Re^2: Speeding permutation counting

Yes, this is the correct interpretation of my question, and thanks for your suggestions. I benchmarked your two suggestions relative to my original, and things are definitely improved:

            Rate      orig substring  blokhead  blokchop
orig      1199/s        --      -45%      -62%      -69%
substring 2162/s       80%        --      -31%      -44%
blokhead  3144/s      162%       45%        --      -18%
blokchop  3840/s      220%       78%       22%        --
[download]

"Substring" is closer to my original (too verbose) style, so some of the gains are also coming from the more succinct, easier to read code. Best so far is the "chop" based suggestion.

Still wondering if there might be other improvements, or some vastly different way to solve this.

Thanks for these suggestions, they really have helped.
-a

Comment on Re^2: Speeding permutation counting Download Code

Replies are listed 'Best First'.
Re^3: Speeding permutation counting by ikegami (Patriarch) on Jul 18, 2007 at 15:44 UTC
Do you have the benchmark code? I consider benchmark results useless without the accompanying benchmark code. It's also stopping us from comparing our own solutions without redoing all your work. If you're worried about the space it takes, use `<readmore>` or `<spoiler>` tags.	[reply] [d/l] [select]
Re^4: Speeding permutation counting by albert (Monk) on Jul 18, 2007 at 16:23 UTC
Yeah, it was a space concern. Took things right from the solutions, and code includes other things I was testing (and your suggestion as well). Details... Read more... (8 kB)	[reply] [d/l]
Re^5: Speeding permutation counting by ikegami (Patriarch) on Jul 18, 2007 at 17:37 UTC
It's already bad enough that the code being benchmark is in a sub. Adding another sub in the call chain hurts that much more. To remove the extra sub call, use `blokchop => \&blokchop, orig => \&orig, browserUK => \&browserUK, ikegami => \&ikegami,` [download]	[reply] [d/l]