Announcement

Collapse
No announcement yet.

A New & Exciting OpenGL 3 Benchmark To Run

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

  • A New & Exciting OpenGL 3 Benchmark To Run

    Phoronix: A New & Exciting OpenGL 3 Benchmark To Run

    There's finally a new and visually exciting OpenGL benchmark to try out for Linux, OS X, and Windows users alike. The benchmark also supports OpenGL 3.x contexts for making testing more exciting with regard to the Linux graphics driver stack...

    http://www.phoronix.com/vr.php?view=MTQwNjE

  • #2
    I've been trying to make latency-oriented benchmarking profiles for PTS, but I couldn't find any documentation on how to write one and looking the existing examples didn't get me very far.
    Is there any documentation on writing test profiles? Maybe a reference manual for the XML definitions?

    If not, where can I at least look up the commonly used values? Maybe on OpenBenchmarking or in the source code?

    What about the environment variables that are set before script is run? (although I guess I could just dump those with "env", but still)
    Last edited by Shnatsel; 07-09-2013, 09:56 AM.

    Comment


    • #3
      Originally posted by Shnatsel View Post
      I've been trying to make latency-oriented benchmarking profiles for PTS, but I couldn't find any documentation on how to write one and looking the existing examples didn't get me very far.
      Is there any documentation on writing test profiles? Maybe a reference manual for the XML definitions?

      If not, where can I at least look up the commonly used values? Maybe on OpenBenchmarking or in the source code?

      What about the environment variables that are set before script is run? (although I guess I could just dump those with "env", but still)
      There's a bit of docs within the documentation/ dir. Within the pts-core/openbenchmarking.org/schemas are the XML/XSL schemas.

      The possible environment variables should be shown by: phoronix-test-suite diagnostics under "Environmental Variables (accessible via test scripts)".

      For the most part though I'd recommend looking at other similar test profiles for reference. I wish there was more/better documentation, but unfortunately with largely developing it all myself and still having a back log of other work to do, there isn't too much time for good docs. If you have any questions feel free to let me know!
      Michael Larabel
      http://www.michaellarabel.com/

      Comment


      • #4
        Thanks! I'll see if I can make profiles for http://maemo.gitorious.org/maemo-too...mem-throughput and http://maemo.gitorious.org/maemo-too...-test-pingpong

        http://lmbench.sourceforge.net/ also looks interesting, it has a pipe latency test among other things.

        Comment


        • #5
          @Michael

          Can you make tests of various media players on Linux displaying various data formats (one movie, various qualities, formats), using various back ends (xine, gstreamer), using various accelerations apis?

          (ofc. in context of power & resources usage)

          Comment


          • #6
            Oh, this reminds me - we don't seem to have any VA-API benchmarks yet.

            Comment


            • #7
              Here are some hints on benchmarking VA-API: http://lists.freedesktop.org/archive...er/012524.html
              I might write a profile based on some other mplayer profile if I get my hands on hardware with VA-API.

              Comment


              • #8
                Originally posted by Shnatsel View Post
                Thanks! I'll see if I can make profiles for http://maemo.gitorious.org/maemo-too...mem-throughput and http://maemo.gitorious.org/maemo-too...-test-pingpong

                http://lmbench.sourceforge.net/ also looks interesting, it has a pipe latency test among other things.
                Many thanks, let me know if you need any help. Would love to try them out and use them upstream.
                Michael Larabel
                http://www.michaellarabel.com/

                Comment


                • #9
                  Originally posted by Shnatsel View Post
                  Here are some hints on benchmarking VA-API: http://lists.freedesktop.org/archive...er/012524.html
                  I might write a profile based on some other mplayer profile if I get my hands on hardware with VA-API.
                  See video-cpu-usage test profile does some mplayer-based benchmarking with different video accel methods... Though right now I think the sample video file is still crap since not many good free high quality videos with good download speeds.
                  Michael Larabel
                  http://www.michaellarabel.com/

                  Comment


                  • #10
                    Tears of Steel is probably a good candidate... it has live footage and a lot of SFX. If we cut a two minute long excerpt from it and re-encode it in with different h264 compression levels, it should make for a good sample video. I have no experience with performing such manipulations, though.
                    Last edited by Shnatsel; 07-09-2013, 04:57 PM. Reason: added link

                    Comment


                    • #11
                      As for hosting, I believe SourceForge would be suitable. It has mirrors all over the world, and I'm sure CC-BY license is fine with them. If not, there's also archive.org

                      Comment


                      • #12
                        so it is just another quake based benchmark.

                        pts is full with that. Pretty one sided. Which aren't other engines used in tests? ut2004 is still around. For example. Vegastrike is VERY demanding, awesome to look at and scriptable. Crystal Space. Ogre. ...

                        Comment


                        • #13
                          Vegastrike is demanding? Can't tell by the look of it. Space simulators are very easy to render because you have very little objects in the frame - just a few ships, no terrain or weather to render. If it's slow, it means it's just very poorly optimized.

                          Ogre demos pack looks interesting, I'll check if it exposes any performance counters.

                          Comment


                          • #14
                            Originally posted by Shnatsel View Post
                            Vegastrike is demanding? Can't tell by the look of it. Space simulators are very easy to render because you have very little objects in the frame - just a few ships, no terrain or weather to render. If it's slow, it means it's just very poorly optimized.

                            Ogre demos pack looks interesting, I'll check if it exposes any performance counters.
                            I've checked Ogre before but as I remember right there was either not any good free demos or there wasn't a way to benchmark it in terms of no timedemo commands, etc.
                            Michael Larabel
                            http://www.michaellarabel.com/

                            Comment


                            • #15
                              Originally posted by Shnatsel View Post
                              Vegastrike is demanding? Can't tell by the look of it. Space simulators are very easy to render because you have very little objects in the frame - just a few ships, no terrain or weather to render. If it's slow, it means it's just very poorly optimized.

                              Ogre demos pack looks interesting, I'll check if it exposes any performance counters.
                              so you never played it.

                              It is not 'just a few ships'.

                              It is hundreds of ships. Reflecting. Planets with ring systems. Asteroid fields with 'real' objects.

                              The game is an incredible big load as soon as you crank up the graphics settings and get into an intensive fight.

                              Comment

                              Working...
                              X