[Haskell-cafe] Haskell performance question

[Xiao-Yong Jin]

"Dan Piponi" <dpiponi at gmail.com> writes:

> My wasn't intended to represent the problem that I'm trying to solve,
> but the approach I want to take. The problems that I do want to solve
> don't lend themselves to this kind of approach.
>
> My real situation is that I want to write code that has both a
> high-level component and a low-level number-crunching component that
> works on large dense and sparse arrays. Idiomatic Haskell is great for
> the high-level component. But the question is whether or not its worth
> trying to write the low-level code in Haskell or whether I should just
> write that code in C and use the FFI. 

I've written a bit number-crunching code in haskell.  My
experience is that usually the haskell code runs about 2 to
5 times slower -- It depends on whether I use the best
optimisation scheme or not -- I do not know if I am doing
the correct thing most of the time, since there is not much
literature on-line about numerical computing in haskell, or
in FP in general.  There is one in ocaml, though.

> There are still advantages to using Haskell even if the
> code is highly unidiomatic: you have the freedom of
> throwing in higher order functions from time to time in
> the low-level code, and writing mixed-language code is a
> pain.  -- Dan

I fully agree.

Xiao-Yong
-- 
    c/*    __o/*
    <\     * (__
    */\      <