 News |
FAQ
|
| Search
|
Home
|
| Getting Started | Documentation | Glish | Learn More | Programming | Contact Us |
|
| Version 1.9 Build 1367 |
|
T.J. Cornwell, NRAO
1999 January 27
The major accomplishment of this quarter was the successful release of the third beta on October 15, 1998. In addition, a stable release was designated in the week of 18 January 1998.
In this quarter, a special emphasis has been placed on the integration and standardization of applications development in preparation for the first public release in 1999. A meeting to address these objectives was held in Socorro in November. A key objective of this initiative is to standardize the framework for AIPS++ applications and to ensure as uniform a user interface as practical. The successful discussions at the meeting in Socorro have been followed up by weekly integration meetings to ensure that the identified items are addressed in the preparations for the public release. Planning in all areas was also reviewed at the time of the integration meeting in November, in preparation for the release.
In Single Dish support, Dish itself remained largely unchanged after the beta release during this quarter. A revised dish plotter has been designed. Dish was demonstrated at the GBT Advisory Committee meeting in October. Garwood and McMullin each presented poster papers at ADASS in November. Garwood attended the aips++ integration meeting at the AOC following ADASS.
Both Bob Garwood and Joe McMullin continue to be involved in supporting the GBT. This includes regular 2 to 3 day visits by at least one of them to Green Bank for most weeks during the quarter. The aips++ position in Green Bank remains unfilled.
The dish plotter and results manager were incorporated into the GBT continuum commissioning tools. Several options for pointing solutions for the GBT were examined. It is likely that we will use the TPOINT software for this task. The GBT filler are being rewritten.
In Synthesis support, the development strategy outlined in the previous quarterly report for synthesis imaging and calibration has been continued in this quarter, with primary emphasis on the expansion of the scientific capabilities and usefulness of the package. Development on a broad front has continued in calibration and imaging, but resources have also been allocated to the "thin-path" capability in the area of data fillers and uv-data visualization, consistent with the overall strategy discussed in earlier reports.
As previously agreed, work on the implementation of the second version of the MeasurementSet format was scheduled in this quarter. This has included a formal change proposal required by the project software engineering guidelines. Work on the implementation has been undertaken by Mark Wieringa (ATNF), and will be well in place before the public release, as previously planned.
Peter Barnes (NRAO), who joined the project in the preceding quarter, has assumed responsibility for the development of uv-data visualization components compatible with the current Display Library. This is a cooperative effort with Athol Kemball (NRAO) and David Barnes (ATNF). A revised MeasurementSet summary format, and data lister has been completed, and an initial layout of display components had been completed along with a plan for their incorporation in the Display Library. This work will continue actively in this quarter.
Work on the VLA filler has been undertaken by Ralph Marson (NRAO), with current efforts focused on tape support and MODCOMP format translators. This target will proceed actively in the next quarter.
Calibration capabilities have been expanded this quarter, with the adoption of the revised and extended calibration table format, and the implementation of basic cross-calibration capabilities. Current efforts in this area involve the extension of interpolation and data selection options. A new initiative in integrating single-dish and interferometry calibration has been undertaken in an initial group comprising Athol Kemball, Tim Cornwell, Joe McMullin and Bob Garwood, with outside expertise provided by Rick Fisher. Two meetings have been held in this regard and the resulting note will be forwarded more widely once it is complete for broader discussion. The aim of this work is to standardize calibration models and data formats between single-dish and interferometric calibration development within the project, to allow joint calibration or calibration transfer.
Jan Noordam (NFRA) has continued work on supporting TMS and WSRT commissioning using customized Glish scripts to assess data quality. Details of this work can be found in the NFRA section of the report.
In imaging, the significant effort in mosaicing started by Mark Holdaway (NRAO) in the last quarter has continued. This has included the expansion of primary beam models supported in the imaging system and their representation. He has also nearly completed the early implementation of the Cornwell-Evans MEM algorithm, as required for subsequent planned development in mosaicing. Tim Cornwell (NRAO) has taken over 3-D wide-field imaging, and has successfully demonstrated a prototype capability in this area. In addition, he has implemented a multi-scale CLEAN deconvolution algorithm, which has shown significant improvements in imaging performance in early trials. General refinements to imager developments have continued during this quarter, including the development of image-plane-only deconvolution capabilities in support of the MEM and multi-scale clean developments.
A summary of work on the parallelization of synthesis code can be found in the NCSA section. A primary emphasis in this area has been the demonstration of a scientifically useful capability in the area of embarrassingly parallel problems. An implementation of this type for spectral line deconvolution has been demonstrated in early imaging of a four-pointing, multi-configuration VLA dataset taken towards M33, which cannot easily be imaged using existing packages. Athol Kemball (NRAO) presented a talk on calibration at the ADASS'98 meeting in November. Tim Cornwell (NRAO) provided a demonstration of AIPS++ at the same meeting. The NCSA funded position at NRAO for parallel synthesis development has been filled, with a current starting date of 1 April 1999. There are no other outstanding positions in synthesis at this time. In Glish, the past quarter has been one of consolidation. The graphical widgets were tentatively moved to a client for testing during the second quarter. At that time, while the default behavior of the widgets was unchanged, the new widget client was made available for testing. This quarter the new client-based widgets became the default widgets. This change revealed a number of problems both in Glish and in the client-based widgets. Fixing these and other problems, as well as adding a few new features were the primary Glish activity of the quarter.
A significant amount of time was spent working on memory leaks arising from self referential records. As a result of records containing functions, these sorts of records are often created. While a solution was found, more work is required to make it efficient before incorporating it into the distributed version of Glish. Other bug fixes included:
In addition to fixing bugs, a few new features were added during the quarter. These new features include:
In Measures, only limited changes were made:
In AIPS++ Infrastructure, we have focussed on integration of the the user interface. This required consolidation and reconcilation of the various interfaces, both graphical and command line, that have been designed. This was a major discussion item in the November integration meeting. The changes in the last quarter are:
Two major additions have been made to the Table system:
A few small bugs have been fixed and some small additions have been done.
A group of 14 Lattices classes has been submitted for review. Their documentation and test programs have been upgraded considerably. It is expected that the review will be finished by the end of January after which they can be moved from trial to aips.
The Lattice and Image region classes have been finished. The LC classes define a region in pixel coordinates. They are used by the Lattice classes (e.g. to create a SubLattice). An LC region consists of a (minimal) bounding box and optionally a mask. The mask tells which pixels in the bounding box belong to the region. The basic classes can be used to define a box, ellipsoid, or polygon. The compound classes make it possible to make the union, difference, complement, etc. of regions. The WC classes define a region in world coordinates. They can also be (partly) defined in pixel or fractional coordinates. When a WC region is used on an Image, it is first converted to an LC region. In this process axes are permutated as needed, so the order of the region axes matches the order of the image axes. Similar to the LC classes there are basic and compound WC region classes.
The LEL (Lattice Expression Language) classes have been changed to handle masked lattices (e.g. a sublattice formed by a circle) correctly. They even handle a lattice without any valid element correctly. Note that LEL is optimized such that no extra tests are done when a lattice has no mask. The LEL grammar has been extended with functions complex (forming a complex from two reals) and length (get length of given axes) and with the [] operator. This operator makes it possible to mask a lattice with a condition.
It appeared that egcs1.0 compiled some code using LatticeExprNode in a strange way (it invoked the Lattice copy constructor). By making that constructor protected, such cases were caught. It looks as if egcs1.1.1 is handling this better.
The class TempImage has been created to make temporary images (in memory or on disk depending on the size). To avoid having too many open files, functionality has been added (also to TempLattice) to temporarily close such an image when on disk. The reopen is done transparently.
In Image Analysis, the main goal was to write an image-plane fitter that could work both interactively and non-interactively. It was to produce a standard componentlist of fitted components. Display was to be accomplished using the Display Library viewer (see below). Neil Killeen writes in regard to this work:
As my next experiment in writing an application in aips++, I started on imagefitter, to fit 2-D models to images. To do this I created infrastructure C++ classes Fit2D and GaussianConvert. I then added a function to the image DO and Glish wrapper to fit a region of an image to a model. I then created an interactive application in Glish using this basic fitting service. At present the displaying/cursoring is done via the pgplotwidget. This will soon be replaced by Display Library capability when David gives it to me (some experiments already done successfully on this).
We expect this, in conjunction with the component processing in the synthesis application imager, will be one of the major deliverables of the first release.
In addition, the image region functionality was completed:
In Display work, work proceeded on a number of fronts all aimed at developing the library and producing some initial applications:
In the System area, we continued some minor work on compilers. In an effort to evaluate the egcs compiler, egcs1.1.1 has been installed on a few machines. In general it seems to work fine, but alas Solaris 2.4 has a bug which prevents it from building shared libraries. It means we cannot switch to egcs1.1.1 until all sites have moved to Solaris2.5 or higher. The strong points of egcs are that it supports almost the full C++ standard and that its exception handling is multi-thread safe. Egcs1.1.1 is stricter than gcc, so the code had to be changed in a few places. A few small tests indicated that code compiled with egcs seems to perform at the same level as code compiled with gcc.
A few general changes have been made to adhere the new C++ standard in a better way:
Towards the end of the quarter building the documentation seemed to be quite stable.
The aipsinit.sh file has been changed such that in general it can set up the environment such that an aipshosts file is not needed. This change has not been put into the system yet. because aipsinit files for other shells have to be changed too. Furthermore it'll be tried to cope with setting LD_LIBRARY_PATH as good as possible.
We considered replacements for our bug-tracking system, the Free Software Foundation GNATS system, which we have now outgrown. None of the free software candidates offer enough functionality and so we are considering buying commercial software. The cost is moderate and would be confined to consortium sites, maybe even to NRAO only. Commerical systems provide a range of functionality ranging from simple defect tracking to full help-desk maintenance, most of which we will need in the near future. Evaluation will continue in the next quarter.
In the area of Parallelization, the major push of the last quarter was a demonstration of the parallel AIPS++ system at the ADASS'98 conference in Urbana the first week of November. Wes Young and Doug Roberts presented a demonstration of the parallel capabilities of the AIPS++ system. Wes Young has implemented a parallel algorithm applicator in the AIPS++ system. This applicator class can be used for algorithms that can carry out significant work across multiple processors with little or no communication between the processors. This demonstration showed the first use of this class to deconvolve a large spectral line cube within AIPS++. The demonstration ran remotely on one of the NCSA Origin2000 systems. The implementation of parallel processing is with the Message Passing Interface (MPI). MPI is a portable system that allows data and instructions to be sent to remote processors (either on the same machine or on different machines).
In addition to the work put into the parallel algorithm applicator which was put into the imager application (and subsequently renamed pimager) significant effort was put into the user interface. Doug Roberts and Wes Young have worked on the user interface to allow users to access the batch systems at NCSA. The NCSA policy is to allow interactive systems for short cpu times (15 minute limit); all other jobs must be submitted to the batch manager lsbatch, which is a load-sharing batch system built on top of the Load Sharing Facility (LSF) system. The interface additions to the object manager allowed the users choices in the GUI to be written to a Glish script file. The Glish script could be submitted to the batch system with an additional click in the GUI. The NCSA enabling technology (ET) team responsible for high performance computing has been involved in an effort to obtain I/O statistics of the AIPS++ code. The Pablo group at the University of Illinois Computer Science department has developed an I/O library that records the I/O usage during the execution of an application. The group also provides a set of tools to analyze the I/O data in the recorded file. The tools can be used to create statistics in several forms, from a table to a 3-D output that can be visualized on an immersive display. Doug Roberts and Wes Young have been involved in assisting Pablo group get a stable AIPS++ installation and linking in the instrumented I/O libraries. The changes to the code have been checked into the AIPS++ system to allow easy instrumentation of code running on other machines. Just before Christmas, we saw the first output from an "imagertest". Additional testing using larger data sets will be required before making conclusions on the I/O performance of the AIPS++ system.
Athol Kemball, the head of the parallelization project circulated a plan for the next two quarters of the parallel project. It includes plans for a small exploration of OpenMP for fine-grain parallelization of gridding, continued work on the algorithm applicator class to extend it to additional processing (beyond Clark CLEAN that is currently parallelized). The group will assist in taking two large datasets (a VLA HI mosaic and a VLBA monitoring experiment) through the parallel AIPS++ system.
In Documentation, we continued improvement of the ``Getting Started in AIPS++'' document. We investigated replacement of our flaky and fragile, home-grown documentation system with other alternatives. A survey of commercial products and free software produced nothing that would meet all our needs. Our plan is to nurse the current system along until after the first public release, at which point, we expect to have to make a change with some concomittant loss of functionality.
In Management, we:
In the Quality Assurance Group, the following has occurred.
The third beta release was made on October 15, 1998. Response from our list of external beta testers has been very poor. We have received about a dozen bug reports, and few comments. A poll of the approximately 20 sites in mid December produced 8 responses, of which most people said that they had installed the system but not used it extensively. A few sites (NRL and JCMT, for example) are using the system in earnest. Many plead overcommitment of time. Lack of fillers is for some an issue, e.g. NRAO 12m.
We expect to make a first release within the first six months of this year. We plan a code-freeze on 15 March 1999. This will be preceded and followed by periods of intensive testing. Our intention is to make a robust, well-documented release in time for the summer meeting of the AAS in Chicago. Most major capabilities expected in the release are now in the system, and require debugging or some relatively minor additions. The key areas yet to be finalized are:
Testing of the system prior to the release is a serious concern. We currently test AIPS++ in four different ways:
To improve project-wide awareness of the Glish script testing, we now perform so-called ``Build and Smoke Test'' three times a week. This practice is commonplace at Microsoft and other commercial software vendors. As described in an IEEE software article (Best Practices, IEEE Software, Vol 13, No. 4, July 1996), the key is to arrange that:
Every file is compiled, linked, and combined into an executable program every day, and the program is then put through a "smoke test," a relatively simple check to see whether the product "smokes" when it runs.
Our smoke test is to run the assay script which executes a number of glish-based test scripts within the typical user environment. For logistic reasons, we do this three times a week instead of daily. The summarized results are then mailed to our aips2-lib mail exploder so that if the system smokes, everyone can see who is responsible. This accomplishes two objectives: first, developers become more aware of the current state of the system as would be experienced by users, and second, peer-pressure is utilized to minimize breakages.
Both directed and un-directed User testing are vital in producing a high quality first release. With the recent loss of our Chief Tester, Jan Noordam, to WSRT and SKA duties, we have no-one currently designated to lead the testing. We must resolve this in the next weeks, even if it means a reassignment away from development. We now believe that the the ineffectiveness of external beta testing is to be expected without an established user base. This means that we will have to rely upon internal testers. At the AOC, we have two testers who devote significant time to testing, and who have a demonstrated history of providing valuable and useable feedback. We must make efforts to find and cultivate similar testers at the other consortium sites. Some help and pressure from the Executive Committee is needed here, particularly in providing help for directed testing.
In Single Dish, the top priority at the start of Q4 is to get the new GBT filler working so that it can be used for a planned holography backend test towards the end of the quarter. The details of the GBT pointing solution and its integration with other aips++ tools will be worked out.
Once the revised fillers are in place, the focus will shift back toward dish. Dish must be made able to interact directly with an aips++ MeasurementSet. A multi-component fitting operation will be added to dish as will an fft operation. The dish documentation will be improved with a focus on describing the command line interface so that users can use other aips++ tools to do things which can not be done with the dish graphical interface.
Towards the end of the quarter we plan on starting to implement some single dish calibration schemes in aips++. Discussions will also begin on a unified approach to holography within aips++.
in Measures, we plan to:
In Image Analysis, we plan to:
In the Display Library, the next quarter will be spent on touching up and documenting the viewer application in time for the First Release. Only two major outstanding issues remain: the PostScript output, which has been accepted as a target by Harold Ravlin (NCSA), and proper axis labelling and annotation methods. One moderate issue remains to support simple region drawing for the imagefitter application being developed by Neil Killeen (ATNF).
In AIPS++ Infrastructure, we plan to complete the revised user interfaces.
In System, we plan to move to shared libraries, as needed for integration of the Display Library into other applications.
This quarter the ATNF had 4 people working in AIPS++. These are Neil Killeen (25-75%, also local manager), Wim Brouw (50%), Mark Wieringa (30%) and David Barnes (75%).
As foreshadowed last time, Wim's availability decreased owing to a temporary secondment to the ATNF's ATOMS project. However, Wim did more aips++ work than expected, and he will continue on reduced aips++ rations (30%) in the first quarter 1999 as he completes his ATOMS work.
Our systems (Solaris/Gnu and Digital Unix/Gnu) were generally stable this quarter. We will start egcs installations in the new year for Solaris, Digital Unix and Linux. The latter on our new Linux server.
David and Neil visited Socorro for two weeks to attend the integration and applications workshop (no doubt described elsewhere by Tim).
For the next quarter, I have lined up three people at the ATNF to test components of the system. They will analyze ATCA and VLA data as part of their testing.
David's time was spent on the Display library as described above. In addition, he attended the Socorro meeting.
Wim has spent his time this quarter on:
This quarter I mainly consolidated regions in the system (although I know that I must do one more substantial revision), generated a lot of reuseable Glish/Tk widgets and started on a new application to fit models to images.
Neil's time went on:
I also widgetized regionmanager (from which much of my stuff has its origins), Wim's measures GUI (non trivial) and the quanta GUI.
As my next experiment in writing an application in aips++, I started on imagefitter, to fit 2-D models to images. To do this I created infrastructure C++ classes Fit2D and GaussianConvert. I then added a function to the image DO and Glish wrapper to fit a region of an image to a model. I then created an interactive application in Glish using this basic fitting service.f At present the displaying/cursoring is done via the pgplotwidget. This will soon be replaced by Display Library capability when David gives it to me (some experiments already done successfully on this)
This quarter Mark's time has been spent on
Planned for 1999Q1:
During the last quarter, we had Harold Ravlin (5%) and Doug Roberts (50%) working on the AIPS++ project. In addition, NCSA is funding the salary of Wes Young at NRAO-Socorro and will shortly fund the salary of a second NRAO person. Accomplishments during the past quarter are described above under Parallelization.
A significant amount of time was spent by our group in preparation for the ADASS'98 conference, which we hosted in Urbana on November 1-4. The conference was also a useful deadline, which prompted the parallel group to finish up some of the parallelization work in order for a presentation at ADASS.
Our native IRIX build is now stable. The builds are first done on an IRIX system at NRAO. When builds are relatively clean there an inhale is done at NCSA. We still have in place a red/black system which allows a build in one while a user is actively using the other. This system seems to be working out fine. NCSA is now connected to the NRAO intranet via Socorro. This has significantly increased the interactivity for Wes Young in Socorro working on the NCSA Origin 2000. This should also allow overnight transfers of moderate sized data sets from the VLA to NCSA.
Several AIPS++ programmers attended the ADASS'98 conference in early November. Our group had significant discussions with Athol Kemball (the parallel project manager) and Wes Young. Wes Young also visited from December 16 through Christmas. Wes and Doug met with the performance engineering group at NCSA and the Pablo group at UIUC CS department. Wes and Doug started processing a large 6k x 6k x 120 HI image of M33 (from Dave Westpfahl, New Mexico Tech).
Doug Roberts visited Houston for a short meeting to encourage interactions between the high performance computing enabling technology team of NCSA and AIPS++. Several possible collaborations were discussed, including the use of a Linux cluster at the Albuquerque High Performance Computing Center. We also discussed the use of MPI-I/O to address our I/O problems.
A significant fraction of the time of the NCSA personnel was taken up with the ADASS'98 conference. Nothing new was added to Aipsview in the last quarter. The parallel work progressed to the point of a successful demonstration at the ADASS conference and progress toward an instrumentation of the I/O performance of the AIPS++ code.
Harold Ravlin has been heavily involved with system administration during the last quarter. Much of our local computer systems were used during the ADASS conference and Harold was in charge of setting the systems up and returning them to normal operation at the conclusion of the conference. Harold spent a small amount of time with Aipsview consulting. Harold investigated getting Aipsview to display on systems with TrueColor visuals. The conclusion was that a robust solution is possible but probably prohibitive in terms of time since Aipsview is not currently being developed.
Doug Roberts spend much of the first part of the quarter preparing for the parallel AIPS++ demonstration at the ADASS conference. He wrote a perl script to submit a Glish script to the batch queue, including the user's environment. With the additions that Wes Young made to the object catalog, the user now sees only a minor difference between an interactive and batch system.
Doug has also installed AIPS++ on a linux cluster at the UIUC CS department and consulted with the Pablo group to enable the instrumentation of the AIPS++ I/O system. So far we just have preliminary results. We have the instrumented code on the Origin 2000 and are accumulating statistics on large data processing. Results on a variety of problems sizes will be obtained in the first quarter of 1999. Further information is available in the Parallel section.
Local project members: Ger van Diepen (GVD), Michael Haller (MXH), Jan Noordam (JEN, local manager).
Michael Haller will leave NFRA at the end of March 1999. Until then, he will continue writing Glish applications for the TMS project. In addition to his role as local AIPS++ system manager, he will participate in the weekly testing of AIPS++.
Bob Campbell (postdoc at NFRA) has made little progress to his work on ionospheric calibration, due to his greater involvement with VLBI work in Bonn.
| Global AIPS++ | Local AIPS++ | Other | |
| Ger van Diepen: | 70% | 10% | 20% |
| Michael Haller: | 10% | 30% | 60% |
| Jan Noordam: | 20% | 50% | 30% |
The WSRT DZB/TMS 10MHz system (i.e. using the first of 8 DZB units) has been declared operational by 19 Dec. It relies heaviliy on AIPS++. The on-line system TMS uses tables, Measures and Glish, and off-line determination of basic instrumental parameters is done in a local AIPS++ 'mini-package' (see below). Much remains to be done.
A clear split has been made between the sofdtware needed for the '10MHz' and the '20MHz' systems. The former are now all available in the well-tested 'mstool' set of applications, which include the uv-data visualiser. The remainder has been implemented in the new 'uvbrick' set of applications, but not yet tested 'in the field'. Before March, all WSRT setup and inspection applications will be based on uvbrick.
A few small changes to ms2scn and fits2ms have been made.
This has been reported separately by GVD above.
TMS Table Browsers: Work on this began in the 3rd quarter. These are mostly finished except for maintenance updates etc.
TMS Specification application: Initial design including a few meetings to discuss this.
On-going administration of aips++ system. On average 4 hours a week. Quite a lot of work here due to upgrade to use of egcs cmpiler. Also a number of (ongoing) changes to the AIPS++ system in Westerbork.
Upgrade of aips++ system on W'bork and JNoordam's laptop. Upgrade/maintenance of linux systems. Aips++ support to several aips++ users. Linux installation help to sysadmin. Help sysadmin is determining problems with a number of installed tools/applications (eg, broken gcc, etc).
The core NRAO AIPS++ group is now Barnes (100%), Cornwell (100%), Garwood (100%), Holdaway(100%), Kemball(100%), Marson (100%), McMullin (100%), Schiebel (100%), Weatherall (50%), and Young (100%, funded by NCSA-NRAO collaboration).
Peter:
1. Received comments from testers about MSSummary format. Made extensive changes to code after consultation with Athol to reflect received suggestions - most accomodated. MSSummary now has 2 main modes of operation, a "verbose" mode which gives more details about the various tables in a Measurement Set, and a "terse" mode which gives fewer details. Both are formatted to read well on the page or screen. Also, I enhanced the code in several other respects to make it more conformant with aips++ recommended style. Kept learning more things about using aips++ libraries and modules.
2. Wrote new class MSLister to produce text-based data listings, based on MSSelector class and what I learned from revising MSSummary. Had to spend quite a bit of time learning more about the existing libraries and how to use them, especially MSSelector and the class MSRange which was split off from it, as well as compile and debug issues, scoping and formatting, etc. Got a nicely working version which will print out all data (default) or data restricted to a given time range from a Measurement Set.
3. Began working on a DisplayLibrary design memo. Using discussion with Athol and David B, enums in MSRange, and MSdefV2 as sources of inspiration (besides my own experience with numerous plotting and astro software packages). Gave Athol a rough draft before leaving for xmas hols. Main inspiration was: for coders, plot styles should be organised according to the underlying dimensionality of the plots, ie spectra and time series are intrinsically 1D, maps etc are 2D, and cubes are cubes. Then particular plot applications become simple extensions of basic plot styles from these families, making the code useful and easily extentible for astronomers, but keeping an elegant OO architecture for coders.
Tim did the following:
Bob Garwood's primary responsibility is to oversee and contribute towards the single dish work in aips++. This work remains focussed on the "dish" environment and the support of the GBT. His contribution in support of the GBT is primarily through the GBT fillers which convert the GBT FITS data files to an aips++ MeasurementSet.
Over the past 3 months (October through December) he has done the following:
Mark Holdaway has written new primary beam/voltage pattern classes which permit the construction of the beams from a number of expandable generic types (Airy pattern, polynomial, etc). Currently, the beams are restricted to be rotationally symmetric, but classes have been left for full 2-D beam images. Likewise, the possibility of different beam patterns for different antennas is also left open. All beams can deal with beam squint. Tests of removing the effects of beam squint from Stokes V and Stokes I data are under way.
In addition, Mark is nearly finished with writing and testing an implementation of the Cornwell and Evans maximum entropy deconvolution algorithm. The polymorphic entropy class currently permits both maximum entropy and maximum emptiness algorithms.
Athol writes:
``In the area of code development, I have continued to implement expanded calibration capabilities in the existing synthesis package. This has included the design and implementation of a new calibration table format, and the implementation of cross-calibration capabilities. I have modified the calibrater DO interface and documentation accordingly and have spent time testing the new calibration features. Significant effort has also been expended on calibration design for current and future development in this area.
``I have worked with a small group to coordinate calibration implementation in single-dish and synthesis, to maximize the overlap between these two efforts, in order to allow joint calibration and calibration transfer. Two meetings have been held in this regard during this quarter. The MS v2.0 data format has been revised slightly to accommodate improved single-dish data handling.
``:I have continued my normal coordination responsibilities in general synthesis development and AIPS++ parallelization. The work in these areas is described in the synthesis and NCSA sections of this report. I have undertaken coordination responsibilities in applications integration during this quarter, as a follow-up to the integration meeting held in Socorro in November. This is described in the synthesis section of the report. My current duties in the technical area and supervision have continued during this quarter, and have included work on procurement.
``I attended the ADASS'98 meeting in Champaign-Urbana, where I presented a talk on data processing models in AIPS++. I also attended the parallel object-oriented meeting ISCOPE'98 in Santa Fe in early December.''
Ralph writes:
``This has been a quite quarter for me because:
``The major thing I worked on this quarter was the Lattice Convolver and the Lattice based clean. The Lattice Convolver is finished and if test programs and documentation were up to scratch could be code copped.
``The Lattice based clean is 80% finished but not checked in yet. This will be completed (excluding test programs and documentation) after we declare a stable release as it involves overhauling the base classes for deconvolution.
``I also spent two weeks this quarter with issues related to the beta release. In particular the mathematics module was split into seperate files and the documentation overhauled. I also undertook some of the testing responsibilities.
``Two weeks was also spent on the AIPS++ integration meetings and associated discussions.
``I spent one week visiting the Navy Protype Optical Interferometer in Arizona. This is instrument is operated by the Naval Research Laboratory (NRL) and they are interested in using AIPS++ for data reduction. During this week I installed the latest AIPS++ beta release (ie., binaries) and a recent code snapshot (ie., source and binaries). This arraingement gives them a stable platform to evaluate AIPS++ as well as a mechanism for compiling code and linking their own code into AIPS++.
``NRL currently has some software for reducing data from their instrument. This software relies on the commercial product PV-WAVE and there is concern for the longer term viability of this development path. The current stratagy is for them to evaluate AIPS++ over the next few months and over the next few years migrate their current software to AIPS++.
``During this week in Arizona I gave a talk on AIPS++ to a group of about 20 astronomers at the NRL optical telescope near Flagstaff.''
Joseph McMullin's main responsibilities are managing the Green Bank installation of AIPS++, supporting AIPS++-related GBT needs, and work on single dish applications within AIPS++.
Darrell Schiebel implemented all of the Glish changes and additions described in the section on progress in the last quarter.
In addition to all of the Glish changes, Darrell worked on a few other items. He wrote the GBT observer table parser which Rick Fisher drives with his GBT Observer interface. This Glish client parses the GBT observer table language which Rick developed, and returns Glish records which are arranged for easy use by Rick's scripts.
Following the departure of Jeff Uphoff, Darrell worked on a number of AIPS++ system/clean up items:
Kate was responsible for all changes made to the AIPS++ Web pages.
of IO statistics from distributed objects. We ran some test cases with imagertest. We expect to collect results from processing the M33 data.
In this section, we give the names of people in the various AIPS++ groups and the nominal fraction of time allocated to AIPS++. These are as expected for the next quarter (and thus do not include the expected replacements for Briggs or Uphoff, or the new hires in Socorro).
The ATNF group is: Neil Killeen (75%, also local manager), Wim Brouw (20%), Mark Wieringa (30%) and David Barnes (75%).
The BIMA/NCSA group is: Harold Ravlin (5%), Doug Roberts (50%).
The NFRA group is: Ger van Diepen, Michael Haller, and Jan Noordam. The allocations are:
| Global AIPS++ | Local AIPS++ | Other | |
| Ger van Diepen: | 70% | 10% | 20% |
| Michael Haller: | 10% | 30% | 60% |
| Jan Noordam: | 20% | 50% | 30% |
The NRAO group is: Tim Cornwell (100%), Bob Garwood (100%), Mark Holdaway (100%), Athol Kemball (100%), Ralph Marson (100%), Joe McMullin (100%), Darrell Schiebel (100%), Kate Weatherall (50%) and Wes Young (100%, funded by NCSA-NRAO collaboration). We have one open position in Green Bank. Kumar Golap of MRT is scheduled to join the Socorro group in early April. Kumar will work on optimization and parallelization of wide-field imaging.
Thus, in aggregate, we have 18 people contributing about 13 FTEs to the AIPS++ Project. Of these, 9 are employed by NRAO (one funded by NCSA), and contribute 8.5 FTEs. The numbers for the other partners are: ATNF 4 and 2, BIMA/NCSA 2 and 0.55, NFRA 3 and 1.9. There are three unfilled positions, one at BIMA/NCSA (Briggs) and two at NRAO (GBT scientist, parallel applications developer: Kumar Golap due in April).
