Skip to content

Please consider the following formatting changes to #14657 #36

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
Sep 6, 2025
Merged

Please consider the following formatting changes to #14657 #36

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
48 changes: 24 additions & 24 deletions GPU/GPUTracking/TPCClusterFinder/GPUTPCNNClusterizer.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,10 +69,10 @@ void* GPUTPCNNClusterizer::setIOPointers(void* mem)
if (mNnClusterizerVerbosity > 2) {
if (mNnClusterizerVerbosity > 3) {
auto fmt = [](size_t bytes) {
std::ostringstream os;
double mb = bytes / (1024.0 * 1024.0);
os << bytes << " bytes (" << std::fixed << std::setprecision(3) << mb << " MB)";
return os.str();
std::ostringstream os;
double mb = bytes / (1024.0 * 1024.0);
os << bytes << " bytes (" << std::fixed << std::setprecision(3) << mb << " MB)";
return os.str();
};

// Element counts (number of array entries, not bytes)
Expand Down Expand Up @@ -101,35 +101,35 @@ void* GPUTPCNNClusterizer::setIOPointers(void* mem)

LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") Pointers set for clusterizer with memoryID " << mMemoryId << " deviceID " << mDeviceId << " and sector " << mISector;
LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") mOutputDataClass pointer: " << mOutputDataClass
<< " | elements=" << elemsOutputDataClass << " (= mNnClusterizerTotalClusters)"
<< " | " << fmt(szOutputDataClass);
<< " | elements=" << elemsOutputDataClass << " (= mNnClusterizerTotalClusters)"
<< " | " << fmt(szOutputDataClass);
LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") mClusterFlags pointer: " << static_cast<const void*>(mClusterFlags)
<< " | elements=" << elemsClusterFlags << " (= 2 * mNnClusterizerBatchedMode)"
<< " | " << fmt(szClusterFlags);
<< " | elements=" << elemsClusterFlags << " (= 2 * mNnClusterizerBatchedMode)"
<< " | " << fmt(szClusterFlags);
LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") mInputData_16 pointer: " << mInputData_16
<< " | elements=" << elemsInput16 << " (= mNnClusterizerBatchedMode * mNnClusterizerElementSize)"
<< " | " << fmt(szInput16);
<< " | elements=" << elemsInput16 << " (= mNnClusterizerBatchedMode * mNnClusterizerElementSize)"
<< " | " << fmt(szInput16);
LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") mModelProbabilities_16 pointer: " << mModelProbabilities_16
<< " | elements=" << elemsProb16 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelClassNumOutputNodes)"
<< " | " << fmt(szProb16);
<< " | elements=" << elemsProb16 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelClassNumOutputNodes)"
<< " | " << fmt(szProb16);
LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") mOutputDataReg1_16 pointer: " << mOutputDataReg1_16
<< " | elements=" << elemsReg1_16 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelReg1NumOutputNodes)"
<< " | " << fmt(szReg1_16);
<< " | elements=" << elemsReg1_16 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelReg1NumOutputNodes)"
<< " | " << fmt(szReg1_16);
LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") mOutputDataReg2_16 pointer: " << mOutputDataReg2_16
<< " | elements=" << elemsReg2_16 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelReg2NumOutputNodes)"
<< " | " << fmt(szReg2_16);
<< " | elements=" << elemsReg2_16 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelReg2NumOutputNodes)"
<< " | " << fmt(szReg2_16);
LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") mInputData_32 pointer: " << mInputData_32
<< " | elements=" << elemsInput32 << " (= mNnClusterizerBatchedMode * mNnClusterizerElementSize)"
<< " | " << fmt(szInput32);
<< " | elements=" << elemsInput32 << " (= mNnClusterizerBatchedMode * mNnClusterizerElementSize)"
<< " | " << fmt(szInput32);
LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") mModelProbabilities_32 pointer: " << mModelProbabilities_32
<< " | elements=" << elemsProb32 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelClassNumOutputNodes)"
<< " | " << fmt(szProb32);
<< " | elements=" << elemsProb32 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelClassNumOutputNodes)"
<< " | " << fmt(szProb32);
LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") mOutputDataReg1_32 pointer: " << mOutputDataReg1_32
<< " | elements=" << elemsReg1_32 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelReg1NumOutputNodes)"
<< " | " << fmt(szReg1_32);
<< " | elements=" << elemsReg1_32 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelReg1NumOutputNodes)"
<< " | " << fmt(szReg1_32);
LOG(info) << "(NNCLUS, GPUTPCNNClusterizer, this=" << this << ") mOutputDataReg2_32 pointer: " << mOutputDataReg2_32
<< " | elements=" << elemsReg2_32 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelReg2NumOutputNodes)"
<< " | " << fmt(szReg2_32);
<< " | elements=" << elemsReg2_32 << " (= mNnClusterizerBatchedMode * mNnClusterizerModelReg2NumOutputNodes)"
<< " | " << fmt(szReg2_32);
}
// Compute allocated bytes (difference between advanced pointer and start pointer)
size_t allocatedBytes = static_cast<size_t>(reinterpret_cast<uintptr_t>(mem) - reinterpret_cast<uintptr_t>(startMem));
Expand Down
76 changes: 38 additions & 38 deletions GPU/GPUTracking/TPCClusterFinder/GPUTPCNNClusterizerKernels.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -275,7 +275,7 @@ GPUdii() void GPUTPCNNClusterizerKernels::Thread<GPUTPCNNClusterizerKernels::det
if (glo_idx + batchStart >= clusterer.mPmemory->counters.nClusters || glo_idx >= clustererNN.mNnClusterizerBatchedMode) {
return;
}
if(clustererNN.mNnClusterizerUseClassification) {
if (clustererNN.mNnClusterizerUseClassification) {
if (dtype == 0) {
clustererNN.mOutputDataClass[glo_idx + batchStart] = (int32_t)((clustererNN.mModelProbabilities_16[glo_idx]).ToFloat() > clustererNN.mNnClassThreshold);
} else if (dtype == 1) {
Expand All @@ -295,7 +295,7 @@ GPUdii() void GPUTPCNNClusterizerKernels::Thread<GPUTPCNNClusterizerKernels::det
if (glo_idx + batchStart >= clusterer.mPmemory->counters.nClusters || glo_idx >= clustererNN.mNnClusterizerBatchedMode) {
return;
}
if(clustererNN.mNnClusterizerUseClassification) {
if (clustererNN.mNnClusterizerUseClassification) {
uint32_t elem_iterator = glo_idx * clustererNN.mNnClusterizerModelClassNumOutputNodes;
float current_max_prob = 0.f; // If the neural network doesn't contain the softmax as a last layer, the outputs can range in [-infty, infty]
uint32_t class_label = 0;
Expand Down Expand Up @@ -401,20 +401,20 @@ GPUdii() void GPUTPCNNClusterizerKernels::Thread<GPUTPCNNClusterizerKernels::pub

if (dtype == 0) {
pc.setFull(central_charge * clustererNN.mOutputDataReg1_16[model_output_index + 4].ToFloat(),
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg1_16[model_output_index].ToFloat(),
notSinglePad ? clustererNN.mOutputDataReg1_16[model_output_index + 2].ToFloat() : 0.f,
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg1_16[model_output_index + 1].ToFloat(),
notSingleTime ? clustererNN.mOutputDataReg1_16[model_output_index + 3].ToFloat() : 0.f,
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg1_16[model_output_index].ToFloat(),
notSinglePad ? clustererNN.mOutputDataReg1_16[model_output_index + 2].ToFloat() : 0.f,
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg1_16[model_output_index + 1].ToFloat(),
notSingleTime ? clustererNN.mOutputDataReg1_16[model_output_index + 3].ToFloat() : 0.f,
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
} else if (dtype == 1) {
pc.setFull(central_charge * clustererNN.mOutputDataReg1_32[model_output_index + 4],
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg1_32[model_output_index],
notSinglePad ? clustererNN.mOutputDataReg1_32[model_output_index + 2] : 0.f,
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg1_32[model_output_index + 1],
notSingleTime ? clustererNN.mOutputDataReg1_32[model_output_index + 3] : 0.f,
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg1_32[model_output_index],
notSinglePad ? clustererNN.mOutputDataReg1_32[model_output_index + 2] : 0.f,
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg1_32[model_output_index + 1],
notSingleTime ? clustererNN.mOutputDataReg1_32[model_output_index + 3] : 0.f,
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
}

tpc::ClusterNative myCluster;
Expand Down Expand Up @@ -511,20 +511,20 @@ GPUdii() void GPUTPCNNClusterizerKernels::Thread<GPUTPCNNClusterizerKernels::pub
// Cluster 1
if (dtype == 0) {
pc.setFull(central_charge * clustererNN.mOutputDataReg2_16[model_output_index + 8].ToFloat(),
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg2_16[model_output_index].ToFloat(),
clustererNN.mOutputDataReg2_16[model_output_index + 4].ToFloat(),
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg2_16[model_output_index + 2].ToFloat(),
clustererNN.mOutputDataReg2_16[model_output_index + 6].ToFloat(),
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg2_16[model_output_index].ToFloat(),
clustererNN.mOutputDataReg2_16[model_output_index + 4].ToFloat(),
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg2_16[model_output_index + 2].ToFloat(),
clustererNN.mOutputDataReg2_16[model_output_index + 6].ToFloat(),
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
} else if (dtype == 1) {
pc.setFull(central_charge * clustererNN.mOutputDataReg2_32[model_output_index + 8],
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg2_32[model_output_index],
clustererNN.mOutputDataReg2_32[model_output_index + 4],
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg2_32[model_output_index + 2],
clustererNN.mOutputDataReg2_32[model_output_index + 6],
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg2_32[model_output_index],
clustererNN.mOutputDataReg2_32[model_output_index + 4],
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg2_32[model_output_index + 2],
clustererNN.mOutputDataReg2_32[model_output_index + 6],
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
}

tpc::ClusterNative myCluster;
Expand Down Expand Up @@ -559,20 +559,20 @@ GPUdii() void GPUTPCNNClusterizerKernels::Thread<GPUTPCNNClusterizerKernels::pub
// Cluster 2
if (dtype == 0) {
pc.setFull(central_charge * clustererNN.mOutputDataReg2_16[model_output_index + 9].ToFloat(),
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg2_16[model_output_index + 1].ToFloat(),
clustererNN.mOutputDataReg2_16[model_output_index + 5].ToFloat(),
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg2_16[model_output_index + 3].ToFloat(),
clustererNN.mOutputDataReg2_16[model_output_index + 7].ToFloat(),
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg2_16[model_output_index + 1].ToFloat(),
clustererNN.mOutputDataReg2_16[model_output_index + 5].ToFloat(),
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg2_16[model_output_index + 3].ToFloat(),
clustererNN.mOutputDataReg2_16[model_output_index + 7].ToFloat(),
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
} else if (dtype == 1) {
pc.setFull(central_charge * clustererNN.mOutputDataReg2_32[model_output_index + 9],
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg2_32[model_output_index + 1],
clustererNN.mOutputDataReg2_32[model_output_index + 5],
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg2_32[model_output_index + 3],
clustererNN.mOutputDataReg2_32[model_output_index + 7],
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
static_cast<float>(peak.pad()) + clustererNN.mOutputDataReg2_32[model_output_index + 1],
clustererNN.mOutputDataReg2_32[model_output_index + 5],
(clusterer.mPmemory->fragment).start + static_cast<float>(peak.time()) + clustererNN.mOutputDataReg2_32[model_output_index + 3],
clustererNN.mOutputDataReg2_32[model_output_index + 7],
clustererNN.mClusterFlags[2 * glo_idx],
clustererNN.mClusterFlags[2 * glo_idx + 1]);
}

rejectCluster = !pc.toNative(peak, central_charge, myCluster, clusterer.Param(), chargeMap);
Expand Down