New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Turning on ClusterRepair for edge hits only #26263
Turning on ClusterRepair for edge hits only #26263
Conversation
The code-checks are being triggered in jenkins. |
+code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-26263/8938
|
A new Pull Request was created by @cmantill (Cristina Ana Mantilla Suarez) for master. It involves the following packages: RecoLocalTracker/SiPixelRecHits @cmsbuild, @perrotta, @slava77 can you please review it and eventually sign? Thanks. cms-bot commands are listed here |
Standard runTheMatrix tests have been performed -> but they are failing due to a global tag error [1] [1] |
Quick update: I repeated the runTheMatrix standard tests on lxplus and most of them were completed (it seems that the error from before was something regarding the connection in the fermilab cluster). The following workflows fail: [2] |
Hi @tocheng, the PR fails because it wants to use the Prompt GTs. PR26281 seems to fix this. Can we do some kind of workaround to pass these tests? |
@tvami I tried your GT: 105X_dataRun2_v8 and this works for all 3 workflows that were failing before. But I tried the PR26281 and it still does not work, e.g for workflow 136.788:
|
@tvami @cmantill |
@tocheng I don't believe that's the problem for issue reported in #26263 (comment), because all the other workflows run fine excepted there 3:
By the way, phase-0 detector reconstruction is not touched intentionally by this PR, see : As far as I can see all of the three failing workflows use the same 2D template tag: SiPixel2DTemplateDBObject_38T_v1_prompt: 136.788 => GT: and all of these GTs use the same tag [0]
an easy workaround would be if you could supply new Global Tag for the prompt-like workflows that instead of the tag [0]
|
Hi, |
Hi @tocheng, I've created |
The code-checks are being triggered in jenkins. |
-code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-26263/9101
Code check has found code style and quality issues which could be resolved by applying a patch in https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-26263/9101/git-diff.patch You can run |
@mmusich @tocheng I am confused by this patch: However, the first lines are reverting the commit we just did to update pixel local reco conditions for 2017. Please let me know how to proceed. |
@cmantill Can you rebase the PR to the most recent IB? |
@perrotta We found a bug in the current implementation to call the 2D fit: For cases where we are in the correct layers (BPIX 2,3,4) but not edgehits, the theClusterParam.recommended2D_ is set to a garbage value (!=0) that later passes the 2D call. Thus, instead of calling 2D reco for 0.5% of the hits, we were calling for 10% of them. We have fixed that in the latest commit by setting theClusterParam.recommended2D_ false by default. I should clarify that this bug should have not affected the hit residuals by a significant amount. For the hits affected by this bug, the 2D Reco would have been running with option edgeY=0, i.e. when CR is not trying to repair anything and is pretty much the same as the 1D Reco. It was only making things much slower as shown by the results below. With the fix introduced in the latest commit we have updated the timing and CPU time tests as follows:
Although the time increases for tracking, the increased time is less than that shown before the bugfix. We think that the overall increase of time is due to random increases of other modules (also this comparison is done to a test with CROFF two days ago). [1] https://cmantill.web.cern.ch/cmantill/tmp/newtiming_afterbugfix.log |
please test |
The tests are being triggered in jenkins. |
Comparison job queued. |
Comparison is ready Comparison Summary:
|
After the last fix, overall timing is quite comparable while using ClusterRepair on a limited number of hits, as in this PR, or having it turned off, as in the baseline IB. I think that the differences you report in #26263 (comment) are mostly due to an overall normalization factor, as I see no relevant increase while running on the usual 100 evt of the TTbar PU sample 11024. The number of hits affected being now rather limited, as you pointed out, also the effects on general tracks justshow up as small deviations from the baseline, see A detailed validation cannot be done with just those small test samples, but some is certainly expected based on pre4, provided this PR can get merged for it |
+1
|
+1 |
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. @davidlange6, @slava77, @smuzaffar, @fabiocos (and backports should be raised in the release meeting by the corresponding L2) |
+1 |
PR description:
This PR turns ON ClusterRepair CPE for edge hits only.
It adds an option to the configuration file so that is possible to choose which pixel sub-detectors/layers to run on. By default this option is: "PXB 2","PXB 3","PXB 4" which refers to layers 2,3 and 4 of PixelBarrel.
The PR further adds a flag to turn on/off CR for truncated clusters - this flag is false by default.
The PR follows from the discussion on the DPG meeting https://indico.cern.ch/event/787658/?showDate=all&showSession=6
The following are updated tags for autoCond:
Update pixel local reco conditions in the offline GTs
Data
New offline GT is
https://cms-conddb.cern.ch/cmsDbBrowser/list/Prod/gts/106X_dataRun2_v4
Diff
https://cms-conddb.cern.ch/cmsDbBrowser/diff/Prod/gts/106X_dataRun2_v4/106X_dataRun2_v1
New offline relval GT is
https://cms-conddb.cern.ch/cmsDbBrowser/list/Prod/gts/106X_dataRun2_relval_v2
Diff
https://cms-conddb.cern.ch/cmsDbBrowser/diff/Prod/gts/106X_dataRun2_relval_v2/106X_dataRun2_relval_1
Only 2017 conditions are updated in GenErr, Lorentz angle and 1D template.
2D template changes in 2016 and 2017, but it was not consumed in the past till this PR.
https://cms-conddb.cern.ch/cmsDbBrowser/diff/Prod/tags/SiPixelGenErrorDBObject_38T_forPR26263/SiPixelGenErrorDBObject_38T_v8_offline
https://cms-conddb.cern.ch/cmsDbBrowser/diff/Prod/tags/SiPixelLorentzAngle_38T_forPR26263/SiPixelLorentzAngle_v10_offline
https://cms-conddb.cern.ch/cmsDbBrowser/diff/Prod/tags/SiPixelTemplateDBObject_38T_forPR26263/SiPixelTemplateDBObject_38T_v13_offline
https://cms-conddb.cern.ch/cmsDbBrowser/diff/Prod/tags/SiPixel2DTemplateDBObject_38T_v2_offline/SiPixel2DTemplateDBObject_38T_v1_offline
PR validation:
It has been tested locally on data - using cmsDriver step3 - (on JetHT and on a run for which a 2D template exists) and verified by printouts that is called only for the layers that are defined in the configuration file.
It has also been tested using standard runTheMatrix tests using 105X_dataRun2_Candidate_2019_04_03_13_37_01 GT and checking step3 independently for runs corresponding to each 2017 IOV.