Announcement

Collapse
No announcement yet.

Wayland Preparing For 1.0 Stable Release

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

  • #51
    Originally posted by Scias View Post
    So, regular users will be able to get the very nice progresses of Wayland over X11, because let's face it (excluding network translucency) X11 is horrible concerning modern Desktop performance.
    While I certainly like Wayland as a modern/slim display server and will use it on my desktops as soon as the programs/toolkits support it under stable conditions, I don't quite understand what you mean with X offering horrible desktop performance and how Wayland will increase such performance. I've had no problems running all sorts of eyecandy on my desktop under X, and using programs like Blender/Maya/Mudbox certainly hasn't been slower than under Windows. If Wayland increases performance then great (I've always thought performance was a driver issue, not bound to X) but it's not as I've ever found X server to be some kind of bottleneck. Am I missing something?

    Comment


    • #52
      Wayland addresses a real need for 99% of users, and that is replacing the awful performance, esoteric API, and legacy bloat of X11, with something modern that's designed for the way people use computers now.
      What on Earth will be different if you remove X?

      At the moment, people program for a toolkit, which renders everything to an offscreen pixmap, then blits it over to the screen using direct rendering and bypassing the X protocol mostly.

      With Wayland, people will program for a toolkit, which renders everything to an offscreen pixmap, then blits it over to the screen using direct rendering. There is no X to bypass.

      The only improvement will be getting rid of lots of X cruft which is rarely used nowadays, but does not get in the way for 99% of the users, because modern toolkits and desktops have a pretty direct way of rendering already. Most of the stuff (such as graphics and input drivers) have moved out of X already and will be used by Wayland in exactly the same way they are used now. I don't see where the amazing speedup will come from. The slow and crufty parts of X are rarely used today, and only used for applications that Wayland can't do anyway.

      So while a minimalist rendering infrastructure would make the whole thing much more maintainable (which is a good thing, since X is quite scary), I'm not sure it will be that much more performant -- nvidida blob outperforms the windows driver already, there is no "X cruft" getting in the way for modern applications.

      Comment


      • #53
        Originally posted by XorEaxEax View Post
        While I certainly like Wayland as a modern/slim display server and will use it on my desktops as soon as the programs/toolkits support it under stable conditions, I don't quite understand what you mean with X offering horrible desktop performance and how Wayland will increase such performance. I've had no problems running all sorts of eyecandy on my desktop under X, and using programs like Blender/Maya/Mudbox certainly hasn't been slower than under Windows. If Wayland increases performance then great (I've always thought performance was a driver issue, not bound to X) but it's not as I've ever found X server to be some kind of bottleneck. Am I missing something?
        We'll find out when the 1.0 stable release comes out and I'm sure Michael will be the first to benchmark it.

        I'm kind of surprised to see posters here promising us that Wayland is going to bury X in terms of performance (which is the implication when they talk about how awful X performs and how Wayland is the savior). I mean, I totally get that X has many shortcomings that piss me off... and that it's huge... and that it's very old... and that the codebase is no doubt downright evil. Plenty of room for improvement in those areas maybe. But the current X scorn is blown so out of proportion. And I think Wayland is way over-hyped. JMO.

        Comment


        • #54
          Originally posted by pingufunkybeat View Post
          So while a minimalist rendering infrastructure would make the whole thing much more maintainable (which is a good thing, since X is quite scary), I'm not sure it will be that much more performant -- nvidida blob outperforms the windows driver already, there is no "X cruft" getting in the way for modern applications.
          I would expect that replacing X with Wayland would eventually lead to more efficient Radeon and Nouveau drivers, bad code and legacy cruft have surely been a detriment to the development of the FOSS drivers, taking precious man-hours away from real work.

          Even though the binary blobs have much better performance than the FOSS drivers, even on X, I would give X a lot of the credit for why the blobs are so buggy and unreliable most of the time compared to the FOSS drivers. None of these components are an island unto themselves.

          Comment


          • #55
            Originally posted by asdx
            Sigh.

            I don't have a limited view of Linux uses, maybe you're the one that have?

            When I was referring about X "holding Linux back" I was remembering about the past. Go back 10 years in time when we didn't had things like KMS.

            I have been using Linux for more than a decade now, and I can't count the times I had helped NEW and NON TECH-SAVVY Linux users to configure their xorg.conf by hand so that X would boot on their system, it was PATHETIC I tell you.

            At that time we didn't had KMS and most of the time X would just NOT boot at all and it would give NEW and NON TECH-SAVVY users a command prompt. How is that not holding Linux back? Those NEW users would try to configure their xorg.conf manually and if that failed, they would just go back to Windows.

            Then what? 1 less Linux user on the desktop. Again, how that didn't hold Linux back? Who is having a limited mindset here? Me? I don't think so.

            Also, what I said has nothing to do with desktop effects, sure they are nice, but what I was said has more to do with having a MODERN display system that JUST WORKS. And X failed us in the past in that department a lot, until KMS came and things started to improve. But still, it's time to move away from X for once.

            Also, network transparency is nice but nothing prevents us from having that on Wayland. And I think X11 will still be around for some time on top of Wayland.
            What you're referring to was more of a universal limitation of linux and unix environments back in those times. From that perspective, everything sucked back then. Quite frankly, ordinary folks just weren't the target of linux / unix back then. So configuring everything wasn't pretty. I mean damn... 10 years ago we had to practically compile everything ourselves (after we resolved all the dependencies and such of course).

            Maybe xorg.conf isn't the nicest configuration file to work with. I wouldn't think you'd need to rewrite a display server from the ground up just to come up with a better configuration system. There's nothing stopping anybody from writing a nice GUI to configure the display settings. The one that NVIDIA has is good enough for my needs, but I'm sure someone could come up with something better. Your criticism really isn't of X, but of a lack of configuration tools.

            A display that fails to configure by itself is a FUCKING FAILURE, and that was X and still is due to FUCKING proprietary blobs like FUCKING NVIDIA. Sure things have changed and things work better now that we have KMS and open source drivers supporting KMS, and we don't have to configure things by hand anymore, unless you are one of those people using the NVIDIA blob that still lacks KMS even today.
            I love my NVIDIA proprietary drivers.

            Comment


            • #56
              Originally posted by leeenux View Post
              I would expect that replacing X with Wayland would eventually lead to more efficient Radeon and Nouveau drivers, bad code and legacy cruft have surely been a detriment to the development of the FOSS drivers, taking precious man-hours away from real work.
              I doubt that. The DDX part of the drivers is relatively small, and only does 2d acceleration and X interfacing. This will likely be replaced by an X state tracker for Gallium in the future. The really nasty bits are in the kernel and Mesa/Gallium3d, almost no connection to X there (other than a few GLX extensions if I'm not mistaken).

              Even though the binary blobs have much better performance than the FOSS drivers, even on X, I would give X a lot of the credit for why the blobs are so buggy and unreliable most of the time compared to the FOSS drivers.
              Why? The blobs pretty much bypass X completely.

              I also think that some of you have no idea how far X has come in the last 5 years or so.

              Comment


              • #57
                Oh, and xorg.conf hasn't been needed in years. You only need it for tweaking non-standard options (and for the blobs).

                The only reason I have xorg.conf is that I'm running a multi-seat configuration.

                Which brings me to the next point -- Wayland can't do multi-seat. There's many things it can't do, and it will take years before it can do everything X can do right now. There's decades of experience and testing in X, and you can't duplicate that overnight, or expect it to be bug-free.
                oo.

                Comment


                • #58
                  Originally posted by schmidtbag View Post
                  i'm really excited to test wayland, and i'd really like to see some comparisons of it vs X. as much as i can tell amd and nvidia dislike working on linux drivers, if wayland is as simplified as stated, it might be EASIER to make drivers for it, and it might perform better.

                  as for kde support, i'd say that kde5 should be the beginning of wayland support. i think that'd be a great way to start a clean slate for kde, and reduce confusion.
                  I am sure nvidia and AMD would get on board with adding wayland support to their binary blobs. As for the other free drivers like nouveau and radeon these should potentially be simplified due to the simpler interfaces in wayland

                  I'm sure there may be support added to even KDE 4.x to add support for wayland. Its about time the display server got a good rewrite these days as a lot of the X interfaces are pretty much deprecated by now so emoving old X code but retaining essential functions would make it more efficient and faster.

                  Comment


                  • #59
                    Over time, the window system has been split apart into multiple pieces - the X server, the window manager, the compositing manager, etc. All of these pieces are linked by complex, asynchronous protocols. Performance suffers as a result; for example, every keystroke must pass through at least three processes: the application, the X server, and the compositing manager. But we don't need to do things that way any more; we can simplify the architecture and improve responsiveness.
                    All of these complex asynchronous protocols are sources of bugs, not just poor performance. And they make adding new features extremely complicated, wasting developer time in a system that has a manpower shortage.

                    That's why Wayland is important. Not because it will add 5fps to a game. Because in the larger picture X is becoming unmaintainable, at least if we want to continue adding the features that users expect from other operating systems.

                    Some people view network transparency as the core feature of X - it's the reason they use Linux. There's nothing wrong with that, except the concept that it must somehow be tied in to X. There is no reason you can't have just as good of a remote rendering workflow using another protocol. In fact, performance is likely to be much better due to the aforementioned simplification. When they Wayland devs say that network transparency won't be a core part of their protocol, it's because they are following the unix tradition of making 1 program to do 1 thing very well. Rendering over the network is the job for a separate part of the stack, that can focus entirely on that task, and do it better than if it had to be tied into the rest of the system.

                    Some people simply can't imagine anything being better than X. This is a lack of vision on their part, and it simply will never go away until they see with their own eyes that something better has arrived. That's fair enough - no one is expecting them to switch to a poorer solution than the one they have. Things will move along in their own time, as the pieces come together and everything starts working.

                    It will be several years before that begins to truly happen, but the time is coming.

                    Comment


                    • #60
                      Originally posted by phred14 View Post
                      Can you furnish a pointer to instructions to set up free NX? I'm in a situation where it needs to be unencumbered, and while I get the impression that that's possible, I've found a "twisty, turny path of directions, all different." I used to use dxpc back in the day, but that has withered, and I understand that that was one of the starting points for NX. I recently tried to dust it off. The source would build, but it crashed before doing anything useful. I've been interested in NX for some time, but never taken the plunge.
                      Not really... When I tried it I needed to edit the config file quite a bit too. But the archlinux wiki is not that bad there:


                      What did not work was "shadowing" sessions (connecting to "normal" desktop sessions), only over a hack with a vnc server which kind of defeats the purpose. I'm not sure if this is really supposed to work.

                      Comment

                      Working...
                      X