Closed mlarson02 closed 4 weeks ago
tomalin
commented
1 month ago @aryd as Mark has now stopped working on this (but is still contactable), could you please post here an explanation of what prevents this being merged? This will help whoever takes over this work.
aryd
commented
1 month ago Ian,
I think this can be merged; I was just saturated with a funding agency review and I’m on holidays this and next week. I did discuss the status of this PR with Mark and I think this should be good to merge.
-Anders
Anders Ryd @.**@.>
On Jul 30, 2024, at 6:27 PM, Ian Tomalin @.**@.>> wrote:
@arydhttps://github.com/aryd as Mark has now stopped working on this (but is still contactable), could you please post here an explanation of what prevents this being merged? This will help whoever takes over this work.
— Reply to this email directly, view it on GitHubhttps://github.com/cms-L1TK/cmssw/pull/276#issuecomment-2258745300, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABI4LTLBOLNABLO2HT3HZPDZO65IDAVCNFSM6AAAAABHU7Y5M2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENJYG42DKMZQGA. You are receiving this because you were mentioned.Message ID: @.***>
Fixes Issue #212 of reducing truncation in the MatchProcessor by adding the ability to duplicate MatchProcessors for configurable (in Settings.h) layers/disks for phi regions B, C where truncation is the worst.
Improvement in efficiency ranges relative to base combined modules ranges from 0.4% to 0.6% depending on which layers/disks are included. With the maximum improvement studied (duplicating for L3, L4, L5, D2, and D3) of 0.60%, the efficiency is still 0.17% below that of the non-combined modules. The average resolution, track quality, etc. remain largely unchanged.
For layers 4 and 5 the projections routed to each duplicated MP are balanced differently (rather than sequentially taking half of the projections to each MP) to account for the higher number of projections from the L1L2 seed.