vp9_rdopt.c

/*
 *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
 *
 *  Use of this source code is governed by a BSD-style license
 *  that can be found in the LICENSE file in the root of the source
 *  tree. An additional intellectual property rights grant can be found
 *  in the file PATENTS.  All contributing project authors may
 *  be found in the AUTHORS file in the root of the source tree.
 */


#include <stdio.h>
#include <math.h>
#include <limits.h>
#include <assert.h>
#include "vp9/common/vp9_pragmas.h"

#include "vp9/encoder/vp9_tokenize.h"
#include "vp9/encoder/vp9_treewriter.h"
#include "vp9/encoder/vp9_onyx_int.h"
#include "vp9/encoder/vp9_modecosts.h"
#include "vp9/encoder/vp9_encodeintra.h"
#include "vp9/common/vp9_entropymode.h"
#include "vp9/common/vp9_reconinter.h"
#include "vp9/common/vp9_reconintra.h"
#include "vp9/common/vp9_findnearmv.h"
#include "vp9/common/vp9_quant_common.h"
#include "vp9/encoder/vp9_encodemb.h"
#include "vp9/encoder/vp9_quantize.h"
#include "vp9/encoder/vp9_variance.h"
#include "vp9/encoder/vp9_mcomp.h"
#include "vp9/encoder/vp9_rdopt.h"
#include "vp9/encoder/vp9_ratectrl.h"
#include "vpx_mem/vpx_mem.h"
#include "vp9/common/vp9_systemdependent.h"
#include "vp9/encoder/vp9_encodemv.h"

#include "vp9/common/vp9_seg_common.h"
#include "vp9/common/vp9_pred_common.h"
#include "vp9/common/vp9_entropy.h"
#include "vp9_rtcd.h"
#include "vp9/common/vp9_mvref_common.h"
#include "vp9/common/vp9_common.h"

#define MAXF(a,b)            (((a) > (b)) ? (a) : (b))

#define INVALID_MV 0x80008000

/* Factor to weigh the rate for switchable interp filters */
#define SWITCHABLE_INTERP_RATE_FACTOR 1

static const int auto_speed_thresh[17] = {
  1000,
  200,
  150,
  130,
  150,
  125,
  120,
  115,
  115,
  115,
  115,
  115,
  115,
  115,
  115,
  115,
  105
};

const MODE_DEFINITION vp9_mode_order[MAX_MODES] = {
  {ZEROMV,    LAST_FRAME,   NONE},
  {DC_PRED,   INTRA_FRAME,  NONE},

  {NEARESTMV, LAST_FRAME,   NONE},
  {NEARMV,    LAST_FRAME,   NONE},

  {ZEROMV,    GOLDEN_FRAME, NONE},
  {NEARESTMV, GOLDEN_FRAME, NONE},

  {ZEROMV,    ALTREF_FRAME, NONE},
  {NEARESTMV, ALTREF_FRAME, NONE},

  {NEARMV,    GOLDEN_FRAME, NONE},
  {NEARMV,    ALTREF_FRAME, NONE},

  {V_PRED,    INTRA_FRAME,  NONE},
  {H_PRED,    INTRA_FRAME,  NONE},
  {D45_PRED,  INTRA_FRAME,  NONE},
  {D135_PRED, INTRA_FRAME,  NONE},
  {D117_PRED, INTRA_FRAME,  NONE},
  {D153_PRED, INTRA_FRAME,  NONE},
  {D27_PRED,  INTRA_FRAME,  NONE},
  {D63_PRED,  INTRA_FRAME,  NONE},

  {TM_PRED,   INTRA_FRAME,  NONE},

  {NEWMV,     LAST_FRAME,   NONE},
  {NEWMV,     GOLDEN_FRAME, NONE},
  {NEWMV,     ALTREF_FRAME, NONE},

  {SPLITMV,   LAST_FRAME,   NONE},
  {SPLITMV,   GOLDEN_FRAME, NONE},
  {SPLITMV,   ALTREF_FRAME, NONE},

  {I4X4_PRED,    INTRA_FRAME,  NONE},
  {I8X8_PRED, INTRA_FRAME,  NONE},

  /* compound prediction modes */
  {ZEROMV,    LAST_FRAME,   GOLDEN_FRAME},
  {NEARESTMV, LAST_FRAME,   GOLDEN_FRAME},
  {NEARMV,    LAST_FRAME,   GOLDEN_FRAME},

  {ZEROMV,    ALTREF_FRAME, LAST_FRAME},
  {NEARESTMV, ALTREF_FRAME, LAST_FRAME},
  {NEARMV,    ALTREF_FRAME, LAST_FRAME},

  {ZEROMV,    GOLDEN_FRAME, ALTREF_FRAME},
  {NEARESTMV, GOLDEN_FRAME, ALTREF_FRAME},
  {NEARMV,    GOLDEN_FRAME, ALTREF_FRAME},

  {NEWMV,     LAST_FRAME,   GOLDEN_FRAME},
  {NEWMV,     ALTREF_FRAME, LAST_FRAME  },
  {NEWMV,     GOLDEN_FRAME, ALTREF_FRAME},

  {SPLITMV,   LAST_FRAME,   GOLDEN_FRAME},
  {SPLITMV,   ALTREF_FRAME, LAST_FRAME  },
  {SPLITMV,   GOLDEN_FRAME, ALTREF_FRAME},

#if CONFIG_COMP_INTERINTRA_PRED
  /* compound inter-intra prediction */
  {ZEROMV,    LAST_FRAME,   INTRA_FRAME},
  {NEARESTMV, LAST_FRAME,   INTRA_FRAME},
  {NEARMV,    LAST_FRAME,   INTRA_FRAME},
  {NEWMV,     LAST_FRAME,   INTRA_FRAME},

  {ZEROMV,    GOLDEN_FRAME,   INTRA_FRAME},
  {NEARESTMV, GOLDEN_FRAME,   INTRA_FRAME},
  {NEARMV,    GOLDEN_FRAME,   INTRA_FRAME},
  {NEWMV,     GOLDEN_FRAME,   INTRA_FRAME},

  {ZEROMV,    ALTREF_FRAME,   INTRA_FRAME},
  {NEARESTMV, ALTREF_FRAME,   INTRA_FRAME},
  {NEARMV,    ALTREF_FRAME,   INTRA_FRAME},
  {NEWMV,     ALTREF_FRAME,   INTRA_FRAME},
#endif
};

static void fill_token_costs(vp9_coeff_count *c,
                             vp9_coeff_probs *p,
                             TX_SIZE tx_size) {
  int i, j, k, l;

  for (i = 0; i < BLOCK_TYPES; i++)
    for (j = 0; j < REF_TYPES; j++)
      for (k = 0; k < COEF_BANDS; k++)
        for (l = 0; l < PREV_COEF_CONTEXTS; l++) {
          vp9_cost_tokens_skip((int *)(c[i][j][k][l]),
                               p[i][j][k][l],
                               vp9_coef_tree);
        }
}

static int rd_iifactor[32] =  { 4, 4, 3, 2, 1, 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, };

// 3* dc_qlookup[Q]*dc_qlookup[Q];

/* values are now correlated to quantizer */
static int sad_per_bit16lut[QINDEX_RANGE];
static int sad_per_bit4lut[QINDEX_RANGE];

void vp9_init_me_luts() {
  int i;

  // Initialize the sad lut tables using a formulaic calculation for now
  // This is to make it easier to resolve the impact of experimental changes
  // to the quantizer tables.
  for (i = 0; i < QINDEX_RANGE; i++) {
    sad_per_bit16lut[i] =
      (int)((0.0418 * vp9_convert_qindex_to_q(i)) + 2.4107);
    sad_per_bit4lut[i] = (int)((0.063 * vp9_convert_qindex_to_q(i)) + 2.742);
  }
}

static int compute_rd_mult(int qindex) {
  const int q = vp9_dc_quant(qindex, 0);
  return (11 * q * q) >> 2;
}

void vp9_initialize_me_consts(VP9_COMP *cpi, int qindex) {
  cpi->mb.sadperbit16 = sad_per_bit16lut[qindex];
  cpi->mb.sadperbit4 = sad_per_bit4lut[qindex];
}


void vp9_initialize_rd_consts(VP9_COMP *cpi, int qindex) {
  int q, i;

  vp9_clear_system_state();  // __asm emms;

  // Further tests required to see if optimum is different
  // for key frames, golden frames and arf frames.
  // if (cpi->common.refresh_golden_frame ||
  //     cpi->common.refresh_alt_ref_frame)
  qindex = (qindex < 0) ? 0 : ((qindex > MAXQ) ? MAXQ : qindex);

  cpi->RDMULT = compute_rd_mult(qindex);
  if (cpi->pass == 2 && (cpi->common.frame_type != KEY_FRAME)) {
    if (cpi->twopass.next_iiratio > 31)
      cpi->RDMULT += (cpi->RDMULT * rd_iifactor[31]) >> 4;
    else
      cpi->RDMULT +=
          (cpi->RDMULT * rd_iifactor[cpi->twopass.next_iiratio]) >> 4;
  }
  cpi->mb.errorperbit = cpi->RDMULT >> 6;
  cpi->mb.errorperbit += (cpi->mb.errorperbit == 0);

  vp9_set_speed_features(cpi);

  q = (int)pow(vp9_dc_quant(qindex, 0) >> 2, 1.25);
  q <<= 2;
  if (q < 8)
    q = 8;

  if (cpi->RDMULT > 1000) {
    cpi->RDDIV = 1;
    cpi->RDMULT /= 100;

    for (i = 0; i < MAX_MODES; i++) {
      if (cpi->sf.thresh_mult[i] < INT_MAX) {
        cpi->rd_threshes[i] = cpi->sf.thresh_mult[i] * q / 100;
      } else {
        cpi->rd_threshes[i] = INT_MAX;
      }
      cpi->rd_baseline_thresh[i] = cpi->rd_threshes[i];
    }
  } else {
    cpi->RDDIV = 100;

    for (i = 0; i < MAX_MODES; i++) {
      if (cpi->sf.thresh_mult[i] < (INT_MAX / q)) {
        cpi->rd_threshes[i] = cpi->sf.thresh_mult[i] * q;
      } else {
        cpi->rd_threshes[i] = INT_MAX;
      }
      cpi->rd_baseline_thresh[i] = cpi->rd_threshes[i];
    }
  }

  fill_token_costs(cpi->mb.token_costs[TX_4X4],
                   cpi->common.fc.coef_probs_4x4, TX_4X4);
  fill_token_costs(cpi->mb.token_costs[TX_8X8],
                   cpi->common.fc.coef_probs_8x8, TX_8X8);
  fill_token_costs(cpi->mb.token_costs[TX_16X16],
                   cpi->common.fc.coef_probs_16x16, TX_16X16);
  fill_token_costs(cpi->mb.token_costs[TX_32X32],
                   cpi->common.fc.coef_probs_32x32, TX_32X32);

  for (i = 0; i < NUM_PARTITION_CONTEXTS; i++)
    vp9_cost_tokens(cpi->mb.partition_cost[i],
                    cpi->common.fc.partition_prob[i],
                    vp9_partition_tree);

  /*rough estimate for costing*/
  cpi->common.kf_ymode_probs_index = cpi->common.base_qindex >> 4;
  vp9_init_mode_costs(cpi);

  if (cpi->common.frame_type != KEY_FRAME) {
    vp9_build_nmv_cost_table(
        cpi->mb.nmvjointcost,
        cpi->mb.e_mbd.allow_high_precision_mv ?
        cpi->mb.nmvcost_hp : cpi->mb.nmvcost,
        &cpi->common.fc.nmvc,
        cpi->mb.e_mbd.allow_high_precision_mv, 1, 1);
  }
}

int vp9_block_error_c(int16_t *coeff, int16_t *dqcoeff, int block_size) {
  int i, error = 0;

  for (i = 0; i < block_size; i++) {
    int this_diff = coeff[i] - dqcoeff[i];
    error += this_diff * this_diff;
  }

  return error;
}

static INLINE int cost_coeffs(VP9_COMMON *const cm, MACROBLOCK *mb,
                              int ib, PLANE_TYPE type,
                              ENTROPY_CONTEXT *A,
                              ENTROPY_CONTEXT *L,
                              TX_SIZE tx_size,
                              int y_blocks) {
  MACROBLOCKD *const xd = &mb->e_mbd;
  MB_MODE_INFO *mbmi = &xd->mode_info_context->mbmi;
  int pt;
  int c = 0;
  int cost = 0, pad;
  const int *scan, *nb;
  const struct plane_block_idx pb_idx = plane_block_idx(y_blocks, ib);
  const int eob = xd->plane[pb_idx.plane].eobs[pb_idx.block];
  const int16_t *qcoeff_ptr = BLOCK_OFFSET(xd->plane[pb_idx.plane].qcoeff,
                                           pb_idx.block, 16);
  const int ref = mbmi->ref_frame != INTRA_FRAME;
  unsigned int (*token_costs)[PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS] =
      mb->token_costs[tx_size][type][ref];
  ENTROPY_CONTEXT above_ec, left_ec;
  TX_TYPE tx_type = DCT_DCT;

#if CONFIG_CODE_ZEROGROUP
  int last_nz_pos[3] = {-1, -1, -1};  // Encoder only
  int is_eoo_list[3] = {0, 0, 0};
  int is_eoo_negative[3] = {0, 0, 0};
  int is_last_zero[3] = {0, 0, 0};
  int o, rc, skip_coef_val;
  vp9_zpc_probs *zpc_probs;
  uint8_t token_cache_full[1024];
#endif
  const int segment_id = xd->mode_info_context->mbmi.segment_id;
  vp9_prob (*coef_probs)[REF_TYPES][COEF_BANDS][PREV_COEF_CONTEXTS]
                        [ENTROPY_NODES];
  int seg_eob, default_eob;
  uint8_t token_cache[1024];

#if CONFIG_CODE_ZEROGROUP
  vpx_memset(token_cache, UNKNOWN_TOKEN, sizeof(token_cache));
#endif

  // Check for consistency of tx_size with mode info
  assert((!type && !pb_idx.plane) || (type && pb_idx.plane));
  if (type == PLANE_TYPE_Y_WITH_DC) {
    assert(xd->mode_info_context->mbmi.txfm_size == tx_size);
  } else {
    TX_SIZE tx_size_uv = get_uv_tx_size(xd);
    assert(tx_size == tx_size_uv);
  }

  switch (tx_size) {
    case TX_4X4: {
      tx_type = (type == PLANE_TYPE_Y_WITH_DC) ?
          get_tx_type_4x4(xd, ib) : DCT_DCT;
      above_ec = A[0] != 0;
      left_ec = L[0] != 0;
      coef_probs = cm->fc.coef_probs_4x4;
      seg_eob = 16;
      scan = get_scan_4x4(tx_type);
#if CONFIG_CODE_ZEROGROUP
      zpc_probs = &cm->fc.zpc_probs_4x4;
#endif
      break;
    }
    case TX_8X8: {
      const BLOCK_SIZE_TYPE sb_type = xd->mode_info_context->mbmi.sb_type;
      const int sz = 1 + b_width_log2(sb_type);
      const int x = ib & ((1 << sz) - 1), y = ib - x;
      TX_TYPE tx_type = (type == PLANE_TYPE_Y_WITH_DC) ?
          get_tx_type_8x8(xd, y + (x >> 1)) : DCT_DCT;
      above_ec = (A[0] + A[1]) != 0;
      left_ec = (L[0] + L[1]) != 0;
      scan = get_scan_8x8(tx_type);
      coef_probs = cm->fc.coef_probs_8x8;
      seg_eob = 64;
#if CONFIG_CODE_ZEROGROUP
      zpc_probs = &cm->fc.zpc_probs_8x8;
#endif
      break;
    }
    case TX_16X16: {
      const BLOCK_SIZE_TYPE sb_type = xd->mode_info_context->mbmi.sb_type;
      const int sz = 2 + b_width_log2(sb_type);
      const int x = ib & ((1 << sz) - 1), y = ib - x;
      TX_TYPE tx_type = (type == PLANE_TYPE_Y_WITH_DC) ?
          get_tx_type_16x16(xd, y + (x >> 2)) : DCT_DCT;
      scan = get_scan_16x16(tx_type);
      coef_probs = cm->fc.coef_probs_16x16;
      seg_eob = 256;
      above_ec = (A[0] + A[1] + A[2] + A[3]) != 0;
      left_ec = (L[0] + L[1] + L[2] + L[3]) != 0;
#if CONFIG_CODE_ZEROGROUP
      zpc_probs = &cm->fc.zpc_probs_16x16;
#endif
      break;
    }
    case TX_32X32:
      scan = vp9_default_zig_zag1d_32x32;
      coef_probs = cm->fc.coef_probs_32x32;
      seg_eob = 1024;
      above_ec = (A[0] + A[1] + A[2] + A[3] + A[4] + A[5] + A[6] + A[7]) != 0;
      left_ec = (L[0] + L[1] + L[2] + L[3] + L[4] + L[5] + L[6] + L[7]) != 0;

#if CONFIG_CODE_ZEROGROUP
      zpc_probs = &cm->fc.zpc_probs_32x32;
#endif
      break;
    default:
      abort();
      break;
  }
  assert(eob <= seg_eob);

  pt = combine_entropy_contexts(above_ec, left_ec);
  nb = vp9_get_coef_neighbors_handle(scan, &pad);
  default_eob = seg_eob;

  if (vp9_segfeature_active(xd, segment_id, SEG_LVL_SKIP))
    seg_eob = 0;

  /* sanity check to ensure that we do not have spurious non-zero q values */
  if (eob < seg_eob)
    assert(qcoeff_ptr[scan[eob]] == 0);

#if CONFIG_CODE_ZEROGROUP
  vpx_memset(token_cache_full, ZERO_TOKEN, sizeof(token_cache_full));
  for (c = 0; c < eob; ++c) {
    rc = scan[c];
    token_cache_full[rc] = vp9_dct_value_tokens_ptr[qcoeff_ptr[rc]].token;
    o = vp9_get_orientation(rc, tx_size);
    if (qcoeff_ptr[rc] != 0)
      last_nz_pos[o] = c;
  }
#endif
  {
    for (c = 0; c < eob; c++) {
      int v = qcoeff_ptr[scan[c]];
      int t = vp9_dct_value_tokens_ptr[v].token;
      int band = get_coef_band(scan, tx_size, c);
      if (c)
        pt = vp9_get_coef_context(scan, nb, pad, token_cache, c, default_eob);
#if CONFIG_CODE_ZEROGROUP
      rc = scan[c];
      o = vp9_get_orientation(rc, tx_size);
      skip_coef_val = (token_cache[rc] == ZERO_TOKEN || is_eoo_list[o]);
      if (!skip_coef_val) {
        cost += token_costs[band][pt][t] + vp9_dct_value_cost_ptr[v];
      } else {
        assert(v == 0);
      }
#else
      cost += token_costs[band][pt][t] + vp9_dct_value_cost_ptr[v];
#endif
      if (!c || token_cache[scan[c - 1]])
        cost += vp9_cost_bit(coef_probs[type][ref][band][pt][0], 1);
      token_cache[scan[c]] = t;
#if CONFIG_CODE_ZEROGROUP
      if (t == ZERO_TOKEN && !skip_coef_val) {
        int eoo = 0, use_eoo;
#if USE_ZPC_EOORIENT == 1
        use_eoo = vp9_use_eoo(c, seg_eob, scan, tx_size,
                              is_last_zero, is_eoo_list);
#else
        use_eoo = 0;
#endif
        if (use_eoo) {
          eoo = vp9_is_eoo(c, eob, scan, tx_size, qcoeff_ptr, last_nz_pos);
          if (eoo && is_eoo_negative[o]) eoo = 0;
          if (eoo) {
            int c_;
            int savings = 0;
            int zsaved = 0;
            savings = vp9_cost_bit((*zpc_probs)[ref]
                                   [coef_to_zpc_band(band)]
                                   [coef_to_zpc_ptok(pt)][0], 1) -
                      vp9_cost_bit((*zpc_probs)[ref]
                                   [coef_to_zpc_band(band)]
                                   [coef_to_zpc_ptok(pt)][0], 0);
            for (c_ = c + 1; c_ < eob; ++c_) {
              if (o == vp9_get_orientation(scan[c_], tx_size)) {
                int pt_ = vp9_get_coef_context(scan, nb, pad,
                                               token_cache_full, c_,
                                               default_eob);
                int band_ = get_coef_band(scan, tx_size, c_);
                assert(token_cache_full[scan[c_]] == ZERO_TOKEN);
                if (!c_ || token_cache_full[scan[c_ - 1]])
                  savings += vp9_cost_bit(
                      coef_probs[type][ref][band_][pt_][0], 1);
                savings += vp9_cost_bit(
                    coef_probs[type][ref][band_][pt_][1], 0);
                zsaved++;
              }
            }
            if (savings < 0) {
            // if (zsaved < ZPC_ZEROSSAVED_EOO) {
              eoo = 0;
              is_eoo_negative[o] = 1;
            }
          }
        }
        if (use_eoo) {
          cost += vp9_cost_bit((*zpc_probs)[ref]
                                           [coef_to_zpc_band(band)]
                                           [coef_to_zpc_ptok(pt)][0], !eoo);
          if (eoo) {
            assert(is_eoo_list[o] == 0);
            is_eoo_list[o] = 1;
          }
        }
      }
      is_last_zero[o] = (t == ZERO_TOKEN);
#endif
    }
    if (c < seg_eob) {
      if (c)
        pt = vp9_get_coef_context(scan, nb, pad, token_cache, c, default_eob);
      cost += mb->token_costs[tx_size][type][ref]
          [get_coef_band(scan, tx_size, c)]
          [pt][DCT_EOB_TOKEN];
    }
  }

  // is eob first coefficient;
  for (pt = 0; pt < (1 << tx_size); pt++) {
    A[pt] = L[pt] = c > 0;
  }

  return cost;
}

static void choose_txfm_size_from_rd(VP9_COMP *cpi, MACROBLOCK *x,
                                     int (*r)[2], int *rate,
                                     int *d, int *distortion,
                                     int *s, int *skip,
                                     int64_t txfm_cache[NB_TXFM_MODES],
                                     TX_SIZE max_txfm_size) {
  VP9_COMMON *const cm = &cpi->common;
  MACROBLOCKD *const xd = &x->e_mbd;
  MB_MODE_INFO *const mbmi = &xd->mode_info_context->mbmi;
  vp9_prob skip_prob = vp9_get_pred_prob(cm, xd, PRED_MBSKIP);
  int64_t rd[TX_SIZE_MAX_SB][2];
  int n, m;
  int s0, s1;

  for (n = TX_4X4; n <= max_txfm_size; n++) {
    r[n][1] = r[n][0];
    for (m = 0; m <= n - (n == max_txfm_size); m++) {
      if (m == n)
        r[n][1] += vp9_cost_zero(cm->prob_tx[m]);
      else
        r[n][1] += vp9_cost_one(cm->prob_tx[m]);
    }
  }

  assert(skip_prob > 0);
  s0 = vp9_cost_bit(skip_prob, 0);
  s1 = vp9_cost_bit(skip_prob, 1);

  for (n = TX_4X4; n <= max_txfm_size; n++) {
    if (s[n]) {
      rd[n][0] = rd[n][1] = RDCOST(x->rdmult, x->rddiv, s1, d[n]);
    } else {
      rd[n][0] = RDCOST(x->rdmult, x->rddiv, r[n][0] + s0, d[n]);
      rd[n][1] = RDCOST(x->rdmult, x->rddiv, r[n][1] + s0, d[n]);
    }
  }

  if (max_txfm_size == TX_32X32 &&
      (cm->txfm_mode == ALLOW_32X32 ||
       (cm->txfm_mode == TX_MODE_SELECT &&
        rd[TX_32X32][1] < rd[TX_16X16][1] && rd[TX_32X32][1] < rd[TX_8X8][1] &&
        rd[TX_32X32][1] < rd[TX_4X4][1]))) {
    mbmi->txfm_size = TX_32X32;
  } else if ( cm->txfm_mode == ALLOW_16X16 ||
             (max_txfm_size == TX_16X16 && cm->txfm_mode == ALLOW_32X32) ||
             (cm->txfm_mode == TX_MODE_SELECT &&
              rd[TX_16X16][1] < rd[TX_8X8][1] &&
              rd[TX_16X16][1] < rd[TX_4X4][1])) {
    mbmi->txfm_size = TX_16X16;
  } else if (cm->txfm_mode == ALLOW_8X8 ||
           (cm->txfm_mode == TX_MODE_SELECT && rd[TX_8X8][1] < rd[TX_4X4][1])) {
    mbmi->txfm_size = TX_8X8;
  } else {
    assert(cm->txfm_mode == ONLY_4X4 || cm->txfm_mode == TX_MODE_SELECT);
    mbmi->txfm_size = TX_4X4;
  }

  *distortion = d[mbmi->txfm_size];
  *rate       = r[mbmi->txfm_size][cm->txfm_mode == TX_MODE_SELECT];
  *skip       = s[mbmi->txfm_size];

  txfm_cache[ONLY_4X4] = rd[TX_4X4][0];
  txfm_cache[ALLOW_8X8] = rd[TX_8X8][0];
  txfm_cache[ALLOW_16X16] = rd[TX_16X16][0];
  txfm_cache[ALLOW_32X32] = rd[max_txfm_size][0];
  if (max_txfm_size == TX_32X32 &&
      rd[TX_32X32][1] < rd[TX_16X16][1] && rd[TX_32X32][1] < rd[TX_8X8][1] &&
      rd[TX_32X32][1] < rd[TX_4X4][1])
    txfm_cache[TX_MODE_SELECT] = rd[TX_32X32][1];
  else if (rd[TX_16X16][1] < rd[TX_8X8][1] && rd[TX_16X16][1] < rd[TX_4X4][1])
    txfm_cache[TX_MODE_SELECT] = rd[TX_16X16][1];
  else
    txfm_cache[TX_MODE_SELECT] = rd[TX_4X4][1] < rd[TX_8X8][1] ?
                                 rd[TX_4X4][1] : rd[TX_8X8][1];
}

static int block_error(int16_t *coeff, int16_t *dqcoeff,
                       int block_size, int shift) {
  int i;
  int64_t error = 0;

  for (i = 0; i < block_size; i++) {
    int this_diff = coeff[i] - dqcoeff[i];
    error += this_diff * this_diff;
  }
  error >>= shift;

  return error > INT_MAX ? INT_MAX : (int)error;
}

static int block_error_sby(MACROBLOCK *x, int block_size, int shift) {
  return block_error(x->plane[0].coeff, x->e_mbd.plane[0].dqcoeff,
                     block_size, shift);
}

static int block_error_sbuv(MACROBLOCK *x, BLOCK_SIZE_TYPE bsize, int shift) {
  const int bwl = b_width_log2(bsize), bhl = b_height_log2(bsize);
  int64_t sum = 0;
  int plane;

  for (plane = 1; plane < MAX_MB_PLANE; plane++) {
    const int subsampling = x->e_mbd.plane[plane].subsampling_x +
                            x->e_mbd.plane[plane].subsampling_y;
    sum += block_error(x->plane[plane].coeff, x->e_mbd.plane[plane].dqcoeff,
                       16 << (bwl + bhl - subsampling), 0);
  }
  sum >>= shift;
  return sum > INT_MAX ? INT_MAX : (int)sum;
}

static int rdcost_sby_4x4(VP9_COMMON *const cm, MACROBLOCK *x,
                          BLOCK_SIZE_TYPE bsize) {
  const int bwl = b_width_log2(bsize), bw = 1 << bwl;
  const int bh = 1 << b_height_log2(bsize);
  int cost = 0, b;
  MACROBLOCKD *const xd = &x->e_mbd;
  ENTROPY_CONTEXT t_above[16], t_left[16];

  vpx_memcpy(&t_above, xd->plane[0].above_context,
             sizeof(ENTROPY_CONTEXT) * bw);
  vpx_memcpy(&t_left,  xd->plane[0].left_context,
             sizeof(ENTROPY_CONTEXT) * bh);

  for (b = 0; b < bw * bh; b++) {
    const int x_idx = b & (bw - 1), y_idx = b >> bwl;
    cost += cost_coeffs(cm, x, b, PLANE_TYPE_Y_WITH_DC,
                        t_above + x_idx, t_left + y_idx,
                        TX_4X4, bw * bh);
  }

  return cost;
}

static void super_block_yrd_4x4(VP9_COMMON *const cm, MACROBLOCK *x,
                                int *rate, int *distortion, int *skippable,
                                BLOCK_SIZE_TYPE bsize) {
  const int bwl = b_width_log2(bsize), bhl = b_height_log2(bsize);
  MACROBLOCKD *const xd = &x->e_mbd;

  xd->mode_info_context->mbmi.txfm_size = TX_4X4;
  vp9_transform_sby_4x4(x, bsize);
  vp9_quantize_sby_4x4(x, bsize);

  *distortion = block_error_sby(x, 16 << (bwl + bhl), 2);
  *rate       = rdcost_sby_4x4(cm, x, bsize);
  *skippable  = vp9_sby_is_skippable(xd, bsize);
}

static int rdcost_sby_8x8(VP9_COMMON *const cm, MACROBLOCK *x,
                          BLOCK_SIZE_TYPE bsize) {
  const int bwl = b_width_log2(bsize) - 1, bw = 1 << bwl;
  const int bh = 1 << (b_height_log2(bsize) - 1);
  int cost = 0, b;
  MACROBLOCKD *const xd = &x->e_mbd;
  ENTROPY_CONTEXT t_above[16], t_left[16];

  vpx_memcpy(&t_above, xd->plane[0].above_context,
             sizeof(ENTROPY_CONTEXT) * 2 * bw);
  vpx_memcpy(&t_left,  xd->plane[0].left_context,
             sizeof(ENTROPY_CONTEXT) * 2 * bh);

  for (b = 0; b < bw * bh; b++) {
    const int x_idx = b & (bw - 1), y_idx = b >> bwl;
    cost += cost_coeffs(cm, x, b * 4, PLANE_TYPE_Y_WITH_DC,
                        t_above + x_idx * 2, t_left + y_idx * 2,
                        TX_8X8, 4 * bw * bh);
  }

  return cost;
}

static void super_block_yrd_8x8(VP9_COMMON *const cm, MACROBLOCK *x,
                                int *rate, int *distortion, int *skippable,
                                BLOCK_SIZE_TYPE bsize) {
  const int bwl = b_width_log2(bsize), bhl = b_height_log2(bsize);
  MACROBLOCKD *const xd = &x->e_mbd;

  xd->mode_info_context->mbmi.txfm_size = TX_8X8;
  vp9_transform_sby_8x8(x, bsize);
  vp9_quantize_sby_8x8(x, bsize);

  *distortion = block_error_sby(x, 16 << (bhl + bwl), 2);
  *rate       = rdcost_sby_8x8(cm, x, bsize);
  *skippable  = vp9_sby_is_skippable(xd, bsize);
}

static int rdcost_sby_16x16(VP9_COMMON *const cm, MACROBLOCK *x,
                            BLOCK_SIZE_TYPE bsize) {
  const int bwl = b_width_log2(bsize) - 2, bw = 1 << bwl;
  const int bh = 1 << (b_height_log2(bsize) - 2);
  int cost = 0, b;
  MACROBLOCKD *const xd = &x->e_mbd;
  ENTROPY_CONTEXT t_above[16], t_left[16];

  vpx_memcpy(&t_above, xd->plane[0].above_context,
             sizeof(ENTROPY_CONTEXT) * 4 * bw);
  vpx_memcpy(&t_left,  xd->plane[0].left_context,
             sizeof(ENTROPY_CONTEXT) * 4 * bh);

  for (b = 0; b < bw * bh; b++) {
    const int x_idx = b & (bw - 1), y_idx = b >> bwl;
    cost += cost_coeffs(cm, x, b * 16, PLANE_TYPE_Y_WITH_DC,
                        t_above + x_idx * 4, t_left + y_idx * 4,
                        TX_16X16, bw * bh * 16);
  }

  return cost;
}

static void super_block_yrd_16x16(VP9_COMMON *const cm, MACROBLOCK *x,
                                  int *rate, int *distortion, int *skippable,
                                  BLOCK_SIZE_TYPE bsize) {
  const int bwl = b_width_log2(bsize), bhl = b_height_log2(bsize);
  MACROBLOCKD *const xd = &x->e_mbd;

  xd->mode_info_context->mbmi.txfm_size = TX_16X16;
  vp9_transform_sby_16x16(x, bsize);
  vp9_quantize_sby_16x16(x, bsize);

  *distortion = block_error_sby(x, 16 << (bwl + bhl), 2);
  *rate       = rdcost_sby_16x16(cm, x, bsize);
  *skippable  = vp9_sby_is_skippable(xd, bsize);
}

static int rdcost_sby_32x32(VP9_COMMON *const cm, MACROBLOCK *x,
                            BLOCK_SIZE_TYPE bsize) {
  const int bwl = b_width_log2(bsize) - 3, bw = 1 << bwl;
  const int bh = 1 << (b_height_log2(bsize) - 3);
  int cost = 0, b;
  MACROBLOCKD * const xd = &x->e_mbd;
  ENTROPY_CONTEXT t_above[16], t_left[16];

  vpx_memcpy(&t_above, xd->plane[0].above_context,
             sizeof(ENTROPY_CONTEXT) * 8 * bw);
  vpx_memcpy(&t_left,  xd->plane[0].left_context,
             sizeof(ENTROPY_CONTEXT) * 8 * bh);

  for (b = 0; b < bw * bh; b++) {
    const int x_idx = b & (bw - 1), y_idx = b >> bwl;
    cost += cost_coeffs(cm, x, b * 64, PLANE_TYPE_Y_WITH_DC,
                        t_above + x_idx * 8, t_left + y_idx * 8,
                        TX_32X32, bw * bh * 64);
  }

  return cost;
}

static void super_block_yrd_32x32(VP9_COMMON *const cm, MACROBLOCK *x,
                                  int *rate, int *distortion, int *skippable,
                                  BLOCK_SIZE_TYPE bsize) {
  const int bwl = b_width_log2(bsize), bhl = b_height_log2(bsize);
  MACROBLOCKD *const xd = &x->e_mbd;

  xd->mode_info_context->mbmi.txfm_size = TX_32X32;
  vp9_transform_sby_32x32(x, bsize);
  vp9_quantize_sby_32x32(x, bsize);

  *distortion = block_error_sby(x, 16 << (bwl + bhl), 0);
  *rate       = rdcost_sby_32x32(cm, x, bsize);
  *skippable  = vp9_sby_is_skippable(xd, bsize);
}

static void super_block_yrd(VP9_COMP *cpi,
                            MACROBLOCK *x, int *rate, int *distortion,
                            int *skip, BLOCK_SIZE_TYPE bs,
                            int64_t txfm_cache[NB_TXFM_MODES]) {
  VP9_COMMON *const cm = &cpi->common;
  int r[TX_SIZE_MAX_SB][2], d[TX_SIZE_MAX_SB], s[TX_SIZE_MAX_SB];

  vp9_subtract_sby(x, bs);

  if (bs >= BLOCK_SIZE_SB32X32)
    super_block_yrd_32x32(cm, x, &r[TX_32X32][0], &d[TX_32X32], &s[TX_32X32],
                          bs);
  super_block_yrd_16x16(cm, x, &r[TX_16X16][0], &d[TX_16X16], &s[TX_16X16], bs);
  super_block_yrd_8x8(cm, x,   &r[TX_8X8][0],   &d[TX_8X8],   &s[TX_8X8],   bs);
  super_block_yrd_4x4(cm, x,   &r[TX_4X4][0],   &d[TX_4X4],   &s[TX_4X4],   bs);

  choose_txfm_size_from_rd(cpi, x, r, rate, d, distortion, s, skip, txfm_cache,
                           TX_32X32 - (bs < BLOCK_SIZE_SB32X32));
}

static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, int ib,
                                     B_PREDICTION_MODE *best_mode,
                                     int *bmode_costs,
                                     ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l,
                                     int *bestrate, int *bestratey,
                                     int *bestdistortion) {
  B_PREDICTION_MODE mode;
  MACROBLOCKD *xd = &x->e_mbd;
  int64_t best_rd = INT64_MAX;
  int rate = 0;
  int distortion;
  VP9_COMMON *const cm = &cpi->common;
  const int src_stride = x->plane[0].src.stride;
  uint8_t* const src =
      raster_block_offset_uint8(xd, BLOCK_SIZE_MB16X16, 0, ib,
                                x->plane[0].src.buf, src_stride);
  int16_t* const src_diff =
      raster_block_offset_int16(xd, BLOCK_SIZE_MB16X16, 0, ib,
                                x->plane[0].src_diff);
  int16_t* const diff =
      raster_block_offset_int16(xd, BLOCK_SIZE_MB16X16, 0, ib,
                                xd->plane[0].diff);
  int16_t* const coeff = BLOCK_OFFSET(x->plane[0].coeff, ib, 16);
  uint8_t* const dst =
      raster_block_offset_uint8(xd, BLOCK_SIZE_MB16X16, 0, ib,
                                xd->plane[0].dst.buf, xd->plane[0].dst.stride);
  ENTROPY_CONTEXT ta = *a, tempa = *a;
  ENTROPY_CONTEXT tl = *l, templ = *l;
  TX_TYPE tx_type = DCT_DCT;
  TX_TYPE best_tx_type = DCT_DCT;
  /*
   * The predictor buffer is a 2d buffer with a stride of 16.  Create
   * a temp buffer that meets the stride requirements, but we are only
   * interested in the left 4x4 block
   * */
  DECLARE_ALIGNED_ARRAY(16, int16_t, best_dqcoeff, 16);

  assert(ib < 16);
#if CONFIG_NEWBINTRAMODES
  xd->mode_info_context->bmi[ib].as_mode.context =
    vp9_find_bpred_context(xd, ib, dst, xd->plane[0].dst.stride);
#endif
  xd->mode_info_context->mbmi.txfm_size = TX_4X4;
  for (mode = B_DC_PRED; mode < LEFT4X4; mode++) {
    int64_t this_rd;
    int ratey;

#if CONFIG_NEWBINTRAMODES
    if (xd->frame_type == KEY_FRAME) {
      if (mode == B_CONTEXT_PRED) continue;
    } else {
      if (mode >= B_CONTEXT_PRED - CONTEXT_PRED_REPLACEMENTS &&
          mode < B_CONTEXT_PRED)
        continue;
    }
#endif

    xd->mode_info_context->bmi[ib].as_mode.first = mode;
#if CONFIG_NEWBINTRAMODES
    rate = bmode_costs[
        mode == B_CONTEXT_PRED ? mode - CONTEXT_PRED_REPLACEMENTS : mode];
#else
    rate = bmode_costs[mode];
#endif

    vp9_intra4x4_predict(xd, ib, mode, dst, xd->plane[0].dst.stride);
    vp9_subtract_block(4, 4, src_diff, 16,
                       src, src_stride,
                       dst, xd->plane[0].dst.stride);

    xd->mode_info_context->bmi[ib].as_mode.first = mode;
    tx_type = get_tx_type_4x4(xd, ib);
    if (tx_type != DCT_DCT) {
      vp9_short_fht4x4(src_diff, coeff, 16, tx_type);
      x->quantize_b_4x4(x, ib, tx_type, 16);
    } else {
      x->fwd_txm4x4(src_diff, coeff, 32);
      x->quantize_b_4x4(x, ib, tx_type, 16);
    }

    tempa = ta;
    templ = tl;

    ratey = cost_coeffs(cm, x, ib,
                        PLANE_TYPE_Y_WITH_DC, &tempa, &templ, TX_4X4, 16);
    rate += ratey;
    distortion = vp9_block_error(coeff,