Dissecting the Apple M1 GPU, the End

Yes I think so. Her resume says she started working for Intel on open source graphics driver this month.

From "draw a triangle" to upstream Vulkan on M1. Practically, this makes the Venus/virtio path viable for guests on Apple Silicon (no passthrough in VZ), which is what many people actually need.

> I wish i had half of the energy of this guy

Trolling will get you banned here, so please don't.

i think it's a woman?

> in lieu of building better stuff from the ground up

To be fair, even if you have the best CPU and GPU designers, it's not as if you can call up TSMC and have them do a run of your shiny new processor on their latest (or even older) process. You can't fab them at home either.

Even with proper documentation, there still would have been loads of work to get M1/M2 GPUs working on Asahi Linux. Writing GPU drivers worth a damn is about as difficult as targeting a compiler to a new CPU architecture. It would not be "in a heartbeat".

Reverse engineering requires a different mindset and somewhat different skill set than “forward” engineering. I’ve met people who were happy to only do reverse engineering (to figure out what make things “tick”) without building anything new.

If it was up to me, 2 years of successful reverse engineering (of a variety of projects/products) would be a requirement to be called an engineer. You learn a lot from working things that you can’t learn from a book (and without having to do the mistakes yourself first…)

Just to make it clear: I am not implying anything about Alyssa - just stating an observation based on my own experience.

Ultimately we didn't need Jon Lech Johansen's work (and Derek Fawcus's, and others') on cracking DVD DRM or cracking Apple's FairPlay DRM, as there have always been alternatives. But their efforts did push the fight against DRM in our favour, and who knows what the world would be like if we had done nothing?

Creating things is a gamble, as mass adoption is almost never by technical merits, but by marketing. So you could make open documented everything but still end up with nobody benefiting from that openness, because a competitor (whether open or not) wipes you out. You saw this happen even in the era where electronic devices were expected to come with full schematics -- there were winners and losers even then.

But, if something has become widespread and well adopted, and it's not open, that's a problem. It absolutely should be opened up and documented. Especially if it's not because the money-grubbing creators of the something are deliberately hiding how it works and locking down control in order to extract more money from everyone else's pockets. The sooner you put an end to that, and the more often you fight against that, the sooner society itself becomes more efficient and fairer for everyone.

Xe HPG is also known as A750 and B580 so yes, it competes with the 3060/4060/5060.

No clue who you are, but real talk she’s achieved more than I will in my entire life. I’ve been in the industry for decades.

Found out from some Reddit discussions that the developers aim to first upstream everything for M1/M2 to the kernel, and as of https://asahilinux.org/2025/08/progress-report-6-16/:

> With Linux 6.16, we also hit a pretty cool milestone. In our first progress report, we mentioned that we were carrying over 1200 patches downstream. After doing a little housekeeping on our branch and upstreaming what we have so far, that number is now below 1000 for the first time in many years, meaning we have managed to upstream a little over 20% of our entire patch set in just under five months. If we discount the DCP and GPU/Rust patches from both figures, that proportion jumps to just under half!

So if the discussions are true, it can take years for the developers to finish M1/M2 upstreaming with all the Linux kernel bureaucracy. That is, unless they decide to start working on M3 before finishing the upstreaming

The M3+ GPU is also very different. So while it may be true that the driver development for M1/M2 is now more or less complete as OP says, future work along the same lines will very much be needed.

I'd pay easily let's say $100-200 a year to have linux running on modern apple laptops with full features. I'm sure I'm not alone. Their hardware, "our" OS would be perfect. Well, except notch and lack of OLED - but, reportedly that's in the works too.

The new leadership team set a short term goal of getting their existing work upstreamed, which seems to be going well.

> 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.

https://asahilinux.org/2025/02/passing-the-torch/

> With Linux 6.16, we also hit a pretty cool milestone. In our first progress report, we mentioned that we were carrying over 1200 patches downstream. After doing a little housekeeping on our branch and upstreaming what we have so far, that number is now below 1000 for the first time in many years, meaning we have managed to upstream a little over 20% of our entire patch set in just under five months. If we discount the DCP and GPU/Rust patches from both figures, that proportion jumps to just under half!

While we still have quite a way to go, this progress has already made rebases significantly less hassle and given us some room to breathe.

https://asahilinux.org/2025/08/progress-report-6-16/

Switching to work on Intel GPUs is not walking away.

It is accepting a new challenge.

There isn't really anything left to do for her - everything missing (including work on the newer graphics chips) can be somewhat easily done by less talented people, building on her work.

She did the challenging stuff she cares about. One aspect of nerd brain often is that you can hyperfocus on challenging stuff, but can't get the motivation to work on stuff you don't care about - and even what would be a 20 minute task can end up taking days because of that. It's great that she has the self awareness to set goals, and step away once they're done.

I didn't have that in that age - and still sometimes struggle. I was lucky enough that my employer back then recognized my issues, and paired other people with me for doing the stuff I was not interested in, and now usually manage to load those issues onto other co-workers by myself.

Maybe she didn’t pass a leetcode interview. :p

Maybe she doesn't want to.

They seem closely aligned with the Free Software Foundation (FSF), so I could very well imagine that being a major ideological reason not to want to work with Apple. Yes, Apple sometimes upstream patches and they do contribute to open source here and there, but they certainly are no FSF poster child. Intel on the other hand are about as open as it gets when it comes to their track record in the graphics space. I personally have nothing but admiration for Rosenzweig's work and I hope they will continue to find environments where they can flourish and do great things in the years to come.

Alyssa's post mentions how lots of the work she's done has at least started as side projects while she's working on something else (Panfrost while at high school, M1 drivers while at Collabora). Obviously I'm not her so I can't say anything specific to her. In general, Apple doesn't allow its developers to work on open source projects on the side while employed at the company. I think this is a stupid idea that costs them a lot of talent, but I doubt Apple cares what I think. I've seen multiple cases where an active open source contributor gets hired by Apple, then their presence in open source communities vanishes. Based on all the open source work she's done so far, I think it would take a lot to make her stop all contributions like that.

Genuine question: is she a once-in-a-generation prodigy? We forget this class of people indeed exists. As fellow professionals are we inspired or deeply ashamed of ourselves?

I believe Walmart has a deal[1] with Apple to sell[2] M1 MacBook Airs. This has been the case for a year or so, so I don't think it's old stock. They have been in stock for since that date, and slowly getting cheaper.

[1]https://9to5mac.com/2024/03/16/walmart-m1-macbook-air-launch...

[2]https://www.walmart.com/ip/Apple-MacBook-Air-13-3-inch-Lapto...

> M1 chip laptops can only be bought second hand at this point

New M1 Macbook Airs are still available at Walmart (maybe elsewhere). But even if not, who cares? People are still writing code for computers that haven't been sold since the 1980s.

When the next gen chip comes out you can usually reuse large amount of stuff.

Why only m1? Just installed stuff on my M2 Max Some texts says m3 also supported...

You can repurpose a decent hardware, in terms of performance and consumption, to run for far more time than Apple is willing to support it?

Why would the product have to be available new for the project to be worth it? There are still many M1 chips out there, and this helps prolong the usefulness of those chips.

The short answer is that it's just a stupid idea (and a waste of money). Asahi only works somewhat ok on M1.

m2 has magsafe

Are you coming from Windows? MacOS is a BSD descendant so it’s quite Unix-y. I never miss Linux on it and I used to only use Linux. Just learn how to get around the minor annoyances (eg the file explorer sucks , I use eMacs for that) and it’s a fine OS. It’s really not worthwhile trying to install anything else on the Mac.

What is this, quora ?

One of the few people who are actually competent.

Although most likely she’s well compensated, and doesn’t have to waste time on useless efforts at work, this level of discipline and striving towards a goal is just very rare in general.

Possibly also no family, limited social life and no other hobbies.