x86/dsputil: add emms to ff_scalarproduct_int16_mmxext()

Also undo the changes to ra144enc.c from previous commits.
Should fix ticket #3429

Signed-off-by: James Almer <jamrial@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
(cherry picked from commit 9e0e1f9067430de1655a7b28536b5afed48bded5)

Conflicts:

	libavcodec/ra144enc.c

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

libavcodec/ra144enc.c

@@ -34,7 +34,6 @@
 #include "celp_filters.h"
 #include "ra144.h"
 
-
 static av_cold int ra144_encode_close(AVCodecContext *avctx)
 {
     RA144Context *ractx = avctx->priv_data;

libavcodec/x86/dsputil.asm

@@ -61,6 +61,9 @@ cglobal scalarproduct_int16, 3,3,3, v1, v2, order
 %endif
     paddd   m2, m0
     movd   eax, m2
+%if mmsize == 8
+    emms
+%endif
     RET
 
 ; int scalarproduct_and_madd_int16(int16_t *v1, int16_t *v2, int16_t *v3, int order, int mul)