Subj : Re: Memory visibility and MS Interlocked instructions
To   : comp.programming.threads
From : Alexander Terekhov
Date : Thu Sep 01 2005 08:23 pm


Sean Kelly wrote:
[...]
> > >                                    I'll admit that until this I was
> > > beginning to think that only msync.acq stores needed a LOCK.
> >
> > That's not true either.
> 
> Help me out here :)  It's been established that loads can be migrated
> above preceding stores if no membars are present, so LOCK must be used
> in some cases to prevent this, correct?  

No must. MFENCE is probably better. ;-) 

I've simply misread what you wrote thinking that you mean acquire 
load (in Joe's Universum with totally out-of-order loads on IA32, it 
does need to LOCK something, fences aside for a moment ;-) ), not 
atomic<T>'s store(T, msync::acq_t).

regards,
alexander.

.