Monday Morning Meeting March 14 2016
- DIAL-IN NUMBERS & PASSCODES:
- IP: 188.8.131.52##8110
- Phone: (434) 817-6524
News / Meetings / Visitors
- Pam will move to at will position.
- Chris Hales will be leaving end of year.
- No meeting next meeting
- Issue on 10.11 logger with x-code
- 4.6 Release issues
- Juergen to open o-tool issue
- Issue with long search times for models insetjy on OS X
- George to move fixes into the release branch
- Documentation for gencal TEC correction
- 4.7 Issue with mosaics being in flat noise.
- (pford) test_plotms failures:
- plotms() calls worked for 4.6.169 (r36313) but ALL failed for 4.7.1 (r36339) and prerelease 4.6.0 on RH5 and RH6. Tests passed on OSX. There were no changes to plotms between these revisions. test_plotms passes manually from 4.7.1 tarball on AOC and CV hosts. Data path for regression tests is valid. I cannot log into build&test server to investigate further.
- Automated testing on the cluster
- Spreadsheet collection of both critical and accepted tests currently running in the framework
- On request of Julian, Darrell added MPI 3 RPMs to el6/el7 repositories. Julian has updated the RPM specs for OpenMPI to 1.10.2 and mpi4py to 2.0.0 to support MPI 3. Note: OpenMPI will not build on el5.
- As part of the integration of MPI in C++ in CASA, Justo made a prototype showing how to create MPI sub-comms directly in C++ with MPI-2.0 and MPI-3.0; MPI 3 being the preferred one because only with version 3 we can create sub-clusters under any circumstances.
- Mark Kittenis working on bringing VLBI support into CASA.
- Pipeline working group distributed cycle 4 maging pipeline requirements document
- NAASC released bench marking imaging and simulation results for comparison with C3R4 pipeline
- pipeline cvel2 issues, some uvcontsub changes required
- Sanjay Bhatnagar
- Sandra Castro
- I have implemented an overwrite sub-parameter in task flagdata when savepars=True (CAS-8391).
- Verified CAS-8398 that flagdata does not flag antenna=X on ALMA bandpass tables whenever X is the reference antenna. I told Todd that for baseline-based cal tables, MS selection should be used with ampersands or the reference table will not be flagged, while for antenna-based tables (when ANTENNA2 is set to -1), not. The inline help of flagdata explains this. I also referred him to another ticket about this CAS-5754. He decreased the priority of the ticket to Minor, but I think it should be closed. This is not a problem in flagdata. Jeff could you verify this ticket and close it if you agree with my comments?
- Have committed a change to flagdata, which will process an MMS as a monolithic MS (in sequential) when CASA is started without mpicasa. Previously, flagdata would process each SubMS separated in sequential, which had the overhead of creating the jobs for each SubMS. With this change, flagdata tests run in half the time in the presence of MMS data, without an MPI cluster.
- Working on a change to the MMS structure, to have links to all sub-tables instead of a copy for each SubMS. I need to change how the history table is written in parallel (CAS-7896).
- Email discussion with Lindsey and Justo about mstransform/cvel2.
- Lindsey Davis
- ALMA imaging pipeline requirements review and development planning
- tclean data selection discussions with Jeff and Urvashi
- cvel in the pipeline discussions with Jeff
- uvcontsub discussions with Jeff and George
- workflow discussions with Jeff
- RHEL5 -> RHEL6 transition, not completely trivial ...
- 3 days vacation
- Pam Ford
- RH7 build for CAS-8383
- CAS-8383 (port casaqt to use Qwt 6.0) for RH7 - in progress, cannot compile 'code' on RH7 until this is fixed
- Release 4_6 build for CAS-8337
- CAS-8337 RTT (spw averaging bug) - committed to 4.6 branch and trunk
- test_plotms failures: see Verification Testing section above
- Kumar Golap
- Noise in Linear and Joint Mosaic
- webinar on optimizing code from Colfax
- Justo González Villalba
- Added a watchdog feature for the servers in the MPI framework to handle sub-process termination with MPI-3.0
- Made a prototype showing how to create MPI sub-comms directly in C++ with MPI-2.0 and MPI-3.0
- Review design document of XCLASS automatic line identification algorithm
- Removed unnecessary Plot Report generation in flagdata summary mode, and review handling of pointers for the field breakdown case
- Jim Jacobs
- Jeff Kern
- Slides for VC
- ALMA Imaging Pipeline discussions
- Susan Loveland
- CAS-8328: Text revision on radio/optical choice (ready to test)
- CAS-8364: Line overlays for high redshifts do not work (ready to test)
- CARTA: Work on reorganizing layers into groups.
- David Mehringer
- (CAS-8405) listobs in 4.6.164 took 13 hours (!!) when it took 1.5 hours (!) in 4.5.2. WIP. Iterating with Julian on submitted casacore PR. Fixed bug, implemented several optimizations so that listobs now runs in 11 minutes on my machine on the half TB dataset.
- Fixed imhistory test failure.
- (CAS-8406) ia.fft() to write history to image products. RTT
- (CAS-8403) ia.decimate() should use standard history writing method. RTT
- (CAS-8402) ia.convolve2d to write history. RTT
- (CAS-8399) ia.continuumsub() to write history. RTT
- (CAS-8397) ia.calcmask() to write history. RTT
- (CAS-8396) ia.calc() to write history. RTT
- George Moellenbrock
- CASA Fringe-fitting discussions with Mark Kettenis (visiting from JIVE)
- CAS-8313: Added nominal direction for Eph objects to Caltables. Found/fixed minor solution timestamp calculation error that was exposed by numerical pathology in calibrated weights.
- Dirk Petry
- offset tables still having issues
- Martin Pokorny
- Host by Rob Lathem
- Parallel Iteration Control now up to date, runs in serial mode.
- Urvashi Rao * Squashed heisenbug (we hope) * Fix to allow single stokes plane during imaging * Selection with a list of MSs (still needs discussion)
- Darrell Schiebel
* Ville Suoranta
* Parallelization webinar; should we push on the intel compiler
- Ville Suoranta
- Julian Taylor
- Takahiro Tsutsumi
- setjy listmodels=T bug fix (CAS-8412)
* uvconsub email exchanges with Justo
* Created CAS-8404. Looking into a bit for how to deal with outframe in tclean for 4.7
- Some science discussions with a ALMA cycle2 PI
---+++++ Friday NAOJ Meetin
* Kanako Sugimoto
- A discussion request by Erik: changing default of with_pointing_correction to True in importasdm. Pointing correction is inevitable in TP especially in fast scanning observations. Does it do any harm in interferometer?
- MS growth by adding CORRECTED_DATA column. sent the MS to Jim. haven't heard back yet. it turned out the new unit tests of initweights task is suffered from the issue.
- Wataru Kawasaki
- Masaya Kuniyoshi
- Takeshi Nakazato
- Renaud Miel
- CAS-7339 calmode=otf :
- clarified edges detection result mismatch between libsakura v2 and ASAP edges detectors
- verified correctness of caltable results on small testcase data (square , declination=60)
- verification of moon observation data ongoing (double-circle pattern)
- CAS-8362 RHEL7 / rpm: setup RHEL 7.0 virtual machine
-- KanaSugimoto - 2016-03-10