search

SegmentLinking / cmssw

CMS Offline Software
http://cms-sw.github.io/
Apache License 2.0
1 stars 1 forks source link

Usage of split TrackCandidate collections for LST objects and fixes/improvements for TrajectorySeed collections #25

Closed VourMa closed 1 month ago

VourMa commented 1 month ago

This PR transfers the machinery for the improvements introduced in Phase 2 HLT for LST and partially transfers these improvements. More specifically, the pT LST objects and the T5 LST objects are now used to create separate TrackCandidate collections, so that a different tracking ID can be applied to each one of them: The default tracking ID for highPtTripletStep for Phase 2 is applied to the pT LST objects and not tracking ID is applied to the T5 LST objects. What can also be tested and applied in a future PR is the utilization of the LST seeds as inputs to a following CKF/mkFit "iteration".

Apart from the above, there are a few fixes/improvements for the TrajectorySeeds collection of LST objects (see PR #24 and the discussion that took place there about the proper length for the hitsForSeed vector and the addition of a toggle to construct non-pLS TrajectorySeeds).

The plots for the two-iteration setup with CKF vs. LST vs. LST after this PR can be found here: https://uaf-10.t2.ucsd.edu/~evourlio/SDL/CMSSWPR25/CKF_LST_LSTAfterAllUpdates/

Below you can find some screenshots for reference: image image image

I will leave some comments below for explanations and to ask for feedback.

VourMa commented 1 month ago

@slava77 I pushed the edits here, just to keep track but I created #27 to 14_1_0_pre3 and that should be the one to merge.

VourMa commented 1 month ago

@slava77 I pushed the edits here, just to keep track but I created #27 to 14_1_0_pre3 and that should be the one to merge.

Well, scratch that, this PR was messed up :)

slava77 commented 1 month ago

@slava77 I pushed the edits here, just to keep track but I created #27 to 14_1_0_pre3 and that should be the one to merge.

Well, scratch that, this PR was messed up :)

it looked like there were no conflicts wrt pre3 variant in #27; you should've kept editing in the pre0 area. Or did I look there after the rebase to pre3?

VourMa commented 1 month ago

There were and I had to rebase, hence the new PR, but it didn't work as cleanly as I thought it would.