Embed Notice

“Why would you want to do that, Audrey?” well, couple reasons that it might be interesting (but probably not worth it and complicated as fuck):
1. Reverse engineering of proprietary stacks? Not sure how possible or relevant/useful this would be
2. Shit like QEMU VirGL? If you could emulate a device to pass through, in theory you could install the normal compute and driver stack but handle it however you want on the host side?
3. Opening up usage of CUDA on non-CUDA devices?
4. Curiosity

No idea how realistic any of that is and whether it’s worth it, but it might be interesting to see if anything exists out there. And if it did allow more flexibility in running software with hardware options, that would be neat.