__atomic_load doesn't match ACQUIRE memory model

[Yvan Roux]

Compiling this code for ARMv7

int v;

int foo()
{
  return __atomic_load_n (&v, __ATOMIC_ACQUIRE);
}

generates a data memory barrier defore the load:

foo:
    movw r3, #:lower16:v
    movt r3, #:upper16:v
    dmb sy
    ldr r0, [r3, #0]
    bx lr

But, the `Acquire’ semantics imply that no reads in the current thread
dependent on the value currently loaded can be reordered before this load.
Thus, the barrier needs to be after the load and not before.

This problem is referenced in GCC PR http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55492

[Yvan Roux (yvan-roux)]

Fix commited upstream r194490 and 194504