Well, blokhead's memoize version take < 10s so that the easiest thing available as a benchmark and it's takes little effort to run it.

I managed to generate all solutions just using Perl in 74 mins on one machine and 2 hours on another. I'd like to be able to compare these results to your 72 mins without having to chew 72 mins of CPU. I just tried it on that machine and killed it after 93 mins. It was compiled with -O2. Have you got a very very fast machine? If so then I think Perl with ugly-memoize beats the C, which is very surprising.

fergal,
I compiled with the following optimizations:

gcc -Wall -march=pentium4 -s -funroll-loops -fomit-frame-pointer -O3 l
+oops.c -o loops.exe
[download]

When you say that you were able to generate all the solutions in Perl in 74 mins on 1 machine was it using blokhead's code without Memoize or your modification to that code? I ask because then we would be comparing apples and oranges. I had a completely different algorithm.

Cheers - L~R

Update: Added second hyphen to -fomit-frame-pointer

My modified code. Blockhead's used memo for counting but not for generation. I suspect that without memoize blokhead's code would take hours or days.

So we're comparing apples to oranges which is perfectly legitimate if you want to find out whether apples are faster than oranges :).

As far as I can tell my Perl is faster than your C. So (for a sanity check) if you run blokhead's memoized counting code on your machine (the machine that takes 72 mins for the C) how long does it take?