Announcement

Collapse
No announcement yet.

EGLStreams XWayland Code Revised Ahead Of X.Org Server 1.20

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

  • #11
    Isn't this a EGL extension to do more stuff with GBM?
    Again I'm not a developer so I might be wrong but it seems like that to me.

    Comment


    • #12
      Originally posted by DanL View Post
      I find it funny that the people complaining about "maintenance burden" act like they're the ones who are stuck with it. The code was written by a RedHat-paid dev and longtime Linux graphics contributor (Adam Jackson).
      Did Adam Jackson add EGLStreams support to KDE, Mate, Cinnamon, and anything else too?
      Because that's the "maintenance burden" the complaints are for.

      Having support for it in X11 or XWayland does not translate in "everyone can use EGLStreams".

      As for GBM, it's not a good open standard since it's tied to Mesa/Gallium. We need something more agnostic.
      EGLStreams is in the same boat, for that matter. It was code-dropped from NVIDIA without asking anyone else's opinion.

      The GBM zealots saying, "F Nvidia and Khronos. We're the standard because we were here first and we're not interested in a better solution." are the ones causing fragmentation.
      Lol no. They weren't interested in NVIDIA's own code-dropped solution, they were open to make a better API where their needs could also be addressed. https://www.phoronix.com/scan.php?pa...Device-Mem-API

      Comment


      • #13
        Originally posted by starshipeleven View Post
        As opposed to NVIDIA actually conforming to standards every once in a while and not requiring devs to add support to their own stuff everywhere if they want to run on NVIDIA cards.

        That's an NVIDIA fragmentation problem and not a community fragmentation problem?

        That's not the point. If everyone bends over backwards to NVIDIA then they will have 0 incentive to actually follow standards, which is one of the reasons Linux is great, many things can be swapped to have different effects and everything will still overall work.

        If it goes on like this, NVIDIA cards will only work on the Standard Desktop (GNOME) and in commercial products using Qt-based compositors when X11 is finally dropped and Wayland becomes the default, as none else is interested (KDE) or has the manpower (everyone else) to deal with EGLStreams.

        Which means it will cause fragmentation on what DEs will support it for no other reason than NVIDIA being run by a bag of dicks that can't just follow the fucking standards even if it is somewhat mildly inconvenient for them.
        Then go use Arcan already if you have nVidia and want a modern display server that's not X.

        Comment


        • #14
          Originally posted by starshipeleven View Post
          Did Adam Jackson add EGLStreams support to KDE, Mate, Cinnamon, and anything else too? Because that's the "maintenance burden" the complaints are for.
          Nope, but that's a moot point. None of the desktops are forced to use it or support it. The point is that it's available for those who do want to use it, and it's not hurting anyone who doesn't. If the devs of a particular desktop don't want to bother, they can mark any feature request or bug report as 'Will Not/Fix Support". As I feel like I said a bazillion times, if you don't like it, don't use it. But quit trying to piss on others' parade just because you hate Nvidia.

          EGLStreams is in the same boat, for that matter. It was code-dropped from NVIDIA without asking anyone else's opinion.
          Nvidia has already acknowledged this and seeks to move forward with a new allocator. Until that comes to fruition, EGLStreams is a stopgap alternative solution.

          Comment


          • #15
            Originally posted by Vistaus View Post
            Then go use Arcan already if you want a modern display server.
            Fixed.

            That's still very much up in the "build your own GUI from scratch" alley though, like people who chose their own WM and compositor and things instead of relying on a fully fledged DE. I don't think common distros are going to package Arcan/Durden DE in the near future, if ever. Of course Arch and Gentoo have it.

            Imho it's a bit of a shame, Arcan is pretty cool and can do things most others can't, regardless of EGLStreams support.

            Comment


            • #16
              Originally posted by DanL View Post
              None of the desktops are forced to use it or support it.
              Hah, none is forced to use or support it you say?

              So RedHat can just say "lol dawg, my Standard Linux DE GNOME which is the default GUI in my RHEL OS I use as source of revenue does not run with the most powerful/well-known/only-ones-supporting-CUDA GPUs on the market with the best performing driver for workstation (or only driver if we talk of CUDA) applications on Linux because they are not following standards"?

              Guess why GNOME is the only one supporting it of the major DEs while everyone else can afford to not care at all and just slap a WONTFIX tag on bug reports about this.

              The point is that it's available for those who do want to use it,
              Those who? The users that want to use NVIDIA cards on Linux with their favourite DE on Wayland? No, of course not. How about you don't cherry-pick arguments to make this look different from reality?

              But quit trying to piss on others' parade just because you hate Nvidia.
              I'll quit when you stop white-knighting for NVIDIA by cherry-picking things.

              Nvidia has already acknowledged this and seeks to move forward with a new allocator. Until that comes to fruition, EGLStreams is a stopgap alternative solution.
              This does remove what little push there is for NVIDIA to develop the new allocator in the first place. Sure they might still eventually do it, but this is more due to their opensource developers being good people, more than NVIDIA's actual priority.

              Comment


              • #17
                Originally posted by starshipeleven View Post
                Hah, none is forced to use or support it you say? So RedHat can just say "lol dawg, my Standard Linux DE GNOME which is the default GUI in my RHEL OS I use as source of revenue does not run with the most powerful/well-known/only-ones-supporting-CUDA GPUs on the market with the best performing driver for workstation (or only driver if we talk of CUDA) applications on Linux because they are not following standards"?
                Gnome is not the standard desktop. (Did GhostofFunkS hack your account or something?) GBM is not a good open standard either as already explained.

                Those who? The users that want to use NVIDIA cards on Linux with their favourite DE on Wayland?
                No, I was clearly referring to desktop devs.

                This does remove what little push there is for NVIDIA to develop the new allocator in the first place. Sure they might still eventually do it, but this is more due to their opensource developers being good people, more than NVIDIA's actual priority.
                The current situation isn't exactly encouraging them.

                Nvidia asks for input/ideas on new allocator and gets little help because of "GBM or die" mentality:
                "ZOMG! Nvidia is expecting the community to do their work for them!"

                Nvidia prototypes and starts development of new allocator:
                "Nvidia is just code-dropping stuff and not collaborating!"

                I'll quit when you stop white-knighting for NVIDIA by cherry-picking things.
                Who's cherry picking? I'm responding to all of your BS.

                Comment


                • #18
                  Originally posted by DanL View Post
                  Gnome is not the standard desktop. (Did GhostofFunkS hack your account or something?)
                  joke ---->
                  You

                  GBM is not a good open standard either as already explained.
                  While this is technically true, I'd like to point out that the only ones complaining about GBM is NVIDIA, everyone else agreed to make/use it. You know, AMD, Intel, and other open source driver devs that actually develop and use Mesa at all.

                  NVIDIA driver does not use Mesa. It replaces Mesa with its own 3D userspace library.

                  No, I was clearly referring to desktop devs.
                  No duh. That's why I said you are cherry-picking. You're purposefully ignoring the big picture for the sake of white-knighting.

                  Nvidia asks for input/ideas on new allocator and gets little help because of "GBM or die" mentality:
                  No, they get little help because the other parties involved are not really interested in it. Did I mention that GBM was ok with mesa devs already?

                  EGLStreams was not interesting enough to justify the switch from GBM, as simple as that.

                  Nvidia prototypes and starts development of new allocator:
                  "Nvidia is just code-dropping stuff and not collaborating!"
                  Here you are mixing up things for the sake of fitting your fictional narrative.

                  EGLStreams is not a prototype but a fully-fledged EGL extension, and it was code-dropped. https://www.khronos.org/registry/EGL...KHR_stream.txt

                  Since as I said everyone else in the driver development has already agreed to use GBM, they see 0 reason to switch all to EGLStreams, is NVIDIA paying them for that work, QA and such? No. Is it better than GBM for their usecase? No.

                  Which is why they asked to come up with "something better than GBM for everyone", not "something better than GBM only for NVIDIA".

                  The new API which is being slowly developed and should be something that has more chances to actually be accepted if they play their cards correctly is here and is called UDMA. https://github.com/cubanismo/allocator

                  Who's cherry picking?
                  You. It's 3 posts already and you still can't get your facts straight.
                  Last edited by starshipeleven; 22 April 2018, 02:45 PM.

                  Comment


                  • #19
                    As far as I can tell, eglstreams support in XWayland has an #if switch, so those distributions which don’t want to deal with it, can just compile without it. Or not?

                    The “new allocator” API seems technically better than GBM even with modifiers (my impression based on xdc 2017 presentations) and it seems to be what NVIDIA wants as an open source solution, so here is a starting point to get nvidia involved with open source.

                    eglstreams is not anymore what nvidia wants, it’s for nvidia users, as a practical part of transition to Wayland. Also, as a small developer, I probably wouldn’t want to develop for Wayland until all major GPUs support it.

                    AMD recently referred to “freedom of choice”, and I support that, however it needs to go each and every way. Vulkan has roots in AMD technology, yet nvidia supports it just fine.

                    Comment


                    • #20
                      Originally posted by starshipeleven View Post
                      joke ---->
                      You
                      Huh? Joke is on me? Care to explain? (Edit: On second thought, forget it. I don't care)

                      NVIDIA driver does not use Mesa. It replaces Mesa with its own 3D userspace library.
                      "No duh." That's why EGLStreams is a good option for Xwayland.

                      You're purposefully ignoring the big picture for the sake of white-knighting... Here you are mixing up things for the sake of fitting your fictional narrative.
                      White knighting? You've got to be kidding. And you are the one with the fictional narrative that merging these patches affects non-Nvidia users and will enslave all Linux users.

                      EGLStreams was not interesting enough to justify the switch from GBM, as simple as that... Since as I said everyone else in the driver development has already agreed to use GBM, they see 0 reason to switch all to EGLStreams,
                      Straw man argument. No one is asking users/devs to switch from GBM to EGLStreams. It would be a nice option for Nvidia users if the desktops offered them that choice, but it's not mandatory.

                      It's 3 posts already and you still can't get your facts straight.
                      The feeing is mutual. It's 3 posts and you still haven't given a good reason why merging the Xwayland patches hurts non-Nvidia users. That is the crux of the issue.
                      Last edited by DanL; 22 April 2018, 06:03 PM.

                      Comment

                      Working...
                      X