[Openmp-commits] [llvm] [openmp] [Offload] Allow to record kernel launch stack traces (PR #100472)

[Johannes Doerfert via Openmp-commits]

================
@@ -3480,6 +3508,30 @@ void *AMDGPUDeviceTy::allocate(size_t Size, void *, TargetAllocTy Kind) {
   return Alloc;
 }
 
+void AMDGPUQueueTy::callbackError(hsa_status_t Status, hsa_queue_t *Source,
+                                  void *Data) {
+  auto &AMDGPUDevice = *reinterpret_cast<AMDGPUDeviceTy *>(Data);
+
+  if (Status == HSA_STATUS_ERROR_EXCEPTION) {
+    auto KernelTraceInfoRecord =
+        AMDGPUDevice.KernelLaunchTraces.getExclusiveAccessor();
+    std::function<bool(__tgt_async_info &)> AsyncInfoWrapperMatcher =
+        [=](__tgt_async_info &AsyncInfo) {
+          auto *Stream = reinterpret_cast<AMDGPUStreamTy *>(AsyncInfo.Queue);
+          if (!Stream)
+            return false;
+          if (!Stream->getQueue())
+            return false;
+          return Stream->getQueue()->Queue == Source;
----------------
jdoerfert wrote:

Will do


https://github.com/llvm/llvm-project/pull/100472