Minutes for LArSoft coordination meeting on April 9, 2019
Present | Vito di Benedetto, Giusseppe Ceeriati, Lynn Garren, Krzytzstof Genser, Herb Greenlee, Robert Hatcher, Kyle Knoepfel, Marc Paterno, Saba Sehrish, Erica Snider, Hans Wenzel, Alex Himmel |
Remote | Tracy Usher, Gianluca Petrillo, Tom Junk |
Release and Project status report [Erica Snider]
- art 3.02 with clhep 2.4, changed default random engine to use MINMAX random number generator
- one special case was found, if G4engine is user selected type, then Geant4 used the default engine and default clhep is provided, which breaks the code
- This is being fixed
- A possible issue: if someone uses G4 before create engine, then will get different results. Risk is low given small phase space
- Expert on how to set up G4 engines (Soon Jun Yung)
Conclusion:
- Do experiments need a test release for this bug fix?
- All experiments are on board with not having such release
- There will be a regular release this week, then one with new art if ready
Exploiting detector symmetries for photon visibility [Gianluca Petrillo]
- Photon Library is currently implemented as a table, which take a lot of memory
- which photon vectors can receive light, and how much light, reflected light, arrival time,
- Workaround to reduce memory impact of the optical library: introduce a new mapping layer that with prroper transformation can only cover part of detector volume
- transform detector space into the (voxelised) one of the library to access it
- transform optical detector numbering to decode the library content
- PhotonvisibilityService provides both mappings - GetVisibility()
- GetAllVisibilities()
- example for ICARUS prrovided
- icaruscode/Light/LibraryMappingTools: starting point for how to implement mapping for different case
- [Question:], does your code handle the opaque barrier beetween the two symetric volumes? You can do that but it is not implemented as part of the symetry. What if there is not a barrier? The recommendation was to keep the space charge separate.
- Alex Himmel commented that we are already moving away from the Photon library
Conclusion:
- test passed: compared with LArSoft v08_14_00, photons from LArG4 are exactly the same (5 ICARUS events, 10 SBND events)
- code is in branches feature/gp_PhotonVisTransformations
- also provided for dunetpc (update) and icaruscode (custom mapping)
- nothing needed for ArgoNeuT, LArIAT, SBND and MicroBooNE
- requesting to merge it into LArSoft - approved
A couple of tweaks in LArSoft to improve CPU efficiency [Kyle Knoepfel]
- None of the message logging facility throws an exception,
- The OpDetBacktrackerRecord was recording an error that consumed 16% of CPU time
- 16% of the job spent in logging, so if it is an error throw an exception, if not then change to DEBUG
- there was some discussion on if this is a real problem
- TODO: follow up with Jason S, Alex Himmel will follow up
- Now non-breaking changes: use of std::pow Vs. std::sqrt, prefer sqrt,
- Prefer compile-time expressions over run-time, already pushed
- Dont use art::Ptr if you dont need to perrsist a reference, already pushed
Conclusion:
- Follow up on the first breaking change presented on the message logging facility
Status of MicroBooNE MCC9 Larsoft Updates [Heerbert Greenlee]
- These are chabnges that have gone into MicroBooNE MCC9 branch, may or may not have gone into LArSoft develop
- Pandora
- asmith_ubooneMCC9_release_v08_05_00_br merged into v08_05_00_04
- Significant update to larpandora.
- Conflicts w develop, so cannot be merged directly
- Pandora team should provide update to community
- Particle Identification updates (Kirsty Duffy)
- breaking change, showed slides from Kirsty's slides (Jan 15)
- waiting on root 16.00 upgrade, problem with IOrule due to root bug, so could not be made backwards compatible
- A breaking change, so a lot of experiment code needed changes
- LArSoft branches to merge cleanly into develop
- Other changes at Jan 15 meeting were implemented
- Kirsty will develop and test the IOrule by hte next LCM
- Trajcluster (Bruce Baller)
- feature/bb_trajcluster_mcc9
- Merges cleanly (except producdt_deps)
- Bruce Baller (the uathor) is working on a new branch that includes updates from this branch. Will be compatible with LArSoft develop
- MCReco
- Gets truth info from SimEnergyDeposit rather than SimChannel
- Also a bug fix for case when energy deposit does not drift to a physical wire
- Change can be cherry-picked to merge cleanly. David Caratelli has done this
- Minor updates and bug fixes
- SpaceOpointSolver (from Tingjun)
- Merges cleanly
- fcl change for MicroBooNE-specific file in larreco
- VertexFitter bug fix (from Giuseppe)
- Merges cleanly, non-breaking
- Genie v3.
- MicroBooNE has a v3 beta release
- LArSoft will not integrate this. Will wait for tagged release
Conclusion:
- Merge requests for this week's integration release.
- larsim: larsim: feature/davidc_mcreco_segfaultfix
- larreco: feature/tjyang_sps
- larreco: feature/cerati_fix_vtx_fit_bug_v08_05_00_br
- approved
There are minutes attached to this event.
Show them.