Re: JavaMemoryModel: Implementing KeepAlive

From: Sylvia Else (sylviae@optushome.com.au)
Date: Sat Apr 24 2004 - 02:19:57 EDT


At 10:58 PM 23/04/2004 -0700, Hans Boehm wrote:
>Yes. That's one reason not to go with the weak interpretation.
>
>I now think there are others. It's possible to break a program by writing
>zero to a volatile variable that is always zero. I think this complicates
>reasoning about programs without much benefit. We decided to go with the
>strong interpretation in the end.

I don't see that. Nothing about the weak interpretation ever suggested that
an optimiser could throw aware happens-before edges that the model
specified as existing. It might throw away the actual write and read, but
only if doing so didn't have a hardware impact that negated the effects
required by the happens-before relationship.

I rather thought the strong v weak decision was still a live one, given
Sarita's last posting on the subject.

Sylvia.

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



This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:01:06 EDT