Skip to content

Fix assignment of CountHepMC for LHE only jobs

Spyros Argyropoulos requested to merge lP into master

Description of bug

CountHepMC had no meaning for LHE-only jobs, since only a single event is written to EVNT. This would lead to wrong results in particular for the CPU extrapolation.

Changes introduced

  • For LHE only jobs assign CountHepMC to be equal to nEventsRequested (the latter is specified by Py:Gen_tf ... Requested output events)
  • Skip check of number of events written to EVNT - essentially we accept any number of events for an LHE file (@cgutscho please confirm this is what we want)

Tests

Before fix

---------------------
Generate transform params:
---------------------
- ecmEnergy =  13000
- nEventsPerJob =  304
- Requested output events =  104
- transform =  Gen_tf
- inputFilesPerJob =  0

---------------------
Performance metrics:
---------------------
- actual CPU (1 events) = 3.46 hrs
- CPU extrapolated to 304 events = 1053.2 hrs
- CPU =  1053.24 hrs  <-- ERROR: Too high CPU time - should be between 6-12h. Adjust nEventsPerJob!
- estimated CPU for CI job = 10.53 hrs
- Virtual memory =  1137.113 Mb

---------------------
 Summary:
---------------------
Errors : 2 , Warnings : 3  -> Errors encountered! Not ready for production!

After fix

---------------------
Generate transform params:
---------------------
- ecmEnergy =  13000
- nEventsPerJob =  304
- Requested output events =  104
- transform =  Gen_tf
- inputFilesPerJob =  0

...

---------------------
Performance metrics:
---------------------
- actual CPU (104 events) = 3.46 hrs
- CPU extrapolated to 304 events = 10.1 hrs
- CPU =  10.13 hrs
- estimated CPU for CI job = 0.10 hrs
- Virtual memory =  1137.113 Mb

---------------------
 Summary:
---------------------
Errors : 0 , Warnings : 3  -> Some warnings encountered, check that these are ok before submitting production!

Issues resolved

Closes #

Edited by Spyros Argyropoulos

Merge request reports