Support passing macros to hipLaunchKernelGGL

This makes hipLaunchKernelGGL take a variable argument list, that will be
expanded before being fed to hipLaunchKernelGGLInternal.

This is different from 4dfef9a25b.

We try to accomodate the case when a kernel template has multiple
type parameters.

Change-Id: I87577d402c92b0f3b51e298f8293f4065e1f6de8


[ROCm/clr commit: 133251a6e1]

projects/clr/hipamd/include/hip/hcc_detail/hip_runtime.h

@@ -435,10 +435,12 @@ void hipLaunchKernelGGL(F kernel, const dim3& numBlocks, const dim3& dimBlocks,
     hipLaunchKernel(k, numBlocks, dimBlocks, _Args, sharedMemBytes, stream);
 }
 #else
-#define hipLaunchKernelGGL(kernelName, numblocks, numthreads, memperblock, streamId, ...)          \
+#define hipLaunchKernelGGLInternal(kernelName, numBlocks, numThreads, memPerBlock, streamId, ...)  \
     do {                                                                                           \
-        kernelName<<<(numblocks), (numthreads), (memperblock), (streamId)>>>(__VA_ARGS__);         \
+        kernelName<<<(numBlocks), (numThreads), (memPerBlock), (streamId)>>>(__VA_ARGS__);         \
     } while (0)
+
+#define hipLaunchKernelGGL(kernelName, ...)  hipLaunchKernelGGLInternal((kernelName), __VA_ARGS__)
 #endif
 
 #include <hip/hip_runtime_api.h>

projects/clr/hipamd/include/hip/nvcc_detail/hip_runtime.h

@@ -31,11 +31,13 @@ THE SOFTWARE.
 
 typedef int hipLaunchParm;
 
-#define hipLaunchKernelGGL(kernelName, numblocks, numthreads, memperblock, streamId, ...)          \
+#define hipLaunchKernelGGLInternal(kernelName, numBlocks, numThreads, memPerBlock, streamId, ...)  \
     do {                                                                                           \
-        kernelName<<<numblocks, numthreads, memperblock, streamId>>>(__VA_ARGS__);                 \
+        kernelName<<<numBlocks, numThreads, memPerBlock, streamId>>>(__VA_ARGS__);                 \
     } while (0)
 
+#define hipLaunchKernelGGL(kernelName, ...)  hipLaunchKernelGGLInternal((kernelName), __VA_ARGS__)
+
 #define hipReadModeElementType cudaReadModeElementType
 
 #ifdef __CUDA_ARCH__

projects/clr/hipamd/tests/src/kernel/hipLaunchParm.cpp

@@ -571,6 +571,10 @@ __global__ void hipLaunchKernelStructFunc21(
 
 __global__ void vAdd(float* a) {}
 
+template<class T1, class T2>
+__global__ void myKernel(T1 a, T2 b) {}
+
+
 //---
 // Some wrapper macro for testing:
 #define WRAP(...) __VA_ARGS__
@@ -913,6 +917,18 @@ int main() {
     hipLaunchKernelGGL(HIP_KERNEL_NAME(vAdd), dim3(1024), 1, 0, 0, Ad);
     hipLaunchKernelGGL(HIP_KERNEL_NAME(vAdd), dim3(1024), dim3(1), 0, 0, Ad);
 
+    // Test: Passing macro to hipLaunchKernelGGL
+#define KERNEL_CONFIG  dim3(1024), dim3(1), 0, 0
+    hipLaunchKernelGGL(HIP_KERNEL_NAME(vAdd), KERNEL_CONFIG, Ad);
+
+    // Test: Same thing with templates:
+    int a;
+    float b;
+    hipLaunchKernelGGL(HIP_KERNEL_NAME(myKernel<int, float>), KERNEL_CONFIG, a, b);
+
+#define TYPE_PARAM_CONFIG int, float
+    hipLaunchKernelGGL(HIP_KERNEL_NAME(myKernel<TYPE_PARAM_CONFIG>), KERNEL_CONFIG, a, b);
+
     // Test: Passing hipLaunchKernelGGL inside another macro:
     float e0;
     MY_LAUNCH_MACRO(hipLaunchKernelGGL(vAdd, dim3(1024),