Skip to content

[SYCL] Fix race that occurs when submitting to single queue in parallel #1872

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
bader merged 14 commits into from
Jul 2, 2020
Merged

[SYCL] Fix race that occurs when submitting to single queue in parallel #1872

Changes from 2 commits
Commits
Show all changes
14 commits
Select commit Hold shift + click to select a range
3ec79ba
[SYCL] Fix race that occurs when submitting to single queue in parallel
alexanderfle Jun 11, 2020
939dc81
[SYCL] Apply comment from review about NewCmd
alexanderfle Jun 11, 2020
edbbcf4
[SYCL] Added distinct lock for enqueue new command
alexanderfle Jun 11, 2020
f1faabe
[SYCL] Apply clang-format
alexanderfle Jun 11, 2020
301a0f1
[SYCL] Apply a single lock in Scheduler::addCG(), change locking in o…
alexanderfle Jun 11, 2020
2870269
[SYCL] Fix hang in Lit tests
alexanderfle Jun 15, 2020
2e4b1d3
[SYCL] Fix hang in Lit with Cuda
alexanderfle Jun 15, 2020
0bd1bf9
[SYCL] Add per-kernel mutex to fix the race when setting kernel param…
alexanderfle Jun 22, 2020
2bdd731
Merge branch 'sycl' into fix_race_submitting_to_single_queue
alexanderfle Jun 22, 2020
fcb4e6c
[SYCL] Use lock_guard to lock mutex
alexanderfle Jun 23, 2020
0dca88f
[SYCL] Use per-kernel mutex for all cases when the kernel is cacheable.
alexanderfle Jun 24, 2020
12502ae
[SYCL] Refactoring of lambda into a separate function
alexanderfle Jun 26, 2020
615ba25
Merge branch 'sycl' into fix_race_submitting_to_single_queue
alexanderfle Jul 2, 2020
486ea37
Merge remote-tracking branch 'origin' into fix_race_submitting_to_sin…
alexanderfle Jul 2, 2020
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
28 changes: 17 additions & 11 deletions sycl/source/detail/scheduler/scheduler.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -63,12 +63,13 @@ void Scheduler::waitForRecordToFinish(MemObjRecord *Record) {

EventImplPtr Scheduler::addCG(std::unique_ptr<detail::CG> CommandGroup,
QueueImplPtr Queue) {
Command *NewCmd = nullptr;
EventImplPtr NewEvent = nullptr;
const bool IsKernel = CommandGroup->getType() == CG::KERNEL;
{
std::unique_lock<std::shared_timed_mutex> Lock(MGraphLock, std::defer_lock);
lockSharedTimedMutex(Lock);

Command *NewCmd = nullptr;
switch (CommandGroup->getType()) {
case CG::UPDATE_HOST:
NewCmd = MGraphBuilder.addCGUpdateHost(std::move(CommandGroup),
Expand All @@ -80,22 +81,27 @@ EventImplPtr Scheduler::addCG(std::unique_ptr<detail::CG> CommandGroup,
default:
NewCmd = MGraphBuilder.addCG(std::move(CommandGroup), std::move(Queue));
}
NewEvent = NewCmd->getEvent();
}

{
std::shared_lock<std::shared_timed_mutex> Lock(MGraphLock);
std::unique_lock<std::shared_timed_mutex> Lock(MGraphLock, std::defer_lock);
lockSharedTimedMutex(Lock);

// TODO: Check if lazy mode.
EnqueueResultT Res;
bool Enqueued = GraphProcessor::enqueueCommand(NewCmd, Res);
if (!Enqueued && EnqueueResultT::SyclEnqueueFailed == Res.MResult)
throw runtime_error("Enqueue process failed.", PI_INVALID_OPERATION);
}
Command *Cmd = static_cast<Command *>(NewEvent->getCommand());
if (Cmd) {
// TODO: Check if lazy mode.
EnqueueResultT Res;
bool Enqueued = GraphProcessor::enqueueCommand(Cmd, Res);
if (!Enqueued && EnqueueResultT::SyclEnqueueFailed == Res.MResult)
throw runtime_error("Enqueue process failed.", PI_INVALID_OPERATION);

if (IsKernel)
((ExecCGCommand *)NewCmd)->flushStreams();
if (IsKernel)
((ExecCGCommand *)Cmd)->flushStreams();
}
}

return NewCmd->getEvent();
return NewEvent;
}

EventImplPtr Scheduler::addCopyBack(Requirement *Req) {
Expand Down