GlueX Offline Meeting, July 24, 2018

From GlueXWiki
Jump to: navigation, search

GlueX Offline Software Meeting
Tuesday, July 24, 2018
2:00 pm EDT
JLab: CEBAF Center A110
BlueJeans: 968 592 007

Agenda

  1. Announcements
    1. MCwrapper 1.16 released
    2. MCwrapper 2.0 released (split-off)
  2. Review of minutes from the July 13 meeting (all)
  3. Report from the HDGeant4 Meeting (Mark)
  4. Splitting up Sim-Recon (Sean, Mark)
  5. Status of Recon Launch (Alex A.)
  6. NERSC Status Update (David)
  7. Detached vertex how-to in GlueX simulations
  8. Retirement of old hddm-XXX utilities (Richard)
  9. What I Learned at CHEP2018 (David)
  10. Review of recent pull requests (all)
  11. Review of recent discussion on the GlueX Software Help List (all)
  12. Action Item Review (all)

Communication Information

Remote Connection

Slides

Talks can be deposited in the directory /group/halld/www/halldweb/html/talks/2018 on the JLab CUE. This directory is accessible from the web at https://halldweb.jlab.org/talks/2018/ .

Minutes

Present:

  • FIU: Mahmoud Kamel
  • FSU: Sean Dobbs
  • JLab: Alex Austregesilo, Thomas Britton, Mark Ito (chair), David Lawrence, Simon Taylor, Beni Zihlmann
  • UConn: Richard Jones
  • W&M: Justin Stevens
  • Yerevan: Hrach Marukyan

There is a recording of this meeting on the BlueJeans site. Use your JLab credentials to access it.

Announcements

  1. MCwrapper 1.16 released. This includes a fully-functional database scheme to accept "orders" via the web and track each job over its lifetime.
  2. MCwrapper 2.0 released (split-off). This is the same code as v1.16, except it is hosted in a new stand-alone repository on GitHub.

Review of minutes from the July 13 meeting

We went over the minutes.

The occasionally missing blocks of 40 events is still a thing. We agreed that it should be tracked as an issue in some way.

Report from the HDGeant4 Meeting

We went over the notes from the meeting. Since the meeting there has been action on several issues. For details see the corresponding issue pages, linked below.

Splitting up Sim-Recon

Mark reported that we have delayed the deadline for pushes to the sim-recon repository, in advance of splitting that repository into halld_sim and halld_recon, three times now. The main cause of delay is the lack of a final tag on the code for the reconstruction launch, and the desire to apply that tag before we do the split. There is still no definite date for application of the tag. We decided to forego deadline announcements in the mean time. Collaborators should be prepared to see the split occur soon after the tag is applied.

Status of Recon Launch

Monitoring launches and other studies are running now. There will be discussion of the results at tomorrow's analysis meeting.

NERSC Status Update

David reported on recent success running jobs at NERSC using the new swif2 tool from SciComp. See his slides for details.

  • We may have to set up group-like accounts at NERSC and/or Globus Online to manage jobs in the future.
  • There are several layers of scripts necessary to run a job, much like launches on the JLab batch farm, just with a couple more layers.
  • Data transfer is handled with Globus Online
  • With the measured rate and the Lab's advertised bandwidth, we should be able to sustain over 500 jobs simultaneously.
  • Each job takes about three hours.

Detached vertex how-to in GlueX simulations

Richard led us through his Guide to Monte Carlo event timing and detached vertices in HDGeant/4 wiki page. He wrote it in response to misunderstandings he has seen in private event generators, especially those with detached secondary vertices. There are two main topics addressed:

  1. How to maintain proper correlation between z and t of the interaction point to correspond to the position and time of photons synced to the machine RF.
  2. How to generate explicit particle decays and avoid double counting, i. e., having HDGeant track both parent particle and daughter particles.

Retirement of old hddm-XXX utilities

Richard has written a new Guide to roll-your-own python hddm transforms. These can be used in place of several compiled C++ programs we have been using to perform various manipulations on HDDM-based data. Now all of that functionality can be written in a very few lines of Python, a form which is much more transparent and much more easily modified and/or extended if the need arises. Several examples are presented and explained. David suggested that the some of these should be made into public scripts for general use.

What I Learned at CHEP2018

David presented a summary of his impressions from the recent CHEP2018 conference in Bulgaria. Please see his slides for all of the topics. Among the highlights:

  • Vectorization in ROOT
  • Deep Learning applied to FPGAs
  • Citable Software (referencing software in archival journals)
  • SPACK, a new package manager
  • Software Training workshop structure