Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!usc!samsung!emory!hubcap!grunwald
From: grunwald@foobar.colorado.edu (Dirk Grunwald)
Newsgroups: comp.parallel
Subject: Re: CrOS (Crystalline OS ?)
Message-ID: <1991Jan23.042208.19094@csn.org>
Date: 23 Jan 91 12:56:20 GMT
References: <12708@hubcap.clemson.edu>
Sender: fpst@hubcap.clemson.edu
Reply-To: grunwald@foobar.colorado.edu
Organization: University of Colorado at Boulder
Lines: 19
Approved: parallel@hubcap.clemson.edu
In-Reply-To: cherwig@hubcap.clemson.edu's message of 21 Jan 91 13:05:36 GMT

re: CrOS et al

In particular, one might note that CrOS et al follow the mindset of
the phsyics researchers at CalTech, that is, that all communication
should be either strictly local (near forces) or global (far forces).

Thus, early versions of CrOS only support communication with adjacent
neighbours and global broadcast. This was important on e.g., the
Mark-III, which was had a store and forward network, but less
important on e.g., the Intel iPSC/2 hypercube, where locality was less
important.

This is not to belittle the CrOS; it was considerably more efficient
for localized communication than e.g., Mercury, another Mark-III O/S.
(Old Ametek S-14 users will note that Mercury==mars and CrOS==whatever
it was that you actually used on the S-14). However, for other than
those special purposes, you had to ``roll your own'' multi-hop
communication primitives.