ARM: optimised MAC64 and MLS64

Originally committed as revision 14245 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/armv4l/mathops.h

@@ -57,6 +57,16 @@ static inline av_const int64_t MUL64(int a, int b)
 }
 #define MUL64 MUL64
 
+static inline av_const int64_t MAC64(int64_t d, int a, int b)
+{
+    union { uint64_t x; unsigned hl[2]; } x = { d };
+    asm ("smlal %0, %1, %2, %3"
+         : "+r"(x.hl[0]), "+r"(x.hl[1]) : "r"(a), "r"(b));
+    return x.x;
+}
+#define MAC64(d, a, b) ((d) = MAC64(d, a, b))
+#define MLS64(d, a, b) MAC64(d, -(a), b)
+
 #if defined(HAVE_ARMV5TE)
 
 /* signed 16x16 -> 32 multiply add accumulate */