Re: JavaMemoryModel: Draft JMM Synopsis

From: Jeremy Manson (jmanson@cs.umd.edu)
Date: Sun Mar 07 2004 - 11:47:51 EST


> I had tried to avoid the term "happens before" here. It is a crummy
> term. It is a good idea to use the term in the spec to clarify
> relations to other memory model work; plus it is mainly expressed in
> symbols anyway. But it seems needlessly confusing to say things like
> "an unlock happens before a subsequent lock".

> But my attempt to use "complete" here was completely(!) wrong, and led
> to lots of misstatements. The best alternative I can think of for "i
> happens before j" is "the effects of i occur before those of j". Which
> seems too verbose, and might still have unwanted connotations for some
> readers. Does anyone have any better suggestions?

We actually avoided "occurs before" because we tend to use it for actions
that occur before others in the the total order over actions. We are very
careful in our use of happens-before to mean synchronization + program
order, and occurs before to mean justification order.

If you really wanted to change it, you could probably come up with some
variant of "ordered before".

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



This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:59 EDT