Core Minutes 10 January 2006

Core Minutes 1/10/2006

Present: Joanne Bogart, Philippe Bruel, Eric Charles, Jim Chiang, Seth Digel, Richard Dubois, Dan Flath, Berrie Giebels, Riccardo Giannitrapani, Tom Glanzman, Navid Golpayegani, Heather Kelly, Michael Kuss, David Landriu, Chuck Patterson, Igor Pavlin, James Peachey, Leon Rochester, Andrew Strong, Tracy Usher

DC2:
- Sky model (Richard) Target date to generate sky model is the 15th; this work is in its final stages. AGNs are done. There is an extra constraint for GRBs: they can't span multiple batch jobs.
- Interleaving (Tom G., reporting on work done or to be done by Dan, Toby and himself)
  - The plan Generate about 5 billion* distinct background events starting at launch+10 (about enough for 5 days) and some number of secret signal events, spanning 30 to 56 days. Software will shuffle these together, reusing and modifying background events for the longer time span. Output will be the massive (~500 Gbyte) merit ntuple for the full span. It will be followed by a post-processing stage to get rid of any trace of MC information and do some additional clean-up.
    *Note: The original target was 1 billion background events, but it appears this will not be enough to avoid artifacts when the events are recycled. Generation of backgrounds is going smoothly and quickly, so we'll just keep making them, probably ending up with about 5 billion.
  - Aspects of implementation The new Interleave package takes care of the recycling and shuffling. Background events are binned by geomagnetic latitude since rate is a function of this variable (there are probably other variables on which rate depends, but they are believed to have at most a small influence, so are ignored here). When Interleave needs one, it will be randomly selected from the appropriate bin and read in, modified as needed, and output.
  - Status All major pieces are done or nearly so. Here is the remaining to-do list:
    1. Redo background geomagnetic latitude rate analysis applying new trigger cut
    2. Fix a problem with event numbering
    3. Tune code to run reliably and in a finite amount of time
    4. Test thoroughly
    5. Demonstrate basic system (yesterday's milestone — not quite ready yet, hopefully today)
    6. Apply to full DC2 sample next week.
  [Thanks to Tom for supplying me with his notes.]
Classification trees (Tracy) He's had to play catch-up with Bill's additions. There is just one more item, affecting only a small subset of events, to be handled. (Richard) Can Bill's 3 classes (see last page of his presentation) be expressed as TCuts? (Tracy) Yes, with some awkwardness.
10% filter (Richard) We're aiming for something which, reliably and quickly, can mark 90% of incoming events as garbage and eliminate them from further consideration or processing. Leon is starting to look into this.

EM (Heather) Working on a new tag, to include

New CAL code, already in HEAD
Updates to ACD from Eric (additions to AcdRecon event classes, requiring related changes to other packages)
xmlGeoDbs tag with correct overlap direction for ACD top tiles
Perhaps use of Double32. Was ok in test TkrRecon, reconRootData.

She would like the EM release to be based on a newer GR tag, but recent GR tags have had some test program failures.

ACD Geometry upgrade (Joanne) A significant discrepancy between the ACD geometry model and the real thing — lack of bent tiles — has been addressed in the latest xmlGeoDbs tag. Many thanks to Mr FRED for making it possible to get this done quickly and colorfully.

ACD thresholds (Heather) would like to fix the logic in the Onboard Trigger to use the correct value (0.3 MIP) as the condition for setting ACDLow. She plans to put the fix in HEAD and perhaps EM, but it won't go into DC2. Steve has asked for a presentation of the details in 2 weeks, backed up with a system test.

Data server issues: (Igor) Several problems have been identified which should all be fixed by some time next week. Let him know if any of these is particularly bothersome; he will bump up its priority accordingly.

Core week Jan 17-20: (Richard) Topics will include

Possible new versions of external libraries. Heather will have recommendations on Gaudi and ROOT. Francesco will speak about G4 and perhaps CLHEP.
Compiler support, especially for Linux. Inevitably, not all users will have the first-choice (used by SLAC batch farm) compiler.
Mac support?
MRvcmt update from Riccardo (remote)

Toby, Heather, Navid, Julie and James Peachey will be among those coming to SLAC.

minutes index

J. Bogart, Last Modified: 01-Jun-2010 15:48:13 -0700