> > There are also happens-before edges from each action to all previous
>> actions in that thread.
>
>So the implication, to me, is that all actions are ordered as per the
>program, unless the semantics of the memory model allow them to be
>reordered - as per the table in figure 3. Is that right? That means we would
>need to specify all allowed reorderings - which is something I thought we
>had already determined we couldn't/wouldn't do??
>
>David
Figure 3 is part of a theorem showing that some reorderings are
valid. Figure 3 is derived from the memory model, it isn't a part of
the memory model.
Does this clarify the issue?
Bill
-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:40 EDT