[Halld-offline] Offline Software Meeting Minutes, March 8, 2017

Wed Mar 8 19:14:13 EST 2017

Folks,

Please find the minutes below and at 
https://halldweb.jlab.org/wiki/index.php/GlueX_Offline_Meeting,_March_8,_2017#Minutes 
.

   -- Mark

________________________________

    Minutes

Present:

  * *CMU*: Naomi Jarvis, Curtis Meyer
  * *JLab*: Alex Austregesilo, Thomas Britton, Brad Cannon, Sean Dobbs,
    Mark Ito (chair), Richard Jones, Mahmoud Kamel, David Lawrence, Paul
    Mattione, Nathan Sparks, Simon Taylor, Beni Zihlmann
  * *MIT*: Cristiano Fanelli
  * *Yerevan*: Hrach Marukyan

There is a recording of this meeting <https://bluejeans.com/s/9X8NZ/> on 
the BlueJeans site.

      Announcements

 1. HDGeant4 repository has moved
    <https://mailman.jlab.org/pipermail/halld-offline/2017-February/002641.html>.
    The authoritative version of the repository is now part of the
    JeffersonLab organization on GitHub. The move was made two weeks
    ago. There is now a tagged release of HDGeant4, version 1.1.0
    <https://github.com/JeffersonLab/HDGeant4/releases/tag/1.1.0>. It
    has been built on all four platforms at JLab (CentOS/RHEL × 6/7).
    Mark will send out an announcement soon.
 2. DocDB-JLab is born!
    <https://mailman.jlab.org/mailman/private/gluex-collaboration/2017-February/004519.html>.
    The DocDB is now hosted at JLab. Thanks to Zisis, David and the
    Computer Center. Mark changed all links on the wiki and the private
    wiki to point to documents in their new location.
 3. New Scicomp Web Portal
    <https://mailman.jlab.org/pipermail/jlab-scicomp-briefs/2017q1/000138.html>.
    The new version of the SciComp webpages has been released. Brad
    pointed out that there is a new page that will track SWIF workflows.
    Sean also reported that Chris Larrieu has started work on a list of
    feature improvements he suggested some time ago.
 4. Software Carpentry Workshop, May 17 -19, 2017
    <https://mailman.jlab.org/pipermail/cuga/2017-March/001791.html>.
    Mark called our attention to this recent announcement. Grad students
    and post-docs may be interested.
 5. sim-recon-2.12.0
    <https://mailman.jlab.org/pipermail/halld-offline/2017-March/002667.html>.
    The last release came out January 23. It was time.
 6. hdpm 0.5.0
    <https://github.com/JeffersonLab/hdpm/releases/tag/v0.5.0>. Nathan
    is just about to release this version. He has incorporated
    suggestions from several people since the last release. There is a
    new feature where sim-recon can be built incrementally; a total
    rebuild from scratch is no longer necessary.

      Review of minutes from the last meeting

We reviewed the minutes from February 22 
<https://halldweb.jlab.org/wiki/index.php/GlueX_Offline_Meeting,_February_22,_2017#Minutes>. 

        OSG Update

The submit node is being installed as we meet.

Richard gave us an update on the OSG All-Hands meeting going on at UCSD.

  * He gave a 15 minute presentation on GlueX status and plans vis-a-vis
    the OSG.
  * There are a lot of non-accelerator based particle physics
    experiments looking to the OSG for computing cycles. Currently
    queues are long.
  * One trend is using Docker containers
    <https://en.wikipedia.org/wiki/Docker_%28software%29> to export
    compiled code to the OSG for running. This eliminates a lot of the
    problems getting locally developed code running in an environment
    where the user has no control of the natively installed software
    base. The idea is to bring your software base with you in the
    container. On a national scale, there are computing projects with
    plans for with huge computational capabilities and this technology
    may allow us to take advantage of crumbs of computing that may come
    our way, despite possibly novel machine architectures. Nathan has
    been using Docker to build his HDPM distributions for the various
    OS's that he supports.
  * If the GlueX VO is able to contribute computing cycles to the OSG,
    our quality of service would improve proportionally. Doing this has
    been on various lists for some time, but so far only UConn and NU
    have offered nodes. Rob Gardner of the OSG is willing to work with
    us to connect university clusters to the Grid. Mark will organize a
    meeting with interested parties to discuss plans.

        MCwrapper

Thomas has a customer: Justin and a student of his. The next step is to 
use HDGeant4 under the hood.

      Mixing real random triggers with simulated events

Thomas reported on progress:

  * Thomas has written an EVIO-to-HDDM converter. The idea is to take
    random triggers, render them in HDDM and pass them to mcsmear along
    with simulated physics events.
  * Richard has agreed to take on combining the hits from these two
    sources in mcsmear.
  * Sean will manage production of HDDM files containing the random
    triggers from raw data.
  * Beni asked if a similar process could be used to combine real PS
    triggers with real physics triggers to simulate running at a
    user-selected background rates. Folks thought that this could be done.

      Review of recent pull requests

We looked at the list 
<https://github.com/JeffersonLab/sim-recon/pulls?q=is%3Aopen+is%3Apr>.

  * Richard described recent work on EM background in hdgeant
    <https://github.com/JeffersonLab/sim-recon/pull/748>. See his email
    <https://mailman.jlab.org/pipermail/halld-offline/2017-March/002668.html>
    for a detailed description. This fixes a bug where spurious
    background photons were being generated, slowing down execution. The
    change also introduced the BGTAGONLY card so that tagger accidentals
    can be generated in addition to the "photon" which creates the event.
  * David put in a change to pull out
    <https://github.com/JeffersonLab/sim-recon/pull/751> the
    locked/not-locked status of F1TDCs in a more explicit form.
  * David has several pull requests aimed at implementing "time
    fiducials" so that events from beam-off periods can be eliminated
    from the analysis and beam-on time can be calculated accurately.
    Time maps of the runs are stored in the CCDB and are now being
    generated automatically. See his log entry
    <https://logbooks.jlab.org/entry/3461150> for details.
      o He is also working on a capability on creating the map from the
        time-density of triggers in the data itself.
      o To facilitate these operations, he is creating a skim of 1%-ish
        of the data. These will be written to tape along with the raw data.
      o Now there is a facility to write out the locations of blocks in
        the raw data. These can be used by EventStore to jump to
        particular locations in the file without having to go through
        the file event-by-event.
  * Paul re-factored and reorganized detector matching routines
    <https://github.com/JeffersonLab/sim-recon/pull/737>.
  * Naomi mentioned a problem with one where CCDB is being interrogated
    with the wrong run number even though the correct run number is
    known to hd_root. She will post a description to the software help list.

      Review of recent discussion on the Gluex Software Help List

We perused some of the recent threads on the List 
<https://groups.google.com/forum/#%21forum/gluex-software>.

  * Paul can generate "flat" TTree's
    <https://groups.google.com/forum/#%21topic/gluex-software/ooaCeUSAwJ8>
    with DSelector now.
  * He also has a method for creating SWIF workflows based on a
    particular DSelector
    <https://groups.google.com/forum/#%21topic/gluex-software/h245qyh_NFM>.

Retrieved from 
"https://halldweb.jlab.org/wiki/index.php?title=GlueX_Offline_Meeting,_March_8,_2017&oldid=81222"

  * This page was last modified on 8 March 2017, at 19:11.

-- 
marki at jlab.org, (757)269-5295

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-offline/attachments/20170308/27ef13f7/attachment-0001.html>