Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!lll-crg!lll-lcc!unisoft!paul
From: paul@unisoft.UUCP (Paul Campbell)
Newsgroups: net.arch
Subject: Re: Speed is the one true performance metric
Message-ID: <161@unisoft.UUCP>
Date: Sat, 8-Nov-86 23:29:49 EST
Article-I.D.: unisoft.161
Posted: Sat Nov  8 23:29:49 1986
Date-Received: Sun, 9-Nov-86 05:59:07 EST
References: <340@euroies.UUCP> <1989@videovax.UUCP> <798@spar.SPAR.SLB.COM> <3576@utcsri.UUCP>
Reply-To: paul@unisoft.UUCP (Paul Campbell)
Organization: UniSoft Systems; Berkeley, CA
Lines: 25

In article <3576@utcsri.UUCP> greg@utcsri.UUCP (Gregory Smith) writes:
>
>This is silly. Broken computers don't give wrong answers. They crash,
>or they log soft errors, or they act flaky. It is almost impossible to
>imagine a hardware fault that would have no visible effect other than
>to make the 'value' (whatever it may be) of the output wrong.
>

	I guess you asked for it ..... here is my war story. At a previous
job our Burroughs 6700 (remember those? they took up a whole room, about as
fast as a 780) had a problem ... one of the people using the stats packages
checked his work by hand (I guess he didn't trust the machine) turns out that
sometimes it was wrong (but always consistantly wrong for the same input data).
It took months to track it down to the MOD instruction that gave wrong answers
only for some input values. Of course the engineer didn't beleive us until we
could give him really hard proof. He said something like


	"This is silly. Broken computers don't give wrong answers"    (:-)




		Paul Campbell
		ucbvax!unisoft!paul