Search:

Type: Posts; User: kylelutz

Search: Search took 0.00 seconds.

  1. Not sure if this is an option for you, but you...

    Not sure if this is an option for you, but you could use the program::link() method from the Boost.Compute library (which provides its own low-level C++ wrapper for OpenCL) which does use the proper...
  2. Replies
    3
    Views
    637

    Congratulations AJ!

    Congratulations AJ!
  3. Replies
    3
    Views
    627

    No, cl::Buffer doesn't directly support...

    No, cl::Buffer doesn't directly support re-sizing. You would have to allocate a new buffer with the new size, copy over the data, and then release the old buffer.

    If you're looking for a...
  4. Replies
    0
    Views
    467

    Boost.Compute v0.1 Released

    I'm proud to announce the initial release (version 0.1) of Boost.Compute!

    Boost.Compute is a C++ library for GPU and parallel-computing based on OpenCL. It provides an STL-like API and implements...
  5. Replies
    2
    Views
    533

    Your for-loop is wrong. Try something like this:...

    Your for-loop is wrong. Try something like this:



    int id = get_global_id(0); // will be 0, 1, 2, or 3
    int start = id * 250; // index of the first element for this block
    for(int i = 0; i <...
  6. Replies
    3
    Views
    724

    I usually just use the cl_ulong type in kernel...

    I usually just use the cl_ulong type in kernel code to represent a size_t.

    Make sure you cast the value to cl_ulong before passing it to the kernel in the off-chance that size_t is greater than...
  7. Replies
    1
    Views
    759

    That does look like a bug. I'm not sure who is...

    That does look like a bug. I'm not sure who is responsible for fixing it. Possibly try to get in contact with the following people: Benedict R. Gaster, Laurent Morichetti and Lee Howes (listed as...
  8. Replies
    1
    Views
    1,197

    You should checkout Boost.Compute [1]. It offers...

    You should checkout Boost.Compute [1]. It offers a C++, STL-like API just like Thrust but is based on OpenCL.

    [1] https://github.com/kylelutz/compute
  9. You can't. The type is implementation defined and...

    You can't. The type is implementation defined and not guaranteed to be accessible from user code.

    But lets take a step back for a second. Why do you need the sizeof _cl_platform_id? What are you...
  10. Replies
    4
    Views
    4,015

    Re: Templates support

    Sorry to dig up an old thread but I just stumbled upon it via google.

    I've been working on a C++ STL-like library for OpenCL called Boost.Compute. It provides some standard containers (vector<T>,...
  11. Replies
    1
    Views
    1,436

    Re: doubt on OpenCL's C++ feature

    Can you give some references? The OpenCL specification makes no mention of C++ kernel support. AMD does have some support for C++ in kernel code, but this is only for their OpenCL implementation and...
  12. Replies
    4
    Views
    1,910

    Re: constant memory issue

    Can you post your OpenCL kernel code, and also show us how you're setting up your buffers?
  13. Replies
    3
    Views
    1,479

    Re: Profiling of kernel code

    I have yet to find a good OpenCL profiler to measure performance within a kernel. My usual approach is to break the kernel into smaller pieces and profile each of those. Not ideal, but it may work...
  14. Replies
    4
    Views
    1,660

    Re: get_global_id is undefined

    OpenCL source files usually end in ".cl" (.cu is used by CUDA). But no, .cl files contain the source code.


    To get a kernel first create a cl_program object with clCreateProgramWithSource(),...
  15. Re: clCreateProgramWithSource - could be a file?

    This is because not everyone loads kernel sources from files.

    If clCreateProgramWithSource() only took a file name it would prevent applications from programmatically creating and using kernel...
Results 1 to 15 of 15