0

I have a mid-2015 15" Macbook Pro. The iGPU is part of Intel's Iris Pro brand, which as I understand allows a VM to access the GPU directly with a native driver through GVT-d.

I would like to run a Linux guest OS in this manner, giving it native access to the iGPU so that it can run a native driver (and thus have OpenGL 4.3 support). I have no idea how to go about doing this, though. I've heard there are forks of QEMU/KVM with GVT included, but I have no idea how to go about setting up a VM with one of those.

  • This is way outside my area of expertise, but wouldn't any OS running in a VM be using the VM's virtualisation of the graphics drivers, not the actual drivers themselves - thereby putting an effective wall between VM & GPU – Tetsujin Sep 14 '15 at 7:12
  • @Tetsujin: API passthrough is a common technique, but some solutions allow direct passthrough to a GPU. Under Linux, for example, it's possible to enable PCI passthrough so that a VM can have private and native access to a GPU or other PCI expansion card, similar to how you can do USB passthrough to run a USB device with another OS' drivers. Intel's GVT portfolio extends this with some more advanced features such as GPU time sharing across multiple VMs, and I was hoping this would allow both a host and guest to natively access the same iGPU simultaneously. – lcmylin Sep 15 '15 at 18:50
1

Some further research has produced somewhat contrary answers. This forum thread seems to suggest that it's possible, but not for a Mac OS host. Conversely, here it's suggested that it's not possible to implement on a type-2 hypervisor like Virtualbox.

I'm not entirely sure what to make of this, except maybe to suspect that it's possible to time-share an Intel GPU across multiple VMs with this method, but not to time-share an Intel GPU across a host and a guest.

Either way, it seems the answer to this question would have to be No for the time being.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .