Follow up to the Alpaka integration in CMSSW

[fwyzard]

PR description:

HeterogeneousCore/AlpakaInterface

Rewrite the work division utilities:

• rewrite the uniform element kernel loops.
• rewrite the independent element kernel loops.
• remove the obsolete alpakatools utilities.

Rewrite the zeroAndInit kernel using the alpakatools utilities.
Rename elements_with_stride to uniform_elements in user code.

RecoLocalTracker/SiPixelClusterizer

Implement fixes that were lost during the CUDA-to-Alpaka migration.

Collect the updates from #39711, #42618, #42977 and #43052, that were missing from the Alpaka implementation:

• skip invalid or corrupted ROCs: make the Alpaka implementation of the pixel unpacker skip spurious ROCs, similar to the legacy and CUDA versions of the unpacker, and store the
  invalid ROC number error (errorType=36);
• fix the decoding of channels with timeout errors (errorType=27);
• synchronise the treatment of pixel errorType=26, 27, 30 with the legacy code;
• disable printf statements at compile time.

RecoTracker/PixelSeeding

Do not call the fishbone for events with pixel hits only in BPIX1 (reimplement #35638).

RecoTracker/MeasurementDet

Add a check on the pixel ROC range, to avoid overflows in the (unsigned) range differences.

DataFormats/TrackingRecHitSoA

Extend TrackingRecHits collections for testing and their unit tests.

CondFormats/SiPixelObjects

Clean up includes and dependencies, and avoid including framework headers in the device compilation.

Geometry/CommonTopologies

Update comments to pixel topologies.

Other changes

Fix include guards after files were moved to different packages.
Clean up include files.
Simplify nested namespace declarations.

PR validation:

Unit tests pass.

Validated running a recent HLT menu over 20k events without issues, both on GPU and on CPU, on data and MC: the alpaka implementation running on CPU or GPU gives the same results as the CUDA implementation, within the same non-reproducibility margins (1 event or up to 1% for high-rate triggers).

Backport status

The backport to CMSSW_14_0_X for 2024 data taking is in #43879.

[cmsbuild]

cms-bot internal usage

[fwyzard]

assign heterogeneous

[cmsbuild]

-code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-43853/38691

ERROR: Build errors found during clang-tidy run.

RecoLocalTracker/SiPixelClusterizer/plugins/alpaka/SiPixelRawToClusterKernel.dev.cc:636:41: error: expected '(' for function-style cast or type construction [clang-diagnostic-error]
  636 |                               CalibDigis<true>{},
      |                               ~~~~~~~~~~^
RecoLocalTracker/SiPixelClusterizer/plugins/alpaka/SiPixelRawToClusterKernel.dev.cc:636:46: error: initializer list cannot be used on the right hand side of operator '>' [clang-diagnostic-error]
  636 |                               CalibDigis<true>{},
      |                                              ^~~
RecoLocalTracker/SiPixelClusterizer/plugins/alpaka/SiPixelRawToClusterKernel.dev.cc:645:41: error: expected '(' for function-style cast or type construction [clang-diagnostic-error]
  645 |                               CalibDigis<false>{},
      |                               ~~~~~~~~~~^
RecoLocalTracker/SiPixelClusterizer/plugins/alpaka/SiPixelRawToClusterKernel.dev.cc:645:47: error: initializer list cannot be used on the right hand side of operator '>' [clang-diagnostic-error]
  645 |                               CalibDigis<false>{},
      |                                               ^~~
Suppressed 571 warnings (571 in non-user code).
--
gmake: *** [config/SCRAM/GMake/Makefile.coderules:129: code-checks] Error 2
gmake: *** [There are compilation/build errors. Please see the detail log above.] Error 2

[fwyzard]

test parameters:

• enable = gpu
• workflows = 12434.402,12434.403,12434.404
• workflows_gpu = 12434.402,12434.403,12434.404
• workflow_opts = -w upgrade
• workflow_opts_gpu = -w upgrade

[fwyzard]

please test

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-43853/38692

• This PR adds an extra 36KB to repository

• There are other open Pull requests which might conflict with changes you have proposed:

  • File RecoLocalTracker/SiPixelClusterizer/plugins/alpaka/SiPixelRawToClusterKernel.dev.cc modified in PR(s): Update Alpaka to version 1.1.0 #43772

[cmsbuild]

New categories assigned: heterogeneous

@fwyzard,@makortel you have been requested to review this Pull request/Issue and eventually sign? Thanks

[cmsbuild]

Pull request #43853 was updated. @jfernan2, @francescobrivio, @subirsarkar, @consuegs, @perrotta, @srimanob, @saumyaphor4252, @mandrenguyen can you please check and sign again.

[srimanob]

+Upgrade

[perrotta]

+alca

[perrotta]

+db

[mandrenguyen]

are the tests stuck?

[smuzaffar]

are the tests stuck?

No @mandrenguyen , tests are running. Jenkins is very busy today that is why tests are waiting for free nodes

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-536c0d/37363/summary.html
COMMIT: c7d3641
CMSSW: CMSSW_14_1_X_2024-02-12-1100/el8_amd64_gcc12
Additional Tests: GPU
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/43853/37363/install.sh to create a dev area with all the needed externals and cmssw changes.

Comparison Summary

Summary:

• You potentially removed 101 lines from the logs
• Reco comparison results: 44 differences found in the comparisons
• DQMHistoTests: Total files compared: 50
• DQMHistoTests: Total histograms compared: 3280699
• DQMHistoTests: Total failures: 5
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 3280672
• DQMHistoTests: Total skipped: 22
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 49 files compared)
• Checked 211 log files, 173 edm output root files, 50 DQM output files
• TriggerResults: no differences found

GPU Comparison Summary

Summary:

• No significant changes to the logs found
• Reco comparison results: 70 differences found in the comparisons
• DQMHistoTests: Total files compared: 5
• DQMHistoTests: Total histograms compared: 71813
• DQMHistoTests: Total failures: 2870
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 68943
• DQMHistoTests: Total skipped: 0
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 4 files compared)
• Checked 19 log files, 22 edm output root files, 5 DQM output files
• TriggerResults: no differences found

[mandrenguyen]

+1

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

[antoniovilela]

+1

[Martin-Grunewald]

@fwyzard
Is there a backport to 14_0 for this PR? Is it #43879 ?

[fwyzard]

Yes, #43879 .
I've updated the description with a link to it.