I don't know that you can say that in general. Some sort algorithms have worst case timings when sorting a pre-sorted (or nearly pre-sorted) list.

That being said, however, Perl may take that into account.

I do not, however, without benchmark data, buy that running sort() inside of a loop is faster than running sort outside of the loop.

Does anyone have a benchmark framework that can test inside and outside sorts with input data that is sorted, reverse-sorted, nearly sorted, and random?