Re: [Gcl-devel] Boyer benchmark results

[Vadim V. Zhytnikov]

Camm Maguire writes:

<snip>


>         b) Shortly after writing this, I will put up at 
>
>                 http://people.debian.org/~camm/gabriel.tgz
>
>            the benchmark suite I used with the slight modifications as
>            mentioned on the web page.  I'd be most appreciative if
>            those in the know could run this suite as well, as well as
>            look it over for mistakes/oversights.

Please do it.  I'd like to repeat your benchmarks on
the Athlon XP 300.  Maybe some tests under considiration are
strongly CPU dependent (Pentium 4 vs Athlon).  At leat I know
some other lisp tests (Reduce CAS test with Portable Standard
Lisp) which dislikes Pentium 4 and favor Pentium III and
especially Athlon - Pentium 4 is 2-3 (!!!) times slower.

<snip>

> Peter Graves <address@hidden> writes:
>
>
> > Hi,
> >
> > There was a little discussion on #lisp yesterday about the benchmark
> > results on this page:
> >
> >    http://people.debian.org/~camm/GCL_2_6_2_tests.html
> >
> > To add another data point, I ran the Boyer benchmark on GCL CVS head
> > (built with --enable-ansi) and several other implementations, using an
> > Athlon XP 2100+ machine running Linux 2.6.0, with the results below.
> >
> > The benchmark I used was extracted from Eric Marsden's cl-bench suite
> > and is available here:
> >
> >    http://armedbear.org/boyer.lisp
> >
> > I added (declaim (optimize speed)) at the top of the file (and turned
> > off *gc-verbose* for CMUCL).
> >
> > For each system, I simply did (compile-file "boyer.lisp"), (load *),
> > and then (test 100) 3 times.
> >
> > GET-INTERNAL-RUN-TIME was used for timing.
> >
> > The results (best of 3 runs for each system):

I confirm these results on Linux 2.4.26,
Athlon XP 3000+, 512Mb

> >    GCL (GNU Common Lisp)  2.7.0 ANSI   Jun 17 2004 17:50:11
> >
> >        3.22 seconds

GCL 2.6.2 ANSI  -  2.78 seconds

> >    CMU Common Lisp CVS release-18e-branch + minimal debian patches
> >
> >        2.87 seconds

CMUCL 18e  -  2.26 seconds

> >    SBCL 0.8.11
> >
> >        2.592 seconds

SBCL 0.8.8  -  2.54 seconds

> >    Allegro CL 6.2 Trial Edition
> >
> >        3.79 seconds

ACL 6.2 Trial  -  3.79 seconds (?)

> >    CLISP 2.33.1 (2004-05-22)
> >
> >        18.059256 seconds


CLISP 2.33.1  -  15.48 seconds

> >    ABCL 0.0.3.15+ (Jun 17 2004 12:21:53) (Java 1.5.0-beta2)
> >
> >        16.786 seconds
> >
> > -Peter


--
     Vadim V. Zhytnikov

     <address@hidden>