Mesa 24.0.2 Brings Many Zink Fixes, More Intel Alder Lake N IDs

[phoronix]

Phoronix: Mesa 24.0.2 Brings Many Zink Fixes, More Intel Alder Lake N IDs

Mesa 24.0 series release manager Eric Engestrom is out with another on-time bi-weekly point release for this set of open-source GPU user-space driver components. There are many fixes, new Intel ADL-N PCI IDs, and other backported updates for this latest stable release...

https://www.phoronix.com/news/Mesa-24.0.2-Released

[Kjell]

Zink benchmarks, when?

[stormcrow]

More bug fixes for Zink are great! I'm already using Zink as I finally discovered where the problem is with the weird alpha layer rendering was coming from in Linux with OpenGL programs using AMD GPUs. I thought it a weird banding problem. Turns out it's a bug in the default AMD OpenGL Mesa stack that flips lighting/shadow shading on alpha/transparency textures (as in, the bright areas are dark, and the shadows are bright). It's fixed with the Zink OpenGL to Vulkan stack. Only took three years to figure it out and track down how to fix it (and sooooo many people saying bug doesn't exist or it was my fault). If anyone else sees this bug on their AMD GPU system using OpenGL with transparency shading put this environment variable

MESA_LOADER_DRIVER_OVERRIDE=zink

either in your CLI or the custom environment field in however you launch your programs.

https://wiki.archlinux.org/title/OpenGL#OpenGL_over_Vulkan_(Zink)

https://docs.mesa3d.org/drivers/zink.html

The docs has more environment variables set, but it seems the relevant one for me was just MESA_LOADER_DRIVER_OVERRIDE and the others were apparently superfluous on Debian 12.x.

Edit to add: Not really seeing significant loss of performance that couldn't be chalked up to system load differences. (less than a 5 FPS variance at over 120 FPS average for the program in question on an RX 6650 XT) Mileage is obviously going to vary depending on program, complexity of scenes being rendered, other programs running, etc.