Search:

Type: Posts; User: LeeHowes

Search: Search took 0.00 seconds.

  1. This is not valid in OpenCL 1.x because the...

    This is not valid in OpenCL 1.x because the cl_mem pointer and the pointer the device code expects to see are not required to be the same (for reasons of portability). The runtime needs additional...
  2. Thread: NULL or !NULL

    by LeeHowes
    Replies
    1
    Views
    790

    Usually in C code you can't make those sorts of...

    Usually in C code you can't make those sorts of assumption and that is the case here for the same reason. buffer will not be initialized and the buffer variable will not be reinitialized after a...
  3. That much I understood :) My point was about...

    That much I understood :) My point was about memory management INSIDE the method. I've just looked again though and of course the CL API doesn't take it as anything resembling pairs anyway so doing...
  4. Some of those suggestions should be very simple...

    Some of those suggestions should be very simple to implement. Some of them have always been frustrating to me, too, but have fairly sensible reasons somewhere. For example, the program source list I...
  5. Replies
    16
    Views
    6,897

    Re: OpenCL 1.2 cl.hpp

    The update to that document is going through the process at the moment. Expect an upload in a month or so.

    Sorry about the delay.
  6. Re: Is float point operation in OpenCL stochastic?

    As notzed said:


    For example, if you add a random sequence:
    2^-30, 2^30, 2^-30, 2^30 ...
    you'll end up getting the same as:
    0, 2^30, 0, 2^30 ...

    because the mantissa of a float isn't big...
  7. Re: Is float point operation in OpenCL stochastic?

    Remember also that the order of floating point operations matters both within a work item (notzed's compiler point) and between work items - so any code where the order of operations is dependent on...
  8. Replies
    16
    Views
    6,897

    Re: OpenCL 1.2 cl.hpp

    Yes, unfortunately when I took over maintaining cl.hpp doxygen content was very poor. We've tried to improve that, with some external input, but as you can see it's incomplete. We have other efforts...
  9. Replies
    16
    Views
    6,897

    Re: OpenCL 1.2 cl.hpp

    That's something we've been working on. You will hopefully see updates very soon.
  10. Replies
    16
    Views
    6,897

    Re: OpenCL 1.2 cl.hpp

    Sorry, I've been a bit slow getting to this. It's up:
    http://www.khronos.org/registry/cl/

    Unfortunately we do need to update the page as it still says that it hasn't been updated. This is not...
  11. Replies
    3
    Views
    1,974

    Re: Lifetime of cl::Event objects?

    cl::Event objects reference count the underlying cl_event - but when you constructed the vector you made *copies*. So updating readImgGenEvent during the enqueue, for example, I think will not update...
  12. Replies
    16
    Views
    6,897

    Re: OpenCL 1.2 cl.hpp

    We came up with a little snag. This year, absolutely. If nobody reports any serious issues with the current code I'd like to think in the next month. There is a preview version in the current AMD...
  13. Replies
    16
    Views
    6,897

    Re: OpenCL 1.2 cl.hpp

    Code is written and it's in a review cycle. We had to think through a good way to deal with feature deprecation but I think we've solved that now. I can't give a clear time line for release but it...
  14. Replies
    12
    Views
    3,466

    Re: Problems with multiple commandQueues

    If the queue execution is lazy then there might be nothing until the finish to trigger anything to be executed at all. By doing a read and then a blocking flush there is nothing to trigger the second...
Results 1 to 14 of 20