Minutes of the Computing meeting held on 28 August 2000

Present: M.Cattaneo, G.Corti, I.Belayev, J.Harvey, K.Harrison, M.Needham, J.Closier,  I.Korolko, O.Callot,
              P.Mato, F.Ranjard, B.Hay,  F.Harris, O.Schneider, I.Videau, A.Jacholkowska 

Minutes : Florence Ranjard

   1. SICBMC v234 and SICBDST v236 (Flr)

• These versions are installed on $LHCBSOFT but cannot be used for production:
  • There is not yet a trigger correctly implemented for this version
  • Current production of DST2 with SICBDST v235 is currently stopped because of too many crashes in RICH code:
    • these crashes are probably due to the huge number of tracks because they happen:
      • every 2 events when making pileup with SICBDST v235 + dbase v229  (extended tracking)
      • less often when making pileup with SICBDST v235 + dbase v228 (no extended tracking)
      • times to times with single events with SICBDST v234r2 + dbase v229
    • Eric will start a DST2 production with SICBDST v235 + dbase v228 when physicists have looked at the first results obtained by the Data Quality Program on some DST2 produced with these versions.
    • From some analysis done by I.Korolko it seems that there is too much material in front of the Calorimeter in the RICH2 region.
      • It would be useful to run a dedicated job to compute the radiation length seen by a geantino in the apparatus everytime there are changes in the geometry. (Flr will look at it).

     2. Brunel v1r3 is released (M.Cattaneo)

• Brunel is now linked with SICBDST v235.
• This version should be checked by all subdetectors:
  • Marco has produced DST with SICBDST and with Brunel for comparison

    3. Discussion on proposed conventions (M.Cattaneo) (slides)

• Units : The proposal is to adopt the CLHEP units (whatever they are)
  • that implies:
    • to input a number you have to multiply it by the unit : P=75.*kg
    • to output a number you have to divide it by the unit : print P/kg  to get 75.
  • jobOptions will understand "*kg".
  • This choice will create a lot of "*1." and "/1." : make sure that compilers are able to avoid such operations.
• Event data model conventions:
  • Some naming conventions are proposed which should apply to permanent data.
  • navigation from simulated raw data bach to the MC Hits truth information will be by inheritance.
  • A strong rule: navigation should not produce "I/O overhead". To avoid such overhead it could be possible to make persistent some intermediate objects built during reconstruction to be used during analysis.
• Access to MC truth:
  • Implement a set of Gaudi Tools, called associators to shield the code from the implementation details of the navigation procedure.
    • Gloria Corti will prepare some examples of such associators.

           ===> A final proposal will be discussed at the computing meeting in 2 weeks: 11 September

===============================================================

next meeting :  3 September 2000 at 14:00