We Have Mir & Wayland, But There Still Could Be X12

[LinuxGamer]

Total system performance will go down as many e.g. Cairo-based applications will run slower or hog CPU or GPU.

really as we only seen lower cpu or gpu usage from Wayland so far

[JS987]

really as we only seen lower cpu or gpu usage from Wayland so far

Wayland applications can have lower usage on some machines, but higher on others.

[AJSB]

RHEL is moving to Wayland in like a year or so from now and they're going to have Xwayland anyways so the user will not see any change at all

Yes, i was forgetting about Xwayland, but...

...AFAIK, performance under Xwayland (and xMir for that matter) is lower than in X...so, they will notice, at minimum , that difference.

[mrugiero]

Wayland applications can have lower usage on some machines, but higher on others.

First, since almost all applications actually render outside the X11 protocol, and not through Xlib, there should be no noticeable change on performance on this area. Second, yes, it will probably be slower than non-composited X on memory constrained machines, because of the extra buffers required for compositing. If memory is limited, there might be paging to disk, slowing down the computer.

On the other hand, rendering toolkits could use acceleration, too, either with an agnostic, specialized 2D driver (kind of DDX, but generic for any toolkit) or via OpenGL, so this possible hog of the CPU sounds like a temporary problem, if any.

[sp82]

Yes, i was forgetting about Xwayland, but...

...AFAIK, performance under Xwayland (and xMir for that matter) is lower than in X...so, they will notice, at minimum , that difference.

I don't think you can see the difference in the single app. The performance are not so bad in Wayland+Xwayland and you have all the benefits that wayland carry on, the smooth composite is in.

For sure you can notice you are in wayland because the window composite process is smooth and perfect.

I think Wayland+Xwayland is a better X.

[dee.]

Yes, i was forgetting about Xwayland, but...

...AFAIK, performance under Xwayland (and xMir for that matter) is lower than in X...so, they will notice, at minimum , that difference.

The performance is lower in XMir because of the way Canonical is using it, running a whole X-based DE on top of it. This is because the compositing of windows is still done in X, but with the added overhead of XMir/Mir in the mix. That's not a smart way of doing things.

XWayland is meant to be used as a rootless application-specific server, in such a way that there's a private, rootless XWayland X-server server for each X application. In this way, the X applications can actually be faster than they are in plain X, as each application can run fullscreen with composite bypass inside its own X-server, and the window composition is done by Wayland, which is more efficient than X.

[skies]

[QUOTE=Daktyl198;361370]Okay, let me rephrase that:
SHM and DRI2 (which are the only two that are used anymore) don't work over a network.
Thus, X.org is "Network capable" but not "Network Transparent".

AFAIK SHM and DRI2 is not part of the X Window System standard. They are used by extensions.

And the XFree86 server is not part of the X11 standard either. It's just one server implementation among many.

DRI2 is Linux specific right? or do AIX, IRIX, Solaris and HP-UX have DRI2?

[LinuxGamer]

Yes, i was forgetting about Xwayland, but...

...AFAIK, performance under Xwayland (and xMir for that matter) is lower than in X...so, they will notice, at minimum , that difference.

Xwayland is not supporting compositing bypass atm is why its just a little slower than Xorg on some apps

[Remote User]

As I understands it this is not real networktransparens. To get real networktransparens you need to move the network layer to your toolkit in this case gtk? Only if you use real xlib based app you get the network transparency. (or is I compleatly off?)

Your understanding is wrong. My client applications do not rely on gtk (or qt) but render the touchscreen GUI through Xlib. It's not difficult to do this and the advantages, for me at least, are both unique and substantial. It is common for people who don't use X in the way that I use it, which is pretty much the traditional way, to assert that network transparency isn't particularly useful or valuable, and that virtually no one uses it as a network transparent display protocol, but I consider them to be uninformed, if not disadvantaged.

[Remote User]

Well, if all you need for remoting is sending dumb images across the network, I think you need not be afraid, it will be achievable with any of new display servers.

No, I do not send 'dumb images' across the network. I use Xlib to render the GUIs. You are making the same assumptions that nearly everyone else makes, that anyone doing remote display is NOT doing it with the X protocol, and they, like you, are wrong in your assumptions and in any conclusions you draw from them.