Subj : Re: Memory Barriers, Compiler Optimizations, etc.
To   : comp.programming.threads
From : Alexander Terekhov
Date : Mon Feb 07 2005 01:46 pm


Joseph Seigh wrote:
> 
> On Mon, 07 Feb 2005 11:17:07 +0100, Alexander Terekhov <terekhov@web.de> wrote:
> 
> >
> > Yes. Note that I said "volatile abuse". Simply put, when you write
> > something like
> >
> > volatile int i = 1;
> >
> > int main() {
> >   return --i;
> > }
> >
> > you better be prepared that smart compiler can tranform it to
> >
> > int main() {
> > }
> >
> 
> The variable i doesn't have an external attribute.  It can't be
> seen anyway.

And since the C/C++ standards don't concern themselves with threads, 
use of "internal" volatiles by multiple threads doesn't constitute 
"external attribute" in your terminology. IOW, C/C++ compiler can 
simply ignore volatile abuse and treat all such variables as 
nonvolatile.

regards,
alexander.

.