Skip to content

[fleet_executor] Fix overlap hang #38024

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Dec 10, 2021
Merged

[fleet_executor] Fix overlap hang #38024

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
90e6af5
fix overlap hang
FeixLiu Dec 9, 2021
76a5b09
add annotation
FeixLiu Dec 9, 2021
File filter

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 5 additions & 0 deletions paddle/fluid/distributed/fleet_executor/carrier.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,6 +75,11 @@ class Carrier final {

bool IsInit() const;

// NOTE: This mutex will be used in interceptor's RunOps function.
// This mutex is used for avoiding forward ops and backward ops run
// simultaneously, which will lead to a random hang for some sync ops.
std::mutex run;

DISABLE_COPY_AND_ASSIGN(Carrier);

private:
Expand Down
3 changes: 3 additions & 0 deletions paddle/fluid/distributed/fleet_executor/compute_interceptor.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,6 +13,7 @@
// limitations under the License.

#include "paddle/fluid/distributed/fleet_executor/compute_interceptor.h"
#include "paddle/fluid/distributed/fleet_executor/carrier.h"

#include "paddle/fluid/distributed/fleet_executor/task_node.h"
#include "paddle/fluid/framework/executor_gc_helper.h"
Expand Down Expand Up @@ -169,6 +170,8 @@ void ComputeInterceptor::ReplyCompletedToUpStream() {
}

void ComputeInterceptor::RunOps() {
Carrier& carrier_instance = Carrier::Instance();
std::unique_lock<std::mutex> lock(carrier_instance.run);
VLOG(3) << "ComputeInterceptor " << interceptor_id_ << " running ops for the "
<< step_ + 1 << " time.";
for (auto op : node_->ops()) {
Expand Down