Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!randvax!news
From: edhall@rand.org (Ed Hall)
Newsgroups: comp.unix.i386
Subject: Re: hardware costs, again...
Message-ID: <1990Jun8.205849.582@rand.org>
Date: 8 Jun 90 20:58:49 GMT
References: <136279@su <1990May29.031904.23465@rfengr.com> <25859@bcsaic.UUCP> <1990Jun8.083211.22300@pegasus.com>
Sender: news@rand.org
Reply-To: edhall@rand.org (Ed Hall)
Organization: The RAND Corporation, Santa Monica, CA
Lines: 19

In article <1990Jun8.083211.22300@pegasus.com> richard@pegasus.com (Richard Foulk) writes:
> . . . .
>And I'm not sure the SLC would "blow the doors off any 386..."
>Where'd you get your data?  My 33MHz 386 is noticeably faster than a
>SparcStation 1 on integer only tasks -- I don't think the networking
>code does much floating point...

The SPARCStation, at least, doesn't have an integer multiply
instruction (though it does have a ``multiply step'' instruction to
make it easy to code a multiplication subroutine).  This skews
benchmark results, since unlike many benchmarks most system software
requires few multiplies except for subscript calculations, and Sun's
compiler does a pretty good job of strength-reduction and other tricks
to eliminate these.

I suspect the SLC is the same.

		-Ed Hall
		edhall@rand.org