Blog from January, 2013

Pass8 GR 20 08 02 Outstanding Problems
Johan BREGEON posted on Jan 21, 2013

Page created for the Pass8 Winter 2013 Workshop in Paris.

Outstanding problems

I chose to list and document here issues found during the Monte Carlo simulation production phase and the data reprocessing with GR-20-08-02, but also a few long standing issues that should now get fixed.

  • General
    • slow: a factor of 2 slower than Pass7 release (not acceptable)
      • can't run more than 5k gammas or 10k bkg events on the long queue
      • can't reprocess more than 5k events per job (need 500 long queue jobs per run)
      • what can be turned off ? (Old profile, Acd1, ?)
      • what can be speed up ? (MST, CalClassify, New profile, Tree Based Tracking ?)
  • Data reprocessing
    • Unguarded sort LPATE-73@JIRA: blocker (can't reprocess full data runs, crash for 1 event out of 500k)
    • other bug ? have to check more log files...
  • MC Production
    • Geant4 exception LPATE-51@JIRA
      • unknown particles
      • alpha with too much energy (still a problem ?) LPATE-50@JIRA
    • Failed to write event to McTree LPATE-7@JIRA
    • other bug ? (some crashes in recent AG, same bug or new one ?)

Other known bugs

  • See LPATE-84@JIRA
    • Errors in TkrEnergyTool in the TkrUtil package LPATE-85@JIRA
      • the UB2val[6] bug (instead of UB2val[9]): it should affect only events converting in layer2
      • the weight used to merge the parametric and profile is inverted. It affects all events.
      • That's why Tkr1ConEne and EvtJointEnergy are different between 1 and 10 GeV in 20-08-02.
    • LPATE-83@JIRA : The merit variable TkrNumTracks can only have values 0,1 or 2
  • LPATE-91@JIRA: we can have CalEnergyCorr==0 when CalEnergyRaw>0.
  • We do not always correctly take the 1st track in the all algorithms.