Updated: March 2009
Guarantees that every previous memory reference, including both load and store memory references, is globally visible before any subsequent memory reference.
On the x64 platform, this routine generates an instruction that is a faster store fence than the sfence instruction. Use this intrinsic instead of _mm_sfence on the x64 platform.
This routine is only available as an intrinsic.