GlueX Offline Meeting, January 25, 2012

From GlueXWiki
Jump to: navigation, search

GlueX Offline Software Meeting
Wednesday, January 25, 2012
1:30 pm EST
JLab: CEBAF Center F326


  1. Announcements
    1. Memory leak in the reconstruction fixed: Simon
    2. Nightly builds fixed: David
  2. Review of minutes from the last meeting: all
  3. Reconstruction sub-group reports
    1. Calorimeters: all
      1. BCAL Update:Will
    2. Tracking: Recent changes up to 1/25/12 Simon, Paul: Slow Pions
    1. PID: Paul M.
  1. Offline Software Review Planning: David
  2. Collaboration Meeting Offline Session.
    1. 8:30 Session IIIa (120) --- Offline Reconstruction - (Organizer: M. Ito) Chair:
      1. 8:30 (30) ---
      2. 9:00 (30) ---
      3. 9:30 (30) ---
      4. 10:00 (30) ---
    2. 10:30 (30) Coffee
  3. HDDM's C++ API: Richard
  4. Action Item Review
  5. Review of recent repository activity: all

Communication Information

Video Conferencing


Talks can be deposited in the directory /group/halld/www/halldweb/html/talks/2012-1Q on the JLab CUE. This directory is accessible from the web at .



  • CMU: Will Levine, Paul Mattione
  • FSU: Nathan Sparks
  • IU: Ryan Mitchel, Kei Moriya
  • JLab: Eugene Chudakov, Mark Ito (chair), David Lawrence, John Leckey, Sascha Somov, Simon Taylor, Ivan Tolstukhin, Elliott Wolin, Beni Zihlmann
  • Regina: Andrei Semenov
  • UConn: Richard Jones


  • Simon announced that the memory explosion (a.k.a. memory leak problem) should not occur anymore. He implemented changes suggested by Richard. Other related issues with track finding raised in earlier meetings are still being addressed.
  • David announced that the nightly builds are completing successfully now. There was an issue with interaction of the make system and the new C++ API for HDDM. Multiple files differing only in their suffix cause problems. Base names were changed to fix the problem. Also certain C++ compilations require a search of a new directory and now this is done.

Review of minutes from the last meeting

We reviewed minutes of the December 14th meeting.

  • The 1-2-3-4 scheme for BCAL read-out the default. Thanks to Will and David.
  • David made a clarification. The EVIO input to JANA code has been made into special version to support the FCAL beam test in Hall B, in particular support for Flash-250 data has been added. This capability will serve as a base for future flash support.
  • David remarked that the features of the HDF5-based DST data scheme presented by Gagik Gavalian might be useful in addressing the issue long-term preservation of data. Richard pointed out that the design of HDDM was motivated by this concern in the first place. Mark reminded us that Gagik's scheme can naturally store auxilliary information, such as calibration constants, so that systems which serve that information do not have to be preserved. Mark also thought that this scheme is something to keep in mind, but is probably not a front-burner issue for us now.
  • Although Dmitry Romanov installed the CCDB system on the CUE before leaving the USA, none of us have played with it much since. We need to organize active testing of the system.
  • Richard asked about deployment of the calibration database on off-site institutions. In particular, he hoped that the use case where calibration sets are being generated and tested repeatedly, as is often the case, was adequately support. David noted that using local files for constants was in the orignal JANA scheme. Mark told us that local files would be supported for the CCDB as well, but the implementation is not there yet.

Reconstruction sub-group reports


Will reported on his recent work in understanding spurious clusters in the BCAL reconstruction. He believes that many of these come from random dark noise hits pulling the time of legitimate photon shower hits such that they are identified as separate clusters. See his wiki page for details. We had an extended discussion about how to deal with this; overlapping issues include:

  • Should the dark hits be introduced in HDGeant or in mcsmear?
  • Should HDGeant generate waveform information to allow realistic inclusion of dark noise?
  • Do we need to revisit the method of dark noise injection in mcsmear to add realism?

David agreed to summarize the situation with respect to dark noise simulation and present it at a future meeting.


  • Simon reviewed recent changes to the tracking code. See his slides for details. They addressed:
    • Recent changes to tracking code
    • Charge from fit
    • Recent changes to Kalman filter
    • Effects of hit pruning in filter
  • Paul presented studies of π+'s with momentum of 150 MeV/c, at 90° in polar angle. These generally curl around back to the target and may do so multiple times. See his wiki page for details of his study. The track multiplicity distributions for the various stages of tracking vary quite a bit from stage to stage and have very wide distributions considering that only a single track was generated. Mark remarked that Paul had chosen a difficult corner of phase space that would likely require special code to deal with, code that is nowhere present in the repository at present. He also mentioned that the identification of these "curlers" was more easily done with global information; depending on local hit-to-hit, step-to-step information could be unreliable. Paul said that he has some ideas on how to address these tracks and that he would pursue them.


Richard started leading us through his notes on the new API. The code was recently checked into the repository.

Unfortunately we had to end the presentation early due to over-running our reserved time in the meeting room at JLab. The Chair promised Richard another shot at a future meeting; in the mean time those interested can look at his notes. Luckily, they are done in a pedagogical style.

New Action Items

  1. Organize active testing of the CCDB. -> Mark
  2. Present a summary of dark noise generation in the BCAL. -> David