Search:

Type: Posts; User: erwin

Search: Search took 0.00 seconds.

  1. Replies
    2
    Views
    2,374

    Re: build static OpenCL library

    It would be good to have a statically linked OpenCL CPU backend, that is always available.

    Perhaps an open source OpenCL implementation can do it. It can coexist with the ICD.
  2. Re: OpenCL device selection for interop with OpenGL.

    You can iterate over every OpenCL device until interop with a OpenGL context succeeds.
  3. Re: Error in CodeXL: Not able to produce resource usage inf

    Do those kernels use barrier? CodeXL doesn't support debugging of kernels with barriers yet.
  4. Thread: OPENCL GDEBUGGER

    by erwin
    Replies
    1
    Views
    1,749

    Re: OPENCL GDEBUGGER

    gDebugger is replaced by CodeXL.

    See http://developer.amd.com/tools/heteroge ... ng/codexl/
  5. Replies
    4
    Views
    2,596

    Re: OpenCL slow compiling on AMD card

    Others experienced similar issues:

    http://wiki.blender.org/index.php/Dev:2 ... les/OpenCL

    Hopefully the AMD OpenCL driver writers know about the issue and fix it at some stage.
  6. Re: How to debug on Apple? [CL_INVALID_COMMAND_QUEUE]

    I fixed my kernel by breaking it up into several smaller kernels, and by replacing some code.

    It looks like local arrays cannot be used to index into global arrays.



    __kernel(__global...
  7. How to debug on Apple? [CL_INVALID_COMMAND_QUEUE]

    I'm trying to compile my kernels on Mac OSX (Mountain Lion). They already work fine under Windows, both for NVIDIA and AMD GPUs.

    Most of the kernels could be fixed by making sure to only use basic...
  8. Replies
    14
    Views
    5,503

    Re: problem passing struct kernel arguments (Apple)

    Although your own structs cannot be passed as arguments to OpenCL kernels,
    you can pass in built-in types such as int4, uint4 by value as a workaround. You can typedef the x,y,z,w fields to...
  9. Replies
    14
    Views
    5,503

    Re: problem passing struct kernel arguments (Apple)

    It looks like passing structures by value to an OpenCL kernel doesn't work on Apple indeed.
  10. Re: cl_ext_atomic_counters_32 vs global memory atomics

    Ah, I didn't realize that each extension is also set as a preprocessor define indeed.

    Thanks for the help!
    Erwin
  11. cl_ext_atomic_counters_32 vs global memory atomics

    OpenCL 1.1 has global atomics that work great on NVIDIA GPUs but slower on AMD GPUs. On AMD GPUs the counters32_t perform better. Here are the two versions:


    //OpenCL 1.1 has atomic_inc build-in...
  12. Replies
    5
    Views
    7,029

    Re: How to compile OpenCL example in MS VC++?

    (by the way, this problem will be fixed in an upcoming OpenCL update on the Khronos website)
  13. Replies
    5
    Views
    7,029

    Re: How to compile OpenCL example in MS VC++?

    It is easy to fix the compilation problems.

    Please check the MiniCL project, a small subset of OpenCL used for the Bullet Physics SDK. It has working OpenCL header files.

    Hopefully this gets...
Results 1 to 13 of 13