Proper bulking of ops not using FCompute (#15272)

* Less syncs * Trigger CI
apache · Jun 20, 2019 · c45d23b · c45d23b
1 parent 2de0db0
commit c45d23b
Showing 1 changed file with 3 additions and 3 deletions.
diff --git a/src/imperative/imperative_utils.h b/src/imperative/imperative_utils.h
@@ -464,7 +464,7 @@ inline void PushFComputeEx(const FComputeEx& fn,
       InvalidateOutputs(outputs, req);
 #endif
       fn(attrs, opctx, inputs, req, outputs);
-      if (ctx.dev_mask() == gpu::kDevMask && exec_type == ExecType::kSync) {
+      if (ctx.dev_mask() == gpu::kDevMask && exec_type == ExecType::kSync && !rctx.is_bulk) {
         rctx.get_stream<gpu>()->Wait();
       }
     };
@@ -512,7 +512,7 @@ inline void PushOperator(const OpStatePtr& state,
 #endif
       fcompute_ex(state, opctx, inputs, req, outputs);
       if (ctx.dev_mask() == gpu::kDevMask && exec_type == ExecType::kSync
-          && rctx.get_stream<gpu>()) {
+          && rctx.get_stream<gpu>() && !rctx.is_bulk) {
         rctx.get_stream<gpu>()->Wait();
       }
     };
@@ -562,7 +562,7 @@ inline void PushOperator(const OpStatePtr& state,
         // post-fcompute fallback, cast to original storage type, if necessary
         CastNonDefaultStorage(post_temp_src, post_temp_dst, opctx, is_gpu);
         if (is_gpu && exec_type == ExecType::kSync
-            && rctx.get_stream<gpu>()) {
+            && rctx.get_stream<gpu>() && !rctx.is_bulk) {
           rctx.get_stream<gpu>()->Wait();
         }
       };