GlueX Software Meeting, January 8, 2019

From GlueXWiki
Revision as of 22:04, 9 January 2019 by Marki (Talk | contribs) (minutes)

Jump to: navigation, search

GlueX Offline Software Meeting
Tuesday, January 8, 2019
3:00 pm EST
JLab: CEBAF Center A110
BlueJeans: 968 592 007

Agenda

  1. Announcements
    1. MCwrapper changes (not a release)
    2. hdds updates to eliminate overlaps - Richard J.
  2. Review of minutes from the December 11 meeting (all)
  3. Report from the December 18 HDGeant4 Meeting (all)
  4. Merging the DIRC branches (Justin)
  5. Developer-friendly, container-based builds (Mark)
  6. Debug and optimized builds of halld_recon and halld_sim (Mark)
  7. Review of recent issues and pull requests:
    1. halld_recon
    2. halld_sim
  8. Review of recent discussion on the GlueX Software Help List (all)
  9. Action Item Review (all)

Slides

Talks can be deposited in the directory /group/halld/www/halldweb/html/talks/2019 on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2018/ .


Minutes

Present:

  • CMU: Naomi Jarvis
  • JLab: Thomas Britton, Ashley Ernst, Mark Ito (chair), Richard Jones, Justin Stevens, Simon Taylor

There is a recording of this meeting on the BlueJeans site. Use your JLab credentials to access it.

Announcements

  1. MCwrapper changes (not a release). Thomas reported on recent changes.
    • MCwrapper submission revamp.
      • The CPP experiment has been added to the submission page as an entity separate from GlueX. Other experiments can be added similarly.
      • The "confirm" button will check for similar MC runs and notify the user if they exist. This should lessen duplication of effort.
    • Self-Service error corrections.
      • If test of submission fails, form is re-presented to user for corrections. Corrections are only allowed if coming from the original user. This feature has been used already by Peter Pauli.
  2. hdds updates to eliminate overlaps. Richard has addressed the geometry overlaps as reported by Maria Patsyuk.
    • There were 98 occurrences of overlap, many of them of the "rounding" variety (insufficient precision in the geometry parameters).
    • The most serious overlap was one between the BCAL and FDC. A tilt was added to the BCAL to make it consistent with survey data. The axis of rotation turned out to be at the upstream end of the BCAL, causing an overlap with the FDC at the downstream end. Richard changed the rotation to be (effectively) around the center of the BCAL, maintaining the amount of tilt.
    • Action Items:
      1. Richard will submit a pull request to the hdds repository.
      2. He will also consult with Sean Dobbs about how to include this change into the CCDB-resident geometry files.

Review of minutes from the December 11 meeting

We went over the minutes.

  • Crashing Monitor Launches?. The cause of the crashes in monitoring launches was traced to a Start Counter plugin by Alex Austregesilo. It was not related to the reconstruction code itself.
  • Record of per file event ranges. We need to ask if progress has been made.

Report from the December 18 HDGeant4 Meeting

We went over these minutes as well.

Richard reported that has fixed the long-standing problem with missing hits along the track in the FDC. There was a off-by-one error in the layer labeling. This caused layers to be rotated incorrectly in HDGeant4.

Merging the DIRC branches

Justin led us through his email proposing a scheme for merging the DIRC branches onto the master branches. Please see the email for a nice description of the scheme, its features, and a comment on tests completed.

For HDGeant4 (hdg4), since the geometry is defined at run time using files from the CCDB, there is no ambiguity about whether the DIRC should be included or not. For hdgeant (hdg3) the decision is made at build time by pointing to an appropriate build of HDDS, one with or without the DIRC as appropriate. We discussed whether we want to even support hdg3 with the DIRC; if not we could define hdg3 as a DIRC-free program.

Richard proposed a mechanism where inclusion of the DIRC is determined by a CCDB-resident constant. All hdg3 builds going forward would include the DIRC, but during initialization, i. e., at run time, if the CCDB says there should not be a DIRC, the DIRC geometry will be replaced by air. Again, this would only be for hdg3. Richard volunteered to code the new initialization step; Justin volunteered to create the new item in the CCDB.

Justin mentioned that initially DIRC hits would be put into the REST format. This is in contrast to our usual practice of storing only reconstructed quantities. The intention is to do this only until the DIRC reconstruction code matures. For now having the hits in the REST events allows for efficient development of the reconstruction. Justin will look into how much this will cost in terms of event size.

No objections to the plan were voiced. Justin plans to submit the pull requests tomorrow (Wednesday).

Developer-friendly, container-based builds

Mark filled in some details in his announcement of Oasis-resident builds that can be used for code development. Please see his slides for the details. They are also available as PDF from GlueX Doc 3855.

Debug and optimized builds of halld_recon and halld_sim

Mark gave further explanation of his announcement of build_scripts 1.44. The most significant change is the introduction of "debug levels"; new combinations of optimization levels and inclusion/exclusion of debugging symbols. See his slides for the details. They are also available as PDF from GlueX Doc 3856)

The idea is to (a) save disk space (b) tailor builds to the task, i. e., debugging, production via additional choices of build parameters.

Review of recent issues and pull requests

We decided to forego this section of the agenda.