Minutes for LArSoft coordination meeting on Dec 04, 2018

Present | Vito Di Benedetto, Giuseppe Cerati, Lynn Garren, Kyle Knoepfel, Saba Sehrish, Erica Snider, Hans Wenzel, Robert Hatcher, Joseph Zennamo |

Remote | Herb Greenlee, Katherine Lato, Gray Putnam, Paul Russo, Tracy Usher |

Project status report and art v3 discussion [Erica Snider]

  • Notes on getting genie v3 tagged for the next LArSoft release (see slides)
  • Migration to art v3, it has support for multi-threading
  • Important to note that the default mode is serial, you have to enable multi-threading, algorithms do not change
  • jobs that relied on accidental behavior may fail, e.g. modules that need to be executed one after another need to be on the same trigger path
  • We would like to proceed with the migration ASAP by adapting to framework changes, without multi-threading. Moving to full support for multi-threading in LArSoft will be a long and gradual process
    • already started with services
  • The sooner migration will help avoid creating multiple branches that are far behind the head
  • The migration process will be the same as first getting a candidate release; currently looking at v08 RC
  • Vote from the experiments: ICARUS: yes, DUNE pixel detector: yes, uboone: yes
  • No rep from DUNE or other experiments. Will contact them via email after the meeting
  • backward compatibility between input/output files will not be an issue.
    • For example signal reco with art v2 and pattern recognition with v3

Conclusion:

  • move ahead with art v3
  • We do not know when genie v3 will be tagged, uboone really want it to be tagged for MCC9, and will directly talk with the genie collaboration

Changes to MCRecoPart [Gray Putnam]

  • This work addresses the following larsoft issues: #21394, #13718
  • The code to define the fiducial volume is implicitly tied to MicroBooNE and not quite correct for that detector either
    • This work is about how fiducial volume is defined and now it includes all the TPCs.
    • alternately we could consider not deleting the points outside the volume
  • [Erica] Does it consider multiple cryostats? (last slide)
  • Design discussion later: have an algoritthm code to define the volume instead of making the data product smarter
  • Following to be checked by [Gray Putnam] per [Joseph Zennamo]
    • Have you checked whether fiducial volume is within the active volume?
    • is it used by MCShower?

Conclusion:

  • the proposed work is already an improvement on what is currently there
    • feature branch to be merged in this weeks release larsim: feature/gputnam_MCSTReco_FV_fix