Xref: utzoo rec.humor:18804 comp.misc:5057
Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!nrl-cmf!ukma!rutgers!apple!voder!pyramid!leadsv!laic!nova!darin
From: darin@nova.laic.uucp (Darin Johnson)
Newsgroups: rec.humor,comp.misc
Subject: Re: Looking for Computer Folklore
Keywords: sieve benchmark microsoft
Message-ID: <447@laic.UUCP>
Date: 11 Feb 89 01:40:36 GMT
References: <6507@boulder.Colorado.EDU>
Sender: news@laic.UUCP
Reply-To: darin@nova.UUCP (Darin Johnson)
Distribution: usa
Organization: Lockheed AI Center, Menlo Park
Lines: 19

In article <6507@boulder.Colorado.EDU> loughry@tramp.Colorado.EDU (J. Loughry) writes:
>	Prime Computer once had a compiler optimize their competitor's
>	benchmark down to a single NOP--and for several years they gleefully
>	used this "performance" figure in their ads.

This sort of stuff used to irk me back in college.  I wanted to see the
assembler output of a Pascal program for my assembly language class.
So I would write programs that were nothing more than assignments,
function calls, etc.  I would then compile them on some arcane system
that used the same chip as the machines in our class.  Trouble was,
when I forgot to actually use what I wrote, the compiler optimized them
away.  I did have a rather large function compile to a return
instruction.  

The optimizer wasn't all that clever though, occasionally it would
remove the useless code, but leave the surrounding loop :-)

Darin Johnson (leadsv!laic!darin@pyramid.pyramid.com)
	Can you "Spot the Looney"?