How to get WSLg working with OpenGL, Vulkan, and CUDA on (Official) Arch Linux and Ubuntu WSL Instances.

wslg-guide.md

WSLg with OpenGL, Vulkan, and CUDA on (Official) Arch Linux and Ubuntu WSL Instances

Note: I haven't documented CUDA here. I'm lazy (and haven't tested on Arch). See https://documentation.ubuntu.com/wsl/en/latest/howto/gpu-cuda for that on Ubuntu.

Context

microsoft/wslg#1312:

So I got this working on both Ubuntu, and Arch. First, start with a fresh install (of archlinux from https://gitlab.archlinux.org/archlinux/archlinux-wsl or Ubuntu from wsl --install Ubuntu or https://apps.microsoft.com/detail/9pdxgncfsczv).

Arch

1. Install mesa and vulkan-dzn.
2. Run echo "L+ /tmp/.X11-unix - - - - /mnt/wslg/.X11-unix" | sudo tee /etc/tmpfiles.d/wslg.conf. (Thanks to microsoft/wslg#43 (comment) for this.)
3. Create your user account as per https://wiki.archlinux.org/title/Users_and_groups#User_management and confgure it as the default user for WSL as per https://wiki.archlinux.org/title/Install_Arch_Linux_on_WSL#Set_default_user.
4. su into the user or open a new terminal tab/window.
5. Run:

cat << EOF
export GALLIUM_DRIVER=d3d12

for i in "/mnt/wslg/runtime-dir/"*; do
  [ "$XDG_RUNTIME_DIR" = "$HOME" ] && XDG_RUNTIME_DIR="/var/run/user/$UID"
  if [ ! -L "$XDG_RUNTIME_DIR$(basename "$i")" ]; then
    [ -d "$XDG_RUNTIME_DIR$(basename "$i")" ] && rm -r "$XDG_RUNTIME_DIR$(basename "$i")"
    ln -s "$i" "$XDG_RUNTIME_DIR$(basename "$i")"
  fi
done
EOF | sudo tee /etc/profile.d/wslg.sh`

6. Restart WSL with wsl --shutdown from CMD/PowerShell and you should see the following (given you install the relevant packages):

$ glxinfo | grep Device
    Device: D3D12 (NVIDIA GeForce RTX 4080 SUPER) (0xffffffff)

$ vulkaninfo | grep "GPU id"
WARNING: dzn is not a conformant Vulkan implementation, testing use only.
                GPU id = 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 4080 SUPER))
                GPU id = 1 (llvmpipe (LLVM 19.1.7, 256 bits))
                GPU id = 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 4080 SUPER))
                GPU id = 1 (llvmpipe (LLVM 19.1.7, 256 bits))
                GPU id = 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 4080 SUPER))
                GPU id = 1 (llvmpipe (LLVM 19.1.7, 256 bits))
GPU id : 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 4080 SUPER)):
GPU id : 1 (llvmpipe (LLVM 19.1.7, 256 bits)):

Ubuntu

1. OpenGL should already be working. We'll add Vulkan support.
2. Go to https://ppa.launchpadcontent.net/kisak/kisak-mesa/ubuntu/pool/main/m/mesa/.
3. Look for mesa-vulkan-drivers_*.deb.
4. You'll see ...<LETTER>_<ARCH>.deb. The letter corresponds to your Ubuntu version codename. In your case, "n" for Noble.
5. Copy the link for the right package, https://ppa.launchpadcontent.net/kisak/kisak-mesa/ubuntu/pool/main/m/mesa/mesa-vulkan-drivers_25.0.1~kisak1~n_amd64.deb in your case.
6. Run wget https://ppa.launchpadcontent.net/kisak/kisak-mesa/ubuntu/pool/main/m/mesa/mesa-vulkan-drivers_25.0.1~kisak1~n_amd64.deb anywhere you have write access to on the Ubuntu WSL instance.
7. Run sudo apt install <PACKAGE> (in your case, sudo apt install mesa-vulkan-drivers_25.0.1~kisak1~n_amd64.deb.
8. Run:

cat << EOF
export GALLIUM_DRIVER=d3d12

for i in "/mnt/wslg/runtime-dir/"*; do
  [ "$XDG_RUNTIME_DIR" = "$HOME" ] && XDG_RUNTIME_DIR="/var/run/user/$UID"
  if [ ! -L "$XDG_RUNTIME_DIR$(basename "$i")" ]; then
    [ -d "$XDG_RUNTIME_DIR$(basename "$i")" ] && rm -r "$XDG_RUNTIME_DIR$(basename "$i")"
    ln -s "$i" "$XDG_RUNTIME_DIR$(basename "$i")"
  fi
done
EOF | sudo tee /etc/profile.d/wslg.sh`

9. You should see the following (given you install mesa-utils and vulkan-tools):

$ glxinfo | grep Device
WARNING: dzn is not a conformant Vulkan implementation, testing use only.
WARNING: Some incorrect rendering might occur because the selected Vulkan device (Microsoft Direct3D12 (NVIDIA GeForce RTX 4080 SUPER)) doesn't support base Zink requirements: feats.features.logicOp have_EXT_custom_border_color have_EXT_line_rasterization
    Device: D3D12 (NVIDIA GeForce RTX 4080 SUPER) (0xffffffff)

$ vulkaninfo | grep "GPU id"
WARNING: dzn is not a conformant Vulkan implementation, testing use only.
                GPU id = 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 4080 SUPER))
                GPU id = 1 (llvmpipe (LLVM 19.1.7, 256 bits))
                GPU id = 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 4080 SUPER))
                GPU id = 1 (llvmpipe (LLVM 19.1.7, 256 bits))
                GPU id = 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 4080 SUPER))
                GPU id = 1 (llvmpipe (LLVM 19.1.7, 256 bits))
GPU id : 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 4080 SUPER)):
GPU id : 1 (llvmpipe (LLVM 19.1.7, 256 bits)):

The reason we're using just the Vulkan driver from Kisak's repo. is that the mesa package causes llvmpipe to be used instead of D3D12. Maybe overring GALLIUM_DRIVER would work - I didn't try.

Updated the script to include adding [ -f "/etc/wsl.conf" ] && [ ! -L "/run/user/$UID/wayland-0" ] && ln -s "/mnt/wslg/runtime-dir/wayland-0" "/run/user/$UID/wayland-0" to your bash startup script. Unfortunately all forms of user-tmpfiles.d doesn't seem to work on WSL (likely due to how WSL logs in the default user) so we have to resort to this.

Hello, I am using WSL, ubuntu24.04, with Geforce RTX5070Ti，cuda=12.8， I have configured according to the method you provided, but still cannot use GPU in OpenGL :(

$ glxinfo -B
name of display: :0
display: :0 screen: 0
direct rendering: Yes
Extended renderer info (GLX_MESA_query_renderer):
Vendor: Mesa (0xffffffff)
Device: llvmpipe (LLVM 17.0.6, 256 bits) (0xffffffff)
Version: 24.3.4
Accelerated: no
Video memory: 23726MB
Unified memory: yes
Preferred profile: core (0x1)
Max core profile version: 4.5
Max compat profile version: 4.5
Max GLES1 profile version: 1.1
Max GLES[23] profile version: 3.2
Memory info (GL_ATI_meminfo):
VBO free memory - total: 31 MB, largest block: 31 MB
VBO free aux. memory - total: 22905 MB, largest block: 22905 MB
Texture free memory - total: 31 MB, largest block: 31 MB
Texture free aux. memory - total: 22905 MB, largest block: 22905 MB
Renderbuffer free memory - total: 31 MB, largest block: 31 MB
Renderbuffer free aux. memory - total: 22905 MB, largest block: 22905 MB
Memory info (GL_NVX_gpu_memory_info):
Dedicated video memory: 4293168161 MB
Total available memory: 4293191887 MB
Currently available dedicated video memory: 31 MB
OpenGL vendor string: Mesa
OpenGL renderer string: llvmpipe (LLVM 17.0.6, 256 bits)
OpenGL core profile version string: 4.5 (Core Profile) Mesa 24.3.4 - kisak-mesa PPA
OpenGL core profile shading language version string: 4.50
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile

@ChuXiaoyuu, I've updated the guide. Can you try on a fresh distro to see if it works? OpenGL should work out of the box on Ubuntu.

EDIT: Updated once again to get around a bug where Pulse sub-folders are sometimes created (what are you doing Microsoft).

@ChuXiaoyuu, I've updated the guide. Can you try on a fresh distro to see if it works?

Thank u for your reply! I followed your guide about Ubuntu, and I can add Vulkan support now, but OpenGL is still using the llvmpipe

$ glxinfo | grep Device
Device: llvmpipe (LLVM 17.0.6, 256 bits) (0xffffffff)

$ vulkaninfo | grep "GPU id"
WARNING: dzn is not a conformant Vulkan implementation, testing use only.
WARNING: dzn is not a conformant Vulkan implementation, testing use only.
GPU id = 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 5070 Ti))
GPU id = 1 (Microsoft Direct3D12 (Intel(R) Graphics))
GPU id = 2 (llvmpipe (LLVM 17.0.6, 256 bits))
GPU id = 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 5070 Ti))
GPU id = 1 (Microsoft Direct3D12 (Intel(R) Graphics))
GPU id = 2 (llvmpipe (LLVM 17.0.6, 256 bits))
GPU id = 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 5070 Ti))
GPU id = 1 (Microsoft Direct3D12 (Intel(R) Graphics))
GPU id = 2 (llvmpipe (LLVM 17.0.6, 256 bits))
GPU id : 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 5070 Ti)):
GPU id : 1 (Microsoft Direct3D12 (Intel(R) Graphics)):
GPU id : 2 (llvmpipe (LLVM 17.0.6, 256 bits)):
GPU id : 0 (Microsoft Direct3D12 (NVIDIA GeForce RTX 5070 Ti)):
GPU id : 1 (Microsoft Direct3D12 (Intel(R) Graphics)):
GPU id : 2 (llvmpipe (LLVM 17.0.6, 256 bits)):

@ChuXiaoyuu, I wonder if it's to do with Optimus...

Try running GALLIUM_DRIVER=d3d12 glxinfo | grep Device.

Though glxinfo should complain about dzn on Ubuntu as zink is supported (and will make Vulkan calls which brings up the dzn error). Did you install a custom mesa package (you should only install mesa-vulkan-drivers from the PPA, not mesa).