Refine atomic operations for Linux/ARM.

Refine atomic operations for Linux/ARM.

This patch comes from recent discussions from the following
review link:

  https://chromiumcodereview.appspot.com/16109010/

In a nutshell:

- Using the "cmp" instruction, instead of "teq" is
  better for recent ARM CPUs.

- Fix a tiny typo.

- Fix the ARMv5 implementation of Release_CompareAndSwap()
  since the old one could return without performing any
  memory barrier when prev_value != old_value.

  It's not clear whether the old behaviour is desirable
  because client code could naively assume the function
  always perform a release operation, even in case of
  failure.

BUG=234215
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=211013

[Mark Mentovai, 2013-07-10 21:49:44]

LGTM

[JF, 2013-07-10 21:51:26]

lgtm

[commit-bot: I haz the power, 2013-07-10 22:01:44]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/digit@chromium.org/18984012/2001

[commit-bot: I haz the power, 2013-07-11 03:52:58]

Change committed as 211013

[Dmitry Vyukov, 2013-07-15 14:42:32]

https://chromiumcodereview.appspot.com/18984012/diff/2001/base/atomicops_inte...
File base/atomicops_internals_arm_gcc.h (right):

https://chromiumcodereview.appspot.com/18984012/diff/2001/base/atomicops_inte...
base/atomicops_internals_arm_gcc.h:245: return Acquire_CompareAndSwap(ptr,
old_value, new_value);
A failing Release_CompareAndSwap() does not need any memory barriers. Release
makes sense only when an acquire operation loads the value stored by side effect
of the release operation. A failing release CAS does not mutate anything, so
nobody can possibly observe it's side effect.

See the relevant part of the C++ standard:

29.6.5/21
Effects: Atomically, compares the contents of the memory pointed to by object or
by this for equality
with that in expected, and if true, replaces the contents of the memory pointed
to by object or by
this with that in desired, and if false, updates the contents of the memory in
expected with the
contents of the memory pointed to by object or by this. Further, if the
comparison is true, memory
is affected according to the value of success, and if the comparison is false,
memory is affected
according to the value of failure. When only one memory_order argument is
supplied, the value of
success is order, and the value of failure is order except that a value of
memory_order_acq_rel
shall be replaced by the value memory_order_acquire and a value of
memory_order_release shall
be replaced by the value memory_order_relaxed. If the operation returns true,
these operations are
atomic read-modify-write operations (1.10). Otherwise, these operations are
atomic load operations.