Skip to content

TrigFastTrackFinder: optimize displaced vertex hit storage

Frank Winklmeier requested to merge fwinkl/athena:ftf_hits into 23.0

CPU optimizations for displaced vertex hit storage:

  • pre-populate DataVector instead of using push_back
  • use Accessor instead of setDetail (2% speedup by itself)

Results in a total speedup of about 13% in TrigFastTrackFinder_jet where this feature is used by default.

Relates to ATR-27846. cc @smh @sutt

Edited by Frank Winklmeier

Merge request reports