I’m honestly surprised the major distros haven’t gotten together to do this, in a shared fork if necessary. It’s outright embarrassing that this attitude still infects the Linux project.
None of the distros are going to get behind your fork because all agree with Linux on this matter, rightfully so. Nvidia can fuck off with their proprietary crap, Linux has no intention of bending over for that garbage.
ATI/AMD still has their own proprietary, closed blob driver that is not in the tree - and which actually runs quite a bit faster than the open source one.
$ cd ~/sources/linux/drivers/gpu/
$ git log . | grep "@amd.com" | wc -l
42418
$ cd ~/sources/mesa
$ git log | grep "@amd.com" | wc -l
4817
It doesn't run faster, this is nonsense. I've personally met the AMDGPU driver developers and if you ask them they'll even tell you that basically no one needs AMDGPU Pro. It's not faster, this is bollucks.
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/lin...
You are probably confusing the situation with nouveau, which is indeed community developed reverse engineered kernel driver for Nvidia.
It is generally slower than Mesa + upstream AMDGPU and has been for more than a year.
Linux needs to get over itself and recognize that it’s part of a larger world, and that its “my way or the highway” attitude is unnecessarily constraining.
The fact that people are coming to you and saying “I’d like to use my Nvidia card with Linux but Linux isn’t supported” does mean it’s a problem—and the problem is Linux’s, not Nvidia’s.
I don't find it embarrassing that Linux maintainers prefer to leave every line of code in the kernel up for improvement, rather than desperately replicating known misbehaviours until a new five-to-ten year waterfall release comes. AMD's drivers work great, I buy AMD because instead of getting upset that they don't get to call all the shots, they put in the work and make peace with the establishment.
NVIDIA would rather throw a pathetic, childish tantrum and blame everyone else for their arrogant position which leaves their customers begging maintainers to comply and work on their own time to enrich NVIDIA.
It works great in MPV, I don't use VLC so I can't comment on that. The major problem right now with video acceleration is that Chromium (or maybe it's my distribution's build of it) seems to have enabled VA-API acceleration, but the Mesa VA-API state tracker (I think, I'd be glad to be corrected on this) I think is missing the API that would tell Chromium which pixel format to expect the decoder's output in; so Chromium assumes that it'll be in a certain format which it turns out not to be, and this leads to hideous corruption.
I suspect that if and when there is any significant amount of HEVC content on the web, this issue will probably be resolved by then. For now there are configuration parameters which can be used to work around the issue by disabling the pixel format Chromium defaults to, and do so just in chromium, but a fix (which will be in Mesa 19) is required for that configuration to work with Chromium (due to its unusual process naming).
I have used the MJPEG UVD decoder through ffmpeg to decode my 4K webcam's output and it works a treat.
1. https://www.phoronix.com/forums/forum/linux-graphics-x-org-d...
In practice though, that's likely the least of their reasons not to open their driver. More probably, they just don't want to expose some embarrassing code mess or non compliant cheating in their graphics stack, as well as don't want to allow anyone decide how to use their GPUs without Nvidia dictating them. If they control accelerated driver - they can dictate and demand more money in some use cases from those who want a more permissive driver. All of that has very little to do with intellectual property and more with market manipulation.