While performing timing measurements for the Phase 2 HLT menu, it was observed that some L1P2GT paths had abnormally high rates. For example, pDoublePuppiTau52_52 in our measurements:
has ~3 times the rate of pSingleTkEle36, while this is not the case in the L1T tables from the 2024 Annual Review.
This issue is meant to keep track of the L1T updates to the P2GT thresholds in the release.
Additionally, giving the bias that such rate differences cause in the Phase 2 HLT timing measurements and the fact that this issue was discovered in the HLT tests by chance (smaller but still wrong rate differences might have gone unnoticed), it would be extremely useful if some kind of mechanism was set up to check the compatibility of the P2GT thresholds with the ones in the standalone L1T MenuTools to catch such issues beforehand at the L1T level.
@Dr15Jones, @antoniovilela, @makortel, @mandrenguyen, @rappoccio, @sextonkennedy, @smuzaffar can you please review it and eventually sign/assign? Thanks.
cmsbuild
commented
2 hours ago
While performing timing measurements for the Phase 2 HLT menu, it was observed that some L1P2GT paths had abnormally high rates. For example,
has ~3 times the rate of
pDoublePuppiTau52_52in our measurements:pSingleTkEle36, while this is not the case in the L1T tables from the 2024 Annual Review.After discussing with the L1T experts, it was realized that the high rates are a result of not updating the P2GT emulator online thresholds to the ones presented in the 2024 L1T Annual Review. The following table shows the differences between the L1P2GT rate and the standalone L1T rate: https://docs.google.com/spreadsheets/d/1VV0MWQaTDzPIPC5vyzqO8lDDMtHoVSRfC7by8Ff_a88/edit?gid=0#gid=0
This issue is meant to keep track of the L1T updates to the P2GT thresholds in the release. Additionally, giving the bias that such rate differences cause in the Phase 2 HLT timing measurements and the fact that this issue was discovered in the HLT tests by chance (smaller but still wrong rate differences might have gone unnoticed), it would be extremely useful if some kind of mechanism was set up to check the compatibility of the P2GT thresholds with the ones in the standalone L1T MenuTools to catch such issues beforehand at the L1T level.
FYI: @rovere @artlbv