Initial attempt to enable CUDA dynamic dispatch codegen (#1549)

* Front-load cuda module loading to fill in RTTI pointers.

* Enable dynamic dispatch codegen for CUDA.

1 files changed, 9 insertions, 1 deletions

diff --git a/source/slang/slang-emit-cuda.cpp b/source/slang/slang-emit-cuda.cpp
index acd913865..ecc3eb68b 100644
--- a/source/slang/slang-emit-cuda.cpp
+++ b/source/slang/slang-emit-cuda.cpp
@@ -300,7 +300,7 @@ String CUDASourceEmitter::generateEntryPointNameImpl(IREntryPointDecoration* ent
     // > The input PTX should include one or more NVIDIA OptiX programs.
     // > The type of program affects how the program can be used during
     // > the execution of the pipeline. These program types are specified
-    // by prefixing the program’s name with the following:
+    // by prefixing the program name with the following:
     //
     // >    Program type        Function name prefix
     CASE(   RayGeneration,      __raygen__);
@@ -322,6 +322,11 @@ String CUDASourceEmitter::generateEntryPointNameImpl(IREntryPointDecoration* ent
     return globalSymbolName;
 }
 
+void CUDASourceEmitter::emitGlobalRTTISymbolPrefix()
+{
+    m_writer->emit("__device__");
+}
+
 void CUDASourceEmitter::emitCall(const HLSLIntrinsic* specOp, IRInst* inst, const IRUse* operands, int numOperands, const EmitOpInfo& inOuterPrec)
 {
     switch (specOp->op)
@@ -740,6 +745,9 @@ void CUDASourceEmitter::emitModuleImpl(IRModule* module)
     // TODO(JS): We may need to generate types (for example for matrices)
 
     CLikeSourceEmitter::emitModuleImpl(module);
+
+    // Emit all witness table definitions.
+    _emitWitnessTableDefinitions();
 }