Announcement

Collapse
No announcement yet.

Former AMD Developer: OpenGL Is Broken

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

  • #91
    Originally posted by Aleve Sicofante View Post
    All I can see from these devs is they have been used for too long to a particular API (D3D) and are just too lazy to learn something new (new for them, that is).
    .
    rofl, the lengths some delusional people will go to defend flaws in an API that they have probably never coded anything serious in themselves. In case you were unaware, both Rich Geldreich and Josh Barczak have worked on real world AAA codebases using OpenGL, I don't think they're in the process of "learning" the API.

    Also, GLuint is not "different", it's just plain horrible.

    Comment


    • #92
      Originally posted by ua=42 View Post
      [*]GLSL is inefficient. Compiles everytime you run the program...
      That's what shader caches are for. AFAIK nvidia has a in-RAM and on-disk cache, no idea about fglrx but Mesa has a in-RAM cache and work has been started for a on-disk cache (see for example http://patchwork.freedesktop.org/patch/18964/ ).
      In other words: Shaders being compiled multiple times is a fault of the program: It could request the compiled shader after the first time and re-use it ( http://www.opengl.org/wiki/GLAPI/glProgramBinary ). Anyway drivers are able to implement a work-around (cache) for programs doing it wrong. So this is not OpenGLs fault.

      Comment


      • #93
        Taxi thats not the point alone. The Problem is also that you have to do the whole compiling process at the first start. With a Bytecode you has only to compile for your hardware and save time.

        Comment


        • #94
          That is an annoying feature because the first time you play a game you have all kinds of jitters from the driver building the GL cache. It's not the best experience and the first time through should be the optimum experience.

          Comment


          • #95
            Originally posted by BlackStar View Post
            No, he is developing for OpenGL. Where the hell did you get he is developing for Linux?
            Oh, he's developing Unity3D in a vacuum, not for Linux, Windows and OS X users to actually use, but just for the fun. I'm such a fool for not understanding that...

            Comment


            • #96
              Originally posted by Ancurio View Post
              rofl, the lengths some delusional people will go to defend flaws in an API that they have probably never coded anything serious in themselves. In case you were unaware, both Rich Geldreich and Josh Barczak have worked on real world AAA codebases using OpenGL, I don't think they're in the process of "learning" the API.

              Also, GLuint is not "different", it's just plain horrible.
              I'm not defending anything. I used GL well before it was called OpenGL but that's not the point. I of course don't know these two guys and I've only said what this looks to me from the outside. The fact that they had a number of replies showing they don't quite understand how the OpenGL works, shows me it doesn't matter how many "AAA codebases" they have worked on.

              Comment


              • #97
                Especially with games in wine there this problem and it sounds like it is the shader cache.

                For example if you play mass effect 2 with d3dstream patches it works with really good performance but after loading a level the first look around suffers from heavy stuttering and if it is a big room you basically can't play for 2-3 seconds. I have read that this is because the engine doesn't preload the graphics stuff and on direct3d/windows this isn't much of a problem, but translated to opengl it apparently is...

                Comment


                • #98
                  Originally posted by TAXI View Post
                  That's what shader caches are for. AFAIK nvidia has a in-RAM and on-disk cache, no idea about fglrx but Mesa has a in-RAM cache and work has been started for a on-disk cache (see for example http://patchwork.freedesktop.org/patch/18964/ ).
                  In other words: Shaders being compiled multiple times is a fault of the program: It could request the compiled shader after the first time and re-use it ( http://www.opengl.org/wiki/GLAPI/glProgramBinary ). Anyway drivers are able to implement a work-around (cache) for programs doing it wrong. So this is not OpenGLs fault.
                  Wtf are you smoking? I want some of that too!

                  A work-around is a work-around, not a solution. glProgramBinary? really!? have you even read about it? The binaries are hardware dependent! Good luck shipping your binary shaders with your games/apps.
                  Are you also going to blame the stupid developers for asking for separate shaders? Perhaps khronos was drunk to obey, since there are work-arounds...

                  A driver's job is to drive the hardware and expose it's functionality, not implement API specific work-arounds for it's mindless design! OpenGL is currently still a fail in many ways, even if it has come a long long way...

                  Comment


                  • #99
                    Why not simply compile all the shaders on first run, every time the driver version changes? I mean, compiling shaders doesn't actually take that long, does? Especially if as we have read the shaders get distributed already "compiled" to a simple subset of GLSL.

                    Comment


                    • Originally posted by ChrisXY View Post
                      Why not simply compile all the shaders on first run, every time the driver version changes? I mean, compiling shaders doesn't actually take that long, does? Especially if as we have read the shaders get distributed already "compiled" to a simple subset of GLSL.
                      Still, loading times of up to 1 minute, even if only for the first run, can be incredibly frustrating for users (especially if this never happens on Windows): https://github.com/ValveSoftware/Dota-2/issues/661

                      Comment

                      Working...
                      X