[cpp-threads] memory model
Alexander Terekhov
alexander.terekhov at gmail.com
Fri Apr 29 14:02:37 BST 2005
On 4/29/05, Doug Lea <dl at cs.oswego.edu> wrote:
[...]
> Right. This is why "volatile" is not the same as "atomic".
> You always get barriers, but loads/stores of only those
> volatile variables smaller than some platform-dependent size are also
> atomic.
Please visit
http://www.google.de/groups?selm=40F4F750.FE8BCD9B%40web.de
http://www.google.de/groups?selm=40F647F8.F59FE644%40web.de
Uhmm,
http://www.google.de/groups?selm=40f77780_1%40news.melbourne.pipenetworks.com
<quote>
So while I do not think that "volatile" should have atomicity properties,
it does appear that with Atomics (and a good VM that intrinsifies
everything) you don't actually need 'volatile' (not withstanding that
the memory semantics of Atomics are defined in terms of 'volatile' :) ).
</quote>
;-)
regards,
alexander.
More information about the cpp-threads
mailing list