GNOME X.Org vs. Wayland Performance + Power Usage On Fedora 32 With AMD Renoir Laptop

[raster]

That's way too much text around some good stats, but you're apparently only comparing clean just-logged-in sessions with nothing running.

IF you had read the text rather than just some numbers, you wouldn't need that reply as I clearly said it's not a brilliant benchmark but an indicator... perhaps you should read a bit more? Like the bits where I say I work on said code so you can be sure this indicator is not just a "on startup" but a general trend over a longer runtime too which I clearly indicate as well as this is just an indicator of where things are going. I actually profile my code regularly in gory detail from a range of angles, and an app is running - a terminal, not empty. You didn't read the bit where I said '"I spent all of 5 minutes on this and i need to go to sleep now" which is why it's not clearly not exhaustive. I clearly qualified and detailed the situation already - thus text to go read. It'd up to a much more involved benchmark to do that and I had and still have many other more important things to do, but I do know Wayland can easily do better than X11.

I've been doing this for over 25 years. Wayland is the first X alternative that has gotten any traction and it is decently designed. I've watched the hopefuls come and go and leave X11 still standing, until now. X11 is still standing but Wayland has not gone... it's growing. I was totally "Oh ANOTHER X11 replacement? Do you want to join the corpses of Y-windows, Berlin and DirectFB and possibly some other lesser knowns? Ummm no thanks" for the first few years because of this. I changed my mind as time went on for good reasons.

[tildearrow]

tildearrow My feelings are not important. So I haven’t shared any.

Yes you have.

[raster]

Any suggestions where and why they might have spent the extra space? Michael's results show a significant increase in memory use.

I can't say. You'd have to ask the Gnome guys. I know I pore over memory footprint for EFL and E regularly from massif through to pmap through to a custom tool I wrote to show all pages in all mapping and if they are read-only, not even mapped in or dirty. I do it often enough to have some idea where the memory goes, but I can't say for Gnome.

[tildearrow]

tildearrow Nobody wants to lose compatibility. Nobody wants to maintain compatibility. Entropy wins. Bye compatibility.

Bye Linux on Desktop.







Now what made you think your feelings are important?! (Note: saying "facts" makes you look like a huge liar, so it will not count)

A typical tactics from Phoronix. He didn't dispel doubts regarding benchmark. As a few people noted it was probably XWayland benchmark. There are dozens of comments, google bots will notice and the revenue will be higher.

Not bots,... All the people, who read comments page by page, and each page shows tons of ads. These ads generate quite an amount of money. And,.. What's the best way to get lots of comments? ... What about making some benchmarks comparing two technologies with the most flame war going on... ?

[tildearrow]

tildearrow What made you think your feelings are important?

I have never said so!

Nobody wants to lose compatibility! Ask companies and Average Joe.

Only some hardcore Linux and Apple fans don't care about compatibility.

[duby229]

The point was Wayland is quite young. It also matters when Gnome, KDE, Firefox started implementing it. X is 30 years with us and it's still far from perfect.

I think the real point is that X11 works now and has worked for decades and Wayland doesn't and never did. And the -only- way to run -almost all- things on Wayland (and with a seriously degraded experience) is -WITH- X11. So yeah, I think I'll settle for "far from perfect".

[Veerappan]

Maybe but they behave quite strange with multi-monitor configs. So I usualy don't use them. Most annoying thing is switching workspace with multiple monitors - mutter switches workspance on one minotor only, and I'm not sure at all how to make it to switch both monitors at once or just switch it on second monitor as necessary.

And this have nothing to do with either X11 or Wayland, since it behave the same on both.

I'm only on page 6 of the comments so far (of 16), but this is configurable in Gnome Tweak Tool (a.k.a. Advanced Settings). Go to the Workspaces vertical tab > Display Handling > select "Workspaces span displays".

I use it to bind my 2 desktop monitors and the laptop screen into one large workspace. Then when I use the workspace switcher, all of the displays switch together.

[duby229]

And what is it that you believe is missing? And please don't say "network transparency"?
Someone mentioned "color management" earlier. That is currently proposed as a protocol extension:

https://www.phoronix.com/scan.php?pa...ager-Calibrate
https://lists.freedesktop.org/archiv...il/040431.html

So that is at least one thing that is currently missing. Anything else?

Well, first of all, there is Hogsberg's own words on the matter...
https://www.linux-magazine.com/Onlin...ayland-So-Long

And then, there is proof in the pudding, so to speak. Take input lag as a perfect example, Gnome's solution to input lag is a cgroup rule that causes the processor to power up and brute force it and even just -moving the mouse curser- powers up the CPU. That's not a solution, that's a disgusting hack. And even then under CPU intensive loads input lag can become unbearable, where moving the mouse curser can have -multiple seconds- of lag.

[johnp117]

it seems to display the Wayland session tests in the 4th/5th charts ending much faster than X11?

Maze solver Wayland was run 3 times, whereas X11 was run 5 times (see the `N` in the benchmark graphs).

PSPDFKit WASM scored higher on Wayland and finished slightly faster, so nothing wrong. I suppose separate test runs can also be slightly differentially delayed based on waiting for the system metrics (e.g. temperature) to become stable between runs.