Skip to content

Reintroduce Kalman Fit in VeloClusterTrackingSIMD

Arthur Marius Hennequin requested to merge ahennequ_velo_fit into master

This is still a WIP, things that needs to be added before merge:

  • tune error to get correct pulls (The resolutions looks correct but the pulls RMS are a little too high)

The slowdown is there but barely noticeable (<1%).

@cattanem @adudziak @rquaglia @ldufour @sponce @sstahl @chasse

Things to not forget:

  • Using the Kalman fit also for the state endVelo resulted to lower UT and FT efficiencies, so I reverted back to use the linear fit for this state. The old algorithms had the option of using the Kalman for this state, but it was always turned off.

Needs LHCb!2038 (merged)

Edited by Marco Cattaneo

Merge request reports