Search:

Type: Posts; User: monarodan

Page 1 of 2 1 2

Search: Search took 0.00 seconds.

  1. Replies
    2
    Views
    1,341

    Re: Callback When Event Already Set?

    I had this exact same query, so I read the spec and was horribly confused.

    Yes, the spec says, "All callbacks registered for an event object must be called.", but it also says "The registered...
  2. Re: Device affinity for command queues and buffers seems at

    I guess it depends on what you think is the correct behavior. If I am trying to minimise latency or have deterministic behavior then silently waiting is behavior that I would need to avoid - but...
  3. Re: Device affinity for command queues and buffers seems at odds

    Let's change the scenario a little - let's say I have N devices and want to make a buffer available to M of those devices, where 1 < M < N. How, in the OpenCL API, can I ensure that all M (and only...
  4. Re: Device affinity for command queues and buffers seems at odds

    Hello Affie,

    Your explanation certainly covers many of the motivations behind the design, however, I still think it is very confusing.

    Given your explanation, I find it very strange that there...
  5. Re: Device affinity for command queues and buffers seems at odds

    Remember that most commands need to be executed in a particular device, hence the convenience of having queues tied to particular devices.

    In-order queues are great while you are prototyping your...
  6. Re: Device affinity for command queues and buffers seems at odds

    Unfortunately that's not in my hands. You need to be a member of Khronos in order to contribute to the discussions for future specs. See http://www.khronos.org/members/. Look in particular at the...
  7. Re: Device affinity for command queues and buffers seems at odds

    Sorry, but you lost me here. What do you mean by that? The dependencies already form a DAG and if your device supports out-of-order execution it may take advantage of the structure of the DAG to...
  8. Re: non-blocking call to clEnqueueNDRangeKernel?

    I've taken the same approach of having a host thread per device, and this is working well, though it does introduce some overheads.

    My concern is with thread safety. I am using a single context...
  9. Re: Device affinity for command queues and buffers seems at odds

    I should have said "what device runs the commands in the queue" (my bad).
    [/quote]

    I wasn't trying to nitpick - your language actually aligns quite well with what the spec says. In this regard,...
  10. Re: Device affinity for command queues and buffers seems at odds

    Yes, by device agnostic I mean a queue that is not bound to one (and only one) device. I could have gone the other way and called it promiscuous as you could say that it services many devices.

    I...
  11. Re: Device affinity for command queues and buffers seems at odds

    True, but you have the same problem trying to share a buffer between multiple NDRange kernel executions inside the same device. I.e., you can't. Different kernels must work on the same buffer with...
  12. Re: Device affinity for command queues and buffers seems at odds

    To be honest I don't remember this particular topic being discussed in the meetings; I don't think anybody saw this asymmetry as a big problem. Leaving aesthetics aside, do you see this being a...
  13. Re: Device affinity for command queues and buffers seems at odds

    That sentence from the spec is stating something that should be expected anyway: modifying a resource in one queue while another queue is making use of it is going to cause trouble.
    [/quote]
    ...
  14. Re: Device affinity for command queues and buffers seems at odds

    That is defined in Appendix A. The quote I provided is only an excerpt.
    [/quote]

    It is not defined nor discussed in any great detail in Appendix A. Perhaps there is little detail required as...
  15. Re: Device affinity for command queues and buffers seems at odds

    [/quote]

    Fine, but this seems irrelevant. Consider the concept of thread local storage - the allocations still happen in the context of the processes heap, but the memory has thread affinity.

    ...
  16. Replies
    99
    Views
    64,947

    Re: OpenCL C++ Bindings

    I agree with this principle. Though when the rubber hits the road this can be difficult to adhere to %100 of the time. For example, the STL does a fantastic job of following this principle. Though...
  17. Replies
    99
    Views
    64,947

    Re: OpenCL C++ Bindings

    I second that!



    I guess there are a couple of design principals that I follow that might help us converge on an API.

    1) Identify attributes that are immutable for the lifetime of the class,...
  18. Re: non-blocking call to clEnqueueNDRangeKernel?

    Hi Markus,

    As per the suggestion above, did you get any feedback from NVidia on this issue? I think that the blocking nature of clEnqueueNDRangeKernel() is ridiculous as it stops a single CPU...
  19. Replies
    99
    Views
    64,947

    Re: OpenCL C++ Bindings

    I thought I'd fix this in my local header by simply passing "events" in place of NULL to enqueueNDRangeKernel(). However, there was an unexpected complication. When invoking a functor with one...
  20. Replies
    99
    Views
    64,947

    Re: OpenCL C++ Bindings

    Hi All,

    Something that all users of the C++ bindings should be aware of is the following. When using the template magic to bind kernel args, you will call a function declared like the following:...
  21. Replies
    99
    Views
    64,947

    Re: OpenCL C++ Bindings

    You right, another class which would be over cl::Buffer (why not cl::BufferPtr) could implement your "one data type" concept.
    But perhaps it's not the c++ OpenCL layer, but another one (a bit more...
  22. Replies
    99
    Views
    64,947

    Re: OpenCL C++ Bindings

    Just a quick correction - I've used the term cl_buffer where I should have used cl_mem in my previous post.

    Dan



    It should be noted that a cl::Buffer is just a facade to a cl_mem object and...
  23. Replies
    99
    Views
    64,947

    Re: OpenCL C++ Bindings

    It should be noted that a cl::Buffer is just a facade to a cl_mem object and you must be aware of this to use it (for example, the cl::Buffer copy constructor does not create a new cl_buffer -...
  24. Replies
    99
    Views
    64,947

    Re: OpenCL C++ Bindings

    Hi Again,

    When I got around to looking at cl::CommandQueue::enqueueReadBuffer() closely, I was surprised to find that it took a pointer to vector of events, with an empty vector being indicated by...
  25. Replies
    99
    Views
    64,947

    Re: OpenCL C++ Bindings

    Hi All,

    Just wondering why the cl::Buffer is not a template class. This would allow better type safety and avoiding the need for multiplying by sizeof( xxx ) everywhere. For example:


    ...
Results 1 to 25 of 28
Page 1 of 2 1 2