Announcement

Collapse
No announcement yet.

Mesa Finishes Up OpenGL 3, Lots Of OpenGL 4 Ahead

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

  • #11
    Originally posted by kaprikawn View Post
    I have no idea what's involved in adding GL compliance. Why does it take so long? Is it just a low priority, are there just not people working on it, or is it really complicated?
    It's really complicated and the driver teams are one tenth the size of what you'd find on Windows (plus a lot of the people are doing this as a hobby, not as paid work.)

    The amazing thing is that by sharing code between the drivers, results are coming much faster than in closed-source drivers. Do you know how long it took Intel and Ati to get proper OpenGL support on Windows? Close to a decade. The open-source drivers got there in half the time using teams 1/10 the size.

    That's the power of Free Software right there.

    Comment


    • #12
      Originally posted by BlackStar View Post
      It's really complicated and the driver teams are one tenth the size of what you'd find on Windows (plus a lot of the people are doing this as a hobby, not as paid work.)

      The amazing thing is that by sharing code between the drivers, results are coming much faster than in closed-source drivers. Do you know how long it took Intel and Ati to get proper OpenGL support on Windows? Close to a decade. The open-source drivers got there in half the time using teams 1/10 the size.

      That's the power of Free Software right there.
      That's why I can't wait for game developer to start having contact with open drivers. The development will probably skyrocket as game devs uncover bugs which affects their games and provide patches to mesa/gallium.

      As a side note, my feeling is that most of mesa developers are paid by vmware/red hat/intel/amd, not hobbists... but i may be wrong there...

      Comment


      • #13
        Originally posted by caligula View Post
        Since Intel is useless POS when is comes to performance, let's hope the driver development for Nvidia and AMD works faster in the future. LLVM compiled, reverse engineered open drivers are the only option for gaming if you don't like commercial quality drivers.
        It's enough for a lot of games, actually (especially more recent IGPs).

        The radeon driver is not the fruit of reverse engineering. Also, who says "commercial" says "bad side effects".

        Comment


        • #14
          Originally posted by BlackStar View Post
          It's really complicated and the driver teams are one tenth the size of what you'd find on Windows (plus a lot of the people are doing this as a hobby, not as paid work.)

          The amazing thing is that by sharing code between the drivers, results are coming much faster than in closed-source drivers. Do you know how long it took Intel and Ati to get proper OpenGL support on Windows? Close to a decade. The open-source drivers got there in half the time using teams 1/10 the size.

          That's the power of Free Software right there.
          You probably mean open source. The display drivers are not GNU and sharing results is open source. Free software is kind of nice but most GNU projects are a bit outdated when it comes to cutting edge technology.

          Comment


          • #15
            Originally posted by caligula View Post
            You probably mean open source. The display drivers are not GNU and sharing results is open source. Free software is kind of nice but most GNU projects are a bit outdated when it comes to cutting edge technology.
            But the kernel most certainly is.

            As a side note, my feeling is that most of mesa developers are paid by vmware/red hat/intel/amd, not hobbists... but i may be wrong there...
            AFAICT, Nouveau is almost completely done by hobbyists. Intel is mostly paid and AMD is somewhere in between (started out as a hobbyist project, but several of the core developers were hired later on.)

            Comment


            • #16
              Originally posted by caligula View Post
              You probably mean open source. The display drivers are not GNU and sharing results is open source. Free software is kind of nice but most GNU projects are a bit outdated when it comes to cutting edge technology.
              I think you have the wrong idea of what Free Software is. The Mesa license is Free Software. GNU is Free Software, but Free Software is more than GNU.

              Comment


              • #17
                Originally posted by caligula View Post
                You probably mean open source.
                No, he probably means Free Software. Mesa, Gallium3d and the kernel are all under Free Software licenses.

                You could also refer to it as "open source" and be right, because they are also OSI-approved licenses.

                Comment


                • #18
                  Originally posted by Nobu View Post
                  Well, seeing as r200 hardware only supports OpenGL 1.3, it can't be that surprising...or were you being faceious?

                  Similarly, Sandy Bridge CPUs (which have HD 2000/3000 graphics), support only OpenGL 3.1 at most (had to search deep for that one). So, yeah, you're sort of stuck there... Edit: Oh, is the hardware actually capable of 3.3? Interesting...
                  Strange, RadeonFeature matrix claims that 1.4 is possible. See http://xorg.freedesktop.org/wiki/RadeonFeature/

                  Comment


                  • #19
                    Originally posted by stikonas View Post
                    Strange, RadeonFeature matrix claims that 1.4 is possible. See http://xorg.freedesktop.org/wiki/RadeonFeature/
                    That doesn't sound quite right. It's been a while since I've worked on r200, but IIRC r200 is missing the functionality of 1.4 for depth textures, that is ARB_depth_texture and ARB_shadow. It should support everything else.
                    I'm not sure if it could be emulated - note not even r300 can actually do ARB_shadow natively, you have to emulate that functionality in the shader - but r200 only allows for very short shader programs so that could get you into trouble. And IIRC you can't actually emulate it since unlike r300 it cannot sample the required depth buffer formats at all.
                    Though I guess software fallback would be possible. But those are useless as hell. (That, of course, didn't stop nvidia announcing GL 1.4 support for GeForce 3/4 cards despite that these chips lacked some of the new blend functionality. Thus forcing apps to use hacks to determine if the cards could really do it as of course the performance hit when using this functionality meant that you needed to avoid this functionality no matter what.)

                    Comment


                    • #20
                      What's the status of GLSL 1.40+ and Geometry shaders for the radeon driver?

                      Comment

                      Working...
                      X