Private Memory Consistency

So far, I’ve understood that all reads/writes from private memory are guaranteed to be completed immediately before the next use. Is there any extension or feature for OpenCL that relaxes this restriction for double buffering optimization?