Merge Conflict Fix CL to lmp-mr1-release for ag/849478

DO NOT MERGE - libvpx: Pull from upstream

Current HEAD: 7105df53d7dc13d5e575bc8df714ec8d1da36b06

BUG=23452792

Change-Id: Ic78176fc369e0bacc71d423e0e2e6075d004aaec

1 files changed, 283 insertions, 320 deletions

diff --git a/libvpx/vp9/common/vp9_reconintra.c b/libvpx/vp9/common/vp9_reconintra.c
index 44951b5..e60eff8 100644
--- a/libvpx/vp9/common/vp9_reconintra.c
+++ b/libvpx/vp9/common/vp9_reconintra.c
@@ -9,12 +9,15 @@
  */
 
 #include "./vpx_config.h"
+#include "./vpx_dsp_rtcd.h"
 
+#if CONFIG_VP9_HIGHBITDEPTH
+#include "vpx_dsp/vpx_dsp_common.h"
+#endif  // CONFIG_VP9_HIGHBITDEPTH
 #include "vpx_mem/vpx_mem.h"
+#include "vpx_ports/mem.h"
 #include "vpx_ports/vpx_once.h"
 
-#include "./vp9_rtcd.h"
-
 #include "vp9/common/vp9_reconintra.h"
 #include "vp9/common/vp9_onyxc_int.h"
 
@@ -31,308 +34,112 @@ const TX_TYPE intra_mode_to_tx_type_lookup[INTRA_MODES] = {
   ADST_ADST,  // TM
 };
 
-#define intra_pred_sized(type, size) \
-  void vp9_##type##_predictor_##size##x##size##_c(uint8_t *dst, \
-                                                  ptrdiff_t stride, \
-                                                  const uint8_t *above, \
-                                                  const uint8_t *left) { \
-    type##_predictor(dst, stride, size, above, left); \
-  }
-
-#define intra_pred_allsizes(type) \
-  intra_pred_sized(type, 4) \
-  intra_pred_sized(type, 8) \
-  intra_pred_sized(type, 16) \
-  intra_pred_sized(type, 32)
-
-static INLINE void d207_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                  const uint8_t *above, const uint8_t *left) {
-  int r, c;
-
-  // first column
-  for (r = 0; r < bs - 1; ++r)
-    dst[r * stride] = ROUND_POWER_OF_TWO(left[r] + left[r + 1], 1);
-  dst[(bs - 1) * stride] = left[bs - 1];
-  dst++;
-
-  // second column
-  for (r = 0; r < bs - 2; ++r)
-    dst[r * stride] = ROUND_POWER_OF_TWO(left[r] + left[r + 1] * 2 +
-                                         left[r + 2], 2);
-  dst[(bs - 2) * stride] = ROUND_POWER_OF_TWO(left[bs - 2] +
-                                              left[bs - 1] * 3, 2);
-  dst[(bs - 1) * stride] = left[bs - 1];
-  dst++;
-
-  // rest of last row
-  for (c = 0; c < bs - 2; ++c)
-    dst[(bs - 1) * stride + c] = left[bs - 1];
-
-  for (r = bs - 2; r >= 0; --r)
-    for (c = 0; c < bs - 2; ++c)
-      dst[r * stride + c] = dst[(r + 1) * stride + c - 2];
-}
-intra_pred_allsizes(d207)
-
-static INLINE void d63_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                 const uint8_t *above, const uint8_t *left) {
-  int r, c;
-  for (r = 0; r < bs; ++r) {
-    for (c = 0; c < bs; ++c)
-      dst[c] = r & 1 ? ROUND_POWER_OF_TWO(above[r/2 + c] +
-                                          above[r/2 + c + 1] * 2 +
-                                          above[r/2 + c + 2], 2)
-                     : ROUND_POWER_OF_TWO(above[r/2 + c] +
-                                          above[r/2 + c + 1], 1);
-    dst += stride;
-  }
-}
-intra_pred_allsizes(d63)
-
-static INLINE void d45_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                 const uint8_t *above, const uint8_t *left) {
-  int r, c;
-  for (r = 0; r < bs; ++r) {
-    for (c = 0; c < bs; ++c)
-      dst[c] = r + c + 2 < bs * 2 ?  ROUND_POWER_OF_TWO(above[r + c] +
-                                                        above[r + c + 1] * 2 +
-                                                        above[r + c + 2], 2)
-                                  : above[bs * 2 - 1];
-    dst += stride;
-  }
-}
-intra_pred_allsizes(d45)
-
-static INLINE void d117_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                  const uint8_t *above, const uint8_t *left) {
-  int r, c;
-
-  // first row
-  for (c = 0; c < bs; c++)
-    dst[c] = ROUND_POWER_OF_TWO(above[c - 1] + above[c], 1);
-  dst += stride;
-
-  // second row
-  dst[0] = ROUND_POWER_OF_TWO(left[0] + above[-1] * 2 + above[0], 2);
-  for (c = 1; c < bs; c++)
-    dst[c] = ROUND_POWER_OF_TWO(above[c - 2] + above[c - 1] * 2 + above[c], 2);
-  dst += stride;
-
-  // the rest of first col
-  dst[0] = ROUND_POWER_OF_TWO(above[-1] + left[0] * 2 + left[1], 2);
-  for (r = 3; r < bs; ++r)
-    dst[(r - 2) * stride] = ROUND_POWER_OF_TWO(left[r - 3] + left[r - 2] * 2 +
-                                               left[r - 1], 2);
-
-  // the rest of the block
-  for (r = 2; r < bs; ++r) {
-    for (c = 1; c < bs; c++)
-      dst[c] = dst[-2 * stride + c - 1];
-    dst += stride;
-  }
-}
-intra_pred_allsizes(d117)
-
-static INLINE void d135_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                  const uint8_t *above, const uint8_t *left) {
-  int r, c;
-  dst[0] = ROUND_POWER_OF_TWO(left[0] + above[-1] * 2 + above[0], 2);
-  for (c = 1; c < bs; c++)
-    dst[c] = ROUND_POWER_OF_TWO(above[c - 2] + above[c - 1] * 2 + above[c], 2);
-
-  dst[stride] = ROUND_POWER_OF_TWO(above[-1] + left[0] * 2 + left[1], 2);
-  for (r = 2; r < bs; ++r)
-    dst[r * stride] = ROUND_POWER_OF_TWO(left[r - 2] + left[r - 1] * 2 +
-                                         left[r], 2);
-
-  dst += stride;
-  for (r = 1; r < bs; ++r) {
-    for (c = 1; c < bs; c++)
-      dst[c] = dst[-stride + c - 1];
-    dst += stride;
-  }
-}
-intra_pred_allsizes(d135)
-
-static INLINE void d153_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                  const uint8_t *above, const uint8_t *left) {
-  int r, c;
-  dst[0] = ROUND_POWER_OF_TWO(above[-1] + left[0], 1);
-  for (r = 1; r < bs; r++)
-    dst[r * stride] = ROUND_POWER_OF_TWO(left[r - 1] + left[r], 1);
-  dst++;
-
-  dst[0] = ROUND_POWER_OF_TWO(left[0] + above[-1] * 2 + above[0], 2);
-  dst[stride] = ROUND_POWER_OF_TWO(above[-1] + left[0] * 2 + left[1], 2);
-  for (r = 2; r < bs; r++)
-    dst[r * stride] = ROUND_POWER_OF_TWO(left[r - 2] + left[r - 1] * 2 +
-                                         left[r], 2);
-  dst++;
-
-  for (c = 0; c < bs - 2; c++)
-    dst[c] = ROUND_POWER_OF_TWO(above[c - 1] + above[c] * 2 + above[c + 1], 2);
-  dst += stride;
-
-  for (r = 1; r < bs; ++r) {
-    for (c = 0; c < bs - 2; c++)
-      dst[c] = dst[-stride + c - 2];
-    dst += stride;
-  }
-}
-intra_pred_allsizes(d153)
-
-static INLINE void v_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                               const uint8_t *above, const uint8_t *left) {
-  int r;
-
-  for (r = 0; r < bs; r++) {
-    vpx_memcpy(dst, above, bs);
-    dst += stride;
-  }
-}
-intra_pred_allsizes(v)
-
-static INLINE void h_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                               const uint8_t *above, const uint8_t *left) {
-  int r;
-
-  for (r = 0; r < bs; r++) {
-    vpx_memset(dst, left[r], bs);
-    dst += stride;
-  }
-}
-intra_pred_allsizes(h)
-
-static INLINE void tm_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                const uint8_t *above, const uint8_t *left) {
-  int r, c;
-  int ytop_left = above[-1];
-
-  for (r = 0; r < bs; r++) {
-    for (c = 0; c < bs; c++)
-      dst[c] = clip_pixel(left[r] + above[c] - ytop_left);
-    dst += stride;
-  }
-}
-intra_pred_allsizes(tm)
-
-static INLINE void dc_128_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                    const uint8_t *above, const uint8_t *left) {
-  int r;
-
-  for (r = 0; r < bs; r++) {
-    vpx_memset(dst, 128, bs);
-    dst += stride;
-  }
-}
-intra_pred_allsizes(dc_128)
-
-static INLINE void dc_left_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                     const uint8_t *above,
-                                     const uint8_t *left) {
-  int i, r, expected_dc, sum = 0;
-
-  for (i = 0; i < bs; i++)
-    sum += left[i];
-  expected_dc = (sum + (bs >> 1)) / bs;
-
-  for (r = 0; r < bs; r++) {
-    vpx_memset(dst, expected_dc, bs);
-    dst += stride;
-  }
-}
-intra_pred_allsizes(dc_left)
-
-static INLINE void dc_top_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                    const uint8_t *above, const uint8_t *left) {
-  int i, r, expected_dc, sum = 0;
-
-  for (i = 0; i < bs; i++)
-    sum += above[i];
-  expected_dc = (sum + (bs >> 1)) / bs;
-
-  for (r = 0; r < bs; r++) {
-    vpx_memset(dst, expected_dc, bs);
-    dst += stride;
-  }
-}
-intra_pred_allsizes(dc_top)
-
-static INLINE void dc_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
-                                const uint8_t *above, const uint8_t *left) {
-  int i, r, expected_dc, sum = 0;
-  const int count = 2 * bs;
-
-  for (i = 0; i < bs; i++) {
-    sum += above[i];
-    sum += left[i];
-  }
-
-  expected_dc = (sum + (count >> 1)) / count;
+enum {
+  NEED_LEFT = 1 << 1,
+  NEED_ABOVE = 1 << 2,
+  NEED_ABOVERIGHT = 1 << 3,
+};
 
-  for (r = 0; r < bs; r++) {
-    vpx_memset(dst, expected_dc, bs);
-    dst += stride;
-  }
-}
-intra_pred_allsizes(dc)
-#undef intra_pred_allsizes
+static const uint8_t extend_modes[INTRA_MODES] = {
+  NEED_ABOVE | NEED_LEFT,       // DC
+  NEED_ABOVE,                   // V
+  NEED_LEFT,                    // H
+  NEED_ABOVERIGHT,              // D45
+  NEED_LEFT | NEED_ABOVE,       // D135
+  NEED_LEFT | NEED_ABOVE,       // D117
+  NEED_LEFT | NEED_ABOVE,       // D153
+  NEED_LEFT,                    // D207
+  NEED_ABOVERIGHT,              // D63
+  NEED_LEFT | NEED_ABOVE,       // TM
+};
 
 typedef void (*intra_pred_fn)(uint8_t *dst, ptrdiff_t stride,
                               const uint8_t *above, const uint8_t *left);
 
-static intra_pred_fn pred[INTRA_MODES][4];
-static intra_pred_fn dc_pred[2][2][4];
-
-static void init_intra_pred_fn_ptrs(void) {
-#define intra_pred_allsizes(l, type) \
-  l[0] = vp9_##type##_predictor_4x4; \
-  l[1] = vp9_##type##_predictor_8x8; \
-  l[2] = vp9_##type##_predictor_16x16; \
-  l[3] = vp9_##type##_predictor_32x32
-
-  intra_pred_allsizes(pred[V_PRED], v);
-  intra_pred_allsizes(pred[H_PRED], h);
-  intra_pred_allsizes(pred[D207_PRED], d207);
-  intra_pred_allsizes(pred[D45_PRED], d45);
-  intra_pred_allsizes(pred[D63_PRED], d63);
-  intra_pred_allsizes(pred[D117_PRED], d117);
-  intra_pred_allsizes(pred[D135_PRED], d135);
-  intra_pred_allsizes(pred[D153_PRED], d153);
-  intra_pred_allsizes(pred[TM_PRED], tm);
-
-  intra_pred_allsizes(dc_pred[0][0], dc_128);
-  intra_pred_allsizes(dc_pred[0][1], dc_top);
-  intra_pred_allsizes(dc_pred[1][0], dc_left);
-  intra_pred_allsizes(dc_pred[1][1], dc);
+static intra_pred_fn pred[INTRA_MODES][TX_SIZES];
+static intra_pred_fn dc_pred[2][2][TX_SIZES];
+
+#if CONFIG_VP9_HIGHBITDEPTH
+typedef void (*intra_high_pred_fn)(uint16_t *dst, ptrdiff_t stride,
+                                   const uint16_t *above, const uint16_t *left,
+                                   int bd);
+static intra_high_pred_fn pred_high[INTRA_MODES][4];
+static intra_high_pred_fn dc_pred_high[2][2][4];
+#endif  // CONFIG_VP9_HIGHBITDEPTH
+
+static void vp9_init_intra_predictors_internal(void) {
+#define INIT_ALL_SIZES(p, type) \
+  p[TX_4X4] = vpx_##type##_predictor_4x4; \
+  p[TX_8X8] = vpx_##type##_predictor_8x8; \
+  p[TX_16X16] = vpx_##type##_predictor_16x16; \
+  p[TX_32X32] = vpx_##type##_predictor_32x32
+
+  INIT_ALL_SIZES(pred[V_PRED], v);
+  INIT_ALL_SIZES(pred[H_PRED], h);
+  INIT_ALL_SIZES(pred[D207_PRED], d207);
+  INIT_ALL_SIZES(pred[D45_PRED], d45);
+  INIT_ALL_SIZES(pred[D63_PRED], d63);
+  INIT_ALL_SIZES(pred[D117_PRED], d117);
+  INIT_ALL_SIZES(pred[D135_PRED], d135);
+  INIT_ALL_SIZES(pred[D153_PRED], d153);
+  INIT_ALL_SIZES(pred[TM_PRED], tm);
+
+  INIT_ALL_SIZES(dc_pred[0][0], dc_128);
+  INIT_ALL_SIZES(dc_pred[0][1], dc_top);
+  INIT_ALL_SIZES(dc_pred[1][0], dc_left);
+  INIT_ALL_SIZES(dc_pred[1][1], dc);
+
+#if CONFIG_VP9_HIGHBITDEPTH
+  INIT_ALL_SIZES(pred_high[V_PRED], highbd_v);
+  INIT_ALL_SIZES(pred_high[H_PRED], highbd_h);
+  INIT_ALL_SIZES(pred_high[D207_PRED], highbd_d207);
+  INIT_ALL_SIZES(pred_high[D45_PRED], highbd_d45);
+  INIT_ALL_SIZES(pred_high[D63_PRED], highbd_d63);
+  INIT_ALL_SIZES(pred_high[D117_PRED], highbd_d117);
+  INIT_ALL_SIZES(pred_high[D135_PRED], highbd_d135);
+  INIT_ALL_SIZES(pred_high[D153_PRED], highbd_d153);
+  INIT_ALL_SIZES(pred_high[TM_PRED], highbd_tm);
+
+  INIT_ALL_SIZES(dc_pred_high[0][0], highbd_dc_128);
+  INIT_ALL_SIZES(dc_pred_high[0][1], highbd_dc_top);
+  INIT_ALL_SIZES(dc_pred_high[1][0], highbd_dc_left);
+  INIT_ALL_SIZES(dc_pred_high[1][1], highbd_dc);
+#endif  // CONFIG_VP9_HIGHBITDEPTH
 
 #undef intra_pred_allsizes
 }
 
-static void build_intra_predictors(const MACROBLOCKD *xd, const uint8_t *ref,
-                                   int ref_stride, uint8_t *dst, int dst_stride,
-                                   MB_PREDICTION_MODE mode, TX_SIZE tx_size,
-                                   int up_available, int left_available,
-                                   int right_available, int x, int y,
-                                   int plane) {
+#if CONFIG_VP9_HIGHBITDEPTH
+static void build_intra_predictors_high(const MACROBLOCKD *xd,
+                                        const uint8_t *ref8,
+                                        int ref_stride,
+                                        uint8_t *dst8,
+                                        int dst_stride,
+                                        PREDICTION_MODE mode,
+                                        TX_SIZE tx_size,
+                                        int up_available,
+                                        int left_available,
+                                        int right_available,
+                                        int x, int y,
+                                        int plane, int bd) {
   int i;
-  DECLARE_ALIGNED_ARRAY(16, uint8_t, left_col, 64);
-  DECLARE_ALIGNED_ARRAY(16, uint8_t, above_data, 128 + 16);
-  uint8_t *above_row = above_data + 16;
-  const uint8_t *const_above_row = above_row;
+  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
+  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
+  DECLARE_ALIGNED(16, uint16_t, left_col[32]);
+  DECLARE_ALIGNED(16, uint16_t, above_data[64 + 16]);
+  uint16_t *above_row = above_data + 16;
+  const uint16_t *const_above_row = above_row;
   const int bs = 4 << tx_size;
   int frame_width, frame_height;
   int x0, y0;
   const struct macroblockd_plane *const pd = &xd->plane[plane];
-
+  //  int base=128;
+  int base = 128 << (bd - 8);
   // 127 127 127 .. 127 127 127 127 127 127
   // 129  A   B  ..  Y   Z
   // 129  C   D  ..  W   X
   // 129  E   F  ..  U   V
   // 129  G   H  ..  S   T   T   T   T   T
-  // ..
-
-  once(init_intra_pred_fn_ptrs);
 
   // Get current frame pointer, width and height.
   if (plane == 0) {
@@ -347,8 +154,6 @@ static void build_intra_predictors(const MACROBLOCKD *xd, const uint8_t *ref,
   x0 = (-xd->mb_to_left_edge >> (3 + pd->subsampling_x)) + x;
   y0 = (-xd->mb_to_top_edge >> (3 + pd->subsampling_y)) + y;
 
-  vpx_memset(left_col, 129, 64);
-
   // left
   if (left_available) {
     if (xd->mb_to_bottom_edge < 0) {
@@ -368,60 +173,207 @@ static void build_intra_predictors(const MACROBLOCKD *xd, const uint8_t *ref,
       for (i = 0; i < bs; ++i)
         left_col[i] = ref[i * ref_stride - 1];
     }
+  } else {
+    // TODO(Peter): this value should probably change for high bitdepth
+    vpx_memset16(left_col, base + 1, bs);
   }
 
   // TODO(hkuang) do not extend 2*bs pixels for all modes.
   // above
   if (up_available) {
-    const uint8_t *above_ref = ref - ref_stride;
+    const uint16_t *above_ref = ref - ref_stride;
     if (xd->mb_to_right_edge < 0) {
       /* slower path if the block needs border extension */
       if (x0 + 2 * bs <= frame_width) {
         if (right_available && bs == 4) {
-          vpx_memcpy(above_row, above_ref, 2 * bs);
+          memcpy(above_row, above_ref, 2 * bs * sizeof(above_row[0]));
         } else {
-          vpx_memcpy(above_row, above_ref, bs);
-          vpx_memset(above_row + bs, above_row[bs - 1], bs);
+          memcpy(above_row, above_ref, bs * sizeof(above_row[0]));
+          vpx_memset16(above_row + bs, above_row[bs - 1], bs);
         }
       } else if (x0 + bs <= frame_width) {
         const int r = frame_width - x0;
         if (right_available && bs == 4) {
-          vpx_memcpy(above_row, above_ref, r);
-          vpx_memset(above_row + r, above_row[r - 1],
-                     x0 + 2 * bs - frame_width);
+          memcpy(above_row, above_ref, r * sizeof(above_row[0]));
+          vpx_memset16(above_row + r, above_row[r - 1],
+                       x0 + 2 * bs - frame_width);
         } else {
-          vpx_memcpy(above_row, above_ref, bs);
-          vpx_memset(above_row + bs, above_row[bs - 1], bs);
+          memcpy(above_row, above_ref, bs * sizeof(above_row[0]));
+          vpx_memset16(above_row + bs, above_row[bs - 1], bs);
         }
       } else if (x0 <= frame_width) {
         const int r = frame_width - x0;
-        if (right_available && bs == 4) {
-          vpx_memcpy(above_row, above_ref, r);
-          vpx_memset(above_row + r, above_row[r - 1],
-                     x0 + 2 * bs - frame_width);
-        } else {
-          vpx_memcpy(above_row, above_ref, r);
-          vpx_memset(above_row + r, above_row[r - 1],
-                     x0 + 2 * bs - frame_width);
-        }
+        memcpy(above_row, above_ref, r * sizeof(above_row[0]));
+        vpx_memset16(above_row + r, above_row[r - 1],
+                       x0 + 2 * bs - frame_width);
       }
-      above_row[-1] = left_available ? above_ref[-1] : 129;
+      // TODO(Peter) this value should probably change for high bitdepth
+      above_row[-1] = left_available ? above_ref[-1] : (base+1);
     } else {
       /* faster path if the block does not need extension */
       if (bs == 4 && right_available && left_available) {
         const_above_row = above_ref;
       } else {
-        vpx_memcpy(above_row, above_ref, bs);
+        memcpy(above_row, above_ref, bs * sizeof(above_row[0]));
         if (bs == 4 && right_available)
-          vpx_memcpy(above_row + bs, above_ref + bs, bs);
+          memcpy(above_row + bs, above_ref + bs, bs * sizeof(above_row[0]));
         else
-          vpx_memset(above_row + bs, above_row[bs - 1], bs);
-        above_row[-1] = left_available ? above_ref[-1] : 129;
+          vpx_memset16(above_row + bs, above_row[bs - 1], bs);
+        // TODO(Peter): this value should probably change for high bitdepth
+        above_row[-1] = left_available ? above_ref[-1] : (base+1);
       }
     }
   } else {
-    vpx_memset(above_row, 127, bs * 2);
-    above_row[-1] = 127;
+    vpx_memset16(above_row, base - 1, bs * 2);
+    // TODO(Peter): this value should probably change for high bitdepth
+    above_row[-1] = base - 1;
+  }
+
+  // predict
+  if (mode == DC_PRED) {
+    dc_pred_high[left_available][up_available][tx_size](dst, dst_stride,
+                                                        const_above_row,
+                                                        left_col, xd->bd);
+  } else {
+    pred_high[mode][tx_size](dst, dst_stride, const_above_row, left_col,
+                             xd->bd);
+  }
+}
+#endif  // CONFIG_VP9_HIGHBITDEPTH
+
+static void build_intra_predictors(const MACROBLOCKD *xd, const uint8_t *ref,
+                                   int ref_stride, uint8_t *dst, int dst_stride,
+                                   PREDICTION_MODE mode, TX_SIZE tx_size,
+                                   int up_available, int left_available,
+                                   int right_available, int x, int y,
+                                   int plane) {
+  int i;
+  DECLARE_ALIGNED(16, uint8_t, left_col[32]);
+  DECLARE_ALIGNED(16, uint8_t, above_data[64 + 16]);
+  uint8_t *above_row = above_data + 16;
+  const uint8_t *const_above_row = above_row;
+  const int bs = 4 << tx_size;
+  int frame_width, frame_height;
+  int x0, y0;
+  const struct macroblockd_plane *const pd = &xd->plane[plane];
+
+  // 127 127 127 .. 127 127 127 127 127 127
+  // 129  A   B  ..  Y   Z
+  // 129  C   D  ..  W   X
+  // 129  E   F  ..  U   V
+  // 129  G   H  ..  S   T   T   T   T   T
+  // ..
+
+  // Get current frame pointer, width and height.
+  if (plane == 0) {
+    frame_width = xd->cur_buf->y_width;
+    frame_height = xd->cur_buf->y_height;
+  } else {
+    frame_width = xd->cur_buf->uv_width;
+    frame_height = xd->cur_buf->uv_height;
+  }
+
+  // Get block position in current frame.
+  x0 = (-xd->mb_to_left_edge >> (3 + pd->subsampling_x)) + x;
+  y0 = (-xd->mb_to_top_edge >> (3 + pd->subsampling_y)) + y;
+
+  // NEED_LEFT
+  if (extend_modes[mode] & NEED_LEFT) {
+    if (left_available) {
+      if (xd->mb_to_bottom_edge < 0) {
+        /* slower path if the block needs border extension */
+        if (y0 + bs <= frame_height) {
+          for (i = 0; i < bs; ++i)
+            left_col[i] = ref[i * ref_stride - 1];
+        } else {
+          const int extend_bottom = frame_height - y0;
+          for (i = 0; i < extend_bottom; ++i)
+            left_col[i] = ref[i * ref_stride - 1];
+          for (; i < bs; ++i)
+            left_col[i] = ref[(extend_bottom - 1) * ref_stride - 1];
+        }
+      } else {
+        /* faster path if the block does not need extension */
+        for (i = 0; i < bs; ++i)
+          left_col[i] = ref[i * ref_stride - 1];
+      }
+    } else {
+      memset(left_col, 129, bs);
+    }
+  }
+
+  // NEED_ABOVE
+  if (extend_modes[mode] & NEED_ABOVE) {
+    if (up_available) {
+      const uint8_t *above_ref = ref - ref_stride;
+      if (xd->mb_to_right_edge < 0) {
+        /* slower path if the block needs border extension */
+        if (x0 + bs <= frame_width) {
+          memcpy(above_row, above_ref, bs);
+        } else if (x0 <= frame_width) {
+          const int r = frame_width - x0;
+          memcpy(above_row, above_ref, r);
+          memset(above_row + r, above_row[r - 1], x0 + bs - frame_width);
+        }
+      } else {
+        /* faster path if the block does not need extension */
+        if (bs == 4 && right_available && left_available) {
+          const_above_row = above_ref;
+        } else {
+          memcpy(above_row, above_ref, bs);
+        }
+      }
+      above_row[-1] = left_available ? above_ref[-1] : 129;
+    } else {
+      memset(above_row, 127, bs);
+      above_row[-1] = 127;
+    }
+  }
+
+  // NEED_ABOVERIGHT
+  if (extend_modes[mode] & NEED_ABOVERIGHT) {
+    if (up_available) {
+      const uint8_t *above_ref = ref - ref_stride;
+      if (xd->mb_to_right_edge < 0) {
+        /* slower path if the block needs border extension */
+        if (x0 + 2 * bs <= frame_width) {
+          if (right_available && bs == 4) {
+            memcpy(above_row, above_ref, 2 * bs);
+          } else {
+            memcpy(above_row, above_ref, bs);
+            memset(above_row + bs, above_row[bs - 1], bs);
+          }
+        } else if (x0 + bs <= frame_width) {
+          const int r = frame_width - x0;
+          if (right_available && bs == 4) {
+            memcpy(above_row, above_ref, r);
+            memset(above_row + r, above_row[r - 1], x0 + 2 * bs - frame_width);
+          } else {
+            memcpy(above_row, above_ref, bs);
+            memset(above_row + bs, above_row[bs - 1], bs);
+          }
+        } else if (x0 <= frame_width) {
+          const int r = frame_width - x0;
+          memcpy(above_row, above_ref, r);
+          memset(above_row + r, above_row[r - 1], x0 + 2 * bs - frame_width);
+        }
+      } else {
+        /* faster path if the block does not need extension */
+        if (bs == 4 && right_available && left_available) {
+          const_above_row = above_ref;
+        } else {
+          memcpy(above_row, above_ref, bs);
+          if (bs == 4 && right_available)
+            memcpy(above_row + bs, above_ref + bs, bs);
+          else
+            memset(above_row + bs, above_row[bs - 1], bs);
+        }
+      }
+      above_row[-1] = left_available ? above_ref[-1] : 129;
+    } else {
+      memset(above_row, 127, bs * 2);
+      above_row[-1] = 127;
+    }
   }
 
   // predict
@@ -433,20 +385,31 @@ static void build_intra_predictors(const MACROBLOCKD *xd, const uint8_t *ref,
   }
 }
 
-void vp9_predict_intra_block(const MACROBLOCKD *xd, int block_idx, int bwl_in,
-                             TX_SIZE tx_size, MB_PREDICTION_MODE mode,
+void vp9_predict_intra_block(const MACROBLOCKD *xd, int bwl_in,
+                             TX_SIZE tx_size, PREDICTION_MODE mode,
                              const uint8_t *ref, int ref_stride,
                              uint8_t *dst, int dst_stride,
                              int aoff, int loff, int plane) {
-  const int bwl = bwl_in - tx_size;
-  const int wmask = (1 << bwl) - 1;
-  const int have_top = (block_idx >> bwl) || xd->up_available;
-  const int have_left = (block_idx & wmask) || xd->left_available;
-  const int have_right = ((block_idx & wmask) != wmask);
+  const int bw = (1 << bwl_in);
+  const int txw = (1 << tx_size);
+  const int have_top = loff || xd->up_available;
+  const int have_left = aoff || xd->left_available;
+  const int have_right = (aoff + txw) < bw;
   const int x = aoff * 4;
   const int y = loff * 4;
 
-  assert(bwl >= 0);
+#if CONFIG_VP9_HIGHBITDEPTH
+  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+    build_intra_predictors_high(xd, ref, ref_stride, dst, dst_stride, mode,
+                                tx_size, have_top, have_left, have_right,
+                                x, y, plane, xd->bd);
+    return;
+  }
+#endif
   build_intra_predictors(xd, ref, ref_stride, dst, dst_stride, mode, tx_size,
                          have_top, have_left, have_right, x, y, plane);
 }
+
+void vp9_init_intra_predictors(void) {
+  once(vp9_init_intra_predictors_internal);
+}