I would like to be able to use mps in my Linux VM (my setup is Mac M1 + Ubuntu 22.04 via VMWare Fusion), however it seems like there are two major barriers in my way/questions that I have:
- Does there exist a Linux + arm64/aarch64 with M1 Pytorch build?
I have not been able to find such a build. From what I’ve seen, most people who are looking for a Linux + arm64/aarch64 build are typically using NVIDIA GPUs.
- Is VMware’s 3D acceleration actually providing “expected” GPU passthrough?
By expected, I mean that I expect the OpenGL values for Renderer to be the same regardless of whether I’m on host (macOS) or guest (Ubuntu). Following Antonin Carette’s tutorial OpenGL, in C++, on m1 mac, I get the following print out for my GPU renderer specs:
Renderer: Apple M1 Max
OpenGL version supported: 4.1 Metal - 86
however when I follow GetIntoGameDev’s tutorial Setting up an OpenGL Project in Ubuntu (and add in the print out lines from the m1 mac tutorial), I get the following print out:
Renderer: SVGA3D; build: RELEASE; LLVM;
OpenGL version supported: 4.3 (Compatibility Profile) Mesa 23.2.1-1ubuntu3.1~22.04.2
I have no idea if the change in renderer names makes a difference, however I thought it was worth mentioning as another possible issue.