This gives quite the same problems as experimental CUDA kernels and for until it's found a root cause of the problem we'd just explicitly uninline the function.
This gives quite the same problems as experimental CUDA kernels and for until it's found a root cause of the problem we'd just explicitly uninline the function.