Re: JavaMemoryModel: No "Spooky errors at a distance"

From: Martin Trotter (martin_trotter@uk.ibm.com)
Date: Fri Mar 08 2002 - 08:20:09 EST

I strongly agree but am concerned that your question contains some subtlety
that I'm missing. If r2 is correctly synchronized then that should
certainly ensure that the initialized version is always seen. To do
otherwise would imply that some part of the synchronization had been
'optimised out', in this case presumably the load barrier in r2. Is that
what you're trying to rule out in the spec ?

Martin Trotter

-------------------------------
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