[cpp-threads] Re: Increment/decrement operators on atomics package

[Alexander Terekhov]

On 4/29/07, Paul E. McKenney <paulmck at linux.vnet.ibm.com> wrote:
> On Sat, Apr 28, 2007 at 01:56:14PM +0200, Alexander Terekhov wrote:
> > On 4/28/07, Raul Silvera <rauls at ca.ibm.com> wrote:
> > [...]
> > >At this point we're fairly confident that SC is achievable on current PPC
> > >implementations by having hwsyncs between each pair of accesses. It is
> > >widely believed that this was the intent of the architecture,
> >
> > This is news for me. Consider:
> >
> > http://citeseer.ist.psu.edu/adve95shared.html
> >
> > "partly supported by an IBM ...  supported by DARPA"
> >
> > ;-)
>
> This paper predates my time at IBM, but they tell me that the "intent
> of the architecture" did change at about that time.  I believe that
> the change added the notion of cumulativity described near the end of
> section 1.7.1 of Book 2 of the architecture.

How does cumulativity help in the IRIW case?

P1: x = 1;
P2: y = 1;
P3: r1 = x; r2 = y;
P4: r3 = y; r4 = x;

regards,
alexander.