Re: JavaMemoryModel: How figure 2 example is forbidden by the original JMM

From: Jeremy Manson (jmanson@cs.umd.edu)
Date: Thu Nov 13 2003 - 23:19:16 EST


> So, is it possible to deduce that a modified JMM where a read can only see
> the latest write in justification order is equivalent to the proposed JMM?

It wouldn't be equivalent to our memory model, because it would kill
forbidden executions. Specifically, each forbidden execution has a read r
that is forbidden (more or less); there must be an alternate, legal
execution with the same prefix of r where that read must see another
write. If the read has to see the last write, you can't see an another
write.

That may just have pushed your question around a little. Feel free to
follow up :)

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



This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:53 EDT