diff --git a/libfixmath/fix16.c b/libfixmath/fix16.c
index a557ef1..309b913 100644
--- a/libfixmath/fix16.c
+++ b/libfixmath/fix16.c
@@ -14,16 +14,16 @@ fix16_t fix16_sadd(fix16_t inArg0, fix16_t inArg1) {
 
 
 
-#if defined(__arm__) || defined(_ARM)
+#if defined(__arm__) || defined(_ARM) && !defined(__thumb__)
 fix16_t fix16_mul(int32_t inArg0, int32_t inArg1) {
 	fix16_t  res;
 	asm(
-	"smull %1, r0, %2, %3\n\t"
+	"smull %0, r0, %1, %2\n\t"
 	#ifndef FIXMATH_NO_ROUNDING
-	"add   %1, %1, #0x8000\n\t"
+	"add   %0, %0, #0x8000\n\t"
 	#endif
-	"mov   %1, %1, LSR #16\n\t"
-	"or    %1, %1, r0, LSL #16"
+	"mov   %0, %0, LSR #16\n\t"
+	"or    %0, %0, r0, LSL #16"
 	: "=r"(res)
 	: "r"(inArg0), "r"(inArg1)
 	: "r0");