[flang-commits] [flang] [flang][cuda] Handle launch of cooperative kernel (PR #124362)

[Valentin Clement バレンタイン クレメン via flang-commits]

================
@@ -151,4 +151,69 @@ void RTDEF(CUFLaunchClusterKernel)(const void *kernel, intptr_t clusterX,
   CUDA_REPORT_IF_ERROR(cudaLaunchKernelExC(&config, kernel, params));
 }
 
+void RTDEF(CUFLaunchCooperativeKernel)(const void *kernel, intptr_t gridX,
+    intptr_t gridY, intptr_t gridZ, intptr_t blockX, intptr_t blockY,
+    intptr_t blockZ, int32_t smem, void **params, void **extra) {
----------------
clementval wrote:

No. Stream is not handled yet. I'll add it later to the three functions when we have proper support. This extra arg is modeled on the mlir runtime entry point. It's currently unused. I would like to keep it for consistency in the 3 functions. If in the end we don't need it I'll remove it from the3 functions at the same time. 

https://github.com/llvm/llvm-project/pull/124362