Announcement

Collapse
No announcement yet.

Replacing X With Wayland On The Raspberry Pi

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

  • #31
    Originally posted by renox
    You're wrong: it was about server side management(allocation) of buffers.

    Originally posted by Thaodan View Post
    This is no reason, since the wayland protocoll spec isn't against it.
    Are you sure? I thought that with Wayland it is the client which has to submit the buffer..
    And if I understood correctly the point of Mir dev was that if you add server-side allocation on a protocol designed for client-side allocation you add a roundtrip which reduces performance (I have no idea whether this is correct or not).

    Comment


    • #32
      Originally posted by Delgarde View Post
      What kind of "average user" are you envisioning, given that the Pi is basically a geek toy - not a device for the computer-using masses?
      Average user meaning anyone who isn't a professional software/hardware developer. Pi may be a geek toy but it can be configured to be very user friendly, hence the reason why it is intended for schools for average people.

      Comment


      • #33
        Originally posted by schmidtbag View Post
        Average user meaning anyone who isn't a professional software/hardware developer. Pi may be a geek toy but it can be configured to be very user friendly, hence the reason why it is intended for schools for average people.
        But as I understands it, it is supposed to be a tool to teach software development so the intended user is hardly total unfamiliarity with technology. I think it is unlikely the physical design on rpi had been approved in EU as a consumer product to untechnological users.

        Comment


        • #34
          Originally posted by renox View Post
          Are you sure? I thought that with Wayland it is the client which has to submit the buffer..
          And if I understood correctly the point of Mir dev was that if you add server-side allocation on a protocol designed for client-side allocation you add a roundtrip which reduces performance (I have no idea whether this is correct or not).
          There's no way to do server-side allocation without a roundtrip, and Mir in fact enforces roundtrips for _all_ allocations, even if they could be done on the client side. I've implemented server-side buffer allocation on Wayland for an ARM-based platform and it worked just great.

          As for the 'but you could just do the same in X11' stuff: no, you can't. Compositing in X11 means an external process giving the server one final flat image to display and nothing else: so you have to use a rendering API like GLES and the copy that involves. The only possible way you can offload compositing to hardware is if you disabled compositing, so you're back to unsynchronised rendering, flashing and generally looking horrible. And even then, it's a hell of a lot of work which hasn't really been done before.

          Comment


          • #35
            By reading these threads(including other threads) it has become clear that wayland is already compromised and allowed for every and any use case people can think of, and is totally confusing. I thought the idea was, client controls composition, so we get prefect frames everytime. Now, from what I've read, it allows server, client, and everything in between. Does anyone how it's "supposed" to be used anymore? Becuase it is not clear to me.

            Comment


            • #36
              Originally posted by dh04000 View Post
              By reading these threads(including other threads) it has become clear that wayland is already compromised and allowed for every and any use case people can think of, and is totally confusing. I thought the idea was, client controls composition, so we get prefect frames everytime. Now, from what I've read, it allows server, client, and everything in between. Does anyone how it's "supposed" to be used anymore? Becuase it is not clear to me.
              You're just confusing things.

              - A client submits complete frames of a whole window atomically; every frame of a window is perfect.
              - A compositor composites these perfect window images such that the whole desktop is consistent every time; every frame on your monitor is perfect.

              Buffer allocation is a very different thing, it is only about how a client gets a buffer it can render into. Later the client gives it to the server for display. Who originally allocates the memory for a buffer does not affect perfectness.

              The only thing clients need to take care of, is to not modify a buffer's contents while it is being read by the server. The Wayland protocol explicitly says when a buffer can be read by the server and when a client can write into it.

              Comment

              Working...
              X