> -----Original Message-----
> From: Sarita Adve [mailto:sadve@cs.uiuc.edu]
> Sent: Wednesday, July 18, 2001 1:43 PM
> (3) A write can become visible to some other processors early.
>
I should add that in this context also, update protocols on NUMA systems
make things difficult - a common solution suggested is to use two phase
protocols. Of course, I don't know of any NUMA update protocols today, but
it is another argument in favor of not *mandating* write atomicity.
Sarita
-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:33 EDT