0

If I want to benefit from Kepler GK110's Hyper-Q mechanism, i.e., to make two streams be put into two different hardware work queues to avoid some false dependencies, is it necessary for me to create the two streams with two CPU threads or the process forementioned will be undertaken automatically by CUDA driver or something else like CUDA work distributor?

Or by what means I can testify the consideration above?

einpoklum
  • 118,144
  • 57
  • 340
  • 684
troore
  • 777
  • 1
  • 6
  • 15
  • Have you taken a look at the `simpleHyperQ` SDK example and at the related [document](http://docs.nvidia.com/cuda/samples/6_Advanced/simpleHyperQ/doc/HyperQ.pdf)? – Vitality Jul 28 '13 at 10:21
  • That is exactly the answer and method I need. Thanks, Jack! – troore Jul 28 '13 at 11:52
  • @JackOLantern could you post this as an answer, I will upvote it. – Robert Crovella Jul 28 '13 at 14:17
  • @RobertCrovella Thanks Robert. I have added a very short answer, as I think that the example and documentation are very self-explanatory and it is always better and enough to refer to the original source of information. – Vitality Jul 28 '13 at 14:47

1 Answers1

3

As for other new CUDA features (e.g., dynamic parallelism, shuffle operations), Hyper-Q is equipped and very well explained and illustrated by the simpleHyperQ SDK example and the related documentation.

Vitality
  • 20,705
  • 4
  • 108
  • 146