Asahi Linux On The Apple M1: "Usable As A Basic Linux Desktop" Sans GPU Acceleration

[phoronix]

Phoronix: Asahi Linux On The Apple M1: "Usable As A Basic Linux Desktop" Sans GPU Acceleration

The Asahi Linux project that has been working nearly the past year on bringing up Apple M1 support under Linux has issued their September 2021 porting and reverse engineering report...

https://www.phoronix.com/scan.php?page=news_item&px=Asahi-Linux-September-2021

[sarmad]

Maybe try to get the original macOS driver working under Linux somehow, along with Metal, then run Vulkan on top of Metal, then run OpenGL on top of Vulkan, and you'll get a hardware accelerated stack that is ugly, but hey, maybe it satisfies Apple fans.

[cytomax55]

here is the loop of emotions when reading this
Truly amazing work... just feel its such as waste of talent... but then again who am i to say what they should or shouldnt do with their time...

[benjiro]

here is the loop of emotions when reading this
Truly amazing work... just feel its such as waste of talent... but then again who am i to say what they should or shouldnt do with their time...

His is paid for the work via donations ... So its just a job.

[Developer12]

Maybe try to get the original macOS driver working under Linux somehow, along with Metal, then run Vulkan on top of Metal, then run OpenGL on top of Vulkan, and you'll get a hardware accelerated stack that is ugly, but hey, maybe it satisfies Apple fans.

This is a terrible idea. Even if it worked the performance would be far worse than a native driver. Then you have to contend with the fact apple's code is incomplete (do they even post the code for their mesa-like userspace components? I don't think so). And that it's written for a completely different kernel. Porting all that garbage would be a nightmare and maintaining it long-term as a distribution would be hell.

If you try to do things fast, you do them wrong. Just like Corellium's M1 port which will NEVER be upstreamed or even touched ever again.

Asahi already have a half-finished userspace driver (shader, assembler, and all that fun stuff) that is upstreamed in mesa and being tested on macOS using the kernel interfaces there. Once the DRM drivers for the Linux kernel are stabilized, they just have to move it over and implement the missing functionality macOS' DRM drivers do but Linux one's don't.

Every step of the way they have thoroughly reverse-engineered, understood, and documented all of their findings. Then they have always sought the least intrusive, most well-organized, most forward-thinking way of implementing things physically possible. That's why they're optimistic this stuff will work on future M1X/M2 chips with minimal changes. It's clean.

The fact they've been able to do this quickly is a testament to their talent.

[hajj_3]

If you guys didn't already know a future microsoft surface laptop will use an AMD arm soc. Microsoft is paying AMD to make them an arm soc with an amd gpu and off the shelf cortex x1 cpu cores. The performance of x1 cores is lower than apple m1 but it should still be decent. It should be a piece of cake to install linux on that unlike this mac.

[Developer12]

If you guys didn't already know a future microsoft surface laptop will use an AMD arm soc. Microsoft is paying AMD to make them an arm soc with an amd gpu and off the shelf cortex x1 cpu cores. The performance of x1 cores is lower than apple m1 but it should still be decent. It should be a piece of cake to install linux on that unlike this mac.

It's not really big news or particularly interesting. Microsoft has made ARM laptops using qualcom chips for years and years. These will be the same old crap with standard ARM cores, except now with AMD providing the GPU for them to bottleneck.

I'd be way more interested if it was actually AMD designing the ARM cores. Before they ran out of money they were working on project skybridge, trying to make socket-compatible ARM and x86 chips. Those would have been real, far more powerful ARM cores. On paper they could take ZEN and strip the x86 translator off it, replacing it with a smaller and faster ARM one, but I'm not holding my breath.

[Sonadow]

If I had an M1 Mac I sure as hell won't put Linux on it.

[sarmad]

This is a terrible idea. Even if it worked the performance would be far worse than a native driver. Then you have to contend with the fact apple's code is incomplete (do they even post the code for their mesa-like userspace components? I don't think so). And that it's written for a completely different kernel. Porting all that garbage would be a nightmare and maintaining it long-term as a distribution would be hell.

If you try to do things fast, you do them wrong. Just like Corellium's M1 port which will NEVER be upstreamed or even touched ever again.

Asahi already have a half-finished userspace driver (shader, assembler, and all that fun stuff) that is upstreamed in mesa and being tested on macOS using the kernel interfaces there. Once the DRM drivers for the Linux kernel are stabilized, they just have to move it over and implement the missing functionality macOS' DRM drivers do but Linux one's don't.

Every step of the way they have thoroughly reverse-engineered, understood, and documented all of their findings. Then they have always sought the least intrusive, most well-organized, most forward-thinking way of implementing things physically possible. That's why they're optimistic this stuff will work on future M1X/M2 chips with minimal changes. It's clean.

The fact they've been able to do this quickly is a testament to their talent.

Looks like you didn't realize that I was being sarcastic, as the whole idea seems useless to me since nobody is going to use an operating system without hardware acceleration. I know it sounds promising to see a half finished driver already, but remember that we had a more mature nVidia open source driver almost 10 years ago and here we are after ten years the driver is still pretty much useless. What makes you think the story will be different with Apple? If anything, it's expected to be worse since nVidia is just selling the hardware and have interest in seeing more software running on their hardware, whereas Apple is selling you an ecosystem that they want to lock you in, so they are more likely to try to complicate things for open source devs.

[Developer12]

Looks like you didn't realize that I was being sarcastic, as the whole idea seems useless to me since nobody is going to use an operating system without hardware acceleration. I know it sounds promising to see a half finished driver already, but remember that we had a more mature nVidia open source driver almost 10 years ago and here we are after ten years the driver is still pretty much useless. What makes you think the story will be different with Apple? If anything, it's expected to be worse since nVidia is just selling the hardware and have interest in seeing more software running on their hardware, whereas Apple is selling you an ecosystem that they want to lock you in, so they are more likely to try to complicate things for open source devs.

You clearly know nothing of the person developing this driver. Go research her.