Skip to content

ITS: GPU report found neighbours #14438

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
mconcas merged 1 commit into from
Jun 24, 2025
Merged

ITS: GPU report found neighbours #14438

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
21 changes: 19 additions & 2 deletions Detectors/ITSMFT/ITS/tracking/GPU/ITStrackingGPU/TimeFrameGPU.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -130,7 +130,9 @@ class TimeFrameGPU : public TimeFrame<nLayers>
// Host-specific getters
gsl::span<int, nLayers - 1> getNTracklets() { return mNTracklets; }
gsl::span<int, nLayers - 2> getNCells() { return mNCells; }
std::array<int, nLayers - 2>& getArrayNCells() { return mNCells; }
auto& getArrayNCells() { return mNCells; }
gsl::span<int, nLayers - 3> getNNeighbours() { return mNNeighbours; }
auto& getArrayNNeighbours() { return mNNeighbours; }

// Host-available device getters
gsl::span<int*> getDeviceTrackletsLUTs() { return mTrackletsLUTDevice; }
Expand All @@ -139,7 +141,9 @@ class TimeFrameGPU : public TimeFrame<nLayers>
gsl::span<CellSeed*> getDeviceCells() { return mCellsDevice; }

// Overridden getters
int getNumberOfCells() const;
int getNumberOfTracklets() const final;
int getNumberOfCells() const final;
int getNumberOfNeighbours() const final;

private:
void allocMemAsync(void**, size_t, Stream*, bool); // Abstract owned and unowned memory allocations
Expand All @@ -149,6 +153,7 @@ class TimeFrameGPU : public TimeFrame<nLayers>
// Host-available device buffer sizes
std::array<int, nLayers - 1> mNTracklets;
std::array<int, nLayers - 2> mNCells;
std::array<int, nLayers - 3> mNNeighbours;

// Device pointers
IndexTableUtils* mIndexTableUtilsDevice;
Expand Down Expand Up @@ -218,12 +223,24 @@ inline std::vector<unsigned int> TimeFrameGPU<nLayers>::getClusterSizes()
return sizes;
}

template <int nLayers>
inline int TimeFrameGPU<nLayers>::getNumberOfTracklets() const
{
return std::accumulate(mNTracklets.begin(), mNTracklets.end(), 0);
}

template <int nLayers>
inline int TimeFrameGPU<nLayers>::getNumberOfCells() const
{
return std::accumulate(mNCells.begin(), mNCells.end(), 0);
}

template <int nLayers>
inline int TimeFrameGPU<nLayers>::getNumberOfNeighbours() const
{
return std::accumulate(mNNeighbours.begin(), mNNeighbours.end(), 0);
}

} // namespace o2::its::gpu

#endif
3 changes: 3 additions & 0 deletions Detectors/ITSMFT/ITS/tracking/GPU/cuda/TimeFrameGPU.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -342,6 +342,9 @@ void TimeFrameGPU<nLayers>::createNeighboursIndexTablesDevice()
LOGP(debug, "gpu-transfer: loading neighbours LUT for {} elements on layer {}, for {} MB.", mNCells[iLayer], iLayer, mNCells[iLayer] * sizeof(CellSeed) / MB);
allocMemAsync(reinterpret_cast<void**>(&mNeighboursIndexTablesDevice[iLayer]), (mNCells[iLayer] + 1) * sizeof(int), nullptr, this->getExtAllocator());
GPUChkErrS(cudaMemsetAsync(mNeighboursIndexTablesDevice[iLayer], 0, (mNCells[iLayer] + 1) * sizeof(int), mGpuStreams[0]->get()));
if (iLayer < nLayers - 3) {
mNNeighbours[iLayer] = 0;
}
}
STOP_GPU_STREAM_TIMER(mGpuStreams[0]->get());
}
Expand Down
11 changes: 6 additions & 5 deletions Detectors/ITSMFT/ITS/tracking/GPU/cuda/TrackerTraitsGPU.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -169,7 +169,7 @@ template <int nLayers>
void TrackerTraitsGPU<nLayers>::findCellsNeighbours(const int iteration)
{
mTimeFrameGPU->createNeighboursIndexTablesDevice();
auto& conf = o2::its::ITSGpuTrackingParamConfig::Instance();
const auto& conf = o2::its::ITSGpuTrackingParamConfig::Instance();
for (int iLayer{0}; iLayer < this->mTrkParams[iteration].CellsPerRoad() - 1; ++iLayer) {
const int nextLayerCellsNum{static_cast<int>(mTimeFrameGPU->getNCells()[iLayer + 1])};

Expand Down Expand Up @@ -208,10 +208,11 @@ void TrackerTraitsGPU<nLayers>::findCellsNeighbours(const int iteration)
conf.nBlocks,
conf.nThreads);

filterCellNeighboursHandler(mTimeFrameGPU->getDeviceNeighbourPairs(iLayer),
mTimeFrameGPU->getDeviceNeighbours(iLayer),
nNeigh,
mTimeFrameGPU->getExternalAllocator());
nNeigh = filterCellNeighboursHandler(mTimeFrameGPU->getDeviceNeighbourPairs(iLayer),
mTimeFrameGPU->getDeviceNeighbours(iLayer),
nNeigh,
mTimeFrameGPU->getExternalAllocator());
mTimeFrameGPU->getArrayNNeighbours()[iLayer] = nNeigh;
}
mTimeFrameGPU->createNeighboursDeviceArray();
mTimeFrameGPU->unregisterRest();
Expand Down
6 changes: 3 additions & 3 deletions Detectors/ITSMFT/ITS/tracking/include/ITStracking/TimeFrame.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -174,9 +174,9 @@ struct TimeFrame {
auto& getVerticesMCRecInfo() { return mVerticesMCRecInfo; }

int getNumberOfClusters() const;
int getNumberOfCells() const;
int getNumberOfTracklets() const;
int getNumberOfNeighbours() const;
virtual int getNumberOfCells() const;
virtual int getNumberOfTracklets() const;
virtual int getNumberOfNeighbours() const;
size_t getNumberOfTracks() const;
size_t getNumberOfUsedClusters() const;
auto getNumberOfExtendedTracks() const { return mNExtendedTracks; }
Expand Down