Merge changes Ife0d8147,I7d469716,Ic9a5615f into experimental

* changes:
  Restore SSSE3 subpixel filters in new convolve framework
  Convert subpixel filters to use convolve framework
  Add 8-tap generic convolver

19 files changed, 671 insertions, 5894 deletions

diff --git a/vp9/common/generic/vp9_systemdependent.c b/vp9/common/generic/vp9_systemdependent.c
index b02f3f083..79092cd0e 100644
--- a/vp9/common/generic/vp9_systemdependent.c
+++ b/vp9/common/generic/vp9_systemdependent.c
@@ -11,8 +11,6 @@
 
 #include "./vpx_config.h"
 #include "vp9_rtcd.h"
-#include "vp9/common/vp9_subpixel.h"
-#include "vp9/common/vp9_loopfilter.h"
 #include "vp9/common/vp9_onyxc_int.h"
 
 void vp9_machine_specific_config(VP9_COMMON *ctx) {
diff --git a/vp9/common/ppc/vp9_systemdependent.c b/vp9/common/ppc/vp9_systemdependent.c
index 106a2b763..02035191f 100644
--- a/vp9/common/ppc/vp9_systemdependent.c
+++ b/vp9/common/ppc/vp9_systemdependent.c
@@ -8,7 +8,6 @@
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
-#include "vp9/common/vp9_subpixel.h"
 #include "vp9/common/vp9_loopfilter.h"
 #include "recon.h"
 #include "vp9/common/vp9_onyxc_int.h"
diff --git a/vp9/common/vp9_blockd.h b/vp9/common/vp9_blockd.h
index c6702ae31..28fafbdc9 100644
--- a/vp9/common/vp9_blockd.h
+++ b/vp9/common/vp9_blockd.h
@@ -16,9 +16,9 @@ void vpx_log(const char *format, ...);
 
 #include "./vpx_config.h"
 #include "vpx_scale/yv12config.h"
+#include "vp9/common/vp9_convolve.h"
 #include "vp9/common/vp9_mv.h"
 #include "vp9/common/vp9_treecoder.h"
-#include "vp9/common/vp9_subpixel.h"
 #include "vpx_ports/mem.h"
 #include "vp9/common/vp9_common.h"
 
@@ -394,15 +394,8 @@ typedef struct macroblockd {
   void (*inv_walsh4x4_1)(int16_t *in, int16_t *out);
   void (*inv_walsh4x4_lossless)(int16_t *in, int16_t *out);
 
+  struct subpix_fn_table  subpix;
 
-  vp9_subpix_fn_t  subpixel_predict4x4;
-  vp9_subpix_fn_t  subpixel_predict8x4;
-  vp9_subpix_fn_t  subpixel_predict8x8;
-  vp9_subpix_fn_t  subpixel_predict16x16;
-  vp9_subpix_fn_t  subpixel_predict_avg4x4;
-  vp9_subpix_fn_t  subpixel_predict_avg8x4;
-  vp9_subpix_fn_t  subpixel_predict_avg8x8;
-  vp9_subpix_fn_t  subpixel_predict_avg16x16;
   int allow_high_precision_mv;
 
   int corrupted;
diff --git a/vp9/common/vp9_convolve.c b/vp9/common/vp9_convolve.c
new file mode 100644
index 000000000..f21f1d84e
--- /dev/null
+++ b/vp9/common/vp9_convolve.c
@@ -0,0 +1,345 @@
+/*
+ *  Copyright (c) 2013 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 <assert.h>
+
+#include "./vpx_config.h"
+#include "./vp9_rtcd.h"
+#include "vp9/common/vp9_common.h"
+#include "vpx/vpx_integer.h"
+
+#define VP9_FILTER_WEIGHT 128
+#define VP9_FILTER_SHIFT  7
+#define ALIGN_FILTERS_256 0
+
+/* Assume a bank of 16 filters to choose from. There are two implementations
+ * for filter wrapping behavior, since we want to be able to pick which filter
+ * to start with. We could either:
+ *
+ * 1) make filter_ a pointer to the base of the filter array, and then add an
+ *    additional offset parameter, to choose the starting filter.
+ * 2) use a pointer to 2 periods worth of filters, so that even if the original
+ *    phase offset is at 15/16, we'll have valid data to read. The filter
+ *    tables become [32][8], and the second half is duplicated.
+ * 3) fix the alignment of the filter tables, so that we know the 0/16 is
+ *    always 256 byte aligned.
+ *
+ * Implementations 2 and 3 are likely preferable, as they avoid an extra 2
+ * parameters, and switching between them is trivial.
+ */
+static void convolve_horiz_c(const uint8_t *src, int src_stride,
+                             uint8_t *dst, int 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) {
+  int x, y, k, sum;
+  const int16_t *filter_x_base = filter_x0;
+
+#if ALIGN_FILTERS_256
+  filter_x_base = (const int16_t *)(((intptr_t)filter_x0) & ~(intptr_t)0xff);
+#endif
+
+  /* Adjust base pointer address for this source line */
+  src -= taps / 2 - 1;
+
+  for (y = 0; y < h; ++y) {
+    /* Pointer to filter to use */
+    const int16_t *filter_x = filter_x0;
+
+    /* Initial phase offset */
+    int x_q4 = (filter_x - filter_x_base) / taps;
+
+    for (x = 0; x < w; ++x) {
+      /* Per-pixel src offset */
+      int src_x = x_q4 >> 4;
+
+      for (sum = 0, k = 0; k < taps; ++k) {
+        sum += src[src_x + k] * filter_x[k];
+      }
+      sum += (VP9_FILTER_WEIGHT >> 1);
+      dst[x] = clip_pixel(sum >> VP9_FILTER_SHIFT);
+
+      /* Adjust source and filter to use for the next pixel */
+      x_q4 += x_step_q4;
+      filter_x = filter_x_base + (x_q4 & 0xf) * taps;
+    }
+    src += src_stride;
+    dst += dst_stride;
+  }
+}
+
+static void convolve_avg_horiz_c(const uint8_t *src, int src_stride,
+                                 uint8_t *dst, int 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) {
+  int x, y, k, sum;
+  const int16_t *filter_x_base = filter_x0;
+
+#if ALIGN_FILTERS_256
+  filter_x_base = (const int16_t *)(((intptr_t)filter_x0) & ~(intptr_t)0xff);
+#endif
+
+  /* Adjust base pointer address for this source line */
+  src -= taps / 2 - 1;
+
+  for (y = 0; y < h; ++y) {
+    /* Pointer to filter to use */
+    const int16_t *filter_x = filter_x0;
+
+    /* Initial phase offset */
+    int x_q4 = (filter_x - filter_x_base) / taps;
+
+    for (x = 0; x < w; ++x) {
+      /* Per-pixel src offset */
+      int src_x = x_q4 >> 4;
+
+      for (sum = 0, k = 0; k < taps; ++k) {
+        sum += src[src_x + k] * filter_x[k];
+      }
+      sum += (VP9_FILTER_WEIGHT >> 1);
+      dst[x] = (dst[x] + clip_pixel(sum >> VP9_FILTER_SHIFT) + 1) >> 1;
+
+      /* Adjust source and filter to use for the next pixel */
+      x_q4 += x_step_q4;
+      filter_x = filter_x_base + (x_q4 & 0xf) * taps;
+    }
+    src += src_stride;
+    dst += dst_stride;
+  }
+}
+
+static void convolve_vert_c(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_y0, int y_step_q4,
+                            int w, int h, int taps) {
+  int x, y, k, sum;
+
+  const int16_t *filter_y_base = filter_y0;
+
+#if ALIGN_FILTERS_256
+  filter_y_base = (const int16_t *)(((intptr_t)filter_y0) & ~(intptr_t)0xff);
+#endif
+
+  /* Adjust base pointer address for this source column */
+  src -= src_stride * (taps / 2 - 1);
+  for (x = 0; x < w; ++x) {
+    /* Pointer to filter to use */
+    const int16_t *filter_y = filter_y0;
+
+    /* Initial phase offset */
+    int y_q4 = (filter_y - filter_y_base) / taps;
+
+    for (y = 0; y < h; ++y) {
+      /* Per-pixel src offset */
+      int src_y = y_q4 >> 4;
+
+      for (sum = 0, k = 0; k < taps; ++k) {
+        sum += src[(src_y + k) * src_stride] * filter_y[k];
+      }
+      sum += (VP9_FILTER_WEIGHT >> 1);
+      dst[y * dst_stride] = clip_pixel(sum >> VP9_FILTER_SHIFT);
+
+      /* Adjust source and filter to use for the next pixel */
+      y_q4 += y_step_q4;
+      filter_y = filter_y_base + (y_q4 & 0xf) * taps;
+    }
+    ++src;
+    ++dst;
+  }
+}
+
+static void convolve_avg_vert_c(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_y0, int y_step_q4,
+                                int w, int h, int taps) {
+  int x, y, k, sum;
+
+  const int16_t *filter_y_base = filter_y0;
+
+#if ALIGN_FILTERS_256
+  filter_y_base = (const int16_t *)(((intptr_t)filter_y0) & ~(intptr_t)0xff);
+#endif
+
+  /* Adjust base pointer address for this source column */
+  src -= src_stride * (taps / 2 - 1);
+  for (x = 0; x < w; ++x) {
+    /* Pointer to filter to use */
+    const int16_t *filter_y = filter_y0;
+
+    /* Initial phase offset */
+    int y_q4 = (filter_y - filter_y_base) / taps;
+
+    for (y = 0; y < h; ++y) {
+      /* Per-pixel src offset */
+      int src_y = y_q4 >> 4;
+
+      for (sum = 0, k = 0; k < taps; ++k) {
+        sum += src[(src_y + k) * src_stride] * filter_y[k];
+      }
+      sum += (VP9_FILTER_WEIGHT >> 1);
+      dst[y * dst_stride] =
+          (dst[y * dst_stride] + clip_pixel(sum >> VP9_FILTER_SHIFT) + 1) >> 1;
+
+      /* Adjust source and filter to use for the next pixel */
+      y_q4 += y_step_q4;
+      filter_y = filter_y_base + (y_q4 & 0xf) * taps;
+    }
+    ++src;
+    ++dst;
+  }
+}
+
+static void convolve_c(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, int taps) {
+  /* Fixed size intermediate buffer places limits on parameters. */
+  uint8_t temp[16 * 23];
+  assert(w <= 16);
+  assert(h <= 16);
+  assert(taps <= 8);
+
+  convolve_horiz_c(src - src_stride * (taps / 2 - 1), src_stride,
+                   temp, 16,
+                   filter_x, x_step_q4, filter_y, y_step_q4,
+                   w, h + taps - 1, taps);
+  convolve_vert_c(temp + 16 * (taps / 2 - 1), 16, dst, dst_stride,
+                  filter_x, x_step_q4, filter_y, y_step_q4,
+                  w, h, taps);
+}
+
+static void convolve_avg_c(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, int taps) {
+  /* Fixed size intermediate buffer places limits on parameters. */
+  uint8_t temp[16 * 23];
+  assert(w <= 16);
+  assert(h <= 16);
+  assert(taps <= 8);
+
+  convolve_horiz_c(src - src_stride * (taps / 2 - 1), src_stride,
+                   temp, 16,
+                   filter_x, x_step_q4, filter_y, y_step_q4,
+                   w, h + taps - 1, taps);
+  convolve_avg_vert_c(temp + 16 * (taps / 2 - 1), 16, dst, dst_stride,
+                      filter_x, x_step_q4, filter_y, y_step_q4,
+                      w, h, taps);
+}
+
+void vp9_convolve8_horiz_c(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) {
+  convolve_horiz_c(src, src_stride, dst, dst_stride,
+                   filter_x, x_step_q4, filter_y, y_step_q4,
+                   w, h, 8);
+}
+
+void vp9_convolve8_avg_horiz_c(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) {
+  convolve_avg_horiz_c(src, src_stride, dst, dst_stride,
+                       filter_x, x_step_q4, filter_y, y_step_q4,
+                       w, h, 8);
+}
+
+void vp9_convolve8_vert_c(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) {
+  convolve_vert_c(src, src_stride, dst, dst_stride,
+                  filter_x, x_step_q4, filter_y, y_step_q4,
+                  w, h, 8);
+}
+
+void vp9_convolve8_avg_vert_c(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) {
+  convolve_avg_vert_c(src, src_stride, dst, dst_stride,
+                      filter_x, x_step_q4, filter_y, y_step_q4,
+                      w, h, 8);
+}
+
+void vp9_convolve8_c(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) {
+  convolve_c(src, src_stride, dst, dst_stride,
+             filter_x, x_step_q4, filter_y, y_step_q4,
+             w, h, 8);
+}
+
+void vp9_convolve8_avg_c(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) {
+  convolve_avg_c(src, src_stride, dst, dst_stride,
+                 filter_x, x_step_q4, filter_y, y_step_q4,
+                 w, h, 8);
+}
+
+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 (h == 16) {
+    vp9_copy_mem16x16(src, src_stride, dst, dst_stride);
+  } else if (h == 8) {
+    vp9_copy_mem8x8(src, src_stride, dst, dst_stride);
+  } else if (w == 8) {
+    vp9_copy_mem8x4(src, src_stride, dst, dst_stride);
+  } else {
+    // 4x4
+    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];
+#else
+      *(uint32_t *)dst = *(const uint32_t *)src;
+#endif
+      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) {
+  int x, y;
+
+  for (y = 0; y < h; ++y) {
+    for (x = 0; x < w; ++x) {
+      dst[x] = (dst[x] + src[x] + 1) >> 1;
+    }
+    src += src_stride;
+    dst += dst_stride;
+  }
+}
diff --git a/vp9/common/vp9_convolve.h b/vp9/common/vp9_convolve.h
new file mode 100644
index 000000000..46c935ab7
--- /dev/null
+++ b/vp9/common/vp9_convolve.h
@@ -0,0 +1,43 @@
+/*
+ *  Copyright (c) 2013 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.
+ */
+#ifndef VP9_COMMON_CONVOLVE_H_
+#define VP9_COMMON_CONVOLVE_H_
+
+#include "vpx/vpx_integer.h"
+
+typedef void (*convolve_fn_t)(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 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 {
+  convolve_fn_t predict[2][2][2];  // horiz, vert, avg
+  const int16_t (*filter_x)[8];
+  const int16_t (*filter_y)[8];
+  int x_step_q4;
+  int y_step_q4;
+};
+
+#endif  // VP9_COMMON_CONVOLVE_H_
diff --git a/vp9/common/vp9_filter.c b/vp9/common/vp9_filter.c
index 07d8a169f..5e425895f 100644
--- a/vp9/common/vp9_filter.c
+++ b/vp9/common/vp9_filter.c
@@ -15,23 +15,23 @@
 #include "vp9_rtcd.h"
 #include "vp9/common/vp9_common.h"
 
-DECLARE_ALIGNED(16, const int16_t, vp9_bilinear_filters[SUBPEL_SHIFTS][2]) = {
-  { 128,   0 },
-  { 120,   8 },
-  { 112,  16 },
-  { 104,  24 },
-  {  96,  32 },
-  {  88,  40 },
-  {  80,  48 },
-  {  72,  56 },
-  {  64,  64 },
-  {  56,  72 },
-  {  48,  80 },
-  {  40,  88 },
-  {  32,  96 },
-  {  24, 104 },
-  {  16, 112 },
-  {   8, 120 }
+DECLARE_ALIGNED(16, const int16_t, vp9_bilinear_filters[SUBPEL_SHIFTS][8]) = {
+  { 0, 0, 0, 128,   0, 0, 0, 0 },
+  { 0, 0, 0, 120,   8, 0, 0, 0 },
+  { 0, 0, 0, 112,  16, 0, 0, 0 },
+  { 0, 0, 0, 104,  24, 0, 0, 0 },
+  { 0, 0, 0,  96,  32, 0, 0, 0 },
+  { 0, 0, 0,  88,  40, 0, 0, 0 },
+  { 0, 0, 0,  80,  48, 0, 0, 0 },
+  { 0, 0, 0,  72,  56, 0, 0, 0 },
+  { 0, 0, 0,  64,  64, 0, 0, 0 },
+  { 0, 0, 0,  56,  72, 0, 0, 0 },
+  { 0, 0, 0,  48,  80, 0, 0, 0 },
+  { 0, 0, 0,  40,  88, 0, 0, 0 },
+  { 0, 0, 0,  32,  96, 0, 0, 0 },
+  { 0, 0, 0,  24, 104, 0, 0, 0 },
+  { 0, 0, 0,  16, 112, 0, 0, 0 },
+  { 0, 0, 0,   8, 120, 0, 0, 0 }
 };
 
 #define FILTER_ALPHA       0
@@ -144,1072 +144,21 @@ DECLARE_ALIGNED(16, const int16_t,
   { 1, -2, -7, 37, 80, 28, -8, -1}
 };
 
-DECLARE_ALIGNED(16, const int16_t, vp9_sub_pel_filters_6[SUBPEL_SHIFTS][6]) = {
-  {0,   0, 128,   0,   0, 0},
-  {1,  -5, 125,   8,  -2, 1},
-  {1,  -8, 122,  17,  -5, 1},
-  {2, -11, 116,  27,  -8, 2},
-  {3, -14, 110,  37, -10, 2},
-  {3, -15, 103,  47, -12, 2},
-  {3, -16,  95,  57, -14, 3},
-  {3, -16,  86,  67, -15, 3},
-  {3, -16,  77,  77, -16, 3},
-  {3, -15,  67,  86, -16, 3},
-  {3, -14,  57,  95, -16, 3},
-  {2, -12,  47, 103, -15, 3},
-  {2, -10,  37, 110, -14, 3},
-  {2,  -8,  27, 116, -11, 2},
-  {1,  -5,  17, 122,  -8, 1},
-  {1,  -2,   8, 125,  -5, 1}
+DECLARE_ALIGNED(16, const int16_t, vp9_sub_pel_filters_6[SUBPEL_SHIFTS][8]) = {
+  {0, 0,   0, 128,   0,   0, 0,  0},
+  {0, 1,  -5, 125,   8,  -2, 1,  0},
+  {0, 1,  -8, 122,  17,  -5, 1,  0},
+  {0, 2, -11, 116,  27,  -8, 2,  0},
+  {0, 3, -14, 110,  37, -10, 2,  0},
+  {0, 3, -15, 103,  47, -12, 2,  0},
+  {0, 3, -16,  95,  57, -14, 3,  0},
+  {0, 3, -16,  86,  67, -15, 3,  0},
+  {0, 3, -16,  77,  77, -16, 3,  0},
+  {0, 3, -15,  67,  86, -16, 3,  0},
+  {0, 3, -14,  57,  95, -16, 3,  0},
+  {0, 2, -12,  47, 103, -15, 3,  0},
+  {0, 2, -10,  37, 110, -14, 3,  0},
+  {0, 2,  -8,  27, 116, -11, 2,  0},
+  {0, 1,  -5,  17, 122,  -8, 1,  0},
+  {0, 1,  -2,   8, 125,  -5, 1,  0}
 };
-
-static void filter_block2d_first_pass_6(uint8_t *src_ptr,
-                                        int *output_ptr,
-                                        unsigned int src_pixels_per_line,
-                                        unsigned int pixel_step,
-                                        unsigned int output_height,
-                                        unsigned int output_width,
-                                        const int16_t *vp9_filter) {
-  unsigned int i, j;
-  int temp;
-
-  for (i = 0; i < output_height; i++) {
-    for (j = 0; j < output_width; j++) {
-      temp = ((int)src_ptr[-2 * (int)pixel_step] * vp9_filter[0]) +
-             ((int)src_ptr[-1 * (int)pixel_step] * vp9_filter[1]) +
-             ((int)src_ptr[0]                    * vp9_filter[2]) +
-             ((int)src_ptr[pixel_step]           * vp9_filter[3]) +
-             ((int)src_ptr[2 * pixel_step]       * vp9_filter[4]) +
-             ((int)src_ptr[3 * pixel_step]       * vp9_filter[5]) +
-             (VP9_FILTER_WEIGHT >> 1);      /* Rounding */
-
-      /* Normalize back to 0-255 */
-      output_ptr[j] = clip_pixel(temp >> VP9_FILTER_SHIFT);
-      src_ptr++;
-    }
-
-    /* Next row... */
-    src_ptr    += src_pixels_per_line - output_width;
-    output_ptr += output_width;
-  }
-}
-
-static void filter_block2d_second_pass_6(int *src_ptr,
-                                         uint8_t *output_ptr,
-                                         int output_pitch,
-                                         unsigned int src_pixels_per_line,
-                                         unsigned int pixel_step,
-                                         unsigned int output_height,
-                                         unsigned int output_width,
-                                         const int16_t *vp9_filter) {
-  unsigned int i, j;
-  int temp;
-
-  for (i = 0; i < output_height; i++) {
-    for (j = 0; j < output_width; j++) {
-      /* Apply filter */
-      temp = ((int)src_ptr[-2 * (int)pixel_step] * vp9_filter[0]) +
-             ((int)src_ptr[-1 * (int)pixel_step] * vp9_filter[1]) +
-             ((int)src_ptr[0]                    * vp9_filter[2]) +
-             ((int)src_ptr[pixel_step]           * vp9_filter[3]) +
-             ((int)src_ptr[2 * pixel_step]         * vp9_filter[4]) +
-             ((int)src_ptr[3 * pixel_step]         * vp9_filter[5]) +
-             (VP9_FILTER_WEIGHT >> 1);   /* Rounding */
-
-      /* Normalize back to 0-255 */
-      output_ptr[j] = clip_pixel(temp >> VP9_FILTER_SHIFT);
-      src_ptr++;
-    }
-
-    /* Start next row */
-    src_ptr    += src_pixels_per_line - output_width;
-    output_ptr += output_pitch;
-  }
-}
-
-/*
- * The only functional difference between filter_block2d_second_pass()
- * and this function is that filter_block2d_second_pass() does a sixtap
- * filter on the input and stores it in the output. This function
- * (filter_block2d_second_pass_avg()) does a sixtap filter on the input,
- * and then averages that with the content already present in the output
- * ((filter_result + dest + 1) >> 1) and stores that in the output.
- */
-static void filter_block2d_second_pass_avg_6(int *src_ptr,
-                                             uint8_t *output_ptr,
-                                             int output_pitch,
-                                             unsigned int src_pixels_per_line,
-                                             unsigned int pixel_step,
-                                             unsigned int output_height,
-                                             unsigned int output_width,
-                                             const int16_t *vp9_filter) {
-  unsigned int i, j;
-  int temp;
-
-  for (i = 0; i < output_height; i++) {
-    for (j = 0; j < output_width; j++) {
-      /* Apply filter */
-      temp = ((int)src_ptr[-2 * (int)pixel_step] * vp9_filter[0]) +
-             ((int)src_ptr[-1 * (int)pixel_step] * vp9_filter[1]) +
-             ((int)src_ptr[0]                    * vp9_filter[2]) +
-             ((int)src_ptr[pixel_step]           * vp9_filter[3]) +
-             ((int)src_ptr[2 * pixel_step]         * vp9_filter[4]) +
-             ((int)src_ptr[3 * pixel_step]         * vp9_filter[5]) +
-             (VP9_FILTER_WEIGHT >> 1);   /* Rounding */
-
-      /* Normalize back to 0-255 */
-      output_ptr[j] = (clip_pixel(temp >> VP9_FILTER_SHIFT) +
-                       output_ptr[j] + 1) >> 1;
-      src_ptr++;
-    }
-
-    /* Start next row */
-    src_ptr    += src_pixels_per_line - output_width;
-    output_ptr += output_pitch;
-  }
-}
-
-#define Interp_Extend 3
-static void filter_block2d_6(uint8_t *src_ptr,
-                             uint8_t *output_ptr,
-                             unsigned int src_pixels_per_line,
-                             int output_pitch,
-                             const int16_t *HFilter,
-                             const int16_t *VFilter) {
-  int FData[(3 + Interp_Extend * 2) * 4]; /* Temp data buffer */
-
-  /* First filter 1-D horizontally... */
-  filter_block2d_first_pass_6(
-      src_ptr - ((Interp_Extend - 1) * src_pixels_per_line), FData,
-      src_pixels_per_line, 1, 3 + Interp_Extend * 2, 4, HFilter);
-
-  /* then filter vertically... */
-  filter_block2d_second_pass_6(FData + 4 * (Interp_Extend - 1), output_ptr,
-                               output_pitch, 4, 4, 4, 4, VFilter);
-}
-
-
-void vp9_sixtap_predict4x4_c(uint8_t *src_ptr,
-                             int src_pixels_per_line,
-                             int xoffset,
-                             int yoffset,
-                             uint8_t *dst_ptr,
-                             int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_sub_pel_filters_6[xoffset];   /* 6 tap */
-  VFilter = vp9_sub_pel_filters_6[yoffset];   /* 6 tap */
-
-  filter_block2d_6(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter,
-                   VFilter);
-}
-
-/*
- * The difference between filter_block2d_6() and filter_block2d_avg_6 is
- * that filter_block2d_6() does a 6-tap filter and stores it in the output
- * buffer, whereas filter_block2d_avg_6() does the same 6-tap filter, and
- * then averages that with the content already present in the output
- * ((filter_result + dest + 1) >> 1) and stores that in the output.
- */
-static void filter_block2d_avg_6(uint8_t *src_ptr,
-                                 uint8_t *output_ptr,
-                                 unsigned int src_pixels_per_line,
-                                 int output_pitch,
-                                 const int16_t *HFilter,
-                                 const int16_t *VFilter) {
-  int FData[(3 + Interp_Extend * 2) * 4]; /* Temp data buffer */
-
-  /* First filter 1-D horizontally... */
-  filter_block2d_first_pass_6(
-      src_ptr - ((Interp_Extend - 1) * src_pixels_per_line), FData,
-      src_pixels_per_line, 1, 3 + Interp_Extend * 2, 4, HFilter);
-
-  /* then filter vertically... */
-  filter_block2d_second_pass_avg_6(FData + 4 * (Interp_Extend - 1), output_ptr,
-                                   output_pitch, 4, 4, 4, 4, VFilter);
-}
-
-void vp9_sixtap_predict_avg4x4_c(uint8_t *src_ptr,
-                                 int src_pixels_per_line,
-                                 int xoffset,
-                                 int yoffset,
-                                 uint8_t *dst_ptr,
-                                 int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_sub_pel_filters_6[xoffset];   /* 6 tap */
-  VFilter = vp9_sub_pel_filters_6[yoffset];   /* 6 tap */
-
-  filter_block2d_avg_6(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch,
-                       HFilter, VFilter);
-}
-
-void vp9_sixtap_predict8x8_c(uint8_t *src_ptr,
-                             int src_pixels_per_line,
-                             int xoffset,
-                             int yoffset,
-                             uint8_t *dst_ptr,
-                             int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-  int FData[(7 + Interp_Extend * 2) * 8]; /* Temp data buffer */
-
-  HFilter = vp9_sub_pel_filters_6[xoffset];   /* 6 tap */
-  VFilter = vp9_sub_pel_filters_6[yoffset];   /* 6 tap */
-
-  /* First filter 1-D horizontally... */
-  filter_block2d_first_pass_6(
-      src_ptr - ((Interp_Extend - 1) * src_pixels_per_line), FData,
-      src_pixels_per_line, 1, 7 + Interp_Extend * 2, 8, HFilter);
-
-  /* then filter vertically... */
-  filter_block2d_second_pass_6(FData + 8 * (Interp_Extend - 1), dst_ptr,
-                               dst_pitch, 8, 8, 8, 8, VFilter);
-
-}
-
-void vp9_sixtap_predict_avg8x8_c(uint8_t *src_ptr,
-                                 int src_pixels_per_line,
-                                 int xoffset,
-                                 int yoffset,
-                                 uint8_t *dst_ptr,
-                                 int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-  int FData[(7 + Interp_Extend * 2) * 8]; /* Temp data buffer */
-
-  HFilter = vp9_sub_pel_filters_6[xoffset];   /* 6 tap */
-  VFilter = vp9_sub_pel_filters_6[yoffset];   /* 6 tap */
-
-  /* First filter 1-D horizontally... */
-  filter_block2d_first_pass_6(
-      src_ptr - ((Interp_Extend - 1) * src_pixels_per_line), FData,
-      src_pixels_per_line, 1, 7 + Interp_Extend * 2, 8, HFilter);
-
-  /* then filter vertically... */
-  filter_block2d_second_pass_avg_6(FData + 8 * (Interp_Extend - 1), dst_ptr,
-                                   dst_pitch, 8, 8, 8, 8, VFilter);
-}
-
-void vp9_sixtap_predict8x4_c(uint8_t *src_ptr,
-                             int src_pixels_per_line,
-                             int xoffset,
-                             int yoffset,
-                             uint8_t *dst_ptr,
-                             int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-  int FData[(3 + Interp_Extend * 2) * 8]; /* Temp data buffer */
-
-  HFilter = vp9_sub_pel_filters_6[xoffset];   /* 6 tap */
-  VFilter = vp9_sub_pel_filters_6[yoffset];   /* 6 tap */
-
-  /* First filter 1-D horizontally... */
-  filter_block2d_first_pass_6(
-      src_ptr - ((Interp_Extend - 1) * src_pixels_per_line), FData,
-      src_pixels_per_line, 1, 3 + Interp_Extend * 2, 8, HFilter);
-
-  /* then filter vertically... */
-  filter_block2d_second_pass_6(FData + 8 * (Interp_Extend - 1), dst_ptr,
-                               dst_pitch, 8, 8, 4, 8, VFilter);
-}
-
-void vp9_sixtap_predict16x16_c(uint8_t *src_ptr,
-                               int src_pixels_per_line,
-                               int xoffset,
-                               int yoffset,
-                               uint8_t *dst_ptr,
-                               int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-  int FData[(15 + Interp_Extend * 2) * 16]; /* Temp data buffer */
-
-  HFilter = vp9_sub_pel_filters_6[xoffset];   /* 6 tap */
-  VFilter = vp9_sub_pel_filters_6[yoffset];   /* 6 tap */
-
-  /* First filter 1-D horizontally... */
-  filter_block2d_first_pass_6(
-      src_ptr - ((Interp_Extend - 1) * src_pixels_per_line), FData,
-      src_pixels_per_line, 1, 15 + Interp_Extend * 2, 16, HFilter);
-
-  /* then filter vertically... */
-  filter_block2d_second_pass_6(FData + 16 * (Interp_Extend - 1), dst_ptr,
-                               dst_pitch, 16, 16, 16, 16, VFilter);
-}
-
-void vp9_sixtap_predict_avg16x16_c(uint8_t *src_ptr,
-                                   int src_pixels_per_line,
-                                   int xoffset,
-                                   int yoffset,
-                                   uint8_t *dst_ptr,
-                                   int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-  int FData[(15 + Interp_Extend * 2) * 16]; /* Temp data buffer */
-
-  HFilter = vp9_sub_pel_filters_6[xoffset];   /* 6 tap */
-  VFilter = vp9_sub_pel_filters_6[yoffset];   /* 6 tap */
-
-  /* First filter 1-D horizontally... */
-  filter_block2d_first_pass_6(
-      src_ptr - ((Interp_Extend - 1) * src_pixels_per_line), FData,
-      src_pixels_per_line, 1, 15 + Interp_Extend * 2, 16, HFilter);
-
-  /* then filter vertically... */
-  filter_block2d_second_pass_avg_6(FData + 16 * (Interp_Extend - 1), dst_ptr,
-                                   dst_pitch, 16, 16, 16, 16, VFilter);
-}
-
-typedef enum {
-  VPX_FILTER_4x4 = 0,
-  VPX_FILTER_8x8 = 1,
-  VPX_FILTER_8x4 = 2,
-  VPX_FILTER_16x16 = 3,
-} filter_size_t;
-
-static const unsigned int filter_size_to_wh[][2] = {
-  {4, 4},
-  {8, 8},
-  {8, 4},
-  {16,16},
-};
-
-static void filter_block2d_8_c(const uint8_t *src_ptr,
-                               const unsigned int src_stride,
-                               const int16_t *HFilter,
-                               const int16_t *VFilter,
-                               const filter_size_t filter_size,
-                               uint8_t *dst_ptr,
-                               unsigned int dst_stride) {
-  const unsigned int output_width = filter_size_to_wh[filter_size][0];
-  const unsigned int output_height = filter_size_to_wh[filter_size][1];
-
-  // Between passes, we use an intermediate buffer whose height is extended to
-  // have enough horizontally filtered values as input for the vertical pass.
-  // This buffer is allocated to be big enough for the largest block type we
-  // support.
-  const int kInterp_Extend = 4;
-  const unsigned int intermediate_height =
-    (kInterp_Extend - 1) +     output_height + kInterp_Extend;
-
-  /* Size of intermediate_buffer is max_intermediate_height * filter_max_width,
-   * where max_intermediate_height = (kInterp_Extend - 1) + filter_max_height
-   *                                 + kInterp_Extend
-   *                               = 3 + 16 + 4
-   *                               = 23
-   * and filter_max_width = 16
-   */
-  uint8_t intermediate_buffer[23 * 16];
-  const int intermediate_next_stride = 1 - intermediate_height * output_width;
-
-  // Horizontal pass (src -> transposed intermediate).
-  {
-    uint8_t *output_ptr = intermediate_buffer;
-    const int src_next_row_stride = src_stride - output_width;
-    unsigned int i, j;
-    src_ptr -= (kInterp_Extend - 1) * src_stride + (kInterp_Extend - 1);
-    for (i = 0; i < intermediate_height; i++) {
-      for (j = 0; j < output_width; j++) {
-        // Apply filter...
-        int temp = ((int)src_ptr[0] * HFilter[0]) +
-                   ((int)src_ptr[1] * HFilter[1]) +
-                   ((int)src_ptr[2] * HFilter[2]) +
-                   ((int)src_ptr[3] * HFilter[3]) +
-                   ((int)src_ptr[4] * HFilter[4]) +
-                   ((int)src_ptr[5] * HFilter[5]) +
-                   ((int)src_ptr[6] * HFilter[6]) +
-                   ((int)src_ptr[7] * HFilter[7]) +
-                   (VP9_FILTER_WEIGHT >> 1); // Rounding
-
-        // Normalize back to 0-255...
-        *output_ptr = clip_pixel(temp >> VP9_FILTER_SHIFT);
-        src_ptr++;
-        output_ptr += intermediate_height;
-      }
-      src_ptr += src_next_row_stride;
-      output_ptr += intermediate_next_stride;
-    }
-  }
-
-  // Vertical pass (transposed intermediate -> dst).
-  {
-    uint8_t *src_ptr = intermediate_buffer;
-    const int dst_next_row_stride = dst_stride - output_width;
-    unsigned int i, j;
-    for (i = 0; i < output_height; i++) {
-      for (j = 0; j < output_width; j++) {
-        // Apply filter...
-        int temp = ((int)src_ptr[0] * VFilter[0]) +
-                   ((int)src_ptr[1] * VFilter[1]) +
-                   ((int)src_ptr[2] * VFilter[2]) +
-                   ((int)src_ptr[3] * VFilter[3]) +
-                   ((int)src_ptr[4] * VFilter[4]) +
-                   ((int)src_ptr[5] * VFilter[5]) +
-                   ((int)src_ptr[6] * VFilter[6]) +
-                   ((int)src_ptr[7] * VFilter[7]) +
-                   (VP9_FILTER_WEIGHT >> 1); // Rounding
-
-        // Normalize back to 0-255...
-        *dst_ptr++ = clip_pixel(temp >> VP9_FILTER_SHIFT);
-        src_ptr += intermediate_height;
-      }
-      src_ptr += intermediate_next_stride;
-      dst_ptr += dst_next_row_stride;
-    }
-  }
-}
-
-void vp9_filter_block2d_4x4_8_c(const uint8_t *src_ptr,
-                                const unsigned int src_stride,
-                                const int16_t *HFilter_aligned16,
-                                const int16_t *VFilter_aligned16,
-                                uint8_t *dst_ptr,
-                                unsigned int dst_stride) {
-  filter_block2d_8_c(src_ptr, src_stride, HFilter_aligned16, VFilter_aligned16,
-                     VPX_FILTER_4x4, dst_ptr, dst_stride);
-}
-
-void vp9_filter_block2d_8x4_8_c(const uint8_t *src_ptr,
-                                const unsigned int src_stride,
-                                const int16_t *HFilter_aligned16,
-                                const int16_t *VFilter_aligned16,
-                                uint8_t *dst_ptr,
-                                unsigned int dst_stride) {
-  filter_block2d_8_c(src_ptr, src_stride, HFilter_aligned16, VFilter_aligned16,
-                     VPX_FILTER_8x4, dst_ptr, dst_stride);
-}
-
-void vp9_filter_block2d_8x8_8_c(const uint8_t *src_ptr,
-                                const unsigned int src_stride,
-                                const int16_t *HFilter_aligned16,
-                                const int16_t *VFilter_aligned16,
-                                uint8_t *dst_ptr,
-                                unsigned int dst_stride) {
-  filter_block2d_8_c(src_ptr, src_stride, HFilter_aligned16, VFilter_aligned16,
-                     VPX_FILTER_8x8, dst_ptr, dst_stride);
-}
-
-void vp9_filter_block2d_16x16_8_c(const uint8_t *src_ptr,
-                                  const unsigned int src_stride,
-                                  const int16_t *HFilter_aligned16,
-                                  const int16_t *VFilter_aligned16,
-                                  uint8_t *dst_ptr,
-                                  unsigned int dst_stride) {
-  filter_block2d_8_c(src_ptr, src_stride, HFilter_aligned16, VFilter_aligned16,
-                     VPX_FILTER_16x16, dst_ptr, dst_stride);
-}
-
-static void block2d_average_c(uint8_t *src,
-                              unsigned int src_stride,
-                              uint8_t *output_ptr,
-                              unsigned int output_stride,
-                              const filter_size_t filter_size) {
-  const unsigned int output_width = filter_size_to_wh[filter_size][0];
-  const unsigned int output_height = filter_size_to_wh[filter_size][1];
-
-  unsigned int i, j;
-  for (i = 0; i < output_height; i++) {
-    for (j = 0; j < output_width; j++) {
-      output_ptr[j] = (output_ptr[j] + src[i * src_stride + j] + 1) >> 1;
-    }
-    output_ptr += output_stride;
-  }
-}
-
-#define block2d_average block2d_average_c
-
-void vp9_eighttap_predict4x4_c(uint8_t *src_ptr,
-                               int src_pixels_per_line,
-                               int xoffset,
-                               int yoffset,
-                               uint8_t *dst_ptr,
-                               int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_sub_pel_filters_8[xoffset];
-  VFilter = vp9_sub_pel_filters_8[yoffset];
-
-  vp9_filter_block2d_4x4_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                           dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict_avg4x4_c(uint8_t *src_ptr,
-                                   int src_pixels_per_line,
-                                   int xoffset,
-                                   int yoffset,
-                                   uint8_t *dst_ptr,
-                                   int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8[yoffset];
-  uint8_t tmp[4 * 4];
-
-  vp9_filter_block2d_4x4_8(src_ptr, src_pixels_per_line, HFilter, VFilter, tmp,
-                           4);
-  block2d_average(tmp, 4, dst_ptr, dst_pitch, VPX_FILTER_4x4);
-}
-
-void vp9_eighttap_predict4x4_sharp_c(uint8_t *src_ptr,
-                                     int src_pixels_per_line,
-                                     int xoffset,
-                                     int yoffset,
-                                     uint8_t *dst_ptr,
-                                     int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_sub_pel_filters_8s[xoffset];
-  VFilter = vp9_sub_pel_filters_8s[yoffset];
-
-  vp9_filter_block2d_4x4_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                           dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict4x4_smooth_c(uint8_t *src_ptr,
-                                      int src_pixels_per_line,
-                                      int xoffset,
-                                      int yoffset,
-                                      uint8_t *dst_ptr,
-                                      int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_sub_pel_filters_8lp[xoffset];
-  VFilter = vp9_sub_pel_filters_8lp[yoffset];
-
-  vp9_filter_block2d_4x4_8(src_ptr, src_pixels_per_line,
-                           HFilter, VFilter,
-                           dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict_avg4x4_sharp_c(uint8_t *src_ptr,
-                                         int src_pixels_per_line,
-                                         int xoffset,
-                                         int yoffset,
-                                         uint8_t *dst_ptr,
-                                         int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8s[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8s[yoffset];
-  uint8_t tmp[4 * 4];
-
-  vp9_filter_block2d_4x4_8(src_ptr, src_pixels_per_line, HFilter, VFilter, tmp,
-                           4);
-  block2d_average(tmp, 4, dst_ptr, dst_pitch, VPX_FILTER_4x4);
-}
-
-void vp9_eighttap_predict_avg4x4_smooth_c(uint8_t *src_ptr,
-                                          int src_pixels_per_line,
-                                          int xoffset,
-                                          int yoffset,
-                                          uint8_t *dst_ptr,
-                                          int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8lp[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8lp[yoffset];
-  uint8_t tmp[4 * 4];
-
-  vp9_filter_block2d_4x4_8(src_ptr, src_pixels_per_line, HFilter, VFilter, tmp,
-                           4);
-  block2d_average(tmp, 4, dst_ptr, dst_pitch, VPX_FILTER_4x4);
-}
-
-
-void vp9_eighttap_predict8x8_c(uint8_t *src_ptr,
-                               int src_pixels_per_line,
-                               int xoffset,
-                               int yoffset,
-                               uint8_t *dst_ptr,
-                               int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8[yoffset];
-
-  vp9_filter_block2d_8x8_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                           dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict8x8_sharp_c(uint8_t *src_ptr,
-                                     int src_pixels_per_line,
-                                     int xoffset,
-                                     int yoffset,
-                                     uint8_t *dst_ptr,
-                                     int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8s[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8s[yoffset];
-
-  vp9_filter_block2d_8x8_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                           dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict8x8_smooth_c(uint8_t *src_ptr,
-                                      int src_pixels_per_line,
-                                      int xoffset,
-                                      int yoffset,
-                                      uint8_t *dst_ptr,
-                                      int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8lp[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8lp[yoffset];
-
-  vp9_filter_block2d_8x8_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                           dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict_avg8x8_c(uint8_t *src_ptr,
-                                   int src_pixels_per_line,
-                                   int xoffset,
-                                   int yoffset,
-                                   uint8_t *dst_ptr,
-                                   int dst_pitch) {
-  uint8_t tmp[8 * 8];
-  const int16_t *HFilter = vp9_sub_pel_filters_8[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8[yoffset];
-
-  vp9_filter_block2d_8x8_8(src_ptr, src_pixels_per_line, HFilter, VFilter, tmp,
-                           8);
-  block2d_average(tmp, 8, dst_ptr, dst_pitch, VPX_FILTER_8x8);
-}
-
-void vp9_eighttap_predict_avg8x8_sharp_c(uint8_t *src_ptr,
-                                         int src_pixels_per_line,
-                                         int xoffset,
-                                         int yoffset,
-                                         uint8_t *dst_ptr,
-                                         int dst_pitch) {
-  uint8_t tmp[8 * 8];
-  const int16_t *HFilter = vp9_sub_pel_filters_8s[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8s[yoffset];
-
-  vp9_filter_block2d_8x8_8(src_ptr, src_pixels_per_line, HFilter, VFilter, tmp,
-                           8);
-  block2d_average(tmp, 8, dst_ptr, dst_pitch, VPX_FILTER_8x8);
-}
-
-void vp9_eighttap_predict_avg8x8_smooth_c(uint8_t *src_ptr,
-                                          int src_pixels_per_line,
-                                          int xoffset,
-                                          int yoffset,
-                                          uint8_t *dst_ptr,
-                                          int dst_pitch) {
-  uint8_t tmp[8 * 8];
-  const int16_t *HFilter = vp9_sub_pel_filters_8lp[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8lp[yoffset];
-
-  vp9_filter_block2d_8x8_8(src_ptr, src_pixels_per_line, HFilter, VFilter, tmp,
-                           8);
-  block2d_average(tmp, 8, dst_ptr, dst_pitch, VPX_FILTER_8x8);
-}
-
-void vp9_eighttap_predict8x4_c(uint8_t *src_ptr,
-                               int src_pixels_per_line,
-                               int xoffset,
-                               int yoffset,
-                               uint8_t *dst_ptr,
-                               int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8[yoffset];
-
-  vp9_filter_block2d_8x4_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                           dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict8x4_sharp_c(uint8_t *src_ptr,
-                                     int src_pixels_per_line,
-                                     int xoffset,
-                                     int yoffset,
-                                     uint8_t *dst_ptr,
-                                     int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8s[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8s[yoffset];
-
-  vp9_filter_block2d_8x4_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                           dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict8x4_smooth_c(uint8_t *src_ptr,
-                                      int src_pixels_per_line,
-                                      int xoffset,
-                                      int yoffset,
-                                      uint8_t *dst_ptr,
-                                      int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8lp[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8lp[yoffset];
-
-  vp9_filter_block2d_8x4_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                           dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict16x16_c(uint8_t *src_ptr,
-                                 int src_pixels_per_line,
-                                 int xoffset,
-                                 int yoffset,
-                                 uint8_t *dst_ptr,
-                                 int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8[yoffset];
-
-  vp9_filter_block2d_16x16_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                             dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict16x16_sharp_c(uint8_t *src_ptr,
-                                       int src_pixels_per_line,
-                                       int xoffset,
-                                       int yoffset,
-                                       uint8_t *dst_ptr,
-                                       int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8s[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8s[yoffset];
-
-  vp9_filter_block2d_16x16_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                             dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict16x16_smooth_c(uint8_t *src_ptr,
-                                        int src_pixels_per_line,
-                                        int xoffset,
-                                        int yoffset,
-                                        uint8_t *dst_ptr,
-                                        int dst_pitch) {
-  const int16_t *HFilter = vp9_sub_pel_filters_8lp[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8lp[yoffset];
-
-  vp9_filter_block2d_16x16_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                             dst_ptr, dst_pitch);
-}
-
-void vp9_eighttap_predict_avg16x16_c(uint8_t *src_ptr,
-                                     int src_pixels_per_line,
-                                     int xoffset,
-                                     int yoffset,
-                                     uint8_t *dst_ptr,
-                                     int dst_pitch) {
-  DECLARE_ALIGNED_ARRAY(16, uint8_t, tmp, 16 * 16);
-  const int16_t *HFilter = vp9_sub_pel_filters_8[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8[yoffset];
-
-  vp9_filter_block2d_16x16_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                             tmp, 16);
-  block2d_average(tmp, 16, dst_ptr, dst_pitch, VPX_FILTER_16x16);
-}
-
-void vp9_eighttap_predict_avg16x16_sharp_c(uint8_t *src_ptr,
-                                           int src_pixels_per_line,
-                                           int xoffset,
-                                           int yoffset,
-                                           uint8_t *dst_ptr,
-                                           int dst_pitch) {
-  DECLARE_ALIGNED_ARRAY(16, uint8_t, tmp, 16 * 16);
-  const int16_t *HFilter = vp9_sub_pel_filters_8s[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8s[yoffset];
-
-  vp9_filter_block2d_16x16_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                             tmp, 16);
-  block2d_average(tmp, 16, dst_ptr, dst_pitch, VPX_FILTER_16x16);
-}
-
-void vp9_eighttap_predict_avg16x16_smooth_c(uint8_t *src_ptr,
-                                            int src_pixels_per_line,
-                                            int xoffset,
-                                            int yoffset,
-                                            uint8_t *dst_ptr,
-                                            int dst_pitch) {
-  DECLARE_ALIGNED_ARRAY(16, uint8_t, tmp, 16 * 16);
-  const int16_t *HFilter = vp9_sub_pel_filters_8lp[xoffset];
-  const int16_t *VFilter = vp9_sub_pel_filters_8lp[yoffset];
-
-  vp9_filter_block2d_16x16_8(src_ptr, src_pixels_per_line, HFilter, VFilter,
-                             tmp, 16);
-  block2d_average(tmp, 16, dst_ptr, dst_pitch, VPX_FILTER_16x16);
-}
-
-/****************************************************************************
- *
- *  ROUTINE       : filter_block2d_bil_first_pass
- *
- *  INPUTS        : uint8_t  *src_ptr    : Pointer to source block.
- *                  uint32_t  src_stride : Stride of source block.
- *                  uint32_t  height     : Block height.
- *                  uint32_t  width      : Block width.
- *                  int32_t  *vp9_filter : Array of 2 bi-linear filter taps.
- *
- *  OUTPUTS       : int32_t  *dst_ptr    : Pointer to filtered block.
- *
- *  RETURNS       : void
- *
- *  FUNCTION      : Applies a 1-D 2-tap bi-linear filter to the source block
- *                  in the horizontal direction to produce the filtered output
- *                  block. Used to implement first-pass of 2-D separable filter.
- *
- *  SPECIAL NOTES : Produces int32_t output to retain precision for next pass.
- *                  Two filter taps should sum to VP9_FILTER_WEIGHT.
- *
- ****************************************************************************/
-static void filter_block2d_bil_first_pass(uint8_t *src_ptr,
-                                          uint16_t *dst_ptr,
-                                          unsigned int src_stride,
-                                          unsigned int height,
-                                          unsigned int width,
-                                          const int16_t *vp9_filter) {
-  unsigned int i, j;
-
-  for (i = 0; i < height; i++) {
-    for (j = 0; j < width; j++) {
-      /* Apply bilinear filter */
-      dst_ptr[j] = (((int)src_ptr[0] * vp9_filter[0]) +
-                    ((int)src_ptr[1] * vp9_filter[1]) +
-                    (VP9_FILTER_WEIGHT / 2)) >> VP9_FILTER_SHIFT;
-      src_ptr++;
-    }
-
-    /* Next row... */
-    src_ptr += src_stride - width;
-    dst_ptr += width;
-  }
-}
-
-/****************************************************************************
- *
- *  ROUTINE       : filter_block2d_bil_second_pass
- *
- *  INPUTS        : int32_t  *src_ptr    : Pointer to source block.
- *                  uint32_t  dst_pitch  : Destination block pitch.
- *                  uint32_t  height     : Block height.
- *                  uint32_t  width      : Block width.
- *                  int32_t  *vp9_filter : Array of 2 bi-linear filter taps.
- *
- *  OUTPUTS       : uint16_t *dst_ptr    : Pointer to filtered block.
- *
- *  RETURNS       : void
- *
- *  FUNCTION      : Applies a 1-D 2-tap bi-linear filter to the source block
- *                  in the vertical direction to produce the filtered output
- *                  block. Used to implement second-pass of 2-D separable filter.
- *
- *  SPECIAL NOTES : Requires 32-bit input as produced by filter_block2d_bil_first_pass.
- *                  Two filter taps should sum to VP9_FILTER_WEIGHT.
- *
- ****************************************************************************/
-static void filter_block2d_bil_second_pass(uint16_t *src_ptr,
-                                           uint8_t *dst_ptr,
-                                           int dst_pitch,
-                                           unsigned int height,
-                                           unsigned int width,
-                                           const int16_t *vp9_filter) {
-  unsigned int i, j;
-  int temp;
-
-  for (i = 0; i < height; i++) {
-    for (j = 0; j < width; j++) {
-      /* Apply filter */
-      temp = ((int)src_ptr[0]     * vp9_filter[0]) +
-             ((int)src_ptr[width] * vp9_filter[1]) +
-             (VP9_FILTER_WEIGHT / 2);
-      dst_ptr[j] = (unsigned int)(temp >> VP9_FILTER_SHIFT);
-      src_ptr++;
-    }
-
-    /* Next row... */
-    dst_ptr += dst_pitch;
-  }
-}
-
-/*
- * As before for filter_block2d_second_pass_avg(), the functional difference
- * between filter_block2d_bil_second_pass() and filter_block2d_bil_second_pass_avg()
- * is that filter_block2d_bil_second_pass() does a bilinear filter on input
- * and stores the result in output; filter_block2d_bil_second_pass_avg(),
- * instead, does a bilinear filter on input, averages the resulting value
- * with the values already present in the output and stores the result of
- * that back into the output ((filter_result + dest + 1) >> 1).
- */
-static void filter_block2d_bil_second_pass_avg(uint16_t *src_ptr,
-                                               uint8_t *dst_ptr,
-                                               int dst_pitch,
-                                               unsigned int height,
-                                               unsigned int width,
-                                               const int16_t *vp9_filter) {
-  unsigned int i, j;
-  int temp;
-
-  for (i = 0; i < height; i++) {
-    for (j = 0; j < width; j++) {
-      /* Apply filter */
-      temp = (((int)src_ptr[0]     * vp9_filter[0]) +
-              ((int)src_ptr[width] * vp9_filter[1]) +
-              (VP9_FILTER_WEIGHT / 2)) >> VP9_FILTER_SHIFT;
-      dst_ptr[j] = (unsigned int)((temp + dst_ptr[j] + 1) >> 1);
-      src_ptr++;
-    }
-
-    /* Next row... */
-    dst_ptr += dst_pitch;
-  }
-}
-
-/****************************************************************************
- *
- *  ROUTINE       : filter_block2d_bil
- *
- *  INPUTS        : uint8_t  *src_ptr          : Pointer to source block.
- *                  uint32_t  src_pitch        : Stride of source block.
- *                  uint32_t  dst_pitch        : Stride of destination block.
- *                  int32_t  *HFilter          : Array of 2 horizontal filter taps.
- *                  int32_t  *VFilter          : Array of 2 vertical filter taps.
- *                  int32_t  Width             : Block width
- *                  int32_t  Height            : Block height
- *
- *  OUTPUTS       : uint16_t *dst_ptr       : Pointer to filtered block.
- *
- *  RETURNS       : void
- *
- *  FUNCTION      : 2-D filters an input block by applying a 2-tap
- *                  bi-linear filter horizontally followed by a 2-tap
- *                  bi-linear filter vertically on the result.
- *
- *  SPECIAL NOTES : The largest block size can be handled here is 16x16
- *
- ****************************************************************************/
-static void filter_block2d_bil(uint8_t *src_ptr,
-                               uint8_t *dst_ptr,
-                               unsigned int src_pitch,
-                               unsigned int dst_pitch,
-                               const int16_t *HFilter,
-                               const int16_t *VFilter,
-                               int Width,
-                               int Height) {
-
-  uint16_t FData[17 * 16];  /* Temp data buffer used in filtering */
-
-  /* First filter 1-D horizontally... */
-  filter_block2d_bil_first_pass(src_ptr, FData, src_pitch, Height + 1, Width, HFilter);
-
-  /* then 1-D vertically... */
-  filter_block2d_bil_second_pass(FData, dst_ptr, dst_pitch, Height, Width, VFilter);
-}
-
-static void filter_block2d_bil_avg(uint8_t *src_ptr,
-                                   uint8_t *dst_ptr,
-                                   unsigned int src_pitch,
-                                   unsigned int dst_pitch,
-                                   const int16_t *HFilter,
-                                   const int16_t *VFilter,
-                                   int Width,
-                                   int Height) {
-  uint16_t FData[17 * 16];  /* Temp data buffer used in filtering */
-
-  /* First filter 1-D horizontally... */
-  filter_block2d_bil_first_pass(src_ptr, FData, src_pitch, Height + 1, Width, HFilter);
-
-  /* then 1-D vertically... */
-  filter_block2d_bil_second_pass_avg(FData, dst_ptr, dst_pitch, Height, Width, VFilter);
-}
-
-void vp9_bilinear_predict4x4_c(uint8_t *src_ptr,
-                               int src_pixels_per_line,
-                               int xoffset,
-                               int yoffset,
-                               uint8_t *dst_ptr,
-                               int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_bilinear_filters[xoffset];
-  VFilter = vp9_bilinear_filters[yoffset];
-
-  filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 4, 4);
-}
-
-void vp9_bilinear_predict_avg4x4_c(uint8_t *src_ptr,
-                                   int src_pixels_per_line,
-                                   int xoffset,
-                                   int yoffset,
-                                   uint8_t *dst_ptr,
-                                   int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_bilinear_filters[xoffset];
-  VFilter = vp9_bilinear_filters[yoffset];
-
-  filter_block2d_bil_avg(src_ptr, dst_ptr, src_pixels_per_line,
-                         dst_pitch, HFilter, VFilter, 4, 4);
-}
-
-void vp9_bilinear_predict8x8_c(uint8_t *src_ptr,
-                               int src_pixels_per_line,
-                               int xoffset,
-                               int yoffset,
-                               uint8_t *dst_ptr,
-                               int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_bilinear_filters[xoffset];
-  VFilter = vp9_bilinear_filters[yoffset];
-
-  filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 8);
-
-}
-
-void vp9_bilinear_predict_avg8x8_c(uint8_t *src_ptr,
-                                   int src_pixels_per_line,
-                                   int xoffset,
-                                   int yoffset,
-                                   uint8_t *dst_ptr,
-                                   int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_bilinear_filters[xoffset];
-  VFilter = vp9_bilinear_filters[yoffset];
-
-  filter_block2d_bil_avg(src_ptr, dst_ptr, src_pixels_per_line,
-                         dst_pitch, HFilter, VFilter, 8, 8);
-}
-
-void vp9_bilinear_predict8x4_c(uint8_t *src_ptr,
-                               int src_pixels_per_line,
-                               int xoffset,
-                               int yoffset,
-                               uint8_t *dst_ptr,
-                               int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_bilinear_filters[xoffset];
-  VFilter = vp9_bilinear_filters[yoffset];
-
-  filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 8, 4);
-
-}
-
-void vp9_bilinear_predict16x16_c(uint8_t *src_ptr,
-                                 int src_pixels_per_line,
-                                 int xoffset,
-                                 int yoffset,
-                                 uint8_t *dst_ptr,
-                                 int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_bilinear_filters[xoffset];
-  VFilter = vp9_bilinear_filters[yoffset];
-
-  filter_block2d_bil(src_ptr, dst_ptr, src_pixels_per_line, dst_pitch, HFilter, VFilter, 16, 16);
-}
-
-void vp9_bilinear_predict_avg16x16_c(uint8_t *src_ptr,
-                                     int src_pixels_per_line,
-                                     int xoffset,
-                                     int yoffset,
-                                     uint8_t *dst_ptr,
-                                     int dst_pitch) {
-  const int16_t *HFilter;
-  const int16_t *VFilter;
-
-  HFilter = vp9_bilinear_filters[xoffset];
-  VFilter = vp9_bilinear_filters[yoffset];
-
-  filter_block2d_bil_avg(src_ptr, dst_ptr, src_pixels_per_line,
-                         dst_pitch, HFilter, VFilter, 16, 16);
-}
diff --git a/vp9/common/vp9_filter.h b/vp9/common/vp9_filter.h
index cd666578d..1ccfdaac2 100644
--- a/vp9/common/vp9_filter.h
+++ b/vp9/common/vp9_filter.h
@@ -21,10 +21,17 @@
 
 #define SUBPEL_SHIFTS 16
 
-extern const int16_t vp9_bilinear_filters[SUBPEL_SHIFTS][2];
-extern const int16_t vp9_sub_pel_filters_6[SUBPEL_SHIFTS][6];
+extern const int16_t vp9_bilinear_filters[SUBPEL_SHIFTS][8];
+extern const int16_t vp9_sub_pel_filters_6[SUBPEL_SHIFTS][8];
 extern const int16_t vp9_sub_pel_filters_8[SUBPEL_SHIFTS][8];
 extern const int16_t vp9_sub_pel_filters_8s[SUBPEL_SHIFTS][8];
 extern const int16_t vp9_sub_pel_filters_8lp[SUBPEL_SHIFTS][8];
 
+// The VP9_BILINEAR_FILTERS_2TAP macro returns a pointer to the bilinear
+// filter kernel as a 2 tap filter.
+#define BF_LENGTH (sizeof(vp9_bilinear_filters[0]) / \
+                   sizeof(vp9_bilinear_filters[0][0]))
+#define BF_OFFSET (BF_LENGTH / 2 - 1)
+#define VP9_BILINEAR_FILTERS_2TAP(x) (vp9_bilinear_filters[x] + BF_OFFSET)
+
 #endif  // VP9_COMMON_VP9_FILTER_H_
diff --git a/vp9/common/vp9_findnearmv.c b/vp9/common/vp9_findnearmv.c
index 9585611dd..f713212a7 100644
--- a/vp9/common/vp9_findnearmv.c
+++ b/vp9/common/vp9_findnearmv.c
@@ -87,8 +87,8 @@ unsigned int vp9_sub_pixel_variance16x2_c(const uint8_t *src_ptr,
   uint8_t temp2[2 * 16];
   const int16_t *HFilter, *VFilter;
 
-  HFilter = vp9_bilinear_filters[xoffset];
-  VFilter = vp9_bilinear_filters[yoffset];
+  HFilter = VP9_BILINEAR_FILTERS_2TAP(xoffset);
+  VFilter = VP9_BILINEAR_FILTERS_2TAP(yoffset);
 
   var_filter_block2d_bil_first_pass(src_ptr, FData3,
                                     src_pixels_per_line, 1, 3, 16, HFilter);
@@ -108,8 +108,8 @@ unsigned int vp9_sub_pixel_variance2x16_c(const uint8_t *src_ptr,
   uint8_t temp2[2 * 16];
   const int16_t *HFilter, *VFilter;
 
-  HFilter = vp9_bilinear_filters[xoffset];
-  VFilter = vp9_bilinear_filters[yoffset];
+  HFilter = VP9_BILINEAR_FILTERS_2TAP(xoffset);
+  VFilter = VP9_BILINEAR_FILTERS_2TAP(yoffset);
 
   var_filter_block2d_bil_first_pass(src_ptr, FData3,
                                     src_pixels_per_line, 1, 17, 2, HFilter);
diff --git a/vp9/common/vp9_reconinter.c b/vp9/common/vp9_reconinter.c
index 20de7b7f1..d4435d872 100644
--- a/vp9/common/vp9_reconinter.c
+++ b/vp9/common/vp9_reconinter.c
@@ -8,66 +8,58 @@
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
+#include <assert.h>
 
 #include "./vpx_config.h"
 #include "vpx/vpx_integer.h"
 #include "vp9/common/vp9_blockd.h"
+#include "vp9/common/vp9_filter.h"
 #include "vp9/common/vp9_reconinter.h"
 #include "vp9/common/vp9_reconintra.h"
 
 void vp9_setup_interp_filters(MACROBLOCKD *xd,
                               INTERPOLATIONFILTERTYPE mcomp_filter_type,
                               VP9_COMMON *cm) {
+  // TODO(agrange): Investigate the best choice of functions to use here
+  // for EIGHTTAP_SMOOTH. Since it is not interpolating, need to choose what
+  // to do at full-pel offsets. The current selection, where the filter is
+  // applied in one direction only, and not at all for 0,0, seems to give the
+  // best quality, but it may be worth trying an additional mode that does
+  // do the filtering on full-pel.
+  xd->subpix.predict[0][0][0] = vp9_convolve_copy;
+  xd->subpix.predict[0][0][1] = vp9_convolve_avg;
+  xd->subpix.predict[0][1][0] = vp9_convolve8_vert;
+  xd->subpix.predict[0][1][1] = vp9_convolve8_avg_vert;
+  xd->subpix.predict[1][0][0] = vp9_convolve8_horiz;
+  xd->subpix.predict[1][0][1] = vp9_convolve8_avg_horiz;
+  xd->subpix.predict[1][1][0] = vp9_convolve8;
+  xd->subpix.predict[1][1][1] = vp9_convolve8_avg;
+
+  xd->subpix.x_step_q4 = 16;
+  xd->subpix.y_step_q4 = 16;
+  switch (mcomp_filter_type) {
+    case EIGHTTAP:
+    case SWITCHABLE:
+      xd->subpix.filter_x = xd->subpix.filter_y = vp9_sub_pel_filters_8;
+      break;
+    case EIGHTTAP_SMOOTH:
+      xd->subpix.filter_x = xd->subpix.filter_y = vp9_sub_pel_filters_8lp;
+      break;
+    case EIGHTTAP_SHARP:
+      xd->subpix.filter_x = xd->subpix.filter_y = vp9_sub_pel_filters_8s;
+      break;
+    case BILINEAR:
+      xd->subpix.filter_x = xd->subpix.filter_y = vp9_bilinear_filters;
+      break;
 #if CONFIG_ENABLE_6TAP
-  if (mcomp_filter_type == SIXTAP) {
-    xd->subpixel_predict4x4     = vp9_sixtap_predict4x4;
-    xd->subpixel_predict8x4     = vp9_sixtap_predict8x4;
-    xd->subpixel_predict8x8     = vp9_sixtap_predict8x8;
-    xd->subpixel_predict16x16   = vp9_sixtap_predict16x16;
-    xd->subpixel_predict_avg4x4 = vp9_sixtap_predict_avg4x4;
-    xd->subpixel_predict_avg8x8 = vp9_sixtap_predict_avg8x8;
-    xd->subpixel_predict_avg16x16 = vp9_sixtap_predict_avg16x16;
-  } else {
+    case SIXTAP:
+      xd->subpix.filter_x = xd->subpix.filter_y = vp9_sub_pel_filters_6;
+      break;
 #endif
-  if (mcomp_filter_type == EIGHTTAP || mcomp_filter_type == SWITCHABLE) {
-    xd->subpixel_predict4x4     = vp9_eighttap_predict4x4;
-    xd->subpixel_predict8x4     = vp9_eighttap_predict8x4;
-    xd->subpixel_predict8x8     = vp9_eighttap_predict8x8;
-    xd->subpixel_predict16x16   = vp9_eighttap_predict16x16;
-    xd->subpixel_predict_avg4x4 = vp9_eighttap_predict_avg4x4;
-    xd->subpixel_predict_avg8x8 = vp9_eighttap_predict_avg8x8;
-    xd->subpixel_predict_avg16x16 = vp9_eighttap_predict_avg16x16;
-  } else if (mcomp_filter_type == EIGHTTAP_SMOOTH) {
-    xd->subpixel_predict4x4     = vp9_eighttap_predict4x4_smooth;
-    xd->subpixel_predict8x4     = vp9_eighttap_predict8x4_smooth;
-    xd->subpixel_predict8x8     = vp9_eighttap_predict8x8_smooth;
-    xd->subpixel_predict16x16   = vp9_eighttap_predict16x16_smooth;
-    xd->subpixel_predict_avg4x4 = vp9_eighttap_predict_avg4x4_smooth;
-    xd->subpixel_predict_avg8x8 = vp9_eighttap_predict_avg8x8_smooth;
-    xd->subpixel_predict_avg16x16 = vp9_eighttap_predict_avg16x16_smooth;
-  } else if (mcomp_filter_type == EIGHTTAP_SHARP) {
-    xd->subpixel_predict4x4     = vp9_eighttap_predict4x4_sharp;
-    xd->subpixel_predict8x4     = vp9_eighttap_predict8x4_sharp;
-    xd->subpixel_predict8x8     = vp9_eighttap_predict8x8_sharp;
-    xd->subpixel_predict16x16   = vp9_eighttap_predict16x16_sharp;
-    xd->subpixel_predict_avg4x4 = vp9_eighttap_predict_avg4x4_sharp;
-    xd->subpixel_predict_avg8x8 = vp9_eighttap_predict_avg8x8_sharp;
-    xd->subpixel_predict_avg16x16 = vp9_eighttap_predict_avg16x16_sharp_c;
-  } else {
-    xd->subpixel_predict4x4     = vp9_bilinear_predict4x4;
-    xd->subpixel_predict8x4     = vp9_bilinear_predict8x4;
-    xd->subpixel_predict8x8     = vp9_bilinear_predict8x8;
-    xd->subpixel_predict16x16   = vp9_bilinear_predict16x16;
-    xd->subpixel_predict_avg4x4 = vp9_bilinear_predict_avg4x4;
-    xd->subpixel_predict_avg8x8 = vp9_bilinear_predict_avg8x8;
-    xd->subpixel_predict_avg16x16 = vp9_bilinear_predict_avg16x16;
-  }
-#if CONFIG_ENABLE_6TAP
   }
-#endif
 }
 
-void vp9_copy_mem16x16_c(uint8_t *src,
+void vp9_copy_mem16x16_c(const uint8_t *src,
                          int src_stride,
                          uint8_t *dst,
                          int dst_stride) {
@@ -93,10 +85,10 @@ void vp9_copy_mem16x16_c(uint8_t *src,
     dst[15] = src[15];
 
 #else
-    ((uint32_t *)dst)[0] = ((uint32_t *)src)[0];
-    ((uint32_t *)dst)[1] = ((uint32_t *)src)[1];
-    ((uint32_t *)dst)[2] = ((uint32_t *)src)[2];
-    ((uint32_t *)dst)[3] = ((uint32_t *)src)[3];
+    ((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;
@@ -104,25 +96,7 @@ void vp9_copy_mem16x16_c(uint8_t *src,
   }
 }
 
-void vp9_avg_mem16x16_c(uint8_t *src,
-                        int src_stride,
-                        uint8_t *dst,
-                        int dst_stride) {
-  int r;
-
-  for (r = 0; r < 16; r++) {
-    int n;
-
-    for (n = 0; n < 16; n++) {
-      dst[n] = (dst[n] + src[n] + 1) >> 1;
-    }
-
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-void vp9_copy_mem8x8_c(uint8_t *src,
+void vp9_copy_mem8x8_c(const uint8_t *src,
                        int src_stride,
                        uint8_t *dst,
                        int dst_stride) {
@@ -139,33 +113,15 @@ void vp9_copy_mem8x8_c(uint8_t *src,
     dst[6] = src[6];
     dst[7] = src[7];
 #else
-    ((uint32_t *)dst)[0] = ((uint32_t *)src)[0];
-    ((uint32_t *)dst)[1] = ((uint32_t *)src)[1];
+    ((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_avg_mem8x8_c(uint8_t *src,
-                      int src_stride,
-                      uint8_t *dst,
-                      int dst_stride) {
-  int r;
-
-  for (r = 0; r < 8; r++) {
-    int n;
-
-    for (n = 0; n < 8; n++) {
-      dst[n] = (dst[n] + src[n] + 1) >> 1;
-    }
-
-    src += src_stride;
-    dst += dst_stride;
-  }
-}
-
-void vp9_copy_mem8x4_c(uint8_t *src,
+void vp9_copy_mem8x4_c(const uint8_t *src,
                        int src_stride,
                        uint8_t *dst,
                        int dst_stride) {
@@ -182,16 +138,16 @@ void vp9_copy_mem8x4_c(uint8_t *src,
     dst[6] = src[6];
     dst[7] = src[7];
 #else
-    ((uint32_t *)dst)[0] = ((uint32_t *)src)[0];
-    ((uint32_t *)dst)[1] = ((uint32_t *)src)[1];
+    ((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_predictors_b(BLOCKD *d, int pitch, vp9_subpix_fn_t sppf) {
-  int r;
+void vp9_build_inter_predictors_b(BLOCKD *d, int pitch,
+                                  struct subpix_fn_table *subpix) {
   uint8_t *ptr_base;
   uint8_t *ptr;
   uint8_t *pred_ptr = d->predictor;
@@ -199,30 +155,14 @@ void vp9_build_inter_predictors_b(BLOCKD *d, int pitch, vp9_subpix_fn_t sppf) {
 
   ptr_base = *(d->base_pre);
   mv.as_int = d->bmi.as_mv.first.as_int;
+  ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride +
+        (mv.as_mv.col >> 3);
 
-  if (mv.as_mv.row & 7 || mv.as_mv.col & 7) {
-    ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride +
-          (mv.as_mv.col >> 3);
-    sppf(ptr, d->pre_stride, (mv.as_mv.col & 7) << 1, (mv.as_mv.row & 7) << 1,
-         pred_ptr, pitch);
-  } else {
-    ptr_base += d->pre + (mv.as_mv.row >> 3) * d->pre_stride +
-                (mv.as_mv.col >> 3);
-    ptr = ptr_base;
-
-    for (r = 0; r < 4; r++) {
-#if !(CONFIG_FAST_UNALIGNED)
-      pred_ptr[0]  = ptr[0];
-      pred_ptr[1]  = ptr[1];
-      pred_ptr[2]  = ptr[2];
-      pred_ptr[3]  = ptr[3];
-#else
-      *(uint32_t *)pred_ptr = *(uint32_t *)ptr;
-#endif
-      pred_ptr     += pitch;
-      ptr         += d->pre_stride;
-    }
-  }
+  subpix->predict[!!(mv.as_mv.col & 7)][!!(mv.as_mv.row & 7)][0](
+      ptr, d->pre_stride, pred_ptr, pitch,
+      subpix->filter_x[(mv.as_mv.col & 7) << 1], subpix->x_step_q4,
+      subpix->filter_y[(mv.as_mv.row & 7) << 1], subpix->y_step_q4,
+      4, 4);
 }
 
 /*
@@ -232,8 +172,7 @@ void vp9_build_inter_predictors_b(BLOCKD *d, int pitch, vp9_subpix_fn_t sppf) {
  * predictor of the second reference frame / motion vector.
  */
 void vp9_build_2nd_inter_predictors_b(BLOCKD *d, int pitch,
-                                      vp9_subpix_fn_t sppf) {
-  int r;
+                                      struct subpix_fn_table *subpix) {
   uint8_t *ptr_base;
   uint8_t *ptr;
   uint8_t *pred_ptr = d->predictor;
@@ -241,26 +180,14 @@ void vp9_build_2nd_inter_predictors_b(BLOCKD *d, int pitch,
 
   ptr_base = *(d->base_second_pre);
   mv.as_int = d->bmi.as_mv.second.as_int;
+  ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride +
+        (mv.as_mv.col >> 3);
 
-  if (mv.as_mv.row & 7 || mv.as_mv.col & 7) {
-    ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride +
-          (mv.as_mv.col >> 3);
-    sppf(ptr, d->pre_stride, (mv.as_mv.col & 7) << 1, (mv.as_mv.row & 7) << 1,
-         pred_ptr, pitch);
-  } else {
-    ptr_base += d->pre + (mv.as_mv.row >> 3) * d->pre_stride +
-                (mv.as_mv.col >> 3);
-    ptr = ptr_base;
-
-    for (r = 0; r < 4; r++) {
-      pred_ptr[0]  = (pred_ptr[0] + ptr[0] + 1) >> 1;
-      pred_ptr[1]  = (pred_ptr[1] + ptr[1] + 1) >> 1;
-      pred_ptr[2]  = (pred_ptr[2] + ptr[2] + 1) >> 1;
-      pred_ptr[3]  = (pred_ptr[3] + ptr[3] + 1) >> 1;
-      pred_ptr    += pitch;
-      ptr         += d->pre_stride;
-    }
-  }
+  subpix->predict[!!(mv.as_mv.col & 7)][!!(mv.as_mv.row & 7)][1](
+      ptr, d->pre_stride, pred_ptr, pitch,
+      subpix->filter_x[(mv.as_mv.col & 7) << 1], subpix->x_step_q4,
+      subpix->filter_y[(mv.as_mv.row & 7) << 1], subpix->y_step_q4,
+      4, 4);
 }
 
 void vp9_build_inter_predictors4b(MACROBLOCKD *xd, BLOCKD *d, int pitch) {
@@ -274,12 +201,11 @@ void vp9_build_inter_predictors4b(MACROBLOCKD *xd, BLOCKD *d, int pitch) {
   ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride +
         (mv.as_mv.col >> 3);
 
-  if (mv.as_mv.row & 7 || mv.as_mv.col & 7) {
-    xd->subpixel_predict8x8(ptr, d->pre_stride, (mv.as_mv.col & 7) << 1,
-                            (mv.as_mv.row & 7) << 1, pred_ptr, pitch);
-  } else {
-    vp9_copy_mem8x8(ptr, d->pre_stride, pred_ptr, pitch);
-  }
+  xd->subpix.predict[!!(mv.as_mv.col & 7)][!!(mv.as_mv.row & 7)][0](
+      ptr, d->pre_stride, pred_ptr, pitch,
+      xd->subpix.filter_x[(mv.as_mv.col & 7) << 1], xd->subpix.x_step_q4,
+      xd->subpix.filter_y[(mv.as_mv.row & 7) << 1], xd->subpix.y_step_q4,
+      8, 8);
 }
 
 /*
@@ -300,12 +226,11 @@ void vp9_build_2nd_inter_predictors4b(MACROBLOCKD *xd,
   ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride +
         (mv.as_mv.col >> 3);
 
-  if (mv.as_mv.row & 7 || mv.as_mv.col & 7) {
-    xd->subpixel_predict_avg8x8(ptr, d->pre_stride, (mv.as_mv.col & 7) << 1,
-                               (mv.as_mv.row & 7) << 1, pred_ptr, pitch);
-  } else {
-    vp9_avg_mem8x8(ptr, d->pre_stride, pred_ptr, pitch);
-  }
+  xd->subpix.predict[!!(mv.as_mv.col & 7)][!!(mv.as_mv.row & 7)][1](
+      ptr, d->pre_stride, pred_ptr, pitch,
+      xd->subpix.filter_x[(mv.as_mv.col & 7) << 1], xd->subpix.x_step_q4,
+      xd->subpix.filter_y[(mv.as_mv.row & 7) << 1], xd->subpix.y_step_q4,
+      8, 8);
 }
 
 static void build_inter_predictors2b(MACROBLOCKD *xd, BLOCKD *d, int pitch) {
@@ -319,12 +244,11 @@ static void build_inter_predictors2b(MACROBLOCKD *xd, BLOCKD *d, int pitch) {
   ptr = ptr_base + d->pre + (mv.as_mv.row >> 3) * d->pre_stride +
         (mv.as_mv.col >> 3);
 
-  if (mv.as_mv.row & 7 || mv.as_mv.col & 7) {
-    xd->subpixel_predict8x4(ptr, d->pre_stride, (mv.as_mv.col & 7) << 1,
-                           (mv.as_mv.row & 7) << 1, pred_ptr, pitch);
-  } else {
-    vp9_copy_mem8x4(ptr, d->pre_stride, pred_ptr, pitch);
-  }
+  xd->subpix.predict[!!(mv.as_mv.col & 7)][!!(mv.as_mv.row & 7)][0](
+      ptr, d->pre_stride, pred_ptr, pitch,
+      xd->subpix.filter_x[(mv.as_mv.col & 7) << 1], xd->subpix.x_step_q4,
+      xd->subpix.filter_y[(mv.as_mv.row & 7) << 1], xd->subpix.y_step_q4,
+      8, 4);
 }
 
 /*encoder only*/
@@ -411,13 +335,13 @@ void vp9_build_inter4x4_predictors_mbuv(MACROBLOCKD *xd) {
     if (d0->bmi.as_mv.first.as_int == d1->bmi.as_mv.first.as_int)
       build_inter_predictors2b(xd, d0, 8);
     else {
-      vp9_build_inter_predictors_b(d0, 8, xd->subpixel_predict4x4);
-      vp9_build_inter_predictors_b(d1, 8, xd->subpixel_predict4x4);
+      vp9_build_inter_predictors_b(d0, 8, &xd->subpix);
+      vp9_build_inter_predictors_b(d1, 8, &xd->subpix);
     }
 
     if (xd->mode_info_context->mbmi.second_ref_frame > 0) {
-      vp9_build_2nd_inter_predictors_b(d0, 8, xd->subpixel_predict_avg4x4);
-      vp9_build_2nd_inter_predictors_b(d1, 8, xd->subpixel_predict_avg4x4);
+      vp9_build_2nd_inter_predictors_b(d0, 8, &xd->subpix);
+      vp9_build_2nd_inter_predictors_b(d1, 8, &xd->subpix);
     }
   }
 }
@@ -475,14 +399,11 @@ void vp9_build_1st_inter16x16_predictors_mby(MACROBLOCKD *xd,
 
   ptr = ptr_base + (ymv.as_mv.row >> 3) * pre_stride + (ymv.as_mv.col >> 3);
 
-    if ((ymv.as_mv.row | ymv.as_mv.col) & 7) {
-      xd->subpixel_predict16x16(ptr, pre_stride,
-                                (ymv.as_mv.col & 7) << 1,
-                                (ymv.as_mv.row & 7) << 1,
-                                dst_y, dst_ystride);
-    } else {
-      vp9_copy_mem16x16(ptr, pre_stride, dst_y, dst_ystride);
-    }
+  xd->subpix.predict[!!(ymv.as_mv.col & 7)][!!(ymv.as_mv.row & 7)][0](
+      ptr, pre_stride, dst_y, dst_ystride,
+      xd->subpix.filter_x[(ymv.as_mv.col & 7) << 1], xd->subpix.x_step_q4,
+      xd->subpix.filter_y[(ymv.as_mv.row & 7) << 1], xd->subpix.y_step_q4,
+      16, 16);
 }
 
 void vp9_build_1st_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
@@ -523,15 +444,19 @@ void vp9_build_1st_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
   uptr = xd->pre.u_buffer + offset;
   vptr = xd->pre.v_buffer + offset;
 
-    if (_o16x16mv.as_int & 0x000f000f) {
-      xd->subpixel_predict8x8(uptr, pre_stride, _o16x16mv.as_mv.col & 15,
-                              _o16x16mv.as_mv.row & 15, dst_u, dst_uvstride);
-      xd->subpixel_predict8x8(vptr, pre_stride, _o16x16mv.as_mv.col & 15,
-                              _o16x16mv.as_mv.row & 15, dst_v, dst_uvstride);
-    } else {
-      vp9_copy_mem8x8(uptr, pre_stride, dst_u, dst_uvstride);
-      vp9_copy_mem8x8(vptr, pre_stride, dst_v, dst_uvstride);
-    }
+  xd->subpix.predict[!!(_o16x16mv.as_mv.col & 15)]
+                    [!!(_o16x16mv.as_mv.row & 15)][0](
+      uptr, pre_stride, dst_u, dst_uvstride,
+      xd->subpix.filter_x[_o16x16mv.as_mv.col & 15], xd->subpix.x_step_q4,
+      xd->subpix.filter_y[_o16x16mv.as_mv.row & 15], xd->subpix.y_step_q4,
+      8, 8);
+
+  xd->subpix.predict[!!(_o16x16mv.as_mv.col & 15)]
+                    [!!(_o16x16mv.as_mv.row & 15)][0](
+      vptr, pre_stride, dst_v, dst_uvstride,
+      xd->subpix.filter_x[_o16x16mv.as_mv.col & 15], xd->subpix.x_step_q4,
+      xd->subpix.filter_y[_o16x16mv.as_mv.row & 15], xd->subpix.y_step_q4,
+      8, 8);
 }
 
 
@@ -714,12 +639,11 @@ void vp9_build_2nd_inter16x16_predictors_mby(MACROBLOCKD *xd,
 
   ptr = ptr_base + (mv_row >> 3) * pre_stride + (mv_col >> 3);
 
-  if ((mv_row | mv_col) & 7) {
-    xd->subpixel_predict_avg16x16(ptr, pre_stride, (mv_col & 7) << 1,
-                                  (mv_row & 7) << 1, dst_y, dst_ystride);
-  } else {
-    vp9_avg_mem16x16(ptr, pre_stride, dst_y, dst_ystride);
-  }
+  xd->subpix.predict[!!(mv_col & 7)][!!(mv_row & 7)][1](
+      ptr, pre_stride, dst_y, dst_ystride,
+      xd->subpix.filter_x[(mv_col & 7) << 1], xd->subpix.x_step_q4,
+      xd->subpix.filter_y[(mv_row & 7) << 1], xd->subpix.y_step_q4,
+      16, 16);
 }
 
 void vp9_build_2nd_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
@@ -758,15 +682,17 @@ void vp9_build_2nd_inter16x16_predictors_mbuv(MACROBLOCKD *xd,
   uptr = xd->second_pre.u_buffer + offset;
   vptr = xd->second_pre.v_buffer + offset;
 
-    if ((omv_row | omv_col) & 15) {
-      xd->subpixel_predict_avg8x8(uptr, pre_stride, omv_col & 15,
-                                  omv_row & 15, dst_u, dst_uvstride);
-      xd->subpixel_predict_avg8x8(vptr, pre_stride, omv_col & 15,
-                                  omv_row & 15, dst_v, dst_uvstride);
-    } else {
-      vp9_avg_mem8x8(uptr, pre_stride, dst_u, dst_uvstride);
-      vp9_avg_mem8x8(vptr, pre_stride, dst_v, dst_uvstride);
-    }
+  xd->subpix.predict[!!(omv_col & 15)][!!(omv_row & 15)][1](
+      uptr, pre_stride, dst_u, dst_uvstride,
+      xd->subpix.filter_x[omv_col & 15], xd->subpix.x_step_q4,
+      xd->subpix.filter_y[omv_row & 15], xd->subpix.y_step_q4,
+      8, 8);
+
+  xd->subpix.predict[!!(omv_col & 15)][!!(omv_row & 15)][1](
+      vptr, pre_stride, dst_v, dst_uvstride,
+      xd->subpix.filter_x[omv_col & 15], xd->subpix.x_step_q4,
+      xd->subpix.filter_y[omv_row & 15], xd->subpix.y_step_q4,
+      8, 8);
 }
 
 void vp9_build_2nd_inter16x16_predictors_mb(MACROBLOCKD *xd,
@@ -835,13 +761,13 @@ static void build_inter4x4_predictors_mb(MACROBLOCKD *xd) {
       if (d0->bmi.as_mv.first.as_int == d1->bmi.as_mv.first.as_int)
         build_inter_predictors2b(xd, d0, 16);
       else {
-        vp9_build_inter_predictors_b(d0, 16, xd->subpixel_predict4x4);
-        vp9_build_inter_predictors_b(d1, 16, xd->subpixel_predict4x4);
+        vp9_build_inter_predictors_b(d0, 16, &xd->subpix);
+        vp9_build_inter_predictors_b(d1, 16, &xd->subpix);
       }
 
       if (mbmi->second_ref_frame > 0) {
-        vp9_build_2nd_inter_predictors_b(d0, 16, xd->subpixel_predict_avg4x4);
-        vp9_build_2nd_inter_predictors_b(d1, 16, xd->subpixel_predict_avg4x4);
+        vp9_build_2nd_inter_predictors_b(d0, 16, &xd->subpix);
+        vp9_build_2nd_inter_predictors_b(d1, 16, &xd->subpix);
       }
     }
   }
@@ -853,13 +779,13 @@ static void build_inter4x4_predictors_mb(MACROBLOCKD *xd) {
     if (d0->bmi.as_mv.first.as_int == d1->bmi.as_mv.first.as_int)
       build_inter_predictors2b(xd, d0, 8);
     else {
-      vp9_build_inter_predictors_b(d0, 8, xd->subpixel_predict4x4);
-      vp9_build_inter_predictors_b(d1, 8, xd->subpixel_predict4x4);
+      vp9_build_inter_predictors_b(d0, 8, &xd->subpix);
+      vp9_build_inter_predictors_b(d1, 8, &xd->subpix);
     }
 
     if (mbmi->second_ref_frame > 0) {
-      vp9_build_2nd_inter_predictors_b(d0, 8, xd->subpixel_predict_avg4x4);
-      vp9_build_2nd_inter_predictors_b(d1, 8, xd->subpixel_predict_avg4x4);
+      vp9_build_2nd_inter_predictors_b(d0, 8, &xd->subpix);
+      vp9_build_2nd_inter_predictors_b(d1, 8, &xd->subpix);
     }
   }
 }
diff --git a/vp9/common/vp9_reconinter.h b/vp9/common/vp9_reconinter.h
index 89868b95e..903bd2e86 100644
--- a/vp9/common/vp9_reconinter.h
+++ b/vp9/common/vp9_reconinter.h
@@ -14,6 +14,8 @@
 #include "vpx/vpx_integer.h"
 #include "vp9/common/vp9_onyxc_int.h"
 
+struct subpix_fn_table;
+
 extern void vp9_build_1st_inter16x16_predictors_mby(MACROBLOCKD *xd,
                                                     uint8_t *dst_y,
                                                     int dst_ystride,
@@ -64,10 +66,10 @@ extern void vp9_build_inter64x64_predictors_sb(MACROBLOCKD *x,
 extern void vp9_build_inter_predictors_mb(MACROBLOCKD *xd);
 
 extern void vp9_build_inter_predictors_b(BLOCKD *d, int pitch,
-                                         vp9_subpix_fn_t sppf);
+                                         struct subpix_fn_table *sppf);
 
 extern void vp9_build_2nd_inter_predictors_b(BLOCKD *d, int pitch,
-                                             vp9_subpix_fn_t sppf);
+                                             struct subpix_fn_table *sppf);
 
 extern void vp9_build_inter_predictors4b(MACROBLOCKD *xd, BLOCKD *d,
                                          int pitch);
diff --git a/vp9/common/vp9_rtcd_defs.sh b/vp9/common/vp9_rtcd_defs.sh
index 5e4d485b5..9259f7d99 100644
--- a/vp9/common/vp9_rtcd_defs.sh
+++ b/vp9/common/vp9_rtcd_defs.sh
@@ -23,21 +23,6 @@ EOF
 }
 forward_decls vp9_common_forward_decls
 
-prototype void vp9_filter_block2d_4x4_8 "const uint8_t *src_ptr, const unsigned int src_stride, const int16_t *HFilter_aligned16, const int16_t *VFilter_aligned16, uint8_t *dst_ptr, unsigned int dst_stride"
-prototype void vp9_filter_block2d_8x4_8 "const uint8_t *src_ptr, const unsigned int src_stride, const int16_t *HFilter_aligned16, const int16_t *VFilter_aligned16, uint8_t *dst_ptr, unsigned int dst_stride"
-prototype void vp9_filter_block2d_8x8_8 "const uint8_t *src_ptr, const unsigned int src_stride, const int16_t *HFilter_aligned16, const int16_t *VFilter_aligned16, uint8_t *dst_ptr, unsigned int dst_stride"
-prototype void vp9_filter_block2d_16x16_8 "const uint8_t *src_ptr, const unsigned int src_stride, const int16_t *HFilter_aligned16, const int16_t *VFilter_aligned16, uint8_t *dst_ptr, unsigned int dst_stride"
-
-# At the very least, MSVC 2008 has compiler bug exhibited by this code; code
-# compiles warning free but a dissassembly of generated code show bugs. To be
-# on the safe side, only enabled when compiled with 'gcc'.
-if [ "$CONFIG_GCC" = "yes" ]; then
-    specialize vp9_filter_block2d_4x4_8 sse4_1 sse2
-fi
-    specialize vp9_filter_block2d_8x4_8 ssse3 #sse4_1 sse2
-    specialize vp9_filter_block2d_8x8_8 ssse3 #sse4_1 sse2
-    specialize vp9_filter_block2d_16x16_8 ssse3 #sse4_1 sse2
-
 #
 # Dequant
 #
@@ -86,27 +71,17 @@ specialize vp9_dequant_idct_add_uv_block_16x16
 #
 # RECON
 #
-prototype void vp9_copy_mem16x16 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
+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 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
+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 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
+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_avg_mem16x16 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
-specialize vp9_avg_mem16x16
-
-prototype void vp9_avg_mem8x8 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
-specialize vp9_avg_mem8x8
-
-prototype void vp9_copy_mem8x4 "uint8_t *src, int src_pitch, uint8_t *dst, int dst_pitch"
-specialize vp9_copy_mem8x4 mmx dspr2
-vp9_copy_mem8x4_dspr2=vp9_copy_mem8x4_dspr2
-
 prototype void vp9_recon_b "uint8_t *pred_ptr, int16_t *diff_ptr, uint8_t *dst_ptr, int stride"
 specialize vp9_recon_b
 
@@ -269,110 +244,23 @@ specialize vp9_sub_pixel_variance16x2 sse2
 #
 # Sub Pixel Filters
 #
-prototype void vp9_eighttap_predict16x16 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict16x16
-
-prototype void vp9_eighttap_predict8x8 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict8x8
-
-prototype void vp9_eighttap_predict_avg16x16 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict_avg16x16
-
-prototype void vp9_eighttap_predict_avg8x8 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict_avg8x8
-
-prototype void vp9_eighttap_predict_avg4x4 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict_avg4x4
-
-prototype void vp9_eighttap_predict8x4 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict8x4
-
-prototype void vp9_eighttap_predict4x4 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict4x4
-
-prototype void vp9_eighttap_predict16x16_sharp "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict16x16_sharp
-
-prototype void vp9_eighttap_predict8x8_sharp "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict8x8_sharp
-
-prototype void vp9_eighttap_predict_avg16x16_sharp "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict_avg16x16_sharp
-
-prototype void vp9_eighttap_predict_avg8x8_sharp "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict_avg8x8_sharp
-
-prototype void vp9_eighttap_predict_avg4x4_sharp "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict_avg4x4_sharp
-
-prototype void vp9_eighttap_predict8x4_sharp "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict8x4_sharp
-
-prototype void vp9_eighttap_predict4x4_sharp "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict4x4_sharp
-
-prototype void vp9_eighttap_predict16x16_smooth "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict16x16_smooth
-
-prototype void vp9_eighttap_predict8x8_smooth "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict8x8_smooth
-
-prototype void vp9_eighttap_predict_avg16x16_smooth "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict_avg16x16_smooth
-
-prototype void vp9_eighttap_predict_avg8x8_smooth "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict_avg8x8_smooth
-
-prototype void vp9_eighttap_predict_avg4x4_smooth "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict_avg4x4_smooth
-
-prototype void vp9_eighttap_predict8x4_smooth "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict8x4_smooth
-
-prototype void vp9_eighttap_predict4x4_smooth "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_eighttap_predict4x4_smooth
-
-prototype void vp9_sixtap_predict16x16 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_sixtap_predict16x16
-
-prototype void vp9_sixtap_predict8x8 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_sixtap_predict8x8
-
-prototype void vp9_sixtap_predict_avg16x16 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_sixtap_predict_avg16x16
-
-prototype void vp9_sixtap_predict_avg8x8 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_sixtap_predict_avg8x8
-
-prototype void vp9_sixtap_predict8x4 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_sixtap_predict8x4
-
-prototype void vp9_sixtap_predict4x4 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_sixtap_predict4x4
-
-prototype void vp9_sixtap_predict_avg4x4 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_sixtap_predict_avg4x4
-
-prototype void vp9_bilinear_predict16x16 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_bilinear_predict16x16 sse2
-
-prototype void vp9_bilinear_predict8x8 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_bilinear_predict8x8 sse2
+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"
+specialize vp9_convolve8 ssse3
 
-prototype void vp9_bilinear_predict_avg16x16 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_bilinear_predict_avg16x16
+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"
+specialize vp9_convolve8_horiz ssse3
 
-prototype void vp9_bilinear_predict_avg8x8 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_bilinear_predict_avg8x8
+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"
+specialize vp9_convolve8_vert ssse3
 
-prototype void vp9_bilinear_predict8x4 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_bilinear_predict8x4
+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"
+specialize vp9_convolve8_avg
 
-prototype void vp9_bilinear_predict4x4 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_bilinear_predict4x4
+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"
+specialize vp9_convolve8_avg_horiz
 
-prototype void vp9_bilinear_predict_avg4x4 "uint8_t *src_ptr, int  src_pixels_per_line, int  xoffset, int  yoffset, uint8_t *dst_ptr, int  dst_pitch"
-specialize vp9_bilinear_predict_avg4x4
+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"
+specialize vp9_convolve8_avg_vert
 
 #
 # dct
diff --git a/vp9/common/vp9_subpixel.h b/vp9/common/vp9_subpixel.h
deleted file mode 100644
index dc4eadfb1..000000000
--- a/vp9/common/vp9_subpixel.h
+++ /dev/null
@@ -1,20 +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.
- */
-
-#ifndef VP9_COMMON_VP9_SUBPIXEL_H_
-#define VP9_COMMON_VP9_SUBPIXEL_H_
-
-#define prototype_subpixel_predict(sym) \
-  void sym(uint8_t *src, int src_pitch, int xofst, int yofst, \
-           uint8_t *dst, int dst_pitch)
-
-typedef prototype_subpixel_predict((*vp9_subpix_fn_t));
-
-#endif  // VP9_COMMON_VP9_SUBPIXEL_H_
diff --git a/vp9/common/x86/vp9_asm_stubs.c b/vp9/common/x86/vp9_asm_stubs.c
index f09e2d78b..3e2346f29 100644
--- a/vp9/common/x86/vp9_asm_stubs.c
+++ b/vp9/common/x86/vp9_asm_stubs.c
@@ -8,91 +8,11 @@
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
+#include <assert.h>
 
 #include "./vpx_config.h"
+#include "./vp9_rtcd.h"
 #include "vpx_ports/mem.h"
-#include "vp9/common/vp9_subpixel.h"
-
-extern const short vp9_six_tap_mmx[8][6 * 8];
-
-extern void vp9_filter_block1d_h6_mmx(unsigned char   *src_ptr,
-                                      unsigned short  *output_ptr,
-                                      unsigned int     src_pixels_per_line,
-                                      unsigned int     pixel_step,
-                                      unsigned int     output_height,
-                                      unsigned int     output_width,
-                                      const short     *vp9_filter);
-
-extern void vp9_filter_block1dc_v6_mmx(unsigned short *src_ptr,
-                                       unsigned char  *output_ptr,
-                                       int             output_pitch,
-                                       unsigned int    pixels_per_line,
-                                       unsigned int    pixel_step,
-                                       unsigned int    output_height,
-                                       unsigned int    output_width,
-                                       const short    *vp9_filter);
-
-extern void vp9_filter_block1d8_h6_sse2(unsigned char  *src_ptr,
-                                        unsigned short *output_ptr,
-                                        unsigned int    src_pixels_per_line,
-                                        unsigned int    pixel_step,
-                                        unsigned int    output_height,
-                                        unsigned int    output_width,
-                                        const short    *vp9_filter);
-
-extern void vp9_filter_block1d16_h6_sse2(unsigned char  *src_ptr,
-                                         unsigned short *output_ptr,
-                                         unsigned int    src_pixels_per_line,
-                                         unsigned int    pixel_step,
-                                         unsigned int    output_height,
-                                         unsigned int    output_width,
-                                         const short    *vp9_filter);
-
-extern void vp9_filter_block1d8_v6_sse2(unsigned short *src_ptr,
-                                        unsigned char *output_ptr,
-                                        int dst_ptich,
-                                        unsigned int pixels_per_line,
-                                        unsigned int pixel_step,
-                                        unsigned int output_height,
-                                        unsigned int output_width,
-                                        const short    *vp9_filter);
-
-extern void vp9_filter_block1d16_v6_sse2(unsigned short *src_ptr,
-                                         unsigned char *output_ptr,
-                                         int dst_ptich,
-                                         unsigned int pixels_per_line,
-                                         unsigned int pixel_step,
-                                         unsigned int output_height,
-                                         unsigned int output_width,
-                                         const short    *vp9_filter);
-
-extern void vp9_unpack_block1d16_h6_sse2(unsigned char  *src_ptr,
-                                         unsigned short *output_ptr,
-                                         unsigned int    src_pixels_per_line,
-                                         unsigned int    output_height,
-                                         unsigned int    output_width);
-
-extern void vp9_filter_block1d8_h6_only_sse2(unsigned char *src_ptr,
-                                             unsigned int   src_pixels_per_line,
-                                             unsigned char *output_ptr,
-                                             int            dst_pitch,
-                                             unsigned int   output_height,
-                                             const short   *vp9_filter);
-
-extern void vp9_filter_block1d16_h6_only_sse2(unsigned char *src_ptr,
-                                              unsigned int   src_pixels_per_lin,
-                                              unsigned char *output_ptr,
-                                              int            dst_pitch,
-                                              unsigned int   output_height,
-                                              const short   *vp9_filter);
-
-extern void vp9_filter_block1d8_v6_only_sse2(unsigned char *src_ptr,
-                                             unsigned int   src_pixels_per_line,
-                                             unsigned char *output_ptr,
-                                             int            dst_pitch,
-                                             unsigned int   output_height,
-                                             const short   *vp9_filter);
-
 ///////////////////////////////////////////////////////////////////////////
 // the mmx function that does the bilinear filtering and var calculation //
 // int one pass                                                          //
@@ -116,389 +36,7 @@ DECLARE_ALIGNED(16, const short, vp9_bilinear_filters_mmx[16][8]) = {
   {   8,  8,  8,  8, 120, 120, 120, 120 }
 };
 
-#if HAVE_MMX
-void vp9_sixtap_predict4x4_mmx(unsigned char  *src_ptr,
-                               int  src_pixels_per_line,
-                               int  xoffset,
-                               int  yoffset,
-                               unsigned char *dst_ptr,
-                               int  dst_pitch) {
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict4x4_mmx\n");
-#endif
-  /* Temp data bufffer used in filtering */
-  DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 16 * 16);
-  const short *hfilter, *vfilter;
-  hfilter = vp9_six_tap_mmx[xoffset];
-  vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line), fdata2,
-                            src_pixels_per_line, 1, 9, 8, hfilter);
-  vfilter = vp9_six_tap_mmx[yoffset];
-  vp9_filter_block1dc_v6_mmx(fdata2 + 8, dst_ptr, dst_pitch,
-                             8, 4, 4, 4, vfilter);
-}
-
-void vp9_sixtap_predict16x16_mmx(unsigned char  *src_ptr,
-                                 int  src_pixels_per_line,
-                                 int  xoffset,
-                                 int  yoffset,
-                                 unsigned char *dst_ptr,
-                                 int dst_pitch) {
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict16x16_mmx\n");
-#endif
-  /* Temp data bufffer used in filtering */
-  DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 24 * 24);
-  const short *hfilter, *vfilter;
-
-  hfilter = vp9_six_tap_mmx[xoffset];
-  vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line),
-                            fdata2,   src_pixels_per_line, 1, 21, 32,
-                            hfilter);
-  vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line) + 4,
-                            fdata2 + 4, src_pixels_per_line, 1, 21, 32,
-                            hfilter);
-  vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line) + 8,
-                            fdata2 + 8, src_pixels_per_line, 1, 21, 32,
-                            hfilter);
-  vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line) + 12,
-                            fdata2 + 12, src_pixels_per_line, 1, 21, 32,
-                            hfilter);
-
-  vfilter = vp9_six_tap_mmx[yoffset];
-  vp9_filter_block1dc_v6_mmx(fdata2 + 32, dst_ptr,      dst_pitch,
-                             32, 16, 16, 16, vfilter);
-  vp9_filter_block1dc_v6_mmx(fdata2 + 36, dst_ptr + 4,  dst_pitch,
-                             32, 16, 16, 16, vfilter);
-  vp9_filter_block1dc_v6_mmx(fdata2 + 40, dst_ptr + 8,  dst_pitch,
-                             32, 16, 16, 16, vfilter);
-  vp9_filter_block1dc_v6_mmx(fdata2 + 44, dst_ptr + 12, dst_pitch,
-                             32, 16, 16, 16, vfilter);
-}
-
-void vp9_sixtap_predict8x8_mmx(unsigned char  *src_ptr,
-                               int  src_pixels_per_line,
-                               int  xoffset,
-                               int  yoffset,
-                               unsigned char *dst_ptr,
-                               int  dst_pitch) {
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict8x8_mmx\n");
-#endif
-  /* Temp data bufffer used in filtering */
-  DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 256);
-  const short *hfilter, *vfilter;
-
-  hfilter = vp9_six_tap_mmx[xoffset];
-  vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line),
-                            fdata2,   src_pixels_per_line, 1, 13, 16,
-                            hfilter);
-  vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line) + 4,
-                            fdata2 + 4, src_pixels_per_line, 1, 13, 16,
-                            hfilter);
-
-  vfilter = vp9_six_tap_mmx[yoffset];
-  vp9_filter_block1dc_v6_mmx(fdata2 + 16, dst_ptr,     dst_pitch,
-                             16, 8, 8, 8, vfilter);
-  vp9_filter_block1dc_v6_mmx(fdata2 + 20, dst_ptr + 4, dst_pitch,
-                             16, 8, 8, 8, vfilter);
-}
-
-void vp9_sixtap_predict8x4_mmx(unsigned char  *src_ptr,
-                               int  src_pixels_per_line,
-                               int  xoffset,
-                               int  yoffset,
-                               unsigned char *dst_ptr,
-                               int  dst_pitch) {
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict8x4_mmx\n");
-#endif
-  /* Temp data bufffer used in filtering */
-  DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 256);
-  const short *hfilter, *vfilter;
-
-  hfilter = vp9_six_tap_mmx[xoffset];
-  vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line),
-                            fdata2,   src_pixels_per_line, 1, 9, 16, hfilter);
-  vp9_filter_block1d_h6_mmx(src_ptr - (2 * src_pixels_per_line) + 4,
-                            fdata2 + 4, src_pixels_per_line, 1, 9, 16, hfilter);
-
-  vfilter = vp9_six_tap_mmx[yoffset];
-  vp9_filter_block1dc_v6_mmx(fdata2 + 16, dst_ptr,     dst_pitch,
-                             16, 8, 4, 8, vfilter);
-  vp9_filter_block1dc_v6_mmx(fdata2 + 20, dst_ptr + 4, dst_pitch,
-                             16, 8, 4, 8, vfilter);
-}
-#endif
-
-#if HAVE_SSE2
-void vp9_sixtap_predict16x16_sse2(unsigned char  *src_ptr,
-                                  int  src_pixels_per_line,
-                                  int  xoffset,
-                                  int  yoffset,
-                                  unsigned char *dst_ptr,
-                                  int  dst_pitch) {
-  /* Temp data bufffer used in filtering */
-  DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 24 * 24);
-  const short *hfilter, *vfilter;
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict16x16_sse2\n");
-#endif
-
-  if (xoffset) {
-    if (yoffset) {
-      hfilter = vp9_six_tap_mmx[xoffset];
-      vp9_filter_block1d16_h6_sse2(src_ptr - (2 * src_pixels_per_line), fdata2,
-                                   src_pixels_per_line, 1, 21, 32, hfilter);
-      vfilter = vp9_six_tap_mmx[yoffset];
-      vp9_filter_block1d16_v6_sse2(fdata2 + 32, dst_ptr, dst_pitch,
-                                   32, 16, 16, dst_pitch, vfilter);
-    } else {
-      /* First-pass only */
-      hfilter = vp9_six_tap_mmx[xoffset];
-      vp9_filter_block1d16_h6_only_sse2(src_ptr, src_pixels_per_line,
-                                        dst_ptr, dst_pitch, 16, hfilter);
-    }
-  } else {
-    /* Second-pass only */
-    vfilter = vp9_six_tap_mmx[yoffset];
-    vp9_unpack_block1d16_h6_sse2(src_ptr - (2 * src_pixels_per_line), fdata2,
-                                 src_pixels_per_line, 21, 32);
-    vp9_filter_block1d16_v6_sse2(fdata2 + 32, dst_ptr, dst_pitch,
-                                 32, 16, 16, dst_pitch, vfilter);
-  }
-}
-
-void vp9_sixtap_predict8x8_sse2(unsigned char  *src_ptr,
-                                int  src_pixels_per_line,
-                                int  xoffset,
-                                int  yoffset,
-                                unsigned char *dst_ptr,
-                                int  dst_pitch) {
-  /* Temp data bufffer used in filtering */
-  DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 256);
-  const short *hfilter, *vfilter;
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict8x8_sse2\n");
-#endif
-
-  if (xoffset) {
-    if (yoffset) {
-      hfilter = vp9_six_tap_mmx[xoffset];
-      vp9_filter_block1d8_h6_sse2(src_ptr - (2 * src_pixels_per_line), fdata2,
-                                  src_pixels_per_line, 1, 13, 16, hfilter);
-      vfilter = vp9_six_tap_mmx[yoffset];
-      vp9_filter_block1d8_v6_sse2(fdata2 + 16, dst_ptr, dst_pitch,
-                                  16, 8, 8, dst_pitch, vfilter);
-    } else {
-      /* First-pass only */
-      hfilter = vp9_six_tap_mmx[xoffset];
-      vp9_filter_block1d8_h6_only_sse2(src_ptr, src_pixels_per_line,
-                                       dst_ptr, dst_pitch, 8, hfilter);
-    }
-  } else {
-    /* Second-pass only */
-    vfilter = vp9_six_tap_mmx[yoffset];
-    vp9_filter_block1d8_v6_only_sse2(src_ptr - (2 * src_pixels_per_line),
-                                     src_pixels_per_line,
-                                     dst_ptr, dst_pitch, 8, vfilter);
-  }
-}
-
-void vp9_sixtap_predict8x4_sse2(unsigned char  *src_ptr,
-                                int  src_pixels_per_line,
-                                int  xoffset,
-                                int  yoffset,
-                                unsigned char *dst_ptr,
-                                int  dst_pitch) {
-  /* Temp data bufffer used in filtering */
-  DECLARE_ALIGNED_ARRAY(16, unsigned short, fdata2, 256);
-  const short *hfilter, *vfilter;
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict8x4_sse2\n");
-#endif
-
-  if (xoffset) {
-    if (yoffset) {
-      hfilter = vp9_six_tap_mmx[xoffset];
-      vp9_filter_block1d8_h6_sse2(src_ptr - (2 * src_pixels_per_line), fdata2,
-                                  src_pixels_per_line, 1, 9, 16, hfilter);
-      vfilter = vp9_six_tap_mmx[yoffset];
-      vp9_filter_block1d8_v6_sse2(fdata2 + 16, dst_ptr, dst_pitch,
-                                  16, 8, 4, dst_pitch, vfilter);
-    } else {
-      /* First-pass only */
-      hfilter = vp9_six_tap_mmx[xoffset];
-      vp9_filter_block1d8_h6_only_sse2(src_ptr, src_pixels_per_line,
-                                       dst_ptr, dst_pitch, 4, hfilter);
-    }
-  } else {
-    /* Second-pass only */
-    vfilter = vp9_six_tap_mmx[yoffset];
-    vp9_filter_block1d8_v6_only_sse2(src_ptr - (2 * src_pixels_per_line),
-                                     src_pixels_per_line,
-                                     dst_ptr, dst_pitch, 4, vfilter);
-  }
-}
-#endif
-
 #if HAVE_SSSE3
-extern void vp9_filter_block1d8_h6_ssse3(unsigned char  *src_ptr,
-                                         unsigned int    src_pixels_per_line,
-                                         unsigned char  *output_ptr,
-                                         unsigned int    output_pitch,
-                                         unsigned int    output_height,
-                                         unsigned int    vp9_filter_index);
-
-extern void vp9_filter_block1d16_h6_ssse3(unsigned char  *src_ptr,
-                                          unsigned int    src_pixels_per_line,
-                                          unsigned char  *output_ptr,
-                                          unsigned int    output_pitch,
-                                          unsigned int    output_height,
-                                          unsigned int    vp9_filter_index);
-
-extern void vp9_filter_block1d16_v6_ssse3(unsigned char *src_ptr,
-                                          unsigned int   src_pitch,
-                                          unsigned char *output_ptr,
-                                          unsigned int   out_pitch,
-                                          unsigned int   output_height,
-                                          unsigned int   vp9_filter_index);
-
-extern void vp9_filter_block1d8_v6_ssse3(unsigned char *src_ptr,
-                                         unsigned int   src_pitch,
-                                         unsigned char *output_ptr,
-                                         unsigned int   out_pitch,
-                                         unsigned int   output_height,
-                                         unsigned int   vp9_filter_index);
-
-extern void vp9_filter_block1d4_h6_ssse3(unsigned char  *src_ptr,
-                                         unsigned int    src_pixels_per_line,
-                                         unsigned char  *output_ptr,
-                                         unsigned int    output_pitch,
-                                         unsigned int    output_height,
-                                         unsigned int    vp9_filter_index);
-
-extern void vp9_filter_block1d4_v6_ssse3(unsigned char *src_ptr,
-                                         unsigned int   src_pitch,
-                                         unsigned char *output_ptr,
-                                         unsigned int   out_pitch,
-                                         unsigned int   output_height,
-                                         unsigned int   vp9_filter_index);
-
-void vp9_sixtap_predict16x16_ssse3(unsigned char  *src_ptr,
-                                   int  src_pixels_per_line,
-                                   int  xoffset,
-                                   int  yoffset,
-                                   unsigned char *dst_ptr,
-                                   int  dst_pitch) {
-  DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 24 * 24);
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict16x16_ssse3\n");
-#endif
-
-  if (xoffset) {
-    if (yoffset) {
-      vp9_filter_block1d16_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
-                                    src_pixels_per_line,
-                                    fdata2, 16, 21, xoffset);
-      vp9_filter_block1d16_v6_ssse3(fdata2, 16, dst_ptr, dst_pitch,
-                                    16, yoffset);
-    } else {
-      /* First-pass only */
-      vp9_filter_block1d16_h6_ssse3(src_ptr, src_pixels_per_line,
-                                    dst_ptr, dst_pitch, 16, xoffset);
-    }
-  } else {
-    /* Second-pass only */
-    vp9_filter_block1d16_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
-                                  src_pixels_per_line,
-                                  dst_ptr, dst_pitch, 16, yoffset);
-  }
-}
-
-void vp9_sixtap_predict8x8_ssse3(unsigned char  *src_ptr,
-                                 int  src_pixels_per_line,
-                                 int  xoffset,
-                                 int  yoffset,
-                                 unsigned char *dst_ptr,
-                                 int  dst_pitch) {
-  DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 256);
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict8x8_ssse3\n");
-#endif
-
-  if (xoffset) {
-    if (yoffset) {
-      vp9_filter_block1d8_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
-                                   src_pixels_per_line, fdata2, 8, 13, xoffset);
-      vp9_filter_block1d8_v6_ssse3(fdata2, 8, dst_ptr, dst_pitch, 8, yoffset);
-    } else {
-      vp9_filter_block1d8_h6_ssse3(src_ptr, src_pixels_per_line,
-                                   dst_ptr, dst_pitch, 8, xoffset);
-    }
-  } else {
-    /* Second-pass only */
-    vp9_filter_block1d8_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
-                                 src_pixels_per_line,
-                                 dst_ptr, dst_pitch, 8, yoffset);
-  }
-}
-
-void vp9_sixtap_predict8x4_ssse3(unsigned char  *src_ptr,
-                                 int  src_pixels_per_line,
-                                 int  xoffset,
-                                 int  yoffset,
-                                 unsigned char *dst_ptr,
-                                 int  dst_pitch) {
-  DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 256);
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict8x4_ssse3\n");
-#endif
-
-  if (xoffset) {
-    if (yoffset) {
-      vp9_filter_block1d8_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
-                                   src_pixels_per_line, fdata2, 8, 9, xoffset);
-      vp9_filter_block1d8_v6_ssse3(fdata2, 8, dst_ptr, dst_pitch, 4, yoffset);
-    } else {
-      /* First-pass only */
-      vp9_filter_block1d8_h6_ssse3(src_ptr, src_pixels_per_line,
-                                   dst_ptr, dst_pitch, 4, xoffset);
-    }
-  } else {
-    /* Second-pass only */
-    vp9_filter_block1d8_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
-                                 src_pixels_per_line,
-                                 dst_ptr, dst_pitch, 4, yoffset);
-  }
-}
-
-void vp9_sixtap_predict4x4_ssse3(unsigned char  *src_ptr,
-                                 int   src_pixels_per_line,
-                                 int  xoffset,
-                                 int  yoffset,
-                                 unsigned char *dst_ptr,
-                                 int dst_pitch) {
-  DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 4 * 9);
-#ifdef ANNOUNCE_FUNCTION
-  printf("vp9_sixtap_predict4x4_ssse3\n");
-#endif
-
-  if (xoffset) {
-    if (yoffset) {
-      vp9_filter_block1d4_h6_ssse3(src_ptr - (2 * src_pixels_per_line),
-                                   src_pixels_per_line, fdata2, 4, 9, xoffset);
-      vp9_filter_block1d4_v6_ssse3(fdata2, 4, dst_ptr, dst_pitch, 4, yoffset);
-    } else {
-      vp9_filter_block1d4_h6_ssse3(src_ptr, src_pixels_per_line,
-                                   dst_ptr, dst_pitch, 4, xoffset);
-    }
-  } else {
-    vp9_filter_block1d4_v6_ssse3(src_ptr - (2 * src_pixels_per_line),
-                                 src_pixels_per_line,
-                                 dst_ptr, dst_pitch, 4, yoffset);
-  }
-}
-
 void vp9_filter_block1d16_v8_ssse3(const unsigned char *src_ptr,
                                    const unsigned int src_pitch,
                                    unsigned char *output_ptr,
@@ -513,30 +51,6 @@ void vp9_filter_block1d16_h8_ssse3(const unsigned char *src_ptr,
                                    unsigned int output_height,
                                    const short *filter);
 
-void vp9_filter_block2d_16x16_8_ssse3(const unsigned char *src_ptr,
-                                      const unsigned int src_stride,
-                                      const short *hfilter_aligned16,
-                                      const short *vfilter_aligned16,
-                                      unsigned char *dst_ptr,
-                                      unsigned int dst_stride) {
-  if (hfilter_aligned16[3] != 128 && vfilter_aligned16[3] != 128) {
-    DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 23 * 16);
-
-    vp9_filter_block1d16_h8_ssse3(src_ptr - (3 * src_stride), src_stride,
-                                  fdata2, 16, 23, hfilter_aligned16);
-    vp9_filter_block1d16_v8_ssse3(fdata2, 16, dst_ptr, dst_stride, 16,
-                                  vfilter_aligned16);
-  } else {
-    if (hfilter_aligned16[3] != 128) {
-      vp9_filter_block1d16_h8_ssse3(src_ptr, src_stride, dst_ptr, dst_stride,
-                                    16, hfilter_aligned16);
-    } else {
-      vp9_filter_block1d16_v8_ssse3(src_ptr - (3 * src_stride), src_stride,
-                                    dst_ptr, dst_stride, 16, vfilter_aligned16);
-    }
-  }
-}
-
 void vp9_filter_block1d8_v8_ssse3(const unsigned char *src_ptr,
                                    const unsigned int src_pitch,
                                    unsigned char *output_ptr,
@@ -551,51 +65,100 @@ void vp9_filter_block1d8_h8_ssse3(const unsigned char *src_ptr,
                                    unsigned int output_height,
                                    const short *filter);
 
-void vp9_filter_block2d_8x8_8_ssse3(const unsigned char *src_ptr,
-                                    const unsigned int src_stride,
-                                    const short *hfilter_aligned16,
-                                    const short *vfilter_aligned16,
-                                    unsigned char *dst_ptr,
-                                    unsigned int dst_stride) {
-  if (hfilter_aligned16[3] != 128 && vfilter_aligned16[3] != 128) {
-    DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 23 * 16);
-
-    vp9_filter_block1d8_h8_ssse3(src_ptr - (3 * src_stride), src_stride,
-                                 fdata2, 16, 15, hfilter_aligned16);
-    vp9_filter_block1d8_v8_ssse3(fdata2, 16, dst_ptr, dst_stride, 8,
-                                 vfilter_aligned16);
-  } else {
-    if (hfilter_aligned16[3] != 128) {
-      vp9_filter_block1d8_h8_ssse3(src_ptr, src_stride, dst_ptr, dst_stride, 8,
-                                   hfilter_aligned16);
-    } else {
-      vp9_filter_block1d8_v8_ssse3(src_ptr - (3 * src_stride), src_stride,
-                                   dst_ptr, dst_stride, 8, vfilter_aligned16);
+void vp9_convolve8_horiz_ssse3(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) {
+  if (x_step_q4 == 16 && filter_x[3] != 128) {
+    while (w >= 16) {
+      vp9_filter_block1d16_h8_ssse3(src, src_stride,
+                                    dst, dst_stride,
+                                    h, filter_x);
+      src += 16;
+      dst += 16;
+      w -= 16;
     }
+    while (w >= 8) {
+      vp9_filter_block1d8_h8_ssse3(src, src_stride,
+                                   dst, dst_stride,
+                                   h, filter_x);
+      src += 8;
+      dst += 8;
+      w -= 8;
+    }
+  }
+  if (w) {
+    vp9_convolve8_horiz_c(src, src_stride, dst, dst_stride,
+                          filter_x, x_step_q4, filter_y, y_step_q4,
+                          w, h);
   }
 }
 
-void vp9_filter_block2d_8x4_8_ssse3(const unsigned char *src_ptr,
-                                    const unsigned int src_stride,
-                                    const short *hfilter_aligned16,
-                                    const short *vfilter_aligned16,
-                                    unsigned char *dst_ptr,
-                                    unsigned int dst_stride) {
-  if (hfilter_aligned16[3] !=128 && vfilter_aligned16[3] != 128) {
-      DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 23 * 16);
+void vp9_convolve8_vert_ssse3(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) {
+  if (y_step_q4 == 16 && filter_y[3] != 128) {
+    while (w >= 16) {
+      vp9_filter_block1d16_v8_ssse3(src - src_stride * 3, src_stride,
+                                    dst, dst_stride,
+                                    h, filter_y);
+      src += 16;
+      dst += 16;
+      w -= 16;
+    }
+    while (w >= 8) {
+      vp9_filter_block1d8_v8_ssse3(src - src_stride * 3, src_stride,
+                                   dst, dst_stride,
+                                   h, filter_y);
+      src += 8;
+      dst += 8;
+      w -= 8;
+    }
+  }
+  if (w) {
+    vp9_convolve8_vert_c(src, src_stride, dst, dst_stride,
+                         filter_x, x_step_q4, filter_y, y_step_q4,
+                         w, h);
+  }
+}
 
-      vp9_filter_block1d8_h8_ssse3(src_ptr - (3 * src_stride), src_stride,
-                                   fdata2, 16, 11, hfilter_aligned16);
-      vp9_filter_block1d8_v8_ssse3(fdata2, 16, dst_ptr, dst_stride, 4,
-                                   vfilter_aligned16);
-  } else {
-    if (hfilter_aligned16[3] != 128) {
-      vp9_filter_block1d8_h8_ssse3(src_ptr, src_stride, dst_ptr, dst_stride, 4,
-                                   hfilter_aligned16);
-    } else {
-      vp9_filter_block1d8_v8_ssse3(src_ptr - (3 * src_stride), src_stride,
-                                   dst_ptr, dst_stride, 4, vfilter_aligned16);
+void vp9_convolve8_ssse3(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) {
+  DECLARE_ALIGNED_ARRAY(16, unsigned char, fdata2, 16*23);
+
+  // check w/h due to fixed size fdata2 array
+  assert(w <= 16);
+  assert(h <= 16);
+
+  if (x_step_q4 == 16 && y_step_q4 == 16 &&
+      filter_x[3] != 128 && filter_y[3] != 128) {
+    if (w == 16) {
+      vp9_filter_block1d16_h8_ssse3(src - 3 * src_stride, src_stride,
+                                    fdata2, 16,
+                                    h + 7, filter_x);
+      vp9_filter_block1d16_v8_ssse3(fdata2, 16,
+                                    dst, dst_stride,
+                                    h, filter_y);
+      return;
+    }
+    if (w == 8) {
+      vp9_filter_block1d8_h8_ssse3(src - 3 * src_stride, src_stride,
+                                   fdata2, 16,
+                                   h + 7, filter_x);
+      vp9_filter_block1d8_v8_ssse3(fdata2, 16,
+                                   dst, dst_stride,
+                                   h, filter_y);
+      return;
     }
   }
+  vp9_convolve8_c(src, src_stride, dst, dst_stride,
+                  filter_x, x_step_q4, filter_y, y_step_q4,
+                  w, h);
 }
 #endif
diff --git a/vp9/common/x86/vp9_filter_sse2.c b/vp9/common/x86/vp9_filter_sse2.c
deleted file mode 100644
index 8e02ac197..000000000
--- a/vp9/common/x86/vp9_filter_sse2.c
+++ /dev/null
@@ -1,290 +0,0 @@
-/*
- *  Copyright (c) 2012 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 <assert.h> // for alignment checks
-#include <emmintrin.h> // SSE2
-#include "vp9/common/vp9_filter.h"
-#include "vpx_ports/emmintrin_compat.h"
-#include "vpx_ports/mem.h" // for DECLARE_ALIGNED
-#include "vp9_rtcd.h"
-
-// TODO(cd): After cleanup, commit faster versions for non 4x4 size. This is
-//           just a quick partial snapshot so that other can already use some
-//           speedup.
-// TODO(cd): Use vectorized 8 tap filtering code as speedup to pure C 6 tap
-//           filtering.
-// TODO(cd): Add some comments, better variable naming.
-// TODO(cd): Maybe use _mm_maddubs_epi16 if smaller filter coeficients (no sum
-//           of positive above 128), or have higher precision filter
-//           coefficients.
-
-DECLARE_ALIGNED(16, static const unsigned int, rounding_c[4]) = {
-  VP9_FILTER_WEIGHT >> 1,
-  VP9_FILTER_WEIGHT >> 1,
-  VP9_FILTER_WEIGHT >> 1,
-  VP9_FILTER_WEIGHT >> 1,
-};
-
-// Creating a macro to do more than four pixels at once to hide instruction
-// latency is actually slower :-(
-#define DO_FOUR_PIXELS(result, src_ptr, offset)                                \
-  {                                                                            \
-  /* Do shifted load to achieve require shuffles through unpacking */          \
-  const __m128i src0  = _mm_loadu_si128((const __m128i *)(src_ptr + offset + 0)); \
-  const __m128i src1  = _mm_loadu_si128((const __m128i *)(src_ptr + offset + 1)); \
-  const __m128i src2  = _mm_loadu_si128((const __m128i *)(src_ptr + offset + 2)); \
-  const __m128i src3  = _mm_loadu_si128((const __m128i *)(src_ptr + offset + 3)); \
-  const __m128i src01 = _mm_unpacklo_epi8(src0, src1);                         \
-  const __m128i src01_16 = _mm_unpacklo_epi8(src01, zero);                     \
-  const __m128i src23 = _mm_unpacklo_epi8(src2, src3);                         \
-  const __m128i src23_16 = _mm_unpacklo_epi8(src23, zero);                     \
-  /* Shit by 4 bytes through suffle to get additional shifted loads */         \
-  const __m128i src4  = _mm_shuffle_epi32(src0, _MM_SHUFFLE(3, 3, 2, 1));      \
-  const __m128i src5  = _mm_shuffle_epi32(src1, _MM_SHUFFLE(3, 3, 2, 1));      \
-  const __m128i src6  = _mm_shuffle_epi32(src2, _MM_SHUFFLE(3, 3, 2, 1));      \
-  const __m128i src7  = _mm_shuffle_epi32(src3, _MM_SHUFFLE(3, 3, 2, 1));      \
-  const __m128i src45 = _mm_unpacklo_epi8(src4, src5);                         \
-  const __m128i src45_16 = _mm_unpacklo_epi8(src45, zero);                     \
-  const __m128i src67 = _mm_unpacklo_epi8(src6, src7);                         \
-  const __m128i src67_16 = _mm_unpacklo_epi8(src67, zero);                     \
-  /* multiply accumulate them */                                               \
-  const __m128i mad01 = _mm_madd_epi16(src01_16, fil01);                       \
-  const __m128i mad23 = _mm_madd_epi16(src23_16, fil23);                       \
-  const __m128i mad45 = _mm_madd_epi16(src45_16, fil45);                       \
-  const __m128i mad67 = _mm_madd_epi16(src67_16, fil67);                       \
-  const __m128i mad0123 = _mm_add_epi32(mad01, mad23);                         \
-  const __m128i mad4567 = _mm_add_epi32(mad45, mad67);                         \
-  __m128i mad_all = _mm_add_epi32(mad0123, mad4567);                           \
-  mad_all = _mm_add_epi32(mad_all, rounding);                                  \
-  result = _mm_srai_epi32(mad_all, VP9_FILTER_SHIFT);                          \
-  }
-
-void vp9_filter_block2d_4x4_8_sse2
-(
- const unsigned char *src_ptr, const unsigned int src_stride,
- const short *HFilter_aligned16, const short *VFilter_aligned16,
- unsigned char *dst_ptr, unsigned int dst_stride
-) {
-  __m128i intermediateA, intermediateB, intermediateC;
-
-  const int kInterp_Extend = 4;
-
-  const __m128i zero = _mm_set1_epi16(0);
-  const __m128i rounding = _mm_load_si128((const __m128i *)rounding_c);
-
-  // check alignment
-  assert(0 == ((long)HFilter_aligned16)%16);
-  assert(0 == ((long)VFilter_aligned16)%16);
-
-  {
-    __m128i transpose3_0;
-    __m128i transpose3_1;
-    __m128i transpose3_2;
-    __m128i transpose3_3;
-
-    // Horizontal pass (src -> intermediate).
-    {
-      const __m128i HFilter = _mm_load_si128((const __m128i *)HFilter_aligned16);
-      // get first two columns filter coefficients
-      __m128i fil01 = _mm_shuffle_epi32(HFilter, _MM_SHUFFLE(0, 0, 0, 0));
-      __m128i fil23 = _mm_shuffle_epi32(HFilter, _MM_SHUFFLE(1, 1, 1, 1));
-      __m128i fil45 = _mm_shuffle_epi32(HFilter, _MM_SHUFFLE(2, 2, 2, 2));
-      __m128i fil67 = _mm_shuffle_epi32(HFilter, _MM_SHUFFLE(3, 3, 3, 3));
-      src_ptr -= (kInterp_Extend - 1) * src_stride + (kInterp_Extend - 1);
-
-      {
-        __m128i mad_all0;
-        __m128i mad_all1;
-        __m128i mad_all2;
-        __m128i mad_all3;
-        DO_FOUR_PIXELS(mad_all0, src_ptr, 0*src_stride)
-        DO_FOUR_PIXELS(mad_all1, src_ptr, 1*src_stride)
-        DO_FOUR_PIXELS(mad_all2, src_ptr, 2*src_stride)
-        DO_FOUR_PIXELS(mad_all3, src_ptr, 3*src_stride)
-        mad_all0 = _mm_packs_epi32(mad_all0, mad_all1);
-        mad_all2 = _mm_packs_epi32(mad_all2, mad_all3);
-        intermediateA = _mm_packus_epi16(mad_all0, mad_all2);
-        // --
-        src_ptr += src_stride*4;
-        // --
-        DO_FOUR_PIXELS(mad_all0, src_ptr, 0*src_stride)
-        DO_FOUR_PIXELS(mad_all1, src_ptr, 1*src_stride)
-        DO_FOUR_PIXELS(mad_all2, src_ptr, 2*src_stride)
-        DO_FOUR_PIXELS(mad_all3, src_ptr, 3*src_stride)
-        mad_all0 = _mm_packs_epi32(mad_all0, mad_all1);
-        mad_all2 = _mm_packs_epi32(mad_all2, mad_all3);
-        intermediateB = _mm_packus_epi16(mad_all0, mad_all2);
-        // --
-        src_ptr += src_stride*4;
-        // --
-        DO_FOUR_PIXELS(mad_all0, src_ptr, 0*src_stride)
-        DO_FOUR_PIXELS(mad_all1, src_ptr, 1*src_stride)
-        DO_FOUR_PIXELS(mad_all2, src_ptr, 2*src_stride)
-        mad_all0 = _mm_packs_epi32(mad_all0, mad_all1);
-        mad_all2 = _mm_packs_epi32(mad_all2, mad_all2);
-        intermediateC = _mm_packus_epi16(mad_all0, mad_all2);
-      }
-    }
-
-    // Transpose result (intermediate -> transpose3_x)
-    {
-      // 00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33
-      // 40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73
-      // 80 81 82 83 90 91 92 93 A0 A1 A2 A3 xx xx xx xx
-      const __m128i transpose0_0 = _mm_unpacklo_epi8(intermediateA, intermediateB);
-      const __m128i transpose0_1 = _mm_unpackhi_epi8(intermediateA, intermediateB);
-      const __m128i transpose0_2 = _mm_unpacklo_epi8(intermediateC, intermediateC);
-      const __m128i transpose0_3 = _mm_unpackhi_epi8(intermediateC, intermediateC);
-      // 00 40 01 41 02 42 03 43 10 50 11 51 12 52 13 53
-      // 20 60 21 61 22 62 23 63 30 70 31 71 32 72 33 73
-      // 80 xx 81 xx 82 xx 83 xx 90 xx 91 xx 92 xx 93 xx
-      // A0 xx A1 xx A2 xx A3 xx xx xx xx xx xx xx xx xx
-      const __m128i transpose1_0 = _mm_unpacklo_epi8(transpose0_0, transpose0_1);
-      const __m128i transpose1_1 = _mm_unpackhi_epi8(transpose0_0, transpose0_1);
-      const __m128i transpose1_2 = _mm_unpacklo_epi8(transpose0_2, transpose0_3);
-      const __m128i transpose1_3 = _mm_unpackhi_epi8(transpose0_2, transpose0_3);
-      // 00 20 40 60 01 21 41 61 02 22 42 62 03 23 43 63
-      // 10 30 50 70 11 31 51 71 12 32 52 72 13 33 53 73
-      // 80 A0 xx xx 81 A1 xx xx 82 A2 xx xx 83 A3 xx xx
-      // 90 xx xx xx 91 xx xx xx 92 xx xx xx 93 xx xx xx
-      const __m128i transpose2_0 = _mm_unpacklo_epi8(transpose1_0, transpose1_1);
-      const __m128i transpose2_1 = _mm_unpackhi_epi8(transpose1_0, transpose1_1);
-      const __m128i transpose2_2 = _mm_unpacklo_epi8(transpose1_2, transpose1_3);
-      const __m128i transpose2_3 = _mm_unpackhi_epi8(transpose1_2, transpose1_3);
-      // 00 10 20 30 40 50 60 70 01 11 21 31 41 51 61 71
-      // 02 12 22 32 42 52 62 72 03 13 23 33 43 53 63 73
-      // 80 90 A0 xx xx xx xx xx 81 91 A1 xx xx xx xx xx
-      // 82 92 A2 xx xx xx xx xx 83 93 A3 xx xx xx xx xx
-      transpose3_0 = _mm_castps_si128(
-                            _mm_shuffle_ps(_mm_castsi128_ps(transpose2_0),
-                                           _mm_castsi128_ps(transpose2_2),
-                                           _MM_SHUFFLE(1, 0, 1, 0)));
-      transpose3_1 = _mm_castps_si128(
-                            _mm_shuffle_ps(_mm_castsi128_ps(transpose2_0),
-                                           _mm_castsi128_ps(transpose2_2),
-                                           _MM_SHUFFLE(3, 2, 3, 2)));
-      transpose3_2 = _mm_castps_si128(
-                            _mm_shuffle_ps(_mm_castsi128_ps(transpose2_1),
-                                           _mm_castsi128_ps(transpose2_3),
-                                           _MM_SHUFFLE(1, 0, 1, 0)));
-      transpose3_3 = _mm_castps_si128(
-                            _mm_shuffle_ps(_mm_castsi128_ps(transpose2_1),
-                                           _mm_castsi128_ps(transpose2_3),
-                                           _MM_SHUFFLE(3, 2, 3, 2)));
-      // 00 10 20 30 40 50 60 70 80 90 A0 xx xx xx xx xx
-      // 01 11 21 31 41 51 61 71 81 91 A1 xx xx xx xx xx
-      // 02 12 22 32 42 52 62 72 82 92 A2 xx xx xx xx xx
-      // 03 13 23 33 43 53 63 73 83 93 A3 xx xx xx xx xx
-    }
-
-    // Vertical pass (transpose3_x -> dst).
-    {
-      const __m128i VFilter = _mm_load_si128((const __m128i *)VFilter_aligned16);
-      // get first two columns filter coefficients
-      __m128i fil01 = _mm_shuffle_epi32(VFilter, _MM_SHUFFLE(0, 0, 0, 0));
-      __m128i fil23 = _mm_shuffle_epi32(VFilter, _MM_SHUFFLE(1, 1, 1, 1));
-      __m128i fil45 = _mm_shuffle_epi32(VFilter, _MM_SHUFFLE(2, 2, 2, 2));
-      __m128i fil67 = _mm_shuffle_epi32(VFilter, _MM_SHUFFLE(3, 3, 3, 3));
-      __m128i col0, col1, col2, col3;
-        DECLARE_ALIGNED(16, unsigned char, temp[32]);
-      {
-        _mm_store_si128((__m128i *)temp, transpose3_0);
-        DO_FOUR_PIXELS(col0, temp, 0);
-      }
-      {
-        _mm_store_si128((__m128i *)temp, transpose3_1);
-        DO_FOUR_PIXELS(col1, temp, 0);
-      }
-      {
-        _mm_store_si128((__m128i *)temp, transpose3_2);
-        DO_FOUR_PIXELS(col2, temp, 0);
-      }
-      {
-        _mm_store_si128((__m128i *)temp, transpose3_3);
-        DO_FOUR_PIXELS(col3, temp, 0);
-      }
-      // transpose
-      {
-        __m128i T0 = _mm_unpacklo_epi32(col0, col1);
-        __m128i T1 = _mm_unpacklo_epi32(col2, col3);
-        __m128i T2 = _mm_unpackhi_epi32(col0, col1);
-        __m128i T3 = _mm_unpackhi_epi32(col2, col3);
-        col0 = _mm_unpacklo_epi64(T0, T1);
-        col1 = _mm_unpackhi_epi64(T0, T1);
-        col2 = _mm_unpacklo_epi64(T2, T3);
-        col3 = _mm_unpackhi_epi64(T2, T3);
-      }
-      // saturate to 8 bit
-      {
-        col0 = _mm_packs_epi32(col0, col0);
-        col0 = _mm_packus_epi16(col0, col0);
-        col1 = _mm_packs_epi32(col1, col1);
-        col1 = _mm_packus_epi16(col1, col1);
-        col2 = _mm_packs_epi32 (col2, col2);
-        col2 = _mm_packus_epi16(col2, col2);
-        col3 = _mm_packs_epi32 (col3, col3);
-        col3 = _mm_packus_epi16(col3, col3);
-      }
-      // store
-      {
-        *((unsigned int *)&dst_ptr[dst_stride * 0]) = _mm_cvtsi128_si32(col0);
-        *((unsigned int *)&dst_ptr[dst_stride * 1]) = _mm_cvtsi128_si32(col1);
-        *((unsigned int *)&dst_ptr[dst_stride * 2]) = _mm_cvtsi128_si32(col2);
-        *((unsigned int *)&dst_ptr[dst_stride * 3]) = _mm_cvtsi128_si32(col3);
-      }
-    }
-  }
-}
-
-void vp9_filter_block2d_8x4_8_sse2
-(
- const unsigned char *src_ptr, const unsigned int src_stride,
- const short *HFilter_aligned16, const short *VFilter_aligned16,
- unsigned char *dst_ptr, unsigned int dst_stride
-) {
-  int j;
-  for (j=0; j<8; j+=4) {
-    vp9_filter_block2d_4x4_8_sse2(src_ptr + j, src_stride,
-                                  HFilter_aligned16, VFilter_aligned16,
-                                  dst_ptr + j, dst_stride);
-  }
-}
-
-void vp9_filter_block2d_8x8_8_sse2
-(
- const unsigned char *src_ptr, const unsigned int src_stride,
- const short *HFilter_aligned16, const short *VFilter_aligned16,
- unsigned char *dst_ptr, unsigned int dst_stride
-) {
-  int i, j;
-  for (i=0; i<8; i+=4) {
-    for (j=0; j<8; j+=4) {
-      vp9_filter_block2d_4x4_8_sse2(src_ptr + j + i*src_stride, src_stride,
-                                    HFilter_aligned16, VFilter_aligned16,
-                                    dst_ptr + j + i*dst_stride, dst_stride);
-    }
-  }
-}
-
-void vp9_filter_block2d_16x16_8_sse2
-(
- const unsigned char *src_ptr, const unsigned int src_stride,
- const short *HFilter_aligned16, const short *VFilter_aligned16,
- unsigned char *dst_ptr, unsigned int dst_stride
-) {
-  int i, j;
-  for (i=0; i<16; i+=4) {
-    for (j=0; j<16; j+=4) {
-      vp9_filter_block2d_4x4_8_sse2(src_ptr + j + i*src_stride, src_stride,
-                                    HFilter_aligned16, VFilter_aligned16,
-                                    dst_ptr + j + i*dst_stride, dst_stride);
-    }
-  }
-}
diff --git a/vp9/common/x86/vp9_filter_sse4.c b/vp9/common/x86/vp9_filter_sse4.c
deleted file mode 100644
index 52c35b296..000000000
--- a/vp9/common/x86/vp9_filter_sse4.c
+++ /dev/null
@@ -1,362 +0,0 @@
-/*
- *  Copyright (c) 2012 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 <assert.h> // for alignment checks
-#include <smmintrin.h> // SSE4.1
-#include "vp9/common/vp9_filter.h"
-#include "vpx_ports/mem.h" // for DECLARE_ALIGNED
-#include "vp9_rtcd.h"
-
-// TODO(cd): After cleanup, commit faster versions for non 4x4 size. This is
-//           just a quick partial snapshot so that other can already use some
-//           speedup.
-// TODO(cd): Use vectorized 8 tap filtering code as speedup to pure C 6 tap
-//           filtering.
-// TODO(cd): Reduce source size by using macros instead of current code
-//           duplication.
-// TODO(cd): Add some comments, better variable naming.
-// TODO(cd): Maybe use _mm_maddubs_epi16 if smaller filter coeficients (no sum
-//           of positive above 128), or have higher precision filter
-//           coefficients.
-
-DECLARE_ALIGNED(16, static const unsigned char, mask0123_c[16]) = {
-  0x00, 0x01,
-  0x01, 0x02,
-  0x02, 0x03,
-  0x03, 0x04,
-  0x02, 0x03,
-  0x03, 0x04,
-  0x04, 0x05,
-  0x05, 0x06,
-};
-DECLARE_ALIGNED(16, static const unsigned char, mask4567_c[16]) = {
-  0x04, 0x05,
-  0x05, 0x06,
-  0x06, 0x07,
-  0x07, 0x08,
-  0x06, 0x07,
-  0x07, 0x08,
-  0x08, 0x09,
-  0x09, 0x0A,
-};
-DECLARE_ALIGNED(16, static const unsigned int, rounding_c[4]) = {
-  VP9_FILTER_WEIGHT >> 1,
-  VP9_FILTER_WEIGHT >> 1,
-  VP9_FILTER_WEIGHT >> 1,
-  VP9_FILTER_WEIGHT >> 1,
-};
-DECLARE_ALIGNED(16, static const unsigned char, transpose_c[16]) = {
-  0, 4,  8, 12,
-  1, 5,  9, 13,
-  2, 6, 10, 14,
-  3, 7, 11, 15
-};
-
-// Creating a macro to do more than four pixels at once to hide instruction
-// latency is actually slower :-(
-#define DO_FOUR_PIXELS(result, offset)                                         \
-  {                                                                            \
-  /*load pixels*/                                                              \
-  __m128i src  = _mm_loadu_si128((const __m128i *)(src_ptr + offset));         \
-  /* extract the ones used for first column */                                 \
-  __m128i src0123 = _mm_shuffle_epi8(src, mask0123);                           \
-  __m128i src4567 = _mm_shuffle_epi8(src, mask4567);                           \
-  __m128i src01_16 = _mm_unpacklo_epi8(src0123, zero);                         \
-  __m128i src23_16 = _mm_unpackhi_epi8(src0123, zero);                         \
-  __m128i src45_16 = _mm_unpacklo_epi8(src4567, zero);                         \
-  __m128i src67_16 = _mm_unpackhi_epi8(src4567, zero);                         \
-  /* multiply accumulate them */                                               \
-  __m128i mad01 = _mm_madd_epi16(src01_16, fil01);                             \
-  __m128i mad23 = _mm_madd_epi16(src23_16, fil23);                             \
-  __m128i mad45 = _mm_madd_epi16(src45_16, fil45);                             \
-  __m128i mad67 = _mm_madd_epi16(src67_16, fil67);                             \
-  __m128i mad0123 = _mm_add_epi32(mad01, mad23);                               \
-  __m128i mad4567 = _mm_add_epi32(mad45, mad67);                               \
-  __m128i mad_all = _mm_add_epi32(mad0123, mad4567);                           \
-  mad_all = _mm_add_epi32(mad_all, rounding);                                  \
-  result = _mm_srai_epi32(mad_all, VP9_FILTER_SHIFT);                          \
-  }
-
-void vp9_filter_block2d_4x4_8_sse4_1
-(
- const unsigned char *src_ptr, const unsigned int src_stride,
- const short *HFilter_aligned16, const short *VFilter_aligned16,
- unsigned char *dst_ptr, unsigned int dst_stride
-) {
-  __m128i intermediateA, intermediateB, intermediateC;
-
-  const int kInterp_Extend = 4;
-
-  const __m128i zero = _mm_set1_epi16(0);
-  const __m128i mask0123 = _mm_load_si128((const __m128i *)mask0123_c);
-  const __m128i mask4567 = _mm_load_si128((const __m128i *)mask4567_c);
-  const __m128i rounding = _mm_load_si128((const __m128i *)rounding_c);
-  const __m128i transpose = _mm_load_si128((const __m128i *)transpose_c);
-
-  // check alignment
-  assert(0 == ((long)HFilter_aligned16)%16);
-  assert(0 == ((long)VFilter_aligned16)%16);
-
-  {
-    __m128i transpose3_0;
-    __m128i transpose3_1;
-    __m128i transpose3_2;
-    __m128i transpose3_3;
-
-    // Horizontal pass (src -> intermediate).
-    {
-      const __m128i HFilter = _mm_load_si128((const __m128i *)HFilter_aligned16);
-      // get first two columns filter coefficients
-      __m128i fil01 = _mm_shuffle_epi32(HFilter, _MM_SHUFFLE(0, 0, 0, 0));
-      __m128i fil23 = _mm_shuffle_epi32(HFilter, _MM_SHUFFLE(1, 1, 1, 1));
-      __m128i fil45 = _mm_shuffle_epi32(HFilter, _MM_SHUFFLE(2, 2, 2, 2));
-      __m128i fil67 = _mm_shuffle_epi32(HFilter, _MM_SHUFFLE(3, 3, 3, 3));
-      src_ptr -= (kInterp_Extend - 1) * src_stride + (kInterp_Extend - 1);
-
-      {
-        __m128i mad_all0;
-        __m128i mad_all1;
-        __m128i mad_all2;
-        __m128i mad_all3;
-        DO_FOUR_PIXELS(mad_all0, 0*src_stride)
-        DO_FOUR_PIXELS(mad_all1, 1*src_stride)
-        DO_FOUR_PIXELS(mad_all2, 2*src_stride)
-        DO_FOUR_PIXELS(mad_all3, 3*src_stride)
-        mad_all0 = _mm_packs_epi32(mad_all0, mad_all1);
-        mad_all2 = _mm_packs_epi32(mad_all2, mad_all3);
-        intermediateA = _mm_packus_epi16(mad_all0, mad_all2);
-        // --
-        src_ptr += src_stride*4;
-        // --
-        DO_FOUR_PIXELS(mad_all0, 0*src_stride)
-        DO_FOUR_PIXELS(mad_all1, 1*src_stride)
-        DO_FOUR_PIXELS(mad_all2, 2*src_stride)
-        DO_FOUR_PIXELS(mad_all3, 3*src_stride)
-        mad_all0 = _mm_packs_epi32(mad_all0, mad_all1);
-        mad_all2 = _mm_packs_epi32(mad_all2, mad_all3);
-        intermediateB = _mm_packus_epi16(mad_all0, mad_all2);
-        // --
-        src_ptr += src_stride*4;
-        // --
-        DO_FOUR_PIXELS(mad_all0, 0*src_stride)
-        DO_FOUR_PIXELS(mad_all1, 1*src_stride)
-        DO_FOUR_PIXELS(mad_all2, 2*src_stride)
-        mad_all0 = _mm_packs_epi32(mad_all0, mad_all1);
-        mad_all2 = _mm_packs_epi32(mad_all2, mad_all2);
-        intermediateC = _mm_packus_epi16(mad_all0, mad_all2);
-      }
-    }
-
-    // Transpose result (intermediate -> transpose3_x)
-    {
-      // 00 01 02 03 10 11 12 13 20 21 22 23 30 31 32 33
-      // 40 41 42 43 50 51 52 53 60 61 62 63 70 71 72 73
-      // 80 81 82 83 90 91 92 93 A0 A1 A2 A3 xx xx xx xx
-      const __m128i transpose1_0 = _mm_shuffle_epi8(intermediateA, transpose);
-      const __m128i transpose1_1 = _mm_shuffle_epi8(intermediateB, transpose);
-      const __m128i transpose1_2 = _mm_shuffle_epi8(intermediateC, transpose);
-      // 00 10 20 30 01 11 21 31 02 12 22 32 03 13 23 33
-      // 40 50 60 70 41 51 61 71 42 52 62 72 43 53 63 73
-      // 80 90 A0 xx 81 91 A1 xx 82 92 A2 xx 83 93 A3 xx
-      const __m128i transpose2_0 = _mm_unpacklo_epi32(transpose1_0, transpose1_1);
-      const __m128i transpose2_1 = _mm_unpackhi_epi32(transpose1_0, transpose1_1);
-      // 00 10 20 30 40 50 60 70 01 11 21 31 41 51 61 71
-      // 02 12 22 32 42 52 62 72 03 13 23 33 43 53 63 73
-      transpose3_0 = _mm_castps_si128(
-                            _mm_shuffle_ps(_mm_castsi128_ps(transpose2_0),
-                                           _mm_castsi128_ps(transpose1_2),
-                                           _MM_SHUFFLE(0, 0, 1, 0)));
-      transpose3_1 = _mm_castps_si128(
-                            _mm_shuffle_ps(_mm_castsi128_ps(transpose2_0),
-                                           _mm_castsi128_ps(transpose1_2),
-                                           _MM_SHUFFLE(1, 1, 3, 2)));
-      transpose3_2 = _mm_castps_si128(
-                            _mm_shuffle_ps(_mm_castsi128_ps(transpose2_1),
-                                           _mm_castsi128_ps(transpose1_2),
-                                           _MM_SHUFFLE(2, 2, 1, 0)));
-      transpose3_3 = _mm_castps_si128(
-                            _mm_shuffle_ps(_mm_castsi128_ps(transpose2_1),
-                                           _mm_castsi128_ps(transpose1_2),
-                                           _MM_SHUFFLE(3, 3, 3, 2)));
-      // 00 10 20 30 40 50 60 70 80 90 A0 xx xx xx xx xx
-      // 01 11 21 31 41 51 61 71 81 91 A1 xx xx xx xx xx
-      // 02 12 22 32 42 52 62 72 82 92 A2 xx xx xx xx xx
-      // 03 13 23 33 43 53 63 73 83 93 A3 xx xx xx xx xx
-    }
-
-    // Vertical pass (transpose3_x -> dst).
-    {
-      const __m128i VFilter = _mm_load_si128((const __m128i *)VFilter_aligned16);
-      // get first two columns filter coefficients
-      __m128i fil01 = _mm_shuffle_epi32(VFilter, _MM_SHUFFLE(0, 0, 0, 0));
-      __m128i fil23 = _mm_shuffle_epi32(VFilter, _MM_SHUFFLE(1, 1, 1, 1));
-      __m128i fil45 = _mm_shuffle_epi32(VFilter, _MM_SHUFFLE(2, 2, 2, 2));
-      __m128i fil67 = _mm_shuffle_epi32(VFilter, _MM_SHUFFLE(3, 3, 3, 3));
-      __m128i col0, col1, col2, col3;
-      {
-        //load pixels
-        __m128i src  = transpose3_0;
-        // extract the ones used for first column
-        __m128i src0123 = _mm_shuffle_epi8(src, mask0123);
-        __m128i src4567 = _mm_shuffle_epi8(src, mask4567);
-        __m128i src01_16 = _mm_unpacklo_epi8(src0123, zero);
-        __m128i src23_16 = _mm_unpackhi_epi8(src0123, zero);
-        __m128i src45_16 = _mm_unpacklo_epi8(src4567, zero);
-        __m128i src67_16 = _mm_unpackhi_epi8(src4567, zero);
-        // multiply accumulate them
-        __m128i mad01 = _mm_madd_epi16(src01_16, fil01);
-        __m128i mad23 = _mm_madd_epi16(src23_16, fil23);
-        __m128i mad45 = _mm_madd_epi16(src45_16, fil45);
-        __m128i mad67 = _mm_madd_epi16(src67_16, fil67);
-        __m128i mad0123 = _mm_add_epi32(mad01, mad23);
-        __m128i mad4567 = _mm_add_epi32(mad45, mad67);
-        __m128i mad_all = _mm_add_epi32(mad0123, mad4567);
-        mad_all = _mm_add_epi32(mad_all, rounding);
-        mad_all = _mm_srai_epi32(mad_all, VP9_FILTER_SHIFT);
-        mad_all = _mm_packs_epi32(mad_all, mad_all);
-        col0 = _mm_packus_epi16(mad_all, mad_all);
-      }
-      {
-        //load pixels
-        __m128i src  = transpose3_1;
-        // extract the ones used for first column
-        __m128i src0123 = _mm_shuffle_epi8(src, mask0123);
-        __m128i src4567 = _mm_shuffle_epi8(src, mask4567);
-        __m128i src01_16 = _mm_unpacklo_epi8(src0123, zero);
-        __m128i src23_16 = _mm_unpackhi_epi8(src0123, zero);
-        __m128i src45_16 = _mm_unpacklo_epi8(src4567, zero);
-        __m128i src67_16 = _mm_unpackhi_epi8(src4567, zero);
-        // multiply accumulate them
-        __m128i mad01 = _mm_madd_epi16(src01_16, fil01);
-        __m128i mad23 = _mm_madd_epi16(src23_16, fil23);
-        __m128i mad45 = _mm_madd_epi16(src45_16, fil45);
-        __m128i mad67 = _mm_madd_epi16(src67_16, fil67);
-        __m128i mad0123 = _mm_add_epi32(mad01, mad23);
-        __m128i mad4567 = _mm_add_epi32(mad45, mad67);
-        __m128i mad_all = _mm_add_epi32(mad0123, mad4567);
-        mad_all = _mm_add_epi32(mad_all, rounding);
-        mad_all = _mm_srai_epi32(mad_all, VP9_FILTER_SHIFT);
-        mad_all = _mm_packs_epi32(mad_all, mad_all);
-        col1 = _mm_packus_epi16(mad_all, mad_all);
-      }
-      {
-        //load pixels
-        __m128i src  = transpose3_2;
-        // extract the ones used for first column
-        __m128i src0123 = _mm_shuffle_epi8(src, mask0123);
-        __m128i src4567 = _mm_shuffle_epi8(src, mask4567);
-        __m128i src01_16 = _mm_unpacklo_epi8(src0123, zero);
-        __m128i src23_16 = _mm_unpackhi_epi8(src0123, zero);
-        __m128i src45_16 = _mm_unpacklo_epi8(src4567, zero);
-        __m128i src67_16 = _mm_unpackhi_epi8(src4567, zero);
-        // multiply accumulate them
-        __m128i mad01 = _mm_madd_epi16(src01_16, fil01);
-        __m128i mad23 = _mm_madd_epi16(src23_16, fil23);
-        __m128i mad45 = _mm_madd_epi16(src45_16, fil45);
-        __m128i mad67 = _mm_madd_epi16(src67_16, fil67);
-        __m128i mad0123 = _mm_add_epi32(mad01, mad23);
-        __m128i mad4567 = _mm_add_epi32(mad45, mad67);
-        __m128i mad_all = _mm_add_epi32(mad0123, mad4567);
-        mad_all = _mm_add_epi32(mad_all, rounding);
-        mad_all = _mm_srai_epi32(mad_all, VP9_FILTER_SHIFT);
-        mad_all = _mm_packs_epi32(mad_all, mad_all);
-        col2 = _mm_packus_epi16(mad_all, mad_all);
-      }
-      {
-        //load pixels
-        __m128i src  = transpose3_3;
-        // extract the ones used for first column
-        __m128i src0123 = _mm_shuffle_epi8(src, mask0123);
-        __m128i src4567 = _mm_shuffle_epi8(src, mask4567);
-        __m128i src01_16 = _mm_unpacklo_epi8(src0123, zero);
-        __m128i src23_16 = _mm_unpackhi_epi8(src0123, zero);
-        __m128i src45_16 = _mm_unpacklo_epi8(src4567, zero);
-        __m128i src67_16 = _mm_unpackhi_epi8(src4567, zero);
-        // multiply accumulate them
-        __m128i mad01 = _mm_madd_epi16(src01_16, fil01);
-        __m128i mad23 = _mm_madd_epi16(src23_16, fil23);
-        __m128i mad45 = _mm_madd_epi16(src45_16, fil45);
-        __m128i mad67 = _mm_madd_epi16(src67_16, fil67);
-        __m128i mad0123 = _mm_add_epi32(mad01, mad23);
-        __m128i mad4567 = _mm_add_epi32(mad45, mad67);
-        __m128i mad_all = _mm_add_epi32(mad0123, mad4567);
-        mad_all = _mm_add_epi32(mad_all, rounding);
-        mad_all = _mm_srai_epi32(mad_all, VP9_FILTER_SHIFT);
-        mad_all = _mm_packs_epi32(mad_all, mad_all);
-        col3 = _mm_packus_epi16(mad_all, mad_all);
-      }
-      {
-        __m128i col01 = _mm_unpacklo_epi8(col0, col1);
-        __m128i col23 = _mm_unpacklo_epi8(col2, col3);
-        __m128i col0123 = _mm_unpacklo_epi16(col01, col23);
-        //TODO(cd): look into Ronald's comment:
-        //    Future suggestion: I believe here, too, you can merge the
-        //    packs_epi32() and pacus_epi16() for the 4 cols above, so that
-        //    you get the data in a single register, and then use pshufb
-        //    (shuffle_epi8()) instead of the unpacks here. Should be
-        //    2+3+2 instructions faster.
-        *((unsigned int *)&dst_ptr[dst_stride * 0]) =
-            _mm_extract_epi32(col0123, 0);
-        *((unsigned int *)&dst_ptr[dst_stride * 1]) =
-            _mm_extract_epi32(col0123, 1);
-        *((unsigned int *)&dst_ptr[dst_stride * 2]) =
-            _mm_extract_epi32(col0123, 2);
-        *((unsigned int *)&dst_ptr[dst_stride * 3]) =
-            _mm_extract_epi32(col0123, 3);
-      }
-    }
-  }
-}
-
-void vp9_filter_block2d_8x4_8_sse4_1
-(
- const unsigned char *src_ptr, const unsigned int src_stride,
- const short *HFilter_aligned16, const short *VFilter_aligned16,
- unsigned char *dst_ptr, unsigned int dst_stride
-) {
-  int j;
-  for (j=0; j<8; j+=4) {
-    vp9_filter_block2d_4x4_8_sse4_1(src_ptr + j, src_stride,
-                                    HFilter_aligned16, VFilter_aligned16,
-                                    dst_ptr + j, dst_stride);
-  }
-}
-
-void vp9_filter_block2d_8x8_8_sse4_1
-(
- const unsigned char *src_ptr, const unsigned int src_stride,
- const short *HFilter_aligned16, const short *VFilter_aligned16,
- unsigned char *dst_ptr, unsigned int dst_stride
-) {
-  int i, j;
-  for (i=0; i<8; i+=4) {
-    for (j=0; j<8; j+=4) {
-      vp9_filter_block2d_4x4_8_sse4_1(src_ptr + j + i*src_stride, src_stride,
-                                      HFilter_aligned16, VFilter_aligned16,
-                                      dst_ptr + j + i*dst_stride, dst_stride);
-    }
-  }
-}
-
-void vp9_filter_block2d_16x16_8_sse4_1
-(
- const unsigned char *src_ptr, const unsigned int src_stride,
- const short *HFilter_aligned16, const short *VFilter_aligned16,
- unsigned char *dst_ptr, unsigned int dst_stride
-) {
-  int i, j;
-  for (i=0; i<16; i+=4) {
-    for (j=0; j<16; j+=4) {
-      vp9_filter_block2d_4x4_8_sse4_1(src_ptr + j + i*src_stride, src_stride,
-                                      HFilter_aligned16, VFilter_aligned16,
-                                      dst_ptr + j + i*dst_stride, dst_stride);
-    }
-  }
-}
diff --git a/vp9/common/x86/vp9_subpixel_mmx.asm b/vp9/common/x86/vp9_subpixel_mmx.asm
deleted file mode 100644
index dee29b8fb..000000000
--- a/vp9/common/x86/vp9_subpixel_mmx.asm
+++ /dev/null
@@ -1,268 +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"
-
-
-%define BLOCK_HEIGHT_WIDTH 4
-%define vp9_filter_weight 128
-%define VP9_FILTER_SHIFT  7
-
-
-;void vp9_filter_block1d_h6_mmx
-;(
-;    unsigned char   *src_ptr,
-;    unsigned short  *output_ptr,
-;    unsigned int    src_pixels_per_line,
-;    unsigned int    pixel_step,
-;    unsigned int    output_height,
-;    unsigned int    output_width,
-;    short           * vp9_filter
-;)
-global sym(vp9_filter_block1d_h6_mmx) PRIVATE
-sym(vp9_filter_block1d_h6_mmx):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rdx,    arg(6) ;vp9_filter
-
-        movq        mm1,    [rdx + 16]             ; do both the negative taps first!!!
-        movq        mm2,    [rdx + 32]         ;
-        movq        mm6,    [rdx + 48]        ;
-        movq        mm7,    [rdx + 64]        ;
-
-        mov         rdi,    arg(1) ;output_ptr
-        mov         rsi,    arg(0) ;src_ptr
-        movsxd      rcx,    dword ptr arg(4) ;output_height
-        movsxd      rax,    dword ptr arg(5) ;output_width      ; destination pitch?
-        pxor        mm0,    mm0              ; mm0 = 00000000
-
-.nextrow:
-        movq        mm3,    [rsi-2]          ; mm3 = p-2..p5
-        movq        mm4,    mm3              ; mm4 = p-2..p5
-        psrlq       mm3,    8                ; mm3 = p-1..p5
-        punpcklbw   mm3,    mm0              ; mm3 = p-1..p2
-        pmullw      mm3,    mm1              ; mm3 *= kernel 1 modifiers.
-
-        movq        mm5,    mm4              ; mm5 = p-2..p5
-        punpckhbw   mm4,    mm0              ; mm5 = p2..p5
-        pmullw      mm4,    mm7              ; mm5 *= kernel 4 modifiers
-        paddsw      mm3,    mm4              ; mm3 += mm5
-
-        movq        mm4,    mm5              ; mm4 = p-2..p5;
-        psrlq       mm5,    16               ; mm5 = p0..p5;
-        punpcklbw   mm5,    mm0              ; mm5 = p0..p3
-        pmullw      mm5,    mm2              ; mm5 *= kernel 2 modifiers
-        paddsw      mm3,    mm5              ; mm3 += mm5
-
-        movq        mm5,    mm4              ; mm5 = p-2..p5
-        psrlq       mm4,    24               ; mm4 = p1..p5
-        punpcklbw   mm4,    mm0              ; mm4 = p1..p4
-        pmullw      mm4,    mm6              ; mm5 *= kernel 3 modifiers
-        paddsw      mm3,    mm4              ; mm3 += mm5
-
-        ; do outer positive taps
-        movd        mm4,    [rsi+3]
-        punpcklbw   mm4,    mm0              ; mm5 = p3..p6
-        pmullw      mm4,    [rdx+80]         ; mm5 *= kernel 0 modifiers
-        paddsw      mm3,    mm4              ; mm3 += mm5
-
-        punpcklbw   mm5,    mm0              ; mm5 = p-2..p1
-        pmullw      mm5,    [rdx]            ; mm5 *= kernel 5 modifiers
-        paddsw      mm3,    mm5              ; mm3 += mm5
-
-        paddsw      mm3,    [GLOBAL(rd)]              ; mm3 += round value
-        psraw       mm3,    VP9_FILTER_SHIFT     ; mm3 /= 128
-        packuswb    mm3,    mm0              ; pack and unpack to saturate
-        punpcklbw   mm3,    mm0              ;
-
-        movq        [rdi],  mm3              ; store the results in the destination
-
-%if ABI_IS_32BIT
-        add         rsi,    dword ptr arg(2) ;src_pixels_per_line ; next line
-        add         rdi,    rax;
-%else
-        movsxd      r8,     dword ptr arg(2) ;src_pixels_per_line
-        add         rdi,    rax;
-
-        add         rsi,    r8               ; next line
-%endif
-
-        dec         rcx                      ; decrement count
-        jnz         .nextrow                 ; next row
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vp9_filter_block1dc_v6_mmx
-;(
-;   short *src_ptr,
-;   unsigned char *output_ptr,
-;    int output_pitch,
-;   unsigned int pixels_per_line,
-;   unsigned int pixel_step,
-;   unsigned int output_height,
-;   unsigned int output_width,
-;   short * vp9_filter
-;)
-global sym(vp9_filter_block1dc_v6_mmx) PRIVATE
-sym(vp9_filter_block1dc_v6_mmx):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 8
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        movq      mm5, [GLOBAL(rd)]
-        push        rbx
-        mov         rbx, arg(7) ;vp9_filter
-        movq      mm1, [rbx + 16]             ; do both the negative taps first!!!
-        movq      mm2, [rbx + 32]         ;
-        movq      mm6, [rbx + 48]        ;
-        movq      mm7, [rbx + 64]        ;
-
-        movsxd      rdx, dword ptr arg(3) ;pixels_per_line
-        mov         rdi, arg(1) ;output_ptr
-        mov         rsi, arg(0) ;src_ptr
-        sub         rsi, rdx
-        sub         rsi, rdx
-        movsxd      rcx, DWORD PTR arg(5) ;output_height
-        movsxd      rax, DWORD PTR arg(2) ;output_pitch      ; destination pitch?
-        pxor        mm0, mm0              ; mm0 = 00000000
-
-
-.nextrow_cv:
-        movq        mm3, [rsi+rdx]        ; mm3 = p0..p8  = row -1
-        pmullw      mm3, mm1              ; mm3 *= kernel 1 modifiers.
-
-
-        movq        mm4, [rsi + 4*rdx]      ; mm4 = p0..p3  = row 2
-        pmullw      mm4, mm7              ; mm4 *= kernel 4 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-        movq        mm4, [rsi + 2*rdx]           ; mm4 = p0..p3  = row 0
-        pmullw      mm4, mm2              ; mm4 *= kernel 2 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-        movq        mm4, [rsi]            ; mm4 = p0..p3  = row -2
-        pmullw      mm4, [rbx]            ; mm4 *= kernel 0 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-
-        add         rsi, rdx              ; move source forward 1 line to avoid 3 * pitch
-        movq        mm4, [rsi + 2*rdx]     ; mm4 = p0..p3  = row 1
-        pmullw      mm4, mm6              ; mm4 *= kernel 3 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-        movq        mm4, [rsi + 4*rdx]    ; mm4 = p0..p3  = row 3
-        pmullw      mm4, [rbx +80]        ; mm4 *= kernel 3 modifiers.
-        paddsw      mm3, mm4              ; mm3 += mm4
-
-
-        paddsw      mm3, mm5               ; mm3 += round value
-        psraw       mm3, VP9_FILTER_SHIFT     ; mm3 /= 128
-        packuswb    mm3, mm0              ; pack and saturate
-
-        movd        [rdi],mm3             ; store the results in the destination
-        ; the subsequent iterations repeat 3 out of 4 of these reads.  Since the
-        ; recon block should be in cache this shouldn't cost much.  Its obviously
-        ; avoidable!!!.
-        lea         rdi,  [rdi+rax] ;
-        dec         rcx                   ; decrement count
-        jnz         .nextrow_cv           ; next row
-
-        pop         rbx
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-SECTION_RODATA
-align 16
-rd:
-    times 4 dw 0x40
-
-align 16
-global HIDDEN_DATA(sym(vp9_six_tap_mmx))
-sym(vp9_six_tap_mmx):
-    times 8 dw 0
-    times 8 dw 0
-    times 8 dw 128
-    times 8 dw 0
-    times 8 dw 0
-    times 8 dw 0
-
-    times 8 dw 0
-    times 8 dw -6
-    times 8 dw 123
-    times 8 dw 12
-    times 8 dw -1
-    times 8 dw 0
-
-    times 8 dw 2
-    times 8 dw -11
-    times 8 dw 108
-    times 8 dw 36
-    times 8 dw -8
-    times 8 dw 1
-
-    times 8 dw 0
-    times 8 dw -9
-    times 8 dw 93
-    times 8 dw 50
-    times 8 dw -6
-    times 8 dw 0
-
-    times 8 dw 3
-    times 8 dw -16
-    times 8 dw 77
-    times 8 dw 77
-    times 8 dw -16
-    times 8 dw 3
-
-    times 8 dw 0
-    times 8 dw -6
-    times 8 dw 50
-    times 8 dw 93
-    times 8 dw -9
-    times 8 dw 0
-
-    times 8 dw 1
-    times 8 dw -8
-    times 8 dw 36
-    times 8 dw 108
-    times 8 dw -11
-    times 8 dw 2
-
-    times 8 dw 0
-    times 8 dw -1
-    times 8 dw 12
-    times 8 dw 123
-    times 8 dw -6
-    times 8 dw 0
-
diff --git a/vp9/common/x86/vp9_subpixel_sse2.asm b/vp9/common/x86/vp9_subpixel_sse2.asm
deleted file mode 100644
index b0c4f1282..000000000
--- a/vp9/common/x86/vp9_subpixel_sse2.asm
+++ /dev/null
@@ -1,1372 +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"
-
-%define BLOCK_HEIGHT_WIDTH 4
-%define VP9_FILTER_WEIGHT 128
-%define VP9_FILTER_SHIFT  7
-
-
-;/************************************************************************************
-; Notes: filter_block1d_h6 applies a 6 tap filter horizontally to the input pixels. The
-; input pixel array has output_height rows. This routine assumes that output_height is an
-; even number. This function handles 8 pixels in horizontal direction, calculating ONE
-; rows each iteration to take advantage of the 128 bits operations.
-;*************************************************************************************/
-;void vp9_filter_block1d8_h6_sse2
-;(
-;    unsigned char  *src_ptr,
-;    unsigned short *output_ptr,
-;    unsigned int    src_pixels_per_line,
-;    unsigned int    pixel_step,
-;    unsigned int    output_height,
-;    unsigned int    output_width,
-;    short           *vp9_filter
-;)
-global sym(vp9_filter_block1d8_h6_sse2) PRIVATE
-sym(vp9_filter_block1d8_h6_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rdx,        arg(6) ;vp9_filter
-        mov         rsi,        arg(0) ;src_ptr
-
-        mov         rdi,        arg(1) ;output_ptr
-
-        movsxd      rcx,        dword ptr arg(4) ;output_height
-        movsxd      rax,        dword ptr arg(2) ;src_pixels_per_line            ; Pitch for Source
-%if ABI_IS_32BIT=0
-        movsxd      r8,         dword ptr arg(5) ;output_width
-%endif
-        pxor        xmm0,       xmm0                        ; clear xmm0 for unpack
-
-.filter_block1d8_h6_rowloop:
-        movq        xmm3,       MMWORD PTR [rsi - 2]
-        movq        xmm1,       MMWORD PTR [rsi + 6]
-
-        prefetcht2  [rsi+rax-2]
-
-        pslldq      xmm1,       8
-        por         xmm1,       xmm3
-
-        movdqa      xmm4,       xmm1
-        movdqa      xmm5,       xmm1
-
-        movdqa      xmm6,       xmm1
-        movdqa      xmm7,       xmm1
-
-        punpcklbw   xmm3,       xmm0                        ; xx05 xx04 xx03 xx02 xx01 xx01 xx-1 xx-2
-        psrldq      xmm4,       1                           ; xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00 -1
-
-        pmullw      xmm3,       XMMWORD PTR [rdx]           ; x[-2] * H[-2]; Tap 1
-        punpcklbw   xmm4,       xmm0                        ; xx06 xx05 xx04 xx03 xx02 xx01 xx00 xx-1
-
-        psrldq      xmm5,       2                           ; xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00
-        pmullw      xmm4,       XMMWORD PTR [rdx+16]        ; x[-1] * H[-1]; Tap 2
-
-
-        punpcklbw   xmm5,       xmm0                        ; xx07 xx06 xx05 xx04 xx03 xx02 xx01 xx00
-        psrldq      xmm6,       3                           ; xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01
-
-        pmullw      xmm5,       [rdx+32]                    ; x[ 0] * H[ 0]; Tap 3
-
-        punpcklbw   xmm6,       xmm0                        ; xx08 xx07 xx06 xx05 xx04 xx03 xx02 xx01
-        psrldq      xmm7,       4                           ; xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02
-
-        pmullw      xmm6,       [rdx+48]                    ; x[ 1] * h[ 1] ; Tap 4
-
-        punpcklbw   xmm7,       xmm0                        ; xx09 xx08 xx07 xx06 xx05 xx04 xx03 xx02
-        psrldq      xmm1,       5                           ; xx xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03
-
-
-        pmullw      xmm7,       [rdx+64]                    ; x[ 2] * h[ 2] ; Tap 5
-
-        punpcklbw   xmm1,       xmm0                        ; xx0a xx09 xx08 xx07 xx06 xx05 xx04 xx03
-        pmullw      xmm1,       [rdx+80]                    ; x[ 3] * h[ 3] ; Tap 6
-
-
-        paddsw      xmm4,       xmm7
-        paddsw      xmm4,       xmm5
-
-        paddsw      xmm4,       xmm3
-        paddsw      xmm4,       xmm6
-
-        paddsw      xmm4,       xmm1
-        paddsw      xmm4,       [GLOBAL(rd)]
-
-        psraw       xmm4,       7
-
-        packuswb    xmm4,       xmm0
-        punpcklbw   xmm4,       xmm0
-
-        movdqa      XMMWORD Ptr [rdi],         xmm4
-        lea         rsi,        [rsi + rax]
-
-%if ABI_IS_32BIT
-        add         rdi,        DWORD Ptr arg(5) ;[output_width]
-%else
-        add         rdi,        r8
-%endif
-        dec         rcx
-
-        jnz         .filter_block1d8_h6_rowloop                ; next row
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vp9_filter_block1d16_h6_sse2
-;(
-;    unsigned char  *src_ptr,
-;    unsigned short *output_ptr,
-;    unsigned int    src_pixels_per_line,
-;    unsigned int    pixel_step,
-;    unsigned int    output_height,
-;    unsigned int    output_width,
-;    short           *vp9_filter
-;)
-;/************************************************************************************
-; Notes: filter_block1d_h6 applies a 6 tap filter horizontally to the input pixels. The
-; input pixel array has output_height rows. This routine assumes that output_height is an
-; even number. This function handles 8 pixels in horizontal direction, calculating ONE
-; rows each iteration to take advantage of the 128 bits operations.
-;*************************************************************************************/
-global sym(vp9_filter_block1d16_h6_sse2) PRIVATE
-sym(vp9_filter_block1d16_h6_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 7
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rdx,        arg(6) ;vp9_filter
-        mov         rsi,        arg(0) ;src_ptr
-
-        mov         rdi,        arg(1) ;output_ptr
-
-        movsxd      rcx,        dword ptr arg(4) ;output_height
-        movsxd      rax,        dword ptr arg(2) ;src_pixels_per_line            ; Pitch for Source
-%if ABI_IS_32BIT=0
-        movsxd      r8,         dword ptr arg(5) ;output_width
-%endif
-
-        pxor        xmm0,       xmm0                        ; clear xmm0 for unpack
-
-.filter_block1d16_h6_sse2_rowloop:
-        movq        xmm3,       MMWORD PTR [rsi - 2]
-        movq        xmm1,       MMWORD PTR [rsi + 6]
-
-        movq        xmm2,       MMWORD PTR [rsi +14]
-        pslldq      xmm2,       8
-
-        por         xmm2,       xmm1
-        prefetcht2  [rsi+rax-2]
-
-        pslldq      xmm1,       8
-        por         xmm1,       xmm3
-
-        movdqa      xmm4,       xmm1
-        movdqa      xmm5,       xmm1
-
-        movdqa      xmm6,       xmm1
-        movdqa      xmm7,       xmm1
-
-        punpcklbw   xmm3,       xmm0                        ; xx05 xx04 xx03 xx02 xx01 xx01 xx-1 xx-2
-        psrldq      xmm4,       1                           ; xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00 -1
-
-        pmullw      xmm3,       XMMWORD PTR [rdx]           ; x[-2] * H[-2]; Tap 1
-        punpcklbw   xmm4,       xmm0                        ; xx06 xx05 xx04 xx03 xx02 xx01 xx00 xx-1
-
-        psrldq      xmm5,       2                           ; xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00
-        pmullw      xmm4,       XMMWORD PTR [rdx+16]        ; x[-1] * H[-1]; Tap 2
-
-
-        punpcklbw   xmm5,       xmm0                        ; xx07 xx06 xx05 xx04 xx03 xx02 xx01 xx00
-        psrldq      xmm6,       3                           ; xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01
-
-        pmullw      xmm5,       [rdx+32]                    ; x[ 0] * H[ 0]; Tap 3
-
-        punpcklbw   xmm6,       xmm0                        ; xx08 xx07 xx06 xx05 xx04 xx03 xx02 xx01
-        psrldq      xmm7,       4                           ; xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02
-
-        pmullw      xmm6,       [rdx+48]                    ; x[ 1] * h[ 1] ; Tap 4
-
-        punpcklbw   xmm7,       xmm0                        ; xx09 xx08 xx07 xx06 xx05 xx04 xx03 xx02
-        psrldq      xmm1,       5                           ; xx xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03
-
-
-        pmullw      xmm7,       [rdx+64]                    ; x[ 2] * h[ 2] ; Tap 5
-
-        punpcklbw   xmm1,       xmm0                        ; xx0a xx09 xx08 xx07 xx06 xx05 xx04 xx03
-        pmullw      xmm1,       [rdx+80]                    ; x[ 3] * h[ 3] ; Tap 6
-
-        paddsw      xmm4,       xmm7
-        paddsw      xmm4,       xmm5
-
-        paddsw      xmm4,       xmm3
-        paddsw      xmm4,       xmm6
-
-        paddsw      xmm4,       xmm1
-        paddsw      xmm4,       [GLOBAL(rd)]
-
-        psraw       xmm4,       7
-
-        packuswb    xmm4,       xmm0
-        punpcklbw   xmm4,       xmm0
-
-        movdqa      XMMWORD Ptr [rdi],         xmm4
-
-        movdqa      xmm3,       xmm2
-        movdqa      xmm4,       xmm2
-
-        movdqa      xmm5,       xmm2
-        movdqa      xmm6,       xmm2
-
-        movdqa      xmm7,       xmm2
-
-        punpcklbw   xmm3,       xmm0                        ; xx05 xx04 xx03 xx02 xx01 xx01 xx-1 xx-2
-        psrldq      xmm4,       1                           ; xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00 -1
-
-        pmullw      xmm3,       XMMWORD PTR [rdx]           ; x[-2] * H[-2]; Tap 1
-        punpcklbw   xmm4,       xmm0                        ; xx06 xx05 xx04 xx03 xx02 xx01 xx00 xx-1
-
-        psrldq      xmm5,       2                           ; xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00
-        pmullw      xmm4,       XMMWORD PTR [rdx+16]        ; x[-1] * H[-1]; Tap 2
-
-
-        punpcklbw   xmm5,       xmm0                        ; xx07 xx06 xx05 xx04 xx03 xx02 xx01 xx00
-        psrldq      xmm6,       3                           ; xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01
-
-        pmullw      xmm5,       [rdx+32]                    ; x[ 0] * H[ 0]; Tap 3
-
-        punpcklbw   xmm6,       xmm0                        ; xx08 xx07 xx06 xx05 xx04 xx03 xx02 xx01
-        psrldq      xmm7,       4                           ; xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02
-
-        pmullw      xmm6,       [rdx+48]                    ; x[ 1] * h[ 1] ; Tap 4
-
-        punpcklbw   xmm7,       xmm0                        ; xx09 xx08 xx07 xx06 xx05 xx04 xx03 xx02
-        psrldq      xmm2,       5                           ; xx xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03
-
-        pmullw      xmm7,       [rdx+64]                    ; x[ 2] * h[ 2] ; Tap 5
-
-        punpcklbw   xmm2,       xmm0                        ; xx0a xx09 xx08 xx07 xx06 xx05 xx04 xx03
-        pmullw      xmm2,       [rdx+80]                    ; x[ 3] * h[ 3] ; Tap 6
-
-
-        paddsw      xmm4,       xmm7
-        paddsw      xmm4,       xmm5
-
-        paddsw      xmm4,       xmm3
-        paddsw      xmm4,       xmm6
-
-        paddsw      xmm4,       xmm2
-        paddsw      xmm4,       [GLOBAL(rd)]
-
-        psraw       xmm4,       7
-
-        packuswb    xmm4,       xmm0
-        punpcklbw   xmm4,       xmm0
-
-        movdqa      XMMWORD Ptr [rdi+16],      xmm4
-
-        lea         rsi,        [rsi + rax]
-%if ABI_IS_32BIT
-        add         rdi,        DWORD Ptr arg(5) ;[output_width]
-%else
-        add         rdi,        r8
-%endif
-
-        dec         rcx
-        jnz         .filter_block1d16_h6_sse2_rowloop                ; next row
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vp9_filter_block1d8_v6_sse2
-;(
-;    short *src_ptr,
-;    unsigned char *output_ptr,
-;    int dst_ptich,
-;    unsigned int pixels_per_line,
-;    unsigned int pixel_step,
-;    unsigned int output_height,
-;    unsigned int output_width,
-;    short * vp9_filter
-;)
-;/************************************************************************************
-; Notes: filter_block1d8_v6 applies a 6 tap filter vertically to the input pixels. The
-; input pixel array has output_height rows.
-;*************************************************************************************/
-global sym(vp9_filter_block1d8_v6_sse2) PRIVATE
-sym(vp9_filter_block1d8_v6_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 8
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rax,        arg(7) ;vp9_filter
-        movsxd      rdx,        dword ptr arg(3) ;pixels_per_line
-
-        mov         rdi,        arg(1) ;output_ptr
-        mov         rsi,        arg(0) ;src_ptr
-
-        sub         rsi,        rdx
-        sub         rsi,        rdx
-
-        movsxd      rcx,        DWORD PTR arg(5) ;[output_height]
-        pxor        xmm0,       xmm0                        ; clear xmm0
-
-        movdqa      xmm7,       XMMWORD PTR [GLOBAL(rd)]
-%if ABI_IS_32BIT=0
-        movsxd      r8,         dword ptr arg(2) ; dst_ptich
-%endif
-
-.vp9_filter_block1d8_v6_sse2_loop:
-        movdqa      xmm1,       XMMWORD PTR [rsi]
-        pmullw      xmm1,       [rax]
-
-        movdqa      xmm2,       XMMWORD PTR [rsi + rdx]
-        pmullw      xmm2,       [rax + 16]
-
-        movdqa      xmm3,       XMMWORD PTR [rsi + rdx * 2]
-        pmullw      xmm3,       [rax + 32]
-
-        movdqa      xmm5,       XMMWORD PTR [rsi + rdx * 4]
-        pmullw      xmm5,       [rax + 64]
-
-        add         rsi,        rdx
-        movdqa      xmm4,       XMMWORD PTR [rsi + rdx * 2]
-
-        pmullw      xmm4,       [rax + 48]
-        movdqa      xmm6,       XMMWORD PTR [rsi + rdx * 4]
-
-        pmullw      xmm6,       [rax + 80]
-
-        paddsw      xmm2,       xmm5
-        paddsw      xmm2,       xmm3
-
-        paddsw      xmm2,       xmm1
-        paddsw      xmm2,       xmm4
-
-        paddsw      xmm2,       xmm6
-        paddsw      xmm2,       xmm7
-
-        psraw       xmm2,       7
-        packuswb    xmm2,       xmm0              ; pack and saturate
-
-        movq        QWORD PTR [rdi], xmm2         ; store the results in the destination
-%if ABI_IS_32BIT
-        add         rdi,        DWORD PTR arg(2) ;[dst_ptich]
-%else
-        add         rdi,        r8
-%endif
-        dec         rcx         ; decrement count
-        jnz         .vp9_filter_block1d8_v6_sse2_loop               ; next row
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vp9_filter_block1d16_v6_sse2
-;(
-;    unsigned short *src_ptr,
-;    unsigned char *output_ptr,
-;    int dst_ptich,
-;    unsigned int pixels_per_line,
-;    unsigned int pixel_step,
-;    unsigned int output_height,
-;    unsigned int output_width,
-;    const short    *vp9_filter
-;)
-;/************************************************************************************
-; Notes: filter_block1d16_v6 applies a 6 tap filter vertically to the input pixels. The
-; input pixel array has output_height rows.
-;*************************************************************************************/
-global sym(vp9_filter_block1d16_v6_sse2) PRIVATE
-sym(vp9_filter_block1d16_v6_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 8
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rax,        arg(7) ;vp9_filter
-        movsxd      rdx,        dword ptr arg(3) ;pixels_per_line
-
-        mov         rdi,        arg(1) ;output_ptr
-        mov         rsi,        arg(0) ;src_ptr
-
-        sub         rsi,        rdx
-        sub         rsi,        rdx
-
-        movsxd      rcx,        DWORD PTR arg(5) ;[output_height]
-%if ABI_IS_32BIT=0
-        movsxd      r8,         dword ptr arg(2) ; dst_ptich
-%endif
-
-.vp9_filter_block1d16_v6_sse2_loop:
-; The order for adding 6-tap is 2 5 3 1 4 6. Read in data in that order.
-        movdqa      xmm1,       XMMWORD PTR [rsi + rdx]       ; line 2
-        movdqa      xmm2,       XMMWORD PTR [rsi + rdx + 16]
-        pmullw      xmm1,       [rax + 16]
-        pmullw      xmm2,       [rax + 16]
-
-        movdqa      xmm3,       XMMWORD PTR [rsi + rdx * 4]       ; line 5
-        movdqa      xmm4,       XMMWORD PTR [rsi + rdx * 4 + 16]
-        pmullw      xmm3,       [rax + 64]
-        pmullw      xmm4,       [rax + 64]
-
-        movdqa      xmm5,       XMMWORD PTR [rsi + rdx * 2]       ; line 3
-        movdqa      xmm6,       XMMWORD PTR [rsi + rdx * 2 + 16]
-        pmullw      xmm5,       [rax + 32]
-        pmullw      xmm6,       [rax + 32]
-
-        movdqa      xmm7,       XMMWORD PTR [rsi]       ; line 1
-        movdqa      xmm0,       XMMWORD PTR [rsi + 16]
-        pmullw      xmm7,       [rax]
-        pmullw      xmm0,       [rax]
-
-        paddsw      xmm1,       xmm3
-        paddsw      xmm2,       xmm4
-        paddsw      xmm1,       xmm5
-        paddsw      xmm2,       xmm6
-        paddsw      xmm1,       xmm7
-        paddsw      xmm2,       xmm0
-
-        add         rsi,        rdx
-
-        movdqa      xmm3,       XMMWORD PTR [rsi + rdx * 2]       ; line 4
-        movdqa      xmm4,       XMMWORD PTR [rsi + rdx * 2 + 16]
-        pmullw      xmm3,       [rax + 48]
-        pmullw      xmm4,       [rax + 48]
-
-        movdqa      xmm5,       XMMWORD PTR [rsi + rdx * 4]       ; line 6
-        movdqa      xmm6,       XMMWORD PTR [rsi + rdx * 4 + 16]
-        pmullw      xmm5,       [rax + 80]
-        pmullw      xmm6,       [rax + 80]
-
-        movdqa      xmm7,       XMMWORD PTR [GLOBAL(rd)]
-        pxor        xmm0,       xmm0                        ; clear xmm0
-
-        paddsw      xmm1,       xmm3
-        paddsw      xmm2,       xmm4
-        paddsw      xmm1,       xmm5
-        paddsw      xmm2,       xmm6
-
-        paddsw      xmm1,       xmm7
-        paddsw      xmm2,       xmm7
-
-        psraw       xmm1,       7
-        psraw       xmm2,       7
-
-        packuswb    xmm1,       xmm2              ; pack and saturate
-        movdqa      XMMWORD PTR [rdi], xmm1       ; store the results in the destination
-%if ABI_IS_32BIT
-        add         rdi,        DWORD PTR arg(2) ;[dst_ptich]
-%else
-        add         rdi,        r8
-%endif
-        dec         rcx         ; decrement count
-        jnz         .vp9_filter_block1d16_v6_sse2_loop              ; next row
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vp9_filter_block1d8_h6_only_sse2
-;(
-;    unsigned char  *src_ptr,
-;    unsigned int    src_pixels_per_line,
-;    unsigned char  *output_ptr,
-;    int dst_ptich,
-;    unsigned int    output_height,
-;    const short    *vp9_filter
-;)
-; First-pass filter only when yoffset==0
-global sym(vp9_filter_block1d8_h6_only_sse2) PRIVATE
-sym(vp9_filter_block1d8_h6_only_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rdx,        arg(5) ;vp9_filter
-        mov         rsi,        arg(0) ;src_ptr
-
-        mov         rdi,        arg(2) ;output_ptr
-
-        movsxd      rcx,        dword ptr arg(4) ;output_height
-        movsxd      rax,        dword ptr arg(1) ;src_pixels_per_line            ; Pitch for Source
-%if ABI_IS_32BIT=0
-        movsxd      r8,         dword ptr arg(3) ;dst_ptich
-%endif
-        pxor        xmm0,       xmm0                        ; clear xmm0 for unpack
-
-.filter_block1d8_h6_only_rowloop:
-        movq        xmm3,       MMWORD PTR [rsi - 2]
-        movq        xmm1,       MMWORD PTR [rsi + 6]
-
-        prefetcht2  [rsi+rax-2]
-
-        pslldq      xmm1,       8
-        por         xmm1,       xmm3
-
-        movdqa      xmm4,       xmm1
-        movdqa      xmm5,       xmm1
-
-        movdqa      xmm6,       xmm1
-        movdqa      xmm7,       xmm1
-
-        punpcklbw   xmm3,       xmm0                        ; xx05 xx04 xx03 xx02 xx01 xx01 xx-1 xx-2
-        psrldq      xmm4,       1                           ; xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00 -1
-
-        pmullw      xmm3,       XMMWORD PTR [rdx]           ; x[-2] * H[-2]; Tap 1
-        punpcklbw   xmm4,       xmm0                        ; xx06 xx05 xx04 xx03 xx02 xx01 xx00 xx-1
-
-        psrldq      xmm5,       2                           ; xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00
-        pmullw      xmm4,       XMMWORD PTR [rdx+16]        ; x[-1] * H[-1]; Tap 2
-
-
-        punpcklbw   xmm5,       xmm0                        ; xx07 xx06 xx05 xx04 xx03 xx02 xx01 xx00
-        psrldq      xmm6,       3                           ; xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01
-
-        pmullw      xmm5,       [rdx+32]                    ; x[ 0] * H[ 0]; Tap 3
-
-        punpcklbw   xmm6,       xmm0                        ; xx08 xx07 xx06 xx05 xx04 xx03 xx02 xx01
-        psrldq      xmm7,       4                           ; xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02
-
-        pmullw      xmm6,       [rdx+48]                    ; x[ 1] * h[ 1] ; Tap 4
-
-        punpcklbw   xmm7,       xmm0                        ; xx09 xx08 xx07 xx06 xx05 xx04 xx03 xx02
-        psrldq      xmm1,       5                           ; xx xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03
-
-
-        pmullw      xmm7,       [rdx+64]                    ; x[ 2] * h[ 2] ; Tap 5
-
-        punpcklbw   xmm1,       xmm0                        ; xx0a xx09 xx08 xx07 xx06 xx05 xx04 xx03
-        pmullw      xmm1,       [rdx+80]                    ; x[ 3] * h[ 3] ; Tap 6
-
-
-        paddsw      xmm4,       xmm7
-        paddsw      xmm4,       xmm5
-
-        paddsw      xmm4,       xmm3
-        paddsw      xmm4,       xmm6
-
-        paddsw      xmm4,       xmm1
-        paddsw      xmm4,       [GLOBAL(rd)]
-
-        psraw       xmm4,       7
-
-        packuswb    xmm4,       xmm0
-
-        movq        QWORD PTR [rdi],   xmm4       ; store the results in the destination
-        lea         rsi,        [rsi + rax]
-
-%if ABI_IS_32BIT
-        add         rdi,        DWORD Ptr arg(3) ;dst_ptich
-%else
-        add         rdi,        r8
-%endif
-        dec         rcx
-
-        jnz         .filter_block1d8_h6_only_rowloop               ; next row
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vp9_filter_block1d16_h6_only_sse2
-;(
-;    unsigned char  *src_ptr,
-;    unsigned int    src_pixels_per_line,
-;    unsigned char  *output_ptr,
-;    int dst_ptich,
-;    unsigned int    output_height,
-;    const short    *vp9_filter
-;)
-; First-pass filter only when yoffset==0
-global sym(vp9_filter_block1d16_h6_only_sse2) PRIVATE
-sym(vp9_filter_block1d16_h6_only_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rdx,        arg(5) ;vp9_filter
-        mov         rsi,        arg(0) ;src_ptr
-
-        mov         rdi,        arg(2) ;output_ptr
-
-        movsxd      rcx,        dword ptr arg(4) ;output_height
-        movsxd      rax,        dword ptr arg(1) ;src_pixels_per_line            ; Pitch for Source
-%if ABI_IS_32BIT=0
-        movsxd      r8,         dword ptr arg(3) ;dst_ptich
-%endif
-
-        pxor        xmm0,       xmm0                        ; clear xmm0 for unpack
-
-.filter_block1d16_h6_only_sse2_rowloop:
-        movq        xmm3,       MMWORD PTR [rsi - 2]
-        movq        xmm1,       MMWORD PTR [rsi + 6]
-
-        movq        xmm2,       MMWORD PTR [rsi +14]
-        pslldq      xmm2,       8
-
-        por         xmm2,       xmm1
-        prefetcht2  [rsi+rax-2]
-
-        pslldq      xmm1,       8
-        por         xmm1,       xmm3
-
-        movdqa      xmm4,       xmm1
-        movdqa      xmm5,       xmm1
-
-        movdqa      xmm6,       xmm1
-        movdqa      xmm7,       xmm1
-
-        punpcklbw   xmm3,       xmm0                        ; xx05 xx04 xx03 xx02 xx01 xx01 xx-1 xx-2
-        psrldq      xmm4,       1                           ; xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00 -1
-
-        pmullw      xmm3,       XMMWORD PTR [rdx]           ; x[-2] * H[-2]; Tap 1
-        punpcklbw   xmm4,       xmm0                        ; xx06 xx05 xx04 xx03 xx02 xx01 xx00 xx-1
-
-        psrldq      xmm5,       2                           ; xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00
-        pmullw      xmm4,       XMMWORD PTR [rdx+16]        ; x[-1] * H[-1]; Tap 2
-
-        punpcklbw   xmm5,       xmm0                        ; xx07 xx06 xx05 xx04 xx03 xx02 xx01 xx00
-        psrldq      xmm6,       3                           ; xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01
-
-        pmullw      xmm5,       [rdx+32]                    ; x[ 0] * H[ 0]; Tap 3
-
-        punpcklbw   xmm6,       xmm0                        ; xx08 xx07 xx06 xx05 xx04 xx03 xx02 xx01
-        psrldq      xmm7,       4                           ; xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02
-
-        pmullw      xmm6,       [rdx+48]                    ; x[ 1] * h[ 1] ; Tap 4
-
-        punpcklbw   xmm7,       xmm0                        ; xx09 xx08 xx07 xx06 xx05 xx04 xx03 xx02
-        psrldq      xmm1,       5                           ; xx xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03
-
-        pmullw      xmm7,       [rdx+64]                    ; x[ 2] * h[ 2] ; Tap 5
-
-        punpcklbw   xmm1,       xmm0                        ; xx0a xx09 xx08 xx07 xx06 xx05 xx04 xx03
-        pmullw      xmm1,       [rdx+80]                    ; x[ 3] * h[ 3] ; Tap 6
-
-        paddsw      xmm4,       xmm7
-        paddsw      xmm4,       xmm5
-
-        paddsw      xmm4,       xmm3
-        paddsw      xmm4,       xmm6
-
-        paddsw      xmm4,       xmm1
-        paddsw      xmm4,       [GLOBAL(rd)]
-
-        psraw       xmm4,       7
-
-        packuswb    xmm4,       xmm0                        ; lower 8 bytes
-
-        movq        QWORD Ptr [rdi],         xmm4           ; store the results in the destination
-
-        movdqa      xmm3,       xmm2
-        movdqa      xmm4,       xmm2
-
-        movdqa      xmm5,       xmm2
-        movdqa      xmm6,       xmm2
-
-        movdqa      xmm7,       xmm2
-
-        punpcklbw   xmm3,       xmm0                        ; xx05 xx04 xx03 xx02 xx01 xx01 xx-1 xx-2
-        psrldq      xmm4,       1                           ; xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00 -1
-
-        pmullw      xmm3,       XMMWORD PTR [rdx]           ; x[-2] * H[-2]; Tap 1
-        punpcklbw   xmm4,       xmm0                        ; xx06 xx05 xx04 xx03 xx02 xx01 xx00 xx-1
-
-        psrldq      xmm5,       2                           ; xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00
-        pmullw      xmm4,       XMMWORD PTR [rdx+16]        ; x[-1] * H[-1]; Tap 2
-
-        punpcklbw   xmm5,       xmm0                        ; xx07 xx06 xx05 xx04 xx03 xx02 xx01 xx00
-        psrldq      xmm6,       3                           ; xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01
-
-        pmullw      xmm5,       [rdx+32]                    ; x[ 0] * H[ 0]; Tap 3
-
-        punpcklbw   xmm6,       xmm0                        ; xx08 xx07 xx06 xx05 xx04 xx03 xx02 xx01
-        psrldq      xmm7,       4                           ; xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03 02
-
-        pmullw      xmm6,       [rdx+48]                    ; x[ 1] * h[ 1] ; Tap 4
-
-        punpcklbw   xmm7,       xmm0                        ; xx09 xx08 xx07 xx06 xx05 xx04 xx03 xx02
-        psrldq      xmm2,       5                           ; xx xx xx xx xx 0d 0c 0b 0a 09 08 07 06 05 04 03
-
-        pmullw      xmm7,       [rdx+64]                    ; x[ 2] * h[ 2] ; Tap 5
-
-        punpcklbw   xmm2,       xmm0                        ; xx0a xx09 xx08 xx07 xx06 xx05 xx04 xx03
-        pmullw      xmm2,       [rdx+80]                    ; x[ 3] * h[ 3] ; Tap 6
-
-        paddsw      xmm4,       xmm7
-        paddsw      xmm4,       xmm5
-
-        paddsw      xmm4,       xmm3
-        paddsw      xmm4,       xmm6
-
-        paddsw      xmm4,       xmm2
-        paddsw      xmm4,       [GLOBAL(rd)]
-
-        psraw       xmm4,       7
-
-        packuswb    xmm4,       xmm0                        ; higher 8 bytes
-
-        movq        QWORD Ptr [rdi+8],      xmm4            ; store the results in the destination
-
-        lea         rsi,        [rsi + rax]
-%if ABI_IS_32BIT
-        add         rdi,        DWORD Ptr arg(3) ;dst_ptich
-%else
-        add         rdi,        r8
-%endif
-
-        dec         rcx
-        jnz         .filter_block1d16_h6_only_sse2_rowloop               ; next row
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vp9_filter_block1d8_v6_only_sse2
-;(
-;    unsigned char *src_ptr,
-;    unsigned int    src_pixels_per_line,
-;    unsigned char *output_ptr,
-;    int dst_ptich,
-;    unsigned int output_height,
-;    const short    *vp9_filter
-;)
-; Second-pass filter only when xoffset==0
-global sym(vp9_filter_block1d8_v6_only_sse2) PRIVATE
-sym(vp9_filter_block1d8_v6_only_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rsi,        arg(0) ;src_ptr
-        mov         rdi,        arg(2) ;output_ptr
-
-        movsxd      rcx,        dword ptr arg(4) ;output_height
-        movsxd      rdx,        dword ptr arg(1) ;src_pixels_per_line
-
-        mov         rax,        arg(5) ;vp9_filter
-
-        pxor        xmm0,       xmm0                        ; clear xmm0
-
-        movdqa      xmm7,       XMMWORD PTR [GLOBAL(rd)]
-%if ABI_IS_32BIT=0
-        movsxd      r8,         dword ptr arg(3) ; dst_ptich
-%endif
-
-.vp9_filter_block1d8_v6_only_sse2_loop:
-        movq        xmm1,       MMWORD PTR [rsi]
-        movq        xmm2,       MMWORD PTR [rsi + rdx]
-        movq        xmm3,       MMWORD PTR [rsi + rdx * 2]
-        movq        xmm5,       MMWORD PTR [rsi + rdx * 4]
-        add         rsi,        rdx
-        movq        xmm4,       MMWORD PTR [rsi + rdx * 2]
-        movq        xmm6,       MMWORD PTR [rsi + rdx * 4]
-
-        punpcklbw   xmm1,       xmm0
-        pmullw      xmm1,       [rax]
-
-        punpcklbw   xmm2,       xmm0
-        pmullw      xmm2,       [rax + 16]
-
-        punpcklbw   xmm3,       xmm0
-        pmullw      xmm3,       [rax + 32]
-
-        punpcklbw   xmm5,       xmm0
-        pmullw      xmm5,       [rax + 64]
-
-        punpcklbw   xmm4,       xmm0
-        pmullw      xmm4,       [rax + 48]
-
-        punpcklbw   xmm6,       xmm0
-        pmullw      xmm6,       [rax + 80]
-
-        paddsw      xmm2,       xmm5
-        paddsw      xmm2,       xmm3
-
-        paddsw      xmm2,       xmm1
-        paddsw      xmm2,       xmm4
-
-        paddsw      xmm2,       xmm6
-        paddsw      xmm2,       xmm7
-
-        psraw       xmm2,       7
-        packuswb    xmm2,       xmm0              ; pack and saturate
-
-        movq        QWORD PTR [rdi], xmm2         ; store the results in the destination
-%if ABI_IS_32BIT
-        add         rdi,        DWORD PTR arg(3) ;[dst_ptich]
-%else
-        add         rdi,        r8
-%endif
-        dec         rcx         ; decrement count
-        jnz         .vp9_filter_block1d8_v6_only_sse2_loop              ; next row
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vp9_unpack_block1d16_h6_sse2
-;(
-;    unsigned char  *src_ptr,
-;    unsigned short *output_ptr,
-;    unsigned int    src_pixels_per_line,
-;    unsigned int    output_height,
-;    unsigned int    output_width
-;)
-global sym(vp9_unpack_block1d16_h6_sse2) PRIVATE
-sym(vp9_unpack_block1d16_h6_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        mov         rsi,        arg(0) ;src_ptr
-        mov         rdi,        arg(1) ;output_ptr
-
-        movsxd      rcx,        dword ptr arg(3) ;output_height
-        movsxd      rax,        dword ptr arg(2) ;src_pixels_per_line            ; Pitch for Source
-
-        pxor        xmm0,       xmm0                        ; clear xmm0 for unpack
-%if ABI_IS_32BIT=0
-        movsxd      r8,         dword ptr arg(4) ;output_width            ; Pitch for Source
-%endif
-
-.unpack_block1d16_h6_sse2_rowloop:
-        movq        xmm1,       MMWORD PTR [rsi]            ; 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00 -1 -2
-        movq        xmm3,       MMWORD PTR [rsi+8]          ; make copy of xmm1
-
-        punpcklbw   xmm3,       xmm0                        ; xx05 xx04 xx03 xx02 xx01 xx01 xx-1 xx-2
-        punpcklbw   xmm1,       xmm0
-
-        movdqa      XMMWORD Ptr [rdi],         xmm1
-        movdqa      XMMWORD Ptr [rdi + 16],    xmm3
-
-        lea         rsi,        [rsi + rax]
-%if ABI_IS_32BIT
-        add         rdi,        DWORD Ptr arg(4) ;[output_width]
-%else
-        add         rdi,        r8
-%endif
-        dec         rcx
-        jnz         .unpack_block1d16_h6_sse2_rowloop               ; next row
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vp9_bilinear_predict16x16_sse2
-;(
-;    unsigned char  *src_ptr,
-;    int   src_pixels_per_line,
-;    int  xoffset,
-;    int  yoffset,
-;    unsigned char *dst_ptr,
-;    int dst_pitch
-;)
-extern sym(vp9_bilinear_filters_mmx)
-global sym(vp9_bilinear_predict16x16_sse2) PRIVATE
-sym(vp9_bilinear_predict16x16_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    ;const short *HFilter = bilinear_filters_mmx[xoffset]
-    ;const short *VFilter = bilinear_filters_mmx[yoffset]
-
-        lea         rcx,        [GLOBAL(sym(vp9_bilinear_filters_mmx))]
-        movsxd      rax,        dword ptr arg(2) ;xoffset
-
-        cmp         rax,        0      ;skip first_pass filter if xoffset=0
-        je          .b16x16_sp_only
-
-        shl         rax,        5
-        add         rax,        rcx    ;HFilter
-
-        mov         rdi,        arg(4) ;dst_ptr
-        mov         rsi,        arg(0) ;src_ptr
-        movsxd      rdx,        dword ptr arg(5) ;dst_pitch
-
-        movdqa      xmm1,       [rax]
-        movdqa      xmm2,       [rax+16]
-
-        movsxd      rax,        dword ptr arg(3) ;yoffset
-
-        cmp         rax,        0      ;skip second_pass filter if yoffset=0
-        je          .b16x16_fp_only
-
-        shl         rax,        5
-        add         rax,        rcx    ;VFilter
-
-        lea         rcx,        [rdi+rdx*8]
-        lea         rcx,        [rcx+rdx*8]
-        movsxd      rdx,        dword ptr arg(1) ;src_pixels_per_line
-
-        pxor        xmm0,       xmm0
-
-%if ABI_IS_32BIT=0
-        movsxd      r8,         dword ptr arg(5) ;dst_pitch
-%endif
-        ; get the first horizontal line done
-        movdqu      xmm3,       [rsi]               ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
-        movdqa      xmm4,       xmm3                 ; make a copy of current line
-
-        punpcklbw   xmm3,       xmm0                 ; xx 00 01 02 03 04 05 06
-        punpckhbw   xmm4,       xmm0
-
-        pmullw      xmm3,       xmm1
-        pmullw      xmm4,       xmm1
-
-        movdqu      xmm5,       [rsi+1]
-        movdqa      xmm6,       xmm5
-
-        punpcklbw   xmm5,       xmm0
-        punpckhbw   xmm6,       xmm0
-
-        pmullw      xmm5,       xmm2
-        pmullw      xmm6,       xmm2
-
-        paddw       xmm3,       xmm5
-        paddw       xmm4,       xmm6
-
-        paddw       xmm3,       [GLOBAL(rd)]        ; xmm3 += round value
-        psraw       xmm3,       VP9_FILTER_SHIFT        ; xmm3 /= 128
-
-        paddw       xmm4,       [GLOBAL(rd)]
-        psraw       xmm4,       VP9_FILTER_SHIFT
-
-        movdqa      xmm7,       xmm3
-        packuswb    xmm7,       xmm4
-
-        add         rsi,        rdx                 ; next line
-.next_row:
-        movdqu      xmm3,       [rsi]               ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
-        movdqa      xmm4,       xmm3                 ; make a copy of current line
-
-        punpcklbw   xmm3,       xmm0                 ; xx 00 01 02 03 04 05 06
-        punpckhbw   xmm4,       xmm0
-
-        pmullw      xmm3,       xmm1
-        pmullw      xmm4,       xmm1
-
-        movdqu      xmm5,       [rsi+1]
-        movdqa      xmm6,       xmm5
-
-        punpcklbw   xmm5,       xmm0
-        punpckhbw   xmm6,       xmm0
-
-        pmullw      xmm5,       xmm2
-        pmullw      xmm6,       xmm2
-
-        paddw       xmm3,       xmm5
-        paddw       xmm4,       xmm6
-
-        movdqa      xmm5,       xmm7
-        movdqa      xmm6,       xmm7
-
-        punpcklbw   xmm5,       xmm0
-        punpckhbw   xmm6,       xmm0
-
-        pmullw      xmm5,       [rax]
-        pmullw      xmm6,       [rax]
-
-        paddw       xmm3,       [GLOBAL(rd)]        ; xmm3 += round value
-        psraw       xmm3,       VP9_FILTER_SHIFT        ; xmm3 /= 128
-
-        paddw       xmm4,       [GLOBAL(rd)]
-        psraw       xmm4,       VP9_FILTER_SHIFT
-
-        movdqa      xmm7,       xmm3
-        packuswb    xmm7,       xmm4
-
-        pmullw      xmm3,       [rax+16]
-        pmullw      xmm4,       [rax+16]
-
-        paddw       xmm3,       xmm5
-        paddw       xmm4,       xmm6
-
-        paddw       xmm3,       [GLOBAL(rd)]        ; xmm3 += round value
-        psraw       xmm3,       VP9_FILTER_SHIFT        ; xmm3 /= 128
-
-        paddw       xmm4,       [GLOBAL(rd)]
-        psraw       xmm4,       VP9_FILTER_SHIFT
-
-        packuswb    xmm3,       xmm4
-        movdqa      [rdi],      xmm3                 ; store the results in the destination
-
-        add         rsi,        rdx                 ; next line
-%if ABI_IS_32BIT
-        add         rdi,        DWORD PTR arg(5) ;dst_pitch
-%else
-        add         rdi,        r8
-%endif
-
-        cmp         rdi,        rcx
-        jne         .next_row
-
-        jmp         .done
-
-.b16x16_sp_only:
-        movsxd      rax,        dword ptr arg(3) ;yoffset
-        shl         rax,        5
-        add         rax,        rcx    ;VFilter
-
-        mov         rdi,        arg(4) ;dst_ptr
-        mov         rsi,        arg(0) ;src_ptr
-        movsxd      rdx,        dword ptr arg(5) ;dst_pitch
-
-        movdqa      xmm1,       [rax]
-        movdqa      xmm2,       [rax+16]
-
-        lea         rcx,        [rdi+rdx*8]
-        lea         rcx,        [rcx+rdx*8]
-        movsxd      rax,        dword ptr arg(1) ;src_pixels_per_line
-
-        pxor        xmm0,       xmm0
-
-        ; get the first horizontal line done
-        movdqu      xmm7,       [rsi]               ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
-
-        add         rsi,        rax                 ; next line
-.next_row_spo:
-        movdqu      xmm3,       [rsi]               ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
-
-        movdqa      xmm5,       xmm7
-        movdqa      xmm6,       xmm7
-
-        movdqa      xmm4,       xmm3                 ; make a copy of current line
-        movdqa      xmm7,       xmm3
-
-        punpcklbw   xmm5,       xmm0
-        punpckhbw   xmm6,       xmm0
-        punpcklbw   xmm3,       xmm0                 ; xx 00 01 02 03 04 05 06
-        punpckhbw   xmm4,       xmm0
-
-        pmullw      xmm5,       xmm1
-        pmullw      xmm6,       xmm1
-        pmullw      xmm3,       xmm2
-        pmullw      xmm4,       xmm2
-
-        paddw       xmm3,       xmm5
-        paddw       xmm4,       xmm6
-
-        paddw       xmm3,       [GLOBAL(rd)]        ; xmm3 += round value
-        psraw       xmm3,       VP9_FILTER_SHIFT        ; xmm3 /= 128
-
-        paddw       xmm4,       [GLOBAL(rd)]
-        psraw       xmm4,       VP9_FILTER_SHIFT
-
-        packuswb    xmm3,       xmm4
-        movdqa      [rdi],      xmm3                 ; store the results in the destination
-
-        add         rsi,        rax                 ; next line
-        add         rdi,        rdx                 ;dst_pitch
-        cmp         rdi,        rcx
-        jne         .next_row_spo
-
-        jmp         .done
-
-.b16x16_fp_only:
-        lea         rcx,        [rdi+rdx*8]
-        lea         rcx,        [rcx+rdx*8]
-        movsxd      rax,        dword ptr arg(1) ;src_pixels_per_line
-        pxor        xmm0,       xmm0
-
-.next_row_fpo:
-        movdqu      xmm3,       [rsi]               ; xx 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14
-        movdqa      xmm4,       xmm3                 ; make a copy of current line
-
-        punpcklbw   xmm3,       xmm0                 ; xx 00 01 02 03 04 05 06
-        punpckhbw   xmm4,       xmm0
-
-        pmullw      xmm3,       xmm1
-        pmullw      xmm4,       xmm1
-
-        movdqu      xmm5,       [rsi+1]
-        movdqa      xmm6,       xmm5
-
-        punpcklbw   xmm5,       xmm0
-        punpckhbw   xmm6,       xmm0
-
-        pmullw      xmm5,       xmm2
-        pmullw      xmm6,       xmm2
-
-        paddw       xmm3,       xmm5
-        paddw       xmm4,       xmm6
-
-        paddw       xmm3,       [GLOBAL(rd)]        ; xmm3 += round value
-        psraw       xmm3,       VP9_FILTER_SHIFT        ; xmm3 /= 128
-
-        paddw       xmm4,       [GLOBAL(rd)]
-        psraw       xmm4,       VP9_FILTER_SHIFT
-
-        packuswb    xmm3,       xmm4
-        movdqa      [rdi],      xmm3                 ; store the results in the destination
-
-        add         rsi,        rax                 ; next line
-        add         rdi,        rdx                 ; dst_pitch
-        cmp         rdi,        rcx
-        jne         .next_row_fpo
-
-.done:
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-;void vp9_bilinear_predict8x8_sse2
-;(
-;    unsigned char  *src_ptr,
-;    int   src_pixels_per_line,
-;    int  xoffset,
-;    int  yoffset,
-;    unsigned char *dst_ptr,
-;    int dst_pitch
-;)
-extern sym(vp9_bilinear_filters_mmx)
-global sym(vp9_bilinear_predict8x8_sse2) PRIVATE
-sym(vp9_bilinear_predict8x8_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    ALIGN_STACK 16, rax
-    sub         rsp, 144                         ; reserve 144 bytes
-
-    ;const short *HFilter = bilinear_filters_mmx[xoffset]
-    ;const short *VFilter = bilinear_filters_mmx[yoffset]
-        lea         rcx,        [GLOBAL(sym(vp9_bilinear_filters_mmx))]
-
-        mov         rsi,        arg(0) ;src_ptr
-        movsxd      rdx,        dword ptr arg(1) ;src_pixels_per_line
-
-    ;Read 9-line unaligned data in and put them on stack. This gives a big
-    ;performance boost.
-        movdqu      xmm0,       [rsi]
-        lea         rax,        [rdx + rdx*2]
-        movdqu      xmm1,       [rsi+rdx]
-        movdqu      xmm2,       [rsi+rdx*2]
-        add         rsi,        rax
-        movdqu      xmm3,       [rsi]
-        movdqu      xmm4,       [rsi+rdx]
-        movdqu      xmm5,       [rsi+rdx*2]
-        add         rsi,        rax
-        movdqu      xmm6,       [rsi]
-        movdqu      xmm7,       [rsi+rdx]
-
-        movdqa      XMMWORD PTR [rsp],            xmm0
-
-        movdqu      xmm0,       [rsi+rdx*2]
-
-        movdqa      XMMWORD PTR [rsp+16],         xmm1
-        movdqa      XMMWORD PTR [rsp+32],         xmm2
-        movdqa      XMMWORD PTR [rsp+48],         xmm3
-        movdqa      XMMWORD PTR [rsp+64],         xmm4
-        movdqa      XMMWORD PTR [rsp+80],         xmm5
-        movdqa      XMMWORD PTR [rsp+96],         xmm6
-        movdqa      XMMWORD PTR [rsp+112],        xmm7
-        movdqa      XMMWORD PTR [rsp+128],        xmm0
-
-        movsxd      rax,        dword ptr arg(2) ;xoffset
-        shl         rax,        5
-        add         rax,        rcx    ;HFilter
-
-        mov         rdi,        arg(4) ;dst_ptr
-        movsxd      rdx,        dword ptr arg(5) ;dst_pitch
-
-        movdqa      xmm1,       [rax]
-        movdqa      xmm2,       [rax+16]
-
-        movsxd      rax,        dword ptr arg(3) ;yoffset
-        shl         rax,        5
-        add         rax,        rcx    ;VFilter
-
-        lea         rcx,        [rdi+rdx*8]
-
-        movdqa      xmm5,       [rax]
-        movdqa      xmm6,       [rax+16]
-
-        pxor        xmm0,       xmm0
-
-        ; get the first horizontal line done
-        movdqa      xmm3,       XMMWORD PTR [rsp]
-        movdqa      xmm4,       xmm3                 ; make a copy of current line
-        psrldq      xmm4,       1
-
-        punpcklbw   xmm3,       xmm0                 ; 00 01 02 03 04 05 06 07
-        punpcklbw   xmm4,       xmm0                 ; 01 02 03 04 05 06 07 08
-
-        pmullw      xmm3,       xmm1
-        pmullw      xmm4,       xmm2
-
-        paddw       xmm3,       xmm4
-
-        paddw       xmm3,       [GLOBAL(rd)]        ; xmm3 += round value
-        psraw       xmm3,       VP9_FILTER_SHIFT        ; xmm3 /= 128
-
-        movdqa      xmm7,       xmm3
-        add         rsp,        16                 ; next line
-.next_row8x8:
-        movdqa      xmm3,       XMMWORD PTR [rsp]               ; 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
-        movdqa      xmm4,       xmm3                 ; make a copy of current line
-        psrldq      xmm4,       1
-
-        punpcklbw   xmm3,       xmm0                 ; 00 01 02 03 04 05 06 07
-        punpcklbw   xmm4,       xmm0                 ; 01 02 03 04 05 06 07 08
-
-        pmullw      xmm3,       xmm1
-        pmullw      xmm4,       xmm2
-
-        paddw       xmm3,       xmm4
-        pmullw      xmm7,       xmm5
-
-        paddw       xmm3,       [GLOBAL(rd)]        ; xmm3 += round value
-        psraw       xmm3,       VP9_FILTER_SHIFT        ; xmm3 /= 128
-
-        movdqa      xmm4,       xmm3
-
-        pmullw      xmm3,       xmm6
-        paddw       xmm3,       xmm7
-
-        movdqa      xmm7,       xmm4
-
-        paddw       xmm3,       [GLOBAL(rd)]        ; xmm3 += round value
-        psraw       xmm3,       VP9_FILTER_SHIFT        ; xmm3 /= 128
-
-        packuswb    xmm3,       xmm0
-        movq        [rdi],      xmm3                 ; store the results in the destination
-
-        add         rsp,        16                 ; next line
-        add         rdi,        rdx
-
-        cmp         rdi,        rcx
-        jne         .next_row8x8
-
-    ;add rsp, 144
-    pop rsp
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-SECTION_RODATA
-align 16
-rd:
-    times 8 dw 0x40
diff --git a/vp9/common/x86/vp9_subpixel_ssse3.asm b/vp9/common/x86/vp9_subpixel_ssse3.asm
deleted file mode 100644
index b260480e0..000000000
--- a/vp9/common/x86/vp9_subpixel_ssse3.asm
+++ /dev/null
@@ -1,1515 +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"
-
-%define BLOCK_HEIGHT_WIDTH 4
-%define VP9_FILTER_WEIGHT 128
-%define VP9_FILTER_SHIFT  7
-
-
-;/************************************************************************************
-; Notes: filter_block1d_h6 applies a 6 tap filter horizontally to the input pixels. The
-; input pixel array has output_height rows. This routine assumes that output_height is an
-; even number. This function handles 8 pixels in horizontal direction, calculating ONE
-; rows each iteration to take advantage of the 128 bits operations.
-;
-; This is an implementation of some of the SSE optimizations first seen in ffvp8
-;
-;*************************************************************************************/
-;void vp9_filter_block1d8_h6_ssse3
-;(
-;    unsigned char  *src_ptr,
-;    unsigned int    src_pixels_per_line,
-;    unsigned char *output_ptr,
-;    unsigned int    output_pitch,
-;    unsigned int    output_height,
-;    unsigned int    vp9_filter_index
-;)
-global sym(vp9_filter_block1d8_h6_ssse3) PRIVATE
-sym(vp9_filter_block1d8_h6_ssse3):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    movsxd      rdx, DWORD PTR arg(5)   ;table index
-    xor         rsi, rsi
-    shl         rdx, 4
-
-    movdqa      xmm7, [GLOBAL(rd)]
-
-    lea         rax, [GLOBAL(k0_k5)]
-    add         rax, rdx
-    mov         rdi, arg(2)             ;output_ptr
-
-    cmp         esi, DWORD PTR [rax]
-    je          vp9_filter_block1d8_h4_ssse3
-
-    movdqa      xmm4, XMMWORD PTR [rax]         ;k0_k5
-    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
-
-    mov         rsi, arg(0)             ;src_ptr
-    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
-    movsxd      rcx, dword ptr arg(4)   ;output_height
-
-    movsxd      rdx, dword ptr arg(3)   ;output_pitch
-
-    sub         rdi, rdx
-;xmm3 free
-.filter_block1d8_h6_rowloop_ssse3:
-    movq        xmm0,   MMWORD PTR [rsi - 2]    ; -2 -1  0  1  2  3  4  5
-
-    movq        xmm2,   MMWORD PTR [rsi + 3]    ;  3  4  5  6  7  8  9 10
-
-    punpcklbw   xmm0,   xmm2                    ; -2  3 -1  4  0  5  1  6  2  7  3  8  4  9  5 10
-
-    movdqa      xmm1,   xmm0
-    pmaddubsw   xmm0,   xmm4
-
-    movdqa      xmm2,   xmm1
-    pshufb      xmm1,   [GLOBAL(shuf2bfrom1)]
-
-    pshufb      xmm2,   [GLOBAL(shuf3bfrom1)]
-    pmaddubsw   xmm1,   xmm5
-
-    lea         rdi,    [rdi + rdx]
-    pmaddubsw   xmm2,   xmm6
-
-    lea         rsi,    [rsi + rax]
-    dec         rcx
-
-    paddsw      xmm0,   xmm1
-    paddsw      xmm2,   xmm7
-
-    paddsw      xmm0,   xmm2
-
-    psraw       xmm0,   7
-
-    packuswb    xmm0,   xmm0
-
-    movq        MMWORD Ptr [rdi], xmm0
-    jnz         .filter_block1d8_h6_rowloop_ssse3
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-vp9_filter_block1d8_h4_ssse3:
-    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
-
-    movdqa      xmm3, XMMWORD PTR [GLOBAL(shuf2bfrom1)]
-    movdqa      xmm4, XMMWORD PTR [GLOBAL(shuf3bfrom1)]
-
-    mov         rsi, arg(0)             ;src_ptr
-
-    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
-    movsxd      rcx, dword ptr arg(4)   ;output_height
-
-    movsxd      rdx, dword ptr arg(3)   ;output_pitch
-
-    sub         rdi, rdx
-
-.filter_block1d8_h4_rowloop_ssse3:
-    movq        xmm0,   MMWORD PTR [rsi - 2]    ; -2 -1  0  1  2  3  4  5
-
-    movq        xmm1,   MMWORD PTR [rsi + 3]    ;  3  4  5  6  7  8  9 10
-
-    punpcklbw   xmm0,   xmm1                    ; -2  3 -1  4  0  5  1  6  2  7  3  8  4  9  5 10
-
-    movdqa      xmm2,   xmm0
-    pshufb      xmm0,   xmm3
-
-    pshufb      xmm2,   xmm4
-    pmaddubsw   xmm0,   xmm5
-
-    lea         rdi,    [rdi + rdx]
-    pmaddubsw   xmm2,   xmm6
-
-    lea         rsi,    [rsi + rax]
-    dec         rcx
-
-    paddsw      xmm0,   xmm7
-
-    paddsw      xmm0,   xmm2
-
-    psraw       xmm0,   7
-
-    packuswb    xmm0,   xmm0
-
-    movq        MMWORD Ptr [rdi], xmm0
-
-    jnz         .filter_block1d8_h4_rowloop_ssse3
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-;void vp9_filter_block1d16_h6_ssse3
-;(
-;    unsigned char  *src_ptr,
-;    unsigned int    src_pixels_per_line,
-;    unsigned char  *output_ptr,
-;    unsigned int    output_pitch,
-;    unsigned int    output_height,
-;    unsigned int    vp9_filter_index
-;)
-global sym(vp9_filter_block1d16_h6_ssse3) PRIVATE
-sym(vp9_filter_block1d16_h6_ssse3):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    movsxd      rdx, DWORD PTR arg(5)           ;table index
-    xor         rsi, rsi
-    shl         rdx, 4      ;
-
-    lea         rax, [GLOBAL(k0_k5)]
-    add         rax, rdx
-
-    mov         rdi, arg(2)                     ;output_ptr
-
-    mov         rsi, arg(0)                     ;src_ptr
-
-    movdqa      xmm4, XMMWORD PTR [rax]         ;k0_k5
-    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
-
-    movsxd      rax, dword ptr arg(1)           ;src_pixels_per_line
-    movsxd      rcx, dword ptr arg(4)           ;output_height
-    movsxd      rdx, dword ptr arg(3)           ;output_pitch
-
-.filter_block1d16_h6_rowloop_ssse3:
-    movq        xmm0,   MMWORD PTR [rsi - 2]    ; -2 -1  0  1  2  3  4  5
-
-    movq        xmm3,   MMWORD PTR [rsi + 3]    ;  3  4  5  6  7  8  9 10
-
-    punpcklbw   xmm0,   xmm3                    ; -2  3 -1  4  0  5  1  6  2  7  3  8  4  9  5 10
-
-    movdqa      xmm1,   xmm0
-    pmaddubsw   xmm0,   xmm4
-
-    movdqa      xmm2,   xmm1
-    pshufb      xmm1,   [GLOBAL(shuf2bfrom1)]
-
-    pshufb      xmm2,   [GLOBAL(shuf3bfrom1)]
-    movq        xmm3,   MMWORD PTR [rsi +  6]
-
-    pmaddubsw   xmm1,   xmm5
-    movq        xmm7,   MMWORD PTR [rsi + 11]
-
-    pmaddubsw   xmm2,   xmm6
-    punpcklbw   xmm3,   xmm7
-
-    paddsw      xmm0,   xmm1
-    movdqa      xmm1,   xmm3
-
-    pmaddubsw   xmm3,   xmm4
-    paddsw      xmm0,   xmm2
-
-    movdqa      xmm2,   xmm1
-    paddsw      xmm0,   [GLOBAL(rd)]
-
-    pshufb      xmm1,   [GLOBAL(shuf2bfrom1)]
-    pshufb      xmm2,   [GLOBAL(shuf3bfrom1)]
-
-    psraw       xmm0,   7
-    pmaddubsw   xmm1,   xmm5
-
-    pmaddubsw   xmm2,   xmm6
-    packuswb    xmm0,   xmm0
-
-    lea         rsi,    [rsi + rax]
-    paddsw      xmm3,   xmm1
-
-    paddsw      xmm3,   xmm2
-
-    paddsw      xmm3,   [GLOBAL(rd)]
-
-    psraw       xmm3,   7
-
-    packuswb    xmm3,   xmm3
-
-    punpcklqdq  xmm0,   xmm3
-
-    movdqa      XMMWORD Ptr [rdi], xmm0
-
-    lea         rdi,    [rdi + rdx]
-    dec         rcx
-    jnz         .filter_block1d16_h6_rowloop_ssse3
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-;void vp9_filter_block1d4_h6_ssse3
-;(
-;    unsigned char  *src_ptr,
-;    unsigned int    src_pixels_per_line,
-;    unsigned char  *output_ptr,
-;    unsigned int    output_pitch,
-;    unsigned int    output_height,
-;    unsigned int    vp9_filter_index
-;)
-global sym(vp9_filter_block1d4_h6_ssse3) PRIVATE
-sym(vp9_filter_block1d4_h6_ssse3):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    movsxd      rdx, DWORD PTR arg(5)   ;table index
-    xor         rsi, rsi
-    shl         rdx, 4      ;
-
-    lea         rax, [GLOBAL(k0_k5)]
-    add         rax, rdx
-    movdqa      xmm7, [GLOBAL(rd)]
-
-    cmp         esi, DWORD PTR [rax]
-    je          .vp9_filter_block1d4_h4_ssse3
-
-    movdqa      xmm4, XMMWORD PTR [rax]         ;k0_k5
-    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
-
-    mov         rsi, arg(0)             ;src_ptr
-    mov         rdi, arg(2)             ;output_ptr
-    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
-    movsxd      rcx, dword ptr arg(4)   ;output_height
-
-    movsxd      rdx, dword ptr arg(3)   ;output_pitch
-
-;xmm3 free
-.filter_block1d4_h6_rowloop_ssse3:
-    movdqu      xmm0,   XMMWORD PTR [rsi - 2]
-
-    movdqa      xmm1, xmm0
-    pshufb      xmm0, [GLOBAL(shuf1b)]
-
-    movdqa      xmm2, xmm1
-    pshufb      xmm1, [GLOBAL(shuf2b)]
-    pmaddubsw   xmm0, xmm4
-    pshufb      xmm2, [GLOBAL(shuf3b)]
-    pmaddubsw   xmm1, xmm5
-
-;--
-    pmaddubsw   xmm2, xmm6
-
-    lea         rsi,    [rsi + rax]
-;--
-    paddsw      xmm0, xmm1
-    paddsw      xmm0, xmm7
-    pxor        xmm1, xmm1
-    paddsw      xmm0, xmm2
-    psraw       xmm0, 7
-    packuswb    xmm0, xmm0
-
-    movd        DWORD PTR [rdi], xmm0
-
-    add         rdi, rdx
-    dec         rcx
-    jnz         .filter_block1d4_h6_rowloop_ssse3
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-.vp9_filter_block1d4_h4_ssse3:
-    movdqa      xmm5, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm6, XMMWORD PTR [rax+128]     ;k1_k3
-    movdqa      xmm0, XMMWORD PTR [GLOBAL(shuf2b)]
-    movdqa      xmm3, XMMWORD PTR [GLOBAL(shuf3b)]
-
-    mov         rsi, arg(0)             ;src_ptr
-    mov         rdi, arg(2)             ;output_ptr
-    movsxd      rax, dword ptr arg(1)   ;src_pixels_per_line
-    movsxd      rcx, dword ptr arg(4)   ;output_height
-
-    movsxd      rdx, dword ptr arg(3)   ;output_pitch
-
-.filter_block1d4_h4_rowloop_ssse3:
-    movdqu      xmm1,   XMMWORD PTR [rsi - 2]
-
-    movdqa      xmm2, xmm1
-    pshufb      xmm1, xmm0 ;;[GLOBAL(shuf2b)]
-    pshufb      xmm2, xmm3 ;;[GLOBAL(shuf3b)]
-    pmaddubsw   xmm1, xmm5
-
-;--
-    pmaddubsw   xmm2, xmm6
-
-    lea         rsi,    [rsi + rax]
-;--
-    paddsw      xmm1, xmm7
-    paddsw      xmm1, xmm2
-    psraw       xmm1, 7
-    packuswb    xmm1, xmm1
-
-    movd        DWORD PTR [rdi], xmm1
-
-    add         rdi, rdx
-    dec         rcx
-    jnz         .filter_block1d4_h4_rowloop_ssse3
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-
-
-;void vp9_filter_block1d16_v6_ssse3
-;(
-;    unsigned char *src_ptr,
-;    unsigned int   src_pitch,
-;    unsigned char *output_ptr,
-;    unsigned int   out_pitch,
-;    unsigned int   output_height,
-;    unsigned int   vp9_filter_index
-;)
-global sym(vp9_filter_block1d16_v6_ssse3) PRIVATE
-sym(vp9_filter_block1d16_v6_ssse3):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    movsxd      rdx, DWORD PTR arg(5)   ;table index
-    xor         rsi, rsi
-    shl         rdx, 4      ;
-
-    lea         rax, [GLOBAL(k0_k5)]
-    add         rax, rdx
-
-    cmp         esi, DWORD PTR [rax]
-    je          .vp9_filter_block1d16_v4_ssse3
-
-    movdqa      xmm5, XMMWORD PTR [rax]         ;k0_k5
-    movdqa      xmm6, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm7, XMMWORD PTR [rax+128]     ;k1_k3
-
-    mov         rsi, arg(0)             ;src_ptr
-    movsxd      rdx, DWORD PTR arg(1)   ;pixels_per_line
-    mov         rdi, arg(2)             ;output_ptr
-
-%if ABI_IS_32BIT=0
-    movsxd      r8, DWORD PTR arg(3)    ;out_pitch
-%endif
-    mov         rax, rsi
-    movsxd      rcx, DWORD PTR arg(4)   ;output_height
-    add         rax, rdx
-
-
-.vp9_filter_block1d16_v6_ssse3_loop:
-    movq        xmm1, MMWORD PTR [rsi]                  ;A
-    movq        xmm2, MMWORD PTR [rsi + rdx]            ;B
-    movq        xmm3, MMWORD PTR [rsi + rdx * 2]        ;C
-    movq        xmm4, MMWORD PTR [rax + rdx * 2]        ;D
-    movq        xmm0, MMWORD PTR [rsi + rdx * 4]        ;E
-
-    punpcklbw   xmm2, xmm4                  ;B D
-    punpcklbw   xmm3, xmm0                  ;C E
-
-    movq        xmm0, MMWORD PTR [rax + rdx * 4]        ;F
-
-    pmaddubsw   xmm3, xmm6
-    punpcklbw   xmm1, xmm0                  ;A F
-    pmaddubsw   xmm2, xmm7
-    pmaddubsw   xmm1, xmm5
-
-    paddsw      xmm2, xmm3
-    paddsw      xmm2, xmm1
-    paddsw      xmm2, [GLOBAL(rd)]
-    psraw       xmm2, 7
-    packuswb    xmm2, xmm2
-
-    movq        MMWORD PTR [rdi], xmm2          ;store the results
-
-    movq        xmm1, MMWORD PTR [rsi + 8]                  ;A
-    movq        xmm2, MMWORD PTR [rsi + rdx + 8]            ;B
-    movq        xmm3, MMWORD PTR [rsi + rdx * 2 + 8]        ;C
-    movq        xmm4, MMWORD PTR [rax + rdx * 2 + 8]        ;D
-    movq        xmm0, MMWORD PTR [rsi + rdx * 4 + 8]        ;E
-
-    punpcklbw   xmm2, xmm4                  ;B D
-    punpcklbw   xmm3, xmm0                  ;C E
-
-    movq        xmm0, MMWORD PTR [rax + rdx * 4 + 8]        ;F
-    pmaddubsw   xmm3, xmm6
-    punpcklbw   xmm1, xmm0                  ;A F
-    pmaddubsw   xmm2, xmm7
-    pmaddubsw   xmm1, xmm5
-
-    add         rsi,  rdx
-    add         rax,  rdx
-;--
-;--
-    paddsw      xmm2, xmm3
-    paddsw      xmm2, xmm1
-    paddsw      xmm2, [GLOBAL(rd)]
-    psraw       xmm2, 7
-    packuswb    xmm2, xmm2
-
-    movq        MMWORD PTR [rdi+8], xmm2
-
-%if ABI_IS_32BIT
-    add         rdi,        DWORD PTR arg(3) ;out_pitch
-%else
-    add         rdi,        r8
-%endif
-    dec         rcx
-    jnz         .vp9_filter_block1d16_v6_ssse3_loop
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-.vp9_filter_block1d16_v4_ssse3:
-    movdqa      xmm6, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm7, XMMWORD PTR [rax+128]     ;k1_k3
-
-    mov         rsi, arg(0)             ;src_ptr
-    movsxd      rdx, DWORD PTR arg(1)   ;pixels_per_line
-    mov         rdi, arg(2)             ;output_ptr
-
-%if ABI_IS_32BIT=0
-    movsxd      r8, DWORD PTR arg(3)    ;out_pitch
-%endif
-    mov         rax, rsi
-    movsxd      rcx, DWORD PTR arg(4)   ;output_height
-    add         rax, rdx
-
-.vp9_filter_block1d16_v4_ssse3_loop:
-    movq        xmm2, MMWORD PTR [rsi + rdx]            ;B
-    movq        xmm3, MMWORD PTR [rsi + rdx * 2]        ;C
-    movq        xmm4, MMWORD PTR [rax + rdx * 2]        ;D
-    movq        xmm0, MMWORD PTR [rsi + rdx * 4]        ;E
-
-    punpcklbw   xmm2, xmm4                  ;B D
-    punpcklbw   xmm3, xmm0                  ;C E
-
-    pmaddubsw   xmm3, xmm6
-    pmaddubsw   xmm2, xmm7
-    movq        xmm5, MMWORD PTR [rsi + rdx + 8]            ;B
-    movq        xmm1, MMWORD PTR [rsi + rdx * 2 + 8]        ;C
-    movq        xmm4, MMWORD PTR [rax + rdx * 2 + 8]        ;D
-    movq        xmm0, MMWORD PTR [rsi + rdx * 4 + 8]        ;E
-
-    paddsw      xmm2, [GLOBAL(rd)]
-    paddsw      xmm2, xmm3
-    psraw       xmm2, 7
-    packuswb    xmm2, xmm2
-
-    punpcklbw   xmm5, xmm4                  ;B D
-    punpcklbw   xmm1, xmm0                  ;C E
-
-    pmaddubsw   xmm1, xmm6
-    pmaddubsw   xmm5, xmm7
-
-    movdqa      xmm4, [GLOBAL(rd)]
-    add         rsi,  rdx
-    add         rax,  rdx
-;--
-;--
-    paddsw      xmm5, xmm1
-    paddsw      xmm5, xmm4
-    psraw       xmm5, 7
-    packuswb    xmm5, xmm5
-
-    punpcklqdq  xmm2, xmm5
-
-    movdqa       XMMWORD PTR [rdi], xmm2
-
-%if ABI_IS_32BIT
-    add         rdi,        DWORD PTR arg(3) ;out_pitch
-%else
-    add         rdi,        r8
-%endif
-    dec         rcx
-    jnz         .vp9_filter_block1d16_v4_ssse3_loop
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-;void vp9_filter_block1d8_v6_ssse3
-;(
-;    unsigned char *src_ptr,
-;    unsigned int   src_pitch,
-;    unsigned char *output_ptr,
-;    unsigned int   out_pitch,
-;    unsigned int   output_height,
-;    unsigned int   vp9_filter_index
-;)
-global sym(vp9_filter_block1d8_v6_ssse3) PRIVATE
-sym(vp9_filter_block1d8_v6_ssse3):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    movsxd      rdx, DWORD PTR arg(5)   ;table index
-    xor         rsi, rsi
-    shl         rdx, 4      ;
-
-    lea         rax, [GLOBAL(k0_k5)]
-    add         rax, rdx
-
-    movsxd      rdx, DWORD PTR arg(1)   ;pixels_per_line
-    mov         rdi, arg(2)             ;output_ptr
-%if ABI_IS_32BIT=0
-    movsxd      r8, DWORD PTR arg(3)    ; out_pitch
-%endif
-    movsxd      rcx, DWORD PTR arg(4)   ;[output_height]
-
-    cmp         esi, DWORD PTR [rax]
-    je          .vp9_filter_block1d8_v4_ssse3
-
-    movdqa      xmm5, XMMWORD PTR [rax]         ;k0_k5
-    movdqa      xmm6, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm7, XMMWORD PTR [rax+128]     ;k1_k3
-
-    mov         rsi, arg(0)             ;src_ptr
-
-    mov         rax, rsi
-    add         rax, rdx
-
-.vp9_filter_block1d8_v6_ssse3_loop:
-    movq        xmm1, MMWORD PTR [rsi]                  ;A
-    movq        xmm2, MMWORD PTR [rsi + rdx]            ;B
-    movq        xmm3, MMWORD PTR [rsi + rdx * 2]        ;C
-    movq        xmm4, MMWORD PTR [rax + rdx * 2]        ;D
-    movq        xmm0, MMWORD PTR [rsi + rdx * 4]        ;E
-
-    punpcklbw   xmm2, xmm4                  ;B D
-    punpcklbw   xmm3, xmm0                  ;C E
-
-    movq        xmm0, MMWORD PTR [rax + rdx * 4]        ;F
-    movdqa      xmm4, [GLOBAL(rd)]
-
-    pmaddubsw   xmm3, xmm6
-    punpcklbw   xmm1, xmm0                  ;A F
-    pmaddubsw   xmm2, xmm7
-    pmaddubsw   xmm1, xmm5
-    add         rsi,  rdx
-    add         rax,  rdx
-;--
-;--
-    paddsw      xmm2, xmm3
-    paddsw      xmm2, xmm1
-    paddsw      xmm2, xmm4
-    psraw       xmm2, 7
-    packuswb    xmm2, xmm2
-
-    movq        MMWORD PTR [rdi], xmm2
-
-%if ABI_IS_32BIT
-    add         rdi,        DWORD PTR arg(3) ;[out_pitch]
-%else
-    add         rdi,        r8
-%endif
-    dec         rcx
-    jnz         .vp9_filter_block1d8_v6_ssse3_loop
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-.vp9_filter_block1d8_v4_ssse3:
-    movdqa      xmm6, XMMWORD PTR [rax+256]     ;k2_k4
-    movdqa      xmm7, XMMWORD PTR [rax+128]     ;k1_k3
-    movdqa      xmm5, [GLOBAL(rd)]
-
-    mov         rsi, arg(0)             ;src_ptr
-
-    mov         rax, rsi
-    add         rax, rdx
-
-.vp9_filter_block1d8_v4_ssse3_loop:
-    movq        xmm2, MMWORD PTR [rsi + rdx]            ;B
-    movq        xmm3, MMWORD PTR [rsi + rdx * 2]        ;C
-    movq        xmm4, MMWORD PTR [rax + rdx * 2]        ;D
-    movq        xmm0, MMWORD PTR [rsi + rdx * 4]        ;E
-
-    punpcklbw   xmm2, xmm4                  ;B D
-    punpcklbw   xmm3, xmm0                  ;C E
-
-    pmaddubsw   xmm3, xmm6
-    pmaddubsw   xmm2, xmm7
-    add         rsi,  rdx
-    add         rax,  rdx
-;--
-;--
-    paddsw      xmm2, xmm3
-    paddsw      xmm2, xmm5
-    psraw       xmm2, 7
-    packuswb    xmm2, xmm2
-
-    movq        MMWORD PTR [rdi], xmm2
-
-%if ABI_IS_32BIT
-    add         rdi,        DWORD PTR arg(3) ;[out_pitch]
-%else
-    add         rdi,        r8
-%endif
-    dec         rcx
-    jnz         .vp9_filter_block1d8_v4_ssse3_loop
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-;void vp9_filter_block1d4_v6_ssse3
-;(
-;    unsigned char *src_ptr,
-;    unsigned int   src_pitch,
-;    unsigned char *output_ptr,
-;    unsigned int   out_pitch,
-;    unsigned int   output_height,
-;    unsigned int   vp9_filter_index
-;)
-global sym(vp9_filter_block1d4_v6_ssse3) PRIVATE
-sym(vp9_filter_block1d4_v6_ssse3):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    movsxd      rdx, DWORD PTR arg(5)   ;table index
-    xor         rsi, rsi
-    shl         rdx, 4      ;
-
-    lea         rax, [GLOBAL(k0_k5)]
-    add         rax, rdx
-
-    movsxd      rdx, DWORD PTR arg(1)   ;pixels_per_line
-    mov         rdi, arg(2)             ;output_ptr
-%if ABI_IS_32BIT=0
-    movsxd      r8, DWORD PTR arg(3)    ; out_pitch
-%endif
-    movsxd      rcx, DWORD PTR arg(4)   ;[output_height]
-
-    cmp         esi, DWORD PTR [rax]
-    je          .vp9_filter_block1d4_v4_ssse3
-
-    movq        mm5, MMWORD PTR [rax]         ;k0_k5
-    movq        mm6, MMWORD PTR [rax+256]     ;k2_k4
-    movq        mm7, MMWORD PTR [rax+128]     ;k1_k3
-
-    mov         rsi, arg(0)             ;src_ptr
-
-    mov         rax, rsi
-    add         rax, rdx
-
-.vp9_filter_block1d4_v6_ssse3_loop:
-    movd        mm1, DWORD PTR [rsi]                  ;A
-    movd        mm2, DWORD PTR [rsi + rdx]            ;B
-    movd        mm3, DWORD PTR [rsi + rdx * 2]        ;C
-    movd        mm4, DWORD PTR [rax + rdx * 2]        ;D
-    movd        mm0, DWORD PTR [rsi + rdx * 4]        ;E
-
-    punpcklbw   mm2, mm4                  ;B D
-    punpcklbw   mm3, mm0                  ;C E
-
-    movd        mm0, DWORD PTR [rax + rdx * 4]        ;F
-
-    movq        mm4, [GLOBAL(rd)]
-
-    pmaddubsw   mm3, mm6
-    punpcklbw   mm1, mm0                  ;A F
-    pmaddubsw   mm2, mm7
-    pmaddubsw   mm1, mm5
-    add         rsi,  rdx
-    add         rax,  rdx
-;--
-;--
-    paddsw      mm2, mm3
-    paddsw      mm2, mm1
-    paddsw      mm2, mm4
-    psraw       mm2, 7
-    packuswb    mm2, mm2
-
-    movd        DWORD PTR [rdi], mm2
-
-%if ABI_IS_32BIT
-    add         rdi,        DWORD PTR arg(3) ;[out_pitch]
-%else
-    add         rdi,        r8
-%endif
-    dec         rcx
-    jnz         .vp9_filter_block1d4_v6_ssse3_loop
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-.vp9_filter_block1d4_v4_ssse3:
-    movq        mm6, MMWORD PTR [rax+256]     ;k2_k4
-    movq        mm7, MMWORD PTR [rax+128]     ;k1_k3
-    movq        mm5, MMWORD PTR [GLOBAL(rd)]
-
-    mov         rsi, arg(0)             ;src_ptr
-
-    mov         rax, rsi
-    add         rax, rdx
-
-.vp9_filter_block1d4_v4_ssse3_loop:
-    movd        mm2, DWORD PTR [rsi + rdx]            ;B
-    movd        mm3, DWORD PTR [rsi + rdx * 2]        ;C
-    movd        mm4, DWORD PTR [rax + rdx * 2]        ;D
-    movd        mm0, DWORD PTR [rsi + rdx * 4]        ;E
-
-    punpcklbw   mm2, mm4                  ;B D
-    punpcklbw   mm3, mm0                  ;C E
-
-    pmaddubsw   mm3, mm6
-    pmaddubsw   mm2, mm7
-    add         rsi,  rdx
-    add         rax,  rdx
-;--
-;--
-    paddsw      mm2, mm3
-    paddsw      mm2, mm5
-    psraw       mm2, 7
-    packuswb    mm2, mm2
-
-    movd        DWORD PTR [rdi], mm2
-
-%if ABI_IS_32BIT
-    add         rdi,        DWORD PTR arg(3) ;[out_pitch]
-%else
-    add         rdi,        r8
-%endif
-    dec         rcx
-    jnz         .vp9_filter_block1d4_v4_ssse3_loop
-
-    ; begin epilog
-    pop rdi
-    pop rsi
-    RESTORE_GOT
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-;void vp9_bilinear_predict16x16_ssse3
-;(
-;    unsigned char  *src_ptr,
-;    int   src_pixels_per_line,
-;    int  xoffset,
-;    int  yoffset,
-;    unsigned char *dst_ptr,
-;    int dst_pitch
-;)
-global sym(vp9_bilinear_predict16x16_ssse3) PRIVATE
-sym(vp9_bilinear_predict16x16_ssse3):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-        lea         rcx,        [GLOBAL(bilinear_filters_ssse3)]
-        movsxd      rax,        dword ptr arg(2)    ; xoffset
-
-        cmp         rax,        0                   ; skip first_pass filter if xoffset=0
-        je          .b16x16_sp_only
-
-        shl         rax,        4
-        lea         rax,        [rax + rcx]         ; HFilter
-
-        mov         rdi,        arg(4)              ; dst_ptr
-        mov         rsi,        arg(0)              ; src_ptr
-        movsxd      rdx,        dword ptr arg(5)    ; dst_pitch
-
-        movdqa      xmm1,       [rax]
-
-        movsxd      rax,        dword ptr arg(3)    ; yoffset
-
-        cmp         rax,        0                   ; skip second_pass filter if yoffset=0
-        je          .b16x16_fp_only
-
-        shl         rax,        4
-        lea         rax,        [rax + rcx]         ; VFilter
-
-        lea         rcx,        [rdi+rdx*8]
-        lea         rcx,        [rcx+rdx*8]
-        movsxd      rdx,        dword ptr arg(1)    ; src_pixels_per_line
-
-        movdqa      xmm2,       [rax]
-
-%if ABI_IS_32BIT=0
-        movsxd      r8,         dword ptr arg(5)    ; dst_pitch
-%endif
-        movq        xmm3,       [rsi]               ; 00 01 02 03 04 05 06 07
-        movq        xmm5,       [rsi+1]             ; 01 02 03 04 05 06 07 08
-
-        punpcklbw   xmm3,       xmm5                ; 00 01 01 02 02 03 03 04 04 05 05 06 06 07 07 08
-        movq        xmm4,       [rsi+8]             ; 08 09 10 11 12 13 14 15
-
-        movq        xmm5,       [rsi+9]             ; 09 10 11 12 13 14 15 16
-
-        lea         rsi,        [rsi + rdx]         ; next line
-
-        pmaddubsw   xmm3,       xmm1                ; 00 02 04 06 08 10 12 14
-
-        punpcklbw   xmm4,       xmm5                ; 08 09 09 10 10 11 11 12 12 13 13 14 14 15 15 16
-        pmaddubsw   xmm4,       xmm1                ; 01 03 05 07 09 11 13 15
-
-        paddw       xmm3,       [GLOBAL(rd)]        ; xmm3 += round value
-        psraw       xmm3,       VP9_FILTER_SHIFT    ; xmm3 /= 128
-
-        paddw       xmm4,       [GLOBAL(rd)]        ; xmm4 += round value
-        psraw       xmm4,       VP9_FILTER_SHIFT    ; xmm4 /= 128
-
-        movdqa      xmm7,       xmm3
-        packuswb    xmm7,       xmm4                ; 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
-
-.next_row:
-        movq        xmm6,       [rsi]               ; 00 01 02 03 04 05 06 07
-        movq        xmm5,       [rsi+1]             ; 01 02 03 04 05 06 07 08
-
-        punpcklbw   xmm6,       xmm5
-        movq        xmm4,       [rsi+8]             ; 08 09 10 11 12 13 14 15
-
-        movq        xmm5,       [rsi+9]             ; 09 10 11 12 13 14 15 16
-        lea         rsi,        [rsi + rdx]         ; next line
-
-        pmaddubsw   xmm6,       xmm1
-
-        punpcklbw   xmm4,       xmm5
-        pmaddubsw   xmm4,       xmm1
-
-        paddw       xmm6,       [GLOBAL(rd)]        ; xmm6 += round value
-        psraw       xmm6,       VP9_FILTER_SHIFT    ; xmm6 /= 128
-
-        paddw       xmm4,       [GLOBAL(rd)]        ; xmm4 += round value
-        psraw       xmm4,       VP9_FILTER_SHIFT    ; xmm4 /= 128
-
-        packuswb    xmm6,       xmm4
-        movdqa      xmm5,       xmm7
-
-        punpcklbw   xmm5,       xmm6
-        pmaddubsw   xmm5,       xmm2
-
-        punpckhbw   xmm7,       xmm6
-        pmaddubsw   xmm7,       xmm2
-
-        paddw       xmm5,       [GLOBAL(rd)]        ; xmm5 += round value
-        psraw       xmm5,       VP9_FILTER_SHIFT    ; xmm5 /= 128
-
-        paddw       xmm7,       [GLOBAL(rd)]        ; xmm7 += round value
-        psraw       xmm7,       VP9_FILTER_SHIFT    ; xmm7 /= 128
-
-        packuswb    xmm5,       xmm7
-        movdqa      xmm7,       xmm6
-
-        movdqa      [rdi],      xmm5                ; store the results in the destination
-%if ABI_IS_32BIT
-        add         rdi,        DWORD PTR arg(5)    ; dst_pitch
-%else
-        add         rdi,        r8
-%endif
-
-        cmp         rdi,        rcx
-        jne         .next_row
-
-        jmp         .done
-
-.b16x16_sp_only:
-        movsxd      rax,        dword ptr arg(3)    ; yoffset
-        shl         rax,        4
-        lea         rax,        [rax + rcx]         ; VFilter
-
-        mov         rdi,        arg(4)              ; dst_ptr
-        mov         rsi,        arg(0)              ; src_ptr
-        movsxd      rdx,        dword ptr arg(5)    ; dst_pitch
-
-        movdqa      xmm1,       [rax]               ; VFilter
-
-        lea         rcx,        [rdi+rdx*8]
-        lea         rcx,        [rcx+rdx*8]
-        movsxd      rax,        dword ptr arg(1)    ; src_pixels_per_line
-
-        ; get the first horizontal line done
-        movq        xmm4,       [rsi]               ; load row 0
-        movq        xmm2,       [rsi + 8]           ; load row 0
-
-        lea         rsi,        [rsi + rax]         ; next line
-.next_row_sp:
-        movq        xmm3,       [rsi]               ; load row + 1
-        movq        xmm5,       [rsi + 8]           ; load row + 1
-
-        punpcklbw   xmm4,       xmm3
-        punpcklbw   xmm2,       xmm5
-
-        pmaddubsw   xmm4,       xmm1
-        movq        xmm7,       [rsi + rax]         ; load row + 2
-
-        pmaddubsw   xmm2,       xmm1
-        movq        xmm6,       [rsi + rax + 8]     ; load row + 2
-
-        punpcklbw   xmm3,       xmm7
-        punpcklbw   xmm5,       xmm6
-
-        pmaddubsw   xmm3,       xmm1
-        paddw       xmm4,       [GLOBAL(rd)]
-
-        pmaddubsw   xmm5,       xmm1
-        paddw       xmm2,       [GLOBAL(rd)]
-
-        psraw       xmm4,       VP9_FILTER_SHIFT
-        psraw       xmm2,       VP9_FILTER_SHIFT
-
-        packuswb    xmm4,       xmm2
-        paddw       xmm3,       [GLOBAL(rd)]
-
-        movdqa      [rdi],      xmm4                ; store row 0
-        paddw       xmm5,       [GLOBAL(rd)]
-
-        psraw       xmm3,       VP9_FILTER_SHIFT
-        psraw       xmm5,       VP9_FILTER_SHIFT
-
-        packuswb    xmm3,       xmm5
-        movdqa      xmm4,       xmm7
-
-        movdqa      [rdi + rdx],xmm3                ; store row 1
-        lea         rsi,        [rsi + 2*rax]
-
-        movdqa      xmm2,       xmm6
-        lea         rdi,        [rdi + 2*rdx]
-
-        cmp         rdi,        rcx
-        jne         .next_row_sp
-
-        jmp         .done
-
-.b16x16_fp_only:
-        lea         rcx,        [rdi+rdx*8]
-        lea         rcx,        [rcx+rdx*8]
-        movsxd      rax,        dword ptr arg(1)    ; src_pixels_per_line
-
-.next_row_fp:
-        movq        xmm2,       [rsi]               ; 00 01 02 03 04 05 06 07
-        movq        xmm4,       [rsi+1]             ; 01 02 03 04 05 06 07 08
-
-        punpcklbw   xmm2,       xmm4
-        movq        xmm3,       [rsi+8]             ; 08 09 10 11 12 13 14 15
-
-        pmaddubsw   xmm2,       xmm1
-        movq        xmm4,       [rsi+9]             ; 09 10 11 12 13 14 15 16
-
-        lea         rsi,        [rsi + rax]         ; next line
-        punpcklbw   xmm3,       xmm4
-
-        pmaddubsw   xmm3,       xmm1
-        movq        xmm5,       [rsi]
-
-        paddw       xmm2,       [GLOBAL(rd)]
-        movq        xmm7,       [rsi+1]
-
-        movq        xmm6,       [rsi+8]
-        psraw       xmm2,       VP9_FILTER_SHIFT
-
-        punpcklbw   xmm5,       xmm7
-        movq        xmm7,       [rsi+9]
-
-        paddw       xmm3,       [GLOBAL(rd)]
-        pmaddubsw   xmm5,       xmm1
-
-        psraw       xmm3,       VP9_FILTER_SHIFT
-        punpcklbw   xmm6,       xmm7
-
-        packuswb    xmm2,       xmm3
-        pmaddubsw   xmm6,       xmm1
-
-        movdqa      [rdi],      xmm2                ; store the results in the destination
-        paddw       xmm5,       [GLOBAL(rd)]
-
-        lea         rdi,        [rdi + rdx]         ; dst_pitch
-        psraw       xmm5,       VP9_FILTER_SHIFT
-
-        paddw       xmm6,       [GLOBAL(rd)]
-        psraw       xmm6,       VP9_FILTER_SHIFT
-
-        packuswb    xmm5,       xmm6
-        lea         rsi,        [rsi + rax]         ; next line
-
-        movdqa      [rdi],      xmm5                ; store the results in the destination
-        lea         rdi,        [rdi + rdx]         ; dst_pitch
-
-        cmp         rdi,        rcx
-
-        jne         .next_row_fp
-
-.done:
-    ; begin epilog
-    pop         rdi
-    pop         rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-;void vp9_bilinear_predict8x8_ssse3
-;(
-;    unsigned char  *src_ptr,
-;    int   src_pixels_per_line,
-;    int  xoffset,
-;    int  yoffset,
-;    unsigned char *dst_ptr,
-;    int dst_pitch
-;)
-global sym(vp9_bilinear_predict8x8_ssse3) PRIVATE
-sym(vp9_bilinear_predict8x8_ssse3):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 6
-    SAVE_XMM 7
-    GET_GOT     rbx
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    ALIGN_STACK 16, rax
-    sub         rsp, 144                         ; reserve 144 bytes
-
-        lea         rcx,        [GLOBAL(bilinear_filters_ssse3)]
-
-        mov         rsi,        arg(0) ;src_ptr
-        movsxd      rdx,        dword ptr arg(1) ;src_pixels_per_line
-
-    ;Read 9-line unaligned data in and put them on stack. This gives a big
-    ;performance boost.
-        movdqu      xmm0,       [rsi]
-        lea         rax,        [rdx + rdx*2]
-        movdqu      xmm1,       [rsi+rdx]
-        movdqu      xmm2,       [rsi+rdx*2]
-        add         rsi,        rax
-        movdqu      xmm3,       [rsi]
-        movdqu      xmm4,       [rsi+rdx]
-        movdqu      xmm5,       [rsi+rdx*2]
-        add         rsi,        rax
-        movdqu      xmm6,       [rsi]
-        movdqu      xmm7,       [rsi+rdx]
-
-        movdqa      XMMWORD PTR [rsp],            xmm0
-
-        movdqu      xmm0,       [rsi+rdx*2]
-
-        movdqa      XMMWORD PTR [rsp+16],         xmm1
-        movdqa      XMMWORD PTR [rsp+32],         xmm2
-        movdqa      XMMWORD PTR [rsp+48],         xmm3
-        movdqa      XMMWORD PTR [rsp+64],         xmm4
-        movdqa      XMMWORD PTR [rsp+80],         xmm5
-        movdqa      XMMWORD PTR [rsp+96],         xmm6
-        movdqa      XMMWORD PTR [rsp+112],        xmm7
-        movdqa      XMMWORD PTR [rsp+128],        xmm0
-
-        movsxd      rax,        dword ptr arg(2)    ; xoffset
-        cmp         rax,        0                   ; skip first_pass filter if xoffset=0
-        je          .b8x8_sp_only
-
-        shl         rax,        4
-        add         rax,        rcx                 ; HFilter
-
-        mov         rdi,        arg(4)              ; dst_ptr
-        movsxd      rdx,        dword ptr arg(5)    ; dst_pitch
-
-        movdqa      xmm0,       [rax]
-
-        movsxd      rax,        dword ptr arg(3)    ; yoffset
-        cmp         rax,        0                   ; skip second_pass filter if yoffset=0
-        je          .b8x8_fp_only
-
-        shl         rax,        4
-        lea         rax,        [rax + rcx]         ; VFilter
-
-        lea         rcx,        [rdi+rdx*8]
-
-        movdqa      xmm1,       [rax]
-
-        ; get the first horizontal line done
-        movdqa      xmm3,       [rsp]               ; 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
-        movdqa      xmm5,       xmm3                ; 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 xx
-
-        psrldq      xmm5,       1
-        lea         rsp,        [rsp + 16]          ; next line
-
-        punpcklbw   xmm3,       xmm5                ; 00 01 01 02 02 03 03 04 04 05 05 06 06 07 07 08
-        pmaddubsw   xmm3,       xmm0                ; 00 02 04 06 08 10 12 14
-
-        paddw       xmm3,       [GLOBAL(rd)]        ; xmm3 += round value
-        psraw       xmm3,       VP9_FILTER_SHIFT    ; xmm3 /= 128
-
-        movdqa      xmm7,       xmm3
-        packuswb    xmm7,       xmm7                ; 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
-
-.next_row:
-        movdqa      xmm6,       [rsp]               ; 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15
-        lea         rsp,        [rsp + 16]          ; next line
-
-        movdqa      xmm5,       xmm6
-
-        psrldq      xmm5,       1
-
-        punpcklbw   xmm6,       xmm5
-        pmaddubsw   xmm6,       xmm0
-
-        paddw       xmm6,       [GLOBAL(rd)]        ; xmm6 += round value
-        psraw       xmm6,       VP9_FILTER_SHIFT    ; xmm6 /= 128
-
-        packuswb    xmm6,       xmm6
-
-        punpcklbw   xmm7,       xmm6
-        pmaddubsw   xmm7,       xmm1
-
-        paddw       xmm7,       [GLOBAL(rd)]        ; xmm7 += round value
-        psraw       xmm7,       VP9_FILTER_SHIFT    ; xmm7 /= 128
-
-        packuswb    xmm7,       xmm7
-
-        movq        [rdi],      xmm7                ; store the results in the destination
-        lea         rdi,        [rdi + rdx]
-
-        movdqa      xmm7,       xmm6
-
-        cmp         rdi,        rcx
-        jne         .next_row
-
-        jmp         .done8x8
-
-.b8x8_sp_only:
-        movsxd      rax,        dword ptr arg(3)    ; yoffset
-        shl         rax,        4
-        lea         rax,        [rax + rcx]         ; VFilter
-
-        mov         rdi,        arg(4) ;dst_ptr
-        movsxd      rdx,        dword ptr arg(5)    ; dst_pitch
-
-        movdqa      xmm0,       [rax]               ; VFilter
-
-        movq        xmm1,       XMMWORD PTR [rsp]
-        movq        xmm2,       XMMWORD PTR [rsp+16]
-
-        movq        xmm3,       XMMWORD PTR [rsp+32]
-        punpcklbw   xmm1,       xmm2
-
-        movq        xmm4,       XMMWORD PTR [rsp+48]
-        punpcklbw   xmm2,       xmm3
-
-        movq        xmm5,       XMMWORD PTR [rsp+64]
-        punpcklbw   xmm3,       xmm4
-
-        movq        xmm6,       XMMWORD PTR [rsp+80]
-        punpcklbw   xmm4,       xmm5
-
-        movq        xmm7,       XMMWORD PTR [rsp+96]
-        punpcklbw   xmm5,       xmm6
-
-        pmaddubsw   xmm1,       xmm0
-        pmaddubsw   xmm2,       xmm0
-
-        pmaddubsw   xmm3,       xmm0
-        pmaddubsw   xmm4,       xmm0
-
-        pmaddubsw   xmm5,       xmm0
-        punpcklbw   xmm6,       xmm7
-
-        pmaddubsw   xmm6,       xmm0
-        paddw       xmm1,       [GLOBAL(rd)]
-
-        paddw       xmm2,       [GLOBAL(rd)]
-        psraw       xmm1,       VP9_FILTER_SHIFT
-
-        paddw       xmm3,       [GLOBAL(rd)]
-        psraw       xmm2,       VP9_FILTER_SHIFT
-
-        paddw       xmm4,       [GLOBAL(rd)]
-        psraw       xmm3,       VP9_FILTER_SHIFT
-
-        paddw       xmm5,       [GLOBAL(rd)]
-        psraw       xmm4,       VP9_FILTER_SHIFT
-
-        paddw       xmm6,       [GLOBAL(rd)]
-        psraw       xmm5,       VP9_FILTER_SHIFT
-
-        psraw       xmm6,       VP9_FILTER_SHIFT
-        packuswb    xmm1,       xmm1
-
-        packuswb    xmm2,       xmm2
-        movq        [rdi],      xmm1
-
-        packuswb    xmm3,       xmm3
-        movq        [rdi+rdx],  xmm2
-
-        packuswb    xmm4,       xmm4
-        movq        xmm1,       XMMWORD PTR [rsp+112]
-
-        lea         rdi,        [rdi + 2*rdx]
-        movq        xmm2,       XMMWORD PTR [rsp+128]
-
-        packuswb    xmm5,       xmm5
-        movq        [rdi],      xmm3
-
-        packuswb    xmm6,       xmm6
-        movq        [rdi+rdx],  xmm4
-
-        lea         rdi,        [rdi + 2*rdx]
-        punpcklbw   xmm7,       xmm1
-
-        movq        [rdi],      xmm5
-        pmaddubsw   xmm7,       xmm0
-
-        movq        [rdi+rdx],  xmm6
-        punpcklbw   xmm1,       xmm2
-
-        pmaddubsw   xmm1,       xmm0
-        paddw       xmm7,       [GLOBAL(rd)]
-
-        psraw       xmm7,       VP9_FILTER_SHIFT
-        paddw       xmm1,       [GLOBAL(rd)]
-
-        psraw       xmm1,       VP9_FILTER_SHIFT
-        packuswb    xmm7,       xmm7
-
-        packuswb    xmm1,       xmm1
-        lea         rdi,        [rdi + 2*rdx]
-
-        movq        [rdi],      xmm7
-
-        movq        [rdi+rdx],  xmm1
-        lea         rsp,        [rsp + 144]
-
-        jmp         .done8x8
-
-.b8x8_fp_only:
-        lea         rcx,        [rdi+rdx*8]
-
-.next_row_fp:
-        movdqa      xmm1,       XMMWORD PTR [rsp]
-        movdqa      xmm3,       XMMWORD PTR [rsp+16]
-
-        movdqa      xmm2,       xmm1
-        movdqa      xmm5,       XMMWORD PTR [rsp+32]
-
-        psrldq      xmm2,       1
-        movdqa      xmm7,       XMMWORD PTR [rsp+48]
-
-        movdqa      xmm4,       xmm3
-        psrldq      xmm4,       1
-
-        movdqa      xmm6,       xmm5
-        psrldq      xmm6,       1
-
-        punpcklbw   xmm1,       xmm2
-        pmaddubsw   xmm1,       xmm0
-
-        punpcklbw   xmm3,       xmm4
-        pmaddubsw   xmm3,       xmm0
-
-        punpcklbw   xmm5,       xmm6
-        pmaddubsw   xmm5,       xmm0
-
-        movdqa      xmm2,       xmm7
-        psrldq      xmm2,       1
-
-        punpcklbw   xmm7,       xmm2
-        pmaddubsw   xmm7,       xmm0
-
-        paddw       xmm1,       [GLOBAL(rd)]
-        psraw       xmm1,       VP9_FILTER_SHIFT
-
-        paddw       xmm3,       [GLOBAL(rd)]
-        psraw       xmm3,       VP9_FILTER_SHIFT
-
-        paddw       xmm5,       [GLOBAL(rd)]
-        psraw       xmm5,       VP9_FILTER_SHIFT
-
-        paddw       xmm7,       [GLOBAL(rd)]
-        psraw       xmm7,       VP9_FILTER_SHIFT
-
-        packuswb    xmm1,       xmm1
-        packuswb    xmm3,       xmm3
-
-        packuswb    xmm5,       xmm5
-        movq        [rdi],      xmm1
-
-        packuswb    xmm7,       xmm7
-        movq        [rdi+rdx],  xmm3
-
-        lea         rdi,        [rdi + 2*rdx]
-        movq        [rdi],      xmm5
-
-        lea         rsp,        [rsp + 4*16]
-        movq        [rdi+rdx],  xmm7
-
-        lea         rdi,        [rdi + 2*rdx]
-        cmp         rdi,        rcx
-
-        jne         .next_row_fp
-
-        lea         rsp,        [rsp + 16]
-
-.done8x8:
-    ;add rsp, 144
-    pop         rsp
-    ; begin epilog
-    pop         rdi
-    pop         rsi
-    RESTORE_GOT
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-SECTION_RODATA
-align 16
-shuf1b:
-    db 0, 5, 1, 6, 2, 7, 3, 8, 4, 9, 5, 10, 6, 11, 7, 12
-shuf2b:
-    db 2, 4, 3, 5, 4, 6, 5, 7, 6, 8, 7, 9, 8, 10, 9, 11
-shuf3b:
-    db 1, 3, 2, 4, 3, 5, 4, 6, 5, 7, 6, 8, 7, 9, 8, 10
-
-align 16
-shuf2bfrom1:
-    db  4, 8, 6, 1, 8, 3, 1, 5, 3, 7, 5, 9, 7,11, 9,13
-align 16
-shuf3bfrom1:
-    db  2, 6, 4, 8, 6, 1, 8, 3, 1, 5, 3, 7, 5, 9, 7,11
-
-align 16
-rd:
-    times 8 dw 0x40
-
-align 16
-k0_k5:
-    times 8 db 0, 0             ;placeholder
-    times 8 db 0, 0
-    times 8 db 2, 1
-    times 8 db 0, 0
-    times 8 db 3, 3
-    times 8 db 0, 0
-    times 8 db 1, 2
-    times 8 db 0, 0
-k1_k3:
-    times 8 db  0,    0         ;placeholder
-    times 8 db  -6,  12
-    times 8 db -11,  36
-    times 8 db  -9,  50
-    times 8 db -16,  77
-    times 8 db  -6,  93
-    times 8 db  -8, 108
-    times 8 db  -1, 123
-k2_k4:
-    times 8 db 128,    0        ;placeholder
-    times 8 db 123,   -1
-    times 8 db 108,   -8
-    times 8 db  93,   -6
-    times 8 db  77,  -16
-    times 8 db  50,   -9
-    times 8 db  36,  -11
-    times 8 db  12,   -6
-align 16
-bilinear_filters_ssse3:
-    times 8 db 128, 0
-    times 8 db 120, 8
-    times 8 db 112, 16
-    times 8 db 104, 24
-    times 8 db 96,  32
-    times 8 db 88,  40
-    times 8 db 80,  48
-    times 8 db 72,  56
-    times 8 db 64,  64
-    times 8 db 56,  72
-    times 8 db 48,  80
-    times 8 db 40,  88
-    times 8 db 32,  96
-    times 8 db 24,  104
-    times 8 db 16,  112
-    times 8 db 8,   120
-
diff --git a/vp9/common/x86/vp9_subpixel_x86.h b/vp9/common/x86/vp9_subpixel_x86.h
deleted file mode 100644
index 25bc26d9b..000000000
--- a/vp9/common/x86/vp9_subpixel_x86.h
+++ /dev/null
@@ -1,109 +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.
- */
-
-
-#ifndef VP9_COMMON_X86_VP9_SUBPIXEL_X86_H_
-#define VP9_COMMON_X86_VP9_SUBPIXEL_X86_H_
-
-/* Note:
- *
- * This platform is commonly built for runtime CPU detection. If you modify
- * any of the function mappings present in this file, be sure to also update
- * them in the function pointer initialization code
- */
-
-#if HAVE_MMX
-extern prototype_subpixel_predict(vp9_sixtap_predict16x16_mmx);
-extern prototype_subpixel_predict(vp9_sixtap_predict8x8_mmx);
-extern prototype_subpixel_predict(vp9_sixtap_predict8x4_mmx);
-extern prototype_subpixel_predict(vp9_sixtap_predict4x4_mmx);
-extern prototype_subpixel_predict(vp9_bilinear_predict16x16_mmx);
-
-#if !CONFIG_RUNTIME_CPU_DETECT
-#undef  vp9_subpix_sixtap16x16
-#define vp9_subpix_sixtap16x16 vp9_sixtap_predict16x16_mmx
-
-#undef  vp9_subpix_sixtap8x8
-#define vp9_subpix_sixtap8x8 vp9_sixtap_predict8x8_mmx
-
-#undef  vp9_subpix_sixtap8x4
-#define vp9_subpix_sixtap8x4 vp9_sixtap_predict8x4_mmx
-
-#undef  vp9_subpix_sixtap4x4
-#define vp9_subpix_sixtap4x4 vp9_sixtap_predict4x4_mmx
-
-#undef  vp9_subpix_bilinear16x16
-#define vp9_subpix_bilinear16x16 vp9_bilinear_predict16x16_mmx
-
-#endif
-#endif
-
-
-#if HAVE_SSE2
-extern prototype_subpixel_predict(vp9_sixtap_predict16x16_sse2);
-extern prototype_subpixel_predict(vp9_sixtap_predict8x8_sse2);
-extern prototype_subpixel_predict(vp9_sixtap_predict8x4_sse2);
-extern prototype_subpixel_predict(vp9_bilinear_predict16x16_sse2);
-extern prototype_subpixel_predict(vp9_bilinear_predict8x8_sse2);
-
-
-#if !CONFIG_RUNTIME_CPU_DETECT
-#undef  vp9_subpix_sixtap16x16
-#define vp9_subpix_sixtap16x16 vp9_sixtap_predict16x16_sse2
-
-#undef  vp9_subpix_sixtap8x8
-#define vp9_subpix_sixtap8x8 vp9_sixtap_predict8x8_sse2
-
-#undef  vp9_subpix_sixtap8x4
-#define vp9_subpix_sixtap8x4 vp9_sixtap_predict8x4_sse2
-
-#undef  vp9_subpix_bilinear16x16
-#define vp9_subpix_bilinear16x16 vp9_bilinear_predict16x16_sse2
-
-#undef  vp9_subpix_bilinear8x8
-#define vp9_subpix_bilinear8x8 vp9_bilinear_predict8x8_sse2
-
-#endif
-#endif
-
-#if HAVE_SSSE3
-extern prototype_subpixel_predict(vp9_sixtap_predict16x16_ssse3);
-extern prototype_subpixel_predict(vp9_sixtap_predict8x8_ssse3);
-extern prototype_subpixel_predict(vp9_sixtap_predict8x4_ssse3);
-extern prototype_subpixel_predict(vp9_sixtap_predict4x4_ssse3);
-extern prototype_subpixel_predict(vp9_bilinear_predict16x16_ssse3);
-extern prototype_subpixel_predict(vp9_bilinear_predict8x8_ssse3);
-
-#if !CONFIG_RUNTIME_CPU_DETECT
-#undef  vp9_subpix_sixtap16x16
-#define vp9_subpix_sixtap16x16 vp9_sixtap_predict16x16_ssse3
-
-#undef  vp9_subpix_sixtap8x8
-#define vp9_subpix_sixtap8x8 vp9_sixtap_predict8x8_ssse3
-
-#undef  vp9_subpix_sixtap8x4
-#define vp9_subpix_sixtap8x4 vp9_sixtap_predict8x4_ssse3
-
-#undef  vp9_subpix_sixtap4x4
-#define vp9_subpix_sixtap4x4 vp9_sixtap_predict4x4_ssse3
-
-
-#undef  vp9_subpix_bilinear16x16
-#define vp9_subpix_bilinear16x16 vp9_bilinear_predict16x16_ssse3
-
-#undef  vp9_subpix_bilinear8x8
-#define vp9_subpix_bilinear8x8 vp9_bilinear_predict8x8_ssse3
-
-#endif
-#endif
-
-
-
-#endif