home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.parallel
- Path: sparky!uunet!gumby!destroyer!gatech!hubcap!fpst
- From: bernardl@SC.Stanford.EDU (Bernard Lee)
- Subject: clarifications on iPSC/DASH/SGI library
- Message-ID: <1992Aug13.091742.22547@leland.Stanford.EDU>
- Keywords: clarifications, iPSC/DASH/SGI library
- Sender: news@leland.Stanford.EDU (Mr News)
- Organization: CS Dept, Stanford University, CA 94305, USA
- Date: Thu, 13 Aug 92 09:17:42 GMT
- Approved: parallel@hubcap.clemson.edu
- Lines: 36
-
- +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
- Clarifications about the status of the iPSC message-passing li-
- brary on the DASH/SGI machines.
- +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
-
- (1) We have successfully integrated FORTNET to the iPSC library,
- but that part is NOT ready to be shipped yet. An early announce-
- ment was made as part of good research etiquette so that other
- researchers can plan ahead. The matter is in the hands of the
- British researchers; the Stanford team is only supporting the in-
- terface. However, Dr. Robert Allan at the SERC Daresbury Lab
- (rja@cxa.dl.ac.uk) will still be glad to hear from you if you
- just want to send in an early registration of your interest. My
- apologies for the confusions in the earlier posting!
-
- (2) Just to follow Ms. Manner's research guidelines, perhaps we
- should give some indications on the direction that we are taking
- for the next major release. We are planning to upgrade the li-
- brary to socket-based I/O, so that it will eventually run on a
- network of 340VGX or 380VGX and simultaneously upgrade the per-
- formance on DASH. The entire idea is to compare the hardware-
- driven performance on DASH to the software-driven performance on
- a cluster of 340VGX under the same protocol. The spin-off bene-
- fit to the CS community is a socket-based iPSC library that, in
- theory, should run on any cluster of SGI's. Since there
- shouldn't be too much architecture-dependent code in the library,
- one can reasonably expect to port the code to other types of
- workstation clusters, although our group will not have the man-
- power to support other architectures. Anyone who is interested
- in supporting other architectures should feel free to send me
- mail so that I can send you early versions of our "scratch" code
- (and put everything under the same release).
-
- (3) We also welcome users to make "do-able" suggestions for the
- library, such as that of expanding the library to OSF/860.
-
-