Skip to content

GPU Workflow: Pop next tf from completion policy queue only when actually running and add sanity checks #14640

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
davidrohr merged 1 commit into from
Aug 30, 2025
Merged

GPU Workflow: Pop next tf from completion policy queue only when actually running and add sanity checks #14640

Show file tree
Hide file tree
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
3 changes: 3 additions & 0 deletions GPU/GPUTracking/Global/GPUChainTrackingClusterizer.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -751,6 +751,9 @@ int32_t GPUChainTracking::RunTPCClusterizer(bool synchronizeOutput)
if (buildNativeGPU) {
AllocateRegisteredMemory(mInputsHost->mResourceClusterNativeBuffer);
}
if (mWaitForFinalInputs && GetProcessingSettings().nTPCClustererLanes > 6) {
GPUFatal("ERROR, mWaitForFinalInputs cannot be called with nTPCClustererLanes > 6");
}
if (buildNativeHost && !(buildNativeGPU && GetProcessingSettings().delayedOutput)) {
if (mWaitForFinalInputs) {
GPUFatal("Cannot use waitForFinalInput callback without delayed output");
Expand Down
14 changes: 9 additions & 5 deletions GPU/Workflow/src/GPUWorkflowPipeline.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -66,11 +66,7 @@ void GPURecoWorkflowSpec::initPipeline(o2::framework::InitContext& ic)
mPolicyOrder = [this](o2::framework::DataProcessingHeader::StartTime timeslice) {
std::unique_lock lk(mPipeline->completionPolicyMutex);
mPipeline->completionPolicyNotify.wait(lk, [pipeline = mPipeline.get()] { return pipeline->pipelineSenderTerminating || !pipeline->completionPolicyQueue.empty(); });
if (mPipeline->completionPolicyQueue.front() == timeslice) {
mPipeline->completionPolicyQueue.pop();
return true;
}
return false;
return !mPipeline->completionPolicyQueue.empty() && mPipeline->completionPolicyQueue.front() == timeslice;
};
mPipeline->receiveThread = std::thread([this]() { RunReceiveThread(); });
for (uint32_t i = 0; i < mPipeline->workers.size(); i++) {
Expand Down Expand Up @@ -175,6 +171,14 @@ int32_t GPURecoWorkflowSpec::handlePipeline(ProcessingContext& pc, GPUTrackingIn
tpcZSmeta = std::move(context->tpcZSmeta);
tpcZS = context->tpcZS;
ptrs.tpcZS = &tpcZS;

{
std::lock_guard lk(mPipeline->completionPolicyMutex);
if (mPipeline->completionPolicyQueue.empty() || mPipeline->completionPolicyQueue.front() != tinfo.timeslice) {
LOG(fatal) << "Time frame processed does not equal the timeframe at the top of the queue, time frames seem out of sync";
}
mPipeline->completionPolicyQueue.pop();
}
}
if (mSpecConfig.enableDoublePipeline == 2) {
auto prepareDummyMessage = pc.outputs().make<DataAllocator::UninitializedVector<char>>(Output{gDataOriginGPU, "PIPELINEPREPARE", 0}, 0u);
Expand Down