Use the intrinsic version of the Interlocked* functions.

base/atomicops_internals_x86_msvc.h

Index: base/atomicops_internals_x86_msvc.h
diff --git a/base/atomicops_internals_x86_msvc.h b/base/atomicops_internals_x86_msvc.h
index 016744c144416f869606881e87ee7d8072f75cdd..0269d894a1fa72d592d3e85b824d84612bb0090f 100644
--- a/base/atomicops_internals_x86_msvc.h
+++ b/base/atomicops_internals_x86_msvc.h
@@ -9,6 +9,8 @@
 
 #include <windows.h>
 
+#include <intrin.h>
+
 #include "base/macros.h"
 
 #if defined(ARCH_CPU_64_BITS)
@@ -26,7 +28,7 @@ namespace subtle {
 inline Atomic32 NoBarrier_CompareAndSwap(volatile Atomic32* ptr,
                                          Atomic32 old_value,
                                          Atomic32 new_value) {
-  LONG result = InterlockedCompareExchange(
+  LONG result = _InterlockedCompareExchange(
       reinterpret_cast<volatile LONG*>(ptr),
       static_cast<LONG>(new_value),
       static_cast<LONG>(old_value));
@@ -35,7 +37,7 @@ inline Atomic32 NoBarrier_CompareAndSwap(volatile Atomic32* ptr,
 
 inline Atomic32 NoBarrier_AtomicExchange(volatile Atomic32* ptr,
                                          Atomic32 new_value) {
-  LONG result = InterlockedExchange(
+  LONG result = _InterlockedExchange(
       reinterpret_cast<volatile LONG*>(ptr),
       static_cast<LONG>(new_value));
   return static_cast<Atomic32>(result);
@@ -43,7 +45,7 @@ inline Atomic32 NoBarrier_AtomicExchange(volatile Atomic32* ptr,
 
 inline Atomic32 Barrier_AtomicIncrement(volatile Atomic32* ptr,
                                         Atomic32 increment) {
-  return InterlockedExchangeAdd(
+  return _InterlockedExchangeAdd(
       reinterpret_cast<volatile LONG*>(ptr),
       static_cast<LONG>(increment)) + increment;
 }