/*
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 */

#include <assert.h>
#include <stdio.h>
#include <limits.h>

#include "./aom_scale_rtcd.h"
#include "./aom_dsp_rtcd.h"
#include "./aom_config.h"

#include "aom/aom_integer.h"
#include "aom_dsp/blend.h"

#include "av1/common/blockd.h"
#include "av1/common/reconinter.h"
#include "av1/common/reconintra.h"
#if CONFIG_MOTION_VAR
#include "av1/common/onyxc_int.h"
#include "av1/common/obmc.h"
#endif  // CONFIG_MOTION_VAR

// This function will determine whether or not to create a warped
// prediction and return the appropriate motion model depending
// on the configuration. Behavior will change with different
// combinations of GLOBAL_MOTION, WARPED_MOTION and MOTION_VAR.
static INLINE int allow_warp(const MODE_INFO *const mi,
                             const WarpTypesAllowed *const warp_types,
                             const WarpedMotionParams *const gm_params,
#if CONFIG_MOTION_VAR
                             int build_for_obmc,
#endif  // CONFIG_MOTION_VAR
                             WarpedMotionParams *final_warp_params) {
  const MB_MODE_INFO *const mbmi = &mi->mbmi;
  *final_warp_params = default_warp_params;

#if CONFIG_WARPED_MOTION
  if (mbmi->wm_params[0].invalid) return 0;
#endif

// Only global motion configured
#if !CONFIG_WARPED_MOTION && !CONFIG_MOTION_VAR
  (void)mbmi;
  if (warp_types->global_warp_allowed) {
    memcpy(final_warp_params, gm_params, sizeof(*final_warp_params));
    return 1;
  }
#endif  // !CONFIG_WARPED_MOTION && !CONFIG_MOTION_VAR

// Only warped motion configured

// Warped and global motion configured
#if CONFIG_WARPED_MOTION && !CONFIG_MOTION_VAR
  // When both are enabled, warped will take priority. The global parameters
  // will only be used to compute projection samples to find the warped model.
  // Note that when a block chooses global, it will not be possible to
  // select WARPED_CAUSAL.
  if (warp_types->local_warp_allowed) {
    memcpy(final_warp_params, &mbmi->wm_params[0], sizeof(*final_warp_params));
    return 1;
  } else if (warp_types->global_warp_allowed) {
    memcpy(final_warp_params, gm_params, sizeof(*final_warp_params));
    return 1;
  }
#endif  // CONFIG_WARPED_MOTION && !CONFIG_MOTION_VAR

// Motion var and global motion configured
#if CONFIG_MOTION_VAR && !CONFIG_WARPED_MOTION
  // We warp if either case is true:
  //   1.) We are predicting a block which uses global motion
  //   2.) We are predicting a neighboring block of a block using OBMC,
  //       the neighboring block uses global motion, and we have enabled
  //       WARP_GM_NEIGHBORS_WITH_OBMC
  (void)mbmi;
  if (warp_types->global_warp_allowed &&
      (WARP_GM_NEIGHBORS_WITH_OBMC || !build_for_obmc)) {
    memcpy(final_warp_params, gm_params, sizeof(*final_warp_params));
    return 1;
  }
#endif  // CONFIG_MOTION_VAR && !CONFIG_WARPED_MOTION

// Motion var and warped motion configured

// Motion var, warped motion and global motion all configured
#if CONFIG_WARPED_MOTION && CONFIG_MOTION_VAR
  if (warp_types->local_warp_allowed) {
    if ((build_for_obmc && WARP_WM_NEIGHBORS_WITH_OBMC) || (!build_for_obmc)) {
      memcpy(final_warp_params, &mbmi->wm_params[0],
             sizeof(*final_warp_params));
      return 1;
    }
  } else if (warp_types->global_warp_allowed &&
             (WARP_GM_NEIGHBORS_WITH_OBMC || !build_for_obmc)) {
    memcpy(final_warp_params, gm_params, sizeof(*final_warp_params));
    return 1;
  }
#endif  // CONFIG_WARPED_MOTION && CONFIG_MOTION_VAR

  return 0;
}

static INLINE void av1_make_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, ConvolveParams *conv_params, InterpFilters interp_filters,
    const WarpTypesAllowed *warp_types, int p_col, int p_row, int plane,
    int ref,
#if CONFIG_MOTION_VAR
    const MODE_INFO *mi, int build_for_obmc,
#endif
    int xs, int ys, const MACROBLOCKD *xd) {
  (void)xd;
#if !CONFIG_MOTION_VAR
  const MODE_INFO *mi = xd->mi[0];
  (void)mi;
#endif  // CONFIG_MOTION_VAR

// Make sure the selected motion mode is valid for this configuration
#if CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION
  assert_motion_mode_valid(mi->mbmi.motion_mode, 0, xd->global_motion,
#if CONFIG_WARPED_MOTION
                           xd,
#endif
                           mi);
#endif  // CONFIG MOTION_VAR || CONFIG_WARPED_MOTION

  WarpedMotionParams final_warp_params;
  const int do_warp =
      (w >= 8 && h >= 8 &&
       allow_warp(mi, warp_types,
#if CONFIG_COMPOUND_SINGLEREF
                  // TODO(zoeliu): To further check the single
                  // ref comp mode to work together with
                  //               global motion.
                  has_second_ref(&mi->mbmi)
                      ? &xd->global_motion[mi->mbmi.ref_frame[ref]]
                      : &xd->global_motion[mi->mbmi.ref_frame[0]],
#else   // !(CONFIG_COMPOUND_SINGLEREF)
                  &xd->global_motion[mi->mbmi.ref_frame[ref]],
#endif  // CONFIG_COMPOUND_SINGLEREF
#if CONFIG_MOTION_VAR
                  build_for_obmc,
#endif  // CONFIG_MOTION_VAR
                  &final_warp_params));
  if (do_warp
#if CONFIG_AMVR
      && xd->cur_frame_force_integer_mv == 0
#endif
      ) {
    const struct macroblockd_plane *const pd = &xd->plane[plane];
    const struct buf_2d *const pre_buf = &pd->pre[ref];
    av1_warp_plane(&final_warp_params,
#if CONFIG_HIGHBITDEPTH
                   xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH, xd->bd,
#endif  // CONFIG_HIGHBITDEPTH
                   pre_buf->buf0, pre_buf->width, pre_buf->height,
                   pre_buf->stride, dst, p_col, p_row, w, h, dst_stride,
                   pd->subsampling_x, pd->subsampling_y, xs, ys, conv_params);
    return;
  }
#if CONFIG_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    highbd_inter_predictor(src, src_stride, dst, dst_stride, subpel_x, subpel_y,
                           sf, w, h, conv_params, interp_filters, xs, ys,
                           xd->bd);
    return;
  }
#endif  // CONFIG_HIGHBITDEPTH
  inter_predictor(src, src_stride, dst, dst_stride, subpel_x, subpel_y, sf, w,
                  h, conv_params, interp_filters, xs, ys);
}

#define NSMOOTHERS 1

// [smoother][negative][direction]
DECLARE_ALIGNED(16, static uint8_t,
                wedge_mask_obl[NSMOOTHERS][2][WEDGE_DIRECTIONS]
                              [MASK_MASTER_SIZE * MASK_MASTER_SIZE]);

DECLARE_ALIGNED(16, static uint8_t,
                wedge_signflip_lookup[BLOCK_SIZES_ALL][MAX_WEDGE_TYPES]);

// 4 * MAX_WEDGE_SQUARE is an easy to compute and fairly tight upper bound
// on the sum of all mask sizes up to an including MAX_WEDGE_SQUARE.
DECLARE_ALIGNED(16, static uint8_t,
                wedge_mask_buf[2 * MAX_WEDGE_TYPES * 4 * MAX_WEDGE_SQUARE]);

static wedge_masks_type wedge_masks[BLOCK_SIZES_ALL][2];

// Some unused wedge codebooks left temporarily to facilitate experiments.
// To be removed when settled.
/*
static wedge_code_type wedge_codebook_8_hgtw[8] = {
  { WEDGE_OBLIQUE27, 4, 4 },  { WEDGE_OBLIQUE63, 4, 4 },
  { WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
  { WEDGE_OBLIQUE27, 4, 2 },  { WEDGE_OBLIQUE27, 4, 6 },
  { WEDGE_OBLIQUE153, 4, 2 }, { WEDGE_OBLIQUE153, 4, 6 },
};

static wedge_code_type wedge_codebook_8_hltw[8] = {
  { WEDGE_OBLIQUE27, 4, 4 },  { WEDGE_OBLIQUE63, 4, 4 },
  { WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
  { WEDGE_OBLIQUE63, 2, 4 },  { WEDGE_OBLIQUE63, 6, 4 },
  { WEDGE_OBLIQUE117, 2, 4 }, { WEDGE_OBLIQUE117, 6, 4 },
};

static wedge_code_type wedge_codebook_8_heqw[8] = {
  { WEDGE_OBLIQUE27, 4, 4 },  { WEDGE_OBLIQUE63, 4, 4 },
  { WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
  { WEDGE_HORIZONTAL, 4, 2 }, { WEDGE_HORIZONTAL, 4, 6 },
  { WEDGE_VERTICAL, 2, 4 },   { WEDGE_VERTICAL, 6, 4 },
};

static const wedge_code_type wedge_codebook_32_hgtw[32] = {
  { WEDGE_OBLIQUE27, 4, 4 },  { WEDGE_OBLIQUE63, 4, 4 },
  { WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
  { WEDGE_HORIZONTAL, 4, 2 }, { WEDGE_HORIZONTAL, 4, 4 },
  { WEDGE_HORIZONTAL, 4, 6 }, { WEDGE_VERTICAL, 4, 4 },
  { WEDGE_OBLIQUE27, 4, 1 },  { WEDGE_OBLIQUE27, 4, 2 },
  { WEDGE_OBLIQUE27, 4, 3 },  { WEDGE_OBLIQUE27, 4, 5 },
  { WEDGE_OBLIQUE27, 4, 6 },  { WEDGE_OBLIQUE27, 4, 7 },
  { WEDGE_OBLIQUE153, 4, 1 }, { WEDGE_OBLIQUE153, 4, 2 },
  { WEDGE_OBLIQUE153, 4, 3 }, { WEDGE_OBLIQUE153, 4, 5 },
  { WEDGE_OBLIQUE153, 4, 6 }, { WEDGE_OBLIQUE153, 4, 7 },
  { WEDGE_OBLIQUE63, 1, 4 },  { WEDGE_OBLIQUE63, 2, 4 },
  { WEDGE_OBLIQUE63, 3, 4 },  { WEDGE_OBLIQUE63, 5, 4 },
  { WEDGE_OBLIQUE63, 6, 4 },  { WEDGE_OBLIQUE63, 7, 4 },
  { WEDGE_OBLIQUE117, 1, 4 }, { WEDGE_OBLIQUE117, 2, 4 },
  { WEDGE_OBLIQUE117, 3, 4 }, { WEDGE_OBLIQUE117, 5, 4 },
  { WEDGE_OBLIQUE117, 6, 4 }, { WEDGE_OBLIQUE117, 7, 4 },
};

static const wedge_code_type wedge_codebook_32_hltw[32] = {
  { WEDGE_OBLIQUE27, 4, 4 },  { WEDGE_OBLIQUE63, 4, 4 },
  { WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
  { WEDGE_VERTICAL, 2, 4 },   { WEDGE_VERTICAL, 4, 4 },
  { WEDGE_VERTICAL, 6, 4 },   { WEDGE_HORIZONTAL, 4, 4 },
  { WEDGE_OBLIQUE27, 4, 1 },  { WEDGE_OBLIQUE27, 4, 2 },
  { WEDGE_OBLIQUE27, 4, 3 },  { WEDGE_OBLIQUE27, 4, 5 },
  { WEDGE_OBLIQUE27, 4, 6 },  { WEDGE_OBLIQUE27, 4, 7 },
  { WEDGE_OBLIQUE153, 4, 1 }, { WEDGE_OBLIQUE153, 4, 2 },
  { WEDGE_OBLIQUE153, 4, 3 }, { WEDGE_OBLIQUE153, 4, 5 },
  { WEDGE_OBLIQUE153, 4, 6 }, { WEDGE_OBLIQUE153, 4, 7 },
  { WEDGE_OBLIQUE63, 1, 4 },  { WEDGE_OBLIQUE63, 2, 4 },
  { WEDGE_OBLIQUE63, 3, 4 },  { WEDGE_OBLIQUE63, 5, 4 },
  { WEDGE_OBLIQUE63, 6, 4 },  { WEDGE_OBLIQUE63, 7, 4 },
  { WEDGE_OBLIQUE117, 1, 4 }, { WEDGE_OBLIQUE117, 2, 4 },
  { WEDGE_OBLIQUE117, 3, 4 }, { WEDGE_OBLIQUE117, 5, 4 },
  { WEDGE_OBLIQUE117, 6, 4 }, { WEDGE_OBLIQUE117, 7, 4 },
};

static const wedge_code_type wedge_codebook_32_heqw[32] = {
  { WEDGE_OBLIQUE27, 4, 4 },  { WEDGE_OBLIQUE63, 4, 4 },
  { WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
  { WEDGE_HORIZONTAL, 4, 2 }, { WEDGE_HORIZONTAL, 4, 6 },
  { WEDGE_VERTICAL, 2, 4 },   { WEDGE_VERTICAL, 6, 4 },
  { WEDGE_OBLIQUE27, 4, 1 },  { WEDGE_OBLIQUE27, 4, 2 },
  { WEDGE_OBLIQUE27, 4, 3 },  { WEDGE_OBLIQUE27, 4, 5 },
  { WEDGE_OBLIQUE27, 4, 6 },  { WEDGE_OBLIQUE27, 4, 7 },
  { WEDGE_OBLIQUE153, 4, 1 }, { WEDGE_OBLIQUE153, 4, 2 },
  { WEDGE_OBLIQUE153, 4, 3 }, { WEDGE_OBLIQUE153, 4, 5 },
  { WEDGE_OBLIQUE153, 4, 6 }, { WEDGE_OBLIQUE153, 4, 7 },
  { WEDGE_OBLIQUE63, 1, 4 },  { WEDGE_OBLIQUE63, 2, 4 },
  { WEDGE_OBLIQUE63, 3, 4 },  { WEDGE_OBLIQUE63, 5, 4 },
  { WEDGE_OBLIQUE63, 6, 4 },  { WEDGE_OBLIQUE63, 7, 4 },
  { WEDGE_OBLIQUE117, 1, 4 }, { WEDGE_OBLIQUE117, 2, 4 },
  { WEDGE_OBLIQUE117, 3, 4 }, { WEDGE_OBLIQUE117, 5, 4 },
  { WEDGE_OBLIQUE117, 6, 4 }, { WEDGE_OBLIQUE117, 7, 4 },
};
*/

static const wedge_code_type wedge_codebook_16_hgtw[16] = {
  { WEDGE_OBLIQUE27, 4, 4 },  { WEDGE_OBLIQUE63, 4, 4 },
  { WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
  { WEDGE_HORIZONTAL, 4, 2 }, { WEDGE_HORIZONTAL, 4, 4 },
  { WEDGE_HORIZONTAL, 4, 6 }, { WEDGE_VERTICAL, 4, 4 },
  { WEDGE_OBLIQUE27, 4, 2 },  { WEDGE_OBLIQUE27, 4, 6 },
  { WEDGE_OBLIQUE153, 4, 2 }, { WEDGE_OBLIQUE153, 4, 6 },
  { WEDGE_OBLIQUE63, 2, 4 },  { WEDGE_OBLIQUE63, 6, 4 },
  { WEDGE_OBLIQUE117, 2, 4 }, { WEDGE_OBLIQUE117, 6, 4 },
};

static const wedge_code_type wedge_codebook_16_hltw[16] = {
  { WEDGE_OBLIQUE27, 4, 4 },  { WEDGE_OBLIQUE63, 4, 4 },
  { WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
  { WEDGE_VERTICAL, 2, 4 },   { WEDGE_VERTICAL, 4, 4 },
  { WEDGE_VERTICAL, 6, 4 },   { WEDGE_HORIZONTAL, 4, 4 },
  { WEDGE_OBLIQUE27, 4, 2 },  { WEDGE_OBLIQUE27, 4, 6 },
  { WEDGE_OBLIQUE153, 4, 2 }, { WEDGE_OBLIQUE153, 4, 6 },
  { WEDGE_OBLIQUE63, 2, 4 },  { WEDGE_OBLIQUE63, 6, 4 },
  { WEDGE_OBLIQUE117, 2, 4 }, { WEDGE_OBLIQUE117, 6, 4 },
};

static const wedge_code_type wedge_codebook_16_heqw[16] = {
  { WEDGE_OBLIQUE27, 4, 4 },  { WEDGE_OBLIQUE63, 4, 4 },
  { WEDGE_OBLIQUE117, 4, 4 }, { WEDGE_OBLIQUE153, 4, 4 },
  { WEDGE_HORIZONTAL, 4, 2 }, { WEDGE_HORIZONTAL, 4, 6 },
  { WEDGE_VERTICAL, 2, 4 },   { WEDGE_VERTICAL, 6, 4 },
  { WEDGE_OBLIQUE27, 4, 2 },  { WEDGE_OBLIQUE27, 4, 6 },
  { WEDGE_OBLIQUE153, 4, 2 }, { WEDGE_OBLIQUE153, 4, 6 },
  { WEDGE_OBLIQUE63, 2, 4 },  { WEDGE_OBLIQUE63, 6, 4 },
  { WEDGE_OBLIQUE117, 2, 4 }, { WEDGE_OBLIQUE117, 6, 4 },
};

const wedge_params_type wedge_params_lookup[BLOCK_SIZES_ALL] = {
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
  { 4, wedge_codebook_16_heqw, wedge_signflip_lookup[BLOCK_8X8], 0,
    wedge_masks[BLOCK_8X8] },
  { 4, wedge_codebook_16_hgtw, wedge_signflip_lookup[BLOCK_8X16], 0,
    wedge_masks[BLOCK_8X16] },
  { 4, wedge_codebook_16_hltw, wedge_signflip_lookup[BLOCK_16X8], 0,
    wedge_masks[BLOCK_16X8] },
  { 4, wedge_codebook_16_heqw, wedge_signflip_lookup[BLOCK_16X16], 0,
    wedge_masks[BLOCK_16X16] },
  { 4, wedge_codebook_16_hgtw, wedge_signflip_lookup[BLOCK_16X32], 0,
    wedge_masks[BLOCK_16X32] },
  { 4, wedge_codebook_16_hltw, wedge_signflip_lookup[BLOCK_32X16], 0,
    wedge_masks[BLOCK_32X16] },
  { 4, wedge_codebook_16_heqw, wedge_signflip_lookup[BLOCK_32X32], 0,
    wedge_masks[BLOCK_32X32] },
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
#if CONFIG_EXT_PARTITION
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
#endif  // CONFIG_EXT_PARTITION
  { 0, wedge_codebook_16_hgtw, wedge_signflip_lookup[BLOCK_4X16], 0,
    wedge_masks[BLOCK_4X16] },
  { 0, wedge_codebook_16_hltw, wedge_signflip_lookup[BLOCK_16X4], 0,
    wedge_masks[BLOCK_16X4] },
  { 4, wedge_codebook_16_hgtw, wedge_signflip_lookup[BLOCK_8X32], 0,
    wedge_masks[BLOCK_8X32] },
  { 4, wedge_codebook_16_hltw, wedge_signflip_lookup[BLOCK_32X8], 0,
    wedge_masks[BLOCK_32X8] },
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
#if CONFIG_EXT_PARTITION
  { 0, NULL, NULL, 0, NULL },
  { 0, NULL, NULL, 0, NULL },
#endif  // CONFIG_EXT_PARTITION
};

static const uint8_t *get_wedge_mask_inplace(int wedge_index, int neg,
                                             BLOCK_SIZE sb_type) {
  const uint8_t *master;
  const int bh = block_size_high[sb_type];
  const int bw = block_size_wide[sb_type];
  const wedge_code_type *a =
      wedge_params_lookup[sb_type].codebook + wedge_index;
  const int smoother = wedge_params_lookup[sb_type].smoother;
  int woff, hoff;
  const uint8_t wsignflip = wedge_params_lookup[sb_type].signflip[wedge_index];

  assert(wedge_index >= 0 &&
         wedge_index < (1 << get_wedge_bits_lookup(sb_type)));
  woff = (a->x_offset * bw) >> 3;
  hoff = (a->y_offset * bh) >> 3;
  master = wedge_mask_obl[smoother][neg ^ wsignflip][a->direction] +
           MASK_MASTER_STRIDE * (MASK_MASTER_SIZE / 2 - hoff) +
           MASK_MASTER_SIZE / 2 - woff;
  return master;
}

const uint8_t *av1_get_soft_mask(int wedge_index, int wedge_sign,
                                 BLOCK_SIZE sb_type, int offset_x,
                                 int offset_y) {
  const uint8_t *mask =
      get_wedge_mask_inplace(wedge_index, wedge_sign, sb_type);
  if (mask) mask -= (offset_x + offset_y * MASK_MASTER_STRIDE);
  return mask;
}

static uint8_t *invert_mask(uint8_t *mask_inv_buffer, const uint8_t *const mask,
                            int h, int w, int stride) {
  int i, j;

  for (i = 0; i < h; ++i)
    for (j = 0; j < w; ++j) {
      mask_inv_buffer[i * stride + j] =
          AOM_BLEND_A64_MAX_ALPHA - mask[i * stride + j];
    }
  return mask_inv_buffer;
}

const uint8_t *av1_get_compound_type_mask_inverse(
    const INTERINTER_COMPOUND_DATA *const comp_data, uint8_t *mask_buffer,
    int h, int w, int stride, BLOCK_SIZE sb_type) {
  assert(is_masked_compound_type(comp_data->interinter_compound_type));
  (void)sb_type;
  switch (comp_data->interinter_compound_type) {
    case COMPOUND_WEDGE:
      return av1_get_contiguous_soft_mask(comp_data->wedge_index,
                                          !comp_data->wedge_sign, sb_type);
    case COMPOUND_SEG:
      return invert_mask(mask_buffer, comp_data->seg_mask, h, w, stride);
    default: assert(0); return NULL;
  }
}

const uint8_t *av1_get_compound_type_mask(
    const INTERINTER_COMPOUND_DATA *const comp_data, BLOCK_SIZE sb_type) {
  assert(is_masked_compound_type(comp_data->interinter_compound_type));
  (void)sb_type;
  switch (comp_data->interinter_compound_type) {
    case COMPOUND_WEDGE:
      return av1_get_contiguous_soft_mask(comp_data->wedge_index,
                                          comp_data->wedge_sign, sb_type);
    case COMPOUND_SEG: return comp_data->seg_mask;
    default: assert(0); return NULL;
  }
}

#if COMPOUND_SEGMENT_TYPE == 0
static void uniform_mask(uint8_t *mask, int which_inverse, BLOCK_SIZE sb_type,
                         int h, int w, int mask_val) {
  int i, j;
  int block_stride = block_size_wide[sb_type];
  for (i = 0; i < h; ++i)
    for (j = 0; j < w; ++j) {
      mask[i * block_stride + j] =
          which_inverse ? AOM_BLEND_A64_MAX_ALPHA - mask_val : mask_val;
    }
}

void build_compound_seg_mask(uint8_t *mask, SEG_MASK_TYPE mask_type,
                             const uint8_t *src0, int src0_stride,
                             const uint8_t *src1, int src1_stride,
                             BLOCK_SIZE sb_type, int h, int w) {
  (void)src0;
  (void)src1;
  (void)src0_stride;
  (void)src1_stride;
  switch (mask_type) {
    case UNIFORM_45: uniform_mask(mask, 0, sb_type, h, w, 45); break;
    case UNIFORM_45_INV: uniform_mask(mask, 1, sb_type, h, w, 45); break;
    default: assert(0);
  }
}

#if CONFIG_HIGHBITDEPTH
void build_compound_seg_mask_highbd(uint8_t *mask, SEG_MASK_TYPE mask_type,
                                    const uint8_t *src0, int src0_stride,
                                    const uint8_t *src1, int src1_stride,
                                    BLOCK_SIZE sb_type, int h, int w, int bd) {
  (void)src0;
  (void)src1;
  (void)src0_stride;
  (void)src1_stride;
  (void)bd;
  switch (mask_type) {
    case UNIFORM_45: uniform_mask(mask, 0, sb_type, h, w, 45); break;
    case UNIFORM_45_INV: uniform_mask(mask, 1, sb_type, h, w, 45); break;
    default: assert(0);
  }
}
#endif  // CONFIG_HIGHBITDEPTH

#elif COMPOUND_SEGMENT_TYPE == 1
#define DIFF_FACTOR 16

#if CONFIG_CONVOLVE_ROUND
static void diffwtd_mask_d32(uint8_t *mask, int which_inverse, int mask_base,
                             const int32_t *src0, int src0_stride,
                             const int32_t *src1, int src1_stride,
                             BLOCK_SIZE sb_type, int h, int w,
                             ConvolveParams *conv_params, int bd) {
  int round =
      2 * FILTER_BITS - conv_params->round_0 - conv_params->round_1 + (bd - 8);
  int i, j, m, diff;
  int block_stride = block_size_wide[sb_type];
  for (i = 0; i < h; ++i) {
    for (j = 0; j < w; ++j) {
      diff = abs(src0[i * src0_stride + j] - src1[i * src1_stride + j]);
      diff = ROUND_POWER_OF_TWO(diff, round);
      m = clamp(mask_base + (diff / DIFF_FACTOR), 0, AOM_BLEND_A64_MAX_ALPHA);
      mask[i * block_stride + j] =
          which_inverse ? AOM_BLEND_A64_MAX_ALPHA - m : m;
    }
  }
}

static void build_compound_seg_mask_d32(uint8_t *mask, SEG_MASK_TYPE mask_type,
                                        const int32_t *src0, int src0_stride,
                                        const int32_t *src1, int src1_stride,
                                        BLOCK_SIZE sb_type, int h, int w,
                                        ConvolveParams *conv_params, int bd) {
  switch (mask_type) {
    case DIFFWTD_38:
      diffwtd_mask_d32(mask, 0, 38, src0, src0_stride, src1, src1_stride,
                       sb_type, h, w, conv_params, bd);
      break;
    case DIFFWTD_38_INV:
      diffwtd_mask_d32(mask, 1, 38, src0, src0_stride, src1, src1_stride,
                       sb_type, h, w, conv_params, bd);
      break;
    default: assert(0);
  }
}
#endif  // CONFIG_CONVOLVE_ROUND

static void diffwtd_mask(uint8_t *mask, int which_inverse, int mask_base,
                         const uint8_t *src0, int src0_stride,
                         const uint8_t *src1, int src1_stride,
                         BLOCK_SIZE sb_type, int h, int w) {
  int i, j, m, diff;
  int block_stride = block_size_wide[sb_type];
  for (i = 0; i < h; ++i) {
    for (j = 0; j < w; ++j) {
      diff =
          abs((int)src0[i * src0_stride + j] - (int)src1[i * src1_stride + j]);
      m = clamp(mask_base + (diff / DIFF_FACTOR), 0, AOM_BLEND_A64_MAX_ALPHA);
      mask[i * block_stride + j] =
          which_inverse ? AOM_BLEND_A64_MAX_ALPHA - m : m;
    }
  }
}

void build_compound_seg_mask(uint8_t *mask, SEG_MASK_TYPE mask_type,
                             const uint8_t *src0, int src0_stride,
                             const uint8_t *src1, int src1_stride,
                             BLOCK_SIZE sb_type, int h, int w) {
  switch (mask_type) {
    case DIFFWTD_38:
      diffwtd_mask(mask, 0, 38, src0, src0_stride, src1, src1_stride, sb_type,
                   h, w);
      break;
    case DIFFWTD_38_INV:
      diffwtd_mask(mask, 1, 38, src0, src0_stride, src1, src1_stride, sb_type,
                   h, w);
      break;
    default: assert(0);
  }
}

#if CONFIG_HIGHBITDEPTH
static void diffwtd_mask_highbd(uint8_t *mask, int which_inverse, int mask_base,
                                const uint16_t *src0, int src0_stride,
                                const uint16_t *src1, int src1_stride,
                                BLOCK_SIZE sb_type, int h, int w, int bd) {
  int i, j, m, diff;
  int block_stride = block_size_wide[sb_type];
  for (i = 0; i < h; ++i) {
    for (j = 0; j < w; ++j) {
      diff = abs((int)src0[i * src0_stride + j] -
                 (int)src1[i * src1_stride + j]) >>
             (bd - 8);
      m = clamp(mask_base + (diff / DIFF_FACTOR), 0, AOM_BLEND_A64_MAX_ALPHA);
      mask[i * block_stride + j] =
          which_inverse ? AOM_BLEND_A64_MAX_ALPHA - m : m;
    }
  }
}

void build_compound_seg_mask_highbd(uint8_t *mask, SEG_MASK_TYPE mask_type,
                                    const uint8_t *src0, int src0_stride,
                                    const uint8_t *src1, int src1_stride,
                                    BLOCK_SIZE sb_type, int h, int w, int bd) {
  switch (mask_type) {
    case DIFFWTD_38:
      diffwtd_mask_highbd(mask, 0, 38, CONVERT_TO_SHORTPTR(src0), src0_stride,
                          CONVERT_TO_SHORTPTR(src1), src1_stride, sb_type, h, w,
                          bd);
      break;
    case DIFFWTD_38_INV:
      diffwtd_mask_highbd(mask, 1, 38, CONVERT_TO_SHORTPTR(src0), src0_stride,
                          CONVERT_TO_SHORTPTR(src1), src1_stride, sb_type, h, w,
                          bd);
      break;
    default: assert(0);
  }
}
#endif  // CONFIG_HIGHBITDEPTH
#endif  // COMPOUND_SEGMENT_TYPE

#if MASK_MASTER_SIZE == 64
static const uint8_t wedge_master_oblique_odd[NSMOOTHERS][MASK_MASTER_SIZE] = {
  {
      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  1,  2,  6,  18,
      37, 53, 60, 63, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
      64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
  }
};
static const uint8_t wedge_master_oblique_even[NSMOOTHERS][MASK_MASTER_SIZE] = {
  {
      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
      0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  1,  4,  11, 27,
      46, 58, 62, 63, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
      64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
  }
};
static const uint8_t wedge_master_vertical[NSMOOTHERS][MASK_MASTER_SIZE] = { {
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
    0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  2,  7,  21,
    43, 57, 62, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
    64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
} };

static void shift_copy(const uint8_t *src, uint8_t *dst, int shift, int width) {
  if (shift >= 0) {
    memcpy(dst + shift, src, width - shift);
    memset(dst, src[0], shift);
  } else {
    shift = -shift;
    memcpy(dst, src + shift, width - shift);
    memset(dst + width - shift, src[width - 1], shift);
  }
}
#else
static const double smoother_param[NSMOOTHERS] = { 3.0 };
#endif  // MASK_MASTER_SIZE == 64

static void init_wedge_master_masks() {
  int i, j, s;
  const int w = MASK_MASTER_SIZE;
  const int h = MASK_MASTER_SIZE;
  const int stride = MASK_MASTER_STRIDE;
  for (s = 0; s < NSMOOTHERS; s++) {
// Note: index [0] stores the masters, and [1] its complement.
#if MASK_MASTER_SIZE == 64
    // Generate prototype by shifting the masters
    int shift = h / 4;
    for (i = 0; i < h; i += 2) {
      shift_copy(wedge_master_oblique_even[s],
                 &wedge_mask_obl[s][0][WEDGE_OBLIQUE63][i * stride], shift,
                 MASK_MASTER_SIZE);
      shift--;
      shift_copy(wedge_master_oblique_odd[s],
                 &wedge_mask_obl[s][0][WEDGE_OBLIQUE63][(i + 1) * stride],
                 shift, MASK_MASTER_SIZE);
      memcpy(&wedge_mask_obl[s][0][WEDGE_VERTICAL][i * stride],
             wedge_master_vertical[s],
             MASK_MASTER_SIZE * sizeof(wedge_master_vertical[s][0]));
      memcpy(&wedge_mask_obl[s][0][WEDGE_VERTICAL][(i + 1) * stride],
             wedge_master_vertical[s],
             MASK_MASTER_SIZE * sizeof(wedge_master_vertical[s][0]));
    }
#else
    const int a[2] = { 2, 1 };
    const double asqrt = sqrt(a[0] * a[0] + a[1] * a[1]);
    for (i = 0; i < h; i++) {
      for (j = 0; j < w; ++j) {
        int x = (2 * j + 1 - w);
        int y = (2 * i + 1 - h);
        double d = (a[0] * x + a[1] * y) / asqrt;
        const int msk = (int)rint((1.0 + tanh(d / smoother_param[s])) * 32);
        wedge_mask_obl[s][0][WEDGE_OBLIQUE63][i * stride + j] = msk;
        const int mskx = (int)rint((1.0 + tanh(x / smoother_param[s])) * 32);
        wedge_mask_obl[s][0][WEDGE_VERTICAL][i * stride + j] = mskx;
      }
    }
#endif  // MASK_MASTER_SIZE == 64
    for (i = 0; i < h; ++i) {
      for (j = 0; j < w; ++j) {
        const int msk = wedge_mask_obl[s][0][WEDGE_OBLIQUE63][i * stride + j];
        wedge_mask_obl[s][0][WEDGE_OBLIQUE27][j * stride + i] = msk;
        wedge_mask_obl[s][0][WEDGE_OBLIQUE117][i * stride + w - 1 - j] =
            wedge_mask_obl[s][0][WEDGE_OBLIQUE153][(w - 1 - j) * stride + i] =
                (1 << WEDGE_WEIGHT_BITS) - msk;
        wedge_mask_obl[s][1][WEDGE_OBLIQUE63][i * stride + j] =
            wedge_mask_obl[s][1][WEDGE_OBLIQUE27][j * stride + i] =
                (1 << WEDGE_WEIGHT_BITS) - msk;
        wedge_mask_obl[s][1][WEDGE_OBLIQUE117][i * stride + w - 1 - j] =
            wedge_mask_obl[s][1][WEDGE_OBLIQUE153][(w - 1 - j) * stride + i] =
                msk;
        const int mskx = wedge_mask_obl[s][0][WEDGE_VERTICAL][i * stride + j];
        wedge_mask_obl[s][0][WEDGE_HORIZONTAL][j * stride + i] = mskx;
        wedge_mask_obl[s][1][WEDGE_VERTICAL][i * stride + j] =
            wedge_mask_obl[s][1][WEDGE_HORIZONTAL][j * stride + i] =
                (1 << WEDGE_WEIGHT_BITS) - mskx;
      }
    }
  }
}

// If the signs for the wedges for various blocksizes are
// inconsistent flip the sign flag. Do it only once for every
// wedge codebook.
static void init_wedge_signs() {
  BLOCK_SIZE sb_type;
  memset(wedge_signflip_lookup, 0, sizeof(wedge_signflip_lookup));
  for (sb_type = BLOCK_4X4; sb_type < BLOCK_SIZES_ALL; ++sb_type) {
    const int bw = block_size_wide[sb_type];
    const int bh = block_size_high[sb_type];
    const wedge_params_type wedge_params = wedge_params_lookup[sb_type];
    const int wbits = wedge_params.bits;
    const int wtypes = 1 << wbits;
    int i, w;
    if (wbits == 0) continue;
    for (w = 0; w < wtypes; ++w) {
      // Get the mask master, i.e. index [0]
      const uint8_t *mask = get_wedge_mask_inplace(w, 0, sb_type);
      int avg = 0;
      for (i = 0; i < bw; ++i) avg += mask[i];
      for (i = 1; i < bh; ++i) avg += mask[i * MASK_MASTER_STRIDE];
      avg = (avg + (bw + bh - 1) / 2) / (bw + bh - 1);
      // Default sign of this wedge is 1 if the average < 32, 0 otherwise.
      // If default sign is 1:
      //   If sign requested is 0, we need to flip the sign and return
      //   the complement i.e. index [1] instead. If sign requested is 1
      //   we need to flip the sign and return index [0] instead.
      // If default sign is 0:
      //   If sign requested is 0, we need to return index [0] the master
      //   if sign requested is 1, we need to return the complement index [1]
      //   instead.
      wedge_params.signflip[w] = (avg < 32);
      // printf("%d[%d] = %d\n", sb_type, w, wedge_params.signflip[w]);
    }
  }
}

static void init_wedge_masks() {
  uint8_t *dst = wedge_mask_buf;
  BLOCK_SIZE bsize;
  memset(wedge_masks, 0, sizeof(wedge_masks));
  for (bsize = BLOCK_4X4; bsize < BLOCK_SIZES_ALL; ++bsize) {
    const uint8_t *mask;
    const int bw = block_size_wide[bsize];
    const int bh = block_size_high[bsize];
    const wedge_params_type *wedge_params = &wedge_params_lookup[bsize];
    const int wbits = wedge_params->bits;
    const int wtypes = 1 << wbits;
    int w;
    if (wbits == 0) continue;
    for (w = 0; w < wtypes; ++w) {
      mask = get_wedge_mask_inplace(w, 0, bsize);
      aom_convolve_copy(mask, MASK_MASTER_STRIDE, dst, bw, NULL, 0, NULL, 0, bw,
                        bh);
      wedge_params->masks[0][w] = dst;
      dst += bw * bh;

      mask = get_wedge_mask_inplace(w, 1, bsize);
      aom_convolve_copy(mask, MASK_MASTER_STRIDE, dst, bw, NULL, 0, NULL, 0, bw,
                        bh);
      wedge_params->masks[1][w] = dst;
      dst += bw * bh;
    }
    assert(sizeof(wedge_mask_buf) >= (size_t)(dst - wedge_mask_buf));
  }
}

// Equation of line: f(x, y) = a[0]*(x - a[2]*w/8) + a[1]*(y - a[3]*h/8) = 0
void av1_init_wedge_masks() {
  init_wedge_master_masks();
  init_wedge_signs();
  init_wedge_masks();
}

#if CONFIG_CONVOLVE_ROUND
static void build_masked_compound_no_round(
    CONV_BUF_TYPE *dst, int dst_stride, const CONV_BUF_TYPE *src0,
    int src0_stride, const CONV_BUF_TYPE *src1, int src1_stride,
    const INTERINTER_COMPOUND_DATA *const comp_data, BLOCK_SIZE sb_type, int h,
    int w) {
  // Derive subsampling from h and w passed in. May be refactored to
  // pass in subsampling factors directly.
  const int subh = (2 << b_height_log2_lookup[sb_type]) == h;
  const int subw = (2 << b_width_log2_lookup[sb_type]) == w;
  const uint8_t *mask = av1_get_compound_type_mask(comp_data, sb_type);
  aom_blend_a64_d32_mask(dst, dst_stride, src0, src0_stride, src1, src1_stride,
                         mask, block_size_wide[sb_type], h, w, subh, subw);
}
#endif  // CONFIG_CONVOLVE_ROUND
static void build_masked_compound(
    uint8_t *dst, int dst_stride, const uint8_t *src0, int src0_stride,
    const uint8_t *src1, int src1_stride,
    const INTERINTER_COMPOUND_DATA *const comp_data, BLOCK_SIZE sb_type, int h,
    int w) {
  // Derive subsampling from h and w passed in. May be refactored to
  // pass in subsampling factors directly.
  const int subh = (2 << b_height_log2_lookup[sb_type]) == h;
  const int subw = (2 << b_width_log2_lookup[sb_type]) == w;
  const uint8_t *mask = av1_get_compound_type_mask(comp_data, sb_type);
  aom_blend_a64_mask(dst, dst_stride, src0, src0_stride, src1, src1_stride,
                     mask, block_size_wide[sb_type], h, w, subh, subw);
}

#if CONFIG_HIGHBITDEPTH
static void build_masked_compound_highbd(
    uint8_t *dst_8, int dst_stride, const uint8_t *src0_8, int src0_stride,
    const uint8_t *src1_8, int src1_stride,
    const INTERINTER_COMPOUND_DATA *const comp_data, BLOCK_SIZE sb_type, int h,
    int w, int bd) {
  // Derive subsampling from h and w passed in. May be refactored to
  // pass in subsampling factors directly.
  const int subh = (2 << b_height_log2_lookup[sb_type]) == h;
  const int subw = (2 << b_width_log2_lookup[sb_type]) == w;
  const uint8_t *mask = av1_get_compound_type_mask(comp_data, sb_type);
  // const uint8_t *mask =
  //     av1_get_contiguous_soft_mask(wedge_index, wedge_sign, sb_type);
  aom_highbd_blend_a64_mask(dst_8, dst_stride, src0_8, src0_stride, src1_8,
                            src1_stride, mask, block_size_wide[sb_type], h, w,
                            subh, subw, bd);
}
#endif  // CONFIG_HIGHBITDEPTH

void av1_make_masked_inter_predictor(
    const uint8_t *pre, int pre_stride, uint8_t *dst, int dst_stride,
    const int subpel_x, const int subpel_y, const struct scale_factors *sf,
    int w, int h, ConvolveParams *conv_params, InterpFilters interp_filters,
    int xs, int ys, int plane, const WarpTypesAllowed *warp_types, int p_col,
    int p_row, int ref, MACROBLOCKD *xd) {
  const MODE_INFO *mi = xd->mi[0];

  const INTERINTER_COMPOUND_DATA comp_data = {
    mi->mbmi.wedge_index, mi->mbmi.wedge_sign, mi->mbmi.mask_type, xd->seg_mask,
    mi->mbmi.interinter_compound_type
  };

// We're going to call av1_make_inter_predictor to generate a prediction into
// a temporary buffer, then will blend that temporary buffer with that from
// the other reference.
//
// With CONFIG_CONVOLVE_ROUND, if the rounding mode is CONVOLVE_OPT_NO_ROUND
// then the predictions are at 32-bits, so we'll need 32 bits per
// pixel. Otherwise, we'll need up to 16 bits per pixel if
// CONFIG_HIGHBITDEPTH or just 8 otherwise.
#if CONFIG_CONVOLVE_ROUND
#define INTER_PRED_BYTES_PER_PIXEL 4
#elif CONFIG_HIGHBITDEPTH
#define INTER_PRED_BYTES_PER_PIXEL 2
#else
#define INTER_PRED_BYTES_PER_PIXEL 1
#endif
  DECLARE_ALIGNED(16, uint8_t,
                  tmp_buf[INTER_PRED_BYTES_PER_PIXEL * MAX_SB_SQUARE]);
#undef INTER_PRED_BYTES_PER_PIXEL

#if CONFIG_HIGHBITDEPTH
  uint8_t *tmp_dst = (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
                         ? CONVERT_TO_BYTEPTR(tmp_buf)
                         : tmp_buf;
#else
  uint8_t *tmp_dst = tmp_buf;
#endif

#if CONFIG_CONVOLVE_ROUND
  const int tmp_buf_stride = MAX_SB_SIZE;
  const int is_conv_no_round = conv_params->round == CONVOLVE_OPT_NO_ROUND;
  CONV_BUF_TYPE *org_dst = conv_params->dst;
  int org_dst_stride = conv_params->dst_stride;
  CONV_BUF_TYPE *tmp_buf32 = (CONV_BUF_TYPE *)tmp_buf;
  if (is_conv_no_round) {
    conv_params->dst = tmp_buf32;
    conv_params->dst_stride = tmp_buf_stride;
    assert(conv_params->do_average == 0);
  }
#endif  // CONFIG_CONVOLVE_ROUND

  // This will generate a prediction in tmp_buf for the second reference
  av1_make_inter_predictor(pre, pre_stride, tmp_dst, MAX_SB_SIZE, subpel_x,
                           subpel_y, sf, w, h, conv_params, interp_filters,
                           warp_types, p_col, p_row, plane, ref,
#if CONFIG_MOTION_VAR
                           mi, 0,
#endif
                           xs, ys, xd);

  if (!plane && comp_data.interinter_compound_type == COMPOUND_SEG) {
#if CONFIG_CONVOLVE_ROUND
    if (is_conv_no_round) {
      build_compound_seg_mask_d32(comp_data.seg_mask, comp_data.mask_type,
                                  org_dst, org_dst_stride, tmp_buf32,
                                  tmp_buf_stride, mi->mbmi.sb_type, h, w,
                                  conv_params, xd->bd);
    } else {
#endif  // CONFIG_CONVOLVE_ROUND
#if CONFIG_HIGHBITDEPTH
      if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
        build_compound_seg_mask_highbd(comp_data.seg_mask, comp_data.mask_type,
                                       dst, dst_stride, tmp_dst, MAX_SB_SIZE,
                                       mi->mbmi.sb_type, h, w, xd->bd);
      } else {
#endif
        build_compound_seg_mask(comp_data.seg_mask, comp_data.mask_type, dst,
                                dst_stride, tmp_dst, MAX_SB_SIZE,
                                mi->mbmi.sb_type, h, w);
#if CONFIG_HIGHBITDEPTH
      }
#endif
#if CONFIG_CONVOLVE_ROUND
    }
#endif
  }

#if CONFIG_CONVOLVE_ROUND
  if (is_conv_no_round) {
    build_masked_compound_no_round(org_dst, org_dst_stride, org_dst,
                                   org_dst_stride, tmp_buf32, tmp_buf_stride,
                                   &comp_data, mi->mbmi.sb_type, h, w);

    const int convolve_rounding_bits =
        FILTER_BITS * 2 - conv_params->round_0 - conv_params->round_1;
#if CONFIG_HIGHBITDEPTH
    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
      av1_highbd_convolve_rounding(org_dst, org_dst_stride, dst, dst_stride, w,
                                   h, convolve_rounding_bits, xd->bd);
    else
#endif
      av1_convolve_rounding(org_dst, org_dst_stride, dst, dst_stride, w, h,
                            convolve_rounding_bits);

    conv_params->do_post_rounding = 0;
  } else {
#endif  // CONFIG_CONVOLVE_ROUND

#if CONFIG_HIGHBITDEPTH
    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
      build_masked_compound_highbd(dst, dst_stride, dst, dst_stride, tmp_dst,
                                   MAX_SB_SIZE, &comp_data, mi->mbmi.sb_type, h,
                                   w, xd->bd);
    else
#endif  // CONFIG_HIGHBITDEPTH
      build_masked_compound(dst, dst_stride, dst, dst_stride, tmp_dst,
                            MAX_SB_SIZE, &comp_data, mi->mbmi.sb_type, h, w);
#if CONFIG_CONVOLVE_ROUND
  }
#endif  // CONFIG_CONVOLVE_ROUND
}

// TODO(sarahparker) av1_highbd_build_inter_predictor and
// av1_build_inter_predictor should be combined with
// av1_make_inter_predictor
#if CONFIG_HIGHBITDEPTH
void av1_highbd_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,
    InterpFilters interp_filters, const WarpTypesAllowed *warp_types, int p_col,
    int p_row, int plane, enum mv_precision precision, int x, int y,
    const MACROBLOCKD *xd) {
  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 = av1_scale_mv(&mv_q4, x, y, sf);
  mv.col += SCALE_EXTRA_OFF;
  mv.row += SCALE_EXTRA_OFF;
  const int subpel_x = mv.col & SCALE_SUBPEL_MASK;
  const int subpel_y = mv.row & SCALE_SUBPEL_MASK;
  ConvolveParams conv_params = get_conv_params(ref, ref, plane);

  src += (mv.row >> SCALE_SUBPEL_BITS) * src_stride +
         (mv.col >> SCALE_SUBPEL_BITS);

  av1_make_inter_predictor(src, src_stride, dst, dst_stride, subpel_x, subpel_y,
                           sf, w, h, &conv_params, interp_filters, warp_types,
                           p_col, p_row, plane, ref,
#if CONFIG_MOTION_VAR
                           xd->mi[0], 0,
#endif
                           sf->x_step_q4, sf->y_step_q4, xd);
}
#endif  // CONFIG_HIGHBITDEPTH

void av1_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,
    ConvolveParams *conv_params, InterpFilters interp_filters,
    const WarpTypesAllowed *warp_types, int p_col, int p_row, int plane,
    int ref, enum mv_precision precision, int x, int y, const MACROBLOCKD *xd) {
  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 = av1_scale_mv(&mv_q4, x, y, sf);
  mv.col += SCALE_EXTRA_OFF;
  mv.row += SCALE_EXTRA_OFF;
  const int subpel_x = mv.col & SCALE_SUBPEL_MASK;
  const int subpel_y = mv.row & SCALE_SUBPEL_MASK;

  src += (mv.row >> SCALE_SUBPEL_BITS) * src_stride +
         (mv.col >> SCALE_SUBPEL_BITS);

  av1_make_inter_predictor(src, src_stride, dst, dst_stride, subpel_x, subpel_y,
                           sf, w, h, conv_params, interp_filters, warp_types,
                           p_col, p_row, plane, ref,
#if CONFIG_MOTION_VAR
                           xd->mi[0], 0,
#endif
                           sf->x_step_q4, sf->y_step_q4, xd);
}

typedef struct SubpelParams {
  int xs;
  int ys;
  int subpel_x;
  int subpel_y;
} SubpelParams;

#if CONFIG_JNT_COMP
static void jnt_comp_weight_assign(const AV1_COMMON *cm,
                                   const MB_MODE_INFO *mbmi,
                                   ConvolveParams *conv_params,
                                   int is_compound) {
  if (is_compound) {
    int bck_idx = cm->frame_refs[mbmi->ref_frame[0] - LAST_FRAME].idx;
    int fwd_idx = cm->frame_refs[mbmi->ref_frame[1] - LAST_FRAME].idx;
    int bck_frame_index = 0, fwd_frame_index = 0;
    int cur_frame_index = cm->cur_frame->cur_frame_offset;

    if (bck_idx >= 0) {
      bck_frame_index = cm->buffer_pool->frame_bufs[bck_idx].cur_frame_offset;
    }

    if (fwd_idx >= 0) {
      fwd_frame_index = cm->buffer_pool->frame_bufs[fwd_idx].cur_frame_offset;
    }

    conv_params->bck_offset = abs(cur_frame_index - bck_frame_index);
    conv_params->fwd_offset = abs(fwd_frame_index - cur_frame_index);

    const double fwd = abs(fwd_frame_index - cur_frame_index);
    const double bck = abs(cur_frame_index - bck_frame_index);
    int order;
    double ratio;

    if (COMPOUND_WEIGHT_MODE == DIST) {
      if (fwd > bck) {
        ratio = (bck != 0) ? fwd / bck : 5.0;
        order = 0;
      } else {
        ratio = (fwd != 0) ? bck / fwd : 5.0;
        order = 1;
      }
      int quant_dist_idx;
      for (quant_dist_idx = 0; quant_dist_idx < 4; ++quant_dist_idx) {
        if (ratio < quant_dist_category[quant_dist_idx]) break;
      }
      conv_params->fwd_offset =
          quant_dist_lookup_table[0][quant_dist_idx][order];
      conv_params->bck_offset =
          quant_dist_lookup_table[0][quant_dist_idx][1 - order];
    } else {
      conv_params->fwd_offset = (DIST_PRECISION >> 1);
      conv_params->bck_offset = (DIST_PRECISION >> 1);
    }

    if (mbmi->compound_idx) {
      conv_params->fwd_offset = -1;
      conv_params->bck_offset = -1;
    }
  } else {
    conv_params->bck_offset = -1;
    conv_params->fwd_offset = -1;
  }
}
#endif  // CONFIG_JNT_COMP

static INLINE void build_inter_predictors(
    const AV1_COMMON *cm, MACROBLOCKD *xd, int plane,
#if CONFIG_MOTION_VAR
    const MODE_INFO *mi, int build_for_obmc,
#endif  // CONFIG_MOTION_VAR
    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];
#if !CONFIG_MOTION_VAR
  const MODE_INFO *mi = xd->mi[0];
#endif  // CONFIG_MOTION_VAR
  int is_compound = has_second_ref(&mi->mbmi);
#if CONFIG_COMPOUND_SINGLEREF
  int is_comp_mode_pred =
      is_compound || is_inter_singleref_comp_mode(mi->mbmi.mode);
#endif  // CONFIG_COMPOUND_SINGLEREF
  int ref;
#if CONFIG_INTRABC
  const int is_intrabc = is_intrabc_block(&mi->mbmi);
  assert(IMPLIES(is_intrabc, !is_compound));
#endif  // CONFIG_INTRABC
  int is_global[2] = { 0, 0 };
  for (ref = 0; ref < 1 + is_compound; ++ref) {
    WarpedMotionParams *const wm = &xd->global_motion[mi->mbmi.ref_frame[ref]];
    is_global[ref] = is_global_mv_block(mi, block, wm->wmtype);
  }
#if CONFIG_COMPOUND_SINGLEREF
  if (!is_compound && is_comp_mode_pred) is_global[1] = is_global[0];
#endif  // CONFIG_COMPOUND_SINGLEREF

  (void)block;
  (void)cm;

  const BLOCK_SIZE bsize = mi->mbmi.sb_type;
  const int ss_x = pd->subsampling_x;
  const int ss_y = pd->subsampling_y;
  int sub8x8_inter = (block_size_wide[bsize] < 8 && ss_x) ||
                     (block_size_high[bsize] < 8 && ss_y);

#if CONFIG_INTRABC
  if (is_intrabc) {
    sub8x8_inter = 0;
  }
#endif

#if CONFIG_MOTION_VAR
  sub8x8_inter = sub8x8_inter && !build_for_obmc;
#endif  // CONFIG_MOTION_VAR
  const int row_start = (block_size_high[bsize] == 4) && ss_y ? -1 : 0;
  const int col_start = (block_size_wide[bsize] == 4) && ss_x ? -1 : 0;

  if (sub8x8_inter) {
    for (int row = row_start; row <= 0 && sub8x8_inter; ++row)
      for (int col = col_start; col <= 0; ++col)
        if (!is_inter_block(&xd->mi[row * xd->mi_stride + col]->mbmi))
          sub8x8_inter = 0;
  }

  if (sub8x8_inter) {
    // block size
    const int b4_w = block_size_wide[bsize] >> ss_x;
    const int b4_h = block_size_high[bsize] >> ss_y;
    const BLOCK_SIZE plane_bsize = scale_chroma_bsize(bsize, ss_x, ss_y);
    const int b8_w = block_size_wide[plane_bsize] >> ss_x;
    const int b8_h = block_size_high[plane_bsize] >> ss_y;
    int idx, idy;

    const int x_base = x;
    const int y_base = y;

    const struct buf_2d orig_pred_buf[2] = { pd->pre[0], pd->pre[1] };

    int row = row_start;
    for (idy = 0; idy < b8_h; idy += b4_h) {
      int col = col_start;
      for (idx = 0; idx < b8_w; idx += b4_w) {
        MB_MODE_INFO *this_mbmi = &xd->mi[row * xd->mi_stride + col]->mbmi;
        is_compound = has_second_ref(this_mbmi);
#if CONFIG_CONVOLVE_ROUND
        DECLARE_ALIGNED(16, int32_t, tmp_dst[8 * 8]);
        int tmp_dst_stride = 8;
        assert(w <= 8 && h <= 8);
#endif  // CONFIG_CONVOLVE_ROUND
#if CONFIG_CONVOLVE_ROUND
        ConvolveParams conv_params =
            get_conv_params_no_round(0, 0, plane, tmp_dst, tmp_dst_stride);
#else
        ConvolveParams conv_params = get_conv_params(0, 0, plane);
#endif
#if CONFIG_JNT_COMP
        conv_params.fwd_offset = -1;
        conv_params.bck_offset = -1;
#endif  // CONFIG_JNT_COMP
        struct buf_2d *const dst_buf = &pd->dst;
        x = x_base + idx;
        y = y_base + idy;
        uint8_t *dst = dst_buf->buf + dst_buf->stride * y + x;

        // TODO(zoeliu): If single ref comp modes are considered here, a
        //               mismatch was caused. Need a further investigation.
        for (ref = 0; ref < 1 + is_compound; ++ref) {
          const RefBuffer *ref_buf =
              &cm->frame_refs[this_mbmi->ref_frame[ref] - LAST_FRAME];

          const int c_offset = (mi_x + MI_SIZE * col_start) >> ss_x;
          const int r_offset = (mi_y + MI_SIZE * row_start) >> ss_y;
          pd->pre[ref].buf0 =
              (plane == 1) ? ref_buf->buf->u_buffer : ref_buf->buf->v_buffer;
          pd->pre[ref].buf =
              pd->pre[ref].buf0 + scaled_buffer_offset(c_offset, r_offset,
                                                       ref_buf->buf->uv_stride,
                                                       &ref_buf->sf);
          pd->pre[ref].width = ref_buf->buf->uv_crop_width;
          pd->pre[ref].height = ref_buf->buf->uv_crop_height;
          pd->pre[ref].stride = ref_buf->buf->uv_stride;

#if CONFIG_INTRABC
          const struct scale_factors *const sf =
              is_intrabc ? &xd->sf_identity : &ref_buf->sf;
          struct buf_2d *const pre_buf = is_intrabc ? dst_buf : &pd->pre[ref];
#else
          const struct scale_factors *const sf = &ref_buf->sf;
          struct buf_2d *const pre_buf = &pd->pre[ref];
#endif  // CONFIG_INTRABC

          const MV mv = this_mbmi->mv[ref].as_mv;

          uint8_t *pre;
          int xs, ys, subpel_x, subpel_y;
          const int is_scaled = av1_is_scaled(sf);
          WarpTypesAllowed warp_types;
          warp_types.global_warp_allowed = is_global[ref];
#if CONFIG_WARPED_MOTION
          warp_types.local_warp_allowed =
              this_mbmi->motion_mode == WARPED_CAUSAL;
#endif  // CONFIG_WARPED_MOTION

          if (is_scaled) {
            int ssx = pd->subsampling_x;
            int ssy = pd->subsampling_y;
            int orig_pos_y = (mi_y << (SUBPEL_BITS - ssy)) + (y << SUBPEL_BITS);
            orig_pos_y += mv.row * (1 << (1 - ssy));
            int orig_pos_x = (mi_x << (SUBPEL_BITS - ssx)) + (x << SUBPEL_BITS);
            orig_pos_x += mv.col * (1 << (1 - ssx));
            int pos_y = sf->scale_value_y(orig_pos_y, sf);
            int pos_x = sf->scale_value_x(orig_pos_x, sf);
            pos_x += SCALE_EXTRA_OFF;
            pos_y += SCALE_EXTRA_OFF;

            const int top = -AOM_LEFT_TOP_MARGIN_SCALED;
            const int left = -AOM_LEFT_TOP_MARGIN_SCALED;
            const int bottom = (pre_buf->height + AOM_INTERP_EXTEND)
                               << SCALE_SUBPEL_BITS;
            const int right = (pre_buf->width + AOM_INTERP_EXTEND)
                              << SCALE_SUBPEL_BITS;
            pos_y = clamp(pos_y, top, bottom);
            pos_x = clamp(pos_x, left, right);

            pre = pre_buf->buf0 +
                  (pos_y >> SCALE_SUBPEL_BITS) * pre_buf->stride +
                  (pos_x >> SCALE_SUBPEL_BITS);
            subpel_x = pos_x & SCALE_SUBPEL_MASK;
            subpel_y = pos_y & SCALE_SUBPEL_MASK;
            xs = sf->x_step_q4;
            ys = sf->y_step_q4;
          } else {
            const MV mv_q4 = clamp_mv_to_umv_border_sb(
                xd, &mv, bw, bh, pd->subsampling_x, pd->subsampling_y);
            xs = ys = SCALE_SUBPEL_SHIFTS;
            subpel_x = (mv_q4.col & SUBPEL_MASK) << SCALE_EXTRA_BITS;
            subpel_y = (mv_q4.row & SUBPEL_MASK) << SCALE_EXTRA_BITS;
            pre = pre_buf->buf +
                  (y + (mv_q4.row >> SUBPEL_BITS)) * pre_buf->stride +
                  (x + (mv_q4.col >> SUBPEL_BITS));
          }

          conv_params.ref = ref;
          conv_params.do_average = ref;
          if (is_masked_compound_type(mi->mbmi.interinter_compound_type)) {
            // masked compound type has its own average mechanism
            conv_params.do_average = 0;
          }
          if (ref && is_masked_compound_type(mi->mbmi.interinter_compound_type))
            av1_make_masked_inter_predictor(
                pre, pre_buf->stride, dst, dst_buf->stride, subpel_x, subpel_y,
                sf, b4_w, b4_h, &conv_params, mi->mbmi.interp_filters, xs, ys,
                plane, &warp_types, (mi_x >> pd->subsampling_x) + x,
                (mi_y >> pd->subsampling_y) + y, ref, xd);
          else
            av1_make_inter_predictor(
                pre, pre_buf->stride, dst, dst_buf->stride, subpel_x, subpel_y,
                sf, b4_w, b4_h, &conv_params, this_mbmi->interp_filters,
                &warp_types, (mi_x >> pd->subsampling_x) + x,
                (mi_y >> pd->subsampling_y) + y, plane, ref,
#if CONFIG_MOTION_VAR
                mi, build_for_obmc,
#endif  // CONFIG_MOTION_VAR
                xs, ys, xd);
        }  // for (ref = 0; ref < 1 + is_compound; ++ref)
#if CONFIG_CONVOLVE_ROUND
        if (conv_params.do_post_rounding) {
#if CONFIG_HIGHBITDEPTH
          if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
            av1_highbd_convolve_rounding(
                tmp_dst, tmp_dst_stride, dst, dst_buf->stride, b4_w, b4_h,
                FILTER_BITS * 2 + is_compound - conv_params.round_0 -
                    conv_params.round_1,
                xd->bd);
          else
#endif  // CONFIG_HIGHBITDEPTH
#if CONFIG_COMPOUND_SINGLEREF
            av1_convolve_rounding(
                tmp_dst, tmp_dst_stride, dst, dst_buf->stride, b4_w, b4_h,
                FILTER_BITS * 2 + is_comp_mode_pred - conv_params.round_0 -
                    conv_params.round_1);
#else   // !(CONFIG_COMPOUND_SINGLEREF)
          av1_convolve_rounding(tmp_dst, tmp_dst_stride, dst, dst_buf->stride,
                                b4_w, b4_h,
                                FILTER_BITS * 2 + is_compound -
                                    conv_params.round_0 - conv_params.round_1);
#endif  // CONFIG_COMPOUND_SINGLEREF
        }
#endif  // CONFIG_CONVOLVE_ROUND
        ++col;
      }
      ++row;
    }

    for (ref = 0; ref < 2; ++ref) pd->pre[ref] = orig_pred_buf[ref];
    return;
  }

  {
    struct buf_2d *const dst_buf = &pd->dst;
    uint8_t *const dst = dst_buf->buf + dst_buf->stride * y + x;
    uint8_t *pre[2];
    SubpelParams subpel_params[2];
#if CONFIG_CONVOLVE_ROUND
    DECLARE_ALIGNED(16, int32_t, tmp_dst[MAX_SB_SIZE * MAX_SB_SIZE]);
#endif  // CONFIG_CONVOLVE_ROUND

#if CONFIG_COMPOUND_SINGLEREF
    for (ref = 0; ref < 1 + is_comp_mode_pred; ++ref)
#else
    for (ref = 0; ref < 1 + is_compound; ++ref)
#endif  // CONFIG_COMPOUND_SINGLEREF
    {
#if CONFIG_INTRABC
      const struct scale_factors *const sf =
          is_intrabc ? &xd->sf_identity : &xd->block_refs[ref]->sf;
      struct buf_2d *const pre_buf = is_intrabc ? dst_buf : &pd->pre[ref];
#else
      const struct scale_factors *const sf = &xd->block_refs[ref]->sf;
      struct buf_2d *const pre_buf = &pd->pre[ref];
#endif  // CONFIG_INTRABC
      const MV mv = mi->mbmi.mv[ref].as_mv;

      const int is_scaled = av1_is_scaled(sf);
      if (is_scaled) {
        // Note: The various inputs here have different units:
        // * mi_x/mi_y are in units of luma pixels
        // * mv is in units of 1/8 luma pixels
        // * x/y are in units of pixels *in the current plane*
        // Here we unify these into a q4-format position within the current
        // plane, then project into the reference frame
        int ssx = pd->subsampling_x;
        int ssy = pd->subsampling_y;
        int orig_pos_y = (mi_y << (SUBPEL_BITS - ssy)) + (y << SUBPEL_BITS);
        orig_pos_y += mv.row * (1 << (1 - ssy));
        int orig_pos_x = (mi_x << (SUBPEL_BITS - ssx)) + (x << SUBPEL_BITS);
        orig_pos_x += mv.col * (1 << (1 - ssx));
        int pos_y = sf->scale_value_y(orig_pos_y, sf);
        int pos_x = sf->scale_value_x(orig_pos_x, sf);
        pos_x += SCALE_EXTRA_OFF;
        pos_y += SCALE_EXTRA_OFF;

        // Clamp against the reference frame borders, with enough extension
        // that we don't force the reference block to be partially onscreen.
        const int top = -AOM_LEFT_TOP_MARGIN_SCALED;
        const int left = -AOM_LEFT_TOP_MARGIN_SCALED;
        const int bottom = (pre_buf->height + AOM_INTERP_EXTEND)
                           << SCALE_SUBPEL_BITS;
        const int right = (pre_buf->width + AOM_INTERP_EXTEND)
                          << SCALE_SUBPEL_BITS;
        pos_y = clamp(pos_y, top, bottom);
        pos_x = clamp(pos_x, left, right);

        pre[ref] = pre_buf->buf0 +
                   (pos_y >> SCALE_SUBPEL_BITS) * pre_buf->stride +
                   (pos_x >> SCALE_SUBPEL_BITS);
        subpel_params[ref].subpel_x = pos_x & SCALE_SUBPEL_MASK;
        subpel_params[ref].subpel_y = pos_y & SCALE_SUBPEL_MASK;
        subpel_params[ref].xs = sf->x_step_q4;
        subpel_params[ref].ys = sf->y_step_q4;
      } else {
        const MV mv_q4 = clamp_mv_to_umv_border_sb(
            xd, &mv, bw, bh, pd->subsampling_x, pd->subsampling_y);
        subpel_params[ref].subpel_x = (mv_q4.col & SUBPEL_MASK)
                                      << SCALE_EXTRA_BITS;
        subpel_params[ref].subpel_y = (mv_q4.row & SUBPEL_MASK)
                                      << SCALE_EXTRA_BITS;
        subpel_params[ref].xs = SCALE_SUBPEL_SHIFTS;
        subpel_params[ref].ys = SCALE_SUBPEL_SHIFTS;
        pre[ref] = pre_buf->buf +
                   (y + (mv_q4.row >> SUBPEL_BITS)) * pre_buf->stride +
                   (x + (mv_q4.col >> SUBPEL_BITS));
      }
    }

#if CONFIG_CONVOLVE_ROUND
    ConvolveParams conv_params =
        get_conv_params_no_round(ref, ref, plane, tmp_dst, MAX_SB_SIZE);
#if CONFIG_JNT_COMP
    jnt_comp_weight_assign(cm, &mi->mbmi, &conv_params, is_compound);
#endif  // CONFIG_JNT_COMP

#else
    ConvolveParams conv_params = get_conv_params(ref, ref, plane);
#endif  // CONFIG_CONVOLVE_ROUND

#if CONFIG_COMPOUND_SINGLEREF
    for (ref = 0; ref < 1 + is_comp_mode_pred; ++ref)
#else
    for (ref = 0; ref < 1 + is_compound; ++ref)
#endif  // CONFIG_COMPOUND_SINGLEREF
    {
#if CONFIG_INTRABC
      const struct scale_factors *const sf =
          is_intrabc ? &xd->sf_identity : &xd->block_refs[ref]->sf;
      struct buf_2d *const pre_buf = is_intrabc ? dst_buf : &pd->pre[ref];
#else
      const struct scale_factors *const sf = &xd->block_refs[ref]->sf;
      struct buf_2d *const pre_buf = &pd->pre[ref];
#endif  // CONFIG_INTRABC
      WarpTypesAllowed warp_types;
      warp_types.global_warp_allowed = is_global[ref];
#if CONFIG_WARPED_MOTION
      warp_types.local_warp_allowed = mi->mbmi.motion_mode == WARPED_CAUSAL;
#endif  // CONFIG_WARPED_MOTION
      conv_params.ref = ref;
      conv_params.do_average = ref;
      if (is_masked_compound_type(mi->mbmi.interinter_compound_type)) {
        // masked compound type has its own average mechanism
        conv_params.do_average = 0;
      }

      if (ref && is_masked_compound_type(mi->mbmi.interinter_compound_type))
        av1_make_masked_inter_predictor(
            pre[ref], pre_buf->stride, dst, dst_buf->stride,
            subpel_params[ref].subpel_x, subpel_params[ref].subpel_y, sf, w, h,
            &conv_params, mi->mbmi.interp_filters, subpel_params[ref].xs,
            subpel_params[ref].ys, plane, &warp_types,
            (mi_x >> pd->subsampling_x) + x, (mi_y >> pd->subsampling_y) + y,
            ref, xd);
      else
        av1_make_inter_predictor(
            pre[ref], pre_buf->stride, dst, dst_buf->stride,
            subpel_params[ref].subpel_x, subpel_params[ref].subpel_y, sf, w, h,
            &conv_params, mi->mbmi.interp_filters, &warp_types,
            (mi_x >> pd->subsampling_x) + x, (mi_y >> pd->subsampling_y) + y,
            plane, ref,
#if CONFIG_MOTION_VAR
            mi, build_for_obmc,
#endif  // CONFIG_MOTION_VAR
            subpel_params[ref].xs, subpel_params[ref].ys, xd);
    }

#if CONFIG_CONVOLVE_ROUND
    // TODO(angiebird): This part needs optimization
    if (conv_params.do_post_rounding) {
#if CONFIG_HIGHBITDEPTH
      if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
        av1_highbd_convolve_rounding(
            tmp_dst, MAX_SB_SIZE, dst, dst_buf->stride, w, h,
            FILTER_BITS * 2 + is_compound - conv_params.round_0 -
                conv_params.round_1,
            xd->bd);
      else
#endif  // CONFIG_HIGHBITDEPTH
#if CONFIG_COMPOUND_SINGLEREF
        av1_convolve_rounding(tmp_dst, MAX_SB_SIZE, dst, dst_buf->stride, w, h,
                              FILTER_BITS * 2 + is_comp_mode_pred -
                                  conv_params.round_0 - conv_params.round_1);
#else   // !(CONFIG_COMPOUND_SINGLEREF)
      av1_convolve_rounding(tmp_dst, MAX_SB_SIZE, dst, dst_buf->stride, w, h,
                            FILTER_BITS * 2 + is_compound -
                                conv_params.round_0 - conv_params.round_1);
#endif  // CONFIG_COMPOUND_SINGLEREF
    }
#endif  // CONFIG_CONVOLVE_ROUND
  }
}

static void build_inter_predictors_for_planes(const AV1_COMMON *cm,
                                              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 struct macroblockd_plane *pd = &xd->plane[plane];
    const int bw = pd->width;
    const int bh = pd->height;

    if (!is_chroma_reference(mi_row, mi_col, bsize, pd->subsampling_x,
                             pd->subsampling_y))
      continue;

    build_inter_predictors(cm, xd, plane,
#if CONFIG_MOTION_VAR
                           xd->mi[0], 0,
#endif  // CONFIG_MOTION_VAR
                           0, bw, bh, 0, 0, bw, bh, mi_x, mi_y);
  }
}

void av1_build_inter_predictors_sby(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                    int mi_row, int mi_col, BUFFER_SET *ctx,
                                    BLOCK_SIZE bsize) {
  build_inter_predictors_for_planes(cm, xd, bsize, mi_row, mi_col, 0, 0);
#if CONFIG_INTERINTRA
  if (is_interintra_pred(&xd->mi[0]->mbmi)) {
    BUFFER_SET default_ctx = { { xd->plane[0].dst.buf, NULL, NULL },
                               { xd->plane[0].dst.stride, 0, 0 } };
    if (!ctx) ctx = &default_ctx;
    av1_build_interintra_predictors_sby(cm, xd, xd->plane[0].dst.buf,
                                        xd->plane[0].dst.stride, ctx, bsize);
  }
#else
  (void)ctx;
#endif  // CONFIG_INTERINTRA
}

void av1_build_inter_predictors_sbuv(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                     int mi_row, int mi_col, BUFFER_SET *ctx,
                                     BLOCK_SIZE bsize) {
  build_inter_predictors_for_planes(cm, xd, bsize, mi_row, mi_col, 1,
                                    MAX_MB_PLANE - 1);
#if CONFIG_INTERINTRA
  if (is_interintra_pred(&xd->mi[0]->mbmi)) {
    BUFFER_SET default_ctx = {
      { NULL, xd->plane[1].dst.buf, xd->plane[2].dst.buf },
      { 0, xd->plane[1].dst.stride, xd->plane[2].dst.stride }
    };
    if (!ctx) ctx = &default_ctx;
    av1_build_interintra_predictors_sbuv(
        cm, xd, xd->plane[1].dst.buf, xd->plane[2].dst.buf,
        xd->plane[1].dst.stride, xd->plane[2].dst.stride, ctx, bsize);
  }
#else
  (void)ctx;
#endif  // CONFIG_INTERINTRA
}

void av1_build_inter_predictors_sb(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                   int mi_row, int mi_col, BUFFER_SET *ctx,
                                   BLOCK_SIZE bsize) {
  av1_build_inter_predictors_sby(cm, xd, mi_row, mi_col, ctx, bsize);
  av1_build_inter_predictors_sbuv(cm, xd, mi_row, mi_col, ctx, bsize);
}

void av1_setup_dst_planes(struct macroblockd_plane *planes, BLOCK_SIZE bsize,
                          const YV12_BUFFER_CONFIG *src, int mi_row,
                          int mi_col) {
  const int widths[MAX_MB_PLANE] = { src->y_crop_width, src->uv_crop_width,
                                     src->uv_crop_width };
  const int heights[MAX_MB_PLANE] = { src->y_crop_height, src->uv_crop_height,
                                      src->uv_crop_height };
  const int strides[MAX_MB_PLANE] = { src->y_stride, src->uv_stride,
                                      src->uv_stride };
  int i;

  for (i = 0; i < MAX_MB_PLANE; ++i) {
    struct macroblockd_plane *const pd = &planes[i];
    setup_pred_plane(&pd->dst, bsize, src->buffers[i], widths[i], heights[i],
                     strides[i], mi_row, mi_col, NULL, pd->subsampling_x,
                     pd->subsampling_y);
  }
}

void av1_setup_pre_planes(MACROBLOCKD *xd, int idx,
                          const YV12_BUFFER_CONFIG *src, int mi_row, int mi_col,
                          const struct scale_factors *sf) {
  if (src != NULL) {
    int i;
    uint8_t *const buffers[MAX_MB_PLANE] = { src->y_buffer, src->u_buffer,
                                             src->v_buffer };
    const int widths[MAX_MB_PLANE] = { src->y_crop_width, src->uv_crop_width,
                                       src->uv_crop_width };
    const int heights[MAX_MB_PLANE] = { src->y_crop_height, src->uv_crop_height,
                                        src->uv_crop_height };
    const int strides[MAX_MB_PLANE] = { src->y_stride, src->uv_stride,
                                        src->uv_stride };
    for (i = 0; i < MAX_MB_PLANE; ++i) {
      struct macroblockd_plane *const pd = &xd->plane[i];
      setup_pred_plane(&pd->pre[idx], xd->mi[0]->mbmi.sb_type, buffers[i],
                       widths[i], heights[i], strides[i], mi_row, mi_col, sf,
                       pd->subsampling_x, pd->subsampling_y);
    }
  }
}

#if CONFIG_MOTION_VAR
// obmc_mask_N[overlap_position]
static const uint8_t obmc_mask_1[1] = { 64 };

static const uint8_t obmc_mask_2[2] = { 45, 64 };

static const uint8_t obmc_mask_4[4] = { 39, 50, 59, 64 };

static const uint8_t obmc_mask_8[8] = { 36, 42, 48, 53, 57, 61, 64, 64 };

static const uint8_t obmc_mask_16[16] = { 34, 37, 40, 43, 46, 49, 52, 54,
                                          56, 58, 60, 61, 64, 64, 64, 64 };

static const uint8_t obmc_mask_32[32] = { 33, 35, 36, 38, 40, 41, 43, 44,
                                          45, 47, 48, 50, 51, 52, 53, 55,
                                          56, 57, 58, 59, 60, 60, 61, 62,
                                          64, 64, 64, 64, 64, 64, 64, 64 };

#if CONFIG_EXT_PARTITION
static const uint8_t obmc_mask_64[64] = {
  33, 34, 35, 35, 36, 37, 38, 39, 40, 40, 41, 42, 43, 44, 44, 44,
  45, 46, 47, 47, 48, 49, 50, 51, 51, 51, 52, 52, 53, 54, 55, 56,
  56, 56, 57, 57, 58, 58, 59, 60, 60, 60, 60, 60, 61, 62, 62, 62,
  62, 62, 63, 63, 63, 63, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64,
};
#endif  // CONFIG_EXT_PARTITION

const uint8_t *av1_get_obmc_mask(int length) {
  switch (length) {
    case 1: return obmc_mask_1;
    case 2: return obmc_mask_2;
    case 4: return obmc_mask_4;
    case 8: return obmc_mask_8;
    case 16: return obmc_mask_16;
    case 32: return obmc_mask_32;
#if CONFIG_EXT_PARTITION
    case 64: return obmc_mask_64;
#endif  // CONFIG_EXT_PARTITION
    default: assert(0); return NULL;
  }
}

#if CONFIG_NCOBMC
// obmc_mask_flipN[overlap_position]
static const uint8_t obmc_mask_flip1[1] = { 55 };

static const uint8_t obmc_mask_flip2[2] = { 62, 45 };

static const uint8_t obmc_mask_flip4[4] = { 64, 59, 50, 39 };

static const uint8_t obmc_mask_flip8[8] = { 64, 63, 61, 57, 53, 48, 42, 36 };

static const uint8_t obmc_mask_flip16[16] = { 64, 64, 64, 63, 61, 60, 58, 56,
                                              54, 52, 49, 46, 43, 40, 37, 34 };

static const uint8_t obmc_mask_flip32[32] = { 64, 64, 64, 64, 64, 63, 63, 62,
                                              62, 61, 60, 60, 59, 58, 57, 56,
                                              55, 53, 52, 51, 50, 48, 47, 45,
                                              44, 43, 41, 40, 38, 36, 35, 33 };

#if CONFIG_EXT_PARTITION
static const uint8_t obmc_mask_flip64[64] = {
  64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 63, 63, 63, 63, 62, 62,
  62, 62, 62, 61, 60, 60, 60, 60, 60, 59, 58, 58, 57, 57, 56, 56,
  56, 55, 54, 53, 52, 52, 51, 51, 51, 50, 49, 48, 47, 47, 46, 45,
  44, 44, 44, 43, 42, 41, 40, 40, 39, 38, 37, 36, 35, 35, 34, 33,
};
#endif  // CONFIG_EXT_PARTITION

const uint8_t *av1_get_obmc_mask_flipped(int length) {
  switch (length) {
    case 1: return obmc_mask_flip1;
    case 2: return obmc_mask_flip2;
    case 4: return obmc_mask_flip4;
    case 8: return obmc_mask_flip8;
    case 16: return obmc_mask_flip16;
    case 32: return obmc_mask_flip32;
#if CONFIG_EXT_PARTITION
    case 64: return obmc_mask_flip64;
#endif  // CONFIG_EXT_PARTITION
    default: assert(0); return NULL;
  }
}
#endif  // CONFIG_NCOBMC

static INLINE void increment_int_ptr(MACROBLOCKD *xd, int rel_mi_rc,
                                     uint8_t mi_hw, MODE_INFO *mi,
                                     void *fun_ctxt) {
  (void)xd;
  (void)rel_mi_rc;
  (void)mi_hw;
  (void)mi;
  ++*(int *)fun_ctxt;
}

void av1_count_overlappable_neighbors(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                      int mi_row, int mi_col) {
  MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;

  mbmi->overlappable_neighbors[0] = 0;
  mbmi->overlappable_neighbors[1] = 0;

  if (!is_motion_variation_allowed_bsize(mbmi->sb_type)) return;

  foreach_overlappable_nb_above(cm, xd, mi_col, INT_MAX, increment_int_ptr,
                                &mbmi->overlappable_neighbors[0]);
  foreach_overlappable_nb_left(cm, xd, mi_row, INT_MAX, increment_int_ptr,
                               &mbmi->overlappable_neighbors[1]);
}

// HW does not support < 4x4 prediction. To limit the bandwidth requirement, if
// block-size of current plane is smaller than 8x8, always only blend with the
// left neighbor(s) (skip blending with the above side).
#define DISABLE_CHROMA_U8X8_OBMC 0  // 0: one-sided obmc; 1: disable

int skip_u4x4_pred_in_obmc(BLOCK_SIZE bsize, const struct macroblockd_plane *pd,
                           int dir) {
  assert(is_motion_variation_allowed_bsize(bsize));

  BLOCK_SIZE bsize_plane =
      ss_size_lookup[bsize][pd->subsampling_x][pd->subsampling_y];
  if (bsize_plane < BLOCK_4X4) return 1;
  switch (bsize_plane) {
#if DISABLE_CHROMA_U8X8_OBMC
    case BLOCK_4X4:
    case BLOCK_8X4:
    case BLOCK_4X8: return 1; break;
#else
    case BLOCK_4X4:
    case BLOCK_8X4:
    case BLOCK_4X8: return dir == 0; break;
#endif
    default: return 0;
  }
}

struct obmc_inter_pred_ctxt {
  uint8_t **adjacent;
  int *adjacent_stride;
};

static INLINE void build_obmc_inter_pred_above(MACROBLOCKD *xd, int rel_mi_col,
                                               uint8_t above_mi_width,
                                               MODE_INFO *above_mi,
                                               void *fun_ctxt) {
  (void)above_mi;
  struct obmc_inter_pred_ctxt *ctxt = (struct obmc_inter_pred_ctxt *)fun_ctxt;
  const BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
#if CONFIG_HIGHBITDEPTH
  const int is_hbd = (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) ? 1 : 0;
#endif  // CONFIG_HIGHBITDEPTH
  const int overlap =
      AOMMIN(block_size_high[bsize], block_size_high[BLOCK_64X64]) >> 1;

  for (int plane = 0; plane < MAX_MB_PLANE; ++plane) {
    const struct macroblockd_plane *pd = &xd->plane[plane];
    const int bw = (above_mi_width * MI_SIZE) >> pd->subsampling_x;
    const int bh = overlap >> pd->subsampling_y;
    const int plane_col = (rel_mi_col * MI_SIZE) >> pd->subsampling_x;

    if (skip_u4x4_pred_in_obmc(bsize, pd, 0)) continue;

    const int dst_stride = pd->dst.stride;
    uint8_t *const dst = &pd->dst.buf[plane_col];
    const int tmp_stride = ctxt->adjacent_stride[plane];
    const uint8_t *const tmp = &ctxt->adjacent[plane][plane_col];
    const uint8_t *const mask = av1_get_obmc_mask(bh);

#if CONFIG_HIGHBITDEPTH
    if (is_hbd)
      aom_highbd_blend_a64_vmask(dst, dst_stride, dst, dst_stride, tmp,
                                 tmp_stride, mask, bh, bw, xd->bd);
    else
#endif  // CONFIG_HIGHBITDEPTH
      aom_blend_a64_vmask(dst, dst_stride, dst, dst_stride, tmp, tmp_stride,
                          mask, bh, bw);
  }
}

static INLINE void build_obmc_inter_pred_left(MACROBLOCKD *xd, int rel_mi_row,
                                              uint8_t left_mi_height,
                                              MODE_INFO *left_mi,
                                              void *fun_ctxt) {
  (void)left_mi;
  struct obmc_inter_pred_ctxt *ctxt = (struct obmc_inter_pred_ctxt *)fun_ctxt;
  const BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
  const int overlap =
      AOMMIN(block_size_wide[bsize], block_size_wide[BLOCK_64X64]) >> 1;
#if CONFIG_HIGHBITDEPTH
  const int is_hbd = (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) ? 1 : 0;
#endif  // CONFIG_HIGHBITDEPTH

  for (int plane = 0; plane < MAX_MB_PLANE; ++plane) {
    const struct macroblockd_plane *pd = &xd->plane[plane];
    const int bw = overlap >> pd->subsampling_x;
    const int bh = (left_mi_height * MI_SIZE) >> pd->subsampling_y;
    const int plane_row = (rel_mi_row * MI_SIZE) >> pd->subsampling_y;

    if (skip_u4x4_pred_in_obmc(bsize, pd, 1)) continue;

    const int dst_stride = pd->dst.stride;
    uint8_t *const dst = &pd->dst.buf[plane_row * dst_stride];
    const int tmp_stride = ctxt->adjacent_stride[plane];
    const uint8_t *const tmp = &ctxt->adjacent[plane][plane_row * tmp_stride];
    const uint8_t *const mask = av1_get_obmc_mask(bw);

#if CONFIG_HIGHBITDEPTH
    if (is_hbd)
      aom_highbd_blend_a64_hmask(dst, dst_stride, dst, dst_stride, tmp,
                                 tmp_stride, mask, bh, bw, xd->bd);
    else
#endif  // CONFIG_HIGHBITDEPTH
      aom_blend_a64_hmask(dst, dst_stride, dst, dst_stride, tmp, tmp_stride,
                          mask, bh, bw);
  }
}

// This function combines motion compensated predictions that are generated by
// top/left neighboring blocks' inter predictors with the regular inter
// prediction. We assume the original prediction (bmc) is stored in
// xd->plane[].dst.buf
void av1_build_obmc_inter_prediction(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                     int mi_row, int mi_col,
                                     uint8_t *above[MAX_MB_PLANE],
                                     int above_stride[MAX_MB_PLANE],
                                     uint8_t *left[MAX_MB_PLANE],
                                     int left_stride[MAX_MB_PLANE]) {
  const BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;

  // handle above row
  struct obmc_inter_pred_ctxt ctxt_above = { above, above_stride };
  foreach_overlappable_nb_above(cm, xd, mi_col,
                                max_neighbor_obmc[b_width_log2_lookup[bsize]],
                                build_obmc_inter_pred_above, &ctxt_above);

  // handle left column
  struct obmc_inter_pred_ctxt ctxt_left = { left, left_stride };
  foreach_overlappable_nb_left(cm, xd, mi_row,
                               max_neighbor_obmc[b_height_log2_lookup[bsize]],
                               build_obmc_inter_pred_left, &ctxt_left);
}

void modify_neighbor_predictor_for_obmc(MB_MODE_INFO *mbmi) {
  if (is_interintra_pred(mbmi)) {
    mbmi->ref_frame[1] = NONE_FRAME;
  } else if (has_second_ref(mbmi) &&
             is_masked_compound_type(mbmi->interinter_compound_type)) {
    mbmi->interinter_compound_type = COMPOUND_AVERAGE;
    mbmi->ref_frame[1] = NONE_FRAME;
#if CONFIG_COMPOUND_SINGLEREF
  } else if (!has_second_ref(mbmi) &&
             is_inter_singleref_comp_mode(mbmi->mode)) {
    // mbmi->mode = compound_ref0_mode(mbmi->mode);
    mbmi->mode = compound_ref1_mode(mbmi->mode);
    assert(is_inter_singleref_mode(mbmi->mode));
    mbmi->mv[0].as_int = mbmi->mv[1].as_int;
#endif  // CONFIG_COMPOUND_SINGLEREF
  }
  if (has_second_ref(mbmi)) mbmi->ref_frame[1] = NONE_FRAME;
  return;
}

struct build_prediction_ctxt {
  const AV1_COMMON *cm;
  int mi_row;
  int mi_col;
  uint8_t **tmp_buf;
  int *tmp_width;
  int *tmp_height;
  int *tmp_stride;
  int mb_to_far_edge;
};

static INLINE void build_prediction_by_above_pred(MACROBLOCKD *xd,
                                                  int rel_mi_col,
                                                  uint8_t above_mi_width,
                                                  MODE_INFO *above_mi,
                                                  void *fun_ctxt) {
  MB_MODE_INFO *above_mbmi = &above_mi->mbmi;
  const BLOCK_SIZE a_bsize = AOMMAX(BLOCK_8X8, above_mbmi->sb_type);
  struct build_prediction_ctxt *ctxt = (struct build_prediction_ctxt *)fun_ctxt;
  const int above_mi_col = ctxt->mi_col + rel_mi_col;

  MB_MODE_INFO backup_mbmi = *above_mbmi;
  modify_neighbor_predictor_for_obmc(above_mbmi);

  for (int j = 0; j < MAX_MB_PLANE; ++j) {
    struct macroblockd_plane *const pd = &xd->plane[j];
    setup_pred_plane(&pd->dst, a_bsize, ctxt->tmp_buf[j], ctxt->tmp_width[j],
                     ctxt->tmp_height[j], ctxt->tmp_stride[j], 0, rel_mi_col,
                     NULL, pd->subsampling_x, pd->subsampling_y);
  }

#if CONFIG_COMPOUND_SINGLEREF
  const int num_refs = 1 + is_inter_anyref_comp_mode(above_mbmi->mode);
#else
  const int num_refs = 1 + has_second_ref(above_mbmi);
#endif

  for (int ref = 0; ref < num_refs; ++ref) {
#if CONFIG_COMPOUND_SINGLEREF
    const MV_REFERENCE_FRAME frame = has_second_ref(above_mbmi)
                                         ? above_mbmi->ref_frame[ref]
                                         : above_mbmi->ref_frame[0];
#else
    const MV_REFERENCE_FRAME frame = above_mbmi->ref_frame[ref];
#endif  // CONFIG_COMPOUND_SINGLEREF

    const RefBuffer *const ref_buf = &ctxt->cm->frame_refs[frame - LAST_FRAME];

    xd->block_refs[ref] = ref_buf;
    if ((!av1_is_valid_scale(&ref_buf->sf)))
      aom_internal_error(xd->error_info, AOM_CODEC_UNSUP_BITSTREAM,
                         "Reference frame has invalid dimensions");
    av1_setup_pre_planes(xd, ref, ref_buf->buf, ctxt->mi_row, above_mi_col,
                         &ref_buf->sf);
  }

  xd->mb_to_left_edge = 8 * MI_SIZE * (-above_mi_col);
  xd->mb_to_right_edge = ctxt->mb_to_far_edge +
                         (xd->n8_w - rel_mi_col - above_mi_width) * MI_SIZE * 8;

  int mi_x = above_mi_col << MI_SIZE_LOG2;
  int mi_y = ctxt->mi_row << MI_SIZE_LOG2;

  const BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;

  for (int j = 0; j < MAX_MB_PLANE; ++j) {
    const struct macroblockd_plane *pd = &xd->plane[j];
    int bw = (above_mi_width * MI_SIZE) >> pd->subsampling_x;
    int bh = clamp(block_size_high[bsize] >> (pd->subsampling_y + 1), 4,
                   block_size_high[BLOCK_64X64] >> (pd->subsampling_y + 1));

    if (skip_u4x4_pred_in_obmc(bsize, pd, 0)) continue;
    build_inter_predictors(ctxt->cm, xd, j, above_mi, 1, 0, bw, bh, 0, 0, bw,
                           bh, mi_x, mi_y);
  }
  *above_mbmi = backup_mbmi;
}

void av1_build_prediction_by_above_preds(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                         int mi_row, int mi_col,
                                         uint8_t *tmp_buf[MAX_MB_PLANE],
                                         int tmp_width[MAX_MB_PLANE],
                                         int tmp_height[MAX_MB_PLANE],
                                         int tmp_stride[MAX_MB_PLANE]) {
  if (!xd->up_available) return;

  // Adjust mb_to_bottom_edge to have the correct value for the OBMC
  // prediction block. This is half the height of the original block,
  // except for 128-wide blocks, where we only use a height of 32.
  int this_height = xd->n8_h * MI_SIZE;
  int pred_height = AOMMIN(this_height / 2, 32);
  xd->mb_to_bottom_edge += (this_height - pred_height) * 8;

  struct build_prediction_ctxt ctxt = { cm,         mi_row,
                                        mi_col,     tmp_buf,
                                        tmp_width,  tmp_height,
                                        tmp_stride, xd->mb_to_right_edge };
  BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
  foreach_overlappable_nb_above(cm, xd, mi_col,
                                max_neighbor_obmc[b_width_log2_lookup[bsize]],
                                build_prediction_by_above_pred, &ctxt);

  xd->mb_to_left_edge = -((mi_col * MI_SIZE) * 8);
  xd->mb_to_right_edge = ctxt.mb_to_far_edge;
  xd->mb_to_bottom_edge -= (this_height - pred_height) * 8;
}

static INLINE void build_prediction_by_left_pred(MACROBLOCKD *xd,
                                                 int rel_mi_row,
                                                 uint8_t left_mi_height,
                                                 MODE_INFO *left_mi,
                                                 void *fun_ctxt) {
  MB_MODE_INFO *left_mbmi = &left_mi->mbmi;
  const BLOCK_SIZE l_bsize = AOMMAX(BLOCK_8X8, left_mbmi->sb_type);
  struct build_prediction_ctxt *ctxt = (struct build_prediction_ctxt *)fun_ctxt;
  const int left_mi_row = ctxt->mi_row + rel_mi_row;

  MB_MODE_INFO backup_mbmi = *left_mbmi;
  modify_neighbor_predictor_for_obmc(left_mbmi);

  for (int j = 0; j < MAX_MB_PLANE; ++j) {
    struct macroblockd_plane *const pd = &xd->plane[j];
    setup_pred_plane(&pd->dst, l_bsize, ctxt->tmp_buf[j], ctxt->tmp_width[j],
                     ctxt->tmp_height[j], ctxt->tmp_stride[j], rel_mi_row, 0,
                     NULL, pd->subsampling_x, pd->subsampling_y);
  }

#if CONFIG_COMPOUND_SINGLEREF
  const int num_refs = 1 + is_inter_anyref_comp_mode(left_mbmi->mode);
#else
  const int num_refs = 1 + has_second_ref(left_mbmi);
#endif

  for (int ref = 0; ref < num_refs; ++ref) {
#if CONFIG_COMPOUND_SINGLEREF
    const MV_REFERENCE_FRAME frame = has_second_ref(left_mbmi)
                                         ? left_mbmi->ref_frame[ref]
                                         : left_mbmi->ref_frame[0];
#else
    const MV_REFERENCE_FRAME frame = left_mbmi->ref_frame[ref];
#endif  // CONFIG_COMPOUND_SINGLEREF

    const RefBuffer *const ref_buf = &ctxt->cm->frame_refs[frame - LAST_FRAME];

    xd->block_refs[ref] = ref_buf;
    if ((!av1_is_valid_scale(&ref_buf->sf)))
      aom_internal_error(xd->error_info, AOM_CODEC_UNSUP_BITSTREAM,
                         "Reference frame has invalid dimensions");
    av1_setup_pre_planes(xd, ref, ref_buf->buf, left_mi_row, ctxt->mi_col,
                         &ref_buf->sf);
  }

  xd->mb_to_top_edge = 8 * MI_SIZE * (-left_mi_row);
  xd->mb_to_bottom_edge =
      ctxt->mb_to_far_edge +
      (xd->n8_h - rel_mi_row - left_mi_height) * MI_SIZE * 8;

  int mi_x = ctxt->mi_col << MI_SIZE_LOG2;
  int mi_y = left_mi_row << MI_SIZE_LOG2;

  const BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;

  for (int j = 0; j < MAX_MB_PLANE; ++j) {
    const struct macroblockd_plane *pd = &xd->plane[j];
    int bw = clamp(block_size_wide[bsize] >> (pd->subsampling_x + 1), 4,
                   block_size_wide[BLOCK_64X64] >> (pd->subsampling_x + 1));
    int bh = (left_mi_height << MI_SIZE_LOG2) >> pd->subsampling_y;

    if (skip_u4x4_pred_in_obmc(bsize, pd, 1)) continue;
    build_inter_predictors(ctxt->cm, xd, j, left_mi, 1, 0, bw, bh, 0, 0, bw, bh,
                           mi_x, mi_y);
  }
  *left_mbmi = backup_mbmi;
}

void av1_build_prediction_by_left_preds(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                        int mi_row, int mi_col,
                                        uint8_t *tmp_buf[MAX_MB_PLANE],
                                        int tmp_width[MAX_MB_PLANE],
                                        int tmp_height[MAX_MB_PLANE],
                                        int tmp_stride[MAX_MB_PLANE]) {
  if (!xd->left_available) return;

  // Adjust mb_to_right_edge to have the correct value for the OBMC
  // prediction block. This is half the width of the original block,
  // except for 128-wide blocks, where we only use a width of 32.
  int this_width = xd->n8_w * MI_SIZE;
  int pred_width = AOMMIN(this_width / 2, 32);
  xd->mb_to_right_edge += (this_width - pred_width) * 8;

  struct build_prediction_ctxt ctxt = { cm,         mi_row,
                                        mi_col,     tmp_buf,
                                        tmp_width,  tmp_height,
                                        tmp_stride, xd->mb_to_bottom_edge };
  BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
  foreach_overlappable_nb_left(cm, xd, mi_row,
                               max_neighbor_obmc[b_height_log2_lookup[bsize]],
                               build_prediction_by_left_pred, &ctxt);

  xd->mb_to_top_edge = -((mi_row * MI_SIZE) * 8);
  xd->mb_to_right_edge -= (this_width - pred_width) * 8;
  xd->mb_to_bottom_edge = ctxt.mb_to_far_edge;
}

void av1_build_obmc_inter_predictors_sb(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                        int mi_row, int mi_col) {
#if CONFIG_HIGHBITDEPTH
  DECLARE_ALIGNED(16, uint8_t, tmp_buf1[2 * MAX_MB_PLANE * MAX_SB_SQUARE]);
  DECLARE_ALIGNED(16, uint8_t, tmp_buf2[2 * MAX_MB_PLANE * MAX_SB_SQUARE]);
#else
  DECLARE_ALIGNED(16, uint8_t, tmp_buf1[MAX_MB_PLANE * MAX_SB_SQUARE]);
  DECLARE_ALIGNED(16, uint8_t, tmp_buf2[MAX_MB_PLANE * MAX_SB_SQUARE]);
#endif  // CONFIG_HIGHBITDEPTH
  uint8_t *dst_buf1[MAX_MB_PLANE], *dst_buf2[MAX_MB_PLANE];
  int dst_stride1[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };
  int dst_stride2[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };
  int dst_width1[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };
  int dst_width2[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };
  int dst_height1[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };
  int dst_height2[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };

#if CONFIG_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    int len = sizeof(uint16_t);
    dst_buf1[0] = CONVERT_TO_BYTEPTR(tmp_buf1);
    dst_buf1[1] = CONVERT_TO_BYTEPTR(tmp_buf1 + MAX_SB_SQUARE * len);
    dst_buf1[2] = CONVERT_TO_BYTEPTR(tmp_buf1 + MAX_SB_SQUARE * 2 * len);
    dst_buf2[0] = CONVERT_TO_BYTEPTR(tmp_buf2);
    dst_buf2[1] = CONVERT_TO_BYTEPTR(tmp_buf2 + MAX_SB_SQUARE * len);
    dst_buf2[2] = CONVERT_TO_BYTEPTR(tmp_buf2 + MAX_SB_SQUARE * 2 * len);
  } else {
#endif  // CONFIG_HIGHBITDEPTH
    dst_buf1[0] = tmp_buf1;
    dst_buf1[1] = tmp_buf1 + MAX_SB_SQUARE;
    dst_buf1[2] = tmp_buf1 + MAX_SB_SQUARE * 2;
    dst_buf2[0] = tmp_buf2;
    dst_buf2[1] = tmp_buf2 + MAX_SB_SQUARE;
    dst_buf2[2] = tmp_buf2 + MAX_SB_SQUARE * 2;
#if CONFIG_HIGHBITDEPTH
  }
#endif  // CONFIG_HIGHBITDEPTH
  av1_build_prediction_by_above_preds(cm, xd, mi_row, mi_col, dst_buf1,
                                      dst_width1, dst_height1, dst_stride1);
  av1_build_prediction_by_left_preds(cm, xd, mi_row, mi_col, dst_buf2,
                                     dst_width2, dst_height2, dst_stride2);
  av1_setup_dst_planes(xd->plane, xd->mi[0]->mbmi.sb_type,
                       get_frame_new_buffer(cm), mi_row, mi_col);
  av1_build_obmc_inter_prediction(cm, xd, mi_row, mi_col, dst_buf1, dst_stride1,
                                  dst_buf2, dst_stride2);
}

#if CONFIG_NCOBMC
void av1_build_prediction_by_bottom_preds(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                          int mi_row, int mi_col,
                                          uint8_t *tmp_buf[MAX_MB_PLANE],
                                          int tmp_width[MAX_MB_PLANE],
                                          int tmp_height[MAX_MB_PLANE],
                                          int tmp_stride[MAX_MB_PLANE]) {
  const TileInfo *const tile = &xd->tile;
#if CONFIG_DEBUG
  BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
#endif
  int i, j, mi_step, ref;
  const int ilimit = AOMMIN(xd->n8_w, cm->mi_cols - mi_col);
  int mb_to_right_edge_base = xd->mb_to_right_edge;

  if (mi_row + xd->n8_h >= tile->mi_row_end ||
      (mi_row + xd->n8_h) % MI_SIZE == 0 || (mi_row + xd->n8_h) >= cm->mi_rows)
    return;
  assert(bsize >= BLOCK_8X8);

  xd->mb_to_top_edge -= xd->n8_h * 32;
  for (i = 0; i < ilimit; i += mi_step) {
    int mi_row_offset = xd->n8_h;
    int mi_col_offset = i;
    int mi_x, mi_y, bw, bh;
    MODE_INFO *mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
    MB_MODE_INFO *mbmi = &mi->mbmi;
    MB_MODE_INFO backup_mbmi;

    mi_step = AOMMIN(xd->n8_w, mi_size_wide[mbmi->sb_type]);

    if (!is_neighbor_overlappable(mbmi)) continue;

    backup_mbmi = *mbmi;
    modify_neighbor_predictor_for_obmc(mbmi);

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      struct macroblockd_plane *const pd = &xd->plane[j];
      setup_pred_plane(&pd->dst, AOMMAX(mbmi->sb_type, BLOCK_8X8), tmp_buf[j],
                       tmp_width[j], tmp_height[j], tmp_stride[j],
                       (xd->n8_h >> 1), i, NULL, pd->subsampling_x,
                       pd->subsampling_y);
    }
    for (ref = 0; ref < 1 + has_second_ref(mbmi); ++ref) {
      const MV_REFERENCE_FRAME frame = mbmi->ref_frame[ref];
      const RefBuffer *const ref_buf = &cm->frame_refs[frame - LAST_FRAME];

      xd->block_refs[ref] = ref_buf;
      if ((!av1_is_valid_scale(&ref_buf->sf)))
        aom_internal_error(xd->error_info, AOM_CODEC_UNSUP_BITSTREAM,
                           "Reference frame has invalid dimensions");
      av1_setup_pre_planes(xd, ref, ref_buf->buf, mi_row + (xd->n8_h >> 1),
                           mi_col + i, &ref_buf->sf);
    }

    xd->mb_to_left_edge = -(((mi_col + i) * MI_SIZE) * 8);
    xd->mb_to_right_edge =
        mb_to_right_edge_base + (xd->n8_w - i - mi_step) * 64;
    mi_x = (mi_col + i) << MI_SIZE_LOG2;
    mi_y = (mi_row << MI_SIZE_LOG2) + xd->n8_h * (MI_SIZE >> 1);

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      const struct macroblockd_plane *pd = &xd->plane[j];
      bw = (mi_step << MI_SIZE_LOG2) >> pd->subsampling_x;
      bh = (xd->n8_h << (MI_SIZE_LOG2 - 1)) >> pd->subsampling_y;
      build_inter_predictors(cm, xd, j, mi, 1, 0, bw, bh, 0,
                             xd->n8_h == 1 ? (4 >> pd->subsampling_y) : 0, bw,
                             bh, mi_x, mi_y);
    }
    *mbmi = backup_mbmi;
  }
  xd->mb_to_left_edge = -((mi_col * MI_SIZE) * 8);
  xd->mb_to_right_edge = mb_to_right_edge_base;
  xd->mb_to_top_edge += xd->n8_h * 32;
}

void av1_build_prediction_by_right_preds(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                         int mi_row, int mi_col,
                                         uint8_t *tmp_buf[MAX_MB_PLANE],
                                         int tmp_width[MAX_MB_PLANE],
                                         int tmp_height[MAX_MB_PLANE],
                                         const int tmp_stride[MAX_MB_PLANE]) {
  const TileInfo *const tile = &xd->tile;
#if CONFIG_DEBUG
  BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
#endif
  int i, j, mi_step, ref;
  const int ilimit = AOMMIN(xd->n8_h, cm->mi_rows - mi_row);
  int mb_to_bottom_edge_base = xd->mb_to_bottom_edge;

  if (mi_col + xd->n8_w >= tile->mi_col_end ||
      (mi_col + xd->n8_w) % MI_SIZE == 0 || (mi_col + xd->n8_w) >= cm->mi_cols)
    return;

  assert(bsize >= BLOCK_8X8);

  xd->mb_to_left_edge -= xd->n8_w / 2 * MI_SIZE * 8;
  for (i = 0; i < ilimit; i += mi_step) {
    int mi_row_offset = i;
    int mi_col_offset = xd->n8_w;
    int mi_x, mi_y, bw, bh;
    MODE_INFO *mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
    MB_MODE_INFO *mbmi = &mi->mbmi;
    MB_MODE_INFO backup_mbmi;

    mi_step = AOMMIN(xd->n8_h, mi_size_high[mbmi->sb_type]);

    if (!is_neighbor_overlappable(mbmi)) continue;

    backup_mbmi = *mbmi;
    modify_neighbor_predictor_for_obmc(mbmi);

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      struct macroblockd_plane *const pd = &xd->plane[j];
      setup_pred_plane(&pd->dst, AOMMAX(mbmi->sb_type, BLOCK_8X8), tmp_buf[j],
                       tmp_width[j], tmp_height[j], tmp_stride[j], i,
                       xd->n8_w >> 1, NULL, pd->subsampling_x,
                       pd->subsampling_y);
    }
    for (ref = 0; ref < 1 + has_second_ref(mbmi); ++ref) {
      const MV_REFERENCE_FRAME frame = mbmi->ref_frame[ref];
      const RefBuffer *const ref_buf = &cm->frame_refs[frame - LAST_FRAME];

      xd->block_refs[ref] = ref_buf;
      if ((!av1_is_valid_scale(&ref_buf->sf)))
        aom_internal_error(xd->error_info, AOM_CODEC_UNSUP_BITSTREAM,
                           "Reference frame has invalid dimensions");
      av1_setup_pre_planes(xd, ref, ref_buf->buf, mi_row + i,
                           mi_col + (xd->n8_w >> 1), &ref_buf->sf);
    }

    xd->mb_to_top_edge = -(((mi_row + i) * MI_SIZE) * 8);
    xd->mb_to_bottom_edge =
        mb_to_bottom_edge_base + (xd->n8_h - i - mi_step) * MI_SIZE * 8;
    mi_x = (mi_col << MI_SIZE_LOG2) + xd->n8_w * (MI_SIZE >> 1);
    mi_y = (mi_row + i) << MI_SIZE_LOG2;

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      const struct macroblockd_plane *pd = &xd->plane[j];
      bw = (xd->n8_w << (MI_SIZE_LOG2 - 1)) >> pd->subsampling_x;
      bh = (mi_step << MI_SIZE_LOG2) >> pd->subsampling_y;
      build_inter_predictors(cm, xd, j, mi, 1, 0, bw, bh,
                             xd->n8_w == 1 ? 4 >> pd->subsampling_x : 0, 0, bw,
                             bh, mi_x, mi_y);
    }
    *mbmi = backup_mbmi;
  }
  xd->mb_to_top_edge = -((mi_row * MI_SIZE) * 8);
  xd->mb_to_bottom_edge = mb_to_bottom_edge_base;
  xd->mb_to_left_edge += xd->n8_w / 2 * MI_SIZE * 8;
}

// This function combines motion compensated predictions that is generated by
// bottom/right neighboring blocks' inter predictors with prediction in dst
// buffer.
void av1_merge_dst_bottom_right_preds(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                      int mi_row, int mi_col,
                                      uint8_t *bottom[MAX_MB_PLANE],
                                      const int bottom_stride[MAX_MB_PLANE],
                                      uint8_t *right[MAX_MB_PLANE],
                                      const int right_stride[MAX_MB_PLANE]) {
  const TileInfo *const tile = &xd->tile;
  BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
  int plane, i, mi_step;
  const int bottom_available = mi_row + xd->n8_h < tile->mi_row_end &&
                               (mi_row + xd->n8_h) % MI_SIZE != 0 &&
                               (mi_row + xd->n8_h) < cm->mi_rows;
#if CONFIG_HIGHBITDEPTH
  int is_hbd = (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) ? 1 : 0;
#endif  // CONFIG_HIGHBITDEPTH

  // handle bottom row
  for (i = 0; bottom_available && i < AOMMIN(xd->n8_w, cm->mi_cols - mi_col);
       i += mi_step) {
    int mi_row_offset = xd->n8_h;
    int mi_col_offset = i;
    MODE_INFO *mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
    MB_MODE_INFO *mbmi = &mi->mbmi;
    int overlap;

    mi_step = AOMMIN(xd->n8_w, mi_size_wide[mbmi->sb_type]);

    if (!is_neighbor_overlappable(mbmi)) continue;

    overlap = num_4x4_blocks_high_lookup[bsize] << 1;

    for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
      const struct macroblockd_plane *pd = &xd->plane[plane];
      const int bw = (mi_step * MI_SIZE) >> pd->subsampling_x;
      const int bh = overlap >> pd->subsampling_y;
      const int dst_stride = pd->dst.stride;
      uint8_t *dst =
          &pd->dst.buf[((i * MI_SIZE) >> pd->subsampling_x) +
                       (((xd->n8_h * MI_SIZE - overlap) * dst_stride) >>
                        pd->subsampling_y)];
      const int tmp_stride = bottom_stride[plane];
      const uint8_t *const tmp =
          &bottom[plane][((i * MI_SIZE) >> pd->subsampling_x) +
                         (((xd->n8_h * MI_SIZE - overlap) * tmp_stride) >>
                          pd->subsampling_y)];
      const uint8_t *const mask = av1_get_obmc_mask_flipped(bh);

#if CONFIG_HIGHBITDEPTH
      if (is_hbd)
        aom_highbd_blend_a64_vmask(dst, dst_stride, dst, dst_stride, tmp,
                                   tmp_stride, mask, bh, bw, xd->bd);
      else
#endif  // CONFIG_HIGHBITDEPTH
        aom_blend_a64_vmask(dst, dst_stride, dst, dst_stride, tmp, tmp_stride,
                            mask, bh, bw);
    }
  }  // each mi in the bottom row

  // handle right column
  if (mi_col + xd->n8_w >= tile->mi_col_end ||
      (mi_col + xd->n8_w) % MI_SIZE == 0 || (mi_col + xd->n8_w) >= cm->mi_cols)
    return;

  for (i = 0; i < AOMMIN(xd->n8_h, cm->mi_rows - mi_row); i += mi_step) {
    int mi_row_offset = i;
    int mi_col_offset = xd->n8_w;
    int overlap;
    MODE_INFO *mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
    MB_MODE_INFO *mbmi = &mi->mbmi;

    mi_step = AOMMIN(xd->n8_h, mi_size_high[mbmi->sb_type]);

    if (!is_neighbor_overlappable(mbmi)) continue;

    overlap = num_4x4_blocks_wide_lookup[bsize] << 1;

    for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
      const struct macroblockd_plane *pd = &xd->plane[plane];
      const int bw = overlap >> pd->subsampling_x;
      const int bh = (mi_step * MI_SIZE) >> pd->subsampling_y;
      const int dst_stride = pd->dst.stride;
      uint8_t *dst =
          &pd->dst.buf[((i * MI_SIZE * dst_stride) >> pd->subsampling_y) +
                       ((xd->n8_w * MI_SIZE - overlap) >> pd->subsampling_x)];
      const int tmp_stride = right_stride[plane];
      const uint8_t *const tmp =
          &right[plane][((i * MI_SIZE * tmp_stride) >> pd->subsampling_y) +
                        ((xd->n8_w * MI_SIZE - overlap) >> pd->subsampling_x)];
      const uint8_t *const mask = av1_get_obmc_mask_flipped(bw);

#if CONFIG_HIGHBITDEPTH
      if (is_hbd)
        aom_highbd_blend_a64_hmask(dst, dst_stride, dst, dst_stride, tmp,
                                   tmp_stride, mask, bh, bw, xd->bd);
      else
#endif  // CONFIG_HIGHBITDEPTH
        aom_blend_a64_hmask(dst, dst_stride, dst, dst_stride, tmp, tmp_stride,
                            mask, bh, bw);
    }
  }  // each mi in the right column
}

// This function generates 4 sided obmc. (1) Prediction blocks generated by
// bottom and right motion vectors are calculated. (2) Combine them with the
// original prediction block (which should be pre-stored in xd->plane[].dst.buf
// before calling this function). The results is updated in xd->plane[].dst.buf
// (3) Call causal obmc prediction function, which will generate left and above
// preds, and then merge them and xd->plane[].dst.buf.
void av1_build_ncobmc_inter_predictors_sb(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                          int mi_row, int mi_col) {
#if CONFIG_HIGHBITDEPTH
  DECLARE_ALIGNED(16, uint8_t, tmp_buf1[2 * MAX_MB_PLANE * MAX_SB_SQUARE]);
  DECLARE_ALIGNED(16, uint8_t, tmp_buf2[2 * MAX_MB_PLANE * MAX_SB_SQUARE]);
#else
  DECLARE_ALIGNED(16, uint8_t, tmp_buf1[MAX_MB_PLANE * MAX_SB_SQUARE]);
  DECLARE_ALIGNED(16, uint8_t, tmp_buf2[MAX_MB_PLANE * MAX_SB_SQUARE]);
#endif  // CONFIG_HIGHBITDEPTH
  uint8_t *dst_buf1[MAX_MB_PLANE], *dst_buf2[MAX_MB_PLANE];
  int dst_stride1[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };
  int dst_stride2[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };
  int dst_width1[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };
  int dst_width2[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };
  int dst_height1[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };
  int dst_height2[MAX_MB_PLANE] = { MAX_SB_SIZE, MAX_SB_SIZE, MAX_SB_SIZE };

#if CONFIG_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    int len = sizeof(uint16_t);
    dst_buf1[0] = CONVERT_TO_BYTEPTR(tmp_buf1);
    dst_buf1[1] = CONVERT_TO_BYTEPTR(tmp_buf1 + MAX_SB_SQUARE * len);
    dst_buf1[2] = CONVERT_TO_BYTEPTR(tmp_buf1 + MAX_SB_SQUARE * 2 * len);
    dst_buf2[0] = CONVERT_TO_BYTEPTR(tmp_buf2);
    dst_buf2[1] = CONVERT_TO_BYTEPTR(tmp_buf2 + MAX_SB_SQUARE * len);
    dst_buf2[2] = CONVERT_TO_BYTEPTR(tmp_buf2 + MAX_SB_SQUARE * 2 * len);
  } else {
#endif  // CONFIG_HIGHBITDEPTH
    dst_buf1[0] = tmp_buf1;
    dst_buf1[1] = tmp_buf1 + MAX_SB_SQUARE;
    dst_buf1[2] = tmp_buf1 + MAX_SB_SQUARE * 2;
    dst_buf2[0] = tmp_buf2;
    dst_buf2[1] = tmp_buf2 + MAX_SB_SQUARE;
    dst_buf2[2] = tmp_buf2 + MAX_SB_SQUARE * 2;
#if CONFIG_HIGHBITDEPTH
  }
#endif  // CONFIG_HIGHBITDEPTH

  const BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
  // TODO(zoeliu): COMPOUND_SINGLEREF has not worked with NCOBMC yet.
  av1_build_prediction_by_bottom_preds(cm, xd, mi_row, mi_col, dst_buf1,
                                       dst_width1, dst_height1, dst_stride1);
  av1_build_prediction_by_right_preds(cm, xd, mi_row, mi_col, dst_buf2,
                                      dst_width2, dst_height2, dst_stride2);
  av1_setup_dst_planes(xd->plane, bsize, get_frame_new_buffer(cm), mi_row,
                       mi_col);
  av1_merge_dst_bottom_right_preds(cm, xd, mi_row, mi_col, dst_buf1,
                                   dst_stride1, dst_buf2, dst_stride2);
  av1_setup_dst_planes(xd->plane, bsize, get_frame_new_buffer(cm), mi_row,
                       mi_col);
  av1_build_obmc_inter_predictors_sb(cm, xd, mi_row, mi_col);
  av1_setup_dst_planes(xd->plane, bsize, get_frame_new_buffer(cm), mi_row,
                       mi_col);
}
#endif  // CONFIG_NCOBMC

#if CONFIG_NCOBMC_ADAPT_WEIGHT
void reset_xd_boundary(MACROBLOCKD *xd, int mi_row, int bh, int mi_col, int bw,
                       int mi_rows, int mi_cols) {
  xd->mb_to_top_edge = -((mi_row * MI_SIZE) * 8);
  xd->mb_to_bottom_edge = ((mi_rows - bh - mi_row) * MI_SIZE) * 8;
  xd->mb_to_left_edge = -((mi_col * MI_SIZE) * 8);
  xd->mb_to_right_edge = ((mi_cols - bw - mi_col) * MI_SIZE) * 8;
}
void set_sb_mi_boundaries(const AV1_COMMON *const cm, MACROBLOCKD *const xd,
                          const int mi_row, const int mi_col) {
  const BLOCK_SIZE sb = cm->sb_size;
  const int num_mi_w = mi_size_wide[sb];
  const int num_mi_h = mi_size_high[sb];

  xd->sb_mi_bd.mi_col_begin = mi_col;
  xd->sb_mi_bd.mi_row_begin = mi_row;
  // points to the last mi
  xd->sb_mi_bd.mi_col_end =
      mi_col + num_mi_w > cm->mi_cols ? cm->mi_cols - 1 : mi_col + num_mi_w - 1;
  xd->sb_mi_bd.mi_row_end =
      mi_row + num_mi_h > cm->mi_rows ? cm->mi_rows - 1 : mi_row + num_mi_h - 1;
}
#endif

#endif  // CONFIG_MOTION_VAR

/* clang-format off */
#if CONFIG_INTERINTRA
#if CONFIG_EXT_PARTITION
static const int ii_weights1d[MAX_SB_SIZE] = {
  60, 58, 56, 54, 52, 50, 48, 47, 45, 44, 42, 41, 39, 38, 37, 35, 34, 33, 32,
  31, 30, 29, 28, 27, 26, 25, 24, 23, 22, 22, 21, 20, 19, 19, 18, 18, 17, 16,
  16, 15, 15, 14, 14, 13, 13, 12, 12, 12, 11, 11, 10, 10, 10,  9,  9,  9,  8,
  8,  8,  8,  7,  7,  7,  7,  6,  6,  6,  6,  6,  5,  5,  5,  5,  5,  4,  4,
  4,  4,  4,  4,  4,  4,  3,  3,  3,  3,  3,  3,  3,  3,  3,  2,  2,  2,  2,
  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  2,  1,  1,  1,  1,  1,  1,  1,  1,
  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1
};
static int ii_size_scales[BLOCK_SIZES_ALL] = {
    32, 32, 32,
    32, 16, 16, 16, 8, 8, 8, 4,
    4,  4,  2,  2,  2, 1, 1, 1,
    16, 16, 8, 8, 4, 4, 2, 2
};
#else
static const int ii_weights1d[MAX_SB_SIZE] = {
  60, 56, 52, 48, 45, 42, 39, 37, 34, 32, 30, 28, 26, 24, 22, 21,
  19, 18, 17, 16, 15, 14, 13, 12, 11, 10, 10,  9,  8,  8,  7,  7,
  6,  6,  6,  5,  5,  4,  4,  4,  4,  3,  3,  3,  3,  3,  2,  2,
  2,  2,  2,  2,  2,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1,  1
};
static int ii_size_scales[BLOCK_SIZES_ALL] = {
    16, 16, 16,
    16, 8, 8, 8, 4, 4, 4,
    2,  2, 2, 1, 1, 1,
    8, 8, 4, 4, 2, 2,
};
/* clang-format on */
#endif  // CONFIG_EXT_PARTITION

static void combine_interintra(INTERINTRA_MODE mode, int use_wedge_interintra,
                               int wedge_index, int wedge_sign,
                               BLOCK_SIZE bsize, BLOCK_SIZE plane_bsize,
                               uint8_t *comppred, int compstride,
                               const uint8_t *interpred, int interstride,
                               const uint8_t *intrapred, int intrastride) {
  const int bw = block_size_wide[plane_bsize];
  const int bh = block_size_high[plane_bsize];
  const int size_scale = ii_size_scales[plane_bsize];
  int i, j;

  if (use_wedge_interintra) {
    if (is_interintra_wedge_used(bsize)) {
      const uint8_t *mask =
          av1_get_contiguous_soft_mask(wedge_index, wedge_sign, bsize);
      const int subw = 2 * num_4x4_blocks_wide_lookup[bsize] == bw;
      const int subh = 2 * num_4x4_blocks_high_lookup[bsize] == bh;
      aom_blend_a64_mask(comppred, compstride, intrapred, intrastride,
                         interpred, interstride, mask, block_size_wide[bsize],
                         bh, bw, subh, subw);
    }
    return;
  }

  switch (mode) {
    case II_V_PRED:
      for (i = 0; i < bh; ++i) {
        for (j = 0; j < bw; ++j) {
          int scale = ii_weights1d[i * size_scale];
          comppred[i * compstride + j] =
              AOM_BLEND_A64(scale, intrapred[i * intrastride + j],
                            interpred[i * interstride + j]);
        }
      }
      break;

    case II_H_PRED:
      for (i = 0; i < bh; ++i) {
        for (j = 0; j < bw; ++j) {
          int scale = ii_weights1d[j * size_scale];
          comppred[i * compstride + j] =
              AOM_BLEND_A64(scale, intrapred[i * intrastride + j],
                            interpred[i * interstride + j]);
        }
      }
      break;

    case II_SMOOTH_PRED:
      for (i = 0; i < bh; ++i) {
        for (j = 0; j < bw; ++j) {
          int scale = ii_weights1d[(i < j ? i : j) * size_scale];
          comppred[i * compstride + j] =
              AOM_BLEND_A64(scale, intrapred[i * intrastride + j],
                            interpred[i * interstride + j]);
        }
      }
      break;

    case II_DC_PRED:
    default:
      for (i = 0; i < bh; ++i) {
        for (j = 0; j < bw; ++j) {
          comppred[i * compstride + j] = AOM_BLEND_AVG(
              intrapred[i * intrastride + j], interpred[i * interstride + j]);
        }
      }
      break;
  }
}

#if CONFIG_HIGHBITDEPTH
static void combine_interintra_highbd(
    INTERINTRA_MODE mode, int use_wedge_interintra, int wedge_index,
    int wedge_sign, BLOCK_SIZE bsize, BLOCK_SIZE plane_bsize,
    uint8_t *comppred8, int compstride, const uint8_t *interpred8,
    int interstride, const uint8_t *intrapred8, int intrastride, int bd) {
  const int bw = block_size_wide[plane_bsize];
  const int bh = block_size_high[plane_bsize];
  const int size_scale = ii_size_scales[plane_bsize];
  int i, j;

  uint16_t *comppred = CONVERT_TO_SHORTPTR(comppred8);
  const uint16_t *interpred = CONVERT_TO_SHORTPTR(interpred8);
  const uint16_t *intrapred = CONVERT_TO_SHORTPTR(intrapred8);

  if (use_wedge_interintra) {
    if (is_interintra_wedge_used(bsize)) {
      const uint8_t *mask =
          av1_get_contiguous_soft_mask(wedge_index, wedge_sign, bsize);
      const int subh = 2 * num_4x4_blocks_high_lookup[bsize] == bh;
      const int subw = 2 * num_4x4_blocks_wide_lookup[bsize] == bw;
      aom_highbd_blend_a64_mask(comppred8, compstride, intrapred8, intrastride,
                                interpred8, interstride, mask,
                                block_size_wide[bsize], bh, bw, subh, subw, bd);
    }
    return;
  }

  switch (mode) {
    case II_V_PRED:
      for (i = 0; i < bh; ++i) {
        for (j = 0; j < bw; ++j) {
          int scale = ii_weights1d[i * size_scale];
          comppred[i * compstride + j] =
              AOM_BLEND_A64(scale, intrapred[i * intrastride + j],
                            interpred[i * interstride + j]);
        }
      }
      break;

    case II_H_PRED:
      for (i = 0; i < bh; ++i) {
        for (j = 0; j < bw; ++j) {
          int scale = ii_weights1d[j * size_scale];
          comppred[i * compstride + j] =
              AOM_BLEND_A64(scale, intrapred[i * intrastride + j],
                            interpred[i * interstride + j]);
        }
      }
      break;

    case II_SMOOTH_PRED:
      for (i = 0; i < bh; ++i) {
        for (j = 0; j < bw; ++j) {
          int scale = ii_weights1d[(i < j ? i : j) * size_scale];
          comppred[i * compstride + j] =
              AOM_BLEND_A64(scale, intrapred[i * intrastride + j],
                            interpred[i * interstride + j]);
        }
      }
      break;

    case II_DC_PRED:
    default:
      for (i = 0; i < bh; ++i) {
        for (j = 0; j < bw; ++j) {
          comppred[i * compstride + j] = AOM_BLEND_AVG(
              interpred[i * interstride + j], intrapred[i * intrastride + j]);
        }
      }
      break;
  }
}
#endif  // CONFIG_HIGHBITDEPTH

void av1_build_intra_predictors_for_interintra(const AV1_COMMON *cm,
                                               MACROBLOCKD *xd,
                                               BLOCK_SIZE bsize, int plane,
                                               BUFFER_SET *ctx, uint8_t *dst,
                                               int dst_stride) {
  struct macroblockd_plane *const pd = &xd->plane[plane];
  BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, &xd->plane[plane]);
  PREDICTION_MODE mode =
      interintra_to_intra_mode[xd->mi[0]->mbmi.interintra_mode];

  av1_predict_intra_block(cm, xd, pd->width, pd->height, plane_bsize, mode,
                          ctx->plane[plane], ctx->stride[plane], dst,
                          dst_stride, 0, 0, plane);
}

void av1_combine_interintra(MACROBLOCKD *xd, BLOCK_SIZE bsize, int plane,
                            const uint8_t *inter_pred, int inter_stride,
                            const uint8_t *intra_pred, int intra_stride) {
  const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, &xd->plane[plane]);
#if CONFIG_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    combine_interintra_highbd(
        xd->mi[0]->mbmi.interintra_mode, xd->mi[0]->mbmi.use_wedge_interintra,
        xd->mi[0]->mbmi.interintra_wedge_index,
        xd->mi[0]->mbmi.interintra_wedge_sign, bsize, plane_bsize,
        xd->plane[plane].dst.buf, xd->plane[plane].dst.stride, inter_pred,
        inter_stride, intra_pred, intra_stride, xd->bd);
    return;
  }
#endif  // CONFIG_HIGHBITDEPTH
  combine_interintra(xd->mi[0]->mbmi.interintra_mode,
                     xd->mi[0]->mbmi.use_wedge_interintra,
                     xd->mi[0]->mbmi.interintra_wedge_index,
                     xd->mi[0]->mbmi.interintra_wedge_sign, bsize, plane_bsize,
                     xd->plane[plane].dst.buf, xd->plane[plane].dst.stride,
                     inter_pred, inter_stride, intra_pred, intra_stride);
}

void av1_build_interintra_predictors_sby(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                         uint8_t *ypred, int ystride,
                                         BUFFER_SET *ctx, BLOCK_SIZE bsize) {
#if CONFIG_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    DECLARE_ALIGNED(16, uint16_t, intrapredictor[MAX_SB_SQUARE]);
    av1_build_intra_predictors_for_interintra(
        cm, xd, bsize, 0, ctx, CONVERT_TO_BYTEPTR(intrapredictor), MAX_SB_SIZE);
    av1_combine_interintra(xd, bsize, 0, ypred, ystride,
                           CONVERT_TO_BYTEPTR(intrapredictor), MAX_SB_SIZE);
    return;
  }
#endif  // CONFIG_HIGHBITDEPTH
  {
    DECLARE_ALIGNED(16, uint8_t, intrapredictor[MAX_SB_SQUARE]);
    av1_build_intra_predictors_for_interintra(cm, xd, bsize, 0, ctx,
                                              intrapredictor, MAX_SB_SIZE);
    av1_combine_interintra(xd, bsize, 0, ypred, ystride, intrapredictor,
                           MAX_SB_SIZE);
  }
}

void av1_build_interintra_predictors_sbc(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                         uint8_t *upred, int ustride,
                                         BUFFER_SET *ctx, int plane,
                                         BLOCK_SIZE bsize) {
#if CONFIG_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    DECLARE_ALIGNED(16, uint16_t, uintrapredictor[MAX_SB_SQUARE]);
    av1_build_intra_predictors_for_interintra(
        cm, xd, bsize, plane, ctx, CONVERT_TO_BYTEPTR(uintrapredictor),
        MAX_SB_SIZE);
    av1_combine_interintra(xd, bsize, plane, upred, ustride,
                           CONVERT_TO_BYTEPTR(uintrapredictor), MAX_SB_SIZE);
    return;
  }
#endif  // CONFIG_HIGHBITDEPTH
  {
    DECLARE_ALIGNED(16, uint8_t, uintrapredictor[MAX_SB_SQUARE]);
    av1_build_intra_predictors_for_interintra(cm, xd, bsize, plane, ctx,
                                              uintrapredictor, MAX_SB_SIZE);
    av1_combine_interintra(xd, bsize, plane, upred, ustride, uintrapredictor,
                           MAX_SB_SIZE);
  }
}

void av1_build_interintra_predictors_sbuv(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                          uint8_t *upred, uint8_t *vpred,
                                          int ustride, int vstride,
                                          BUFFER_SET *ctx, BLOCK_SIZE bsize) {
  av1_build_interintra_predictors_sbc(cm, xd, upred, ustride, ctx, 1, bsize);
  av1_build_interintra_predictors_sbc(cm, xd, vpred, vstride, ctx, 2, bsize);
}

void av1_build_interintra_predictors(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                     uint8_t *ypred, uint8_t *upred,
                                     uint8_t *vpred, int ystride, int ustride,
                                     int vstride, BUFFER_SET *ctx,
                                     BLOCK_SIZE bsize) {
  av1_build_interintra_predictors_sby(cm, xd, ypred, ystride, ctx, bsize);
  av1_build_interintra_predictors_sbuv(cm, xd, upred, vpred, ustride, vstride,
                                       ctx, bsize);
}
#endif  // CONFIG_INTERINTRA

// Builds the inter-predictor for the single ref case
// for use in the encoder to search the wedges efficiently.
static void build_inter_predictors_single_buf(MACROBLOCKD *xd, int plane,
                                              int block, int bw, int bh, int x,
                                              int y, int w, int h, int mi_x,
                                              int mi_y, int ref,
                                              uint8_t *const ext_dst,
                                              int ext_dst_stride) {
  struct macroblockd_plane *const pd = &xd->plane[plane];
  const MODE_INFO *mi = xd->mi[0];

  const struct scale_factors *const sf = &xd->block_refs[ref]->sf;
  struct buf_2d *const pre_buf = &pd->pre[ref];
#if CONFIG_HIGHBITDEPTH
  uint8_t *const dst =
      (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH ? CONVERT_TO_BYTEPTR(ext_dst)
                                                   : ext_dst) +
      ext_dst_stride * y + x;
#else
  uint8_t *const dst = ext_dst + ext_dst_stride * y + x;
#endif
  const MV mv = mi->mbmi.sb_type < BLOCK_8X8
                    ? average_split_mvs(pd, mi, ref, block)
                    : mi->mbmi.mv[ref].as_mv;

  uint8_t *pre;
  int xs, ys, subpel_x, subpel_y;
  const int is_scaled = av1_is_scaled(sf);
  ConvolveParams conv_params = get_conv_params(ref, 0, plane);
  WarpTypesAllowed warp_types;
#if CONFIG_COMPOUND_SINGLEREF
  WarpedMotionParams *const wm =
      mi->mbmi.ref_frame[ref] > 0 ? &xd->global_motion[mi->mbmi.ref_frame[ref]]
                                  : &xd->global_motion[mi->mbmi.ref_frame[0]];
#else   // !(CONFIG_COMPOUND_SINGLEREF)
  WarpedMotionParams *const wm = &xd->global_motion[mi->mbmi.ref_frame[ref]];
#endif  // CONFIG_COMPOUND_SINGLEREF
  warp_types.global_warp_allowed = is_global_mv_block(mi, block, wm->wmtype);
#if CONFIG_WARPED_MOTION
  warp_types.local_warp_allowed = mi->mbmi.motion_mode == WARPED_CAUSAL;
#endif  // CONFIG_WARPED_MOTION

  if (is_scaled) {
    int ssx = pd->subsampling_x;
    int ssy = pd->subsampling_y;
    int orig_pos_y = (mi_y << (SUBPEL_BITS - ssy)) + (y << SUBPEL_BITS);
    orig_pos_y += mv.row * (1 << (1 - ssy));
    int orig_pos_x = (mi_x << (SUBPEL_BITS - ssx)) + (x << SUBPEL_BITS);
    orig_pos_x += mv.col * (1 << (1 - ssx));
    int pos_y = sf->scale_value_y(orig_pos_y, sf);
    int pos_x = sf->scale_value_x(orig_pos_x, sf);
    pos_x += SCALE_EXTRA_OFF;
    pos_y += SCALE_EXTRA_OFF;

    const int top = -AOM_LEFT_TOP_MARGIN_SCALED;
    const int left = -AOM_LEFT_TOP_MARGIN_SCALED;
    const int bottom = (pre_buf->height + AOM_INTERP_EXTEND)
                       << SCALE_SUBPEL_BITS;
    const int right = (pre_buf->width + AOM_INTERP_EXTEND) << SCALE_SUBPEL_BITS;
    pos_y = clamp(pos_y, top, bottom);
    pos_x = clamp(pos_x, left, right);

    pre = pre_buf->buf0 + (pos_y >> SCALE_SUBPEL_BITS) * pre_buf->stride +
          (pos_x >> SCALE_SUBPEL_BITS);
    subpel_x = pos_x & SCALE_SUBPEL_MASK;
    subpel_y = pos_y & SCALE_SUBPEL_MASK;
    xs = sf->x_step_q4;
    ys = sf->y_step_q4;
  } else {
    const MV mv_q4 = clamp_mv_to_umv_border_sb(
        xd, &mv, bw, bh, pd->subsampling_x, pd->subsampling_y);
    xs = ys = SCALE_SUBPEL_SHIFTS;
    subpel_x = (mv_q4.col & SUBPEL_MASK) << SCALE_EXTRA_BITS;
    subpel_y = (mv_q4.row & SUBPEL_MASK) << SCALE_EXTRA_BITS;
    pre = pre_buf->buf + (y + (mv_q4.row >> SUBPEL_BITS)) * pre_buf->stride +
          (x + (mv_q4.col >> SUBPEL_BITS));
  }

  av1_make_inter_predictor(pre, pre_buf->stride, dst, ext_dst_stride, subpel_x,
                           subpel_y, sf, w, h, &conv_params,
                           mi->mbmi.interp_filters, &warp_types,
                           (mi_x >> pd->subsampling_x) + x,
                           (mi_y >> pd->subsampling_y) + y, plane, ref,
#if CONFIG_MOTION_VAR
                           mi, 0,
#endif
                           xs, ys, xd);
}

void av1_build_inter_predictors_for_planes_single_buf(
    MACROBLOCKD *xd, BLOCK_SIZE bsize, int plane_from, int plane_to, int mi_row,
    int mi_col, int ref, uint8_t *ext_dst[3], int ext_dst_stride[3]) {
  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 bw = block_size_wide[plane_bsize];
    const int bh = block_size_high[plane_bsize];
    build_inter_predictors_single_buf(xd, plane, 0, bw, bh, 0, 0, bw, bh, mi_x,
                                      mi_y, ref, ext_dst[plane],
                                      ext_dst_stride[plane]);
  }
}

static void build_wedge_inter_predictor_from_buf(
    MACROBLOCKD *xd, int plane, int x, int y, int w, int h, uint8_t *ext_dst0,
    int ext_dst_stride0, uint8_t *ext_dst1, int ext_dst_stride1) {
  MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
  const int is_compound = has_second_ref(mbmi);
  MACROBLOCKD_PLANE *const pd = &xd->plane[plane];
  struct buf_2d *const dst_buf = &pd->dst;
  uint8_t *const dst = dst_buf->buf + dst_buf->stride * y + x;
  const INTERINTER_COMPOUND_DATA comp_data = { mbmi->wedge_index,
                                               mbmi->wedge_sign,
                                               mbmi->mask_type, xd->seg_mask,
                                               mbmi->interinter_compound_type };

#if CONFIG_COMPOUND_SINGLEREF
  if ((is_compound || is_inter_singleref_comp_mode(mbmi->mode)) &&
      is_masked_compound_type(mbmi->interinter_compound_type))
#else   // !CONFIG_COMPOUND_SINGLEREF
  if (is_compound && is_masked_compound_type(mbmi->interinter_compound_type))
#endif  // CONFIG_COMPOUND_SINGLEREF
  {
    if (!plane && comp_data.interinter_compound_type == COMPOUND_SEG) {
#if CONFIG_HIGHBITDEPTH
      if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
        build_compound_seg_mask_highbd(
            comp_data.seg_mask, comp_data.mask_type,
            CONVERT_TO_BYTEPTR(ext_dst0), ext_dst_stride0,
            CONVERT_TO_BYTEPTR(ext_dst1), ext_dst_stride1, mbmi->sb_type, h, w,
            xd->bd);
      else
#endif  // CONFIG_HIGHBITDEPTH
        build_compound_seg_mask(comp_data.seg_mask, comp_data.mask_type,
                                ext_dst0, ext_dst_stride0, ext_dst1,
                                ext_dst_stride1, mbmi->sb_type, h, w);
    }

#if CONFIG_HIGHBITDEPTH
    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
      build_masked_compound_highbd(
          dst, dst_buf->stride, CONVERT_TO_BYTEPTR(ext_dst0), ext_dst_stride0,
          CONVERT_TO_BYTEPTR(ext_dst1), ext_dst_stride1, &comp_data,
          mbmi->sb_type, h, w, xd->bd);
    else
#endif  // CONFIG_HIGHBITDEPTH
      build_masked_compound(dst, dst_buf->stride, ext_dst0, ext_dst_stride0,
                            ext_dst1, ext_dst_stride1, &comp_data,
                            mbmi->sb_type, h, w);
  } else {
#if CONFIG_HIGHBITDEPTH
    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
      aom_highbd_convolve_copy(CONVERT_TO_BYTEPTR(ext_dst0), ext_dst_stride0,
                               dst, dst_buf->stride, NULL, 0, NULL, 0, w, h,
                               xd->bd);
    else
#endif  // CONFIG_HIGHBITDEPTH
      aom_convolve_copy(ext_dst0, ext_dst_stride0, dst, dst_buf->stride, NULL,
                        0, NULL, 0, w, h);
  }
}

void av1_build_wedge_inter_predictor_from_buf(MACROBLOCKD *xd, BLOCK_SIZE bsize,
                                              int plane_from, int plane_to,
                                              uint8_t *ext_dst0[3],
                                              int ext_dst_stride0[3],
                                              uint8_t *ext_dst1[3],
                                              int ext_dst_stride1[3]) {
  int plane;
  for (plane = plane_from; plane <= plane_to; ++plane) {
    const BLOCK_SIZE plane_bsize =
        get_plane_block_size(bsize, &xd->plane[plane]);
    const int bw = block_size_wide[plane_bsize];
    const int bh = block_size_high[plane_bsize];
    build_wedge_inter_predictor_from_buf(
        xd, plane, 0, 0, bw, bh, ext_dst0[plane], ext_dst_stride0[plane],
        ext_dst1[plane], ext_dst_stride1[plane]);
  }
}
#if CONFIG_NCOBMC_ADAPT_WEIGHT

void alloc_ncobmc_pred_buffer(MACROBLOCKD *const xd) {
  int i;
  // allocate interpolated prediction buffer
  for (i = 0; i < MAX_MB_PLANE; ++i) {
    xd->ncobmc_pred_buf[i] = (uint8_t *)malloc(sizeof(uint8_t) * MAX_SB_SQUARE);
    av1_zero_array(xd->ncobmc_pred_buf[i], MAX_SB_SQUARE);
    xd->ncobmc_pred_buf_stride[i] = MAX_SB_SIZE;
  }
}

void free_ncobmc_pred_buffer(MACROBLOCKD *const xd) {
  for (int i = 0; i < MAX_MB_PLANE; ++i) free(xd->ncobmc_pred_buf[i]);
}

void get_pred_from_intrpl_buf(MACROBLOCKD *xd, int mi_row, int mi_col,
                              BLOCK_SIZE bsize, int plane) {
  uint8_t *dst = xd->plane[plane].dst.buf;
  int ds = xd->plane[plane].dst.stride;
  int ss_x = xd->plane[plane].subsampling_x;
  int ss_y = xd->plane[plane].subsampling_y;

  const int ip_wide = mi_size_wide[bsize] * MI_SIZE >> ss_x;
  const int ip_high = mi_size_high[bsize] * MI_SIZE >> ss_y;
  // relative coordinates of this MI in the superblock
  int row_rlt = (mi_row - xd->sb_mi_bd.mi_row_begin) * MI_SIZE >> ss_y;
  int col_rlt = (mi_col - xd->sb_mi_bd.mi_col_begin) * MI_SIZE >> ss_x;
  int s = xd->ncobmc_pred_buf_stride[plane];
  int r, c;

  for (r = 0; r < ip_high; ++r) {
    for (c = 0; c < ip_wide; ++c) {
      dst[r * ds + c] =
          xd->ncobmc_pred_buf[plane][(r + row_rlt) * s + c + col_rlt];
    }
  }
}
// scaling factors for ncobmc kernels
#define KERNEL_SCALE_LOG 14

void build_ncobmc_intrpl_pred(const AV1_COMMON *const cm, MACROBLOCKD *xd,
                              int plane, int pxl_row, int pxl_col,
                              BLOCK_SIZE bsize, uint8_t *preds[][MAX_MB_PLANE],
                              int stride[MAX_MB_PLANE],  // pred buffer strides
                              int mode) {
  const ADAPT_OVERLAP_BLOCK ao_block = adapt_overlap_block_lookup[bsize];
  const NCOBMC_KERNELS *const knls = &cm->ncobmc_kernels[ao_block][mode];
  const int wide = mi_size_wide[bsize] * MI_SIZE;
  const int high = mi_size_high[bsize] * MI_SIZE;
  const int s = stride[plane];
  const int ss_x = xd->plane[plane].subsampling_x;
  const int ss_y = xd->plane[plane].subsampling_y;
  int row_offset = (pxl_row - xd->sb_mi_bd.mi_row_begin * MI_SIZE) >> ss_y;
  int col_offset = (pxl_col - xd->sb_mi_bd.mi_col_begin * MI_SIZE) >> ss_x;
  int dst_stride = xd->ncobmc_pred_buf_stride[plane];
  int dst_offset = row_offset * dst_stride + col_offset;

#if CONFIG_HIGHBITDEPTH
  const int is_hbd = (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) ? 1 : 0;
#else
  const int is_hbd = 0;
#endif  // CONFIG_HIGHBITDEPTH

  int r, c, k_r, k_c;
  int64_t tmp;

  for (r = 0; r < (high >> ss_x); ++r) {
    for (c = 0; c < (wide >> ss_y); ++c) {
      int pos = r * s + c;
      int q_tmp;
      uint8_t val;

      // TODO(weitinglin): find out the optimal sub-sampling patterns for
      //                   chroma
      k_r = (r << ss_y) + ss_y;
      k_c = (c << ss_x) + ss_x;
      if (ss_y && k_r >= high) k_r -= 1;
      if (ss_x && k_c >= wide) k_c -= 1;

      if (!is_hbd) {
        uint8_t *tmp_p[4];
        int i;
        for (i = 0; i < 4; ++i) tmp_p[i] = preds[i][plane];

        tmp = 0;
        for (i = 0; i < 4; ++i)
          tmp += knls->KERNEL[i][k_r][k_c] * tmp_p[i][pos];

      } else {
        uint16_t *tmp_p[4];
        int i;
        for (i = 0; i < 4; ++i) tmp_p[i] = CONVERT_TO_SHORTPTR(preds[i][plane]);

        tmp = 0;
        for (i = 0; i < 4; ++i)
          tmp += knls->KERNEL[i][k_r][k_c] * tmp_p[i][pos];
      }

      q_tmp = (tmp <= 0) ? 0 : ROUND_POWER_OF_TWO(tmp, KERNEL_SCALE_LOG);
      val = clip_pixel(q_tmp);

      xd->ncobmc_pred_buf[plane][r * dst_stride + c + dst_offset] = val;

      assert(r * dst_stride + c + dst_offset < MAX_SB_SQUARE);
    }
  }
}

void get_pred_by_horz_neighbor(const AV1_COMMON *cm, MACROBLOCKD *xd, int bsize,
                               int mi_row, int mi_col,
                               uint8_t *dst_buf[MAX_MB_PLANE],
                               int dst_stride[MAX_MB_PLANE]) {
  const TileInfo *const tile = &xd->tile;
  const int mb_to_bottom_edge_base = xd->mb_to_bottom_edge;
  const int mb_to_top_edge_base = xd->mb_to_top_edge;
  const int mb_to_left_edge_base = xd->mb_to_left_edge;
  const int mb_to_right_edge_base = xd->mb_to_right_edge;
  int overlappable_offset = -1;
  const int mi_nums = AOMMIN(mi_size_high[bsize], cm->mi_rows - mi_row);

  int i, j, mi_step, ref;

  xd->mb_to_right_edge += mi_size_wide[bsize] * MI_SIZE * 4;

  // build from left neighbors
  for (i = 0; i < mi_nums; i += mi_step) {
    int mi_row_offset = i;
    int mi_col_offset = -1;
    int mi_x, mi_y, bw, bh;
    MODE_INFO *left_mi;
    MB_MODE_INFO *left_mbmi, backup_mbmi;
    BLOCK_SIZE l_bsize;

    // create the original prediction if offset exceeds the boundary
    if (mi_col == 0 || (mi_col - 1 < tile->mi_col_start)) mi_col_offset = 0;

    left_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
    left_mbmi = &left_mi->mbmi;
    l_bsize = AOMMAX(left_mbmi->sb_type, BLOCK_8X8);

    mi_step = AOMMIN(xd->n8_h, mi_size_high[l_bsize]);

    // reset the mi if it is not overlappble
    if (!is_neighbor_overlappable(left_mbmi)) {
      // use left_mbmi->sb_type instead of l_bsize to handle
      // sub8x8 cases
      int search_mi_step = mi_size_high[left_mbmi->sb_type];
      while (!is_neighbor_overlappable(left_mbmi)) {
        mi_row_offset += search_mi_step;
        if (mi_row_offset < mi_nums) {
          left_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
          left_mbmi = &left_mi->mbmi;
          search_mi_step = mi_size_high[left_mbmi->sb_type];
        } else {
          if (overlappable_offset >= 0) {
            mi_row_offset = overlappable_offset;
          } else {
            mi_row_offset = 0;
            mi_col_offset = 0;
          }
          left_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
          left_mbmi = &left_mi->mbmi;
          break;
        }
      }
    } else {
      // update the available overlappable mi
      overlappable_offset = mi_row_offset;
    }

    backup_mbmi = *left_mbmi;
    modify_neighbor_predictor_for_obmc(left_mbmi);

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      struct macroblockd_plane *const pd = &xd->plane[j];
      setup_pred_plane(&pd->dst, l_bsize, dst_buf[j], MAX_SB_SIZE, MAX_SB_SIZE,
                       dst_stride[j], i, 0, NULL, pd->subsampling_x,
                       pd->subsampling_y);
    }
#if CONFIG_COMPOUND_SINGLEREF
    for (ref = 0; ref < 1 + (is_inter_anyref_comp_mode(left_mbmi->mode));
         ++ref) {
      const MV_REFERENCE_FRAME frame = has_second_ref(left_mbmi)
                                           ? left_mbmi->ref_frame[ref]
                                           : left_mbmi->ref_frame[0];
#else   // !(CONFIG_COMPOUND_SINGLEREF)
    for (ref = 0; ref < 1 + has_second_ref(left_mbmi); ++ref) {
      const MV_REFERENCE_FRAME frame = left_mbmi->ref_frame[ref];
#endif  // CONFIG_COMPOUND_SINGLEREF
      const RefBuffer *const ref_buf = &cm->frame_refs[frame - LAST_FRAME];

      xd->block_refs[ref] = ref_buf;
      if ((!av1_is_valid_scale(&ref_buf->sf)))
        aom_internal_error(xd->error_info, AOM_CODEC_UNSUP_BITSTREAM,
                           "Reference frame has invalid dimensions");
      av1_setup_pre_planes(xd, ref, ref_buf->buf, mi_row + i, mi_col,
                           &ref_buf->sf);
    }
    xd->mb_to_top_edge = -((mi_row + i) * MI_SIZE * 8);
    xd->mb_to_bottom_edge =
        mb_to_bottom_edge_base + (mi_nums - i - mi_step) * MI_SIZE * 8;
    mi_x = mi_col << MI_SIZE_LOG2;
    mi_y = (mi_row + i) << MI_SIZE_LOG2;

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      const struct macroblockd_plane *pd = &xd->plane[j];
      bw = mi_size_wide[bsize] << (MI_SIZE_LOG2 - 1) >> pd->subsampling_x;
      bh = (mi_step << MI_SIZE_LOG2) >> pd->subsampling_y;

      build_inter_predictors(cm, xd, j, left_mi, 1, 0, bw, bh, 0, 0, bw, bh,
                             mi_x, mi_y);
    }
    *left_mbmi = backup_mbmi;
  }

  // build from right neighbors
  xd->mb_to_right_edge = mb_to_right_edge_base;
  xd->mb_to_left_edge -= mi_size_wide[bsize] * MI_SIZE * 4;

  overlappable_offset = -1;

  for (i = 0; i < mi_nums; i += mi_step) {
    int mi_row_offset = i;
    int mi_col_offset = mi_size_wide[bsize];
    int mi_x, mi_y, bw, bh;
    int mi_col_shift = mi_size_wide[bsize] >> 1;
    MODE_INFO *right_mi;
    MB_MODE_INFO *right_mbmi, backup_mbmi;
    BLOCK_SIZE r_bsize;

    // create the original prediction if offset exceeds the boundary
    if (mi_col + mi_col_offset > xd->sb_mi_bd.mi_col_end) mi_col_offset = 0;

    right_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
    right_mbmi = &right_mi->mbmi;
    r_bsize = AOMMAX(right_mbmi->sb_type, BLOCK_8X8);

    mi_step = AOMMIN(mi_nums, mi_size_high[r_bsize]);

    if (!is_neighbor_overlappable(right_mbmi)) {
      int search_mi_step = mi_size_high[right_mbmi->sb_type];
      while (!is_neighbor_overlappable(right_mbmi)) {
        mi_row_offset += search_mi_step;
        if (mi_row_offset < mi_nums) {
          right_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
          right_mbmi = &right_mi->mbmi;
          search_mi_step = mi_size_high[right_mbmi->sb_type];
        } else {
          if (overlappable_offset >= 0) {
            mi_row_offset = overlappable_offset;
          } else {
            mi_row_offset = 0;
            mi_col_offset = 0;
          }
          right_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
          right_mbmi = &right_mi->mbmi;
          break;
        }
      }
    } else {
      overlappable_offset = mi_row_offset;
    }

    backup_mbmi = *right_mbmi;
    modify_neighbor_predictor_for_obmc(right_mbmi);

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      struct macroblockd_plane *const pd = &xd->plane[j];
      setup_pred_plane(&pd->dst, r_bsize, dst_buf[j], MAX_SB_SIZE, MAX_SB_SIZE,
                       dst_stride[j], i, mi_col_shift, NULL, pd->subsampling_x,
                       pd->subsampling_y);
    }
#if CONFIG_COMPOUND_SINGLEREF
    for (ref = 0; ref < 1 + (is_inter_anyref_comp_mode(right_mbmi->mode));
         ++ref) {
      const MV_REFERENCE_FRAME frame = has_second_ref(right_mbmi)
                                           ? right_mbmi->ref_frame[ref]
                                           : right_mbmi->ref_frame[0];
#else   // !(CONFIG_COMPOUND_SINGLEREF)
    for (ref = 0; ref < 1 + has_second_ref(right_mbmi); ++ref) {
      const MV_REFERENCE_FRAME frame = right_mbmi->ref_frame[ref];
#endif  // CONFIG_COMPOUND_SINGLEREF
      const RefBuffer *const ref_buf = &cm->frame_refs[frame - LAST_FRAME];
      xd->block_refs[ref] = ref_buf;
      if ((!av1_is_valid_scale(&ref_buf->sf)))
        aom_internal_error(xd->error_info, AOM_CODEC_UNSUP_BITSTREAM,
                           "Reference frame has invalid dimensions");
      av1_setup_pre_planes(xd, ref, ref_buf->buf, mi_row + i,
                           mi_col + mi_col_shift, &ref_buf->sf);
    }
    xd->mb_to_top_edge = -((mi_row + i) * MI_SIZE * 8);
    xd->mb_to_bottom_edge =
        mb_to_bottom_edge_base + (mi_nums - i - mi_step) * MI_SIZE * 8;
    mi_x = (mi_col + mi_col_shift) << MI_SIZE_LOG2;
    mi_y = (mi_row + i) << MI_SIZE_LOG2;

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      const struct macroblockd_plane *pd = &xd->plane[j];
      bw = mi_size_wide[bsize] << (MI_SIZE_LOG2 - 1) >> pd->subsampling_x;
      bh = (mi_step << MI_SIZE_LOG2) >> pd->subsampling_y;

      build_inter_predictors(cm, xd, j, right_mi, 1, 0, bw, bh, 0, 0, bw, bh,
                             mi_x, mi_y);
    }

    *right_mbmi = backup_mbmi;
  }

  // restore the boundaries
  xd->mb_to_top_edge = mb_to_top_edge_base;
  xd->mb_to_bottom_edge = mb_to_bottom_edge_base;
  xd->mb_to_left_edge = mb_to_left_edge_base;
  xd->mb_to_right_edge = mb_to_right_edge_base;
}

void get_pred_by_vert_neighbor(const AV1_COMMON *cm, MACROBLOCKD *xd, int bsize,
                               int mi_row, int mi_col,
                               uint8_t *dst_buf[MAX_MB_PLANE],
                               int dst_stride[MAX_MB_PLANE]) {
  const TileInfo *const tile = &xd->tile;
  const int mb_to_bottom_edge_base = xd->mb_to_bottom_edge;
  const int mb_to_top_edge_base = xd->mb_to_top_edge;
  const int mb_to_left_edge_base = xd->mb_to_left_edge;
  const int mb_to_right_edge_base = xd->mb_to_right_edge;
  int overlappable_offset = -1;
  const int mi_nums = AOMMIN(mi_size_wide[bsize], cm->mi_cols - mi_col);

  int i, j, mi_step, ref;

  xd->mb_to_bottom_edge += mi_nums * MI_SIZE * 4;

  // build from above neighbors
  for (i = 0; i < mi_nums; i += mi_step) {
    int mi_row_offset = -1;
    int mi_col_offset = i;
    int mi_x, mi_y, bw, bh;
    MODE_INFO *above_mi;
    MB_MODE_INFO *above_mbmi, backup_mbmi;
    BLOCK_SIZE a_bsize;

    // create the original prediction if offset exceeds the boundary
    if (mi_row <= tile->mi_row_start) mi_row_offset = 0;

    above_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
    above_mbmi = &above_mi->mbmi;
    a_bsize = AOMMAX(above_mbmi->sb_type, BLOCK_8X8);

    mi_step = AOMMIN(mi_nums, mi_size_high[a_bsize]);

    // reset the mi if it is not overlappble
    if (!is_neighbor_overlappable(above_mbmi)) {
      int search_mi_step = mi_size_high[above_mbmi->sb_type];
      // backward search
      while (!is_neighbor_overlappable(above_mbmi)) {
        mi_col_offset += search_mi_step;
        if (mi_col_offset < mi_nums) {
          above_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
          above_mbmi = &above_mi->mbmi;
          search_mi_step = mi_size_high[above_mbmi->sb_type];
        } else {
          if (overlappable_offset >= 0) {
            mi_col_offset = overlappable_offset;
          } else {
            mi_row_offset = 0;
            mi_col_offset = 0;
          }
          above_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
          above_mbmi = &above_mi->mbmi;
          break;
        }
      }
    } else {
      // update the available overlappable mi
      overlappable_offset = mi_col_offset;
    }

    backup_mbmi = *above_mbmi;
    modify_neighbor_predictor_for_obmc(above_mbmi);

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      struct macroblockd_plane *const pd = &xd->plane[j];
      setup_pred_plane(&pd->dst, a_bsize, dst_buf[j], MAX_SB_SIZE, MAX_SB_SIZE,
                       dst_stride[j], 0, i, NULL, pd->subsampling_x,
                       pd->subsampling_y);
    }
#if CONFIG_COMPOUND_SINGLEREF
    for (ref = 0; ref < 1 + (is_inter_anyref_comp_mode(above_mbmi->mode));
         ++ref) {
      const MV_REFERENCE_FRAME frame = has_second_ref(above_mbmi)
                                           ? above_mbmi->ref_frame[ref]
                                           : above_mbmi->ref_frame[0];
#else   // !(CONFIG_COMPOUND_SINGLEREF)
    for (ref = 0; ref < 1 + has_second_ref(above_mbmi); ++ref) {
      const MV_REFERENCE_FRAME frame = above_mbmi->ref_frame[ref];
#endif  // CONFIG_COMPOUND_SINGLEREF
      const RefBuffer *const ref_buf = &cm->frame_refs[frame - LAST_FRAME];

      xd->block_refs[ref] = ref_buf;
      if ((!av1_is_valid_scale(&ref_buf->sf)))
        aom_internal_error(xd->error_info, AOM_CODEC_UNSUP_BITSTREAM,
                           "Reference frame has invalid dimensions");
      av1_setup_pre_planes(xd, ref, ref_buf->buf, mi_row, mi_col + i,
                           &ref_buf->sf);
    }

    xd->mb_to_left_edge = -(((mi_col + i) * MI_SIZE) * 8);
    xd->mb_to_right_edge =
        mb_to_right_edge_base + (mi_nums - i - mi_step) * MI_SIZE * 8;
    mi_x = (mi_col + i) << MI_SIZE_LOG2;
    mi_y = mi_row << MI_SIZE_LOG2;

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      const struct macroblockd_plane *pd = &xd->plane[j];

      bh = mi_size_high[bsize] << (MI_SIZE_LOG2 - 1) >> pd->subsampling_x;
      bw = (mi_step << MI_SIZE_LOG2) >> pd->subsampling_y;

      build_inter_predictors(cm, xd, j, above_mi, 1, 0, bw, bh, 0, 0, bw, bh,
                             mi_x, mi_y);
    }

    *above_mbmi = backup_mbmi;
  }

  // build from bottom neighbors
  xd->mb_to_bottom_edge = mb_to_bottom_edge_base;
  xd->mb_to_top_edge -= mi_size_high[bsize] * MI_SIZE * 4;

  overlappable_offset = -1;

  for (i = 0; i < mi_nums; i += mi_step) {
    int mi_row_offset = mi_size_high[bsize];
    int mi_col_offset = i;
    int mi_x, mi_y, bw, bh;
    int mi_row_shift = mi_size_high[bsize] >> 1;
    MODE_INFO *bottom_mi;
    MB_MODE_INFO *bottom_mbmi, backup_mbmi;
    BLOCK_SIZE b_bsize;

    // create the original prediction if offset exceeds the boundary
    if (mi_row + mi_row_offset > xd->sb_mi_bd.mi_row_end) mi_row_offset = 0;

    bottom_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
    bottom_mbmi = &bottom_mi->mbmi;
    b_bsize = AOMMAX(bottom_mbmi->sb_type, BLOCK_8X8);

    mi_step = AOMMIN(mi_nums, mi_size_high[b_bsize]);

    // reset the mi if it is not overlappble
    if (!is_neighbor_overlappable(bottom_mbmi)) {
      int search_mi_step = mi_size_high[bottom_mbmi->sb_type];
      while (!is_neighbor_overlappable(bottom_mbmi)) {
        mi_col_offset += search_mi_step;
        if (mi_col_offset < mi_nums) {
          bottom_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
          bottom_mbmi = &bottom_mi->mbmi;
          search_mi_step = mi_size_high[bottom_mbmi->sb_type];
        } else {
          if (overlappable_offset >= 0) {
            mi_col_offset = overlappable_offset;
          } else {
            mi_col_offset = 0;
            mi_row_offset = 0;
          }
          bottom_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
          bottom_mbmi = &bottom_mi->mbmi;
          break;
        }
      }
    } else {
      // update the available overlappable mi
      overlappable_offset = mi_col_offset;
    }

    backup_mbmi = *bottom_mbmi;
    modify_neighbor_predictor_for_obmc(bottom_mbmi);

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      struct macroblockd_plane *const pd = &xd->plane[j];
      setup_pred_plane(&pd->dst, b_bsize, dst_buf[j], MAX_SB_SIZE, MAX_SB_SIZE,
                       dst_stride[j], mi_row_shift, i, NULL, pd->subsampling_x,
                       pd->subsampling_y);
    }
#if CONFIG_COMPOUND_SINGLEREF
    for (ref = 0; ref < 1 + (is_inter_anyref_comp_mode(bottom_mbmi->mode));
         ++ref) {
      const MV_REFERENCE_FRAME frame = has_second_ref(bottom_mbmi)
                                           ? bottom_mbmi->ref_frame[ref]
                                           : bottom_mbmi->ref_frame[0];
#else   // !(CONFIG_COMPOUND_SINGLEREF)
    for (ref = 0; ref < 1 + has_second_ref(bottom_mbmi); ++ref) {
      const MV_REFERENCE_FRAME frame = bottom_mbmi->ref_frame[ref];
#endif  // CONFIG_COMPOUND_SINGLEREF
      const RefBuffer *const ref_buf = &cm->frame_refs[frame - LAST_FRAME];
      xd->block_refs[ref] = ref_buf;
      if ((!av1_is_valid_scale(&ref_buf->sf)))
        aom_internal_error(xd->error_info, AOM_CODEC_UNSUP_BITSTREAM,
                           "Reference frame has invalid dimensions");
      av1_setup_pre_planes(xd, ref, ref_buf->buf, mi_row + mi_row_shift,
                           mi_col + i, &ref_buf->sf);
    }

    xd->mb_to_left_edge = -(((mi_col + i) * MI_SIZE) * 8);
    xd->mb_to_right_edge =
        mb_to_right_edge_base + (mi_nums - i - mi_step) * MI_SIZE * 8;
    mi_x = (mi_col + i) << MI_SIZE_LOG2;
    mi_y = (mi_row + mi_row_shift) << MI_SIZE_LOG2;

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      const struct macroblockd_plane *pd = &xd->plane[j];

      bh = mi_size_high[bsize] << (MI_SIZE_LOG2 - 1) >> pd->subsampling_x;
      bw = (mi_step << MI_SIZE_LOG2) >> pd->subsampling_y;

      build_inter_predictors(cm, xd, j, bottom_mi, 1, 0, bw, bh, 0, 0, bw, bh,
                             mi_x, mi_y);
    }

    *bottom_mbmi = backup_mbmi;
  }
  // restore the boundaries
  xd->mb_to_top_edge = mb_to_top_edge_base;
  xd->mb_to_bottom_edge = mb_to_bottom_edge_base;
  xd->mb_to_left_edge = mb_to_left_edge_base;
  xd->mb_to_right_edge = mb_to_right_edge_base;
}

void get_pred_by_corner_neighbor(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                 int bsize, int mi_row, int mi_col,
                                 uint8_t *dst_buf[MAX_MB_PLANE],
                                 int dst_stride[MAX_MB_PLANE]) {
  const TileInfo *const tile = &xd->tile;
  const int mb_to_bottom_edge_base = xd->mb_to_bottom_edge;
  const int mb_to_top_edge_base = xd->mb_to_top_edge;
  const int mb_to_left_edge_base = xd->mb_to_left_edge;
  const int mb_to_right_edge_base = xd->mb_to_right_edge;
  const int mi_wide = mi_size_wide[bsize];
  const int mi_high = mi_size_high[bsize];

  // location of four mi sources
  const int mi_row_offsets[4] = { -1, -1, mi_high, mi_high };
  const int mi_col_offsets[4] = { -1, mi_wide, -1, mi_wide };

  MB_MODE_INFO backup_mbmi;
  int mi_x, mi_y, bh, bw;
  int i, j, ref;

  assert(bsize >= BLOCK_8X8);

  for (i = 0; i < 4; ++i) {
    int mi_row_offset = mi_row_offsets[i];
    int mi_col_offset = mi_col_offsets[i];
    MODE_INFO *corner_mi;
    MB_MODE_INFO *corner_mbmi;

    if (mi_col + mi_col_offset < tile->mi_col_start ||
        mi_col + mi_col_offset > xd->sb_mi_bd.mi_col_end)
      mi_col_offset = 0;

    if (mi_row + mi_row_offset < tile->mi_row_start ||
        mi_row + mi_row_offset > xd->sb_mi_bd.mi_row_end)
      mi_row_offset = 0;

    corner_mi = xd->mi[mi_col_offset + mi_row_offset * xd->mi_stride];
    corner_mbmi = &corner_mi->mbmi;

    // reset the mi if it is not overlappble
    if (!is_neighbor_overlappable(corner_mbmi)) {
      mi_row_offset = 0;
      mi_col_offset = 0;
      corner_mi = xd->mi[0];
      corner_mbmi = &corner_mi->mbmi;
    }

    backup_mbmi = *corner_mbmi;
    modify_neighbor_predictor_for_obmc(corner_mbmi);

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      struct macroblockd_plane *const pd = &xd->plane[j];
      setup_pred_plane(&pd->dst, BLOCK_8X8, dst_buf[j], MAX_SB_SIZE,
                       MAX_SB_SIZE, dst_stride[j], (i / 2) * (mi_high >> 1),
                       (i % 2) * (mi_wide >> 1), NULL, pd->subsampling_x,
                       pd->subsampling_y);
    }

#if CONFIG_COMPOUND_SINGLEREF
    for (ref = 0; ref < 1 + (is_inter_anyref_comp_mode(corner_mbmi->mode));
         ++ref) {
      const MV_REFERENCE_FRAME frame = has_second_ref(corner_mbmi)
                                           ? corner_mbmi->ref_frame[ref]
                                           : corner_mbmi->ref_frame[0];
#else
    for (ref = 0; ref < 1 + has_second_ref(corner_mbmi); ++ref) {
      const MV_REFERENCE_FRAME frame = corner_mbmi->ref_frame[ref];
#endif
      const RefBuffer *const ref_buf = &cm->frame_refs[frame - LAST_FRAME];
      xd->block_refs[ref] = ref_buf;

      if ((!av1_is_valid_scale(&ref_buf->sf)))
        aom_internal_error(xd->error_info, AOM_CODEC_UNSUP_BITSTREAM,
                           "Reference frame has invalid dimensions");
      av1_setup_pre_planes(xd, ref, ref_buf->buf,
                           mi_row + (i / 2) * (mi_high >> 1),
                           mi_col + (i % 2) * (mi_wide >> 1), &ref_buf->sf);
    }
    // adjust mi boundaries of this block
    xd->mb_to_bottom_edge =
        mb_to_bottom_edge_base + (1 - (i / 2)) * mi_high * MI_SIZE * 4;
    xd->mb_to_top_edge = mb_to_top_edge_base - (i / 2) * mi_high * MI_SIZE * 4;
    xd->mb_to_right_edge =
        mb_to_right_edge_base + (1 - (i % 2)) * mi_wide * MI_SIZE * 4;
    xd->mb_to_left_edge =
        mb_to_left_edge_base - (i % 2) * mi_wide * MI_SIZE * 4;

    mi_x = (mi_col + (i % 2) * mi_wide / 2) << MI_SIZE_LOG2;
    mi_y = (mi_row + (i / 2) * mi_high / 2) << MI_SIZE_LOG2;

    for (j = 0; j < MAX_MB_PLANE; ++j) {
      const struct macroblockd_plane *pd = &xd->plane[j];
      bh = mi_high << MI_SIZE_LOG2 >> (pd->subsampling_x + 1);
      bw = mi_wide << MI_SIZE_LOG2 >> (pd->subsampling_y + 1);
      build_inter_predictors(cm, xd, j, corner_mi, 1, 0, bw, bh, 0, 0, bw, bh,
                             mi_x, mi_y);
    }
    *corner_mbmi = backup_mbmi;
  }
  // restore the boundaries
  xd->mb_to_bottom_edge = mb_to_bottom_edge_base;
  xd->mb_to_top_edge = mb_to_top_edge_base;
  xd->mb_to_right_edge = mb_to_right_edge_base;
  xd->mb_to_left_edge = mb_to_left_edge_base;
}

// get the stitched extra prediction for this block
void av1_get_ext_blk_preds(const AV1_COMMON *cm, MACROBLOCKD *xd, int bsize,
                           int mi_row, int mi_col,
                           uint8_t *dst_buf[][MAX_MB_PLANE],
                           int dst_stride[MAX_MB_PLANE]) {
  get_pred_by_corner_neighbor(cm, xd, bsize, mi_row, mi_col, dst_buf[0],
                              dst_stride);
  get_pred_by_vert_neighbor(cm, xd, bsize, mi_row, mi_col, dst_buf[1],
                            dst_stride);
  get_pred_by_horz_neighbor(cm, xd, bsize, mi_row, mi_col, dst_buf[2],
                            dst_stride);
}

void av1_get_ori_blk_pred(const AV1_COMMON *cm, MACROBLOCKD *xd, int bsize,
                          int mi_row, int mi_col,
                          uint8_t *dst_buf[MAX_MB_PLANE],
                          int dst_stride[MAX_MB_PLANE]) {
  MODE_INFO *const mi = xd->mi[0];
  MB_MODE_INFO *const mbmi = &mi->mbmi;
  int mi_x = mi_col << MI_SIZE_LOG2;
  int mi_y = mi_row << MI_SIZE_LOG2;
  int bw = block_size_wide[bsize];
  int bh = block_size_high[bsize];
  int i, ref;

  for (i = 0; i < MAX_MB_PLANE; ++i) {
    struct macroblockd_plane *const pd = &xd->plane[i];
    setup_pred_plane(&pd->dst, BLOCK_8X8, dst_buf[i], MAX_SB_SIZE, MAX_SB_SIZE,
                     dst_stride[i], 0, 0, NULL, pd->subsampling_x,
                     pd->subsampling_y);
  }

  for (ref = 0; ref < 1 + has_second_ref(mbmi); ++ref) {
    const MV_REFERENCE_FRAME frame = mbmi->ref_frame[ref];
    const RefBuffer *const ref_buf = &cm->frame_refs[frame - LAST_FRAME];
    xd->block_refs[ref] = ref_buf;

    if (!av1_is_valid_scale(&ref_buf->sf))
      aom_internal_error(xd->error_info, AOM_CODEC_UNSUP_BITSTREAM,
                         "Reference frame has invalid dimensions");

    av1_setup_pre_planes(xd, ref, ref_buf->buf, mi_row, mi_col, &ref_buf->sf);
  }

  for (i = 0; i < MAX_MB_PLANE; ++i) {
    const struct macroblockd_plane *pd = &xd->plane[i];
    build_inter_predictors(
        cm, xd, i, mi, 1, 0, bw >> pd->subsampling_x, bh >> pd->subsampling_y,
        0, 0, bw >> pd->subsampling_x, bh >> pd->subsampling_y, mi_x, mi_y);
  }
}

#endif