[cpp-threads] Causality on more than two processors, write atomicity
Alexander Terekhov
alexander.terekhov at gmail.com
Wed Sep 7 09:33:14 BST 2005
On 9/7/05, Doug Lea <dl at cs.oswego.edu> wrote:
[...]
> with so much other chaos that the lack of transitivity
I think the question is whether msync::stuff shall be extended with
an extra mode for load() which will turn load() into dummy RMW (providing
illusion of remote write atomicity) and/or atomic<T> itself shall be
turned into atomic<T, bool remote_write_atomicity = false>. atomic<T,
true> in Itanic's WB memory could use st.rel for store() instead of
dummy RMW inside load().
I think both could be useful.
regards,
alexander.
More information about the cpp-threads
mailing list