JavaMemoryModel: Performing speculative writes ahead of loops that may not terminate

From: Eliot Moss (moss@cs.umass.edu)
Date: Fri May 28 2004 - 12:24:33 EDT

I recall global instruction scheduling work that I think would allow the
x=1 to be moved. If it is executed, it is executed under the same control
conditions as code immediately preceding the while loop. However, if x is
volatile and there are any volatile accesses in the loop, then it would not
be a legal transformation, right?

(I think the paper in question was Bernstein and Rodeh; of course they obey
data dependences as well, but here there are none.)

-- Eliot
-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel

This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:01:08 EDT