Monday Morning Meeting 9 December 2015

• DIAL-IN NUMBERS & PASSCODES:
• USA Toll Free Number: 866-901-8266
• USA Toll Number: +1-203-566-3863
• PARTICIPANT PASSCODE: 725760

Attendance

• Socorro: Jeff, Susan, Lindsey, Kumar, Tak, Martin, Pam, Jim, George
• CV: Andy, Darrell, Akeem, Ville,
• Garching: Dirk, Sandra, Justo, Julian
• SCO:

News / Meetings / Visitors

• ALMA CPM5 Meeting
• ngVLA Meeting

Build, Release

• RefImager- Having issues.
• Make App should be fixed, need devel distribution, have a single version now.
• OSX 10.11 Hardware in place, can start testing on it this week.

Verification Testing

• M100 Regression tests: runtime between normal (3 hours) and HPC (4 hours) tests are contradictory, since HPC should run faster
  • Serial and HPC have same speed in Garching

Validation Testing

Architecture

• Technical Forums:
• Warnings: Getting there
• Testing

HPC

• Sent an email to all interested people to have an HPC telecon this Thursday at 16:00 UTC. This is to go through the list of development items for 2016. The agenda is available in https://safe.nrao.edu/wiki/bin/view/HPC/Dec1015
• I/O Optimization Project

Development

• Remove parallel_go,

Pipeline

AOB

Developer Reports

Monday Meeting

• Sanjay Bhatnagar
• Sandra Castro
  • Had a team meeting with Justo and Julian to consolidate the list of development items for the HPC project.
  • Started to define the project scope for the I/O Performance Tuning project. Kick-off meeting should also be this Thursday after the HPC telecon.
• Lindsey Davis
• Pam Ford
  • Investigated test_plotms failure
    • change in VIVB2 caused exception in plotms; Jim fixed TransformingVi2 for this use case
    • still fails for plotms averaging (baseline & antenna) - array conformance error when averaging many rows into one. Emailed Jim.
  • CAS-2502 (highres=False in plotms does not work)
    • RTT - tested successfully by B&T group
    • Updated documentation XML files for CAS-2502 (remove highres "Always true for jpg and png") and CAS-7733 (add "must set selfscale=True" when using sharedaxis)
    • Lindsey reported iteration plots are blank except first one - rescheduled ticket
    • Fixed exporting iteration plots at screen resolution - BUT didn't work for grid - reverted code to do high-res exports only for now
  • CAS-1977 RTT (Set dpi, width x height for plotms exports)
• Kumar Golap
  • Looking into an issue of msuvbin where sources outside the field image causes larger ripple
  • checked in a working naclean in deconvolver
  • looked at CAS-8160..which there were suspicion of mem alloc issues in Virtual model column...could not reproduce.
  • Tested the MakeApp from Darrell, make be fixed.
• Justo González Villalba
  • Removed warnings from mstransform module
  • Fixed issue in mpi4casa related with iPython initialization at the servers
  • Implemented basic run-time re-configuration capabilities of OpenMP params
  • Provided feedback for TVI restructuring
  • Review ARTIST/XCLASS documentation
• Jim Jacobs
• Jeff Kern
  • ALMA CPM5
• Susan Loveland: Carta Testing, Selenium Test Writing, and Bug fixes for layered views.
• David Mehringer
• George Moellenbrock
• Dirk Petry
• Martin Pokorny
• Urvashi Rao
  • ALMA large cube memory use
    • Studied tclean memory profile + looked at code + discussions with KG.
    • Modified ordering of loops for channel chunking vs data I/O, eliminated a few ImageInterface ->get() calls from framework to prevent full image loads.
    • Found/fixed bug in memory release section of FTMachines. Channel chunking now has the expected memory profile.
    • TODO : Plumbing for correct modelvis prediction, study minor cycle mem profiles.
  • Some non-CASA work.
• Darrell Schiebel
  • finished linux make application script
  • merged linux script into original OS X "make-app" script
  • modified script to create casaviewer distribution
  • started back with linux Structure101 (but license had expired)
• Ville Suoranta
  • Working on the OSX10.11
• Julian Taylor
• Takahiro Tsutsumi
  • Worked on C++ unit tests. Converted some existing ones (and expanding test coverage) to Google tests
  • Looked at the tclean issue reported for the pipeline (CAS-8093). The problem may be in the data itself.

Friday NAOJ Meeting

• Kanako Sugimoto
• Wataru Kawasaki
• Masaya Kuniyoshi
• Takeshi Nakazato
• Renaud Miel