You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have a standard Fedora 36 installation with NVIDIA drivers installed (x64 and i686 libraries installed) and both the GL and GL32 flatpak application installed. Running vulkaninfo on the host shows a single GPU which matches reality. When running a Vulkan application inside flatpak however, the application sees the same device twice.
It shows indeed two NVIDIA related Vulkan icd files.
TingPing
changed the title
Vulkan applications running in flatpak seeing same device twice
Having both GL and GL32 extension results in duplicated Vulkan nvidia_icd.json files
May 19, 2022
I confirm it still is happening with latest drivers, 535.104.05 at the time of writing.
Bottles now have a workaround for this specific issue: bottlesdevs/Bottles@5627b5b
Hi there,
I have a standard Fedora 36 installation with NVIDIA drivers installed (x64 and i686 libraries installed) and both the
GL
andGL32
flatpak application installed. Runningvulkaninfo
on the host shows a single GPU which matches reality. When running a Vulkan application inside flatpak however, the application sees the same device twice.I'm not familiar how the driver setup in flatpak works, but this looks very similar to negativo17/nvidia-driver#95 where both the x64 and i686 nvidia driver put an Vulkan icd file with the same content into
/usr/share/vulkan/icd.d/
which made Vulkan to load the same device twice. So my completely uneducated guess would be that this lines https://github.com/flathub/org.freedesktop.Platform.GL.nvidia/blob/master/nvidia-apply-extra.c#L112 and https://github.com/flathub/org.freedesktop.Platform.GL.nvidia/blob/master/nvidia-apply-extra.c#L463 (probably also https://github.com/flathub/org.freedesktop.Platform.GL.nvidia/blob/master/nvidia-apply-extra.c#L117) should only apply to the GL/x64 driver package. I could be wrong though.Kind regards and thanks a lot for all your hard work!
The text was updated successfully, but these errors were encountered: