Asahi Linux Enables Early Apple GPU Driver Support - WIP OpenGL 2.1 + GLES 2.0

[phoronix]

Phoronix: Asahi Linux Enables Early Apple GPU Driver Support - WIP OpenGL 2.1 + GLES 2.0

The Asahi Linux distribution is now shipping an early, alpha quality graphics driver stack for the Apple M1/M2 SoCs. This work-in-progress driver consists of their experimental Rust-written DRM kernel driver and then the AGX Gallium3D code in Mesa that is currently targeting OpenGL 2.1 and OpenGL ES 2.0 support...

https://www.phoronix.com/news/Asahi-Linux-Enables-Apple-GPU

[hyperchaotic]

This is a huge step forward, getting GPU assisted GNOME and KDE on Wayland and X going makes a big difference for the desktop, and the use of Rust for the driver has really definitively proven to be the right choice.

[bemerk]

I think the project idea is amazing. This seems to be founded development from the community for the community.
Should be done more often as it speeds up progress a lot.

[cooperate]

What’s wrong with using Zink to translate OpenGL? The performance is pretty much identical to an OpenGL driver. It’d be better if they focus all their efforts into making the best Vulkan driver they can than making subpar Vulkan and OpenGL drivers.

[Tomin]

What’s wrong with using Zink to translate OpenGL? The performance is pretty much identical to an OpenGL driver. It’d be better if they focus all their efforts into making the best Vulkan driver they can than making subpar Vulkan and OpenGL drivers.

This makes me feel that you didn't even read the article. It was mentioned in the quote on this very Phoronix article that they estimated they would get basic OpenGL driver done much quicker than a Vulkan driver capable of running those apps. It's further explained in the original blog post. A short version of it is that it's actually more complicated matter to support any level of OpenGL via a new Vulkan driver than one might think.

[Eirikr1848]

Really wish Apple would contribute to the open source development work for their own hardware.

Mesa works on NetBSD/Darwin/Linux right?

So this would also work for their own open source Darwin OS (which seems to be basically a minimal effort project outside of its use for macOS?)

[cooperate]

This makes me feel that you didn't even read the article. It was mentioned in the quote on this very Phoronix article that they estimated they would get basic OpenGL driver done much quicker than a Vulkan driver capable of running those apps. It's further explained in the original blog post. A short version of it is that it's actually more complicated matter to support any level of OpenGL via a new Vulkan driver than one might think.

I read it and that’s why I commented. I’m saying why waste time from the beginning on making the OpenGL driver when you can start with Vulkan right away? It’ll take a little longer to see results, but at least they’re not wasting effort. Maybe some numbers will help you understand. If, let’s say, it’ll take 1 month for the gl driver and 2 months for the vulkan one, then skipping OpenGL would allow to finish vulkan in 1.5 months. You see where I’m coming from?

[NobodyXu]

I read it and that’s why I commented. I’m saying why waste time from the beginning on making the OpenGL driver when you can start with Vulkan right away? It’ll take a little longer to see results, but at least they’re not wasting effort. Maybe some numbers will help you understand. If, let’s say, it’ll take 1 month for the gl driver and 2 months for the vulkan one, then skipping OpenGL would allow to finish vulkan in 1.5 months. You see where I’m coming from?

Are you sure the vulkan driver can be done in 2 months?

It's also mentioned in the blog that to make the vulkan driver useful, they need to implement some extensions too.

So not just the basic vulkan protocol but also exts, I doubt it can be done in 2 months.

Besides, I think it takes them longer than 1 months for the opengl driver and having a working desktop env earlier would help them better understand the M1 GPU plus find the subtle bugs, which will also improve their understanding of M1.

[cooperate]

Are you sure the vulkan driver can be done in 2 months?

It's also mentioned in the blog that to make the vulkan driver useful, they need to implement some extensions too.

So not just the basic vulkan protocol but also exts, I doubt it can be done in 2 months.

Besides, I think it takes them longer than 1 months for the opengl driver and having a working desktop env earlier would help them better understand the M1 GPU plus find the subtle bugs, which will also improve their understanding of M1.

… are you doing this on purpose? Those were just numbers I made up to let you picture it. I said IF.

[NobodyXu]

… are you doing this on purpose? Those were just numbers I made up to let you picture it. I said IF.

So what if the time for the vulkan driver to be functional is wayyyy longer than the time required for opengl?

That would mean the initial demo would take way longer and the Asahi Linux users would have to use software rendering until then.

And since vulkan is more complex, they could have a bug and not sure which code is to blame. It could be in the kernel GPU driver (wrong mmap) or in the userspace vulkan one.

A simple demo like opengl is less complex, means easier to debug and with earlier rollout, they can find out the bugs earlier.