LHCb trigger twiki pages

General information

Planning

Information for Hlt Piquets and WG liaisons

The HLT piquets should refer to the HLT Piquet Guide

Moore testing

It is expected that anybody developing HLT code, including Hlt2 line authors, should keep an eye on these automatic tests, and report any issues.

Nightly qmtests Look here under lhcb-head (Moore and MooreOnline). In order to run all tests, you can do: make test ARGS ="-j 12"
Throughput tests with a farm node Pick the latest entry from here Seee also LBCORE-1007, and these instructions to test Moore in the online environment on a dedicated node.
HLT rates See these instructions to use the Moore_RateTest script yourself.
Tests are run on a daily basis within the lhcbpr framework, using 30k L0 filtered events from the TestFileDB.
Tests are run on samples corresponding to a few different L0 TCKs, and on different nightly build slots.
The results are published here. See also gitlab-repo, and these slides.
The test runs over 30k events by default to run over more events specify the --evtmax option.
Hlt1 independence test Under MooreOnline in the nightlies
There is some older documentation here.

Developing Hlt2 lines

2018 startup deadlines: presentation of new lines to be included in 2018 is February 2nd. Deadline for committing new lines will be ~6 weeks later (March 16th, subject to changes in the LHC schedule).

If you wish to introduce a new Hlt2 line, or retune an existing one, please follow the steps listed in the table below.
The aim is to have a turnaround time of a few weeks for a new or updated Hlt2 line to enter the data taking. This means there are no specific dead lines and we try to be as flexible as possible.
This turnaround time does not apply at the beginning of a year when potentially many new lines are added at once.

1 Contact your WG liason Discuss with them any questions you have on the following steps
2 Determine the signal rate in the LHCb acceptance This should be without any selection or reconstruction cuts (so you don't need MC)
3 Carefully consider need for Turbo(++) or FULL Please keep your liaison involved
4 Get signal MC Request or make sure you have already a small signal MC sample with raw banks
5 Read the general guidelines Make sure you are prepared to follow these instructions, if not speak with your liason.
6 First HLT ops presentation After completing the above steps present on the general plan for the new line and tell us your signal rate For new lines intended for the 2018 startup, this must be done by Feb 2nd.
7 Create an LBHLT JIRA task See for example LBHLT-124
8 Write your trigger line Follow these instructions.
9 Instrument your line with at least one monitoring histogram detailed instructions here.
10 Measure the rates of your line(s) Using with at least 50000 HLT1 accepted events. A dataset is provided in the TestFileDB and there is an example on how to process it with the Moore_RateTest on this page.
11 Measure the efficiency of your lines(s) Again using the standard Moore_RateTest script, with juggled signal MC, or data, where possible.
12 Estimate the purity of your line(s) For rare/forbidden/unobserved modes, it is suggested to use e.g. the current world's best limit (cross section x) BF, or if there is no limit, you should use a limit that would be considered "interesting".
13 Independence test Run the Moore_RateTest script without your line included and use the logfiles (and ntuples) to check that all other lines give the same inclusive rate.
14 Tuning knobs We need to see the purity and efficiency as a function of rate for a tunable parameter (a cut you're happy to tighten or at worst a prescale)
15 Second HLT ops presentation It should include the results of all of the above steps. The deadline for this presentation for new lines for the 2018 startup is March 16th.
16 Merge request Can happen before the second presentation. Mark as WIP if still work in progress. The target should be 2018-patches
17 Create JIRA subtask for new TCK  
18 Check the next nightly tests  
19 Release manager accepts merge request  
20 Update Tesla configuration In case a new Turbo or TurCal line is updated, the Tesla production step needs to be updated to know about the new line, otherwise the productions will crash. The HLT should inform the production team about this change in advance.
21 HLT piquet to prepare new HLT2 TCK  
22 PPG/OPG sign off for deployment  
23 TCK deployment  
24 Validation of data Line authors are expected to look at the monitoring histograms for the lines, and at the output data, and report back to HLT ops within 1-2 weeks of first data with the new TCK.

The Tesla project

HLT Monitoring

Level-0

  • TCKs: All released ones as listed here. See also slide-9.
  • Bandwidth division: More details on this page
  • Testing L0 TCKs: See these instructions to run L0App and print the rates.
  • Software: See this page on the L0App application, and these details on L0Muon emulation.

Trigger performance studies

  • Timing Studies: Checking the impact of increasing the number of Kalman iterations on Hlt execution time
  • Hlt2Rates: Rates of Hlt2 lines
  • HltCorrelations: Correlations of L0, Hlt and Stripping lines

Upgrade Trigger development

Physics working group specific pages

About the HLT

Old pages

A list of historic, and probably obsolete, pages on the LHCbTrigger is here.

Topic attachments
I Attachment History Action Size Date Who Comment
PNGpng LHCb_2012_Deferred_Trigger_Plot.001.png r1 manage 90.0 K 2013-10-23 - 11:45 VladimirGligorov  
PDFpdf LHCb_2012_Deferred_Trigger_Plot.pdf r1 manage 19.2 K 2013-10-23 - 11:25 VladimirGligorov  
PDFpdf LHCb_Deferred_Trigger_Plot.key.pdf r1 manage 19.8 K 2013-09-24 - 14:52 VladimirGligorov  
PDFpdf LHCb_Split_Trigger_Plot.key.pdf r1 manage 19.5 K 2013-09-24 - 14:51 VladimirGligorov  
Unknown file formatgzip chepPosterNeubert.tar.gzip r1 manage 2734.3 K 2013-10-15 - 14:53 SebastianNeubert Sources for the poster on HLT Architecture beyond LS1 presented at CHEP2013

This topic: LHCb > LHCbTrigger
Topic revision: r122 - 2018-11-02 - AlexPearce
 
This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback