[concurrency-interest] Stricter read ordering

Andrew Haley aph at redhat.com
Tue Apr 29 06:22:28 EDT 2014


On 04/24/2014 12:15 AM, David Holmes wrote:

> The discussion of CAS being like a "volatile load plus volatile
> store" did expose that that description is inadequate to fully
> convey the required ordering constraints.

Actually it didn't, unless I missed that message.  What are the
ordering constraints required by?  All I remember was "We don't
really know."  Only longer.  :-)

> The actual implementations in the Oracle JDK do not allow external
> accesses to leak into the atomic region.

OK.  If that is really what is required.

Andrew.



More information about the Concurrency-interest mailing list