GNOME 3.26: Wayland vs. X.Org Performance - Boot Times, Power Use, Memory Use & Gaming

[sdack]

the video shows the same oepratin made in the two different environment. Visually you can see the difference:

https://youtu.be/Zsz7Shbnb9c

Here the technical explanation about the benefit of wayland.

I'm not asking for technical explanations on Wayland. Everything looks good on paper, but it needs to look good in practise, too. So I'm asking why it doesn't show in the Phoronix benchmarks. Shouldn't it have benchmarks where it shows a difference?

[Britoid]

Oh, also, I wish there was a way to tell whether something is running natively or through XWayland. Right now they are too seamless!

Run xeyes.

The eyes will only be able to track your cursor over apps running under Xwayland, additionally it also shows the terrible security design of X at the same time.

[profoundWHALE]

I'm not asking for technical explanations on Wayland. Everything looks good on paper, but it needs to look good in practise, too. So I'm asking why it doesn't show in the Phoronix benchmarks. Shouldn't it have benchmarks where it shows a difference?

Being on par with xserver (not including the one game benchmark) is extremely impressive. Both xmir and xwayland got to that point. Both had better performance when an application was written with the intent of it being used with that particular display server/protocol.

The issue is not that xwayland is at 95% compatibility as a drop-in replacement for xserver. The issue is that most applications still are expecting an xserver type display server, not a Wayland display server.

Until that changes, just like with Vulkan and OpenGL, performance will be around the same.

sdack, you really come across as arrogant when you say (paraphrased) "It hasn't convinced me of its usefulness so it is dead-on-arrival."

[leipero]

not being able to launch broken software is nice

The only problem with gparted are users, not software itself. Users got a habit to run it as root with "sudo gparted", that is not the way to run gui application, "gksu gparted" is. So, on some (if not all now?) distributions when you run it with sudo, doing something with some USB devices will result in "Operation completed bla bla bla.., but failed to inform core/kernel bla bla bla...", I don't know why, but using gksu instead would work properly without those errors. And having on mind that those applications are ment to be run with gksu instead of sudo, it's not broken application, but user error.

[sdack]

sdack, you really come across as arrogant when you say (paraphrased) "It hasn't convinced me of its usefulness so it is dead-on-arrival."

So I'm arrogant. I'm sure you can handle it. But when you spend a lot of time on something and it isn't better than what you've already got, then you've only wasted time, and this is not the definition of progress, but the definition of stagnation. Now I still believe it's worth trying even when it leads to failure, but for it to become a success does it need to convince, or it's born dead. I don't see this as arrogant. It's a reminder of what is at stake.

When you want more people to support Wayland then you need to have benchmarks showing its strengths. I believe that's why Phoronix has been benchmarking Wayland, to show where it's at and to put the news out there. If you think Wayland can do better then you should start discussing how to benchmark this.

[GrayShade]

it is experience with one specific implementation

it was not wayland comparison, but gnome-shell comparison. there are other wayland compositors

From my understanding, GNOME Shell is the best of the lot. Do you have other suggestions?

then both of you could continue using x

In other posts I mentioned I was running Wayland until it started crashing on me. I use X forwarding from time to time, but I didn't say I care about it. It would be insane, however, to claim that nobody uses it. There are plenty of voices complaining that Wayland has no alternative (although as I've said in this thread, there's nothing that prevents Wayland from getting this as an extension).

if they had enough devs, they wouldn't do ui in javascript in the first place

That's interesting, I've never seen anyone say it. But if you compare KDE's and GNOME's budgets, GNOME doesn't do too bad: http://www.linux-magazine.com/Online...ls-and-culture.

One more thing: even if you hate repeating the same arguments, please try to take a kinder attitude to the others. Being aggressive doesn't usually help getting one's point across.

[RussianNeuroMancer]

So to your question, "Is there a way to implement this with Wayland in mind?" is the answer still yes. You can take the existing code and make your own modification and implement it in anyway you want.

If your question is, "Is there a way for somebody to implement this for me and exactly the way I need it?" then I don't know the answer. You'll have to find somebody willing to do this for you and likely discuss a form of payment.

Well, the point is - for X11 there is existing code in X.Org Server that allow implementation of such workflow. Moreover, there is working products that using this code, like Ulteo. In Mutter and KWin case Wayland API for app delivery is even doesn't there, so I don't think anyone will ever start working on this (unless someone like Red Hat or Micro Focus will try to deliver product for such workflow, which is probably very unlikely since this days it's considered legacy in front of CEF-based web apps that run on end point hardware).

And by the way, having different remote desktop API per compositor is very unfortunate. But, well, maybe it's ok, since before Wayland there was Vino for Gnome Shell and krfb for KDE.

[GrayShade]

I'm not asking for technical explanations on Wayland. Everything looks good on paper, but it needs to look good in practise, too. So I'm asking why it doesn't show in the Phoronix benchmarks. Shouldn't it have benchmarks where it shows a difference?

Maturity, I suppose, keeping in mind that being written in JavaScript isn't always conductive to good performance. Also, Xorg has a lot of driver specific 2D fast paths that the Wayland compositors chose to throw away, preferring OpenGL instead.

I might be wrong, but I think GNOME Shell doesn't stop redirecting full-screen windows, which is a must for good performance in games.

[GreatEmerald]

Run xeyes.

The eyes will only be able to track your cursor over apps running under Xwayland, additionally it also shows the terrible security design of X at the same time.

Huh, that's hilarious. Never thought xeyes would have a use case!

[pal666]

They did say it will be faster.

and it is faster even in this benchmark. if you think they said it will be more faster, you have to back it up with quote

Why are you a fan of Wayland?

i have allergy to idiots