One of the goals of my project is to transparently run multi-threaded Java
program on a cluster of computers, where threads can be distributed or
migrated to different nodes of cluster to achieve speedup and solve larger
size problem. So Java memory model has large impact on the performance of
such a system.
Java threads' memory sharing behavior inspires to extend it to distributed
shared memory environment. From this viewpoint, I wish future Java memory
model can adopt release consistency model, since sequential consistency and
other not relaxed consistency model have already been proven to result poor
performance in a distributed environment. Will the lazy release consistency
model adopted in the proposed JMM be changed afterwards?
Regards,
Weijian
> The current stage is what is being discussed on this list. Jeremy and
> I announced a draft formal proposal last November, although it wasn't
> really in the format that would be needed for the JCP process. It has
> been difficult to get much feedback on it, and it is only now that
> some of the experts on the topic of memory models have been getting
> back to us with comments (or promises of comments).
>
> To force things to move along, the expert group will be voting in a
> few weeks on what formal model to adopt as the basis for the Java
> memory model.
>
>
>
> Bill
> -------------------------------
> JavaMemoryModel mailing list -
http://www.cs.umd.edu/~pugh/java/memoryModel
>
-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:38 EDT