80 likes | 260 Views
2. Well attended (~10 participants)Tutorial of running athenaMT with the calorimeter slice (egamma, tau, jet) by Xin Wu Tutorial of running athenaPT with the calorimeter slice (egamma) by Imma RiuTutorial of running athenaMT with online monitoring histograms with the calorimeter slice (egamma)
E N D
1. 1 HLT Algorithm Integration Workshop Summary and Future Plan
2. 2 Well attended (~10 participants)
Tutorial of running athenaMT with the calorimeter slice (egamma, tau, jet) by Xin Wu
Tutorial of running athenaPT with the calorimeter slice (egamma) by Imma Riu
Tutorial of running athenaMT with online monitoring histograms with the calorimeter slice (egamma) by Till Eifert
Demonstration of creating and running an online partition by Andre dos Anjos and Per Werner
Demonstration of running an online EF-monitoring partition by Serge Sushkov
Feedbacks received are generally good
Could be repeated in the future, probably with some new features:
With release 12?
With database configuration?
Combined with tutorials on running online partitions?
Shifters for Large Scale Test
Offsite testbeds for algorithm integration
3. 3 Attended mainly by algorithm developers (~10 people)
Some consensus reached on algorithm online integration
HLT algorithms developed in the offline framework have made great progresses and time is ripe for integration tests
First step of integration is best organized as trigger slices
Egamma, muon, tau, jet/etmiss, B-physics, …
Each slice has 1-2 people responsible for online integration
People identified
Egamma: Xin Wu, Imma Riu
Muon: Alessandro Di Mattia, Diana Scannicchio
Tau: Pilar Casado
Jet/Etmiss: Ignacio Aracena, John Idarraga
Need to have person(s) to serve as centralized interface to the HLT integration team and to coordinate slice integration
Follow HLT releases, provide examples, follow up feedbacks and requests, produce combined jo’s and BS files, …
People identified (for the next few months …)
Imma Riu, Xin Wu
4. 4 HLT algorithms developed in the offline framework
Read MC pool RDO files and use transient BS
Run together with Reconstruction
Well suited for trigger performance studies (efficiency/rate, physics)
Steps tried in algorithm online integration with 10.0.6
Setup a slice running in the offline framework
Derive a minimal jobOption to run with BS input
Reduce to minimum couplings to high level jo files (flags) for offline reconstruction and offline trigger studies
Test the jo with athenaMT
Compatibility mode, single thread, multithread
Combine slices to a single jo
Produce BS input file including different L1 RoI type
Test the combined jo with athenaMT
Pass the slice and combined jo to HLT integration team
Tried with muon (Mufast+InDet) and egamma (T2Calo+InDet)
Also try standalone athenaPT and EF integration
5. 5 Follow the similar strategy as 10.0.6
Ask people from slices to setup slice integration
Use the tutorial from the workshop as starting point
Produce a jo of a full slice (LVL2+EF) with BS input
Produce a BS input file from CSC production
Test it with athenaMT and athenaPT
If possible include LVL2 algorithm monitoring histogram
Same jo should work both online and offline
Default is full slice offline
Use flag doMT, doPT, doLVL2Only, doEFOnly
Keep an working example in 11.0.6 nightlies
Egamma slice will be ready in a few of days
Will run in HLT nightly test
Other slices are in good progress and will be added when ready
Aim for egamma, muon, tau and jet slice integration
Also try to have combined integrations
Eg: egamma+muon, egamma+tau+jet, …
Encourage algorithm developers to participate in pre-series test
6. 6 HLT algorithm integration was well treated during the workshop
Tutorial well appreciated
Parallel discussion reached consensus on integration strategy
A team of people from the algorithm slices are working for integration with 11.0.6
First to integrate some individual slices
Egamma: almost ready
Jiri Masik integrated the calo slice (T2Calo+TrigCaloRec) of 11.0.5 into HLT image in short time after the release was out
Muon: LVL2 almost ready, EF in good progress
Tau and Jet: LVL2 almost ready, EF starting
Then try combining slices
If the current approach works will try integration with release 12
Ensure strong participation of algorithm developers in the next Large Scale Test
Also in P1 pre-series tests