avcodec/x86/cavsdsp: remove unneeded tmp

This is faster and simpler as well

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
(cherry picked from commit d79f7bf0d63a81ee66026ee92a6946a7303d04bd)

Conflicts:

	libavcodec/x86/cavsdsp.c

1 files changed, 11 insertions, 13 deletions

diff --git a/libavcodec/x86/cavsdsp.c b/libavcodec/x86/cavsdsp.c
index d155fb20bc..6199a33209 100644
--- a/libavcodec/x86/cavsdsp.c
+++ b/libavcodec/x86/cavsdsp.c
@@ -142,8 +142,6 @@ static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride)
     DECLARE_ALIGNED(8, int16_t, b2)[64];
 
     for(i=0; i<2; i++){
-        DECLARE_ALIGNED(8, uint64_t, tmp);
-
         cavs_idct8_1d(block+4*i, ff_pw_4.a);
 
         __asm__ volatile(
@@ -155,19 +153,19 @@ static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride)
             "psraw     $3, %%mm2  \n\t"
             "psraw     $3, %%mm1  \n\t"
             "psraw     $3, %%mm0  \n\t"
-            "movq   %%mm7,    %0   \n\t"
+            "movq   %%mm7,   (%0) \n\t"
             TRANSPOSE4( %%mm0, %%mm2, %%mm4, %%mm6, %%mm7 )
-            "movq   %%mm0,  8(%1)  \n\t"
-            "movq   %%mm6, 24(%1)  \n\t"
-            "movq   %%mm7, 40(%1)  \n\t"
-            "movq   %%mm4, 56(%1)  \n\t"
-            "movq    %0,    %%mm7  \n\t"
+            "movq   %%mm0,  8(%0)  \n\t"
+            "movq   %%mm6, 24(%0)  \n\t"
+            "movq   %%mm7, 40(%0)  \n\t"
+            "movq   %%mm4, 56(%0)  \n\t"
+            "movq    (%0),  %%mm7  \n\t"
             TRANSPOSE4( %%mm7, %%mm5, %%mm3, %%mm1, %%mm0 )
-            "movq   %%mm7,   (%1)  \n\t"
-            "movq   %%mm1, 16(%1)  \n\t"
-            "movq   %%mm0, 32(%1)  \n\t"
-            "movq   %%mm3, 48(%1)  \n\t"
-            : "=m"(tmp)
+            "movq   %%mm7,   (%0)  \n\t"
+            "movq   %%mm1, 16(%0)  \n\t"
+            "movq   %%mm0, 32(%0)  \n\t"
+            "movq   %%mm3, 48(%0)  \n\t"
+            :
             : "r"(b2+32*i)
             : "memory"
         );