[Bug target/55673] Reversed before/after handling in sparc_emit_membar_for_model

[rth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55673

Richard Henderson <rth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |amacleod at redhat dot com

--- Comment #2 from Richard Henderson <rth at gcc dot gnu.org> 2012-12-13 17:31:23 UTC ---
The patch looks correct to me.

And indeed while investigating this I notice that we have a similar problem
with the fully generic code in optabs.c.

Andrew, have a look at expand_atomic_load placing the barrier before the load.
There's also a bit of incorrect commentary cut and paste between there and
expand_atomic_store.  It appears as if the only platform for which we get all
this correct is powerpc, and there because it implements its own atomic_load
and atomic_store patterns.