[OpenMP][libomptarget] Fixed an issue that device sync is skipped if the kernel doesn't have any argument

Currently if there is not kernel argument, device synchronization will
be skipped. This can lead to two issues:
1. If there is any device error, it will not be captured;
2. The target region might end before the kernel is done, which is not spec
   conformant.

The test added in this patch only runs on NVPTX platform, although it will not
be executed by Phab at all. It also requires `not` which is not available on most
systems.

Reviewed By: jdoerfert

Differential Revision: https://reviews.llvm.org/D96067

openmp/libomptarget/src/omptarget.cpp

@@ -451,6 +451,17 @@ struct DeallocTgtPtrInfo {
       : HstPtrBegin(HstPtr), DataSize(Size), ForceDelete(ForceDelete),
         HasCloseModifier(HasCloseModifier) {}
 };
+
+/// Synchronize device
+static int syncDevice(DeviceTy &Device, __tgt_async_info *AsyncInfo) {
+  assert(AsyncInfo && AsyncInfo->Queue && "Invalid AsyncInfo");
+  if (Device.synchronize(AsyncInfo) != OFFLOAD_SUCCESS) {
+    REPORT("Failed to synchronize device.\n");
+    return OFFLOAD_FAIL;
+  }
+
+  return OFFLOAD_SUCCESS;
+}
 } // namespace
 
 /// Internal function to undo the mapping and retrieve the data from the device.
@@ -631,11 +642,9 @@ int targetDataEnd(ident_t *loc, DeviceTy &Device, int32_t ArgNum,
   // AsyncInfo->Queue will not be nullptr, so again, we don't need to
   // synchronize.
   if (AsyncInfo && AsyncInfo->Queue) {
-    Ret = Device.synchronize(AsyncInfo);
-    if (Ret != OFFLOAD_SUCCESS) {
-      REPORT("Failed to synchronize device.\n");
+    Ret = syncDevice(Device, AsyncInfo);
+    if (Ret != OFFLOAD_SUCCESS)
       return OFFLOAD_FAIL;
-    }
   }
 
   // Deallocate target pointer
@@ -1307,6 +1316,11 @@ int target(ident_t *loc, int64_t DeviceId, void *HostPtr, int32_t ArgNum,
       REPORT("Failed to process data after launching the kernel.\n");
       return OFFLOAD_FAIL;
     }
+  } else if (AsyncInfo.Queue) {
+    // If ArgNum is zero, but AsyncInfo.Queue is valid, then the kernel doesn't
+    // hava any argument, and the device supports async operations, so we need a
+    // sync at this point.
+    return syncDevice(Device, &AsyncInfo);
   }
 
   return OFFLOAD_SUCCESS;

openmp/libomptarget/test/offloading/assert.cpp

@@ -0,0 +1,8 @@
+// RUN: %libomptarget-compilexx-nvptx64-nvidia-cuda && %libomptarget-run-fail-nvptx64-nvidia-cuda
+
+int main(int argc, char *argv[]) {
+#pragma omp target
+  { __builtin_trap(); }
+
+  return 0;
+}