[Arm-netbook] mali gpu reverse engineering lkcl may ignore

David Niklas doark at mail.com
Sat Jun 17 14:39:36 BST 2017

On Wed, 31 May 2017 00:16:42 +0200
sam via arm-netbook <arm-netbook at lists.phcomp.co.uk> wrote:
> On 30/05/17 22:48, doark at mail.com wrote:
> > Well, I'd put ten dollars to a campaign like this without a HW reward.
> > I'm assuming that beings that there are so many Mali GPUs and hacker
> > boards out there that other people would also be very interested in
> > this. What would I need to know to do this?
> > Quick, point me to the books!
> > No, really, I would do such a thing, I don't have a lot to loose,
> > though for free I'd be taking my time...
> > But still, I'd need an education.
> >
> >  
> Start from the existing code http://limadriver.org/

I understand that I understand very little about this process so please
read this with hearty laugh prepared.

Whereas other older people have gone from simple reverse engineering
projects to more difficult ones I have come into the game late when all
the projects are the most difficult.
Let me assume that the GPU is a RISC model and uses 8-bit instructions.
Then it would have a total of 255 instructions (the 256th would be all
zeros and be a no-op because the wires on the line need a way to tell if
they have an instruction on them and that is the most power conservative
I can think of).
Now let use assume that if the signed bit is set that the GPU receives an
instruction to set an internal option.
All that I could probably learn from the lima driver and also some idea
of how the 2D rendering engine works and what it's instructions are.
Now the questions come up:
1. What are the options for the 3D engine?
2. What are the instructions for the 3D engine?
3. How do the 2D, 3D, and video (de|en)code engine fit together?

To sum it up, I don't think it's as simple as downloading the code,
signing up for the mailing list, and coding. It might be, someone could
have left full specs laying around waiting to be turned into mock-up code
and then real code; but I doubt it.
That's not to say I will not try, but I just don't see this as a very
productive path.

I suffer from the black box discouragement effect. Someone builds a black
box, then a bigger black box, then an even larger black box; eventually
no one knows how it works inside, even the people who designed it
understand only a relatively small part.


More information about the arm-netbook mailing list