Core Minutes 1/15/2008

• Big run: (Richard) The latest GR tag, v13r9 includes Tracy's OBF enhancements and a last ACD bug fix from Eric C. Last of validation runs, 20,000 orbit seconds of background, is in progress [complete as of 10:30 this morning]. Next order of business is Interleave. Meanwhile, Sky Model is undergoing final touches. When it's ready, will make 8 days of backgrounds + Sky Model for OpsSim, subtract off Sky Model for Big Run. Follow progress on the Big Run Checklist.

  (Philippe) Still sees a difference in ACD +Y and -Y sides. Tracy's track fix was effective.

• More GlastRelease: (Heather) What's the status on Interleave work? (Tracy) Same place it was before. (Heather) The question should perhaps include CEL. (David) has been investigating the use of external friend chains when reading back in CELs, but is not confident it is robust enough. Investigation will go on, but in the short term he plans to change the current implementation of Gleam reading mechanisms, replacing the use of external friend chains with the "shallow read" or "deep read" features which were recently added to the CEL class.  He'll have the time as long as there is no immediate need for more Skimmer features. (Richard) Should be ok; the deadline for CEL+Interleave functionality is about a month from now.

• ACD Review (Heather, Leon) A two-pronged review is in the works:

  1. Standard code review. Package owner does a walk-through of code. Also a time to check for compliance with coding and documentation standards. Timing for this largely depends on how soon Eric C. will be ready. To be followed by
  2. A more strategic approach. Can some of the elaborate geometry-related logic be extracted (to, e.g., G4Propagator?) and made more general?

  (Richard, Heather) Alex Moiseev and Dave Thompson should be involved in this process.

• SCons status: (Heather for Navid) The new RM is in test. First-pass development is done; it may need some wringing-out.

  Tags in the new format are being automatically created by RM; e.g., manual creation of tag v1r9p5 of celestialSources/genericSources results in the additional tag genericSources-01-09-05. (Toby) These tags don't seem to be used for anything. Do we need them? (Heather) They are preferred because they can be sorted and ultimately will be required to impose uniqueness. [What a memory! Here is a quote from Navid's Dec. 17 email on the subject:

  Hi all,
     I'm hoping we can come up with a new tagging convention in this
  email before the meeting tomorrow so I can announce tomorrow that the
  RM will start converting old tagging formats applied to builds to our
  new agreed upon format. This is a transition away from the old tagging
  convention.
  
  Anyways I don't have any strong preference over one format or another
  as long as it has two features:
  1) Easily sortable when comparing two tags from the same package
  2) Unique across all packages
  
  The first point is so that the RM can finally reliably sort the
  packages in order in which they should be displayed. The second one is
  for SCons and our new CVS layout style I explained in the summer.
  

  ed.]

• Resources: (Richard) In this time of straitened budgets we need to minimize expenditures on hardware. We originally slated to purchase an additional 400 cores; these will instead come out of the SLAC farm. Disk is harder to conjure up. We were expecting to obtain an additional 150 TByte for real data and 100 TByte for MC. (Richard) will be looking into ways to delay the purchase. Possibilities include reusing old disks and looking into ways to streamline output. TkrRecon is a prime candidate for the latter. While we're in a conserving mood, we should also think about speeding up execution. (Toby) We could look into OpenMP. From the Wikipedia article:

  The OpenMP (Open Multi-Processing) is an application programming interface (API) that supports multi-platform shared memory multiprocessing programming in C/C++ and Fortran on many architectures, including Unix and Microsoft Windows platforms. It consists of a set of compiler directives, library routines, and environment variables that influence run-time behavior.

  VS 2005 supports it as well as gcc [gcc 4.2, that is, not yet one of our supported compilers. ed.]

• New meeting plan: Approximately the same set of people has been attending the Tuesday Core Meeting and the Thursday General Software meeting; would be nice to somehow compress into a single meeting. The following was agreeable to everyone present:

  There will be a single meeting which will take place on Tuesdays. It will start with the summary reports given at the Thursday meeting (Science Tools, Data Handling, etc.) followed by the more open-ended discussion typical of Core meetings.

  [May I ask that those giving summary reports either talk from a presentation which may be linked in or send me remarks in plain text afterwards? In fact, it's a good idea for anyone giving an extended report to provide something for the minutes: it guarantees that you will be represented accurately. Thanks.]

• AOB:
  • (Toby) Can someone explain where the output at the bottom of this log file comes from? Do we need it?
  • (Leon) Concerning the question last week about continuing OBF support on Windows: have been talking about using output of Onboard Filter as part of a Tracker or Event prefilter. So OBF would play an integral role in event processing.
  • Heather gave a quick run-down on status of some of the unresolved issues of interest just before Hip 2.0.
  • (Richard) Julie is about to have all four — all four impacted — wisdom teeth removed. She probably won't be too responsive this week. Best wishes for a quick recovery, Julie!
  • ..and the same to Navid, who just broke a leg!

previous

minutes index

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