Core Minutes 1/23/2007

• MRStudio: (Richard, Joanne for Riccardo) As promised at the last meeting, on Windows he has added creation of Visual Studio solution files. On Linux he fixed a bug causing a crash when one attempted to save a configuration. It now crashes when one tries to select a package from the package list. Once the Linux development environment on his new laptop is fully in place he should be able to fix this one.
• ROOT upgrade: (Richard for Heather) She is consulting with ROOT wizards to understand and eliminate a warning message coming from AcdDigi.
• Hadronic physics list: (Francesco Longo email) While working on an analysis of Beamtest proton data Johan Bregon discovered an error in the setting of job options for G4Generator so that hadronic models were not included, as of GR v9r20. (Richard) It's hard to believe no one caught this earlier, but apparently no one did.
• Irreducible backgrounds/MC output mods: (Tracy) See his comprehensive explanation, complete with pictures, of what he proposes and what effects it will have. In summary, it will
  • address the original request for tools to identify irreducible backgrounds
  • (Leon) allow Fred to show a much better representation of events when read back in, something which has caused confusion in the past
  • take somewhat more disk space on average for gammas but less for background events. Most likely we'll win overall.

  In the process of doing this work, Tracy found and fixed a memory leak. It would be nice to know just how bad it was. The whole thing should be ready in about a week.

• Level 0 processing: (Tom) [from his written notes, with a little augmentation here and there] What we have now:
  • 1 min of partial background
  • ~30 seconds of full background + full sky (except pulsars).

  The 1 minute background test sample has been processed through the Level 0.5 steps and produced a so-called ".evt" file which has been handed over to Larry Wai. (I've even processed this .evt file myself to create a digi.root file which, at least superficially, looks good - it has the correct structure and number of events.)

  I'm using Pipeline I to configure (and debug) the Gleam steps and Pipeline II to set up final task. There are issues to be resolved for both pipelines. Concerning P1:

  • Pulsars cause Gleam crash - still being investigated by experts
  • Memory
    • Moon takes ~650MB right off the bat. New Sun and Moon sources "in the works" [see Toby's remarks below]
    • Memory leak somewhere (growth of >500MB in 6 orbit-sec sim). (Richard) Maybe it's the leak Tracy found in G4Generator.
  • Optimized build needed (currently takes 6-9 hours to sim 6 orbit seconds!)

  Concerning P2:

  • Would like full environment variable access (e.g. to variables set by previous job steps, needed for concatenation)
  • Need a way to augment tasks with new runs without losing benefits of P2
  • Directory clean-up mechanism for redos. For Pipeline 1, Tom did this manually; it's more complicated for P2.
  • Working directory management (pipeline_wrapper/pipeline_summary issue). There are some aspects of P2 which make it difficult to reuse old working directories.

  (Toby) We had implemented the sun and moon as a sequence of point sources (735 for the sun and 8761 for the moon out of a total of about 16,000 sources) but it doesn't have to be done that way. With the new tag of flux the sun and moon won't be such memory hogs.

  (Larry) has done some processing of the .evt file Tom produced; so far (ccsds to digi), so good. He and Warren are working on different parts of the processing. (Richard) Have you looked at the file in detail to see if the bits are reasonable? (Larry) Not yet. That will happen as a matter of course by running the Digi Report, which is something that has to be done anyway. (Richard) When will this be done? In time for the End-to-end test? (Larry) We will get it done as fast as we can. The End-to-end test is a good goal.

• Filter status word: (Richard, Tracy) There was some discussion at the C & A meeting about use of the Filter Status word. Currently various pieces of code are comparing the top (16-bit) word to zero to see if an event has been vetoed. The information of interest is really contained in just the most significant bit of the word. There has been talk of masking off other bits, but then we lose the possibility of accessing original value of those bits. (Tracy) JJ cannot promise that the meaning for all the bits in this word will stay constant, however the most significant bit will always be the summary veto bit; that is the one to check to see if an event is vetoed. [Outside of the OnboardFilter and OnboardFilterTds packages, there are 10 *.cxx files and 7 *.h files in GlastRelease packages containing the string "FilterStatus". These are the files which might be testing for veto improperly. How hard can it be to fix this? Unfortunately, as Toby has pointed out, the issue is a little more complicated because it affects analyses of ROOT data beyond the GlastRelease c++ code. ed.]  
• Disk space: (Richard) The 50 Tbyte order has gone to Purchasing and they have put it up for bid. Typically it is 2 months from the time a vendor is selected until the disk is actually available. It's going to be tight.

previous

minutes index

next

J. Bogart, Last Modified: 01-Jun-2010 15:45:56 -0700