Skip to content

Please consider the following formatting changes to #14510 #27

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
ChSonnabend merged 1 commit into from
Jul 13, 2025
Merged

Please consider the following formatting changes to #14510 #27

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
26 changes: 14 additions & 12 deletions GPU/GPUTracking/TPCClusterFinder/GPUTPCNNClusterizerHost.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -93,9 +93,9 @@ void GPUTPCNNClusterizerHost::initClusterizer(const GPUSettingsProcessingNNclust
clustererNN.mNnClusterizerElementSize = clustererNN.mNnClusterizerChargeArraySize + (settings.nnClusterizerAddIndexData ? 3 : 0);
clustererNN.mBoundaryMapSizeRow = 3 * clustererNN.mNnClusterizerSizeInputRow + o2::tpc::constants::MAXGLOBALPADROW;
clustererNN.mBoundaryPadding = 11; // padding on each side to account for pad_offset. N=11 since then mIsBoundary = 24320 ~< (1.5 x 2^14 = 24576) && N must be bigger than (NPads[row(end_iroc + 1)] - NPads[row(end_iroc)])/2 (=6) for pad_offset to work
clustererNN.mBoundaryMapSizePadsPerRow = GPUTPCGeometry::NPads(o2::tpc::constants::MAXGLOBALPADROW) + 2*clustererNN.mBoundaryPadding;
clustererNN.mBoundaryMapSize = clustererNN.mBoundaryMapSizeRow*clustererNN.mBoundaryMapSizePadsPerRow;
clustererNN.mIndexLookupSize = 3*clustererNN.mNnClusterizerChargeArraySize; // local row, pad, time shift from flat index
clustererNN.mBoundaryMapSizePadsPerRow = GPUTPCGeometry::NPads(o2::tpc::constants::MAXGLOBALPADROW) + 2 * clustererNN.mBoundaryPadding;
clustererNN.mBoundaryMapSize = clustererNN.mBoundaryMapSizeRow * clustererNN.mBoundaryMapSizePadsPerRow;
clustererNN.mIndexLookupSize = 3 * clustererNN.mNnClusterizerChargeArraySize; // local row, pad, time shift from flat index
clustererNN.mNnClusterizerAddIndexData = settings.nnClusterizerAddIndexData;
clustererNN.mNnClusterizerBatchedMode = settings.nnClusterizerBatchedMode;
clustererNN.mNnClusterizerBoundaryFillValue = settings.nnClusterizerBoundaryFillValue;
Expand Down Expand Up @@ -124,34 +124,36 @@ void GPUTPCNNClusterizerHost::initClusterizer(const GPUSettingsProcessingNNclust
}
}

void GPUTPCNNClusterizerHost::createBoundary(GPUTPCNNClusterizer& clustererNN) {
void GPUTPCNNClusterizerHost::createBoundary(GPUTPCNNClusterizer& clustererNN)
{
// Call after init of the clustererNN elements
for(int r = 0; r < clustererNN.mBoundaryMapSizeRow; r++) {
for (int r = 0; r < clustererNN.mBoundaryMapSizeRow; r++) {
int8_t skipCheckInRow = 0;
for (int p = 0; p < clustererNN.mBoundaryMapSizePadsPerRow; p++) {
int32_t i = r * clustererNN.mBoundaryMapSizePadsPerRow + p;
clustererNN.mIsBoundary[i] = 1;
if (!skipCheckInRow && (p >= clustererNN.mBoundaryPadding || r >= clustererNN.mNnClusterizerSizeInputRow)) {
if (r < (GPUTPCGeometry::EndIROC() + clustererNN.mNnClusterizerSizeInputRow)) {
clustererNN.mIsBoundary[i] = (int32_t)((p - clustererNN.mBoundaryPadding) >= static_cast<int>(GPUTPCGeometry::NPads(r - clustererNN.mNnClusterizerSizeInputRow)));
} else if (r >= (GPUTPCGeometry::EndIROC() + 2*clustererNN.mNnClusterizerSizeInputRow) && r < (o2::tpc::constants::MAXGLOBALPADROW + 2*clustererNN.mNnClusterizerSizeInputRow)) {
clustererNN.mIsBoundary[i] = (int32_t)((p - clustererNN.mBoundaryPadding) >= static_cast<int>(GPUTPCGeometry::NPads(r - 2*clustererNN.mNnClusterizerSizeInputRow)));
} else if (r >= (GPUTPCGeometry::EndIROC() + 2 * clustererNN.mNnClusterizerSizeInputRow) && r < (o2::tpc::constants::MAXGLOBALPADROW + 2 * clustererNN.mNnClusterizerSizeInputRow)) {
clustererNN.mIsBoundary[i] = (int32_t)((p - clustererNN.mBoundaryPadding) >= static_cast<int>(GPUTPCGeometry::NPads(r - 2 * clustererNN.mNnClusterizerSizeInputRow)));
}
skipCheckInRow = (clustererNN.mIsBoundary[i] == 1); // No need to check further pads in this row
}
}
}
}

void GPUTPCNNClusterizerHost::createIndexLookup(GPUTPCNNClusterizer& clustererNN) {
for(int32_t i = 0; i < clustererNN.mNnClusterizerChargeArraySize; i++){
void GPUTPCNNClusterizerHost::createIndexLookup(GPUTPCNNClusterizer& clustererNN)
{
for (int32_t i = 0; i < clustererNN.mNnClusterizerChargeArraySize; i++) {
int32_t r = CAMath::Floor(i / ((2 * clustererNN.mNnClusterizerSizeInputPad + 1) * (2 * clustererNN.mNnClusterizerSizeInputTime + 1))) - clustererNN.mNnClusterizerSizeInputRow;
int32_t rest_1 = i % ((2 * clustererNN.mNnClusterizerSizeInputPad + 1) * (2 * clustererNN.mNnClusterizerSizeInputTime + 1));
int32_t p = CAMath::Floor(rest_1 / (2 * clustererNN.mNnClusterizerSizeInputTime + 1)) - clustererNN.mNnClusterizerSizeInputPad;
int32_t t = (rest_1 % (2 * clustererNN.mNnClusterizerSizeInputTime + 1)) - clustererNN.mNnClusterizerSizeInputTime;
clustererNN.mIndexLookup[3*i] = r;
clustererNN.mIndexLookup[3*i + 1] = p;
clustererNN.mIndexLookup[3*i + 2] = t;
clustererNN.mIndexLookup[3 * i] = r;
clustererNN.mIndexLookup[3 * i + 1] = p;
clustererNN.mIndexLookup[3 * i + 2] = t;
}
}

Expand Down
6 changes: 3 additions & 3 deletions GPU/GPUTracking/TPCClusterFinder/GPUTPCNNClusterizerKernels.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -75,7 +75,7 @@ GPUdii() void GPUTPCNNClusterizerKernels::Thread<GPUTPCNNClusterizerKernels::fil
int32_t isBoundaryIndex = (row_pos + row_offset + clustererNN.mNnClusterizerSizeInputRow) * clustererNN.mBoundaryMapSizePadsPerRow + pad_pos + clustererNN.mBoundaryPadding;
if (!clustererNN.mIsBoundary[isBoundaryIndex] && (time_pos >= 0) && (time_pos < TPC_MAX_FRAGMENT_LEN_GPU)) {
CfChargePos tmp_pos(row_pos, pad_pos, time_pos);
if (!clustererNN.mNnClusterizerSetDeconvolutionFlags) { // Only if deconvolution flags are not set
if (!clustererNN.mNnClusterizerSetDeconvolutionFlags) { // Only if deconvolution flags are not set
if (r == 0 && !clustererNN.mClusterFlags[2 * glo_idx] && CAMath::Abs(p) < 3 && CAMath::Abs(t) < 3 && p != 0 && t != 0) { // ordering is done for short circuit optimization
clustererNN.mClusterFlags[2 * glo_idx] += CfUtils::isPeak(isPeakMap[tmp_pos]);
clustererNN.mClusterFlags[2 * glo_idx + 1] = clustererNN.mClusterFlags[2 * glo_idx];
Expand Down Expand Up @@ -162,7 +162,7 @@ GPUdii() void GPUTPCNNClusterizerKernels::Thread<GPUTPCNNClusterizerKernels::fil
}
} else if ((int32_t)transient_index < clustererNN.mNnClusterizerChargeArraySize) {
int32_t time = static_cast<int>(peak.time());
int32_t idxLookup = 3*transient_index;
int32_t idxLookup = 3 * transient_index;
int32_t r = clustererNN.mIndexLookup[idxLookup] + row, p = clustererNN.mIndexLookup[idxLookup + 1] + pad, t = clustererNN.mIndexLookup[idxLookup + 2] + time;
int32_t row_offset = GPUTPCNNClusterizerKernels::rowOffset(row, clustererNN.mNnClusterizerSizeInputRow);
int32_t pad_offset = GPUTPCNNClusterizerKernels::padOffset(row, r);
Expand Down Expand Up @@ -501,7 +501,7 @@ GPUdii() void GPUTPCNNClusterizerKernels::Thread<GPUTPCNNClusterizerKernels::pub
// THe following arithmetic is done because the network is trained with a split between IROC and OROC boundary
GPUd() int32_t GPUTPCNNClusterizerKernels::padOffset(int32_t row_ref, int32_t row_current)
{
if(row_current < 0 || row_current > o2::tpc::constants::MAXGLOBALPADROW) {
if (row_current < 0 || row_current > o2::tpc::constants::MAXGLOBALPADROW) {
return 0; // Short-circuit for negative rows
} else {
return (int)((GPUTPCGeometry::NPads(row_current) - GPUTPCGeometry::NPads(row_ref)) / 2);
Expand Down