Improve MultiTrackValidator performance #8988
Conversation
…ulting bool later
- Remove string comparisons - Remove copying of ValueMaps - Tidy up monitoring implementation Using the same ValueMap for all input track collections works for track validation, as all input collections are Refs to the original generalTracks.
- Break as soon as know that the return value is false - Reorder tests such that lighter ones are checked first - Reduce repeated calculations (minor)
Reduce repeated computations in - MTVHistoProducerAlgoForTracker::fill_simAssociated_recoTrack_histos() - MTVHistoProducerAlgoForTracker::fill_generic_recoTrack_histos() - MTVHistoProducerAlgoForTracker::fill_ResoAndPull_recoTrack_histos() Streamline MTVHistoProducerAlgoForTracker::fill_generic_recoTrack_histos(), plus other cleanups.
This reduces the amount of repeated calculations of those. Eliminating all repeated calculations would require fancier caching, or precalculating them for all TPs (such that they would be accessible through Refs, taking into account that label_tp_effic and label_tp_fake may point to different TP collections...).
|
A new Pull Request was created by @makortel (Matti Kortelainen) for CMSSW_7_5_X. Improve MultiTrackValidator performance It involves the following packages: SimDataFormats/TrackingAnalysis @civanch, @nclopezo, @danduggan, @mdhildreth, @cmsbuild, @deguio can you please review it and eventually sign? Thanks. |
|
please test |
|
this is very nice.. :) |
|
The tests are being triggered in jenkins. |
|
+1 |
|
This pull request is fully signed and it will be integrated in one of the next CMSSW_7_5_X IBs unless changes (tests are also fine). This pull request requires discussion in the ORP meeting before it's merged. @davidlange6, @nclopezo, @smuzaffar |
|
+1 |
Improve MultiTrackValidator performance
This PR reduces the time spent in MultiTrackValidator in RAW2DIGI,RECO,VALIDATION,DQM jobs using TTbar + 35@25ns PU events by a factor of 25. The main sources of improvement are (in descending order, roughly)
analyze()A side effect is that now the dEdx-histograms are filled also for the subcollections of generalTracks, and the histograms are not booked for MTV instances other than the one for FullSim pp validation.
Tested in CMSSW_7_5_X_2015-05-03-1100+#8888 and rebased (+tested with limited matrix) on top of CMSSW_7_5_X_2015-05-07-1100 for IB containing #8888. No changes are expected in already-filled histograms.
@rovere @VinInn