Make ff_pw_4 128 bits

Originally committed as revision 24207 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/x86/cavsdsp_mmx.c

@@ -118,7 +118,7 @@ static void cavs_idct8_add_mmx(uint8_t *dst, int16_t *block, int stride)
     for(i=0; i<2; i++){
         DECLARE_ALIGNED(8, uint64_t, tmp);
 
-        cavs_idct8_1d(block+4*i, ff_pw_4);
+        cavs_idct8_1d(block+4*i, ff_pw_4.a);
 
         __asm__ volatile(
             "psraw     $3, %%mm7  \n\t"

libavcodec/x86/dsputil_h264_template_ssse3.c

@@ -43,7 +43,7 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
             "pshuflw $0, %%xmm7, %%xmm7 \n\t"
             "movlhps %%xmm6, %%xmm6 \n\t"
             "movlhps %%xmm7, %%xmm7 \n\t"
-            :: "r"(255*(x+y)+8), "m"(*(rnd?&ff_pw_4:&ff_pw_3))
+            :: "r"(255*(x+y)+8), "m"(*(rnd?&ff_pw_4.a:&ff_pw_3))
         );
 
         if(x) {

libavcodec/x86/dsputil_mmx.c

@@ -47,7 +47,7 @@ DECLARE_ALIGNED(16, const uint64_t, ff_pdw_80000000)[2] =
 {0x8000000080000000ULL, 0x8000000080000000ULL};
 
 DECLARE_ALIGNED(8,  const uint64_t, ff_pw_3  ) = 0x0003000300030003ULL;
-DECLARE_ALIGNED(8,  const uint64_t, ff_pw_4  ) = 0x0004000400040004ULL;
+DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_4  ) = {0x0004000400040004ULL, 0x0004000400040004ULL};
 DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_5  ) = {0x0005000500050005ULL, 0x0005000500050005ULL};
 DECLARE_ALIGNED(16, const xmm_reg,  ff_pw_8  ) = {0x0008000800080008ULL, 0x0008000800080008ULL};
 DECLARE_ALIGNED(8,  const uint64_t, ff_pw_15 ) = 0x000F000F000F000FULL;

libavcodec/x86/dsputil_mmx.h

@@ -33,7 +33,7 @@ extern const uint64_t ff_wtwo;
 extern const uint64_t ff_pdw_80000000[2];
 
 extern const uint64_t ff_pw_3;
-extern const uint64_t ff_pw_4;
+extern const xmm_reg  ff_pw_4;
 extern const xmm_reg  ff_pw_5;
 extern const xmm_reg  ff_pw_8;
 extern const uint64_t ff_pw_15;