h264pred_template.c 41 KB
Newer Older
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
/*
 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
 * Copyright (c) 2003-2011 Michael Niedermayer <michaelni@gmx.at>
 *
 * This file is part of Libav.
 *
 * Libav is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * Libav is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with Libav; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file
 * H.264 / AVC / MPEG4 part10 prediction functions.
 * @author Michael Niedermayer <michaelni@gmx.at>
 */

28 29
#include "libavutil/intreadwrite.h"

30
#include "mathops.h"
31 32

#include "bit_depth_template.c"
33

34 35 36
static void FUNCC(pred4x4_vertical)(uint8_t *_src, const uint8_t *topright,
                                    ptrdiff_t _stride)
{
37 38
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
39 40 41 42 43 44
    const pixel4 a= AV_RN4PA(src-stride);

    AV_WN4PA(src+0*stride, a);
    AV_WN4PA(src+1*stride, a);
    AV_WN4PA(src+2*stride, a);
    AV_WN4PA(src+3*stride, a);
45 46
}

47 48 49
static void FUNCC(pred4x4_horizontal)(uint8_t *_src, const uint8_t *topright,
                                      ptrdiff_t _stride)
{
50 51
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
52 53 54 55
    AV_WN4PA(src+0*stride, PIXEL_SPLAT_X4(src[-1+0*stride]));
    AV_WN4PA(src+1*stride, PIXEL_SPLAT_X4(src[-1+1*stride]));
    AV_WN4PA(src+2*stride, PIXEL_SPLAT_X4(src[-1+2*stride]));
    AV_WN4PA(src+3*stride, PIXEL_SPLAT_X4(src[-1+3*stride]));
56 57
}

58 59 60
static void FUNCC(pred4x4_dc)(uint8_t *_src, const uint8_t *topright,
                              ptrdiff_t _stride)
{
61 62
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
63 64
    const int dc= (  src[-stride] + src[1-stride] + src[2-stride] + src[3-stride]
                   + src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 4) >>3;
65
    const pixel4 a = PIXEL_SPLAT_X4(dc);
66

67 68 69 70
    AV_WN4PA(src+0*stride, a);
    AV_WN4PA(src+1*stride, a);
    AV_WN4PA(src+2*stride, a);
    AV_WN4PA(src+3*stride, a);
71 72
}

73 74 75
static void FUNCC(pred4x4_left_dc)(uint8_t *_src, const uint8_t *topright,
                                   ptrdiff_t _stride)
{
76 77
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
78
    const int dc= (  src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 2) >>2;
79
    const pixel4 a = PIXEL_SPLAT_X4(dc);
80

81 82 83 84
    AV_WN4PA(src+0*stride, a);
    AV_WN4PA(src+1*stride, a);
    AV_WN4PA(src+2*stride, a);
    AV_WN4PA(src+3*stride, a);
85 86
}

87 88 89
static void FUNCC(pred4x4_top_dc)(uint8_t *_src, const uint8_t *topright,
                                  ptrdiff_t _stride)
{
90 91
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
92
    const int dc= (  src[-stride] + src[1-stride] + src[2-stride] + src[3-stride] + 2) >>2;
93
    const pixel4 a = PIXEL_SPLAT_X4(dc);
94

95 96 97 98
    AV_WN4PA(src+0*stride, a);
    AV_WN4PA(src+1*stride, a);
    AV_WN4PA(src+2*stride, a);
    AV_WN4PA(src+3*stride, a);
99 100
}

101 102 103
static void FUNCC(pred4x4_128_dc)(uint8_t *_src, const uint8_t *topright,
                                  ptrdiff_t _stride)
{
104 105
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
106 107 108 109 110 111
    const pixel4 a = PIXEL_SPLAT_X4(1<<(BIT_DEPTH-1));

    AV_WN4PA(src+0*stride, a);
    AV_WN4PA(src+1*stride, a);
    AV_WN4PA(src+2*stride, a);
    AV_WN4PA(src+3*stride, a);
112 113
}

114 115 116
static void FUNCC(pred4x4_127_dc)(uint8_t *_src, const uint8_t *topright,
                                  ptrdiff_t _stride)
{
117 118
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
119 120 121 122 123 124
    const pixel4 a = PIXEL_SPLAT_X4((1<<(BIT_DEPTH-1))-1);

    AV_WN4PA(src+0*stride, a);
    AV_WN4PA(src+1*stride, a);
    AV_WN4PA(src+2*stride, a);
    AV_WN4PA(src+3*stride, a);
125 126
}

127 128 129
static void FUNCC(pred4x4_129_dc)(uint8_t *_src, const uint8_t *topright,
                                  ptrdiff_t _stride)
{
130 131
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
132 133 134 135 136 137
    const pixel4 a = PIXEL_SPLAT_X4((1<<(BIT_DEPTH-1))+1);

    AV_WN4PA(src+0*stride, a);
    AV_WN4PA(src+1*stride, a);
    AV_WN4PA(src+2*stride, a);
    AV_WN4PA(src+3*stride, a);
138 139 140 141
}


#define LOAD_TOP_RIGHT_EDGE\
142 143 144 145
    const unsigned av_unused t4 = topright[0];\
    const unsigned av_unused t5 = topright[1];\
    const unsigned av_unused t6 = topright[2];\
    const unsigned av_unused t7 = topright[3];\
146 147

#define LOAD_DOWN_LEFT_EDGE\
148 149 150 151
    const unsigned av_unused l4 = src[-1+4*stride];\
    const unsigned av_unused l5 = src[-1+5*stride];\
    const unsigned av_unused l6 = src[-1+6*stride];\
    const unsigned av_unused l7 = src[-1+7*stride];\
152 153

#define LOAD_LEFT_EDGE\
154 155 156 157
    const unsigned av_unused l0 = src[-1+0*stride];\
    const unsigned av_unused l1 = src[-1+1*stride];\
    const unsigned av_unused l2 = src[-1+2*stride];\
    const unsigned av_unused l3 = src[-1+3*stride];\
158 159

#define LOAD_TOP_EDGE\
160 161 162 163
    const unsigned av_unused t0 = src[ 0-1*stride];\
    const unsigned av_unused t1 = src[ 1-1*stride];\
    const unsigned av_unused t2 = src[ 2-1*stride];\
    const unsigned av_unused t3 = src[ 3-1*stride];\
164

165 166 167
static void FUNCC(pred4x4_down_right)(uint8_t *_src, const uint8_t *topright,
                                      ptrdiff_t _stride)
{
168 169
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
    const int lt= src[-1-1*stride];
    LOAD_TOP_EDGE
    LOAD_LEFT_EDGE

    src[0+3*stride]=(l3 + 2*l2 + l1 + 2)>>2;
    src[0+2*stride]=
    src[1+3*stride]=(l2 + 2*l1 + l0 + 2)>>2;
    src[0+1*stride]=
    src[1+2*stride]=
    src[2+3*stride]=(l1 + 2*l0 + lt + 2)>>2;
    src[0+0*stride]=
    src[1+1*stride]=
    src[2+2*stride]=
    src[3+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
    src[1+0*stride]=
    src[2+1*stride]=
    src[3+2*stride]=(lt + 2*t0 + t1 + 2)>>2;
    src[2+0*stride]=
    src[3+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
    src[3+0*stride]=(t1 + 2*t2 + t3 + 2)>>2;
}

192 193 194
static void FUNCC(pred4x4_down_left)(uint8_t *_src, const uint8_t *_topright,
                                     ptrdiff_t _stride)
{
195 196 197
    pixel *src = (pixel*)_src;
    const pixel *topright = (const pixel*)_topright;
    int stride = _stride/sizeof(pixel);
198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219
    LOAD_TOP_EDGE
    LOAD_TOP_RIGHT_EDGE
//    LOAD_LEFT_EDGE

    src[0+0*stride]=(t0 + t2 + 2*t1 + 2)>>2;
    src[1+0*stride]=
    src[0+1*stride]=(t1 + t3 + 2*t2 + 2)>>2;
    src[2+0*stride]=
    src[1+1*stride]=
    src[0+2*stride]=(t2 + t4 + 2*t3 + 2)>>2;
    src[3+0*stride]=
    src[2+1*stride]=
    src[1+2*stride]=
    src[0+3*stride]=(t3 + t5 + 2*t4 + 2)>>2;
    src[3+1*stride]=
    src[2+2*stride]=
    src[1+3*stride]=(t4 + t6 + 2*t5 + 2)>>2;
    src[3+2*stride]=
    src[2+3*stride]=(t5 + t7 + 2*t6 + 2)>>2;
    src[3+3*stride]=(t6 + 3*t7 + 2)>>2;
}

220 221 222 223
static void FUNCC(pred4x4_vertical_right)(uint8_t *_src,
                                          const uint8_t *topright,
                                          ptrdiff_t _stride)
{
224 225
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247
    const int lt= src[-1-1*stride];
    LOAD_TOP_EDGE
    LOAD_LEFT_EDGE

    src[0+0*stride]=
    src[1+2*stride]=(lt + t0 + 1)>>1;
    src[1+0*stride]=
    src[2+2*stride]=(t0 + t1 + 1)>>1;
    src[2+0*stride]=
    src[3+2*stride]=(t1 + t2 + 1)>>1;
    src[3+0*stride]=(t2 + t3 + 1)>>1;
    src[0+1*stride]=
    src[1+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
    src[1+1*stride]=
    src[2+3*stride]=(lt + 2*t0 + t1 + 2)>>2;
    src[2+1*stride]=
    src[3+3*stride]=(t0 + 2*t1 + t2 + 2)>>2;
    src[3+1*stride]=(t1 + 2*t2 + t3 + 2)>>2;
    src[0+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
    src[0+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
}

248 249 250 251
static void FUNCC(pred4x4_vertical_left)(uint8_t *_src,
                                         const uint8_t *_topright,
                                         ptrdiff_t _stride)
{
252 253 254
    pixel *src = (pixel*)_src;
    const pixel *topright = (const pixel*)_topright;
    int stride = _stride/sizeof(pixel);
255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275
    LOAD_TOP_EDGE
    LOAD_TOP_RIGHT_EDGE

    src[0+0*stride]=(t0 + t1 + 1)>>1;
    src[1+0*stride]=
    src[0+2*stride]=(t1 + t2 + 1)>>1;
    src[2+0*stride]=
    src[1+2*stride]=(t2 + t3 + 1)>>1;
    src[3+0*stride]=
    src[2+2*stride]=(t3 + t4+ 1)>>1;
    src[3+2*stride]=(t4 + t5+ 1)>>1;
    src[0+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
    src[1+1*stride]=
    src[0+3*stride]=(t1 + 2*t2 + t3 + 2)>>2;
    src[2+1*stride]=
    src[1+3*stride]=(t2 + 2*t3 + t4 + 2)>>2;
    src[3+1*stride]=
    src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
    src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
}

276 277 278
static void FUNCC(pred4x4_horizontal_up)(uint8_t *_src, const uint8_t *topright,
                                         ptrdiff_t _stride)
{
279 280
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300
    LOAD_LEFT_EDGE

    src[0+0*stride]=(l0 + l1 + 1)>>1;
    src[1+0*stride]=(l0 + 2*l1 + l2 + 2)>>2;
    src[2+0*stride]=
    src[0+1*stride]=(l1 + l2 + 1)>>1;
    src[3+0*stride]=
    src[1+1*stride]=(l1 + 2*l2 + l3 + 2)>>2;
    src[2+1*stride]=
    src[0+2*stride]=(l2 + l3 + 1)>>1;
    src[3+1*stride]=
    src[1+2*stride]=(l2 + 2*l3 + l3 + 2)>>2;
    src[3+2*stride]=
    src[1+3*stride]=
    src[0+3*stride]=
    src[2+2*stride]=
    src[2+3*stride]=
    src[3+3*stride]=l3;
}

301 302 303 304
static void FUNCC(pred4x4_horizontal_down)(uint8_t *_src,
                                           const uint8_t *topright,
                                           ptrdiff_t _stride)
{
305 306
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328
    const int lt= src[-1-1*stride];
    LOAD_TOP_EDGE
    LOAD_LEFT_EDGE

    src[0+0*stride]=
    src[2+1*stride]=(lt + l0 + 1)>>1;
    src[1+0*stride]=
    src[3+1*stride]=(l0 + 2*lt + t0 + 2)>>2;
    src[2+0*stride]=(lt + 2*t0 + t1 + 2)>>2;
    src[3+0*stride]=(t0 + 2*t1 + t2 + 2)>>2;
    src[0+1*stride]=
    src[2+2*stride]=(l0 + l1 + 1)>>1;
    src[1+1*stride]=
    src[3+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
    src[0+2*stride]=
    src[2+3*stride]=(l1 + l2+ 1)>>1;
    src[1+2*stride]=
    src[3+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
    src[0+3*stride]=(l2 + l3 + 1)>>1;
    src[1+3*stride]=(l1 + 2*l2 + l3 + 2)>>2;
}

329 330
static void FUNCC(pred16x16_vertical)(uint8_t *_src, ptrdiff_t _stride)
{
331
    int i;
332 333
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
334 335 336 337
    const pixel4 a = AV_RN4PA(((pixel4*)(src-stride))+0);
    const pixel4 b = AV_RN4PA(((pixel4*)(src-stride))+1);
    const pixel4 c = AV_RN4PA(((pixel4*)(src-stride))+2);
    const pixel4 d = AV_RN4PA(((pixel4*)(src-stride))+3);
338 339

    for(i=0; i<16; i++){
340 341 342 343
        AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
        AV_WN4PA(((pixel4*)(src+i*stride))+2, c);
        AV_WN4PA(((pixel4*)(src+i*stride))+3, d);
344 345 346
    }
}

347 348
static void FUNCC(pred16x16_horizontal)(uint8_t *_src, ptrdiff_t stride)
{
349
    int i;
350 351
    pixel *src = (pixel*)_src;
    stride /= sizeof(pixel);
352 353

    for(i=0; i<16; i++){
354 355 356 357 358 359
        const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);

        AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
        AV_WN4PA(((pixel4*)(src+i*stride))+2, a);
        AV_WN4PA(((pixel4*)(src+i*stride))+3, a);
360 361 362
    }
}

363 364
#define PREDICT_16x16_DC(v)\
    for(i=0; i<16; i++){\
365 366 367 368
        AV_WN4PA(src+ 0, v);\
        AV_WN4PA(src+ 4, v);\
        AV_WN4PA(src+ 8, v);\
        AV_WN4PA(src+12, v);\
369 370 371
        src += stride;\
    }

372 373
static void FUNCC(pred16x16_dc)(uint8_t *_src, ptrdiff_t stride)
{
374
    int i, dc=0;
375 376 377
    pixel *src = (pixel*)_src;
    pixel4 dcsplat;
    stride /= sizeof(pixel);
378 379 380 381 382 383 384 385 386

    for(i=0;i<16; i++){
        dc+= src[-1+i*stride];
    }

    for(i=0;i<16; i++){
        dc+= src[i-stride];
    }

387 388
    dcsplat = PIXEL_SPLAT_X4((dc+16)>>5);
    PREDICT_16x16_DC(dcsplat);
389 390
}

391 392
static void FUNCC(pred16x16_left_dc)(uint8_t *_src, ptrdiff_t stride)
{
393
    int i, dc=0;
394 395 396
    pixel *src = (pixel*)_src;
    pixel4 dcsplat;
    stride /= sizeof(pixel);
397 398 399 400 401

    for(i=0;i<16; i++){
        dc+= src[-1+i*stride];
    }

402 403
    dcsplat = PIXEL_SPLAT_X4((dc+8)>>4);
    PREDICT_16x16_DC(dcsplat);
404 405
}

406 407
static void FUNCC(pred16x16_top_dc)(uint8_t *_src, ptrdiff_t stride)
{
408
    int i, dc=0;
409 410 411
    pixel *src = (pixel*)_src;
    pixel4 dcsplat;
    stride /= sizeof(pixel);
412 413 414 415 416

    for(i=0;i<16; i++){
        dc+= src[i-stride];
    }

417 418
    dcsplat = PIXEL_SPLAT_X4((dc+8)>>4);
    PREDICT_16x16_DC(dcsplat);
419 420
}

421
#define PRED16x16_X(n, v) \
422 423
static void FUNCC(pred16x16_##n##_dc)(uint8_t *_src, ptrdiff_t stride)\
{\
424 425 426 427
    int i;\
    pixel *src = (pixel*)_src;\
    stride /= sizeof(pixel);\
    PREDICT_16x16_DC(PIXEL_SPLAT_X4(v));\
428 429
}

430 431 432
PRED16x16_X(127, (1<<(BIT_DEPTH-1))-1)
PRED16x16_X(128, (1<<(BIT_DEPTH-1))+0)
PRED16x16_X(129, (1<<(BIT_DEPTH-1))+1)
433

434 435 436 437 438
static inline void FUNCC(pred16x16_plane_compat)(uint8_t *_src,
                                                 ptrdiff_t _stride,
                                                 const int svq3,
                                                 const int rv40)
{
439 440
  int i, j, k;
  int a;
441 442 443 444 445 446
  INIT_CLIP
  pixel *src = (pixel*)_src;
  int stride = _stride/sizeof(pixel);
  const pixel * const src0 = src +7-stride;
  const pixel *       src1 = src +8*stride-1;
  const pixel *       src2 = src1-2*stride;    // == src+6*stride-1;
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
  int H = src0[1] - src0[-1];
  int V = src1[0] - src2[ 0];
  for(k=2; k<=8; ++k) {
    src1 += stride; src2 -= stride;
    H += k*(src0[k] - src0[-k]);
    V += k*(src1[0] - src2[ 0]);
  }
  if(svq3){
    H = ( 5*(H/4) ) / 16;
    V = ( 5*(V/4) ) / 16;

    /* required for 100% accuracy */
    i = H; H = V; V = i;
  }else if(rv40){
    H = ( H + (H>>2) ) >> 4;
    V = ( V + (V>>2) ) >> 4;
  }else{
    H = ( 5*H+32 ) >> 6;
    V = ( 5*V+32 ) >> 6;
  }

  a = 16*(src1[0] + src2[16] + 1) - 7*(V+H);
  for(j=16; j>0; --j) {
    int b = a;
    a += V;
    for(i=-16; i<0; i+=4) {
473 474 475 476
      src[16+i] = CLIP((b    ) >> 5);
      src[17+i] = CLIP((b+  H) >> 5);
      src[18+i] = CLIP((b+2*H) >> 5);
      src[19+i] = CLIP((b+3*H) >> 5);
477 478 479 480 481 482
      b += 4*H;
    }
    src += stride;
  }
}

483 484
static void FUNCC(pred16x16_plane)(uint8_t *src, ptrdiff_t stride)
{
485
    FUNCC(pred16x16_plane_compat)(src, stride, 0, 0);
486 487
}

488 489
static void FUNCC(pred8x8_vertical)(uint8_t *_src, ptrdiff_t _stride)
{
490
    int i;
491 492
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
493 494
    const pixel4 a= AV_RN4PA(((pixel4*)(src-stride))+0);
    const pixel4 b= AV_RN4PA(((pixel4*)(src-stride))+1);
495 496

    for(i=0; i<8; i++){
497 498
        AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
499 500 501
    }
}

502 503
static void FUNCC(pred8x16_vertical)(uint8_t *_src, ptrdiff_t _stride)
{
504 505 506 507 508 509 510 511 512 513 514 515
    int i;
    pixel *src = (pixel*)_src;
    int stride = _stride>>(sizeof(pixel)-1);
    const pixel4 a= AV_RN4PA(((pixel4*)(src-stride))+0);
    const pixel4 b= AV_RN4PA(((pixel4*)(src-stride))+1);

    for(i=0; i<16; i++){
        AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
    }
}

516 517
static void FUNCC(pred8x8_horizontal)(uint8_t *_src, ptrdiff_t stride)
{
518
    int i;
519 520
    pixel *src = (pixel*)_src;
    stride /= sizeof(pixel);
521 522

    for(i=0; i<8; i++){
523 524 525
        const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
        AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
526 527 528
    }
}

529 530
static void FUNCC(pred8x16_horizontal)(uint8_t *_src, ptrdiff_t stride)
{
531 532 533 534 535 536 537 538 539 540
    int i;
    pixel *src = (pixel*)_src;
    stride >>= sizeof(pixel)-1;
    for(i=0; i<16; i++){
        const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
        AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
    }
}

541
#define PRED8x8_X(n, v)\
542 543
static void FUNCC(pred8x8_##n##_dc)(uint8_t *_src, ptrdiff_t stride)\
{\
544
    int i;\
545
    const pixel4 a = PIXEL_SPLAT_X4(v);\
546 547 548
    pixel *src = (pixel*)_src;\
    stride /= sizeof(pixel);\
    for(i=0; i<8; i++){\
549 550
        AV_WN4PA(((pixel4*)(src+i*stride))+0, a);\
        AV_WN4PA(((pixel4*)(src+i*stride))+1, a);\
551
    }\
552 553
}

554 555 556
PRED8x8_X(127, (1<<(BIT_DEPTH-1))-1)
PRED8x8_X(128, (1<<(BIT_DEPTH-1))+0)
PRED8x8_X(129, (1<<(BIT_DEPTH-1))+1)
557

558 559
static void FUNCC(pred8x16_128_dc)(uint8_t *_src, ptrdiff_t stride)
{
560 561 562 563
    FUNCC(pred8x8_128_dc)(_src, stride);
    FUNCC(pred8x8_128_dc)(_src+8*stride, stride);
}

564 565
static void FUNCC(pred8x8_left_dc)(uint8_t *_src, ptrdiff_t stride)
{
566 567
    int i;
    int dc0, dc2;
568 569 570
    pixel4 dc0splat, dc2splat;
    pixel *src = (pixel*)_src;
    stride /= sizeof(pixel);
571 572 573 574 575 576

    dc0=dc2=0;
    for(i=0;i<4; i++){
        dc0+= src[-1+i*stride];
        dc2+= src[-1+(i+4)*stride];
    }
577 578
    dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
    dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
579 580

    for(i=0; i<4; i++){
581 582
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc0splat);
583 584
    }
    for(i=4; i<8; i++){
585 586
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc2splat);
587 588 589
    }
}

590 591
static void FUNCC(pred8x16_left_dc)(uint8_t *_src, ptrdiff_t stride)
{
592 593 594 595
    FUNCC(pred8x8_left_dc)(_src, stride);
    FUNCC(pred8x8_left_dc)(_src+8*stride, stride);
}

596 597
static void FUNCC(pred8x8_top_dc)(uint8_t *_src, ptrdiff_t stride)
{
598 599
    int i;
    int dc0, dc1;
600 601 602
    pixel4 dc0splat, dc1splat;
    pixel *src = (pixel*)_src;
    stride /= sizeof(pixel);
603 604 605 606 607 608

    dc0=dc1=0;
    for(i=0;i<4; i++){
        dc0+= src[i-stride];
        dc1+= src[4+i-stride];
    }
609 610
    dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
    dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
611 612

    for(i=0; i<4; i++){
613 614
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
615 616
    }
    for(i=4; i<8; i++){
617 618
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
619 620 621
    }
}

622 623
static void FUNCC(pred8x16_top_dc)(uint8_t *_src, ptrdiff_t stride)
{
624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643
    int i;
    int dc0, dc1;
    pixel4 dc0splat, dc1splat;
    pixel *src = (pixel*)_src;
    stride >>= sizeof(pixel)-1;

    dc0=dc1=0;
    for(i=0;i<4; i++){
        dc0+= src[i-stride];
        dc1+= src[4+i-stride];
    }
    dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
    dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);

    for(i=0; i<16; i++){
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
    }
}

644 645
static void FUNCC(pred8x8_dc)(uint8_t *_src, ptrdiff_t stride)
{
646
    int i;
647 648 649 650
    int dc0, dc1, dc2;
    pixel4 dc0splat, dc1splat, dc2splat, dc3splat;
    pixel *src = (pixel*)_src;
    stride /= sizeof(pixel);
651 652 653 654 655 656 657

    dc0=dc1=dc2=0;
    for(i=0;i<4; i++){
        dc0+= src[-1+i*stride] + src[i-stride];
        dc1+= src[4+i-stride];
        dc2+= src[-1+(i+4)*stride];
    }
658 659 660 661
    dc0splat = PIXEL_SPLAT_X4((dc0 + 4)>>3);
    dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
    dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
    dc3splat = PIXEL_SPLAT_X4((dc1 + dc2 + 4)>>3);
662 663

    for(i=0; i<4; i++){
664 665
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
666 667
    }
    for(i=4; i<8; i++){
668 669
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc3splat);
670 671 672
    }
}

673 674
static void FUNCC(pred8x16_dc)(uint8_t *_src, ptrdiff_t stride)
{
675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715
    int i;
    int dc0, dc1, dc2, dc3, dc4;
    pixel4 dc0splat, dc1splat, dc2splat, dc3splat, dc4splat, dc5splat, dc6splat, dc7splat;
    pixel *src = (pixel*)_src;
    stride >>= sizeof(pixel)-1;

    dc0=dc1=dc2=dc3=dc4=0;
    for(i=0;i<4; i++){
        dc0+= src[-1+i*stride] + src[i-stride];
        dc1+= src[4+i-stride];
        dc2+= src[-1+(i+4)*stride];
        dc3+= src[-1+(i+8)*stride];
        dc4+= src[-1+(i+12)*stride];
    }
    dc0splat = PIXEL_SPLAT_X4((dc0 + 4)>>3);
    dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
    dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
    dc3splat = PIXEL_SPLAT_X4((dc1 + dc2 + 4)>>3);
    dc4splat = PIXEL_SPLAT_X4((dc3 + 2)>>2);
    dc5splat = PIXEL_SPLAT_X4((dc1 + dc3 + 4)>>3);
    dc6splat = PIXEL_SPLAT_X4((dc4 + 2)>>2);
    dc7splat = PIXEL_SPLAT_X4((dc1 + dc4 + 4)>>3);

    for(i=0; i<4; i++){
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
    }
    for(i=4; i<8; i++){
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc3splat);
    }
    for(i=8; i<12; i++){
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc4splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc5splat);
    }
    for(i=12; i<16; i++){
        AV_WN4PA(((pixel4*)(src+i*stride))+0, dc6splat);
        AV_WN4PA(((pixel4*)(src+i*stride))+1, dc7splat);
    }
}

716 717
static void FUNC(pred8x8_mad_cow_dc_l0t)(uint8_t *src, ptrdiff_t stride)
{
718 719
    FUNCC(pred8x8_top_dc)(src, stride);
    FUNCC(pred4x4_dc)(src, NULL, stride);
720 721
}

722 723
static void FUNC(pred8x16_mad_cow_dc_l0t)(uint8_t *src, ptrdiff_t stride)
{
724 725 726 727
    FUNCC(pred8x16_top_dc)(src, stride);
    FUNCC(pred4x4_dc)(src, NULL, stride);
}

728 729
static void FUNC(pred8x8_mad_cow_dc_0lt)(uint8_t *src, ptrdiff_t stride)
{
730 731
    FUNCC(pred8x8_dc)(src, stride);
    FUNCC(pred4x4_top_dc)(src, NULL, stride);
732 733
}

734 735
static void FUNC(pred8x16_mad_cow_dc_0lt)(uint8_t *src, ptrdiff_t stride)
{
736 737 738 739
    FUNCC(pred8x16_dc)(src, stride);
    FUNCC(pred4x4_top_dc)(src, NULL, stride);
}

740 741
static void FUNC(pred8x8_mad_cow_dc_l00)(uint8_t *src, ptrdiff_t stride)
{
742 743 744
    FUNCC(pred8x8_left_dc)(src, stride);
    FUNCC(pred4x4_128_dc)(src + 4*stride                  , NULL, stride);
    FUNCC(pred4x4_128_dc)(src + 4*stride + 4*sizeof(pixel), NULL, stride);
745 746
}

747 748
static void FUNC(pred8x16_mad_cow_dc_l00)(uint8_t *src, ptrdiff_t stride)
{
749 750 751 752 753
    FUNCC(pred8x16_left_dc)(src, stride);
    FUNCC(pred4x4_128_dc)(src + 4*stride                  , NULL, stride);
    FUNCC(pred4x4_128_dc)(src + 4*stride + 4*sizeof(pixel), NULL, stride);
}

754 755
static void FUNC(pred8x8_mad_cow_dc_0l0)(uint8_t *src, ptrdiff_t stride)
{
756 757 758
    FUNCC(pred8x8_left_dc)(src, stride);
    FUNCC(pred4x4_128_dc)(src                  , NULL, stride);
    FUNCC(pred4x4_128_dc)(src + 4*sizeof(pixel), NULL, stride);
759 760
}

761 762
static void FUNC(pred8x16_mad_cow_dc_0l0)(uint8_t *src, ptrdiff_t stride)
{
763 764 765 766 767
    FUNCC(pred8x16_left_dc)(src, stride);
    FUNCC(pred4x4_128_dc)(src                  , NULL, stride);
    FUNCC(pred4x4_128_dc)(src + 4*sizeof(pixel), NULL, stride);
}

768 769
static void FUNCC(pred8x8_plane)(uint8_t *_src, ptrdiff_t _stride)
{
770 771
  int j, k;
  int a;
772 773 774 775 776 777
  INIT_CLIP
  pixel *src = (pixel*)_src;
  int stride = _stride/sizeof(pixel);
  const pixel * const src0 = src +3-stride;
  const pixel *       src1 = src +4*stride-1;
  const pixel *       src2 = src1-2*stride;    // == src+2*stride-1;
778 779 780 781 782 783 784 785 786 787 788 789 790 791
  int H = src0[1] - src0[-1];
  int V = src1[0] - src2[ 0];
  for(k=2; k<=4; ++k) {
    src1 += stride; src2 -= stride;
    H += k*(src0[k] - src0[-k]);
    V += k*(src1[0] - src2[ 0]);
  }
  H = ( 17*H+16 ) >> 5;
  V = ( 17*V+16 ) >> 5;

  a = 16*(src1[0] + src2[8]+1) - 3*(V+H);
  for(j=8; j>0; --j) {
    int b = a;
    a += V;
792 793 794 795 796
    src[0] = CLIP((b    ) >> 5);
    src[1] = CLIP((b+  H) >> 5);
    src[2] = CLIP((b+2*H) >> 5);
    src[3] = CLIP((b+3*H) >> 5);
    src[4] = CLIP((b+4*H) >> 5);
797 798 799 800 801 802 803
    src[5] = CLIP((b+5*H) >> 5);
    src[6] = CLIP((b+6*H) >> 5);
    src[7] = CLIP((b+7*H) >> 5);
    src += stride;
  }
}

804 805
static void FUNCC(pred8x16_plane)(uint8_t *_src, ptrdiff_t _stride)
{
806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838
  int j, k;
  int a;
  INIT_CLIP
  pixel *src = (pixel*)_src;
  int stride = _stride>>(sizeof(pixel)-1);
  const pixel * const src0 = src +3-stride;
  const pixel *       src1 = src +8*stride-1;
  const pixel *       src2 = src1-2*stride;    // == src+6*stride-1;
  int H = src0[1] - src0[-1];
  int V = src1[0] - src2[ 0];

  for (k = 2; k <= 4; ++k) {
      src1 += stride; src2 -= stride;
      H += k*(src0[k] - src0[-k]);
      V += k*(src1[0] - src2[ 0]);
  }
  for (; k <= 8; ++k) {
      src1 += stride; src2 -= stride;
      V += k*(src1[0] - src2[0]);
  }

  H = (17*H+16) >> 5;
  V = (5*V+32) >> 6;

  a = 16*(src1[0] + src2[8] + 1) - 7*V - 3*H;
  for(j=16; j>0; --j) {
    int b = a;
    a += V;
    src[0] = CLIP((b    ) >> 5);
    src[1] = CLIP((b+  H) >> 5);
    src[2] = CLIP((b+2*H) >> 5);
    src[3] = CLIP((b+3*H) >> 5);
    src[4] = CLIP((b+4*H) >> 5);
839 840 841
    src[5] = CLIP((b+5*H) >> 5);
    src[6] = CLIP((b+6*H) >> 5);
    src[7] = CLIP((b+7*H) >> 5);
842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878
    src += stride;
  }
}

#define SRC(x,y) src[(x)+(y)*stride]
#define PL(y) \
    const int l##y = (SRC(-1,y-1) + 2*SRC(-1,y) + SRC(-1,y+1) + 2) >> 2;
#define PREDICT_8x8_LOAD_LEFT \
    const int l0 = ((has_topleft ? SRC(-1,-1) : SRC(-1,0)) \
                     + 2*SRC(-1,0) + SRC(-1,1) + 2) >> 2; \
    PL(1) PL(2) PL(3) PL(4) PL(5) PL(6) \
    const int l7 av_unused = (SRC(-1,6) + 3*SRC(-1,7) + 2) >> 2

#define PT(x) \
    const int t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
#define PREDICT_8x8_LOAD_TOP \
    const int t0 = ((has_topleft ? SRC(-1,-1) : SRC(0,-1)) \
                     + 2*SRC(0,-1) + SRC(1,-1) + 2) >> 2; \
    PT(1) PT(2) PT(3) PT(4) PT(5) PT(6) \
    const int t7 av_unused = ((has_topright ? SRC(8,-1) : SRC(7,-1)) \
                     + 2*SRC(7,-1) + SRC(6,-1) + 2) >> 2

#define PTR(x) \
    t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
#define PREDICT_8x8_LOAD_TOPRIGHT \
    int t8, t9, t10, t11, t12, t13, t14, t15; \
    if(has_topright) { \
        PTR(8) PTR(9) PTR(10) PTR(11) PTR(12) PTR(13) PTR(14) \
        t15 = (SRC(14,-1) + 3*SRC(15,-1) + 2) >> 2; \
    } else t8=t9=t10=t11=t12=t13=t14=t15= SRC(7,-1);

#define PREDICT_8x8_LOAD_TOPLEFT \
    const int lt = (SRC(-1,0) + 2*SRC(-1,-1) + SRC(0,-1) + 2) >> 2

#define PREDICT_8x8_DC(v) \
    int y; \
    for( y = 0; y < 8; y++ ) { \
879 880
        AV_WN4PA(((pixel4*)src)+0, v); \
        AV_WN4PA(((pixel4*)src)+1, v); \
881 882 883
        src += stride; \
    }

884 885
static void FUNCC(pred8x8l_128_dc)(uint8_t *_src, int has_topleft,
                                   int has_topright, ptrdiff_t _stride)
886
{
887 888 889 890
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);

    PREDICT_8x8_DC(PIXEL_SPLAT_X4(1<<(BIT_DEPTH-1)));
891
}
892 893
static void FUNCC(pred8x8l_left_dc)(uint8_t *_src, int has_topleft,
                                    int has_topright, ptrdiff_t _stride)
894
{
895 896 897
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);

898
    PREDICT_8x8_LOAD_LEFT;
899
    const pixel4 dc = PIXEL_SPLAT_X4((l0+l1+l2+l3+l4+l5+l6+l7+4) >> 3);
900 901
    PREDICT_8x8_DC(dc);
}
902 903
static void FUNCC(pred8x8l_top_dc)(uint8_t *_src, int has_topleft,
                                   int has_topright, ptrdiff_t _stride)
904
{
905 906 907
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);

908
    PREDICT_8x8_LOAD_TOP;
909
    const pixel4 dc = PIXEL_SPLAT_X4((t0+t1+t2+t3+t4+t5+t6+t7+4) >> 3);
910 911
    PREDICT_8x8_DC(dc);
}
912 913
static void FUNCC(pred8x8l_dc)(uint8_t *_src, int has_topleft,
                               int has_topright, ptrdiff_t _stride)
914
{
915 916 917
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);

918 919
    PREDICT_8x8_LOAD_LEFT;
    PREDICT_8x8_LOAD_TOP;
920 921
    const pixel4 dc = PIXEL_SPLAT_X4((l0+l1+l2+l3+l4+l5+l6+l7
                                     +t0+t1+t2+t3+t4+t5+t6+t7+8) >> 4);
922 923
    PREDICT_8x8_DC(dc);
}
924 925
static void FUNCC(pred8x8l_horizontal)(uint8_t *_src, int has_topleft,
                                       int has_topright, ptrdiff_t _stride)
926
{
927 928
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
929
    pixel4 a;
930

931
    PREDICT_8x8_LOAD_LEFT;
932 933 934
#define ROW(y) a = PIXEL_SPLAT_X4(l##y); \
               AV_WN4PA(src+y*stride, a); \
               AV_WN4PA(src+y*stride+4, a);
935 936 937
    ROW(0); ROW(1); ROW(2); ROW(3); ROW(4); ROW(5); ROW(6); ROW(7);
#undef ROW
}
938 939
static void FUNCC(pred8x8l_vertical)(uint8_t *_src, int has_topleft,
                                     int has_topright, ptrdiff_t _stride)
940 941
{
    int y;
942 943
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
944
    pixel4 a, b;
945

946 947 948 949 950 951 952 953 954
    PREDICT_8x8_LOAD_TOP;
    src[0] = t0;
    src[1] = t1;
    src[2] = t2;
    src[3] = t3;
    src[4] = t4;
    src[5] = t5;
    src[6] = t6;
    src[7] = t7;
955 956
    a = AV_RN4PA(((pixel4*)src)+0);
    b = AV_RN4PA(((pixel4*)src)+1);
957
    for( y = 1; y < 8; y++ ) {
958 959
        AV_WN4PA(((pixel4*)(src+y*stride))+0, a);
        AV_WN4PA(((pixel4*)(src+y*stride))+1, b);
960
    }
961
}
962 963
static void FUNCC(pred8x8l_down_left)(uint8_t *_src, int has_topleft,
                                      int has_topright, ptrdiff_t _stride)
964
{
965 966
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984
    PREDICT_8x8_LOAD_TOP;
    PREDICT_8x8_LOAD_TOPRIGHT;
    SRC(0,0)= (t0 + 2*t1 + t2 + 2) >> 2;
    SRC(0,1)=SRC(1,0)= (t1 + 2*t2 + t3 + 2) >> 2;
    SRC(0,2)=SRC(1,1)=SRC(2,0)= (t2 + 2*t3 + t4 + 2) >> 2;
    SRC(0,3)=SRC(1,2)=SRC(2,1)=SRC(3,0)= (t3 + 2*t4 + t5 + 2) >> 2;
    SRC(0,4)=SRC(1,3)=SRC(2,2)=SRC(3,1)=SRC(4,0)= (t4 + 2*t5 + t6 + 2) >> 2;
    SRC(0,5)=SRC(1,4)=SRC(2,3)=SRC(3,2)=SRC(4,1)=SRC(5,0)= (t5 + 2*t6 + t7 + 2) >> 2;
    SRC(0,6)=SRC(1,5)=SRC(2,4)=SRC(3,3)=SRC(4,2)=SRC(5,1)=SRC(6,0)= (t6 + 2*t7 + t8 + 2) >> 2;
    SRC(0,7)=SRC(1,6)=SRC(2,5)=SRC(3,4)=SRC(4,3)=SRC(5,2)=SRC(6,1)=SRC(7,0)= (t7 + 2*t8 + t9 + 2) >> 2;
    SRC(1,7)=SRC(2,6)=SRC(3,5)=SRC(4,4)=SRC(5,3)=SRC(6,2)=SRC(7,1)= (t8 + 2*t9 + t10 + 2) >> 2;
    SRC(2,7)=SRC(3,6)=SRC(4,5)=SRC(5,4)=SRC(6,3)=SRC(7,2)= (t9 + 2*t10 + t11 + 2) >> 2;
    SRC(3,7)=SRC(4,6)=SRC(5,5)=SRC(6,4)=SRC(7,3)= (t10 + 2*t11 + t12 + 2) >> 2;
    SRC(4,7)=SRC(5,6)=SRC(6,5)=SRC(7,4)= (t11 + 2*t12 + t13 + 2) >> 2;
    SRC(5,7)=SRC(6,6)=SRC(7,5)= (t12 + 2*t13 + t14 + 2) >> 2;
    SRC(6,7)=SRC(7,6)= (t13 + 2*t14 + t15 + 2) >> 2;
    SRC(7,7)= (t14 + 3*t15 + 2) >> 2;
}
985 986
static void FUNCC(pred8x8l_down_right)(uint8_t *_src, int has_topleft,
                                       int has_topright, ptrdiff_t _stride)
987
{
988 989
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008
    PREDICT_8x8_LOAD_TOP;
    PREDICT_8x8_LOAD_LEFT;
    PREDICT_8x8_LOAD_TOPLEFT;
    SRC(0,7)= (l7 + 2*l6 + l5 + 2) >> 2;
    SRC(0,6)=SRC(1,7)= (l6 + 2*l5 + l4 + 2) >> 2;
    SRC(0,5)=SRC(1,6)=SRC(2,7)= (l5 + 2*l4 + l3 + 2) >> 2;
    SRC(0,4)=SRC(1,5)=SRC(2,6)=SRC(3,7)= (l4 + 2*l3 + l2 + 2) >> 2;
    SRC(0,3)=SRC(1,4)=SRC(2,5)=SRC(3,6)=SRC(4,7)= (l3 + 2*l2 + l1 + 2) >> 2;
    SRC(0,2)=SRC(1,3)=SRC(2,4)=SRC(3,5)=SRC(4,6)=SRC(5,7)= (l2 + 2*l1 + l0 + 2) >> 2;
    SRC(0,1)=SRC(1,2)=SRC(2,3)=SRC(3,4)=SRC(4,5)=SRC(5,6)=SRC(6,7)= (l1 + 2*l0 + lt + 2) >> 2;
    SRC(0,0)=SRC(1,1)=SRC(2,2)=SRC(3,3)=SRC(4,4)=SRC(5,5)=SRC(6,6)=SRC(7,7)= (l0 + 2*lt + t0 + 2) >> 2;
    SRC(1,0)=SRC(2,1)=SRC(3,2)=SRC(4,3)=SRC(5,4)=SRC(6,5)=SRC(7,6)= (lt + 2*t0 + t1 + 2) >> 2;
    SRC(2,0)=SRC(3,1)=SRC(4,2)=SRC(5,3)=SRC(6,4)=SRC(7,5)= (t0 + 2*t1 + t2 + 2) >> 2;
    SRC(3,0)=SRC(4,1)=SRC(5,2)=SRC(6,3)=SRC(7,4)= (t1 + 2*t2 + t3 + 2) >> 2;
    SRC(4,0)=SRC(5,1)=SRC(6,2)=SRC(7,3)= (t2 + 2*t3 + t4 + 2) >> 2;
    SRC(5,0)=SRC(6,1)=SRC(7,2)= (t3 + 2*t4 + t5 + 2) >> 2;
    SRC(6,0)=SRC(7,1)= (t4 + 2*t5 + t6 + 2) >> 2;
    SRC(7,0)= (t5 + 2*t6 + t7 + 2) >> 2;
}
1009 1010
static void FUNCC(pred8x8l_vertical_right)(uint8_t *_src, int has_topleft,
                                           int has_topright, ptrdiff_t _stride)
1011
{
1012 1013
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
    PREDICT_8x8_LOAD_TOP;
    PREDICT_8x8_LOAD_LEFT;
    PREDICT_8x8_LOAD_TOPLEFT;
    SRC(0,6)= (l5 + 2*l4 + l3 + 2) >> 2;
    SRC(0,7)= (l6 + 2*l5 + l4 + 2) >> 2;
    SRC(0,4)=SRC(1,6)= (l3 + 2*l2 + l1 + 2) >> 2;
    SRC(0,5)=SRC(1,7)= (l4 + 2*l3 + l2 + 2) >> 2;
    SRC(0,2)=SRC(1,4)=SRC(2,6)= (l1 + 2*l0 + lt + 2) >> 2;
    SRC(0,3)=SRC(1,5)=SRC(2,7)= (l2 + 2*l1 + l0 + 2) >> 2;
    SRC(0,1)=SRC(1,3)=SRC(2,5)=SRC(3,7)= (l0 + 2*lt + t0 + 2) >> 2;
    SRC(0,0)=SRC(1,2)=SRC(2,4)=SRC(3,6)= (lt + t0 + 1) >> 1;
    SRC(1,1)=SRC(2,3)=SRC(3,5)=SRC(4,7)= (lt + 2*t0 + t1 + 2) >> 2;
    SRC(1,0)=SRC(2,2)=SRC(3,4)=SRC(4,6)= (t0 + t1 + 1) >> 1;
    SRC(2,1)=SRC(3,3)=SRC(4,5)=SRC(5,7)= (t0 + 2*t1 + t2 + 2) >> 2;
    SRC(2,0)=SRC(3,2)=SRC(4,4)=SRC(5,6)= (t1 + t2 + 1) >> 1;
    SRC(3,1)=SRC(4,3)=SRC(5,5)=SRC(6,7)= (t1 + 2*t2 + t3 + 2) >> 2;
    SRC(3,0)=SRC(4,2)=SRC(5,4)=SRC(6,6)= (t2 + t3 + 1) >> 1;
    SRC(4,1)=SRC(5,3)=SRC(6,5)=SRC(7,7)= (t2 + 2*t3 + t4 + 2) >> 2;
    SRC(4,0)=SRC(5,2)=SRC(6,4)=SRC(7,6)= (t3 + t4 + 1) >> 1;
    SRC(5,1)=SRC(6,3)=SRC(7,5)= (t3 + 2*t4 + t5 + 2) >> 2;
    SRC(5,0)=SRC(6,2)=SRC(7,4)= (t4 + t5 + 1) >> 1;
    SRC(6,1)=SRC(7,3)= (t4 + 2*t5 + t6 + 2) >> 2;
    SRC(6,0)=SRC(7,2)= (t5 + t6 + 1) >> 1;
    SRC(7,1)= (t5 + 2*t6 + t7 + 2) >> 2;
    SRC(7,0)= (t6 + t7 + 1) >> 1;
}
1040 1041
static void FUNCC(pred8x8l_horizontal_down)(uint8_t *_src, int has_topleft,
                                            int has_topright, ptrdiff_t _stride)
1042
{
1043 1044
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070
    PREDICT_8x8_LOAD_TOP;
    PREDICT_8x8_LOAD_LEFT;
    PREDICT_8x8_LOAD_TOPLEFT;
    SRC(0,7)= (l6 + l7 + 1) >> 1;
    SRC(1,7)= (l5 + 2*l6 + l7 + 2) >> 2;
    SRC(0,6)=SRC(2,7)= (l5 + l6 + 1) >> 1;
    SRC(1,6)=SRC(3,7)= (l4 + 2*l5 + l6 + 2) >> 2;
    SRC(0,5)=SRC(2,6)=SRC(4,7)= (l4 + l5 + 1) >> 1;
    SRC(1,5)=SRC(3,6)=SRC(5,7)= (l3 + 2*l4 + l5 + 2) >> 2;
    SRC(0,4)=SRC(2,5)=SRC(4,6)=SRC(6,7)= (l3 + l4 + 1) >> 1;
    SRC(1,4)=SRC(3,5)=SRC(5,6)=SRC(7,7)= (l2 + 2*l3 + l4 + 2) >> 2;
    SRC(0,3)=SRC(2,4)=SRC(4,5)=SRC(6,6)= (l2 + l3 + 1) >> 1;
    SRC(1,3)=SRC(3,4)=SRC(5,5)=SRC(7,6)= (l1 + 2*l2 + l3 + 2) >> 2;
    SRC(0,2)=SRC(2,3)=SRC(4,4)=SRC(6,5)= (l1 + l2 + 1) >> 1;
    SRC(1,2)=SRC(3,3)=SRC(5,4)=SRC(7,5)= (l0 + 2*l1 + l2 + 2) >> 2;
    SRC(0,1)=SRC(2,2)=SRC(4,3)=SRC(6,4)= (l0 + l1 + 1) >> 1;
    SRC(1,1)=SRC(3,2)=SRC(5,3)=SRC(7,4)= (lt + 2*l0 + l1 + 2) >> 2;
    SRC(0,0)=SRC(2,1)=SRC(4,2)=SRC(6,3)= (lt + l0 + 1) >> 1;
    SRC(1,0)=SRC(3,1)=SRC(5,2)=SRC(7,3)= (l0 + 2*lt + t0 + 2) >> 2;
    SRC(2,0)=SRC(4,1)=SRC(6,2)= (t1 + 2*t0 + lt + 2) >> 2;
    SRC(3,0)=SRC(5,1)=SRC(7,2)= (t2 + 2*t1 + t0 + 2) >> 2;
    SRC(4,0)=SRC(6,1)= (t3 + 2*t2 + t1 + 2) >> 2;
    SRC(5,0)=SRC(7,1)= (t4 + 2*t3 + t2 + 2) >> 2;
    SRC(6,0)= (t5 + 2*t4 + t3 + 2) >> 2;
    SRC(7,0)= (t6 + 2*t5 + t4 + 2) >> 2;
}
1071 1072
static void FUNCC(pred8x8l_vertical_left)(uint8_t *_src, int has_topleft,
                                          int has_topright, ptrdiff_t _stride)
1073
{
1074 1075
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
    PREDICT_8x8_LOAD_TOP;
    PREDICT_8x8_LOAD_TOPRIGHT;
    SRC(0,0)= (t0 + t1 + 1) >> 1;
    SRC(0,1)= (t0 + 2*t1 + t2 + 2) >> 2;
    SRC(0,2)=SRC(1,0)= (t1 + t2 + 1) >> 1;
    SRC(0,3)=SRC(1,1)= (t1 + 2*t2 + t3 + 2) >> 2;
    SRC(0,4)=SRC(1,2)=SRC(2,0)= (t2 + t3 + 1) >> 1;
    SRC(0,5)=SRC(1,3)=SRC(2,1)= (t2 + 2*t3 + t4 + 2) >> 2;
    SRC(0,6)=SRC(1,4)=SRC(2,2)=SRC(3,0)= (t3 + t4 + 1) >> 1;
    SRC(0,7)=SRC(1,5)=SRC(2,3)=SRC(3,1)= (t3 + 2*t4 + t5 + 2) >> 2;
    SRC(1,6)=SRC(2,4)=SRC(3,2)=SRC(4,0)= (t4 + t5 + 1) >> 1;
    SRC(1,7)=SRC(2,5)=SRC(3,3)=SRC(4,1)= (t4 + 2*t5 + t6 + 2) >> 2;
    SRC(2,6)=SRC(3,4)=SRC(4,2)=SRC(5,0)= (t5 + t6 + 1) >> 1;
    SRC(2,7)=SRC(3,5)=SRC(4,3)=SRC(5,1)= (t5 + 2*t6 + t7 + 2) >> 2;
    SRC(3,6)=SRC(4,4)=SRC(5,2)=SRC(6,0)= (t6 + t7 + 1) >> 1;
    SRC(3,7)=SRC(4,5)=SRC(5,3)=SRC(6,1)= (t6 + 2*t7 + t8 + 2) >> 2;
    SRC(4,6)=SRC(5,4)=SRC(6,2)=SRC(7,0)= (t7 + t8 + 1) >> 1;
    SRC(4,7)=SRC(5,5)=SRC(6,3)=SRC(7,1)= (t7 + 2*t8 + t9 + 2) >> 2;
    SRC(5,6)=SRC(6,4)=SRC(7,2)= (t8 + t9 + 1) >> 1;
    SRC(5,7)=SRC(6,5)=SRC(7,3)= (t8 + 2*t9 + t10 + 2) >> 2;
    SRC(6,6)=SRC(7,4)= (t9 + t10 + 1) >> 1;
    SRC(6,7)=SRC(7,5)= (t9 + 2*t10 + t11 + 2) >> 2;
    SRC(7,6)= (t10 + t11 + 1) >> 1;
    SRC(7,7)= (t10 + 2*t11 + t12 + 2) >> 2;
}
1101 1102
static void FUNCC(pred8x8l_horizontal_up)(uint8_t *_src, int has_topleft,
                                          int has_topright, ptrdiff_t _stride)
1103
{
1104 1105
    pixel *src = (pixel*)_src;
    int stride = _stride/sizeof(pixel);
1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
    PREDICT_8x8_LOAD_LEFT;
    SRC(0,0)= (l0 + l1 + 1) >> 1;
    SRC(1,0)= (l0 + 2*l1 + l2 + 2) >> 2;
    SRC(0,1)=SRC(2,0)= (l1 + l2 + 1) >> 1;
    SRC(1,1)=SRC(3,0)= (l1 + 2*l2 + l3 + 2) >> 2;
    SRC(0,2)=SRC(2,1)=SRC(4,0)= (l2 + l3 + 1) >> 1;
    SRC(1,2)=SRC(3,1)=SRC(5,0)= (l2 + 2*l3 + l4 + 2) >> 2;
    SRC(0,3)=SRC(2,2)=SRC(4,1)=SRC(6,0)= (l3 + l4 + 1) >> 1;
    SRC(1,3)=SRC(3,2)=SRC(5,1)=SRC(7,0)= (l3 + 2*l4 + l5 + 2) >> 2;
    SRC(0,4)=SRC(2,3)=SRC(4,2)=SRC(6,1)= (l4 + l5 + 1) >> 1;
    SRC(1,4)=SRC(3,3)=SRC(5,2)=SRC(7,1)= (l4 + 2*l5 + l6 + 2) >> 2;
    SRC(0,5)=SRC(2,4)=SRC(4,3)=SRC(6,2)= (l5 + l6 + 1) >> 1;
    SRC(1,5)=SRC(3,4)=SRC(5,3)=SRC(7,2)= (l5 + 2*l6 + l7 + 2) >> 2;
    SRC(0,6)=SRC(2,5)=SRC(4,4)=SRC(6,3)= (l6 + l7 + 1) >> 1;
    SRC(1,6)=SRC(3,5)=SRC(5,4)=SRC(7,3)= (l6 + 3*l7 + 2) >> 2;
    SRC(0,7)=SRC(1,7)=SRC(2,6)=SRC(2,7)=SRC(3,6)=
    SRC(3,7)=SRC(4,5)=SRC(4,6)=SRC(4,7)=SRC(5,5)=
    SRC(5,6)=SRC(5,7)=SRC(6,4)=SRC(6,5)=SRC(6,6)=
    SRC(6,7)=SRC(7,4)=SRC(7,5)=SRC(7,6)=SRC(7,7)= l7;
}
#undef PREDICT_8x8_LOAD_LEFT
#undef PREDICT_8x8_LOAD_TOP
#undef PREDICT_8x8_LOAD_TOPLEFT
#undef PREDICT_8x8_LOAD_TOPRIGHT
#undef PREDICT_8x8_DC
#undef PTR
#undef PT
#undef PL
#undef SRC

1136
static void FUNCC(pred4x4_vertical_add)(uint8_t *_pix, int16_t *_block,
1137 1138
                                        ptrdiff_t stride)
{
1139
    int i;
1140 1141 1142
    pixel *pix = (pixel*)_pix;
    const dctcoef *block = (const dctcoef*)_block;
    stride /= sizeof(pixel);
1143 1144
    pix -= stride;
    for(i=0; i<4; i++){
1145
        pixel v = pix[0];
1146 1147 1148 1149 1150 1151 1152
        pix[1*stride]= v += block[0];
        pix[2*stride]= v += block[4];
        pix[3*stride]= v += block[8];
        pix[4*stride]= v +  block[12];
        pix++;
        block++;
    }
1153 1154

    memset(_block, 0, sizeof(dctcoef) * 16);
1155 1156
}

1157
static void FUNCC(pred4x4_horizontal_add)(uint8_t *_pix, int16_t *_block,
1158 1159
                                          ptrdiff_t stride)
{
1160
    int i;
1161 1162 1163
    pixel *pix = (pixel*)_pix;
    const dctcoef *block = (const dctcoef*)_block;
    stride /= sizeof(pixel);
1164
    for(i=0; i<4; i++){
1165
        pixel v = pix[-1];
1166 1167 1168 1169 1170 1171 1172
        pix[0]= v += block[0];
        pix[1]= v += block[1];
        pix[2]= v += block[2];
        pix[3]= v +  block[3];
        pix+= stride;
        block+= 4;
    }
1173 1174

    memset(_block, 0, sizeof(dctcoef) * 16);
1175 1176
}

1177
static void FUNCC(pred8x8l_vertical_add)(uint8_t *_pix, int16_t *_block,
1178 1179
                                         ptrdiff_t stride)
{
1180
    int i;
1181 1182 1183
    pixel *pix = (pixel*)_pix;
    const dctcoef *block = (const dctcoef*)_block;
    stride /= sizeof(pixel);
1184 1185
    pix -= stride;
    for(i=0; i<8; i++){
1186
        pixel v = pix[0];
1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
        pix[1*stride]= v += block[0];
        pix[2*stride]= v += block[8];
        pix[3*stride]= v += block[16];
        pix[4*stride]= v += block[24];
        pix[5*stride]= v += block[32];
        pix[6*stride]= v += block[40];
        pix[7*stride]= v += block[48];
        pix[8*stride]= v +  block[56];
        pix++;
        block++;
    }
1198 1199

    memset(_block, 0, sizeof(dctcoef) * 64);
1200 1201
}

1202
static void FUNCC(pred8x8l_horizontal_add)(uint8_t *_pix, int16_t *_block,
1203 1204
                                           ptrdiff_t stride)
{
1205
    int i;
1206 1207 1208
    pixel *pix = (pixel*)_pix;
    const dctcoef *block = (const dctcoef*)_block;
    stride /= sizeof(pixel);
1209
    for(i=0; i<8; i++){
1210
        pixel v = pix[-1];
1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221
        pix[0]= v += block[0];
        pix[1]= v += block[1];
        pix[2]= v += block[2];
        pix[3]= v += block[3];
        pix[4]= v += block[4];
        pix[5]= v += block[5];
        pix[6]= v += block[6];
        pix[7]= v +  block[7];
        pix+= stride;
        block+= 8;
    }
1222 1223

    memset(_block, 0, sizeof(dctcoef) * 64);
1224 1225
}

1226
static void FUNCC(pred16x16_vertical_add)(uint8_t *pix, const int *block_offset,
1227
                                          int16_t *block,
1228 1229
                                          ptrdiff_t stride)
{
1230 1231
    int i;
    for(i=0; i<16; i++)
1232
        FUNCC(pred4x4_vertical_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1233 1234
}

1235 1236
static void FUNCC(pred16x16_horizontal_add)(uint8_t *pix,
                                            const int *block_offset,
1237
                                            int16_t *block,
1238 1239
                                            ptrdiff_t stride)
{
1240 1241
    int i;
    for(i=0; i<16; i++)
1242
        FUNCC(pred4x4_horizontal_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1243 1244
}

1245
static void FUNCC(pred8x8_vertical_add)(uint8_t *pix, const int *block_offset,
1246
                                        int16_t *block, ptrdiff_t stride)
1247
{
1248 1249
    int i;
    for(i=0; i<4; i++)
1250
        FUNCC(pred4x4_vertical_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1251 1252
}

1253
static void FUNCC(pred8x16_vertical_add)(uint8_t *pix, const int *block_offset,
1254
                                         int16_t *block, ptrdiff_t stride)
1255
{
1256 1257 1258 1259 1260 1261 1262
    int i;
    for(i=0; i<4; i++)
        FUNCC(pred4x4_vertical_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
    for(i=4; i<8; i++)
        FUNCC(pred4x4_vertical_add)(pix + block_offset[i+4], block + i*16*sizeof(pixel), stride);
}

1263
static void FUNCC(pred8x8_horizontal_add)(uint8_t *pix, const int *block_offset,
1264
                                          int16_t *block,
1265 1266
                                          ptrdiff_t stride)
{
1267 1268
    int i;
    for(i=0; i<4; i++)
1269
        FUNCC(pred4x4_horizontal_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1270
}
1271

1272 1273
static void FUNCC(pred8x16_horizontal_add)(uint8_t *pix,
                                           const int *block_offset,
1274
                                           int16_t *block, ptrdiff_t stride)
1275
{
1276 1277 1278 1279 1280 1281
    int i;
    for(i=0; i<4; i++)
        FUNCC(pred4x4_horizontal_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
    for(i=4; i<8; i++)
        FUNCC(pred4x4_horizontal_add)(pix + block_offset[i+4], block + i*16*sizeof(pixel), stride);
}