuffer, nullptr, printBufferViewDesc);
 
     ITransientResourceHeap::Desc transientResourceHeapDesc = {};
     transientResourceHeapDesc.constantBufferSize = 256;
diff --git a/examples/model-viewer/main.cpp b/examples/model-viewer/main.cpp
index dcc1a975f..3b278fede 100644
--- a/examples/model-viewer/main.cpp
+++ b/examples/model-viewer/main.cpp
@@ -108,7 +108,7 @@ struct RendererContext
         // We can create a `gfx::IShaderProgram` object from `composedProgram`
         // so it may be used by the graphics layer.
         gfx::IShaderProgram::Desc programDesc = {};
-        programDesc.slangProgram = composedProgram.get();
+        programDesc.slangGlobalScope = composedProgram.get();
 
         shaderProgram = device->createProgram(programDesc);
 
diff --git a/examples/ray-tracing-pipeline/main.cpp b/examples/ray-tracing-pipeline/main.cpp
index 9b01d251c..9e00ccf67 100644
--- a/examples/ray-tracing-pipeline/main.cpp
+++ b/examples/ray-tracing-pipeline/main.cpp
@@ -201,7 +201,7 @@ gfx::Result loadShaderProgram(
     SLANG_RETURN_ON_FAIL(result);
 
     gfx::IShaderProgram::Desc programDesc = {};
-    programDesc.slangProgram = linkedProgram;
+    programDesc.slangGlobalScope = linkedProgram;
     SLANG_RETURN_ON_FAIL(device->createProgram(programDesc, outProgram));
 
     return SLANG_OK;
@@ -327,7 +327,7 @@ Slang::Result initialize()
     primitiveSRVDesc.format = Format::Unknown;
     primitiveSRVDesc.type = IResourceView::Type::ShaderResource;
     primitiveSRVDesc.bufferElementSize = sizeof(Primitive);
-    gPrimitiveBufferSRV = gDevice->createBufferView(gPrimitiveBuffer, primitiveSRVDesc);
+    gPrimitiveBufferSRV = gDevice->createBufferView(gPrimitiveBuffer, nullptr, primitiveSRVDesc);
 
     IBufferResource::Desc transformBufferDesc;
     transformBufferDesc.type = IResource::Type::Buffer;
diff --git a/examples/ray-tracing/main.cpp b/examples/ray-tracing/main.cpp
index 9d6a638d4..82ad01c95 100644
--- a/examples/ray-tracing/main.cpp
+++ b/examples/ray-tracing/main.cpp
@@ -193,7 +193,7 @@ gfx::Result loadShaderProgram(
     SLANG_RETURN_ON_FAIL(result);
 
     gfx::IShaderProgram::Desc programDesc = {};
-    programDesc.slangProgram = linkedProgram;
+    programDesc.slangGlobalScope = linkedProgram;
     SLANG_RETURN_ON_FAIL(device->createProgram(programDesc, outProgram));
 
     return SLANG_OK;
@@ -318,7 +318,7 @@ Slang::Result initialize()
     primitiveSRVDesc.format = Format::Unknown;
     primitiveSRVDesc.type = IResourceView::Type::ShaderResource;
     primitiveSRVDesc.bufferElementSize = sizeof(Primitive);
-    gPrimitiveBufferSRV = gDevice->createBufferView(gPrimitiveBuffer, primitiveSRVDesc);
+    gPrimitiveBufferSRV = gDevice->createBufferView(gPrimitiveBuffer, nullptr, primitiveSRVDesc);
 
     IBufferResource::Desc transformBufferDesc;
     transformBufferDesc.type = IResource::Type::Buffer;
diff --git a/examples/shader-object/main.cpp b/examples/shader-object/main.cpp
index 58c6abbfd..7c72194d3 100644
--- a/examples/shader-object/main.cpp
+++ b/examples/shader-object/main.cpp
@@ -120,7 +120,7 @@ Result loadShaderProgram(
     // We can create a `gfx::IShaderProgram` object from `composedProgram`
     // so it may be used by the graphics layer.
     gfx::IShaderProgram::Desc programDesc = {};
-    programDesc.slangProgram = composedProgram.get();
+    programDesc.slangGlobalScope = composedProgram.get();
 
     auto shaderProgram = device->createProgram(programDesc);
 
@@ -183,7 +183,8 @@ int main()
     gfx::IResourceView::Desc viewDesc = {};
     viewDesc.type = gfx::IResourceView::Type::UnorderedAccess;
     viewDesc.format = gfx::Format::Unknown;
-    SLANG_RETURN_ON_FAIL(device->createBufferView(numbersBuffer, viewDesc, bufferView.writeRef()));
+    SLANG_RETURN_ON_FAIL(
+        device->createBufferView(numbersBuffer, nullptr, viewDesc, bufferView.writeRef()));
 
     // We have done all the set up work, now it is time to start recording a command buffer for
     // GPU execution.
diff --git a/examples/shader-toy/main.cpp b/examples/shader-toy/main.cpp
index 879330aca..ec161aee8 100644
--- a/examples/shader-toy/main.cpp
+++ b/examples/shader-toy/main.cpp
@@ -263,7 +263,7 @@ Result loadShaderProgram(gfx::IDevice* device, ComPtr<gfx::IShaderProgram>& outS
     SLANG_RETURN_ON_FAIL(result);
 
     gfx::IShaderProgram::Desc programDesc = {};
-    programDesc.slangProgram = linkedProgram.get();
+    programDesc.slangGlobalScope = linkedProgram.get();
     auto shaderProgram = device->createProgram(programDesc);
     outShaderProgram = shaderProgram;
     return SLANG_OK;
diff --git a/examples/triangle/main.cpp b/examples/triangle/main.cpp
index a81a105ab..bd6542b38 100644
--- a/examples/triangle/main.cpp
+++ b/examples/triangle/main.cpp
@@ -181,7 +181,7 @@ gfx::Result loadShaderProgram(
     // program representation.
     //
     gfx::IShaderProgram::Desc programDesc = {};
-    programDesc.slangProgram = linkedProgram;
+    programDesc.slangGlobalScope = linkedProgram;
     SLANG_RETURN_ON_FAIL(device->createProgram(programDesc, outProgram));
 
     return SLANG_OK;
diff --git a/slang-gfx.h b/slang-gfx.h
index 37eb64935..d0e80d923 100644
--- a/slang-gfx.h
+++ b/slang-gfx.h
@@ -110,9 +110,33 @@ class ITransientResourceHeap;
 class IShaderProgram: public ISlangUnknown
 {
 public:
+    // Defines how linking should be performed for a shader program.
+    enum class LinkingStyle
+    {
+        // Compose all entry-points in a single program, then compile all entry-points together with the same
+        // set of root shader arguments.
+        SingleProgram,
+
+        // Link and compile each entry-point individually, potentially with different specializations.
+        SeparateEntryPointCompilation
+    };
+
     struct Desc
     {
-        slang::IComponentType*  slangProgram;
+        // The linking style of this program.
+        LinkingStyle linkingStyle = LinkingStyle::SingleProgram;
+
+        // The global scope or a Slang composite component that represents the entire program.
+        slang::IComponentType*  slangGlobalScope;
+
+        // Number of separate entry point components in the `slangEntryPoints` array to link in.
+        // If set to 0, then `slangGlobalScope` must contain Slang EntryPoint components.
+        // If not 0, then `slangGlobalScope` must not contain any EntryPoint components.
+        uint32_t entryPointCount = 0;
+
+        // An array of Slang entry points. The size of the array must be `entryPointCount`.
+        // Each element must define only 1 Slang EntryPoint.
+        slang::IComponentType** slangEntryPoints = nullptr;
     };
 };
 #define SLANG_UUID_IShaderProgram                                                       \
@@ -558,8 +582,8 @@ struct ClearValue
 
 struct BufferRange
 {
-    uint32_t firstElement;
-    uint32_t elementCount;
+    uint64_t firstElement;
+    uint64_t elementCount;
 };
 
 enum class TextureAspect : uint32_t
@@ -1278,11 +1302,12 @@ struct HitGroupDesc
 struct RayTracingPipelineStateDesc
 {
     IShaderProgram* program = nullptr;
-    int32_t hitGroupCount;
-    const HitGroupDesc* hitGroups;
-    int maxRecursion;
-    int maxRayPayloadSize;
-    RayTracingPipelineFlags::Enum flags;
+    int32_t hitGroupCount = 0;
+    const HitGroupDesc* hitGroups = nullptr;
+    int maxRecursion = 0;
+    int maxRayPayloadSize = 0;
+    int maxAttributeSizeInBytes = 8;
+    RayTracingPipelineFlags::Enum flags = RayTracingPipelineFlags::None;
 };
 
 class IShaderTable : public ISlangUnknown
@@ -2104,12 +2129,16 @@ public:
     }
 
     virtual SLANG_NO_THROW Result SLANG_MCALL createBufferView(
-        IBufferResource* buffer, IResourceView::Desc const& desc, IResourceView** outView) = 0;
+        IBufferResource* buffer,
+        IBufferResource* counterBuffer,
+        IResourceView::Desc const& desc,
+        IResourceView** outView) = 0;
 
-    inline ComPtr<IResourceView> createBufferView(IBufferResource* buffer, IResourceView::Desc const& desc)
+    inline ComPtr<IResourceView> createBufferView(
+        IBufferResource* buffer, IBufferResource* counterBuffer, IResourceView::Desc const& desc)
     {
         ComPtr<IResourceView> view;
-        SLANG_RETURN_NULL_ON_FAIL(createBufferView(buffer, desc, view.writeRef()));
+        SLANG_RETURN_NULL_ON_FAIL(createBufferView(buffer, counterBuffer, desc, view.writeRef()));
         return view;
     }
 
diff --git a/slang.h b/slang.h
index 4d9effad3..cdc0daeab 100644
--- a/slang.h
+++ b/slang.h
@@ -2114,6 +2114,9 @@ extern "C"
     SLANG_API char const* spReflectionEntryPoint_getName(
         SlangReflectionEntryPoint* entryPoint);
 
+    SLANG_API char const* spReflectionEntryPoint_getNameOverride(
+        SlangReflectionEntryPoint* entryPoint);
+
     SLANG_API unsigned spReflectionEntryPoint_getParameterCount(
         SlangReflectionEntryPoint* entryPoint);
 
@@ -2880,6 +2883,11 @@ namespace slang
             return spReflectionEntryPoint_getName((SlangReflectionEntryPoint*) this);
         }
 
+        char const* getNameOverride()
+        {
+            return spReflectionEntryPoint_getNameOverride((SlangReflectionEntryPoint*)this);
+        }
+
         unsigned getParameterCount()
         {
             return spReflectionEntryPoint_getParameterCount((SlangReflectionEntryPoint*) this);
diff --git a/slang.sln b/slang.sln
index 8b7bf1cf5..b4b30eeee 100644
--- a/slang.sln
+++ b/slang.sln
@@ -35,10 +35,6 @@ Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "gpu-printing", "build\visua
 EndProject
 Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "hello-world", "build\visual-studio\hello-world\hello-world.vcxproj", "{010BE414-ED5B-CF56-16C0-BD18027062C0}"
 EndProject
-Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "heterogeneous-first-gen", "build\visual-studio\heterogeneous-hello-world\heterogeneous-first-gen.vcxproj", "{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}"
-EndProject
-Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "heterogeneous-hello-world", "build\visual-studio\heterogeneous-hello-world\heterogeneous-hello-world.vcxproj", "{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}"
-EndProject
 Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "model-viewer", "build\visual-studio\model-viewer\model-viewer.vcxproj", "{2F8724C6-1BC3-2730-84D5-3F277030D04A}"
 EndProject
 Project("{8BC9CEB8-8B4A-11D0-8D11-00A0C91BC942}") = "ray-tracing", "build\visual-studio\ray-tracing\ray-tracing.vcxproj", "{71AC0F50-5DFD-FA91-8661-E95372118EFB}"
@@ -271,30 +267,6 @@ Global
 		{010BE414-ED5B-CF56-16C0-BD18027062C0}.Release|Win32.Build.0 = Release|Win32
 		{010BE414-ED5B-CF56-16C0-BD18027062C0}.Release|x64.ActiveCfg = Release|x64
 		{010BE414-ED5B-CF56-16C0-BD18027062C0}.Release|x64.Build.0 = Release|x64
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Debug|aarch64.ActiveCfg = Debug aarch64|ARM
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Debug|aarch64.Build.0 = Debug aarch64|ARM
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Debug|Win32.ActiveCfg = Debug|Win32
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Debug|Win32.Build.0 = Debug|Win32
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Debug|x64.ActiveCfg = Debug|x64
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Debug|x64.Build.0 = Debug|x64
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Release|aarch64.ActiveCfg = Release aarch64|ARM
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Release|aarch64.Build.0 = Release aarch64|ARM
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Release|Win32.ActiveCfg = Release|Win32
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Release|Win32.Build.0 = Release|Win32
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Release|x64.ActiveCfg = Release|x64
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B}.Release|x64.Build.0 = Release|x64
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Debug|aarch64.ActiveCfg = Debug aarch64|ARM
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Debug|aarch64.Build.0 = Debug aarch64|ARM
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Debug|Win32.ActiveCfg = Debug|Win32
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Debug|Win32.Build.0 = Debug|Win32
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Debug|x64.ActiveCfg = Debug|x64
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Debug|x64.Build.0 = Debug|x64
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Release|aarch64.ActiveCfg = Release aarch64|ARM
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Release|aarch64.Build.0 = Release aarch64|ARM
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Release|Win32.ActiveCfg = Release|Win32
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Release|Win32.Build.0 = Release|Win32
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Release|x64.ActiveCfg = Release|x64
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49}.Release|x64.Build.0 = Release|x64
 		{2F8724C6-1BC3-2730-84D5-3F277030D04A}.Debug|aarch64.ActiveCfg = Debug aarch64|ARM
 		{2F8724C6-1BC3-2730-84D5-3F277030D04A}.Debug|aarch64.Build.0 = Debug aarch64|ARM
 		{2F8724C6-1BC3-2730-84D5-3F277030D04A}.Debug|Win32.ActiveCfg = Debug|Win32
@@ -493,8 +465,6 @@ Global
 		{37BED5B5-23FA-D81F-8C0C-F1167867813A} = {EB5FC2C6-D72D-B6CC-C0C1-26F3AC2E9231}
 		{57C81DD3-4304-213D-AC16-39349871C957} = {EB5FC2C6-D72D-B6CC-C0C1-26F3AC2E9231}
 		{010BE414-ED5B-CF56-16C0-BD18027062C0} = {EB5FC2C6-D72D-B6CC-C0C1-26F3AC2E9231}
-		{3E8C6F8B-AAC3-B722-B3CD-1B461F84DE1B} = {EB5FC2C6-D72D-B6CC-C0C1-26F3AC2E9231}
-		{150CAA5A-0177-6A66-AA92-CFCB96DC2D49} = {EB5FC2C6-D72D-B6CC-C0C1-26F3AC2E9231}
 		{2F8724C6-1BC3-2730-84D5-3F277030D04A} = {EB5FC2C6-D72D-B6CC-C0C1-26F3AC2E9231}
 		{71AC0F50-5DFD-FA91-8661-E95372118EFB} = {EB5FC2C6-D72D-B6CC-C0C1-26F3AC2E9231}
 		{17BA8E32-034E-84DA-6C12-DE8E58C5BECC} = {EB5FC2C6-D72D-B6CC-C0C1-26F3AC2E9231}
diff --git a/source/core/slang-chunked-list.h b/source/core/slang-chunked-list.h
index fced9834a..36f97cedd 100644
--- a/source/core/slang-chunked-list.h
+++ b/source/core/slang-chunked-list.h
@@ -243,6 +243,7 @@ private:
         {
             auto nextChunk = chunk->next;
             freeChunk(chunk);
+            chunk = nextChunk;
         }
         m_firstChunk.next = 0;
         m_firstChunk.size = 0;
diff --git a/source/slang/slang-parameter-binding.cpp b/source/slang/slang-parameter-binding.cpp
index 48e637047..abbc010f8 100644
--- a/source/slang/slang-parameter-binding.cpp
+++ b/source/slang/slang-parameter-binding.cpp
@@ -2477,6 +2477,7 @@ static void removePerEntryPointParameterKinds(
 static RefPtr<EntryPointLayout> collectEntryPointParameters(
     ParameterBindingContext*                    context,
     EntryPoint*                                 entryPoint,
+    String                                      entryPointNameOverride,
     EntryPoint::EntryPointSpecializationInfo*   specializationInfo)
 {
     auto astBuilder = context->getASTBuilder();
@@ -2487,6 +2488,7 @@ static RefPtr<EntryPointLayout> collectEntryPointParameters(
     RefPtr<EntryPointLayout> entryPointLayout = new EntryPointLayout();
     entryPointLayout->profile = entryPoint->getProfile();
     entryPointLayout->name = entryPoint->getName();
+    entryPointLayout->nameOverride = entryPointNameOverride;
 
     // The entry point layout must be added to the output
     // program layout so that it can be accessed by reflection.
@@ -2836,6 +2838,7 @@ struct CollectParametersVisitor : ComponentTypeVisitor
     {}
 
     ParameterBindingContext* m_context;
+    String m_currentEntryPointNameOverride;
 
     void visitComposite(CompositeComponentType* composite, CompositeComponentType::CompositeSpecializationInfo* specializationInfo) SLANG_OVERRIDE
     {
@@ -2892,15 +2895,18 @@ struct CollectParametersVisitor : ComponentTypeVisitor
         ParameterBindingContext contextData = *m_context;
         auto context = &contextData;
         context->stage = entryPoint->getStage();
-
-        collectEntryPointParameters(context, entryPoint, specializationInfo);
+        collectEntryPointParameters(
+            context, entryPoint, m_currentEntryPointNameOverride, specializationInfo);
     }
 
     void visitRenamedEntryPoint(
         RenamedEntryPointComponentType* renamedEntryPoint,
         EntryPoint::EntryPointSpecializationInfo* specializationInfo) SLANG_OVERRIDE
     {
+        auto lastNameOverride = m_currentEntryPointNameOverride;
+        m_currentEntryPointNameOverride = renamedEntryPoint->getEntryPointNameOverride(0);
         renamedEntryPoint->getBase()->acceptVisitor(this, specializationInfo);
+        m_currentEntryPointNameOverride = lastNameOverride;
     }
 
     void visitModule(Module* module, Module::ModuleSpecializationInfo* specializationInfo) SLANG_OVERRIDE
diff --git a/source/slang/slang-reflection-api.cpp b/source/slang/slang-reflection-api.cpp
index 17ceb6842..714dc292f 100644
--- a/source/slang/slang-reflection-api.cpp
+++ b/source/slang/slang-reflection-api.cpp
@@ -728,6 +728,8 @@ SLANG_API SlangReflectionType * spReflection_FindTypeByName(SlangReflection * re
     try
     {
         Type* result = program->getTypeFromString(name, &sink);
+        if (as<ErrorType>(result))
+            return nullptr;
         return (SlangReflectionType*)result;
     }
     catch( ... )
@@ -2567,6 +2569,19 @@ SLANG_API char const* spReflectionEntryPoint_getName(
     return entryPointLayout ? getCstr(entryPointLayout->name) : nullptr;
 }
 
+SLANG_API char const* spReflectionEntryPoint_getNameOverride(SlangReflectionEntryPoint* inEntryPoint)
+{
+    auto entryPointLayout = convert(inEntryPoint);
+    if (entryPointLayout)
+    {
+        if (entryPointLayout->nameOverride.getLength())
+            return entryPointLayout->nameOverride.getBuffer();
+        else
+            return getCstr(entryPointLayout->name);
+    }
+    return nullptr;
+}
+
 SLANG_API unsigned spReflectionEntryPoint_getParameterCount(
     SlangReflectionEntryPoint* inEntryPoint)
 {
diff --git a/source/slang/slang-type-layout.h b/source/slang/slang-type-layout.h
index 8c50b4eb3..82d333fa9 100644
--- a/source/slang/slang-type-layout.h
+++ b/source/slang/slang-type-layout.h
@@ -761,6 +761,9 @@ public:
     // The name of the entry point. Always available even if entryPoint is nullptr (for example when it came from a library)
     Name* name = nullptr;
 
+    // The overrided name of the entry point.
+    String nameOverride;
+
     // Layout for any results of the entry point
     RefPtr<VarLayout> resultLayout;
 
diff --git a/tools/gfx-unit-test/buffer-barrier-test.cpp b/tools/gfx-unit-test/buffer-barrier-test.cpp
index fb3d62f8c..b2692f6fd 100644
--- a/tools/gfx-unit-test/buffer-barrier-test.cpp
+++ b/tools/gfx-unit-test/buffer-barrier-test.cpp
@@ -48,7 +48,8 @@ namespace gfx_test
         viewDesc.type = unorderedAccess ? IResourceView::Type::UnorderedAccess : IResourceView::Type::ShaderResource;
         viewDesc.format = Format::Unknown;
         viewDesc.bufferElementSize = sizeof(float);
-        GFX_CHECK_CALL_ABORT(device->createBufferView(outBuffer.buffer, viewDesc, outBuffer.view.writeRef()));
+        GFX_CHECK_CALL_ABORT(device->createBufferView(
+            outBuffer.buffer, nullptr, viewDesc, outBuffer.view.writeRef()));
     }
 
     void barrierTestImpl(IDevice* device, UnitTestContext* context)
diff --git a/tools/gfx-unit-test/compute-smoke.cpp b/tools/gfx-unit-test/compute-smoke.cpp
index 185e8aa99..68925f1e0 100644
--- a/tools/gfx-unit-test/compute-smoke.cpp
+++ b/tools/gfx-unit-test/compute-smoke.cpp
@@ -51,7 +51,8 @@ namespace gfx_test
         IResourceView::Desc viewDesc = {};
         viewDesc.type = IResourceView::Type::UnorderedAccess;
         viewDesc.format = Format::Unknown;
-        GFX_CHECK_CALL_ABORT(device->createBufferView(numbersBuffer, viewDesc, bufferView.writeRef()));
+        GFX_CHECK_CALL_ABORT(
+            device->createBufferView(numbersBuffer, nullptr, viewDesc, bufferView.writeRef()));
 
         // We have done all the set up work, now it is time to start recording a command buffer for
         // GPU execution.
diff --git a/tools/gfx-unit-test/create-buffer-from-handle.cpp b/tools/gfx-unit-test/create-buffer-from-handle.cpp
index 22612dbf0..4e2261465 100644
--- a/tools/gfx-unit-test/create-buffer-from-handle.cpp
+++ b/tools/gfx-unit-test/create-buffer-from-handle.cpp
@@ -57,7 +57,8 @@ namespace gfx_test
         IResourceView::Desc viewDesc = {};
         viewDesc.type = IResourceView::Type::UnorderedAccess;
         viewDesc.format = Format::Unknown;
-        GFX_CHECK_CALL_ABORT(device->createBufferView(numbersBuffer, viewDesc, bufferView.writeRef()));
+        GFX_CHECK_CALL_ABORT(
+            device->createBufferView(numbersBuffer, nullptr, viewDesc, bufferView.writeRef()));
 
         // We have done all the set up work, now it is time to start recording a command buffer for
         // GPU execution.
diff --git a/tools/gfx-unit-test/existing-device-handle-test.cpp b/tools/gfx-unit-test/existing-device-handle-test.cpp
index cf13fe090..d6f66af84 100644
--- a/tools/gfx-unit-test/existing-device-handle-test.cpp
+++ b/tools/gfx-unit-test/existing-device-handle-test.cpp
@@ -51,7 +51,8 @@ namespace gfx_test
         IResourceView::Desc viewDesc = {};
         viewDesc.type = IResourceView::Type::UnorderedAccess;
         viewDesc.format = Format::Unknown;
-        GFX_CHECK_CALL_ABORT(device->createBufferView(numbersBuffer, viewDesc, bufferView.writeRef()));
+        GFX_CHECK_CALL_ABORT(
+            device->createBufferView(numbersBuffer, nullptr, viewDesc, bufferView.writeRef()));
 
         // We have done all the set up work, now it is time to start recording a command buffer for
         // GPU execution.
diff --git a/tools/gfx-unit-test/format-unit-tests.cpp b/tools/gfx-unit-test/format-unit-tests.cpp
index 55aa2768b..1a885174f 100644
--- a/tools/gfx-unit-test/format-unit-tests.cpp
+++ b/tools/gfx-unit-test/format-unit-tests.cpp
@@ -153,7 +153,8 @@ namespace gfx_test
         IResourceView::Desc viewDesc = {};
         viewDesc.type = IResourceView::Type::UnorderedAccess;
         viewDesc.format = Format::Unknown;
-        GFX_CHECK_CALL_ABORT(device->createBufferView(outBuffer, viewDesc, bufferView.writeRef()));
+        GFX_CHECK_CALL_ABORT(
+            device->createBufferView(outBuffer, nullptr, viewDesc, bufferView.writeRef()));
         return bufferView;
     }
 
diff --git a/tools/gfx-unit-test/gfx-test-util.cpp b/tools/gfx-unit-test/gfx-test-util.cpp
index c95c3d0e5..1133e3842 100644
--- a/tools/gfx-unit-test/gfx-test-util.cpp
+++ b/tools/gfx-unit-test/gfx-test-util.cpp
@@ -57,7 +57,7 @@ namespace gfx_test
         slangReflection = composedProgram->getLayout();
 
         gfx::IShaderProgram::Desc programDesc = {};
-        programDesc.slangProgram = composedProgram.get();
+        programDesc.slangGlobalScope = composedProgram.get();
 
         auto shaderProgram = device->createProgram(programDesc);
 
@@ -105,7 +105,7 @@ namespace gfx_test
         slangReflection = composedProgram->getLayout();
 
         gfx::IShaderProgram::Desc programDesc = {};
-        programDesc.slangProgram = composedProgram.get();
+        programDesc.slangGlobalScope = composedProgram.get();
 
         auto shaderProgram = device->createProgram(programDesc);
 
diff --git a/tools/gfx-unit-test/mutable-shader-object.cpp b/tools/gfx-unit-test/mutable-shader-object.cpp
index 28ec8fb45..cc9707643 100644
--- a/tools/gfx-unit-test/mutable-shader-object.cpp
+++ b/tools/gfx-unit-test/mutable-shader-object.cpp
@@ -51,7 +51,8 @@ namespace gfx_test
         IResourceView::Desc viewDesc = {};
         viewDesc.type = IResourceView::Type::UnorderedAccess;
         viewDesc.format = Format::Unknown;
-        GFX_CHECK_CALL_ABORT(device->createBufferView(numbersBuffer, viewDesc, bufferView.writeRef()));
+        GFX_CHECK_CALL_ABORT(
+            device->createBufferView(numbersBuffer, nullptr, viewDesc, bufferView.writeRef()));
 
         {
             slang::TypeReflection* addTransformerType =
diff --git a/tools/gfx-unit-test/nested-parameter-block.cpp b/tools/gfx-unit-test/nested-parameter-block.cpp
index 2917218f1..b21b0bcf2 100644
--- a/tools/gfx-unit-test/nested-parameter-block.cpp
+++ b/tools/gfx-unit-test/nested-parameter-block.cpp
@@ -70,7 +70,7 @@ namespace gfx_test
             srvDesc.bufferElementSize = sizeof(uint32_t) * 4;
             srvDesc.bufferRange.elementCount = 4;
             srvDesc.bufferRange.firstElement = 0;
-            srvs.add(device->createBufferView(srvBuffers[i], srvDesc));
+            srvs.add(device->createBufferView(srvBuffers[i], nullptr, srvDesc));
         }
         Slang::ComPtr<IBufferResource> resultBuffer =
             createBuffer(device, 0, gfx::ResourceState::UnorderedAccess);
@@ -82,7 +82,7 @@ namespace gfx_test
         resultBufferViewDesc.bufferRange.firstElement = 0;
         Slang::ComPtr<IResourceView> resultBufferView;
         SLANG_CHECK(SLANG_SUCCEEDED(device->createBufferView(
-            resultBuffer, resultBufferViewDesc, resultBufferView.writeRef())));
+            resultBuffer, nullptr, resultBufferViewDesc, resultBufferView.writeRef())));
 
         Slang::ComPtr<IShaderObject> materialObject;
         SLANG_CHECK(SLANG_SUCCEEDED(device->createMutableShaderObject(
diff --git a/tools/gfx-unit-test/root-mutable-shader-object.cpp b/tools/gfx-unit-test/root-mutable-shader-object.cpp
index 43c5fbd1f..10079b62a 100644
--- a/tools/gfx-unit-test/root-mutable-shader-object.cpp
+++ b/tools/gfx-unit-test/root-mutable-shader-object.cpp
@@ -51,7 +51,8 @@ namespace gfx_test
         IResourceView::Desc viewDesc = {};
         viewDesc.type = IResourceView::Type::UnorderedAccess;
         viewDesc.format = Format::Unknown;
-        GFX_CHECK_CALL_ABORT(device->createBufferView(numbersBuffer, viewDesc, bufferView.writeRef()));
+        GFX_CHECK_CALL_ABORT(
+            device->createBufferView(numbersBuffer, nullptr, viewDesc, bufferView.writeRef()));
 
         ComPtr<IShaderObject> rootObject;
         device->createMutableRootShaderObject(shaderProgram, rootObject.writeRef());
diff --git a/tools/gfx-unit-test/shared-buffers-tests.cpp b/tools/gfx-unit-test/shared-buffers-tests.cpp
index 915da798b..ccd223fc0 100644
--- a/tools/gfx-unit-test/shared-buffers-tests.cpp
+++ b/tools/gfx-unit-test/shared-buffers-tests.cpp
@@ -70,7 +70,8 @@ namespace gfx_test
         IResourceView::Desc viewDesc = {};
         viewDesc.type = IResourceView::Type::UnorderedAccess;
         viewDesc.format = Format::Unknown;
-        GFX_CHECK_CALL_ABORT(dstDevice->createBufferView(dstBuffer, viewDesc, bufferView.writeRef()));
+        GFX_CHECK_CALL_ABORT(
+            dstDevice->createBufferView(dstBuffer, nullptr, viewDesc, bufferView.writeRef()));
 
         {
             ICommandQueue::Desc queueDesc = { ICommandQueue::QueueType::Graphics };
diff --git a/tools/gfx-unit-test/shared-textures-tests.cpp b/tools/gfx-unit-test/shared-textures-tests.cpp
index b83f58a78..831296c72 100644
--- a/tools/gfx-unit-test/shared-textures-tests.cpp
+++ b/tools/gfx-unit-test/shared-textures-tests.cpp
@@ -130,7 +130,8 @@ namespace gfx_test
         IResourceView::Desc viewDesc = {};
         viewDesc.type = IResourceView::Type::UnorderedAccess;
         viewDesc.format = Format::Unknown;
-        GFX_CHECK_CALL_ABORT(device->createBufferView(outBuffer, viewDesc, bufferView.writeRef()));
+        GFX_CHECK_CALL_ABORT(
+            device->createBufferView(outBuffer, nullptr, viewDesc, bufferView.writeRef()));
         return bufferView;
     }
 
diff --git a/tools/gfx/cpu/render-cpu.cpp b/tools/gfx/cpu/render-cpu.cpp
index 38b72e28b..a20ed3ebe 100644
--- a/tools/gfx/cpu/render-cpu.cpp
+++ b/tools/gfx/cpu/render-cpu.cpp
@@ -1091,7 +1091,7 @@ private:
 
         ComPtr<ISlangSharedLibrary> sharedLibrary;
         ComPtr<ISlangBlob> diagnostics;
-        auto compileResult = program->slangProgram->getEntryPointHostCallable(
+        auto compileResult = program->slangGlobalScope->getEntryPointHostCallable(
             entryPointIndex, targetIndex, sharedLibrary.writeRef(), diagnostics.writeRef());
         if (diagnostics)
         {
@@ -1205,7 +1205,10 @@ public:
     }
 
     virtual SLANG_NO_THROW Result SLANG_MCALL createBufferView(
-        IBufferResource* inBuffer, IResourceView::Desc const& desc, IResourceView** outView) override
+        IBufferResource* inBuffer,
+        IBufferResource* counterBuffer,
+        IResourceView::Desc const& desc,
+        IResourceView** outView) override
     {
         auto buffer = static_cast<CPUBufferResource*>(inBuffer);
         RefPtr<CPUBufferView> view = new CPUBufferView(desc, buffer);
@@ -1266,15 +1269,11 @@ public:
         ISlangBlob** outDiagnosticBlob) override
     {
         RefPtr<CPUShaderProgram> cpuProgram = new CPUShaderProgram();
-
-        // TODO: stuff?
-
-        auto slangProgram = desc.slangProgram;
-        if( slangProgram )
+        cpuProgram->init(desc);
+        auto slangGlobalScope = cpuProgram->linkedProgram;
+        if( slangGlobalScope )
         {
-            cpuProgram->slangProgram = slangProgram;
-
-            auto slangProgramLayout = slangProgram->getLayout();
+            auto slangProgramLayout = slangGlobalScope->getLayout();
             if(!slangProgramLayout)
                 return SLANG_FAIL;
 
diff --git a/tools/gfx/cuda/render-cuda.cpp b/tools/gfx/cuda/render-cuda.cpp
index cb8f779f4..bacaa42ee 100644
--- a/tools/gfx/cuda/render-cuda.cpp
+++ b/tools/gfx/cuda/render-cuda.cpp
@@ -2112,7 +2112,10 @@ public:
     }
 
     virtual SLANG_NO_THROW Result SLANG_MCALL createBufferView(
-        IBufferResource* buffer, IResourceView::Desc const& desc, IResourceView** outView) override
+        IBufferResource* buffer,
+        IBufferResource* counterBuffer,
+        IResourceView::Desc const& desc,
+        IResourceView** outView) override
     {
         RefPtr<CUDAResourceView> view = new CUDAResourceView();
         view->m_desc = desc;
@@ -2181,18 +2184,18 @@ public:
         // don't actually create any kernels. This program will be specialized later when we know
         // the shader object bindings.
         RefPtr<CUDAShaderProgram> cudaProgram = new CUDAShaderProgram();
-        cudaProgram->slangProgram = desc.slangProgram;
+        cudaProgram->init(desc);
         cudaProgram->cudaContext = m_context;
-        if (desc.slangProgram->getSpecializationParamCount() != 0)
+        if (desc.slangGlobalScope->getSpecializationParamCount() != 0)
         {
-            cudaProgram->layout = new CUDAProgramLayout(this, desc.slangProgram->getLayout());
+            cudaProgram->layout = new CUDAProgramLayout(this, desc.slangGlobalScope->getLayout());
             returnComPtr(outProgram, cudaProgram);
             return SLANG_OK;
         }
 
         ComPtr<ISlangBlob> kernelCode;
         ComPtr<ISlangBlob> diagnostics;
-        auto compileResult = desc.slangProgram->getEntryPointCode(
+        auto compileResult = desc.slangGlobalScope->getEntryPointCode(
             (SlangInt)0, 0, kernelCode.writeRef(), diagnostics.writeRef());
         if (diagnostics)
         {
@@ -2206,16 +2209,16 @@ public:
         SLANG_RETURN_ON_FAIL(compileResult);
         
         SLANG_CUDA_RETURN_ON_FAIL(cuModuleLoadData(&cudaProgram->cudaModule, kernelCode->getBufferPointer()));
-        cudaProgram->kernelName = desc.slangProgram->getLayout()->getEntryPointByIndex(0)->getName();
+        cudaProgram->kernelName = desc.slangGlobalScope->getLayout()->getEntryPointByIndex(0)->getName();
         SLANG_CUDA_RETURN_ON_FAIL(cuModuleGetFunction(
             &cudaProgram->cudaKernel, cudaProgram->cudaModule, cudaProgram->kernelName.getBuffer()));
 
-        auto slangProgram = desc.slangProgram;
-        if( slangProgram )
+        auto slangGlobalScope = desc.slangGlobalScope;
+        if( slangGlobalScope )
         {
-            cudaProgram->slangProgram = slangProgram;
+            cudaProgram->slangGlobalScope = slangGlobalScope;
 
-            auto slangProgramLayout = slangProgram->getLayout();
+            auto slangProgramLayout = slangGlobalScope->getLayout();
             if(!slangProgramLayout)
                 return SLANG_FAIL;
 
diff --git a/tools/gfx/d3d11/render-d3d11.cpp b/tools/gfx/d3d11/render-d3d11.cpp
index 2e3963c30..04399700e 100644
--- a/tools/gfx/d3d11/render-d3d11.cpp
+++ b/tools/gfx/d3d11/render-d3d11.cpp
@@ -93,6 +93,7 @@ public:
 
     virtual SLANG_NO_THROW Result SLANG_MCALL createBufferView(
         IBufferResource* buffer,
+        IBufferResource* counterBuffer,
         IResourceView::Desc const& desc,
         IResourceView** outView) override;
 
@@ -3008,7 +3009,11 @@ Result D3D11Device::createTextureView(ITextureResource* texture, IResourceView::
     }
 }
 
-Result D3D11Device::createBufferView(IBufferResource* buffer, IResourceView::Desc const& desc, IResourceView** outView)
+Result D3D11Device::createBufferView(
+    IBufferResource* buffer,
+    IBufferResource* counterBuffer,
+    IResourceView::Desc const& desc,
+    IResourceView** outView)
 {
     auto resourceImpl = (BufferResourceImpl*) buffer;
     auto resourceDesc = *resourceImpl->getDesc();
@@ -3483,21 +3488,21 @@ void D3D11Device::drawIndexedInstanced(
 Result D3D11Device::createProgram(
     const IShaderProgram::Desc& desc, IShaderProgram** outProgram, ISlangBlob** outDiagnosticBlob)
 {
-    SLANG_ASSERT(desc.slangProgram);
+    SLANG_ASSERT(desc.slangGlobalScope);
 
-    if (desc.slangProgram->getSpecializationParamCount() != 0)
+    if (desc.slangGlobalScope->getSpecializationParamCount() != 0)
     {
         // For a specializable program, we don't invoke any actual slang compilation yet.
         RefPtr<ShaderProgramImpl> shaderProgram = new ShaderProgramImpl();
-        shaderProgram->slangProgram = desc.slangProgram;
+        shaderProgram->init(desc);
         returnComPtr(outProgram, shaderProgram);
         return SLANG_OK;
     }
 
     // If the program is already specialized, compile and create shader kernels now.
     SlangInt targetIndex = 0;
-    auto slangProgram = desc.slangProgram;
-    auto programLayout = slangProgram->getLayout(targetIndex);
+    auto slangGlobalScope = desc.slangGlobalScope;
+    auto programLayout = slangGlobalScope->getLayout(targetIndex);
     if (!programLayout)
         return SLANG_FAIL;
     SlangUInt entryPointCount = programLayout->getEntryPointCount();
@@ -3505,7 +3510,7 @@ Result D3D11Device::createProgram(
         return SLANG_FAIL;
 
     RefPtr<ShaderProgramImpl> shaderProgram = new ShaderProgramImpl();
-    shaderProgram->slangProgram = desc.slangProgram;
+    shaderProgram->slangGlobalScope = desc.slangGlobalScope;
 
     ScopeNVAPI scopeNVAPI;
     SLANG_RETURN_ON_FAIL(scopeNVAPI.init(this, 0));
@@ -3514,7 +3519,7 @@ Result D3D11Device::createProgram(
         ComPtr<ISlangBlob> kernelCode;
         ComPtr<ISlangBlob> diagnostics;
 
-        auto compileResult = slangProgram->getEntryPointCode(
+        auto compileResult = slangGlobalScope->getEntryPointCode(
             (SlangInt)i, 0, kernelCode.writeRef(), diagnostics.writeRef());
 
         if (diagnostics)
@@ -3728,7 +3733,7 @@ Result D3D11Device::createRootShaderObject(IShaderProgram* program, ShaderObject
     RefPtr<RootShaderObjectImpl> shaderObject;
     RefPtr<RootShaderObjectLayoutImpl> rootLayout;
     SLANG_RETURN_ON_FAIL(RootShaderObjectLayoutImpl::create(
-        this, programImpl->slangProgram, programImpl->slangProgram->getLayout(), rootLayout.writeRef()));
+        this, programImpl->slangGlobalScope, programImpl->slangGlobalScope->getLayout(), rootLayout.writeRef()));
     SLANG_RETURN_ON_FAIL(RootShaderObjectImpl::create(
         this, rootLayout.Ptr(), shaderObject.writeRef()));
     returnRefPtrMove(outObject, shaderObject);
diff --git a/tools/gfx/d3d12/descriptor-heap-d3d12.h b/tools/gfx/d3d12/descriptor-heap-d3d12.h
index 26b9c6a16..2718f09e1 100644
--- a/tools/gfx/d3d12/descriptor-heap-d3d12.h
+++ b/tools/gfx/d3d12/descriptor-heap-d3d12.h
@@ -91,8 +91,7 @@ class D3D12GeneralDescriptorHeap : public Slang::RefObject
     Slang::VirtualObjectPool m_allocator;
 
 public:
-    D3D12GeneralDescriptorHeap()
-    {}
+    int getSize() { return m_chunkSize; }
 
     Slang::Result init(ID3D12Device* device, int chunkSize, D3D12_DESCRIPTOR_HEAP_TYPE type, D3D12_DESCRIPTOR_HEAP_FLAGS flag)
     {
@@ -151,16 +150,140 @@ public:
     }
 };
 
+class D3D12GeneralExpandingDescriptorHeap : public Slang::RefObject
+{
+    ID3D12Device* m_device;
+    D3D12_DESCRIPTOR_HEAP_TYPE m_type;
+    D3D12_DESCRIPTOR_HEAP_FLAGS m_flag;
+    int m_chunkSize;
+    Slang::List<Slang::RefPtr<D3D12GeneralDescriptorHeap>> m_subHeaps;
+    Slang::List<int> m_subHeapStartingIndex;
+
+public:
+    Slang::Result newSubHeap()
+    {
+        Slang::RefPtr<D3D12GeneralDescriptorHeap> subHeap = new D3D12GeneralDescriptorHeap();
+        SLANG_RETURN_ON_FAIL(subHeap->init(m_device, m_chunkSize, m_type, m_flag));
+        m_subHeaps.add(subHeap);
+        if (m_subHeapStartingIndex.getCount())
+        {
+            m_subHeapStartingIndex.add(m_subHeapStartingIndex.getLast() + m_subHeaps.getLast()->getSize());
+        }
+        else
+        {
+            m_subHeapStartingIndex.add(0);
+        }
+        return SLANG_OK;
+    }
+
+    int getSubHeapIndex(int descriptorIndex) const
+    {
+        Slang::Index l = 0;
+        Slang::Index r = m_subHeapStartingIndex.getCount();
+        while (l < r - 1)
+        {
+            Slang::Index m = l + (r - l) / 2;
+            if (m_subHeapStartingIndex[m] < descriptorIndex)
+                l = m;
+            else if (m_subHeapStartingIndex[m] > descriptorIndex)
+                r = m;
+            else
+                return (int)m;
+        }
+        assert(
+            m_subHeapStartingIndex[l] <= descriptorIndex &&
+            m_subHeapStartingIndex[l] + m_subHeaps[l]->getSize() > descriptorIndex);
+        return (int)l;
+    }
+
+    Slang::Result init(
+        ID3D12Device* device,
+        int chunkSize,
+        D3D12_DESCRIPTOR_HEAP_TYPE type,
+        D3D12_DESCRIPTOR_HEAP_FLAGS flag)
+    {
+        m_device = device;
+        m_chunkSize = chunkSize;
+        m_type = type;
+        m_flag = flag;
+
+        return newSubHeap();
+    }
+
+    SLANG_FORCE_INLINE D3D12_CPU_DESCRIPTOR_HANDLE getCpuHandle(int index) const
+    {
+        auto subHeapIndex = getSubHeapIndex(index);
+        return m_subHeaps[subHeapIndex]->getCpuHandle(index - m_subHeapStartingIndex[subHeapIndex]);
+    }
+
+    SLANG_FORCE_INLINE D3D12_GPU_DESCRIPTOR_HANDLE getGpuHandle(int index) const
+    {
+        auto subHeapIndex = getSubHeapIndex(index);
+        return m_subHeaps[subHeapIndex]->getGpuHandle(index - m_subHeapStartingIndex[subHeapIndex]);
+    }
+
+    int allocate(int count)
+    {
+        auto result = m_subHeaps.getLast()->allocate(count);
+        if (result == -1)
+        {
+            newSubHeap();
+            return allocate(count);
+        }
+        return result + m_subHeapStartingIndex.getLast();
+    }
+
+    Slang::Result allocate(D3D12Descriptor* outDescriptor)
+    {
+        int index = allocate(1);
+        if (index < 0)
+        {
+            assert(!"descriptor allocation failed");
+            return SLANG_FAIL;
+        }
+
+        D3D12Descriptor descriptor;
+        descriptor.cpuHandle = getCpuHandle(index);
+
+        *outDescriptor = descriptor;
+        return SLANG_OK;
+    }
+
+    void free(int index, int count)
+    {
+        auto subHeapIndex = getSubHeapIndex(index);
+        m_subHeaps[subHeapIndex]->free(index - m_subHeapStartingIndex[subHeapIndex], count);
+    }
+
+    void free(D3D12Descriptor descriptor)
+    {
+        for (auto& subHeap : m_subHeaps)
+        {
+            if (descriptor.cpuHandle.ptr >= subHeap->getCpuHandle(0).ptr)
+            {
+                auto subIndex = descriptor.cpuHandle.ptr - subHeap->getCpuHandle(0).ptr;
+                if (subIndex < subHeap->getSize())
+                {
+                    subHeap->free(descriptor);
+                    break;
+                }
+            }
+            
+        }
+    }
+};
+
 struct DescriptorHeapReference
 {
     enum class Type
     {
-        Linear, General
+        Linear, General, ExpandingGeneral
     };
     union Ptr
     {
         D3D12DescriptorHeap* linearHeap;
         D3D12GeneralDescriptorHeap* generalHeap;
+        D3D12GeneralExpandingDescriptorHeap* generalExpandingHeap;
     };
     Type type;
     Ptr ptr;
@@ -175,14 +298,21 @@ struct DescriptorHeapReference
         type = Type::General;
         ptr.generalHeap = heap;
     }
+    DescriptorHeapReference(D3D12GeneralExpandingDescriptorHeap* heap)
+    {
+        type = Type::ExpandingGeneral;
+        ptr.generalExpandingHeap = heap;
+    }
     D3D12_CPU_DESCRIPTOR_HANDLE getCpuHandle(int index) const
     {
         switch (type)
         {
         case Type::Linear:
             return ptr.linearHeap->getCpuHandle(index);
-        default:
+        case Type::General:
             return ptr.generalHeap->getCpuHandle(index);
+        default:
+            return ptr.generalExpandingHeap->getCpuHandle(index);
         }
     }
     D3D12_GPU_DESCRIPTOR_HANDLE getGpuHandle(int index) const
@@ -191,8 +321,10 @@ struct DescriptorHeapReference
         {
         case Type::Linear:
             return ptr.linearHeap->getGpuHandle(index);
-        default:
+        case Type::General:
             return ptr.generalHeap->getGpuHandle(index);
+        default:
+            return ptr.generalExpandingHeap->getGpuHandle(index);
         }
     }
     int allocate(int numDescriptors)
@@ -201,8 +333,10 @@ struct DescriptorHeapReference
         {
         case Type::Linear:
             return ptr.linearHeap->allocate(numDescriptors);
-        default:
+        case Type::General:
             return ptr.generalHeap->allocate(numDescriptors);
+        default:
+            return ptr.generalExpandingHeap->allocate(numDescriptors);
         }
     }
     void free(int index, int count)
@@ -212,8 +346,10 @@ struct DescriptorHeapReference
         case Type::Linear:
             SLANG_ASSERT(!"Linear heap does not support free().");
             break;
-        default:
+        case Type::General:
             return ptr.generalHeap->free(index, count);
+        default:
+            return ptr.generalExpandingHeap->free(index, count);
         }
     }
     void freeIfSupported(int index, int count)
@@ -222,8 +358,10 @@ struct DescriptorHeapReference
         {
         case Type::Linear:
             return;
-        default:
+        case Type::General:
             return ptr.generalHeap->free(index, count);
+        default:
+            return ptr.generalExpandingHeap->free(index, count);
         }
     }
 };
diff --git a/tools/gfx/d3d12/render-d3d12.cpp b/tools/gfx/d3d12/render-d3d12.cpp
index a6d02cdc3..e5a1d1876 100644
--- a/tools/gfx/d3d12/render-d3d12.cpp
+++ b/tools/gfx/d3d12/render-d3d12.cpp
@@ -110,7 +110,10 @@ public:
         IResourceView::Desc const& desc,
         IResourceView** outView) override;
     virtual SLANG_NO_THROW Result SLANG_MCALL createBufferView(
-        IBufferResource* buffer, IResourceView::Desc const& desc, IResourceView** outView) override;
+        IBufferResource* buffer,
+        IBufferResource* counterBuffer,
+        IResourceView::Desc const& desc,
+        IResourceView** outView) override;
 
     virtual SLANG_NO_THROW Result SLANG_MCALL
         createFramebuffer(IFramebuffer::Desc const& desc, IFramebuffer** outFrameBuffer) override;
@@ -377,7 +380,7 @@ public:
     {
     public:
         D3D12Descriptor m_descriptor;
-        Slang::RefPtr<D3D12GeneralDescriptorHeap> m_allocator;
+        Slang::RefPtr<D3D12GeneralExpandingDescriptorHeap> m_allocator;
         ~SamplerStateImpl()
         {
             m_allocator->free(m_descriptor);
@@ -394,7 +397,7 @@ public:
     {
     public:
         D3D12Descriptor m_descriptor;
-        RefPtr<D3D12GeneralDescriptorHeap> m_allocator;
+        RefPtr<D3D12GeneralExpandingDescriptorHeap> m_allocator;
         ~ResourceViewInternalImpl() { m_allocator->free(m_descriptor); }
     };
 
@@ -2279,6 +2282,7 @@ public:
     struct ShaderBinary
     {
         SlangStage stage;
+        slang::EntryPointReflection* entryPointInfo;
         List<uint8_t> code;
     };
 
@@ -3934,7 +3938,24 @@ public:
             virtual SLANG_NO_THROW void SLANG_MCALL
                 dispatchComputeIndirect(IBufferResource* argBuffer, uint64_t offset) override
             {
-                SLANG_UNIMPLEMENTED_X("dispatchComputeIndirect");
+                // Submit binding for compute
+                {
+                    ComputeSubmitter submitter(m_d3dCmdList);
+                    RefPtr<PipelineStateBase> newPipeline;
+                    if (SLANG_FAILED(_bindRenderState(&submitter, newPipeline)))
+                    {
+                        assert(!"Failed to bind render state");
+                    }
+                }
+                auto argBufferImpl = static_cast<BufferResourceImpl*>(argBuffer);
+
+                m_d3dCmdList->ExecuteIndirect(
+                    m_renderer->dispatchIndirectCmdSignature,
+                    1,
+                    argBufferImpl->m_resource,
+                    offset,
+                    nullptr,
+                    0);
             }
         };
 
@@ -4233,11 +4254,13 @@ public:
                     }
                     auto rowSize = (footprint.Footprint.Width + formatInfo.blockWidth - 1) /
                                    formatInfo.blockWidth * formatInfo.blockSizeInBytes;
+                    auto rowCount = (footprint.Footprint.Height + formatInfo.blockHeight - 1) /
+                                    formatInfo.blockHeight;
                     footprint.Footprint.RowPitch = (UINT)D3DUtil::calcAligned(
                         rowSize, (uint32_t)D3D12_TEXTURE_DATA_PITCH_ALIGNMENT);
 
-                    auto bufferSize = footprint.Footprint.RowPitch * footprint.Footprint.Height *
-                                      footprint.Footprint.Depth;
+                    auto bufferSize =
+                        footprint.Footprint.RowPitch * rowCount * footprint.Footprint.Depth;
 
                     IBufferResource* stagingBuffer;
                     m_commandBuffer->m_transientHeap->allocateStagingBuffer(
@@ -4249,11 +4272,10 @@ public:
                     bufferImpl->m_resource.getResource()->Map(0, &mapRange, (void**)&bufferData);
                     for (uint32_t z = 0; z < footprint.Footprint.Depth; z++)
                     {
-                        auto imageStart = bufferData + footprint.Footprint.RowPitch *
-                                                           footprint.Footprint.Height * (size_t)z;
+                        auto imageStart = bufferData + footprint.Footprint.RowPitch * rowCount * (size_t)z;
                         auto srcData =
                             (uint8_t*)subResourceData->data + subResourceData->strideZ * z;
-                        for (uint32_t row = 0; row < footprint.Footprint.Height; row++)
+                        for (uint32_t row = 0; row < rowCount; row++)
                         {
                             memcpy(
                                 imageStart + row * (size_t)footprint.Footprint.RowPitch,
@@ -4979,14 +5001,14 @@ public:
     RefPtr<CommandQueueImpl> m_resourceCommandQueue;
     RefPtr<TransientResourceHeapImpl> m_resourceCommandTransientHeap;
 
-    RefPtr<D3D12GeneralDescriptorHeap> m_rtvAllocator;
-    RefPtr<D3D12GeneralDescriptorHeap> m_dsvAllocator;
+    RefPtr<D3D12GeneralExpandingDescriptorHeap> m_rtvAllocator;
+    RefPtr<D3D12GeneralExpandingDescriptorHeap> m_dsvAllocator;
     // Space in the GPU-visible heaps is precious, so we will also keep
     // around CPU-visible heaps for storing descriptors in a format
     // that is ready for copying into the GPU-visible heaps as needed.
     //
-    RefPtr<D3D12GeneralDescriptorHeap> m_cpuViewHeap; ///< Cbv, Srv, Uav
-    RefPtr<D3D12GeneralDescriptorHeap> m_cpuSamplerHeap; ///< Heap for samplers
+    RefPtr<D3D12GeneralExpandingDescriptorHeap> m_cpuViewHeap; ///< Cbv, Srv, Uav
+    RefPtr<D3D12GeneralExpandingDescriptorHeap> m_cpuSamplerHeap; ///< Heap for samplers
 
     // Dll entry points
     PFN_D3D12_GET_DEBUG_INTERFACE m_D3D12GetDebugInterface = nullptr;
@@ -4999,6 +5021,7 @@ public:
     // as well as the command type to be used (DrawInstanced and DrawIndexedInstanced, in this case).
     ComPtr<ID3D12CommandSignature> drawIndirectCmdSignature;
     ComPtr<ID3D12CommandSignature> drawIndexedIndirectCmdSignature;
+    ComPtr<ID3D12CommandSignature> dispatchIndirectCmdSignature;
 };
 
 SLANG_NO_THROW Result SLANG_MCALL D3D12Device::TransientResourceHeapImpl::synchronizeAndReset()
@@ -5189,18 +5212,34 @@ static void _initSrvDesc(
     {
         switch (desc.Dimension)
         {
-            case D3D12_RESOURCE_DIMENSION_TEXTURE1D:  descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE1D; break;
-            case D3D12_RESOURCE_DIMENSION_TEXTURE2D:  descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2D; break;
-            case D3D12_RESOURCE_DIMENSION_TEXTURE3D:  descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE3D; break;
-            default: assert(!"Unknown dimension");
+        case D3D12_RESOURCE_DIMENSION_TEXTURE1D:
+            descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE1D;
+            descOut.Texture1D.MipLevels = subresourceRange.mipLevelCount == 0
+                                              ? desc.MipLevels - subresourceRange.mipLevel
+                                              : subresourceRange.mipLevelCount;
+            descOut.Texture1D.MostDetailedMip = subresourceRange.mipLevel;
+            break;
+        case D3D12_RESOURCE_DIMENSION_TEXTURE2D:
+            descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2D;
+            descOut.Texture2D.PlaneSlice =
+                D3DUtil::getPlaneSlice(descOut.Format, subresourceRange.aspectMask);
+            descOut.Texture2D.ResourceMinLODClamp = 0.0f;
+            descOut.Texture2D.MipLevels = subresourceRange.mipLevelCount == 0
+                                              ? desc.MipLevels - subresourceRange.mipLevel
+                                              : subresourceRange.mipLevelCount;
+            descOut.Texture2D.MostDetailedMip = subresourceRange.mipLevel;
+            break;
+        case D3D12_RESOURCE_DIMENSION_TEXTURE3D:
+            descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE3D;
+            descOut.Texture3D.MipLevels = subresourceRange.mipLevelCount == 0
+                                              ? desc.MipLevels - subresourceRange.mipLevel
+                                              : subresourceRange.mipLevelCount;
+            descOut.Texture3D.MostDetailedMip = subresourceRange.mipLevel;
+            break;
+        default:
+            assert(!"Unknown dimension");
         }
 
-        descOut.Texture2D.MipLevels =
-            subresourceRange.mipLevelCount == 0 ? desc.MipLevels : subresourceRange.mipLevelCount;
-        descOut.Texture2D.MostDetailedMip = subresourceRange.mipLevel;
-        descOut.Texture2D.PlaneSlice =
-            D3DUtil::getPlaneSlice(descOut.Format, subresourceRange.aspectMask);
-        descOut.Texture2D.ResourceMinLODClamp = 0.0f;
     }
     else if (resourceType == IResource::Type::TextureCube)
     {
@@ -5213,7 +5252,7 @@ static void _initSrvDesc(
                                                     : subresourceRange.layerCount / 6;
             descOut.TextureCubeArray.First2DArrayFace = subresourceRange.baseArrayLayer;
             descOut.TextureCubeArray.MipLevels = subresourceRange.mipLevelCount == 0
-                                                     ? desc.MipLevels
+                                                     ? desc.MipLevels - subresourceRange.mipLevel
                                                      : subresourceRange.mipLevelCount;
             descOut.TextureCubeArray.MostDetailedMip = subresourceRange.mipLevel;
             descOut.TextureCubeArray.ResourceMinLODClamp = 0;
@@ -5223,7 +5262,7 @@ static void _initSrvDesc(
             descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURECUBE;
 
             descOut.TextureCube.MipLevels = subresourceRange.mipLevelCount == 0
-                                                ? desc.MipLevels
+                                                ? desc.MipLevels - subresourceRange.mipLevel
                                                 : subresourceRange.mipLevelCount;
             descOut.TextureCube.MostDetailedMip = subresourceRange.mipLevel;
             descOut.TextureCube.ResourceMinLODClamp = 0;
@@ -5235,22 +5274,46 @@ static void _initSrvDesc(
 
         switch (desc.Dimension)
         {
-            case D3D12_RESOURCE_DIMENSION_TEXTURE1D:  descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE1DARRAY; break;
-            case D3D12_RESOURCE_DIMENSION_TEXTURE2D:  descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2DARRAY; break;
-            case D3D12_RESOURCE_DIMENSION_TEXTURE3D:  descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE3D; break;
+        case D3D12_RESOURCE_DIMENSION_TEXTURE1D:
+            descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE1DARRAY;
+            descOut.Texture1D.MostDetailedMip = subresourceRange.mipLevel;
+            descOut.Texture1D.MipLevels = subresourceRange.mipLevelCount == 0
+                                              ? desc.MipLevels
+                                              : subresourceRange.mipLevelCount;
+            descOut.Texture1DArray.ArraySize = subresourceRange.layerCount == 0
+                                                   ? desc.DepthOrArraySize
+                                                   : subresourceRange.layerCount;
+            descOut.Texture1DArray.FirstArraySlice = subresourceRange.baseArrayLayer;
+            descOut.Texture1DArray.ResourceMinLODClamp = 0;
+            descOut.Texture1DArray.MostDetailedMip = subresourceRange.mipLevel;
+            descOut.Texture1DArray.MipLevels = subresourceRange.mipLevelCount == 0
+                                                   ? desc.MipLevels - subresourceRange.mipLevel
+                                                   : subresourceRange.mipLevelCount;
+            break;
+        case D3D12_RESOURCE_DIMENSION_TEXTURE2D:
+            descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE2DARRAY;
+            descOut.Texture2DArray.ArraySize =
+                subresourceRange.layerCount == 0 ? desc.DepthOrArraySize : subresourceRange.layerCount;
+            descOut.Texture2DArray.FirstArraySlice = subresourceRange.baseArrayLayer;
+            descOut.Texture2DArray.PlaneSlice =
+                D3DUtil::getPlaneSlice(descOut.Format, subresourceRange.aspectMask);
+            descOut.Texture2DArray.ResourceMinLODClamp = 0;
+            descOut.Texture2DArray.MostDetailedMip = subresourceRange.mipLevel;
+            descOut.Texture2DArray.MipLevels = subresourceRange.mipLevelCount == 0
+                                                   ? desc.MipLevels - subresourceRange.mipLevel
+                                                   : subresourceRange.mipLevelCount;
+            break;
+        case D3D12_RESOURCE_DIMENSION_TEXTURE3D:
+            descOut.ViewDimension = D3D12_SRV_DIMENSION_TEXTURE3D;
+            descOut.Texture3D.MostDetailedMip = subresourceRange.mipLevel;
+            descOut.Texture3D.MipLevels = subresourceRange.mipLevelCount == 0
+                                                   ? desc.MipLevels
+                                                   : subresourceRange.mipLevelCount;
+            break;
 
-            default: assert(!"Unknown dimension");
+        default:
+            assert(!"Unknown dimension");
         }
-
-        descOut.Texture2DArray.ArraySize =
-            subresourceRange.layerCount == 0 ? desc.DepthOrArraySize : subresourceRange.layerCount;
-        descOut.Texture2DArray.MostDetailedMip = subresourceRange.mipLevel;
-        descOut.Texture2DArray.MipLevels =
-            subresourceRange.mipLevelCount == 0 ? desc.MipLevels : subresourceRange.mipLevelCount;
-        descOut.Texture2DArray.FirstArraySlice = subresourceRange.baseArrayLayer;
-        descOut.Texture2DArray.PlaneSlice =
-            D3DUtil::getPlaneSlice(descOut.Format, subresourceRange.aspectMask);
-        descOut.Texture2DArray.ResourceMinLODClamp = 0;
     }
 }
 
@@ -5822,20 +5885,20 @@ Result D3D12Device::initialize(const Desc& desc)
     // since this object is already owned by `D3D12Device`.
     m_resourceCommandTransientHeap->breakStrongReferenceToDevice();
 
-    m_cpuViewHeap = new D3D12GeneralDescriptorHeap();
+    m_cpuViewHeap = new D3D12GeneralExpandingDescriptorHeap();
     SLANG_RETURN_ON_FAIL(m_cpuViewHeap->init(
         m_device,
         1024 * 1024,
         D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV,
         D3D12_DESCRIPTOR_HEAP_FLAG_NONE));
-    m_cpuSamplerHeap = new D3D12GeneralDescriptorHeap();
+    m_cpuSamplerHeap = new D3D12GeneralExpandingDescriptorHeap();
     SLANG_RETURN_ON_FAIL(m_cpuSamplerHeap->init(
         m_device, 2048, D3D12_DESCRIPTOR_HEAP_TYPE_SAMPLER, D3D12_DESCRIPTOR_HEAP_FLAG_NONE));
 
-    m_rtvAllocator = new D3D12GeneralDescriptorHeap();
+    m_rtvAllocator = new D3D12GeneralExpandingDescriptorHeap();
     SLANG_RETURN_ON_FAIL(m_rtvAllocator->init(
         m_device, 16 * 1024, D3D12_DESCRIPTOR_HEAP_TYPE_RTV, D3D12_DESCRIPTOR_HEAP_FLAG_NONE));
-    m_dsvAllocator = new D3D12GeneralDescriptorHeap();
+    m_dsvAllocator = new D3D12GeneralExpandingDescriptorHeap();
     SLANG_RETURN_ON_FAIL(m_dsvAllocator->init(
         m_device, 1024, D3D12_DESCRIPTOR_HEAP_TYPE_DSV, D3D12_DESCRIPTOR_HEAP_FLAG_NONE));
 
@@ -5936,6 +5999,21 @@ Result D3D12Device::initialize(const Desc& desc)
         SLANG_RETURN_ON_FAIL(m_device->CreateCommandSignature(&desc, nullptr, IID_PPV_ARGS(drawIndexedIndirectCmdSignature.writeRef())));
     }
 
+    // Allocate a D3D12 "command signature" object that matches the behavior
+    // of a D3D11-style `Dispatch` operation.
+    {
+        D3D12_INDIRECT_ARGUMENT_DESC args;
+        args.Type = D3D12_INDIRECT_ARGUMENT_TYPE_DISPATCH;
+
+        D3D12_COMMAND_SIGNATURE_DESC desc;
+        desc.ByteStride = sizeof(D3D12_DISPATCH_ARGUMENTS);
+        desc.NumArgumentDescs = 1;
+        desc.pArgumentDescs = &args;
+        desc.NodeMask = 0;
+
+        SLANG_RETURN_ON_FAIL(m_device->CreateCommandSignature(
+            &desc, nullptr, IID_PPV_ARGS(dispatchIndirectCmdSignature.writeRef())));
+    }
     m_isInitialized = true;
     return SLANG_OK;
 }
@@ -6046,6 +6124,31 @@ static D3D12_RESOURCE_DIMENSION _calcResourceDimension(IResource::Type type)
     }
 }
 
+DXGI_FORMAT getTypelessFormatFromDepthFormat(Format format)
+{
+    switch (format)
+    {
+    case Format::D16_UNORM:
+        return DXGI_FORMAT_R16_TYPELESS;
+    case Format::D32_FLOAT:
+        return DXGI_FORMAT_R32_TYPELESS;
+    default:
+        return D3DUtil::getMapFormat(format);
+    }
+}
+
+BOOL isTypelessDepthFormat(DXGI_FORMAT format)
+{
+    switch (format)
+    {
+    case DXGI_FORMAT_R16_TYPELESS:
+    case DXGI_FORMAT_R32_TYPELESS:
+        return true;
+    default:
+        return false;
+    }
+}
+
 Result setupResourceDesc(D3D12_RESOURCE_DESC& resourceDesc, const ITextureResource::Desc& srcDesc)
 {
     const DXGI_FORMAT pixelFormat = D3DUtil::getMapFormat(srcDesc.format);
@@ -6080,6 +6183,13 @@ Result setupResourceDesc(D3D12_RESOURCE_DESC& resourceDesc, const ITextureResour
 
     resourceDesc.Alignment = 0;
 
+    if (isDepthFormat(srcDesc.format) &&
+        (srcDesc.allowedStates.contains(ResourceState::ShaderResource) ||
+         srcDesc.allowedStates.contains(ResourceState::UnorderedAccess)))
+    {
+        resourceDesc.Format = getTypelessFormatFromDepthFormat(srcDesc.format);
+    }
+
     return SLANG_OK;
 }
 
@@ -6129,6 +6239,10 @@ Result D3D12Device::createTextureResource(const ITextureResource::Desc& descIn,
         {
             clearValuePtr = nullptr;
         }
+        if (isTypelessDepthFormat(resourceDesc.Format))
+        {
+            clearValuePtr = nullptr;
+        }
         clearValue.Format = resourceDesc.Format;
         memcpy(clearValue.Color, &descIn.optimalClearValue.color, sizeof(clearValue.Color));
         clearValue.DepthStencil.Depth = descIn.optimalClearValue.depthStencil.depth;
@@ -6632,9 +6746,7 @@ Result D3D12Device::createTextureView(ITextureResource* texture, IResourceView::
                 d3d12desc.ViewDimension = D3D12_UAV_DIMENSION_TEXTURE3D;
                 d3d12desc.Texture3D.MipSlice = desc.subresourceRange.mipLevel;
                 d3d12desc.Texture3D.FirstWSlice = desc.subresourceRange.baseArrayLayer;
-                d3d12desc.Texture3D.WSize = desc.subresourceRange.layerCount == 0
-                                                ? resourceDesc.size.depth
-                                                : desc.subresourceRange.layerCount;
+                d3d12desc.Texture3D.WSize = resourceDesc.size.depth;
                 break;
             default:
                 return SLANG_FAIL;
@@ -6719,7 +6831,11 @@ Result D3D12Device::getFormatSupportedResourceStates(Format format, ResourceStat
     return SLANG_OK;
 }
 
-Result D3D12Device::createBufferView(IBufferResource* buffer, IResourceView::Desc const& desc, IResourceView** outView)
+Result D3D12Device::createBufferView(
+    IBufferResource* buffer,
+    IBufferResource* counterBuffer,
+    IResourceView::Desc const& desc,
+    IResourceView** outView)
 {
     auto resourceImpl = (BufferResourceImpl*) buffer;
     auto resourceDesc = *resourceImpl->getDesc();
@@ -6745,14 +6861,14 @@ Result D3D12Device::createBufferView(IBufferResource* buffer, IResourceView::Des
                 uavDesc.Buffer.NumElements =
                     desc.bufferRange.elementCount == 0
                         ? UINT(resourceDesc.sizeInBytes / desc.bufferElementSize)
-                        : desc.bufferRange.elementCount;
+                        : (UINT)desc.bufferRange.elementCount;
             }
             else if(desc.format == Format::Unknown)
             {
                 uavDesc.Format = DXGI_FORMAT_R32_TYPELESS;
                 uavDesc.Buffer.NumElements = desc.bufferRange.elementCount == 0
                                                  ? UINT(resourceDesc.sizeInBytes / 4)
-                                                 : desc.bufferRange.elementCount / 4;
+                                                 : UINT(desc.bufferRange.elementCount / 4);
                 uavDesc.Buffer.Flags |= D3D12_BUFFER_UAV_FLAG_RAW;
             }
             else
@@ -6763,16 +6879,16 @@ Result D3D12Device::createBufferView(IBufferResource* buffer, IResourceView::Des
                 uavDesc.Buffer.NumElements =
                     desc.bufferRange.elementCount == 0
                         ? UINT(resourceDesc.sizeInBytes / sizeInfo.blockSizeInBytes)
-                        : desc.bufferRange.elementCount;
+                        : (UINT)desc.bufferRange.elementCount;
             }
-
-
-            // TODO: need to support the separate "counter resource" for the case
-            // of append/consume buffers with attached counters.
-
+            auto counterResourceImpl = static_cast<BufferResourceImpl*>(counterBuffer);
             SLANG_RETURN_ON_FAIL(m_cpuViewHeap->allocate(&viewImpl->m_descriptor));
             viewImpl->m_allocator = m_cpuViewHeap;
-            m_device->CreateUnorderedAccessView(resourceImpl->m_resource, nullptr, &uavDesc, viewImpl->m_descriptor.cpuHandle);
+            m_device->CreateUnorderedAccessView(
+                resourceImpl->m_resource,
+                counterResourceImpl ? counterResourceImpl->m_resource.getResource() : nullptr,
+                &uavDesc,
+                viewImpl->m_descriptor.cpuHandle);
         }
         break;
 
@@ -6790,14 +6906,14 @@ Result D3D12Device::createBufferView(IBufferResource* buffer, IResourceView::Des
                 srvDesc.Buffer.NumElements =
                     desc.bufferRange.elementCount == 0
                         ? UINT(resourceDesc.sizeInBytes / desc.bufferElementSize)
-                        : desc.bufferRange.elementCount;
+                        : (UINT)desc.bufferRange.elementCount;
             }
             else if (desc.format == Format::Unknown)
             {
                 srvDesc.Format = DXGI_FORMAT_R32_TYPELESS;
                 srvDesc.Buffer.NumElements = desc.bufferRange.elementCount == 0
                                                  ? UINT(resourceDesc.sizeInBytes / 4)
-                                                 : desc.bufferRange.elementCount / 4;
+                                                 : UINT(desc.bufferRange.elementCount / 4);
                 srvDesc.Buffer.Flags |= D3D12_BUFFER_SRV_FLAG_RAW;
             }
             else
@@ -6808,7 +6924,7 @@ Result D3D12Device::createBufferView(IBufferResource* buffer, IResourceView::Des
                 srvDesc.Buffer.NumElements =
                     desc.bufferRange.elementCount == 0
                         ? UINT(resourceDesc.sizeInBytes / sizeInfo.blockSizeInBytes)
-                        : desc.bufferRange.elementCount;
+                        : (UINT)desc.bufferRange.elementCount;
             }
 
             SLANG_RETURN_ON_FAIL(m_cpuViewHeap->allocate(&viewImpl->m_descriptor));
@@ -7021,12 +7137,12 @@ Result D3D12Device::readBufferResource(
 Result D3D12Device::createProgram(const IShaderProgram::Desc& desc, IShaderProgram** outProgram, ISlangBlob** outDiagnosticBlob)
 {
     RefPtr<ShaderProgramImpl> shaderProgram = new ShaderProgramImpl();
-    shaderProgram->slangProgram = desc.slangProgram;
+    shaderProgram->init(desc);
     ComPtr<ID3DBlob> d3dDiagnosticBlob;
     auto rootShaderLayoutResult = RootShaderObjectLayoutImpl::create(
         this,
-        desc.slangProgram,
-        desc.slangProgram->getLayout(),
+        shaderProgram->linkedProgram,
+        shaderProgram->linkedProgram->getLayout(),
         shaderProgram->m_rootObjectLayout.writeRef(),
         d3dDiagnosticBlob.writeRef());
     if (!SLANG_SUCCEEDED(rootShaderLayoutResult))
@@ -7039,22 +7155,22 @@ Result D3D12Device::createProgram(const IShaderProgram::Desc& desc, IShaderProgr
         }
         return rootShaderLayoutResult;
     }
-    if (desc.slangProgram->getSpecializationParamCount() != 0)
+    if (shaderProgram->isSpecializable())
     {
         // For a specializable program, we don't invoke any actual slang compilation yet.
         returnComPtr(outProgram, shaderProgram);
         return SLANG_OK;
     }
     // For a fully specialized program, read and store its kernel code in `shaderProgram`.
-    auto programReflection = desc.slangProgram->getLayout();
-    for (SlangUInt i = 0; i < programReflection->getEntryPointCount(); i++)
+    auto compileShader = [&](slang::EntryPointReflection* entryPointInfo,
+                              slang::IComponentType* entryPointComponent,
+                              SlangInt entryPointIndex)
     {
-        auto entryPointInfo = programReflection->getEntryPointByIndex(i);
         auto stage = entryPointInfo->getStage();
         ComPtr<ISlangBlob> kernelCode;
         ComPtr<ISlangBlob> diagnostics;
-        auto compileResult = desc.slangProgram->getEntryPointCode(
-            (SlangInt)i, 0, kernelCode.writeRef(), diagnostics.writeRef());
+        auto compileResult = entryPointComponent->getEntryPointCode(
+            entryPointIndex, 0, kernelCode.writeRef(), diagnostics.writeRef());
         if (diagnostics)
         {
             getDebugCallback()->handleMessage(
@@ -7067,10 +7183,35 @@ Result D3D12Device::createProgram(const IShaderProgram::Desc& desc, IShaderProgr
         SLANG_RETURN_ON_FAIL(compileResult);
         ShaderBinary shaderBin;
         shaderBin.stage = stage;
+        shaderBin.entryPointInfo = entryPointInfo;
         shaderBin.code.addRange(
             reinterpret_cast<const uint8_t*>(kernelCode->getBufferPointer()),
             (Index)kernelCode->getBufferSize());
         shaderProgram->m_shaders.add(_Move(shaderBin));
+        return SLANG_OK;
+    };
+
+    if (shaderProgram->linkedEntryPoints.getCount() == 0)
+    {
+        // If the user does not explicitly specify entry point components, find them from `linkedEntryPoints`.
+        auto programReflection = shaderProgram->linkedProgram->getLayout();
+        for (SlangUInt i = 0; i < programReflection->getEntryPointCount(); i++)
+        {
+            SLANG_RETURN_ON_FAIL(compileShader(
+                programReflection->getEntryPointByIndex(i),
+                shaderProgram->linkedProgram,
+                (SlangInt)i));
+        }
+    }
+    else
+    {
+        // If the user specifies entry point components via the separated entry point array, compile code
+        // from there.
+        for (auto& entryPoint : shaderProgram->linkedEntryPoints)
+        {
+            SLANG_RETURN_ON_FAIL(
+                compileShader(entryPoint->getLayout()->getEntryPointByIndex(0), entryPoint, 0));
+        }
     }
     returnComPtr(outProgram, shaderProgram);
     return SLANG_OK;
@@ -7150,6 +7291,7 @@ Result D3D12Device::createGraphicsPipelineState(const GraphicsPipelineStateDesc&
     D3D12_GRAPHICS_PIPELINE_STATE_DESC psoDesc = {};
 
     psoDesc.pRootSignature = programImpl->m_rootObjectLayout->m_rootSignature;
+
     for (auto& shaderBin : programImpl->m_shaders)
     {
         switch (shaderBin.stage)
@@ -7283,7 +7425,7 @@ Result D3D12Device::createGraphicsPipelineState(const GraphicsPipelineStateDesc&
     if (m_pipelineCreationAPIDispatcher)
     {
         SLANG_RETURN_ON_FAIL(m_pipelineCreationAPIDispatcher->createGraphicsPipelineState(
-            this, programImpl->slangProgram.get(), &psoDesc, (void**)pipelineState.writeRef()));
+            this, programImpl->linkedProgram.get(), &psoDesc, (void**)pipelineState.writeRef()));
     }
     else
     {
@@ -7312,7 +7454,7 @@ Result D3D12Device::createComputePipelineState(const ComputePipelineStateDesc& i
 
     // Only actually create a D3D12 pipeline state if the pipeline is fully specialized.
     ComPtr<ID3D12PipelineState> pipelineState;
-    if (!programImpl->slangProgram || programImpl->slangProgram->getSpecializationParamCount() == 0)
+    if (!programImpl->isSpecializable())
     {
         // Describe and create the compute pipeline state object
         D3D12_COMPUTE_PIPELINE_STATE_DESC computeDesc = {};
@@ -7361,7 +7503,7 @@ Result D3D12Device::createComputePipelineState(const ComputePipelineStateDesc& i
             {
                 SLANG_RETURN_ON_FAIL(m_pipelineCreationAPIDispatcher->createComputePipelineState(
                     this,
-                    programImpl->slangProgram.get(),
+                    programImpl->linkedProgram.get(),
                     &computeDesc,
                     (void**)pipelineState.writeRef()));
             }
@@ -7835,8 +7977,8 @@ Result D3D12Device::createRayTracingPipelineState(const RayTracingPipelineStateD
     pipelineStateImpl->init(inDesc);
 
     auto program = static_cast<ShaderProgramImpl*>(inDesc.program);
-    auto slangProgram = program->slangProgram;
-    auto programLayout = slangProgram->getLayout();
+    auto slangGlobalScope = program->linkedProgram;
+    auto programLayout = slangGlobalScope->getLayout();
 
     if (!program->m_rootObjectLayout->m_rootSignature)
     {
@@ -7847,13 +7989,24 @@ Result D3D12Device::createRayTracingPipelineState(const RayTracingPipelineStateD
     ChunkedList<D3D12_DXIL_LIBRARY_DESC> dxilLibraries;
     ChunkedList<D3D12_HIT_GROUP_DESC> hitGroups;
     ChunkedList<ComPtr<ISlangBlob>> codeBlobs;
+    ChunkedList<D3D12_EXPORT_DESC> exports;
+    ChunkedList<const wchar_t*> strPtrs;
+
     ComPtr<ISlangBlob> diagnostics;
     ChunkedList<OSString> stringPool;
-    for (SlangUInt i = 0; i < programLayout->getEntryPointCount(); i++)
+    auto getWStr = [&](const char* name)
+    {
+        String str = String(name);
+        auto wstr = str.toWString();
+        return stringPool.add(wstr)->begin();
+    };
+    auto compileShader = [&](slang::EntryPointLayout* entryPointInfo,
+                             slang::IComponentType* component,
+                             SlangInt entryPointIndex)
     {
         ComPtr<ISlangBlob> codeBlob;
-        auto compileResult =
-            slangProgram->getEntryPointCode(i, 0, codeBlob.writeRef(), diagnostics.writeRef());
+        auto compileResult = component->getEntryPointCode(
+            entryPointIndex, 0, codeBlob.writeRef(), diagnostics.writeRef());
         if (diagnostics.get())
         {
             getDebugCallback()->handleMessage(
@@ -7864,20 +8017,38 @@ Result D3D12Device::createRayTracingPipelineState(const RayTracingPipelineStateD
         SLANG_RETURN_ON_FAIL(compileResult);
         codeBlobs.add(codeBlob);
         D3D12_DXIL_LIBRARY_DESC library = {};
-        library.DXILLibrary.BytecodeLength = codeBlob->getBufferSize();;
+        library.DXILLibrary.BytecodeLength = codeBlob->getBufferSize();
         library.DXILLibrary.pShaderBytecode = codeBlob->getBufferPointer();
+        library.NumExports = 1;
+        D3D12_EXPORT_DESC exportDesc = {};
+        exportDesc.Name = getWStr(entryPointInfo->getNameOverride());
+        exportDesc.ExportToRename = getWStr(entryPointInfo->getNameOverride());
+        exportDesc.Flags = D3D12_EXPORT_FLAG_NONE;
+        library.pExports = exports.add(exportDesc);
 
         D3D12_STATE_SUBOBJECT dxilSubObject = {};
         dxilSubObject.Type = D3D12_STATE_SUBOBJECT_TYPE_DXIL_LIBRARY;
         dxilSubObject.pDesc = dxilLibraries.add(library);
         subObjects.add(dxilSubObject);
+        return SLANG_OK;
+    };
+    if (program->linkedEntryPoints.getCount() == 0)
+    {
+        for (SlangUInt i = 0; i < programLayout->getEntryPointCount(); i++)
+        {
+            SLANG_RETURN_ON_FAIL(compileShader(
+                programLayout->getEntryPointByIndex(i), program->linkedProgram, (SlangInt)i));
+        }
     }
-    auto getWStr = [&](const char* name)
+    else
     {
-        String str = String(name);
-        auto wstr = str.toWString();
-        return stringPool.add(wstr)->begin();
-    };
+        for (auto& entryPoint : program->linkedEntryPoints)
+        {
+            SLANG_RETURN_ON_FAIL(
+                compileShader(entryPoint->getLayout()->getEntryPointByIndex(0), entryPoint, 0));
+        }
+    }
+    
     for (int i = 0; i < inDesc.hitGroupCount; i++)
     {
         auto hitGroup = inDesc.hitGroups[i];
@@ -7909,7 +8080,7 @@ Result D3D12Device::createRayTracingPipelineState(const RayTracingPipelineStateD
     D3D12_RAYTRACING_SHADER_CONFIG shaderConfig = {};
     // According to DXR spec, fixed function triangle intersections must use float2 as ray attributes
     // that defines the barycentric coordinates at intersection.
-    shaderConfig.MaxAttributeSizeInBytes = sizeof(float) * 2;
+    shaderConfig.MaxAttributeSizeInBytes = inDesc.maxAttributeSizeInBytes;
     shaderConfig.MaxPayloadSizeInBytes = inDesc.maxRayPayloadSize;
     D3D12_STATE_SUBOBJECT shaderConfigSubObject = {};
     shaderConfigSubObject.Type = D3D12_STATE_SUBOBJECT_TYPE_RAYTRACING_SHADER_CONFIG;
@@ -7932,7 +8103,7 @@ Result D3D12Device::createRayTracingPipelineState(const RayTracingPipelineStateD
 
     if (m_pipelineCreationAPIDispatcher)
     {
-        m_pipelineCreationAPIDispatcher->beforeCreateRayTracingState(this, slangProgram);
+        m_pipelineCreationAPIDispatcher->beforeCreateRayTracingState(this, slangGlobalScope);
     }
 
     D3D12_STATE_OBJECT_DESC rtpsoDesc = {};
@@ -7943,7 +8114,7 @@ Result D3D12Device::createRayTracingPipelineState(const RayTracingPipelineStateD
 
     if (m_pipelineCreationAPIDispatcher)
     {
-        m_pipelineCreationAPIDispatcher->afterCreateRayTracingState(this, slangProgram);
+        m_pipelineCreationAPIDispatcher->afterCreateRayTracingState(this, slangGlobalScope);
     }
 
     returnComPtr(outState, pipelineStateImpl);
diff --git a/tools/gfx/debug-layer.cpp b/tools/gfx/debug-layer.cpp
index 484729068..eed6afdb0 100644
--- a/tools/gfx/debug-layer.cpp
+++ b/tools/gfx/debug-layer.cpp
@@ -464,6 +464,7 @@ Result DebugDevice::createTextureView(
 
 Result DebugDevice::createBufferView(
     IBufferResource* buffer,
+    IBufferResource* counterBuffer,
     IResourceView::Desc const& desc,
     IResourceView** outView)
 {
@@ -472,6 +473,7 @@ Result DebugDevice::createBufferView(
     RefPtr<DebugResourceView> outObject = new DebugResourceView();
     auto result = baseObject->createBufferView(
         static_cast<DebugBufferResource*>(buffer)->baseObject,
+        counterBuffer ? static_cast<DebugBufferResource*>(counterBuffer)->baseObject : nullptr,
         desc,
         outObject->baseObject.writeRef());
     if (SLANG_FAILED(result))
@@ -1909,7 +1911,7 @@ Result DebugFence::setCurrentValue(uint64_t value)
 
 DebugShaderProgram::DebugShaderProgram(const IShaderProgram::Desc& desc)
 {
-    m_slangProgram = desc.slangProgram;
+    m_slangProgram = desc.slangGlobalScope;
 }
 
 Result DebugPipelineState::getNativeHandle(InteropHandle* outHandle)
diff --git a/tools/gfx/debug-layer.h b/tools/gfx/debug-layer.h
index 069895f9b..cb894dfd4 100644
--- a/tools/gfx/debug-layer.h
+++ b/tools/gfx/debug-layer.h
@@ -85,6 +85,7 @@ public:
         IResourceView** outView) override;
     virtual SLANG_NO_THROW Result SLANG_MCALL createBufferView(
         IBufferResource* buffer,
+        IBufferResource* counterBuffer,
         IResourceView::Desc const& desc,
         IResourceView** outView) override;
     virtual SLANG_NO_THROW Result SLANG_MCALL getAccelerationStructurePrebuildInfo(
diff --git a/tools/gfx/mutable-shader-object.h b/tools/gfx/mutable-shader-object.h
index 54fdf87c2..a9562f200 100644
--- a/tools/gfx/mutable-shader-object.h
+++ b/tools/gfx/mutable-shader-object.h
@@ -249,7 +249,7 @@ namespace gfx
         MutableRootShaderObject(RendererBase* device, Slang::RefPtr<ShaderProgramBase> program)
         {
             this->m_device = device;
-            auto programLayout = program->slangProgram->getLayout();
+            auto programLayout = program->slangGlobalScope->getLayout();
             SlangInt entryPointCount = programLayout->getEntryPointCount();
             for (SlangInt e = 0; e < entryPointCount; ++e)
             {
diff --git a/tools/gfx/open-gl/render-gl.cpp b/tools/gfx/open-gl/render-gl.cpp
index 63c20fbea..b407fa000 100644
--- a/tools/gfx/open-gl/render-gl.cpp
+++ b/tools/gfx/open-gl/render-gl.cpp
@@ -130,7 +130,10 @@ public:
     virtual SLANG_NO_THROW Result SLANG_MCALL createTextureView(
         ITextureResource* texture, IResourceView::Desc const& desc, IResourceView** outView) override;
     virtual SLANG_NO_THROW Result SLANG_MCALL createBufferView(
-        IBufferResource* buffer, IResourceView::Desc const& desc, IResourceView** outView) override;
+        IBufferResource* buffer,
+        IBufferResource* counterBuffer,
+        IResourceView::Desc const& desc,
+        IResourceView** outView) override;
 
     virtual SLANG_NO_THROW Result SLANG_MCALL createInputLayout(
         IInputLayout::Desc const& desc,
@@ -2546,7 +2549,10 @@ SLANG_NO_THROW Result SLANG_MCALL GLDevice::createTextureView(
 }
 
 SLANG_NO_THROW Result SLANG_MCALL GLDevice::createBufferView(
-    IBufferResource* buffer, IResourceView::Desc const& desc, IResourceView** outView)
+    IBufferResource* buffer,
+    IBufferResource* counterBuffer,
+    IResourceView::Desc const& desc,
+    IResourceView** outView)
 {
     auto resourceImpl = (BufferResourceImpl*) buffer;
 
@@ -2759,23 +2765,23 @@ void GLDevice::dispatchCompute(int x, int y, int z)
 Result GLDevice::createProgram(
     const IShaderProgram::Desc& desc, IShaderProgram** outProgram, ISlangBlob** outDiagnosticBlob)
 {
-    if (desc.slangProgram->getSpecializationParamCount() != 0)
+    if (desc.slangGlobalScope->getSpecializationParamCount() != 0)
     {
         // For a specializable program, we don't invoke any actual slang compilation yet.
         RefPtr<ShaderProgramImpl> shaderProgram = new ShaderProgramImpl(m_weakRenderer, 0);
-        shaderProgram->slangProgram = desc.slangProgram;
+        shaderProgram->init(desc);
         returnComPtr(outProgram, shaderProgram);
         return SLANG_OK;
     }
 
     auto programID = glCreateProgram();
-    auto programLayout = desc.slangProgram->getLayout();
+    auto programLayout = desc.slangGlobalScope->getLayout();
     ShortList<GLuint> shaderIDs;
     for (SlangUInt i = 0; i < programLayout->getEntryPointCount(); i++)
     {
         ComPtr<ISlangBlob> kernelCode;
         ComPtr<ISlangBlob> diagnostics;
-        auto compileResult = desc.slangProgram->getEntryPointCode(
+        auto compileResult = desc.slangGlobalScope->getEntryPointCode(
             i, 0, kernelCode.writeRef(), diagnostics.writeRef());
         if (diagnostics)
         {
@@ -2844,7 +2850,7 @@ Result GLDevice::createProgram(
     }
 
     RefPtr<ShaderProgramImpl> program = new ShaderProgramImpl(m_weakRenderer, programID);
-    program->slangProgram = desc.slangProgram;
+    program->slangGlobalScope = desc.slangGlobalScope;
     returnComPtr(outProgram, program);
     return SLANG_OK;
 }
@@ -2913,7 +2919,7 @@ Result GLDevice::createRootShaderObject(IShaderProgram* program, ShaderObjectBas
     RefPtr<RootShaderObjectImpl> shaderObject;
     RefPtr<RootShaderObjectLayoutImpl> rootLayout;
     SLANG_RETURN_ON_FAIL(RootShaderObjectLayoutImpl::create(
-        this, programImpl->slangProgram, programImpl->slangProgram->getLayout(), rootLayout.writeRef()));
+        this, programImpl->slangGlobalScope, programImpl->slangGlobalScope->getLayout(), rootLayout.writeRef()));
     SLANG_RETURN_ON_FAIL(RootShaderObjectImpl::create(
         this, rootLayout.Ptr(), shaderObject.writeRef()));
     returnRefPtrMove(outObject, shaderObject);
diff --git a/tools/gfx/renderer-shared.cpp b/tools/gfx/renderer-shared.cpp
index d1cedaa61..f9ffe8dfd 100644
--- a/tools/gfx/renderer-shared.cpp
+++ b/tools/gfx/renderer-shared.cpp
@@ -298,8 +298,17 @@ void PipelineStateBase::initializeBase(const PipelineStateDesc& inDesc)
 
     auto program = desc.getProgram();
     m_program = program;
-    isSpecializable = (program->slangProgram && program->slangProgram->getSpecializationParamCount() != 0);
-
+    isSpecializable = false;
+    if (program->slangGlobalScope && program->slangGlobalScope->getSpecializationParamCount() != 0)
+        isSpecializable = true;
+    for (auto& entryPoint : program->slangEntryPoints)
+    {
+        if (entryPoint->getSpecializationParamCount() != 0)
+        {
+            isSpecializable = true;
+            break;
+        }
+    }
     // Hold a strong reference to inputLayout and framebufferLayout objects to prevent it from
     // destruction.
     if (inDesc.type == PipelineType::Graphics)
@@ -716,12 +725,12 @@ ResourceViewBase* SimpleShaderObjectData::getResourceView(
         viewDesc.format = Format::Unknown;
         viewDesc.type = IResourceView::Type::ShaderResource;
         SLANG_RETURN_NULL_ON_FAIL(device->createBufferView(
-            bufferResource.get(), viewDesc, resourceView.writeRef()));
+            bufferResource.get(), nullptr, viewDesc, resourceView.writeRef()));
         m_structuredBufferView = static_cast<ResourceViewBase*>(resourceView.get());
         viewDesc.type = IResourceView::Type::UnorderedAccess;
         SLANG_RETURN_NULL_ON_FAIL(
             device->createBufferView(
-            bufferResource.get(), viewDesc, resourceView.writeRef()));
+            bufferResource.get(), nullptr, viewDesc, resourceView.writeRef()));
         m_rwStructuredBufferView = static_cast<ResourceViewBase*>(resourceView.get());
     }
 
@@ -737,6 +746,57 @@ ResourceViewBase* SimpleShaderObjectData::getResourceView(
     }
 }
 
+void ShaderProgramBase::init(const IShaderProgram::Desc& inDesc)
+{
+    desc = inDesc;
+
+    slangGlobalScope = desc.slangGlobalScope;
+    for (uint32_t i = 0; i < desc.entryPointCount; i++)
+    {
+        slangEntryPoints.add(ComPtr<slang::IComponentType>(desc.slangEntryPoints[i]));
+    }
+
+    auto session = desc.slangGlobalScope ? desc.slangGlobalScope->getSession() : nullptr;
+    if (desc.linkingStyle == IShaderProgram::LinkingStyle::SingleProgram)
+    {
+        List<slang::IComponentType*> components;
+        if (desc.slangGlobalScope)
+        {
+            components.add(desc.slangGlobalScope);
+        }
+        for (uint32_t i = 0; i < desc.entryPointCount; i++)
+        {
+            if (!session)
+            {
+                session = desc.slangEntryPoints[i]->getSession();
+            }
+            components.add(desc.slangEntryPoints[i]);
+        }
+        session->createCompositeComponentType(
+            components.getBuffer(), components.getCount(), linkedProgram.writeRef());
+    }
+    else
+    {
+        for (uint32_t i = 0; i < desc.entryPointCount; i++)
+        {
+            if (desc.slangGlobalScope)
+            {
+                slang::IComponentType* entryPointComponents[2] = {
+                    desc.slangGlobalScope, desc.slangEntryPoints[i]};
+                ComPtr<slang::IComponentType> linkedEntryPoint;
+                session->createCompositeComponentType(
+                    entryPointComponents, 2, linkedEntryPoint.writeRef());
+                linkedEntryPoints.add(linkedEntryPoint);
+            }
+            else
+            {
+                linkedEntryPoints.add(ComPtr<slang::IComponentType>(desc.slangEntryPoints[i]));
+            }
+        }
+        linkedProgram = desc.slangGlobalScope;
+    }
+}
+
 Result RendererBase::maybeSpecializePipeline(
     PipelineStateBase* currentPipeline,
     ShaderObjectBase* rootObject,
@@ -766,11 +826,11 @@ Result RendererBase::maybeSpecializePipeline(
             auto unspecializedProgram = static_cast<ShaderProgramBase*>(pipelineType == PipelineType::Compute
                 ? currentPipeline->desc.compute.program
                 : currentPipeline->desc.graphics.program);
-            auto unspecializedProgramLayout = unspecializedProgram->slangProgram->getLayout();
+            auto unspecializedProgramLayout = unspecializedProgram->linkedProgram->getLayout();
 
             ComPtr<slang::IComponentType> specializedComponentType;
             ComPtr<slang::IBlob> diagnosticBlob;
-            auto compileRs = unspecializedProgram->slangProgram->specialize(
+            auto compileRs = unspecializedProgram->linkedProgram->specialize(
                 specializationArgs.components.getArrayView().getBuffer(),
                 specializationArgs.getCount(),
                 specializedComponentType.writeRef(),
@@ -784,10 +844,18 @@ Result RendererBase::maybeSpecializePipeline(
             }
             SLANG_RETURN_ON_FAIL(compileRs);
 
-            // Now create specialized shader program using compiled binaries.
+            // Now create the specialized shader program using compiled binaries.
             ComPtr<IShaderProgram> specializedProgram;
-            IShaderProgram::Desc specializedProgramDesc = {};
-            specializedProgramDesc.slangProgram = specializedComponentType;
+            IShaderProgram::Desc specializedProgramDesc = unspecializedProgram->desc;
+            specializedProgramDesc.slangGlobalScope = specializedComponentType;
+
+            if (specializedProgramDesc.linkingStyle == IShaderProgram::LinkingStyle::SingleProgram)
+            {
+                // When linking style is GraphicsCompute, the specialized global scope already contains
+                // entry-points, so we do not need to supply them again when creating the specialized
+                // pipeline.
+                specializedProgramDesc.entryPointCount = 0;
+            }
             SLANG_RETURN_ON_FAIL(createProgram(specializedProgramDesc, specializedProgram.writeRef()));
 
             // Create specialized pipeline state.
@@ -904,5 +972,17 @@ Result ShaderTableBase::init(const IShaderTable::Desc& desc)
     return SLANG_OK;
 }
 
+bool isDepthFormat(Format format)
+{
+    switch (format)
+    {
+    case Format::D16_UNORM:
+    case Format::D32_FLOAT:
+        return true;
+    default:
+        return false;
+    }
+}
+
 } // namespace gfx
 
diff --git a/tools/gfx/renderer-shared.h b/tools/gfx/renderer-shared.h
index a23cf9dfe..e6f54ae10 100644
--- a/tools/gfx/renderer-shared.h
+++ b/tools/gfx/renderer-shared.h
@@ -973,7 +973,35 @@ public:
     SLANG_COM_OBJECT_IUNKNOWN_ALL
     IShaderProgram* getInterface(const Slang::Guid& guid);
 
-    ComPtr<slang::IComponentType> slangProgram;
+    Desc desc;
+
+    Slang::ComPtr<slang::IComponentType> slangGlobalScope;
+    Slang::List<ComPtr<slang::IComponentType>> slangEntryPoints;
+
+    // Linked program when linkingStyle is GraphicsCompute, or the original global scope
+    // when linking style is RayTracing.
+    Slang::ComPtr<slang::IComponentType> linkedProgram;
+
+    // Linked program for each entry point when linkingStyle is RayTracing.
+    Slang::List<Slang::ComPtr<slang::IComponentType>> linkedEntryPoints;
+
+    void init(const IShaderProgram::Desc& desc);
+
+    bool isSpecializable()
+    {
+        if (slangGlobalScope->getSpecializationParamCount() != 0)
+        {
+            return true;
+        }
+        for (auto& entryPoint : slangEntryPoints)
+        {
+            if (entryPoint->getSpecializationParamCount() != 0)
+            {
+                return true;
+            }
+        }
+        return false;
+    }
 };
 
 class InputLayoutBase
@@ -1369,6 +1397,8 @@ public:
     Slang::Dictionary<slang::TypeReflection*, Slang::RefPtr<ShaderObjectLayoutBase>> m_shaderObjectLayoutCache;
 };
 
+bool isDepthFormat(Format format);
+
 IDebugCallback*& _getDebugCallback();
 IDebugCallback* _getNullDebugCallback();
 inline IDebugCallback* getDebugCallback()
diff --git a/tools/gfx/vulkan/render-vk.cpp b/tools/gfx/vulkan/render-vk.cpp
index 7b6b31366..80084ae7b 100644
--- a/tools/gfx/vulkan/render-vk.cpp
+++ b/tools/gfx/vulkan/render-vk.cpp
@@ -94,7 +94,10 @@ public:
     virtual SLANG_NO_THROW Result SLANG_MCALL createTextureView(
         ITextureResource* texture, IResourceView::Desc const& desc, IResourceView** outView) override;
     virtual SLANG_NO_THROW Result SLANG_MCALL createBufferView(
-        IBufferResource* buffer, IResourceView::Desc const& desc, IResourceView** outView) override;
+        IBufferResource* buffer,
+        IBufferResource* counterBuffer,
+        IResourceView::Desc const& desc,
+        IResourceView** outView) override;
 
     virtual SLANG_NO_THROW Result SLANG_MCALL createInputLayout(
         IInputLayout::Desc const& desc,
@@ -8075,7 +8078,11 @@ Result VKDevice::getFormatSupportedResourceStates(Format format, ResourceStateSe
     return SLANG_OK;
 }
 
-Result VKDevice::createBufferView(IBufferResource* buffer, IResourceView::Desc const& desc, IResourceView** outView)
+Result VKDevice::createBufferView(
+    IBufferResource* buffer,
+    IBufferResource* counterBuffer,
+    IResourceView::Desc const& desc,
+    IResourceView** outView)
 {
     auto resourceImpl = (BufferResourceImpl*) buffer;
 
@@ -8244,15 +8251,16 @@ Result VKDevice::createProgram(
     const IShaderProgram::Desc& desc, IShaderProgram** outProgram, ISlangBlob** outDiagnosticBlob)
 {
     RefPtr<ShaderProgramImpl> shaderProgram = new ShaderProgramImpl(this);
-    shaderProgram->slangProgram = desc.slangProgram;
+    shaderProgram->init(desc);
+
     m_deviceObjectsWithPotentialBackReferences.add(shaderProgram);
 
     RootShaderObjectLayout::create(
         this,
-        desc.slangProgram,
-        desc.slangProgram->getLayout(),
+        shaderProgram->linkedProgram,
+        shaderProgram->linkedProgram->getLayout(),
         shaderProgram->m_rootObjectLayout.writeRef());
-    if (desc.slangProgram->getSpecializationParamCount() != 0)
+    if (shaderProgram->isSpecializable())
     {
         // For a specializable program, we don't invoke any actual slang compilation yet.
         returnComPtr(outProgram, shaderProgram);
@@ -8260,15 +8268,15 @@ Result VKDevice::createProgram(
     }
 
     // For a fully specialized program, create `VkShaderModule`s for each shader stage.
-    auto programReflection = desc.slangProgram->getLayout();
-    for (SlangUInt i = 0; i < programReflection->getEntryPointCount(); i++)
+    auto compileShader = [&](slang::EntryPointReflection* entryPointInfo,
+                             slang::IComponentType* component,
+                             SlangInt entryPointIndex)
     {
-        auto entryPointInfo = programReflection->getEntryPointByIndex(i);
         auto stage = entryPointInfo->getStage();
         ComPtr<ISlangBlob> kernelCode;
         ComPtr<ISlangBlob> diagnostics;
-        auto compileResult = desc.slangProgram->getEntryPointCode(
-            (SlangInt)i, 0, kernelCode.writeRef(), diagnostics.writeRef());
+        auto compileResult = component->getEntryPointCode(
+            entryPointIndex, 0, kernelCode.writeRef(), diagnostics.writeRef());
         if (diagnostics)
         {
             getDebugCallback()->handleMessage(
@@ -8295,6 +8303,27 @@ Result VKDevice::createProgram(
             (VkShaderStageFlagBits)VulkanUtil::getShaderStage(stage),
             shaderModule));
         shaderProgram->m_modules.add(shaderModule);
+        return SLANG_OK;
+    };
+    if (shaderProgram->linkedEntryPoints.getCount() == 0)
+    {
+        // If the user does not explicitly specify entry point components, find them from
+        // `linkedEntryPoints`.
+        auto programReflection = shaderProgram->linkedProgram->getLayout();
+        for (SlangUInt i = 0; i < programReflection->getEntryPointCount(); i++)
+        {
+            auto entryPointInfo = programReflection->getEntryPointByIndex(i);
+            SLANG_RETURN_ON_FAIL(compileShader(entryPointInfo, shaderProgram->linkedProgram, (SlangInt)i));
+        }
+    }
+    else
+    {
+        // If the user specifies entry point components via the separated entry point array, compile
+        // code from there.
+        for (auto& entryPoint : shaderProgram->linkedEntryPoints)
+        {
+            SLANG_RETURN_ON_FAIL(compileShader(entryPoint->getLayout()->getEntryPointByIndex(0), entryPoint, 0));
+        }
     }
     returnComPtr(outProgram, shaderProgram);
     return SLANG_OK;
diff --git a/tools/platform/gui.cpp b/tools/platform/gui.cpp
index 37e69114a..84d458e71 100644
--- a/tools/platform/gui.cpp
+++ b/tools/platform/gui.cpp
@@ -98,7 +98,7 @@ GUI::GUI(
 #if 0
     gfx::IShaderProgram::Desc programDesc = {};
     programDesc.pipelineType = gfx::PipelineType::Graphics;
-    programDesc.slangProgram = slangProgram;
+    programDesc.slangGlobalScope = slangGlobalScope;
     program = device->createProgram(programDesc);
 #endif
     InputElementDesc inputElements[] = {
diff --git a/tools/render-test/render-test-main.cpp b/tools/render-test/render-test-main.cpp
index d36ee8a86..4b46f8115 100644
--- a/tools/render-test/render-test-main.cpp
+++ b/tools/render-test/render-test-main.cpp
@@ -213,9 +213,7 @@ struct AssignValsFromLayoutContext
         viewDesc.type = IResourceView::Type::UnorderedAccess;
         viewDesc.format = srcBuffer.format;
         viewDesc.bufferElementSize = srcVal->bufferDesc.stride;
-        auto bufferView = device->createBufferView(
-            bufferResource,
-            viewDesc);
+        auto bufferView = device->createBufferView(bufferResource, nullptr, viewDesc);
         dstCursor.setResource(bufferView);
         maybeAddOutput(dstCursor, srcVal, bufferResource);
 
diff --git a/tools/render-test/slang-support.cpp b/tools/render-test/slang-support.cpp
index 8240212ec..ed7bb752e 100644
--- a/tools/render-test/slang-support.cpp
+++ b/tools/render-test/slang-support.cpp
@@ -54,13 +54,13 @@ void ShaderCompilerUtil::Output::set(
     slang::IComponentType*              inSlangProgram)
 {
     slangProgram = inSlangProgram;
-    desc.slangProgram = inSlangProgram;
+    desc.slangGlobalScope = inSlangProgram;
 }
 
 void ShaderCompilerUtil::Output::reset()
 {
     {
-        desc.slangProgram = nullptr;
+        desc.slangGlobalScope = nullptr;
     }
 
     session = nullptr;
@@ -313,7 +313,7 @@ void ShaderCompilerUtil::Output::reset()
         SLANG_RETURN_ON_FAIL(_compileProgramImpl(session, options, input, request, out));
 
         out.m_extraRequestForReflection = slangOutput.getRequestForReflection();
-        out.desc.slangProgram = slangOutput.desc.slangProgram;
+        out.desc.slangGlobalScope = slangOutput.desc.slangGlobalScope;
         slangOutput.m_requestForKernels = nullptr;
 
         return SLANG_OK;
diff --git a/tools/slang-unit-test/unit-test-chunked-list.cpp b/tools/slang-unit-test/unit-test-chunked-list.cpp
new file mode 100644
index 000000000..af4f597b2
--- /dev/null
+++ b/tools/slang-unit-test/unit-test-chunked-list.cpp
@@ -0,0 +1,41 @@
+// unit-test-path.cpp
+
+#include "source/core/slang-basic.h"
+#include "source/core/slang-chunked-list.h"
+#include "tools/unit-test/slang-unit-test.h"
+
+using namespace Slang;
+
+template<typename T>
+static bool _checkArrayView(ArrayView<T> v0, ArrayView<T> v1)
+{
+    if (v0.getCount() != v1.getCount())
+        return false;
+    for (Index i = 0; i < v0.getCount(); i++)
+        if (v0[i] != v1[i])
+            return false;
+    return true;
+}
+
+SLANG_UNIT_TEST(chunkedList)
+{
+    {
+        ChunkedList<String> stringList;
+        List<String*> ptrs;
+        for (int i = 0; i < 256; i++)
+        {
+            ptrs.add(stringList.add(String(i)));
+        }
+        SLANG_CHECK(stringList.getCount() == 256);
+        SLANG_CHECK(*(ptrs[128]) == "128");
+
+        stringList.clearAndDeallocate();
+        ptrs.clear();
+        for (int i = 0; i < 64; i++)
+        {
+            ptrs.add(stringList.add(String(i)));
+        }
+        SLANG_CHECK(stringList.getCount() == 64);
+        SLANG_CHECK(*(ptrs[32]) == "32");
+    }
+}
author	Yong He <yonghe@outlook.com>	2022-02-09 15:30:38 -0800
committer	GitHub <noreply@github.com>	2022-02-09 15:30:38 -0800
commit	b8982fcf43b86c1e39dcc3dd19bff2821633eda6 (patch)
tree	0d66dbf46b50e760cce4aee232bd6a020976e6fb
parent	59f3fdc0a372d19ce4e989514ee3e9ecbcbf234c (diff)