Path: utzoo!attcan!uunet!ncrlnk!ncrcae!hubcap!wen-king
From: wen-king@csvax.caltech.edu (Wen-King Su)
Newsgroups: comp.parallel
Subject: please post this follow-up
Message-ID: <7152@hubcap.clemson.edu>
Date: 21 Nov 89 13:37:55 GMT
Sender: fpst@hubcap.clemson.edu
Lines: 26
Approved: parallel@hubcap.clemson.edu

Subject: Re: IPSC Communications

In article <7130@hubcap.clemson.edu> brooks@maddog.llnl.gov writes:
<                         It is interesting to note that the orginal message
>passing machine at Caltech was proposed as a grid, not a hypercube.  The
<Computer scientists there thought that hypercubes were prettier at the time
>and forced the physicists to learn Grey codes to get their physics problem
<mappings done.

The message-routing technology available at the time precludes the use
of grids.  The machine has to make do with a slow store-and-forward
routing that uses an 8-bytes at-a-time, interrupt-driven, programed
transfer with a 5 megahertz 8086 CPU.  Routing on a grid would be
unimaginablely slow.  The physics wanted grids because their programs
at that time fit on a grid and they do not need non-local messages.
However, the computer scientists have a broader vision and went with
cubes.  It is not until we made the first worm-hole message-routing
chip and put it to trial in Symult's machines, has grid connectivity
for MIMD message-passing machines become a superior alternative.  The
real problem now is that the term 'hypercube' has been over-sold by
the marketing people and many people have been conditioned to believe
that cube is the only viable connectivity.

/*------------------------------------------------------------------------*\
| Wen-King Su  wen-king@vlsi.caltech.edu  Caltech Corp of Cosmic Engineers |
\*------------------------------------------------------------------------*/