Accumulate results in output register explicitly

Author: arnej27959

kernel/arm64/dot_kernel_asimd.c

@@ -134,7 +134,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 	"	fadd	v4.4s, v4.4s, v6.4s		\n"	\
 	"	fadd	v0.4s, v0.4s, v4.4s		\n"	\
 	"	faddp	v0.4s, v0.4s, v0.4s		\n"	\
-	"	faddp	v0.4s, v0.4s, v0.4s		\n"
+	"	faddp	"OUT", v0.2s			\n"
 
 #else /* !defined(DSDOT) */
 #define KERNEL_F1						\