The Future for Tyr, a Rust GPU Driver for Arm Mali Hardware
lwn.net108 points by todsacerdoti 8 hours ago
108 points by todsacerdoti 8 hours ago
> Dave Airlie just announced in the Maintainers Summit that the DRM subsystem is only ""about a year away"" from disallowing new drivers written in C and requiring the use of Rust.
wow
When the C absolutist maintainers fought for control over the ability to keep Rust out of their ballpark, I didn't expect the reverse to happen.
Still, I think it makes a lot of sense. Completely new GPU drivers are quite rare and the macOS drivers from Asahi are a showcase proving that Rust and GPU drivers work together well. If there's any subcomponent switching to Rust-first for new contributions, it makes sense for it to be the one that had already been proven to be Rust-compatible.
I would envision to see some more GPU drivers from Chinese companies like MooreThreads
Asahi project looks barely alive, almost abandoned. I know that their explanation of low activity is that they are being active elsewhere, supposedly pushing all their work upstream, but this has been happening for months and they don't give any reports about their progress, so I'm worried it will all die soon. And given that the project barely brought some Linux compatibility for m1 and m2 hardware and no prospects for bringing similar compatibility for newer generations - I fear it all will be kinda useless in the end.
> this has been happening for months and they don't give any reports about their progress
This seems a bit exaggerated, their latest progress report is barely two months old: https://asahilinux.org/2025/12/progress-report-6-18/
They inarguably have slowed down, but this should be expected as the project matures. It has also inevitably now faced the time when new generations of contributors are needed as existing ones retire/ move to other projects.
>as the project matures
How can it be mature if it can't even boot on newer MacBooks. The slowness does not seem to be due to running out of impactful work that needs to be done.
The new leadership team blogged last year that their priority would be on upstreaming their existing work.
> Our priority is kernel upstreaming. Our downstream Linux tree contains over 1000 patches required for Apple Silicon that are not yet in upstream Linux. The upstream kernel moves fast, requiring us to constantly rebase our changes on top of upstream while battling merge conflicts and regressions. Janne, Neal, and marcan have rebased our tree for years, but it is laborious with so many patches. Before adding more, we need to reduce our patch stack to remain sustainable long-term...
Where do the M3 and M4 fit in? Until upstreaming and CI progress, the core team cannot prioritize new hardware.
https://asahilinux.org/2025/02/passing-the-torch/
I think the majority of that upstreaming work (that isn't on hold until the kernal is ready for the Rust graphics driver to land) has happened and additional features like DP alt mode for USB C have been demoed.
The next update from the team should land on their blog after 6.19 ships
Activity has died down as a result of general Linux kernel developer drama, petty in-fighting, and other factors, but that doesn't change the results they did produce during their most prolific phase so far.
Without proper support from upstream like AMD, Intel, and Qualcomm (to some extent) are doing, Linux will never work as well on Apple's hardware as it does on normal hardware.
Wasn't it just a couple of weeks ago that they started supporting M3? That smells like progress to me.
One can start working on creation of a teleportation device. Doesn't mean we have it.
https://asahilinux.org/docs/platform/feature-support/m3/
What do you see as progress here? Nothing is supported, everything is "to be announced" (i.e. unsupported).
they likely meant this progress post showing a desktop booting on an M3 mac: https://www.reddit.com/r/AsahiLinux/comments/1qnddjd/m3_now_... albeit with software graphics
Weren't the old Linux kernel developers promised the opposite by Linus Torvalds? That they would be able to continue writing in C?
https://lkml.org/lkml/2025/2/20/2066
> The document claims no subsystem is forced to take Rust
Dave Airlie is saying that the subsystem maintainers are themselves choosing to move to Rust.
Means that other platforms need to allow Rust in the kernel too in order to use future drivers.
What do you mean other platforms?
Also they can just expose c bindings to these rust libraries no?
The old drivers are mostly dual GPL or MIT licenced, they have been used in all the BSD variants.
> One simply cannot deploy a driver that [...] crashes and takes the user's work with it.
Somebody needs to tell whoever wrote the drivers in the PC where I'm writing this.
Can't wait to write a Rust driver for my eink tablet <3
Interesting to see the building blocks come together. I hope that they can lay foundations that last.
Tyr is a Danish metal band. Period. :-)
I thought Tyr was the Norse god of War & Justice.
Considering that the Mali GPUs were developed by ARM Norway, and this driver is Just, I would say this is one aptly named driver.
T-Y-R is also the root in semitic languages (eg, Hebrew and Arabic) related to flying! Maybe not on purpose, but I really like that incidental connection, given the combined reputation of rust and GPU operations for being fast.
As a Hebrew speaker I cannot understand how you came into this conclusion. The closest I can think of is ת-י-ר, which is the root of being in a trip.
Technically they are from the Faroe Islands. Great band, seen them live many times.