Announcement

Collapse
No announcement yet.

Intel 2D GLAMOR vs. UXA vs. SNA On Ubuntu 14.04

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

  • #11
    Originally posted by entropy View Post
    From my experience it's definitely worth it - performance wise.

    However, gen4 *features* some (apparently) broken designs that may cause corruptions with SNA:

    https://bugs.freedesktop.org/show_bug.cgi?id=55500
    Thank you very much! This why I like phoronix! No one has bothered to answer my question on 01.org.

    It's weird though. I've got GM45 and the only issue I've seen ware some isolated garbled fonts. Barely noticeable.

    I'm asking because it's not easy to test 2D performance. Gtkperf is kind of... synthetic.

    Comment


    • #12
      hm

      is improving not bad.

      xwayland? what the point? i want to see wayland not some hybrid with x.

      Comment


      • #13
        hm, so you don't want to use _any_ legacy x11 program?

        Comment


        • #14
          Originally posted by entropy View Post
          Chris Wilson does an amazing job, no doubt.

          I wonder what the motivation for intel is, to invest man power into
          GLAMOR as they already have the superior SNA...
          SNA has to be handtuned on a per-generation basis. While very effective, I can't imagine its very clean. Also I'm sure there's other things that Chris would prefer to be working on than micro-optimizing SNA instructions. With GLAMOR its all OpenGL, so you just optimize for a common set of optimizations then let the driver worry about it from there.
          All opinions are my own not those of my employer if you know who they are.

          Comment


          • #15
            Michael,could you make some sna-uxa-glamor power consumption tests?

            Comment


            • #16
              hm

              Originally posted by degasus View Post
              hm, so you don't want to use _any_ legacy x11 program?

              i want to see things be porting to wayland

              Comment


              • #17
                Originally posted by Veerappan View Post
                I believe that glamor will be necessary for accelerated 2D in XWayland
                If that's the case why the focus on X primitives (from keithp blog)?
                Once again I'll say that it is of that Wayland is putting the burden of 2d on the toolkits thus really leaving the folks who don't want/can't use gtk/qt out in the cold.
                Also, I completely understand the desire to not be burdened with an out dated drawing model but offering something like xlib for people to target directly would at least provide the option of a very long term/stable api.
                Last edited by liam; 12 March 2014, 02:01 AM.

                Comment


                • #18
                  Originally posted by liam View Post
                  If that's the case why the focus on X primitives (from keithp blog)?
                  Huh? That was talking about accelerating xrender - the 2d acceleration X provides, which is exactly what XWayland will require. (To accelerate X apps inside wayland)

                  Once again I'll say that it is of that Wayland is putting the burden of 2d on the toolkits thus really leaving the folks who don't want/can't use gtk/qt out in the cold.
                  It's not really toolkits. You just have to target a 2d library, like Cairo. You can use that with any toolkit you want, or with none at all if you prefer.

                  Also, I completely understand the desire to not be burdened with an out dated drawing model but offering something like xlib for people to target directly would at least provide the option of a very long term/stable api.
                  Nobody who has worked with xlib wants anything of the sort. Ever again. You can target cairo to get something semi-stable if that's what you want.

                  Comment


                  • #19
                    Originally posted by smitty3268 View Post
                    Huh? That was talking about accelerating xrender - the 2d acceleration X provides, which is exactly what XWayland will require. (To accelerate X apps inside wayland)
                    Not just xrender but x drawing in general. Regardless, the point was that they seem to be focusing on accelerating X rather than looking at ways to solve the gl 2d problem in general.



                    It's not really toolkits. You just have to target a 2d library, like Cairo. You can use that with any toolkit you want, or with none at all if you prefer.
                    The problem is that Cairo isn't the de facto acceleration library. Gnome uses it but wxwidgets, qt(though they can), EFL, etc don't. Moreover Cairo development seems to have slowed down greatly. It might be a good idea for gnome to adopt skis since that does accelerate well I'm addition to bring one of the fastest rasterizers around.

                    Nobody who has worked with xlib wants anything of the sort. Ever again. You can target cairo to get something semi-stable if that's what you want.
                    That's why I didn't say copy it. I ONLY meant some good 2d library (preferably that also handles input) that any Wayland desktop can be expected to have.
                    Linux is going to be utterly alone in not having something like that. Yes, it makes things very lean and future-proof but it raises the bar for developers.

                    Comment


                    • #20
                      SNA may be fast, but...

                      I keep reading about SNA's superiority, but in a dual monitor setup using SNA, everything is choppy. You can really see the SNA choppiness if you have wobbly windows enabled (with dual monitors).

                      Comment

                      Working...
                      X