Search:

Type: Posts; User: Dithermaster

Page 1 of 7 1 2 3 4

Search: Search took 0.00 seconds.

  1. It's the second. __local is shared local...

    It's the second.

    __local is shared local memory, so it's a single piece of memory that every work item in a work group can access. It's essentially a programmer-managed cache, and frequent used...
  2. No, swap steps 1 and 2 so it becomes: 1. run...

    No, swap steps 1 and 2 so it becomes:

    1. run the kernel
    2. map the clbuffer
    3. use the pointer retuned from maping to go through the buffer and print each item
    4: Unmap the buffer

    Mapping...
  3. You don't use the contents after...

    You don't use the contents after clEnqueueUnmapMemObject.

    Instead of clEnqueueReadBuffer, do a clEnqueueMapBuffer (with blocking), use the pointer returned to access the buffer, then...
  4. For many applications, yes. You can certainly try...

    For many applications, yes. You can certainly try to write a function that calculates an optimal work group size, but it will be a challenge. Alternatively, you can benchmark all sizes on the user's...
  5. I'd love to be proven wrong, but in my opinion...

    I'd love to be proven wrong, but in my opinion and based on my experience, it's a black art.

    It varies by hardware vendor, and I've even seen where non-multiples of...
  6. Replies
    5
    Views
    432

    As I said, look it up in cl.h: #define...

    As I said, look it up in cl.h:

    #define CL_INVALID_VALUE -30

    Then look in the OpenCL specification for the API that is returning that error to see what it means.

    On...
  7. Replies
    5
    Views
    432

    It's always helpful to look at the error code you...

    It's always helpful to look at the error code you get back from OpenCL APIs. For example, what code to you get back from clGetDeviceIDs? Look it up in cl.h to get a clue as to what is happening.
  8. Replies
    4
    Views
    395

    OpenCL 1.x doesn't have a continuous data...

    OpenCL 1.x doesn't have a continuous data streaming mode so you'll need to chop up your data into blocks and upload them one by one, process them, and download results. On modern hardware you'll be...
  9. Replies
    6
    Views
    674

    > If you specify a work group size larger than...

    > If you specify a work group size larger than your hardware or kernel supports, the clEnqueueNDRange call should fail and return an error code.
    That would be nice but I don't think you can reliably...
  10. Replies
    7
    Views
    608

    Just an observation: The GlobalWorkSize is not an...

    Just an observation: The GlobalWorkSize is not an integer multiple of the WorkGroupSize (216 is not evenly divisible by 16). In OpenCL 1.x, if you specify the work group size then the global size...
  11. Replies
    2
    Views
    344

    Use a "buffer" in OpenCL global device memory.

    Use a "buffer" in OpenCL global device memory.
  12. Not yet because AMD has not shipped an OpenCL 2.0...

    Not yet because AMD has not shipped an OpenCL 2.0 driver. When (or if) they do, it is up to AMD as to whether it will support the Radeon HD 6670 or only newer hardware.
  13. Replies
    1
    Views
    400

    Use an OpenCL Buffer object. It will retain value...

    Use an OpenCL Buffer object. It will retain value between kernel invocations and can be accessed with a pointer.
  14. Replies
    1
    Views
    592

    Not yet. The conformance tests were only recently...

    Not yet. The conformance tests were only recently completed.
  15. Replies
    7
    Views
    845

    Oh, that's Java? Sorry, didn't catch that. Well,...

    Oh, that's Java? Sorry, didn't catch that. Well, I can't help you much then except to still say that something is leaking host memory. Do you see Task Manager memory usage grow as your application...
  16. Replies
    7
    Views
    845

    You are leaking host memory in initGPU. If you...

    You are leaking host memory in initGPU. If you run Task Manager you'll see that your memory usage just keeps growing. Everything that has a "new" must have a matching "delete".
  17. Is this a quiz or a mind reading exercise? What...

    Is this a quiz or a mind reading exercise? What pre-release OpenCL platform are you using (since none have shipped yet), what results do you expect, and what errors or incorrect results are you...
  18. Replies
    3
    Views
    541

    AJ's suggestion is great. What I've done is...

    AJ's suggestion is great. What I've done is comment out various parts of the kernel, along the lines of "if this part was 'free' how fast would it run?". You can do this separately for reads, compute...
  19. Replies
    3
    Views
    582

    In OpenCL 1.x buffer and image data needs to be...

    In OpenCL 1.x buffer and image data needs to be contiguous. OpenCL 2.x SVM could support your pointer-based non-contiguous data structure but there are no shipping implementations yet. In the...
  20. Replies
    2
    Views
    442

    It varies, but some devices are capable of...

    It varies, but some devices are capable of running hundreds or thousands of work items in parallel. You should design kernels to be launched over thousands or millions of work items to best utilize...
  21. Replies
    4
    Views
    500

    Perhaps http://portablecl.org/ is what you're...

    Perhaps http://portablecl.org/ is what you're looking for? It's not for AMD GPUs though. For that, you need AMD's implementation of OpenCL, which is only available in binary (not source) form.
  22. Replies
    3
    Views
    582

    Yes, of course you can do that. Two ways: Buffer...

    Yes, of course you can do that. Two ways: Buffer and Image.
    Buffer: As on the host, dimension is just an abstraction since all memory is linear. Use Y * pitch + X to find a location.
    Image: OpenCL...
  23. Replies
    6
    Views
    750

    Yes, in OpenCL 1.x you must use...

    Yes, in OpenCL 1.x you must use clEnqueueMapBuffer before accessing on host side, and clEnqueueUnmapMemObject before using on device side.
  24. Replies
    6
    Views
    750

    Yes, this is possible. Please see:...

    Yes, this is possible. Please see:
    https://software.intel.com/en-us/node/505841
    and
    https://software.intel.com/sites/products/documentation/ioclsdk/2013/OG/Sharing_Resources_Efficiently.htm
  25. Only two things can stop a kernel: The kernel...

    Only two things can stop a kernel: The kernel completes, or the OS kills it because it takes more than a few seconds to run. That's it.

    The kind of "interactive" thing you're looking for is not...
Results 1 to 25 of 153
Page 1 of 7 1 2 3 4