Skip to content
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

Forward-merge branch-24.10 into branch-24.12 #16854

Merged
merged 1 commit into from
Sep 19, 2024
Merged

Forward-merge branch-24.10 into branch-24.12 #16854

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 12 additions & 4 deletions cpp/src/utilities/stream_pool.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -132,6 +132,13 @@ struct cuda_event {
cuda_event() { CUDF_CUDA_TRY(cudaEventCreateWithFlags(&e_, cudaEventDisableTiming)); }
virtual ~cuda_event() { CUDF_ASSERT_CUDA_SUCCESS(cudaEventDestroy(e_)); }

// Moveable but not copyable.
cuda_event(const cuda_event&) = delete;
cuda_event& operator=(const cuda_event&) = delete;

cuda_event(cuda_event&&) = default;
cuda_event& operator=(cuda_event&&) = default;

operator cudaEvent_t() { return e_; }

private:
Expand All @@ -147,11 +154,12 @@ struct cuda_event {
*/
cudaEvent_t event_for_thread()
{
thread_local std::vector<std::unique_ptr<cuda_event>> thread_events(get_num_cuda_devices());
// The program may crash if this function is called from the main thread and user application
// subsequently calls cudaDeviceReset().
// As a workaround, here we intentionally disable RAII and leak cudaEvent_t.
thread_local std::vector<cuda_event*> thread_events(get_num_cuda_devices());
auto const device_id = get_current_cuda_device();
if (not thread_events[device_id.value()]) {
thread_events[device_id.value()] = std::make_unique<cuda_event>();
}
if (not thread_events[device_id.value()]) { thread_events[device_id.value()] = new cuda_event(); }
return *thread_events[device_id.value()];
}

Expand Down
Loading