Scheduling of many processes on NVIDIA GPU

I am intersted in scheduling policies of GPUs (in my case NVIDIA GPUs).
More specifically, I want to learn the scheduling policies that the GPU (Tesla v100) will schedule the many processes that many different users will apply (CUDA or Opencl).

Could you please explain me or provide me with documentations/papers/… that will describe how the GPU schedules different processes in detail?

up up up