If I have an Array created using afcpu and one on say a cuda device using afcuda then I would expect the following to throw an exception (or err code in C):

af_mul(a, b); //where a is cpu & b is cuda

It currently just SEGFAULT's which I think is terrible behavior.

To do this there needs to be some sort of internal flag in the Array, eg: array.backend() & the library needs to handle this logic internally.