Draft: Update SMOG2 KS trackeff lines in Turcal

This MR aims to add some updates and improvements to the Ks->pipi tracking efficiency lines using the VELO-Long method with Ks candidates coming from SMOG2. They mostly follow what is already in the equivalent lines for pp (https://gitlab.cern.ch/lhcb/Moore/-/blob/master/Hlt/Hlt2Conf/python/Hlt2Conf/lines/trackeff/KSVeloLong.py?ref_type=heads).

To do:

• Add tis filter for HLT1 unbiasing
• Add selective persistency of matched track
• Remove RAW banks (not needed with selective persistency)
• Readjust selection to reduce background and improve kinematic converage, update prescales
• Unifiy code with KSVeloLong (with pp lines, following Sasha and Laurent suggestion)
• Remove Long track containers (keep only MatchedLongPions and MatchedUpstreamPions)
• Test new Bandwidht with SMOG2 data, compare with previous one (expect a reduction thanks to moving to Turbo strategy)
• Check if I need to change the line names, or if further changes are needed in hlt2/sprucing sequence (do lines need Turbo in the name now?)

Local Bandwidth tests results (on a representative pp+pAr sample from Block 1, the same as in !4500 (merged) ). Note that this data is from block 1 (mu pp 4.4), so the SMOG2 bandwidth is representative while the bandwidth of pp lines should be below that one in mu=5.3 conditions.

Before any changes (baseline, as in master)

| | Line | Total Retention (%) | Rate (kHz) | Exclusive Retention(%) | Exclusive Rate (kHz) | Avg Total Event Size (kB) | Total Bandwidth (GB/s) | Avg DstData Size (kB) | DstData Bandwidth (GB/s) |

| 6 | SpruceTurCalTrackEff_SMOG2KshortVeloLong_HighPTDecision | 1.19 | 0.108 | 1.19 | 0.108 | 44.1 | 0.00476 | 1.84 | 0.000199 |

| 9 | SpruceTurCalTrackEff_SMOG2KshortVeloLong_LowPTDecision | 0.927 | 0.084 | 0.927 | 0.084 | 40.4 | 0.00339 | 2.42 | 0.000203

| 11 | SpruceTurCalTrackEff_SMOG2KshortVeloLong_VeryHighPTDecision | 0.662 | 0.06 | 0.662 | 0.06 | 51.9 | 0.00312 | 1.8 | 0.000108 |

 

| | Line | Total Retention (%) | Rate (kHz) | Exclusive Retention(%) | Exclusive Rate (kHz) | Avg Total Event Size (kB) | Total Bandwidth (GB/s) | Avg DstData Size (kB) | DstData Bandwidth (GB/s) |
| 0 | Hlt2TurboVelo2Long_KshortHardDecision | 0.03 | 0.36 | 0.03 | 0.36 | 52.4 | 0.0189 | 51.9 | 0.0187 |
| 1 | Hlt2TurboVelo2Long_KshortVHardDecision | 0.016 | 0.192 | 0.016 | 0.192 | 64.8 | 0.0124 | 64.3 | 0.0123 |
| 2 | Hlt2TurboVelo2Long_KshortDecision | 0.014 | 0.168 | 0.014 | 0.168 | 66 | 0.0111 | 65.6 | 0.011 |
| 3 | Hlt2TurboVelo2Long_KshortSoftDecision | 0.012 | 0.144 | 0.012 | 0.144 | 70.8 | 0.0102 | 70.3 | 0.0101 |
| 4 | Hlt2TurboVelo2Long_KshortVSoftDecision | 0.007 | 0.084 | 0.007 | 0.084 | 61.2 | 0.00514 | 60.8 | 0.00511 |

(note that here Long and Upstream containers were persisted, as changes from

As with the proposed changes

TRACKEFF

Other| | Line | Total Retention (%) | Rate (kHz) | Exclusive Retention(%) | Exclusive Rate (kHz) | Avg Total Event Size (kB) | Total Bandwidth (GB/s) | Avg DstData Size (kB) | DstData Bandwidth (GB/s) |
| 1 | Hlt2TurboVelo2Long_SMOG2KshortSoftDecision | 0.885 | 0.0475 | 0.885 | 0.0475 | 2.58 | 0.000123 | 1.99 | 9.44e-05 |
| 2 | Hlt2TurboVelo2Long_SMOG2KshortHardDecision | 0.295 | 0.0158 | 0.295 | 0.0158 | 2.24 | 3.55e-05 | 1.84 | 2.91e-05 |
| 3 | Hlt2TurboVelo2Long_SMOG2KshortDecision | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |

TrackEff| | Line | Total Retention (%) | Rate (kHz) | Exclusive Retention(%) | Exclusive Rate (kHz) | Avg Total Event Size (kB) | Total Bandwidth (GB/s) | Avg DstData Size (kB) | DstData Bandwidth (GB/s) |
| 0 | Hlt2TurboVelo2Long_KshortHardDecision | 13.3 | 0.713 | 13 | 0.697 | 2.28 | 0.00163 | 1.78 | 0.00127 |
| 1 | Hlt2TurboVelo2Long_KshortSoftDecision | 10.6 | 0.57 | 10.3 | 0.554 | 2.42 | 0.00138 | 1.92 | 0.00109 |
| 2 | Hlt2TurboVelo2Long_KshortVHardDecision | 7.96 | 0.428 | 7.67 | 0.412 | 2.23 | 0.000952 | 1.71 | 0.000733 |
| 4 | Hlt2TurboVelo2Long_KshortDecision | 7.08 | 0.38 | 6.78 | 0.364 | 2.34 | 0.000889 | 1.88 | 0.000715 |
| 5 | Hlt2TurboVelo2Long_KshortVSoftDecision | 4.72 | 0.253 | 4.72 | 0.253 | 2.61 | 0.000661 | 2.07 | 0.000526 |

 

changed milestone to %RTA 2025 Data taking

added RTA WP4 hlt2 labels

assigned to @oboenteg

changed the description

added RTA label

Is there a need to have these lines in a different module? Then we could avoid copy paste.

Hi Sasha, for these lines we need to modify the momentum cuts (as KS from smog2 have lower pT in general), add the SMOG2 HLT1 lines in the list of lines for the tis filter, and add a baseline requirement that the KS comes from a SMOG2 BestPV. Since these imply many changes in hard-coded values with respect to the KSVeloLong lines for pp, it seems a bit simpler to me to keep them in a different module like they are now, let me know what you think.

I think it makes sense to add more flexibility in the standard module, instead of copy/pasting the code and having to maintain the same code twice.

I think we should try to reuse the main functions. So that one can profit from bugfixes. We can certainly make things more flexible if needed for your use case.

This will be certainly more ideal, yes. I hope we are not much out of time, I can try this week to change this to use the functions in https://gitlab.cern.ch/lhcb/Moore/-/blob/master/Hlt/Hlt2Conf/python/Hlt2Conf/lines/trackeff/KSVeloLong.py?ref_type=heads, hopefully I don't need to make many changes in there. At minimum, I need to modify this line so for SMOG2 lines it uses the SMOG2 filter, add options for these vertex cuts (specially this one) and add the smog2 lines to this list, and be able to lower the momentum cuts on the tag kaon and add this line as an option (requiring KS from pv in SMOG2). Do you think these changes are feasible to be made in the KSVeloLong file?

That's fine. Might not be super nice but we could add some if statements to make the modifications for the SMOG use case.

Hi @sstahl and @ldufour, as discussed, I have tried to combine the functions of these lines to be used both for pp and SMOG2 lines. Please, let me know what you think, if you have comments on the structure, etc. I have added also the changes from !4407 (merged) as I guess those will get merged first.

I have also added some bandwidth test results from local tests in the MR description, the numbers of the pp lines before and after the changes seem reasonable to me, but please check. I get very small rates for the SMOG2 lines now, so I might need to double check the selection & prescales.

I think it's mostly fine, instead of if statements we could make a few things more configurable and pass down the values different for SMOG and PP but this would require changes in the pp lines then.

Very good! I apply some small modifications to the cuts, I am retesting the rates but then I think it is more or less ready.

!4407 (merged)

-> Have a look at this MR. One can save a lot of bandwidth.

Thank you for pointing that out! I will apply this, it should help to release the prescales as well here.

changed this line in version 2 of the diff

marked the checklist item Add tis filter for HLT1 unbiasing as completed

marked the checklist item Add selective persistency of matched track as completed

marked the checklist item Remove RAW banks (not needed with selective persistency) as completed

changed the description

marked the checklist item Readjust selection to reduce background and improve kinematic converage, update prescales as completed

added 18 commits

• d40d0d8b...7cfe4bb4 - 16 commits from branch master
• 9a0b1d01 - Add Tos filter, and selective persistency
• 7b2a6152 - Change line names to Turbo convention and fix typos

Compare with previous version

added 4 commits

• 36278b2c - Add Tos filter, and selective persistency
• b8236bbe - Change line names to Turbo convention and fix typos
• c6915e8f - Unify SMOG2 and pp Ks VeloLong builders
• d7778b22 - Adjust P requirements for SMOG2

Compare with previous version

added 1 commit

• bcf4943b - fix bug in fromSMOG2 argument definition

Compare with previous version

marked the checklist item Unifiy code with KSVeloLong (with pp lines, following Sasha and Laurent suggestion) as completed

marked the checklist item Remove Long track containers (keep only MatchedLongPions and MatchedUpstreamPions) as completed