Announcement

Collapse
No announcement yet.

Beignet Is Working On OpenCL 2.0 Open-Source Linux Support

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

  • Beignet Is Working On OpenCL 2.0 Open-Source Linux Support

    Phoronix: Beignet Is Working On OpenCL 2.0 Open-Source Linux Support

    While writing this morning about Intel Cherryview support being added to Beignet, I also noticed Intel developers have been quietly fleshing out OpenCL 2.0 support for Linux...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    Nevertheless, it's great that Intel is continuing to invest in open-source OpenCL support with multiple developers working on Beignet while the overall open-source/Linux OpenCL scene remains in relatively sad shape
    And that is only because Intel has been intentionally fragmenting mesa. They've had plenty of time to move on to gallium, but they have chosen not to. It's not due to technical reasons, just corporate sabotage.

    Of course it is a good thing that Intel is releasing code, even though it's fragmenting existing infrastructure.

    Comment


    • #3
      Ofcourse, Fragmenting mesa code is not good idea, however CLover still is incomplete OpenCL 1.1 implementation and it would be rewritten.

      Comment


      • #4
        Hey guys,

        The impact of disabling batch buffer security for Intel Haswell
        platform on Linux for OpenCL kernels that need local shared memory
        access is massive.

        On a native kernel with and without the patch, the performance boost
        is massive, especially ijn the Lu decomposition benchmark.

        Do the math for yourself here. Note that the latest result is labelled
        "Asus GTX 860M" and its' at the bottom of the list.

        The test labelled i7 4700HQ should be identical to the Asus GTX 860M,
        it was simply mis-labelled in another previous test.

        The patchwork is here:


        The results of the benchmark is right here:
        OpenBenchmarking.org, Phoronix Test Suite, Linux benchmarking, automated benchmarking, benchmarking results, benchmarking repository, open source benchmarking, benchmarking test profiles


        The performance was noted for both the Intel IGP + the discrete NVIDIA GPU as both OpenCL ICDs (Intel's Beignet + NVIDIA CUDA's OCL) are installed on the system. Currently running Arch Linux.

        Regards,

        Brainiarc7

        Comment

        Working...
        X