(j3.2006) (SC22WG5.3744) Question about ATOMIC_DEFINE

[Aleksandar Donev]

Van Snyder wrote:
> Assuming the ATOM argument is a coarray, coindexed object, or
> subcomponent of a coindexed object, Is there ever a time when it should
> not be?
Sure, if you are defining multiple locks, you may want to call SYNC 
MEMORY some time later, when you are done with all the work that does 
not depend on segment boundaries.
I don't see a point in mixing things up here by including the effects of 
image control statements with atomic operations. I also don't see any 
reason to treat co-indexed objects differently.
Best,
Aleks