[cpp-threads] Alternatives to SC

[Doug Lea]

Thanks for the typo fix in example derivation.
I think everything you wrote is correct, including...


Paul E. McKenney wrote:
> 
> 
> Another way of putting this (I think) is that a read can establish
> a dependency, but that dependency matters only if that read either:
> 
> 1.	was executed by the focus task (T3 in this case), or
> 
> 2.	force a dependency involving a write event that was
> 	executed on the same processor as was the read.
> 
> Is this correct?
> 

Yes, I think these are the only cases involving reads
that can affect cycle test.

(While the propagation rule is pretty intuitive
(to me anyway :) it took a few tries to get it right.
Thanks mainly to Victor Luchangco for helping debug it.)

-Doug