Present: Vito Di Benedetto, Lynn Garren, Herbert Greenlee, Marc Paterno, Gianluca Petrillo, Erica Snider
Remote: Christopher Backhouse, Robert Hatcher, Katherine Lato, Ben Morgan, Gleb Sinev
Project status report [Erica Snider]
- new GENIE version (2.12.8) with optional nucleon correlation in cascade model
- physics update requested by DUNE, planning for inclusion in this week release
- the second tier of continuous integration test is available now
- runs on the grid, takes a few hours
- could be run nightly
- new projects starting: profiling/optimisation (Soon Yung Jun) and vectorisation (Guilherme Lima)
- 3D reconstruction workshop has triggered enough interest
- so far, best solution is to attach it to DUNE collaboration meeting at CERN (February 2018)
- the CERN venue will enable the participation of some LHC experiment reconstruction experts
- pull request model was requested
- requires a strong commitment by all the experiments
- [Jason Stock] the current model has shown to be liable of merging failures (commits have been inadvertedly overwritten)
- many details need to be addressed
Conclusions
- will pull offline leads about updating GENIE
- the discussion on the 3D reconstruction workshop will be
- there will be follow up
UPS v6 [Marc Paterno]
- UPS uses the kernel version to determine the "flavour" of the platform
- this is not effective in containers that report the kernel version of the host, which is not representative of the execution environment in the container
- UPS 6 addresses this shortcoming
- if a script explicitly uses a flavour, this may need to be fixed
- for Linux flavours,
ups flavor -4 will be required
- for OSX flavours,
ups flavor -2 is probably already used
- more information in the documentation linked in the slides
- CET tools (
mrb, cetbuildtools, ...) are already updated
- most of the changes required for UPS 6 can be deployed already on UPS 5
Conclusions
- users with their own UPS-based build scripts will need to be updated by their maintainers
Tracking data product review report [Giuseppe Cerati, Gianluca Petrillo (presenter)]
- the scribe was talking and he did not take notes
- the question to be answered is whether the current track-producing modules will produce
recob::TrackTrajectory, or also a fully featured recob::Track, and whether these tracks would be complete fits or just trajectories dressed with uncertainties
Conclusions
- an operative plan of transition will be outlined and proposed
- infrastructure for updating modules producing tracks is ready
- experiments will be involved to answer the question of which path to undertake
- infrastructure for update of track usage ("proxy") is also ready
- the proxy is not transparent to the choice of
recob::Track vs. recob::TrackTrajectory: users must explicitly code which one they want
BackTracker and PhotonBackTracker lazy overhaul [Jason Stock]
BackTracker rebuild stage is expensive and it happens at the beginning of every event
- including where it does not have any sense (e.g. when truth information is not available yet)
PhotonBackTracker has the same issues, and it also competes with BackTracker for resources- this is a proposal to overcome these and other issues
- a new service is being introduced (
ParticleInventory), which is lazy and does work only when necessary
- some name and interface fixing is also pulled in
- also proposed a change in
OpDetBackTrackerRecord data product to accommodate optical detector channel
- breaking changes
- a simple service configuration need to be added wherever
BackTracker is configured
- mechanical changes are required for interface changes
OpDetBackTrackerRecord
- [Herbert Greenlee] are you going to phase the changes? [A]
ParticleInventory addition is independent
- [Erica Snider] is this gallery-compatible? [A] probably not, since Jason has no experience on that; but he's willing to give it a shot
Conclusions
ParticleInventory will be merged as soon as it's ready, with the approval of the present experiments- the mechanical changes need to come with feature branches for all experiments
- need to be sorted out what happens with old files which would have the old
OpDetBackTrackerRecord
There are minutes attached to this event.
Show them.