The root of optimization is finding a solid algorithm, then implementing it, then speeding the implementation. You algorithm is solid (I don't know of a faster one . . .monks?) and your implementation is simple, meaning not much room for speed gain. Offhand though, C style for loops might be slightly faster (perl does an implicit ++ and check when you call the range operator, hence the rationale that doing it directly might be faster).

Cheers,
Erik