Can't find any new developments on that
> Besides, you shouldn't be using CUDA lock-in to begin with.
For some teams, not choosing CUDA is a luxury they don't have.