> Thanks. That's nice and just as fast as the original. It's interesting > that when I tried it with SML/NJ, there was a very small slowdown. It's almost certainly because SML/NJ doesn't duplicate the "f" function 6 times, and instead uses function calls.