NVIDIA Confirms Sway Wayland Compositor Works Fine With Their New GBM Driver Support

[oiaohm]

Cant say that I fully agree with you there. Yes Nvidia has been obstinate and belligerent but so have quite a number of wayland devs (aka, sway, gnome, etc.). Nvidia did have a lets say "functional" alternative solution that would let wayland work and most devs threw a hissyfit and said never! KDE for example has had support for wayland on Nvidia for a while now (perhaps for loose definitions of functional).

Really no the kde lead developer mandated that Nvidia provided the developer to KDE for Wayland support then when asked for the future features to complete out wayland support like restart compositor features Nvidia developered end up screwed. This is what leads to GBM support. KDE lead developer has really done a hissyfit just not in the normal way.

KDE lead developer version of hissyfit.
1) Mandate the only party that could implement Wayland support in KDE for Nvidia had to be a Nvidia paid developer.(yes this is lead developer power)
2) Being the lead developer use that power to mandate what features the Nvidia lead developer has to implement.
3) This checkmates the Nvidia developer with a feature they cannot implement and gets Nvidia to implement what KDE lead developer wanted all along as in GBM and DMA BUF.

The reality here the KDE lead developer knew Nvidia did not have a functional solution. Its in many bug reports one is good where the KDE lead developer is asking the Nvidia lead developer how to implement compositor restart and the Nvidia funded developer end up asking for KDE lead developer to restrict KDE functionality until he can talk to the driver developer team on fixing the problem. Yes this is what leads to GBM and DMA BUF support.

KDE lead developer version of a hissyfit is being belligerent as well. KDE had enough market share that Nvidia could not ignore the hissy fit of KDE lead developer saying we are not allowing anyone other than Nvidia to implement this. Yes the KDE lead developer also said he would not waste any time implementing eglstream support.

KDE version of a hissyfit worked to get us to where we are today where Nvidia is now starting to support other Wayland compositors that have not implement eglstreams. Yes this is another part of KDE lead developer being a pain asking for third party compositor proof of Nvidia functional code.

Getting on KDE lead developers bad side long term never end up good.

[MadCatX]

I love the smell of GBM in the morning. It smells like victory!

I also enjoy the sudden shift from "GBM is never gonna work with our driver" to "Sway works just fine out of the box with our driver". Nvidia really has been full of shit about this alleged GBM incompatibility the entire time.

[RecursiveRose]

I love the smell of GBM in the morning. It smells like victory!

I also enjoy the sudden shift from "GBM is never gonna work with our driver" to "Sway works just fine out of the box with our driver". Nvidia really has been full of shit about this alleged GBM incompatibility the entire time.

Every-time actual open community standards (not astroturfed ones like EGLStreams) win against device manufacturers trying to push their own bullshit on everyone else, it's a victory for us all.

[Paradigm Shifter]

I recently found out that nvidia's embedded products including the latest Xavier NX from last year are provided with their L4T distro based off Ubuntu 18.04 , the latest release is August 2021, but still no upgrade to a newer distro, the kernel is stuck on 4.9 still... They have good hardware in that space, it's just sad that they fumble so much on open-source that they can't even support a modern kernel or packages for their platforms.

Don't get me started. I like my little Jetson Nano, but support for it is entirely dependent on the community. A very enterprising chap put together an Ubuntu 20.04 based version, but I really get the feeling that the developer hardware nVidia releases is basically an afterthought.

I've tried a couple of times to get the nVidia GPU/CUDA working with ArchLinuxARM, but it has not been a resounding success. It was more than a year ago, though, so... maybe I should try again.

[Myownfriend]

...not astroturfed ones like EGLStreams...

That's a great way to put it.

[scirocco]

Nice add Sway to the list of functional wayland desktops, now there is 2 gnome and sway, I just wish they could add kde to that list soon.

[dragon321]

EGL Streams was never adopted by anyone as far as i know outside nVidia(and maybe some forgotten ARM project) for the same reason stated for every developer that tried, so it will all die tightly in-house.

EGL Streams wasn't even created for Wayland. As far I know (although I'm not sure about it) EGL Streams was created to exchange buffers between different APIs. It can serve Wayland needs to some extent but it's not perfect match. According to Sway developers EGL Streams cannot provide atomicity guarantees and removes possibility for direct scan out. It was easier for Nvidia to pickup it because they didn't had to put any significant work to support it.

100% false. The current X server project, XOrg was a complete revamp and largely a rewrite of the original X11 project whose last major release was X11R6. Released in ~2004. You think XOrg is bad, X11 was a nightmare

XOrg is implementation of X11 protocol. X11 itself is just like Wayland - protocol specification. XOrg is implementation of such protocol. It wasn't the only exisiting X11 implementation but it became most popular and de facto standard implementation. Of course X11 protocol and XOrg got extensions and improvements to support modern things (Wayland is getting extensions as well) but that doesn't change the fact that it uses extended protocol which first version was released in 80's.

[justdanyul]

Good for Nvidia, but my take on it - too little, too late. They are decades late and it's basically DOA.

With Intel joining high end GPU market especially, Nvidia will become irrelevant on Linux desktop due to lack of upstreamed driver. It will be just AMD and Intel.

If you are talking gaming only, sure. But... CUDA is borderline a monopoly in the ML space, which unfortunately is pretty important dimension of the high-end GPU segment. In short, NVIDIA dominates here, and ROCm is nowere near a challenger unfortunately.

I'm totally unbiased in this btw, I don't give a hoot if its team green, team red or team suspicious brown, i built a Radeon 7 rig to keep myself informed on TensorFlow and PyTorch progress under ROCm, and the fact is just, as it stands, CUDA vs. ROCm is as much a fight as Tyson Fury going up against a 6th grader.

This article, while old, gives you a good starting point. And, if you should do a bit of research, you'll see, the world haven't changed. CUDA is king, everything else is an afterthought from all the important projects in the ML space.

https://towardsdatascience.com/on-the-state-of-deep-learning-outside-of-cudas-walled-garden-d88c8bbb4342

I can't see intel moving this a lot, hopefully they will join forces with AMD and try and put more weight behind ROCm, and give more incentive to Google, Facebook and friends to support ROCm better. But, yeah, it might be too little too late, NVIDIA seriously took the ball and ran with it here.