Elapsed time of my app almost equal spent CPU time while it calls pretty big kernels.
So, question is:where is better to put Sleep() calls to free some CPU cycles for other apps?
For example: few input streams read data from memory. According to Brook manual this is async operation. Then kernel called. Again, according manual kernel will wait all input streams to finish they reads. Will it consume CPU in that awaiting? That is, is it worth to put Sleep() call between stream read and kernel call?
[ And what about stream.domain operation? Is it sync or async ? Is it worth to add Sleep() between sub-stream creation and write from sub-stream to host memory ? ]