EP R&D Software Working Group Meeting

Name: EP R&D Software Working Group Meeting
Start: 2021-10-20T10:00:00+02:00
End: 2021-10-20T11:00:00+02:00
Location: The Zoomiverse

Wednesday 20 Oct 2021, 10:00 → 11:00 Europe/Zurich

The Zoomiverse

62893115044

Graeme A Stewart

Andre Sailer

Hide

Software R&D Working Meeting Minutes

Andi - can run on GPU, so is the EDM GPU friendly? e.g., in CMS

Yes, the data is GPU friendly
In Key4hep relevant data is just extracted from EDM4hep
- In ATLAS the data model woudl have to he handled in the same way (not device friendly)
Felice - create an SOA on the GPU, use an index to link back to the original EDM; for Key4hep might also have a position or energy calibration producer on the GPU
Andi - Vecmem library is interesting for this, gives a host and device view of a container
- Presentaion in CAF meeting in December (also linking to a GPU friendly geometry layer)
- Could this be connected to PODIO?
  - Yes, it encapsulates just this part of data representation (in ACTS, but really it’s “standalone”)
- https://github.com/acts-project/vecmem

Pere - single repository or multiple versions of the code?

Multiple at the moment - there is a fork that needs to be maintained (albeit that at the moment it’s stable code)
Single repo would offer a lot of advantages
- It’s under discussion
- Graeme - given the range of applicability this really merits effort

Physics studies started with single particle, then went to multi-particle, ttbar, then adding pile-up

Marco: HLT TDR for Phase2 of CMS has the most recent results, physics oriented using TICL and CLUE, but inevitably that will fold in other software on top of TICL/CLUE

There are minutes attached to this event. Show them.