Monday Morning Meeting 11 July 2016

  • DIAL-IN NUMBERS & PASSCODES:
  • IP: 192.33.117.12##8110
  • Phone: (434) 817-6524

Attendance

  • Socorro: Jeff, Anand, Susan, Sanjay, Urvashi, Kumar, Martin, Tak
  • CV: Andy, Akeem, Darrell
  • Garching: Justo, Julien, Sandra
  • SCO:

News / Meetings / Visitors

  • ASIAA CASA Development Center
    • located in taiwan
    • 4 members team to begin.
    • 3 year pilot. At end of period should be up to 8-10 FTEs.
    • Interested in visualization and HPC. Will start with visualization, specifically CARTA.
    • Team members will be in Socorro first week of August.
  • Pipeline Cycle 4R2 release
    • Went out on time -- July 1st.
    • Internal release targeted for pipeline WG -- to start testing functionality.
  • 2 Open Positions in CASA Team
    • CASA Liaison
    • Algorithm scientist

Build, Release

  • test_tsdbaseline continues to fail on RHEL7
  • CASA 4.7 Release
    • Feature freeze Aug 1
    • Code freeze Aug 15
    • Target Branch Aug 29

Verification Testing

  • Poor testing performance of VMs could be due to the VM slices being created with qcow2 instead of raw devices (to make for easier expansion of disks)
    • Julian is trying this at ESO to see if the performance declines there in a similar way
  • ALMA first look casaguides are in the test framework
  • General discussions on the state and direction casaguides are going in
    • More recently calibrated data will be kept internally (not uploaded to the guides), and be used to validate test results (tests currently failing due to differences in calibrated data values)
  • Planning on how testing will be done in the new Atlassian CI system - related to the test inventory spreadsheet updated recently

Validation Testing

  • Jen currently AOD in Chile. Trying to stay up to date while there.

Architecture

  • Warnings Summary
  • Continuing to modify CASA to work with CasaCore being in the "casacore" rather than "casa" namespace. It's going somewhat slowly.

HPC

  • Continued testing the pipeline in parallel. Run into different tclean issues. Some problems are due to files not being found by the pipeline in parallel, when restart=True. The latest news from Dirk is that ia.statistics can not handle the virtually concatenated cube.
  • Verified the seg fault from gencal when running on MMS in the EVLA pipeline. George suspects an incompatibility of the TableIterator and the MMS. My debugging pointed me to GenSort in casacore. This problem only happens with this dataset!
  • Committed change to MMS structure that makes all sub-tables of the sub-MSs symbolic links to the ones in the first sub-MS (except for the SOURCE and HISTORY tables). This caused a degradation in performance in flagdata autocorr=True, which was due to a bug that was accessing the FLAG_ROW column in the PROCESSOR table in write-mode. Working to fix this now.

Development

Pipeline

AOB

Developer Reports

Monday Meeting
  • Sanjay Bhatnagar
  • Sandra Castro
    • Responded to CAS-7518 (flagdata extend growfreq reports wrong value), which seems to be a misunderstanding on how the extend mode is automatically used for the auto-flagging modes.
    • Lots of non-CASA related work: interviews for new TelCal lead position, etc.
  • Lindsey Davis
  • Pam Ford
    • Investigated CAS-8018 and CAS-8787, reassigned to Justo for possible bugs in underlying MSTransform code
    • Reviewed plotcal code for adding functionality to plotms
  • Kumar Golap
  • Justo Gonz├ílez Villalba
    • Validated uvcontsubTVI vs uvcontsub3 for simple cases like constant model visibilities with non-trivial channelized flags, implemented Google unit test for it
    • Participated in Data Analysis Tools mid-project review: XCLASS seems to be in the good track, with unit tests working, including Line Identification test cases, but the integration of ARTIST-LIME into CASA is a bit behind of schedule.
    • Look at some MSTransformIterator issues reported by Pam
  • Jim Jacobs
    • Moved virtual model logic into VI implementation so that it will play better with the VI framework. Will flush in mods this a.m.
    • Aded method to VB to allow setting model data as needed by synthesis.
  • Jeff Kern
    • Mostly Holiday last week, some meetings.
    • Prep for ALMA Inf. Pipeline Review
  • Susan Loveland
    • Worked on integrating Pavol's CyberSka spectral fitting code into CARTA.
  • David Mehringer
    • (CAS-6907) multisource description in exportuvfits documentation is ambiguous: Resolved. Updated documentation to the best of my ability given that the details behind the intent of this parameter seem to have been irretrievably lost.
    • (CAS-8796) modify ImageTask::_removeExistingFileIfNecessary() to not remove file if present in table cache Resolved.
    • (CAS-8775) imfit segfaults casa during a multi-component fit with constraints RTT. now logs warning message in the case where overfitting produces nonsensical results and attempts to continue.
    • (CAS-8715) imsubimage 'output file exists' message improvment: resolved. Message improved.
    • (CAS-8741) exportuvfits seems to not set CRVAL of FREQ axis right: Resolved. Fixed case of single channel reading past end of vector.
    • (CAS-8783) add parallel support to imstat/ia.statistics(): WIP. Spent most of my time on this. Working closely with Julian. Making headway and refactoring and improving existing code.
  • George Moellenbrock
  • Dirk Petry
  • Martin Pokorny
  • Urvashi Rao
    • CAS-6675 : Several small feature edits to clean up pbcor implementation in tclean
    • CAS-8789 : Fixed a problem that chanchunks had with model column writes. Discussed with KG about a cleanup/merge of methods.
    • Started adding EVLA 1D polynomial beam models to PBMath, so that they can be used in the widebandpbcor task.
  • Darrell Schiebel
    • scattered work on bamboo, jira, understand and structure101
    • jenkins update after server update to RHEL6
  • Ville Suoranta
  • Julian Taylor
    • working on fixing the casa write cache and its implications: gh-457
    • removed 15% of memory allocations in applycal (gh-458)
    • added more extern templates to casacore to improve compilation time (gh-459)
    • reviewed parallel statistics branch (gh-455)
    • some performed inconclusive experiments to try to reproduce poor m100 regression performance in cv's VM's
  • Takahiro Tsutsumi
    • Worked mostly on making further changes for automask and testing it. It now includes pruning of masks based on size and number of peaks. Also, made a non-binning version of automask available.
    • Some fixes to parallel cube tclean error reported by Dirk M
Friday NAOJ Meeting
  • Kanako Sugimoto
  • Wataru Kawasaki
  • Masaya Kuniyoshi
  • Takeshi Nakazato
  • Renaud Miel
Topic revision: r17 - 2016-07-11, JaredCrossley
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding NRAO Public Wiki? Send feedback