Present: Joanne Bogart, Toby Burnett, Xin Chen, Richard Dubois, Heather Kelly, Michael Kuss, Sean Robinson, Leon Rochester, Alex Schlessinger, Tracy Usher, Karl Young
ACD performance review: Preceded Core Meeting proper. See Heather's presentation here.
Gleam status: As of tonight Gleam requirements file should reference package versions we expect to be in v3. We'll have a chance to examine the results of the nightly build tomorrow before actually making the release. "Making a release" includes making binaries for both platforms publicly accessible. Once the release is made we'll need to make sure we understand how to use CVS branch tags to keep straight patches to the release and longer-term development.
Issues discussed from Toby's detailed report include
Composite sources. (Toby) Very recently this was a significant hole, but new work has probably taken care of it.
TkrRecon cpu/time hog. Tracy has narrowed the problem down to the propagator and thinks he can fix it soon. Meanwhile, Leon discovered a bug which, one would think, one or the other compiler would have disallowed: in statements of the form
const i = 1.1; // no explicit type!
a default type of int is used. Non-integer constants are truncated. [g++ complains, but, with our current set of compiler flags, only issues a warning. ed.]
Bari digi alg is still not quite ready to go. (Leon) This is at least partly because enhancements to the simple digi alg need to be propagated to Bari (or the whole collection of code could be made more modular).
(Toby) Would like to see demonstration that the new use of random seeds really does give us reproducibility as advertised. (Karl) All necessary pieces exist, just need to be properly written up.
Gleam User's Guide: (Heather) A couple of the heretofore-missing significant chunks have been added: a section on using the gui (Joanne) and automated extraction of job options information (Karl). [However, since then Toby has improved the gui. The documentation hasn't yet been upgraded to reflect this. ed.] Still missing: userAlg description (Eduardo?), description of geometry (Riccardo).
Conversion to g++3 and Gaudi 11: (Michael) You can read about the whole adventure to date. At this point, other than some complaints about the use of std::hash_map in the xmlUtil package, remaining problems most likely are coming from Gaudi. (Joanne) Use of std::hash_map is just an optimization for Linux which is making no significant difference in performance. On Windows std::map is used since hash_map is not supported. It would be best to just use std::map on both platforms to avoid obscuring the more serious problems [done as of this afternoon]. Toby spoke for everybody when he thanked Michael and Johann for all the work they've done. We're practically there!
previous | minutes index | next |
J. Bogart Last Modified: 01-Jun-2010 15:46:50 -0700