Subj : Re: volatile or mutex
To   : comp.programming.threads
From : Alexander Terekhov
Date : Thu May 12 2005 04:55 am

"Scott McPhillips [MVP]" wrote:
[...]
> volatile has nothing to do with synchronization.  It only assures that
> the compiler will actually read the variable's memory when you tell it to.

Apart from mutable locals affected with respect to setjmp/longjmp() and 
static sig_atomic_t mess, with respect to objects with standard C/C++ 
storage duration, volatile has no meaning whatsoever. Abstract C/C++ 
machine defines neither "read" nor "variable's memory" (apart from 
accesses to volatile stuff residing outside and thus "observable" just 
like standard I/O -- both constitute "observable behavior" ). Volatiles 
are brain-dead (including threads relevant Java & .Net volatile beasts). 
As for better abstractions... just wait for C++0X (09 hopefully). Well, 
that's apart from brand new TR1's <iohw.h> and <hardware> for memory 
mapped registers and alike stuff under (implied) SC for the outer world, 
so to speak).

regards,
alexander.

.