adds sync to flaky test_events_multi_gpu_query (#26231)

Mike Ruberry · facebook-github-bot · commit 4160b8cd77c6 · 2019-09-14T00:34:44.000-07:00
Summary: This test can sometimes fail in CI. I suspect this flakiness is because the test asks a CUDA stream to record an event, fails to synchronize the CPU with that stream, then checks if the event is recorded on the CPU. There is no guarantee this will have happened. This one-line change preserves the intent of the test while ensuring the GPU has recorded the event before the CPU queries it. Pull Request resolved: #26231 Differential Revision: D17382110 Pulled By: mruberry fbshipit-source-id: 35b701f87f41c24b208aafde48bf10e1a54de059
diff --git a/test/test_cuda.py b/test/test_cuda.py
@@ -1962,6 +1962,7 @@ def test_events_multi_gpu_query(self):
         with torch.cuda.device(d0):
             s0 = torch.cuda.current_stream()
             e0 = s0.record_event()
+            s0.synchronize()
 
         with torch.cuda.device(d1):
             s1 = torch.cuda.current_stream()
