libvpx: Pull from upstream

source/libvpx/vp9/common/vp9_reconinter.c

 /*
  *  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.
  */
 
@@ skipping 45 matching lines @@
       memset(dst + left + copy, ref_row[w - 1], right);
 
     dst += dst_stride;
     ++y;
 
     if (y > 0 && y < h)
       ref_row += src_stride;
   } while (--b_h);
 }
 
+#if CONFIG_VP9_HIGHBITDEPTH
+static void high_build_mc_border(const uint8_t *src8, int src_stride,
+                                 uint16_t *dst, int dst_stride,
+                                 int x, int y, int b_w, int b_h,
+                                 int w, int h) {
+  // Get a pointer to the start of the real data for this row.
+  const uint16_t *src = CONVERT_TO_SHORTPTR(src8);
+  const uint16_t *ref_row = src - x - y * src_stride;
+
+  if (y >= h)
+    ref_row += (h - 1) * src_stride;
+  else if (y > 0)
+    ref_row += y * src_stride;
+
+  do {
+    int right = 0, copy;
+    int left = x < 0 ? -x : 0;
+
+    if (left > b_w)
+      left = b_w;
+
+    if (x + b_w > w)
+      right = x + b_w - w;
+
+    if (right > b_w)
+      right = b_w;
+
+    copy = b_w - left - right;
+
+    if (left)
+      vpx_memset16(dst, ref_row[0], left);
+
+    if (copy)
+      memcpy(dst + left, ref_row + x + left, copy * sizeof(uint16_t));
+
+    if (right)
+      vpx_memset16(dst + left + copy, ref_row[w - 1], right);
+
+    dst += dst_stride;
+    ++y;
+
+    if (y > 0 && y < h)
+      ref_row += src_stride;
+  } while (--b_h);
+}
+#endif  // CONFIG_VP9_HIGHBITDEPTH
+
 static void inter_predictor(const uint8_t *src, int src_stride,
                             uint8_t *dst, int dst_stride,
                             const int subpel_x,
                             const int subpel_y,
                             const struct scale_factors *sf,
                             int w, int h, int ref,
                             const InterpKernel *kernel,
                             int xs, int ys) {
   sf->predict[subpel_x != 0][subpel_y != 0][ref](
       src, src_stride, dst, dst_stride,
@@ skipping 14 matching lines @@
   MV32 mv = vp9_scale_mv(&mv_q4, x, y, sf);
   const int subpel_x = mv.col & SUBPEL_MASK;
   const int subpel_y = mv.row & SUBPEL_MASK;
 
   src += (mv.row >> SUBPEL_BITS) * src_stride + (mv.col >> SUBPEL_BITS);
 
   inter_predictor(src, src_stride, dst, dst_stride, subpel_x, subpel_y,
                   sf, w, h, ref, kernel, sf->x_step_q4, sf->y_step_q4);
 }
 
+#if CONFIG_VP9_HIGHBITDEPTH
+static void high_inter_predictor(const uint8_t *src, int src_stride,
+                                 uint8_t *dst, int dst_stride,
+                                 const int subpel_x,
+                                 const int subpel_y,
+                                 const struct scale_factors *sf,
+                                 int w, int h, int ref,
+                                 const InterpKernel *kernel,
+                                 int xs, int ys, int bd) {
+  sf->high_predict[subpel_x != 0][subpel_y != 0][ref](
+      src, src_stride, dst, dst_stride,
+      kernel[subpel_x], xs, kernel[subpel_y], ys, w, h, bd);
+}
+
+void vp9_high_build_inter_predictor(const uint8_t *src, int src_stride,
+                                    uint8_t *dst, int dst_stride,
+                                    const MV *src_mv,
+                                    const struct scale_factors *sf,
+                                    int w, int h, int ref,
+                                    const InterpKernel *kernel,
+                                    enum mv_precision precision,
+                                    int x, int y, int bd) {
+  const int is_q4 = precision == MV_PRECISION_Q4;
+  const MV mv_q4 = { is_q4 ? src_mv->row : src_mv->row * 2,
+                     is_q4 ? src_mv->col : src_mv->col * 2 };
+  MV32 mv = vp9_scale_mv(&mv_q4, x, y, sf);
+  const int subpel_x = mv.col & SUBPEL_MASK;
+  const int subpel_y = mv.row & SUBPEL_MASK;
+
+  src += (mv.row >> SUBPEL_BITS) * src_stride + (mv.col >> SUBPEL_BITS);
+
+  high_inter_predictor(src, src_stride, dst, dst_stride, subpel_x, subpel_y,
+                       sf, w, h, ref, kernel, sf->x_step_q4, sf->y_step_q4, bd);
+}
+#endif  // CONFIG_VP9_HIGHBITDEPTH
+
 static INLINE int round_mv_comp_q4(int value) {
   return (value < 0 ? value - 2 : value + 2) / 4;
 }
 
 static MV mi_mv_pred_q4(const MODE_INFO *mi, int idx) {
   MV res = { round_mv_comp_q4(mi->bmi[0].as_mv[idx].as_mv.row +
                               mi->bmi[1].as_mv[idx].as_mv.row +
                               mi->bmi[2].as_mv[idx].as_mv.row +
                               mi->bmi[3].as_mv[idx].as_mv.row),
              round_mv_comp_q4(mi->bmi[0].as_mv[idx].as_mv.col +
@@ skipping 62 matching lines @@
       assert(ss_idx <= 3 || ss_idx >= 0);
   }
   return res;
 }
 
 static void build_inter_predictors(MACROBLOCKD *xd, int plane, int block,
                                    int bw, int bh,
                                    int x, int y, int w, int h,
                                    int mi_x, int mi_y) {
   struct macroblockd_plane *const pd = &xd->plane[plane];
-  const MODE_INFO *mi = xd->mi[0];
+  const MODE_INFO *mi = xd->mi[0].src_mi;
   const int is_compound = has_second_ref(&mi->mbmi);
   const InterpKernel *kernel = vp9_get_interp_kernel(mi->mbmi.interp_filter);
   int ref;
 
   for (ref = 0; ref < 1 + is_compound; ++ref) {
     const struct scale_factors *const sf = &xd->block_refs[ref]->sf;
     struct buf_2d *const pre_buf = &pd->pre[ref];
     struct buf_2d *const dst_buf = &pd->dst;
     uint8_t *const dst = dst_buf->buf + dst_buf->stride * y + x;
     const MV mv = mi->mbmi.sb_type < BLOCK_8X8
@@ skipping 22 matching lines @@
       pre = pre_buf->buf + (y * pre_buf->stride + x);
       scaled_mv.row = mv_q4.row;
       scaled_mv.col = mv_q4.col;
       xs = ys = 16;
     }
     subpel_x = scaled_mv.col & SUBPEL_MASK;
     subpel_y = scaled_mv.row & SUBPEL_MASK;
     pre += (scaled_mv.row >> SUBPEL_BITS) * pre_buf->stride
            + (scaled_mv.col >> SUBPEL_BITS);
 
+#if CONFIG_VP9_HIGHBITDEPTH
+    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+      high_inter_predictor(pre, pre_buf->stride, dst, dst_buf->stride,
+                           subpel_x, subpel_y, sf, w, h, ref, kernel, xs, ys,
+                           xd->bd);
+    } else {
+      inter_predictor(pre, pre_buf->stride, dst, dst_buf->stride,
+                      subpel_x, subpel_y, sf, w, h, ref, kernel, xs, ys);
+    }
+#else
     inter_predictor(pre, pre_buf->stride, dst, dst_buf->stride,
                     subpel_x, subpel_y, sf, w, h, ref, kernel, xs, ys);
+#endif  // CONFIG_VP9_HIGHBITDEPTH
   }
 }
 
 static void build_inter_predictors_for_planes(MACROBLOCKD *xd, BLOCK_SIZE bsize,
                                               int mi_row, int mi_col,
                                               int plane_from, int plane_to) {
   int plane;
   const int mi_x = mi_col * MI_SIZE;
   const int mi_y = mi_row * MI_SIZE;
   for (plane = plane_from; plane <= plane_to; ++plane) {
     const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize,
                                                         &xd->plane[plane]);
     const int num_4x4_w = num_4x4_blocks_wide_lookup[plane_bsize];
     const int num_4x4_h = num_4x4_blocks_high_lookup[plane_bsize];
     const int bw = 4 * num_4x4_w;
     const int bh = 4 * num_4x4_h;
 
-    if (xd->mi[0]->mbmi.sb_type < BLOCK_8X8) {
+    if (xd->mi[0].src_mi->mbmi.sb_type < BLOCK_8X8) {
       int i = 0, x, y;
       assert(bsize == BLOCK_8X8);
       for (y = 0; y < num_4x4_h; ++y)
         for (x = 0; x < num_4x4_w; ++x)
            build_inter_predictors(xd, plane, i++, bw, bh,
                                   4 * x, 4 * y, 4, 4, mi_x, mi_y);
     } else {
       build_inter_predictors(xd, plane, 0, bw, bh,
                              0, 0, bw, bh, mi_x, mi_y);
     }
@@ skipping 15 matching lines @@
                                     MAX_MB_PLANE - 1);
 }
 
 // TODO(jingning): This function serves as a placeholder for decoder prediction
 // using on demand border extension. It should be moved to /decoder/ directory.
 static void dec_build_inter_predictors(MACROBLOCKD *xd, int plane, int block,
                                        int bw, int bh,
                                        int x, int y, int w, int h,
                                        int mi_x, int mi_y) {
   struct macroblockd_plane *const pd = &xd->plane[plane];
-  const MODE_INFO *mi = xd->mi[0];
+  const MODE_INFO *mi = xd->mi[0].src_mi;
   const int is_compound = has_second_ref(&mi->mbmi);
   const InterpKernel *kernel = vp9_get_interp_kernel(mi->mbmi.interp_filter);
   int ref;
 
   for (ref = 0; ref < 1 + is_compound; ++ref) {
     const struct scale_factors *const sf = &xd->block_refs[ref]->sf;
     struct buf_2d *const pre_buf = &pd->pre[ref];
     struct buf_2d *const dst_buf = &pd->dst;
     uint8_t *const dst = dst_buf->buf + dst_buf->stride * y + x;
     const MV mv = mi->mbmi.sb_type < BLOCK_8X8
@@ skipping 95 matching lines @@
         y0 -= VP9_INTERP_EXTEND - 1;
         y1 += VP9_INTERP_EXTEND;
         y_pad = 1;
       }
 
       // Skip border extension if block is inside the frame.
       if (x0 < 0 || x0 > frame_width - 1 || x1 < 0 || x1 > frame_width - 1 ||
           y0 < 0 || y0 > frame_height - 1 || y1 < 0 || y1 > frame_height - 1) {
         uint8_t *buf_ptr1 = ref_frame + y0 * pre_buf->stride + x0;
         // Extend the border.
-        build_mc_border(buf_ptr1, pre_buf->stride, xd->mc_buf, x1 - x0 + 1,
-                        x0, y0, x1 - x0 + 1, y1 - y0 + 1, frame_width,
+#if CONFIG_VP9_HIGHBITDEPTH
+        if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+          high_build_mc_border(buf_ptr1,
+                               pre_buf->stride,
+                               xd->mc_buf_high,
+                               x1 - x0 + 1,
+                               x0,
+                               y0,
+                               x1 - x0 + 1,
+                               y1 - y0 + 1,
+                               frame_width,
+                               frame_height);
+          buf_stride = x1 - x0 + 1;
+          buf_ptr = CONVERT_TO_BYTEPTR(xd->mc_buf_high) +
+              y_pad * 3 * buf_stride + x_pad * 3;
+        } else {
+          build_mc_border(buf_ptr1,
+                          pre_buf->stride,
+                          xd->mc_buf,
+                          x1 - x0 + 1,
+                          x0,
+                          y0,
+                          x1 - x0 + 1,
+                          y1 - y0 + 1,
+                          frame_width,
+                          frame_height);
+          buf_stride = x1 - x0 + 1;
+          buf_ptr = xd->mc_buf + y_pad * 3 * buf_stride + x_pad * 3;
+        }
+#else
+        build_mc_border(buf_ptr1,
+                        pre_buf->stride,
+                        xd->mc_buf,
+                        x1 - x0 + 1,
+                        x0,
+                        y0,
+                        x1 - x0 + 1,
+                        y1 - y0 + 1,
+                        frame_width,
                         frame_height);
         buf_stride = x1 - x0 + 1;
         buf_ptr = xd->mc_buf + y_pad * 3 * buf_stride + x_pad * 3;
+#endif  // CONFIG_VP9_HIGHBITDEPTH
       }
     }
 
+#if CONFIG_VP9_HIGHBITDEPTH
+    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
+      high_inter_predictor(buf_ptr, buf_stride, dst, dst_buf->stride, subpel_x,
+                           subpel_y, sf, w, h, ref, kernel, xs, ys, xd->bd);
+    } else {
+      inter_predictor(buf_ptr, buf_stride, dst, dst_buf->stride, subpel_x,
+                    subpel_y, sf, w, h, ref, kernel, xs, ys);
+    }
+#else
     inter_predictor(buf_ptr, buf_stride, dst, dst_buf->stride, subpel_x,
                     subpel_y, sf, w, h, ref, kernel, xs, ys);
+#endif  // CONFIG_VP9_HIGHBITDEPTH
   }
 }
 
 void vp9_dec_build_inter_predictors_sb(MACROBLOCKD *xd, int mi_row, int mi_col,
                                        BLOCK_SIZE bsize) {
   int plane;
   const int mi_x = mi_col * MI_SIZE;
   const int mi_y = mi_row * MI_SIZE;
   for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
     const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize,
                                                         &xd->plane[plane]);
     const int num_4x4_w = num_4x4_blocks_wide_lookup[plane_bsize];
     const int num_4x4_h = num_4x4_blocks_high_lookup[plane_bsize];
     const int bw = 4 * num_4x4_w;
     const int bh = 4 * num_4x4_h;
 
-    if (xd->mi[0]->mbmi.sb_type < BLOCK_8X8) {
+    if (xd->mi[0].src_mi->mbmi.sb_type < BLOCK_8X8) {
       int i = 0, x, y;
       assert(bsize == BLOCK_8X8);
       for (y = 0; y < num_4x4_h; ++y)
         for (x = 0; x < num_4x4_w; ++x)
           dec_build_inter_predictors(xd, plane, i++, bw, bh,
                                      4 * x, 4 * y, 4, 4, mi_x, mi_y);
     } else {
       dec_build_inter_predictors(xd, plane, 0, bw, bh,
                                  0, 0, bw, bh, mi_x, mi_y);
     }
@@ skipping 27 matching lines @@
     const int strides[4] = {src->y_stride, src->uv_stride, src->uv_stride,
                             src->alpha_stride};
 
     for (i = 0; i < MAX_MB_PLANE; ++i) {
       struct macroblockd_plane *const pd = &xd->plane[i];
       setup_pred_plane(&pd->pre[idx], buffers[i], strides[i], mi_row, mi_col,
                        sf, pd->subsampling_x, pd->subsampling_y);
     }
   }
 }