Improvements around C++ code generation (#1396)

* * Remove UniformState and UniformEntryPointParams types
* Put all output C++ source in an anonymous namespace
* If SLANG_PRELUDE_NAMESPACE is set, make what it defines available in generated file.

* Fix signature issue in performance-profile.slang

* Context -> KernelContext to avoid ambiguity.

* Fix issues around dynamic dispatch and anonymous namespace.

* Fix typo.

7 files changed, 43 insertions, 28 deletions

diff --git a/examples/cpu-hello-world/main.cpp b/examples/cpu-hello-world/main.cpp
index 6e06f64bc..4302e6069 100644
--- a/examples/cpu-hello-world/main.cpp
+++ b/examples/cpu-hello-world/main.cpp
@@ -184,7 +184,7 @@ static SlangResult _innerMain(int argc, char** argv)
     // We don't have any entry point parameters so that's passed as NULL
     // We need to cast our definition of the uniform state to the undefined CPPPrelude::UniformState as
     // that type is just a name to indicate what kind of thing needs to be passed in.
-    func(&varyingInput, NULL, (CPPPrelude::UniformState*)&uniformState);
+    func(&varyingInput, NULL, &uniformState);
 
     // bufferContents holds the output
 
diff --git a/prelude/slang-cpp-prelude.h b/prelude/slang-cpp-prelude.h
index 2e680f824..7da613024 100644
--- a/prelude/slang-cpp-prelude.h
+++ b/prelude/slang-cpp-prelude.h
@@ -38,6 +38,4 @@
 #   define SLANG_UNROLL
 #endif
 
-#endif
-
-struct Context;
\ No newline at end of file
+#endif
\ No newline at end of file
diff --git a/prelude/slang-cpp-types.h b/prelude/slang-cpp-types.h
index 311fa667d..7e951fd63 100644
--- a/prelude/slang-cpp-types.h
+++ b/prelude/slang-cpp-types.h
@@ -822,14 +822,11 @@ struct ComputeVaryingInput
     uint3 endGroupID;       ///< Non inclusive end groupID
 };
 
-/* Type that defines the uniform entry point params. The actual content of this type is dependent on the entry point parameters, and can be
-found via reflection or defined such that it matches the shader appropriately.
-*/
-struct UniformEntryPointParams;
-struct UniformState;
+// The uniformEntryPointParams and uniformState must be set to structures that match layout that the kernel expects.
+// This can be determined via reflection for example.
 
-typedef void(*ComputeThreadFunc)(ComputeThreadVaryingInput* varyingInput, UniformEntryPointParams* uniformEntryPointParams, UniformState* uniformState);
-typedef void(*ComputeFunc)(ComputeVaryingInput* varyingInput, UniformEntryPointParams* uniformEntryPointParams, UniformState* uniformState);
+typedef void(*ComputeThreadFunc)(ComputeThreadVaryingInput* varyingInput, void* uniformEntryPointParams, void* uniformState);
+typedef void(*ComputeFunc)(ComputeVaryingInput* varyingInput, void* uniformEntryPointParams, void* uniformState);
 
 #ifdef SLANG_PRELUDE_NAMESPACE
 }
diff --git a/source/slang/slang-emit-cpp.cpp b/source/slang/slang-emit-cpp.cpp
index c31ef3bc7..53a110632 100644
--- a/source/slang/slang-emit-cpp.cpp
+++ b/source/slang/slang-emit-cpp.cpp
@@ -1612,7 +1612,7 @@ void CPPSourceEmitter::_emitWitnessTableDefinitions()
                     m_writer->emit(",\n");
                 else
                     isFirstEntry = false;
-                m_writer->emit("&Context::");
+                m_writer->emit("&KernelContext::");
                 m_writer->emit(getName(funcVal));
             }
             else
@@ -1660,7 +1660,7 @@ void CPPSourceEmitter::_maybeEmitWitnessTableTypeDefinition(
             else
                 isFirstEntry = false;
             emitType(funcVal->getResultType());
-            m_writer->emit(" (Context::*");
+            m_writer->emit(" (KernelContext::*");
             m_writer->emit(getName(entry->requirementKey.get()));
             m_writer->emit(")");
             m_writer->emit("(");
@@ -2134,6 +2134,20 @@ void CPPSourceEmitter::emitPreprocessorDirectivesImpl()
 
     writer->emit("\n");
 
+    
+    if (m_target == CodeGenTarget::CPPSource)
+    {
+        // Put all into an anonymous namespace
+        // This includes any generated types, and generated intrinsics
+
+        m_writer->emit("namespace { // anonymous \n\n");
+        m_writer->emit("#ifdef SLANG_PRELUDE_NAMESPACE\n");
+        m_writer->emit("using namespace SLANG_PRELUDE_NAMESPACE;\n");
+        m_writer->emit("#endif\n\n");
+
+        m_writer->emit("struct KernelContext;\n\n");
+    }
+
     if (m_target == CodeGenTarget::CSource)
     {
         // For C output we need to emit type definitions.
@@ -2169,7 +2183,6 @@ void CPPSourceEmitter::emitPreprocessorDirectivesImpl()
         {
             _maybeEmitSpecializedOperationDefinition(intrinsic);
         }
-        
     }
 }
 
@@ -2293,14 +2306,14 @@ void CPPSourceEmitter::_emitEntryPointDefinitionStart(IRFunc* func, IRGlobalPara
 
     m_writer->emit("(");
     m_writer->emit(varyingTypeName);
-    m_writer->emit("* varyingInput, UniformEntryPointParams* params, UniformState* uniformState)");
+    m_writer->emit("* varyingInput, void* params, void* uniformState)");
     emitSemantics(func);
     m_writer->emit("\n{\n");
 
     m_writer->indent();
     // Initialize when constructing so that globals are zeroed
-    m_writer->emit("Context context = {};\n");
-    m_writer->emit("context.uniformState = uniformState;\n");
+    m_writer->emit("KernelContext context = {};\n");
+    m_writer->emit("context.uniformState = (UniformState*)uniformState;\n");
     
     if (entryPointGlobalParams)
     {
@@ -2590,11 +2603,11 @@ void CPPSourceEmitter::emitModuleImpl(IRModule* module)
 
     List<EmitAction> actions;
     computeEmitActions(module, actions);
-
+    
     _emitForwardDeclarations(actions);
 
     IRGlobalParam* entryPointGlobalParams = nullptr;
-
+    
     // Output the global parameters in a 'UniformState' structure
     {
         m_writer->emit("struct UniformState\n{\n");
@@ -2605,15 +2618,14 @@ void CPPSourceEmitter::emitModuleImpl(IRModule* module)
         m_writer->dedent();
         m_writer->emit("\n};\n\n");
     }
-
+    
     // Output the 'Context' which will be used for execution
     {
-        m_writer->emit("struct Context\n{\n");
+        m_writer->emit("struct KernelContext\n{\n");
         m_writer->indent();
 
         m_writer->emit("UniformState* uniformState;\n");
 
-        
         m_writer->emit("uint3 dispatchThreadID;\n");
 
         //if (m_semanticUsedFlags & SemanticUsedFlag::GroupID)
@@ -2659,12 +2671,19 @@ void CPPSourceEmitter::emitModuleImpl(IRModule* module)
         }
 
         m_writer->dedent();
-        m_writer->emit("};\n\n");
+        m_writer->emit("};\n\n");   
     }
 
     // Emit all witness table definitions.
     _emitWitnessTableDefinitions();
 
+    if (m_target == CodeGenTarget::CPPSource)
+    {
+        // Need to close the anonymous namespace when outputting for C++
+
+        m_writer->emit("} // anonymous\n\n");
+    }
+
      // Finally we need to output dll entry points
 
     for (auto action : actions)
diff --git a/source/slang/slang-emit-cuda.cpp b/source/slang/slang-emit-cuda.cpp
index 25b06027d..06bbbea96 100644
--- a/source/slang/slang-emit-cuda.cpp
+++ b/source/slang/slang-emit-cuda.cpp
@@ -866,10 +866,10 @@ void CUDASourceEmitter::emitModuleImpl(IRModule* module)
                     //
                     // At the binary level, our generated CUDA compute kernels will take
                     // two pointer parameters: the first points to the per-entry-point
-                    // `uniform` parameter data, and the second poinst to the global-scope
+                    // `uniform` parameter data, and the second points to the global-scope
                     // parameter data (if any).
                     //
-                    m_writer->emit("(UniformEntryPointParams* entryPointShaderParameters, UniformState* uniformState)");
+                    m_writer->emit("(void* entryPointShaderParameters, void* uniformState)");
                 }
                 else
                 {
diff --git a/tests/compute/performance-profile.slang b/tests/compute/performance-profile.slang
index bf67a4478..22f73e075 100644
--- a/tests/compute/performance-profile.slang
+++ b/tests/compute/performance-profile.slang
@@ -44,8 +44,9 @@ static void _calc(const RWStructuredBuffer<float>& buf, int start, int end)
 }
 
 SLANG_PRELUDE_EXPORT
-void computeMain(ComputeVaryingInput* varyingInput, UniformEntryPointParams* params, LocalUniformState* uniformState)
+void computeMain(ComputeVaryingInput* varyingInput, void* inParams, void* inUniformState)
 {
+    LocalUniformState* uniformState = (LocalUniformState*)inUniformState;
     _calc(uniformState->outputBuffer_0, varyingInput->startGroupID.x * 16, varyingInput->endGroupID.x * 16);
 }
  
diff --git a/tools/render-test/cpu-compute-util.cpp b/tools/render-test/cpu-compute-util.cpp
index e8b9e8b32..dc3ea1afa 100644
--- a/tools/render-test/cpu-compute-util.cpp
+++ b/tools/render-test/cpu-compute-util.cpp
@@ -619,8 +619,8 @@ static SlangResult _newTexture(const InputTextureDesc& desc, slang::TypeLayoutRe
 
 /* static */SlangResult CPUComputeUtil::execute(const ExecuteInfo& info)
 {
-    CPPPrelude::UniformState* uniformState = (CPPPrelude::UniformState*)info.m_uniformState;
-    CPPPrelude::UniformEntryPointParams* uniformEntryPointParams = (CPPPrelude::UniformEntryPointParams*)info.m_uniformEntryPointParams;
+    void* uniformState = info.m_uniformState;
+    void* uniformEntryPointParams = info.m_uniformEntryPointParams;
 
     switch (info.m_style)
     {