dsputil_mmx: prefer xmm registers below xmm6 when they are available

Originally committed as revision 25606 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/x86/dsputil_mmx.c

@@ -2012,8 +2012,8 @@ static void vorbis_inverse_coupling_sse(float *mag, float *ang, int blocksize)
         "1: \n"\
         "movaps  (%3,%0), %%xmm0 \n"\
  stereo("movaps   %%xmm0, %%xmm1 \n")\
-        "mulps    %%xmm6, %%xmm0 \n"\
- stereo("mulps    %%xmm7, %%xmm1 \n")\
+        "mulps    %%xmm4, %%xmm0 \n"\
+ stereo("mulps    %%xmm5, %%xmm1 \n")\
         "lea 1024(%3,%0), %1 \n"\
         "mov %5, %2 \n"\
         "2: \n"\
@@ -2051,12 +2051,12 @@ static void ac3_downmix_sse(float (*samples)[256], float (*matrix)[2], int out_c
         __asm__ volatile(
             "1: \n"
             "sub $8, %0 \n"
-            "movss     (%2,%0), %%xmm6 \n"
-            "movss    4(%2,%0), %%xmm7 \n"
-            "shufps $0, %%xmm6, %%xmm6 \n"
-            "shufps $0, %%xmm7, %%xmm7 \n"
-            "movaps %%xmm6,   (%1,%0,4) \n"
-            "movaps %%xmm7, 16(%1,%0,4) \n"
+            "movss     (%2,%0), %%xmm4 \n"
+            "movss    4(%2,%0), %%xmm5 \n"
+            "shufps $0, %%xmm4, %%xmm4 \n"
+            "shufps $0, %%xmm5, %%xmm5 \n"
+            "movaps %%xmm4,   (%1,%0,4) \n"
+            "movaps %%xmm5, 16(%1,%0,4) \n"
             "jg 1b \n"
             :"+&r"(j)
             :"r"(matrix_simd), "r"(matrix)

libavcodec/x86/dsputilenc_mmx.c

@@ -61,16 +61,16 @@ static void get_pixels_mmx(DCTELEM *block, const uint8_t *pixels, int line_size)
 static void get_pixels_sse2(DCTELEM *block, const uint8_t *pixels, int line_size)
 {
     __asm__ volatile(
-        "pxor %%xmm7,      %%xmm7         \n\t"
+        "pxor %%xmm4,      %%xmm4         \n\t"
         "movq (%0),        %%xmm0         \n\t"
         "movq (%0, %2),    %%xmm1         \n\t"
         "movq (%0, %2,2),  %%xmm2         \n\t"
         "movq (%0, %3),    %%xmm3         \n\t"
         "lea (%0,%2,4), %0                \n\t"
-        "punpcklbw %%xmm7, %%xmm0         \n\t"
-        "punpcklbw %%xmm7, %%xmm1         \n\t"
-        "punpcklbw %%xmm7, %%xmm2         \n\t"
-        "punpcklbw %%xmm7, %%xmm3         \n\t"
+        "punpcklbw %%xmm4, %%xmm0         \n\t"
+        "punpcklbw %%xmm4, %%xmm1         \n\t"
+        "punpcklbw %%xmm4, %%xmm2         \n\t"
+        "punpcklbw %%xmm4, %%xmm3         \n\t"
         "movdqa %%xmm0,      (%1)         \n\t"
         "movdqa %%xmm1,    16(%1)         \n\t"
         "movdqa %%xmm2,    32(%1)         \n\t"
@@ -79,10 +79,10 @@ static void get_pixels_sse2(DCTELEM *block, const uint8_t *pixels, int line_size
         "movq (%0, %2),    %%xmm1         \n\t"
         "movq (%0, %2,2),  %%xmm2         \n\t"
         "movq (%0, %3),    %%xmm3         \n\t"
-        "punpcklbw %%xmm7, %%xmm0         \n\t"
-        "punpcklbw %%xmm7, %%xmm1         \n\t"
-        "punpcklbw %%xmm7, %%xmm2         \n\t"
-        "punpcklbw %%xmm7, %%xmm3         \n\t"
+        "punpcklbw %%xmm4, %%xmm0         \n\t"
+        "punpcklbw %%xmm4, %%xmm1         \n\t"
+        "punpcklbw %%xmm4, %%xmm2         \n\t"
+        "punpcklbw %%xmm4, %%xmm3         \n\t"
         "movdqa %%xmm0,    64(%1)         \n\t"
         "movdqa %%xmm1,    80(%1)         \n\t"
         "movdqa %%xmm2,    96(%1)         \n\t"