Present: James Amudson, Lynn Garren, Libo Jiang, Chris Jones, Marc Paterno, Gianluca Petrillo, Ruth Pordes, Saba Sehrish, Erica Snider, Tingjun Yang
Remote: David Adams?, Matt Bass, Vito Di Benedetto, Thomas Junk
LArSoft status report [Erica Snider]
-------------------------------------
- we will cut a release including only DUNE's feature branch including a bug fix from Robert Sulej
- this release will be marked production for DUNE's request
- ROOT 6 does not support mathematical expressions in GDML files: GDML files will need to be updated, possibly the creating scripts fixed
- proposing a revision of the repository organization with the addition of one level of directory
- an e-mail will be sent to collect comments and feedback
Conclusions:
- an e-mail will be sent to collect feedback on repository refactoring
- DUNE will communicate to LArSoft release manager
LArSoft/LArLite interoperability project report [Chris Jones]
--------------------------------------------------------------
- support the development model: develop algorithms in LArLite, move them to LArSoft
- allow reading LArSoft files directly in LArLite
- plan: split LArSoft/nutools/art products in lite part and the rest
- [Gianluca Petrillo] LArLite system will need to deal with UPS products; LArLite is not required to use UPS itself
- [Lynn Garren] there will be two copies of code originating from LArLite: one owned by the user(s), one by LArSoft
- LArSoft will pull users' code into the LArSoft-controlled forked repositories
- [Erica Snider] although the current focus is on sharing data products, anything following the "right" protocol (e.g. algorithms) can be shared
- [Marc Paterno] we expect the amount of sharing not to be total: there will be part of LArSoft that won't be available in LArLite
- [Marc Paterno] if the original user's GitHub repository disappears, LArSoft fork will be still available
- three use cases; developer uses LArSoft data products in LArLite, developer uses LArLite to create/maintain LArSoft algorithm and scientists use LArSoft art/ROOT file outside of art.
- on going work: demonstrators of splitting ups products and support art::ptr
- changes (to LArLite data classes) have been proposed to "LArLite people"
-full status of project available at redmine project site
Conclusion:
- more discussion will follow (MicroBooNE, for example, has no representative in the room)
Interfacing CORSIKA with LArSoft [Matt Bass]
---------------------------------------------
- GHEISHA is embedded, FLUKA is optional
- relies on database files that are large
- files currently in a MicroBooNE area; question is where to keep them to be shared more widely
- [Erica Snider] we might need a PNFS area for LArSoft
- [Lynn Garren] there is no request to have CORSIKA as a UPS file; but in the future people might want to generate cosmic rays anew
- added new parameters and modified LArG4 to avoid processing particles outside the detector
- [Lynn Garren] please put it in a feature branch rather than in develop
- [Erica Snider] if we receive a document, we will put it in a proper ares
- [Gianluca Petrillo] is there a measurement or estimation of the effect on time execution for non-CORSIKA jobs? A: not expected to affect, but will do checks
- [Gianluca Petrillo] have an analyser in LArSoft that produces some of the plots we have seen in the presentation. A: plots were generated by a ROOT macro running on AnalysisTree-produced files
Conclusions:
- need to identify a place where to keep large shared data like the CORSIKA pregenerated database
- consider distributing CORSIKA (and FLUKA) as UPS products
- publish modified LArSoft code in a feature branch
- waiting for a document to put in LArSoft-shared area
[Vito Di Benedetto]
-------------------------------------
- [Vito Di Benedetto] one MicroBooNE test still fails, results are not reproducible (possibly pseudo-randomness is involved)
- [Erica Snider] are MicroBooNE tests going? A: the test above is still failing
- [Erica Snider] tried ninja? A: tried, differences seems to be at the level of fluctuations
- [Ruth Pordes] the system should be advertised and we should be very heedful for failures
- [Marc Paterno] the time difference between debug and prof using make test and ci_test should be advertised
Conclusions:
- temporary disable the failing test and deploy
- track down the issues with the failing test
Report on Projection Matching Algorithm [Robert Sulej]
-------------------------------------------------------
Robert Sulej showed slides about improvement and fixes that need to go into LArSoft before DUNE MCC5 starts.
There are minutes attached to this event.
Show them.