Announcement

Collapse
No announcement yet.

The First Wayland Benchmarks From Fedora 20 Show Great Promise

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Originally posted by Ericg View Post
    Depends on how much overhead you mean. Displays can only display SO much so if the overhead means the display can push at max 295fps, instead of 300fps, I don't see a problem there. Especially since most displays will cap at 60fps, and gaming displays cap around 120fps. I get your point about overhead, and we should look to find ways to minimize said overhead, but enough abstraction as necessary to provide multiple backends is likely a good idea
    Yes, it is now, that there are reasons to expect multiple backends. But there really wasn't any, back when xfwm4 (or almost any of the other X window managers) was written.

    Comment


    • #32
      Core i7-4750HQ has 4 Cores

      Just for the record: The Intel Core i7-4750HQ has 4 cores and is capable to run 8 threads. But you can, of course, force Intel to correct the ARK entry to 8 cores.

      http://ark.intel.com/products/76087/

      Comment


      • #33
        Originally posted by xeekei View Post
        How nice! A fellow LAS fan.

        On-topic:
        I almost get goosebumps when I think about how close we are to a X-less desktop. Unfortunately I'll probably have to leave Xfce for GNOME. Let's hope Classic works.
        X11 dependencies are in 69 out of 601 *.c files in the XFCE 4.1 sources (the top tarball expanded, then the sub-tarballs expanded from that one). I did not grep the *.h files. The rest of it is very much GTK-centric....

        So while it looks 'bad', it's a manageable effort to port one of the sweeter-looking DEs completely over to GTK, if they really chose to do it....100% XFCE on Wayland/GTK... definitely has a nice ring to it.

        This also re-iterates the importance of getting XWayland stable/merged, hopefully into X 1.15. Ignoring XWayland will only hinder Wayland/Weston/et al adoption.
        Last edited by MartinN; 10-08-2013, 04:17 AM.

        Comment


        • #34
          Didn't the Wayland devs claim XWayland would be faster than native X.Org by passing complex things?
          It is slower!

          I hope further Wayland and XWayland tweaks makes it faster.

          Would also be nice to see a comparison against XMir.

          Comment


          • #35
            Originally posted by uid313 View Post
            Didn't the Wayland devs claim XWayland would be faster than native X.Org by passing complex things?
            It is slower!
            Originally posted by phoronix
            This testing is similar in nature to our original Ubuntu XMir testing of comparing Ubuntu Linux with a Mir/XMir instance running versus an X.Org Server without anything else. The current XMir performance has improved thanks in large part to the composite bypass support, but already the test results in this article show the XWayland performance with common Linux-native OpenGL workloads being quite good.
            The reason is mentioned in the article, you are commenting on...

            Comment


            • #36
              Originally posted by uid313 View Post
              Didn't the Wayland devs claim XWayland would be faster than native X.Org by passing complex things?
              It is slower!

              I hope further Wayland and XWayland tweaks makes it faster.

              Would also be nice to see a comparison against XMir.
              Canonical is running everything under XMir, for little benefit. So XMir performance is really important, as everything slows down. For Fedora/GNOME the intention is to make everything run under Wayland, so priority is on that. XWayland should still be quick, but I rather have the focus on making sure things run directly under Wayland without anything inbetween. There is a big list of things that needs addressing. Most importantly it should not crash that easily! Then functionality (drag and drop, etc), then performance of XWayland.

              Comment


              • #37
                Originally posted by dee. View Post
                Looking good for Wayland. I know someone who's going to wear a monkey suit...
                I haven't watched them in a while but it's incredible just how ignorant he was about the whole mess. Wonder if he'll eat humble-pie or dance around the issue in a manner that would make a certain resident troll blush.

                Originally posted by Marc Driftmeyer View Post
                In 18 months 90% of the community will be dumping Xorg and using Wayland/Weston.
                The timescale should be about right for the competent and established distros but I'm afraid that it won't be enough for Canonical/Mint/lots of minor DEs to get their heads out of their... and start porting stuff.

                Comment


                • #38
                  Originally posted by Kostas View Post
                  I haven't watched them in a while but it's incredible just how ignorant he was about the whole mess. Wonder if he'll eat humble-pie or dance around the issue in a manner that would make a certain resident troll blush.
                  I don't really watch the show religiously, but I do check out episodes that look like they have interesting stuff in them. And it does seem like both have changed their tune quite a bit since that episode. Guess they caught a bit of flak from the viewers, and recently they don't seem to be so Ubuntu-centric anymore as they used to be.

                  Comment


                  • #39
                    Originally posted by uid313 View Post
                    Didn't the Wayland devs claim XWayland would be faster than native X.Org by passing complex things?
                    It is slower!

                    I hope further Wayland and XWayland tweaks makes it faster.

                    Would also be nice to see a comparison against XMir.
                    As I pointed out about uid, it theoretically CAN be faster, yes. In practice vs in theory though are very different and ontop of that this is still a work in progress-- wiait until everyone's had a while to play around with the code bases and can optimize things.

                    Comment


                    • #40
                      promise? I prefer facts and reality.

                      I prefer use X11. is more faster XD

                      Comment


                      • #41
                        Originally posted by edoantonioco View Post
                        I just hope than wayland support will be as easy as a simple new feature in an update of the game, and not to port the games almost from 0 again to be able to play it in wayland, in that case, valve would not be happy, because all their work is in the currently state of linux (and the same for the other people than had ported games to linux).
                        As far as I am aware this should mostly be handled by the abstraction layers being used, such as SDL.

                        Comment


                        • #42
                          the difference being this is barely usable, even OP stated that it crashed and it was hell to get it to work


                          for Xmir I can't say I had a single crash or show stopping bug

                          Comment


                          • #43
                            Originally posted by Pallidus View Post
                            the difference being this is barely usable, even OP stated that it crashed and it was hell to get it to work


                            for Xmir I can't say I had a single crash or show stopping bug
                            But then, you hadn't a Mir shell and compositor either, so what's the point?

                            Comment


                            • #44
                              The reasons to replace Xorg are well know, listed here.
                              Some months yet and I will use a well done display server.

                              Comment


                              • #45
                                As an additional comment, we found that Weston causes a useless copy (X compositing) because the X decoration window it uses for X application has different visual than the application. That means X has to do an additional copy. (That's going to be fixed)

                                After looking closer at the figures, the performance hit of Gnome XWayland is not due to a similar issue. It isn't an overhead for every frame, but a constant overhead per second.
                                drago01 comments tells that Gnome doesn't support yet bypassing Wayland compositing. It makes sense to say the performance hit observed is entirely due to this: all benchmarks run >= 60 fps, causing Wayland to composite 60 times per second (= constant overhead).

                                Currently work is done to support vsync in XWayland (almost done), and (harder) respect the Wayland buffer release semantic, in the cases in which we can do it without impacting performance, to suppress tearings.
                                It is also possible that performance increases if we manage to implement a DDX feature proposed by Chris Wilson a few years ago, implemented in the intel DDX, but whose patch to use the feature has not been merged in X yet (but maybe it will be merged with XWayland, who knows?).
                                Last edited by mannerov; 10-12-2013, 08:18 AM.

                                Comment

                                Working...
                                X