Explicit GPU Synchronization Merged For XWayland

[phoronix]

Phoronix: Explicit GPU Synchronization Merged For XWayland

One year in the making, NVIDIA's code for explicit GPU synchronization in XWayland along with the X.Org Server DRI3 and Present extensions has now been merged! This is a big culmination of all the recent work around Wayland explicit synchronization and notably takes care of a number of NVIDIA driver problems on Wayland in the process...

https://www.phoronix.com/news/Explicit-GPU-Sync-XWayland-Go

[M.Bahr]

This merge request comes with a high risk to break things in the linux graphics stack and desktop environment. I am still not convinced if this is worth the risk.

PS: There seem to be some confusion in the replies to this. I am not talking about explicit sync but about code changes to x11 in general. Usually it is very risky to make code changes to x11 as it is very old and can break things in the linux graphics stack. Also some people seem to mistakenly think that xwayland was wayland. But Xwayland is an X server that runs under Wayland for compatibility reasons to make old x11 apps run under wayland. Xwayland is not wayland and this merge request is about xwayland therefore it concerns the embedded x11 part. Crazy how so many people here in this forum of all places got it wrong.

[damentz]

This merge request comes with a high risk to break things in the linux graphics stack and desktop environment. I am still not convinced if this is worth the risk.

Not only is it worth the risk, but it'll bring immediate benefits that help current use cases. This quote from Eric:

Beyond that, I believe these patches do offer at least one immediate, tangible benefit. With implicit sync (or poorly implemented explicit sync, for that matter), if the server is compositing from multiple clients, its framerate will be limited by the slowest-rendering one. With explicit sync, it's free to composite from whichever clients are ready while it waits for any slower ones. As an example, spinning a complex model in a Blender window won't make your entire desktop lag.

That doesn't just apply to Xwayland, of course. Wayland compositors could also reap similar benefits from explicit sync. I intentionally tried to make this implementation easily compatible with the proposed Wayland extension for that reason.

Pretty much if you're a power user of any type using more than one application at once, you benefit. Example, playing an intensive game on one display that dips below 60fps regularly, while watching a podcast on another. This is one of those last features that'll suddenly open up use cases that linux desktop really struggles with.

[pharmasolin]

Xorg fans, what else is wrong with a Wayland?

[M.Bahr]

With explicit sync, it's free to composite from whichever clients are ready while it waits for any slower ones. As an example, spinning a complex model in a Blender window won't make your entire desktop lag..

I was not talking about the frame synchronization benefits of explicit sync in general but about the complexity of x11. The code is an ancient mess and just slight changes can break the DE experience. That's the reason why the wayland aka x11 devs weigh up the pros and cons and usually choose to not alter it. X11 is a mining field and nobody wants a broken desktop environment.
By the way explicit sync is in the making and planning by the mesa devs way longer before this merge request. I think the time and effort would have been better spent focusing on wayland and not on x11.

[Jumbotron]

This merge request comes with a high risk to break things in the linux graphics stack and desktop environment. I am still not convinced if this is worth the risk.

And yet it must be done to move forward. X11 is dead. This is the 21st Century. Unix is gone replaced by Linux, which as a reminder, is NOT Unix nor was it ever going to be even at its creation by Linus. Along with the death of Unix and X11 so must also go other paradigms that Unix was created in and for, along with the attendant frameworks and ways of thinking. This is not the 1960s. We are now in the age of science fiction with a computer in everyone’s pocket or wrist or even finger ( think ring health monitors ) not to mention A.I. Almost everything is visual computing and that’s not even taking into account VR and AR. We have extended 1960s computing long enough. If things break, well, isn’t that one of Open Source’s supposed strengths ? To find those breakages even more quickly with more eyes on the problems with even more knowledgeable input than closed source so that progress can, well, progress even faster ?

[Sethox]

Xorg fans, what else is wrong with a Wayland?

Those are opinionated ones that usually argues for workflow and not the reasoning why it's not implemented (not a well structured argument)..
But there are plenty of wrongs with Wayland, but for some reason people cling to the "screen share" (and some others) that bi-products, e.g. pipewire, has been developed as a solution (yes, technically Wayland doesn't support screen share in of itself but it's a bad faith argument).

[Jumbotron]

Xorg fans, what else is wrong with a Wayland?

For the X11 zealots, simply that Wayland exists.

[RealNC]

Xorg fans, what else is wrong with a Wayland?

Its trolls.

[schmidtbag]

Xorg fans, what else is wrong with a Wayland?

I'm a Wayland fan but from what I understand, multi-monitor seems to be the most problematic feature regardless of DE. Not sure if that's still true since I haven't used Linux and multiple extended displays in years (I've cloned displays in the past year but that's not really the same thing).