RecoLocalTracker/SiPixelRecHits: changes to vectorize loops in VVIObjF::VVIObjF constructor

[gartung]

Using Intel Advisor to profile workflow r39634.21 with a CMSSW_12_5_LTO integration build release showed the top time consuming loop with floating point operation is in VVIObjF::VVIObjF constructor.

"1381";"[loop in VVIObjF::VVIObjF at log.h:179]";"41.230s";"28.288s";"Scalar";"";"";"";"";"log.h:179";"libRecoLocalTrackerSiPixelRecHits.so"

With this change the loop calculations are vectorized and the time is reduced.

"1772";"[loop in VVIObjF::VVIObjF at log.h:179]";"3.824s";"3.824s";"Vectorized (Body)";"";"SSE2";"";"";"log.h:179";"libRecoLocalTrackerSiPixelRecHits.so"

Further profiling is being done with CMSSW_12_5_X integration build release.

[cmsbuild]

-code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-39218/31837

• This PR adds an extra 16KB to repository

Code check has found code style and quality issues which could be resolved by applying following patch(s)

• code-format:
  https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-39218/31837/code-format.patch
  e.g. curl -k https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-39218/31837/code-format.patch | patch -p1
  You can also run scram build code-format to apply code format directly

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-39218/31838

• This PR adds an extra 16KB to repository

[cmsbuild]

A new Pull Request was created by @gartung (Patrick Gartung) for master.

It involves the following packages:

• RecoLocalTracker/SiPixelRecHits (reconstruction)

@jpata, @cmsbuild, @mandrenguyen, @clacaputo can you please review it and eventually sign? Thanks.
@mtosi, @VourMa, @felicepantaleo, @GiacomoSguazzoni, @JanFSchulte, @rovere, @VinInn, @OzAmram, @ferencek, @dkotlins, @gpetruc, @mmusich, @threus, @tvami 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

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-39218/31840

• This PR adds an extra 20KB to repository

[cmsbuild]

Pull request #39218 was updated. @jpata, @cmsbuild, @mandrenguyen, @clacaputo can you please check and sign again.

[gartung]

enable profiling

[gartung]

please test

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-9899c9/27166/summary.html
COMMIT: d042492
CMSSW: CMSSW_12_6_X_2022-08-29-1100/el8_amd64_gcc10
Additional Tests: PROFILING
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/39218/27166/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: 4 differences found in the comparisons
• DQMHistoTests: Total files compared: 51
• DQMHistoTests: Total histograms compared: 3695708
• DQMHistoTests: Total failures: 21
• DQMHistoTests: Total nulls: 1
• DQMHistoTests: Total successes: 3695664
• DQMHistoTests: Total skipped: 22
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.004 KiB( 50 files compared)
• DQMHistoSizes: changed ( 312.0 ): 0.004 KiB MessageLogger/Warnings
• Checked 212 log files, 49 edm output root files, 51 DQM output files
• TriggerResults: no differences found

[mandrenguyen]

+reconstruction
Technical. Diffs in reco comparisons appears negligible

[cmsbuild]

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, @rappoccio (and backports should be raised in the release meeting by the corresponding L2)

[tvami]

assign trk-dpg

[cmsbuild]

New categories assigned: trk-dpg

@connorpa,@mmusich,@tsusa you have been requested to review this Pull request/Issue and eventually sign? Thanks

[mmusich]

you have been requested to review this Pull request/Issue and eventually sign

the last round of tests (also the logs) look totally clean and we have in principle monitoring of the cluster probability in the offline DQM.
Not sure if some specific validation would still be required here, but I would be in favor of just signing.

[gartung]

This is the branch I used for the comparison of the arrays calculated with the iterative loop and with the vectorized loop. It produces a lot of output so it should only be used on one event.
master...gartung:cmssw:gartung-vviobjf-vectorize-test

[mmusich]

+1

• based on RecoLocalTracker/SiPixelRecHits: changes to vectorize loops in VVIObjF::VVIObjF constructor #39218 (comment)

[cmsbuild]

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, @rappoccio (and backports should be raised in the release meeting by the corresponding L2)

[perrotta]

+1

• I understand that some further optimization is still possible, based on RecoLocalTracker/SiPixelRecHits: changes to vectorize loops in VVIObjF::VVIObjF constructor #39218 (comment) and RecoLocalTracker/SiPixelRecHits: changes to vectorize loops in VVIObjF::VVIObjF constructor #39218 (comment)