It would be useful to implement a version of the propagator using OpenCL for systems where CUDA is not availiable.
