Newsgroups: comp.parallel
From: mjr@netcom.com (Mark Rosenbaum)
Subject: Re: SMP vs. MPP
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
Date: Mon, 16 Jan 1995 12:52:59 GMT
Message-ID: <mjrD2GEz4.7xK@netcom.com>

In article <3et4h3$ol8@runner.ccr-p.ida.org>,
David desJardins <desj@ccr-p.ida.org> wrote:
>Mark Rosenbaum <mjr@netcom.com> writes:
>> Finally the $64K question. Does anyone have an algorithm that 
>> a) actually users writable shared memory for anything other than locks.
>> b) is faster than other algorithms that don't.
>
>I'm not sure I understand the question.  Generally programs, not
>algorithms, use shared memory.  But anyway, on the Cray T3D writing into
>remote shared memory is generally the fastest method of communication.
>So it gets used a lot.
>
Given the number of questions about this statement it was obviously worded
very poorly.

I was refering to Concurrent Read Concurrent Write (CRCW) instead of the more
common Concurrent Read Exclusive Write (CREW). CRCW would be several processors
all writing to the same location at once.

I believe alogroithms can require shared memory.


Mark Rosenbaum				Director of Engineering
mjr@netcom.com				Otey-Rosenbaum & Frazier
(703) 536-9464				Consultants in High Performance 
Washington D. C. 			and Scalable Computing 
metro area				and Applications


