From 47740ace289721e61489f6b2b5c196f26250aa3f Mon Sep 17 00:00:00 2001
From: Yancey1989 <yanxu05@baidu.com>
Date: Fri, 7 Dec 2018 17:18:45 +0800
Subject: [PATCH] fix performance

---
 paddle/fluid/framework/details/all_reduce_op_handle.cc | 1 +
 1 file changed, 1 insertion(+)

diff --git a/paddle/fluid/framework/details/all_reduce_op_handle.cc b/paddle/fluid/framework/details/all_reduce_op_handle.cc
index ae203387462..6b7bbf9003a 100644
--- a/paddle/fluid/framework/details/all_reduce_op_handle.cc
+++ b/paddle/fluid/framework/details/all_reduce_op_handle.cc
@@ -107,6 +107,7 @@ void AllReduceOpHandle::RunImpl() {
           PADDLE_ENFORCE(platform::dynload::ncclAllReduce(
               buffer, buffer, numel, static_cast<ncclDataType_t>(dtype),
               ncclSum, comm, stream));
+          if (!nccl_ctxs_->need_group_call_) cudaStreamSynchronize(stream);
         });
       }
 
-- 
GitLab