cms-sw
mkFit updates aiming to improve pixelLess iteration and related
Motivated by observed inefficiency in the pixelLess iteration track building for low-pt displaced tracks, several features/bugs were fixed and improvements to the algorithm were added:
- consistently sort seeds in iterations without seed merging
- fix handling of candidates with overlapping hits
- make sure that Kalman updates in the barrel stay on the update plane
- fix a propagation condition check bug in propagation to radius calculation of several Jacobian elements
- fix functions to count the number of matched hits and layers (used in pixelLess and detachedTriplet trajectory filtering)
- explicitly account for the layer thickness in the endcap to determine the phi window to preselect hits (active only for the pixelLess iteration now)
- relax the number of hit requirement for displaced track candidate filter in the pixelLess (endcaps only)
- disable track candidate DNN selection in the pixelLess iteration
- modify seed merging logic for the lowPtQuad iteration for the future developments (this iteration is not yet in the mkFit defaul production setup)
MTV plots, including timing, are available for ttbar relval with pileup [for the standard DQM selections](https://slava77.web.cern.ch/slava77/mic/CMSSW_12_5_0_pre4-pixmkf /RelValTTbar_14TeV/mtv-mkf5-mkf6-orig-Aug15POG-tr102-08be050) (signal tracks have pt>0.9 GeV in the efficiency plots) and separately [for signal simtracks selected with pt<0.9 GeV](https://slava77.web.cern.ch/slava77/mic/CMSSW_12_5_0_pre4-pixmkf/RelValTTbar_14TeV/mtv-mkf5-mkf6-orig-Aug15POG-tr102-08be050-lopt for low-pt)
Expectations:
- slightly longer tracks in all mkFit iterations
- a significant increase in efficiency of reconstructing displaced tracks in the pixelLess iteration
- tracking time goes up by about 3% mainly from longer track fitting time downstream of mkFit, a large fraction of which is from disabling of the candidate DNN selection in the pixelLess iteration
Incremental comparisons are available in https://github.com/trackreco/cmssw/pull/102
Presentation in the Aug 15 TRK POG meeting https://indico.cern.ch/event/1189133/#3-mkfit-updates
Please consider for 12_5_0_pre5
+code-checks
Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-39055/31557
-
This PR adds an extra 84KB to repository
-
There are other open Pull requests which might conflict with changes you have proposed:
- File RecoTracker/MkFitCMS/src/MkStdSeqs.cc modified in PR(s): #38683
A new Pull Request was created by @slava77 (Slava Krutelyov) for master.
It involves the following packages:
- RecoTracker/IterativeTracking (reconstruction)
- RecoTracker/MkFitCMS (reconstruction)
- RecoTracker/MkFitCore (reconstruction)
@jpata, @cmsbuild, @mandrenguyen, @clacaputo can you please review it and eventually sign? Thanks. @VourMa, @makortel, @felicepantaleo, @GiacomoSguazzoni, @JanFSchulte, @rovere, @VinInn, @ebrondol, @gpetruc, @mmusich, @mtosi, @dgulhan this is something you requested to watch as well. @perrotta, @dpiparo, @qliphy, @rappoccio you are the release manager for this.
cms-bot commands are listed here
+1
Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-edbc39/26807/summary.html
COMMIT: 08be05004bdf2d2cbf31a54b8a689cb27ff2a78c
CMSSW: CMSSW_12_5_X_2022-08-14-0000/el8_amd64_gcc10
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/39055/26807/install.sh to create a dev area with all the needed externals and cmssw changes.
Comparison Summary
Summary:
- No significant changes to the logs found
- Reco comparison results: 20473 differences found in the comparisons
- DQMHistoTests: Total files compared: 51
- DQMHistoTests: Total histograms compared: 3692476
- DQMHistoTests: Total failures: 33628
- DQMHistoTests: Total nulls: 4
- DQMHistoTests: Total successes: 3658822
- DQMHistoTests: Total skipped: 22
- DQMHistoTests: Total Missing objects: 0
- DQMHistoSizes: Histogram memory added: -0.012 KiB( 50 files compared)
- DQMHistoSizes: changed ( 138.4 ): -0.004 KiB JetMET/SUSYDQM
- DQMHistoSizes: changed ( 138.5 ): -0.008 KiB JetMET/SUSYDQM
- Checked 212 log files, 49 edm output root files, 51 DQM output files
- TriggerResults: found differences in 3 / 50 workflows
@cms-sw/tracking-pog-l2 were there any outstanding questions you would like to be addressed from the meeting yesterday? (I did not have that impression, but just to check)
were there any outstanding questions you would like to be addressed from the meeting yesterday?
I don't think there was any specific objection to let this PR be integrated, provided it wasn't reactivating mkFit on the pixelLess iteration by default (that would require more work). Having said that, I haven't yet looked at the code changes if that's a review that is requested.
Thanks. It was more of a generic question than a request for review (which is of course always welcome).
+reconstruction
- bugfixes for mkFit
- consequently, changes in tracks and downstream quantities (see the slides in the description for details)
- a slowdown of ~3% in tracking is expected from this
This pull request is fully signed and it will be integrated in one of the next master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @perrotta, @dpiparo, @qliphy, @rappoccio (and backports should be raised in the release meeting by the corresponding L2)
