Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!decuac!cvl!harwood
From: harwood@cvl.umd.edu (David Harwood)
Newsgroups: comp.arch
Subject: Re: comp.arch.par - Free gifts- Read
Message-ID: <2311@cvl.umd.edu>
Date: Wed, 1-Jul-87 10:44:43 EDT
Article-I.D.: cvl.2311
Posted: Wed Jul  1 10:44:43 1987
Date-Received: Thu, 2-Jul-87 05:25:36 EDT
References: <3810041@nucsrl.UUCP> <232@hubcap.UUCP> <7278@mimsy.UUCP>
Reply-To: harwood@cvl.UUCP (David Harwood)
Distribution: world
Organization: Center for Automation Research, Univ. of Md.
Lines: 29

In article <7278@mimsy.UUCP> lls@mimsy.UUCP (Lauren L. Smith) writes:

[...]

>I think that comp.hypercube will be very useful, especially since there
>are several hypercube machines on the market and there is alot of interest
>in various aspects of hypercubes - software, architecture and algorithms.
>
>I think there is need for comp.arch.par to discuss other parallel 
>architectures, including current "paper ones".  I would like to see 
>discussion on the Connection Machine, Butterfly/Monarch, etc...

	We have a couple of parallel machines in our lab - a 128 processor
BBN "butterfly" (banyan switch) shared memory MIMD machine, and a smaller,
fault tolerant MIMD machine for interface. These are linked with a dozen 
assorted Vaxes, Suns, Symbolics, Grinnells, and Vicom image processor. These 
are used in development of parallel algorithms for image analysis and robotic 
navigation. I'll be using the "butterfly" in one project for development 
of a concurrent hybrid logic programming system in the Fall. We also will 
soon have a Connection machine and Warp systolic array processor which might
be integrated usefully. I would very much like to hear about mixed 
architectures for similar complex symbolic/numerical applications before
we start development here.

David Harwood
Computer Vision Lab
Center for Automation Research
University of Maryland, College Park 20742
harwood@cvl.umd.edu