This is low priority, but would it be possible to support numpy.gradient (and to a lesser degree numpy.diff ). Those are easy to write in terms of simple arithmetic operations and slicing, but for performance it would be beneficial to have a single kernel for these operators. In addition, numpy does not include an out argument for these operators, but it would be beneficial to have one.
This is low priority, but would it be possible to support numpy.gradient (and to a lesser degree numpy.diff ). Those are easy to write in terms of simple arithmetic operations and slicing, but for performance it would be beneficial to have a single kernel for these operators. In addition, numpy does not include an
out
argument for these operators, but it would be beneficial to have one.