Skip to content

[ATLIDTRKCP-188] Move ID tracking hole search to pattern recognition step

Maximilian Emanuel Goblirsch-Kolb requested to merge goblirsc/athena:implementHolesFromPattern into master

This MR implmenents the ability to run the hole search directly following the pattern recognition stage, avoiding the costly additional hole search run later during the reconstruction.

The changes and their motivation are documented in more detail at https://indico.cern.ch/event/968904/contributions/4080296/attachments/2131590/3589975/2020-10-28-IDSW_Holes_Updated.pdf).

In summary, this MR does the following:

  • Introduce new outcome flags to the IBoundaryCheckTool.
  • Update the InDetBoundaryCheckTool's logic to work with these flags.
  • Adapt the hole search tool to understand the additional flags and treat them correctly
  • Slightly tune the SiTrajectoryElement_xk, in order to allow for larger multiple scattering effects in the forward pixel region
  • Teach the SiTrajectory_xk to perform the required hole & dead module counting at the end of track finding. Note that during the track finding itelf, we keep using the old hole definition of the pattern, which is better suited to the early phase where uncertainties are still large.
  • Add an additional hole cut based on the resulting count to filter the tracks early
  • Allow the combi track finder event data to store the results of the pattern hole search
  • Teach the SiSpSeededTrackFinder to write the holes into the track summary
  • Make the dense ambi not forget said track summary after a refit
  • Add a flag in the InDetFlags to steer the entire procedure and use it in the config

The MR was updated following feedback received at IDSW on Monday this week.

The performance impact of this MR is non-trivial:

  • We expect to significantly speed up track reconstruction, by around 15% compared to master. The main effect is on the ambiguity solver, which speeds up by a factor 2
  • Fakes are increased, as is the number of total tracks. The fake rate is still well below the one we had with cut level 16, and 70% below release 21. The increase is smaller when only looking at 'TightPrimary' tracks
  • The efficiency, in particular at low pt and high eta, is slightly improved.

Changes are under final validation before this MR can move forward.

Adding @sswatman @zschilla @npetters and @sroe + @christos. Also adding @elsing, @asalzbur, @gavrilen.

Edited by Maximilian Emanuel Goblirsch-Kolb

Merge request reports