Long discussion about recent allGamma and Background Monte Carlo running
experience, what we need to proceed with the first 100M background
sample, and where we wish to go in the near future before the second
100M background sample.
CURRENT STATUS
There have been several recent attempts to generate 2M allGamma
and 10M background (pdr) datasets. Some fraction of the runs
crash (segmentation violation). See, for example,
allGamma-GR-HEAD1.622 and backgndavgpdr-GR-HEAD1.622, of which
5% and 19% of the runs failed, respectively.
Julie reports failures in the recent system tests. She and Toby
to look for causes.
Heather and others have been working to fix bugs and add new
functionality. Toby reports fixes/improvements to CalRecon,
AnalysisNtuple, code to ease background source running from
environmental variables. Some of this work is in the most
recent new tag: GR-v7r0p1.
later today. 10M background events based on this code will also
be generated sometime today once the allGammas are certified.
Early look at the v7r0p1 indicates there are still some
problems, such as:
* seg faults (even without core dumps)
* memory leak(s)
* Heather reports a bug in CalValsTool::calculate
NEW CODE
There is also new code being considered for inclusion prior to
the 100M event background run, including ~8 ACD geometry
packages from
Joanne. But Joanne is out of town for another week... Experts
seemed to think that more testing was needed before this code
was used in
production, such as hermiticity and digi tests.
BACKGROUND MODEL
Toby continues to run his DC2 background model at U.Washington.
Richard made a quick attempt to produce these events at SLAC but
not
yet with success. There are some details with the job options
to be worked out. Richard and Tom have discussed and the latter
will work
with Toby to see if this can be commissioned soon. This first
test task is called CrTotalMix-GR-HEAD1.622 and is still
ongoing.
THE PLAN
* There is some urgency to get 100M background events produced
ASAP
* More work on v7r0p1 needed ==> new release for 100M run
* New ACD code is not yet ready for prime time, but should be
tested in parallel [will need a tag for this testing to proceed]
* Continue to use old "PDR" background model for the moment, but
pursue getting Toby's code to work at SLAC