The message below is from "Y. S. Ramakrishna" <Y.S.Ramakrishna@Sun.COM>:
-----
> This isn't a complete semantics, because is says nothing at all about
> programs that contain even a single data race. But this is one of the
> base elements that people can use to reason about whether a program
> is correctly synchronized.
>
> Bill
>
> Is there really more to add than "the values seen from a data race are
> unspecified"?
>
> Bowen
Might that be too draconian in that one might like to be able to reason
about programs that have "benign races," for instance of the sort
where (supported by some notion of write atomicity) each conflicting
write installs the same value, or where one of the conflicting
writes succeeds, and the race is used to break ties. But I suppose
that's one thing that the memory model would tell us -- how possibly
conflicting writes might be "resolved." A model that ducks that
question by mapping all such conflicts to bottom may thus be too
restrictive.
-- ramki.
-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:35 EDT