Announcement

Collapse
No announcement yet.

GNOME Merge Requests Opened That Would Drop X.Org Session Support

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

  • Originally posted by qarium View Post
    in python when i use PyTorch it reports my ROCm Vega64 and AMD PRO W7900 as CUDA card.
    Ah, I was talking more about training/learning for models rather than just using them.

    Comment


    • Originally posted by kmansoft View Post

      Possibly related - when running under Wayland, apps like web browsers are unable to save / restore their window location / size. Not a deal breaker but really annoying and is a "this in 2023, really?" moment. Affects both Brave Browser and Firefox for me.
      Yes that's one of the problems stemming from it. There's many more: you can't have your own scripts to reposition windows as you want, nor query them at least (for auto hotkeys for example), which is a tragedy for power users.

      And there's many more examples in scientific apps, someone did make an issue about it, but it will never be accepted, or accepted in 2050: https://gitlab.freedesktop.org/wayla...e_requests/247

      People who use their PCs as kiosks, toys or just for fullscreen gaming don't find an issue (but Wayland has many other issues with gaming, such as latency, albeit some of them are "fixed" now but it's still cringe how crippled it started by design).

      Comment


      • Originally posted by avis View Post

        Any sufficiently complex software contains vulnerabilities and can be hacked. It does not mean it's insecure.
        Any sufficiently complex software contains vulnerabilities and can be hacked. It does not mean it's insecure.
        Any sufficiently complex software contains vulnerabilities and can be hacked. It does not mean it's insecure.

        And lastly you cannot stop calling others idiots. This is the last time I've seen or replied to your messages.
        Yes, any software that has known vulnerabilities are inherently insecure, there is no other way to describe it other than "you".
        Some of those software are not feasible to fix due to reasons described below.
        Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

        Here is just one quote from a real developer, who knows what he is talking about
        "GLX is a horrible demotivator! 80,000 lines of sheer terror." and "In the past couple of months I've found 120 bugs there, and I'm not close to done."

        You still didn't get it, right? Oh I forgot you were not responding (hopefully) anymore.

        Comment


        • Software contains vulnarabilities and can be hacked?
          Nobody alive wants to fix 80,000 lines of sheer terror?
          I rather want to keep my mouse macros depending on non-isolation of ui applications, with a change being a tragedy for power users?

          Last edited by reba; 12 October 2023, 12:36 PM.

          Comment


          • Originally posted by Weasel View Post
            And there's many more examples in scientific apps, someone did make an issue about it, but it will never be accepted, or accepted in 2050: https://gitlab.freedesktop.org/wayla...e_requests/247

            For a variety of cases it's desirable to have a method for negotiating the restoration of previously-used states for a client's windows. This helps for e.g., a compositor/client...

            The reality at this point 247 might be never accepted one of the other proposed methods may end up getting up instead.

            Even the person behind 247 admits this is not for the majority of the applications. This is a horrible one. Wayland relative positioning is beneficial 90% of applications by reducing cpu usage and removing render/drawing issues. You know when you move a window and a popup should be above the window and it comes up where it window was because of absolute position in application and this results in user moves window goes to work on the application they just got clear view of and opps they clicked on pop up from another application causing something random to happen. Like it or not there is advantage to less advanced users having the default relative position. Its the old problem of the last 10% that is the hard bit.

            The fact that relative position is beneficial for interaction with 90% of applications kind of explains the hard line problem why should we have absolute position at all.

            Do remember absolute position application working in relative position as wine proved in their Wayland tests 90% of the time this works if you call the top left coner of the window the 0.0 point instead of true absolute position to screen. So making 90% of all desktop applications operate in a relative positioning world is simple.

            Can you see a problem now. People asking for absolute position mostly have been asking for it universally. Reality is the experiments with relative position done with Wayland you don't want absolute position universally in fact most of the time you don't want absolute position as 90% as absolutely position only allowing bad things to happen that would have been prevented using relative position.

            Originally posted by Weasel View Post
            People who use their PCs as kiosks, toys or just for fullscreen gaming don't find an issue (but Wayland has many other issues with gaming, such as latency, albeit some of them are "fixed" now but it's still cringe how crippled it started by design).
            There are a lot of general webbrowser, office suite, email client kind of users who are not finding major any issues with Wayland.

            Weasel you need to think again if you think Wayland performance is crippled by Wayland design with games.

            Every time someone like retroarch and so on has spent the time to track down gaming performance issues with Wayland the answer has never been the Wayland protocol with the problem.

            Lets look at a X11 compositor for a second.
            1) X11 compositor can only use 1 thread to interface with X11 server or things go badly wrong.
            2) X11 compositor can only have a single refresh rate for all monitors doing multi refresh rate X11 server goes belly up.

            Wayland protocol design input process and output rendering can in fact be in different threads. Yes each monitor can be rendered by it own thread. The socket system is able to handle multi threads putting different packet in and not getting the mangled with each other.

            The worse performance problems with gnome mutter with Wayland in fact comes from legacy X11 compositor design that has not been stripped out yet.

            Also even that the Wayland protocol allows multi threads for input and outputs not all Wayland compositors have implemented this because doing multi thread code is not what you can simple all the time.

            Yes Wayland compositor implementation issues mostly caused by legacy X11 compositor code designs and Video card driver issues mostly scheduler(if you are not using Nvidia if using Nvidia darn stack of broken driver stuff +scheduler issues) and CPU scheduler turn out to be the gaming latency problems more than anything else. All of these causing more latency problems than we don't allow tearing.

            Mutter reason for not being able to do VRR with Wayland is stacks of X11 old compositor code that does not work with VRR.

            https://discuss.kde.org/t/wayland-la...r-enabled/2352 Lot of lag issues are like this one. Yes under X11 you can optional enable KMS mouse cursor drawing but it off by default but under Wayland compositor this has habit of being on by default. The option being on exposes a driver defect. This is also why you at times see the reverse where person says Wayland compositor has less lag than X11. Difference in default used features in graphics drivers brings out different bugs of course.

            Comment


            • Originally posted by kmansoft View Post
              I wanted to use that but they did not have a version which supported my non lts Ubuntu. I don't run lts, preferring more fresh software.
              Just imagine if on windows they only supported windows 10 but not 11.
              I did manage to bypass the distro check but then the driver failed to compile.
              I paid for their hardware, it's supposed to support Linux, I'm even running perhaps the most popular Linux distro of all - still it would not work.
              ​​​​​​Just how much was I supposed to endure for the money I paid, for the amd hardware?
              Oh and there was no indication that the freezes bug was not present in the amd pro driver anyway.
              To me the LTS version is the single one reason to use Ubuntu at all. On servers. I consider the versions between the LTS releases to be development snapshots one shouldn't use without specific reasons to do so. They are not totally fresh and also miss the enterprise level stability and support.
              So use Ubuntu LTS and install drivers manually or use Fedora for the latest open source drivers.

              Comment


              • Originally posted by Neuro-Chef View Post
                To me the LTS version is the single one reason to use Ubuntu at all. On servers. I consider the versions between the LTS releases to be development snapshots one shouldn't use without specific reasons to do so. They are not totally fresh and also miss the enterprise level stability and support.
                So use Ubuntu LTS and install drivers manually or use Fedora for the latest open source drivers.
                Hmm seems too complicated - so I'd have to switch distros just to use the hardware.

                And I'll reiterate - there is no indication that the bug was fixed in amd pro drivers.

                I simply switched to nvidia and it just worked.

                If amd wants my money they'll have to stop trying to make me work so hard.

                Comment


                • Originally posted by kmansoft View Post
                  Hmm seems too complicated - so I'd have to switch distros just to use the hardware.
                  Finding the right distribution for your needs can be a journey of years
                  Originally posted by kmansoft View Post
                  And I'll reiterate - there is no indication that the bug was fixed in amd pro drivers.
                  Quite possible, but again, the amd pro drivers are just the last option nowadays.
                  Originally posted by kmansoft View Post
                  I simply switched to nvidia and it just worked.
                  It's question of preference. I like updated systems but I hate having to fix stuff after updating. It can't be 100% avoided with Linux in general, but not relying on separately maintained drivers that are not open source software helps me a lot.
                  Originally posted by kmansoft View Post
                  If amd wants my money they'll have to stop trying to make me work so hard.
                  It's not so much AMD fault, just the ways of the Linux world. You just can not test and support every possible distribution in all configurations.
                  EDIT: But as a hardware vendor you can contribute to the mainline kernel and projects like mesa, that basically all them use!
                  Last edited by Neuro-Chef; 12 October 2023, 07:51 PM.

                  Comment


                  • Originally posted by Weasel View Post
                    Ah, I was talking more about training/learning for models rather than just using them.
                    does this mean you claim that it only works for PyTorch Inference mode ????

                    no of course training and learning models also work

                    Phantom circuit Sequence Reducer Dyslexia

                    Comment


                    • Originally posted by kmansoft View Post
                      Possibly related - when running under Wayland, apps like web browsers are unable to save / restore their window location / size. Not a deal breaker but really annoying and is a "this in 2023, really?" moment. Affects both Brave Browser and Firefox for me.
                      First of all, they aren't, and secondly, this feature isn't part of the Xorg implementation either. It's totally up to the WM to either obey or ignore a window's request to reposition or resize itself. So just like it is down to the WM on Wayland as well.

                      Apps shouldn't be allowed to change their own geometry anyway.It's an absolute nightmare when some apps remember their position but not the size or vice versa, or when their "restore" functionality is buggy and mostly fucks up.

                      The window manager should remember window geometries and restore them after a restart -- and preferably it should be user-configurable. At least back in the Xorg days KDE's window manager had all these tools built-in and I was using them all the time. (Sometimes only to circumvent the app-specific broken implementation!)

                      Comment

                      Working...
                      X