Re: Troll-o-Matic
- From: "bradb" <brad.beveridge@xxxxxxxxx>
- Date: 16 Feb 2006 10:40:25 -0800
funkyj wrote:
<snip>
====
Going a little off topic (or perhaps going more on topic) ...
http://www.ffconsultancy.com/free/ray_tracer/languages.html
I recently read about the benchmark above and boy did ocaml (and
stalin) kick Lisp's ass 3 ways to sunday.
You'd think Lisp could at least be competitive with Java on the
verbosity front if not beat it on performance. Is ocaml really that
much better than lisp? What major language features, if any are
present in CL and ocaml and not in the other (e.g. macros)?
Also, what is the secret of Stalin's success? Perhaps SBCL can be
converted to despotism and reap similar rewards.
I suspect that Lisp's runtime type system is the problem here. ocaml,
Java and C++ are (as far as I am aware) largely statically typed
systems. ocaml doesn't often require type declarations as it infers
them where possible. In this extremely math heavy example, static
typing is a big win. Stalin is an agressive Scheme compiler that does
whole program analysis (that is presumably why it takes so long for
even a trivial example), so I suspect that Stalin is removing all
runtime type checking because that is possible in this example.
I suspect the Lisp speed could be further increased at the cost of
verbosity.
Also, you should be aware that the author of the study shows a bias
toward statically typed languages - I don't know if that will skew
results.
But yes, in this floating point intensive example, Lisp (at least SBCL)
gets beaten like a red-headed stepchild.
Cheers
Brad
.
- Follow-Ups:
- Re: Troll-o-Matic
- From: Juho Snellman
- Re: Troll-o-Matic
- References:
- Re: Troll-o-Matic
- From: Tin Gherdanarra
- Troll-o-Matic
- From: Tin Gherdanarra
- Re: Troll-o-Matic
- From: funkyj
- Re: Troll-o-Matic
- Prev by Date: Re: Troll-o-Matic
- Next by Date: Re: rotatef-if: ¿my possible first macro?
- Previous by thread: Re: Troll-o-Matic
- Next by thread: Re: Troll-o-Matic
- Index(es):
Relevant Pages
|