From: Tom Sawyer <transami@...>
Date: 2002-08-08T07:46:11+09:00
Subject: Re: Coding challenge (on Ruby Garden) [long]

On Wed, 2002-08-07 at 16:04, Christoph wrote:
> In a variation of an old statistics saying "there are lies, dammed lies and
> benchmarks";-).  On the other hand these types of benchmarks  do have
> their place.  If you positively have to verify that a given 1000 x 1000
> matrix is in an  ``one_in_each''  relation to a 1000 elements item array
> or more likely have to actually generate the corresponding bijection
> you'll be dammed  (i.e. you will never finish in a reasonable amount
> of time) if you asymptotically (worst case or not) slower algorithm.

speaking of which, turned out there was a bad bug in my script and after
fixing it, it was dog slow. of course i took the brute force approach
and determined every purmutation and then looked to see if any of them
fit the criteria.

well, disappointed with that. i tried my hand at a few other methods
with the goal of creating a very fast version no matter the "coding
cost". but everything i tried turned out slower. at one point i had a
thread running for every array to array-of-array node! my machine about
keeled over. :-) threads were not helpful at all. so i have been trying
some huristics. essentially reducing the problem down, after eliminating
a number of unnecessaries. still, its turned out to be quite a task and
i haven't finished yet.

just FYI.

~transami