Basic algebra with Numpy

I’m having a bit of trouble with some basic numpy operations in gufunc and cuda device functions. If S is a (n,2) array and P a (1,2) array, how can I effectively do :
.A = S - P (S[:] = S - P doesn’t work since S and P don’t have the same size)
.A = S rotated by an angle theta (I can’t create a rotation matrix and use since I can’t use np.array)

I tried to implement my own functions but the result is now slower than a regular @jit with a 1070.


Related discussion on Gitter: numba/numba - Gitter