Core Minutes 5/9/2006  

Present:    Joanne Bogart, Toby Burnett, Rob Cameron, David Chamont, Jim Chiang, Richard Dubois, Warren Focke, Berrie Giebels, Tom Glanzman, Navid Golpayegani, Heather Kelly, David Landriu, Bryson Lee, Francesco Longo, Michael Kuss, Igor Pavlin, James Peachey, Leon Rochester, Tom Stephens, Tracy Usher

• RM enhancement proposal deferred: (Toby) There was some email discussion of a proposal made by Toby and Jim to provide support for testing and exercising new packages without going so far as to add them to LATEST. It is tabled for now since we don't have the resources to implement it.
• Vectors and aliases: (Toby, Heather) Bill's Insightful Miner interface can't handle vectors within the summary ntuple; one way or another, they need to be flattened for his consumption. Toby suggested creating a separate redundant Branch with three leaves, one for each component of the vector. Heather believes we can handle this with TTree::SetAlias. Either way, this has to be done before the final write during a Gleam job since we desire to have these two modes of access for the same data stored in the tuple. Bill's code may need some modificiation to ignore vectors (and just use the individual component name aliases).
• MRStudio: Riccardo didn't get a chance to make a report at the meeting today, but you can read about current status. MRStudio is in CVS now. No gui yet, but what is there works on both Linux and Windows..still testing, of course.
• Windows builds: (Richard) When we increased the number of allowed concurrent jobs from 3 to 6, individual builds took significantly longer. Now that we have set it back to 3, build times have returned to previous (better, but not acceptable) levels. Meanwhile, Richard is continuing to discuss this and broader issues of Windows support with SCS. Richard Mount ws concerned about the poor disk performance. With regard to overall support: a new Windows guy or gal will be joining SCS soon and will be located in Central Lab, which should help. Conceivably we'll have to resort to maintaining our own cluster if SCS can't do better.
• LAT hand-off support: (Richard) We'd like to get two more software upgrades in: better modeling of blanket, and improved flux model. (Heather) Between them, Heather (blanket material) and Joanne (top blanket with crown) have made the necessary changes to xmlGeoDbs HEAD; need to verify they're ok. Additionally, before this can be released Heather needs to finish upgrades to AcdDigi needed to handle the two-piece "bent" tiles. (Toby) Flux changes are part-way done. One of the sources — secondary protons — has not been checked in yet. (Richard) Once this is all done, we'll do a 200 million event run. If this is ok, larger all-gamma and background runs will follow to redo background analysis.
• Beamtest support: (Francesco) Beamtest has been successfully built against GlastRelease v9r2. Will check to see if problems with old runs are now fixed. Will probably re-run electrons; unlike the old, the new build has the correct geometry.

  (Heather) is working on an online LDF reader that can speak to Gleam, allowing us to run Recon on the fly or use FRED as an event display. So far she's been able to get LATTE code (obtained from Ric Claus) to run; next she has to convince it to spit out LDF data — in the right direction.

• Pfiles problem: (Richard) When running concurrent ScienceTools jobs with parameters specified on the command line, each one is looking for a parameter file with the same name, but they should, in general be different. Apparently one way to deal with this it to make as many directories as there are concurrent jobs, which in practice can be as many as 1000. Is there a better technique? (James) He is in the process of rewriting pil and may be able to address this. But, if the parameters are all specified on the command line, there shouldn't be a real conflict in any case.

previous

minutes index

next

J. Bogart, Last Modified: 01-Jun-2010 15:48:10 -0700