GlueX Offline Meeting, April 30, 2014

From GlueXWiki
Jump to: navigation, search

GlueX Offline Software Meeting
Wednesday, April 30, 2014
1:30 pm EDT
JLab: CEBAF Center F326/327


  1. Announcements
    1. Dmitry's visit completed
    2. Nightly Build running with Scons
      • b1pi and single-track tests working
  2. Review of minutes from the last meeting: all
  3. Data Challenge 2
    1. Event Tally Board
    2. Site Status Updates as appropriate
      1. JLab
      2. OSG
      3. CMU: All of the CMU-generated data has been uploaded to NU. File #0's were renamed to the next available (last) file number.
    3. DC-3 planning
      • Update on raw data generation and reconstruction?
      • Electromagnetic background 2.0?
  4. Event Skimming
    1. PID & Skim Cuts Update (Paul)
    2. Update on EventStore Indices (Sean)
  5. Tagger Reconstruction
  6. Smooth exit from reconstruction errors
  7. Git, Github, Halls A, B, C
  8. C++-11
  9. AOT

Communication Information

  1. To join via Polycom room system go to the IP Address: ( and enter the meeting ID: 292649009.
  2. To join via a Web Browser, go to the page
  3. To join via phone, use one of the following numbers and the Conference ID: 292649009
    • US or Canada: +1 408 740 7256 or
    • US or Canada: +1 888 240 2560


Talks can be deposited in the directory /group/halld/www/halldweb/html/talks/2014-2Q on the JLab CUE. This directory is accessible from the web at .



  • CMU: Will Levine, Paul Mattione, Curtis Meyer, Reinhard Schumacher
  • FSU: Aristeidis Tsaris
  • IU: Matt Shepherd
  • JLab: Mark Ito (chair), David Lawrence, Mike Staib, Simon Taylor, Beni Zihlmann
  • NU: Sean Dobbs


  1. Dmitry's visit is over. He is still reachable via email and is happy to answer questions.
  2. The nightly build running with using SCons for the build. David has a wiki page describing the system. Note that the environment for using these builds is different than that which worked for the BMS make system. Both the b1pi and the single-track tests are now working with the new scheme. Mark will use this for the next tagged release.
  3. Beni has successfully built the GlueX software suite on an Ubuntu 14.04 box.
  4. David has successfully built JANA on a Raspberry Pi board.

Review of minutes from the last meeting

We reviewed the minutes from the April 16 meeting.

  • Paul clarified that currently, kinematic fits are done using the thrown value of the beam photon energy.
  • Mark clarified that at present the online group is using the one-and-only GlueX Subversion repository.
  • David gave the wiki software update a go. We will wait for an upgrade of the web server before trying it again.

Data Challenge

Even Tally Board

We looked at the Event Tally Board. We are at 6.6 billion events. Running continues on the OSG.

Site Status Updates

  • At JLab:
    • Doing clean-up of missing jobs.
    • Archiving hd_root files to tape. This was not done during the running of the jobs.
    • Mark has been using the SRM successfully. There remains a problem with some of the SRM servers at UConn not being white-listed by the JLab firewall in the initial network set-up some years ago. We also realized that this means that NU is not white-listed at all.
  • All of the CMU-generated data has been uploaded to NU.
  • On the OSG:
    • Over the past few days there was a problem with SRM transfers not succeeding on the OSG. This was due to incompatible SRM client on remote nodes. Now Richard including our SRM client with the remote jobs to insure compatibility, and will start up submitting jobs again.
    • The OSG will start mixing in some 9002 and 9003 running.
    • Richard will continue until we get a solid stretch of full-capacity running. After that he will shut it down at his discretion.

DC-3 Planning

Mark mentioned two topics that we need to address before the next data challenge:

Raw data generation and reconstruction capability

David will give a report on status at the next meeting. There is still some work to do on the calibration layer, e. g., converting raw ADC values to energy. Sean has been looking at this recently.

Generating electromagnetic background more efficiently

David had started work on a scheme for mixing in pure electromagnetic background events in with events of interest. There are several questions:

  1. Is this the right scheme? (Likely yes.)
  2. How big does the EM background library have to be?
  3. How much realism is needed in combining background hits with physics event hits?
  4. Do we wait for Geant4 before working on this in earnest?

We agreed to have David present the status of his work at a future meeting and use that as a starting point for discussion.

Event Skimming

PID & Skim Cuts Update

Paul presented updated results on his skim studies focusing on \gamma p\to \pi ^{+}\pi ^{+}\pi ^{-}(n). See his wiki page for details. An earlier stage of this work was presented at the last Physics Meeting. Some items from the presentation and discussion:

  • The PID combined PID confidence-level cut seems to cost a lot in terms of efficiency. This cut is on the combined CL of time-of-flight measurements for all tracks. These measurements may come from the TOF, the Start Counter, the BCAL, FCAL, or the tracking chambers.
  • Some of these measurements look biased and/or non-Gaussian, some with puzzling features in the distributions.
  • Matt suggested looking a lower level quantities to do the skims, such as charged track multiplicity or total energy.
  • Paul told us that the number of tracks was generally pretty high, even for these 3pi events and for this study the requirement was a minimum of 3 charged particles (++-) with no upper bound. And the neutral shower count can be high as well. That makes event classification by topology difficult. We agreed that the situation would be a lot better if this could be done more reliably and that it was worth looking into.

BCAL Timing

Will has studied the timing distributions used in the PID cut for the BCAL. See his slides for details.

He sees a long tail going to late times, out to several nanoseconds, for pions but not for photons. He tried using a truncated mean of hits contributing to the cluster, using only the earliest 50% by time. This did cut down on the tail, but introduces a bias to early times as might be expected. He looked at correlations of the late tail with other cluster properties, but did find any strong ones.

Matt suggested trying the study with muons to see what happens.

Update on EventStore Indices

Sean presented an update on his EventStore development. See his slides for details.

During Sean's presentation, Mark finally came to the understanding that the framework is essentially a system for managing event lists based on a common complete or nearly-complete set of events contained in a well-defined set of files. Various skims are then realized as event lists on that data set. The value added is mainly in that users can easily access multiple skims performed by the collaboration or individuals since there is only one set of data files. The event analyzer, JANA in our case, need only be instrumented to access and select events based on the lists. The data itself needs no special hooks or modification. In addition, there can be multiple target data sets, either from different run periods or or from higher level processing passes on a single set. These choices are available and documented within the framework. At least that is his understanding.

Other Agenda Items

We deferred the following to a future meeting.

  • Tagger Reconstruction
  • Smooth exit from reconstruction errors
  • Git, Github, Halls A, B, C
  • C++-11