NRAO AIPS++ Users Group Meeting Date: 2004-3-24 (Wednesday) Time: 1300 MST Video Hub: SOC-conf (CV should call in to Socorro) Rooms: SOC317/CV311 AIPS++ Threat Level is: Yellow = Elevated [ALMA TST1 concluded, prepare for TST2...] Draft Minutes: 1. NAUG News o SS5 has been delayed, hopefully a successful stable will occur this week. *Joe: There is still one critical problem in imageprofile that needs to be dealt with. The plan for SS6 is up to date and visible at the Project Office. There will also be an intermediate stable snapshot SS5.5 scheduled for April 16 to be available for profiling for the Visiting Committee. o The items that need NAUG review/testing can be found at: http://projectoffice.aips2.nrao.edu/testing.html (also accessed from the Module Testing link on front page). Joe would like a review of: Imager - multi-spectral window imaging Calibrater - interpolation Viewer - blinking Viewer - statistical editing of data - use the adjust menus to select rms, etc. for editing (in MS and Visibility Selection of Adjust Panel). The aim is for the NAUG to sign off (or indicate problems with) these functions. Comments should be in the form of: functionality: usability: documentation: See examples of Debra's reports under the "NAUG Comments" links under the first two items. *Joe: has updated the page to include the test comments. *John: has posted some test logs, and has a few more to send in. o Testing reports *Steve: testing polarization calibration using a script that is appropriate for regression, as well as processing the VLA/VLBA Polarization Calibration observations. *Steve: Used the viewer for MS editing for antenna-based edits, and noticed a quirk on trying to edit out the last antenna in the list (VLA 28) where had to click on the undisplayed 28-28 autocorrelation to edit. Dave mentioned that a way to ask for a specific antenna (e.g. AIPS TVFLG) might be desirable. 2. AIPS++/ISD Status Report o The "Gang of Four" (Cotton, Greisen, Tody, Glendenning) concluded their several day meeting last week on getting a roadmap to the new framework. AIPS++ will (presumably) be relying on this framework so timely progress in this area is critical. *Joe: timescales for the roadmap are being set. It sounds like Corba has been decided on, and the the CCM implementation has been deemed too immature (though Darrel is investigating). The need for a light framework seems to have been agreed upon also. o Summer School: *Debra: the carbon star IRAM dataset looks like the most interesting to use, as it has interesting line structure. It was noted that we will need to get permission (from Rober Lucas?) to use this in the summer school (Action Item for Debra). 3. ALMA o ALMA TST2 is on the horizon. It has been pushed back to August in order to afford time for development, which has been hard to schedule due to the pressures of testing. For TST2 info, see: http://aips2.nrao.edu/projectoffice/almatst2/ALMA_TST2.html One of the focuses of TST2 is mosaicing. We have some possible test datasets in hand, and will begin evaluating these. *ACTION: Joe and Steve will take first crack at the mosaicing data. Joe has unpacked the BIMASONG data, and will soon try loading into AIPS++. o Benchmark Datasets: *Debra: Noted that G192 reduction using MFS in imager was slow and asked when this would be included in the benchmarking. Joe said that this would not occur for a while since MFS is not a critical deliverable for TST2 or CDR2. Kumar noted that the cost should indeed be dominated by gridding and gridding N channels should take N times longer at that step, so a factor of 10 (as reported by Debra) for 128 channels is not surprising. Steve noted that for G192 (22GHz VLA) that MFS is not really necessary, and that frequency independent gridding would speed up things greatly. 4. EVLA o No known EVLA issues at this time. *It was noted that Gustaaf had mentioned that it was time to start thinking about planning for the post-summer (post-CDR2) activities. 5. Main Event - Presentations and Discussion Items o George gave a description of his plans for calibrater improvements, including stuff that will be available soon and also targets further down the road. Details (well, a short list) can be found at: http://aips2.nrao.edu/projectoffice/cal_perfplan.html Summary: - cal transfer: across spwids (e.g. in gspline currently) should be straightforward and available soon - issues in working on cal transfer and other SS5 targets overlap with performance target work for SS6 and beyond After this, there is a shift toward perfomance issues. In rough order: - caltable i/o: currently rowwise operations are slow, big gain from moving to more columnwise operations - model_data column: use of trivial model should give noticable improvements for cases where point models appropriate - cumulative calibration: currently baseline outer products formed at each step (e.g. T, G, D, P?) from antenna based 2x2 matricies. Move to matmul followed by single outer product. Has implications for interpolation and provides the setup for incremental calbration (a later target). - code cleanup in core solvers: performance gains (e.g. remove extra object copies) possible Some issues for discussion came up: - how far should we go to optimize for continuum cases? *Need use cases (and relative importances of cases) *continuum optimization may be independent of other optimizations (e.g. for many-channel line obs) so could be done anyway. Needs investigation. - channelized versus pseudo-continuum terms *e.g. G and B (B is just a channelized G), in general will want a channelized D. Should the pseudo-continuum (G and D currently) just be an instance of a general channelized G (now B) and D with scope to transfer over a channel range? Food for thought... - channel zero for spectral line: *when do we need to form it, at fill? on-the-fly? It was noted that in some cases it needs to be formed after correcting the line channels from which it is formed (is this true for most cases?). *definitely needed for editing. *could this be virtual (e.g. the scratch column change proposal)? *for IRAM data we form this in the filler as a separate spectral window so do that in general (only costs as an extra channel)? It was noted that there was a place in the MS for this anyway (details?). o Possible future topics: - Change proposal: scratch columns and the tool pyramid reorganization (Joe, Steve, Kumar, George, Sanjay) - Wide-field imaging, w-projection, mosaicing (Tim & Sanjay) - The ALMA proto-pipeline (Lindsey) - Framework technologies and plans (Doug?) - Data reduction demos? (the summer school dataset?) 6. AIPS++ Developments - see latest targets and info at Project Office: http://projectoffice.aips2.nrao.edu/ o SS5 Targets -- see Project Office page for development status and testing info. o SS6 Targets -- see Project Office page for development status and testing info. 7. Upcoming meetings and deadlines: o 2004 Mar 30 AIPS++ SS5 (from Mar 16, late) o 2004 Apr 1 ALMA Release 1.1 o 2004 Apr 16 AIPS++ SS5.5 (for VC profiling) o 2004 Apr 26-28 AUI Visiting Committee (SOC) o 2004 Aug 1 ALMA CDR2 = Judgement Day! (docs Jun, rev in Jul) o 2004 Aug 1 ALMA TST2 begins?? The agendas for past NAUG meetings are archived at: http://www.aoc.nrao.edu/~smyers/aips++/agenda/ The minutes for past NAUG meetings are archived at: http://www.aoc.nrao.edu/~smyers/aips++/minutes/