uvg266/src/filter.c

/*****************************************************************************
 * This file is part of Kvazaar HEVC encoder.
 *
 * Copyright (C) 2013-2015 Tampere University of Technology and others (see
 * COPYING file).
 *
 * Kvazaar is free software: you can redistribute it and/or modify it under
 * the terms of the GNU Lesser General Public License as published by the
 * Free Software Foundation; either version 2.1 of the License, or (at your
 * option) any later version.
 *
 * Kvazaar is distributed in the hope that it will be useful, but WITHOUT ANY
 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 * FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public License for
 * more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with Kvazaar.  If not, see <http://www.gnu.org/licenses/>.
 ****************************************************************************/

#include "filter.h"

#include <stdlib.h>

#include "cu.h"
#include "encoder.h"
#include "kvazaar.h"
#include "transform.h"
#include "videoframe.h"


//////////////////////////////////////////////////////////////////////////
// INITIALIZATIONS
const uint16_t kvz_g_tc_table_8x8[66] =
{
  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,   0,   0,   0,   0,   0,   0,   0,   0,   3,   4,   4,   4,
  4,  5,  5,  5,  5,  7,  7,  8,  9,  10,  10,  11,  13,  14,  15,  17,  19,  21,  24,  25,  29,  33,
  36, 41, 45, 51, 57, 64, 71, 80, 89, 100, 112, 125, 141, 157, 177, 198, 222, 250, 280, 314, 352, 395
};

const uint8_t kvz_g_beta_table_8x8[64] =
{
  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,
  6,  7,  8,  9,  10, 11, 12, 13, 14, 15, 16, 17, 18, 20, 22, 24,
  26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 48, 50, 52, 54, 56,
  58, 60, 62, 64, 66, 68, 70, 72, 74, 76, 78, 80, 82, 84, 86, 88
};

const int8_t kvz_g_luma_filter[4][8] =
{
  {  0, 0,   0, 64,  0,   0, 0,  0 },
  { -1, 4, -10, 58, 17,  -5, 1,  0 },
  { -1, 4, -11, 40, 40, -11, 4, -1 },
  {  0, 1,  -5, 17, 58, -10, 4, -1 }
};

const int8_t kvz_g_chroma_filter[8][4] =
{
  {  0, 64,  0,  0 },
  { -2, 58, 10, -2 },
  { -4, 54, 16, -2 },
  { -6, 46, 28, -4 },
  { -4, 36, 36, -4 },
  { -4, 28, 46, -6 },
  { -2, 16, 54, -4 },
  { -2, 10, 58, -2 }
};

//////////////////////////////////////////////////////////////////////////
// FUNCTIONS

/**
 * \brief Perform in strong luma filtering in place.
 * \param line  line of 8 pixels, with center at index 4
 * \param tc  tc treshold
 * \return  Reach of the filter starting from center.
 */
static INLINE int kvz_filter_deblock_luma_strong(
    kvz_pixel *line,
    int32_t tc)
{
  const kvz_pixel m0 = line[0];
  const kvz_pixel m1 = line[1];
  const kvz_pixel m2 = line[2];
  const kvz_pixel m3 = line[3];
  const kvz_pixel m4 = line[4];
  const kvz_pixel m5 = line[5];
  const kvz_pixel m6 = line[6];
  const kvz_pixel m7 = line[7];
  const uint8_t tcW[3] = { 3, 2, 1 }; //Wheights for tc

  line[1] = CLIP(m1 - tcW[3]*tc, m1 + tcW[3]*tc, (2*m0 + 3*m1 +   m2 +   m3 +   m4 + 4) >> 3);
  line[2] = CLIP(m2 - tcW[1]*tc, m2 + tcW[1]*tc, (  m1 +   m2 +   m3 +   m4        + 2) >> 2);
  line[3] = CLIP(m3 - tcW[0]*tc, m3 + tcW[0]*tc, (  m1 + 2*m2 + 2*m3 + 2*m4 +   m5 + 4) >> 3);
  line[4] = CLIP(m4 - tcW[0]*tc, m4 + tcW[0]*tc, (  m2 + 2*m3 + 2*m4 + 2*m5 +   m6 + 4) >> 3);
  line[5] = CLIP(m5 - tcW[1]*tc, m5 + tcW[1]*tc, (  m3 +   m4 +   m5 +   m6        + 2) >> 2);
  line[6] = CLIP(m6 - tcW[3]*tc, m6 + tcW[3]*tc, (  m3 +   m4 +   m5 + 3*m6 + 2*m7 + 4) >> 3);

  return 3;
}

/**
 * \brief Perform in weak luma filtering in place.
 * \param encoder  Encoder
 * \param line  Line of 8 pixels, with center at index 4
 * \param tc  The tc treshold
 * \param p_2nd  Whether to filter the 2nd line of P
 * \param q_2nd  Whether to filter the 2nd line of Q
 */
static INLINE int kvz_filter_deblock_luma_weak(
    const encoder_control_t * const encoder,
    kvz_pixel *line,
    int32_t tc,
    bool p_2nd,
    bool q_2nd)
{
  const kvz_pixel m1 = line[1];
  const kvz_pixel m2 = line[2];
  const kvz_pixel m3 = line[3];
  const kvz_pixel m4 = line[4];
  const kvz_pixel m5 = line[5];
  const kvz_pixel m6 = line[6];

  int32_t delta = (9 * (m4 - m3) - 3 * (m5 - m2) + 8) >> 4;

  if (abs(delta) >= tc * 10) {
    return 0;
  } else {
    int32_t tc2 = tc >> 1;
    delta = CLIP(-tc, tc, delta);
    line[3] = CLIP(0, (1 << encoder->bitdepth) - 1, (m3 + delta));
    line[4] = CLIP(0, (1 << encoder->bitdepth) - 1, (m4 - delta));

    if (p_2nd) {
      int32_t delta1 = CLIP(-tc2, tc2, (((m1 + m3 + 1) >> 1) - m2 + delta) >> 1);
      line[2] = CLIP(0, (1 << encoder->bitdepth) - 1, m2 + delta1);
    }
    if (q_2nd) {
      int32_t delta2 = CLIP(-tc2, tc2, (((m6 + m4 + 1) >> 1) - m5 - delta) >> 1);
      line[5] = CLIP(0, (1 << encoder->bitdepth) - 1, m5 + delta2);
    }
    
    if (p_2nd || q_2nd) {
      return 2;
    } else {
      return 1;
    }
  }
}

/**
 * \brief
 */
static INLINE void kvz_filter_deblock_chroma(const encoder_control_t * const encoder,
                                             kvz_pixel *src,
                                             int32_t offset,
                                             int32_t tc,
                                             int8_t part_P_nofilter,
                                             int8_t part_Q_nofilter)
{
  int32_t delta;
  int16_t m2 = src[-offset * 2];
  int16_t m3 = src[-offset];
  int16_t m4 = src[0];
  int16_t m5 = src[offset];

  delta = CLIP(-tc,tc, (((m4 - m3) * 4) + m2 - m5 + 4 ) >> 3);
  if(!part_P_nofilter) {
    src[-offset] = CLIP(0, (1 << encoder->bitdepth) - 1, m3 + delta);
  }
  if(!part_Q_nofilter) {
    src[0] = CLIP(0, (1 << encoder->bitdepth) - 1, m4 - delta);
  }
}


/**
 * \brief Check whether an edge is a TU boundary.
 *
 * \param state   encoder state
 * \param x       x-coordinate of the scu in pixels
 * \param y       y-coordinate of the scu in pixels
 * \param dir     direction of the edge to check
 * \return        true, if the edge is a TU boundary, otherwise false
 */
static bool is_tu_boundary(const encoder_state_t *const state,
                           int32_t x,
                           int32_t y,
                           edge_dir dir)
{
  const cu_info_t *const scu =
    kvz_cu_array_at_const(state->tile->frame->cu_array, x, y);
  const int tu_width = LCU_WIDTH >> scu->tr_depth;

  if (dir == EDGE_HOR) {
    return (y & (tu_width - 1)) == 0;
  } else {
    return (x & (tu_width - 1)) == 0;
  }
}


/**
 * \brief Check whether an edge is a PU boundary.
 *
 * \param state   encoder state
 * \param x       x-coordinate of the scu in pixels
 * \param y       y-coordinate of the scu in pixels
 * \param dir     direction of the edge to check
 * \return        true, if the edge is a TU boundary, otherwise false
 */
static bool is_pu_boundary(const encoder_state_t *const state,
                           int32_t x,
                           int32_t y,
                           edge_dir dir)
{
  const cu_info_t *const scu =
    kvz_cu_array_at_const(state->tile->frame->cu_array, x, y);
  // Get the containing CU.
  const int32_t cu_width = LCU_WIDTH >> scu->depth;
  const int32_t x_cu = x & ~(cu_width - 1);
  const int32_t y_cu = y & ~(cu_width - 1);
  const cu_info_t *const cu =
    kvz_cu_array_at_const(state->tile->frame->cu_array, x_cu, y_cu);

  const int num_pu = kvz_part_mode_num_parts[cu->part_size];
  for (int i = 0; i < num_pu; i++) {
    if (dir == EDGE_HOR) {
      int y_pu = PU_GET_Y(cu->part_size, cu_width, y_cu, i);
      if (y_pu == y) return true;

    } else {
      int x_pu = PU_GET_X(cu->part_size, cu_width, x_cu, i);
      if (x_pu == x) return true;
    }
  }

  return false;
}


/**
 * \brief Check whether an edge is aligned on a 8x8 grid.
 *
 * \param x     x-coordinate of the edge
 * \param y     y-coordinate of the edge
 * \param dir   direction of the edge
 * \return      true, if the edge is aligned on a 8x8 grid, otherwise false
 */
static bool is_on_8x8_grid(int x, int y, edge_dir dir)
{
  if (dir == EDGE_HOR) {
    return (y & 7) == 0;
  } else {
    return (x & 7) == 0;
  }
}

static int8_t get_qp_y_pred(const encoder_state_t* state, int x, int y, edge_dir dir)
{
  if (state->encoder_control->max_qp_delta_depth < 0) {
    return state->qp;
  }

  int32_t qp_p;
  if (dir == EDGE_HOR && y > 0) {
    qp_p = kvz_cu_array_at_const(state->tile->frame->cu_array, x, y - 1)->qp;
  } else if (dir == EDGE_VER && x > 0) {
    qp_p = kvz_cu_array_at_const(state->tile->frame->cu_array, x - 1, y)->qp;
  } else {
    // TODO: This seems to be dead code. Investigate.
    qp_p = state->encoder_control->cfg.set_qp_in_cu ? 26 : state->frame->QP;
  }

  const int32_t qp_q =
    kvz_cu_array_at_const(state->tile->frame->cu_array, x, y)->qp;

  return (qp_p + qp_q + 1) >> 1;
}

/**
 * \brief Gather pixels needed for deblocking
 */
static INLINE void gather_deblock_pixels(
    const kvz_pixel *src,
    int step, 
    int stride,
    int reach,
    kvz_pixel *dst)
{
  for (int i = -reach; i < +reach; ++i) {
    dst[i + 4] = src[i * step + stride];
  }
}

/**
* \brief Scatter pixels
*/
static INLINE void scatter_deblock_pixels(
    const kvz_pixel *src,
    int step, 
    int stride,
    int reach,
    kvz_pixel *dst)
{
  for (int i = -reach; i < +reach; ++i) {
    dst[i * step + stride] = src[i + 4];
  }
}

//TODO: Implement
static INLINE int  kvz_filter_deblock_large_block(kvz_pixel *line, kvz_pixel *lineL, const int32_t tc,
                                                  const uint8_t filter_length_P, const uint8_t filter_length_Q)
{
  return 3;
}

/**
* \brief Determine if strong or weak filtering should be used
*/
static INLINE bool use_strong_filtering(const kvz_pixel const* b0, const kvz_pixel const* b3,
                                        const kvz_pixel const* b0L, const kvz_pixel const* b3L,
                                        const int_fast32_t dp0, const int_fast32_t dq0,
                                        const int_fast32_t dp3, const int_fast32_t dq3,
                                        const int32_t tc, const int32_t beta,
                                        const bool is_side_P_large, const bool is_side_Q_large,
                                        const uint8_t max_filter_length_P, const uint8_t max_filter_length_Q)
{
  if (is_side_P_large || is_side_Q_large) {
    int_fast32_t sp0 = abs(b0[0] - b0[3]);
    int_fast32_t sq0 = abs(b0[4] - b0[7]);
    int_fast32_t sp3 = abs(b3[0] - b3[3]);
    int_fast32_t sq3 = abs(b3[4] - b3[7]);
    kvz_pixel tmp0, tmp3;
    if (is_side_P_large) {
      if (max_filter_length_P == 7) {
        tmp0 = b0L[0];
        tmp3 = b3L[0];
        sp0 = sp0 + abs(b0L[3] - b0L[2] - b0L[1] + tmp0);
        sp3 = sp3 + abs(b3L[3] - b3L[2] - b3L[1] + tmp3);
      }
      else {
        tmp0 = b0L[2];
        tmp3 = b3L[2];
      }
      sp0 = (sp0 + abs(b0[0] - tmp0) + 1) >> 1;
      sp3 = (sp3 + abs(b3[0] - tmp3) + 1) >> 1;
    }
    if (is_side_Q_large) {
      if (max_filter_length_Q == 7) {
        tmp0 = b0L[7];
        tmp3 = b3L[7];
        sq0 = sq0 + abs(b0L[4] - b0L[5] - b0L[6] + tmp0);
        sq3 = sq3 + abs(b3L[4] - b3L[5] - b3L[6] + tmp3);
      } else {
        tmp0 = b0L[5];
        tmp3 = b3L[5];
      }
      sq0 = (sq0 + abs(tmp0 - b0[7]) + 1) >> 1;
      sq3 = (sq3 + abs(tmp3 - b3[7]) + 1) >> 1;
    }
    return 2 * (dp0 + dq0) < beta >> 4 &&
           2 * (dp3 + dq3) < beta >> 4 &&
           abs(b0[3] - b0[4]) < (5 * tc + 1) >> 1 &&
           abs(b3[3] - b3[4]) < (5 * tc + 1) >> 1 &&
           sp0 + sq0 < (beta * 3 >> 5) &&
           sp3 + sq3 < (beta * 3 >> 5);
  } 
  else {
    return 2 * (dp0 + dq0) < beta >> 2 &&
           2 * (dp3 + dq3) < beta >> 2 &&
           abs(b0[3] - b0[4]) < (5 * tc + 1) >> 1 &&
           abs(b3[3] - b3[4]) < (5 * tc + 1) >> 1 &&
           abs(b0[0] - b0[3]) + abs(b0[4] - b0[7]) < beta >> 3 &&
           abs(b3[0] - b3[3]) + abs(b3[4] - b3[7]) < beta >> 3;
  }
}

/**
 * \brief Apply the deblocking filter to luma pixels on a single edge.
 *
 * The caller should check that the edge is a TU boundary or a PU boundary.
 *
 \verbatim

         .-- filter this edge if dir == EDGE_HOR
         v
     +--------+
     |o <-- pixel at (x, y)
     |        |
     |<-- filter this edge if dir == EDGE_VER
     |        |
     +--------+

 \endverbatim
 *
 * \param state     encoder state
 * \param x         x-coordinate in pixels (see above)
 * \param y         y-coordinate in pixels (see above)
 * \param length    length of the edge in pixels
 * \param dir       direction of the edge to filter
 * \param tu_boundary   whether the edge is a TU boundary
 */
static void filter_deblock_edge_luma(encoder_state_t * const state,
                                     int32_t x,
                                     int32_t y,
                                     int32_t length,
                                     edge_dir dir,
                                     bool tu_boundary)
{
  videoframe_t * const frame = state->tile->frame;
  const encoder_control_t * const encoder = state->encoder_control;

  {
    int32_t stride = frame->rec->stride;
    int32_t beta_offset_div2 = encoder->cfg.deblock_beta;
    int32_t tc_offset_div2   = encoder->cfg.deblock_tc;
    // TODO: support 10+bits
    kvz_pixel *orig_src = &frame->rec->y[x + y*stride];
    kvz_pixel *src = orig_src;

    const int32_t qp = get_qp_y_pred(state, x, y, dir);

    const int MAX_QP = 63; //TODO: Make DEFAULT_INTRA_TC_OFFSET(=2) a define?
    const int8_t lumaBitdepth = encoder->bitdepth;

    int8_t strength = 0;
    int32_t bitdepth_scale  = 1 << (lumaBitdepth - 8);
    int32_t b_index         = CLIP(0, MAX_QP, qp + (beta_offset_div2 << 1));
    int32_t beta            = kvz_g_beta_table_8x8[b_index] * bitdepth_scale;
    int32_t side_threshold  = (beta + (beta >>1 )) >> 3;
    int32_t tc_index;
    int32_t tc;

    //Deblock adapted to halve pixel mvd. TODO: Tie into actual number of fractional mv bits
    const int16_t mvdThreashold = 2; //(1 << (MV_INTERNAL_FRACTIONAL_BITS - 1))

    uint32_t num_4px_parts  = length / 4;

    // Transpose the image by swapping x and y strides when doing horizontal
    // edges.
    const int32_t x_stride = (dir == EDGE_VER) ? 1 : stride;
    const int32_t y_stride = (dir == EDGE_VER) ? stride : 1;

    // TODO: add CU based QP calculation

    // For each 4-pixel part in the edge
    for (uint32_t block_idx = 0; block_idx < num_4px_parts; ++block_idx) {
      {
        // CUs on both sides of the edge
        cu_info_t *cu_p;
        cu_info_t *cu_q;
        if (dir == EDGE_VER) {
          int32_t y_coord = y + 4 * block_idx;
          cu_p = kvz_cu_array_at(frame->cu_array, x - 1, y_coord);
          cu_q = kvz_cu_array_at(frame->cu_array, x,     y_coord);

        } else {
          int32_t x_coord = x + 4 * block_idx;
          cu_p = kvz_cu_array_at(frame->cu_array, x_coord, y - 1);
          cu_q = kvz_cu_array_at(frame->cu_array, x_coord, y    );
        }

        bool nonzero_coeffs = cbf_is_set(cu_q->cbf, cu_q->tr_depth, COLOR_Y)
                           || cbf_is_set(cu_p->cbf, cu_p->tr_depth, COLOR_Y);

        // Filter strength
        strength = 0;
        if (cu_q->type == CU_INTRA || cu_p->type == CU_INTRA) {
          strength = 2;
        } else if (tu_boundary && nonzero_coeffs) {
          // Non-zero residual/coeffs and transform boundary
          // Neither CU is intra so tr_depth <= MAX_DEPTH.
          strength = 1;       
        } else if (cu_p->inter.mv_dir != 3 && cu_q->inter.mv_dir != 3 &&
                 ((abs(cu_q->inter.mv[cu_q->inter.mv_dir - 1][0] - cu_p->inter.mv[cu_p->inter.mv_dir - 1][0]) >= mvdThreashold) ||
                  (abs(cu_q->inter.mv[cu_q->inter.mv_dir - 1][1] - cu_p->inter.mv[cu_p->inter.mv_dir - 1][1]) >= mvdThreashold))) {
          // Absolute motion vector diff between blocks >= 0.5 (Integer pixel)
          strength = 1;
        } else if (cu_p->inter.mv_dir != 3 && cu_q->inter.mv_dir != 3 &&
                   cu_q->inter.mv_ref[cu_q->inter.mv_dir - 1] != cu_p->inter.mv_ref[cu_p->inter.mv_dir - 1]) {
          strength = 1;
        }
        
        // B-slice related checks
        if(!strength && state->frame->slicetype == KVZ_SLICE_B) {

          // Zero all undefined motion vectors for easier usage
          if(!(cu_q->inter.mv_dir & 1)) {
            cu_q->inter.mv[0][0] = 0;
            cu_q->inter.mv[0][1] = 0;
          }
          if(!(cu_q->inter.mv_dir & 2)) {
            cu_q->inter.mv[1][0] = 0;
            cu_q->inter.mv[1][1] = 0;
          }

          if(!(cu_p->inter.mv_dir & 1)) {
            cu_p->inter.mv[0][0] = 0;
            cu_p->inter.mv[0][1] = 0;
          }
          if(!(cu_p->inter.mv_dir & 2)) {
            cu_p->inter.mv[1][0] = 0;
            cu_p->inter.mv[1][1] = 0;
          }
          const int refP0 = (cu_p->inter.mv_dir & 1) ? state->frame->ref_LX[0][cu_p->inter.mv_ref[0]] : -1;
          const int refP1 = (cu_p->inter.mv_dir & 2) ? state->frame->ref_LX[1][cu_p->inter.mv_ref[1]] : -1;
          const int refQ0 = (cu_q->inter.mv_dir & 1) ? state->frame->ref_LX[0][cu_q->inter.mv_ref[0]] : -1;
          const int refQ1 = (cu_q->inter.mv_dir & 2) ? state->frame->ref_LX[1][cu_q->inter.mv_ref[1]] : -1;
          const int16_t* mvQ0 = cu_q->inter.mv[0];
          const int16_t* mvQ1 = cu_q->inter.mv[1];

          const int16_t* mvP0 = cu_p->inter.mv[0];
          const int16_t* mvP1 = cu_p->inter.mv[1];

          if(( refP0 == refQ0 &&  refP1 == refQ1 ) || ( refP0 == refQ1 && refP1==refQ0 ))
          {
            // Different L0 & L1
            if ( refP0 != refP1 ) {          
              if ( refP0 == refQ0 ) {
                strength  = ((abs(mvQ0[0] - mvP0[0]) >= mvdThreashold) ||
                             (abs(mvQ0[1] - mvP0[1]) >= mvdThreashold) ||
                             (abs(mvQ1[0] - mvP1[0]) >= mvdThreashold) ||
                             (abs(mvQ1[1] - mvP1[1]) >= mvdThreashold)) ? 1 : 0;
              } else {
                strength  = ((abs(mvQ1[0] - mvP0[0]) >= mvdThreashold) ||
                             (abs(mvQ1[1] - mvP0[1]) >= mvdThreashold) ||
                             (abs(mvQ0[0] - mvP1[0]) >= mvdThreashold) ||
                             (abs(mvQ0[1] - mvP1[1]) >= mvdThreashold)) ? 1 : 0;
              }
            // Same L0 & L1
            } else {  
              strength  = ((abs(mvQ0[0] - mvP0[0]) >= mvdThreashold) ||
                           (abs(mvQ0[1] - mvP0[1]) >= mvdThreashold) ||
                           (abs(mvQ1[0] - mvP1[0]) >= mvdThreashold) ||
                           (abs(mvQ1[1] - mvP1[1]) >= mvdThreashold)) &&
                          ((abs(mvQ1[0] - mvP0[0]) >= mvdThreashold) ||
                           (abs(mvQ1[1] - mvP0[1]) >= mvdThreashold) ||
                           (abs(mvQ0[0] - mvP1[0]) >= mvdThreashold) ||
                           (abs(mvQ0[1] - mvP1[1]) >= mvdThreashold)) ? 1 : 0;
            }
          } else {
            strength = 1;
          }
        }

        tc_index        = CLIP(0, MAX_QP + 2, (int32_t)(qp + 2*(strength - 1) + (tc_offset_div2 << 1)));
        tc              = lumaBitdepth < 10 ? ((kvz_g_tc_table_8x8[tc_index] + (1 << (9 - lumaBitdepth))) >> (10 - lumaBitdepth))
                                            : ((kvz_g_tc_table_8x8[tc_index] << (lumaBitdepth - 10)));
      }

      if (strength == 0) continue;

      bool is_side_P_large = false;
      bool is_side_Q_large = false;
      uint8_t max_filter_length_P = 7; //TODO: Get actual max filter length
      uint8_t max_filter_length_Q = 7;

      if (max_filter_length_P > 3) {
        is_side_P_large = dir == EDGE_HOR && y % LCU_WIDTH == 0 ? false : true;
        //TODO: Add affine/ATMVP related stuff
        /*if (max_filter_length_P > 5 && cu_p->affine) {
          max_filter_length_P = MIN(max_filter_length_P, 5);
        }*/
      }
      if (max_filter_length_Q > 3) {
        is_side_Q_large = true;
      }
      //                   +-- edge_src
      //                   v
      // line0 p3 p2 p1 p0 q0 q1 q2 q3
      kvz_pixel *edge_src = &src[block_idx * 4 * y_stride];

      // Gather the lines of pixels required for the filter on/off decision.
      //TODO: May need to limit reach in small blocks?
      kvz_pixel b[4][8];
      gather_deblock_pixels(edge_src, x_stride, 0 * y_stride, 4, &b[0][0]);
      gather_deblock_pixels(edge_src, x_stride, 3 * y_stride, 4, &b[3][0]);

      int_fast32_t dp0 = abs(b[0][1] - 2 * b[0][2] + b[0][3]);
      int_fast32_t dq0 = abs(b[0][4] - 2 * b[0][5] + b[0][6]);
      int_fast32_t dp3 = abs(b[3][1] - 2 * b[3][2] + b[3][3]);
      int_fast32_t dq3 = abs(b[3][4] - 2 * b[3][5] + b[3][6]);
      int_fast32_t dp = dp0 + dp3;
      int_fast32_t dq = dq0 + dq3;

      bool sw = false;

      if (is_side_P_large || is_side_Q_large) {
        int_fast32_t dp0L = dp0;
        int_fast32_t dq0L = dq0;
        int_fast32_t dp3L = dp3;
        int_fast32_t dq3L = dq3;
        
        //bL:
        //line0 p7 p6 p5 p4 q4 q5 q6 q7
        kvz_pixel bL[4][8];

        if (is_side_P_large) {
          gather_deblock_pixels(edge_src - 6 * x_stride, x_stride, 0 * y_stride, 2, &bL[0][0]);
          gather_deblock_pixels(edge_src - 6 * x_stride, x_stride, 3 * y_stride, 2, &bL[3][0]);
          dp0L = (dp0L + abs(bL[0][1] - 2 * bL[0][2] + bL[0][3]) + 1) >> 1;
          dp3L = (dp3L + abs(bL[3][1] - 2 * bL[3][2] + bL[3][3]) + 1) >> 1;
        }
        if (is_side_Q_large) {
          gather_deblock_pixels(edge_src + 6 * x_stride, x_stride, 0 * y_stride, 2, &bL[0][4]);
          gather_deblock_pixels(edge_src + 6 * x_stride, x_stride, 3 * y_stride, 2, &bL[3][4]);
          dq0L = (dq0L + abs(bL[0][4] - 2 * bL[0][5] + bL[0][6]) + 1) >> 1;
          dq3L = (dq3L + abs(bL[3][4] - 2 * bL[3][5] + bL[3][6]) + 1) >> 1;
        }
        
        int_fast32_t dpL = dp0L + dp3L;
        int_fast32_t dqL = dq0L + dq3L;

        if (dpL + dqL < beta) {
          sw = use_strong_filtering(&b[0][0], &b[3][0], &bL[0][0], &bL[3][0],
                                    dp0L, dq0L, dp3L, dq3L, tc, beta,
                                    is_side_P_large, is_side_Q_large,
                                    max_filter_length_P, max_filter_length_Q);
          if (sw) {
            gather_deblock_pixels(edge_src, x_stride, 1 * y_stride, 4, &b[1][0]);
            gather_deblock_pixels(edge_src, x_stride, 2 * y_stride, 4, &b[2][0]);
            if (is_side_P_large)
            {
              gather_deblock_pixels(edge_src - 6 * x_stride, x_stride, 1 * y_stride, 2, &bL[1][0]);
              gather_deblock_pixels(edge_src - 6 * x_stride, x_stride, 2 * y_stride, 2, &bL[2][0]);
            }
            if (is_side_Q_large)
            {
              gather_deblock_pixels(edge_src + 6 * x_stride, x_stride, 1 * y_stride, 2, &bL[1][4]);
              gather_deblock_pixels(edge_src + 6 * x_stride, x_stride, 2 * y_stride, 2, &bL[2][4]);
            }

            for (int i = 0; i < 4; ++i) {
              int filter_reach;
              filter_reach = kvz_filter_deblock_large_block(&b[i][0], &bL[i][0], tc,
                                                            is_side_P_large ? max_filter_length_P : 3, 
                                                            is_side_Q_large ? max_filter_length_Q : 3);
              scatter_deblock_pixels(&b[i][0], x_stride, i * y_stride, filter_reach, edge_src);
              if (is_side_P_large) {
                const int diff_reach = (max_filter_length_P - filter_reach) >> 1;
                scatter_deblock_pixels(&bL[i][0] - diff_reach, x_stride, i * y_stride, diff_reach, edge_src - filter_reach - diff_reach);
              }
              if (is_side_Q_large) {
                const int diff_reach = (max_filter_length_P - filter_reach) >> 1;
                scatter_deblock_pixels(&bL[i][0] + diff_reach, x_stride, i * y_stride, diff_reach, edge_src + filter_reach + diff_reach);
              }
            }
          } 
          else {
            is_side_P_large = false;
            is_side_Q_large = false;
          }
        }
      }

      if (!sw)
      {
        if (dp + dq < beta) {
          if (max_filter_length_P > 2 && max_filter_length_Q > 2) {
            // Strong filtering flag checking.
            sw = use_strong_filtering(b[0], b[3], NULL, NULL,
                                      dp0, dq0, dp3, dq3, tc, beta,
                                      is_side_P_large, is_side_Q_large,
                                      max_filter_length_P, max_filter_length_Q);
          }

          // Read lines 1 and 2. Weak filtering doesn't use the outermost pixels
          // but let's give them anyway to simplify control flow.
          gather_deblock_pixels(edge_src, x_stride, 1 * y_stride, 4, &b[1][0]);
          gather_deblock_pixels(edge_src, x_stride, 2 * y_stride, 4, &b[2][0]);

          for (int i = 0; i < 4; ++i) {
            int filter_reach;
            if (sw) {
              filter_reach = kvz_filter_deblock_luma_strong(&b[i][0], tc);
            } else {
              bool p_2nd = false;
              bool q_2nd = false;
              if (max_filter_length_P > 1 && max_filter_length_Q > 1) {
                p_2nd = dp < side_threshold;
                q_2nd = dq < side_threshold;
              }
              filter_reach = kvz_filter_deblock_luma_weak(encoder, &b[i][0], tc, p_2nd, q_2nd);
            }
            scatter_deblock_pixels(&b[i][0], x_stride, i * y_stride, filter_reach, edge_src);
          }
        }
      }
    }
  }
}

/**
 * \brief Apply the deblocking filter to chroma pixels on a single edge.
 *
 * The caller should check that the edge is a TU boundary or a PU boundary.
 *
 \verbatim

         .-- filter this edge if dir == EDGE_HOR
         v
     +--------+
     |o <-- pixel at (x, y)
     |        |
     |<-- filter this edge if dir == EDGE_VER
     |        |
     +--------+

 \endverbatim
 *
 * \param state         encoder state
 * \param x             x-coordinate in chroma pixels (see above)
 * \param y             y-coordinate in chroma pixels (see above)
 * \param length        length of the edge in chroma pixels
 * \param dir           direction of the edge to filter
 * \param tu_boundary   whether the edge is a TU boundary
 */
static void filter_deblock_edge_chroma(encoder_state_t * const state,
                                       int32_t x,
                                       int32_t y,
                                       int32_t length,
                                       edge_dir dir,
                                       bool tu_boundary)
{
  const encoder_control_t * const encoder = state->encoder_control;
  const videoframe_t * const frame = state->tile->frame;

  // For each subpart
  {
    int32_t stride = frame->rec->stride >> 1;
    int32_t tc_offset_div2 = encoder->cfg.deblock_tc;
    // TODO: support 10+bits
    kvz_pixel *src[] = {
      &frame->rec->u[x + y*stride],
      &frame->rec->v[x + y*stride],
    };
    int8_t strength = 2;

    const int32_t luma_qp  = get_qp_y_pred(state, x << 1, y << 1, dir);
    int32_t QP             = kvz_g_chroma_scale[luma_qp];
    int32_t bitdepth_scale = 1 << (encoder->bitdepth-8);
    int32_t TC_index       = CLIP(0, 51+2, (int32_t)(QP + 2*(strength-1) + (tc_offset_div2 << 1)));
    int32_t Tc             = kvz_g_tc_table_8x8[TC_index]*bitdepth_scale;

    const uint32_t num_4px_parts = length / 4;

    const int32_t offset = (dir == EDGE_HOR) ? stride :      1;
    const int32_t step   = (dir == EDGE_HOR) ?      1 : stride;

    for (uint32_t blk_idx = 0; blk_idx < num_4px_parts; ++blk_idx)
    {
      // CUs on both sides of the edge
      cu_info_t *cu_p;
      cu_info_t *cu_q;
      if (dir == EDGE_VER) {
        int32_t y_coord = (y + 4 * blk_idx) << 1;
        cu_p = kvz_cu_array_at(frame->cu_array, (x - 1) << 1, y_coord);
        cu_q = kvz_cu_array_at(frame->cu_array,  x      << 1, y_coord);

      } else {
        int32_t x_coord = (x + 4 * blk_idx) << 1;
        cu_p = kvz_cu_array_at(frame->cu_array, x_coord, (y - 1) << 1);
        cu_q = kvz_cu_array_at(frame->cu_array, x_coord, (y    ) << 1);
      }

      // Only filter when strenght == 2 (one of the blocks is intra coded)
      if (cu_q->type == CU_INTRA || cu_p->type == CU_INTRA) {
        for (int component = 0; component < 2; component++) {
          for (int i = 0; i < 4; i++) {
            kvz_filter_deblock_chroma(encoder, src[component] + step * (4*blk_idx + i), offset, Tc, 0, 0);
          }
        }
      }
    }
  }
}


/**
 * \brief Filter edge of a single PU or TU
 *
 * \param state     encoder state
 * \param x         block x-position in pixels
 * \param y         block y-position in pixels
 * \param width     block width in pixels
 * \param height    block height in pixels
 * \param dir       direction of the edges to filter
 * \param tu_boundary   whether the edge is a TU boundary
 */
static void filter_deblock_unit(encoder_state_t * const state,
                                int x,
                                int y,
                                int width,
                                int height,
                                edge_dir dir,
                                bool tu_boundary)
{
  // no filtering on borders (where filter would use pixels outside the picture)
  if (x == 0 && dir == EDGE_VER) return;
  if (y == 0 && dir == EDGE_HOR) return;

  // Length of luma and chroma edges.
  int32_t length;
  int32_t length_c;

  if (dir == EDGE_HOR) {
    const videoframe_t * const frame = state->tile->frame;
    const int32_t x_right             = x + width;
    const bool rightmost_4px_of_lcu   = x_right % LCU_WIDTH == 0;
    const bool rightmost_4px_of_frame = x_right == frame->width;

    if (rightmost_4px_of_lcu && !rightmost_4px_of_frame) {
      // The last 4 pixels will be deblocked when processing the next LCU.
      length   = width - 4;
      length_c = (width >> 1) - 4;

    } else {
      length   = width;
      length_c = width >> 1;
    }
  } else {
    length   = height;
    length_c = height >> 1;
  }

  filter_deblock_edge_luma(state, x, y, length, dir, tu_boundary);

  // Chroma pixel coordinates.
  const int32_t x_c = x >> 1;
  const int32_t y_c = y >> 1;
  if (state->encoder_control->chroma_format != KVZ_CSP_400 && is_on_8x8_grid(x_c, y_c, dir)) {
    filter_deblock_edge_chroma(state, x_c, y_c, length_c, dir, tu_boundary);
  }
}


/**
 * \brief Deblock PU and TU boundaries inside an LCU.
 *
 * \param state     encoder state
 * \param x_px      block x-position in pixels
 * \param y_px      block y-position in pixels
 * \param dir       direction of the edges to filter
 *
 * Recursively traverse the CU/TU quadtree. At the lowest level, apply the
 * deblocking filter to the left edge (when dir == EDGE_VER) or the top edge
 * (when dir == EDGE_HOR) as needed. Both luma and chroma are filtered.
 */
static void filter_deblock_lcu_inside(encoder_state_t * const state,
                                      int32_t x,
                                      int32_t y,
                                      edge_dir dir)
{
  const int end_x = MIN(x + LCU_WIDTH, state->tile->frame->width);
  const int end_y = MIN(y + LCU_WIDTH, state->tile->frame->height);

  for (int edge_y = y; edge_y < end_y; edge_y += 8) {
    for (int edge_x = x; edge_x < end_x; edge_x += 8) {
      bool tu_boundary = is_tu_boundary(state, edge_x, edge_y, dir);
      if (tu_boundary || is_pu_boundary(state, edge_x, edge_y, dir)) {
        filter_deblock_unit(state, edge_x, edge_y, 8, 8, dir, tu_boundary);
      }
    }
  }
}


/**
 * \brief Filter rightmost 4 pixels of the horizontal egdes of an LCU.
 *
 * \param state     encoder state
 * \param x_px      x-coordinate of the *right* edge of the LCU in pixels
 * \param y_px      y-coordinate of the top edge of the LCU in pixels
 */
static void filter_deblock_lcu_rightmost(encoder_state_t * const state,
                                         int32_t x_px,
                                         int32_t y_px)
{
  // Luma
  const int x = x_px - 4;
  const int end = MIN(y_px + LCU_WIDTH, state->tile->frame->height);
  for (int y = y_px; y < end; y += 8) {
    // The top edge of the whole frame is not filtered.
    bool tu_boundary = is_tu_boundary(state, x, y, EDGE_HOR);
    bool pu_boundary = is_pu_boundary(state, x, y, EDGE_HOR);
    if (y > 0 && (tu_boundary || pu_boundary)) {
      filter_deblock_edge_luma(state, x, y, 4, EDGE_HOR, tu_boundary);
    }
  }

  // Chroma
  if (state->encoder_control->chroma_format != KVZ_CSP_400) {
    const int x_px_c = x_px >> 1;
    const int y_px_c = y_px >> 1;
    const int x_c = x_px_c - 4;
    const int end_c = MIN(y_px_c + LCU_WIDTH_C, state->tile->frame->height >> 1);
    for (int y_c = y_px_c; y_c < end_c; y_c += 8) {
      // The top edge of the whole frame is not filtered.
      bool tu_boundary = is_tu_boundary(state, x_c << 1, y_c << 1, EDGE_HOR);
      bool pu_boundary = is_pu_boundary(state, x_c << 1, y_c << 1, EDGE_HOR);
      if (y_c > 0 && (tu_boundary || pu_boundary)) {
        filter_deblock_edge_chroma(state, x_c, y_c, 4, EDGE_HOR, tu_boundary);
      }
    }
  }
}


/**
 * \brief Deblock a single LCU without using data from right or down.
 *
 * Filter the following vertical edges (horizontal filtering):
 *  1. The left edge of the LCU.
 *  2. All vertical edges within the LCU.
 *
 * Filter the following horizontal edges (vertical filtering):
 *  1. The rightmost 4 pixels of the top edge of the LCU to the left.
 *  2. The rightmost 4 pixels of all horizontal edges within the LCU to the
 *     left.
 *  3. The top edge and all horizontal edges within the LCU, excluding the
 *     rightmost 4 pixels. If the LCU is the rightmost LCU of the frame, the
 *     last 4 pixels are also filtered.
 *
 * What is not filtered:
 *  - The rightmost 4 pixels of the top edge and all horizontal edges within
 *    the LCU, unless the LCU is the rightmost LCU of the frame.
 *  - The bottom edge of the LCU.
 *  - The right edge of the LCU.
 *
 * \param state   encoder state
 * \param x_px    x-coordinate of the left edge of the LCU in pixels
 * \param y_px    y-coordinate of the top edge of the LCU in pixels
 */
 //TODO: Things to check/fix for VVC:
// - Strength calculation to include average Luma level (Luma Adaptive Deblocing Filter LADF) (optional)
// - Stronger Luma and chroma filters (i.e. large block filtering)
// - Deblocking strength for CIIP and IBC modes (CIIP/IBC not currently used)
// - Handle new prediction modes (i.e. PLT) (PLT not currently used)
// - Luma deblocking on a 4x4 grid
// - Deblocking filter for subblock boundaries
// - Account for small blocks (in chroma) and max filter lengths
// - Allow loop filtering across slice/tile boundaries?
void kvz_filter_deblock_lcu(encoder_state_t * const state, int x_px, int y_px)
{
  assert(!state->encoder_control->cfg.lossless);

  filter_deblock_lcu_inside(state, x_px, y_px, EDGE_VER);
  if (x_px > 0) {
    filter_deblock_lcu_rightmost(state, x_px, y_px);
  }
  filter_deblock_lcu_inside(state, x_px, y_px, EDGE_HOR);
}