Announcement

Collapse
No announcement yet.

Understanding The Chrome OS Graphics Pipeline With Intel's Driver

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

  • Understanding The Chrome OS Graphics Pipeline With Intel's Driver

    Phoronix: Understanding The Chrome OS Graphics Pipeline With Intel's Driver

    If you are curious how the Chrome OS graphics rendering pipeline works, Tiago Vignatti and Dongseong Hwang of Intel's Open-Source Technology Center has written an interesting article about how they achieve zero-copy texture uploads in Chrome OS and other details of their low-level graphics implementation...

    http://www.phoronix.com/scan.php?pag...S-Gfx-Pipeline

  • #2
    To be honest, is adding GL_INTEL_map_texture to mesa more complex? So all applications could use this way to stream textures.

    Comment


    • #3
      Google, no, the future of the desktop is not in our browser...

      Comment


      • #4
        Originally posted by Scellow View Post
        Google, no, the future of the desktop is not in our browser...
        What are you saying? All these people who work hard to bring you ChromeOS, they do their work solely in the browser.

        Comment


        • #5
          Originally posted by Scellow View Post
          Google, no, the future of the desktop is not in our browser...

          It is though.
          If only to bring more people into the fold.

          There are far more talented HTML5 devs than there are decent C/C++ programmers.

          In fact, with the help of some toolkits/frameworks, web development has become so easy people can learn it in mere weeks.

          Comment


          • #6
            Users with Chrome can play around with the latest Intel open-source driver advancements by launching Chrome with the --enable-native-gpu-memory-buffers --enable-zero-copy arguments.
            except it requires vgem support for dmabuf/prime, which is missing in upstream kernel until someone figures out a good API for dealing w/ cpu access to dmabuf's in all cases.. not hard, but just something that someone needs to get around to doing

            Comment


            • #7
              isn't this just the opengl4.4 persistently mapped buffers ?

              Comment


              • #8
                So whole idea is to draw(webkit/skia) directly to texture. so you don't need to do glTexImage2D for everything what webkit/skia draw.
                It is good idea but some gpu-s rearrange texture in glTexImage2D call to internal memmory alignment for example qualcomm adreno GPU split texture to 64x32 pixels "tiles" in zigzag format.
                It is maybe for better GPU to cache access

                Comment


                • #9
                  Originally posted by unixfan2001 View Post
                  There are far more talented HTML5 devs than there are decent C/C++ programmers.

                  In fact, with the help of some toolkits/frameworks, web development has become so easy people can learn it in mere weeks.
                  That's a good explanation of why everything's a trainwreck.

                  It is cheaper to produce, though. That is one selling point.

                  Comment


                  • #10
                    Originally posted by johnc View Post

                    That's a good explanation of why everything's a trainwreck.

                    It is cheaper to produce, though. That is one selling point.
                    If everything was a trainwreck, it would be due to the fundamentalists.
                    Doing everything in plain C or C++ (preferably without even using an actual toolkit) just isn't very smart or sane in the long run.

                    Heck. There is no logical reason not to use HTML5 . Hardware is efficient enough to handle it and there are many fairly mature libraries for it.
                    MDM even uses it for its alternative greeters.

                    Comment

                    Working...
                    X