Path: utzoo!mnetor!uunet!cos!smith
From: smith@COS.COM (Steve Smith)
Newsgroups: comp.lang.misc
Subject: Poor Algorithms (was Re: First Languages (yet again) )
Message-ID: <1010@cos.COM>
Date: 24 Feb 88 23:40:01 GMT
References: <4022@ames.arpa> <170500013@uiucdcsb>
Reply-To: smith@cos.UUCP (Steve Smith)
Organization: Corporation for Open Systems, McLean, VA
Lines: 27

In article <170500013@uiucdcsb> robison@uiucdcsb.cs.uiuc.edu writes:

>Some friends and I conjecture that a lot of supercomputer time 
>is spent running poor algorithms.  Anyone have any empirical
>evidence or anecdotes?

>Arch D. Robison


Well, a DEC 10 ain't a supercomputer, but the point is still valid.

A friend of mine ran into this one while he was working in a university
computer center.  Seems they had an Oceanography professor who did a lot
of flow modeling.  Like most such programs, his used a big xyz array to
handle flow elements.  As it turned out, one xy plane fitted nicely into
one virtual memory page.  Unfortunately, for no good reason, he had
chosen to scan it first in the x direction and second in the z
direction.  Result - page faults up the wazoo!  When the people at the
computer center pointed this out and suggested changing the scanning
order, he got downright abusive.  After all, those geeks in the computer
center know nothing about Oceanography!

For some reason (:-), his priority tended to get bumped down.
-- 
                -- Steve
(smith@cos.com)    ({uunet sundc decuac hqda-ai hadron}!cos!smith)
"Truth is stranger than fiction because fiction has to make sense."