Draft: Tuning of KShhLines for 2025
Tuning the BdsToKShhLine since it has tight cuts on the momenta of the hh pair. Edit reduces sizeably BW while trying to keep eff and rates as close as possible to master. Current status of this MR coincides with the rate latest commit with asumpt 3000 modified ks builderscolumn of the following tables
Table summarising effect on BW, rates and Hlt2 eff on simulation:
| Line | rate bnoc_run3 | rate latest commit with asumpt 3000 | rate latest commit with asumpt 3000 modified ks builders |
rate latest commit with asumpt 3000 modified ks builders + DOCA < 3 mm for KS DD builders |
rate latest commit with asumpt 3000 custom ks builders |
| Hlt2BnoC_BdsToKSKpKm_DD | 0.01211103 +- 0.00217507)% | 0.01328307 +- 0.00227788 | 0.01289239 +- 0.00224413 | 0.009376282 +- 0.00191384 | 0.01054832 +- 0.00202992 |
| Hlt2BnoC_BdsToKSKpKm_LL | 0.007813568 +- 0.00174710)% | 0.006641533 +- 0.00161075 | 0.006641533 +- 0.00161075 | 0.006641533 +- 0.00161075 | 0.005469498 +- 0.00146174 |
| Hlt2BnoC_BdsToKSKpPim_DD | (0.02461274 +- 0.00310053)% | 0.02500342 +- 0.00312504 | 0.02422206 +- 0.00307583 | 0.01992460 +- 0.00278972 | 0.02383138 +- 0.00305093 |
| Hlt2BnoC_BdsToKSKpPim_LL | (0.01484578 +- 0.00240812)% | 0.01406442 +- 0.00234391 | 0.01406442 +- 0.00234391 | 0.01406442 +- 0.00234391 | 0.01289239 +- 0.00224413 |
| Hlt2BnoC_BdsToKSPipPim_DD | (0.02031528 +- 0.00281694)% | 0.01914324 +- 0.00273449 | 0.01797121 +- 0.00264947 | 0.01640849 +- 0.00253168 | 0.01640849 +- 0.00253168 |
| Hlt2BnoC_BdsToKSPipPim_LL | 0.007813568 +- 0.00174710)% | 0.01172035 +- 0.00213971 | 0.01172035 +- 0.00213971 | 0.01172035 +- 0.00213971 | 0.01054832 +- 0.00202992 |
| BW (GB/s) | 0.2066 | 0.09575 | 0.09071 | 0.0806 | 0.0856 |
|
eff K*KS simulation (11104151) |
master (%) | latest commit asumpt 3000 modified KS builder |
latest commit asumpt 3000 modified KS builder + DOCA < 3 mm for KS DD builders |
rate latest commit with asumpt 3000 custom ks builders |
| Hlt2BnoC_BdsToKSKpPim_DD | 0.9175000 +- 0.0476729 | 0.8800000 +- 0.0466973 | 0.8250000 +- 0.0452270 | 0.8500000 +- 0.0459014 |
| Hlt2BnoC_BdsToKSKpPim_LL | 0.3025000 +- 0.0274584 | 0.4025000 +- 0.0316575 | 0.4025000 +- 0.0316575 | 0.3900000 +- 0.0311640 |
| eff K+pi-KSsqDalitz simulation (13304106) | master (%) | latest commit asumpt 3000 modified KS builder |
latest commit asumpt 3000 modified KS builder + DOCA < 3 mm for KS DD builders |
rate latest commit with asumpt 3000 custom ks builders |
| Hlt2BnoC_BdsToKSKpPim_DD | 5.562738 +- 0.133691 | 5.464630 +- 0.132582 | 4.950663 +- 0.126534 | 5.270000 +- 0.157992 |
| Hlt2BnoC_BdsToKSKpPim_LL | 2.572128 +- 0.0923365 | 2.834394 +- 0.0968041 | 2.834394 +- 0.0968041 | 2.755000 +- 0.115739 |
Merge request reports
Activity
requested review from @msaur
Resolved by Davide LancieriniUnfortunately, I am not able to check on signal MC what efficiency of the existing line is on this mode without simulation. This however has been requested with high priority hopefully it will be ready soon. Pardon the naiveness but with significant overlap, wouldn't the increase in BW be small?
Edited by Davide Lancierini- Last reply by Davide Lancierini
mentioned in merge request !4313
added 1 commit
- 020020d0 - modifying KShh combiner to relax maxp, maxpt cuts
-
Including @tlatham @peilian in the thread. With this configuration we have moved closer to the Lb builders and removed the max(ph1,ph2) and max(pth1,pth2) cuts. We also relaxed the minptsum cut from 4400 MeV to 4000 (see !4526 (diffs)) and removed the ksvtz separation from the DD. With these changes:
master:
BW 0.03175 GB/s LAZY_AND: Hlt2BnoC_BdsToKSKpPim_DDDecisionWithOutput #=255965 Sum=56 Eff=|(0.02187799 +- 0.00292325)%| LAZY_AND: Hlt2BnoC_BdsToKSKpPim_DD #=255965 Sum=56 Eff=|(0.02187799 +- 0.00292325)%| LAZY_AND: Hlt2BnoC_BdsToKSKpPim_LLDecisionWithOutput #=255965 Sum=21 Eff=|(0.008204247 +- 0.00179024)%| LAZY_AND: Hlt2BnoC_BdsToKSKpPim_LL #=255965 Sum=21 Eff=|(0.008204247 +- 0.00179024)%|vs
this edit:
BW 0.03729 GB/s LAZY_AND: Hlt2BnoC_BdsToKSKpPim_DDDecisionWithOutput #=255965 Sum=57 Eff=|(0.02226867 +- 0.00294923)%| LAZY_AND: Hlt2BnoC_BdsToKSKpPim_DD #=255965 Sum=57 Eff=|(0.02226867 +- 0.00294923)%| LAZY_AND: Hlt2BnoC_BdsToKSKpPim_LLDecisionWithOutput #=255965 Sum=34 Eff=|(0.01328307 +- 0.00227788)%| LAZY_AND: Hlt2BnoC_BdsToKSKpPim_LL #=255965 Sum=34 Eff=|(0.01328307 +- 0.00227788)%|Ran on 255965 evts
CAVEAT: this is only looking into the KSKpPim Line, plan to look effect on all KShh
Edited by Davide Lancierini -
Next:
- Will run now a BW test with minpt sum of all daughters to 4200 to see what impact it has as it seemed to have a big impact on BW
-
Can't seem to be able to tweak the lines and run HltEffChecker within DaVinci to produce simulation ntuples with different cuts wrt the the ones on master, trying to get help from @fswystun on this, but more help would be appreciated.EDIT: I am able to look at the effect on simulation of changing the cuts, so will now investigate that as well
Edited by Davide Lancierini added 1 commit
- 3db04c49 - moved to med_bpvipchi2_min between 2 of the 3 daughters, relaxed asumpt cut
-
With latest edit:
dlancier@lxplus991/afs/cern.ch/work/d/dlancier/private/stack $ grep "LAZY_AND: Hlt2BnoC_BdsToKSKpPim_DD" hlt2_bnoc_data_bandwidth_example_master_removing_asumpt_maxp_maxpt_cuts.log LAZY_AND: Hlt2BnoC_BdsToKSKpPim_DDDecisionWithOutput #=255965 Sum=50 Eff=|(0.01953392 +- 0.00276224)%| LAZY_AND: Hlt2BnoC_BdsToKSKpPim_DD #=255965 Sum=50 Eff=|(0.01953392 +- 0.00276224)%| dlancier@lxplus991/afs/cern.ch/work/d/dlancier/private/stack $ grep "LAZY_AND: Hlt2BnoC_BdsToKSKpPim_LL" hlt2_bnoc_data_bandwidth_example_master_removing_asumpt_maxp_maxpt_cuts.log LAZY_AND: Hlt2BnoC_BdsToKSKpPim_LLDecisionWithOutput #=255965 Sum=25 Eff=|(0.009766960 +- 0.00195330)%| LAZY_AND: Hlt2BnoC_BdsToKSKpPim_LL #=255965 Sum=25 Eff=|(0.009766960 +- 0.00195330)%|And BW is reduced to:
0.03275 GB/s