Arch Linux

julesmazur

Member

Registered: 2016-11-14

Posts: 22

Getting CUDA to work on a Retina MBP with Weyland

I've got Arch running on a Retina MBP (2012 model-year) with a GNOME 3.22 GUI using Wayland. On the backend, I've got the mesa package driving my graphics, which are hybrid: an Intel Graphics HD 4000 onboard chip, and an nVidia 650M discrete chip.

I've had trouble with my graphics setup before (on a higher level; mostly just a glitchy desktop UI), but I'm now interested in working on some code which benefits greatly from GPU acceleration, and has a built-in backend allowing for GPU offloading via CUDA. Unfortunately, I can't do that when I have no nvidia drivers installed, which is easily confirmed:

me@machine$ lsmod | ag nvidia

me@machine$ 

Installing the nvidia package alone causes my computer to hang at startup, and at that point my best option is to boot into single-user and pacman -R nvidia. I remember there being a separate nouveau package, but that seems to have been merged into the mesa package itself.

Has anyone here had any success running CUDA code on this kind of machine? Does anyone have any tips to get this running?

x33a

Forum Fellow

Registered: 2009-08-15

Posts: 4,587

Re: Getting CUDA to work on a Retina MBP with Weyland

Mod note: Since this is not an actual programming question, I'll move this thread to Laptop Issues.