Re: [Caml-list] Functors

[brogoff <brogoff@speakeasy.net>]

On Thu, 6 May 2004, Julien Signoles wrote:
> > In any case, the absence of a defunctorization step means that we often
> > have a choice between performance and a functorized programming style, which
> > stinks.
>
> I don't think you have a stinking choice.

> My opinion is: always choose a
> functorized programming style and, if this style significantly reduces
> performance, then use a defunctorizer like ocamldefun. For example, see
> the Jean-Christophe Filliâtre's contribution to this thread
> (http://caml.inria.fr/archives/200405/msg00087.html).

JCF said he manually defunctorized. I'd prefer not to have to butcher my code
to achieve performance, and I'd rather have the compiler do it, if it can.
If ocamldefun can do the work, then I'd rather that there is a compiler switch
to call it transparently rather than have to add more gunk to my makefiles.
Of course, if ocamldefun can do it, modifying one of the existing build
systems to add a defunctorization pass is also an option.

> > MLton began as an SML defunctorizer if I'm not mistaken, but has evolved
> > into a whole program optimizing compiler.
>
> You're not mistaken :-). See http://www.mlton.org/history.html.

So, are you planning to evolve ocamldefun into a whole program optimizing
OCaml compiler? :-)

-- Brian

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners