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
2023-02-15
Apologies: André,
Room: Graeme S, Armin N, Leonard, Dalila, Erica, Juan, Pere, Benedikt, Paul, Swathi, Piush, Javier, Anna
Remote: Marco Alberto, Juraj Simesko, Alberto Ribon, Felice Pantaleo, Valentin Volkl, Brieuc Francois, Jakob Blomer, Andreas Pfeiffer, Gerri Ganis
Introduction (Graeme)
The text for each task in the R&D 2022 report should be 3/4 of a page, i.e., 0.75 pages!
Tracking (Paul)
Pere: GDML->Projection, then can use Geant4 with either full or surface geometry?
- Disentangle material simplification effects from fast reconstruction effects. Eventually can use G4 based volume material simulation.
Marco: what about passive material?
- Yes, this has to be incorporated down the line..
Graeme: where does the experiment EDM come in?
- The backend should do a conversion once, so it's fast and convenient, and very quick at the end as no further conversion is needed.
Benedikt: EDM, which is agnostic to backend, what does the experiment container need to provide?
- Certain methods need to be there, and there at compile time (templates). Could use C++20 concepts later.
- No assumptions on memory layout - containers have to have indexes, but do not necessarily need to be contiguous.
Pere - could you have used EDM4hep?
- Representation of track state isn't ACTS compatible, e.g., in the way that the covarience is stored. Also, ACTS has only ever local coordinates.
- No need to convert the ATLAS model - adopted the concepts, different implementation.
- Could work with EDM4hep to understand everything that's needed. ACTS needs geometry information, but this is not in EDM4hep, which assumes the geometry is not there...
- Benedikt: work on a template based adaptor
Benedikt: what's the size of the code base?
- Not sure, it's growing, but still well organised!
Graeme: Maintaining CUDA and SYCL at the moment.
- Yes, the CUDA outperforms the SYCL and is also currently easier to program (SYCL has lots of gotchas still)
- Graeme: different questions, what's the ultimate performance? what's the sustainable code base?
AOB
Next meeting 19 April, Calorimetery Reco focus.
There are minutes attached to this event.
Show them.
