<html>
  <head>

    <meta http-equiv="content-type" content="text/html; charset=utf-8">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <p>Gluons,</p>
    <p>Please find the minutes below and at</p>
    <p> 
<a class="moz-txt-link-freetext" href="https://halldweb.jlab.org/wiki/index.php/GlueX_Offline_Meeting,_February_9,_2018">https://halldweb.jlab.org/wiki/index.php/GlueX_Offline_Meeting,_February_9,_2018</a>
      .</p>
    <p>  -- Mark</p>
    <p>_________________</p>
    <p>
    </p>
    <div id="globalWrapper">
      <div id="column-content">
        <div id="content" class="mw-body" role="main">
          <h2 id="firstHeading" class="firstHeading" lang="en"><span
              dir="auto">GlueX Offline Meeting, February 9, 2018</span><span
              class="mw-headline" id="Minutes">, Minutes</span></h2>
          <div id="bodyContent" class="mw-body-content">
            <div id="mw-content-text" dir="ltr" class="mw-content-ltr"
              lang="en">
              <p>Present:
              </p>
              <ul>
                <li> <b> CMU: </b> Curtis Meyer</li>
                <li> <b> FSU: </b> Sean Dobbs</li>
                <li> <b> JLab: </b> Alex Austregesilo, Thomas Britton,
                  Hovanes Egiyan, Mark Ito (chair), David Lawrence, Beni
                  Zihlmann</li>
                <li> <b> Yerevan: </b> Hrach Marukyan</li>
              </ul>
              <p>There is a <a rel="nofollow" class="external text"
                  href="https://bluejeans.com/s/oJ0JA/">recording of
                  this meeting</a> on the BlueJeans site. Use your JLab
                credentials to access it.
              </p>
              <h3><span class="mw-headline" id="Announcements">Announcements</span></h3>
              <ol>
                <li> <a rel="nofollow" class="external text"
href="https://mailman.jlab.org/pipermail/halld-offline/2018-February/003101.html">Reclamation
                    of halld-scratch volume set</a>. 19 of our tapes are
                  about to be erased.</li>
                <li> <b>New top-level directory: /mss/halld/detectors</b>.
                  A new tape directory for data related to specific
                  detectors, e. g., /mss/halld/detectors/DIRC.</li>
                <li> <a rel="nofollow" class="external text"
href="https://mailman.jlab.org/pipermail/halld-offline/2018-February/003106.html">New
                    sim-recon release: version 2.23.0</a>. This release
                  includes Simon's newly tuned parameters for track
                  matching to calorimeter clusters.</li>
                <li> <b>New simple email list: online_calibrations</b>.
                  Sean has created a <a
href="https://halldweb.jlab.org/wiki/index.php/Email_Lists#Simple_Email_Lists"
                    title="Email Lists">Simple Email List</a> for those
                  interested in keeping track of the latest calibration
                  results.</li>
                <li> <b>Launches</b>. Thomas caught us up.
                  <ul>
                    <li> There is a monitoring launch that will start
                      soon.</li>
                    <li> The reconstruction launch with recent software
                      on Spring 2016 data has started.</li>
                    <li> Several people are looking at the anomaly that
                      Beni has identified, visible in some TOF
                      monitoring plots (though the TOF is blameless).</li>
                  </ul>
                </li>
              </ol>
              <h3><span class="mw-headline"
                  id="Review_of_minutes_from_the_January_26_meeting">Review
                  of minutes from the January 26 meeting</span></h3>
              <p>We went over the <a
href="https://halldweb.jlab.org/wiki/index.php/GlueX_Offline_Meeting,_January_26,_2018#Minutes"
                  title="GlueX Offline Meeting, January 26, 2018">minutes</a>.
              </p>
              <h4><span class="mw-headline" id="Optional_Packages">Optional
                  Packages</span></h4>
              <p>In the context of discussing the recent changes with
                using the RCDB in the offline software, we started a
                more general discussion of whether certain packages, and
                RCDB in particular should be optional. David explained
                that there are certain packages (among them RCDB, the ET
                system) that have been maintained as optional. The
                mechanism:
              </p>
              <ul>
                <li>
                  <ol>
                    <li> If the "home" environment variable (e. g.,
                      RCDB_HOME) is not defined, then at build time, the
                      build system (i. e., SCons) does not include the
                      corresponding include paths and libraries in the
                      build commands. No dependence on the package is
                      built into the resulting code. A warning that this
                      is happening is printed during the build. This
                      does not halt the build.</li>
                    <li> If any program requires the package in
                      question, then it is built so that when the
                      program is run, it exits immediately with an error
                      informing the user that the environment variables
                      for the missing package needs to be defined and
                      that package built. Then the build of the program
                      can be redone.
                      <ul>
                        <li> To implement this behavior appropriate
                          C-preprocessor directives need to be included
                          in the source code so that the resulting
                          program has the appropriate behavior depending
                          on the setting or (non-setting) of the home
                          environment variable.</li>
                      </ul>
                    </li>
                  </ol>
                </li>
              </ul>
              <p>There was a lot of discussion on whether this is a
                useful feature that should be maintained, mainly between
                two of us. An incomplete summary:
              </p>
              <ul>
                <li> <b> David </b>. Having the ability to exclude
                  packages whose functionality is completely irrelevant
                  to the software builder is very convenient: extraneous
                  packages need not be built and the resulting disk
                  footprint of the code is reduced.</li>
                <li> <b> Mark </b>. Having this flexibility requires
                  coding and creates trap doors that code software
                  builders can fall into. If a needed environment
                  variable is missing, then the warning at build time
                  can easily be missed and the error at run time may
                  leave the user (who may not be the builder) at a loss
                  as to how to proceed.</li>
              </ul>
              <p>Mark will put the issue on the agenda of a future
                meeting.
              </p>
              <h4><span class="mw-headline"
                  id="Release_Management_Thoughts">Release Management
                  Thoughts</span></h4>
              <p>In the course of discussing Sean's presentation from
                last time, Thomas brought up the idea of breaking up the
                sim-recon repository into two repositories. In a moment
                of inspiration he came up with a concept for the split:
                one repository for simulation and one for
                reconstruction. This would simplify the task of "release
                management" (as defined last time). The technical
                advantage is that simulation code can be versioned
                independently of the reconstruction code. Right now,
                Sean has to maintain a
                reconstruction-fixed-simulation-changing branch of
                sim-recon to get the right behavior. If the two
                functions were versioned separately, this
                recon-fixed-sim-changing property would be manifest.
              </p>
              <p>We noted that the two sides (sim and recon) are
                functionally closely tied together. The question is the
                degree of independence of the development streams on the
                two sides. E. g. if drift time in tracking chambers is
                improved in simulation, then does the tracking
                reconstruction have to change? Likely not, the sim side
                can go ahead independently. On the other hand, if there
                is reconstruction code that does one thing for real data
                and another for simulation, and the simulation is
                improved so that unequal treatment is no longer
                necessary, then both sides have to change together. In
                the latter case it would be easier if both sides were in
                the same repository.
              </p>
              <p>Mark will put the issue on the agenda of a future
                meeting.
              </p>
              <h4><span class="mw-headline"
                  id="Not-the-TOF_Anomaly_in_Monitoring_Histograms">Not-the-TOF
                  Anomaly in Monitoring Histograms</span></h4>
              <p>Alex noted that the problem first appeared several
                months ago when the material maps were changed in the
                CCDB. Beni has reported that the most recent version of
                the code does not exhibit the problem, so the current
                mystery is how the problem could have possibly fixed
                itself.
              </p>
              <h3><span class="mw-headline" id="GlueX_.2B_NERSC">GlueX +
                  NERSC</span></h3>
              <p>David has succeeded in running GlueX reconstruction
                jobs on two of the <a rel="nofollow" class="external
                  text" href="http://www.nersc.gov/">NERSC</a>
                supercomputers.
              </p>
              <ul>
                <li> Cori I: Haswell (comparable to the JLab farm)</li>
                <li> Cori II: Knight's Landing (KNL)</li>
              </ul>
              <p>He analyzed two runs on both architectures. See the
                results on <a rel="nofollow" class="external text"
href="https://docs.google.com/presentation/d/1FcOhSprdD-YPCZnagvaJl-oicK7y3zoSAuB-Nmo_vSc/edit?usp=sharing">his
                  slide</a>. He notes that the KNL jobs run 2.4 times
                slower than the Haswell even though they are using four
                times the number of threads, i. e., ten times slower on
                a per thread basis.
              </p>
              <h3><span class="mw-headline" id="Collaboration_Meeting">Collaboration
                  Meeting</span></h3>
              <p>Sean has put together a nice little session for us on
                the <a
href="https://halldweb.jlab.org/wiki/index.php/GlueX-Collaboration-Feb-2018#Saturday_February_24.2C_2018"
                  title="GlueX-Collaboration-Feb-2018">Collaboration
                  Meeting agenda</a>.
              </p>
              <h3><span class="mw-headline" id="SciComp_Meeting_Report">SciComp
                  Meeting Report</span></h3>
              <p>Mark reported items from the Scientific Computing
                meeting held Thursday, February 1.
              </p>
              <ol>
                <li> <a rel="nofollow" class="external text"
href="https://mailman.jlab.org/pipermail/halld-offline/2018-February/003093.html">Change
                    to fair share allocations...</a>. The change was
                  discussed at the meeting.</li>
                <li> <a rel="nofollow" class="external text"
href="https://mailman.jlab.org/pipermail/halld-offline/2018-February/003092.html">ENP
                    consumption of disk space under /work</a>.
                  <ul>
                    <li> The second shelf of traditional raid is up and
                      running. Our work disk quota has been increased to
                      110 TB from 66 TB.</li>
                    <li> Hall B migration to the new work disk is
                      underway. That should free up 85 TB of cache
                      space. Mark emphasized that Hall D needs more
                      cache space.</li>
                  </ul>
                </li>
                <li> Hall B needs 40 GB per job for the Java virtual
                  machine. This is causing cores to go idle as nodes are
                  running in a memory-limited mode.</li>
                <li> '<i>New tape drives, playing with four new LT07
                    drives and four new LT08 drives.</i> They are
                  planning the upgrade path.</li>
              </ol>
              <h3><span class="mw-headline" id="AMD_benchmark_results">AMD
                  benchmark results</span></h3>
              <p>Sean has purchased a box with new AMD EPYC processors
                and ran benchmarks of hd_root on it. For comparison he
                ran the same tests on gluon119 which has Intel Xeon
                processors. See <a rel="nofollow" class="external text"
href="https://halldweb.jlab.org/wiki/images/f/fe/Sdobbs_Offline_020918.pdf">his
                  slide</a> for details and results. Scaling for the two
                systems is comparable as the number of threads is
                increased, though the AMD processors come a one-third
                the price (on the CPU package itself).
              </p>
              <h3><span class="mw-headline" id="Meeting_on_Containers">Meeting
                  on Containers</span></h3>
              <p>Mark announced a meeting later in the day to discuss
                use of containers (Docker, Singularity) in various
                computing contexts (NERSC, OSG, JLab farm, personal
                laptops). There is a lot of ground to cover here; a
                series of meetings is likely.
              </p>
            </div>
            <br>
          </div>
        </div>
      </div>
    </div>
  </body>
</html>