Announcement

Collapse
No announcement yet.

Some Fresh Radeon EXA/GLAMOR 2D Acceleration Benchmarks

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

  • Some Fresh Radeon EXA/GLAMOR 2D Acceleration Benchmarks

    Phoronix: Some Fresh Radeon EXA/GLAMOR 2D Acceleration Benchmarks

    It's been a while since last publishing any 2D-focused benchmarks since overall the state of Linux 2D acceleration architectures like GLAMOR and Intel's SNA have become good enough that it's no longer a frequent concern by desktop users. However, as some recently have been inquiring about updated numbers, I ran some fresh Linux 2D benchmarks on different Radeon GPUs using the open-source driver stack...

    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

  • #2
    GLAMOR, SNA, EXA, XAA...
    Am I the only person seeing it stupid to reinvent 2d acceleration architecture once in a while?

    Comment


    • #3
      Does that table show which is faster EXA or GLAMOUR? It just looks like you're saying which card is better

      Comment


      • #4
        Originally posted by reavertm View Post
        GLAMOR, SNA, EXA, XAA...
        Am I the only person seeing it stupid to reinvent 2d acceleration architecture once in a while?
        Yes you are!
        Hardware changes, use cases change, the surrounding infrastructure changes, ... why should 2D acceleration be set in stone?

        Comment


        • #5
          Originally posted by FireBurn View Post
          Does that table show which is faster EXA or GLAMOUR? It just looks like you're saying which card is better
          Because he is. He didn't compare the two technologies at all on that graph.

          Comment


          • #6
            Originally posted by Daktyl198 View Post

            Because he is. He didn't compare the two technologies at all on that graph.

            I don't get the point of the article then, it's just another bunch of random benchmarks shoved up again. Pointless

            Comment


            • #7
              Yeah, this isn't quite what I had in mind.
              For one, I'm interested in how sna compares to glamour (I'd imagine sna is still faster but glamour had been the one getting the attention lately from the Intel devs). However, I think Michael has said he's going to wait for the release of the next x server before testing again. If guess this was just a quick preview.
              Regarding 2d rasterization driven by 3d APIs, have any of you heard of TyGL? I just heard about it yesterday when reading an old summit report by the Servo team regarding their 2d drawing options. The work was done by a Hungarian university in Szveged and was sponsored by ARM ( https://github.com/szeged/TyGL).
              So, it's intended for browser 2d (presumably an emphasis on Canvas) as it's a fork of webkit and they claim that (on a Chromebook with cortex a 15 and Mali graphics) it is at least twice as fast as CPU rasterization, and up to 11x faster.
              As far as I can tell, no one has done anything with this work, and no one else has a COMPLETE GPU driven 2d rasterizer (including direct2d).

              Comment


              • #8
                Originally posted by reavertm View Post
                GLAMOR, SNA, EXA, XAA...
                Am I the only person seeing it stupid to reinvent 2d acceleration architecture once in a while?
                If it had happened over a few years then yeah probably stupid, but we're really talking about evolution over maybe 15 years, an underlying shift of acceleration HW from 2D to 3D, and an evolution of typical apps from using primarily 2D operations to primarily rendering with toolkits over GL.

                The XAA arch was focused on traditional 2D; EXA added some functions that needed 3D hardware on most accelerators but permitted much nicer looking desktops, and SNA was (AFAIK) further optimized for the specific characteristics of Intel iGPU hardware. GLAMOR had been around for a long time but HW and GL drivers finally reached a point where it could be broadly used without slowing everything down.
                Test signature

                Comment


                • #9
                  Originally posted by FireBurn View Post
                  Does that table show which is faster EXA or GLAMOUR? It just looks like you're saying which card is better
                  Yes it does.

                  Michael,
                  As you mentioned in your article, we are interested in how GLAMOR compares to Intel's SNA in terms of performance and power usage. But this doesn't really tell us anything.

                  Comment


                  • #10
                    Originally posted by reavertm View Post
                    GLAMOR, SNA, EXA, XAA...
                    Am I the only person seeing it stupid to reinvent 2d acceleration architecture once in a while?
                    That's exactly the point of GLAMOR: to use the driver's OpenGL for the Xorg 2D operations rather than custom hardware-specific paths like XAA/EXA so this insanity should end at last.
                    Fortunately, in the Wayland or xf86-video-modesetting world there's no more 2D/DDX driver at all.
                    Last edited by Scias; 12 February 2016, 01:05 AM.

                    Comment

                    Working...
                    X