None whatsoever apart from to say that it's acceptably fast for whatever data set I happen to throw at it. Another poster here mentioned in an email that he'd tried it against arrays of ~50000 object structures and it was fine.