On Thu, 30 Mar 2006, Chen, Kenneth W wrote: > > > I know, I'm saying since it doesn't make any difference from API point of > > > view whether it is acq, rel, or no ordering, then just make them rel as a > > > "preferred" Operation on ia64. > > > > That would make the behavior of clear_bit different from other bitops and > > references to volatile pointers. I'd like to have this as consistent as > > possible. > > Yeah, but we just agreed that caller shouldn't be thinking clear_bit has > memory ordering at all. In general yes the caller should not be thinking about clear_bit having any memory ordering at all. However for IA64 arch specific code the bit operations must have a certain ordering semantic and it would be best that these are also consistent. clear_bit is not a lock operation and may f.e. be used for locking something. - : send the line "unsubscribe linux-ia64" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
![]() |