experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[Executor] Avoid unnecessary heap allocation in PropagateOutputs().

Browse Source 
PiperOrigin-RevId: 292600542
Change-Id: Ice9602ac275817624ad8cc41bde77ad910fd485f
This commit is contained in:
Derek Murray 2020-01-31 13:01:50 -08:00 committed by TensorFlower Gardener
parent 15fa2a49da
commit 688c4c2842
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
tensorflow/core/common_runtime/executor.cc
View File

@ -2159,7 +2159,7 @@ Status ExecutorState::ProcessOutputs(const NodeItem& item, OpKernelContext* ctx,
void ExecutorState::PropagateOutputs(const TaggedNode& tagged_node, void ExecutorState::PropagateOutputs(const TaggedNode& tagged_node,
const NodeItem* item, EntryVector* outputs, const NodeItem* item, EntryVector* outputs,
TaggedNodeSeq* ready) { TaggedNodeSeq* ready) {
auto activity_handle = absl::make_unique<profiler::TraceMe>( profiler::TraceMe activity(
[&]() { [&]() {
return strings::StrCat( return strings::StrCat(
"ExecutorPropagateOutputs:", item->kernel->name_view(), "ExecutorPropagateOutputs:", item->kernel->name_view(),