I earlier said:
> The key question then is how to formalize when a write is
> responsible for
> another write. A naïve definition would trace through
> traditional control
> and flow dependences for this purpose. This would be a simple
> constructive
> definition, but unfortunately, prohibits optimizations such
>>>>>>>>> as test 3. To be
> more aggressive, we resort to a more abstract definition:
Above, test 3 should be replaced with test 18.
Sarita
-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:57 EDT