libvpx: Pull from upstream

source/libvpx/vp9/common/vp9_reconintra.c

Index: source/libvpx/vp9/common/vp9_reconintra.c
===================================================================
--- source/libvpx/vp9/common/vp9_reconintra.c	(revision 292072)
+++ source/libvpx/vp9/common/vp9_reconintra.c	(working copy)
@@ -40,12 +40,290 @@
     type##_predictor(dst, stride, size, above, left); \
   }
 
+#if CONFIG_VP9_HIGHBITDEPTH
+#define intra_pred_high_sized(type, size) \
+  void vp9_high_##type##_predictor_##size##x##size##_c( \
+      uint16_t *dst, ptrdiff_t stride, const uint16_t *above, \
+      const uint16_t *left, int bd) { \
+    high_##type##_predictor(dst, stride, size, above, left, bd); \
+  }
+
 #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) \
+  intra_pred_high_sized(type, 4) \
+  intra_pred_high_sized(type, 8) \
+  intra_pred_high_sized(type, 16) \
+  intra_pred_high_sized(type, 32)
+
+#else
+
+#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)
+#endif  // CONFIG_VP9_HIGHBITDEPTH
 
+#if CONFIG_VP9_HIGHBITDEPTH
+static INLINE void high_d207_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
+                                       const uint16_t *above,
+                                       const uint16_t *left, int bd) {
+  int r, c;
+  (void) above;
+  (void) bd;
+
+  // 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];
+  }
+}
+
+static INLINE void high_d63_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
+                                      const uint16_t *above,
+                                      const uint16_t *left, int bd) {
+  int r, c;
+  (void) left;
+  (void) bd;
+  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;
+  }
+}
+
+static INLINE void high_d45_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
+                                      const uint16_t *above,
+                                      const uint16_t *left, int bd) {
+  int r, c;
+  (void) left;
+  (void) bd;
+  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;
+  }
+}
+
+static INLINE void high_d117_predictor(uint16_t *dst, ptrdiff_t stride,
+                                       int bs, const uint16_t *above,
+                                       const uint16_t *left, int bd) {
+  int r, c;
+  (void) bd;
+
+  // 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;
+  }
+}
+
+static INLINE void high_d135_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
+                                       const uint16_t *above,
+                                       const uint16_t *left, int bd) {
+  int r, c;
+  (void) bd;
+  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;
+  }
+}
+
+static INLINE void high_d153_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
+                                       const uint16_t *above,
+                                       const uint16_t *left, int bd) {
+  int r, c;
+  (void) bd;
+  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;
+  }
+}
+
+static INLINE void high_v_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
+                                    const uint16_t *above,
+                                    const uint16_t *left, int bd) {
+  int r;
+  (void) left;
+  (void) bd;
+  for (r = 0; r < bs; r++) {
+    vpx_memcpy(dst, above, bs * sizeof(uint16_t));
+    dst += stride;
+  }
+}
+
+static INLINE void high_h_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
+                                    const uint16_t *above, const uint16_t *left,
+                                    int bd) {
+  int r;
+  (void) above;
+  (void) bd;
+  for (r = 0; r < bs; r++) {
+    vpx_memset16(dst, left[r], bs);
+    dst += stride;
+  }
+}
+
+static INLINE void high_tm_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
+                                     const uint16_t *above,
+                                     const uint16_t *left, int bd) {
+  int r, c;
+  int ytop_left = above[-1];
+  (void) bd;
+
+  for (r = 0; r < bs; r++) {
+    for (c = 0; c < bs; c++)
+      dst[c] = clip_pixel_high(left[r] + above[c] - ytop_left, bd);
+    dst += stride;
+  }
+}
+
+static INLINE void high_dc_128_predictor(uint16_t *dst, ptrdiff_t stride,
+                                         int bs, const uint16_t *above,
+                                         const uint16_t *left, int bd) {
+  int r;
+  (void) above;
+  (void) left;
+
+  for (r = 0; r < bs; r++) {
+    vpx_memset16(dst, 128 << (bd - 8), bs);
+    dst += stride;
+  }
+}
+
+static INLINE void high_dc_left_predictor(uint16_t *dst, ptrdiff_t stride,
+                                          int bs, const uint16_t *above,
+                                          const uint16_t *left, int bd) {
+  int i, r, expected_dc, sum = 0;
+  (void) above;
+  (void) bd;
+
+  for (i = 0; i < bs; i++)
+    sum += left[i];
+  expected_dc = (sum + (bs >> 1)) / bs;
+
+  for (r = 0; r < bs; r++) {
+    vpx_memset16(dst, expected_dc, bs);
+    dst += stride;
+  }
+}
+
+static INLINE void high_dc_top_predictor(uint16_t *dst, ptrdiff_t stride,
+                                         int bs, const uint16_t *above,
+                                         const uint16_t *left, int bd) {
+  int i, r, expected_dc, sum = 0;
+  (void) left;
+  (void) bd;
+
+  for (i = 0; i < bs; i++)
+    sum += above[i];
+  expected_dc = (sum + (bs >> 1)) / bs;
+
+  for (r = 0; r < bs; r++) {
+    vpx_memset16(dst, expected_dc, bs);
+    dst += stride;
+  }
+}
+
+static INLINE void high_dc_predictor(uint16_t *dst, ptrdiff_t stride,
+                                     int bs, const uint16_t *above,
+                                     const uint16_t *left, int bd) {
+  int i, r, expected_dc, sum = 0;
+  const int count = 2 * bs;
+  (void) bd;
+
+  for (i = 0; i < bs; i++) {
+    sum += above[i];
+    sum += left[i];
+  }
+
+  expected_dc = (sum + (count >> 1)) / count;
+
+  for (r = 0; r < bs; r++) {
+    vpx_memset16(dst, expected_dc, bs);
+    dst += stride;
+  }
+}
+#endif  // CONFIG_VP9_HIGHBITDEPTH
+
 static INLINE void d207_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                                   const uint8_t *above, const uint8_t *left) {
   int r, c;
@@ -293,6 +571,14 @@
 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
+
 void vp9_init_intra_predictors() {
 #define INIT_ALL_SIZES(p, type) \
   p[TX_4X4] = vp9_##type##_predictor_4x4; \
@@ -315,9 +601,164 @@
   INIT_ALL_SIZES(dc_pred[1][0], dc_left);
   INIT_ALL_SIZES(dc_pred[1][1], dc);
 
-#undef INIT_ALL_SIZES
+#if CONFIG_VP9_HIGHBITDEPTH
+  INIT_ALL_SIZES(pred_high[V_PRED], high_v);
+  INIT_ALL_SIZES(pred_high[H_PRED], high_h);
+  INIT_ALL_SIZES(pred_high[D207_PRED], high_d207);
+  INIT_ALL_SIZES(pred_high[D45_PRED], high_d45);
+  INIT_ALL_SIZES(pred_high[D63_PRED], high_d63);
+  INIT_ALL_SIZES(pred_high[D117_PRED], high_d117);
+  INIT_ALL_SIZES(pred_high[D135_PRED], high_d135);
+  INIT_ALL_SIZES(pred_high[D153_PRED], high_d153);
+  INIT_ALL_SIZES(pred_high[TM_PRED], high_tm);
+
+  INIT_ALL_SIZES(dc_pred_high[0][0], high_dc_128);
+  INIT_ALL_SIZES(dc_pred_high[0][1], high_dc_top);
+  INIT_ALL_SIZES(dc_pred_high[1][0], high_dc_left);
+  INIT_ALL_SIZES(dc_pred_high[1][1], high_dc);
+#endif  // CONFIG_VP9_HIGHBITDEPTH
+
+#undef intra_pred_allsizes
 }
 
+#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;
+  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
+  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
+  DECLARE_ALIGNED_ARRAY(16, uint16_t, left_col, 64);
+  DECLARE_ALIGNED_ARRAY(16, uint16_t, above_data, 128 + 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
+
+  // 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;
+
+  // 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 {
+    // 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 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 * sizeof(uint16_t));
+        } else {
+          vpx_memcpy(above_row, above_ref, bs * sizeof(uint16_t));
+          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 * sizeof(uint16_t));
+          vpx_memset16(above_row + r, above_row[r - 1],
+                       x0 + 2 * bs - frame_width);
+        } else {
+          vpx_memcpy(above_row, above_ref, bs * sizeof(uint16_t));
+          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 * sizeof(uint16_t));
+          vpx_memset16(above_row + r, above_row[r - 1],
+                       x0 + 2 * bs - frame_width);
+        } else {
+          vpx_memcpy(above_row, above_ref, r * sizeof(uint16_t));
+          vpx_memset16(above_row + r, above_row[r - 1],
+                       x0 + 2 * bs - frame_width);
+        }
+      }
+      // 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 * sizeof(uint16_t));
+        if (bs == 4 && right_available)
+          vpx_memcpy(above_row + bs, above_ref + bs, bs * sizeof(uint16_t));
+        else
+          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_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,
@@ -454,6 +895,14 @@
   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);
 }