Request for new machines to be purchased in: 3 mac minis, 1 linux machine
Andy to circulate wiki page
Just during the last week: Prerelease is not being tested for RH5 and RH6 accepted tests due to long test times and I/O errors
We will be looking at these problems this week
OSX 10.10 accepted tests have not been getting reported due to a reboot of the test system, and not reconnecting to cvfiler
Will be fixed this week
Cluster work from Akeem:
Make progress
Validation Testing
CAS-8313:
Possible issue with 7-m
Amanda testing on manual side.
Crystal today to discuss 4.6
Architecture
Technical Forums:
HPC
Development
Pipeline
New imaging pipeline work flow and heuristics
Split out target data, mstransform with reindex=False
UV continuum fitting with CASA cb tool and automatically detected continuum regions
Efficiency issues in cb.open
More callibrary efficiency improvements for VLASS survey data
Low SNR calibrator observation heuristics for EVLA pipeline
Target flagging heuristics for EVLA spectral lines projects
AOB
Developer Reports
Monday Meeting
Sanjay Bhatnagar
Sandra Castro
Lindsey Davis
Evaluated CASA UV continuum fitting tools for use in pipeline
Implemented CASA style uvcontfit task for pipeline
Worked on new imaging pipeline work flow: implemented hif_mstransform, hif_uvcontfit tasks
Drafted CASA news letter article
Pam Ford
Refactored plotms averaging. CAS-8337 and CAS-8394 (spw averaging) RTT, CAS-8456 (baseline averaging) reported and resolved by George.
CAS-8374 (port casaqt to Qwt 6) - Not trivial: Qwt 6 removed/added classes; changed method names, signatures, and return values; removed default and copy constructors. But I finally got it to compile on RH7. Need to test that I didn't break Qwt 5/RH6 compile then will commit.
Kumar Golap
Experimenting with adding a SF to mosaic beam to reduce ringing
Found that FFT of a beam takes as much as 24s on an 8 core machine
22s for FFTShift, 2s for FFT
wrote a FFTshift which takes around 0.8s for 10000x10000 array
Found that FFTs are defaulting to FFTPack now and not FFTW
Using threaded FFTW and FFTshift should now allow full polarization mosaic with 2-D beams to happen marginally slower than a just mosaic with a 1 D beam
Justo González Villalba
Jim Jacobs
Jeff Kern
ASKAP Software Readiness review
Susan Loveland
David Mehringer
George Moellenbrock
Dirk Petry
Martin Pokorny
Urvashi Rao
Darrell Schiebel
Ville Suoranta
CARTA Packaging
Julian Taylor
Takahiro Tsutsumi
Looked at CAS-8444 (tclean mask regridding for cubes not working optimally) - the request was tclean to have the similar behavior as clean which ignores regridding of the frequency axis for the input mask when the number of channels matches. May need yet another parameter to disable/enable spectral axis regridding for the input mask to support other use cases as well as this.
Automask fixes/modifications (CAS-7924)
A bug fix in the mask resolution determination, changed a factor in threshold, and cleaned up log messages (trunk, 4.6branch)
Maskresolution to take a multiplier of the beam (trunk only)