Skip to content

GPU: Parallize track merger resolve kernel. #3676

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
Jun 1, 2020
Merged

GPU: Parallize track merger resolve kernel. #3676

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
d547275
GPU: Parallize track merger resolve kernel.
fweig May 29, 2020
71f2125
Merge branch 'dev' into parallel-resolve
davidrohr Jun 1, 2020
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
23 changes: 19 additions & 4 deletions GPU/Common/GPUDefGPUParameters.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,7 +45,10 @@
#define GPUCA_LB_GPUTPCGMMergerSliceRefit 256
#define GPUCA_LB_GPUTPCGMMergerUnpackResetIds 256
#define GPUCA_LB_GPUTPCGMMergerUnpackGlobal 256
#define GPUCA_LB_GPUTPCGMMergerResolve 256
#define GPUCA_LB_GPUTPCGMMergerResolve_step0 256
#define GPUCA_LB_GPUTPCGMMergerResolve_step1 256
#define GPUCA_LB_GPUTPCGMMergerResolve_step2 256
#define GPUCA_LB_GPUTPCGMMergerResolve_step3 256
#define GPUCA_LB_GPUTPCGMMergerClearLinks 256
#define GPUCA_LB_GPUTPCGMMergerMergeWithinPrepare 256
#define GPUCA_LB_GPUTPCGMMergerMergeSlicesPrepare 256
Expand Down Expand Up @@ -90,7 +93,10 @@
#define GPUCA_LB_GPUTPCGMMergerSliceRefit 64
#define GPUCA_LB_GPUTPCGMMergerUnpackResetIds 256
#define GPUCA_LB_GPUTPCGMMergerUnpackGlobal 256
#define GPUCA_LB_GPUTPCGMMergerResolve 256
#define GPUCA_LB_GPUTPCGMMergerResolve_step0 256
#define GPUCA_LB_GPUTPCGMMergerResolve_step1 256
#define GPUCA_LB_GPUTPCGMMergerResolve_step2 256
#define GPUCA_LB_GPUTPCGMMergerResolve_step3 256
#define GPUCA_LB_GPUTPCGMMergerClearLinks 256
#define GPUCA_LB_GPUTPCGMMergerMergeWithinPrepare 256
#define GPUCA_LB_GPUTPCGMMergerMergeSlicesPrepare 256, 2
Expand Down Expand Up @@ -186,8 +192,17 @@
#ifndef GPUCA_LB_GPUTPCGMMergerUnpackGlobal
#define GPUCA_LB_GPUTPCGMMergerUnpackGlobal 256
#endif
#ifndef GPUCA_LB_GPUTPCGMMergerResolve
#define GPUCA_LB_GPUTPCGMMergerResolve 256
#ifndef GPUCA_LB_GPUTPCGMMergerResolve_step0
#define GPUCA_LB_GPUTPCGMMergerResolve_step0 256
#endif
#ifndef GPUCA_LB_GPUTPCGMMergerResolve_step1
#define GPUCA_LB_GPUTPCGMMergerResolve_step1 256
#endif
#ifndef GPUCA_LB_GPUTPCGMMergerResolve_step2
#define GPUCA_LB_GPUTPCGMMergerResolve_step2 256
#endif
#ifndef GPUCA_LB_GPUTPCGMMergerResolve_step3
#define GPUCA_LB_GPUTPCGMMergerResolve_step3 256
#endif
#ifndef GPUCA_LB_GPUTPCGMMergerClearLinks
#define GPUCA_LB_GPUTPCGMMergerClearLinks 256
Expand Down
5 changes: 4 additions & 1 deletion GPU/GPUTracking/Base/GPUReconstructionKernels.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,10 @@ GPUCA_KRNL_LB((GPUTPCGMMergerUnpackResetIds ), (simple), (, int
GPUCA_KRNL_LB((GPUTPCGMMergerSliceRefit ), (simple), (, int iSlice), (, iSlice))
GPUCA_KRNL_LB((GPUTPCGMMergerUnpackGlobal ), (simple), (, int iSlice), (, iSlice))
GPUCA_KRNL(( GPUTPCGMMergerUnpackSaveNumber ), (simple), (, int id), (, id))
GPUCA_KRNL_LB((GPUTPCGMMergerResolve ), (simple), (, char useOrigTrackParam, char mergeAll), (, useOrigTrackParam, mergeAll))
GPUCA_KRNL_LB((GPUTPCGMMergerResolve, step0 ), (simple), (), ())
GPUCA_KRNL_LB((GPUTPCGMMergerResolve, step1 ), (simple), (), ())
GPUCA_KRNL_LB((GPUTPCGMMergerResolve, step2 ), (simple), (), ())
GPUCA_KRNL_LB((GPUTPCGMMergerResolve, step3 ), (simple), (, char useOrigTrackParam, char mergeAll), (, useOrigTrackParam, mergeAll))
GPUCA_KRNL_LB((GPUTPCGMMergerClearLinks ), (simple), (, char nOutput), (, nOutput))
GPUCA_KRNL_LB((GPUTPCGMMergerMergeWithinPrepare ), (simple), (), ())
GPUCA_KRNL_LB((GPUTPCGMMergerMergeSlicesPrepare ), (simple), (, int border0, int border1, char useOrigTrackParam), (, border0, border1, useOrigTrackParam))
Expand Down
17 changes: 13 additions & 4 deletions GPU/GPUTracking/Global/GPUChainTracking.cxx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1643,6 +1643,15 @@ void GPUChainTracking::RunTPCTrackingMerger_MergeBorderTracks(char withinSlice,
mRec->ReturnVolatileDeviceMemory();
}

void GPUChainTracking::RunTPCTrackingMerger_Resolve(char useOrigTrackParam, char mergeAll, GPUReconstruction::krnlDeviceType deviceType)
{
runKernel<GPUTPCGMMergerResolve, 0>(GetGridBlk(BlockCount(), 0, deviceType), krnlRunRangeNone, krnlEventNone);
runKernel<GPUTPCGMMergerResolve, 1>(GetGridBlk(BlockCount(), 0, deviceType), krnlRunRangeNone, krnlEventNone);
runKernel<GPUTPCGMMergerResolve, 2>(GetGridBlk(BlockCount(), 0, deviceType), krnlRunRangeNone, krnlEventNone);
// TODO: Determine number of blocks from block size to fully occupy gpu
runKernel<GPUTPCGMMergerResolve, 3>(GetGridBlk(BlockCount() * 4, 0, deviceType), krnlRunRangeNone, krnlEventNone, useOrigTrackParam, mergeAll);
}

int GPUChainTracking::RunTPCTrackingMerger(bool synchronizeOutput)
{
if (GetDeviceProcessingSettings().debugLevel >= 6 && GetDeviceProcessingSettings().comparableDebutOutput && param().rec.mergerReadFromTrackerDirectly) {
Expand Down Expand Up @@ -1701,22 +1710,22 @@ int GPUChainTracking::RunTPCTrackingMerger(bool synchronizeOutput)
runKernel<GPUMemClean16>({1, -WarpSize(), 0, deviceType, RecoStep::TPCMerging}, krnlRunRangeNone, {}, MergerShadowAll.TmpCounter(), NSLICES * sizeof(*MergerShadowAll.TmpCounter()));
runKernel<GPUTPCGMMergerMergeWithinPrepare>(GetGridBlk(numBlocks, 0, deviceType), krnlRunRangeNone, krnlEventNone);
RunTPCTrackingMerger_MergeBorderTracks(1, 0, deviceType);
runKernel<GPUTPCGMMergerResolve>(GetGridBlk(numBlocks, 0, deviceType), krnlRunRangeNone, krnlEventNone, 0, 1);
RunTPCTrackingMerger_Resolve(0, 1, deviceType);
DoDebugAndDump(RecoStep::TPCMerging, 0, doGPUall, Merger, &GPUTPCGMMerger::DumpMergedWithinSlices, mDebugFile);

runKernel<GPUTPCGMMergerClearLinks>(GetGridBlk(numBlocks, 0, deviceType), krnlRunRangeNone, krnlEventNone, 0);
runKernel<GPUMemClean16>({1, -WarpSize(), 0, deviceType, RecoStep::TPCMerging}, krnlRunRangeNone, {}, MergerShadowAll.TmpCounter(), 2 * NSLICES * sizeof(*MergerShadowAll.TmpCounter()));
runKernel<GPUTPCGMMergerMergeSlicesPrepare>(GetGridBlk(std::max(2u, numBlocks), 0, deviceType), krnlRunRangeNone, krnlEventNone, 2, 3, 0);
RunTPCTrackingMerger_MergeBorderTracks(0, 0, deviceType);
runKernel<GPUTPCGMMergerResolve>(GetGridBlk(numBlocks, 0, deviceType), krnlRunRangeNone, krnlEventNone, 0, 0);
RunTPCTrackingMerger_Resolve(0, 1, deviceType);
runKernel<GPUMemClean16>({1, -WarpSize(), 0, deviceType, RecoStep::TPCMerging}, krnlRunRangeNone, {}, MergerShadowAll.TmpCounter(), 2 * NSLICES * sizeof(*MergerShadowAll.TmpCounter()));
runKernel<GPUTPCGMMergerMergeSlicesPrepare>(GetGridBlk(std::max(2u, numBlocks), 0, deviceType), krnlRunRangeNone, krnlEventNone, 0, 1, 0);
RunTPCTrackingMerger_MergeBorderTracks(0, 0, deviceType);
runKernel<GPUTPCGMMergerResolve>(GetGridBlk(numBlocks, 0, deviceType), krnlRunRangeNone, krnlEventNone, 0, 0);
RunTPCTrackingMerger_Resolve(0, 1, deviceType);
runKernel<GPUMemClean16>({1, -WarpSize(), 0, deviceType, RecoStep::TPCMerging}, krnlRunRangeNone, {}, MergerShadowAll.TmpCounter(), 2 * NSLICES * sizeof(*MergerShadowAll.TmpCounter()));
runKernel<GPUTPCGMMergerMergeSlicesPrepare>(GetGridBlk(std::max(2u, numBlocks), 0, deviceType), krnlRunRangeNone, krnlEventNone, 0, 1, 1);
RunTPCTrackingMerger_MergeBorderTracks(0, -1, deviceType);
runKernel<GPUTPCGMMergerResolve>(GetGridBlk(numBlocks, 0, deviceType), krnlRunRangeNone, krnlEventNone, 1, 0);
RunTPCTrackingMerger_Resolve(0, 1, deviceType);
DoDebugAndDump(RecoStep::TPCMerging, 0, doGPUall, Merger, &GPUTPCGMMerger::DumpMergedBetweenSlices, mDebugFile);

runKernel<GPUMemClean16>({1, -WarpSize(), 0, deviceType, RecoStep::TPCMerging}, krnlRunRangeNone, {}, MergerShadowAll.TmpCounter(), 2 * NSLICES * sizeof(*MergerShadowAll.TmpCounter()));
Expand Down
1 change: 1 addition & 0 deletions GPU/GPUTracking/Global/GPUChainTracking.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -250,6 +250,7 @@ class GPUChainTracking : public GPUChain, GPUReconstructionHelpers::helperDelega
void RunTPCClusterizer_compactPeaks(GPUTPCClusterFinder& clusterer, GPUTPCClusterFinder& clustererShadow, int stage, bool doGPU, int lane);
std::pair<unsigned int, unsigned int> TPCClusterizerDecodeZSCount(unsigned int iSlice, unsigned int minTime, unsigned int maxTime);
void RunTPCTrackingMerger_MergeBorderTracks(char withinSlice, char mergeMode, GPUReconstruction::krnlDeviceType deviceType);
void RunTPCTrackingMerger_Resolve(char useOrigTrackParam, char mergeAll, GPUReconstruction::krnlDeviceType deviceType);

std::atomic_flag mLockAtomic = ATOMIC_FLAG_INIT;

Expand Down
Loading