Skip to content
Snippets Groups Projects

AthContainers: Switch to using ThinningInfo for the lossy float compression inside copyAuxStoreThinned

This MR builds on !35444 (merged). With this MR we're switching to using ThinningInfo to determine whether lossy float compression should be applied to a given AuxID and apply it in copyAuxStoreThinned. Parts that refer to the now-defunct IAuxStoreCompression are completely removed. The next step is to completely get rid of this interface and its implementation in AuxContainerBase.

One key item that was previously missing was a regression test that ensures the lossy float compression is applied correctly. copyAuxStoreThinned_test is extended to address that problem. Since by default we don't yet compress anything, the changes should be transparent to existing workflows. However, I'm asking for full-unit-tests to make sure there are no unintended side effects.

The next step will be cleaning up the previously implemented infrastructure as noted above.

cc: @ssnyder @mnowak @gemmeren

Merge request reports

Loading
Loading

Activity

Filter activity
  • Approvals
  • Assignees & reviewers
  • Comments (from bots)
  • Comments (from users)
  • Commits & branches
  • Edits
  • Labels
  • Lock status
  • Mentions
  • Merge request status
  • Tracking
  • Johannes Junggeburth
  • added 1 commit

    Compare with previous version

  • This merge request affects 1 package:

    • Control/AthContainers

    Adding @ssnyder as watcher

  • Alaettin Serhan Mete resolved all threads

    resolved all threads

  • :negative_squared_cross_mark: CI Result FAILURE (hash 3677d9fa)

    Athena AthSimulation AthGeneration AnalysisBase
    externals :white_check_mark: :white_check_mark: :white_check_mark: :white_check_mark:
    cmake :white_check_mark: :white_check_mark: :white_check_mark: :white_check_mark:
    make :white_check_mark: :white_check_mark: :white_check_mark: :white_check_mark:
    required tests :o: :white_check_mark: :o: :o:
    optional tests :white_check_mark: :white_check_mark: :white_check_mark: :white_check_mark:

    Full details available on this CI monitor view
    :white_check_mark: Athena: number of compilation errors 0, warnings 0
    :white_check_mark: AthSimulation: number of compilation errors 0, warnings 0
    :white_check_mark: AthGeneration: number of compilation errors 0, warnings 0
    :white_check_mark: AnalysisBase: number of compilation errors 0, warnings 0
    :pencil: For experts only: Jenkins output [CI-MERGE-REQUEST-CC7 18653]

  • Looking at the last results, I see some additional tests (compared to the latest nightly) are failing w/:

    -Py:PropertyProxy  WARNING Found 4 duplicates among the 1067 genConfDb files :
    -Py:PropertyProxy  WARNING --------------------------------------------------
    -Py:PropertyProxy  WARNING   -<component name>: <module> - [ <duplicates> ]
    -Py:PropertyProxy  WARNING --------------------------------------------------
    -Py:PropertyProxy  WARNING   -HLTTest__TestHypoAlg: DecisionHandling.DecisionHandlingConf - ['TrigUpgradeTest.TrigUpgradeTestConf']
    -Py:PropertyProxy  WARNING   -HLTTest__TestHypoTool: DecisionHandling.DecisionHandlingConf - ['TrigUpgradeTest.TrigUpgradeTestConf']
    -Py:PropertyProxy  WARNING   -HLTTest__TestInputMaker: DecisionHandling.DecisionHandlingConf - ['TrigUpgradeTest.TrigUpgradeTestConf']
    -Py:PropertyProxy  WARNING   -HLTTest__TestRecoAlg: DecisionHandling.DecisionHandlingConf - ['TrigUpgradeTest.TrigUpgradeTestConf']
    -Py:PropertyProxy  WARNING Fix your cmt/requirements file !!

    which is probably an incremental build issue. The only fishy one is TileTPCnv_TileDigitsCnv_p2_test_ctest which segfaults. I ran all the tests in TileTPCnv locally w/ my changes and all succeeds so I'm pretty sure that's not related to my changes either.

    I'll be partially offline next week. So, it would be great to get this done before that if possible. Please let me know if there are any concrete suggestions. Thanks.

  • Loading
  • Loading
  • Loading
  • Loading
  • Loading
  • Loading
  • Loading
  • Loading
  • Loading
  • Loading
  • Please register or sign in to reply
    Loading