Search:

Type: Posts; User: Amnon

Search: Search took 0.00 seconds.

  1. Re: Reliable distributed computing w/ OpenCL, anyone doing t

    Now two years later, is there anyone working on a checkpoint/restart facility for OpenCL?

    How about adding a checkpoint/restart feature to the OpenCL standard?

    The idea is of a function call...
  2. Re: Device affinity for command queues and buffers seems at

    Solution 1: Allow NULL as the queue parameter in clEnqueueReadBuffer(), clEnqueueWriteBuffer(), etc.
    Solution 2: Have functions clReadBuffer(), clWriteBuffer(), etc.
    Solution 3: Use a dummy queue,...
  3. Replies
    7
    Views
    4,592

    Re: Needed: information about kernel parameters

    Our cluster-wide OpenCL platform handles 3rd-party OpenCL applications, so it has no advance knowledge about kernel arguments and their type (and we do not have our own parser either to try to...
  4. Replies
    3
    Views
    1,202

    Re: Nvidia bug in clCreateKernelsInProgram

    Oops, sorry for the typo - the correct error-code that I get is "CL_INVALID_KERNEL_DEFINITION" (-47)


    But I can get kernel objects using clCreateKernel, just not using clCreateKernelsInProgram.
    ...
  5. Replies
    7
    Views
    4,592

    Needed: information about kernel parameters

    What we need is to be able to look at a given kernel (cl_kernel) and tell not only how many parameters it has using
    clGetKernelInfo(param_name=CL_KERNEL_NUM_ARGS)

    but for each parameter to...
  6. Replies
    3
    Views
    1,202

    Nvidia bug in clCreateKernelsInProgram

    When a program was created for two Nvidia devices, but built for only one of them, Nvidia's implementation of clCreateKernelsInProgram() returns the error CL_INVALID_PROGRAM_DEFINITION.

    Needless...
  7. Two suggestions for supporting a multi-process environment

    1. Devices of each platform should be numbered.

    When two different processes call "clGetDeviceIDs()", the OpenCL standard does not (currently) guarantee that the physical devices will be listed in...
  8. Replies
    5
    Views
    2,414

    Re: Looking for multi-GPU applications

    Dear David,

    I am very aware of it, which is why we wrote an OpenCL library-platform that provides access to OpenCL devices (CPUs/GPUs) on as many computers across the network as you want. It...
  9. Replies
    5
    Views
    2,414

    Re: Looking for multi-GPU applications

    Thank you very much, we need this in order to evaluate the performance and network-overheads of running real-life applications on GPUs of remote computers, eg. on a cluster of computers with GPU(s),...
  10. Replies
    5
    Views
    2,414

    Looking for multi-GPU applications

    We are looking for OpenCL applications that were written with the intention of running them on several GPU devices in parallel and can run on Linux.

    Any suggestions/pointers please?

    Thank you...
  11. Replies
    4
    Views
    2,436

    Re: One device, many contexts

    Very interesting.

    I just wondered, since GPUs with MMUs are starting to appear, one important feature that goes with it is the ability to preempt a kernel: what if two users use the same GPU and...
  12. Replies
    4
    Views
    2,436

    Re: One device, many contexts

    Thank you, this was very informative.

    I can now see why using the same device with multiple contexts is important.

    I think that using the same device with multiple processes "simultaneously" is...
  13. Replies
    4
    Views
    2,436

    One device, many contexts

    Is it the intent of the OpenCL specification to allow a device to be part of more than one context, or should a device that is already part of an existing context be marked as "unavailable" and cause...
  14. Replies
    0
    Views
    3,227

    Asynchronous Events

    It will be very helpful to be able to be notified asynchronously when specific OpenCL events have completed.

    Note that clWaitForEvents() has several deficiencies:

    1) One cannot wait for any of...
Results 1 to 14 of 15