Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!cornell!uw-beaver!uw-june!david
From: david@cs.washington.edu (David Callahan)
Newsgroups: comp.arch
Subject: Re: parallel systems
Message-ID: <9531@june.cs.washington.edu>
Date: 19 Oct 89 14:43:50 GMT
References: <35825@lll-winken.LLNL.GOV> <20336@princeton.Princeton.EDU> <7651@bunny.GTE.COM>
Reply-To: david@june.cs.washington.edu.cs.washington.edu (David Callahan)
Organization: Tera Computer Co., Seattle WA
Lines: 45

In article <7651@bunny.GTE.COM> hhd0@GTE.COM (Horace Dediu) writes:

>Who cares about shared memory?  Distributed is the only way to scale.

Perhaps you forgot a smiley? Or perhaps when you say "shared" you mean
"centralized"?

"Shared" memory is part of the virtual machine and clearly can be
implemented on a machine with "distributed" packaging of memory with
processors. The BBN Butterfly and the RP3 are both "distributed"
memory machines in the sense that memory is packaged with processsors
and the hardware takes care of building "messages" for every memory
request.

From a programing point of view, machines like the NCUBE have three
(IMHO) serious faults: message passing is done in software and is
therefore has orders of magnitude more latency than a "shared" memory
machine; data movement now requires software-controlled cooperation on
both processors; and finally, the programmer must determine the
location of the "most recent" value of every variable and which
processor was the last to write it or next to use it.

I care about shared memory --- its makes parallel machines much easier
to program.

>Everybody realizes this since it can be proven.
>The only reason shared memory machines exist is because we don't yet know
>how to make good distributed machines.  (Yeah, right! tell that to Ncube) 
>IMHO shared memory is a hack using available bus technology while waiting for 
>the real parallel machines to come.  (they're already here)

Shared memory has nothing to do with busses --- it has to do with
programming. 

>Horace Dediu \"That's the nature of research--you don't know |GTE Laboratories
>(617) 466-4111\  what in hell you're doing." `Doc' Edgerton  |40 Sylvan Road
>UUCP:  ...!harvard!bunny!hhd0................................|Waltham, MA 02254
>Internet: hhd0@gte.com or hhd0%gte.com@relay.cs.net..........|U. S. A.

Disclaimer: I work for a company designing a multiprocessor that
supports shared memory programming.

-- 
David Callahan  (david@tera.com, david@june.cs.washington.edu,david@rice.edu)
Tera Computer Co. 	400 North 34th Street  		Seattle WA, 98103