Small improvement to InterlockedExchange on ARM

Use movs instead of mov when setting the result in r0. This way the Z
flag will be set for the calling function which might allow some smaller
optimizations later on. It does not affect current code in any way,
because flags are not expected to be used across function calls.

git-svn-id: trunk@22033 -

rtl/arm/arm.inc

@@ -511,7 +511,7 @@ asm
   strex r2, r1, [r0]
   cmp r2, #0
   bne .Lloop
-  mov r0, r1
+  movs r0, r1
   bx  lr
 {$else}
 {$if defined(LINUX) and defined(CPUARMEL)}
@@ -535,7 +535,9 @@ asm
   sub r1, r0, #1 // Decrement value
   blx r3	 // Call kuser_cmpxchg, sets C-Flag on success
 
-  movcs r0, r1	 // We expect that to work most of the time so keep it pipeline friendly
+  // MOVS sets the Z flag when the result reaches zero, this can be used later on
+  // The C-Flag will not be modified by this because we're not doing any shifting
+  movcss r0, r1	 // We expect that to work most of the time so keep it pipeline friendly
   ldmcsfd r13!, {pc}
   b .Latomic_dec_loop // kuser_cmpxchg sets C flag on error
 
@@ -551,7 +553,7 @@ asm
   ldr r1, [r0]
   sub r1, r1, #1
   str r1, [r0]
-  mov r0, r1
+  movs r0, r1
 // unlock and return
   str r2, [r3]
   bx  lr