Using a Debian 8 host, I'm interested in setting up a Windows VM guest which gets direct access to my dedicated/discrete graphics card, to see what sort of performance I can get out of it for games (possibly replacing dual boot if good enough).
I've been particularly looking at this guide: Multiheaded NVIDIA Gaming using Ubuntu 14.04 + KVM
It states the host OS and guest cannot use the GPU at the same time, and so the GPU is "blacklisted" from the host OS by binding the GPU(s) to the pci_stub by modifying the initramfs. Is it possible to move the GPU between the host and guest while the system is running? I want to be able to use the dedicated GPU on Debian when the Windows VM isn't running or set to use it. (Without having to modify the initramfs and restart host each time)
I was also hoping someone could explain exactly what the pci_stub is; why binding the GPU(GPU's driver? Module containing its driver?) to the pci_stub prevents its use by the host; and how exactly is the pci_stub used to pass the GPU to a VM?