Update: Following the update I gave today on the joint AMG-DAOD meeting, the proposed additional containers has been blessed by the AMG and TauCP group. I am removing the draft identifier. The migration to the Component Accumulator scheduling is attempted. The code works but it is quite ad-hoc, tagging @jcatmore for help. Also, tagging AMG experts @vimartin and @hrussell for technical help. Thank you all in advance!
Following !50531 (merged), the alternative reconstruction of the hadronically decay tau lepton at AOD level, we propose several tau-related new containers in the DAOD_PHYS derivation. These new objects enables us to tackle the TauID inefficiency when the muon from a nearby leptonically decaying tau overlap with the tau_seed by simply remove the muon tracks and clusters. More detail can be found in ATLTAU-1762. An internal note which documents the results has recently been uploaded. http://cds.cern.ch/record/2807409
Here is a summary of the footprint of these new containers in the worst-case-scenario with ttbar events:
- CPU: 5 ms/evt in average, which is a 0.7% increase in runtime, bringing the total CPU time to 645 ms/evt on my local machine.
- Taus: ~8% additional tau candidates were saved.
- Disk: 0.3 KB/evt more disk usage for tau, from 1.89 KB/evt to 2.19 KB/evt. ~1% size increase overall.
For completeness, here is the footprint in data:
- CPU: 4 ms/evt in average, which is a 0.7% increase in runtime, bringing the total CPU time to 689 ms/evt on my local machine.
- Taus: ~1.1% additional tau candidates were saved.
- Disk: 0.16 KB/evt more disk usage for tau, bringing the overall size to 26.135 KB/evt, ~0.6% size increase.