Search:

Type: Posts; User: david.garcia

Page 1 of 20 1 2 3 4

Search: Search took 0.02 seconds.

  1. Replies
    2
    Views
    1,454

    Re: clEnqueueNDRangeKernel offset parameter

    The question isn't well formed.

    Passing a global offset when you enqueue an NDRange kernel simply affects the values returned by get_global_id(), as explained in section 3.2 of the specification....
  2. Replies
    1
    Views
    1,075

    Re: Do I have coalsecing access?

    Coalesced memory accesses are simple to understand: if consecutive work-items access consecutive memory locations then that memory access is coalesced.

    This does not appear to be the case in the...
  3. Replies
    2
    Views
    2,381

    Re: OpenCL slower than openGL.

    Maybe the question you need to ask yourself is: how is it possible that running on the same hardware one version of my program is 8 times slower than the other? Could the problem be how the slower...
  4. Replies
    2
    Views
    1,439

    Re: clCreateProgramWithBinary on Mac OSX

    The contents of a precompiled binary are implementation-dependent. Maybe Apple chose them to be paths to temporary files.

    All that OpenCL requires is that whatever you obtain when you call...
  5. Replies
    2
    Views
    1,690

    Re: Quadro FX 2700M Compatibility issue

    That may well be the problem. Reinstall the Nvidia drivers.
  6. Replies
    5
    Views
    2,135

    Re: Bug in NV OpenCL compiler

    Protip: if you want to get the attention of developers, write a 20-line application that reproduces the problem and include the source directly in the post.
  7. Replies
    4
    Views
    2,413

    Re: The support of out-of-order mode ?

    Is there a particular reason why you think you would benefit from out-of-order execution?
  8. Replies
    7
    Views
    2,469

    Re: memory buffer question

    No, there's no performance benefit in doing this in two steps. You can call clCreateBuffer() with the CL_MEM_COPY_HOST_PTR flag and do both creation and copy in one step.
  9. Replies
    14
    Views
    3,805

    Re: How to avoid double allocation on CPU

    cl_mem device_input = clCreateBuffer(context, CL_MEM_READ_WRITE | CL_MEM_USE_HOST_PTR, size, original_input, NULL);
    cl_mem device_output = clCreateBuffer(context, CL_MEM_READ_WRITE |...
  10. Replies
    1
    Views
    1,037

    Re: Error when building OpenCl

    Declare variable "i" as "size_t" instead of "unsigned int".
  11. Replies
    2
    Views
    1,016

    Re: strange output

    The problem appears to be this:




    ret = clEnqueueWriteBuffer(command_queue, startpt_obj, CL_TRUE, 0,NVARS*MAX_VECTORS * sizeof(double), startpt, 0, NULL, NULL);

    [...]

    for(j = 0; j <...
  12. Replies
    4
    Views
    2,382

    Re: Array of image_Xt objects as parameter?

    Sorry I didn't reply to you earlier. As you may have already seen, image arrays are a new feature of OpenCL 1.2. I couldn't disclose this information at the time.
  13. Replies
    5
    Views
    1,497

    Re: Mapping Data: remap to change?

    That's not the right way to do it. The right way is: map, read/write to the mapped pointer, then unmap. Reusing an old pointer you got in a previous map call is not a good idea at all.
  14. Replies
    6
    Views
    4,032

    Re: kd-tree nearest neighbour

    If you don't have access to doi.org you can read the same paper from Arxiv: http://arxiv.org/pdf/1112.4539v1.pdf
  15. Replies
    3
    Views
    2,794

    Re: OpenCL pointer to pointer

    That is not correct. The variable declaration "local float *rows[5]" means "This is an array of 5 elements located in the default address space, which is __private (section 6.5). Each element of the...
  16. Re: OpenCL-openGL interoperablity in multi-threaded environm

    If it works with the same thread and it fails with different threads, that may be a driver issue. If you can provide your vendor with a very short application that reproduces the issue they are more...
  17. Re: OpenCL-openGL interoperablity in multi-threaded environm

    Did you follow the instructions in section 9.8.6.1 of the CL 1.1. spec? Here's an excerpt:
  18. Replies
    1
    Views
    840

    Re: [Xavier Faure] OpenCL to GL

    You don't need to read back the data from GPU to CPU. Use Vertex Buffer Objects (VBO) instead so that the data stays in the GPU.

    For that you will need to use the OpenGL/OpenCL interoperability...
  19. Re: clSetKernelArg in loop CL_OUT_OF_RESSOURCES

    Is it possible that you are forgetting to release some resources in that loop? For example, did you remember to call clReleaseEvent() for those events you waited for?
  20. Re: OpenCL-openGL interoperablity in multi-threaded environm

    That should work fine in OpenCL 1.1. onwards. There is only limited thread safety in OpenCL 1.0.



    That should not be necessary.
  21. Re: How to get the number of cores per computing unit

    Can you describe what you mean by "batch size"?

    As for "core", it is a marketing term with no clear definition from an engineering viewpoint. What is called a "core" on one device may not have any...
  22. Re: typedefs (cl_int etc) for mulitple platforms

    Type cl_int on the host is guaranteed to match type "int" on the device. This is true for other types as well with the notable exception of bool and size_t. See section 6.1.1. in the OpenCL 1.2....
  23. Re: Can Native Kernels Enqueue Non-Native Kernels?

    Notice that we are talking about native kernels.
  24. Replies
    1
    Views
    1,011

    Re: Heterogeneous migratable kernels

    Friendly advice: if you are going to write a master thesis on a subject, it is a good idea to make sure that your assumptions are correct before you start.



    You can simply list them as...
  25. Re: OpenCL slow compiling of multiple kernel code

    It is recommended to build your kernels just once when the application is installed on the user's computer and after that rely on prebuilt program binaries.

    See clGetProgramInfo(...,...
Results 1 to 25 of 500
Page 1 of 20 1 2 3 4