Switch on timers on CPU for the first four lanes

Author: ChSonnabend

GPU/GPUTracking/Global/GPUChainTrackingClusterizer.cxx

@@ -1052,7 +1052,7 @@ int32_t GPUChainTracking::RunTPCClusterizer(bool synchronizeOutput)
 
             // NN evaluations
             if(clustererNNShadow.mNnClusterizerUseClassification) {
-              if(GetProcessingSettings().debugLevel >= 1 && doGPU) { nnTimers[3*lane]->Start(); }
+              if(GetProcessingSettings().debugLevel >= 1 && (doGPU || lane < 4)) { nnTimers[3*lane]->Start(); }
               if (clustererNNShadow.mNnInferenceInputDType == 0) {
                 if (clustererNNShadow.mNnInferenceOutputDType == 0) {
                   (nnApplication.mModelClass).inference(clustererNNShadow.mInputData_16, iSize, clustererNNShadow.mModelProbabilities_16);
@@ -1066,13 +1066,13 @@ int32_t GPUChainTracking::RunTPCClusterizer(bool synchronizeOutput)
                   (nnApplication.mModelClass).inference(clustererNNShadow.mInputData_32, iSize, clustererNNShadow.mModelProbabilities_32);
                 }
               }
-              if(GetProcessingSettings().debugLevel >= 1 && doGPU) { nnTimers[3*lane]->Stop(); }
+              if(GetProcessingSettings().debugLevel >= 1 && (doGPU || lane < 4)) { nnTimers[3*lane]->Stop(); } // doGPU || lane<4 -> only for GPU or first 4 CPU lanes (to limit number of concurrent timers). At least gives some statistics for CPU time...
               if (nn_settings.nnClusterizerVerbosity > 3) {
                 LOG(info) << "(NNCLUS, GPUChainTrackingClusterizer, this=" << this << ") Done with NN classification inference. Loop=" << batch << ". (clustererNN=" << &clustererNN << ", clustererNNShadow=" << &clustererNNShadow << ")";
               }
             }
             if (!clustererNNShadow.mNnClusterizerUseCfRegression) {
-              if(GetProcessingSettings().debugLevel >= 1 && doGPU) { nnTimers[3*lane + 1]->Start(); }
+              if(GetProcessingSettings().debugLevel >= 1 && (doGPU || lane < 4)) { nnTimers[3*lane + 1]->Start(); }
               if (clustererNNShadow.mNnInferenceInputDType == 0) {
                 if (clustererNNShadow.mNnInferenceOutputDType == 0) {
                   (nnApplication.mModelReg1).inference(clustererNNShadow.mInputData_16, iSize, clustererNNShadow.mOutputDataReg1_16);
@@ -1086,9 +1086,9 @@ int32_t GPUChainTracking::RunTPCClusterizer(bool synchronizeOutput)
                   (nnApplication.mModelReg1).inference(clustererNNShadow.mInputData_32, iSize, clustererNNShadow.mOutputDataReg1_32);
                 }
               }
-              if(GetProcessingSettings().debugLevel >= 1 && doGPU) { nnTimers[3*lane + 1]->Stop(); }
+              if(GetProcessingSettings().debugLevel >= 1 && (doGPU || lane < 4)) { nnTimers[3*lane + 1]->Stop(); }
               if (nnApplication.mModelClass.getNumOutputNodes()[0][1] > 1 && nnApplication.mModelReg2.isInitialized()) {
-                if(GetProcessingSettings().debugLevel >= 1 && doGPU) { nnTimers[3*lane + 2]->Start(); }
+                if(GetProcessingSettings().debugLevel >= 1 && (doGPU || lane < 4)) { nnTimers[3*lane + 2]->Start(); }
                 if (clustererNNShadow.mNnInferenceInputDType == 0) {
                   if (clustererNNShadow.mNnInferenceOutputDType == 0) {
                     (nnApplication.mModelReg2).inference(clustererNNShadow.mInputData_16, iSize, clustererNNShadow.mOutputDataReg2_16);
@@ -1102,7 +1102,7 @@ int32_t GPUChainTracking::RunTPCClusterizer(bool synchronizeOutput)
                     (nnApplication.mModelReg2).inference(clustererNNShadow.mInputData_32, iSize, clustererNNShadow.mOutputDataReg2_32);
                   }
                 }
-                if(GetProcessingSettings().debugLevel >= 1 && doGPU) { nnTimers[3*lane + 2]->Stop(); }
+                if(GetProcessingSettings().debugLevel >= 1 && (doGPU || lane < 4)) { nnTimers[3*lane + 2]->Stop(); }
               }
               if (nn_settings.nnClusterizerVerbosity > 3) {
                 LOG(info) << "(NNCLUS, GPUChainTrackingClusterizer, this=" << this << ") Done with NN regression inference. Loop=" << batch << ". (clustererNN=" << &clustererNN << ", clustererNNShadow=" << &clustererNNShadow << ")";