Present: Ursula Berthon, Joanne Bogart, Toby Burnett, Xin Chen, James Chiang, Richard Dubois, Riccardo Giannitrapani, Berrie Giebels, Navid Golpayegani, Traudl Hansl-Kozanecka, Heather Kelly, Michael Kuss and/or Johann Cohen-Tanugi, Dirk Petry, Sean Robinson, Leon Rochester, Alex Schlessinger, Tom Stephens, Tracy Usher
GlastRelease: (Toby) We are about to make a, perhaps the last, pre-DC1 release of GlastRelease. Since the previous release 24 packages have been updated! One of these is Event, which means we need a new release to develop against. Major work which will go into this release includes: full support for ACD ribbons (see below), Leon's ToT work, barely-dry celestial geometry work by Sean, late version of Insightful Miner variables, and a fix to G4Generator.
merit contains a clone of FT1 information. There was a problem that, for events without tracks, the FT1 output was nonsensical and there was no easy way to detect this case. Toby's initial proposal to just not write these events at all was met with disfavor, so instead he has added a new implementation of the INtupleWriterSvc, RootTupleSvc, which provides additional flexibility in what gets written. In particular, one may decide on a per-event basis whether to write out the FT1 TTree, without discarding all merit output.
In the process of other work, Toby happened upon several events in which TkrRecon gets hung up. He's working with Tracy on this.
ACD ribbons: (Heather) Everything necessary for sensitive ACD ribbons is now in. No problems have turned up yet, but more testing and examination is necessary, and perhaps another round of discussion on what to do about active distance/DOCA is in order. Current implementation is called Active Distance although in some respects it behaves more like a DOCA.
Optimized Linux build: (Richard, Toby, Leon, Matt,...) Navid got it all working while he was at SLAC last week and made a run of 100,000 gammas for comparison with non-optimized. In detail, the runs are not at all the same since random number sequences quickly go their separate ways. For most system tests the results are statistically equivalent, but there are a couple possible exceptions, most notably TKRTRKNHITS, for which the optimized run has a statistically significant larger ratio of tracks with an even number of hits to those with an odd number. Navid had to make some changes to code to get it to build optimized, but, given the nature of the changes and the subset of code changed, these changes are very unlikely to have had any effect. In fact there is no reason to believe the unoptimized run is somehow "better" than the optimized (or the other way around): either one, viewed in isolation, looks plausible.
The optimized run was 2.5 times faster than unoptimized so we should make it more generally available.This will take some procedural improvements to the build process.
Release Manager: (Alex) He has been tied up with Opus work, but a new version of the Release Manager with several improvements is close to ready to go; he hopes to get to it later this week. These include
Windows build. While at SLAC, Navid got Soap Service for launching builds working on the Windows machine in Alex's office, so in shape for use by the Release Manager.
Redhat 9 gcc 3.2 build
Ability to monitor builds in progress from the Web.
DC1 data handling at SLAC (Richard) He presented current thinking on what sort of access to (n-tuple) data at SLAC we might provide, including a sketch of web interfaces for data access and for monitoring the OPUS pipeline. The user will be able to choose among FITS, n-tuple or Root tree output. The web form for data access provides 3 alternative query styles: fetch data using standard SSC-style cuts, fetch using cuts on merit ntuple columns, or fetch by (Run id, Event id). The last two are well in hand; the SSC-style access is on hold until more information about the SSC web interface is available.
ROOT converter: (Ursula) Progress is slow, but there is progress. She is working on a ROOT server which will encapsulate ROOT functionality not specific to the Conversion Service.
Údine visit: (Joanne) This worked out well in all respects. We (Riccardo, Marco, Joanne) got two projects going: a diagnostic detector display using FRED, and an XML + gui approach to the calibration metadata database. The first is already up and running with most of the design functionality; implementation is just beginning for the second. See this trip report for a full description of these projects.
Fred: (Riccardo) He plans to look into loading heprep files into Fred via http, thus obviating the need for Corba for some applications. He is also working on a build of Corba for Redhat 7.2/gcc 2.95.3.
Thursday meeting: ..is a maybe at this point due to a conflict with a meeting of the GLAST Users Committee at Goddard. An announcement one way or the other will be made to Softlist.
previous | minutes index | next |
J. Bogart Last Modified: 04-Aug-2004 15:39:46 -0700