Testing the recently posted benchmarks on my Mac II w/MACL 1.3.2 I found
(not surprisingly) that it had 10% or even less the performance of an XL
400 on most benchmarks. But there was one important example: The
cons-test, which was almost the same speed. This confirms a suspicion I
have that lisp machines in general behave badly when it comes to REAL
lisp: I.e. real recursive functions and lots of consing. Why is it so?
And more important: What do these benchmarks really measure? The posted
benchmarks didn't (as far as I could tell from my quick glance through
them) measure recursive function call performance. And the results of the
cons-test were only reported in two of the five included test series. My
suspicion is that benchmark performance corresponds to the performance of
fine-tuned destructive lisp programs, not to quickly written
non-destructive programs (which are so typical for everyday lisp
programming...).
If you are about to buy a new lisp, maybe the only reasonably safe speed
comparision test is just to measure the performance of compiling and
running your favourite program(s).
--------------------------------------------------------------
Espen J. Vestre, ·····@coli.uni-sb.de
Universitaet des Saarlandes,
Computerlinguistik, Gebaeude 17.2
Im Stadtwald,
D-6600 SAARBRUECKEN, Germany tel. +49 (681) 302 4501
--------------------------------------------------------------
In article <·····················@coli.uni-sb.de> ·····@coli.uni-sb.de
(Espen J. Vestre) writes:
> Testing the recently posted benchmarks on my Mac II w/MACL 1.3.2 I found
> (not surprisingly) that it had 10% or even less the performance of an XL
> 400 on most benchmarks. But there was one important example: The
Oops - sorry, those benchmarks weren't posted here in comp.lang.lisp!
Nevertheless, I still would like to get opinions on what a good lisp
benchmark should look like.
--------------------------------------------------------------
Espen J. Vestre, ·····@coli.uni-sb.de
Universitaet des Saarlandes,
Computerlinguistik, Gebaeude 17.2
Im Stadtwald,
D-6600 SAARBRUECKEN, Germany tel. +49 (681) 302 4501
--------------------------------------------------------------
From: Brian Monahan
Subject: Re: What do benchmarks measure?
Date:
Message-ID: <3221@m1.cs.man.ac.uk>
In article <·····················@coli.uni-sb.de>, ·····@coli.uni-sb.de (Espen J. Vestre) writes:
|> In article <·····················@coli.uni-sb.de> ·····@coli.uni-sb.de
|> (Espen J. Vestre) writes:
|> > Testing the recently posted benchmarks on my Mac II w/MACL 1.3.2 I found
|> > (not surprisingly) that it had 10% or even less the performance of an XL
|> > 400 on most benchmarks. But there was one important example: The
|>
|> Oops - sorry, those benchmarks weren't posted here in comp.lang.lisp!
Where were they posted then?
--
///////////////////////////////////////////////////////////////////////
/
Dr. Brian Monahan /
/
Department of Computer Science / E-mail : ······@cs.man.ac.uk
The University of Manchester / ···@cs.man.ac.uk
Oxford Road / Phone : (+44)-61-275-6137
Manchester M13 9PL / (New) FAX : (+44)-61-275-6236
/
///////////////////////////////////////////////////////////////////////