Replace copy_memNxM functions with a generic copy/avg function.

Change-Id: I3ce849452ed4f08527de9565a9914d5ee36170aa

10 files changed, 221 insertions, 571 deletions

diff --git a/test/convolve_test.cc b/test/convolve_test.cc
index fd2bd36c1..6e5002fe9 100644
--- a/test/convolve_test.cc
+++ b/test/convolve_test.cc
@@ -22,8 +22,8 @@ extern "C" {
 }
 
 namespace {
-typedef void (*convolve_fn_t)(const uint8_t *src, int src_stride,
-                              uint8_t *dst, int dst_stride,
+typedef void (*convolve_fn_t)(const uint8_t *src, ptrdiff_t src_stride,
+                              uint8_t *dst, ptrdiff_t dst_stride,
                               const int16_t *filter_x, int filter_x_stride,
                               const int16_t *filter_y, int filter_y_stride,
                               int w, int h);
diff --git a/vp9/common/vp9_convolve.c b/vp9/common/vp9_convolve.c
index 914afa702..6f1e41866 100644
--- a/vp9/common/vp9_convolve.c
+++ b/vp9/common/vp9_convolve.c
@@ -38,8 +38,8 @@
  */
  #define ALIGN_FILTERS_256 1
 
-static void convolve_horiz_c(const uint8_t *src, int src_stride,
-                             uint8_t *dst, int dst_stride,
+static void convolve_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
+                             uint8_t *dst, ptrdiff_t dst_stride,
                              const int16_t *filter_x0, int x_step_q4,
                              const int16_t *filter_y, int y_step_q4,
                              int w, int h, int taps) {
@@ -80,8 +80,8 @@ static void convolve_horiz_c(const uint8_t *src, int src_stride,
   }
 }
 
-static void convolve_avg_horiz_c(const uint8_t *src, int src_stride,
-                                 uint8_t *dst, int dst_stride,
+static void convolve_avg_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
+                                 uint8_t *dst, ptrdiff_t dst_stride,
                                  const int16_t *filter_x0, int x_step_q4,
                                  const int16_t *filter_y, int y_step_q4,
                                  int w, int h, int taps) {
@@ -122,8 +122,8 @@ static void convolve_avg_horiz_c(const uint8_t *src, int src_stride,
   }
 }
 
-static void convolve_vert_c(const uint8_t *src, int src_stride,
-                            uint8_t *dst, int dst_stride,
+static void convolve_vert_c(const uint8_t *src, ptrdiff_t src_stride,
+                            uint8_t *dst, ptrdiff_t dst_stride,
                             const int16_t *filter_x, int x_step_q4,
                             const int16_t *filter_y0, int y_step_q4,
                             int w, int h, int taps) {
@@ -164,8 +164,8 @@ static void convolve_vert_c(const uint8_t *src, int src_stride,
   }
 }
 
-static void convolve_avg_vert_c(const uint8_t *src, int src_stride,
-                                uint8_t *dst, int dst_stride,
+static void convolve_avg_vert_c(const uint8_t *src, ptrdiff_t src_stride,
+                                uint8_t *dst, ptrdiff_t dst_stride,
                                 const int16_t *filter_x, int x_step_q4,
                                 const int16_t *filter_y0, int y_step_q4,
                                 int w, int h, int taps) {
@@ -207,8 +207,8 @@ static void convolve_avg_vert_c(const uint8_t *src, int src_stride,
   }
 }
 
-static void convolve_c(const uint8_t *src, int src_stride,
-                       uint8_t *dst, int dst_stride,
+static void convolve_c(const uint8_t *src, ptrdiff_t src_stride,
+                       uint8_t *dst, ptrdiff_t dst_stride,
                        const int16_t *filter_x, int x_step_q4,
                        const int16_t *filter_y, int y_step_q4,
                        int w, int h, int taps) {
@@ -237,8 +237,8 @@ static void convolve_c(const uint8_t *src, int src_stride,
                   w, h, taps);
 }
 
-static void convolve_avg_c(const uint8_t *src, int src_stride,
-                           uint8_t *dst, int dst_stride,
+static void convolve_avg_c(const uint8_t *src, ptrdiff_t src_stride,
+                           uint8_t *dst, ptrdiff_t dst_stride,
                            const int16_t *filter_x, int x_step_q4,
                            const int16_t *filter_y, int y_step_q4,
                            int w, int h, int taps) {
@@ -267,8 +267,8 @@ static void convolve_avg_c(const uint8_t *src, int src_stride,
                       w, h, taps);
 }
 
-void vp9_convolve8_horiz_c(const uint8_t *src, int src_stride,
-                           uint8_t *dst, int dst_stride,
+void vp9_convolve8_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
+                           uint8_t *dst, ptrdiff_t dst_stride,
                            const int16_t *filter_x, int x_step_q4,
                            const int16_t *filter_y, int y_step_q4,
                            int w, int h) {
@@ -277,8 +277,8 @@ void vp9_convolve8_horiz_c(const uint8_t *src, int src_stride,
                    w, h, 8);
 }
 
-void vp9_convolve8_avg_horiz_c(const uint8_t *src, int src_stride,
-                               uint8_t *dst, int dst_stride,
+void vp9_convolve8_avg_horiz_c(const uint8_t *src, ptrdiff_t src_stride,
+                               uint8_t *dst, ptrdiff_t dst_stride,
                                const int16_t *filter_x, int x_step_q4,
                                const int16_t *filter_y, int y_step_q4,
                                int w, int h) {
@@ -287,8 +287,8 @@ void vp9_convolve8_avg_horiz_c(const uint8_t *src, int src_stride,
                        w, h, 8);
 }
 
-void vp9_convolve8_vert_c(const uint8_t *src, int src_stride,
-                          uint8_t *dst, int dst_stride,
+void vp9_convolve8_vert_c(const uint8_t *src, ptrdiff_t src_stride,
+                          uint8_t *dst, ptrdiff_t dst_stride,
                           const int16_t *filter_x, int x_step_q4,
                           const int16_t *filter_y, int y_step_q4,
                           int w, int h) {
@@ -297,8 +297,8 @@ void vp9_convolve8_vert_c(const uint8_t *src, int src_stride,
                   w, h, 8);
 }
 
-void vp9_convolve8_avg_vert_c(const uint8_t *src, int src_stride,
-                              uint8_t *dst, int dst_stride,
+void vp9_convolve8_avg_vert_c(const uint8_t *src, ptrdiff_t src_stride,
+                              uint8_t *dst, ptrdiff_t dst_stride,
                               const int16_t *filter_x, int x_step_q4,
                               const int16_t *filter_y, int y_step_q4,
                               int w, int h) {
@@ -307,8 +307,8 @@ void vp9_convolve8_avg_vert_c(const uint8_t *src, int src_stride,
                       w, h, 8);
 }
 
-void vp9_convolve8_c(const uint8_t *src, int src_stride,
-                     uint8_t *dst, int dst_stride,
+void vp9_convolve8_c(const uint8_t *src, ptrdiff_t src_stride,
+                     uint8_t *dst, ptrdiff_t dst_stride,
                      const int16_t *filter_x, int x_step_q4,
                      const int16_t *filter_y, int y_step_q4,
                      int w, int h) {
@@ -317,8 +317,8 @@ void vp9_convolve8_c(const uint8_t *src, int src_stride,
              w, h, 8);
 }
 
-void vp9_convolve8_avg_c(const uint8_t *src, int src_stride,
-                         uint8_t *dst, int dst_stride,
+void vp9_convolve8_avg_c(const uint8_t *src, ptrdiff_t src_stride,
+                         uint8_t *dst, ptrdiff_t dst_stride,
                          const int16_t *filter_x, int x_step_q4,
                          const int16_t *filter_y, int y_step_q4,
                          int w, int h) {
@@ -339,33 +339,25 @@ void vp9_convolve8_avg_c(const uint8_t *src, int src_stride,
                    w, h);
 }
 
-void vp9_convolve_copy(const uint8_t *src, int src_stride,
-                       uint8_t *dst, int dst_stride,
-                       const int16_t *filter_x, int filter_x_stride,
-                       const int16_t *filter_y, int filter_y_stride,
-                       int w, int h) {
-  if (w == 16 && h == 16) {
-    vp9_copy_mem16x16(src, src_stride, dst, dst_stride);
-  } else if (w == 8 && h == 8) {
-    vp9_copy_mem8x8(src, src_stride, dst, dst_stride);
-  } else if (w == 8 && h == 4) {
-    vp9_copy_mem8x4(src, src_stride, dst, dst_stride);
-  } else {
-    int r;
-
-    for (r = h; r > 0; --r) {
-      memcpy(dst, src, w);
-      src += src_stride;
-      dst += dst_stride;
-    }
+void vp9_convolve_copy_c(const uint8_t *src, ptrdiff_t src_stride,
+                         uint8_t *dst, ptrdiff_t dst_stride,
+                         const int16_t *filter_x, int filter_x_stride,
+                         const int16_t *filter_y, int filter_y_stride,
+                         int w, int h) {
+  int r;
+
+  for (r = h; r > 0; --r) {
+    memcpy(dst, src, w);
+    src += src_stride;
+    dst += dst_stride;
   }
 }
 
-void vp9_convolve_avg(const uint8_t *src, int src_stride,
-                      uint8_t *dst, int dst_stride,
-                      const int16_t *filter_x, int filter_x_stride,
-                      const int16_t *filter_y, int filter_y_stride,
-                      int w, int h) {
+void vp9_convolve_avg_c(const uint8_t *src, ptrdiff_t src_stride,
+                        uint8_t *dst, ptrdiff_t dst_stride,
+                        const int16_t *filter_x, int filter_x_stride,
+                        const int16_t *filter_y, int filter_y_stride,
+                        int w, int h) {
   int x, y;
 
   for (y = 0; y < h; ++y) {
diff --git a/vp9/common/vp9_convolve.h b/vp9/common/vp9_convolve.h
index 0596080c0..3de81119f 100644
--- a/vp9/common/vp9_convolve.h
+++ b/vp9/common/vp9_convolve.h
@@ -13,26 +13,12 @@
 #include "./vpx_config.h"
 #include "vpx/vpx_integer.h"
 
-typedef void (*convolve_fn_t)(const uint8_t *src, int src_stride,
-                              uint8_t *dst, int dst_stride,
+typedef void (*convolve_fn_t)(const uint8_t *src, ptrdiff_t src_stride,
+                              uint8_t *dst, ptrdiff_t dst_stride,
                               const int16_t *filter_x, int x_step_q4,
                               const int16_t *filter_y, int y_step_q4,
                               int w, int h);
 
-// Not a convolution, a block copy conforming to the convolution prototype
-void vp9_convolve_copy(const uint8_t *src, int src_stride,
-                       uint8_t *dst, int dst_stride,
-                       const int16_t *filter_x, int x_step_q4,
-                       const int16_t *filter_y, int y_step_q4,
-                       int w, int h);
-
-// Not a convolution, a block average conforming to the convolution prototype
-void vp9_convolve_avg(const uint8_t *src, int src_stride,
-                      uint8_t *dst, int dst_stride,
-                      const int16_t *filter_x, int x_step_q4,
-                      const int16_t *filter_y, int y_step_q4,
-                      int w, int h);
-
 struct subpix_fn_table {
   const int16_t (*filter_x)[8];
   const int16_t (*filter_y)[8];
diff --git a/vp9/common/vp9_reconinter.c b/vp9/common/vp9_reconinter.c
index 265a19ab1..c29fd147e 100644
--- a/vp9/common/vp9_reconinter.c
+++ b/vp9/common/vp9_reconinter.c
@@ -194,93 +194,6 @@ void vp9_setup_interp_filters(MACROBLOCKD *xd,
   assert(((intptr_t)xd->subpix.filter_x & 0xff) == 0);
 }
 
-void vp9_copy_mem16x16_c(const uint8_t *src,
-                         int src_stride,
-                         uint8_t *dst,
-                         int dst_stride) {
-  int r;
-
-  for (r = 0; r < 16; r++) {
-#if !(CONFIG_FAST_UNALIGNED)
-    dst[0] = src[0];
-    dst[1] = src[1];
-    dst[2] = src[2];
-    dst[3] = src[3];
-    dst[4] = src[4];
-    dst[5] = src[5];
-    dst[6] = src[6];
-    dst[7] = src[7];
-    dst[8] = src[8];
-    dst[9] = src[9];
-    dst[10] = src[10];
-    dst[11] = src[11];
-    dst[12] = src[12];
-    dst[13] = src[13];
-    dst[14] = src[14];
-    dst[15] = src[15];
-
-#else
-    ((uint32_t *)dst)[0] = ((const uint32_t *)src)[0];
-    ((uint32_t *)dst)[1] = ((const uint32_t *)src)[1];
-    ((uint32_t *)dst)[2] = ((const uint32_t *)src)[2];
-    ((uint32_t *)dst)[3] = ((const uint32_t *)src)[3];
-
-#endif
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-void vp9_copy_mem8x8_c(const uint8_t *src,
-                       int src_stride,
-                       uint8_t *dst,
-                       int dst_stride) {
-  int r;
-
-  for (r = 0; r < 8; r++) {
-#if !(CONFIG_FAST_UNALIGNED)
-    dst[0] = src[0];
-    dst[1] = src[1];
-    dst[2] = src[2];
-    dst[3] = src[3];
-    dst[4] = src[4];
-    dst[5] = src[5];
-    dst[6] = src[6];
-    dst[7] = src[7];
-#else
-    ((uint32_t *)dst)[0] = ((const uint32_t *)src)[0];
-    ((uint32_t *)dst)[1] = ((const uint32_t *)src)[1];
-#endif
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-void vp9_copy_mem8x4_c(const uint8_t *src,
-                       int src_stride,
-                       uint8_t *dst,
-                       int dst_stride) {
-  int r;
-
-  for (r = 0; r < 4; r++) {
-#if !(CONFIG_FAST_UNALIGNED)
-    dst[0] = src[0];
-    dst[1] = src[1];
-    dst[2] = src[2];
-    dst[3] = src[3];
-    dst[4] = src[4];
-    dst[5] = src[5];
-    dst[6] = src[6];
-    dst[7] = src[7];
-#else
-    ((uint32_t *)dst)[0] = ((const uint32_t *)src)[0];
-    ((uint32_t *)dst)[1] = ((const uint32_t *)src)[1];
-#endif
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
 void vp9_build_inter_predictor(const uint8_t *src, int src_stride,
                                uint8_t *dst, int dst_stride,
                                const int_mv *src_mv,
diff --git a/vp9/common/vp9_rtcd_defs.sh b/vp9/common/vp9_rtcd_defs.sh
index d861a7a5e..459336e4d 100644
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -43,17 +43,6 @@ specialize vp9_idct_add_32x32
 #
 # RECON
 #
-prototype void vp9_copy_mem16x16 "const uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
-specialize vp9_copy_mem16x16 mmx sse2 dspr2
-vp9_copy_mem16x16_dspr2=vp9_copy_mem16x16_dspr2
-
-prototype void vp9_copy_mem8x8 "const uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
-specialize vp9_copy_mem8x8 mmx dspr2
-vp9_copy_mem8x8_dspr2=vp9_copy_mem8x8_dspr2
-
-prototype void vp9_copy_mem8x4 "const uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
-specialize vp9_copy_mem8x4 mmx
-
 prototype void vp9_d27_predictor_4x4 "uint8_t *ypred_ptr, ptrdiff_t y_stride, uint8_t *yabove_row, uint8_t *yleft_col"
 specialize vp9_d27_predictor_4x4
 
@@ -275,22 +264,28 @@ specialize vp9_blend_b
 #
 # Sub Pixel Filters
 #
-prototype void vp9_convolve8 "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
+prototype void vp9_convolve_copy "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
+specialize vp9_convolve_copy sse2
+
+prototype void vp9_convolve_avg "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
+specialize vp9_convolve_avg sse2
+
+prototype void vp9_convolve8 "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
 specialize vp9_convolve8 ssse3
 
-prototype void vp9_convolve8_horiz "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
+prototype void vp9_convolve8_horiz "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
 specialize vp9_convolve8_horiz ssse3
 
-prototype void vp9_convolve8_vert "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
+prototype void vp9_convolve8_vert "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
 specialize vp9_convolve8_vert ssse3
 
-prototype void vp9_convolve8_avg "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
+prototype void vp9_convolve8_avg "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
 specialize vp9_convolve8_avg ssse3
 
-prototype void vp9_convolve8_avg_horiz "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
+prototype void vp9_convolve8_avg_horiz "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
 specialize vp9_convolve8_avg_horiz ssse3
 
-prototype void vp9_convolve8_avg_vert "const uint8_t *src, int src_stride, uint8_t *dst, int dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
+prototype void vp9_convolve8_avg_vert "const uint8_t *src, ptrdiff_t src_stride, uint8_t *dst, ptrdiff_t dst_stride, const int16_t *filter_x, int x_step_q4, const int16_t *filter_y, int y_step_q4, int w, int h"
 specialize vp9_convolve8_avg_vert ssse3
 
 #
diff --git a/vp9/common/x86/vp9_asm_stubs.c b/vp9/common/x86/vp9_asm_stubs.c
index 2b66834a7..98fc4dc83 100644
--- a/vp9/common/x86/vp9_asm_stubs.c
+++ b/vp9/common/x86/vp9_asm_stubs.c
@@ -121,8 +121,8 @@ void vp9_filter_block1d4_h8_avg_ssse3(const unsigned char *src_ptr,
                                      unsigned int output_height,
                                      const short *filter);
 
-void vp9_convolve8_horiz_ssse3(const uint8_t *src, int src_stride,
-                               uint8_t *dst, int dst_stride,
+void vp9_convolve8_horiz_ssse3(const uint8_t *src, ptrdiff_t src_stride,
+                               uint8_t *dst, ptrdiff_t dst_stride,
                                const int16_t *filter_x, int x_step_q4,
                                const int16_t *filter_y, int y_step_q4,
                                int w, int h) {
@@ -159,8 +159,8 @@ void vp9_convolve8_horiz_ssse3(const uint8_t *src, int src_stride,
   }
 }
 
-void vp9_convolve8_vert_ssse3(const uint8_t *src, int src_stride,
-                              uint8_t *dst, int dst_stride,
+void vp9_convolve8_vert_ssse3(const uint8_t *src, ptrdiff_t src_stride,
+                              uint8_t *dst, ptrdiff_t dst_stride,
                               const int16_t *filter_x, int x_step_q4,
                               const int16_t *filter_y, int y_step_q4,
                               int w, int h) {
@@ -197,8 +197,8 @@ void vp9_convolve8_vert_ssse3(const uint8_t *src, int src_stride,
   }
 }
 
-void vp9_convolve8_avg_horiz_ssse3(const uint8_t *src, int src_stride,
-                               uint8_t *dst, int dst_stride,
+void vp9_convolve8_avg_horiz_ssse3(const uint8_t *src, ptrdiff_t src_stride,
+                               uint8_t *dst, ptrdiff_t dst_stride,
                                const int16_t *filter_x, int x_step_q4,
                                const int16_t *filter_y, int y_step_q4,
                                int w, int h) {
@@ -235,8 +235,8 @@ void vp9_convolve8_avg_horiz_ssse3(const uint8_t *src, int src_stride,
   }
 }
 
-void vp9_convolve8_avg_vert_ssse3(const uint8_t *src, int src_stride,
-                              uint8_t *dst, int dst_stride,
+void vp9_convolve8_avg_vert_ssse3(const uint8_t *src, ptrdiff_t src_stride,
+                              uint8_t *dst, ptrdiff_t dst_stride,
                               const int16_t *filter_x, int x_step_q4,
                               const int16_t *filter_y, int y_step_q4,
                               int w, int h) {
@@ -273,8 +273,8 @@ void vp9_convolve8_avg_vert_ssse3(const uint8_t *src, int src_stride,
   }
 }
 
-void vp9_convolve8_ssse3(const uint8_t *src, int src_stride,
-                         uint8_t *dst, int dst_stride,
+void vp9_convolve8_ssse3(const uint8_t *src, ptrdiff_t src_stride,
+                         uint8_t *dst, ptrdiff_t dst_stride,
                          const int16_t *filter_x, int x_step_q4,
                          const int16_t *filter_y, int y_step_q4,
                          int w, int h) {
@@ -294,8 +294,8 @@ void vp9_convolve8_ssse3(const uint8_t *src, int src_stride,
   }
 }
 
-void vp9_convolve8_avg_ssse3(const uint8_t *src, int src_stride,
-                         uint8_t *dst, int dst_stride,
+void vp9_convolve8_avg_ssse3(const uint8_t *src, ptrdiff_t src_stride,
+                         uint8_t *dst, ptrdiff_t dst_stride,
                          const int16_t *filter_x, int x_step_q4,
                          const int16_t *filter_y, int y_step_q4,
                          int w, int h) {
diff --git a/vp9/common/x86/vp9_copy_sse2.asm b/vp9/common/x86/vp9_copy_sse2.asm
new file mode 100644
index 000000000..dd522c698
--- /dev/null
+++ b/vp9/common/x86/vp9_copy_sse2.asm
@@ -0,0 +1,152 @@
+;
+;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+;
+;  Use of this source code is governed by a BSD-style license
+;  that can be found in the LICENSE file in the root of the source
+;  tree. An additional intellectual property rights grant can be found
+;  in the file PATENTS.  All contributing project authors may
+;  be found in the AUTHORS file in the root of the source tree.
+;
+
+%include "third_party/x86inc/x86inc.asm"
+
+SECTION .text
+
+%macro convolve_fn 1
+INIT_XMM sse2
+cglobal convolve_%1, 4, 7, 4, src, src_stride, dst, dst_stride, \
+                              fx, fxs, fy, fys, w, h
+  mov r4d, dword wm
+  cmp r4d, 4
+  je .w4
+  cmp r4d, 8
+  je .w8
+  cmp r4d, 16
+  je .w16
+  cmp r4d, 32
+  je .w32
+
+  mov                    r4d, dword hm
+.loop64:
+  movu                    m0, [srcq]
+  movu                    m1, [srcq+16]
+  movu                    m2, [srcq+32]
+  movu                    m3, [srcq+48]
+  add                   srcq, src_strideq
+%ifidn %1, avg
+  pavgb                   m0, [dstq]
+  pavgb                   m1, [dstq+16]
+  pavgb                   m2, [dstq+32]
+  pavgb                   m3, [dstq+48]
+%endif
+  mova             [dstq   ], m0
+  mova             [dstq+16], m1
+  mova             [dstq+32], m2
+  mova             [dstq+48], m3
+  add                   dstq, dst_strideq
+  dec                    r4d
+  jnz .loop64
+  RET
+
+.w32:
+  mov                    r4d, dword hm
+.loop32:
+  movu                    m0, [srcq]
+  movu                    m1, [srcq+16]
+  movu                    m2, [srcq+src_strideq]
+  movu                    m3, [srcq+src_strideq+16]
+  lea                   srcq, [srcq+src_strideq*2]
+%ifidn %1, avg
+  pavgb                   m0, [dstq]
+  pavgb                   m1, [dstq            +16]
+  pavgb                   m2, [dstq+dst_strideq]
+  pavgb                   m3, [dstq+dst_strideq+16]
+%endif
+  mova [dstq               ], m0
+  mova [dstq            +16], m1
+  mova [dstq+dst_strideq   ], m2
+  mova [dstq+dst_strideq+16], m3
+  lea                   dstq, [dstq+dst_strideq*2]
+  sub                    r4d, 2
+  jnz .loop32
+  RET
+
+.w16:
+  mov                    r4d, dword hm
+  lea                    r5q, [src_strideq*3]
+  lea                    r6q, [dst_strideq*3]
+.loop16:
+  movu                    m0, [srcq]
+  movu                    m1, [srcq+src_strideq]
+  movu                    m2, [srcq+src_strideq*2]
+  movu                    m3, [srcq+r5q]
+  lea                   srcq, [srcq+src_strideq*4]
+%ifidn %1, avg
+  pavgb                   m0, [dstq]
+  pavgb                   m1, [dstq+dst_strideq]
+  pavgb                   m2, [dstq+dst_strideq*2]
+  pavgb                   m3, [dstq+r6q]
+%endif
+  mova  [dstq              ], m0
+  mova  [dstq+dst_strideq  ], m1
+  mova  [dstq+dst_strideq*2], m2
+  mova  [dstq+r6q          ], m3
+  lea                   dstq, [dstq+dst_strideq*4]
+  sub                    r4d, 4
+  jnz .loop16
+  RET
+
+INIT_MMX sse
+.w8:
+  mov                    r4d, dword hm
+  lea                    r5q, [src_strideq*3]
+  lea                    r6q, [dst_strideq*3]
+.loop8:
+  movu                    m0, [srcq]
+  movu                    m1, [srcq+src_strideq]
+  movu                    m2, [srcq+src_strideq*2]
+  movu                    m3, [srcq+r5q]
+  lea                   srcq, [srcq+src_strideq*4]
+%ifidn %1, avg
+  pavgb                   m0, [dstq]
+  pavgb                   m1, [dstq+dst_strideq]
+  pavgb                   m2, [dstq+dst_strideq*2]
+  pavgb                   m3, [dstq+r6q]
+%endif
+  mova  [dstq              ], m0
+  mova  [dstq+dst_strideq  ], m1
+  mova  [dstq+dst_strideq*2], m2
+  mova  [dstq+r6q          ], m3
+  lea                   dstq, [dstq+dst_strideq*4]
+  sub                    r4d, 4
+  jnz .loop8
+  RET
+
+.w4:
+  mov                    r4d, dword hm
+  lea                    r5q, [src_strideq*3]
+  lea                    r6q, [dst_strideq*3]
+.loop4:
+  movh                    m0, [srcq]
+  movh                    m1, [srcq+src_strideq]
+  movh                    m2, [srcq+src_strideq*2]
+  movh                    m3, [srcq+r5q]
+  lea                   srcq, [srcq+src_strideq*4]
+%ifidn %1, avg
+  pavgb                   m0, [dstq]
+  pavgb                   m1, [dstq+dst_strideq]
+  pavgb                   m2, [dstq+dst_strideq*2]
+  pavgb                   m3, [dstq+r6q]
+%endif
+  movh  [dstq              ], m0
+  movh  [dstq+dst_strideq  ], m1
+  movh  [dstq+dst_strideq*2], m2
+  movh  [dstq+r6q          ], m3
+  lea                   dstq, [dstq+dst_strideq*4]
+  sub                    r4d, 4
+  jnz .loop4
+  RET
+%endmacro
+
+convolve_fn copy
+convolve_fn avg
diff --git a/vp9/common/x86/vp9_recon_mmx.asm b/vp9/common/x86/vp9_recon_mmx.asm
deleted file mode 100644
index 6fbbe48cb..000000000
--- a/vp9/common/x86/vp9_recon_mmx.asm
+++ /dev/null
@@ -1,272 +0,0 @@
-;
-;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-;void copy_mem8x8_mmx(
-;    unsigned char *src,
-;    int src_stride,
-;    unsigned char *dst,
-;    int dst_stride
-;    )
-global sym(vp9_copy_mem8x8_mmx) PRIVATE
-sym(vp9_copy_mem8x8_mmx):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rsi,        arg(0) ;src;
-        movq        mm0,        [rsi]
-
-        movsxd      rax,        dword ptr arg(1) ;src_stride;
-        mov         rdi,        arg(2) ;dst;
-
-        movq        mm1,        [rsi+rax]
-        movq        mm2,        [rsi+rax*2]
-
-        movsxd      rcx,        dword ptr arg(3) ;dst_stride
-        lea         rsi,        [rsi+rax*2]
-
-        movq        [rdi],      mm0
-        add         rsi,        rax
-
-        movq        [rdi+rcx],      mm1
-        movq        [rdi+rcx*2],    mm2
-
-
-        lea         rdi,        [rdi+rcx*2]
-        movq        mm3,        [rsi]
-
-        add         rdi,        rcx
-        movq        mm4,        [rsi+rax]
-
-        movq        mm5,        [rsi+rax*2]
-        movq        [rdi],      mm3
-
-        lea         rsi,        [rsi+rax*2]
-        movq        [rdi+rcx],  mm4
-
-        movq        [rdi+rcx*2],    mm5
-        lea         rdi,        [rdi+rcx*2]
-
-        movq        mm0,        [rsi+rax]
-        movq        mm1,        [rsi+rax*2]
-
-        movq        [rdi+rcx],  mm0
-        movq        [rdi+rcx*2],mm1
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void copy_mem8x4_mmx(
-;    unsigned char *src,
-;    int src_stride,
-;    unsigned char *dst,
-;    int dst_stride
-;    )
-global sym(vp9_copy_mem8x4_mmx) PRIVATE
-sym(vp9_copy_mem8x4_mmx):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rsi,        arg(0) ;src;
-        movq        mm0,        [rsi]
-
-        movsxd      rax,        dword ptr arg(1) ;src_stride;
-        mov         rdi,        arg(2) ;dst;
-
-        movq        mm1,        [rsi+rax]
-        movq        mm2,        [rsi+rax*2]
-
-        movsxd      rcx,        dword ptr arg(3) ;dst_stride
-        lea         rsi,        [rsi+rax*2]
-
-        movq        [rdi],      mm0
-        movq        [rdi+rcx],      mm1
-
-        movq        [rdi+rcx*2],    mm2
-        lea         rdi,        [rdi+rcx*2]
-
-        movq        mm3,        [rsi+rax]
-        movq        [rdi+rcx],      mm3
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void copy_mem16x16_mmx(
-;    unsigned char *src,
-;    int src_stride,
-;    unsigned char *dst,
-;    int dst_stride
-;    )
-global sym(vp9_copy_mem16x16_mmx) PRIVATE
-sym(vp9_copy_mem16x16_mmx):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rsi,        arg(0) ;src;
-        movsxd      rax,        dword ptr arg(1) ;src_stride;
-
-        mov         rdi,        arg(2) ;dst;
-        movsxd      rcx,        dword ptr arg(3) ;dst_stride
-
-        movq        mm0,            [rsi]
-        movq        mm3,            [rsi+8];
-
-        movq        mm1,            [rsi+rax]
-        movq        mm4,            [rsi+rax+8]
-
-        movq        mm2,            [rsi+rax*2]
-        movq        mm5,            [rsi+rax*2+8]
-
-        lea         rsi,            [rsi+rax*2]
-        add         rsi,            rax
-
-        movq        [rdi],          mm0
-        movq        [rdi+8],        mm3
-
-        movq        [rdi+rcx],      mm1
-        movq        [rdi+rcx+8],    mm4
-
-        movq        [rdi+rcx*2],    mm2
-        movq        [rdi+rcx*2+8],  mm5
-
-        lea         rdi,            [rdi+rcx*2]
-        add         rdi,            rcx
-
-        movq        mm0,            [rsi]
-        movq        mm3,            [rsi+8];
-
-        movq        mm1,            [rsi+rax]
-        movq        mm4,            [rsi+rax+8]
-
-        movq        mm2,            [rsi+rax*2]
-        movq        mm5,            [rsi+rax*2+8]
-
-        lea         rsi,            [rsi+rax*2]
-        add         rsi,            rax
-
-        movq        [rdi],          mm0
-        movq        [rdi+8],        mm3
-
-        movq        [rdi+rcx],      mm1
-        movq        [rdi+rcx+8],    mm4
-
-        movq        [rdi+rcx*2],    mm2
-        movq        [rdi+rcx*2+8],  mm5
-
-        lea         rdi,            [rdi+rcx*2]
-        add         rdi,            rcx
-
-        movq        mm0,            [rsi]
-        movq        mm3,            [rsi+8];
-
-        movq        mm1,            [rsi+rax]
-        movq        mm4,            [rsi+rax+8]
-
-        movq        mm2,            [rsi+rax*2]
-        movq        mm5,            [rsi+rax*2+8]
-
-        lea         rsi,            [rsi+rax*2]
-        add         rsi,            rax
-
-        movq        [rdi],          mm0
-        movq        [rdi+8],        mm3
-
-        movq        [rdi+rcx],      mm1
-        movq        [rdi+rcx+8],    mm4
-
-        movq        [rdi+rcx*2],    mm2
-        movq        [rdi+rcx*2+8],  mm5
-
-        lea         rdi,            [rdi+rcx*2]
-        add         rdi,            rcx
-
-        movq        mm0,            [rsi]
-        movq        mm3,            [rsi+8];
-
-        movq        mm1,            [rsi+rax]
-        movq        mm4,            [rsi+rax+8]
-
-        movq        mm2,            [rsi+rax*2]
-        movq        mm5,            [rsi+rax*2+8]
-
-        lea         rsi,            [rsi+rax*2]
-        add         rsi,            rax
-
-        movq        [rdi],          mm0
-        movq        [rdi+8],        mm3
-
-        movq        [rdi+rcx],      mm1
-        movq        [rdi+rcx+8],    mm4
-
-        movq        [rdi+rcx*2],    mm2
-        movq        [rdi+rcx*2+8],  mm5
-
-        lea         rdi,            [rdi+rcx*2]
-        add         rdi,            rcx
-
-        movq        mm0,            [rsi]
-        movq        mm3,            [rsi+8];
-
-        movq        mm1,            [rsi+rax]
-        movq        mm4,            [rsi+rax+8]
-
-        movq        mm2,            [rsi+rax*2]
-        movq        mm5,            [rsi+rax*2+8]
-
-        lea         rsi,            [rsi+rax*2]
-        add         rsi,            rax
-
-        movq        [rdi],          mm0
-        movq        [rdi+8],        mm3
-
-        movq        [rdi+rcx],      mm1
-        movq        [rdi+rcx+8],    mm4
-
-        movq        [rdi+rcx*2],    mm2
-        movq        [rdi+rcx*2+8],  mm5
-
-        lea         rdi,            [rdi+rcx*2]
-        add         rdi,            rcx
-
-        movq        mm0,            [rsi]
-        movq        mm3,            [rsi+8];
-
-        movq        [rdi],          mm0
-        movq        [rdi+8],        mm3
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
diff --git a/vp9/common/x86/vp9_recon_sse2.asm b/vp9/common/x86/vp9_recon_sse2.asm
deleted file mode 100644
index f7cc611f4..000000000
--- a/vp9/common/x86/vp9_recon_sse2.asm
+++ /dev/null
@@ -1,115 +0,0 @@
-;
-;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-%include "vpx_ports/x86_abi_support.asm"
-;void copy_mem16x16_sse2(
-;    unsigned char *src,
-;    int src_stride,
-;    unsigned char *dst,
-;    int dst_stride
-;    )
-global sym(vp9_copy_mem16x16_sse2) PRIVATE
-sym(vp9_copy_mem16x16_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 4
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rsi,        arg(0) ;src;
-        movdqu      xmm0,       [rsi]
-
-        movsxd      rax,        dword ptr arg(1) ;src_stride;
-        mov         rdi,        arg(2) ;dst;
-
-        movdqu      xmm1,       [rsi+rax]
-        movdqu      xmm2,       [rsi+rax*2]
-
-        movsxd      rcx,        dword ptr arg(3) ;dst_stride
-        lea         rsi,        [rsi+rax*2]
-
-        movdqa      [rdi],      xmm0
-        add         rsi,        rax
-
-        movdqa      [rdi+rcx],  xmm1
-        movdqa      [rdi+rcx*2],xmm2
-
-        lea         rdi,        [rdi+rcx*2]
-        movdqu      xmm3,       [rsi]
-
-        add         rdi,        rcx
-        movdqu      xmm4,       [rsi+rax]
-
-        movdqu      xmm5,       [rsi+rax*2]
-        lea         rsi,        [rsi+rax*2]
-
-        movdqa      [rdi],  xmm3
-        add         rsi,        rax
-
-        movdqa      [rdi+rcx],  xmm4
-        movdqa      [rdi+rcx*2],xmm5
-
-        lea         rdi,        [rdi+rcx*2]
-        movdqu      xmm0,       [rsi]
-
-        add         rdi,        rcx
-        movdqu      xmm1,       [rsi+rax]
-
-        movdqu      xmm2,       [rsi+rax*2]
-        lea         rsi,        [rsi+rax*2]
-
-        movdqa      [rdi],      xmm0
-        add         rsi,        rax
-
-        movdqa      [rdi+rcx],  xmm1
-
-        movdqa      [rdi+rcx*2],    xmm2
-        movdqu      xmm3,       [rsi]
-
-        movdqu      xmm4,       [rsi+rax]
-        lea         rdi,        [rdi+rcx*2]
-
-        add         rdi,        rcx
-        movdqu      xmm5,       [rsi+rax*2]
-
-        lea         rsi,        [rsi+rax*2]
-        movdqa      [rdi],  xmm3
-
-        add         rsi,        rax
-        movdqa      [rdi+rcx],  xmm4
-
-        movdqa      [rdi+rcx*2],xmm5
-        movdqu      xmm0,       [rsi]
-
-        lea         rdi,        [rdi+rcx*2]
-        movdqu      xmm1,       [rsi+rax]
-
-        add         rdi,        rcx
-        movdqu      xmm2,       [rsi+rax*2]
-
-        lea         rsi,        [rsi+rax*2]
-        movdqa      [rdi],      xmm0
-
-        movdqa      [rdi+rcx],  xmm1
-        movdqa      [rdi+rcx*2],xmm2
-
-        movdqu      xmm3,       [rsi+rax]
-        lea         rdi,        [rdi+rcx*2]
-
-        movdqa      [rdi+rcx],  xmm3
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
diff --git a/vp9/vp9_common.mk b/vp9/vp9_common.mk
index 1079e2048..ee744d541 100644
--- a/vp9/vp9_common.mk
+++ b/vp9/vp9_common.mk
@@ -75,10 +75,9 @@ VP9_COMMON_SRCS-$(ARCH_X86)$(ARCH_X86_64) += common/x86/vp9_asm_stubs.c
 VP9_COMMON_SRCS-$(ARCH_X86)$(ARCH_X86_64) += common/x86/vp9_loopfilter_intrin_sse2.c
 VP9_COMMON_SRCS-$(CONFIG_POSTPROC) += common/vp9_postproc.h
 VP9_COMMON_SRCS-$(CONFIG_POSTPROC) += common/vp9_postproc.c
-VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_recon_mmx.asm
 VP9_COMMON_SRCS-$(HAVE_MMX) += common/x86/vp9_loopfilter_mmx.asm
+VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_copy_sse2.asm
 VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_loopfilter_sse2.asm
-VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_recon_sse2.asm
 VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_intrapred_sse2.asm
 VP9_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/vp9_intrapred_ssse3.asm
 VP9_COMMON_SRCS-$(HAVE_SSSE3) += common/x86/vp9_subpixel_8t_ssse3.asm