Is general CUDA development permitted, or can I only use this for deep learning?
There's a simulation problem that I encounter often at work which should parallelize very well, and I was considering getting my feet wet with CUDA by writing a solver for it.
