Re: [Caml-list] Re: Why OCaml sucks

[Kuba Ober <ober.14@osu.edu>]

On Monday 12 May 2008, you wrote:
> > Yet, if you look at things in the light of "optimization is
> > depessimization",
> > you'd much rather have easier to read code, than code which is ugly
> > because
> > you preoptimized it by hand. This is why, for me, Ocaml has a long way to
> > go
> > to make it useful for run-of-the-mill production code. My pet peev is
> > performance penalty paid for writing in functional style where it
> > actually makes sense -- say passing an arithmetic operator to a map-style
> > function.
>
> What do you mean by this?  What language would not incur this kind of
> performance hit?  Is F# able to optimize this out or were you referring to
> something else?

It's not much about the language, but about implementation. IIRC some Lisps
can do this kind of an optimization, gcc could do it as well -- whenever
the value of the argument to a function is known, a potentially
call-site-specific version of the function can be generated, and in such cases
it'd be much simpler, emitted-code-wise, than the version which explicitly
emits the arguments and calls the operator.

Cheers, Kuba