Roundtable
Alex
No news on the ACTS side, mainly University work.
Doga
Busy with the dataset. Kinematic of secondary vertices is not well understood yet. Some corrections have to be made, especially for phi. Need to find a way to convert deltaR between secondary vertex and beginning.
Preparing for ACAT-poster.
Struggling with non-working ACTS-Athena-CI.
Carlo
Track-conversion under review now for traccc.
Paul
Put a lot of work in the plot for the pipeline report.
Ambiguity resolution is "gone" because it got inline to track finding. It is still there but not a separate step.
Wrote pipeline report and happy with the results.
Improvements to copying. Currently we copy trackstates 1by1 which is a bit slow.
We can't guarantee what the backend looks like, that's why not choosing SOA directly.
Attila: They could even look like SOA in memory but aren't which leads to crashes.
Started to implement range based copies. Still the user needs to promise that the track states are not out of order. Started in the meantime by dropping the eigen vor direct copies. Also tried for the jagged vector for the measurement.
Anton
Make work flower easier and try different models. Plan to speak with supervisor in Helsinki.
Dimitra
Gen3 implentetation in Athena. MR got merged.Already set up and extrapolation test. It goes through, but cannot find the multilayers. Currently we don;t divide chambers into sub chambers. That's why we need to check a lot of them. With the try-all it finds the chambers but not the multilayer inside. The wholes volume cannot be found.
Paul: start with try all for everything. Then check the assumption on the initial search parameters. Then male conditional on what we are looking for (portals/cells).
Plan to check this out during this week.
Started migrating the pattern recognition algorithm for the small muon wheel from Athena to ACTS.
Attila
Helped with the report.
Revived older developments. Looked into traccc data model to make it competitive with ACTS. Currently, not fancy but first step to describe fitted tracks with the same infrastructure as ACTS.
Planning to discuss with PF on some Athena developments. PF and Spyros looked into exchanging some parts.
Benjamin
Worked on the report. Now more time for other stuff again.
Continue GSF development, like using the sympy stepper and doing covariance inflation.
Joana
Continued looking at the KF. Blew up mass tolerances. Trying to understand if we land on a surfaces because of the tolerance to properly tag it as a hole.
Looked a t task parallel propagation. Started with nvidia example. It can already do some things with warps. Added navigation there but already deadlocks and similar are occuring. Not so many tracks in memory make it really fast already.
Andreas Stefl
Working on speeding up the EF tracking pipeline for C100 and in general the pipeline. Surface grid in barrel was larger than necessary.
Seeding: still not sure why ACTS is 25% slower than Athena even though it should be a direct copy.
Attila: Are we still using the same algorithm? -> Very similar one Athena as a few to choose from. ACTS is a tiny bit more general taking in all the cuts.
Problem is that the matching cannot be understoof before the filtering stage. Since we do inner loops it is quite tight.
Time measurement: Planned with real profiler. Athena does not work so well because no real functions. Now just using timing barriers inside the code.
Attila: maybe vtune with seconds could be a bit better?
Investigate filtering in Athena a bit more and try to generalise it further. Maybe better to have an experiment specific filter?
