shithub: libvpx

Download patch

ref: 0462172ccf91a2e16651092d98345d8ed7147e98
parent: 13888e0eef7c6d9337de2ad6eef86aedffe317a7
parent: 54d48955f6b80de5d58e8da52dba97813296b713
author: Ronald S. Bultje <[email protected]>
date: Wed Sep 30 20:57:37 EDT 2015

Merge "vp8: change build_intra_predictors_mby_s to use vpx_dsp."

--- a/test/intrapred_test.cc
+++ b/test/intrapred_test.cc
@@ -216,95 +216,6 @@
   int num_planes_;
 };
 
-typedef void (*IntraPredYFunc)(MACROBLOCKD *x,
-                               uint8_t *yabove_row,
-                               uint8_t *yleft,
-                               int left_stride,
-                               uint8_t *ypred_ptr,
-                               int y_stride);
-
-class IntraPredYTest
-    : public IntraPredBase,
-      public ::testing::TestWithParam<IntraPredYFunc> {
- public:
-  static void SetUpTestCase() {
-    mb_ = reinterpret_cast<MACROBLOCKD*>(
-        vpx_memalign(32, sizeof(MACROBLOCKD)));
-    mi_ = reinterpret_cast<MODE_INFO*>(
-        vpx_memalign(32, sizeof(MODE_INFO)));
-    data_array_ = reinterpret_cast<uint8_t*>(
-        vpx_memalign(kDataAlignment, kDataBufferSize));
-  }
-
-  static void TearDownTestCase() {
-    vpx_free(data_array_);
-    vpx_free(mi_);
-    vpx_free(mb_);
-    data_array_ = NULL;
-  }
-
- protected:
-  static const int kBlockSize = 16;
-  static const int kDataAlignment = 16;
-  static const int kStride = kBlockSize * 3;
-  // We use 48 so that the data pointer of the first pixel in each row of
-  // each macroblock is 16-byte aligned, and this gives us access to the
-  // top-left and top-right corner pixels belonging to the top-left/right
-  // macroblocks.
-  // We use 17 lines so we have one line above us for top-prediction.
-  static const int kDataBufferSize = kStride * (kBlockSize + 1);
-
-  virtual void SetUp() {
-    pred_fn_ = GetParam();
-    SetupMacroblock(mb_, mi_, data_array_, kBlockSize, kStride, 1);
-  }
-
-  virtual void Predict(MB_PREDICTION_MODE mode) {
-    mbptr_->mode_info_context->mbmi.mode = mode;
-    ASM_REGISTER_STATE_CHECK(pred_fn_(mbptr_,
-                                      data_ptr_[0] - kStride,
-                                      data_ptr_[0] - 1, kStride,
-                                      data_ptr_[0], kStride));
-  }
-
-  IntraPredYFunc pred_fn_;
-  static uint8_t* data_array_;
-  static MACROBLOCKD * mb_;
-  static MODE_INFO *mi_;
-};
-
-MACROBLOCKD* IntraPredYTest::mb_ = NULL;
-MODE_INFO* IntraPredYTest::mi_ = NULL;
-uint8_t* IntraPredYTest::data_array_ = NULL;
-
-TEST_P(IntraPredYTest, IntraPredTests) {
-  RunTest();
-}
-
-INSTANTIATE_TEST_CASE_P(C, IntraPredYTest,
-                        ::testing::Values(
-                            vp8_build_intra_predictors_mby_s_c));
-#if HAVE_SSE2
-INSTANTIATE_TEST_CASE_P(SSE2, IntraPredYTest,
-                        ::testing::Values(
-                            vp8_build_intra_predictors_mby_s_sse2));
-#endif
-#if HAVE_SSSE3
-INSTANTIATE_TEST_CASE_P(SSSE3, IntraPredYTest,
-                        ::testing::Values(
-                            vp8_build_intra_predictors_mby_s_ssse3));
-#endif
-#if HAVE_NEON
-INSTANTIATE_TEST_CASE_P(NEON, IntraPredYTest,
-                        ::testing::Values(
-                            vp8_build_intra_predictors_mby_s_neon));
-#endif
-#if HAVE_MSA
-INSTANTIATE_TEST_CASE_P(MSA, IntraPredYTest,
-                        ::testing::Values(
-                            vp8_build_intra_predictors_mby_s_msa));
-#endif
-
 typedef void (*IntraPredUvFunc)(MACROBLOCKD *x,
                                 uint8_t *uabove_row,
                                 uint8_t *vabove_row,
--- a/vp8/common/arm/neon/reconintra_neon.c
+++ b/vp8/common/arm/neon/reconintra_neon.c
@@ -12,92 +12,6 @@
 
 #include "vp8/common/blockd.h"
 
-void vp8_build_intra_predictors_mby_s_neon(MACROBLOCKD *x,
-                                           unsigned char * yabove_row,
-                                           unsigned char * yleft,
-                                           int left_stride,
-                                           unsigned char * ypred_ptr,
-                                           int y_stride) {
-  const int mode = x->mode_info_context->mbmi.mode;
-  int i;
-
-  switch (mode) {
-    case DC_PRED:
-    {
-      int shift = x->up_available + x->left_available;
-      uint8x16_t v_expected_dc = vdupq_n_u8(128);
-
-      if (shift) {
-        unsigned int average = 0;
-        int expected_dc;
-        if (x->up_available) {
-          const uint8x16_t v_above = vld1q_u8(yabove_row);
-          const uint16x8_t a = vpaddlq_u8(v_above);
-          const uint32x4_t b = vpaddlq_u16(a);
-          const uint64x2_t c = vpaddlq_u32(b);
-          const uint32x2_t d = vadd_u32(vreinterpret_u32_u64(vget_low_u64(c)),
-                                        vreinterpret_u32_u64(vget_high_u64(c)));
-          average = vget_lane_u32(d, 0);
-        }
-        if (x->left_available) {
-          for (i = 0; i < 16; ++i) {
-              average += yleft[0];
-              yleft += left_stride;
-          }
-        }
-        shift += 3;
-        expected_dc = (average + (1 << (shift - 1))) >> shift;
-        v_expected_dc = vmovq_n_u8((uint8_t)expected_dc);
-      }
-      for (i = 0; i < 16; ++i) {
-        vst1q_u8(ypred_ptr, v_expected_dc);
-        ypred_ptr += y_stride;
-      }
-    }
-    break;
-    case V_PRED:
-    {
-      const uint8x16_t v_above = vld1q_u8(yabove_row);
-      for (i = 0; i < 16; ++i) {
-        vst1q_u8(ypred_ptr, v_above);
-        ypred_ptr += y_stride;
-      }
-    }
-    break;
-    case H_PRED:
-    {
-      for (i = 0; i < 16; ++i) {
-        const uint8x16_t v_yleft = vmovq_n_u8((uint8_t)yleft[0]);
-        yleft += left_stride;
-        vst1q_u8(ypred_ptr, v_yleft);
-        ypred_ptr += y_stride;
-      }
-    }
-    break;
-    case TM_PRED:
-    {
-      const uint16x8_t v_ytop_left = vmovq_n_u16((int16_t)yabove_row[-1]);
-      const uint8x16_t v_above = vld1q_u8(yabove_row);
-      for (i = 0; i < 16; ++i) {
-        const uint8x8_t v_yleft = vmov_n_u8((int8_t)yleft[0]);
-        const uint16x8_t a_lo = vaddl_u8(vget_low_u8(v_above), v_yleft);
-        const uint16x8_t a_hi = vaddl_u8(vget_high_u8(v_above), v_yleft);
-        const int16x8_t b_lo = vsubq_s16(vreinterpretq_s16_u16(a_lo),
-                                         vreinterpretq_s16_u16(v_ytop_left));
-        const int16x8_t b_hi = vsubq_s16(vreinterpretq_s16_u16(a_hi),
-                                         vreinterpretq_s16_u16(v_ytop_left));
-        const uint8x8_t pred_lo = vqmovun_s16(b_lo);
-        const uint8x8_t pred_hi = vqmovun_s16(b_hi);
-
-        vst1q_u8(ypred_ptr, vcombine_u8(pred_lo, pred_hi));
-        ypred_ptr += y_stride;
-        yleft += left_stride;
-      }
-    }
-    break;
-  }
-}
-
 void vp8_build_intra_predictors_mbuv_s_neon(MACROBLOCKD *x,
                                             unsigned char * uabove_row,
                                             unsigned char * vabove_row,
--- a/vp8/common/mips/msa/reconintra_msa.c
+++ b/vp8/common/mips/msa/reconintra_msa.c
@@ -22,16 +22,6 @@
     SD4(out, out, out, out, dst, dst_stride);
 }
 
-static void intra_predict_vert_16x16_msa(uint8_t *src, uint8_t *dst,
-                                         int32_t dst_stride)
-{
-    v16u8 out = LD_UB(src);
-
-    ST_UB8(out, out, out, out, out, out, out, out, dst, dst_stride);
-    dst += (8 * dst_stride);
-    ST_UB8(out, out, out, out, out, out, out, out, dst, dst_stride);
-}
-
 static void intra_predict_horiz_8x8_msa(uint8_t *src, int32_t src_stride,
                                         uint8_t *dst, int32_t dst_stride)
 {
@@ -51,34 +41,6 @@
     SD4(out4, out5, out6, out7, dst, dst_stride);
 }
 
-static void intra_predict_horiz_16x16_msa(uint8_t *src, int32_t src_stride,
-                                          uint8_t *dst, int32_t dst_stride)
-{
-    uint32_t row;
-    uint8_t inp0, inp1, inp2, inp3;
-    v16u8 src0, src1, src2, src3;
-
-    for (row = 4; row--;)
-    {
-        inp0 = src[0];
-        src += src_stride;
-        inp1 = src[0];
-        src += src_stride;
-        inp2 = src[0];
-        src += src_stride;
-        inp3 = src[0];
-        src += src_stride;
-
-        src0 = (v16u8)__msa_fill_b(inp0);
-        src1 = (v16u8)__msa_fill_b(inp1);
-        src2 = (v16u8)__msa_fill_b(inp2);
-        src3 = (v16u8)__msa_fill_b(inp3);
-
-        ST_UB4(src0, src1, src2, src3, dst, dst_stride);
-        dst += (4 * dst_stride);
-    }
-}
-
 static void intra_predict_dc_8x8_msa(uint8_t *src_top, uint8_t *src_left,
                                      int32_t src_stride_left,
                                      uint8_t *dst, int32_t dst_stride,
@@ -138,128 +100,6 @@
     SD4(out, out, out, out, dst, dst_stride);
     dst += (4 * dst_stride);
     SD4(out, out, out, out, dst, dst_stride);
-}
-
-static void intra_predict_dc_16x16_msa(uint8_t *src_top, uint8_t *src_left,
-                                       int32_t src_stride_left,
-                                       uint8_t *dst, int32_t dst_stride,
-                                       uint8_t is_above, uint8_t is_left)
-{
-    uint32_t row;
-    uint32_t addition = 0;
-    v16u8 src_above, out;
-    v8u16 sum_above;
-    v4u32 sum_top;
-    v2u64 sum;
-
-    if (is_left && is_above)
-    {
-        src_above = LD_UB(src_top);
-
-        sum_above = __msa_hadd_u_h(src_above, src_above);
-        sum_top = __msa_hadd_u_w(sum_above, sum_above);
-        sum = __msa_hadd_u_d(sum_top, sum_top);
-        sum_top = (v4u32)__msa_pckev_w((v4i32)sum, (v4i32)sum);
-        sum = __msa_hadd_u_d(sum_top, sum_top);
-        addition = __msa_copy_u_w((v4i32)sum, 0);
-
-        for (row = 0; row < 16; ++row)
-        {
-            addition += src_left[row * src_stride_left];
-        }
-
-        addition = (addition + 16) >> 5;
-        out = (v16u8)__msa_fill_b(addition);
-    }
-    else if (is_left)
-    {
-        for (row = 0; row < 16; ++row)
-        {
-            addition += src_left[row * src_stride_left];
-        }
-
-        addition = (addition + 8) >> 4;
-        out = (v16u8)__msa_fill_b(addition);
-    }
-    else if (is_above)
-    {
-        src_above = LD_UB(src_top);
-
-        sum_above = __msa_hadd_u_h(src_above, src_above);
-        sum_top = __msa_hadd_u_w(sum_above, sum_above);
-        sum = __msa_hadd_u_d(sum_top, sum_top);
-        sum_top = (v4u32)__msa_pckev_w((v4i32)sum, (v4i32)sum);
-        sum = __msa_hadd_u_d(sum_top, sum_top);
-        sum = (v2u64)__msa_srari_d((v2i64)sum, 4);
-        out = (v16u8)__msa_splati_b((v16i8)sum, 0);
-    }
-    else
-    {
-        out = (v16u8)__msa_ldi_b(128);
-    }
-
-    ST_UB8(out, out, out, out, out, out, out, out, dst, dst_stride);
-    dst += (8 * dst_stride);
-    ST_UB8(out, out, out, out, out, out, out, out, dst, dst_stride);
-}
-
-void vp8_build_intra_predictors_mby_s_msa(struct macroblockd *x,
-                                          unsigned char *yabove_row,
-                                          unsigned char *yleft,
-                                          int left_stride,
-                                          unsigned char *ypred_ptr,
-                                          int y_stride)
-{
-    uint32_t row, col;
-    uint8_t ytop_left = yabove_row[-1];
-
-    switch (x->mode_info_context->mbmi.mode)
-    {
-        case DC_PRED:
-            intra_predict_dc_16x16_msa(yabove_row, yleft, left_stride,
-                                       ypred_ptr, y_stride,
-                                       x->up_available, x->left_available);
-            break;
-
-        case V_PRED:
-            intra_predict_vert_16x16_msa(yabove_row, ypred_ptr, y_stride);
-            break;
-
-        case H_PRED:
-            intra_predict_horiz_16x16_msa(yleft, left_stride, ypred_ptr,
-                                          y_stride);
-            break;
-
-        case TM_PRED:
-            for (row = 0; row < 16; ++row)
-            {
-                for (col = 0; col < 16; ++col)
-                {
-                    int pred = yleft[row * left_stride] + yabove_row[col] -
-                               ytop_left;
-
-                    if (pred < 0)
-                        pred = 0;
-
-                    if (pred > 255)
-                        pred = 255;
-
-                    ypred_ptr[col] = pred;
-                }
-
-                ypred_ptr += y_stride;
-            }
-            break;
-
-        case B_PRED:
-        case NEARESTMV:
-        case NEARMV:
-        case ZEROMV:
-        case NEWMV:
-        case SPLITMV:
-        case MB_MODE_COUNT:
-            break;
-    }
 }
 
 void vp8_build_intra_predictors_mbuv_s_msa(struct macroblockd *x,
--- a/vp8/common/reconintra.c
+++ b/vp8/common/reconintra.c
@@ -9,133 +9,57 @@
  */
 
 
-#include "vpx_config.h"
-#include "vp8_rtcd.h"
+#include "./vpx_config.h"
+#include "./vpx_dsp_rtcd.h"
+#include "./vp8_rtcd.h"
 #include "vpx_mem/vpx_mem.h"
+#include "vpx_ports/vpx_once.h"
 #include "blockd.h"
+#include "vp8/common/reconintra.h"
 
-void vp8_build_intra_predictors_mby_s_c(MACROBLOCKD *x,
-                                          unsigned char * yabove_row,
-                                          unsigned char * yleft,
-                                          int left_stride,
-                                          unsigned char * ypred_ptr,
-                                          int y_stride)
-{
-    unsigned char yleft_col[16];
-    unsigned char ytop_left = yabove_row[-1];
-    int r, c, i;
+typedef void (*intra_pred_fn)(uint8_t *dst, ptrdiff_t stride,
+                              const uint8_t *above, const uint8_t *left);
 
-    for (i = 0; i < 16; i++)
-    {
-        yleft_col[i] = yleft[i* left_stride];
-    }
+static intra_pred_fn pred[4];
+static intra_pred_fn dc_pred[2][2];
 
-    /* for Y */
-    switch (x->mode_info_context->mbmi.mode)
-    {
-    case DC_PRED:
-    {
-        int expected_dc;
-        int shift;
-        int average = 0;
-
-
-        if (x->up_available || x->left_available)
-        {
-            if (x->up_available)
-            {
-                for (i = 0; i < 16; i++)
-                {
-                    average += yabove_row[i];
-                }
-            }
+static void vp8_init_intra_predictors_internal(void)
+{
+    pred[V_PRED] = vpx_v_predictor_16x16;
+    pred[H_PRED] = vpx_h_predictor_16x16;
+    pred[TM_PRED] = vpx_tm_predictor_16x16;
+
+    dc_pred[0][0] = vpx_dc_128_predictor_16x16;
+    dc_pred[0][1] = vpx_dc_top_predictor_16x16;
+    dc_pred[1][0] = vpx_dc_left_predictor_16x16;
+    dc_pred[1][1] = vpx_dc_predictor_16x16;
+}
 
-            if (x->left_available)
-            {
-
-                for (i = 0; i < 16; i++)
-                {
-                    average += yleft_col[i];
-                }
+void vp8_build_intra_predictors_mby_s(MACROBLOCKD *x,
+                                      unsigned char * yabove_row,
+                                      unsigned char * yleft,
+                                      int left_stride,
+                                      unsigned char * ypred_ptr,
+                                      int y_stride)
+{
+    MB_PREDICTION_MODE mode = x->mode_info_context->mbmi.mode;
+    unsigned char yleft_col[16];
+    int i;
 
-            }
-
-
-
-            shift = 3 + x->up_available + x->left_available;
-            expected_dc = (average + (1 << (shift - 1))) >> shift;
-        }
-        else
-        {
-            expected_dc = 128;
-        }
-
-        /*memset(ypred_ptr, expected_dc, 256);*/
-        for (r = 0; r < 16; r++)
-        {
-            memset(ypred_ptr, expected_dc, 16);
-            ypred_ptr += y_stride;
-        }
-    }
-    break;
-    case V_PRED:
+    for (i = 0; i < 16; i++)
     {
-
-        for (r = 0; r < 16; r++)
-        {
-
-            ((int *)ypred_ptr)[0] = ((int *)yabove_row)[0];
-            ((int *)ypred_ptr)[1] = ((int *)yabove_row)[1];
-            ((int *)ypred_ptr)[2] = ((int *)yabove_row)[2];
-            ((int *)ypred_ptr)[3] = ((int *)yabove_row)[3];
-            ypred_ptr += y_stride;
-        }
+        yleft_col[i] = yleft[i* left_stride];
     }
-    break;
-    case H_PRED:
-    {
 
-        for (r = 0; r < 16; r++)
-        {
-
-            memset(ypred_ptr, yleft_col[r], 16);
-            ypred_ptr += y_stride;
-        }
-
+    if (mode == DC_PRED)
+    {
+        dc_pred[x->left_available][x->up_available](ypred_ptr, y_stride,
+                                                    yabove_row, yleft_col);
     }
-    break;
-    case TM_PRED:
+    else
     {
-
-        for (r = 0; r < 16; r++)
-        {
-            for (c = 0; c < 16; c++)
-            {
-                int pred =  yleft_col[r] + yabove_row[ c] - ytop_left;
-
-                if (pred < 0)
-                    pred = 0;
-
-                if (pred > 255)
-                    pred = 255;
-
-                ypred_ptr[c] = pred;
-            }
-
-            ypred_ptr += y_stride;
-        }
-
+        pred[mode](ypred_ptr, y_stride, yabove_row, yleft_col);
     }
-    break;
-    case B_PRED:
-    case NEARESTMV:
-    case NEARMV:
-    case ZEROMV:
-    case NEWMV:
-    case SPLITMV:
-    case MB_MODE_COUNT:
-        break;
-    }
 }
 
 void vp8_build_intra_predictors_mbuv_s_c(MACROBLOCKD *x,
@@ -277,4 +201,9 @@
     case MB_MODE_COUNT:
         break;
     }
+}
+
+void vp8_init_intra_predictors(void)
+{
+    once(vp8_init_intra_predictors_internal);
 }
--- /dev/null
+++ b/vp8/common/reconintra.h
@@ -1,0 +1,34 @@
+/*
+ *  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 VP8_COMMON_RECONINTRA_H_
+#define VP8_COMMON_RECONINTRA_H_
+
+#include "vp8/common/blockd.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void vp8_build_intra_predictors_mby_s(MACROBLOCKD *x,
+                                      unsigned char *yabove_row,
+                                      unsigned char *yleft,
+                                      int left_stride,
+                                      unsigned char *ypred_ptr,
+                                      int y_stride);
+
+void vp8_init_intra_predictors(void);
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+
+#endif  // VP8_COMMON_RECONINTRA_H_
--- a/vp8/common/rtcd_defs.pl
+++ b/vp8/common/rtcd_defs.pl
@@ -152,9 +152,6 @@
 $vp8_copy_mem8x4_media=vp8_copy_mem8x4_v6;
 $vp8_copy_mem8x4_dspr2=vp8_copy_mem8x4_dspr2;
 
-add_proto qw/void vp8_build_intra_predictors_mby_s/, "struct macroblockd *x, unsigned char * yabove_row, unsigned char * yleft, int left_stride, unsigned char * ypred_ptr, int y_stride";
-specialize qw/vp8_build_intra_predictors_mby_s sse2 ssse3 neon msa/;
-
 add_proto qw/void vp8_build_intra_predictors_mbuv_s/, "struct macroblockd *x, unsigned char * uabove_row, unsigned char * vabove_row,  unsigned char *uleft, unsigned char *vleft, int left_stride, unsigned char * upred_ptr, unsigned char * vpred_ptr, int pred_stride";
 specialize qw/vp8_build_intra_predictors_mbuv_s sse2 ssse3 neon msa/;
 
--- a/vp8/common/x86/recon_sse2.asm
+++ b/vp8/common/x86/recon_sse2.asm
@@ -593,519 +593,12 @@
 vp8_intra_pred_uv_ho mmx2
 vp8_intra_pred_uv_ho ssse3
 
-;void vp8_intra_pred_y_dc_sse2(
-;    unsigned char *dst,
-;    int dst_stride
-;    unsigned char *above,
-;    unsigned char *left,
-;    int left_stride
-;    )
-global sym(vp8_intra_pred_y_dc_sse2) PRIVATE
-sym(vp8_intra_pred_y_dc_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    ; from top
-    mov         rdi,        arg(2) ;above
-    mov         rsi,        arg(3) ;left
-    movsxd      rax,        dword ptr arg(4) ;left_stride;
-
-    pxor        xmm0,       xmm0
-    movdqa      xmm1,       [rdi]
-    psadbw      xmm1,       xmm0
-    movq        xmm2,       xmm1
-    punpckhqdq  xmm1,       xmm1
-    paddw       xmm1,       xmm2
-
-    ; from left
-    lea         rdi,        [rax*3]
-
-    movzx       ecx,        byte [rsi]
-    movzx       edx,        byte [rsi+rax]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax*2]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rdi]
-    add         ecx,        edx
-    lea         rsi,        [rsi+rax*4]
-
-    movzx       edx,        byte [rsi]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax*2]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rdi]
-    add         ecx,        edx
-    lea         rsi,        [rsi+rax*4]
-
-    movzx       edx,        byte [rsi]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax*2]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rdi]
-    add         ecx,        edx
-    lea         rsi,        [rsi+rax*4]
-
-    movzx       edx,        byte [rsi]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax*2]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rdi]
-    add         ecx,        edx
-
-    ; add up
-    pextrw      edx,        xmm1, 0x0
-    lea         edx,        [edx+ecx+16]
-    sar         edx,        5
-    movd        xmm1,       edx
-    ; FIXME use pshufb for ssse3 version
-    pshuflw     xmm1,       xmm1, 0x0
-    punpcklqdq  xmm1,       xmm1
-    packuswb    xmm1,       xmm1
-
-    ; write out
-    mov         rsi,        2
-    mov         rdi,        arg(0) ;dst;
-    movsxd      rcx,        dword ptr arg(1) ;dst_stride
-    lea         rax,        [rcx*3]
-
-.label
-    movdqa [rdi      ],     xmm1
-    movdqa [rdi+rcx  ],     xmm1
-    movdqa [rdi+rcx*2],     xmm1
-    movdqa [rdi+rax  ],     xmm1
-    lea         rdi,        [rdi+rcx*4]
-    movdqa [rdi      ],     xmm1
-    movdqa [rdi+rcx  ],     xmm1
-    movdqa [rdi+rcx*2],     xmm1
-    movdqa [rdi+rax  ],     xmm1
-    lea         rdi,        [rdi+rcx*4]
-    dec         rsi
-    jnz .label
-
-    ; begin epilog
-    pop         rdi
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-;void vp8_intra_pred_y_dctop_sse2(
-;    unsigned char *dst,
-;    int dst_stride
-;    unsigned char *above,
-;    unsigned char *left,
-;    int left_stride
-;    )
-global sym(vp8_intra_pred_y_dctop_sse2) PRIVATE
-sym(vp8_intra_pred_y_dctop_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    GET_GOT     rbx
-    ; end prolog
-
-    ;arg(3), arg(4) not used
-
-    ; from top
-    mov         rcx,        arg(2) ;above;
-    pxor        xmm0,       xmm0
-    movdqa      xmm1,       [rcx]
-    psadbw      xmm1,       xmm0
-    movdqa      xmm2,       xmm1
-    punpckhqdq  xmm1,       xmm1
-    paddw       xmm1,       xmm2
-
-    ; add up
-    paddw       xmm1,       [GLOBAL(dc_8)]
-    psraw       xmm1,       4
-    ; FIXME use pshufb for ssse3 version
-    pshuflw     xmm1,       xmm1, 0x0
-    punpcklqdq  xmm1,       xmm1
-    packuswb    xmm1,       xmm1
-
-    ; write out
-    mov         rsi,        2
-    mov         rdx,        arg(0) ;dst;
-    movsxd      rcx,        dword ptr arg(1) ;dst_stride
-    lea         rax,        [rcx*3]
-
-.label
-    movdqa [rdx      ],     xmm1
-    movdqa [rdx+rcx  ],     xmm1
-    movdqa [rdx+rcx*2],     xmm1
-    movdqa [rdx+rax  ],     xmm1
-    lea         rdx,        [rdx+rcx*4]
-    movdqa [rdx      ],     xmm1
-    movdqa [rdx+rcx  ],     xmm1
-    movdqa [rdx+rcx*2],     xmm1
-    movdqa [rdx+rax  ],     xmm1
-    lea         rdx,        [rdx+rcx*4]
-    dec         rsi
-    jnz .label
-
-    ; begin epilog
-    RESTORE_GOT
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-;void vp8_intra_pred_y_dcleft_sse2(
-;    unsigned char *dst,
-;    int dst_stride
-;    unsigned char *above,
-;    unsigned char *left,
-;    int left_stride
-;    )
-global sym(vp8_intra_pred_y_dcleft_sse2) PRIVATE
-sym(vp8_intra_pred_y_dcleft_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    push        rdi
-    ; end prolog
-
-    ;arg(2) not used
-
-    ; from left
-    mov         rsi,        arg(3) ;left;
-    movsxd      rax,        dword ptr arg(4) ;left_stride;
-
-    lea         rdi,        [rax*3]
-    movzx       ecx,        byte [rsi]
-    movzx       edx,        byte [rsi+rax]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax*2]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rdi]
-    add         ecx,        edx
-    lea         rsi,        [rsi+rax*4]
-    movzx       edx,        byte [rsi]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax*2]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rdi]
-    add         ecx,        edx
-    lea         rsi,        [rsi+rax*4]
-    movzx       edx,        byte [rsi]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax*2]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rdi]
-    add         ecx,        edx
-    lea         rsi,        [rsi+rax*4]
-    movzx       edx,        byte [rsi]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rax*2]
-    add         ecx,        edx
-    movzx       edx,        byte [rsi+rdi]
-    lea         edx,        [ecx+edx+8]
-
-    ; add up
-    shr         edx,        4
-    movd        xmm1,       edx
-    ; FIXME use pshufb for ssse3 version
-    pshuflw     xmm1,       xmm1, 0x0
-    punpcklqdq  xmm1,       xmm1
-    packuswb    xmm1,       xmm1
-
-    ; write out
-    mov         rsi,        2
-    mov         rdi,        arg(0) ;dst;
-    movsxd      rcx,        dword ptr arg(1) ;dst_stride
-    lea         rax,        [rcx*3]
-
-.label
-    movdqa [rdi      ],     xmm1
-    movdqa [rdi+rcx  ],     xmm1
-    movdqa [rdi+rcx*2],     xmm1
-    movdqa [rdi+rax  ],     xmm1
-    lea         rdi,        [rdi+rcx*4]
-    movdqa [rdi      ],     xmm1
-    movdqa [rdi+rcx  ],     xmm1
-    movdqa [rdi+rcx*2],     xmm1
-    movdqa [rdi+rax  ],     xmm1
-    lea         rdi,        [rdi+rcx*4]
-    dec         rsi
-    jnz .label
-
-    ; begin epilog
-    pop         rdi
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-;void vp8_intra_pred_y_dc128_sse2(
-;    unsigned char *dst,
-;    int dst_stride
-;    unsigned char *above,
-;    unsigned char *left,
-;    int left_stride
-;    )
-global sym(vp8_intra_pred_y_dc128_sse2) PRIVATE
-sym(vp8_intra_pred_y_dc128_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    GET_GOT     rbx
-    ; end prolog
-
-    ;arg(2), arg(3), arg(4) not used
-
-    ; write out
-    mov         rsi,        2
-    movdqa      xmm1,       [GLOBAL(dc_128)]
-    mov         rax,        arg(0) ;dst;
-    movsxd      rdx,        dword ptr arg(1) ;dst_stride
-    lea         rcx,        [rdx*3]
-
-.label
-    movdqa [rax      ],     xmm1
-    movdqa [rax+rdx  ],     xmm1
-    movdqa [rax+rdx*2],     xmm1
-    movdqa [rax+rcx  ],     xmm1
-    lea         rax,        [rax+rdx*4]
-    movdqa [rax      ],     xmm1
-    movdqa [rax+rdx  ],     xmm1
-    movdqa [rax+rdx*2],     xmm1
-    movdqa [rax+rcx  ],     xmm1
-    lea         rax,        [rax+rdx*4]
-    dec         rsi
-    jnz .label
-
-    ; begin epilog
-    RESTORE_GOT
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-;void vp8_intra_pred_y_tm_sse2(
-;    unsigned char *dst,
-;    int dst_stride
-;    unsigned char *above,
-;    unsigned char *left,
-;    int left_stride
-;    )
-%macro vp8_intra_pred_y_tm 1
-global sym(vp8_intra_pred_y_tm_%1) PRIVATE
-sym(vp8_intra_pred_y_tm_%1):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    SAVE_XMM 7
-    push        rsi
-    push        rdi
-    push        rbx
-    GET_GOT     rbx
-    ; end prolog
-
-    ; read top row
-    mov         edx,        8
-    mov         rsi,        arg(2) ;above
-    movsxd      rax,        dword ptr arg(4) ;left_stride;
-    pxor        xmm0,       xmm0
-%ifidn %1, ssse3
-    movdqa      xmm3,       [GLOBAL(dc_1024)]
-%endif
-    movdqa      xmm1,       [rsi]
-    movdqa      xmm2,       xmm1
-    punpcklbw   xmm1,       xmm0
-    punpckhbw   xmm2,       xmm0
-
-    ; set up left ptrs ans subtract topleft
-    movd        xmm4,       [rsi-1]
-    mov         rsi,        arg(3) ;left
-%ifidn %1, sse2
-    punpcklbw   xmm4,       xmm0
-    pshuflw     xmm4,       xmm4, 0x0
-    punpcklqdq  xmm4,       xmm4
-%else
-    pshufb      xmm4,       xmm3
-%endif
-    psubw       xmm1,       xmm4
-    psubw       xmm2,       xmm4
-
-    ; set up dest ptrs
-    mov         rdi,        arg(0) ;dst;
-    movsxd      rcx,        dword ptr arg(1) ;dst_stride
-vp8_intra_pred_y_tm_%1_loop:
-    mov         bl,         [rsi]
-    movd        xmm4,       ebx
-
-    mov         bl,         [rsi+rax]
-    movd        xmm5,       ebx
-%ifidn %1, sse2
-    punpcklbw   xmm4,       xmm0
-    punpcklbw   xmm5,       xmm0
-    pshuflw     xmm4,       xmm4, 0x0
-    pshuflw     xmm5,       xmm5, 0x0
-    punpcklqdq  xmm4,       xmm4
-    punpcklqdq  xmm5,       xmm5
-%else
-    pshufb      xmm4,       xmm3
-    pshufb      xmm5,       xmm3
-%endif
-    movdqa      xmm6,       xmm4
-    movdqa      xmm7,       xmm5
-    paddw       xmm4,       xmm1
-    paddw       xmm6,       xmm2
-    paddw       xmm5,       xmm1
-    paddw       xmm7,       xmm2
-    packuswb    xmm4,       xmm6
-    packuswb    xmm5,       xmm7
-    movdqa [rdi    ],       xmm4
-    movdqa [rdi+rcx],       xmm5
-    lea         rsi,        [rsi+rax*2]
-    lea         rdi,        [rdi+rcx*2]
-    dec         edx
-    jnz vp8_intra_pred_y_tm_%1_loop
-
-    ; begin epilog
-    RESTORE_GOT
-    pop         rbx
-    pop         rdi
-    pop         rsi
-    RESTORE_XMM
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-%endmacro
-
-vp8_intra_pred_y_tm sse2
-vp8_intra_pred_y_tm ssse3
-
-;void vp8_intra_pred_y_ve_sse2(
-;    unsigned char *dst,
-;    int dst_stride
-;    unsigned char *above,
-;    unsigned char *left,
-;    int left_stride
-;    )
-global sym(vp8_intra_pred_y_ve_sse2) PRIVATE
-sym(vp8_intra_pred_y_ve_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    ; end prolog
-
-    ;arg(3), arg(4) not used
-
-    mov         rax,        arg(2) ;above;
-    mov         rsi,        2
-    movsxd      rdx,        dword ptr arg(1) ;dst_stride
-
-    ; read from top
-    movdqa      xmm1,       [rax]
-
-    ; write out
-    mov         rax,        arg(0) ;dst;
-    lea         rcx,        [rdx*3]
-
-.label
-    movdqa [rax      ],     xmm1
-    movdqa [rax+rdx  ],     xmm1
-    movdqa [rax+rdx*2],     xmm1
-    movdqa [rax+rcx  ],     xmm1
-    lea         rax,        [rax+rdx*4]
-    movdqa [rax      ],     xmm1
-    movdqa [rax+rdx  ],     xmm1
-    movdqa [rax+rdx*2],     xmm1
-    movdqa [rax+rcx  ],     xmm1
-    lea         rax,        [rax+rdx*4]
-    dec         rsi
-    jnz .label
-
-    ; begin epilog
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
-;void vp8_intra_pred_y_ho_sse2(
-;    unsigned char *dst,
-;    int dst_stride
-;    unsigned char *above,
-;    unsigned char *left,
-;    int left_stride,
-;    )
-global sym(vp8_intra_pred_y_ho_sse2) PRIVATE
-sym(vp8_intra_pred_y_ho_sse2):
-    push        rbp
-    mov         rbp, rsp
-    SHADOW_ARGS_TO_STACK 5
-    push        rsi
-    push        rdi
-    push        rbx
-    ; end prolog
-
-    ;arg(2) not used
-
-    ; read from left and write out
-    mov         edx,        8
-    mov         rsi,        arg(3) ;left;
-    movsxd      rax,        dword ptr arg(4) ;left_stride;
-    mov         rdi,        arg(0) ;dst;
-    movsxd      rcx,        dword ptr arg(1) ;dst_stride
-
-vp8_intra_pred_y_ho_sse2_loop:
-    mov         bl,         [rsi]
-    movd        xmm0,       ebx
-    mov         bl,         [rsi+rax]
-    movd        xmm1,       ebx
-
-    ; FIXME use pshufb for ssse3 version
-    punpcklbw   xmm0,       xmm0
-    punpcklbw   xmm1,       xmm1
-    pshuflw     xmm0,       xmm0, 0x0
-    pshuflw     xmm1,       xmm1, 0x0
-    punpcklqdq  xmm0,       xmm0
-    punpcklqdq  xmm1,       xmm1
-    movdqa [rdi    ],       xmm0
-    movdqa [rdi+rcx],       xmm1
-    lea         rsi,        [rsi+rax*2]
-    lea         rdi,        [rdi+rcx*2]
-    dec         edx
-    jnz vp8_intra_pred_y_ho_sse2_loop
-
-    ; begin epilog
-    pop         rbx
-    pop         rdi
-    pop         rsi
-    UNSHADOW_ARGS
-    pop         rbp
-    ret
-
 SECTION_RODATA
 align 16
 dc_128:
-    times 16 db 128
+    times 8 db 128
 dc_4:
     times 4 dw 4
-align 16
-dc_8:
-    times 8 dw 8
 align 16
 dc_1024:
     times 8 dw 0x400
--- a/vp8/common/x86/recon_wrapper_sse2.c
+++ b/vp8/common/x86/recon_wrapper_sse2.c
@@ -109,78 +109,3 @@
                                         vp8_intra_pred_uv_tm_ssse3,
                                         vp8_intra_pred_uv_ho_ssse3);
 }
-
-#define build_intra_predictors_mby_prototype(sym) \
-    void sym(unsigned char *dst, int dst_stride, \
-             const unsigned char *above, \
-             const unsigned char *left, int left_stride)
-typedef build_intra_predictors_mby_prototype((*build_intra_predictors_mby_fn_t));
-
-extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_dc_sse2);
-extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_dctop_sse2);
-extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_dcleft_sse2);
-extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_dc128_sse2);
-extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_ho_sse2);
-extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_ve_sse2);
-extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_tm_sse2);
-extern build_intra_predictors_mby_prototype(vp8_intra_pred_y_tm_ssse3);
-
-static void vp8_build_intra_predictors_mby_x86(MACROBLOCKD *x,
-                                               unsigned char * yabove_row,
-                                               unsigned char *dst_y,
-                                               int dst_stride,
-                                               unsigned char * yleft,
-                                               int left_stride,
-                                               build_intra_predictors_mby_fn_t tm_func)
-{
-    int mode = x->mode_info_context->mbmi.mode;
-    build_intra_predictors_mbuv_fn_t fn;
-
-    switch (mode) {
-        case  V_PRED: fn = vp8_intra_pred_y_ve_sse2; break;
-        case  H_PRED: fn = vp8_intra_pred_y_ho_sse2; break;
-        case TM_PRED: fn = tm_func; break;
-        case DC_PRED:
-            if (x->up_available) {
-                if (x->left_available) {
-                    fn = vp8_intra_pred_y_dc_sse2; break;
-                } else {
-                    fn = vp8_intra_pred_y_dctop_sse2; break;
-                }
-            } else if (x->left_available) {
-                fn = vp8_intra_pred_y_dcleft_sse2; break;
-            } else {
-                fn = vp8_intra_pred_y_dc128_sse2; break;
-            }
-            break;
-        default: return;
-    }
-
-    fn(dst_y, dst_stride, yabove_row, yleft, left_stride);
-    return;
-}
-
-void vp8_build_intra_predictors_mby_s_sse2(MACROBLOCKD *x,
-                                           unsigned char * yabove_row,
-                                           unsigned char * yleft,
-                                           int left_stride,
-                                           unsigned char * ypred_ptr,
-                                           int y_stride)
-{
-    vp8_build_intra_predictors_mby_x86(x, yabove_row, ypred_ptr,
-                                       y_stride, yleft, left_stride,
-                                       vp8_intra_pred_y_tm_sse2);
-}
-
-void vp8_build_intra_predictors_mby_s_ssse3(MACROBLOCKD *x,
-                                            unsigned char * yabove_row,
-                                            unsigned char * yleft,
-                                            int left_stride,
-                                            unsigned char * ypred_ptr,
-                                            int y_stride)
-{
-    vp8_build_intra_predictors_mby_x86(x, yabove_row, ypred_ptr,
-                                     y_stride, yleft, left_stride,
-                                       vp8_intra_pred_y_tm_ssse3);
-
-}
--- a/vp8/decoder/decodeframe.c
+++ b/vp8/decoder/decodeframe.c
@@ -23,6 +23,7 @@
 #include "vp8/common/entropymode.h"
 #include "vp8/common/quant_common.h"
 #include "vpx_scale/vpx_scale.h"
+#include "vp8/common/reconintra.h"
 #include "vp8/common/setupintrarecon.h"
 
 #include "decodemv.h"
--- a/vp8/decoder/onyxd_if.c
+++ b/vp8/decoder/onyxd_if.c
@@ -25,9 +25,12 @@
 #include <assert.h>
 
 #include "vp8/common/quant_common.h"
+#include "vp8/common/reconintra.h"
+#include "./vpx_dsp_rtcd.h"
 #include "./vpx_scale_rtcd.h"
 #include "vpx_scale/vpx_scale.h"
 #include "vp8/common/systemdependent.h"
+#include "vpx_ports/vpx_once.h"
 #include "vpx_ports/vpx_timer.h"
 #include "detokenize.h"
 #if CONFIG_ERROR_CONCEALMENT
@@ -42,6 +45,17 @@
 static int get_free_fb (VP8_COMMON *cm);
 static void ref_cnt_fb (int *buf, int *idx, int new_idx);
 
+static void initialize_dec(void) {
+    static volatile int init_done = 0;
+
+    if (!init_done)
+    {
+        vpx_dsp_rtcd();
+        vp8_init_intra_predictors();
+        init_done = 1;
+    }
+}
+
 static void remove_decompressor(VP8D_COMP *pbi)
 {
 #if CONFIG_ERROR_CONCEALMENT
@@ -104,6 +118,8 @@
     pbi->independent_partitions = 0;
 
     vp8_setup_block_dptrs(&pbi->mb);
+
+    once(initialize_dec);
 
     return pbi;
 }
--- a/vp8/decoder/threading.c
+++ b/vp8/decoder/threading.c
@@ -24,6 +24,7 @@
 #include "detokenize.h"
 #include "vp8/common/reconintra4x4.h"
 #include "vp8/common/reconinter.h"
+#include "vp8/common/reconintra.h"
 #include "vp8/common/setupintrarecon.h"
 #if CONFIG_ERROR_CONCEALMENT
 #include "error_concealment.h"
--- a/vp8/encoder/encodeintra.c
+++ b/vp8/encoder/encodeintra.c
@@ -13,6 +13,7 @@
 #include "vp8_rtcd.h"
 #include "./vpx_dsp_rtcd.h"
 #include "vp8/encoder/quantize.h"
+#include "vp8/common/reconintra.h"
 #include "vp8/common/reconintra4x4.h"
 #include "encodemb.h"
 #include "vp8/common/invtrans.h"
--- a/vp8/encoder/onyx_if.c
+++ b/vp8/encoder/onyx_if.c
@@ -31,6 +31,7 @@
 #include "vp8/common/postproc.h"
 #endif
 #include "vpx_mem/vpx_mem.h"
+#include "vp8/common/reconintra.h"
 #include "vp8/common/swapyv12buffer.h"
 #include "vp8/common/threading.h"
 #include "vpx_ports/vpx_timer.h"
@@ -422,6 +423,16 @@
 
 static void dealloc_raw_frame_buffers(VP8_COMP *cpi);
 
+void vp8_initialize_enc(void)
+{
+    static volatile int init_done = 0;
+
+    if (!init_done) {
+        vpx_dsp_rtcd();
+        vp8_init_intra_predictors();
+        init_done = 1;
+    }
+}
 
 static void dealloc_compressor_data(VP8_COMP *cpi)
 {
--- a/vp8/encoder/onyx_int.h
+++ b/vp8/encoder/onyx_int.h
@@ -716,6 +716,8 @@
     } rd_costs;
 } VP8_COMP;
 
+void vp8_initialize_enc(void);
+
 void vp8_alloc_compressor_data(VP8_COMP *cpi);
 int vp8_reverse_trans(int x);
 void vp8_new_framerate(VP8_COMP *cpi, double framerate);
--- a/vp8/encoder/pickinter.c
+++ b/vp8/encoder/pickinter.c
@@ -21,6 +21,7 @@
 #include "vp8/common/findnearmv.h"
 #include "encodemb.h"
 #include "vp8/common/reconinter.h"
+#include "vp8/common/reconintra.h"
 #include "vp8/common/reconintra4x4.h"
 #include "vpx_dsp/variance.h"
 #include "mcomp.h"
--- a/vp8/encoder/rdopt.c
+++ b/vp8/encoder/rdopt.c
@@ -24,6 +24,7 @@
 #include "pickinter.h"
 #include "vp8/common/entropymode.h"
 #include "vp8/common/reconinter.h"
+#include "vp8/common/reconintra.h"
 #include "vp8/common/reconintra4x4.h"
 #include "vp8/common/findnearmv.h"
 #include "vp8/common/quant_common.h"
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -45,6 +45,7 @@
 VP8_COMMON_SRCS-yes += common/onyxc_int.h
 VP8_COMMON_SRCS-yes += common/quant_common.h
 VP8_COMMON_SRCS-yes += common/reconinter.h
+VP8_COMMON_SRCS-yes += common/reconintra.h
 VP8_COMMON_SRCS-yes += common/reconintra4x4.h
 VP8_COMMON_SRCS-yes += common/rtcd.c
 VP8_COMMON_SRCS-yes += common/rtcd_defs.pl
--- a/vp8/vp8_cx_iface.c
+++ b/vp8/vp8_cx_iface.c
@@ -17,6 +17,7 @@
 #include "vpx/internal/vpx_codec_internal.h"
 #include "vpx_version.h"
 #include "vpx_mem/vpx_mem.h"
+#include "vpx_ports/vpx_once.h"
 #include "vp8/encoder/onyx_int.h"
 #include "vpx/vp8cx.h"
 #include "vp8/encoder/firstpass.h"
@@ -692,6 +693,8 @@
             ctx->priv->enc.total_encoders   = mr_cfg->mr_total_resolutions;
         else
             ctx->priv->enc.total_encoders   = 1;
+
+        once(vp8_initialize_enc);
 
         res = validate_config(priv, &priv->cfg, &priv->vp8_cfg, 0);
 
--- a/vpx_dsp/vpx_dsp.mk
+++ b/vpx_dsp/vpx_dsp.mk
@@ -36,7 +36,6 @@
 endif
 
 # intra predictions
-ifneq ($(filter yes,$(CONFIG_VP9) $(CONFIG_VP10)),)
 DSP_SRCS-yes += intrapred.c
 
 ifeq ($(CONFIG_USE_X86INC),yes)
@@ -59,7 +58,6 @@
 DSP_SRCS-$(HAVE_DSPR2)  += mips/intrapred4_dspr2.c
 DSP_SRCS-$(HAVE_DSPR2)  += mips/intrapred8_dspr2.c
 DSP_SRCS-$(HAVE_DSPR2)  += mips/intrapred16_dspr2.c
-endif  # CONFIG_VP9 || CONFIG_VP10
 
 DSP_SRCS-$(HAVE_DSPR2)  += mips/common_dspr2.h
 DSP_SRCS-$(HAVE_DSPR2)  += mips/common_dspr2.c
--- a/vpx_dsp/vpx_dsp_rtcd_defs.pl
+++ b/vpx_dsp/vpx_dsp_rtcd_defs.pl
@@ -54,322 +54,320 @@
 # Intra prediction
 #
 
-if ((vpx_config("CONFIG_VP9") eq "yes") || (vpx_config("CONFIG_VP10") eq "yes")) {
-  add_proto qw/void vpx_d207_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d207_predictor_4x4/, "$ssse3_x86inc";
+add_proto qw/void vpx_d207_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d207_predictor_4x4/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d45_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d45_predictor_4x4 neon/, "$ssse3_x86inc";
+add_proto qw/void vpx_d45_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d45_predictor_4x4 neon/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d63_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d63_predictor_4x4/, "$ssse3_x86inc";
+add_proto qw/void vpx_d63_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d63_predictor_4x4/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_h_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_h_predictor_4x4 neon dspr2 msa/, "$ssse3_x86inc";
+add_proto qw/void vpx_h_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_h_predictor_4x4 neon dspr2 msa/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d117_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d117_predictor_4x4/;
+add_proto qw/void vpx_d117_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d117_predictor_4x4/;
 
-  add_proto qw/void vpx_d135_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d135_predictor_4x4 neon/;
+add_proto qw/void vpx_d135_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d135_predictor_4x4 neon/;
 
-  add_proto qw/void vpx_d153_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d153_predictor_4x4/, "$ssse3_x86inc";
+add_proto qw/void vpx_d153_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d153_predictor_4x4/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_v_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_v_predictor_4x4 neon msa/, "$sse_x86inc";
+add_proto qw/void vpx_v_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_v_predictor_4x4 neon msa/, "$sse_x86inc";
 
-  add_proto qw/void vpx_tm_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_tm_predictor_4x4 neon dspr2 msa/, "$sse_x86inc";
+add_proto qw/void vpx_tm_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_tm_predictor_4x4 neon dspr2 msa/, "$sse_x86inc";
 
-  add_proto qw/void vpx_dc_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_predictor_4x4 dspr2 msa neon/, "$sse_x86inc";
+add_proto qw/void vpx_dc_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_predictor_4x4 dspr2 msa neon/, "$sse_x86inc";
 
-  add_proto qw/void vpx_dc_top_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_top_predictor_4x4 msa neon/, "$sse_x86inc";
+add_proto qw/void vpx_dc_top_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_top_predictor_4x4 msa neon/, "$sse_x86inc";
 
-  add_proto qw/void vpx_dc_left_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_left_predictor_4x4 msa neon/, "$sse_x86inc";
+add_proto qw/void vpx_dc_left_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_left_predictor_4x4 msa neon/, "$sse_x86inc";
 
-  add_proto qw/void vpx_dc_128_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_128_predictor_4x4 msa neon/, "$sse_x86inc";
+add_proto qw/void vpx_dc_128_predictor_4x4/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_128_predictor_4x4 msa neon/, "$sse_x86inc";
 
-  add_proto qw/void vpx_d207_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d207_predictor_8x8/, "$ssse3_x86inc";
+add_proto qw/void vpx_d207_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d207_predictor_8x8/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d45_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d45_predictor_8x8 neon/, "$ssse3_x86inc";
+add_proto qw/void vpx_d45_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d45_predictor_8x8 neon/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d63_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d63_predictor_8x8/, "$ssse3_x86inc";
+add_proto qw/void vpx_d63_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d63_predictor_8x8/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_h_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_h_predictor_8x8 neon dspr2 msa/, "$ssse3_x86inc";
+add_proto qw/void vpx_h_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_h_predictor_8x8 neon dspr2 msa/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d117_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d117_predictor_8x8/;
+add_proto qw/void vpx_d117_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d117_predictor_8x8/;
 
-  add_proto qw/void vpx_d135_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d135_predictor_8x8/;
+add_proto qw/void vpx_d135_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d135_predictor_8x8/;
 
-  add_proto qw/void vpx_d153_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d153_predictor_8x8/, "$ssse3_x86inc";
+add_proto qw/void vpx_d153_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d153_predictor_8x8/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_v_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_v_predictor_8x8 neon msa/, "$sse_x86inc";
+add_proto qw/void vpx_v_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_v_predictor_8x8 neon msa/, "$sse_x86inc";
 
-  add_proto qw/void vpx_tm_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_tm_predictor_8x8 neon dspr2 msa/, "$sse2_x86inc";
+add_proto qw/void vpx_tm_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_tm_predictor_8x8 neon dspr2 msa/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_dc_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_predictor_8x8 dspr2 neon msa/, "$sse_x86inc";
+add_proto qw/void vpx_dc_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_predictor_8x8 dspr2 neon msa/, "$sse_x86inc";
 
-  add_proto qw/void vpx_dc_top_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_top_predictor_8x8 neon msa/, "$sse_x86inc";
+add_proto qw/void vpx_dc_top_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_top_predictor_8x8 neon msa/, "$sse_x86inc";
 
-  add_proto qw/void vpx_dc_left_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_left_predictor_8x8 neon msa/, "$sse_x86inc";
+add_proto qw/void vpx_dc_left_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_left_predictor_8x8 neon msa/, "$sse_x86inc";
 
-  add_proto qw/void vpx_dc_128_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_128_predictor_8x8 neon msa/, "$sse_x86inc";
+add_proto qw/void vpx_dc_128_predictor_8x8/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_128_predictor_8x8 neon msa/, "$sse_x86inc";
 
-  add_proto qw/void vpx_d207_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d207_predictor_16x16/, "$ssse3_x86inc";
+add_proto qw/void vpx_d207_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d207_predictor_16x16/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d45_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d45_predictor_16x16 neon/, "$ssse3_x86inc";
+add_proto qw/void vpx_d45_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d45_predictor_16x16 neon/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d63_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d63_predictor_16x16/, "$ssse3_x86inc";
+add_proto qw/void vpx_d63_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d63_predictor_16x16/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_h_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_h_predictor_16x16 neon dspr2 msa/, "$ssse3_x86inc";
+add_proto qw/void vpx_h_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_h_predictor_16x16 neon dspr2 msa/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d117_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d117_predictor_16x16/;
+add_proto qw/void vpx_d117_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d117_predictor_16x16/;
 
-  add_proto qw/void vpx_d135_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d135_predictor_16x16/;
+add_proto qw/void vpx_d135_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d135_predictor_16x16/;
 
-  add_proto qw/void vpx_d153_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d153_predictor_16x16/, "$ssse3_x86inc";
+add_proto qw/void vpx_d153_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d153_predictor_16x16/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_v_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_v_predictor_16x16 neon msa/, "$sse2_x86inc";
+add_proto qw/void vpx_v_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_v_predictor_16x16 neon msa/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_tm_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_tm_predictor_16x16 neon msa/, "$sse2_x86inc";
+add_proto qw/void vpx_tm_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_tm_predictor_16x16 neon msa/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_dc_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_predictor_16x16 dspr2 neon msa/, "$sse2_x86inc";
+add_proto qw/void vpx_dc_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_predictor_16x16 dspr2 neon msa/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_dc_top_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_top_predictor_16x16 neon msa/, "$sse2_x86inc";
+add_proto qw/void vpx_dc_top_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_top_predictor_16x16 neon msa/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_dc_left_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_left_predictor_16x16 neon msa/, "$sse2_x86inc";
+add_proto qw/void vpx_dc_left_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_left_predictor_16x16 neon msa/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_dc_128_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_128_predictor_16x16 neon msa/, "$sse2_x86inc";
+add_proto qw/void vpx_dc_128_predictor_16x16/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_128_predictor_16x16 neon msa/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_d207_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d207_predictor_32x32/, "$ssse3_x86inc";
+add_proto qw/void vpx_d207_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d207_predictor_32x32/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d45_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d45_predictor_32x32/, "$ssse3_x86inc";
+add_proto qw/void vpx_d45_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d45_predictor_32x32/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d63_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d63_predictor_32x32/, "$ssse3_x86inc";
+add_proto qw/void vpx_d63_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d63_predictor_32x32/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_h_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_h_predictor_32x32 neon msa/, "$ssse3_x86inc";
+add_proto qw/void vpx_h_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_h_predictor_32x32 neon msa/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_d117_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d117_predictor_32x32/;
+add_proto qw/void vpx_d117_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d117_predictor_32x32/;
 
-  add_proto qw/void vpx_d135_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d135_predictor_32x32/;
+add_proto qw/void vpx_d135_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d135_predictor_32x32/;
 
-  add_proto qw/void vpx_d153_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_d153_predictor_32x32/, "$ssse3_x86inc";
+add_proto qw/void vpx_d153_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_d153_predictor_32x32/, "$ssse3_x86inc";
 
-  add_proto qw/void vpx_v_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_v_predictor_32x32 neon msa/, "$sse2_x86inc";
+add_proto qw/void vpx_v_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_v_predictor_32x32 neon msa/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_tm_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_tm_predictor_32x32 neon msa/, "$sse2_x86_64_x86inc";
+add_proto qw/void vpx_tm_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_tm_predictor_32x32 neon msa/, "$sse2_x86_64_x86inc";
 
-  add_proto qw/void vpx_dc_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_predictor_32x32 msa neon/, "$sse2_x86inc";
+add_proto qw/void vpx_dc_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_predictor_32x32 msa neon/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_dc_top_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_top_predictor_32x32 msa neon/, "$sse2_x86inc";
+add_proto qw/void vpx_dc_top_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_top_predictor_32x32 msa neon/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_dc_left_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_left_predictor_32x32 msa neon/, "$sse2_x86inc";
+add_proto qw/void vpx_dc_left_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_left_predictor_32x32 msa neon/, "$sse2_x86inc";
 
-  add_proto qw/void vpx_dc_128_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
-  specialize qw/vpx_dc_128_predictor_32x32 msa neon/, "$sse2_x86inc";
+add_proto qw/void vpx_dc_128_predictor_32x32/, "uint8_t *dst, ptrdiff_t y_stride, const uint8_t *above, const uint8_t *left";
+specialize qw/vpx_dc_128_predictor_32x32 msa neon/, "$sse2_x86inc";
 
 # High bitdepth functions
-  if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
-    add_proto qw/void vpx_highbd_d207_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d207_predictor_4x4/;
+if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") {
+  add_proto qw/void vpx_highbd_d207_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d207_predictor_4x4/;
 
-    add_proto qw/void vpx_highbd_d45_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d45_predictor_4x4/;
+  add_proto qw/void vpx_highbd_d45_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d45_predictor_4x4/;
 
-    add_proto qw/void vpx_highbd_d63_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d63_predictor_4x4/;
+  add_proto qw/void vpx_highbd_d63_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d63_predictor_4x4/;
 
-    add_proto qw/void vpx_highbd_h_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_h_predictor_4x4/;
+  add_proto qw/void vpx_highbd_h_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_h_predictor_4x4/;
 
-    add_proto qw/void vpx_highbd_d117_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d117_predictor_4x4/;
+  add_proto qw/void vpx_highbd_d117_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d117_predictor_4x4/;
 
-    add_proto qw/void vpx_highbd_d135_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d135_predictor_4x4/;
+  add_proto qw/void vpx_highbd_d135_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d135_predictor_4x4/;
 
-    add_proto qw/void vpx_highbd_d153_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d153_predictor_4x4/;
+  add_proto qw/void vpx_highbd_d153_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d153_predictor_4x4/;
 
-    add_proto qw/void vpx_highbd_v_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_v_predictor_4x4/, "$sse_x86inc";
+  add_proto qw/void vpx_highbd_v_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_v_predictor_4x4/, "$sse_x86inc";
 
-    add_proto qw/void vpx_highbd_tm_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_tm_predictor_4x4/, "$sse_x86inc";
+  add_proto qw/void vpx_highbd_tm_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_tm_predictor_4x4/, "$sse_x86inc";
 
-    add_proto qw/void vpx_highbd_dc_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_predictor_4x4/, "$sse_x86inc";
+  add_proto qw/void vpx_highbd_dc_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_predictor_4x4/, "$sse_x86inc";
 
-    add_proto qw/void vpx_highbd_dc_top_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_top_predictor_4x4/;
+  add_proto qw/void vpx_highbd_dc_top_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_top_predictor_4x4/;
 
-    add_proto qw/void vpx_highbd_dc_left_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_left_predictor_4x4/;
+  add_proto qw/void vpx_highbd_dc_left_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_left_predictor_4x4/;
 
-    add_proto qw/void vpx_highbd_dc_128_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_128_predictor_4x4/;
+  add_proto qw/void vpx_highbd_dc_128_predictor_4x4/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_128_predictor_4x4/;
 
-    add_proto qw/void vpx_highbd_d207_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d207_predictor_8x8/;
+  add_proto qw/void vpx_highbd_d207_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d207_predictor_8x8/;
 
-    add_proto qw/void vpx_highbd_d45_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d45_predictor_8x8/;
+  add_proto qw/void vpx_highbd_d45_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d45_predictor_8x8/;
 
-    add_proto qw/void vpx_highbd_d63_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d63_predictor_8x8/;
+  add_proto qw/void vpx_highbd_d63_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d63_predictor_8x8/;
 
-    add_proto qw/void vpx_highbd_h_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_h_predictor_8x8/;
+  add_proto qw/void vpx_highbd_h_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_h_predictor_8x8/;
 
-    add_proto qw/void vpx_highbd_d117_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d117_predictor_8x8/;
+  add_proto qw/void vpx_highbd_d117_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d117_predictor_8x8/;
 
-    add_proto qw/void vpx_highbd_d135_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d135_predictor_8x8/;
+  add_proto qw/void vpx_highbd_d135_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d135_predictor_8x8/;
 
-    add_proto qw/void vpx_highbd_d153_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d153_predictor_8x8/;
+  add_proto qw/void vpx_highbd_d153_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d153_predictor_8x8/;
 
-    add_proto qw/void vpx_highbd_v_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_v_predictor_8x8/, "$sse2_x86inc";
+  add_proto qw/void vpx_highbd_v_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_v_predictor_8x8/, "$sse2_x86inc";
 
-    add_proto qw/void vpx_highbd_tm_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_tm_predictor_8x8/, "$sse2_x86inc";
+  add_proto qw/void vpx_highbd_tm_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_tm_predictor_8x8/, "$sse2_x86inc";
 
-    add_proto qw/void vpx_highbd_dc_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_predictor_8x8/, "$sse2_x86inc";;
+  add_proto qw/void vpx_highbd_dc_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_predictor_8x8/, "$sse2_x86inc";;
 
-    add_proto qw/void vpx_highbd_dc_top_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_top_predictor_8x8/;
+  add_proto qw/void vpx_highbd_dc_top_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_top_predictor_8x8/;
 
-    add_proto qw/void vpx_highbd_dc_left_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_left_predictor_8x8/;
+  add_proto qw/void vpx_highbd_dc_left_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_left_predictor_8x8/;
 
-    add_proto qw/void vpx_highbd_dc_128_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_128_predictor_8x8/;
+  add_proto qw/void vpx_highbd_dc_128_predictor_8x8/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_128_predictor_8x8/;
 
-    add_proto qw/void vpx_highbd_d207_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d207_predictor_16x16/;
+  add_proto qw/void vpx_highbd_d207_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d207_predictor_16x16/;
 
-    add_proto qw/void vpx_highbd_d45_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d45_predictor_16x16/;
+  add_proto qw/void vpx_highbd_d45_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d45_predictor_16x16/;
 
-    add_proto qw/void vpx_highbd_d63_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d63_predictor_16x16/;
+  add_proto qw/void vpx_highbd_d63_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d63_predictor_16x16/;
 
-    add_proto qw/void vpx_highbd_h_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_h_predictor_16x16/;
+  add_proto qw/void vpx_highbd_h_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_h_predictor_16x16/;
 
-    add_proto qw/void vpx_highbd_d117_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d117_predictor_16x16/;
+  add_proto qw/void vpx_highbd_d117_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d117_predictor_16x16/;
 
-    add_proto qw/void vpx_highbd_d135_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d135_predictor_16x16/;
+  add_proto qw/void vpx_highbd_d135_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d135_predictor_16x16/;
 
-    add_proto qw/void vpx_highbd_d153_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d153_predictor_16x16/;
+  add_proto qw/void vpx_highbd_d153_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d153_predictor_16x16/;
 
-    add_proto qw/void vpx_highbd_v_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_v_predictor_16x16/, "$sse2_x86inc";
+  add_proto qw/void vpx_highbd_v_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_v_predictor_16x16/, "$sse2_x86inc";
 
-    add_proto qw/void vpx_highbd_tm_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_tm_predictor_16x16/, "$sse2_x86_64_x86inc";
+  add_proto qw/void vpx_highbd_tm_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_tm_predictor_16x16/, "$sse2_x86_64_x86inc";
 
-    add_proto qw/void vpx_highbd_dc_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_predictor_16x16/, "$sse2_x86inc";
+  add_proto qw/void vpx_highbd_dc_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_predictor_16x16/, "$sse2_x86inc";
 
-    add_proto qw/void vpx_highbd_dc_top_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_top_predictor_16x16/;
+  add_proto qw/void vpx_highbd_dc_top_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_top_predictor_16x16/;
 
-    add_proto qw/void vpx_highbd_dc_left_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_left_predictor_16x16/;
+  add_proto qw/void vpx_highbd_dc_left_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_left_predictor_16x16/;
 
-    add_proto qw/void vpx_highbd_dc_128_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_128_predictor_16x16/;
+  add_proto qw/void vpx_highbd_dc_128_predictor_16x16/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_128_predictor_16x16/;
 
-    add_proto qw/void vpx_highbd_d207_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d207_predictor_32x32/;
+  add_proto qw/void vpx_highbd_d207_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d207_predictor_32x32/;
 
-    add_proto qw/void vpx_highbd_d45_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d45_predictor_32x32/;
+  add_proto qw/void vpx_highbd_d45_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d45_predictor_32x32/;
 
-    add_proto qw/void vpx_highbd_d63_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d63_predictor_32x32/;
+  add_proto qw/void vpx_highbd_d63_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d63_predictor_32x32/;
 
-    add_proto qw/void vpx_highbd_h_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_h_predictor_32x32/;
+  add_proto qw/void vpx_highbd_h_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_h_predictor_32x32/;
 
-    add_proto qw/void vpx_highbd_d117_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d117_predictor_32x32/;
+  add_proto qw/void vpx_highbd_d117_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d117_predictor_32x32/;
 
-    add_proto qw/void vpx_highbd_d135_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d135_predictor_32x32/;
+  add_proto qw/void vpx_highbd_d135_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d135_predictor_32x32/;
 
-    add_proto qw/void vpx_highbd_d153_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_d153_predictor_32x32/;
+  add_proto qw/void vpx_highbd_d153_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_d153_predictor_32x32/;
 
-    add_proto qw/void vpx_highbd_v_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_v_predictor_32x32/, "$sse2_x86inc";
+  add_proto qw/void vpx_highbd_v_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_v_predictor_32x32/, "$sse2_x86inc";
 
-    add_proto qw/void vpx_highbd_tm_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_tm_predictor_32x32/, "$sse2_x86_64_x86inc";
+  add_proto qw/void vpx_highbd_tm_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_tm_predictor_32x32/, "$sse2_x86_64_x86inc";
 
-    add_proto qw/void vpx_highbd_dc_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_predictor_32x32/, "$sse2_x86_64_x86inc";
+  add_proto qw/void vpx_highbd_dc_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_predictor_32x32/, "$sse2_x86_64_x86inc";
 
-    add_proto qw/void vpx_highbd_dc_top_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_top_predictor_32x32/;
+  add_proto qw/void vpx_highbd_dc_top_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_top_predictor_32x32/;
 
-    add_proto qw/void vpx_highbd_dc_left_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_left_predictor_32x32/;
+  add_proto qw/void vpx_highbd_dc_left_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_left_predictor_32x32/;
 
-    add_proto qw/void vpx_highbd_dc_128_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
-    specialize qw/vpx_highbd_dc_128_predictor_32x32/;
-  }  # CONFIG_VP9_HIGHBITDEPTH
-}  # CONFIG_VP9 || CONFIG_VP10
+  add_proto qw/void vpx_highbd_dc_128_predictor_32x32/, "uint16_t *dst, ptrdiff_t y_stride, const uint16_t *above, const uint16_t *left, int bd";
+  specialize qw/vpx_highbd_dc_128_predictor_32x32/;
+}  # CONFIG_VP9_HIGHBITDEPTH
 
 #
 # Sub Pixel Filters