Atomic swap
Daniel Eischen
eischen at vigrid.com
Thu Aug 7 12:31:57 PDT 2003
On Thu, 7 Aug 2003, Marcel Moolenaar wrote:
> On Thu, Aug 07, 2003 at 03:06:33PM -0400, Portante, Peter wrote:
> > Marcel,
> >
> > > atomic_swap_long(volatile long *dst, long val, long *res)
> > > {
> > > __asm ( "1: ldq_l t0,%0\n"
> > > " mov %1,t1\n"
> > > " stq_c t1,%0\n"
> > > " beq t1,1b\n"
> > > " stq t0,%3\n"
> ^^^^^^^^^^^^^^^^^^^^
>
> Whoops, typo: %3 should be %2.
Yup, I caught that but didn't want to complain when something
is handed to me on a silver platter ;-)
> > > :: "m"(*dst), "r"(val), "m"(*res) : "memory");
> > > }
> > >
> > A word of caution on performing that stq without an MB before it:
> > another processor cannot read that location and the destination
> > location and assume anything about their contents based on what
> > they read unless an MB is between them.
>
> Good point. I don't think we have to worry about it, though. This
> function has a specific usage (see src/lib/libpthread/sys/lock.c).
> If I understand the code correctly, *res is never used other than
> the caller of atomic_swap_long that writes to *res.
Yes, perhaps the interface should be:
long atomic_swap_long(volatile long *dst, long val)
instead of having the result as the 3rd argument. But it's
already used this way in libpthread, so I didn't want to
change it. I can always do that later.
Hmm, it would be nice just to have atomic_swap (or atomic_xchg,
or whatever you want to call it) in <machine/atomic.h>.
--
Dan Eischen
More information about the freebsd-alpha
mailing list