[Openmp-commits] [PATCH] D99443: [OpenMP] Reset async stream properly upon failure

[Joseph Huber via Phabricator via Openmp-commits]

jhuber6 created this revision.
jhuber6 added a reviewer: jdoerfert.
jhuber6 added a project: OpenMP.
Herald added subscribers: guansong, yaxunl.
jhuber6 requested review of this revision.
Herald added subscribers: openmp-commits, sstefan1.

If the call to `synchronize` fails, it will currently block the stream indefinitely if execution is continued from this point. Additionally, if the program exits it will trigger an assertion on the non-null value of the async queue and prevent the runtime from printing debugging information.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D99443

Files:
  openmp/libomptarget/plugins/cuda/src/rtl.cpp


Index: openmp/libomptarget/plugins/cuda/src/rtl.cpp
===================================================================
--- openmp/libomptarget/plugins/cuda/src/rtl.cpp
+++ openmp/libomptarget/plugins/cuda/src/rtl.cpp
@@ -1035,13 +1035,6 @@
   int synchronize(const int DeviceId, __tgt_async_info *AsyncInfo) const {
     CUstream Stream = reinterpret_cast<CUstream>(AsyncInfo->Queue);
     CUresult Err = cuStreamSynchronize(Stream);
-    if (Err != CUDA_SUCCESS) {
-      REPORT("Error when synchronizing stream. stream = " DPxMOD
-             ", async info ptr = " DPxMOD "\n",
-             DPxPTR(Stream), DPxPTR(AsyncInfo));
-      CUDA_ERR_STRING(Err);
-      return OFFLOAD_FAIL;
-    }
 
     // Once the stream is synchronized, return it to stream pool and reset
     // AsyncInfo. This is to make sure the synchronization only works for its
@@ -1050,7 +1043,13 @@
                                 reinterpret_cast<CUstream>(AsyncInfo->Queue));
     AsyncInfo->Queue = nullptr;
 
-    return OFFLOAD_SUCCESS;
+    if (Err != CUDA_SUCCESS) {
+      REPORT("Error when synchronizing stream. stream = " DPxMOD
+             ", async info ptr = " DPxMOD "\n",
+             DPxPTR(Stream), DPxPTR(AsyncInfo));
+      CUDA_ERR_STRING(Err);
+    }
+    return (Err == CUDA_SUCCESS) ? OFFLOAD_SUCCESS : OFFLOAD_FAIL;
   }
 };
 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D99443.333630.patch
Type: text/x-patch
Size: 1332 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/openmp-commits/attachments/20210326/b47e2f00/attachment.bin>