Tune Si tracking pattern recognition for mu=60
This MR implements a preliminary tuning of the pattern recognition in the Si tracking as documented in https://its.cern.ch/jira/browse/ATLIDTRKCP-260. At this point, the intention is to make this available in master for testing by a larger group of experts and clients.
Two new cut levels are defined:
- Cut level 18 represents the baseline tune, which we are considering for use in run-3 and the reprocessing
- Cut level 17 represents a backup tune, which relies less heavily on overlapping seeds in the tracker at the cost of ~10% of track finding speed, and may be interesting to compare for studies taking into account end-of-run-3 radiation damage.
The cut level is expected to result in a major speed up of the track finding stage at mu=60 by at least a factor 2, and also the ambiguity resolution by 10-20%. Total tracking CPU should decrease in the ballpark of 25-30% compared to master.
Regarding tracking performance, this MR is expected to affect reconstruction output in a visible but slight way. As documented on the Jira linked above, we expect a decrease in the total number of reconstructed tracks. Most of the reduction is from fakes, which are expected to decrease by between 30-40%. The efficiency should decrease very slightly for very soft (pT < 1 GeV), forward (|eta| > 2) pions from ttbar at the level of up to 1-3%. For electrons and muons as well as for high-pt activity, the efficiency is expected to remain stable with only marginal degradation.
For this reason, we also update the xAOD digest reference files.
In addition to these tweaks, this MR introduces one additional histogram to InDetPhysValMonitoring, which is used to validate the efficiency impact of tracking changes at very low pT.