Announcement

Collapse
No announcement yet.

Gallium3D, GEM Ported To Genode OS

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

  • Gallium3D, GEM Ported To Genode OS

    Phoronix: Gallium3D, GEM Ported To Genode OS

    Genode OS, a framework created by German developers for building special-purpose operating systems and provides a choice of six different kernels with targeted uses for high-security computing, dependable computing, automotive applications, and mobile devices, has advanced a great deal with its just-released 10.08 version. Genode OS 10.08 brings support for Qt 4.6, a new ATAPI block driver, IS0 9660 file-system support, MadWiFi wireless networking support, and dynamic linking on ARM EABI platforms, but most interesting in this release are the hardware-accelerated graphics improvements. Genode developers have ported Gallium3D and the Graphics Execution Manager to the Genode OS...

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

  • #2
    LLVM probably doesn't support all the hardware they want Genode to run on, although it would be nice to turn it on for those that do.

    Comment


    • #3
      Originally posted by smitty3268 View Post
      LLVM probably doesn't support all the hardware they want Genode to run on, although it would be nice to turn it on for those that do.
      Isn't LLVM being used for JIT compiling the programmable pipeline shaders? And isn't LLVM just checking on the fly what resources are available on the available CPU and then just compiling for that?

      Or am I consufed with the way JIT compiling is going on in OpenCL if no devices are specified on the code?

      Comment


      • #4
        Typo

        Originally posted by V!NCENT View Post
        Isn't LLVM being used for JIT compiling the programmable pipeline shaders? And isn't LLVM just checking on the fly what resources are available on the available CPU and then just compiling for that?

        Or am I confused with the way JIT compiling is going on in OpenCL if no devices are specified on the code?
        Stupid one minute time edit.

        Comment


        • #5
          Originally posted by V!NCENT View Post
          And isn't LLVM just checking on the fly what resources are available on the available CPU and then just compiling for that?
          Right, but LLVM has to have a backend for each architecture is supports. This is fairly easy to write (http://llvm.org/releases/2.2/docs/Wr...VMBackend.html), but coverage isn't that great yet and since Genode is specifically meant to be used in all sorts of embedded situations where the cpu being used may be completely custom, they may not want to make LLVM a hard dependency yet.

          Comment


          • #6
            Originally posted by smitty3268 View Post
            Right, but LLVM has to have a backend for each architecture is supports. This is fairly easy to write (http://llvm.org/releases/2.2/docs/Wr...VMBackend.html), but coverage isn't that great yet and since Genode is specifically meant to be used in all sorts of embedded situations where the cpu being used may be completely custom, they may not want to make LLVM a hard dependency yet.
            Well, if you're working on an OS, why not code these backends too? Especialy in the embedded sector you want to get as much performance out of the available compute units as possible. Maybe even increase speed to reduce the costs, (a little slower CPU in there?) save power, or both.

            Comment

            Working...
            X