Closed mariadalfonso closed 2 years ago
The BTagSFProducer.cc is very slow (see pie charts) https://github.com/cms-sw/cmssw/blob/master/PhysicsTools/NanoAOD/plugins/BTagSFProducer.cc https://github.com/cms-sw/cmssw/blob/master/PhysicsTools/NanoAOD/python/btagWeightTable_cff.py
In addition thes two floats here are using an outdated SF set 94X CSV files https://cms-nanoaod-integration.web.cern.ch/integration/cms-swmaster/mc106Xul16_doc.html#btagWeight
15% of theCPU is consumed in reading the BTag csv files https://cms-reco-profiling.web.cern.ch/cms-reco-profiling/cgi-bin/igprof-navigator/releases/CMSSW_12_1_0_pre4/11834.21/step5/cpu_endjob/37
Short term plan:
[ ] remove this from the Run3 sequence (can be applied anyway a posteriori when correct calibration is available)
[ ] update the BTagSFProducer reading the json format if deemed necessary
@mondalspandan
speed up in cms-sw/35856
we can close this issues, now the BTags weights are applied as json and with correctionlib
I think the btag weight is still being produced, I see it in the v9 docs and here... shouldn't we just remove it?
The BTagSFProducer.cc is very slow (see pie charts) https://github.com/cms-sw/cmssw/blob/master/PhysicsTools/NanoAOD/plugins/BTagSFProducer.cc https://github.com/cms-sw/cmssw/blob/master/PhysicsTools/NanoAOD/python/btagWeightTable_cff.py
In addition thes two floats here are using an outdated SF set 94X CSV files https://cms-nanoaod-integration.web.cern.ch/integration/cms-swmaster/mc106Xul16_doc.html#btagWeight
15% of theCPU is consumed in reading the BTag csv files https://cms-reco-profiling.web.cern.ch/cms-reco-profiling/cgi-bin/igprof-navigator/releases/CMSSW_12_1_0_pre4/11834.21/step5/cpu_endjob/37
Short term plan:
[ ] remove this from the Run3 sequence (can be applied anyway a posteriori when correct calibration is available)
[ ] update the BTagSFProducer reading the json format if deemed necessary