Fix for Windows atomics

include/atomic/generic-msvc.h

@@ -37,19 +37,17 @@
 #else
 C_MODE_START
 /*Visual Studio 2003 and earlier do not have prototypes for atomic intrinsics*/
-LONG _InterlockedExchange (LONG volatile *Target,LONG Value);
 LONG _InterlockedCompareExchange (LONG volatile *Target, LONG Value, LONG Comp);
-LONG _InterlockedExchangeAdd (LONG volatile *Addend, LONG Value);
+LONGLONG _InterlockedCompareExchange64 (LONGLONG volatile *Target,
+                                        LONGLONG Value, LONGLONG Comp);
 C_MODE_END
 
-#pragma intrinsic(_InterlockedExchangeAdd)
 #pragma intrinsic(_InterlockedCompareExchange)
-#pragma intrinsic(_InterlockedExchange)
+#pragma intrinsic(_InterlockedCompareExchange64)
 #endif
 
-#define InterlockedExchange        _InterlockedExchange
-#define InterlockedExchangeAdd     _InterlockedExchangeAdd
 #define InterlockedCompareExchange _InterlockedCompareExchange
+#define InterlockedCompareExchange64 _InterlockedCompareExchange64
 /*
  No need to do something special for InterlockedCompareExchangePointer
  as it is a #define to InterlockedCompareExchange. The same applies to
@@ -58,33 +56,39 @@ C_MODE_END
 #endif /*_M_IX86*/
 
 #define MY_ATOMIC_MODE "msvc-intrinsics"
-#define IL_EXCHG_ADD32(X,Y)     \
-  InterlockedExchangeAdd((volatile LONG *)(X),(Y))
-#define IL_EXCHG_ADD64(X,Y)     \
-  InterlockedExchangeAdd64((volatile LONGLONG *)(X),(LONGLONG)(Y))
+/* Implement using CAS on WIN32 */
 #define IL_COMP_EXCHG32(X,Y,Z)  \
   InterlockedCompareExchange((volatile LONG *)(X),(Y),(Z))
 #define IL_COMP_EXCHG64(X,Y,Z)  \
   InterlockedCompareExchange64((volatile LONGLONG *)(X), \
                                (LONGLONG)(Y),(LONGLONG)(Z))
 #define IL_COMP_EXCHGptr        InterlockedCompareExchangePointer
+
+#define make_atomic_cas_body(S)                                 \
+  int ## S initial_cmp= *cmp;                                   \
+  int ## S initial_a= IL_COMP_EXCHG ## S (a, set, initial_cmp); \
+  if (!(ret= (initial_a == initial_cmp))) *cmp= initial_a;
+
+#ifndef _M_IX86
+/* Use full set of optimised functions on WIN64 */
+#define IL_EXCHG_ADD32(X,Y)     \
+  InterlockedExchangeAdd((volatile LONG *)(X),(Y))
+#define IL_EXCHG_ADD64(X,Y)     \
+  InterlockedExchangeAdd64((volatile LONGLONG *)(X),(LONGLONG)(Y))
 #define IL_EXCHG32(X,Y)         \
   InterlockedExchange((volatile LONG *)(X),(Y))
 #define IL_EXCHG64(X,Y)         \
   InterlockedExchange64((volatile LONGLONG *)(X),(LONGLONG)(Y))
 #define IL_EXCHGptr             InterlockedExchangePointer
+
 #define make_atomic_add_body(S) \
   v= IL_EXCHG_ADD ## S (a, v)
-#define make_atomic_cas_body(S)                                 \
-  int ## S initial_cmp= *cmp;                                   \
-  int ## S initial_a= IL_COMP_EXCHG ## S (a, set, initial_cmp); \
-  if (!(ret= (initial_a == initial_cmp))) *cmp= initial_a;
 #define make_atomic_swap_body(S) \
   v= IL_EXCHG ## S (a, v)
 #define make_atomic_load_body(S)       \
   ret= 0; /* avoid compiler warning */ \
   ret= IL_COMP_EXCHG ## S (a, ret, ret);
-
+#endif
 /*
   my_yield_processor (equivalent of x86 PAUSE instruction) should be used
   to improve performance on hyperthreaded CPUs. Intel recommends to use it in