This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch for PR libgomp/37938, IA64 specific.
On Thu, Nov 06, 2008 at 03:57:04PM -0800, H.J. Lu wrote:
> On Thu, Nov 6, 2008 at 3:47 PM, Steve Ellcey <sje@cup.hp.com> wrote:
> > On Fri, 2008-11-07 at 00:24 +0100, Jakub Jelinek wrote:
> >
> > I really think that using __sync_synchronize in config/linux/mutex.h and
> >> have all targets but ia64 override this header is worse than having
> >> just special config/linux/ia64/mutex.h.
> >>
> >> Jakub
> >
> > Yes, but I felt I had to protest at least a little. I will test this
> > libgomp patch overnight to verify that it works OK and has no
> > regressions. The only change from this mutex.h and the standard one is
> > the call to __sync_synchronize in gomp_mutex_unlock (and the comments
> > for that routine).
> >
>
> FWIW, ia64 has __sync_lock_release.
Sure, but we need to (atomically) read the previous value before storing 0
to find out if futex_wake is needed or not.
Jakub