Path: utzoo!mnetor!uunet!pcrat!rick
From: rick@pcrat.UUCP (Rick Richardson)
Newsgroups: comp.arch
Subject: "Life" benchmarks
Message-ID: <438@pcrat.UUCP>
Date: 29 Dec 87 06:50:32 GMT
References: <1078@quacky.UUCP> <8809@sgi.SGI.COM> <6964@apple.UUCP> <1156@winchester.UUCP> <1295@sugar.UUCP>
Reply-To: rick@pcrat.UUCP (Rick Richardson)
Organization: PC Research, Inc., Tinton Falls, NJ
Lines: 26

In article <1295@sugar.UUCP> peter@sugar.UUCP (Peter da Silva) writes:
>Ah, benchmarks.
>
>The only non-graphics algorithm that does anywhere near this that I have
>heard of is a 20-generation-per-second 320-by-200 LIFE program.

	[ 1.2 million life pixels/second ]

This reminds me of an interesting machine I worked on circa 1977.  It
was (is?) called the Cytocomputer - a cellular image processor.  The
machine was a long pipeline of identical MSI processors.  I think
we had around 87 processors in the prototype.  The image (1024x1024x8)
was pumped through the pipeline at 1 Mhz.  The prototype required
two processors to perform one generation of life (one for deaths,
one for births).  Thus, we could do 43 generations/second, neglecting
pipeline fill time.  Thats about 45 million life pixels / second.
Of course, you couldn't see the intermediate results.  A later
design was supposed to clock at 10Mhz and would only require
one processor per life generation.

I wonder what current processors are capable of on the "life" benchmark?

-- 
	Rick Richardson, President, PC Research, Inc.
(201) 542-3734 (voice, nights)   OR   (201) 834-1378 (voice, days)
		seismo!uunet!pcrat!rick