This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [v3 RFC] PATCH to memory asm ops in atomicity.h
On Wed, 17 Dec 2003 04:05:51 -0800, Richard Henderson <rth@redhat.com> wrote:
> On Tue, Dec 16, 2003 at 09:00:01PM -0500, Jason Merrill wrote:
>> They're used for atomic access to a particular memory location. Why isn't
>> it enough to specify that location directly?
>
> If that is *all* they are used for, then fine. But if an
> atomic increment is used to implement a mutex, then it is
> not fine.
They aren't; code that needs a mutex uses the gthread API. These functions
are used for reference counting. Actually, a mutex is used to implement
atomic add on targets that don't have suitable primitives.
>> Note also that the asms are marked volatile, which should prevent
>> scheduling across them.
>
> That only applies to scheduling. Not to cse or load-store motion or
> any other optimization that shouldn't cross a thread synchronization
> point.
Good to know, thanks.
Jason