controlling how blocking is implemented.
when using cuda the programmer can control how the wait is implemented in blocking commands. the wait mechanism can be implemented in several way: busy wait , events , etc . (CU_CTX_SCHED_SPIN,CU_CTX_SCHED_YIELD,CU_CTX_BLOCK ING_SYNC ...)
is there an equivalent in OpenCL ?
Re: controlling how blocking is implemented.
There's no equivalent. It is assumed that it will be some sort of non-busy wait, but to the best of my knowledge the spec leaves this as an implementation detail.