Announcement

Collapse
No announcement yet.

New Open-Source CUDA / OpenCL Benchmarks Added

Collapse
X
Collapse
 
  • Page of 1
  • Filter
  • Time
  • Show
Clear All
new posts
Previous template Next
  • #1

    New Open-Source CUDA / OpenCL Benchmarks Added

    Phoronix: New Open-Source CUDA / OpenCL Benchmarks Added

    There are a couple new OpenCL/CUDA tests now present via the Phoronix Test Suite and OpenBenchmarking.org...

    New Open-Source CUDA / OpenCL Benchmarks Added - Phoronix
    http://www.phoronix.com/scan.php?page=news_item&px=New-OpenCL-CUDA-GPGPU
    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
    Tags: None
  • #2
    Originally posted by phoronix View Post
    Phoronix: New Open-Source CUDA / OpenCL Benchmarks Added

    There are a couple new OpenCL/CUDA tests now present via the Phoronix Test Suite and OpenBenchmarking.org...

    http://www.phoronix.com/scan.php?pag...nCL-CUDA-GPGPU
    $ phoronix-test-suite benchmark shoc

    Code:
    ../../../../config/targets.mk:24: recipe for target 'BusSpeedReadback.o' failed
make[4]: *** [BusSpeedReadback.o] Error 1
nvcc fatal   : Unsupported gpu architecture 'compute_12'

    Comment

    • #3
      Originally posted by << ⚛ >> View Post

      $ phoronix-test-suite benchmark shoc

      Code:
      ../../../../config/targets.mk:24: recipe for target 'BusSpeedReadback.o' failed
make[4]: *** [BusSpeedReadback.o] Error 1
nvcc fatal : Unsupported gpu architecture 'compute_12'
      Are you running it from PTS? If you are, it shouldn't have that error. As you can see from the test profile (http://openbenchmarking.org/innhold/...9679af7c269404) it actually defaults to compute_37 unless you set the $CUDA_CPPFLAGS environment variable.
      Michael Larabel
      https://www.michaellarabel.com/

      Comment

      • #4
        Originally posted by Michael View Post
        Are you running it from PTS? If you are, it shouldn't have that error. As you can see from the test profile (http://openbenchmarking.org/innhold/...9679af7c269404) it actually defaults to compute_37 unless you set the $CUDA_CPPFLAGS environment variable.
        On my machine, Cuda is installed in /opt/cuda. I would suggest to replace
        Code:
        if [ -d /usr/local/cuda ] ...
        with
        Code:
        if $(nvcc --version >& /dev/null)...
        in test-profiles/pts/shoc-1.0.0/install.sh.

        Comment

        • #5
          Not really sure what is called "opensource". CUDA only works on GPUs from single vendor and requires proprietary driver. So much "openness", dammit. Snake oil of opensource looks like an "open" program which is useless without huge, blob-only runtime.

          Comment

          • #6
            Originally posted by << ⚛ >> View Post

            On my machine, Cuda is installed in /opt/cuda. I would suggest to replace
            Code:
            if [ -d /usr/local/cuda ] ...
            with
            Code:
            if $(nvcc --version >& /dev/null)...
            in test-profiles/pts/shoc-1.0.0/install.sh.
            pts/shoc-1.0.1 should improve things for you then.
            Michael Larabel
            https://www.michaellarabel.com/

            Comment

            • #7
              Originally posted by Michael View Post
              pts/shoc-1.0.1 should improve things for you then.
              It works. Thanks.

              Comment

              Previous template Next
              Working...
              X