Core Minutes 10/15/2013

• ScienceTools: (Jim) Dave Davis requested a build for testing the new irf handling, so I made ST-HEAD-1-1031, which Joe has downloaded and built at FSSC.

  • Joe checked in and tagged changes to rspgen.
  • I checked in and tagged chages to pyExposure.
  • Jeremy did some testing and revealed a couple of issues that I will address this week.

  In other news,

  • NADIR option has been added to astro and flux to support EarthPhenom.
  • A group met last Tuesday to discuss Pass 8 event type changes and decided to add EVENT_TYPE column to FT1 to support. Matthew will prepare a confluence page describing the kinds of event class and event type combinations and will provide example irfs to use for development.

  See also Science Tools Development Notes.

• Pass7 comparison: (Tracy) sees different results in calculations of a determinant in the moments analysis depending exactly which compilation flags (-g, -O2, both, etc.) are used — this on rhel5 but using gcc34, the native rhel4 compiler. The sign of the result determines which branch in the code is followed, and that can change. This is likely connected with known behavior, especially of optimized code, on some machines where floating point registers may carry more than the specified number of bits (for float or double). See gnu documentation on options controlling optimization, especially -float-store and -fexcess-precision, about 2/3 of the way down. By making some variables class variables (rather than temporary local) he can prevent them from being stored in registers. This eliminates differences in sign of the result.

  (Richard) Is the plan now to use MC simulations to see if this effect is neglibible? (Tracy) Yes. at the C&A meeting there was talk of using data as well as simulation runs using GR 17-35-24-lp60 to run comparisons and look for any differences. He believes that the number of events affected are tenths of a percent.

• Pass7 reprocessing: (Tom G.) Backfill is complete through Friday, Oct. 4. [Since the meeting there has been a request for another, which will go up to Oct. 15.]

• Pass8 reprocessing: (Tom G.) is halted pending minimal changes to use new ACD calibrations (which go into effect starting Sept. 9, 2012). We are at 78% complete, processing data from May, 2012. The pipeline has been drained and Tom ran 10 test jobs using Leon's new job options (using L1current flavor). They will look over log files to make sure there are no changes in the output. Then Tom will reprocess one run which had already been reprocessed with old jo. Finally, once reprocessing catches up to the new ACD calibrations, we must, again, carefully chec the log files.

• Hardware: (Richard) The new disk arrays have arrived and are installed. (Tom G.) has not heard from Wilko or SCCS but the disk array does respond to ping.

• Core talk:

  • GR tags (Heather to Jim) Should she incorporate the updates made for EarthPhenom — specifically, newer tags of astro and evtUtils — into GR? (Jim) Only if P8 is interested in EarthPhenom.

    (Heather) is unclear about what needs to go into jo (especially calibration.txt and readigi_runrecon.txt) for the planned GR P8 tag 20-09-02-rp02 which Tom needs for testing. She'll discuss with Leon offline.

  • Macs (Heather) We have been in touch with Yemi who notes that SCS wants to remove bldmac01 and bldmac02 (our Snow Leopard machines) ASAP. We're offering "end of the year".

    (Tom S.) It's slow going as he works through missing libraries which need to be installed on the Mac Pro. He will be contacting Yemi soon about installing SSH keys for glastrm.

  • Release diffs (Tom S.) No news.

previous

minutes index

next

J. Bogart, Last Modified: