[hack] Force one GPU queue per framework stream

[fwyzard]

PR description:

Force one GPU queue per framework stream.

PR validation:

It kind of works, but it's definitely not production ready.

[cmsbuild]

cms-bot internal usage

[cmsbuild]

-code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-49547/47059

• There are other open Pull requests which might conflict with changes you have proposed:
  • File HeterogeneousCore/AlpakaCore/interface/alpaka/EDMetadataAcquireSentry.h modified in PR(s): Reuse the device queue from produce in transform #49545
  • File HeterogeneousCore/AlpakaCore/interface/alpaka/ProducerBase.h modified in PR(s): Reuse the device queue from produce in transform #49545
  • File HeterogeneousCore/AlpakaCore/src/alpaka/EDMetadataAcquireSentry.cc modified in PR(s): Reuse the device queue from produce in transform #49545

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-49547/47059/code-format.patch
  e.g. curl -k https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-49547/47059/code-format.patch | patch -p1
  You can also run scram build code-format to apply code format directly

[fwyzard]

please test for el9_amd64_gcc13

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-49547/47060

• There are other open Pull requests which might conflict with changes you have proposed:
  • File HeterogeneousCore/AlpakaCore/interface/alpaka/EDMetadataAcquireSentry.h modified in PR(s): Reuse the device queue from produce in transform #49545
  • File HeterogeneousCore/AlpakaCore/interface/alpaka/ProducerBase.h modified in PR(s): Reuse the device queue from produce in transform #49545
  • File HeterogeneousCore/AlpakaCore/src/alpaka/EDMetadataAcquireSentry.cc modified in PR(s): Reuse the device queue from produce in transform #49545

[cmsbuild]

Pull request #49547 was updated.

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-b862d4/49803/summary.html
COMMIT: 01db5b2
CMSSW: CMSSW_16_0_X_2025-12-03-2300/el9_amd64_gcc13
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/49547/49803/install.sh to create a dev area with all the needed externals and cmssw changes.

Comparison Summary

Summary:

• You potentially added 276 lines to the logs
• ROOTFileChecks: Some differences in event products or their sizes found
• Reco comparison results: 72836 differences found in the comparisons
• Reco comparison had 4 failed jobs
• DQMHistoTests: Total files compared: 53
• DQMHistoTests: Total histograms compared: 4273241
• DQMHistoTests: Total failures: 368608
• DQMHistoTests: Total nulls: 301
• DQMHistoTests: Total successes: 3904312
• DQMHistoTests: Total skipped: 20
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: -3.6079999999999997 KiB( 52 files compared)
• DQMHistoSizes: changed ( 10224.0 ): -0.127 KiB SiStrip/MechanicalView
• DQMHistoSizes: changed ( 13034.0 ): 0.293 KiB SiStrip/MechanicalView
• DQMHistoSizes: changed ( 17034.0 ): 0.298 KiB SiStrip/MechanicalView
• DQMHistoSizes: changed ( 18634.0 ): 1.305 KiB SiStrip/MechanicalView
• DQMHistoSizes: changed ( 2024.0040001 ): 0.012 KiB JetMET/SUSYDQM
• DQMHistoSizes: changed ( 250202.181 ): -0.077 KiB SiStrip/MechanicalView
• DQMHistoSizes: changed ( 25202.0 ): 0.288 KiB SiStrip/MechanicalView
• DQMHistoSizes: changed ( 7.3 ): -5.600 KiB SiStrip/MechanicalView
• Checked 227 log files, 198 edm output root files, 53 DQM output files

[cmsbuild]

Milestone for this pull request has been moved to CMSSW_16_1_X. Please open a backport if it should also go in to CMSSW_16_0_X.