Bump dates to 2014
[x262.git] / common / pixel.c
index 979c731..2e82812 100644 (file)
@@ -1,7 +1,7 @@
 /*****************************************************************************
  * pixel.c: pixel metrics
  *****************************************************************************
- * Copyright (C) 2003-2011 x264 project
+ * Copyright (C) 2003-2014 x264 project
  *
  * Authors: Loren Merritt <lorenm@u.washington.edu>
  *          Laurent Aimar <fenrir@via.ecp.fr>
@@ -29,6 +29,7 @@
 
 #if HAVE_MMX
 #   include "x86/pixel.h"
+#   include "x86/predict.h"
 #endif
 #if ARCH_PPC
 #   include "ppc/pixel.h"
@@ -45,8 +46,8 @@
  * pixel_sad_WxH
  ****************************************************************************/
 #define PIXEL_SAD_C( name, lx, ly ) \
-static int name( pixel *pix1, int i_stride_pix1,  \
-                 pixel *pix2, int i_stride_pix2 ) \
+static int name( pixel *pix1, intptr_t i_stride_pix1,  \
+                 pixel *pix2, intptr_t i_stride_pix2 ) \
 {                                                   \
     int i_sum = 0;                                  \
     for( int y = 0; y < ly; y++ )                   \
@@ -67,16 +68,16 @@ PIXEL_SAD_C( x264_pixel_sad_16x8,  16,  8 )
 PIXEL_SAD_C( x264_pixel_sad_8x16,   8, 16 )
 PIXEL_SAD_C( x264_pixel_sad_8x8,    8,  8 )
 PIXEL_SAD_C( x264_pixel_sad_8x4,    8,  4 )
+PIXEL_SAD_C( x264_pixel_sad_4x16,   4, 16 )
 PIXEL_SAD_C( x264_pixel_sad_4x8,    4,  8 )
 PIXEL_SAD_C( x264_pixel_sad_4x4,    4,  4 )
 
-
 /****************************************************************************
  * pixel_ssd_WxH
  ****************************************************************************/
 #define PIXEL_SSD_C( name, lx, ly ) \
-static int name( pixel *pix1, int i_stride_pix1,  \
-                 pixel *pix2, int i_stride_pix2 ) \
+static int name( pixel *pix1, intptr_t i_stride_pix1,  \
+                 pixel *pix2, intptr_t i_stride_pix2 ) \
 {                                                   \
     int i_sum = 0;                                  \
     for( int y = 0; y < ly; y++ )                   \
@@ -97,10 +98,12 @@ PIXEL_SSD_C( x264_pixel_ssd_16x8,  16,  8 )
 PIXEL_SSD_C( x264_pixel_ssd_8x16,   8, 16 )
 PIXEL_SSD_C( x264_pixel_ssd_8x8,    8,  8 )
 PIXEL_SSD_C( x264_pixel_ssd_8x4,    8,  4 )
+PIXEL_SSD_C( x264_pixel_ssd_4x16,   4, 16 )
 PIXEL_SSD_C( x264_pixel_ssd_4x8,    4,  8 )
 PIXEL_SSD_C( x264_pixel_ssd_4x4,    4,  4 )
 
-uint64_t x264_pixel_ssd_wxh( x264_pixel_function_t *pf, pixel *pix1, int i_pix1, pixel *pix2, int i_pix2, int i_width, int i_height )
+uint64_t x264_pixel_ssd_wxh( x264_pixel_function_t *pf, pixel *pix1, intptr_t i_pix1,
+                             pixel *pix2, intptr_t i_pix2, int i_width, int i_height )
 {
     uint64_t i_ssd = 0;
     int y;
@@ -140,7 +143,8 @@ uint64_t x264_pixel_ssd_wxh( x264_pixel_function_t *pf, pixel *pix1, int i_pix1,
     return i_ssd;
 }
 
-static void pixel_ssd_nv12_core( pixel *pixuv1, int stride1, pixel *pixuv2, int stride2, int width, int height, uint64_t *ssd_u, uint64_t *ssd_v )
+static void pixel_ssd_nv12_core( pixel *pixuv1, intptr_t stride1, pixel *pixuv2, intptr_t stride2,
+                                 int width, int height, uint64_t *ssd_u, uint64_t *ssd_v )
 {
     *ssd_u = 0, *ssd_v = 0;
     for( int y = 0; y < height; y++, pixuv1+=stride1, pixuv2+=stride2 )
@@ -153,7 +157,8 @@ static void pixel_ssd_nv12_core( pixel *pixuv1, int stride1, pixel *pixuv2, int
         }
 }
 
-void x264_pixel_ssd_nv12( x264_pixel_function_t *pf, pixel *pix1, int i_pix1, pixel *pix2, int i_pix2, int i_width, int i_height, uint64_t *ssd_u, uint64_t *ssd_v )
+void x264_pixel_ssd_nv12( x264_pixel_function_t *pf, pixel *pix1, intptr_t i_pix1, pixel *pix2, intptr_t i_pix2,
+                          int i_width, int i_height, uint64_t *ssd_u, uint64_t *ssd_v )
 {
     pf->ssd_nv12_core( pix1, i_pix1, pix2, i_pix2, i_width&~7, i_height, ssd_u, ssd_v );
     if( i_width&7 )
@@ -168,11 +173,11 @@ void x264_pixel_ssd_nv12( x264_pixel_function_t *pf, pixel *pix1, int i_pix1, pi
 /****************************************************************************
  * pixel_var_wxh
  ****************************************************************************/
-#define PIXEL_VAR_C( name, w ) \
-static uint64_t name( pixel *pix, int i_stride ) \
+#define PIXEL_VAR_C( name, w, h ) \
+static uint64_t name( pixel *pix, intptr_t i_stride ) \
 {                                             \
     uint32_t sum = 0, sqr = 0;                \
-    for( int y = 0; y < w; y++ )              \
+    for( int y = 0; y < h; y++ )              \
     {                                         \
         for( int x = 0; x < w; x++ )          \
         {                                     \
@@ -184,32 +189,37 @@ static uint64_t name( pixel *pix, int i_stride ) \
     return sum + ((uint64_t)sqr << 32);       \
 }
 
-PIXEL_VAR_C( x264_pixel_var_16x16, 16 )
-PIXEL_VAR_C( x264_pixel_var_8x8,    8 )
+PIXEL_VAR_C( x264_pixel_var_16x16, 16, 16 )
+PIXEL_VAR_C( x264_pixel_var_8x16,   8, 16 )
+PIXEL_VAR_C( x264_pixel_var_8x8,    8,  8 )
 
 /****************************************************************************
  * pixel_var2_wxh
  ****************************************************************************/
-static int pixel_var2_8x8( pixel *pix1, int i_stride1, pixel *pix2, int i_stride2, int *ssd )
-{
-    uint32_t var = 0, sum = 0, sqr = 0;
-    for( int y = 0; y < 8; y++ )
-    {
-        for( int x = 0; x < 8; x++ )
-        {
-            int diff = pix1[x] - pix2[x];
-            sum += diff;
-            sqr += diff * diff;
-        }
-        pix1 += i_stride1;
-        pix2 += i_stride2;
-    }
-    sum = abs(sum);
-    var = sqr - ((uint64_t)sum * sum >> 6);
-    *ssd = sqr;
-    return var;
+#define PIXEL_VAR2_C( name, w, h, shift ) \
+static int name( pixel *pix1, intptr_t i_stride1, pixel *pix2, intptr_t i_stride2, int *ssd ) \
+{ \
+    uint32_t var = 0, sum = 0, sqr = 0; \
+    for( int y = 0; y < h; y++ ) \
+    { \
+        for( int x = 0; x < w; x++ ) \
+        { \
+            int diff = pix1[x] - pix2[x]; \
+            sum += diff; \
+            sqr += diff * diff; \
+        } \
+        pix1 += i_stride1; \
+        pix2 += i_stride2; \
+    } \
+    sum = abs(sum); \
+    var = sqr - ((uint64_t)sum * sum >> shift); \
+    *ssd = sqr; \
+    return var; \
 }
 
+PIXEL_VAR2_C( x264_pixel_var2_8x16, 8, 16, 7 )
+PIXEL_VAR2_C( x264_pixel_var2_8x8,  8,  8, 6 )
+
 #if BIT_DEPTH > 8
     typedef uint32_t sum_t;
     typedef uint64_t sum2_t;
@@ -242,7 +252,7 @@ static ALWAYS_INLINE sum2_t abs2( sum2_t a )
  * pixel_satd_WxH: sum of 4x4 Hadamard transformed differences
  ****************************************************************************/
 
-static NOINLINE int x264_pixel_satd_4x4( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
+static NOINLINE int x264_pixel_satd_4x4( pixel *pix1, intptr_t i_pix1, pixel *pix2, intptr_t i_pix2 )
 {
     sum2_t tmp[4][2];
     sum2_t a0, a1, a2, a3, b0, b1;
@@ -267,7 +277,7 @@ static NOINLINE int x264_pixel_satd_4x4( pixel *pix1, int i_pix1, pixel *pix2, i
     return sum >> 1;
 }
 
-static NOINLINE int x264_pixel_satd_8x4( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
+static NOINLINE int x264_pixel_satd_8x4( pixel *pix1, intptr_t i_pix1, pixel *pix2, intptr_t i_pix2 )
 {
     sum2_t tmp[4][4];
     sum2_t a0, a1, a2, a3;
@@ -289,7 +299,7 @@ static NOINLINE int x264_pixel_satd_8x4( pixel *pix1, int i_pix1, pixel *pix2, i
 }
 
 #define PIXEL_SATD_C( w, h, sub )\
-static int x264_pixel_satd_##w##x##h( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )\
+static int x264_pixel_satd_##w##x##h( pixel *pix1, intptr_t i_pix1, pixel *pix2, intptr_t i_pix2 )\
 {\
     int sum = sub( pix1, i_pix1, pix2, i_pix2 )\
             + sub( pix1+4*i_pix1, i_pix1, pix2+4*i_pix2, i_pix2 );\
@@ -308,10 +318,10 @@ PIXEL_SATD_C( 16, 16, x264_pixel_satd_8x4 )
 PIXEL_SATD_C( 16, 8,  x264_pixel_satd_8x4 )
 PIXEL_SATD_C( 8,  16, x264_pixel_satd_8x4 )
 PIXEL_SATD_C( 8,  8,  x264_pixel_satd_8x4 )
+PIXEL_SATD_C( 4,  16, x264_pixel_satd_4x4 )
 PIXEL_SATD_C( 4,  8,  x264_pixel_satd_4x4 )
 
-
-static NOINLINE int sa8d_8x8( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
+static NOINLINE int sa8d_8x8( pixel *pix1, intptr_t i_pix1, pixel *pix2, intptr_t i_pix2 )
 {
     sum2_t tmp[8][4];
     sum2_t a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3;
@@ -345,13 +355,13 @@ static NOINLINE int sa8d_8x8( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
     return sum;
 }
 
-static int x264_pixel_sa8d_8x8( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
+static int x264_pixel_sa8d_8x8( pixel *pix1, intptr_t i_pix1, pixel *pix2, intptr_t i_pix2 )
 {
     int sum = sa8d_8x8( pix1, i_pix1, pix2, i_pix2 );
     return (sum+2)>>2;
 }
 
-static int x264_pixel_sa8d_16x16( pixel *pix1, int i_pix1, pixel *pix2, int i_pix2 )
+static int x264_pixel_sa8d_16x16( pixel *pix1, intptr_t i_pix1, pixel *pix2, intptr_t i_pix2 )
 {
     int sum = sa8d_8x8( pix1, i_pix1, pix2, i_pix2 )
             + sa8d_8x8( pix1+8, i_pix1, pix2+8, i_pix2 )
@@ -360,8 +370,7 @@ static int x264_pixel_sa8d_16x16( pixel *pix1, int i_pix1, pixel *pix2, int i_pi
     return (sum+2)>>2;
 }
 
-
-static NOINLINE uint64_t pixel_hadamard_ac( pixel *pix, int stride )
+static NOINLINE uint64_t pixel_hadamard_ac( pixel *pix, intptr_t stride )
 {
     sum2_t tmp[32];
     sum2_t a0, a1, a2, a3, dc;
@@ -399,7 +408,7 @@ static NOINLINE uint64_t pixel_hadamard_ac( pixel *pix, int stride )
 }
 
 #define HADAMARD_AC(w,h) \
-static uint64_t x264_pixel_hadamard_ac_##w##x##h( pixel *pix, int stride )\
+static uint64_t x264_pixel_hadamard_ac_##w##x##h( pixel *pix, intptr_t stride )\
 {\
     uint64_t sum = pixel_hadamard_ac( pix, stride );\
     if( w==16 )\
@@ -420,13 +429,15 @@ HADAMARD_AC( 8, 8 )
  * pixel_sad_x4
  ****************************************************************************/
 #define SAD_X( size ) \
-static void x264_pixel_sad_x3_##size( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, int i_stride, int scores[3] )\
+static void x264_pixel_sad_x3_##size( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2,\
+                                      intptr_t i_stride, int scores[3] )\
 {\
     scores[0] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix0, i_stride );\
     scores[1] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix1, i_stride );\
     scores[2] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix2, i_stride );\
 }\
-static void x264_pixel_sad_x4_##size( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, pixel *pix3, int i_stride, int scores[4] )\
+static void x264_pixel_sad_x4_##size( pixel *fenc, pixel *pix0, pixel *pix1,pixel *pix2, pixel *pix3,\
+                                      intptr_t i_stride, int scores[4] )\
 {\
     scores[0] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix0, i_stride );\
     scores[1] = x264_pixel_sad_##size( fenc, FENC_STRIDE, pix1, i_stride );\
@@ -457,13 +468,15 @@ SAD_X( 8x8_vis )
  ****************************************************************************/
 
 #define SATD_X( size, cpu ) \
-static void x264_pixel_satd_x3_##size##cpu( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, int i_stride, int scores[3] )\
+static void x264_pixel_satd_x3_##size##cpu( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2,\
+                                            intptr_t i_stride, int scores[3] )\
 {\
     scores[0] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix0, i_stride );\
     scores[1] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix1, i_stride );\
     scores[2] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix2, i_stride );\
 }\
-static void x264_pixel_satd_x4_##size##cpu( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, pixel *pix3, int i_stride, int scores[4] )\
+static void x264_pixel_satd_x4_##size##cpu( pixel *fenc, pixel *pix0, pixel *pix1, pixel *pix2, pixel *pix3,\
+                                            intptr_t i_stride, int scores[4] )\
 {\
     scores[0] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix0, i_stride );\
     scores[1] = x264_pixel_satd_##size##cpu( fenc, FENC_STRIDE, pix1, i_stride );\
@@ -483,11 +496,14 @@ SATD_X( 4x4, cpu )
 
 SATD_X_DECL7()
 #if HAVE_MMX
-SATD_X_DECL7( _mmxext )
+SATD_X_DECL7( _mmx2 )
 #if !HIGH_BIT_DEPTH
 SATD_X_DECL6( _sse2 )
 SATD_X_DECL7( _ssse3 )
+SATD_X_DECL6( _ssse3_atom )
 SATD_X_DECL7( _sse4 )
+SATD_X_DECL7( _avx )
+SATD_X_DECL7( _xop )
 #endif // !HIGH_BIT_DEPTH
 #endif
 
@@ -497,64 +513,93 @@ SATD_X_DECL7( _neon )
 #endif
 #endif // !HIGH_BIT_DEPTH
 
-#define INTRA_MBCMP_8x8( mbcmp, cpu )\
-void x264_intra_##mbcmp##_x3_8x8##cpu( pixel *fenc, pixel edge[33], int res[3] )\
+#define INTRA_MBCMP_8x8( mbcmp, cpu, cpu2 )\
+void x264_intra_##mbcmp##_x3_8x8##cpu( pixel *fenc, pixel edge[36], int res[3] )\
 {\
     ALIGNED_ARRAY_16( pixel, pix, [8*FDEC_STRIDE] );\
-    x264_predict_8x8_v_c( pix, edge );\
+    x264_predict_8x8_v##cpu2( pix, edge );\
     res[0] = x264_pixel_##mbcmp##_8x8##cpu( pix, FDEC_STRIDE, fenc, FENC_STRIDE );\
-    x264_predict_8x8_h_c( pix, edge );\
+    x264_predict_8x8_h##cpu2( pix, edge );\
     res[1] = x264_pixel_##mbcmp##_8x8##cpu( pix, FDEC_STRIDE, fenc, FENC_STRIDE );\
-    x264_predict_8x8_dc_c( pix, edge );\
+    x264_predict_8x8_dc##cpu2( pix, edge );\
     res[2] = x264_pixel_##mbcmp##_8x8##cpu( pix, FDEC_STRIDE, fenc, FENC_STRIDE );\
 }
 
-INTRA_MBCMP_8x8( sad, )
-INTRA_MBCMP_8x8(sa8d, )
+INTRA_MBCMP_8x8( sad,, _c )
+INTRA_MBCMP_8x8(sa8d,, _c )
 #if HIGH_BIT_DEPTH && HAVE_MMX
-INTRA_MBCMP_8x8( sad, _mmxext)
-INTRA_MBCMP_8x8( sad, _sse2  )
-INTRA_MBCMP_8x8( sad, _ssse3 )
-INTRA_MBCMP_8x8(sa8d, _sse2  )
+#define x264_predict_8x8_v_sse2 x264_predict_8x8_v_sse
+INTRA_MBCMP_8x8( sad, _mmx2,  _c )
+INTRA_MBCMP_8x8(sa8d, _sse2,  _sse2 )
 #endif
 
-#define INTRA_MBCMP( mbcmp, size, pred1, pred2, pred3, chroma, cpu )\
-void x264_intra_##mbcmp##_x3_##size##x##size##chroma##cpu( pixel *fenc, pixel *fdec, int res[3] )\
+#define INTRA_MBCMP( mbcmp, size, pred1, pred2, pred3, chroma, cpu, cpu2 )\
+void x264_intra_##mbcmp##_x3_##size##chroma##cpu( pixel *fenc, pixel *fdec, int res[3] )\
 {\
-    x264_predict_##size##x##size##chroma##_##pred1##_c( fdec );\
-    res[0] = x264_pixel_##mbcmp##_##size##x##size##cpu( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
-    x264_predict_##size##x##size##chroma##_##pred2##_c( fdec );\
-    res[1] = x264_pixel_##mbcmp##_##size##x##size##cpu( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
-    x264_predict_##size##x##size##chroma##_##pred3##_c( fdec );\
-    res[2] = x264_pixel_##mbcmp##_##size##x##size##cpu( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
+    x264_predict_##size##chroma##_##pred1##cpu2( fdec );\
+    res[0] = x264_pixel_##mbcmp##_##size##cpu( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
+    x264_predict_##size##chroma##_##pred2##cpu2( fdec );\
+    res[1] = x264_pixel_##mbcmp##_##size##cpu( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
+    x264_predict_##size##chroma##_##pred3##cpu2( fdec );\
+    res[2] = x264_pixel_##mbcmp##_##size##cpu( fdec, FDEC_STRIDE, fenc, FENC_STRIDE );\
 }
 
-INTRA_MBCMP( sad,  4,  v, h, dc,  , )
-INTRA_MBCMP(satd,  4,  v, h, dc,  , )
-INTRA_MBCMP( sad,  8, dc, h,  v, c, )
-INTRA_MBCMP(satd,  8, dc, h,  v, c, )
-INTRA_MBCMP( sad, 16,  v, h, dc,  , )
-INTRA_MBCMP(satd, 16,  v, h, dc,  , )
+INTRA_MBCMP( sad,  4x4,   v, h, dc,  ,, _c )
+INTRA_MBCMP(satd,  4x4,   v, h, dc,  ,, _c )
+INTRA_MBCMP( sad,  8x8,  dc, h,  v, c,, _c )
+INTRA_MBCMP(satd,  8x8,  dc, h,  v, c,, _c )
+INTRA_MBCMP( sad,  8x16, dc, h,  v, c,, _c )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c,, _c )
+INTRA_MBCMP( sad, 16x16,  v, h, dc,  ,, _c )
+INTRA_MBCMP(satd, 16x16,  v, h, dc,  ,, _c )
 
-#if HIGH_BIT_DEPTH && HAVE_MMX
-INTRA_MBCMP( sad,  4,  v, h, dc,  , _mmxext)
-INTRA_MBCMP(satd,  4,  v, h, dc,  , _mmxext)
-INTRA_MBCMP( sad,  8, dc, h,  v, c, _mmxext)
-INTRA_MBCMP(satd,  8, dc, h,  v, c, _mmxext)
-INTRA_MBCMP( sad, 16,  v, h, dc,  , _mmxext)
-INTRA_MBCMP(satd, 16,  v, h, dc,  , _mmxext)
-INTRA_MBCMP( sad,  8, dc, h,  v, c, _sse2  )
-INTRA_MBCMP( sad, 16,  v, h, dc,  , _sse2  )
-INTRA_MBCMP( sad,  8, dc, h,  v, c, _ssse3 )
-INTRA_MBCMP( sad, 16,  v, h, dc,  , _ssse3 )
+#if HAVE_MMX
+#if HIGH_BIT_DEPTH
+#define x264_predict_8x8c_v_mmx2 x264_predict_8x8c_v_mmx
+#define x264_predict_8x16c_v_mmx2 x264_predict_8x16c_v_c
+#define x264_predict_8x8c_v_sse2 x264_predict_8x8c_v_sse
+#define x264_predict_8x16c_v_sse2 x264_predict_8x16c_v_sse
+#define x264_predict_16x16_v_sse2 x264_predict_16x16_v_sse
+INTRA_MBCMP( sad,  4x4,   v, h, dc,  , _mmx2, _c )
+INTRA_MBCMP( sad,  8x8,  dc, h,  v, c, _mmx2, _mmx2 )
+INTRA_MBCMP( sad,  8x16, dc, h,  v, c, _mmx2, _mmx2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _mmx2, _mmx2 )
+INTRA_MBCMP( sad, 16x16,  v, h, dc,  , _mmx2, _mmx2 )
+INTRA_MBCMP( sad,  8x8,  dc, h,  v, c, _sse2, _sse2 )
+INTRA_MBCMP( sad,  8x16, dc, h,  v, c, _sse2, _sse2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _sse2, _sse2 )
+INTRA_MBCMP( sad, 16x16,  v, h, dc,  , _sse2, _sse2 )
+INTRA_MBCMP( sad,  8x8,  dc, h,  v, c, _ssse3, _sse2 )
+INTRA_MBCMP( sad,  8x16, dc, h,  v, c, _ssse3, _sse2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _ssse3, _sse2 )
+INTRA_MBCMP( sad, 16x16,  v, h, dc,  , _ssse3, _sse2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _sse4, _sse2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _avx, _sse2 )
+#else
+#define x264_predict_8x16c_v_mmx2 x264_predict_8x16c_v_mmx
+INTRA_MBCMP( sad,  8x16, dc, h,  v, c, _mmx2, _mmx2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _mmx2, _mmx2 )
+INTRA_MBCMP( sad,  8x16, dc, h,  v, c, _sse2, _mmx2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _sse2, _mmx2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _ssse3, _mmx2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _sse4, _mmx2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _avx, _mmx2 )
+INTRA_MBCMP(satd,  8x16, dc, h,  v, c, _xop, _mmx2 )
 #endif
+#endif
+
+// No C implementation of intra_satd_x9. See checkasm for its behavior,
+// or see x264_mb_analyse_intra for the entirely different algorithm we
+// use when lacking an asm implementation of it.
+
+
 
 /****************************************************************************
  * structural similarity metric
  ****************************************************************************/
-static void ssim_4x4x2_core( const pixel *pix1, int stride1,
-                             const pixel *pix2, int stride2,
-                             int sums[2][4])
+static void ssim_4x4x2_core( const pixel *pix1, intptr_t stride1,
+                             const pixel *pix2, intptr_t stride2,
+                             int sums[2][4] )
 {
     for( int z = 0; z < 2; z++ )
     {
@@ -616,9 +661,9 @@ static float ssim_end4( int sum0[5][4], int sum1[5][4], int width )
 }
 
 float x264_pixel_ssim_wxh( x264_pixel_function_t *pf,
-                           pixel *pix1, int stride1,
-                           pixel *pix2, int stride2,
-                           int width, int height, void *buf )
+                           pixel *pix1, intptr_t stride1,
+                           pixel *pix2, intptr_t stride2,
+                           int width, int height, void *buf, int *cnt )
 {
     int z = 0;
     float ssim = 0.0;
@@ -637,9 +682,49 @@ float x264_pixel_ssim_wxh( x264_pixel_function_t *pf,
         for( int x = 0; x < width-1; x += 4 )
             ssim += pf->ssim_end4( sum0+x, sum1+x, X264_MIN(4,width-x-1) );
     }
+    *cnt = (height-1) * (width-1);
     return ssim;
 }
 
+static int pixel_vsad( pixel *src, intptr_t stride, int height )
+{
+    int score = 0;
+    for( int i = 1; i < height; i++, src += stride )
+        for( int j = 0; j < 16; j++ )
+            score += abs(src[j] - src[j+stride]);
+    return score;
+}
+
+int x264_field_vsad( x264_t *h, int mb_x, int mb_y )
+{
+    int score_field, score_frame;
+    int stride = h->fenc->i_stride[0];
+    int mb_stride = h->mb.i_mb_stride;
+    pixel *fenc = h->fenc->plane[0] + 16 * (mb_x + mb_y * stride);
+    int mb_xy = mb_x + mb_y*mb_stride;
+
+    /* We don't want to analyze pixels outside the frame, as it gives inaccurate results. */
+    int mbpair_height = X264_MIN( h->param.i_height - mb_y * 16, 32 );
+    score_frame  = h->pixf.vsad( fenc,          stride, mbpair_height );
+    score_field  = h->pixf.vsad( fenc,        stride*2, mbpair_height >> 1 );
+    score_field += h->pixf.vsad( fenc+stride, stride*2, mbpair_height >> 1 );
+
+    if( mb_x > 0 )
+        score_field += 512 - h->mb.field[mb_xy        -1]*1024;
+    if( mb_y > 0 )
+        score_field += 512 - h->mb.field[mb_xy-mb_stride]*1024;
+
+    return (score_field < score_frame);
+}
+
+static int pixel_asd8( pixel *pix1, intptr_t stride1, pixel *pix2, intptr_t stride2, int height )
+{
+    int sum = 0;
+    for( int y = 0; y < height; y++, pix1 += stride1, pix2 += stride2 )
+        for( int x = 0; x < 8; x++ )
+            sum += pix1[x] - pix2[x];
+    return abs( sum );
+}
 
 /****************************************************************************
  * successive elimination
@@ -714,23 +799,27 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
 #define INIT7_NAME( name1, name2, cpu ) \
     INIT6_NAME( name1, name2, cpu ) \
     pixf->name1[PIXEL_4x4]   = x264_pixel_##name2##_4x4##cpu;
+#define INIT8_NAME( name1, name2, cpu ) \
+    INIT7_NAME( name1, name2, cpu ) \
+    pixf->name1[PIXEL_4x16]  = x264_pixel_##name2##_4x16##cpu;
 #define INIT2( name, cpu ) INIT2_NAME( name, name, cpu )
 #define INIT4( name, cpu ) INIT4_NAME( name, name, cpu )
 #define INIT5( name, cpu ) INIT5_NAME( name, name, cpu )
 #define INIT6( name, cpu ) INIT6_NAME( name, name, cpu )
 #define INIT7( name, cpu ) INIT7_NAME( name, name, cpu )
+#define INIT8( name, cpu ) INIT8_NAME( name, name, cpu )
 
 #define INIT_ADS( cpu ) \
     pixf->ads[PIXEL_16x16] = x264_pixel_ads4##cpu;\
     pixf->ads[PIXEL_16x8] = x264_pixel_ads2##cpu;\
     pixf->ads[PIXEL_8x8] = x264_pixel_ads1##cpu;
 
-    INIT7( sad, );
-    INIT7_NAME( sad_aligned, sad, );
+    INIT8( sad, );
+    INIT8_NAME( sad_aligned, sad, );
     INIT7( sad_x3, );
     INIT7( sad_x4, );
-    INIT7( ssd, );
-    INIT7( satd, );
+    INIT8( ssd, );
+    INIT8( satd, );
     INIT7( satd_x3, );
     INIT7( satd_x4, );
     INIT4( hadamard_ac, );
@@ -739,12 +828,16 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
     pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16;
     pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8;
     pixf->var[PIXEL_16x16] = x264_pixel_var_16x16;
+    pixf->var[PIXEL_8x16]  = x264_pixel_var_8x16;
     pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8;
+    pixf->var2[PIXEL_8x16]  = x264_pixel_var2_8x16;
+    pixf->var2[PIXEL_8x8]   = x264_pixel_var2_8x8;
 
     pixf->ssd_nv12_core = pixel_ssd_nv12_core;
     pixf->ssim_4x4x2_core = ssim_4x4x2_core;
     pixf->ssim_end4 = ssim_end4;
-    pixf->var2_8x8 = pixel_var2_8x8;
+    pixf->vsad = pixel_vsad;
+    pixf->asd8 = pixel_asd8;
 
     pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4;
     pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4;
@@ -752,52 +845,66 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
     pixf->intra_sa8d_x3_8x8   = x264_intra_sa8d_x3_8x8;
     pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c;
     pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c;
+    pixf->intra_sad_x3_8x16c  = x264_intra_sad_x3_8x16c;
+    pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c;
     pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16;
     pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16;
 
 #if HIGH_BIT_DEPTH
 #if HAVE_MMX
-    if( cpu&X264_CPU_MMXEXT )
+    if( cpu&X264_CPU_MMX2 )
     {
-        INIT7( sad, _mmxext );
-        INIT7( sad_x3, _mmxext );
-        INIT7( sad_x4, _mmxext );
-        INIT7( satd, _mmxext );
-        INIT7( satd_x3, _mmxext );
-        INIT7( satd_x4, _mmxext );
-        INIT4( hadamard_ac, _mmxext );
-        INIT7( ssd, _mmxext );
-        INIT_ADS( _mmxext );
-
-        pixf->ssd_nv12_core = x264_pixel_ssd_nv12_core_mmxext;
-        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_mmxext;
-        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_mmxext;
-        pixf->var2_8x8 = x264_pixel_var2_8x8_mmxext;
-
-        pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4_mmxext;
-        pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_mmxext;
-        pixf->intra_sad_x3_8x8    = x264_intra_sad_x3_8x8_mmxext;
-        pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_mmxext;
-        pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_mmxext;
-        pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_mmxext;
-        pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_mmxext;
+        INIT7( sad, _mmx2 );
+        INIT7_NAME( sad_aligned, sad, _mmx2 );
+        INIT7( sad_x3, _mmx2 );
+        INIT7( sad_x4, _mmx2 );
+        INIT8( satd, _mmx2 );
+        INIT7( satd_x3, _mmx2 );
+        INIT7( satd_x4, _mmx2 );
+        INIT4( hadamard_ac, _mmx2 );
+        INIT8( ssd, _mmx2 );
+        INIT_ADS( _mmx2 );
+
+        pixf->ssd_nv12_core = x264_pixel_ssd_nv12_core_mmx2;
+        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_mmx2;
+        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_mmx2;
+#if ARCH_X86
+        pixf->var2[PIXEL_8x8]  = x264_pixel_var2_8x8_mmx2;
+        pixf->var2[PIXEL_8x16] = x264_pixel_var2_8x16_mmx2;
+#endif
+
+        pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4_mmx2;
+        pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_mmx2;
+        pixf->intra_sad_x3_8x8    = x264_intra_sad_x3_8x8_mmx2;
+        pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_mmx2;
+        pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_mmx2;
+        pixf->intra_sad_x3_8x16c  = x264_intra_sad_x3_8x16c_mmx2;
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_mmx2;
+        pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_mmx2;
+        pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_mmx2;
     }
     if( cpu&X264_CPU_SSE2 )
     {
         INIT4_NAME( sad_aligned, sad, _sse2_aligned );
         INIT5( ssd, _sse2 );
+        INIT6( satd, _sse2 );
+        pixf->satd[PIXEL_4x16] = x264_pixel_satd_4x16_sse2;
 
         pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_sse2;
         pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_sse2;
 #if ARCH_X86_64
         pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_sse2;
+        pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_sse2;
 #endif
+        pixf->intra_sad_x3_4x4  = x264_intra_sad_x3_4x4_sse2;
         pixf->ssd_nv12_core = x264_pixel_ssd_nv12_core_sse2;
         pixf->ssim_4x4x2_core  = x264_pixel_ssim_4x4x2_core_sse2;
         pixf->ssim_end4        = x264_pixel_ssim_end4_sse2;
         pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_sse2;
-        pixf->var[PIXEL_8x8] = x264_pixel_var_8x8_sse2;
-        pixf->var2_8x8 = x264_pixel_var2_8x8_sse2;
+        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_sse2;
+        pixf->var2[PIXEL_8x8]  = x264_pixel_var2_8x8_sse2;
+        pixf->var2[PIXEL_8x16] = x264_pixel_var2_8x16_sse2;
+        pixf->intra_sad_x3_8x8 = x264_intra_sad_x3_8x8_sse2;
     }
     if( (cpu&X264_CPU_SSE2) && !(cpu&X264_CPU_SSE2_IS_SLOW) )
     {
@@ -810,9 +917,12 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         {
             INIT4( hadamard_ac, _sse2 );
         }
-
+        pixf->vsad = x264_pixel_vsad_sse2;
+        pixf->asd8 = x264_pixel_asd8_sse2;
         pixf->intra_sad_x3_8x8    = x264_intra_sad_x3_8x8_sse2;
         pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_sse2;
+        pixf->intra_sad_x3_8x16c  = x264_intra_sad_x3_8x16c_sse2;
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_sse2;
         pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_sse2;
     }
     if( cpu&X264_CPU_SSE2_IS_FAST )
@@ -828,82 +938,153 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
     if( cpu&X264_CPU_SSSE3 )
     {
         INIT4_NAME( sad_aligned, sad, _ssse3_aligned );
+        pixf->sad_aligned[PIXEL_4x4] = x264_pixel_sad_4x4_ssse3;
+        pixf->sad_aligned[PIXEL_4x8] = x264_pixel_sad_4x8_ssse3;
         INIT7( sad, _ssse3 );
         INIT7( sad_x3, _ssse3 );
         INIT7( sad_x4, _ssse3 );
         INIT_ADS( _ssse3 );
+        INIT6( satd, _ssse3 );
+        pixf->satd[PIXEL_4x16] = x264_pixel_satd_4x16_ssse3;
 
         if( !(cpu&X264_CPU_STACK_MOD4) )
         {
             INIT4( hadamard_ac, _ssse3 );
         }
-
+        pixf->vsad = x264_pixel_vsad_ssse3;
+        pixf->asd8 = x264_pixel_asd8_ssse3;
+        pixf->intra_sad_x3_4x4  = x264_intra_sad_x3_4x4_ssse3;
         pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_ssse3;
         pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_ssse3;
+#if ARCH_X86_64
+        pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_ssse3;
+#endif
+        pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4_ssse3;
         pixf->intra_sad_x3_8x8    = x264_intra_sad_x3_8x8_ssse3;
         pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_ssse3;
+        pixf->intra_sad_x3_8x16c  = x264_intra_sad_x3_8x16c_ssse3;
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_ssse3;
         pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_ssse3;
     }
+    if( cpu&X264_CPU_SSE4 )
+    {
+        INIT6( satd, _sse4 );
+        pixf->satd[PIXEL_4x16] = x264_pixel_satd_4x16_sse4;
+        if( !(cpu&X264_CPU_STACK_MOD4) )
+        {
+            INIT4( hadamard_ac, _sse4 );
+        }
+        pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_sse4;
+        pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_sse4;
+#if ARCH_X86_64
+        pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_sse4;
+#endif
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_sse4;
+    }
+    if( cpu&X264_CPU_AVX )
+    {
+        INIT5_NAME( sad_aligned, sad, _ssse3 ); /* AVX-capable CPUs doesn't benefit from an aligned version */
+        INIT_ADS( _avx );
+        INIT6( satd, _avx );
+        pixf->satd[PIXEL_4x16] = x264_pixel_satd_4x16_avx;
+        if( !(cpu&X264_CPU_STACK_MOD4) )
+        {
+            INIT4( hadamard_ac, _avx );
+        }
+        pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4_avx;
+        pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_avx;
+        pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_avx;
+        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_avx;
+        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_avx;
+        pixf->ssd_nv12_core    = x264_pixel_ssd_nv12_core_avx;
+        pixf->ssim_4x4x2_core  = x264_pixel_ssim_4x4x2_core_avx;
+        pixf->ssim_end4        = x264_pixel_ssim_end4_avx;
+#if ARCH_X86_64
+        pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_avx;
+#endif
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_avx;
+    }
+    if( cpu&X264_CPU_XOP )
+    {
+        pixf->vsad = x264_pixel_vsad_xop;
+        pixf->asd8 = x264_pixel_asd8_xop;
+    }
+    if( cpu&X264_CPU_AVX2 )
+    {
+        INIT2( ssd, _avx2 );
+        INIT2( sad, _avx2 );
+        INIT2_NAME( sad_aligned, sad, _avx2 );
+        INIT2( sad_x3, _avx2 );
+        INIT2( sad_x4, _avx2 );
+        pixf->vsad = x264_pixel_vsad_avx2;
+        pixf->ssd_nv12_core = x264_pixel_ssd_nv12_core_avx2;
+        pixf->intra_sad_x3_8x8 = x264_intra_sad_x3_8x8_avx2;
+    }
 #endif // HAVE_MMX
 #else // !HIGH_BIT_DEPTH
 #if HAVE_MMX
     if( cpu&X264_CPU_MMX )
     {
-        INIT7( ssd, _mmx );
+        INIT8( ssd, _mmx );
     }
 
-    if( cpu&X264_CPU_MMXEXT )
+    if( cpu&X264_CPU_MMX2 )
     {
-        INIT7( sad, _mmxext );
-        INIT7_NAME( sad_aligned, sad, _mmxext );
-        INIT7( sad_x3, _mmxext );
-        INIT7( sad_x4, _mmxext );
-        INIT7( satd, _mmxext );
-        INIT7( satd_x3, _mmxext );
-        INIT7( satd_x4, _mmxext );
-        INIT4( hadamard_ac, _mmxext );
-        INIT_ADS( _mmxext );
-        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_mmxext;
-        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_mmxext;
-        pixf->ssd_nv12_core    = x264_pixel_ssd_nv12_core_mmxext;
+        INIT8( sad, _mmx2 );
+        INIT8_NAME( sad_aligned, sad, _mmx2 );
+        INIT7( sad_x3, _mmx2 );
+        INIT7( sad_x4, _mmx2 );
+        INIT8( satd, _mmx2 );
+        INIT7( satd_x3, _mmx2 );
+        INIT7( satd_x4, _mmx2 );
+        INIT4( hadamard_ac, _mmx2 );
+        INIT_ADS( _mmx2 );
+        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_mmx2;
+        pixf->var[PIXEL_8x16]  = x264_pixel_var_8x16_mmx2;
+        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_mmx2;
+        pixf->ssd_nv12_core    = x264_pixel_ssd_nv12_core_mmx2;
 #if ARCH_X86
-        pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_mmxext;
-        pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_mmxext;
-        pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_mmxext;
-        pixf->ssim_4x4x2_core  = x264_pixel_ssim_4x4x2_core_mmxext;
-        pixf->var2_8x8 = x264_pixel_var2_8x8_mmxext;
+        pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_mmx2;
+        pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_mmx2;
+        pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_mmx2;
+        pixf->ssim_4x4x2_core = x264_pixel_ssim_4x4x2_core_mmx2;
+        pixf->var2[PIXEL_8x8] = x264_pixel_var2_8x8_mmx2;
+        pixf->var2[PIXEL_8x16] = x264_pixel_var2_8x16_mmx2;
+        pixf->vsad = x264_pixel_vsad_mmx2;
 
         if( cpu&X264_CPU_CACHELINE_32 )
         {
-            INIT5( sad, _cache32_mmxext );
-            INIT4( sad_x3, _cache32_mmxext );
-            INIT4( sad_x4, _cache32_mmxext );
+            INIT5( sad, _cache32_mmx2 );
+            INIT4( sad_x3, _cache32_mmx2 );
+            INIT4( sad_x4, _cache32_mmx2 );
         }
-        else if( cpu&X264_CPU_CACHELINE_64 )
+        else if( cpu&X264_CPU_CACHELINE_64 && !(cpu&X264_CPU_SLOW_ATOM) )
         {
-            INIT5( sad, _cache64_mmxext );
-            INIT4( sad_x3, _cache64_mmxext );
-            INIT4( sad_x4, _cache64_mmxext );
+            INIT5( sad, _cache64_mmx2 );
+            INIT4( sad_x3, _cache64_mmx2 );
+            INIT4( sad_x4, _cache64_mmx2 );
         }
 #else
-        if( cpu&X264_CPU_CACHELINE_64 )
+        if( cpu&X264_CPU_CACHELINE_64 && !(cpu&X264_CPU_SLOW_ATOM) )
         {
-            pixf->sad[PIXEL_8x16] = x264_pixel_sad_8x16_cache64_mmxext;
-            pixf->sad[PIXEL_8x8]  = x264_pixel_sad_8x8_cache64_mmxext;
-            pixf->sad[PIXEL_8x4]  = x264_pixel_sad_8x4_cache64_mmxext;
-            pixf->sad_x3[PIXEL_8x16] = x264_pixel_sad_x3_8x16_cache64_mmxext;
-            pixf->sad_x3[PIXEL_8x8]  = x264_pixel_sad_x3_8x8_cache64_mmxext;
-            pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_cache64_mmxext;
-            pixf->sad_x4[PIXEL_8x8]  = x264_pixel_sad_x4_8x8_cache64_mmxext;
+            pixf->sad[PIXEL_8x16] = x264_pixel_sad_8x16_cache64_mmx2;
+            pixf->sad[PIXEL_8x8]  = x264_pixel_sad_8x8_cache64_mmx2;
+            pixf->sad[PIXEL_8x4]  = x264_pixel_sad_8x4_cache64_mmx2;
+            pixf->sad_x3[PIXEL_8x16] = x264_pixel_sad_x3_8x16_cache64_mmx2;
+            pixf->sad_x3[PIXEL_8x8]  = x264_pixel_sad_x3_8x8_cache64_mmx2;
+            pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_cache64_mmx2;
+            pixf->sad_x4[PIXEL_8x8]  = x264_pixel_sad_x4_8x8_cache64_mmx2;
         }
 #endif
-        pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_mmxext;
-        pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_mmxext;
-        pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_mmxext;
-        pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_mmxext;
-        pixf->intra_sad_x3_8x8    = x264_intra_sad_x3_8x8_mmxext;
-        pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_mmxext;
-        pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4_mmxext;
+        pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_mmx2;
+        pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_mmx2;
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_mmx2;
+        pixf->intra_sad_x3_8x16c  = x264_intra_sad_x3_8x16c_mmx2;
+        pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_mmx2;
+        pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_mmx2;
+        pixf->intra_sad_x3_8x8    = x264_intra_sad_x3_8x8_mmx2;
+        pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_mmx2;
+        pixf->intra_sad_x3_4x4    = x264_intra_sad_x3_4x4_mmx2;
     }
 
     if( cpu&X264_CPU_SSE2 )
@@ -918,8 +1099,12 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         pixf->sa8d[PIXEL_8x8]   = x264_pixel_sa8d_8x8_sse2;
 #if ARCH_X86_64
         pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_sse2;
+        pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_sse2;
 #endif
-        pixf->var2_8x8 = x264_pixel_var2_8x8_sse2;
+        pixf->var2[PIXEL_8x8]   = x264_pixel_var2_8x8_sse2;
+        pixf->var2[PIXEL_8x16]  = x264_pixel_var2_8x16_sse2;
+        pixf->vsad = x264_pixel_vsad_sse2;
+        pixf->asd8 = x264_pixel_asd8_sse2;
     }
 
     if( (cpu&X264_CPU_SSE2) && !(cpu&X264_CPU_SSE2_IS_SLOW) )
@@ -928,15 +1113,16 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         INIT2( sad_x3, _sse2 );
         INIT2( sad_x4, _sse2 );
         INIT6( satd, _sse2 );
+        pixf->satd[PIXEL_4x16]   = x264_pixel_satd_4x16_sse2;
         INIT6( satd_x3, _sse2 );
         INIT6( satd_x4, _sse2 );
-        if( !(cpu&X264_CPU_STACK_MOD4) )
-        {
-            INIT4( hadamard_ac, _sse2 );
-        }
+        INIT4( hadamard_ac, _sse2 );
         INIT_ADS( _sse2 );
         pixf->var[PIXEL_8x8] = x264_pixel_var_8x8_sse2;
+        pixf->var[PIXEL_8x16] = x264_pixel_var_8x16_sse2;
         pixf->intra_sad_x3_16x16 = x264_intra_sad_x3_16x16_sse2;
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_sse2;
+        pixf->intra_sad_x3_8x16c  = x264_intra_sad_x3_8x16c_sse2;
         if( cpu&X264_CPU_CACHELINE_64 )
         {
             INIT2( ssd, _sse2); /* faster for width 16 on p4 */
@@ -951,12 +1137,6 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
                pixf->sad_x4[PIXEL_8x16] = x264_pixel_sad_x4_8x16_cache64_sse2;
            }
         }
-
-        if( cpu&X264_CPU_SSE_MISALIGN )
-        {
-            INIT2( sad_x3, _sse2_misalign );
-            INIT2( sad_x4, _sse2_misalign );
-        }
     }
 
     if( cpu&X264_CPU_SSE2_IS_FAST && !(cpu&X264_CPU_CACHELINE_64) )
@@ -980,36 +1160,63 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
 
     if( cpu&X264_CPU_SSSE3 )
     {
+        INIT4( hadamard_ac, _ssse3 );
         if( !(cpu&X264_CPU_STACK_MOD4) )
         {
-            INIT4( hadamard_ac, _ssse3 );
+            pixf->intra_sad_x9_4x4  = x264_intra_sad_x9_4x4_ssse3;
+            pixf->intra_satd_x9_4x4 = x264_intra_satd_x9_4x4_ssse3;
+            pixf->intra_sad_x9_8x8  = x264_intra_sad_x9_8x8_ssse3;
+#if ARCH_X86_64
+            pixf->intra_sa8d_x9_8x8 = x264_intra_sa8d_x9_8x8_ssse3;
+#endif
         }
         INIT_ADS( _ssse3 );
-        if( !(cpu&X264_CPU_SLOW_ATOM) )
+        if( cpu&X264_CPU_SLOW_ATOM )
         {
-            INIT7( ssd, _ssse3 );
+            pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_ssse3_atom;
+            pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_ssse3_atom;
+            INIT6( satd, _ssse3_atom );
+            pixf->satd[PIXEL_4x16]  = x264_pixel_satd_4x16_ssse3_atom;
+            INIT6( satd_x3, _ssse3_atom );
+            INIT6( satd_x4, _ssse3_atom );
+            INIT4( hadamard_ac, _ssse3_atom );
+#if ARCH_X86_64
+            pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_ssse3_atom;
+#endif
+        }
+        else
+        {
+            INIT8( ssd, _ssse3 );
             pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_ssse3;
             pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_ssse3;
-            INIT7( satd, _ssse3 );
+            INIT8( satd, _ssse3 );
             INIT7( satd_x3, _ssse3 );
             INIT7( satd_x4, _ssse3 );
+#if ARCH_X86_64
+            pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_ssse3;
+#endif
         }
         pixf->intra_satd_x3_16x16 = x264_intra_satd_x3_16x16_ssse3;
-        pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_ssse3;
+        if( !(cpu&X264_CPU_SLOW_PSHUFB) )
+            pixf->intra_sad_x3_16x16  = x264_intra_sad_x3_16x16_ssse3;
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_ssse3;
         pixf->intra_satd_x3_8x8c  = x264_intra_satd_x3_8x8c_ssse3;
         pixf->intra_sad_x3_8x8c   = x264_intra_sad_x3_8x8c_ssse3;
-        pixf->intra_satd_x3_4x4   = x264_intra_satd_x3_4x4_ssse3;
-#if ARCH_X86_64
-        pixf->intra_sa8d_x3_8x8 = x264_intra_sa8d_x3_8x8_ssse3;
-#endif
-        pixf->var2_8x8 = x264_pixel_var2_8x8_ssse3;
+        pixf->var2[PIXEL_8x8] = x264_pixel_var2_8x8_ssse3;
+        pixf->var2[PIXEL_8x16] = x264_pixel_var2_8x16_ssse3;
+        pixf->asd8 = x264_pixel_asd8_ssse3;
         if( cpu&X264_CPU_CACHELINE_64 )
         {
             INIT2( sad, _cache64_ssse3 );
             INIT2( sad_x3, _cache64_ssse3 );
             INIT2( sad_x4, _cache64_ssse3 );
         }
-        if( cpu&X264_CPU_SLOW_ATOM || !(cpu&X264_CPU_SHUFFLE_IS_FAST) )
+        else
+        {
+            INIT2( sad_x3, _ssse3 );
+            INIT5( sad_x4, _ssse3 );
+        }
+        if( (cpu&X264_CPU_SLOW_ATOM) || (cpu&X264_CPU_SLOW_SHUFFLE) )
         {
             INIT5( ssd, _sse2 ); /* on conroe, sse2 is faster for width8/16 */
         }
@@ -1017,18 +1224,104 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
 
     if( cpu&X264_CPU_SSE4 )
     {
-        INIT7( satd, _sse4 );
+        INIT8( satd, _sse4 );
         INIT7( satd_x3, _sse4 );
         INIT7( satd_x4, _sse4 );
+        INIT4( hadamard_ac, _sse4 );
         if( !(cpu&X264_CPU_STACK_MOD4) )
         {
-            INIT4( hadamard_ac, _sse4 );
+            pixf->intra_sad_x9_4x4  = x264_intra_sad_x9_4x4_sse4;
+            pixf->intra_satd_x9_4x4 = x264_intra_satd_x9_4x4_sse4;
+            pixf->intra_sad_x9_8x8  = x264_intra_sad_x9_8x8_sse4;
+#if ARCH_X86_64
+            pixf->intra_sa8d_x9_8x8 = x264_intra_sa8d_x9_8x8_sse4;
+#endif
         }
         pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_sse4;
         pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_sse4;
-        pixf->intra_sad_x3_4x4 = x264_intra_sad_x3_4x4_sse4;
-        /* Slower on Conroe, so only enable under SSE4 */
-        pixf->intra_sad_x3_8x8  = x264_intra_sad_x3_8x8_ssse3;
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_sse4;
+#if ARCH_X86_64
+        pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_sse4;
+#endif
+    }
+
+    if( cpu&X264_CPU_AVX )
+    {
+        INIT2_NAME( sad_aligned, sad, _sse2 ); /* AVX-capable CPUs doesn't benefit from an aligned version */
+        INIT2( sad_x3, _avx );
+        INIT2( sad_x4, _avx );
+        INIT8( satd, _avx );
+        INIT7( satd_x3, _avx );
+        INIT7( satd_x4, _avx );
+        INIT_ADS( _avx );
+        INIT4( hadamard_ac, _avx );
+        if( !(cpu&X264_CPU_STACK_MOD4) )
+        {
+            pixf->intra_sad_x9_4x4  = x264_intra_sad_x9_4x4_avx;
+            pixf->intra_satd_x9_4x4 = x264_intra_satd_x9_4x4_avx;
+            pixf->intra_sad_x9_8x8  = x264_intra_sad_x9_8x8_avx;
+#if ARCH_X86_64
+            pixf->intra_sa8d_x9_8x8 = x264_intra_sa8d_x9_8x8_avx;
+#endif
+        }
+        INIT5( ssd, _avx );
+        pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_avx;
+        pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_avx;
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_avx;
+        pixf->ssd_nv12_core    = x264_pixel_ssd_nv12_core_avx;
+        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_avx;
+        pixf->var[PIXEL_8x16]  = x264_pixel_var_8x16_avx;
+        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_avx;
+        pixf->ssim_4x4x2_core  = x264_pixel_ssim_4x4x2_core_avx;
+        pixf->ssim_end4        = x264_pixel_ssim_end4_avx;
+#if ARCH_X86_64
+        pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_avx;
+#endif
+    }
+
+    if( cpu&X264_CPU_XOP )
+    {
+        INIT7( satd, _xop );
+        INIT7( satd_x3, _xop );
+        INIT7( satd_x4, _xop );
+        INIT4( hadamard_ac, _xop );
+        if( !(cpu&X264_CPU_STACK_MOD4) )
+        {
+            pixf->intra_satd_x9_4x4 = x264_intra_satd_x9_4x4_xop;
+        }
+        INIT5( ssd, _xop );
+        pixf->sa8d[PIXEL_16x16]= x264_pixel_sa8d_16x16_xop;
+        pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_xop;
+        pixf->intra_satd_x3_8x16c = x264_intra_satd_x3_8x16c_xop;
+        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_xop;
+        pixf->var[PIXEL_8x16]  = x264_pixel_var_8x16_xop;
+        pixf->var[PIXEL_8x8]   = x264_pixel_var_8x8_xop;
+        pixf->var2[PIXEL_8x8] = x264_pixel_var2_8x8_xop;
+        pixf->var2[PIXEL_8x16] = x264_pixel_var2_8x16_xop;
+#if ARCH_X86_64
+        pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_xop;
+#endif
+    }
+
+    if( cpu&X264_CPU_AVX2 )
+    {
+        INIT2( ssd, _avx2 );
+        INIT2( sad_x3, _avx2 );
+        INIT2( sad_x4, _avx2 );
+        INIT4( satd, _avx2 );
+        INIT2( hadamard_ac, _avx2 );
+        INIT_ADS( _avx2 );
+        pixf->sa8d[PIXEL_8x8]  = x264_pixel_sa8d_8x8_avx2;
+        pixf->var[PIXEL_16x16] = x264_pixel_var_16x16_avx2;
+        pixf->var2[PIXEL_8x16]  = x264_pixel_var2_8x16_avx2;
+        pixf->var2[PIXEL_8x8]   = x264_pixel_var2_8x8_avx2;
+        pixf->intra_sad_x3_16x16 = x264_intra_sad_x3_16x16_avx2;
+        pixf->intra_sad_x9_8x8  = x264_intra_sad_x9_8x8_avx2;
+        pixf->intra_sad_x3_8x8c = x264_intra_sad_x3_8x8c_avx2;
+        pixf->ssd_nv12_core = x264_pixel_ssd_nv12_core_avx2;
+#if ARCH_X86_64
+        pixf->sa8d_satd[PIXEL_16x16] = x264_pixel_sa8d_satd_16x16_avx2;
+#endif
     }
 #endif //HAVE_MMX
 
@@ -1055,7 +1348,7 @@ void x264_pixel_init( int cpu, x264_pixel_function_t *pixf )
         pixf->sa8d[PIXEL_16x16] = x264_pixel_sa8d_16x16_neon;
         pixf->var[PIXEL_8x8]    = x264_pixel_var_8x8_neon;
         pixf->var[PIXEL_16x16]  = x264_pixel_var_16x16_neon;
-        pixf->var2_8x8          = x264_pixel_var2_8x8_neon;
+        pixf->var2[PIXEL_8x8]   = x264_pixel_var2_8x8_neon;
 
         pixf->ssim_4x4x2_core   = x264_pixel_ssim_4x4x2_core_neon;
         pixf->ssim_end4         = x264_pixel_ssim_end4_neon;