On 17/05/16 19:23, Paolo Bonzini wrote: > > On 17/05/2016 18:15, Sergey Fedorov wrote: >> On 14/05/16 06:34, Emilio G. Cota wrote: >>> This new helper expands to __atomic_test_and_set with acquire semantics >>> where available; otherwise it expands to __sync_test_and_set, which >>> has acquire semantics. >> Why don't also add atomic_clear_release() for completeness? > The previous patch adds atomic_set_release.
Yes, but it would take the advantage of __sync_lock_release() being just a release barrier rather than a full barrier of smp_mb() in atomic_set_release(). But that's only the case for legacy __sync built-ins (before GCC 4.7.0), though. Kind regards, Sergey
