reconintra.c 46.8 KB
Newer Older
Jingning Han's avatar
Jingning Han committed
1 2 3 4 5 6 7 8 9 10
/*
 *  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.
 */

hui su's avatar
hui su committed
11 12
#include <math.h>

Jingning Han's avatar
Jingning Han committed
13 14 15 16 17 18 19 20 21 22
#include "./vpx_config.h"
#include "./vpx_dsp_rtcd.h"

#if CONFIG_VP9_HIGHBITDEPTH
#include "vpx_dsp/vpx_dsp_common.h"
#endif  // CONFIG_VP9_HIGHBITDEPTH
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/mem.h"
#include "vpx_ports/vpx_once.h"

23 24
#include "vp10/common/reconintra.h"
#include "vp10/common/onyxc_int.h"
Jingning Han's avatar
Jingning Han committed
25

26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45
enum {
  NEED_LEFT = 1 << 1,
  NEED_ABOVE = 1 << 2,
  NEED_ABOVERIGHT = 1 << 3,
  NEED_ABOVELEFT = 1 << 4,
  NEED_BOTTOMLEFT = 1 << 5,
};

static const uint8_t extend_modes[INTRA_MODES] = {
  NEED_ABOVE | NEED_LEFT,                   // DC
  NEED_ABOVE,                               // V
  NEED_LEFT,                                // H
  NEED_ABOVE | NEED_ABOVERIGHT,             // D45
  NEED_LEFT | NEED_ABOVE | NEED_ABOVELEFT,  // D135
  NEED_LEFT | NEED_ABOVE | NEED_ABOVELEFT,  // D117
  NEED_LEFT | NEED_ABOVE | NEED_ABOVELEFT,  // D153
  NEED_LEFT | NEED_BOTTOMLEFT,              // D207
  NEED_ABOVE | NEED_ABOVERIGHT,             // D63
  NEED_LEFT | NEED_ABOVE | NEED_ABOVELEFT,  // TM
};
Jingning Han's avatar
Jingning Han committed
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 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105
static const uint8_t orders_64x64[1] = { 0 };
static const uint8_t orders_64x32[2] = { 0, 1 };
static const uint8_t orders_32x64[2] = { 0, 1 };
static const uint8_t orders_32x32[4] = {
  0, 1,
  2, 3,
};
static const uint8_t orders_32x16[8] = {
  0, 2,
  1, 3,
  4, 6,
  5, 7,
};
static const uint8_t orders_16x32[8] = {
  0, 1, 2, 3,
  4, 5, 6, 7,
};
static const uint8_t orders_16x16[16] = {
  0,   1,  4,  5,
  2,   3,  6,  7,
  8,   9, 12, 13,
  10, 11, 14, 15,
};
static const uint8_t orders_16x8[32] = {
  0,   2,  8, 10,
  1,   3,  9, 11,
  4,   6, 12, 14,
  5,   7, 13, 15,
  16, 18, 24, 26,
  17, 19, 25, 27,
  20, 22, 28, 30,
  21, 23, 29, 31,
};
static const uint8_t orders_8x16[32] = {
  0,   1,  2,  3,  8,  9, 10, 11,
  4,   5,  6,  7, 12, 13, 14, 15,
  16, 17, 18, 19, 24, 25, 26, 27,
  20, 21, 22, 23, 28, 29, 30, 31,
};
static const uint8_t orders_8x8[64] = {
  0,   1,  4,  5, 16, 17, 20, 21,
  2,   3,  6,  7, 18, 19, 22, 23,
  8,   9, 12, 13, 24, 25, 28, 29,
  10, 11, 14, 15, 26, 27, 30, 31,
  32, 33, 36, 37, 48, 49, 52, 53,
  34, 35, 38, 39, 50, 51, 54, 55,
  40, 41, 44, 45, 56, 57, 60, 61,
  42, 43, 46, 47, 58, 59, 62, 63,
};
static const uint8_t *const orders[BLOCK_SIZES] = {
  orders_8x8, orders_8x8, orders_8x8, orders_8x8,
  orders_8x16, orders_16x8, orders_16x16,
  orders_16x32, orders_32x16, orders_32x32,
  orders_32x64, orders_64x32, orders_64x64,
};

static int vp10_has_right(BLOCK_SIZE bsize, int mi_row, int mi_col,
                          int right_available,
                          TX_SIZE txsz, int y, int x, int ss_x) {
hui su's avatar
hui su committed
106 107 108 109
  const int wl = mi_width_log2_lookup[bsize];
  const int w = VPXMAX(num_4x4_blocks_wide_lookup[bsize] >> ss_x, 1);
  const int step = 1 << txsz;

110
  if (y == 0) {
hui su's avatar
hui su committed
111
    const int hl = mi_height_log2_lookup[bsize];
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
    const uint8_t *order = orders[bsize];
    int my_order, tr_order;

    if (x + step < w)
      return 1;

    mi_row = (mi_row & 7) >> hl;
    mi_col = (mi_col & 7) >> wl;

    if (mi_row == 0)
      return right_available;

    if (((mi_col + 1) << wl) >= 8)
      return 0;

    my_order = order[((mi_row + 0) << (3 - wl)) + mi_col + 0];
    tr_order = order[((mi_row - 1) << (3 - wl)) + mi_col + 1];

    return my_order > tr_order && right_available;
  } else {
    return x + step < w;
  }
}

static int vp10_has_bottom(BLOCK_SIZE bsize, int mi_row, int mi_col,
                           int bottom_available, TX_SIZE txsz,
                           int y, int x, int ss_y) {
  if (x == 0) {
hui su's avatar
hui su committed
140 141 142 143
    const int wl = mi_width_log2_lookup[bsize];
    const int hl = mi_height_log2_lookup[bsize];
    const int h = 1 << (hl + 1 - ss_y);
    const int step = 1 << txsz;
144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167
    const uint8_t *order = orders[bsize];
    int my_order, bl_order;

    mi_row = (mi_row & 7) >> hl;
    mi_col = (mi_col & 7) >> wl;

    if (mi_col == 0)
      return bottom_available &&
             (mi_row << (hl + !ss_y)) + y + step < (8 << !ss_y);

    if (((mi_row + 1) << hl) >= 8)
      return 0;

    if (y + step < h)
      return 1;

    my_order = order[((mi_row + 0) << (3 - wl)) + mi_col + 0];
    bl_order = order[((mi_row + 1) << (3 - wl)) + mi_col - 1];

    return bl_order < my_order && bottom_available;
  } else {
    return 0;
  }
}
Jingning Han's avatar
Jingning Han committed
168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183

typedef void (*intra_pred_fn)(uint8_t *dst, ptrdiff_t stride,
                              const uint8_t *above, const uint8_t *left);

static intra_pred_fn pred[INTRA_MODES][TX_SIZES];
static intra_pred_fn dc_pred[2][2][TX_SIZES];

#if CONFIG_VP9_HIGHBITDEPTH
typedef void (*intra_high_pred_fn)(uint16_t *dst, ptrdiff_t stride,
                                   const uint16_t *above, const uint16_t *left,
                                   int bd);
static intra_high_pred_fn pred_high[INTRA_MODES][4];
static intra_high_pred_fn dc_pred_high[2][2][4];
#endif  // CONFIG_VP9_HIGHBITDEPTH

static void vp10_init_intra_predictors_internal(void) {
184
#define INIT_NO_4X4(p, type) \
Jingning Han's avatar
Jingning Han committed
185 186 187 188
  p[TX_8X8] = vpx_##type##_predictor_8x8; \
  p[TX_16X16] = vpx_##type##_predictor_16x16; \
  p[TX_32X32] = vpx_##type##_predictor_32x32

189 190 191 192
#define INIT_ALL_SIZES(p, type) \
  p[TX_4X4] = vpx_##type##_predictor_4x4; \
  INIT_NO_4X4(p, type)

Jingning Han's avatar
Jingning Han committed
193 194
  INIT_ALL_SIZES(pred[V_PRED], v);
  INIT_ALL_SIZES(pred[H_PRED], h);
195 196 197
  INIT_ALL_SIZES(pred[D207_PRED], d207e);
  INIT_ALL_SIZES(pred[D45_PRED], d45e);
  INIT_ALL_SIZES(pred[D63_PRED], d63e);
Jingning Han's avatar
Jingning Han committed
198 199 200 201 202 203 204 205 206 207 208 209 210
  INIT_ALL_SIZES(pred[D117_PRED], d117);
  INIT_ALL_SIZES(pred[D135_PRED], d135);
  INIT_ALL_SIZES(pred[D153_PRED], d153);
  INIT_ALL_SIZES(pred[TM_PRED], tm);

  INIT_ALL_SIZES(dc_pred[0][0], dc_128);
  INIT_ALL_SIZES(dc_pred[0][1], dc_top);
  INIT_ALL_SIZES(dc_pred[1][0], dc_left);
  INIT_ALL_SIZES(dc_pred[1][1], dc);

#if CONFIG_VP9_HIGHBITDEPTH
  INIT_ALL_SIZES(pred_high[V_PRED], highbd_v);
  INIT_ALL_SIZES(pred_high[H_PRED], highbd_h);
211 212 213
  INIT_ALL_SIZES(pred_high[D207_PRED], highbd_d207e);
  INIT_ALL_SIZES(pred_high[D45_PRED], highbd_d45e);
  INIT_ALL_SIZES(pred_high[D63_PRED], highbd_d63e);
Jingning Han's avatar
Jingning Han committed
214 215 216 217 218 219 220 221 222 223 224 225 226 227
  INIT_ALL_SIZES(pred_high[D117_PRED], highbd_d117);
  INIT_ALL_SIZES(pred_high[D135_PRED], highbd_d135);
  INIT_ALL_SIZES(pred_high[D153_PRED], highbd_d153);
  INIT_ALL_SIZES(pred_high[TM_PRED], highbd_tm);

  INIT_ALL_SIZES(dc_pred_high[0][0], highbd_dc_128);
  INIT_ALL_SIZES(dc_pred_high[0][1], highbd_dc_top);
  INIT_ALL_SIZES(dc_pred_high[1][0], highbd_dc_left);
  INIT_ALL_SIZES(dc_pred_high[1][1], highbd_dc);
#endif  // CONFIG_VP9_HIGHBITDEPTH

#undef intra_pred_allsizes
}

hui su's avatar
hui su committed
228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245
#if CONFIG_EXT_INTRA
#define PI 3.14159265
#define FILTER_INTRA_PREC_BITS 10
#define FILTER_INTRA_ROUND_VAL 511

static const uint8_t ext_intra_extend_modes[FILTER_INTRA_MODES] = {
  NEED_LEFT | NEED_ABOVE,      // FILTER_DC
  NEED_LEFT | NEED_ABOVE,      // FILTER_V
  NEED_LEFT | NEED_ABOVE,      // FILTER_H
  NEED_LEFT | NEED_ABOVE,      // FILTER_D45
  NEED_LEFT | NEED_ABOVE,      // FILTER_D135
  NEED_LEFT | NEED_ABOVE,      // FILTER_D117
  NEED_LEFT | NEED_ABOVE,      // FILTER_D153
  NEED_LEFT | NEED_ABOVE,      // FILTER_D207
  NEED_LEFT | NEED_ABOVE,      // FILTER_D63
  NEED_LEFT | NEED_ABOVE,      // FILTER_TM
};

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
static int intra_subpel_interp(int base, int shift, const uint8_t *ref,
                               int ref_start_idx, int ref_end_idx,
                               INTRA_FILTER filter_type) {
  int val, k, idx, filter_idx = 0;
  const int16_t *filter = NULL;

  if (filter_type == INTRA_FILTER_LINEAR) {
    val = ref[base] * (256 - shift) + ref[base + 1] * shift;
    val = ROUND_POWER_OF_TWO(val, 8);
  } else {
    filter_idx = ROUND_POWER_OF_TWO(shift, 8 - SUBPEL_BITS);
    filter = vp10_intra_filter_kernels[filter_type][filter_idx];

    if (filter_idx < (1 << SUBPEL_BITS)) {
      val = 0;
      for (k = 0; k < SUBPEL_TAPS; ++k) {
        idx = base + 1 - (SUBPEL_TAPS / 2) + k;
        idx = VPXMAX(VPXMIN(idx, ref_end_idx), ref_start_idx);
        val += ref[idx] * filter[k];
      }
      val = ROUND_POWER_OF_TWO(val, FILTER_BITS);
    } else {
      val = ref[base + 1];
    }
  }

  return val;
}

hui su's avatar
hui su committed
275 276 277
// Directional prediction, zone 1: 0 < angle < 90
static void dr_prediction_z1(uint8_t *dst, ptrdiff_t stride, int bs,
                             const uint8_t *above, const uint8_t *left,
278
                             int dx, int dy, INTRA_FILTER filter_type) {
hui su's avatar
hui su committed
279
  int r, c, x, base, shift, val;
hui su's avatar
hui su committed
280 281 282 283 284 285

  (void)left;
  (void)dy;
  assert(dy == 1);
  assert(dx < 0);

hui su's avatar
hui su committed
286 287 288 289 290 291 292 293 294 295 296 297 298 299 300
  if (filter_type != INTRA_FILTER_LINEAR) {
    const int pad_size = SUBPEL_TAPS >> 1;
    int len;
    DECLARE_ALIGNED(16, uint8_t, buf[SUBPEL_SHIFTS][64]);
    DECLARE_ALIGNED(16, uint8_t, src[64 + SUBPEL_TAPS]);
    uint8_t flags[SUBPEL_SHIFTS];

    memset(flags, 0, SUBPEL_SHIFTS * sizeof(flags[0]));
    memset(src, above[0], pad_size * sizeof(above[0]));
    memcpy(src + pad_size, above, 2 * bs * sizeof(above[0]));
    memset(src + pad_size + 2 * bs, above[2 * bs - 1],
           pad_size * sizeof(above[0]));
    flags[0] = 1;
    x = -dx;
    for (r = 0; r < bs; ++r, dst += stride, x -= dx) {
hui su's avatar
hui su committed
301
      base = x >> 8;
hui su's avatar
hui su committed
302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359
      shift = x & 0xFF;
      shift = ROUND_POWER_OF_TWO(shift, 8 - SUBPEL_BITS);
      if (shift == SUBPEL_SHIFTS) {
        base += 1;
        shift = 0;
      }
      len = VPXMIN(bs, 2 * bs - 1 - base);
      if (len <= 0) {
        int i;
        for (i = r; i < bs; ++i) {
          memset(dst, above[2 * bs - 1], bs * sizeof(dst[0]));
          dst += stride;
        }
        return;
      }

      if (len <= (bs >> 1) && !flags[shift]) {
        base = x >> 8;
        shift = x & 0xFF;
        for (c = 0; c < len; ++c) {
          val = intra_subpel_interp(base, shift, above, 0, 2 * bs - 1,
                                    filter_type);
          dst[c] = clip_pixel(val);
          ++base;
        }
      } else {
        if (!flags[shift]) {
          vpx_convolve8_horiz(src + pad_size, 2 * bs, buf[shift], 2 * bs,
                              vp10_intra_filter_kernels[filter_type][shift], 16,
                              NULL, 16, 2 * bs, 2 * bs < 16 ? 2 : 1);
          flags[shift] = 1;
        }
        memcpy(dst, shift == 0 ? src + pad_size + base : &buf[shift][base],
            len * sizeof(dst[0]));
      }

      if (len < bs)
        memset(dst + len, above[2 * bs - 1], (bs - len) * sizeof(dst[0]));
    }
    return;
  }

  // For linear filter, C code is faster.
  x = -dx;
  for (r = 0; r < bs; ++r, dst += stride, x -= dx) {
    base = x >> 8;
    shift = x & 0xFF;

    if (base >= 2 * bs - 1) {
      int i;
      for (i = r; i < bs; ++i) {
        memset(dst, above[2 * bs - 1], bs * sizeof(dst[0]));
        dst += stride;
      }
      return;
    }

    for (c = 0; c < bs; ++c, ++base) {
hui su's avatar
hui su committed
360
      if (base < 2 * bs - 1) {
hui su's avatar
hui su committed
361 362
        val = above[base] * (256 - shift) + above[base + 1] * shift;
        val = ROUND_POWER_OF_TWO(val, 8);
hui su's avatar
hui su committed
363 364 365 366 367 368 369 370 371 372 373
        dst[c] = clip_pixel(val);
      } else {
        dst[c] = above[2 * bs - 1];
      }
    }
  }
}

// Directional prediction, zone 2: 90 < angle < 180
static void dr_prediction_z2(uint8_t *dst, ptrdiff_t stride, int bs,
                             const uint8_t *above, const uint8_t *left,
374
                             int dx, int dy, INTRA_FILTER filter_type) {
Debargha Mukherjee's avatar
Debargha Mukherjee committed
375
  int r, c, x, y, shift1, shift2, val, base1, base2;
hui su's avatar
hui su committed
376 377 378 379

  assert(dx > 0);
  assert(dy > 0);

hui su's avatar
hui su committed
380 381 382 383 384 385 386 387 388
  x = -dx;
  for (r = 0; r < bs; ++r, x -= dx, dst += stride) {
    base1 = x >> 8;
    y = (r << 8) - dy;
    for (c = 0; c < bs; ++c, ++base1, y -= dy) {
      if (base1 >= -1) {
        shift1 = x & 0xFF;
        val = intra_subpel_interp(base1, shift1, above, -1, bs - 1,
                                  filter_type);
hui su's avatar
hui su committed
389
      } else {
hui su's avatar
hui su committed
390 391 392 393 394
        base2 = y >> 8;
        if (base2 >= 0) {
          shift2 = y & 0xFF;
          val = intra_subpel_interp(base2, shift2, left, 0, bs - 1,
                                    filter_type);
hui su's avatar
hui su committed
395
        } else {
396
          val = left[0];
hui su's avatar
hui su committed
397 398 399 400 401 402 403 404 405 406
        }
      }
      dst[c] = clip_pixel(val);
    }
  }
}

// Directional prediction, zone 3: 180 < angle < 270
static void dr_prediction_z3(uint8_t *dst, ptrdiff_t stride, int bs,
                             const uint8_t *above, const uint8_t *left,
407
                             int dx, int dy, INTRA_FILTER filter_type) {
hui su's avatar
hui su committed
408
  int r, c, y, base, shift, val;
hui su's avatar
hui su committed
409 410 411

  (void)above;
  (void)dx;
412

hui su's avatar
hui su committed
413 414 415
  assert(dx == 1);
  assert(dy < 0);

hui su's avatar
hui su committed
416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432
  if (filter_type != INTRA_FILTER_LINEAR) {
    const int pad_size = SUBPEL_TAPS >> 1;
    int len, i;
    DECLARE_ALIGNED(16, uint8_t, buf[64][4 * SUBPEL_SHIFTS]);
    DECLARE_ALIGNED(16, uint8_t, src[(64 + SUBPEL_TAPS) * 4]);
    uint8_t flags[SUBPEL_SHIFTS];

    memset(flags, 0, SUBPEL_SHIFTS * sizeof(flags[0]));
    for (i = 0; i < pad_size; ++i)
      src[4 * i] = left[0];
    for (i = 0; i < 2 * bs; ++i)
      src[4 * (i + pad_size)] = left[i];
    for (i = 0; i < pad_size; ++i)
      src[4 * (i + 2 * bs + pad_size)] = left[2 * bs - 1];
    flags[0] = 1;
    y = -dy;
    for (c = 0; c < bs; ++c, y -= dy) {
hui su's avatar
hui su committed
433
      base = y >> 8;
hui su's avatar
hui su committed
434 435 436 437 438 439 440 441 442
      shift = y & 0xFF;
      shift = ROUND_POWER_OF_TWO(shift, 8 - SUBPEL_BITS);
      if (shift == SUBPEL_SHIFTS) {
        base += 1;
        shift = 0;
      }
      len = VPXMIN(bs, 2 * bs - 1 - base);

      if (len <= 0) {
Debargha Mukherjee's avatar
Debargha Mukherjee committed
443 444
        for (r = 0; r < bs; ++r) {
          dst[r * stride + c] = left[ 2 * bs - 1];
hui su's avatar
hui su committed
445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493
        }
        continue;
      }

      if (len <= (bs >> 1) && !flags[shift]) {
        base = y >> 8;
        shift = y & 0xFF;
        for (r = 0; r < len; ++r) {
          val = intra_subpel_interp(base, shift, left, 0, 2 * bs - 1,
                                    filter_type);
          dst[r * stride + c] = clip_pixel(val);
          ++base;
        }
      } else {
        if (!flags[shift]) {
          vpx_convolve8_vert(src + 4 * pad_size, 4,
                             buf[0] + 4 * shift, 4 * SUBPEL_SHIFTS, NULL, 16,
                             vp10_intra_filter_kernels[filter_type][shift], 16,
                             2 * bs < 16 ? 4 : 4, 2 * bs);
          flags[shift] = 1;
        }

        if (shift == 0) {
          for (r = 0; r < len; ++r) {
            dst[r * stride + c] = left[r + base];
          }
        } else {
          for (r = 0; r < len; ++r) {
            dst[r * stride + c] = buf[r + base][4 * shift];
          }
        }
      }

      if (len < bs) {
        for (r = len; r < bs; ++r) {
          dst[r * stride + c] = left[ 2 * bs - 1];
        }
      }
    }
    return;
  }

  // For linear filter, C code is faster.
  y = -dy;
  for (c = 0; c < bs; ++c, y -= dy) {
    base = y >> 8;
    shift = y & 0xFF;

    for (r = 0; r < bs; ++r, ++base) {
494
      if (base < 2 * bs - 1) {
hui su's avatar
hui su committed
495 496 497
        val = left[base] * (256 - shift) + left[base + 1] * shift;
        val = ROUND_POWER_OF_TWO(val, 8);
        dst[r * stride + c] = clip_pixel(val);
hui su's avatar
hui su committed
498
      } else {
hui su's avatar
hui su committed
499 500 501
        for (; r < bs; ++r)
          dst[r * stride + c] = left[2 * bs - 1];
        break;
hui su's avatar
hui su committed
502 503 504 505 506
      }
    }
  }
}

hui su's avatar
hui su committed
507
static void dr_predictor(uint8_t *dst, ptrdiff_t stride, TX_SIZE tx_size,
508 509
                         const uint8_t *above, const uint8_t *left, int angle,
                         INTRA_FILTER filter_type) {
hui su's avatar
hui su committed
510 511
  double t = 0;
  int dx, dy;
hui su's avatar
hui su committed
512
  int bs = 4 << tx_size;
hui su's avatar
hui su committed
513 514 515 516 517 518

  if (angle != 90 && angle != 180)
    t = tan(angle * PI / 180.0);
  if (angle > 0 && angle < 90) {
    dx = -((int)(256 / t));
    dy = 1;
519
    dr_prediction_z1(dst, stride, bs, above, left, dx, dy, filter_type);
hui su's avatar
hui su committed
520 521 522 523
  } else if (angle > 90 && angle < 180) {
    t = -t;
    dx = (int)(256 / t);
    dy = (int)(256 * t);
524
    dr_prediction_z2(dst, stride, bs, above, left, dx, dy, filter_type);
hui su's avatar
hui su committed
525 526 527
  } else if (angle > 180 && angle < 270) {
    dx = 1;
    dy = -((int)(256 * t));
528
    dr_prediction_z3(dst, stride, bs, above, left, dx, dy, filter_type);
hui su's avatar
hui su committed
529
  } else if (angle == 90) {
hui su's avatar
hui su committed
530
    pred[V_PRED][tx_size](dst, stride, above, left);
hui su's avatar
hui su committed
531
  } else if (angle == 180) {
hui su's avatar
hui su committed
532
    pred[H_PRED][tx_size](dst, stride, above, left);
hui su's avatar
hui su committed
533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692
  }
}

static int filter_intra_taps_4[TX_SIZES][INTRA_MODES][4] = {
    {
        {735, 881, -537, -54},
        {1005, 519, -488, -11},
        {383, 990, -343, -6},
        {442, 805, -542, 319},
        {658, 616, -133, -116},
        {875, 442, -141, -151},
        {386, 741, -23, -80},
        {390, 1027, -446, 51},
        {679, 606, -523, 262},
        {903, 922, -778, -23},
    },
    {
        {648, 803, -444, 16},
        {972, 620, -576, 7},
        {561, 967, -499, -5},
        {585, 762, -468, 144},
        {596, 619, -182, -9},
        {895, 459, -176, -153},
        {557, 722, -126, -129},
        {601, 839, -523, 105},
        {562, 709, -499, 251},
        {803, 872, -695, 43},
    },
    {
        {423, 728, -347, 111},
        {963, 685, -665, 23},
        {281, 1024, -480, 216},
        {640, 596, -437, 78},
        {429, 669, -259, 99},
        {740, 646, -415, 23},
        {568, 771, -346, 40},
        {404, 833, -486, 209},
        {398, 712, -423, 307},
        {939, 935, -887, 17},
    },
    {
        {477, 737, -393, 150},
        {881, 630, -546, 67},
        {506, 984, -443, -20},
        {114, 459, -270, 528},
        {433, 528, 14, 3},
        {837, 470, -301, -30},
        {181, 777, 89, -107},
        {-29, 716, -232, 259},
        {589, 646, -495, 255},
        {740, 884, -728, 77},
    },
};

static void filter_intra_predictors_4tap(uint8_t *dst, ptrdiff_t stride, int bs,
                                         const uint8_t *above,
                                         const uint8_t *left,
                                         int mode) {
  int k, r, c;
  int pred[33][65];
  int mean, ipred;
  const TX_SIZE tx_size = (bs == 32) ? TX_32X32 :
      ((bs == 16) ? TX_16X16 : ((bs == 8) ? TX_8X8 : (TX_4X4)));
  const int c0 = filter_intra_taps_4[tx_size][mode][0];
  const int c1 = filter_intra_taps_4[tx_size][mode][1];
  const int c2 = filter_intra_taps_4[tx_size][mode][2];
  const int c3 = filter_intra_taps_4[tx_size][mode][3];

  k = 0;
  mean = 0;
  while (k < bs) {
    mean = mean + (int)left[k];
    mean = mean + (int)above[k];
    k++;
  }
  mean = (mean + bs) / (2 * bs);

  for (r = 0; r < bs; ++r)
    pred[r + 1][0] = (int)left[r] - mean;

  for (c = 0; c < 2 * bs + 1; ++c)
    pred[0][c] = (int)above[c - 1] - mean;

  for (r = 1; r < bs + 1; ++r)
    for (c = 1; c < 2 * bs + 1 - r; ++c) {
      ipred = c0 * pred[r - 1][c] + c1 * pred[r][c - 1] +
          c2 * pred[r - 1][c - 1] + c3 * pred[r - 1][c + 1];
      pred[r][c] = ipred < 0 ?
          -((-ipred + FILTER_INTRA_ROUND_VAL) >> FILTER_INTRA_PREC_BITS) :
          ((ipred + FILTER_INTRA_ROUND_VAL) >> FILTER_INTRA_PREC_BITS);
    }

  for (r = 0; r < bs; ++r) {
    for (c = 0; c < bs; ++c) {
      ipred = pred[r + 1][c + 1] + mean;
      dst[c] = clip_pixel(ipred);
    }
    dst += stride;
  }
}

static void dc_filter_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                               const uint8_t *above, const uint8_t *left) {
  filter_intra_predictors_4tap(dst, stride, bs, above, left, DC_PRED);
}

static void v_filter_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                               const uint8_t *above, const uint8_t *left) {
  filter_intra_predictors_4tap(dst, stride, bs, above, left, V_PRED);
}

static void h_filter_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                               const uint8_t *above, const uint8_t *left) {
  filter_intra_predictors_4tap(dst, stride, bs, above, left, H_PRED);
}

static void d45_filter_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                                 const uint8_t *above, const uint8_t *left) {
  filter_intra_predictors_4tap(dst, stride, bs, above, left, D45_PRED);
}

static void d135_filter_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                                  const uint8_t *above, const uint8_t *left) {
  filter_intra_predictors_4tap(dst, stride, bs, above, left, D135_PRED);
}

static void d117_filter_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                                  const uint8_t *above, const uint8_t *left) {
  filter_intra_predictors_4tap(dst, stride, bs, above, left, D117_PRED);
}

static void d153_filter_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                                  const uint8_t *above, const uint8_t *left) {
  filter_intra_predictors_4tap(dst, stride, bs, above, left, D153_PRED);
}

static void d207_filter_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                                  const uint8_t *above, const uint8_t *left) {
  filter_intra_predictors_4tap(dst, stride, bs, above, left, D207_PRED);
}

static void d63_filter_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                                 const uint8_t *above, const uint8_t *left) {
  filter_intra_predictors_4tap(dst, stride, bs, above, left, D63_PRED);
}

static void tm_filter_predictor(uint8_t *dst, ptrdiff_t stride, int bs,
                                const uint8_t *above, const uint8_t *left) {
  filter_intra_predictors_4tap(dst, stride, bs, above, left, TM_PRED);
}

static void (*filter_intra_predictors[EXT_INTRA_MODES])(uint8_t *dst,
    ptrdiff_t stride, int bs, const uint8_t *above, const uint8_t *left) = {
        dc_filter_predictor, v_filter_predictor, h_filter_predictor,
        d45_filter_predictor, d135_filter_predictor, d117_filter_predictor,
        d153_filter_predictor, d207_filter_predictor, d63_filter_predictor,
        tm_filter_predictor,
};

#if CONFIG_VP9_HIGHBITDEPTH
693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721
static int highbd_intra_subpel_interp(int base, int shift, const uint16_t *ref,
                                      int ref_start_idx, int ref_end_idx,
                                      INTRA_FILTER filter_type) {
  int val, k, idx, filter_idx = 0;
  const int16_t *filter = NULL;

  if (filter_type == INTRA_FILTER_LINEAR) {
    val = ref[base] * (256 - shift) + ref[base + 1] * shift;
    val = ROUND_POWER_OF_TWO(val, 8);
  } else {
    filter_idx = ROUND_POWER_OF_TWO(shift, 8 - SUBPEL_BITS);
    filter = vp10_intra_filter_kernels[filter_type][filter_idx];

    if (filter_idx < (1 << SUBPEL_BITS)) {
      val = 0;
      for (k = 0; k < SUBPEL_TAPS; ++k) {
        idx = base + 1 - (SUBPEL_TAPS / 2) + k;
        idx = VPXMAX(VPXMIN(idx, ref_end_idx), ref_start_idx);
        val += ref[idx] * filter[k];
      }
      val = ROUND_POWER_OF_TWO(val, FILTER_BITS);
    } else {
      val = ref[base + 1];
    }
  }

  return val;
}

hui su's avatar
hui su committed
722 723 724
// Directional prediction, zone 1: 0 < angle < 90
static void highbd_dr_prediction_z1(uint16_t *dst, ptrdiff_t stride, int bs,
                                    const uint16_t *above, const uint16_t *left,
725 726
                                    int dx, int dy, int bd,
                                    INTRA_FILTER filter_type) {
hui su's avatar
hui su committed
727 728 729 730 731 732 733 734 735 736
  int r, c, x, y, base, shift, val;

  (void)left;
  (void)dy;
  assert(dy == 1);
  assert(dx < 0);

  for (r = 0; r < bs; ++r) {
    y = r + 1;
    for (c = 0; c < bs; ++c) {
737
      x = (c << 8) - y * dx;
hui su's avatar
hui su committed
738
      base = x >> 8;
739
      shift = x - (base << 8);
hui su's avatar
hui su committed
740
      if (base < 2 * bs - 1) {
741 742
        val = highbd_intra_subpel_interp(base, shift, above, 0, 2 * bs - 1,
                                         filter_type);
hui su's avatar
hui su committed
743 744 745 746 747 748 749 750 751 752 753 754
        dst[c] = clip_pixel_highbd(val, bd);
      } else {
        dst[c] = above[2 * bs - 1];
      }
    }
    dst += stride;
  }
}

// Directional prediction, zone 2: 90 < angle < 180
static void highbd_dr_prediction_z2(uint16_t *dst, ptrdiff_t stride, int bs,
                                    const uint16_t *above, const uint16_t *left,
755 756 757
                                    int dx, int dy, int bd,
                                    INTRA_FILTER filter_type) {
  int r, c, x, y, shift, val, base;
hui su's avatar
hui su committed
758 759 760 761 762 763 764

  assert(dx > 0);
  assert(dy > 0);

  for (r = 0; r < bs; ++r) {
    for (c = 0; c < bs; ++c) {
      y = r + 1;
765 766 767 768 769 770
      x = (c << 8) - y * dx;
      base = x >> 8;
      if (base >= -1) {
        shift = x - (base << 8);
        val = highbd_intra_subpel_interp(base, shift, above, -1, bs - 1,
                                         filter_type);
hui su's avatar
hui su committed
771 772
      } else {
        x = c + 1;
773
        y = (r << 8) - x * dy;
hui su's avatar
hui su committed
774 775
        base = y >> 8;
        if (base >= 0) {
776 777 778
          shift = y - (base  << 8);
          val = highbd_intra_subpel_interp(base, shift, left, 0, bs - 1,
                                           filter_type);
hui su's avatar
hui su committed
779
        } else {
780
          val = left[0];
hui su's avatar
hui su committed
781 782 783 784 785 786 787 788 789 790 791
        }
      }
      dst[c] = clip_pixel_highbd(val, bd);
    }
    dst += stride;
  }
}

// Directional prediction, zone 3: 180 < angle < 270
static void highbd_dr_prediction_z3(uint16_t *dst, ptrdiff_t stride, int bs,
                                    const uint16_t *above, const uint16_t *left,
792 793
                                    int dx, int dy, int bd,
                                    INTRA_FILTER filter_type) {
hui su's avatar
hui su committed
794 795 796 797 798 799 800 801 802 803
  int r, c, x, y, base, shift, val;

  (void)above;
  (void)dx;
  assert(dx == 1);
  assert(dy < 0);

  for (r = 0; r < bs; ++r) {
    for (c = 0; c < bs; ++c) {
      x = c + 1;
804
      y = (r << 8) - x * dy;
hui su's avatar
hui su committed
805
      base = y >> 8;
806
      shift = y - (base << 8);
807
      if (base < 2 * bs - 1) {
808 809
        val = highbd_intra_subpel_interp(base, shift, left, 0, 2 * bs - 1,
                                         filter_type);
hui su's avatar
hui su committed
810 811
        dst[c] = clip_pixel_highbd(val, bd);
      } else {
812
        dst[c] = left[2 * bs - 1];
hui su's avatar
hui su committed
813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844
      }
    }
    dst += stride;
  }
}

static INLINE void highbd_v_predictor(uint16_t *dst, ptrdiff_t stride,
                                      int bs, const uint16_t *above,
                                      const uint16_t *left, int bd) {
  int r;
  (void) left;
  (void) bd;
  for (r = 0; r < bs; r++) {
    memcpy(dst, above, bs * sizeof(uint16_t));
    dst += stride;
  }
}

static INLINE void highbd_h_predictor(uint16_t *dst, ptrdiff_t stride,
                                      int bs, const uint16_t *above,
                                      const uint16_t *left, int bd) {
  int r;
  (void) above;
  (void) bd;
  for (r = 0; r < bs; r++) {
    vpx_memset16(dst, left[r], bs);
    dst += stride;
  }
}

static void highbd_dr_predictor(uint16_t *dst, ptrdiff_t stride, int bs,
                                const uint16_t *above, const uint16_t *left,
845
                                int angle, int bd, INTRA_FILTER filter) {
hui su's avatar
hui su committed
846 847 848 849 850 851 852 853
  double t = 0;
  int dx, dy;

  if (angle != 90 && angle != 180)
    t = tan(angle * PI / 180.0);
  if (angle > 0 && angle < 90) {
    dx = -((int)(256 / t));
    dy = 1;
854
    highbd_dr_prediction_z1(dst, stride, bs, above, left, dx, dy, bd, filter);
hui su's avatar
hui su committed
855 856 857 858
  } else if (angle > 90 && angle < 180) {
    t = -t;
    dx = (int)(256 / t);
    dy = (int)(256 * t);
859
    highbd_dr_prediction_z2(dst, stride, bs, above, left, dx, dy, bd, filter);
hui su's avatar
hui su committed
860 861 862
  } else if (angle > 180 && angle < 270) {
    dx = 1;
    dy = -((int)(256 * t));
863
    highbd_dr_prediction_z3(dst, stride, bs, above, left, dx, dy, bd, filter);
hui su's avatar
hui su committed
864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999
  } else if (angle == 90) {
    highbd_v_predictor(dst, stride, bs, above, left, bd);
  } else if (angle == 180) {
    highbd_h_predictor(dst, stride, bs, above, left, bd);
  }
}

static void highbd_filter_intra_predictors_4tap(uint16_t *dst, ptrdiff_t stride,
                                                int bs, const uint16_t *above,
                                                const uint16_t *left, int mode,
                                                int bd) {
  int k, r, c;
  int pred[33][65];
  int mean, ipred;
  const TX_SIZE tx_size = (bs == 32) ? TX_32X32 :
      ((bs == 16) ? TX_16X16 : ((bs == 8) ? TX_8X8 : (TX_4X4)));
  const int c0 = filter_intra_taps_4[tx_size][mode][0];
  const int c1 = filter_intra_taps_4[tx_size][mode][1];
  const int c2 = filter_intra_taps_4[tx_size][mode][2];
  const int c3 = filter_intra_taps_4[tx_size][mode][3];

  k = 0;
  mean = 0;
  while (k < bs) {
    mean = mean + (int)left[k];
    mean = mean + (int)above[k];
    k++;
  }
  mean = (mean + bs) / (2 * bs);

  for (r = 0; r < bs; ++r)
    pred[r + 1][0] = (int)left[r] - mean;

  for (c = 0; c < 2 * bs + 1; ++c)
    pred[0][c] = (int)above[c - 1] - mean;

  for (r = 1; r < bs + 1; ++r)
    for (c = 1; c < 2 * bs + 1 - r; ++c) {
      ipred = c0 * pred[r - 1][c] + c1 * pred[r][c - 1] +
          c2 * pred[r - 1][c - 1] + c3 * pred[r - 1][c + 1];
      pred[r][c] = ipred < 0 ?
          -((-ipred + FILTER_INTRA_ROUND_VAL) >> FILTER_INTRA_PREC_BITS) :
          ((ipred + FILTER_INTRA_ROUND_VAL) >> FILTER_INTRA_PREC_BITS);
    }

  for (r = 0; r < bs; ++r) {
    for (c = 0; c < bs; ++c) {
      ipred = pred[r + 1][c + 1] + mean;
      dst[c] = clip_pixel_highbd(ipred, bd);
    }
    dst += stride;
  }
}

static void highbd_dc_filter_predictor(uint16_t *dst, ptrdiff_t stride,
                                       int bs, const uint16_t *above,
                                       const uint16_t *left, int bd) {
  highbd_filter_intra_predictors_4tap(dst, stride, bs, above, left, DC_PRED,
                                      bd);
}

static void highbd_v_filter_predictor(uint16_t *dst, ptrdiff_t stride,
                                      int bs, const uint16_t *above,
                                      const uint16_t *left, int bd) {
  highbd_filter_intra_predictors_4tap(dst, stride, bs, above, left, V_PRED,
                                      bd);
}

static void highbd_h_filter_predictor(uint16_t *dst, ptrdiff_t stride,
                                      int bs, const uint16_t *above,
                                      const uint16_t *left, int bd) {
  highbd_filter_intra_predictors_4tap(dst, stride, bs, above, left, H_PRED,
                                      bd);
}

static void highbd_d45_filter_predictor(uint16_t *dst, ptrdiff_t stride,
                                        int bs, const uint16_t *above,
                                        const uint16_t *left, int bd) {
  highbd_filter_intra_predictors_4tap(dst, stride, bs, above, left, D45_PRED,
                                      bd);
}

static void highbd_d135_filter_predictor(uint16_t *dst, ptrdiff_t stride,
                                         int bs, const uint16_t *above,
                                         const uint16_t *left, int bd) {
  highbd_filter_intra_predictors_4tap(dst, stride, bs, above, left, D135_PRED,
                                      bd);
}

static void highbd_d117_filter_predictor(uint16_t *dst, ptrdiff_t stride,
                                         int bs, const uint16_t *above,
                                         const uint16_t *left, int bd) {
  highbd_filter_intra_predictors_4tap(dst, stride, bs, above, left, D117_PRED,
                                      bd);
}

static void highbd_d153_filter_predictor(uint16_t *dst, ptrdiff_t stride,
                                         int bs, const uint16_t *above,
                                         const uint16_t *left, int bd) {
  highbd_filter_intra_predictors_4tap(dst, stride, bs, above, left, D153_PRED,
                                      bd);
}

static void highbd_d207_filter_predictor(uint16_t *dst, ptrdiff_t stride,
                                         int bs, const uint16_t *above,
                                         const uint16_t *left, int bd) {
  highbd_filter_intra_predictors_4tap(dst, stride, bs, above, left, D207_PRED,
                                      bd);
}

static void highbd_d63_filter_predictor(uint16_t *dst, ptrdiff_t stride,
                                        int bs, const uint16_t *above,
                                        const uint16_t *left, int bd) {
  highbd_filter_intra_predictors_4tap(dst, stride, bs, above, left, D63_PRED,
                                      bd);
}

static void highbd_tm_filter_predictor(uint16_t *dst, ptrdiff_t stride,
                                       int bs, const uint16_t *above,
                                       const uint16_t *left, int bd) {
  highbd_filter_intra_predictors_4tap(dst, stride, bs, above, left, TM_PRED,
                                      bd);
}

static void (*highbd_filter_intra_predictors[EXT_INTRA_MODES])(uint16_t *dst,
    ptrdiff_t stride, int bs, const uint16_t *above, const uint16_t *left,
    int bd) = {
        highbd_dc_filter_predictor, highbd_v_filter_predictor,
        highbd_h_filter_predictor, highbd_d45_filter_predictor,
        highbd_d135_filter_predictor, highbd_d117_filter_predictor,
        highbd_d153_filter_predictor, highbd_d207_filter_predictor,
        highbd_d63_filter_predictor, highbd_tm_filter_predictor,
};
#endif  // CONFIG_VP9_HIGHBITDEPTH
#endif  // CONFIG_EXT_INTRA

Jingning Han's avatar
Jingning Han committed
1000 1001 1002 1003 1004 1005 1006 1007
#if CONFIG_VP9_HIGHBITDEPTH
static void build_intra_predictors_high(const MACROBLOCKD *xd,
                                        const uint8_t *ref8,
                                        int ref_stride,
                                        uint8_t *dst8,
                                        int dst_stride,
                                        PREDICTION_MODE mode,
                                        TX_SIZE tx_size,
1008 1009
                                        int n_top_px, int n_topright_px,
                                        int n_left_px, int n_bottomleft_px,
hui su's avatar
hui su committed
1010
                                        int plane) {
Jingning Han's avatar
Jingning Han committed
1011 1012 1013
  int i;
  uint16_t *dst = CONVERT_TO_SHORTPTR(dst8);
  uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
1014
  DECLARE_ALIGNED(16, uint16_t, left_col[64]);
Jingning Han's avatar
Jingning Han committed
1015 1016 1017 1018
  DECLARE_ALIGNED(16, uint16_t, above_data[64 + 16]);
  uint16_t *above_row = above_data + 16;
  const uint16_t *const_above_row = above_row;
  const int bs = 4 << tx_size;
hui su's avatar
hui su committed
1019 1020
  int need_left = extend_modes[mode] & NEED_LEFT;
  int need_above = extend_modes[mode] & NEED_ABOVE;
1021
  const uint16_t *above_ref = ref - ref_stride;
hui su's avatar
hui su committed
1022
  int base = 128 << (xd->bd - 8);
Jingning Han's avatar
Jingning Han committed
1023 1024 1025 1026 1027 1028
  // 127 127 127 .. 127 127 127 127 127 127
  // 129  A   B  ..  Y   Z
  // 129  C   D  ..  W   X
  // 129  E   F  ..  U   V
  // 129  G   H  ..  S   T   T   T   T   T

hui su's avatar
hui su committed
1029 1030 1031 1032 1033
#if CONFIG_EXT_INTRA
  const EXT_INTRA_MODE_INFO *ext_intra_mode_info =
      &xd->mi[0]->mbmi.ext_intra_mode_info;
  const EXT_INTRA_MODE ext_intra_mode =
      ext_intra_mode_info->ext_intra_mode[plane != 0];
hui su's avatar
hui su committed
1034
  int p_angle = 0;
hui su's avatar
hui su committed
1035

hui su's avatar
hui su committed
1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
  if (mode != DC_PRED && mode != TM_PRED &&
      xd->mi[0]->mbmi.sb_type >= BLOCK_8X8) {
    p_angle = mode_to_angle_map[mode] +
        xd->mi[0]->mbmi.angle_delta[plane != 0] * ANGLE_STEP;
    if (p_angle <= 90)
      need_above = 1, need_left = 0;
    else if (p_angle < 180)
      need_above = 1, need_left = 1;
    else
      need_above = 0, need_left = 1;
  }

  if (ext_intra_mode_info->use_ext_intra_mode[plane != 0]) {
    EXT_INTRA_MODE ext_intra_mode =
        ext_intra_mode_info->ext_intra_mode[plane != 0];
    need_left = ext_intra_extend_modes[ext_intra_mode] & NEED_LEFT;
    need_above = ext_intra_extend_modes[ext_intra_mode] & NEED_ABOVE;
hui su's avatar
hui su committed
1053 1054 1055
  }
#endif  // CONFIG_EXT_INTRA

1056 1057 1058
  (void) plane;

  // NEED_LEFT
hui su's avatar
hui su committed
1059 1060 1061 1062 1063
  if (need_left) {
#if CONFIG_EXT_INTRA
    int need_bottom;
    if (ext_intra_mode_info->use_ext_intra_mode[plane != 0]) {
        need_bottom = 0;
hui su's avatar
hui su committed
1064 1065 1066
    } else if (mode != DC_PRED && mode != TM_PRED &&
        xd->mi[0]->mbmi.sb_type >= BLOCK_8X8) {
        need_bottom = p_angle > 180;
hui su's avatar
hui su committed
1067 1068 1069 1070
    } else {
      need_bottom = !!(extend_modes[mode] & NEED_BOTTOMLEFT);
    }
#else
1071
    const int need_bottom = !!(extend_modes[mode] & NEED_BOTTOMLEFT);
hui su's avatar
hui su committed
1072
#endif  // CONFIG_EXT_INTRA
1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
    i = 0;
    if (n_left_px > 0) {
      for (; i < n_left_px; i++)
        left_col[i] = ref[i * ref_stride - 1];
      if (need_bottom && n_bottomleft_px > 0) {
        assert(i == bs);
        for (; i < bs + n_bottomleft_px; i++)
          left_col[i] = ref[i * ref_stride - 1];
      }
      if (i < (bs << need_bottom))
Yaowu Xu's avatar
Yaowu Xu committed
1083
        vpx_memset16(&left_col[i], left_col[i - 1], (bs << need_bottom) - i);
1084
    } else {
Yaowu Xu's avatar
Yaowu Xu committed
1085
      vpx_memset16(left_col, base + 1, bs << need_bottom);
1086 1087 1088 1089
    }
  }

  // NEED_ABOVE
hui su's avatar
hui su committed
1090 1091 1092 1093
  if (need_above) {
#if CONFIG_EXT_INTRA
    int need_right;
    if (ext_intra_mode_info->use_ext_intra_mode[plane != 0]) {
hui su's avatar
hui su committed
1094 1095 1096 1097
      need_right = 1;
    } else if (mode != DC_PRED && mode != TM_PRED &&
        xd->mi[0]->mbmi.sb_type >= BLOCK_8X8) {
      need_right = p_angle < 90;
hui su's avatar
hui su committed
1098 1099 1100 1101
    } else {
      need_right = !!(extend_modes[mode] & NEED_ABOVERIGHT);
    }
#else
1102
    const int need_right = !!(extend_modes[mode] & NEED_ABOVERIGHT);
hui su's avatar
hui su committed
1103
#endif  // CONFIG_EXT_INTRA
1104 1105 1106 1107 1108 1109 1110 1111 1112
    if (n_top_px > 0) {
      memcpy(above_row, above_ref, n_top_px * 2);
      i = n_top_px;
      if (need_right && n_topright_px > 0) {
        assert(n_top_px == bs);
        memcpy(above_row + bs, above_ref + bs, n_topright_px * 2);
        i += n_topright_px;
      }
      if (i < (bs << need_right))
Yaowu Xu's avatar
Yaowu Xu committed
1113
        vpx_memset16(&above_row[i], above_row[i - 1], (bs << need_right) - i);
1114
    } else {
Yaowu Xu's avatar
Yaowu Xu committed
1115
      vpx_memset16(above_row, base - 1, bs << need_right);
1116 1117 1118
    }
  }

hui su's avatar
hui su committed
1119 1120
#if CONFIG_EXT_INTRA
  if (ext_intra_mode_info->use_ext_intra_mode[plane != 0] ||
hui su's avatar
hui su committed
1121 1122 1123
      (extend_modes[mode] & NEED_ABOVELEFT) ||
      (mode != DC_PRED && mode != TM_PRED &&
        xd->mi[0]->mbmi.sb_type >= BLOCK_8X8)) {
1124 1125 1126
    above_row[-1] = n_top_px > 0 ?
        (n_left_px > 0 ? above_ref[-1] : base + 1) : base - 1;
  }
hui su's avatar
hui su committed
1127 1128
#else
  if ((extend_modes[mode] & NEED_ABOVELEFT)) {
1129 1130 1131
    above_row[-1] = n_top_px > 0 ?
        (n_left_px > 0 ? above_ref[-1] : base + 1) : base - 1;
  }
hui su's avatar
hui su committed
1132
#endif  // CONFIG_EXT_INTRA
Jingning Han's avatar
Jingning Han committed
1133

hui su's avatar
hui su committed
1134 1135
#if CONFIG_EXT_INTRA
  if (ext_intra_mode_info->use_ext_intra_mode[plane != 0]) {
hui su's avatar
hui su committed
1136
    highbd_filter_intra_predictors[ext_intra_mode](dst, dst_stride, bs,
hui su's avatar
hui su committed
1137
        const_above_row, left_col, xd->bd);
hui su's avatar
hui su committed
1138 1139 1140 1141 1142
    return;
  }

  if (mode != DC_PRED && mode != TM_PRED &&
      xd->mi[0]->mbmi.sb_type >= BLOCK_8X8) {
1143 1144 1145
    INTRA_FILTER filter = INTRA_FILTER_LINEAR;
    if (plane == 0 && pick_intra_filter(p_angle))
      filter = xd->mi[0]->mbmi.intra_filter;
hui su's avatar
hui su committed
1146
    highbd_dr_predictor(dst, dst_stride, bs, const_above_row, left_col,
1147
                        p_angle, xd->bd, filter);
hui su's avatar
hui su committed
1148 1149 1150 1151
    return;
  }
#endif  // CONFIG_EXT_INTRA

Jingning Han's avatar
Jingning Han committed
1152 1153
  // predict
  if (mode == DC_PRED) {
1154 1155 1156
    dc_pred_high[n_left_px > 0][n_top_px > 0][tx_size](dst, dst_stride,
                                                       const_above_row,
                                                       left_col, xd->bd);
Jingning Han's avatar
Jingning Han committed
1157 1158 1159 1160 1161 1162 1163 1164 1165 1166
  } else {
    pred_high[mode][tx_size](dst, dst_stride, const_above_row, left_col,
                             xd->bd);
  }
}
#endif  // CONFIG_VP9_HIGHBITDEPTH

static void build_intra_predictors(const MACROBLOCKD *xd, const uint8_t *ref,
                                   int ref_stride, uint8_t *dst, int dst_stride,
                                   PREDICTION_MODE mode, TX_SIZE tx_size,
1167 1168
                                   int n_top_px, int n_topright_px,
                                   int n_left_px, int n_bottomleft_px,
hui su's avatar
hui su committed
1169
                                   int plane) {
Jingning Han's avatar
Jingning Han committed
1170
  int i;
1171 1172
  DECLARE_ALIGNED(16, uint8_t, left_col[64]);
  const uint8_t *above_ref = ref - ref_stride;
Jingning Han's avatar
Jingning Han committed
1173 1174 1175 1176
  DECLARE_ALIGNED(16, uint8_t, above_data[64 + 16]);
  uint8_t *above_row = above_data + 16;
  const uint8_t *const_above_row = above_row;
  const int bs = 4 << tx_size;
hui su's avatar
hui su committed
1177 1178 1179 1180 1181 1182 1183
  int need_left = extend_modes[mode] & NEED_LEFT;
  int need_above = extend_modes[mode] & NEED_ABOVE;
#if CONFIG_EXT_INTRA
  const EXT_INTRA_MODE_INFO *ext_intra_mode_info =
      &xd->mi[0]->mbmi.ext_intra_mode_info;
  const EXT_INTRA_MODE ext_intra_mode =
      ext_intra_mode_info->ext_intra_mode[plane != 0];
hui su's avatar
hui su committed
1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
  int p_angle = 0;

  if (mode != DC_PRED && mode != TM_PRED &&
      xd->mi[0]->mbmi.sb_type >= BLOCK_8X8) {
    p_angle = mode_to_angle_map[mode] +
        xd->mi[0]->mbmi.angle_delta[plane != 0] * ANGLE_STEP;
    if (p_angle <= 90)
      need_above = 1, need_left = 0;
    else if (p_angle < 180)
      need_above = 1, need_left = 1;
    else
      need_above = 0, need_left = 1;
  }

  if (ext_intra_mode_info->use_ext_intra_mode[plane != 0]) {
    EXT_INTRA_MODE ext_intra_mode =
        ext_intra_mode_info->ext_intra_mode[plane != 0];
    need_left = ext_intra_extend_modes[ext_intra_mode] & NEED_LEFT;
    need_above = ext_intra_extend_modes[ext_intra_mode] & NEED_ABOVE;
hui su's avatar
hui su committed
1203 1204
  }
#endif  // CONFIG_EXT_INTRA
Jingning Han's avatar
Jingning Han committed
1205 1206 1207 1208 1209 1210 1211 1212

  // 127 127 127 .. 127 127 127 127 127 127
  // 129  A   B  ..  Y   Z
  // 129  C   D  ..  W   X
  // 129  E   F  ..  U   V
  // 129  G   H  ..  S   T   T   T   T   T
  // ..

1213 1214 1215 1216 1217 1218
  (void) xd;
  (void) plane;
  assert(n_top_px >= 0);
  assert(n_topright_px >= 0);
  assert(n_left_px >= 0);
  assert(n_bottomleft_px >= 0);
Jingning Han's avatar
Jingning Han committed
1219 1220

  // NEED_LEFT
hui su's avatar
hui su committed
1221 1222 1223 1224
  if (need_left) {
#if CONFIG_EXT_INTRA
    int need_bottom;
    if (ext_intra_mode_info->use_ext_intra_mode[plane != 0]) {
hui su's avatar
hui su committed
1225 1226 1227 1228
      need_bottom = 0;
    } else if (mode != DC_PRED && mode != TM_PRED &&
        xd->mi[0]->mbmi.sb_type >= BLOCK_8X8) {
      need_bottom = p_angle > 180;
hui su's avatar
hui su committed
1229 1230 1231 1232
    } else {
      need_bottom = !!(extend_modes[mode] & NEED_BOTTOMLEFT);
    }
#else
1233
    const int need_bottom = !!(extend_modes[mode] & NEED_BOTTOMLEFT);
hui su's avatar
hui su committed
1234
#endif  // CONFIG_EXT_INTRA
1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248
    i = 0;
    if (n_left_px > 0) {
      for (; i < n_left_px; i++)
        left_col[i] = ref[i * ref_stride - 1];
      if (need_bottom && n_bottomleft_px > 0) {
        assert(i == bs);
        for (; i < bs + n_bottomleft_px; i++)
          left_col[i] = ref[i * ref_stride - 1];
      }
      if (i < (bs << need_bottom))
        memset(&left_col[i], left_col[i - 1], (bs << need_bottom) - i);
    } else {
      memset(left_col, 129, bs << need_bottom);
    }
Jingning Han's avatar
Jingning Han committed
1249 1250 1251
  }

  // NEED_ABOVE
hui su's avatar
hui su committed
1252 1253 1254 1255
  if (need_above) {
#if CONFIG_EXT_INTRA
    int need_right;
    if (ext_intra_mode_info->use_ext_intra_mode[plane != 0]) {
hui su's avatar
hui su committed
1256 1257 1258 1259
      need_right = 1;
    } else if (mode != DC_PRED && mode != TM_PRED &&
        xd->mi[0]->mbmi.sb_type >= BLOCK_8X8) {
      need_right = p_angle < 90;
hui su's avatar
hui su committed
1260 1261 1262 1263
    } else {
      need_right = !!(extend_modes[mode] & NEED_ABOVERIGHT);
    }
#else
1264
    const int need_right = !!(extend_modes[mode] & NEED_ABOVERIGHT);
hui su's avatar
hui su committed
1265
#endif  // CONFIG_EXT_INTRA
1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278
    if (n_top_px > 0) {
      memcpy(above_row, above_ref, n_top_px);
      i = n_top_px;
      if (need_right && n_topright_px > 0) {
        assert(n_top_px == bs);
        memcpy(above_row + bs, above_ref + bs, n_topright_px);
        i += n_topright_px;
      }
      if (i < (bs << need_right))
        memset(&above_row[i], above_row[i - 1], (bs << need_right) - i);
    } else {
      memset(above_row, 127, bs << need_right);
    }
Jingning Han's avatar
Jingning Han committed
1279 1280
  }

hui su's avatar
hui su committed
1281 1282
#if CONFIG_EXT_INTRA
  if (ext_intra_mode_info->use_ext_intra_mode[plane != 0] ||
hui su's avatar
hui su committed
1283 1284 1285
      (extend_modes[mode] & NEED_ABOVELEFT) ||
      (mode != DC_PRED && mode != TM_PRED &&
          xd->mi[0]->mbmi.sb_type >= BLOCK_8X8)) {
1286 1287
    above_row[-1] = n_top_px > 0 ? (n_left_px > 0 ? above_ref[-1] : 129) : 127;
  }
hui su's avatar
hui su committed
1288 1289
#else
  if ((extend_modes[mode] & NEED_ABOVELEFT)) {
1290 1291
    above_row[-1] = n_top_px > 0 ? (n_left_px > 0 ? above_ref[-1] : 129) : 127;
  }
hui su's avatar
hui su committed
1292
#endif  // CONFIG_EXT_INTRA
Jingning Han's avatar
Jingning Han committed
1293

hui su's avatar
hui su committed
1294 1295
#if CONFIG_EXT_INTRA
  if (ext_intra_mode_info->use_ext_intra_mode[plane != 0]) {
hui su's avatar
hui su committed
1296 1297 1298 1299 1300 1301 1302
    filter_intra_predictors[ext_intra_mode](dst, dst_stride, bs,
        const_above_row, left_col);
    return;
  }

  if (mode != DC_PRED && mode != TM_PRED &&
      xd->mi[0]->mbmi.sb_type >= BLOCK_8X8) {
1303 1304 1305 1306 1307
    INTRA_FILTER filter = INTRA_FILTER_LINEAR;
    if (plane == 0 && pick_intra_filter(p_angle))
      filter = xd->mi[0]->mbmi.intra_filter;
    dr_predictor(dst, dst_stride, tx_size, const_above_row, left_col, p_angle,
                 filter);
hui su's avatar
hui su committed
1308 1309 1310 1311
    return;
  }
#endif  // CONFIG_EXT_INTRA

Jingning Han's avatar
Jingning Han committed
1312 1313
  // predict
  if (mode == DC_PRED) {
1314 1315
    dc_pred[n_left_px > 0][n_top_px > 0][tx_size](dst, dst_stride,
                                                  const_above_row, left_col);
Jingning Han's avatar
Jingning Han committed
1316 1317 1318 1319 1320
  } else {
    pred[mode][tx_size](dst, dst_stride, const_above_row, left_col);
  }
}

1321
void vp10_predict_intra_block(const MACROBLOCKD *xd, int bwl_in, int bhl_in,
hui su's avatar
hui su committed
1322 1323 1324 1325
                              TX_SIZE tx_size, PREDICTION_MODE mode,
                              const uint8_t *ref, int ref_stride,
                              uint8_t *dst, int dst_stride,
                              int aoff, int loff, int plane) {
Jingning Han's avatar
Jingning Han committed
1326 1327 1328 1329 1330
  const int txw = (1 << tx_size);
  const int have_top = loff || xd->up_available;
  const int have_left = aoff || xd->left_available;
  const int x = aoff * 4;
  const int y = loff * 4;
1331 1332
  const int bw = VPXMAX(2, 1 << bwl_in);
  const int bh = VPXMAX(2, 1 << bhl_in);
hui su's avatar
hui su committed
1333 1334
  const int mi_row = -xd->mb_to_top_edge >> (3 + MI_SIZE_LOG2);
  const int mi_col = -xd->mb_to_left_edge >> (3 + MI_SIZE_LOG2);
1335 1336
  const BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
  const struct macroblockd_plane *const pd = &xd->plane[plane];
1337 1338
  const int right_available =
      mi_col + (bw >> !pd->subsampling_x) < xd->tile.mi_col_end;
1339
  const int have_right = vp10_has_right(bsize, mi_row, mi_col,
1340
                                        right_available,
1341 1342 1343 1344 1345 1346 1347 1348 1349
                                        tx_size, loff, aoff,
                                        pd->subsampling_x);
  const int have_bottom = vp10_has_bottom(bsize, mi_row, mi_col,
                                          xd->mb_to_bottom_edge > 0,
                                          tx_size, loff, aoff,
                                          pd->subsampling_y);
  const int wpx = 4 * bw;
  const int hpx = 4 * bh;
  const int txpx = 4 * txw;
hui su's avatar
hui su committed
1350 1351 1352 1353 1354 1355 1356 1357
  // Distance between the right edge of this prediction block to
  // the frame right edge
  const int xr = (xd->mb_to_right_edge >> (3 + pd->subsampling_x)) +
      (wpx - x - txpx);
  // Distance between the bottom edge of this prediction block to
  // the frame bottom edge
  const int yd = (xd->mb_to_bottom_edge >> (3 + pd->subsampling_y)) +
      (hpx - y - txpx);
Jingning Han's avatar
Jingning Han committed
1358

hui su's avatar
hui su committed
1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
  if (xd->mi[0]->mbmi.palette_mode_info.palette_size[plane != 0] > 0) {
    const int bs = 4 * (1 << tx_size);
    const int stride = 4 * (1 << bwl_in);
    int r, c;
    uint8_t *map = NULL;
#if CONFIG_VP9_HIGHBITDEPTH
    uint16_t *palette = xd->mi[0]->mbmi.palette_mode_info.palette_colors +
        plane * PALETTE_MAX_SIZE;
#else
    uint8_t *palette = xd->mi[0]->mbmi.palette_mode_info.palette_colors +
        plane * PALETTE_MAX_SIZE;
#endif  // CONFIG_VP9_HIGHBITDEPTH

    map = xd->plane[plane != 0].color_index_map;

#if CONFIG_VP9_HIGHBITDEPTH
    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
      uint16_t *dst16 = CONVERT_TO_SHORTPTR(dst);
      for (r = 0; r < bs; ++r)
        for (c = 0; c < bs; ++c)
          dst16[r * dst_stride + c] =
              palette[map[(r + y) * stride + c + x]];
    } else {
      for (r = 0; r < bs; ++r)
        for (c = 0; c < bs; ++c)
Yaowu Xu's avatar
Yaowu Xu committed
1384 1385
          dst[r * dst_stride + c] =
              (uint8_t)(palette[map[(r + y) * stride + c + x]]);
hui su's avatar
hui su committed
1386
    }
Yaowu Xu's avatar
Yaowu Xu committed
1387 1388 1389 1390
#else
    for (r = 0; r < bs; ++r)
      for (c = 0; c < bs; ++c)
        dst[r * dst_stride + c] = palette[map[(r + y) * stride + c + x]];
hui su's avatar
hui su committed
1391 1392 1393 1394
#endif  // CONFIG_VP9_HIGHBITDEPTH
    return;
  }

1395 1396 1397 1398 1399 1400 1401 1402
#if CONFIG_VP9_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    build_intra_predictors_high(xd, ref, ref_stride, dst, dst_stride, mode,
                                tx_size,
                                have_top ? VPXMIN(txpx, xr + txpx) : 0,
                                have_top && have_right ? VPXMIN(txpx, xr) : 0,
                                have_left ? VPXMIN(txpx, yd + txpx) : 0,
                                have_bottom && have_left ? VPXMIN(txpx, yd) : 0,
hui su's avatar
hui su committed
1403
                                plane);
1404 1405 1406 1407 1408 1409 1410 1411 1412
    return;
  }
#endif
  build_intra_predictors(xd, ref, ref_stride, dst, dst_stride, mode,
                         tx_size,
                         have_top ? VPXMIN(txpx, xr + txpx) : 0,
                         have_top && have_right ? VPXMIN(txpx, xr) : 0,
                         have_left ? VPXMIN(txpx, yd + txpx) : 0,
                         have_bottom && have_left ? VPXMIN(txpx, yd) : 0,