L1 filters using P2GT for tau paths

[agrubercms]

PR description:

This PR adds an L1 filter based on the L1P2GT for the Phase-2 tau HLT paths.
As reported in HLT Upgrade, we see lower efficiency upon including the L1 filter, but it achieves a great speed-up in the HLT menu timing, and an L1 filter will be needed eventually in any case (Also, the integration of new L1 tau seeds, once available, will be straightforward after this PR).

PR validation:

[cmsbuild]

cms-bot internal usage

[cmsbuild]

+code-checks

Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-45041/40345

• This PR adds an extra 40KB to repository

[cmsbuild]

A new Pull Request was created by @agrubercms for master.

It involves the following packages:

• HLTrigger/Configuration (hlt)

@Martin-Grunewald, @mmusich, @cmsbuild can you please review it and eventually sign? Thanks.
@missirol, @silviodonato, @rovere, @Martin-Grunewald, @SohamBhattacharya this is something you requested to watch as well.
@antoniovilela, @rappoccio, @sextonkennedy you are the release manager for this.

cms-bot commands are listed here

[rovere]

Ciao @smuzaffar I'd like to test this PR using this.
I see it is merged: does that mean it's already available?

[mmusich]

enable hlt_p2_timing

[mmusich]

@cmsbuild please test

[mmusich]

does that mean it's already available?

Apparently so.

[rovere]

Noo!!! You have stolen my first run!!! 😭

[mmusich]

You have stolen my first run!!!

So sorry!

[cmsbuild]

+1

Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-d92aed/39519/summary.html
COMMIT: 7aeeda2
CMSSW: CMSSW_14_1_X_2024-05-24-1100/el8_amd64_gcc12
Additional Tests: HLT_P2_TIMING
User test area: For local testing, you can use /cvmfs/cms-ci.cern.ch/week0/cms-sw/cmssw/45041/39519/install.sh to create a dev area with all the needed externals and cmssw changes.

• HLT P2 Timing: chart

Comparison Summary

Summary:

• You potentially removed 2 lines from the logs
• Reco comparison results: 10 differences found in the comparisons
• DQMHistoTests: Total files compared: 48
• DQMHistoTests: Total histograms compared: 3338862
• DQMHistoTests: Total failures: 6
• DQMHistoTests: Total nulls: 0
• DQMHistoTests: Total successes: 3338836
• DQMHistoTests: Total skipped: 20
• DQMHistoTests: Total Missing objects: 0
• DQMHistoSizes: Histogram memory added: 0.0 KiB( 47 files compared)
• Checked 202 log files, 165 edm output root files, 48 DQM output files
• TriggerResults: found differences in 5 / 46 workflows

[rovere]

-1.5s, not bad! -22% of (useless) timing.

[rovere]

@agrubercms please remove the Draft label from this PR as soon as you have verified the efficiency. Any ETA for this?

[agrubercms]

Hi @rovere,
I should be able to produce results this week still - however, I'm afraid that they will show a significant impact on the efficiencies of the tau paths, which should probably be discussed before merging this PR.
I could aim to present results in next week's HLT Upgrade meeting for a discussion, would that work for you?

[mmusich]

hold

• pending clarifications on L1 filters using P2GT for tau paths #45041 (comment)

[cmsbuild]

Pull request has been put on hold by @mmusich
They need to issue an unhold command to remove the hold state or L1 can unhold it for all

[rovere]

Hi @rovere, I should be able to produce results this week still - however, I'm afraid that they will show a significant impact on the efficiencies of the tau paths, which should probably be discussed before merging this PR. I could aim to present results in next week's HLT Upgrade meeting for a discussion, would that work for you?

That would be a surprise: in all slides presented by L1T, the turn-on for taus is quite sharp and the plateau is at 90+% or so see here, page 19.

Let's wait for your studies. Next week at the HLT Upgrade meeting is ok.
Thanks.

[artlbv]

Hi there, just saw this (from the tsg slides).
Do I understand correctly, that you haven’t used any L1 decision to seed this path previously? It would be odd but I would like to clarify..

[artlbv]

@rovere while you are right about the higher efficiency in the plateau, note that for low pt taus the efficiency is not that high. And from these HLT paths it seems the thresholds are below the expected gen pt plateau (52 GeV as in the L1 seed name).

[agrubercms]

Hi @artlbv,
That's correct - the HLT tau paths for Phase-2 do not use any L1 seed yet (that's also why there's such a big improvement in timing by a rather small change)

[agrubercms]

Report regarding the efficiency available in HLT Upgrade.
We see lower efficiency in the tau paths after integrating the L1 filter - however we also see a great improvement of the HLT menu timing, which is why we decided to continue with this PR (while also continuing to investigate the efficiency of our tau paths).
@mmusich

[mmusich]

unhold

[mmusich]

+hlt

• see L1 filters using P2GT for tau paths #45041 (comment)

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @rappoccio, @antoniovilela, @sextonkennedy (and backports should be raised in the release meeting by the corresponding L2)

[rappoccio]

+1