Core Minutes 11/16/2010

• Pass6 Reprocessing: (Tom G.) All data through last Thursday has been reprocessed and sent to FSSC, where it has been ingested. The remaining step is the cut-over from the "old server" to the "new server". Once that happens (probably tomorrow), visitors to FSSC web will have access to the new data.

  GCR work is also done.

• Pass8: (Tracy) reports via email:

  GR v18r8p0 is big news for the week. We need to fix the recon output issue for the MST params, it seems the Pisans believe they need this to do some studies. Otherwise, I think we are ok for now.

  Am making lots of progress on tree based tracking... have developed a scheme for extracting more than one track from any given tree, have refined several of the key pieces of the code to great improve the returned trees, etc. And have been attacking timing issues found on linux. Its getting close!

• Disk news: (Richard) Our final order for thors, 5 of them, has been placed with Oracle. We'll find out delivery date soon. 3-4 week is typical.

  Over the week-end there was another user disk incident. We plan to make over one of the older thors into the user disk; it should be better behaved.

• ScienceTools: (Jim) There are no new tags to report on but there still two ongoing issues with regard to the exposure calculations in the binned likelihood analysis:

  1. Seth performed some tests on the gtexpcube output and compared the maps that are produced when phi-dependence is considered and when the phi-averaged effective area is used explicitly. For sufficiently long observations (Seth considered 1 day), the expectation would be that the differences in exposure on a per pixel basis for an all-sky map would be distributed symmetrically about zero. For the gtexpcube output, the map computed with phi-dependence is systematically larger everywhere than the map computed using the phi-averaged effective area. The effect is around 5-10%. Performing the same test with the binned exposure calculation that was in gtsrcmaps results in delta-exposures that are symmetrically distributed around zero, as expected.
  2. gtexpcube adds LONPOLE and LATPOLE keywords to the image headers of the exposure maps it produces. These keywords give the coordinates of the celestial pole. The gtexpcube code (via the astro::SkyProj class) sets these keywords incorrectly. For certain map projections, AIT in particular, this results in incorrect interpretation of the pixel coordinates and therefore incorrect results when performing a binned analysis. The STG projection seems not to be affected by the incorrect keywords. Unfortunately, the default projection in gtbin (that gets propagated to downstream tools) is AIT.

  Johann has volunteered to try to fix the gtexpcube code, but it may simply be easier to extract the binned exposure calculation from gtsrcmaps and make a new tool.

  [Thanks to Jim for the above contribution. ed]

• ADASS Notes: (Richard) The World-Wide Telescope guy was there. Long ago we had given them our 24-hour sky map. I talked to him about providing something more recent.

  There will be a more comprehensive recap at this week's SCA meeting (10 AM on Friday in Ballam or via EVO).

• Documentation: (Chuck) sends the following:

  I've finished incorporating Elizabeth Ferrara's changes to the unbinned and binned likelihood tutorials, making them SLAC-centric in the process, and now need to actually perform both on SLAC Central Linux. These changes also include the addition of several new support pages, which provide more context for performing the likelihood analyses. All pages can then be posted to the web as soon as Jim and Seth give the go ahead.

• Core talk:

  • Odds and ends (Anders) We're looking into a possible memory problem with acd digis.

    (Heather) is working on the ability to skip events by GEM id.

  • ROOT (Heather) L1, Pass8 and ScienceTools are all at ROOT 5.26 but Pass7 is still at 5.20. Would like to advance it as well. (Tom G.) would also since the newer version works better with xroot.

  • Windows boxes (Heather) glast06 may be dead or dying. It's been problematic for some time anyway (e.g., can't log onto it using the glast account). Meanwhile the new box is approaching ready.

  • Snow Leopard (Tom S.) No news. Can make builds by hand, but jobs don't trigger in lsf.

  • New Gaudi, New Windows OS, No luck (Joanne) is stuck at GlastSvc. She can build it properly and run test_GlastSvc with new gaudi and old (vc 7.1, VS 2003) compiler, but not with vc 9/VS 2008. Error messages are of the unhelpful crash and burn variety. As part of this exercise she made a native VS 2008 build of Xerces (we had been using VS 2005 binaries, the most recent distribution available for Xerces 2.8.0) but it hasn't made any difference.

  • SCons meeting (Heather) ..will take place on Wednesday at 10:30, as usual.

previous

minutes index

next

J. Bogart, Last Modified: