I'm going through the wiki as well as README, issues, and issues and I cannot get a clear picture regarding support for decoding/encoding capabilities of vGPU, especially when used with "unsupported" cards & vgpu_unlock.

This isn't an uncommon scenario to face, where needing multiple VMs video "mangling" support:

1. Is it even possible to use NVDEV/NVENC on virtual GPU instanced? Here, I'm referring to arbitrary stream manipulation rather than framebuffer encoding (e.g. in Looking Glass scenario)
2. How are encoding/decoding resources being divided?
3. Are there any limitations when using vGPU for decoding/encoding vs. using the host GPU directly?
4. Is this scenario documented anywhere (even on NVIDIA side), or should the wiki be updated?