XWayland 21.1.2 Released With NVIDIA Hardware Acceleration Support

[treba]

You got it mostly right, but Nvidia isn't the bad guy here.
GBM is Mesa/Linux only, whereas EGLStreams is a standard OpenGL (OpenGL ES?) extension, thus cross-platform. It fits better Nvidia's cross-platform driver. For bonus points, GBM and EGLStreams are not exactly equivalent to each other and both have their downsides.

AFAIK the idea of EGLStreams being a standard is mostly a myth. Yes, Nvidia used their Khronos membership to standardize their technology. But AFAICS nobody else ever implemented it anywhere and retrospectively it is seen as a mistake by many to standardize it. GBM in turn was baked by everyone else in the Linux community, making it the actual standard.

[binarybanana]

I wonder if at some point in the future someone makes a stand-alone version of Xorg based on XWayland, given that RedHat/IBM have retarded development and even stopped reviewing patches from contributors.

I don't know enough about XWayland to know what exactly is missing from it, but if the Xorg codebase is really as bad as they say that might be a way to keep it alive. Dropping support for anything but libinput and modesetting would probably simplify what and how much would have to be written from scratch.

[oiaohm]

Also, a good chunk of the pains of moving to Wayland is not about GBM vs EGLStreams. It's about Wayland doing away with a lot of functionality that was deemed insecure and implementers having to reinvent the wheel. And since there's no "server" to be shared anymore, everyone gets to reinvent their own wheel

There are still shared libraries options. wlroots. To be correct you are right its not GBM vs EGLStreams either. GBM is only part of it. You have Nvidia not support DMA-BUF and KMS as well.

Reality with Nvidia finally coming around to support KMS, GBM and DMA-BUF this make doing a Wayland compositor lot more simple. This will remove the problem of having to chose between the likes of wlroots and Nvidia graphics support.

bug77 the recent changes by Nvidia will take away a major reason why someone implementing a Wayland compositor will consider doing their own implementation instead of using a shared solution like wlroots or QT5.

Removing the "server" does not mean everyone gets to reinvent their own wheel. Please note as well bug77 historic disputes on how GPU were to be interfaced with back in the 1990s had us with 4 different X11 servers. Yes GBM/KMS/DMA-BUF vs EGLSTREAMS was equal problem to what happened in the 1990s that had us that you had to pay for closed source X11 servers so you could use particular GPUs this time has not go well but it has not gone as bad as the 1990s.

https://packages.debian.org/buster/weirdx

This is something else to remember bug77 as well. There is more than 1 X11 server in most of major Linux distributions. Yes a lot of X11 windows managers and desktops only work with x.org X11 server implementation. Of course this is not particular fun for users if they decide to attempt to try alternative options to x.org X11 and it fails either. The shared server has it fair share of problems not just security ones.

[bug77]

While I believe all of that is true, I think part of the kerfuffle was how long Nvidia waited to say they weren't going to support GBM. There were discussions about what thing to support and they kept quiet until everyone else agreed on GBM. I don't think Nvidia was being malicious, but they simply didn't care enough.

Absolutely. When two parties are involved in something as complex as this, it would childish to think only one side carries the blame.
I only meant, contrary to public opinion Nvidia (or Wayland devs fwiw) are not intentionally evil here. Nvidia is usually stingy when it comes to open source collaboration, Wayland devs needed a plan to move forward, Nvidia discovered Wayland's plans don't jive well with their cross-platform driver. And here we are today.

[JPFSanders]

What most people fail to understand re the Nvidia situation is that Nvidia does their own thing and don't follow the modern Linux standardized way of implementing a graphics card.

This is part due to historic reasons back from the time when the entire graphics stack was a mess, what Nvidia did back then was to implement a full stack of kernel video driver + X Driver + Userspace OpenGL driver, they did this to overcome most of the Linux limitations of the day and how they did it was by implementing their windows stack on top of their blobs. And back in the day it was commendable considering the sad state of Linux graphic card support.

Time passed, the situation is completely different now, the Linux kernel has decent support for graphics cards nowadays. However Nvidia still insist on developing their driver as its own island not integrated with the Linux kernel and graphics ecosystem unlike most of the other graphics card manufacturers do nowadays, Note that this is the same case with some ARM SOCs, Nvidia is just the most obviously blatant case because "muh market share"

If you still do not get it imagine Nvidia imposing onto Microsoft a new driver model that is incompatible with the current windows driver model just because Nvidia can't be bothered to use Microsoft's standard interfaces, and then blaming MS for any problem their implementation has because muh market share advantage.

Nvidia is being incredibly stupid here, they could:

1) license their binary blobs for distribution in the Linux kernel (they're signed by Nvidia, they can't be tampered with)

2) Produce a standard open source KMS driver for the linux kernel, similar to what AMDGPU or Intel do.

3) From there keep their proprietary stack like similar to AMDPRO, keeping everything as closed-source as they want.

Nobody cares about what's inside Nvidia's OpenGL or Vulkan stacks, no one, we have ours now that perform pretty well, and strictly they do not need to open source those.

Cuda can be implemented in the same way, they do not need to open source anything there either.

Nothing prevents them from doing what I describe other than "muh market share attitude", and in the end they will go that route sooner or later.

[bple2137]

How is it going to be? I have GTX 970 card and performance-wise I have zero reasons to upgrade. However I work with order generation of drivers as 470 drivers work unstably for my (after 30minutes-few days the acceleration seems to disappear from some of the applications such as Kicad 3D viewer) graphic cards. It would suck to have to buy a new graphics card just for the desktop environment compatiblity.

I belive that stable version of 470 wasn't released yet. If the driver is still compatible with you GPU, it will be usable for pretty long time as the driver is supposed to be an LTS (and even if it's no longer supported, NVIDIA's old drivers for now unsupported GPUs are still in use by people with old hardware without issues).

I suppose X11 will stop to be packaged pretty soon with major Linux distributions.

Don't worry - whether we like it or not, Xorg sessions will stay with us for a long long time.

[kon14]

How is it going to be? I have GTX 970 card and performance-wise I have zero reasons to upgrade. However I work with order generation of drivers as 470 drivers work unstably for my (after 30minutes-few days the acceleration seems to disappear from some of the applications such as Kicad 3D viewer) graphic cards. It would suck to have to buy a new graphics card just for the desktop environment compatiblity. I suppose X11 will stop to be packaged pretty soon with major Linux distributions.

I've been using 470.42.01 on Fedora Silverblue Rawhide with a GTX 970 ever since the 470 drivers got released. Performance is fine, as is stability on Gnome.
There's really no reason not to try Wayland out, except for app-specific incompatibilities ofc, as long as your distro of choice catches up with the required deps in their repos or you don't mind installing these yourself.

I doubt xorg-server will be removed from any major distro's repos any time soon, but it'll probably not be preinstalled for long given how Nvidia already works on EGL Streams supportive compositors and is soon receiving GBM support anyway.

[oiaohm]

If you still do not get it imagine Nvidia imposing onto Microsoft a new driver model that is incompatible with the current windows driver model just because Nvidia can't be bothered to use Microsoft's standard interfaces, and then blaming MS for any problem their implementation has because muh market share advantage.

We don't need to imagine this. Nvidia really did this with Windows Vista. It was about 2 years after the release Windows Vista before Nvidia decided they had to toe the line because Microsoft was not going to change what they were doing. Yes this is why if you install original release windows vista today on a Nvidia system its a lot better than at release windows vista as long as you use update drivers not the original ones that Nvidia tried to push their own direction with. Microsoft did change some things for AMD and Intel this does make sense when you wake up they have the dominate market share.

Nvidia also did this with game consoles and apple result in losing the market share from game consoles and apple. Nvidia kind of had this problem for quite some time and it been progressively costing Nvidia market share.

[remenic]

I've been using 470.42.01 on Fedora Silverblue Rawhide with a GTX 970 ever since the 470 drivers got released. Performance is fine, as is stability on Gnome.
There's really no reason not to try Wayland out, except for app-specific incompatibilities ofc, as long as your distro of choice catches up with the required deps in their repos or you don't mind installing these yourself.

My laptop has two outputs connected to the NVidia GPU, will I be able to use those as well as the laptop display connected to the Intel iGPU?

Last time I tried it on KDE Wayland it didn't work.

[Azrael5]

How long for 470 Nvidia drivers?