Question:

I have bunch of common functions like swap, min, max etc. How to mark them so I could call them both from the host and from the device?

You use __host__ __device__ before the function declaration, like:

__host__ __device__ int min(int A, int B) {return (A<B)?A:B;}

This is documented in the C programming guide <a href="http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#function-type-qualifiers" rel="nofollow">here</a>.