* Note: This is code uses synchronous atomic operations because that is all that x86 provides.
* Every atomic operation is also an implicit full memory barrier. The upshot is that it simplifies
* the code a bit, but it won't be as fast as it could be on platforms that provide weaker
* Note: This is code uses synchronous atomic operations because that is all that x86 provides.
* Every atomic operation is also an implicit full memory barrier. The upshot is that it simplifies
* the code a bit, but it won't be as fast as it could be on platforms that provide weaker