On Mon, 08 Dec 2003 14:07:56 +0100, Laurent Eschenauer <pepite.be at laurent> wrote:

> Did anyone try the benchmark of CL-PPCRE on ACL ?
>
> I just tried the benchmark on ACL 6.2 and CMUCL and was shocked by
> the results !  Either ACL is awfully slow or I am missing
> something. Anyone tried something similar ?
>
> ACL 6.2:
> $mlisp
> cl-user(1): (load "load.lisp")
> cl-user(2): (cl-ppcreptest:test :file-name "timedata")
>
> 1: 4.0115 (1000000 repetitions, Perl: 3.4945 seconds, CL-PPCRE: 14.0180 seconds)
> 2: 2.9020 (1000000 repetitions, Perl: 3.8404 seconds, CL-PPCRE: 11.1450 seconds)
>
> CMU-CL:
> $lisp
> * (load "load.lisp")
> * (cl-ppcreptest:test :file-name "timedata")
> 1: 0.5637 (1000000 repetitions, Perl: 3.4945 seconds, CL-PPCRE: 1.9700 seconds)
> 2: 0.4296 (1000000 repetitions, Perl: 3.8404 seconds, CL-PPCRE: 1.6500 seconds)
>
> Any insight on this would be appreciated !

Well, I developed CL-PPCRE using CMUCL and optimized it using CMUCL
compiler messages. I'll happily accept patches to make it faster on
other Lisps.

I wouldn't say that Allegro is awfully slow but rather that CMUCL is
extremely fast in some cases.

Edi.