Uses the wolfsort benchmark, which uses a time-based seed for the pseudo-random-number generator "rand". These are therefore unreproducible results (for the randomized inputs).

The results are very interesting and generally follow the rule that bigger hand-optimized code can be faster code.