You could add that the difference with Numpy comes when you have an expression like x * x * x. Numpy would create a temporary array, compute the first operation using this temporary, and then doing a second scan to compute the second operation.def sqrv(x): Â Â Â Â return x * x I didn't know pythran would produce an element-wise iteration over the whole array from this code.That's valid numpy, nothing specific to Pythran.
Pythran would do only *one* iteration over the array. Mehdi