Project

General

Profile

Meeting on May 29 2015

9:45am (Buenos Aires)
7:45am (Chicago)

(Attendant: Rodrigo, Lucio, Nico, Soon)

1. initial performance comparison
- running the default configuration of N02 gave the ending particle time at 65ns (i.e, to travel the maximum track length set in the macro)
wall-clock time: Geant4 ~ 4 sec (including I/O ~ 15 sec) vs. PD ~ 0.5 sec (using the delta quanta = delta one step in Geant4 ( 1.0e-6 nm))
- Soon suggested to measure the Geant4 elapsed time between the beginning of event and the end of event
To see an example of the G4 timer implementation:

    wget https://oink.fnal.gov/perfanalysis/g4p/admin/application/cmsExpMT.tgz 
    tar -xzf cmsExpMT
see eventTimer in cmsExpMT/src/cmsExpEventAction.cc (i.e., at BeginOfEventAction and EndOfEventAction)

2. action items
- add the Geant4 timer into N02to and update the code (Nico, Lucio)
- improve performance measurements
- understand how to control the error (delta quanta vs. the maximal geant4 stepper error or errors(?))
- make kerberos work (Lucio)

3. potential items (not urgent as we should focus on the current example to understand details, but something that we should start to think)
- try the more complicated (finely segmented) geometry in order to test the boundary crossing
- adopt a parametrized field to see/study a potential deviation between PD and Geant4
(i.e., global error control of PD vs. per-step-basis error control of Geant4)