volatile and barriers (especially on PPC)

[Doug Lea]

> If we allow volatile to have ordering 
> semantics weaker than sequential consistency, we might as well go all 
> the way and separate volatility from memory ordering (which I prefer).
> 

That's the second time today that someone has said this to me,
and both times, I have no idea what it means! :-)

Could you explain?

(The only thing I think you could mean here is to propose that ordinary
programmers writing ordinary things like double-check will need to take
responsibility for ordering by manually correctly inserting barrier
instructions? If so, you will need a very compelling defense for
handing a new stack of razor blades to C++ programmers.)

-Doug