smith@ANL-MCS.ARPA (Brian T. Smith) (09/19/86)
[I sent this out once before; however, it is unclear whether anyone actually saw it or not, probably due to the massive local changes here at Purdue to support domains. ---Rsk] We have had VADS on a 12 processor Sequent 8000 since March of this year. Within a week or so, it will be running on a Balance 21000 (24 processor Sequent machine). We originally ordered the multiprocessing version of VADS but so far have not received it. Currently, we are evaluating the product using simulated multitasking. We were originally a Beta test site for the VADS product but have now purchased the product for our Advanced Computing Research Facility (ACRF). As a supported product in that facility, we have proposals from several outside researchers to use it but these projects are just starting at this point. Our experience (Ken Dritz and myself) has been good. We have found a few bugs which have been reported to Sequent and Verdix (I cannot give you any information on how prompt they been in addressing them as we have had some difficulty in establishing our maintenance contract -- the reports have only last week reached the appropriate company rep, Mary Kay, and just today I received mail indicating that they are looking into them). We have found a few bugs in their documentation and support tools as well but so far none of these have been serious. Overall, we are impressed with the quality of the compiler and development system from such a new language, system, and in some respects company. We have tested the VADS products on several hundred lines of our own code and several thousand lines of code from various sources. It is reasonably fast at compilation and the diagnostics are good in the sense that they point to the appropriate violation in the language reference manual. As we are essentially learning the language, we find it difficult to understand the diagnostics in many cases. However, VADS is very helpful in many respects. One caveat -- neither of us can be considered experts in Ada. We have the product for essentially two reasons -- investigation of the multitasking capabilities of Ada, particularly compared with the multitasking capabilities in other languages we have in the ACRF (mainly C and Fortran) and investigation of Ada as a language to write numerical software, particularly in collaboration with colleagues on the Ada Europe Numerics Working Group. These projects are just starting. Brian T. Smith smith@ANL-MCS.ARPA