Merge "Avoid add overflow in summation in AMR-WB decoder" am: a3354a21

    __inline  int32 fxp_mac_16by16(int16 var1,  int16 var2, int32 L_add)

    {

        L_add += (int32)var1 * var2;

        int32 l_orig = L_add;

        if (__builtin_add_overflow( (int32)var1 * var2, l_orig, &L_add)) {

            // needs saturation

            if (l_orig > 0) L_add = MAX_32;

            else            L_add = MIN_32;

        }

        return L_add;

    }