[cpp-threads] Alternatives to SC (prohibiting data races, etc.)

[Paul E. McKenney]

On Wed, Jan 31, 2007 at 06:58:03PM -0800, Lawrence Crowl wrote:
> On 1/31/07, Paul E. McKenney <paulmck at linux.vnet.ibm.com> wrote:
> >OK -- but I didn't see load_raw() in N2145.  This is something you
> >are proposing adding -- or did I get the wrong document?
> 
> Look under core functions, for the definition:
> 
> inline ${TYPENAME} atomic_load_${READ}( atomic_${TYPEKEY} volatile * __a__ )
> 
> After choosing appropriate values for the variables, you have
> 
> int atomic_load_relaxed( atomic_int volatile * __a__ )
> 
> The use would be
> 
>    extern atomic_int av;
>    .... int v = atomic_load_relaxed( &av ); ....

OK, so the real name is atomic_load_relaxed() as opposed to load_raw()?
And this will do a load without memory barriers or atomics, but with no
danger of subsequent refetches?  And, if applied to a lock_free() datum,
the load is atomic in that it sees the value from a particular prior store
(or the initial value)?

> >So atomic<char>::lock_free() would normally (always?) return true, while
> >atomic<struct foo>::lock_free() would normally return false, right?
> 
> Right, presuming struct foo is big enough.  If it is small, it could
> be lock free.

Good point!

						Thanx, Paul