> "loop unrollings should be allowed to fuse
> unrolled segments" looks like it contradicts Rule 2.
Well, if you squint at the rule from the right direction, it doesn't
apply to unrollings since no segment is fused with "itself". Instead,
one part is fused with something that just so happens to look the
same.
I admit that the freedom to abuse such interpretations in any way we
like probably reveals a weakness in how I phrased the rule. Any ideas
for improvements would be appreciated.
-Doug
-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel
This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:29 EDT