sad_av1.c 8.31 KB
Newer Older
Sebastien Alaiwan's avatar
Sebastien Alaiwan committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
/*
 * Copyright (c) 2017, 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 <stdlib.h>

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

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

#if CONFIG_AV1
static INLINE unsigned int masked_sad(const uint8_t *src, int src_stride,
                                      const uint8_t *a, int a_stride,
                                      const uint8_t *b, int b_stride,
                                      const uint8_t *m, int m_stride, int width,
                                      int height) {
  int y, x;
  unsigned int sad = 0;
  for (y = 0; y < height; y++) {
    for (x = 0; x < width; x++) {
      const int16_t pred = AOM_BLEND_A64(m[x], a[x], b[x]);
      sad += abs(pred - src[x]);
    }
    src += src_stride;
    a += a_stride;
    b += b_stride;
    m += m_stride;
  }
  sad = (sad + 31) >> 6;
  return sad;
}

#define MASKSADMxN(m, n)                                                       \
  unsigned int aom_masked_sad##m##x##n##_c(                                    \
      const uint8_t *src, int src_stride, const uint8_t *ref, int ref_stride,  \
      const uint8_t *second_pred, const uint8_t *msk, int msk_stride,          \
      int invert_mask) {                                                       \
    if (!invert_mask)                                                          \
      return masked_sad(src, src_stride, ref, ref_stride, second_pred, m, msk, \
                        msk_stride, m, n);                                     \
    else                                                                       \
      return masked_sad(src, src_stride, second_pred, m, ref, ref_stride, msk, \
                        msk_stride, m, n);                                     \
  }

/* clang-format off */
#if CONFIG_EXT_PARTITION
MASKSADMxN(128, 128)
MASKSADMxN(128, 64)
MASKSADMxN(64, 128)
#endif  // CONFIG_EXT_PARTITION
MASKSADMxN(64, 64)
MASKSADMxN(64, 32)
MASKSADMxN(32, 64)
MASKSADMxN(32, 32)
MASKSADMxN(32, 16)
MASKSADMxN(16, 32)
MASKSADMxN(16, 16)
MASKSADMxN(16, 8)
MASKSADMxN(8, 16)
MASKSADMxN(8, 8)
MASKSADMxN(8, 4)
MASKSADMxN(4, 8)
MASKSADMxN(4, 4)

#if CONFIG_EXT_PARTITION_TYPES
MASKSADMxN(4, 16)
MASKSADMxN(16, 4)
MASKSADMxN(8, 32)
MASKSADMxN(32, 8)
MASKSADMxN(16, 64)
MASKSADMxN(64, 16)
MASKSADMxN(32, 128)
MASKSADMxN(128, 32)
#endif
86
    /* clang-format on */
Sebastien Alaiwan's avatar
Sebastien Alaiwan committed
87

88
    static INLINE
Sebastien Alaiwan's avatar
Sebastien Alaiwan committed
89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216
    unsigned int highbd_masked_sad(const uint8_t *src8, int src_stride,
                                   const uint8_t *a8, int a_stride,
                                   const uint8_t *b8, int b_stride,
                                   const uint8_t *m, int m_stride, int width,
                                   int height) {
  int y, x;
  unsigned int sad = 0;
  const uint16_t *src = CONVERT_TO_SHORTPTR(src8);
  const uint16_t *a = CONVERT_TO_SHORTPTR(a8);
  const uint16_t *b = CONVERT_TO_SHORTPTR(b8);

  for (y = 0; y < height; y++) {
    for (x = 0; x < width; x++) {
      const uint16_t pred = AOM_BLEND_A64(m[x], a[x], b[x]);
      sad += abs(pred - src[x]);
    }

    src += src_stride;
    a += a_stride;
    b += b_stride;
    m += m_stride;
  }
  sad = (sad + 31) >> 6;

  return sad;
}

#define HIGHBD_MASKSADMXN(m, n)                                         \
  unsigned int aom_highbd_masked_sad##m##x##n##_c(                      \
      const uint8_t *src8, int src_stride, const uint8_t *ref8,         \
      int ref_stride, const uint8_t *second_pred8, const uint8_t *msk,  \
      int msk_stride, int invert_mask) {                                \
    if (!invert_mask)                                                   \
      return highbd_masked_sad(src8, src_stride, ref8, ref_stride,      \
                               second_pred8, m, msk, msk_stride, m, n); \
    else                                                                \
      return highbd_masked_sad(src8, src_stride, second_pred8, m, ref8, \
                               ref_stride, msk, msk_stride, m, n);      \
  }

#if CONFIG_EXT_PARTITION
HIGHBD_MASKSADMXN(128, 128)
HIGHBD_MASKSADMXN(128, 64)
HIGHBD_MASKSADMXN(64, 128)
#endif  // CONFIG_EXT_PARTITION
HIGHBD_MASKSADMXN(64, 64)
HIGHBD_MASKSADMXN(64, 32)
HIGHBD_MASKSADMXN(32, 64)
HIGHBD_MASKSADMXN(32, 32)
HIGHBD_MASKSADMXN(32, 16)
HIGHBD_MASKSADMXN(16, 32)
HIGHBD_MASKSADMXN(16, 16)
HIGHBD_MASKSADMXN(16, 8)
HIGHBD_MASKSADMXN(8, 16)
HIGHBD_MASKSADMXN(8, 8)
HIGHBD_MASKSADMXN(8, 4)
HIGHBD_MASKSADMXN(4, 8)
HIGHBD_MASKSADMXN(4, 4)

#if CONFIG_AV1 && CONFIG_EXT_PARTITION_TYPES
HIGHBD_MASKSADMXN(4, 16)
HIGHBD_MASKSADMXN(16, 4)
HIGHBD_MASKSADMXN(8, 32)
HIGHBD_MASKSADMXN(32, 8)
HIGHBD_MASKSADMXN(16, 64)
HIGHBD_MASKSADMXN(64, 16)
HIGHBD_MASKSADMXN(32, 128)
HIGHBD_MASKSADMXN(128, 32)
#endif

// pre: predictor being evaluated
// wsrc: target weighted prediction (has been *4096 to keep precision)
// mask: 2d weights (scaled by 4096)
static INLINE unsigned int obmc_sad(const uint8_t *pre, int pre_stride,
                                    const int32_t *wsrc, const int32_t *mask,
                                    int width, int height) {
  int y, x;
  unsigned int sad = 0;

  for (y = 0; y < height; y++) {
    for (x = 0; x < width; x++)
      sad += ROUND_POWER_OF_TWO(abs(wsrc[x] - pre[x] * mask[x]), 12);

    pre += pre_stride;
    wsrc += width;
    mask += width;
  }

  return sad;
}

#define OBMCSADMxN(m, n)                                                     \
  unsigned int aom_obmc_sad##m##x##n##_c(const uint8_t *ref, int ref_stride, \
                                         const int32_t *wsrc,                \
                                         const int32_t *mask) {              \
    return obmc_sad(ref, ref_stride, wsrc, mask, m, n);                      \
  }

/* clang-format off */
#if CONFIG_EXT_PARTITION
OBMCSADMxN(128, 128)
OBMCSADMxN(128, 64)
OBMCSADMxN(64, 128)
#endif  // CONFIG_EXT_PARTITION
OBMCSADMxN(64, 64)
OBMCSADMxN(64, 32)
OBMCSADMxN(32, 64)
OBMCSADMxN(32, 32)
OBMCSADMxN(32, 16)
OBMCSADMxN(16, 32)
OBMCSADMxN(16, 16)
OBMCSADMxN(16, 8)
OBMCSADMxN(8, 16)
OBMCSADMxN(8, 8)
OBMCSADMxN(8, 4)
OBMCSADMxN(4, 8)
OBMCSADMxN(4, 4)

#if CONFIG_AV1 && CONFIG_EXT_PARTITION_TYPES
OBMCSADMxN(4, 16)
OBMCSADMxN(16, 4)
OBMCSADMxN(8, 32)
OBMCSADMxN(32, 8)
OBMCSADMxN(16, 64)
OBMCSADMxN(64, 16)
OBMCSADMxN(32, 128)
OBMCSADMxN(128, 32)
#endif
217
    /* clang-format on */
Sebastien Alaiwan's avatar
Sebastien Alaiwan committed
218

219
    static INLINE
Sebastien Alaiwan's avatar
Sebastien Alaiwan committed
220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277
    unsigned int highbd_obmc_sad(const uint8_t *pre8, int pre_stride,
                                 const int32_t *wsrc, const int32_t *mask,
                                 int width, int height) {
  int y, x;
  unsigned int sad = 0;
  const uint16_t *pre = CONVERT_TO_SHORTPTR(pre8);

  for (y = 0; y < height; y++) {
    for (x = 0; x < width; x++)
      sad += ROUND_POWER_OF_TWO(abs(wsrc[x] - pre[x] * mask[x]), 12);

    pre += pre_stride;
    wsrc += width;
    mask += width;
  }

  return sad;
}

#define HIGHBD_OBMCSADMXN(m, n)                                \
  unsigned int aom_highbd_obmc_sad##m##x##n##_c(               \
      const uint8_t *ref, int ref_stride, const int32_t *wsrc, \
      const int32_t *mask) {                                   \
    return highbd_obmc_sad(ref, ref_stride, wsrc, mask, m, n); \
  }

/* clang-format off */
#if CONFIG_EXT_PARTITION
HIGHBD_OBMCSADMXN(128, 128)
HIGHBD_OBMCSADMXN(128, 64)
HIGHBD_OBMCSADMXN(64, 128)
#endif  // CONFIG_EXT_PARTITION
HIGHBD_OBMCSADMXN(64, 64)
HIGHBD_OBMCSADMXN(64, 32)
HIGHBD_OBMCSADMXN(32, 64)
HIGHBD_OBMCSADMXN(32, 32)
HIGHBD_OBMCSADMXN(32, 16)
HIGHBD_OBMCSADMXN(16, 32)
HIGHBD_OBMCSADMXN(16, 16)
HIGHBD_OBMCSADMXN(16, 8)
HIGHBD_OBMCSADMXN(8, 16)
HIGHBD_OBMCSADMXN(8, 8)
HIGHBD_OBMCSADMXN(8, 4)
HIGHBD_OBMCSADMXN(4, 8)
HIGHBD_OBMCSADMXN(4, 4)

#if CONFIG_AV1 && CONFIG_EXT_PARTITION_TYPES
HIGHBD_OBMCSADMXN(4, 16)
HIGHBD_OBMCSADMXN(16, 4)
HIGHBD_OBMCSADMXN(8, 32)
HIGHBD_OBMCSADMXN(32, 8)
HIGHBD_OBMCSADMXN(16, 64)
HIGHBD_OBMCSADMXN(64, 16)
HIGHBD_OBMCSADMXN(32, 128)
HIGHBD_OBMCSADMXN(128, 32)
#endif
/* clang-format on */
#endif  // CONFIG_AV1