[Halld-offline] Minutes, GlueX Offline Meeting, Feb. 9 '11

[Mark M. Ito]

Find the minutes at 
http://www.jlab.org/Hall-D/software/wiki/index.php/GlueX_Offline_Meeting%2C_February_9%2C_2011#Minutes 
and below as text.

   ==

GlueX Offline Meeting, February 9, 2011

Minutes

    Present:
      * CMU: Curtis Meyer
      * IU: Ryan Mitchell, Kei Moriya
      * JLab: Hovanes Egiyan, Mark Ito (chair), David Lawrence, Yi Qiang,
        Elton Smith, Simon Taylor, Elliott Wolin, Beni Zihlmann


Announcements

     1. New JANA release. David is about to release a new version of JANA.
        It will include new features and bug fixes accumulated since the
        last release. He will give us more details at the next meeting.
     2. New sim-recon release: sim-recon-2011-02-02. Mark is preparing
        another release. The last one was at the end of November, so we are
        due.


Review of minutes from the last meeting

    We went over [38]minutes from the January 26 meeting.

    David asked the Regina folks about the status of work on the BCAL
    reconstruction code. Both Andrei Semenov and Zisis Papandreou
    responded. Andrei and Irina Semenova are working on it. Their first
    code is to understand the present algorithm. Eugene Chudakov has
    commented that if there is a problem with split-off, we need to know
    about it sooner rather than later and Elton's understanding is that the
    current focus is on deciding the summing scheme in the BCAL rather than
    in suppressing extraneous showers. David will follow up on the
    split-off issue.

Collaboration Meeting Review

    We discussed an item that came up at last week's collaboration meeting:
    the IT Review and our computing estimates. Simon has done some work
    estimating the time it takes to reconstruct bggen events with our
    current code. Mark is working on the estimates. He will circulate a
    version soon. There will be an IT Steering committee meeting Thursday
    to outline the topics to be addressed at the the review. Elton pointed
    out that we need a note on the estimate, explaining where the numbers
    come from and what assumptions were made. In particular, we need to
    revisit the issue of storing Monte Carlo generated data. If that data
    volume is comparable to (or exceeds) that of raw data then our
    computing model would have to be changed significantly.

Reconstruction Parameters vs. Calibration Constants

    We had a discussion of how to manage reconstruction parameters versus
    calibration constants. An example of the former (parameters) would be
    the minimum number of hits in a charged track, an example of the latter
    (constants) would be the minimum drift time for those hits. Some
    parameters are currently stored in the calibration files (SVN:
    trunk/calib) and thus will migrate to the calibration database. Mark
    thought that they should be managed as configuration parameters to
    JANA, since different parameters might be used for different analysis
    of the same data and calibration constants had more of the character of
    "best for all occasions". Others thought that putting them in the
    database had the advantage of "everything in one place".

    We did not come to a conclusion, but David agreed to give a talk on the
    JANA configuration scheme at a future meeting so we could have a more
    informed discussion.

Compiler Warnings

    Beni discussed [42]his recent email pointing out that sim-recon gives
    warnings with the latest version of the GNU compilers. He has given
    Richard Jones an account on his machine so that Richard can have a look
    at the HDGRANT-related warnings. There are some warnings from the BCAL
    that should be looked at as well.

Using Profiling Tools to Guide Optimization

    Dmitry walked us through a recent exercise he went through to speed up
    the CCDB code. He used profiling tools in two contexts: the C++ code
    and the MySQL server. Profiling the C++ code attemting to read a large
    volume of constants revealed that the bulk of the time was being taken
    in database queries. Subsequently, the MySQL profiling tool allowed him
    to identify an appropriate index to create in the database. He also
    implemented some pre-stored queries. This resulted in almost a factor
    of seven improvement in performance. He noted that using the tools
    allowed him to focus on improvements that really made a difference, not
    just on those that may sound good. See [44]his slides for details.

Timing Resolution: Kei

    Kei presented work on looking at timing resolution in the BCAL and TOF
    and how it affects particle identification. He found that using the
    "T0" that comes from tracking as a starting time degrades performance;
    one does better by leaving it out altogether. See [46]his slides for
    details. T0 is a start counter time; the true starting time should come
    from picking out the correct RF bucket of the electron beam. David
    suggested that perhaps we should add a time to the DVertex class that
    could serve as a place-holder for this quantity. It could be set to
    zero for now and filled in with the appropriate value as that analysis
    code becomes available.

Action Item Review

    We skipped a detailed review. Doesn't mean there aren't some items out
    there.

New Action Items

     1. Circulate a computing resources estimate to the collaboration. ->
        Mark
     2. Give a talk on the JANA configuration scheme. -> David
     3. Deal with warnings in sim-recon from the latest GNU compilers. ->
        Beni

    Retrieved from
    
"http://www.jlab.org/Hall-D/software/wiki/index.php/GlueX_Offline_Meeting%2C_February_9%2C_2011"

References

   38. 
http://www.jlab.org/Hall-D/software/wiki/index.php/GlueX_Offline_Meeting%2C_January_26%2C_2011#Minutes
   42. 
https://mailman.jlab.org/pipermail/halld-offline/2011-February/000491.html
   44. 
https://halldweb1.jlab.org/talks/2011-1Q/ccdb_profile_optimization.pptx
   46. 
https://halldweb1.jlab.org/talks/2011-1Q/offline-software.Kei-Moriya.2011-02-09.pdf