Core Minutes 2/1/2011

• ScienceTools: (Jim) couldn't be at the meeting but sends the following:

  • Tagged ScienceTools-09-21-00 on Sunday. This release contains the P6_V11_DIFFUSE IRFs in caldb. It also contains some fixes for the binned likelihood analysis that are noted in the ScienceTools Development update notes.
  • With Johann, I've added the capability of specifying individual prior distributions for model parameters in the Likelihood package. Johann has been doing some testing in the context of the dwarf spheroidal stacking analysis and reports that he is finding reasonable-looking results. This work is checked into the HEADs of optimizers, Likelihood and pyLikelihood, but none of it is tagged yet.
  • For the Mac builds, I disabled the parts of the test code in the optimizers, Likelihood and sane packages that use the f2c'd optimization code (Minuit). The unit tests for all three packages seem to work ok now (modulo disk access problems on bldmac02.) A similar change needs to be made for the burstFit package, but otherwise, the snow leopard builds should be good to go in this regard.

• FSSC: (Eric) They're getting ready for a release in May. There will be an announcement. Meanwhile, he has imported v9r21. It is not trouble-free but no serious problems are expected in supporting it.

• Documentation: (Chuck) sends the following:

  There have been a substantial number of updates to the workbook this past week, including updates to a "gtexpcube2" as well as number of the other Science Tools pages, "Using the SLAC Batch Farm" (to repair broken Ganglia links), and a major restructuring of the navigation bars at the top levels. With that in mind, please take a look at the Change Log to see if any of the changes listed there affect pages you are interested in.

  I'm in the process of updating CVS and will be creating a new tag this week, which will enable Berrie to update the mirror site.

• GR tasks: (Tom G.) The GlastRandomSvc currently rolls over at 1,000,000. This is a limitation for current MC tasks running at Lyon and SLAC. Michael Kuss has worked up a new scheme to allow for 10e9 seeds. This new code needs to be incorporated into a new GR. The currently running MC tasks were inadvertently set up to duplicate random number seed (one seed per run and blocks were defined from 0-999999, and 10000000-up) so ~200k runs (streams) are formally duplicates. However, a close look at the events do not show duplication. Speculation is that different hardware and OS/compiler versions could introduce different floating point issues which can make identical code diverge. Still under investigation.

  Last week 10,917 Fermi runs were reprocessed to generate new FITS files (FT1, LS1, ELECTRONFT1) using the Pass7.4 event classification scheme. These data span the period 4 Aug 2008 through 31 July 2010. Starting today, these data are being ingested into the astroserver, a process that is expected to take at least one day to complete.

• Pass8: (Tracy) It turned out v18r8p5 was not perfect, after all. We're aiming for p6 by COB today which will include/address the following:

  • AcdRecon was not protected against bad input, hence would crash. Eric has fixed this.
  • Tracking would on occasion produce negative chi squares (hence bad input for AcdRecon). Tracy fixed this one.
  • More CalRecon improvements
  • When the McIntegratingHits structure was modified last fall to include crystal end effects in the crystals it meant a new set of data structures were introduced, one for each diode. Unfortunately, this was not included in the overlay, which places the cal information into the McIntegratingHits in order to have it available for the trigger. Since the new digitization scheme for the cal relies on these McIntegratingHit data structures, in the case of hit crystals which are pure overlay, the digitization was, basically, reading random memory locations often resulting in TeV type energies. Now fixed.
  • Use of TMineExt (new external version of TMine) rather than the old package.

• Core talk:

  • Snow Leopard (Tom S.) Since Jim made changes [to eliminate use of f2c for this platform only] the builds run, but there are still some issues to be addressed:

    • Loss of AFS token. This happens on occasion. When it does, the job can't read grits.conf, hence can't connect to the database. The work-around is for Tom (or someone) to log in on the machine, thereby getting a new token. (Tom G.) This shouldn't happen. unix-admin should be informed.
    • RM-started processes don't time out as they should.

  • Windows and nfs (Heather) Windows boxes have had trouble accessing nfs disks in the usual manner. The work-around is to access them via samba01 rather than an alias.

  • Glastwin04 (Tom S.) It's ready for use. He will turn on Release and HEAD builds.

  • Passwordless login (Heather) We now believe all issues have been addressed (e.g. write access to GLAST_EXT area, access to grits.conf file) and all concerned have adequate privilege to do what they need to do.

  • TMine external (Joanne) Builds exist and are now installed for Linux and Windows. [Update: CMT requirements file for the new interface package had some bugs, preventing use from GlastRelease, but, as of late Tuesday afternoon, they're fixed and all is well.]

  • Fred (Heather) Tony, the new hire, Tracy, Leon and Heather will be having a meeting soon about Fred.

  • Skipping events (Heather) is working on Warren's request. It's close to done.

  • SCons meeting Back to the usual time: this week Wednesday at 10:30.

previous

minutes index

next

J. Bogart, Last Modified: