pow function error with 1 float, 1 int input

In a function running on the device:

pow(3.0f, 2);
returns 3.0

pow(3.0f, 2.0f);
returns 9.0

Fixed in 0.9?

This bug (Linux only) is still in 1.0, unfortunately, although it’s a really easy fix: Lines 515 and 533 in CUDA\include\math_function.h should be changed from “r = r * r;” to “a = a * a;”.

Needless to say it is going to get fixed in the next release ;-)