[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Gcl-devel] Re: Shootout
From: |
Camm Maguire |
Subject: |
[Gcl-devel] Re: Shootout |
Date: |
26 Feb 2006 15:51:04 -0500 |
User-agent: |
Gnus/5.09 (Gnus v5.9.0) Emacs/21.2 |
Greetings, and thanks for the pointer to the shootout!
We have been approached for a GCL entry therein, and I intend to work
on this once 2.7.0 is finalized.
What I would really love is a set of concise benchmarks that point out
critical performance characteristics of the existing applications,
acl2/nqthm, maxima and axiom. This way I can tell quickly when I've
spoiled the performance, etc. I have a benchmarks subdirectory for
this purpose. Right now it has the gabriel benchmarks, one of which
has your name attached. Are these still key? For example, GCL does a
poor job on flet right now -- there is no inlining, and a compiled
closure is built -- primarily because it is not critical to anyone as
far as I can see. It should be quite simple however to replace with a
(funcall (lambda ...)) at the lisp level for full inlining, though
I've not tested this.
Take care,
Robert Boyer <address@hidden> writes:
> > If anyone ever comes across an example function which shows this weakness
> > clearly, I'd be most interested.
>
> I have never built or used OCAML so you can distrust practically anything I
> might say about it, but the results posted at the Debian "shootout" are
> remarkable to me, considering that OCAML is a version of ML.
>
> The following URL is one showing OCAML quite a bit better than SBCL on a
> prime sieve.
>
>
> http://shootout.alioth.debian.org/gp4/benchmark.php?test=all&lang=ocaml&lang2=sbcl
>
> A lot of languages and implementations are compared, as you can see at:
>
>
> http://shootout.alioth.debian.org/old/benchmark.php?test=all&lang=all&xfullcpu=1&xmem=1&xloc=1&ackermann=1&ary=3&echo=5&except=1&fibo=2&hash=1&hash2=4&heapsort=4&hello=0&lists=3&matrix=3&methcall=5&moments=2&nestedloop=0&objinst=5&prodcons=1&random=3®exmatch=4&reversefile=4&sieve=4&spellcheck=4&strcat=2&sumcol=3&wc=3&wordfreq=5&calc=Calculate
>
> OCAML seems to be up there with GCC and Intel's C compiler, considerably
> better than CMU, SBCL, Scheme, and several Lisps I don't know about.
>
> I don't see any GCL entries there today. I seem to recall, there were some a
> few months ago in this vicinity. Or maybe I'm just confusing it with CMU. A
> nice number of examples of short little programs to play with, e.g.,
>
> ;;; The Great Computer Language Shootout
> ;;; http://shootout.alioth.debian.org/
> ;;; contributed by Dima Dorfman, 2005
>
> (defun nsievebits (m)
> (declare (fixnum m))
> (let ((a (make-array m :initial-element 1
> :element-type 'bit)))
> (flet ((clear (i)
> (loop for j fixnum from (+ i i) to (1- m) by i
> do (setf (aref a j) 0))))
> (loop for i fixnum from 2 to (1- m)
> when (= 1 (aref a i)) do (clear i)
> count (= 1 (aref a i))))))
>
> (defun test (n)
> (let ((m (* 10000 (expt 2 n))))
> (format t "Primes up to~T~8<~d~>~T~8<~d~>~%" m (nsievebits m))))
>
> (defun main ()
> (let* ((args #+sbcl sb-ext:*posix-argv*
> #+cmu extensions:*command-line-strings*
> #+gcl si::*command-args*)
> (n (parse-integer (car (last args)))))
> (when (>= n 0) (test n))
> (when (>= n 1) (test (- n 1)))
> (when (>= n 2) (test (- n 2)))))
>
>
>
--
Camm Maguire address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens." -- Baha'u'llah
[Prev in Thread] |
Current Thread |
[Next in Thread] |
- [Gcl-devel] Re: Shootout,
Camm Maguire <=