Present: Joanne Bogart, Toby Burnett, Xin Chen, Johann Cohen-Tanugi, Richard Dubois, Berrie Giebels, Navid Golpayegani, Traudl Hansl-Kozanecka, Michael Kuss, Francesco Longo, Julie McEnery, Sean Robinson, Leon Rochester, Alex Schlessinger, Tracy Usher, Karl Young
Integrate filter code: (Richard) It's time to re-integrate JJ's on-board filter code into Gleam and make TDS classes to hold the results. Navid and David Wren will be looking into this.
System test run: (Karl) He expects to be able to do a run for v2r2 later today.
G4 multiple scattering investigation: (Johann, Francesco) They ran tests analogous to the G4 standalone test of multiple scattering, but in the context of Gleam. Bottom line is that behavior is reasonably consistent (within 20-24%) with Moliere theory prediction, although G4 does not use Moliere theory. See the full report in your choice of formats: Powerpoint original, postscript (largish file), or pdf (truncates some plot labels a bit). Both the G4 standalone test and the Gleam version belong in our System Tests.
For these and similar tests it would be useful to have older versions of G4 available for comparison. In fact the "pdr version" is available on SLAC Linux, but the only binaries were made with the old (incompatible) compiler. Alex will see about rebuilding it with the current compiler.
EM: (Leon) He was able to read and make sense of digis after fixing a bug or two. The data was made by injecting charge into the electronics. At this level most of the channels are live. On some planes the controllers are reversed (so that the 2nd of the two controllers is reading out the lower-numbered strips rather than the other way around); this is clearly in the hardware. There is a mislabeling of planes going on somewhere so that plane numbers are in the range 28 - 35 rather than 0 - 7. The best guess currently is that this problem is in the software decoding rather than in the hardware. Joanne will be familiarizing herself with the relevant code and with the electronics/geometry mapping in order to track this down.
ROOT, Go4: (Berrie) Previously there was a bug in Go4 so that we couldn't use it to look at our ROOT trees. The Go4 folk have now fixed it; we can run it successfully. René Brun warns that QT is about to start charging for Linux as well as Windows licences. So far this is unconfirmed.
Geometry numbers: (Richard) Last week's complaint than we have no reliable and consistent source for geometry numbers is having some results. System engineering folk will take responsibility for corralling subsystem people and getting them to sign off on the numbers we need (first approximation is the quantities in the "DimPrim" sections of the html constants table). We should have the results by mid-August.
Infrastructure improvements: (Navid)
GlastRelease summary page in progress
Waiting for SCS to bless our use of CGI
Add to system monitoring. We already check on disk space and some other things. Will add check that Web server, MySQL and Oracle are all up. Meanwhile we'll be looking into open source monitoring tools.
Valgrind experiences: Leon found and fixed one memory leak. The Valgrind output also uncovered an unitialized variable which was a true bug. Overall the experience was valuable.
Joanne looked into the main complaint Valgrind had about xmlUtil. It is not a true leak, but a case of allocation during initialization with no very natural time to give back the memory. Probably this could somehow be done differently to avoid the Valgrind complaints. Is it worth it? Depends on how much effort it would be.
Navid will automate the process of running Valgrind a bit further (so he doesn't have to do it a package at a time) and run it. We might then choose to run it for every new release. Navid also plans to make it possible for anyone to invoke it (for a specific package?) on demand.
Insightful Miner: Tracy and Leon each have a copy. Tracy bogs down, probably due to memory shortage, on his laptop. Bill has written S+ functions which ask to read in all the data at once, however this is not intrinsic to the way Insightful Miner operates so could perhaps be fixed.
There are a list of variables which need to be added to the analysis ntuple. A dictionary is created from the ntuple so that values may be looked up by name. Then the IM file has to be regenerated to reference the new variables.
previous | minutes index | next |
J. Bogart Last Modified: 04-Aug-2004 15:41:16 -0700