BnoC: algorithm for calculating and persisting related info (e.g. isolation information) example implementation in Moore (!4319) · Merge requests · LHCb / Moore

Merged Francesca Swystun requested to merge isolation_reltables into master 2 months ago

Summary

Use of algorithm implemented in Rec to add isolation relation tables to the extra_outputs of a HLT2/Spruce line.

Modifies Hlt/Hlt2Conf/python/Hlt2Conf/isolation.py to add extra_outputs_for_relinfo_isolation() which directly uses the FunctorsToRelTable algorithm
Updates BnoC code to include relatables method of persisting isolation extra_outputs within WG code
Add set of reltable_kwargs to BdsToPhiPhi line that is used in integration testing, DaVinci examples & validation

Dependencies

Needs Rec!4251 (merged), LHCb!4921 (merged)

Downstream dependencies/testing

For example of tupling the output see DaVinci!1229 (merged), TestFileDB for testing of DaVinciExamples added in lhcb-datapkg/PRConfig!535 (merged)

Moore->DaVinci chain tested in LHCbIntegrationTests!95 (merged)

Also tested here, TurboSpruce persistence adding P2InfoRelations into extra_outputs of the threebody_line used in the turbospruce unit test

See !4558 (comment 9224558) for adding into unit test, see !4319 (ed734530)

Testing Summary

762808d6 - Update to apply some default options across BnoC lines (where specific requests for cone angles weren't made). Persist for BnoC isolation lines + some requested in WG survey:

Cone isolation: Relation tables for the reference particle with respect to particles (make_photons() for "NeutralIso", make_long_pions() for "LongTrackIso") in a cone of deltaR=1.0 & deltaR=1.5 around the reference particle. Includes isolation information of:
- Cone's multiplicity (number of selected particles in cone)
- Cone's P[ X, Y, Z],
- Cone Asymmetry in PT,
- Cone Delta PHI
Vertex isolation: Caluclates the minimum IPCHI2 wrt the reference particle's decay vertex among a set of long tracks (equiv. of SMALLEST_DELTACHI2)
Propose these options as a baseline for conducting some initial throughput & bandwidth testing of the implementation
- Suggest any global update to BnoC options to go after this MR (these changes have been reverted)
Test results in test triggered in: !4319 (comment 9259506)
- Summarise here to inform future MR
Reductions in overall HLT2, Spruce (full) bandwidth in BnoC lines as expected, no significant decrease in throughput seen
- 21 Turbo HLT2 isolation lines in master (+ 19 at authors request)
- 37 Sprucing isolation lines in master (+ 3 at authors request)
Spruce(turbo)
Spruce(full)