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

[DeviceASAN][bugfix] Not allow concurrent kernel launches across different queue #2679

Merged
merged 4 commits into from
Feb 10, 2025
+9 −0
Merged

[DeviceASAN][bugfix] Not allow concurrent kernel launches across different queue #2679

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
8e01889
add mutex
yingcong-wu Feb 8, 2025
c882de6
comment & format
yingcong-wu Feb 8, 2025
fd60baf
trigger ci
yingcong-wu Feb 8, 2025
508e6b9
move mutex
yingcong-wu Feb 10, 2025
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
7 changes: 7 additions & 0 deletions source/loader/layers/sanitizer/asan/asan_ddi.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -474,6 +474,13 @@ __urdlllocal ur_result_t UR_APICALL urEnqueueKernelLaunch(
/// [out][optional] return an event object that identifies this
/// particular kernel execution instance.
ur_event_handle_t *phEvent) {

// This mutex is to prevent concurrent kernel launches across different queues
// as the DeviceASAN local/private shadow memory does not support concurrent
// kernel launches now.
std::scoped_lock<ur_shared_mutex> Guard(
getAsanInterceptor()->KernelLaunchMutex);

auto pfnKernelLaunch = getContext()->urDdiTable.Enqueue.pfnKernelLaunch;

if (nullptr == pfnKernelLaunch) {
Expand Down
2 changes: 2 additions & 0 deletions source/loader/layers/sanitizer/asan/asan_interceptor.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -354,6 +354,8 @@ class AsanInterceptor {
std::shared_ptr<ShadowMemory>
getOrCreateShadowMemory(ur_device_handle_t Device, DeviceType Type);

ur_shared_mutex KernelLaunchMutex;

private:
ur_result_t updateShadowMemory(std::shared_ptr<ContextInfo> &ContextInfo,
std::shared_ptr<DeviceInfo> &DeviceInfo,
Expand Down
Loading