- Mainly focussed on collating the needs of people for analysing 2017 data/MC & what updated need to happen in the code.
- Currently the JERC members are using CMSSW 92X for their studies, using 2017 data (JetHT & SingleMu datasets), along with 2016 MC.
— Jets & muons are good, other objects may not be up to date
- Upgrade studies need 93X
- People doing proper analyses (eg for Moriond ’18) as well as POG people will need data & MC in 94X.
eg
— Top tagging SF
— Dijets analysis
— JERC
— Long-lived analysis
- Samples envisaged so far: data = JetHT, SingleMu; MC = TTbar, WJets, Single Top
- Moving NTuple code to 93X/94X is not easy - GCC changes from 5.3.0 to 6.3.0, and various jet config file in CMSSW no longer exist/have moved e.g. the jet collection producers.
— Try to port once 940 released (ETA ~ 3rd Nov)
- Would like to move away from Thomas' public installation of fastjet, especially since 2 versions required for different GCC versions. But:
— CMSSW version too old, API changes mean we need >=3.2.1. Wll have to stick with central installed version for now.
— HOTVR / XCone should be a fastjet-contrib. Some small changes required first (Roman)
- Possible tidying up of (jet) collections - need to review which jet collections are required by which people
— JERC people want lower pT threshold on e.g. AK8 jets so need to redo clustering when making ntuples
— Btagging - no one explicitly working on it in the room, but Matteo Defranchis will be
- Would be good to have a "contact person" per object to keep track of & help implement updates
— Arne already B2G Muon contact
- Performance of code:
— Why low number of events/job (otherwise timeout)
— Occasional memory leaks in some events (affects <~1% of jobs)
— Check speed - can we cut out any unnecessary modules?
- Keep an eye on MINIAOD content
— https://twiki.cern.ch/twiki/bin/view/CMSPublic/WorkBookMiniAOD2017 is being updated, the jets bit should be correct, others may not be
— e.g. slimmedJetsAK8 will be by default with PUPPI not CHS
— Other things that need updating/adding in e.g. electron HEEP ID
- CI: tried on travis, but stuck due to fastjet issue
— ETH had theirs on gitlab with gitlabci - maybe worth investigating as easy to setup
- PRs should be applied to master as well as release-specific branches (as necessary)
- Another meeting to happen when 940 released (with some MC samples to test on)
There are minutes attached to this event.
Show them.
