[Haskell-cafe] Re: Memoizing longest-common-subsequence

[Bulat Ziganshin]

Hello Alistair,

Wednesday, August 23, 2006, 1:43:30 PM, you wrote:

> I've found the folder in which I did some of this testing, and GNU diff
> has no problem with the input files; they're only 7M, My program spends
> 70% of its time doing String-IO (so 30% in the algorithm), and peaks at
> about 350M of memory (which seems quite high, but then it does convert
> the String input into STArrays). The algorithm itself takes about 18secs
> on this input (wall-clock time), but the profile says a lot less CPU
> time is used.

> I think using ByteStrings would be a big improvement; maybe I'll find
> time to try that later.

as variant, you can try streams library - it's several times faster
for line-oriented I/O. another solution that can also speed up your
code, although not so much, is to use hGetContents and then 'lines'

http://haskell.org/haskellwiki/Library/Streams
http://www.haskell.org/library/StreamsBeta.tar.gz

-- 
Best regards,
 Bulat                            mailto:Bulat.Ziganshin at gmail.com