Nvidia documentation states that unified memory is not supported from inside a VM with vGPU (only via pass-through mode).
I am trying to understand:
- Why this restriction exists and whether or not there are plans to support UVM-vGPU in the future.
- Can an application access the nvidia-uvm device froma virtual machine at all?
- What will happen under the hood if a VM will access nvidia-uvm from one virtual machine with certainty that no other virtual machine accesses it concurrently.