Newsgroups: comp.lang.lisp
Path: cantaloupe.srv.cs.cmu.edu!das-news.harvard.edu!news2.near.net!MathWorks.Com!yeshua.marcam.com!zip.eecs.umich.edu!newsxfer.itd.umich.edu!gatech!nntp.msstate.edu!saimiri.primate.wisc.edu!aplcenmp!hall
From: hall@aplcenmp.apl.jhu.edu (Marty Hall)
Subject: Re: Fannkuch revisited or Benchmarking is hard
Message-ID: <Cwqt17.KA8@aplcenmp.apl.jhu.edu>
Organization: JHU/APL AI Lab, Hopkins P/T CS Faculty
References: <KANDERSO.94Sep23150911@wheaton.bbn.com>
Date: Mon, 26 Sep 1994 15:15:07 GMT
Lines: 30

In article <KANDERSO.94Sep23150911@wheaton.bbn.com> kanderso@wheaton.bbn.com (Ken Anderson) writes:
[...]
>This benchmark came out of a thread on comp.lang.lisp in Sept 1994
>originated by Bruno Haible (haible@ma2s2.mathematik.uni-karlsruhe.de).  The
>original post introduced the language Beta to the news group and in passing
>mentioned an "integer hacking" benchmark that indicated that at least some
>Lisp implementations were much slower (50 to 100 times) on the benchmark
>than C.
[...]
>ANALYSIS:
>
>Good benchmarking is harder than you think. 
>
>While the original Lisp version of the fannkuch function is about 10 times
>slower than optimized C, the optimized version takes 1.75 times the
>optimized C time, which is comparable to the time for unoptimized C.

Good going, Ken. This is a truly helpful and informative posting.

Instead of continual reruns of Lisp-vs-C wars, a thread on optimizing
would be great. Ken's example is relevant for the expert hacker, but I
think that it would also be helpful to give some simpler examples of
code that beginning/intermediate programmer's had trouble with, and
what benefits resulted from certain changes. 

If I get a chance in the next couple of days, I'll post a couple of
examples from students and from projects at work.

						- Marty
(proclaim '(inline skates))
