Skip to content

GPU: Add option to free TPC GPU memory, to be used when multiple detectors use the GPU in the workflow. #14084

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 2 commits into from
Mar 18, 2025
Merged

GPU: Add option to free TPC GPU memory, to be used when multiple detectors use the GPU in the workflow. #14084

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3f0bfef
GPU: Improve memory usage debug printout
davidrohr Mar 17, 2025
2560b11
GPU TPC: Add option to clear all non-external-output TPC memory at en…
davidrohr Mar 17, 2025
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
15 changes: 7 additions & 8 deletions GPU/GPUTracking/Base/GPUReconstruction.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -811,11 +811,9 @@ void GPUReconstruction::PopNonPersistentMemory(RecoStep step, uint64_t tag)
GPUFatal("Tag mismatch when popping non persistent memory from stack : pop %s vs on stack %s", qTag2Str(tag).c_str(), qTag2Str(std::get<3>(mNonPersistentMemoryStack.back())).c_str());
}
if ((mProcessingSettings.debugLevel >= 3 || mProcessingSettings.allocDebugLevel) && (IsGPU() || mProcessingSettings.forceHostMemoryPoolSize)) {
if (IsGPU()) {
printf("Allocated Device memory after %30s (%8s): %'13zd (non temporary %'13zd, blocked %'13zd)\n", GPUDataTypes::RECO_STEP_NAMES[getRecoStepNum(step, true)], qTag2Str(std::get<3>(mNonPersistentMemoryStack.back())).c_str(), ptrDiff(mDeviceMemoryPool, mDeviceMemoryBase) + ptrDiff((char*)mDeviceMemoryBase + mDeviceMemorySize, mDeviceMemoryPoolEnd), ptrDiff(mDeviceMemoryPool, mDeviceMemoryBase), mDeviceMemoryPoolBlocked ? ptrDiff((char*)mDeviceMemoryBase + mDeviceMemorySize, mDeviceMemoryPoolBlocked) : 0);
}
printf("Allocated Host memory after %30s (%8s): %'13zd (non temporary %'13zd, blocked %'13zd)\n", GPUDataTypes::RECO_STEP_NAMES[getRecoStepNum(step, true)], qTag2Str(std::get<3>(mNonPersistentMemoryStack.back())).c_str(), ptrDiff(mHostMemoryPool, mHostMemoryBase) + ptrDiff((char*)mHostMemoryBase + mHostMemorySize, mHostMemoryPoolEnd), ptrDiff(mHostMemoryPool, mHostMemoryBase), mHostMemoryPoolBlocked ? ptrDiff((char*)mHostMemoryBase + mHostMemorySize, mHostMemoryPoolBlocked) : 0);
printf("%16s", "");
printf("Allocated memory after %30s (%8s) (Stack %zu): ", GPUDataTypes::RECO_STEP_NAMES[getRecoStepNum(step, true)], qTag2Str(std::get<3>(mNonPersistentMemoryStack.back())).c_str(), mNonPersistentMemoryStack.size());
PrintMemoryOverview();
printf("%76s", "");
PrintMemoryMax();
}
mHostMemoryPoolEnd = std::get<0>(mNonPersistentMemoryStack.back());
Expand Down Expand Up @@ -888,9 +886,10 @@ void GPUReconstruction::PrintMemoryMax()
void GPUReconstruction::PrintMemoryOverview()
{
if (mProcessingSettings.memoryAllocationStrategy == GPUMemoryResource::ALLOCATION_GLOBAL) {
printf("Memory Allocation: Host %'zd / %'zu (Permanent %'zd), Device %'zd / %'zu, (Permanent %'zd) %zu chunks\n",
ptrDiff(mHostMemoryPool, mHostMemoryBase) + ptrDiff((char*)mHostMemoryBase + mHostMemorySize, mHostMemoryPoolEnd), mHostMemorySize, ptrDiff(mHostMemoryPermanent, mHostMemoryBase),
ptrDiff(mDeviceMemoryPool, mDeviceMemoryBase) + ptrDiff((char*)mDeviceMemoryBase + mDeviceMemorySize, mDeviceMemoryPoolEnd), mDeviceMemorySize, ptrDiff(mDeviceMemoryPermanent, mDeviceMemoryBase), mMemoryResources.size());
printf("Memory Allocation: Host %'13zd / %'13zu (Permanent %'13zd, Data %'13zd, Scratch %'13zd), Device %'13zd / %'13zu, (Permanent %'13zd, Data %'13zd, Scratch %'13zd) %zu chunks\n",
ptrDiff(mHostMemoryPool, mHostMemoryBase) + ptrDiff((char*)mHostMemoryBase + mHostMemorySize, mHostMemoryPoolEnd), mHostMemorySize, ptrDiff(mHostMemoryPermanent, mHostMemoryBase), ptrDiff(mHostMemoryPool, mHostMemoryPermanent), ptrDiff((char*)mHostMemoryBase + mHostMemorySize, mHostMemoryPoolEnd),
ptrDiff(mDeviceMemoryPool, mDeviceMemoryBase) + ptrDiff((char*)mDeviceMemoryBase + mDeviceMemorySize, mDeviceMemoryPoolEnd), mDeviceMemorySize, ptrDiff(mDeviceMemoryPermanent, mDeviceMemoryBase), ptrDiff(mDeviceMemoryPool, mDeviceMemoryPermanent), ptrDiff((char*)mDeviceMemoryBase + mDeviceMemorySize, mDeviceMemoryPoolEnd),
mMemoryResources.size());
}
}

Expand Down
11 changes: 11 additions & 0 deletions GPU/GPUTracking/Base/GPUReconstructionCPU.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -215,6 +215,10 @@ int32_t GPUReconstructionCPU::RunChains()
mStatNEvents++;
mNEventsProcessed++;

if (mProcessingSettings.debugLevel >= 3 || mProcessingSettings.allocDebugLevel) {
printf("Allocated memory when starting processing %34s", "");
PrintMemoryOverview();
}
mTimerTotal.Start();
const std::clock_t cpuTimerStart = std::clock();
if (mProcessingSettings.doublePipeline) {
Expand All @@ -232,9 +236,16 @@ int32_t GPUReconstructionCPU::RunChains()
return retVal;
}
}
if (GetProcessingSettings().tpcFreeAllocatedMemoryAfterProcessing) {
ClearAllocatedMemory();
}
}
mTimerTotal.Stop();
mStatCPUTime += (double)(std::clock() - cpuTimerStart) / CLOCKS_PER_SEC;
if (mProcessingSettings.debugLevel >= 3 || mProcessingSettings.allocDebugLevel) {
printf("Allocated memory when ending processing %36s", "");
PrintMemoryOverview();
}

mStatWallTime = (mTimerTotal.GetElapsedTime() * 1000000. / mStatNEvents);
std::string nEventReport;
Expand Down
1 change: 1 addition & 0 deletions GPU/GPUTracking/Definitions/GPUSettingsList.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -326,6 +326,7 @@ AddOption(oclPlatformNum, int32_t, -1, "", 0, "Platform to use, in case the back
AddOption(oclCompileFromSources, bool, false, "", 0, "Compile OpenCL binary from included source code instead of using included spirv code")
AddOption(oclOverrideSourceBuildFlags, std::string, "", "", 0, "Override OCL build flags for compilation from source, put a space for empty options")
AddOption(printSettings, bool, false, "", 0, "Print all settings when initializing")
AddOption(tpcFreeAllocatedMemoryAfterProcessing, bool, false, "", 0, "Clean all memory allocated by TPC when TPC processing done, only data written to external output resources will remain")
AddVariable(eventDisplay, o2::gpu::GPUDisplayFrontendInterface*, nullptr)
AddSubConfig(GPUSettingsProcessingRTC, rtc)
AddSubConfig(GPUSettingsProcessingParam, param)
Expand Down