🚀 go-pugleaf

On Sat, 10 Mar 2001 22:43:52 +0200, Yonatan Taub wrote:

> As I understand, threads create a local copy of variables which
> allows them to execute faster.  They retrieve the master copy
> only after exiting or entering a synchronized block or the
> thread itself.
    :
> The variable "value" was NOT declared volatile but it acts this
> way.  If it wasn't volatile, the inner thread should have NOT
> noticed the change in variable "value" and would continue to
> execute indefinitly.

You are referring to the "memory barrier" that is guaranteed to exist 
when a lock is released.  There is some very interesting discussion of 
memory barriers and their effect on synchronization in the following 
articles:

    http://www.javaworld.com/jw-02-2001/jw-0209-double.html
    http://www.javaworld.com/jw-02-2001/jw-0209-toolbox.html

The reason that your code does not behave as you think it should is 
that the "volatile" keyword imposes additional restrictions on how the 
variable should be handled, but the absence of the "volatile" keyword 
does not.  In other words, the update to the variable from one thread 
must be seen in the other thread if volatile is used.  But the absence 
of the "volatile" keyword does not guarantee that the update will not 
be seen in the other thread.

-- 
Regards,

John McGrath [TeamB]

---------------------------------------------------
Before sending me e-mail, please read:
    http://www.JPMcGrath.net/newsgroups/e-mail.html