Roundtable
Luthien
Doga
CSC
Benjamin
Andreas Salzburger
Noemi
Seeding evaluation for EF-Tracking development. Still a few issues left from Carlo and presented last Thursday. PF presented in open ditac last Tuesday on the clustering and the rest of the chain. Still trying to understand how difference between the different seedings (legacy and current) diverges. Duplet production seems to be in line though. But there is a difference of 25%. They roughly create the same number of seeds.
The phi-bins in the grid have different extend, so we are not putting the same number of spacepoints in every bin. But we don't see how this could make a 25% change. So it needs to come from somewhere else.
Now comparing duplets only in both cases because Athena is not modular enough.
Testing some things from Carlo as well. When moving between strip layers, compatibility between cluster and prediction. Tried some time ago to map the material in the back of the material. Helped to get some compatibility and smoothed variance between layers. At that step, we have already all measurements from the pixels. Athena legacy version of the code managed to find these events.
Dimitra
CSC
FInalising tracking geometry in Athena. No large changes. The material accumulation had to be changed because it wasn't really nice. Added a builder also for the interface. Also added for the ITk strips.
Paul: It doesn't need to be more general than necessary. But how should we structure the construction, that it wraps correctly? e.g. slotting the calorimeter in between, when we need it. Marc is working on it a bit.
Carlo
Spacepoint tuning: removing some in the outer layer and a few in the inner layers.
We are now 25% faster but lost some efficiency. Still the same as legacy but we were better before.
Noemi: maybe contact Corantine on clustering seeds and evaluating the quality.
space points in some region are cut out in the detector because of the way they are calculated the edges are missing.
Attila: Big picture for GPU: we are missing the person to do it. Just applying cuts is easy but for the rest we need an actual GPU expert.
Attila
(in discussion above)
Hackathon last week with G100 went quite well. But this is also mainly because so many seeds are roduced and therefore bound by ACTS. C200 is a bit behind reference but already in competition.
Alex
Catching up with stuff after the vacation.
Andreas Stefl
Working mostly on seeding code and getting EDM discussion done. Looking with PF into profiling and hopefully nail down the 25% we are losing.
Gbts is probably outdated in the code. Tim said that someone works on 2.0 but we need to have a look how to proceed with that one.
