Subj : Re: sync vs. eieio on powerpc To : comp.programming.threads From : Joe Seigh Date : Tue Mar 01 2005 07:45 am On Tue, 01 Mar 2005 13:07:33 +0100, Alexander Terekhov wrote: > > Joe Seigh wrote: >> >> On Tue, 01 Mar 2005 12:04:10 +0100, Alexander Terekhov wrote: >> > >> And some pretty specific overall logic >> at that. > > I.e reservation and branching. No, in the example of a lock, the specific lock release logic matters. > > >> In general, it's incorrect to characterize it as an acquire >> with anything but a prior load. > > > Nope. You can have ordinary loads in between LR and SC+isync. They > are unordered (data and control dependencies aside for a moment). And > think of moving isync above SC (which is a store, not a load). > That's what I said. It's not a store/* barrier. -- Joe Seigh .