Choose timezone
Your profile timezone:
EP R&D Software Working Group Meeting
→
Europe/Zurich
32/1-A24 (CERN)
EP R&D Software Working Group Meeting
Zoom Meeting ID
62893115044
Host
Graeme A Stewart
Alternative host
Andre Sailer
Useful links
Join via phone
Zoom URL
Software R&D Working Meeting Minutes
2023-05-17
Apologies: Andi. Marco
Room: Graeme, André, Leonhard, Swathi, Juan, Vincenzo, Piyush
Remote: Olivier, Juraj, Erica, Alberto, Pere, Valentin, Brieuc, Dalila, Alvaro, Dmitri, Benedikt
Introduction (Andre)
Vincenzo has moved from doctoral student to fellow, continuing to work on RNTuple and distributed analysis.
2022 R&D report is posted to CDS: https://cds.cern.ch/record/2852748.
The MTP is being decided on today, so we expect news very soon on EP R&D II.
Key4hep (Juan)
New visualisation generated with dot - readability could be improved.
Why not code generate the Python the move code from PODIO, rather than use cppyy? It would be great to expose the Python to IDEs properly? Discussion - this should be possible, via PODIO or maybe using PyBind11.
Note that after the move the coll is in an undefined state, which is not well modeled in Pythonic ownership. ROOT see the same issues, tend to use const refs. Could use a weak reference on the Python side, which do have established semantics.
LD_LIBRARY_PATH is used a lot for plugin loading. This is all still quite delicate. This is very important for ACTS development.
Build caches are still broken - Spack will relocate (RPATH) all of the binaries, but there are issues with things where the binaries are in non-standard locations (git) or Python is soft linked. For the moment, don't try to switch the directories between build and deployment.
How are the histogram comparisons made? It's something that we wrote. ACTS have similar validation. And all the experiments have similar infrastructure for physics validation. Should look into how this is done and can we share?
Paul - EDM/ACTS interplay (slide 13). This is now coming together, with a PR in the works to have full PODIO/EDM4hep integration, without any information loss (LCIO parameterisation lost local coordinate information, which is needed for further ACTS use, e.g., for fitting against an electron hypothesis). PODIO is very useful for doing this, as it can use any EDM.
Valentin - what system packages are used from key4hep builds? For CentOS7, nothing as it's all too old. Alma + Ubuntu use the system compiler.
Timescales for FCC studies, looking to the next EPPSU? Recent production with 50M events with Delphes, this took 1 weekend. Last major missing feature is frames. And also with that schema evolution will be there. After that things will become more incremental for specific detector concept studies.
If a nightly fails tests, it does not get deployed. e.g., there's a current issue with Whizad in Ubuntu 22.04 (missing library).
AOB
Don't forget the EP R&D seminar. After that we reconvene WG activities after the summer break.
There are minutes attached to this event.
Show them.
