From: Gregory Collins <greg@gregorycollins.net>

On Mon, Apr 11, 2011 at 3:55 PM, Serguei Son <serguei.son@gmail.com> wrote:
> So if I must use a safe function returning IO a,
> there is no way to improve its performance? To give you
> a benchmark, calling gsl_ran_ugaussian a million times
> in pure C takes only a second or two on my system.

In the C version, are you also producing a linked list containing all
of the values? Because that's what mapM does. Your test is mostly
measuring the cost of allocating and filling ~3 million machine words
on the heap. Try mapM_ instead.

I've done some rough tests of a few versions, and I believe Gregory's analysis is correct. These are for calculating a sum of sines of 1..10^7 with ghc-7.0.3.

version execution time

IO, safe - 4.08s

Pure, safe - 3.07s

IO, unsafe - 2.77s

Pure, unsafe - 1.95s

For the IO versions, I used the following:

main = foldM (\ !acc n -> fmap (acc +) $ c_sin_io n) [1..10^7] >>= print

(this can possibly be improved; it was the first thing I thought of)

The most important factor (as often the case with Haskell) is to pay attention to the space properties of your algorithm. Instead of using mapM, which forces the spine of the list, it's much better to interleave the fold step with the other IO.

Note that the "acc" parameter is strict, this is necessary to avoid building a thunk.

John L.