Asahi Linux May Pursue Writing Apple Silicon GPU Driver In Rust

[Vistaus]

You can claim bad code all you want, but it is inevitable that humans will write flawed code. No one, no matter how big their ego is, ever writes software more than "Hello World!" without errors.

Except for the people that write "Hello Wolrd!".

[Vistaus]

C language and GCC is a proven tool for device driver development. Rust is not a proven tool for device driver development, in fact there is no kernel code at all, except some very trivial example.

Writing a driver for undocumented hardware is challenging, doing so with an experimental compiler is ... hobby-ism?

Hobby-ism, really? You do know that C and GCC weren't "proven" right from the get-go either, right?

[ferry]

Hobby-ism, really? You do know that C and GCC weren't "proven" right from the get-go either, right?

What I mean is if you are investing money into something and you are introducing 2 large risks, you are likely to loose your money. Now, as a hobbyist you don't care as you put in mostly your spare time. That is not a bad thing. Now, indeed GCC and linux itself were at the time considered hobbyist projects.

At that time there were commercial unix'es (HP-UX comes to mind) and associated c compilers. Someone, at the beginning, took the risk of developing hardware, OS and compilers/assemblers at the same time. And succeeded, congrats. But that was at the time when you needed to punch in the machine code instructions to boot your PDP8.

The first version of unix were written in assembly. And only later rewritten in C. All step-by-step. And even then, the work done by labs (like Bell) and universities (like MIT).

[Developer12]

This doesn't replace Mac OSX in the internal SSD? Is that a choice or Apple prevents you from doing this?

The prototype driver is a hack that requires two computers tied together over usb. The actual asahi install has always been installable alongside macos, just as you can install multiple versions of macos at once.

Edit: the ability to install multiple macos/linux/whatever OSs on the same SSD in the same system is something that apple *explicitly* engineered. They spend considerable time and effort producing a system that enables parallel install of "trusted" and "untrusted" OSs and this is an officially supported feature. (not to mention removing it would break backwards compatibility with previous versions of macos)

[Developer12]

I don't like to learn new things so hate Rust, but actually have a technical reason on why Rust is the wrong choice: other operating systems working on the M1 or M2. OpenBSD has a port for the Mac M1 and I think NetBSD does too. They will never accept a rust driver into their OS so why not stick to C so everybody can benefit?

BSD etc will have to rewrite the kernel driver anyway, as usual. Still, the linux implementation (and it's copious documentation) will be educational for them.

[mdedetrich]

What I mean is if you are investing money into something and you are introducing 2 large risks, you are likely to loose your money. Now, as a hobbyist you don't care as you put in mostly your spare time. That is not a bad thing. Now, indeed GCC and linux itself were at the time considered hobbyist projects.

At that time there were commercial unix'es (HP-UX comes to mind) and associated c compilers. Someone, at the beginning, took the risk of developing hardware, OS and compilers/assemblers at the same time. And succeeded, congrats. But that was at the time when you needed to punch in the machine code instructions to boot your PDP8.

The first version of unix were written in assembly. And only later rewritten in C. All step-by-step. And even then, the work done by labs (like Bell) and universities (like MIT).

If that's your standard for being proven then you shouldn't have any issues with Rust.

[andrei_me]

You mean Hector Martin?

Not in this case, the GPU side are spearheaded by Alyssa Rosenzweig (https://twitter.com/alyssarzg) and Asahi Lina (https://twitter.com/LinaAsahi)

The interesting part is that Lina livestream the development sessions and they are recorded so anyone can watch the +100h of work spent on this and follow the train of thought, decisions along the way etc

[erniv2]

Isn´t this topic getting "Rusty"

[rclark]

Yep getting Rusty, but, if you ask me, this driver sounds like the best place for the newest fad language Rust to be tried. If doesn't work out no biggie as hardly anyone would notice Always revert back to standard C.... which should be fairly quick as you'd have a sample driver already laid out to draw from. .

[brad0]

BSD etc will have to rewrite the kernel driver anyway, as usual. Still, the linux implementation (and it's copious documentation) will be educational for them.

There is no rewriting of the drivers usually, so no it's not as usual.