enforce memory order to avoid crash on Ampere

[VinInn]

title says all.
very technical.
no regression expected (but no crash on A100)

[innocent@workergpu53 patatrack-scripts]$ ./benchmark ../pixelraw/gpuBench.py
2 CPUs:
  0: Intel(R) Xeon(R) Platinum 8268 CPU @ 2.90GHz (24 cores, 24 threads)
  1: Intel(R) Xeon(R) Platinum 8268 CPU @ 2.90GHz (24 cores, 24 threads)

1 visible NVIDIA GPUs:
  0: A100-PCIE-40GB (UUID: GPU-c60c7fb1-8112-8f34-7cbf-e3cef8cd2c12)

Warming up

Running 4 times over 4600 events with 4 jobs, each with 8 threads, 8 streams and 1 GPUs
  1303.3 ±   0.5 ev/s (4400 events, 97.6% overlap)
  1308.7 ±   0.8 ev/s (4400 events, 97.8% overlap)
  1288.4 ±   1.0 ev/s (4400 events, 98.0% overlap)
  1304.9 ±   1.0 ev/s (4400 events, 99.2% overlap)
 --------------------
  1301.3 ±   8.9 ev/s

[innocent@workergpu53 patatrack-scripts]$ ./benchmark ../pixelraw/quadBench.py
2 CPUs:
  0: Intel(R) Xeon(R) Platinum 8268 CPU @ 2.90GHz (24 cores, 24 threads)
  1: Intel(R) Xeon(R) Platinum 8268 CPU @ 2.90GHz (24 cores, 24 threads)

1 visible NVIDIA GPUs:
  0: A100-PCIE-40GB (UUID: GPU-c60c7fb1-8112-8f34-7cbf-e3cef8cd2c12)

Warming up

Running 4 times over 4600 events with 4 jobs, each with 8 threads, 8 streams and 1 GPUs
  2424.2 ±   2.8 ev/s (4400 events, 98.2% overlap)
  2428.2 ±   3.2 ev/s (4400 events, 98.4% overlap)
  2403.7 ±   2.5 ev/s (4400 events, 95.3% overlap)
  2422.7 ±   4.6 ev/s (4400 events, 96.8% overlap)
 --------------------
  2419.7 ±  10.9 ev/s

[VinInn]

@cmsbuild , please test

[VinInn]

enable gpu

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-34386/23776

• This PR adds an extra 16KB to repository

[cmsbuild]

A new Pull Request was created by @VinInn (Vincenzo Innocente) for master.

It involves the following packages:

RecoPixelVertexing/PixelTriplets

@perrotta, @jpata, @slava77 can you please review it and eventually sign? Thanks.
@makortel, @felicepantaleo, @GiacomoSguazzoni, @JanFSchulte, @rovere, @VinInn, @mmusich, @mtosi, @dgulhan this is something you requested to watch as well.
@silviodonato, @dpiparo, @qliphy you are the release manager for this.

cms-bot commands are listed here

[VinInn]

@fwyzard

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-9a8a26/16555/summary.html
COMMIT: f5c01ea
CMSSW: CMSSW_12_0_X_2021-07-06-2300/slc7_amd64_gcc900
Additional Tests: GPU
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/34386/16555/install.sh to create a dev area with all the needed externals and cmssw changes.

GPU Comparison Summary

Summary:

• No significant changes to the logs found
• Reco comparison results: 8 differences found in the comparisons
• DQMHistoTests: Total files compared: 4
• DQMHistoTests: Total histograms compared: 19723
• DQMHistoTests: Total failures: 697
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 19026
• DQMHistoTests: Total skipped: 0
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 3 files compared)
• Checked 12 log files, 9 edm output root files, 4 DQM output files
• TriggerResults: no differences found

Comparison Summary

Summary:

• No significant changes to the logs found
• Reco comparison results: 3 differences found in the comparisons
• DQMHistoTests: Total files compared: 38
• DQMHistoTests: Total histograms compared: 2785924
• DQMHistoTests: Total failures: 6
• DQMHistoTests: Total nulls: 1
• DQMHistoTests: Total successes: 2785895
• DQMHistoTests: Total skipped: 22
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.004 KiB( 37 files compared)
• DQMHistoSizes: changed ( 312.0 ): 0.004 KiB MessageLogger/Warnings
• Checked 160 log files, 37 edm output root files, 38 DQM output files
• TriggerResults: no differences found

[slava77]

+reconstruction

for #34386 f5c01ea

• the code changes are technical, in line with the PR description
• jenkins tests pass and comparisons with the baseline show small differences in the GPU test of pixelTracking in wf 10824.502
  • from the past GPU pixel tracking code reviews it is expected that the results are not strictly repeatable
  • perhaps incidentally, the changes seem to be for tracks with fairly large chi2
    

[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. @silviodonato, @dpiparo, @qliphy (and backports should be raised in the release meeting by the corresponding L2)

[qliphy]

+1