[cpp-threads] Slightly revised memory model proposal (D2300)

[Paul E. McKenney]

On Wed, Jun 13, 2007 at 09:12:00AM +0100, Nick Maclaren wrote:
> "Paul E. McKenney" <paulmck at linux.vnet.ibm.com> wrote:
> >
> > > Fine.  But let's see a proper, semi-mathematical specification.
> > > I can see that being viable, and I can see it being ambiguous.
> > 
> > Looks to me like Hans's latest revision to N2171 (AKA D2300) covers
> > this.  ...
> 
> Oops.  Sorry.  I have been concentrating on other tasks.  I will
> read that with more care.
> 
> I think that we are agreed that there are dozens of viable solutions,
> and arguments in favour of most of them.  I see the biggest danger
> being a "committee compromise" on the basic model - it don't matter
> for most of the API, but the model must be consistent.

At this point, I believe that D2300 is pretty close -- if we are willing
to allow Sarita's example to fail.  Given that, the main issue that I
see remaining with D2300 is defining what happens with atomic operations.

						Thanx, Paul