Comments on: Haskell Performance: Array Creation

By: tim

tim — Tue, 31 Mar 2009 05:32:44 +0000

@an: You’re absolutely right — even a standard -ccopt -O2 becomes faster when you zero out the array in a for loop. I’m amazed at the difference between that and Array.fill. Updated the original article, thanks for the info!

By: an

an — Mon, 30 Mar 2009 19:01:37 +0000

For what is worth, OCaml is faster than the fourth Haskell version
if you remember to compile with -unsafe -nodynlink -inline 100, use a
pre-allocated array and zero it out for each
line (like in Haskell), with for i = 0 to 255 do freq.(i) <- 0 done. It doesn’t take hours to
get there ;)

By: tim

tim — Mon, 30 Mar 2009 17:58:07 +0000

@dons: Well, I use a STUArray in the innermost loop… it converts to a UArray on return but from what I can tell in the docs this should be an in-place modification, no copy necessary.

By: dons

dons — Mon, 30 Mar 2009 06:17:38 +0000

Did you consider just using a mutable array (e.g. STUArray) ?
These are the primary array structure with identical behaviour/representation to what you’d use in an imperative language. They should have identical performance.