The Slides Announcing The New "AMDGPU" Kernel Driver

[dungeon]

No, there are fewer blobs. Catalyst has always been there, you just weren't thinking about it before

There we are, we are starting to learn about it and raising awareness that it exist, not in one but in two variants and with more posibilities for future . If you ask some opensource driver user... what is diference between non-pro and pro amdgpu driver? Obvious answer will be: None, both are blobs .

[whitecat]

As prevoius said, i call this something like - unified opensource solution around blobs .

AMD is developing a new kernel driver that will be able to run with TWO "user space driver" : one free, one closed.
So, why did you choose to call it "unified opensource solution around blobs" and not "unified opensource solution around open-source" ?

That is polemic about, developers and users does not think the same . You said "more open", but opensource users see "more blobs"

YOU may be the only one to see more blobs.
AMD Catalyst driver will rely on the kernel instead of it's own "kernel driver" (previously using closed code, now using open one).
AMD Catalyst's users will now have a smaller blob to download on the AMD website.
AMD will pay some developers to work on open-source code while they were working on closed code before.

[dungeon]

AMD is developing a new kernel driver that will be able to run with TWO "user space driver" : one free, one closed.
So, why did you choose to call it "unified opensource solution around blobs" and not "unified opensource solution around open-source" ?

amdgpu driver have model like that, it is primarly designed to run userspace driver blobs . There is not any advantage for opensource driver in comparason to existing opensource radeon driver model, so we can simply say all advantages of new model goes to the blob users

[bridgman]

amdgpu driver have model like that, it is primarly designed to run userspace driver blobs .

amdgpu driver have model like that, it is primarly made of raw chicken parts .

The above two statements are equally true.

YOU may be the only one to see more blobs.

[agd5f]

amdgpu driver have model like that, it is primarly designed to run userspace driver blobs . There is not any advantage for opensource driver in comparason to existing opensource radeon driver model, so we can simply say all advantages of new model goes to the blob users

I don't see why this is any more of an advantage to the closed source driver than it is to the open source driver. Both driver stacks will now have more developers working on the shared components. More developers working on modesetting, memory management, power management, etc.

[whitecat]

There is not any advantage for opensource driver in comparason to existing opensource radeon driver model, so we can simply say all advantages of new model goes to the blob users

I don't share this POV.
I'm very happy that AMD applies this new strategy of synergy. This is exactly what all open-sources enthusiasts wanted for years.

[MartinN]

That is the point - "quickly" cames from relative words category, so it can easely mean also - nearly never .

There is no secret, this is how things will be . Well you will see, there are always bugs in userspace (different for both, with no easy workaround) and AMD will also wants blob to be best served of course , so default will be blob preferably and in the end most users who does not care is it opensource or not will start to use that

So 'dungeon', do you have an alternative to capitalism? I'd like to hear it. Because if by 'open' you mean AMD should be giving away its IP so that some Chinese hack can reverse engineer it and release it as their own faster than two shakes of a lamb's tail, that is most likely never going to happen, unless of course, you have a way of reinventing capitalism into something else. Could something go wrong with the closed source "blob"? Of course. But what you're buying here isn't a "blob", but AMD's -COMMITMENT- to supporting Linux. Commitments are not bound by timelines... or blobs. They are indefinite, just as Linux is.

If someone from AMD is reading here, please disregard this kind of drivel. We are interested in a win-win (sorry to sound trite with these words) where Linux remains king of the OS on the desktop and the server, AMD can contribute improvement/patches/updates/new releases to the 'amggpu' driver if something goes awry and AMD's unwavering commitment toward Linux in general, and more specifically Wayland and its many display server implementations. We want Linux to become the de facto standard for a business desktop use, the server (which it already is) and possibly even gaming, if Valve's et al.'s efforts bear fruit, as we're all hoping and praying here.

All the reasons to remain locked in a operating system dominated by any singular entity (i.e. Microsoft or Apple) are slowly evaporating. The world is bigger than Microsoft or Apple, and the only thing that can serve it is Linux - precisely because no one owns it and it is the most unique, amazing collaborative effort of the last 20 years that I intend to see live into the future far longer than any one of us here.

[CTown]

Will the open source driver use the new kernel space driver for new hardware only as well? Or will it just be ported to one kernel space driver for ease of maintaining? Also, did AMD have a hard time convincing the open source developers to port their user space parts to AMDGPU or did they just get the benefits right away?

[bridgman]

Will the open source driver use the new kernel space driver for new hardware only as well? Or will it just be ported to one kernel space driver for ease of maintaining? Also, did AMD have a hard time convincing the open source developers to port their user space parts to AMDGPU or did they just get the benefits right away?

Initial implementation will be for newer hardware only (basically anything not already supported by radeon). Apart from only wanting one kernel driver upstream for each HW device, there are some architectural changes going into the new code to let us work more effectively with the HW teams going forward (basically organizing the code around individual HW block versions rather than entire chips).

The overall idea was driven initially from the open source side, so there was no "convincing" required. The plan doesn't work without a good open source userspace.

[xeekei]

Will this new driver come with Wayland/Mir support?