GPU: Add throughput to pad filter kernel timer.

fweig · fweig · commit eda4c6cc1d31 · 2025-12-02T14:24:23.000+01:00
diff --git a/GPU/GPUTracking/Global/GPUChainTrackingClusterizer.cxx b/GPU/GPUTracking/Global/GPUChainTrackingClusterizer.cxx
@@ -964,6 +964,7 @@ int32_t GPUChainTracking::RunTPCClusterizer(bool synchronizeOutput)
           int32_t nBlocks = TPC_PADS_IN_SECTOR / GPUTPCCFCheckPadBaseline::PadsPerCacheline;
 
           runKernel<GPUTPCCFCheckPadBaseline>({GetGridBlk(nBlocks, lane), {iSector}});
+          getKernelTimer<GPUTPCCFCheckPadBaseline>(RecoStep::TPCClusterFinding, iSector, TPC_PADS_IN_SECTOR * fragment.lengthWithoutOverlap() * sizeof(PackedCharge), false);
         }
 
         runKernel<GPUTPCCFPeakFinder>({GetGrid(clusterer.mPmemory->counters.nPositions, lane), {iSector}});

Original file line number	Diff line number	Diff line change
`@@ -964,6 +964,7 @@ int32_t GPUChainTracking::RunTPCClusterizer(bool synchronizeOutput)`
`964`	`964`	`int32_t nBlocks = TPC_PADS_IN_SECTOR / GPUTPCCFCheckPadBaseline::PadsPerCacheline;`
`965`	`965`
`966`	`966`	`runKernel<GPUTPCCFCheckPadBaseline>({GetGridBlk(nBlocks, lane), {iSector}});`
	`967`	`+ getKernelTimer<GPUTPCCFCheckPadBaseline>(RecoStep::TPCClusterFinding, iSector, TPC_PADS_IN_SECTOR * fragment.lengthWithoutOverlap() * sizeof(PackedCharge), false);`
`967`	`968`	`}`
`968`	`969`
`969`	`970`	`runKernel<GPUTPCCFPeakFinder>({GetGrid(clusterer.mPmemory->counters.nPositions, lane), {iSector}});`