Why is numpy.power 60x slower than in-lining?

Question

Maybe I\'m doing something odd, but maybe found a surprising performance loss when using numpy, seems consistent regardless of the power used. For instance when x is a random 10