Search:

Type: Posts; User: Dithermaster

Page 1 of 5 1 2 3 4

Search: Search took 0.00 seconds.

  1. Replies
    3
    Views
    272

    The pinned memory read/write thing is unique to...

    The pinned memory read/write thing is unique to NVIDIA. Check on the Intel and AMD sites for their best practices / programming guidelines. A good order of operations is to understand OpenCL based on...
  2. Replies
    3
    Views
    272

    You can use any host memory with...

    You can use any host memory with clEnqueueRead/WriteBuffer. On NVIDIA hardware, the operations will go faster if the source or destination memory was allocated as pinned memory (using clCreateBuffer...
  3. Replies
    7
    Views
    233

    Yes, clCreateBuffer will create a GPU buffer...

    Yes, clCreateBuffer will create a GPU buffer without allocating host memory (as far as you know; an implementation could if it wanted). I'd suggest starting with some of the OpenCL examples to get a...
  4. Sticky: A couple of typographical issues: p.14: "For a...

    A couple of typographical issues:

    p.14: "For a kernel to access local memory on a device, the user can either create a dynamically-sized local accessor object to the kernel as a parameter." --...
  5. Replies
    4
    Views
    225

    Yes, that's what I meant. Perhaps more complex...

    Yes, that's what I meant. Perhaps more complex that you'd like, but using #include files in OpenCL can be equally tricky.
  6. Replies
    4
    Views
    225

    The OpenCL C compiler has no access to your...

    The OpenCL C compiler has no access to your struct definition. While it does have some #include capabilities, I'd start by just repeating your struct definition inside the kernel source. Also, you...
  7. Sorry, this is under Windows. I don't know how to...

    Sorry, this is under Windows. I don't know how to do it under Linux. I assume it is possible though. OpenCL should use the Installable Client Driver (ICD) and each would show up as a platform.
  8. Replies
    2
    Views
    232

    A key to fast parallel code is to reduce the...

    A key to fast parallel code is to reduce the dependencies. If your "previous pixel" dependencies are always "to the left in the same row" then make your work item process a whole row left to right so...
  9. It could be other GPU operations are getting...

    It could be other GPU operations are getting "caught" by your clFinish and you're also timing those. Things like OpenGL drawing your screen. Try creating an OpenCL 'event' for each kernel and get the...
  10. Yes, I have both NVIDIA and AMD GPUs in my...

    Yes, I have both NVIDIA and AMD GPUs in my computer. You can hook monitors to either or both, and you can use both for GPU compute. They each show up at a "platform" in OpenCL, and a "device" within...
  11. Replies
    1
    Views
    326

    Seems reasonable. As long as your class that...

    Seems reasonable.

    As long as your class that holds the context holds it until everyone is done, then you don't need additional retain/release calls since the singleton class holds a reference...
  12. Replies
    6
    Views
    487

    No, getting CL_SUCCESS from clFinish only means...

    No, getting CL_SUCCESS from clFinish only means that command worked.

    If you install the (free) Intel OpenCL SDK then you have another OpenCL compiler available that may give better error messages.
  13. Your actual runtime is 3 seconds? Most platforms...

    Your actual runtime is 3 seconds? Most platforms will terminate an OpenCL kernel if it runs for a while. Can you design your system to instead run as a series of faster kernels?
  14. Replies
    6
    Views
    487

    good ideas kunze. It sounds like a mystery....

    good ideas kunze. It sounds like a mystery. @sajis997, can you try your code on a different platform (if you don't have a GPU from a different vendor, perhaps use the Intel CPU driver). It might give...
  15. Without seeing the code I can only guess, but it...

    Without seeing the code I can only guess, but it sounds like you are going over the amount of private or shared memory.
  16. Replies
    3
    Views
    456

    Start by using only one in-order command queue....

    Start by using only one in-order command queue. Enqueue the data upload, the compute, and the data download. Only the download needs to use the "blocking" flag set to TRUE. You don't need clFinish,...
  17. An OpenCL tookit or SDK is just header files;...

    An OpenCL tookit or SDK is just header files; your executable can use any OpenCL compliant driver once built.

    Therefore you could use the headers that were in CUDA 4.2 or you could AMD or Intel's...
  18. Buffers and Images (and other cl_mem objects on...

    Buffers and Images (and other cl_mem objects on newer versions of OpenCL) passed to kernels are just handles to the memory object. Therefore they are very fast and you can use the same Buffer in...
  19. Replies
    3
    Views
    622

    Not possible. In fact, on most hardware if the...

    Not possible. In fact, on most hardware if the kernel runs for more than a short time the OS will kill it. OpenCL is not designed for continuously running kernels. What you can do instead is...
  20. Most drivers don't implement out-of-order command...

    Most drivers don't implement out-of-order command queues.

    The typical way to do what you want is to have separate command queues for the data transfer and compute and use events to create...
  21. Replies
    5
    Views
    788

    Yes. You'd create a command queue for each GPU...

    Yes.

    You'd create a command queue for each GPU and feed it work. You could also use OpenCL to enqueue work for the CPUs.
  22. Replies
    6
    Views
    487

    Search the specification for all references to...

    Search the specification for all references to CL_INVALID_COMMAND_QUEUE; each API lists the conditions it might return this error code. You might find that one of those conditions exists in your code.
  23. Replies
    3
    Views
    625

    In OpenCL you iterate the Platforms, then within...

    In OpenCL you iterate the Platforms, then within each Platform, iterate the Devices that are available. Filter to those that you wish to support (e.g., perhaps GPU only) then select one or more that...
  24. Replies
    3
    Views
    625

    In order to interoperate between OpenGL and...

    In order to interoperate between OpenGL and OpenCL, the OpenCL context must be built from the OpenGL context. If you don't get an error when trying to do this across devices (GL on one, CL on the...
  25. Replies
    5
    Views
    750

    For a single frame, efficiency is moot, but...

    For a single frame, efficiency is moot, but becomes important if you have a number of frames to capture.

    This is no longer an OpenCL topic, so should migrate from this board.

    Capturing from the...
Results 1 to 25 of 125
Page 1 of 5 1 2 3 4