Early Work Is Underway On Reverse-Engineering The Apple M1 GPU

[microcode]

well apple likes lawsuites when someone uses their devices for something they didnt intended

Well, that's not how property works, and not how IP law works. Apple may seek to harass ignorant people with no voice, but Rosenzweig would be a bad target, especially with a suit over a well-established lawful activity conducted in the open.

[wizard69]

Stop helping apple.

She is helping Linux. It isn't her fault that Apple was the first company to deliver a decent ARM based laptop. I bought one just to benefit from the advantages the M1 clearly provides. An ARM based laptop, running Linux, from Lenovo or Dell, would have been better but the idiots running these companies don't want to invest in such a beast.

[Speedator]

Wouldn't it be better and easier to do a Vulkan driver? Together with Collabora's Zink we get both an OpenGL and Vulkan driver.

[libv]

Well, I have no clue what happened within ARM, but as far as I can remember from Alyssa's blog (or Collabora's blog, I don't remember), it came because the reverse engineering had already produced very good results (it was already very impressive on the RK3399). This lead to renaming various elements in the code for clarity since there was suddenly a better understanding of the whole thing. But again, this is just based on my understanding of what I read on the blogs of Collabora / Alyssa... I'm just a RockPro64 user here!

We produced solid results with lima, and we showed some stellar progress back in 2012-2015. ARM would've preferred if it did not exist, is how they would phrase it, the kindest statements were along the lines of "this is producing documentation that we wish was not public".

Jem Davies is now gone. Nvidia (you know, the company with the biggest GPU market share) bought ARM. Those are the prime reasons for teppid support of this open source Mali driver, which seems to be mostly "let's at least keep our own internal naming consistent with what everyone else works with".

If you want to believe Collabora, then go right ahead. And i truly do hope that Collabora will support this new venture, but i know how that one will pan out. I have been through such stories too often myself.

[bavay]

So ARM "contributing" was simply due to the projected having advanced enough? [...] According to you, it has nothing to do with open source haters being moved aside inside ARM in the meantime? [...] And it definitely cannot have anything to do with NVidia buying up ARM [...]

Well, I have no clue what happened within ARM, but as far as I can remember from Alyssa's blog (or Collabora's blog, I don't remember), it came because the reverse engineering had already produced very good results (it was already very impressive on the RK3399). This lead to renaming various elements in the code for clarity since there was suddenly a better understanding of the whole thing. But again, this is just based on my understanding of what I read on the blogs of Collabora / Alyssa... I'm just a RockPro64 user here!

[libv]

But then, if collabora wants to pay for keeping Alyssa on the payroll for a long time, while all her braincycles are spent on the M1, and not on fixing the issues that collabora's customers pay for, nothing will stand in the way of this.

Your move collabora.

[Michael_S]

I wouldn't break wind to help the most consumer unfriendly company in history.

Wait, Comcast makes this thing?

[libv]

shmerl ... (reverse engineering of the GPU to the point that Arm started helping by contributing documentation, the code now works very well), ...

So ARM "contributing" was simply due to the projected having advanced enough? A concept that worked sooo well a few years earlier...

According to you, it has nothing to do with open source haters being moved aside inside ARM in the meantime?

And it definitely cannot have anything to do with NVidia buying up ARM, effectively making ARM MPD superfluous, and everyone in MPD now working hard on cleaning up their resumes and making sure that they are hireable in future?

[libv]

I can never understand why people waste their time on trying support something that was designed to not be supported like stuff from Apple and Nvidia.
If users want to use hardware with Linux, they should just buy Linux and open source friendly hardware.

If everyone thought like that, you would have no hardware working with linux today, definitely not graphics hardware, especially on ARM.

I do see this specific action as pointless though. The M1 is likely derived from PowerVR, but this time only shipping on tightly controlled hw, for a tightly controlled and very close minded userbase.

With solid technical reasons, i singled out PowerVR when i started lima in 2011. And i managed to convince everyone (that same everyone that copied my methods of exposing ARM GPUs, and the code that came with it, with and without accreditation) of the futility of such an endeavour with said solid technical reasons, stating that if we free everything but PowerVR the market will force PowerVR open.

PowerVR had, and has (amazingly) a sizeable userbase, outside of the tightly controlled apple ecosystem still today, and a valid linux "friendly" userbase, and yet my reasoning still holds firm.

I got met with sufficient political and economic opposition when i did lima, to make me unemployed for several years, until i was finally unable to stop kidding myself, and was forced to go do something else entirely. And where there was no opposition, there was a vast amount of indifference. And this was for highly popular hardware with a huge linux userbase, then and now.

Doing this for the Apple M1, with its PowerVR genetics, for a closed userbase, is an excercise in futility, and probably just hubris.

[discordian]

I wouldn't break wind to help the most consumer unfriendly company in history.

Well, getting the most consumer unfriendly desktop to run on it would be a perfect match.