dsputil.c 173 KB
Newer Older
Fabrice Bellard's avatar
Fabrice Bellard committed
1 2
/*
 * DSP utils
3
 * Copyright (c) 2000, 2001 Fabrice Bellard
4
 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
Fabrice Bellard's avatar
Fabrice Bellard committed
5
 *
6 7
 * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer <michaelni@gmx.at>
 *
8 9 10
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
11 12
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
13
 * version 2.1 of the License, or (at your option) any later version.
Fabrice Bellard's avatar
Fabrice Bellard committed
14
 *
15
 * FFmpeg is distributed in the hope that it will be useful,
Fabrice Bellard's avatar
Fabrice Bellard committed
16
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 18
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
Fabrice Bellard's avatar
Fabrice Bellard committed
19
 *
20
 * You should have received a copy of the GNU Lesser General Public
21
 * License along with FFmpeg; if not, write to the Free Software
22
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Fabrice Bellard's avatar
Fabrice Bellard committed
23
 */
24

Michael Niedermayer's avatar
Michael Niedermayer committed
25
/**
26
 * @file libavcodec/dsputil.c
Michael Niedermayer's avatar
Michael Niedermayer committed
27 28
 * DSP utils
 */
29

Fabrice Bellard's avatar
Fabrice Bellard committed
30 31
#include "avcodec.h"
#include "dsputil.h"
32
#include "simple_idct.h"
33
#include "faandct.h"
34
#include "faanidct.h"
35
#include "mathops.h"
36
#include "h263.h"
37
#include "snow.h"
38

39 40 41
/* snow.c */
void ff_spatial_dwt(int *buffer, int width, int height, int stride, int type, int decomposition_count);

42 43 44
/* vorbis.c */
void vorbis_inverse_coupling(float *mag, float *ang, int blocksize);

Loren Merritt's avatar
Loren Merritt committed
45 46 47
/* ac3dec.c */
void ff_ac3_downmix_c(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);

48 49
/* lpc.c */
void ff_lpc_compute_autocorr(const int32_t *data, int len, int lag, double *autoc);
50

51 52 53
/* pngdec.c */
void ff_add_png_paeth_prediction(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp);

54 55 56
/* eaidct.c */
void ff_ea_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block);

57
uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP] = {0, };
58
uint32_t ff_squareTbl[512] = {0, };
Fabrice Bellard's avatar
Fabrice Bellard committed
59

60 61 62
// 0x7f7f7f7f or 0x7f7f7f7f7f7f7f7f or whatever, depending on the cpu's native arithmetic size
#define pb_7f (~0UL/255 * 0x7f)
#define pb_80 (~0UL/255 * 0x80)
63

64
const uint8_t ff_zigzag_direct[64] = {
65 66
    0,   1,  8, 16,  9,  2,  3, 10,
    17, 24, 32, 25, 18, 11,  4,  5,
67
    12, 19, 26, 33, 40, 48, 41, 34,
68
    27, 20, 13,  6,  7, 14, 21, 28,
69 70 71 72 73 74
    35, 42, 49, 56, 57, 50, 43, 36,
    29, 22, 15, 23, 30, 37, 44, 51,
    58, 59, 52, 45, 38, 31, 39, 46,
    53, 60, 61, 54, 47, 55, 62, 63
};

75 76 77 78 79 80 81 82 83 84 85 86 87
/* Specific zigzag scan for 248 idct. NOTE that unlike the
   specification, we interleave the fields */
const uint8_t ff_zigzag248_direct[64] = {
     0,  8,  1,  9, 16, 24,  2, 10,
    17, 25, 32, 40, 48, 56, 33, 41,
    18, 26,  3, 11,  4, 12, 19, 27,
    34, 42, 49, 57, 50, 58, 35, 43,
    20, 28,  5, 13,  6, 14, 21, 29,
    36, 44, 51, 59, 52, 60, 37, 45,
    22, 30,  7, 15, 23, 31, 38, 46,
    53, 61, 54, 62, 39, 47, 55, 63,
};

88
/* not permutated inverse zigzag_direct + 1 for MMX quantizer */
89
DECLARE_ALIGNED_16(uint16_t, inv_zigzag_direct16[64]);
90

91
const uint8_t ff_alternate_horizontal_scan[64] = {
92
    0,  1,   2,  3,  8,  9, 16, 17,
93
    10, 11,  4,  5,  6,  7, 15, 14,
94
    13, 12, 19, 18, 24, 25, 32, 33,
95
    26, 27, 20, 21, 22, 23, 28, 29,
96
    30, 31, 34, 35, 40, 41, 48, 49,
97
    42, 43, 36, 37, 38, 39, 44, 45,
98
    46, 47, 50, 51, 56, 57, 58, 59,
99 100 101
    52, 53, 54, 55, 60, 61, 62, 63,
};

102
const uint8_t ff_alternate_vertical_scan[64] = {
103
    0,  8,  16, 24,  1,  9,  2, 10,
104
    17, 25, 32, 40, 48, 56, 57, 49,
105
    41, 33, 26, 18,  3, 11,  4, 12,
106
    19, 27, 34, 42, 50, 58, 35, 43,
107
    51, 59, 20, 28,  5, 13,  6, 14,
108
    21, 29, 36, 44, 52, 60, 37, 45,
109
    53, 61, 22, 30,  7, 15, 23, 31,
110 111 112
    38, 46, 54, 62, 39, 47, 55, 63,
};

113 114 115
/* a*inverse[b]>>32 == a/b for all 0<=a<=16909558 && 2<=b<=256
 * for a>16909558, is an overestimate by less than 1 part in 1<<24 */
const uint32_t ff_inverse[257]={
116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146
         0, 4294967295U,2147483648U,1431655766, 1073741824,  858993460,  715827883,  613566757,
 536870912,  477218589,  429496730,  390451573,  357913942,  330382100,  306783379,  286331154,
 268435456,  252645136,  238609295,  226050911,  214748365,  204522253,  195225787,  186737709,
 178956971,  171798692,  165191050,  159072863,  153391690,  148102321,  143165577,  138547333,
 134217728,  130150525,  126322568,  122713352,  119304648,  116080198,  113025456,  110127367,
 107374183,  104755300,  102261127,   99882961,   97612894,   95443718,   93368855,   91382283,
  89478486,   87652394,   85899346,   84215046,   82595525,   81037119,   79536432,   78090315,
  76695845,   75350304,   74051161,   72796056,   71582789,   70409300,   69273667,   68174085,
  67108864,   66076420,   65075263,   64103990,   63161284,   62245903,   61356676,   60492498,
  59652324,   58835169,   58040099,   57266231,   56512728,   55778797,   55063684,   54366675,
  53687092,   53024288,   52377650,   51746594,   51130564,   50529028,   49941481,   49367441,
  48806447,   48258060,   47721859,   47197443,   46684428,   46182445,   45691142,   45210183,
  44739243,   44278014,   43826197,   43383509,   42949673,   42524429,   42107523,   41698712,
  41297763,   40904451,   40518560,   40139882,   39768216,   39403370,   39045158,   38693400,
  38347923,   38008561,   37675152,   37347542,   37025581,   36709123,   36398028,   36092163,
  35791395,   35495598,   35204650,   34918434,   34636834,   34359739,   34087043,   33818641,
  33554432,   33294321,   33038210,   32786010,   32537632,   32292988,   32051995,   31814573,
  31580642,   31350127,   31122952,   30899046,   30678338,   30460761,   30246249,   30034737,
  29826162,   29620465,   29417585,   29217465,   29020050,   28825284,   28633116,   28443493,
  28256364,   28071682,   27889399,   27709467,   27531842,   27356480,   27183338,   27012373,
  26843546,   26676816,   26512144,   26349493,   26188825,   26030105,   25873297,   25718368,
  25565282,   25414008,   25264514,   25116768,   24970741,   24826401,   24683721,   24542671,
  24403224,   24265352,   24129030,   23994231,   23860930,   23729102,   23598722,   23469767,
  23342214,   23216040,   23091223,   22967740,   22845571,   22724695,   22605092,   22486740,
  22369622,   22253717,   22139007,   22025474,   21913099,   21801865,   21691755,   21582751,
  21474837,   21367997,   21262215,   21157475,   21053762,   20951060,   20849356,   20748635,
  20648882,   20550083,   20452226,   20355296,   20259280,   20164166,   20069941,   19976593,
  19884108,   19792477,   19701685,   19611723,   19522579,   19434242,   19346700,   19259944,
  19173962,   19088744,   19004281,   18920561,   18837576,   18755316,   18673771,   18592933,
  18512791,   18433337,   18354562,   18276457,   18199014,   18122225,   18046082,   17970575,
  17895698,   17821442,   17747799,   17674763,   17602325,   17530479,   17459217,   17388532,
147
  17318417,   17248865,   17179870,   17111424,   17043522,   16976156,   16909321,   16843010,
148
  16777216
149 150
};

151 152
/* Input permutation for the simple_idct_mmx */
static const uint8_t simple_mmx_permutation[64]={
153 154 155 156 157 158 159 160
        0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
        0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
        0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
        0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
        0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
        0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
        0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
        0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
161 162
};

163 164
static const uint8_t idct_sse2_row_perm[8] = {0, 4, 1, 5, 2, 6, 3, 7};

165 166 167 168 169 170 171 172 173 174
void ff_init_scantable(uint8_t *permutation, ScanTable *st, const uint8_t *src_scantable){
    int i;
    int end;

    st->scantable= src_scantable;

    for(i=0; i<64; i++){
        int j;
        j = src_scantable[i];
        st->permutated[i] = permutation[j];
175
#if ARCH_PPC
176 177 178 179 180 181 182 183 184 185 186 187 188
        st->inverse[j] = i;
#endif
    }

    end=-1;
    for(i=0; i<64; i++){
        int j;
        j = st->permutated[i];
        if(j>end) end=j;
        st->raster_end[i]= end;
    }
}

189
static int pix_sum_c(uint8_t * pix, int line_size)
190 191 192 193 194
{
    int s, i, j;

    s = 0;
    for (i = 0; i < 16; i++) {
195 196 197 198 199 200 201 202 203 204 205 206
        for (j = 0; j < 16; j += 8) {
            s += pix[0];
            s += pix[1];
            s += pix[2];
            s += pix[3];
            s += pix[4];
            s += pix[5];
            s += pix[6];
            s += pix[7];
            pix += 8;
        }
        pix += line_size - 16;
207 208 209 210
    }
    return s;
}

211
static int pix_norm1_c(uint8_t * pix, int line_size)
212 213
{
    int s, i, j;
214
    uint32_t *sq = ff_squareTbl + 256;
215 216 217

    s = 0;
    for (i = 0; i < 16; i++) {
218
        for (j = 0; j < 16; j += 8) {
219
#if 0
220 221 222 223 224 225 226 227
            s += sq[pix[0]];
            s += sq[pix[1]];
            s += sq[pix[2]];
            s += sq[pix[3]];
            s += sq[pix[4]];
            s += sq[pix[5]];
            s += sq[pix[6]];
            s += sq[pix[7]];
228 229
#else
#if LONG_MAX > 2147483647
230 231 232 233 234
            register uint64_t x=*(uint64_t*)pix;
            s += sq[x&0xff];
            s += sq[(x>>8)&0xff];
            s += sq[(x>>16)&0xff];
            s += sq[(x>>24)&0xff];
235 236 237 238 239
            s += sq[(x>>32)&0xff];
            s += sq[(x>>40)&0xff];
            s += sq[(x>>48)&0xff];
            s += sq[(x>>56)&0xff];
#else
240 241 242 243 244
            register uint32_t x=*(uint32_t*)pix;
            s += sq[x&0xff];
            s += sq[(x>>8)&0xff];
            s += sq[(x>>16)&0xff];
            s += sq[(x>>24)&0xff];
245 246 247 248 249 250 251
            x=*(uint32_t*)(pix+4);
            s += sq[x&0xff];
            s += sq[(x>>8)&0xff];
            s += sq[(x>>16)&0xff];
            s += sq[(x>>24)&0xff];
#endif
#endif
252 253 254
            pix += 8;
        }
        pix += line_size - 16;
255 256 257 258
    }
    return s;
}

259
static void bswap_buf(uint32_t *dst, const uint32_t *src, int w){
Michael Niedermayer's avatar
Michael Niedermayer committed
260
    int i;
261

Michael Niedermayer's avatar
Michael Niedermayer committed
262 263 264 265 266 267 268 269 270 271 272 273 274 275
    for(i=0; i+8<=w; i+=8){
        dst[i+0]= bswap_32(src[i+0]);
        dst[i+1]= bswap_32(src[i+1]);
        dst[i+2]= bswap_32(src[i+2]);
        dst[i+3]= bswap_32(src[i+3]);
        dst[i+4]= bswap_32(src[i+4]);
        dst[i+5]= bswap_32(src[i+5]);
        dst[i+6]= bswap_32(src[i+6]);
        dst[i+7]= bswap_32(src[i+7]);
    }
    for(;i<w; i++){
        dst[i+0]= bswap_32(src[i+0]);
    }
}
276

277 278 279
static int sse4_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h)
{
    int s, i;
280
    uint32_t *sq = ff_squareTbl + 256;
281 282 283 284 285 286 287 288 289 290 291 292 293

    s = 0;
    for (i = 0; i < h; i++) {
        s += sq[pix1[0] - pix2[0]];
        s += sq[pix1[1] - pix2[1]];
        s += sq[pix1[2] - pix2[2]];
        s += sq[pix1[3] - pix2[3]];
        pix1 += line_size;
        pix2 += line_size;
    }
    return s;
}

294
static int sse8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h)
Michael Niedermayer's avatar
Michael Niedermayer committed
295 296
{
    int s, i;
297
    uint32_t *sq = ff_squareTbl + 256;
Michael Niedermayer's avatar
Michael Niedermayer committed
298 299

    s = 0;
300
    for (i = 0; i < h; i++) {
Michael Niedermayer's avatar
Michael Niedermayer committed
301 302 303 304 305 306 307 308 309 310 311 312 313 314
        s += sq[pix1[0] - pix2[0]];
        s += sq[pix1[1] - pix2[1]];
        s += sq[pix1[2] - pix2[2]];
        s += sq[pix1[3] - pix2[3]];
        s += sq[pix1[4] - pix2[4]];
        s += sq[pix1[5] - pix2[5]];
        s += sq[pix1[6] - pix2[6]];
        s += sq[pix1[7] - pix2[7]];
        pix1 += line_size;
        pix2 += line_size;
    }
    return s;
}

315
static int sse16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
316
{
317
    int s, i;
318
    uint32_t *sq = ff_squareTbl + 256;
319 320

    s = 0;
321
    for (i = 0; i < h; i++) {
322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337
        s += sq[pix1[ 0] - pix2[ 0]];
        s += sq[pix1[ 1] - pix2[ 1]];
        s += sq[pix1[ 2] - pix2[ 2]];
        s += sq[pix1[ 3] - pix2[ 3]];
        s += sq[pix1[ 4] - pix2[ 4]];
        s += sq[pix1[ 5] - pix2[ 5]];
        s += sq[pix1[ 6] - pix2[ 6]];
        s += sq[pix1[ 7] - pix2[ 7]];
        s += sq[pix1[ 8] - pix2[ 8]];
        s += sq[pix1[ 9] - pix2[ 9]];
        s += sq[pix1[10] - pix2[10]];
        s += sq[pix1[11] - pix2[11]];
        s += sq[pix1[12] - pix2[12]];
        s += sq[pix1[13] - pix2[13]];
        s += sq[pix1[14] - pix2[14]];
        s += sq[pix1[15] - pix2[15]];
338

339 340
        pix1 += line_size;
        pix2 += line_size;
341 342 343 344
    }
    return s;
}

345

346
#if CONFIG_SNOW_ENCODER //dwt is in snow.c
347
static inline int w_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int w, int h, int type){
348 349
    int s, i, j;
    const int dec_count= w==8 ? 3 : 4;
350
    int tmp[32*32];
351
    int level, ori;
352
    static const int scale[2][2][4][4]={
353 354
      {
        {
355
            // 9/7 8x8 dec=3
356 357 358 359
            {268, 239, 239, 213},
            {  0, 224, 224, 152},
            {  0, 135, 135, 110},
        },{
360
            // 9/7 16x16 or 32x32 dec=4
361 362 363 364 365 366
            {344, 310, 310, 280},
            {  0, 320, 320, 228},
            {  0, 175, 175, 136},
            {  0, 129, 129, 102},
        }
      },{
367 368
        {
            // 5/3 8x8 dec=3
369 370 371 372
            {275, 245, 245, 218},
            {  0, 230, 230, 156},
            {  0, 138, 138, 113},
        },{
373
            // 5/3 16x16 or 32x32 dec=4
374 375 376 377 378 379 380 381 382 383
            {352, 317, 317, 286},
            {  0, 328, 328, 233},
            {  0, 180, 180, 140},
            {  0, 132, 132, 105},
        }
      }
    };

    for (i = 0; i < h; i++) {
        for (j = 0; j < w; j+=4) {
384 385 386 387
            tmp[32*i+j+0] = (pix1[j+0] - pix2[j+0])<<4;
            tmp[32*i+j+1] = (pix1[j+1] - pix2[j+1])<<4;
            tmp[32*i+j+2] = (pix1[j+2] - pix2[j+2])<<4;
            tmp[32*i+j+3] = (pix1[j+3] - pix2[j+3])<<4;
388 389 390 391
        }
        pix1 += line_size;
        pix2 += line_size;
    }
392

393
    ff_spatial_dwt(tmp, w, h, 32, type, dec_count);
394 395

    s=0;
396
    assert(w==h);
397 398
    for(level=0; level<dec_count; level++){
        for(ori= level ? 1 : 0; ori<4; ori++){
399 400 401
            int size= w>>(dec_count-level);
            int sx= (ori&1) ? size : 0;
            int stride= 32<<(dec_count-level);
402
            int sy= (ori&2) ? stride>>1 : 0;
403

404 405 406
            for(i=0; i<size; i++){
                for(j=0; j<size; j++){
                    int v= tmp[sx + sy + i*stride + j] * scale[type][dec_count-3][level][ori];
407
                    s += FFABS(v);
408 409 410 411
                }
            }
        }
    }
412
    assert(s>=0);
413
    return s>>9;
414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431
}

static int w53_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size,  8, h, 1);
}

static int w97_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size,  8, h, 0);
}

static int w53_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size, 16, h, 1);
}

static int w97_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size, 16, h, 0);
}

Måns Rullgård's avatar
Måns Rullgård committed
432
int w53_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
433 434 435
    return w_c(v, pix1, pix2, line_size, 32, h, 1);
}

Måns Rullgård's avatar
Måns Rullgård committed
436
int w97_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
437 438
    return w_c(v, pix1, pix2, line_size, 32, h, 0);
}
439
#endif
440

441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469
/* draw the edges of width 'w' of an image of size width, height */
//FIXME check that this is ok for mpeg4 interlaced
static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w)
{
    uint8_t *ptr, *last_line;
    int i;

    last_line = buf + (height - 1) * wrap;
    for(i=0;i<w;i++) {
        /* top and bottom */
        memcpy(buf - (i + 1) * wrap, buf, width);
        memcpy(last_line + (i + 1) * wrap, last_line, width);
    }
    /* left and right */
    ptr = buf;
    for(i=0;i<height;i++) {
        memset(ptr - w, ptr[0], w);
        memset(ptr + width, ptr[width-1], w);
        ptr += wrap;
    }
    /* corners */
    for(i=0;i<w;i++) {
        memset(buf - (i + 1) * wrap - w, buf[0], w); /* top left */
        memset(buf - (i + 1) * wrap + width, buf[width-1], w); /* top right */
        memset(last_line + (i + 1) * wrap - w, last_line[0], w); /* top left */
        memset(last_line + (i + 1) * wrap + width, last_line[width-1], w); /* top right */
    }
}

470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540
/**
 * Copies a rectangular area of samples to a temporary buffer and replicates the boarder samples.
 * @param buf destination buffer
 * @param src source buffer
 * @param linesize number of bytes between 2 vertically adjacent samples in both the source and destination buffers
 * @param block_w width of block
 * @param block_h height of block
 * @param src_x x coordinate of the top left sample of the block in the source buffer
 * @param src_y y coordinate of the top left sample of the block in the source buffer
 * @param w width of the source buffer
 * @param h height of the source buffer
 */
void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize, int block_w, int block_h,
                                    int src_x, int src_y, int w, int h){
    int x, y;
    int start_y, start_x, end_y, end_x;

    if(src_y>= h){
        src+= (h-1-src_y)*linesize;
        src_y=h-1;
    }else if(src_y<=-block_h){
        src+= (1-block_h-src_y)*linesize;
        src_y=1-block_h;
    }
    if(src_x>= w){
        src+= (w-1-src_x);
        src_x=w-1;
    }else if(src_x<=-block_w){
        src+= (1-block_w-src_x);
        src_x=1-block_w;
    }

    start_y= FFMAX(0, -src_y);
    start_x= FFMAX(0, -src_x);
    end_y= FFMIN(block_h, h-src_y);
    end_x= FFMIN(block_w, w-src_x);

    // copy existing part
    for(y=start_y; y<end_y; y++){
        for(x=start_x; x<end_x; x++){
            buf[x + y*linesize]= src[x + y*linesize];
        }
    }

    //top
    for(y=0; y<start_y; y++){
        for(x=start_x; x<end_x; x++){
            buf[x + y*linesize]= buf[x + start_y*linesize];
        }
    }

    //bottom
    for(y=end_y; y<block_h; y++){
        for(x=start_x; x<end_x; x++){
            buf[x + y*linesize]= buf[x + (end_y-1)*linesize];
        }
    }

    for(y=0; y<block_h; y++){
       //left
        for(x=0; x<start_x; x++){
            buf[x + y*linesize]= buf[start_x + y*linesize];
        }

       //right
        for(x=end_x; x<block_w; x++){
            buf[x + y*linesize]= buf[end_x - 1 + y*linesize];
        }
    }
}

541
static void get_pixels_c(DCTELEM *restrict block, const uint8_t *pixels, int line_size)
Fabrice Bellard's avatar
Fabrice Bellard committed
542 543 544 545 546
{
    int i;

    /* read the pixels */
    for(i=0;i<8;i++) {
547 548 549 550 551 552 553 554 555 556
        block[0] = pixels[0];
        block[1] = pixels[1];
        block[2] = pixels[2];
        block[3] = pixels[3];
        block[4] = pixels[4];
        block[5] = pixels[5];
        block[6] = pixels[6];
        block[7] = pixels[7];
        pixels += line_size;
        block += 8;
Fabrice Bellard's avatar
Fabrice Bellard committed
557 558 559
    }
}

560
static void diff_pixels_c(DCTELEM *restrict block, const uint8_t *s1,
561
                          const uint8_t *s2, int stride){
562 563 564 565
    int i;

    /* read the pixels */
    for(i=0;i<8;i++) {
566 567 568 569 570 571 572 573
        block[0] = s1[0] - s2[0];
        block[1] = s1[1] - s2[1];
        block[2] = s1[2] - s2[2];
        block[3] = s1[3] - s2[3];
        block[4] = s1[4] - s2[4];
        block[5] = s1[5] - s2[5];
        block[6] = s1[6] - s2[6];
        block[7] = s1[7] - s2[7];
574 575
        s1 += stride;
        s2 += stride;
576
        block += 8;
577 578 579 580
    }
}


581
static void put_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict pixels,
582
                                 int line_size)
Fabrice Bellard's avatar
Fabrice Bellard committed
583 584
{
    int i;
585
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
586

Fabrice Bellard's avatar
Fabrice Bellard committed
587 588
    /* read the pixels */
    for(i=0;i<8;i++) {
589 590 591 592 593 594 595 596 597 598 599
        pixels[0] = cm[block[0]];
        pixels[1] = cm[block[1]];
        pixels[2] = cm[block[2]];
        pixels[3] = cm[block[3]];
        pixels[4] = cm[block[4]];
        pixels[5] = cm[block[5]];
        pixels[6] = cm[block[6]];
        pixels[7] = cm[block[7]];

        pixels += line_size;
        block += 8;
Fabrice Bellard's avatar
Fabrice Bellard committed
600 601 602
    }
}

603
static void put_pixels_clamped4_c(const DCTELEM *block, uint8_t *restrict pixels,
604
                                 int line_size)
605 606
{
    int i;
607
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
608

609 610 611 612 613 614 615 616 617 618 619 620
    /* read the pixels */
    for(i=0;i<4;i++) {
        pixels[0] = cm[block[0]];
        pixels[1] = cm[block[1]];
        pixels[2] = cm[block[2]];
        pixels[3] = cm[block[3]];

        pixels += line_size;
        block += 8;
    }
}

621
static void put_pixels_clamped2_c(const DCTELEM *block, uint8_t *restrict pixels,
622
                                 int line_size)
623 624
{
    int i;
625
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
626

627 628 629 630 631 632 633 634 635 636
    /* read the pixels */
    for(i=0;i<2;i++) {
        pixels[0] = cm[block[0]];
        pixels[1] = cm[block[1]];

        pixels += line_size;
        block += 8;
    }
}

637
static void put_signed_pixels_clamped_c(const DCTELEM *block,
638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657
                                        uint8_t *restrict pixels,
                                        int line_size)
{
    int i, j;

    for (i = 0; i < 8; i++) {
        for (j = 0; j < 8; j++) {
            if (*block < -128)
                *pixels = 0;
            else if (*block > 127)
                *pixels = 255;
            else
                *pixels = (uint8_t)(*block + 128);
            block++;
            pixels++;
        }
        pixels += (line_size - 8);
    }
}

658
static void add_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict pixels,
659
                          int line_size)
Fabrice Bellard's avatar
Fabrice Bellard committed
660 661
{
    int i;
662
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
663

Fabrice Bellard's avatar
Fabrice Bellard committed
664 665
    /* read the pixels */
    for(i=0;i<8;i++) {
666 667 668 669 670 671 672 673 674 675
        pixels[0] = cm[pixels[0] + block[0]];
        pixels[1] = cm[pixels[1] + block[1]];
        pixels[2] = cm[pixels[2] + block[2]];
        pixels[3] = cm[pixels[3] + block[3]];
        pixels[4] = cm[pixels[4] + block[4]];
        pixels[5] = cm[pixels[5] + block[5]];
        pixels[6] = cm[pixels[6] + block[6]];
        pixels[7] = cm[pixels[7] + block[7]];
        pixels += line_size;
        block += 8;
Fabrice Bellard's avatar
Fabrice Bellard committed
676 677
    }
}
678 679 680 681 682

static void add_pixels_clamped4_c(const DCTELEM *block, uint8_t *restrict pixels,
                          int line_size)
{
    int i;
683
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
684

685 686 687 688 689 690 691 692 693 694
    /* read the pixels */
    for(i=0;i<4;i++) {
        pixels[0] = cm[pixels[0] + block[0]];
        pixels[1] = cm[pixels[1] + block[1]];
        pixels[2] = cm[pixels[2] + block[2]];
        pixels[3] = cm[pixels[3] + block[3]];
        pixels += line_size;
        block += 8;
    }
}
695 696 697 698 699

static void add_pixels_clamped2_c(const DCTELEM *block, uint8_t *restrict pixels,
                          int line_size)
{
    int i;
700
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
701

702 703 704 705 706 707 708 709
    /* read the pixels */
    for(i=0;i<2;i++) {
        pixels[0] = cm[pixels[0] + block[0]];
        pixels[1] = cm[pixels[1] + block[1]];
        pixels += line_size;
        block += 8;
    }
}
Loren Merritt's avatar
Loren Merritt committed
710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740

static void add_pixels8_c(uint8_t *restrict pixels, DCTELEM *block, int line_size)
{
    int i;
    for(i=0;i<8;i++) {
        pixels[0] += block[0];
        pixels[1] += block[1];
        pixels[2] += block[2];
        pixels[3] += block[3];
        pixels[4] += block[4];
        pixels[5] += block[5];
        pixels[6] += block[6];
        pixels[7] += block[7];
        pixels += line_size;
        block += 8;
    }
}

static void add_pixels4_c(uint8_t *restrict pixels, DCTELEM *block, int line_size)
{
    int i;
    for(i=0;i<4;i++) {
        pixels[0] += block[0];
        pixels[1] += block[1];
        pixels[2] += block[2];
        pixels[3] += block[3];
        pixels += line_size;
        block += 4;
    }
}

741 742 743 744 745 746 747 748
static int sum_abs_dctelem_c(DCTELEM *block)
{
    int sum=0, i;
    for(i=0; i<64; i++)
        sum+= FFABS(block[i]);
    return sum;
}

749 750 751
#if 0

#define PIXOP2(OPNAME, OP) \
Michael Niedermayer's avatar
Michael Niedermayer committed
752
static void OPNAME ## _pixels(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
753 754 755
{\
    int i;\
    for(i=0; i<h; i++){\
756
        OP(*((uint64_t*)block), AV_RN64(pixels));\
757 758 759 760 761
        pixels+=line_size;\
        block +=line_size;\
    }\
}\
\
762
static void OPNAME ## _no_rnd_pixels_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
763 764 765
{\
    int i;\
    for(i=0; i<h; i++){\
766 767
        const uint64_t a= AV_RN64(pixels  );\
        const uint64_t b= AV_RN64(pixels+1);\
768 769 770 771 772 773
        OP(*((uint64_t*)block), (a&b) + (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\
        pixels+=line_size;\
        block +=line_size;\
    }\
}\
\
774
static void OPNAME ## _pixels_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
775 776 777
{\
    int i;\
    for(i=0; i<h; i++){\
778 779
        const uint64_t a= AV_RN64(pixels  );\
        const uint64_t b= AV_RN64(pixels+1);\
780 781 782 783 784 785
        OP(*((uint64_t*)block), (a|b) - (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\
        pixels+=line_size;\
        block +=line_size;\
    }\
}\
\
786
static void OPNAME ## _no_rnd_pixels_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
787 788 789
{\
    int i;\
    for(i=0; i<h; i++){\
790 791
        const uint64_t a= AV_RN64(pixels          );\
        const uint64_t b= AV_RN64(pixels+line_size);\
792 793 794 795 796 797
        OP(*((uint64_t*)block), (a&b) + (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\
        pixels+=line_size;\
        block +=line_size;\
    }\
}\
\
798
static void OPNAME ## _pixels_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
799 800 801
{\
    int i;\
    for(i=0; i<h; i++){\
802 803
        const uint64_t a= AV_RN64(pixels          );\
        const uint64_t b= AV_RN64(pixels+line_size);\
804 805 806 807 808 809
        OP(*((uint64_t*)block), (a|b) - (((a^b)&0xFEFEFEFEFEFEFEFEULL)>>1));\
        pixels+=line_size;\
        block +=line_size;\
    }\
}\
\
810
static void OPNAME ## _pixels_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
811 812
{\
        int i;\
813 814
        const uint64_t a= AV_RN64(pixels  );\
        const uint64_t b= AV_RN64(pixels+1);\
815 816 817 818 819 820 821 822 823
        uint64_t l0=  (a&0x0303030303030303ULL)\
                    + (b&0x0303030303030303ULL)\
                    + 0x0202020202020202ULL;\
        uint64_t h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
                   + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
        uint64_t l1,h1;\
\
        pixels+=line_size;\
        for(i=0; i<h; i+=2){\
824 825
            uint64_t a= AV_RN64(pixels  );\
            uint64_t b= AV_RN64(pixels+1);\
826 827 828 829 830 831 832
            l1=  (a&0x0303030303030303ULL)\
               + (b&0x0303030303030303ULL);\
            h1= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
              + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
            OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\
            pixels+=line_size;\
            block +=line_size;\
833 834
            a= AV_RN64(pixels  );\
            b= AV_RN64(pixels+1);\
835 836 837 838 839 840 841 842 843 844 845
            l0=  (a&0x0303030303030303ULL)\
               + (b&0x0303030303030303ULL)\
               + 0x0202020202020202ULL;\
            h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
              + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
            OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\
            pixels+=line_size;\
            block +=line_size;\
        }\
}\
\
846
static void OPNAME ## _no_rnd_pixels_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
847 848
{\
        int i;\
849 850
        const uint64_t a= AV_RN64(pixels  );\
        const uint64_t b= AV_RN64(pixels+1);\
851 852 853 854 855 856 857 858 859
        uint64_t l0=  (a&0x0303030303030303ULL)\
                    + (b&0x0303030303030303ULL)\
                    + 0x0101010101010101ULL;\
        uint64_t h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
                   + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
        uint64_t l1,h1;\
\
        pixels+=line_size;\
        for(i=0; i<h; i+=2){\
860 861
            uint64_t a= AV_RN64(pixels  );\
            uint64_t b= AV_RN64(pixels+1);\
862 863 864 865 866 867 868
            l1=  (a&0x0303030303030303ULL)\
               + (b&0x0303030303030303ULL);\
            h1= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
              + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
            OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\
            pixels+=line_size;\
            block +=line_size;\
869 870
            a= AV_RN64(pixels  );\
            b= AV_RN64(pixels+1);\
871 872 873 874 875 876 877 878 879 880 881
            l0=  (a&0x0303030303030303ULL)\
               + (b&0x0303030303030303ULL)\
               + 0x0101010101010101ULL;\
            h0= ((a&0xFCFCFCFCFCFCFCFCULL)>>2)\
              + ((b&0xFCFCFCFCFCFCFCFCULL)>>2);\
            OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\
            pixels+=line_size;\
            block +=line_size;\
        }\
}\
\
882 883 884 885 886 887 888
CALL_2X_PIXELS(OPNAME ## _pixels16_c    , OPNAME ## _pixels_c    , 8)\
CALL_2X_PIXELS(OPNAME ## _pixels16_x2_c , OPNAME ## _pixels_x2_c , 8)\
CALL_2X_PIXELS(OPNAME ## _pixels16_y2_c , OPNAME ## _pixels_y2_c , 8)\
CALL_2X_PIXELS(OPNAME ## _pixels16_xy2_c, OPNAME ## _pixels_xy2_c, 8)\
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_x2_c , OPNAME ## _no_rnd_pixels_x2_c , 8)\
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_y2_c , OPNAME ## _no_rnd_pixels_y2_c , 8)\
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_xy2_c, OPNAME ## _no_rnd_pixels_xy2_c, 8)
889 890 891 892 893

#define op_avg(a, b) a = ( ((a)|(b)) - ((((a)^(b))&0xFEFEFEFEFEFEFEFEULL)>>1) )
#else // 64 bit variant

#define PIXOP2(OPNAME, OP) \
894 895 896
static void OPNAME ## _pixels2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
    int i;\
    for(i=0; i<h; i++){\
897
        OP(*((uint16_t*)(block  )), AV_RN16(pixels  ));\
898 899 900 901
        pixels+=line_size;\
        block +=line_size;\
    }\
}\
902 903 904
static void OPNAME ## _pixels4_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
    int i;\
    for(i=0; i<h; i++){\
905
        OP(*((uint32_t*)(block  )), AV_RN32(pixels  ));\
906 907 908 909
        pixels+=line_size;\
        block +=line_size;\
    }\
}\
910
static void OPNAME ## _pixels8_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
911 912
    int i;\
    for(i=0; i<h; i++){\
913 914
        OP(*((uint32_t*)(block  )), AV_RN32(pixels  ));\
        OP(*((uint32_t*)(block+4)), AV_RN32(pixels+4));\
915 916 917 918
        pixels+=line_size;\
        block +=line_size;\
    }\
}\
919 920
static inline void OPNAME ## _no_rnd_pixels8_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
    OPNAME ## _pixels8_c(block, pixels, line_size, h);\
Michael Niedermayer's avatar
Michael Niedermayer committed
921
}\
922
\
Michael Niedermayer's avatar
Michael Niedermayer committed
923 924
static inline void OPNAME ## _no_rnd_pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
                                                int src_stride1, int src_stride2, int h){\
925 926
    int i;\
    for(i=0; i<h; i++){\
Michael Niedermayer's avatar
Michael Niedermayer committed
927
        uint32_t a,b;\
928 929
        a= AV_RN32(&src1[i*src_stride1  ]);\
        b= AV_RN32(&src2[i*src_stride2  ]);\
Michael Niedermayer's avatar
Michael Niedermayer committed
930
        OP(*((uint32_t*)&dst[i*dst_stride  ]), no_rnd_avg32(a, b));\
931 932
        a= AV_RN32(&src1[i*src_stride1+4]);\
        b= AV_RN32(&src2[i*src_stride2+4]);\
Michael Niedermayer's avatar
Michael Niedermayer committed
933
        OP(*((uint32_t*)&dst[i*dst_stride+4]), no_rnd_avg32(a, b));\
934 935 936
    }\
}\
\
Michael Niedermayer's avatar
Michael Niedermayer committed
937 938
static inline void OPNAME ## _pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
                                                int src_stride1, int src_stride2, int h){\
939 940
    int i;\
    for(i=0; i<h; i++){\
Michael Niedermayer's avatar
Michael Niedermayer committed
941
        uint32_t a,b;\
942 943
        a= AV_RN32(&src1[i*src_stride1  ]);\
        b= AV_RN32(&src2[i*src_stride2  ]);\
Michael Niedermayer's avatar
Michael Niedermayer committed
944
        OP(*((uint32_t*)&dst[i*dst_stride  ]), rnd_avg32(a, b));\
945 946
        a= AV_RN32(&src1[i*src_stride1+4]);\
        b= AV_RN32(&src2[i*src_stride2+4]);\
Michael Niedermayer's avatar
Michael Niedermayer committed
947
        OP(*((uint32_t*)&dst[i*dst_stride+4]), rnd_avg32(a, b));\
948 949 950
    }\
}\
\
951 952 953 954 955
static inline void OPNAME ## _pixels4_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
                                                int src_stride1, int src_stride2, int h){\
    int i;\
    for(i=0; i<h; i++){\
        uint32_t a,b;\
956 957
        a= AV_RN32(&src1[i*src_stride1  ]);\
        b= AV_RN32(&src2[i*src_stride2  ]);\
Michael Niedermayer's avatar
Michael Niedermayer committed
958
        OP(*((uint32_t*)&dst[i*dst_stride  ]), rnd_avg32(a, b));\
959 960 961
    }\
}\
\
962 963 964 965 966
static inline void OPNAME ## _pixels2_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
                                                int src_stride1, int src_stride2, int h){\
    int i;\
    for(i=0; i<h; i++){\
        uint32_t a,b;\
967 968
        a= AV_RN16(&src1[i*src_stride1  ]);\
        b= AV_RN16(&src2[i*src_stride2  ]);\
969 970 971 972
        OP(*((uint16_t*)&dst[i*dst_stride  ]), rnd_avg32(a, b));\
    }\
}\
\
Michael Niedermayer's avatar
Michael Niedermayer committed
973 974 975 976 977 978 979 980 981 982 983 984
static inline void OPNAME ## _pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
                                                int src_stride1, int src_stride2, int h){\
    OPNAME ## _pixels8_l2(dst  , src1  , src2  , dst_stride, src_stride1, src_stride2, h);\
    OPNAME ## _pixels8_l2(dst+8, src1+8, src2+8, dst_stride, src_stride1, src_stride2, h);\
}\
\
static inline void OPNAME ## _no_rnd_pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \
                                                int src_stride1, int src_stride2, int h){\
    OPNAME ## _no_rnd_pixels8_l2(dst  , src1  , src2  , dst_stride, src_stride1, src_stride2, h);\
    OPNAME ## _no_rnd_pixels8_l2(dst+8, src1+8, src2+8, dst_stride, src_stride1, src_stride2, h);\
}\
\
985
static inline void OPNAME ## _no_rnd_pixels8_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
Michael Niedermayer's avatar
Michael Niedermayer committed
986 987 988
    OPNAME ## _no_rnd_pixels8_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\
}\
\
989
static inline void OPNAME ## _pixels8_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
Michael Niedermayer's avatar
Michael Niedermayer committed
990 991 992
    OPNAME ## _pixels8_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\
}\
\
993
static inline void OPNAME ## _no_rnd_pixels8_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
Michael Niedermayer's avatar
Michael Niedermayer committed
994 995 996
    OPNAME ## _no_rnd_pixels8_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\
}\
\
997
static inline void OPNAME ## _pixels8_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
Michael Niedermayer's avatar
Michael Niedermayer committed
998 999 1000 1001 1002
    OPNAME ## _pixels8_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\
}\
\
static inline void OPNAME ## _pixels8_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\
                 int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
1003 1004
    int i;\
    for(i=0; i<h; i++){\
Michael Niedermayer's avatar
Michael Niedermayer committed
1005
        uint32_t a, b, c, d, l0, l1, h0, h1;\
1006 1007 1008 1009
        a= AV_RN32(&src1[i*src_stride1]);\
        b= AV_RN32(&src2[i*src_stride2]);\
        c= AV_RN32(&src3[i*src_stride3]);\
        d= AV_RN32(&src4[i*src_stride4]);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
        l0=  (a&0x03030303UL)\
           + (b&0x03030303UL)\
           + 0x02020202UL;\
        h0= ((a&0xFCFCFCFCUL)>>2)\
          + ((b&0xFCFCFCFCUL)>>2);\
        l1=  (c&0x03030303UL)\
           + (d&0x03030303UL);\
        h1= ((c&0xFCFCFCFCUL)>>2)\
          + ((d&0xFCFCFCFCUL)>>2);\
        OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
1020 1021 1022 1023
        a= AV_RN32(&src1[i*src_stride1+4]);\
        b= AV_RN32(&src2[i*src_stride2+4]);\
        c= AV_RN32(&src3[i*src_stride3+4]);\
        d= AV_RN32(&src4[i*src_stride4+4]);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
        l0=  (a&0x03030303UL)\
           + (b&0x03030303UL)\
           + 0x02020202UL;\
        h0= ((a&0xFCFCFCFCUL)>>2)\
          + ((b&0xFCFCFCFCUL)>>2);\
        l1=  (c&0x03030303UL)\
           + (d&0x03030303UL);\
        h1= ((c&0xFCFCFCFCUL)>>2)\
          + ((d&0xFCFCFCFCUL)>>2);\
        OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
1034 1035
    }\
}\
1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
\
static inline void OPNAME ## _pixels4_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
    OPNAME ## _pixels4_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\
}\
\
static inline void OPNAME ## _pixels4_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
    OPNAME ## _pixels4_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\
}\
\
static inline void OPNAME ## _pixels2_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
    OPNAME ## _pixels2_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\
}\
\
static inline void OPNAME ## _pixels2_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
    OPNAME ## _pixels2_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\
}\
\
Michael Niedermayer's avatar
Michael Niedermayer committed
1053 1054
static inline void OPNAME ## _no_rnd_pixels8_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\
                 int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
1055 1056
    int i;\
    for(i=0; i<h; i++){\
Michael Niedermayer's avatar
Michael Niedermayer committed
1057
        uint32_t a, b, c, d, l0, l1, h0, h1;\
1058 1059 1060 1061
        a= AV_RN32(&src1[i*src_stride1]);\
        b= AV_RN32(&src2[i*src_stride2]);\
        c= AV_RN32(&src3[i*src_stride3]);\
        d= AV_RN32(&src4[i*src_stride4]);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1062 1063 1064 1065 1066 1067 1068 1069 1070 1071
        l0=  (a&0x03030303UL)\
           + (b&0x03030303UL)\
           + 0x01010101UL;\
        h0= ((a&0xFCFCFCFCUL)>>2)\
          + ((b&0xFCFCFCFCUL)>>2);\
        l1=  (c&0x03030303UL)\
           + (d&0x03030303UL);\
        h1= ((c&0xFCFCFCFCUL)>>2)\
          + ((d&0xFCFCFCFCUL)>>2);\
        OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
1072 1073 1074 1075
        a= AV_RN32(&src1[i*src_stride1+4]);\
        b= AV_RN32(&src2[i*src_stride2+4]);\
        c= AV_RN32(&src3[i*src_stride3+4]);\
        d= AV_RN32(&src4[i*src_stride4+4]);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
        l0=  (a&0x03030303UL)\
           + (b&0x03030303UL)\
           + 0x01010101UL;\
        h0= ((a&0xFCFCFCFCUL)>>2)\
          + ((b&0xFCFCFCFCUL)>>2);\
        l1=  (c&0x03030303UL)\
           + (d&0x03030303UL);\
        h1= ((c&0xFCFCFCFCUL)>>2)\
          + ((d&0xFCFCFCFCUL)>>2);\
        OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
1086 1087
    }\
}\
Michael Niedermayer's avatar
Michael Niedermayer committed
1088 1089 1090 1091 1092 1093 1094 1095 1096 1097
static inline void OPNAME ## _pixels16_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\
                 int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
    OPNAME ## _pixels8_l4(dst  , src1  , src2  , src3  , src4  , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
    OPNAME ## _pixels8_l4(dst+8, src1+8, src2+8, src3+8, src4+8, dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
}\
static inline void OPNAME ## _no_rnd_pixels16_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\
                 int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\
    OPNAME ## _no_rnd_pixels8_l4(dst  , src1  , src2  , src3  , src4  , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
    OPNAME ## _no_rnd_pixels8_l4(dst+8, src1+8, src2+8, src3+8, src4+8, dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\
}\
1098
\
1099 1100 1101 1102 1103 1104 1105 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
static inline void OPNAME ## _pixels2_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
{\
        int i, a0, b0, a1, b1;\
        a0= pixels[0];\
        b0= pixels[1] + 2;\
        a0 += b0;\
        b0 += pixels[2];\
\
        pixels+=line_size;\
        for(i=0; i<h; i+=2){\
            a1= pixels[0];\
            b1= pixels[1];\
            a1 += b1;\
            b1 += pixels[2];\
\
            block[0]= (a1+a0)>>2; /* FIXME non put */\
            block[1]= (b1+b0)>>2;\
\
            pixels+=line_size;\
            block +=line_size;\
\
            a0= pixels[0];\
            b0= pixels[1] + 2;\
            a0 += b0;\
            b0 += pixels[2];\
\
            block[0]= (a1+a0)>>2;\
            block[1]= (b1+b0)>>2;\
            pixels+=line_size;\
            block +=line_size;\
        }\
}\
\
static inline void OPNAME ## _pixels4_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
{\
        int i;\
1135 1136
        const uint32_t a= AV_RN32(pixels  );\
        const uint32_t b= AV_RN32(pixels+1);\
1137 1138 1139 1140 1141 1142 1143 1144 1145
        uint32_t l0=  (a&0x03030303UL)\
                    + (b&0x03030303UL)\
                    + 0x02020202UL;\
        uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\
                   + ((b&0xFCFCFCFCUL)>>2);\
        uint32_t l1,h1;\
\
        pixels+=line_size;\
        for(i=0; i<h; i+=2){\
1146 1147
            uint32_t a= AV_RN32(pixels  );\
            uint32_t b= AV_RN32(pixels+1);\
1148 1149 1150 1151 1152 1153 1154
            l1=  (a&0x03030303UL)\
               + (b&0x03030303UL);\
            h1= ((a&0xFCFCFCFCUL)>>2)\
              + ((b&0xFCFCFCFCUL)>>2);\
            OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
            pixels+=line_size;\
            block +=line_size;\
1155 1156
            a= AV_RN32(pixels  );\
            b= AV_RN32(pixels+1);\
1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167
            l0=  (a&0x03030303UL)\
               + (b&0x03030303UL)\
               + 0x02020202UL;\
            h0= ((a&0xFCFCFCFCUL)>>2)\
              + ((b&0xFCFCFCFCUL)>>2);\
            OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
            pixels+=line_size;\
            block +=line_size;\
        }\
}\
\
1168
static inline void OPNAME ## _pixels8_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
1169 1170 1171 1172
{\
    int j;\
    for(j=0; j<2; j++){\
        int i;\
1173 1174
        const uint32_t a= AV_RN32(pixels  );\
        const uint32_t b= AV_RN32(pixels+1);\
1175 1176 1177 1178 1179 1180 1181 1182 1183
        uint32_t l0=  (a&0x03030303UL)\
                    + (b&0x03030303UL)\
                    + 0x02020202UL;\
        uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\
                   + ((b&0xFCFCFCFCUL)>>2);\
        uint32_t l1,h1;\
\
        pixels+=line_size;\
        for(i=0; i<h; i+=2){\
1184 1185
            uint32_t a= AV_RN32(pixels  );\
            uint32_t b= AV_RN32(pixels+1);\
1186 1187 1188 1189 1190 1191 1192
            l1=  (a&0x03030303UL)\
               + (b&0x03030303UL);\
            h1= ((a&0xFCFCFCFCUL)>>2)\
              + ((b&0xFCFCFCFCUL)>>2);\
            OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
            pixels+=line_size;\
            block +=line_size;\
1193 1194
            a= AV_RN32(pixels  );\
            b= AV_RN32(pixels+1);\
1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208
            l0=  (a&0x03030303UL)\
               + (b&0x03030303UL)\
               + 0x02020202UL;\
            h0= ((a&0xFCFCFCFCUL)>>2)\
              + ((b&0xFCFCFCFCUL)>>2);\
            OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
            pixels+=line_size;\
            block +=line_size;\
        }\
        pixels+=4-line_size*(h+1);\
        block +=4-line_size*h;\
    }\
}\
\
1209
static inline void OPNAME ## _no_rnd_pixels8_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\
1210 1211 1212 1213
{\
    int j;\
    for(j=0; j<2; j++){\
        int i;\
1214 1215
        const uint32_t a= AV_RN32(pixels  );\
        const uint32_t b= AV_RN32(pixels+1);\
1216 1217 1218 1219 1220 1221 1222 1223 1224
        uint32_t l0=  (a&0x03030303UL)\
                    + (b&0x03030303UL)\
                    + 0x01010101UL;\
        uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\
                   + ((b&0xFCFCFCFCUL)>>2);\
        uint32_t l1,h1;\
\
        pixels+=line_size;\
        for(i=0; i<h; i+=2){\
1225 1226
            uint32_t a= AV_RN32(pixels  );\
            uint32_t b= AV_RN32(pixels+1);\
1227 1228 1229 1230 1231 1232 1233
            l1=  (a&0x03030303UL)\
               + (b&0x03030303UL);\
            h1= ((a&0xFCFCFCFCUL)>>2)\
              + ((b&0xFCFCFCFCUL)>>2);\
            OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
            pixels+=line_size;\
            block +=line_size;\
1234 1235
            a= AV_RN32(pixels  );\
            b= AV_RN32(pixels+1);\
1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
            l0=  (a&0x03030303UL)\
               + (b&0x03030303UL)\
               + 0x01010101UL;\
            h0= ((a&0xFCFCFCFCUL)>>2)\
              + ((b&0xFCFCFCFCUL)>>2);\
            OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\
            pixels+=line_size;\
            block +=line_size;\
        }\
        pixels+=4-line_size*(h+1);\
        block +=4-line_size*h;\
    }\
}\
\
1250 1251 1252 1253 1254 1255 1256 1257
CALL_2X_PIXELS(OPNAME ## _pixels16_c  , OPNAME ## _pixels8_c  , 8)\
CALL_2X_PIXELS(OPNAME ## _pixels16_x2_c , OPNAME ## _pixels8_x2_c , 8)\
CALL_2X_PIXELS(OPNAME ## _pixels16_y2_c , OPNAME ## _pixels8_y2_c , 8)\
CALL_2X_PIXELS(OPNAME ## _pixels16_xy2_c, OPNAME ## _pixels8_xy2_c, 8)\
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_c  , OPNAME ## _pixels8_c         , 8)\
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_x2_c , OPNAME ## _no_rnd_pixels8_x2_c , 8)\
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_y2_c , OPNAME ## _no_rnd_pixels8_y2_c , 8)\
CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_xy2_c, OPNAME ## _no_rnd_pixels8_xy2_c, 8)\
Michael Niedermayer's avatar
Michael Niedermayer committed
1258

Michael Niedermayer's avatar
Michael Niedermayer committed
1259
#define op_avg(a, b) a = rnd_avg32(a, b)
1260 1261 1262 1263 1264 1265 1266 1267
#endif
#define op_put(a, b) a = b

PIXOP2(avg, op_avg)
PIXOP2(put, op_put)
#undef op_avg
#undef op_put

Fabrice Bellard's avatar
Fabrice Bellard committed
1268 1269 1270
#define avg2(a,b) ((a+b+1)>>1)
#define avg4(a,b,c,d) ((a+b+c+d+2)>>2)

1271 1272 1273 1274 1275 1276 1277
static void put_no_rnd_pixels16_l2_c(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h){
    put_no_rnd_pixels16_l2(dst, a, b, stride, stride, stride, h);
}

static void put_no_rnd_pixels8_l2_c(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h){
    put_no_rnd_pixels8_l2(dst, a, b, stride, stride, stride, h);
}
1278

1279
static void gmc1_c(uint8_t *dst, uint8_t *src, int stride, int h, int x16, int y16, int rounder)
Michael Niedermayer's avatar
Michael Niedermayer committed
1280 1281 1282 1283 1284 1285 1286 1287 1288
{
    const int A=(16-x16)*(16-y16);
    const int B=(   x16)*(16-y16);
    const int C=(16-x16)*(   y16);
    const int D=(   x16)*(   y16);
    int i;

    for(i=0; i<h; i++)
    {
Michael Niedermayer's avatar
Michael Niedermayer committed
1289 1290 1291 1292 1293 1294 1295 1296 1297 1298
        dst[0]= (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + rounder)>>8;
        dst[1]= (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + rounder)>>8;
        dst[2]= (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + rounder)>>8;
        dst[3]= (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + rounder)>>8;
        dst[4]= (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + rounder)>>8;
        dst[5]= (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + rounder)>>8;
        dst[6]= (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + rounder)>>8;
        dst[7]= (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + rounder)>>8;
        dst+= stride;
        src+= stride;
Michael Niedermayer's avatar
Michael Niedermayer committed
1299 1300 1301
    }
}

1302
void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
1303 1304 1305 1306
                  int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height)
{
    int y, vx, vy;
    const int s= 1<<shift;
1307

1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324
    width--;
    height--;

    for(y=0; y<h; y++){
        int x;

        vx= ox;
        vy= oy;
        for(x=0; x<8; x++){ //XXX FIXME optimize
            int src_x, src_y, frac_x, frac_y, index;

            src_x= vx>>16;
            src_y= vy>>16;
            frac_x= src_x&(s-1);
            frac_y= src_y&(s-1);
            src_x>>=shift;
            src_y>>=shift;
1325

1326 1327 1328 1329 1330 1331 1332 1333 1334
            if((unsigned)src_x < width){
                if((unsigned)src_y < height){
                    index= src_x + src_y*stride;
                    dst[y*stride + x]= (  (  src[index         ]*(s-frac_x)
                                           + src[index       +1]*   frac_x )*(s-frac_y)
                                        + (  src[index+stride  ]*(s-frac_x)
                                           + src[index+stride+1]*   frac_x )*   frac_y
                                        + r)>>(shift*2);
                }else{
1335
                    index= src_x + av_clip(src_y, 0, height)*stride;
1336
                    dst[y*stride + x]= ( (  src[index         ]*(s-frac_x)
1337 1338 1339 1340 1341
                                          + src[index       +1]*   frac_x )*s
                                        + r)>>(shift*2);
                }
            }else{
                if((unsigned)src_y < height){
1342
                    index= av_clip(src_x, 0, width) + src_y*stride;
1343
                    dst[y*stride + x]= (  (  src[index         ]*(s-frac_y)
1344 1345 1346
                                           + src[index+stride  ]*   frac_y )*s
                                        + r)>>(shift*2);
                }else{
1347
                    index= av_clip(src_x, 0, width) + av_clip(src_y, 0, height)*stride;
1348 1349 1350
                    dst[y*stride + x]=    src[index         ];
                }
            }
1351

1352 1353 1354 1355 1356 1357 1358
            vx+= dxx;
            vy+= dyx;
        }
        ox += dxy;
        oy += dyy;
    }
}
1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372

static inline void put_tpel_pixels_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    switch(width){
    case 2: put_pixels2_c (dst, src, stride, height); break;
    case 4: put_pixels4_c (dst, src, stride, height); break;
    case 8: put_pixels8_c (dst, src, stride, height); break;
    case 16:put_pixels16_c(dst, src, stride, height); break;
    }
}

static inline void put_tpel_pixels_mc10_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1373
        dst[j] = (683*(2*src[j] + src[j+1] + 1)) >> 11;
1374 1375 1376 1377 1378 1379 1380 1381 1382 1383
      }
      src += stride;
      dst += stride;
    }
}

static inline void put_tpel_pixels_mc20_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1384
        dst[j] = (683*(src[j] + 2*src[j+1] + 1)) >> 11;
1385 1386 1387 1388 1389
      }
      src += stride;
      dst += stride;
    }
}
1390

1391 1392 1393 1394
static inline void put_tpel_pixels_mc01_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1395
        dst[j] = (683*(2*src[j] + src[j+stride] + 1)) >> 11;
1396 1397 1398 1399 1400
      }
      src += stride;
      dst += stride;
    }
}
1401

1402 1403 1404 1405
static inline void put_tpel_pixels_mc11_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1406
        dst[j] = (2731*(4*src[j] + 3*src[j+1] + 3*src[j+stride] + 2*src[j+stride+1] + 6)) >> 15;
1407 1408 1409 1410 1411 1412 1413 1414 1415 1416
      }
      src += stride;
      dst += stride;
    }
}

static inline void put_tpel_pixels_mc12_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1417
        dst[j] = (2731*(3*src[j] + 2*src[j+1] + 4*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15;
1418 1419 1420 1421 1422 1423 1424 1425 1426 1427
      }
      src += stride;
      dst += stride;
    }
}

static inline void put_tpel_pixels_mc02_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1428
        dst[j] = (683*(src[j] + 2*src[j+stride] + 1)) >> 11;
1429 1430 1431 1432 1433 1434 1435 1436 1437 1438
      }
      src += stride;
      dst += stride;
    }
}

static inline void put_tpel_pixels_mc21_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1439
        dst[j] = (2731*(3*src[j] + 4*src[j+1] + 2*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15;
1440 1441 1442 1443 1444 1445 1446 1447 1448 1449
      }
      src += stride;
      dst += stride;
    }
}

static inline void put_tpel_pixels_mc22_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1450
        dst[j] = (2731*(2*src[j] + 3*src[j+1] + 3*src[j+stride] + 4*src[j+stride+1] + 6)) >> 15;
1451 1452 1453 1454 1455
      }
      src += stride;
      dst += stride;
    }
}
1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469

static inline void avg_tpel_pixels_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    switch(width){
    case 2: avg_pixels2_c (dst, src, stride, height); break;
    case 4: avg_pixels4_c (dst, src, stride, height); break;
    case 8: avg_pixels8_c (dst, src, stride, height); break;
    case 16:avg_pixels16_c(dst, src, stride, height); break;
    }
}

static inline void avg_tpel_pixels_mc10_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1470
        dst[j] = (dst[j] + ((683*(2*src[j] + src[j+1] + 1)) >> 11) + 1) >> 1;
1471 1472 1473 1474 1475 1476 1477 1478 1479 1480
      }
      src += stride;
      dst += stride;
    }
}

static inline void avg_tpel_pixels_mc20_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1481
        dst[j] = (dst[j] + ((683*(src[j] + 2*src[j+1] + 1)) >> 11) + 1) >> 1;
1482 1483 1484 1485 1486
      }
      src += stride;
      dst += stride;
    }
}
1487

1488 1489 1490 1491
static inline void avg_tpel_pixels_mc01_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1492
        dst[j] = (dst[j] + ((683*(2*src[j] + src[j+stride] + 1)) >> 11) + 1) >> 1;
1493 1494 1495 1496 1497
      }
      src += stride;
      dst += stride;
    }
}
1498

1499 1500 1501 1502
static inline void avg_tpel_pixels_mc11_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1503
        dst[j] = (dst[j] + ((2731*(4*src[j] + 3*src[j+1] + 3*src[j+stride] + 2*src[j+stride+1] + 6)) >> 15) + 1) >> 1;
1504 1505 1506 1507 1508 1509 1510 1511 1512 1513
      }
      src += stride;
      dst += stride;
    }
}

static inline void avg_tpel_pixels_mc12_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1514
        dst[j] = (dst[j] + ((2731*(3*src[j] + 2*src[j+1] + 4*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15) + 1) >> 1;
1515 1516 1517 1518 1519 1520 1521 1522 1523 1524
      }
      src += stride;
      dst += stride;
    }
}

static inline void avg_tpel_pixels_mc02_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1525
        dst[j] = (dst[j] + ((683*(src[j] + 2*src[j+stride] + 1)) >> 11) + 1) >> 1;
1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
      }
      src += stride;
      dst += stride;
    }
}

static inline void avg_tpel_pixels_mc21_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1536
        dst[j] = (dst[j] + ((2731*(3*src[j] + 4*src[j+1] + 2*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15) + 1) >> 1;
1537 1538 1539 1540 1541 1542 1543 1544 1545 1546
      }
      src += stride;
      dst += stride;
    }
}

static inline void avg_tpel_pixels_mc22_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){
    int i,j;
    for (i=0; i < height; i++) {
      for (j=0; j < width; j++) {
1547
        dst[j] = (dst[j] + ((2731*(2*src[j] + 3*src[j+1] + 3*src[j+stride] + 4*src[j+stride+1] + 6)) >> 15) + 1) >> 1;
1548 1549 1550 1551 1552
      }
      src += stride;
      dst += stride;
    }
}
1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574
#if 0
#define TPEL_WIDTH(width)\
static void put_tpel_pixels ## width ## _mc00_c(uint8_t *dst, const uint8_t *src, int stride, int height){\
    void put_tpel_pixels_mc00_c(dst, src, stride, width, height);}\
static void put_tpel_pixels ## width ## _mc10_c(uint8_t *dst, const uint8_t *src, int stride, int height){\
    void put_tpel_pixels_mc10_c(dst, src, stride, width, height);}\
static void put_tpel_pixels ## width ## _mc20_c(uint8_t *dst, const uint8_t *src, int stride, int height){\
    void put_tpel_pixels_mc20_c(dst, src, stride, width, height);}\
static void put_tpel_pixels ## width ## _mc01_c(uint8_t *dst, const uint8_t *src, int stride, int height){\
    void put_tpel_pixels_mc01_c(dst, src, stride, width, height);}\
static void put_tpel_pixels ## width ## _mc11_c(uint8_t *dst, const uint8_t *src, int stride, int height){\
    void put_tpel_pixels_mc11_c(dst, src, stride, width, height);}\
static void put_tpel_pixels ## width ## _mc21_c(uint8_t *dst, const uint8_t *src, int stride, int height){\
    void put_tpel_pixels_mc21_c(dst, src, stride, width, height);}\
static void put_tpel_pixels ## width ## _mc02_c(uint8_t *dst, const uint8_t *src, int stride, int height){\
    void put_tpel_pixels_mc02_c(dst, src, stride, width, height);}\
static void put_tpel_pixels ## width ## _mc12_c(uint8_t *dst, const uint8_t *src, int stride, int height){\
    void put_tpel_pixels_mc12_c(dst, src, stride, width, height);}\
static void put_tpel_pixels ## width ## _mc22_c(uint8_t *dst, const uint8_t *src, int stride, int height){\
    void put_tpel_pixels_mc22_c(dst, src, stride, width, height);}
#endif

1575 1576 1577 1578 1579 1580 1581 1582 1583 1584
#define H264_CHROMA_MC(OPNAME, OP)\
static void OPNAME ## h264_chroma_mc2_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
    const int A=(8-x)*(8-y);\
    const int B=(  x)*(8-y);\
    const int C=(8-x)*(  y);\
    const int D=(  x)*(  y);\
    int i;\
    \
    assert(x<8 && y<8 && x>=0 && y>=0);\
\
1585
    if(D){\
1586
        for(i=0; i<h; i++){\
Michael Niedermayer's avatar
Michael Niedermayer committed
1587 1588 1589 1590 1591
            OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\
            OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\
            dst+= stride;\
            src+= stride;\
        }\
1592 1593 1594
    }else{\
        const int E= B+C;\
        const int step= C ? stride : 1;\
1595
        for(i=0; i<h; i++){\
1596 1597 1598 1599 1600 1601
            OP(dst[0], (A*src[0] + E*src[step+0]));\
            OP(dst[1], (A*src[1] + E*src[step+1]));\
            dst+= stride;\
            src+= stride;\
        }\
    }\
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612
}\
\
static void OPNAME ## h264_chroma_mc4_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
    const int A=(8-x)*(8-y);\
    const int B=(  x)*(8-y);\
    const int C=(8-x)*(  y);\
    const int D=(  x)*(  y);\
    int i;\
    \
    assert(x<8 && y<8 && x>=0 && y>=0);\
\
1613
    if(D){\
1614
        for(i=0; i<h; i++){\
Michael Niedermayer's avatar
Michael Niedermayer committed
1615 1616 1617 1618 1619 1620 1621
            OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\
            OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\
            OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3]));\
            OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4]));\
            dst+= stride;\
            src+= stride;\
        }\
1622 1623 1624
    }else{\
        const int E= B+C;\
        const int step= C ? stride : 1;\
1625
        for(i=0; i<h; i++){\
1626 1627 1628 1629 1630 1631 1632 1633
            OP(dst[0], (A*src[0] + E*src[step+0]));\
            OP(dst[1], (A*src[1] + E*src[step+1]));\
            OP(dst[2], (A*src[2] + E*src[step+2]));\
            OP(dst[3], (A*src[3] + E*src[step+3]));\
            dst+= stride;\
            src+= stride;\
        }\
    }\
1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644
}\
\
static void OPNAME ## h264_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
    const int A=(8-x)*(8-y);\
    const int B=(  x)*(8-y);\
    const int C=(8-x)*(  y);\
    const int D=(  x)*(  y);\
    int i;\
    \
    assert(x<8 && y<8 && x>=0 && y>=0);\
\
1645
    if(D){\
1646
        for(i=0; i<h; i++){\
Michael Niedermayer's avatar
Michael Niedermayer committed
1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657
            OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\
            OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\
            OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3]));\
            OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4]));\
            OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5]));\
            OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6]));\
            OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7]));\
            OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8]));\
            dst+= stride;\
            src+= stride;\
        }\
1658 1659 1660
    }else{\
        const int E= B+C;\
        const int step= C ? stride : 1;\
1661
        for(i=0; i<h; i++){\
1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673
            OP(dst[0], (A*src[0] + E*src[step+0]));\
            OP(dst[1], (A*src[1] + E*src[step+1]));\
            OP(dst[2], (A*src[2] + E*src[step+2]));\
            OP(dst[3], (A*src[3] + E*src[step+3]));\
            OP(dst[4], (A*src[4] + E*src[step+4]));\
            OP(dst[5], (A*src[5] + E*src[step+5]));\
            OP(dst[6], (A*src[6] + E*src[step+6]));\
            OP(dst[7], (A*src[7] + E*src[step+7]));\
            dst+= stride;\
            src+= stride;\
        }\
    }\
1674 1675 1676 1677 1678 1679 1680 1681 1682 1683
}

#define op_avg(a, b) a = (((a)+(((b) + 32)>>6)+1)>>1)
#define op_put(a, b) a = (((b) + 32)>>6)

H264_CHROMA_MC(put_       , op_put)
H264_CHROMA_MC(avg_       , op_avg)
#undef op_avg
#undef op_put

1684
static void put_no_rnd_vc1_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){
1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707
    const int A=(8-x)*(8-y);
    const int B=(  x)*(8-y);
    const int C=(8-x)*(  y);
    const int D=(  x)*(  y);
    int i;

    assert(x<8 && y<8 && x>=0 && y>=0);

    for(i=0; i<h; i++)
    {
        dst[0] = (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6;
        dst[1] = (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6;
        dst[2] = (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6;
        dst[3] = (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6;
        dst[4] = (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + 32 - 4) >> 6;
        dst[5] = (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + 32 - 4) >> 6;
        dst[6] = (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + 32 - 4) >> 6;
        dst[7] = (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + 32 - 4) >> 6;
        dst+= stride;
        src+= stride;
    }
}

1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731
static void avg_no_rnd_vc1_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){
    const int A=(8-x)*(8-y);
    const int B=(  x)*(8-y);
    const int C=(8-x)*(  y);
    const int D=(  x)*(  y);
    int i;

    assert(x<8 && y<8 && x>=0 && y>=0);

    for(i=0; i<h; i++)
    {
        dst[0] = avg2(dst[0], ((A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + 32 - 4) >> 6));
        dst[1] = avg2(dst[1], ((A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6));
        dst[2] = avg2(dst[2], ((A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6));
        dst[3] = avg2(dst[3], ((A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6));
        dst[4] = avg2(dst[4], ((A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + 32 - 4) >> 6));
        dst[5] = avg2(dst[5], ((A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + 32 - 4) >> 6));
        dst[6] = avg2(dst[6], ((A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + 32 - 4) >> 6));
        dst[7] = avg2(dst[7], ((A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + 32 - 4) >> 6));
        dst+= stride;
        src+= stride;
    }
}

Michael Niedermayer's avatar
Michael Niedermayer committed
1732
#define QPEL_MC(r, OPNAME, RND, OP) \
1733
static void OPNAME ## mpeg4_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
1734
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
Michael Niedermayer's avatar
Michael Niedermayer committed
1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748
    int i;\
    for(i=0; i<h; i++)\
    {\
        OP(dst[0], (src[0]+src[1])*20 - (src[0]+src[2])*6 + (src[1]+src[3])*3 - (src[2]+src[4]));\
        OP(dst[1], (src[1]+src[2])*20 - (src[0]+src[3])*6 + (src[0]+src[4])*3 - (src[1]+src[5]));\
        OP(dst[2], (src[2]+src[3])*20 - (src[1]+src[4])*6 + (src[0]+src[5])*3 - (src[0]+src[6]));\
        OP(dst[3], (src[3]+src[4])*20 - (src[2]+src[5])*6 + (src[1]+src[6])*3 - (src[0]+src[7]));\
        OP(dst[4], (src[4]+src[5])*20 - (src[3]+src[6])*6 + (src[2]+src[7])*3 - (src[1]+src[8]));\
        OP(dst[5], (src[5]+src[6])*20 - (src[4]+src[7])*6 + (src[3]+src[8])*3 - (src[2]+src[8]));\
        OP(dst[6], (src[6]+src[7])*20 - (src[5]+src[8])*6 + (src[4]+src[8])*3 - (src[3]+src[7]));\
        OP(dst[7], (src[7]+src[8])*20 - (src[6]+src[8])*6 + (src[5]+src[7])*3 - (src[4]+src[6]));\
        dst+=dstStride;\
        src+=srcStride;\
    }\
Michael Niedermayer's avatar
Michael Niedermayer committed
1749 1750
}\
\
1751
static void OPNAME ## mpeg4_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
Michael Niedermayer's avatar
Michael Niedermayer committed
1752
    const int w=8;\
1753
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
Michael Niedermayer's avatar
Michael Niedermayer committed
1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778
    int i;\
    for(i=0; i<w; i++)\
    {\
        const int src0= src[0*srcStride];\
        const int src1= src[1*srcStride];\
        const int src2= src[2*srcStride];\
        const int src3= src[3*srcStride];\
        const int src4= src[4*srcStride];\
        const int src5= src[5*srcStride];\
        const int src6= src[6*srcStride];\
        const int src7= src[7*srcStride];\
        const int src8= src[8*srcStride];\
        OP(dst[0*dstStride], (src0+src1)*20 - (src0+src2)*6 + (src1+src3)*3 - (src2+src4));\
        OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*6 + (src0+src4)*3 - (src1+src5));\
        OP(dst[2*dstStride], (src2+src3)*20 - (src1+src4)*6 + (src0+src5)*3 - (src0+src6));\
        OP(dst[3*dstStride], (src3+src4)*20 - (src2+src5)*6 + (src1+src6)*3 - (src0+src7));\
        OP(dst[4*dstStride], (src4+src5)*20 - (src3+src6)*6 + (src2+src7)*3 - (src1+src8));\
        OP(dst[5*dstStride], (src5+src6)*20 - (src4+src7)*6 + (src3+src8)*3 - (src2+src8));\
        OP(dst[6*dstStride], (src6+src7)*20 - (src5+src8)*6 + (src4+src8)*3 - (src3+src7));\
        OP(dst[7*dstStride], (src7+src8)*20 - (src6+src8)*6 + (src5+src7)*3 - (src4+src6));\
        dst++;\
        src++;\
    }\
}\
\
1779
static void OPNAME ## mpeg4_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
1780
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
Michael Niedermayer's avatar
Michael Niedermayer committed
1781
    int i;\
Michael Niedermayer's avatar
Michael Niedermayer committed
1782
    \
Michael Niedermayer's avatar
Michael Niedermayer committed
1783 1784 1785 1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805
    for(i=0; i<h; i++)\
    {\
        OP(dst[ 0], (src[ 0]+src[ 1])*20 - (src[ 0]+src[ 2])*6 + (src[ 1]+src[ 3])*3 - (src[ 2]+src[ 4]));\
        OP(dst[ 1], (src[ 1]+src[ 2])*20 - (src[ 0]+src[ 3])*6 + (src[ 0]+src[ 4])*3 - (src[ 1]+src[ 5]));\
        OP(dst[ 2], (src[ 2]+src[ 3])*20 - (src[ 1]+src[ 4])*6 + (src[ 0]+src[ 5])*3 - (src[ 0]+src[ 6]));\
        OP(dst[ 3], (src[ 3]+src[ 4])*20 - (src[ 2]+src[ 5])*6 + (src[ 1]+src[ 6])*3 - (src[ 0]+src[ 7]));\
        OP(dst[ 4], (src[ 4]+src[ 5])*20 - (src[ 3]+src[ 6])*6 + (src[ 2]+src[ 7])*3 - (src[ 1]+src[ 8]));\
        OP(dst[ 5], (src[ 5]+src[ 6])*20 - (src[ 4]+src[ 7])*6 + (src[ 3]+src[ 8])*3 - (src[ 2]+src[ 9]));\
        OP(dst[ 6], (src[ 6]+src[ 7])*20 - (src[ 5]+src[ 8])*6 + (src[ 4]+src[ 9])*3 - (src[ 3]+src[10]));\
        OP(dst[ 7], (src[ 7]+src[ 8])*20 - (src[ 6]+src[ 9])*6 + (src[ 5]+src[10])*3 - (src[ 4]+src[11]));\
        OP(dst[ 8], (src[ 8]+src[ 9])*20 - (src[ 7]+src[10])*6 + (src[ 6]+src[11])*3 - (src[ 5]+src[12]));\
        OP(dst[ 9], (src[ 9]+src[10])*20 - (src[ 8]+src[11])*6 + (src[ 7]+src[12])*3 - (src[ 6]+src[13]));\
        OP(dst[10], (src[10]+src[11])*20 - (src[ 9]+src[12])*6 + (src[ 8]+src[13])*3 - (src[ 7]+src[14]));\
        OP(dst[11], (src[11]+src[12])*20 - (src[10]+src[13])*6 + (src[ 9]+src[14])*3 - (src[ 8]+src[15]));\
        OP(dst[12], (src[12]+src[13])*20 - (src[11]+src[14])*6 + (src[10]+src[15])*3 - (src[ 9]+src[16]));\
        OP(dst[13], (src[13]+src[14])*20 - (src[12]+src[15])*6 + (src[11]+src[16])*3 - (src[10]+src[16]));\
        OP(dst[14], (src[14]+src[15])*20 - (src[13]+src[16])*6 + (src[12]+src[16])*3 - (src[11]+src[15]));\
        OP(dst[15], (src[15]+src[16])*20 - (src[14]+src[16])*6 + (src[13]+src[15])*3 - (src[12]+src[14]));\
        dst+=dstStride;\
        src+=srcStride;\
    }\
}\
\
1806
static void OPNAME ## mpeg4_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
1807
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
Michael Niedermayer's avatar
Michael Niedermayer committed
1808
    int i;\
Michael Niedermayer's avatar
Michael Niedermayer committed
1809
    const int w=16;\
Michael Niedermayer's avatar
Michael Niedermayer committed
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849
    for(i=0; i<w; i++)\
    {\
        const int src0= src[0*srcStride];\
        const int src1= src[1*srcStride];\
        const int src2= src[2*srcStride];\
        const int src3= src[3*srcStride];\
        const int src4= src[4*srcStride];\
        const int src5= src[5*srcStride];\
        const int src6= src[6*srcStride];\
        const int src7= src[7*srcStride];\
        const int src8= src[8*srcStride];\
        const int src9= src[9*srcStride];\
        const int src10= src[10*srcStride];\
        const int src11= src[11*srcStride];\
        const int src12= src[12*srcStride];\
        const int src13= src[13*srcStride];\
        const int src14= src[14*srcStride];\
        const int src15= src[15*srcStride];\
        const int src16= src[16*srcStride];\
        OP(dst[ 0*dstStride], (src0 +src1 )*20 - (src0 +src2 )*6 + (src1 +src3 )*3 - (src2 +src4 ));\
        OP(dst[ 1*dstStride], (src1 +src2 )*20 - (src0 +src3 )*6 + (src0 +src4 )*3 - (src1 +src5 ));\
        OP(dst[ 2*dstStride], (src2 +src3 )*20 - (src1 +src4 )*6 + (src0 +src5 )*3 - (src0 +src6 ));\
        OP(dst[ 3*dstStride], (src3 +src4 )*20 - (src2 +src5 )*6 + (src1 +src6 )*3 - (src0 +src7 ));\
        OP(dst[ 4*dstStride], (src4 +src5 )*20 - (src3 +src6 )*6 + (src2 +src7 )*3 - (src1 +src8 ));\
        OP(dst[ 5*dstStride], (src5 +src6 )*20 - (src4 +src7 )*6 + (src3 +src8 )*3 - (src2 +src9 ));\
        OP(dst[ 6*dstStride], (src6 +src7 )*20 - (src5 +src8 )*6 + (src4 +src9 )*3 - (src3 +src10));\
        OP(dst[ 7*dstStride], (src7 +src8 )*20 - (src6 +src9 )*6 + (src5 +src10)*3 - (src4 +src11));\
        OP(dst[ 8*dstStride], (src8 +src9 )*20 - (src7 +src10)*6 + (src6 +src11)*3 - (src5 +src12));\
        OP(dst[ 9*dstStride], (src9 +src10)*20 - (src8 +src11)*6 + (src7 +src12)*3 - (src6 +src13));\
        OP(dst[10*dstStride], (src10+src11)*20 - (src9 +src12)*6 + (src8 +src13)*3 - (src7 +src14));\
        OP(dst[11*dstStride], (src11+src12)*20 - (src10+src13)*6 + (src9 +src14)*3 - (src8 +src15));\
        OP(dst[12*dstStride], (src12+src13)*20 - (src11+src14)*6 + (src10+src15)*3 - (src9 +src16));\
        OP(dst[13*dstStride], (src13+src14)*20 - (src12+src15)*6 + (src11+src16)*3 - (src10+src16));\
        OP(dst[14*dstStride], (src14+src15)*20 - (src13+src16)*6 + (src12+src16)*3 - (src11+src15));\
        OP(dst[15*dstStride], (src15+src16)*20 - (src14+src16)*6 + (src13+src15)*3 - (src12+src14));\
        dst++;\
        src++;\
    }\
}\
\
1850
static void OPNAME ## qpel8_mc00_c (uint8_t *dst, uint8_t *src, int stride){\
1851
    OPNAME ## pixels8_c(dst, src, stride, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1852 1853
}\
\
1854 1855
static void OPNAME ## qpel8_mc10_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t half[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1856 1857
    put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
    OPNAME ## pixels8_l2(dst, src, half, stride, stride, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1858 1859
}\
\
1860
static void OPNAME ## qpel8_mc20_c(uint8_t *dst, uint8_t *src, int stride){\
Michael Niedermayer's avatar
Michael Niedermayer committed
1861
    OPNAME ## mpeg4_qpel8_h_lowpass(dst, src, stride, stride, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1862 1863
}\
\
1864 1865
static void OPNAME ## qpel8_mc30_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t half[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1866 1867
    put ## RND ## mpeg4_qpel8_h_lowpass(half, src, 8, stride, 8);\
    OPNAME ## pixels8_l2(dst, src+1, half, stride, stride, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1868 1869
}\
\
1870 1871 1872
static void OPNAME ## qpel8_mc01_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t half[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1873
    copy_block9(full, src, 16, stride, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1874
    put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1875
    OPNAME ## pixels8_l2(dst, full, half, stride, 16, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1876 1877
}\
\
1878 1879
static void OPNAME ## qpel8_mc02_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1880
    copy_block9(full, src, 16, stride, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1881
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, full, stride, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1882 1883
}\
\
1884 1885 1886
static void OPNAME ## qpel8_mc03_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t half[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1887
    copy_block9(full, src, 16, stride, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1888
    put ## RND ## mpeg4_qpel8_v_lowpass(half, full, 8, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1889
    OPNAME ## pixels8_l2(dst, full+16, half, stride, 16, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1890
}\
1891 1892 1893 1894 1895
void ff_ ## OPNAME ## qpel8_mc11_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
    uint8_t halfV[64];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1896 1897
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1898 1899
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1900
    OPNAME ## pixels8_l4(dst, full, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1901
}\
1902 1903 1904 1905
static void OPNAME ## qpel8_mc11_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1906 1907 1908 1909 1910 1911
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
    put ## RND ## pixels8_l2(halfH, halfH, full, 8, 8, 16, 9);\
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
    OPNAME ## pixels8_l2(dst, halfH, halfHV, stride, 8, 8, 8);\
}\
1912 1913 1914 1915 1916
void ff_ ## OPNAME ## qpel8_mc31_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
    uint8_t halfV[64];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1917 1918
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1919 1920
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1921
    OPNAME ## pixels8_l4(dst, full+1, halfH, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1922
}\
1923 1924 1925 1926
static void OPNAME ## qpel8_mc31_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1927 1928 1929 1930 1931 1932
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
    put ## RND ## pixels8_l2(halfH, halfH, full+1, 8, 8, 16, 9);\
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
    OPNAME ## pixels8_l2(dst, halfH, halfHV, stride, 8, 8, 8);\
}\
1933 1934 1935 1936 1937
void ff_ ## OPNAME ## qpel8_mc13_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
    uint8_t halfV[64];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1938 1939
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1940 1941
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1942
    OPNAME ## pixels8_l4(dst, full+16, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1943
}\
1944 1945 1946 1947
static void OPNAME ## qpel8_mc13_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1948 1949 1950 1951 1952 1953
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
    put ## RND ## pixels8_l2(halfH, halfH, full, 8, 8, 16, 9);\
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
    OPNAME ## pixels8_l2(dst, halfH+8, halfHV, stride, 8, 8, 8);\
}\
1954 1955 1956 1957 1958
void ff_ ## OPNAME ## qpel8_mc33_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
    uint8_t halfV[64];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1959 1960
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full  , 8, 16, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1961 1962
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1963
    OPNAME ## pixels8_l4(dst, full+17, halfH+8, halfV, halfHV, stride, 16, 8, 8, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1964
}\
1965 1966 1967 1968
static void OPNAME ## qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1969 1970 1971 1972 1973 1974
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
    put ## RND ## pixels8_l2(halfH, halfH, full+1, 8, 8, 16, 9);\
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
    OPNAME ## pixels8_l2(dst, halfH+8, halfHV, stride, 8, 8, 8);\
}\
1975 1976 1977
static void OPNAME ## qpel8_mc21_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t halfH[72];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1978
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1979
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1980
    OPNAME ## pixels8_l2(dst, halfH, halfHV, stride, 8, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1981
}\
1982 1983 1984
static void OPNAME ## qpel8_mc23_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t halfH[72];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1985
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1986
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1987
    OPNAME ## pixels8_l2(dst, halfH+8, halfHV, stride, 8, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1988
}\
1989 1990 1991 1992 1993
void ff_ ## OPNAME ## qpel8_mc12_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
    uint8_t halfV[64];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
1994 1995
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1996 1997
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full, 8, 16);\
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1998
    OPNAME ## pixels8_l2(dst, halfV, halfHV, stride, 8, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
1999
}\
2000 2001 2002
static void OPNAME ## qpel8_mc12_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2003 2004 2005 2006 2007
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
    put ## RND ## pixels8_l2(halfH, halfH, full, 8, 8, 16, 9);\
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
}\
2008 2009 2010 2011 2012
void ff_ ## OPNAME ## qpel8_mc32_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
    uint8_t halfV[64];\
    uint8_t halfHV[64];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2013 2014
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2015 2016
    put ## RND ## mpeg4_qpel8_v_lowpass(halfV, full+1, 8, 16);\
    put ## RND ## mpeg4_qpel8_v_lowpass(halfHV, halfH, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2017
    OPNAME ## pixels8_l2(dst, halfV, halfHV, stride, 8, 8, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2018
}\
2019 2020 2021
static void OPNAME ## qpel8_mc32_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[16*9];\
    uint8_t halfH[72];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2022 2023 2024 2025 2026
    copy_block9(full, src, 16, stride, 9);\
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, full, 8, 16, 9);\
    put ## RND ## pixels8_l2(halfH, halfH, full+1, 8, 8, 16, 9);\
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
}\
2027 2028
static void OPNAME ## qpel8_mc22_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t halfH[72];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2029
    put ## RND ## mpeg4_qpel8_h_lowpass(halfH, src, 8, stride, 9);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2030
    OPNAME ## mpeg4_qpel8_v_lowpass(dst, halfH, stride, 8);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2031
}\
2032
static void OPNAME ## qpel16_mc00_c (uint8_t *dst, uint8_t *src, int stride){\
2033
    OPNAME ## pixels16_c(dst, src, stride, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2034 2035
}\
\
2036 2037
static void OPNAME ## qpel16_mc10_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t half[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2038 2039 2040 2041
    put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
    OPNAME ## pixels16_l2(dst, src, half, stride, stride, 16, 16);\
}\
\
2042
static void OPNAME ## qpel16_mc20_c(uint8_t *dst, uint8_t *src, int stride){\
Michael Niedermayer's avatar
Michael Niedermayer committed
2043
    OPNAME ## mpeg4_qpel16_h_lowpass(dst, src, stride, stride, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2044
}\
Michael Niedermayer's avatar
Michael Niedermayer committed
2045
\
2046 2047
static void OPNAME ## qpel16_mc30_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t half[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2048 2049 2050 2051
    put ## RND ## mpeg4_qpel16_h_lowpass(half, src, 16, stride, 16);\
    OPNAME ## pixels16_l2(dst, src+1, half, stride, stride, 16, 16);\
}\
\
2052 2053 2054
static void OPNAME ## qpel16_mc01_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t half[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2055
    copy_block17(full, src, 24, stride, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2056
    put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2057 2058 2059
    OPNAME ## pixels16_l2(dst, full, half, stride, 24, 16, 16);\
}\
\
2060 2061
static void OPNAME ## qpel16_mc02_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2062
    copy_block17(full, src, 24, stride, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2063
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, full, stride, 24);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2064 2065
}\
\
2066 2067 2068
static void OPNAME ## qpel16_mc03_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t half[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2069
    copy_block17(full, src, 24, stride, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2070
    put ## RND ## mpeg4_qpel16_v_lowpass(half, full, 16, 24);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2071 2072
    OPNAME ## pixels16_l2(dst, full+24, half, stride, 24, 16, 16);\
}\
2073 2074 2075 2076 2077
void ff_ ## OPNAME ## qpel16_mc11_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
    uint8_t halfV[256];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2078 2079
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2080 2081
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2082 2083
    OPNAME ## pixels16_l4(dst, full, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
}\
2084 2085 2086 2087
static void OPNAME ## qpel16_mc11_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2088 2089 2090 2091 2092 2093
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
    put ## RND ## pixels16_l2(halfH, halfH, full, 16, 16, 24, 17);\
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
    OPNAME ## pixels16_l2(dst, halfH, halfHV, stride, 16, 16, 16);\
}\
2094 2095 2096 2097 2098
void ff_ ## OPNAME ## qpel16_mc31_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
    uint8_t halfV[256];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2099 2100
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2101 2102
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2103 2104
    OPNAME ## pixels16_l4(dst, full+1, halfH, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
}\
2105 2106 2107 2108
static void OPNAME ## qpel16_mc31_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2109 2110 2111 2112 2113 2114
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
    put ## RND ## pixels16_l2(halfH, halfH, full+1, 16, 16, 24, 17);\
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
    OPNAME ## pixels16_l2(dst, halfH, halfHV, stride, 16, 16, 16);\
}\
2115 2116 2117 2118 2119
void ff_ ## OPNAME ## qpel16_mc13_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
    uint8_t halfV[256];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2120 2121
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2122 2123
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2124 2125
    OPNAME ## pixels16_l4(dst, full+24, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
}\
2126 2127 2128 2129
static void OPNAME ## qpel16_mc13_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2130 2131 2132 2133 2134 2135
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
    put ## RND ## pixels16_l2(halfH, halfH, full, 16, 16, 24, 17);\
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
    OPNAME ## pixels16_l2(dst, halfH+16, halfHV, stride, 16, 16, 16);\
}\
2136 2137 2138 2139 2140
void ff_ ## OPNAME ## qpel16_mc33_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
    uint8_t halfV[256];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2141 2142
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full  , 16, 24, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2143 2144
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2145 2146
    OPNAME ## pixels16_l4(dst, full+25, halfH+16, halfV, halfHV, stride, 24, 16, 16, 16, 16);\
}\
2147 2148 2149 2150
static void OPNAME ## qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2151 2152 2153 2154 2155 2156
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
    put ## RND ## pixels16_l2(halfH, halfH, full+1, 16, 16, 24, 17);\
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
    OPNAME ## pixels16_l2(dst, halfH+16, halfHV, stride, 16, 16, 16);\
}\
2157 2158 2159
static void OPNAME ## qpel16_mc21_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t halfH[272];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2160
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2161
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2162 2163
    OPNAME ## pixels16_l2(dst, halfH, halfHV, stride, 16, 16, 16);\
}\
2164 2165 2166
static void OPNAME ## qpel16_mc23_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t halfH[272];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2167
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2168
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2169 2170
    OPNAME ## pixels16_l2(dst, halfH+16, halfHV, stride, 16, 16, 16);\
}\
2171 2172 2173 2174 2175
void ff_ ## OPNAME ## qpel16_mc12_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
    uint8_t halfV[256];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2176 2177
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2178 2179
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full, 16, 24);\
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2180 2181
    OPNAME ## pixels16_l2(dst, halfV, halfHV, stride, 16, 16, 16);\
}\
2182 2183 2184
static void OPNAME ## qpel16_mc12_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2185 2186 2187 2188 2189
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
    put ## RND ## pixels16_l2(halfH, halfH, full, 16, 16, 24, 17);\
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
}\
2190 2191 2192 2193 2194
void ff_ ## OPNAME ## qpel16_mc32_old_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
    uint8_t halfV[256];\
    uint8_t halfHV[256];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2195 2196
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2197 2198
    put ## RND ## mpeg4_qpel16_v_lowpass(halfV, full+1, 16, 24);\
    put ## RND ## mpeg4_qpel16_v_lowpass(halfHV, halfH, 16, 16);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2199 2200
    OPNAME ## pixels16_l2(dst, halfV, halfHV, stride, 16, 16, 16);\
}\
2201 2202 2203
static void OPNAME ## qpel16_mc32_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[24*17];\
    uint8_t halfH[272];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2204 2205 2206 2207 2208
    copy_block17(full, src, 24, stride, 17);\
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, full, 16, 24, 17);\
    put ## RND ## pixels16_l2(halfH, halfH, full+1, 16, 16, 24, 17);\
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
}\
2209 2210
static void OPNAME ## qpel16_mc22_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t halfH[272];\
Michael Niedermayer's avatar
Michael Niedermayer committed
2211
    put ## RND ## mpeg4_qpel16_h_lowpass(halfH, src, 16, stride, 17);\
Michael Niedermayer's avatar
Michael Niedermayer committed
2212
    OPNAME ## mpeg4_qpel16_v_lowpass(dst, halfH, stride, 16);\
2213
}
Michael Niedermayer's avatar
Michael Niedermayer committed
2214

Michael Niedermayer's avatar
Michael Niedermayer committed
2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227
#define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1)
#define op_avg_no_rnd(a, b) a = (((a)+cm[((b) + 15)>>5])>>1)
#define op_put(a, b) a = cm[((b) + 16)>>5]
#define op_put_no_rnd(a, b) a = cm[((b) + 15)>>5]

QPEL_MC(0, put_       , _       , op_put)
QPEL_MC(1, put_no_rnd_, _no_rnd_, op_put_no_rnd)
QPEL_MC(0, avg_       , _       , op_avg)
//QPEL_MC(1, avg_no_rnd , _       , op_avg)
#undef op_avg
#undef op_avg_no_rnd
#undef op_put
#undef op_put_no_rnd
Michael Niedermayer's avatar
Michael Niedermayer committed
2228

2229 2230
#if 1
#define H264_LOWPASS(OPNAME, OP, OP2) \
2231
static av_unused void OPNAME ## h264_qpel2_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
2232
    const int h=2;\
2233
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
2234 2235 2236 2237 2238 2239 2240 2241 2242 2243
    int i;\
    for(i=0; i<h; i++)\
    {\
        OP(dst[0], (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]));\
        OP(dst[1], (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]));\
        dst+=dstStride;\
        src+=srcStride;\
    }\
}\
\
2244
static av_unused void OPNAME ## h264_qpel2_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
2245
    const int w=2;\
2246
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263
    int i;\
    for(i=0; i<w; i++)\
    {\
        const int srcB= src[-2*srcStride];\
        const int srcA= src[-1*srcStride];\
        const int src0= src[0 *srcStride];\
        const int src1= src[1 *srcStride];\
        const int src2= src[2 *srcStride];\
        const int src3= src[3 *srcStride];\
        const int src4= src[4 *srcStride];\
        OP(dst[0*dstStride], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\
        OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\
        dst++;\
        src++;\
    }\
}\
\
2264
static av_unused void OPNAME ## h264_qpel2_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
2265 2266
    const int h=2;\
    const int w=2;\
2267
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292
    int i;\
    src -= 2*srcStride;\
    for(i=0; i<h+5; i++)\
    {\
        tmp[0]= (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]);\
        tmp[1]= (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]);\
        tmp+=tmpStride;\
        src+=srcStride;\
    }\
    tmp -= tmpStride*(h+5-2);\
    for(i=0; i<w; i++)\
    {\
        const int tmpB= tmp[-2*tmpStride];\
        const int tmpA= tmp[-1*tmpStride];\
        const int tmp0= tmp[0 *tmpStride];\
        const int tmp1= tmp[1 *tmpStride];\
        const int tmp2= tmp[2 *tmpStride];\
        const int tmp3= tmp[3 *tmpStride];\
        const int tmp4= tmp[4 *tmpStride];\
        OP2(dst[0*dstStride], (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));\
        OP2(dst[1*dstStride], (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));\
        dst++;\
        tmp++;\
    }\
}\
2293 2294
static void OPNAME ## h264_qpel4_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    const int h=4;\
2295
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309
    int i;\
    for(i=0; i<h; i++)\
    {\
        OP(dst[0], (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]));\
        OP(dst[1], (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]));\
        OP(dst[2], (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5]));\
        OP(dst[3], (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6]));\
        dst+=dstStride;\
        src+=srcStride;\
    }\
}\
\
static void OPNAME ## h264_qpel4_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    const int w=4;\
2310
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334
    int i;\
    for(i=0; i<w; i++)\
    {\
        const int srcB= src[-2*srcStride];\
        const int srcA= src[-1*srcStride];\
        const int src0= src[0 *srcStride];\
        const int src1= src[1 *srcStride];\
        const int src2= src[2 *srcStride];\
        const int src3= src[3 *srcStride];\
        const int src4= src[4 *srcStride];\
        const int src5= src[5 *srcStride];\
        const int src6= src[6 *srcStride];\
        OP(dst[0*dstStride], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\
        OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\
        OP(dst[2*dstStride], (src2+src3)*20 - (src1+src4)*5 + (src0+src5));\
        OP(dst[3*dstStride], (src3+src4)*20 - (src2+src5)*5 + (src1+src6));\
        dst++;\
        src++;\
    }\
}\
\
static void OPNAME ## h264_qpel4_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
    const int h=4;\
    const int w=4;\
2335
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369
    int i;\
    src -= 2*srcStride;\
    for(i=0; i<h+5; i++)\
    {\
        tmp[0]= (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]);\
        tmp[1]= (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]);\
        tmp[2]= (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5]);\
        tmp[3]= (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6]);\
        tmp+=tmpStride;\
        src+=srcStride;\
    }\
    tmp -= tmpStride*(h+5-2);\
    for(i=0; i<w; i++)\
    {\
        const int tmpB= tmp[-2*tmpStride];\
        const int tmpA= tmp[-1*tmpStride];\
        const int tmp0= tmp[0 *tmpStride];\
        const int tmp1= tmp[1 *tmpStride];\
        const int tmp2= tmp[2 *tmpStride];\
        const int tmp3= tmp[3 *tmpStride];\
        const int tmp4= tmp[4 *tmpStride];\
        const int tmp5= tmp[5 *tmpStride];\
        const int tmp6= tmp[6 *tmpStride];\
        OP2(dst[0*dstStride], (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));\
        OP2(dst[1*dstStride], (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));\
        OP2(dst[2*dstStride], (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));\
        OP2(dst[3*dstStride], (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));\
        dst++;\
        tmp++;\
    }\
}\
\
static void OPNAME ## h264_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    const int h=8;\
2370
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388
    int i;\
    for(i=0; i<h; i++)\
    {\
        OP(dst[0], (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3 ]));\
        OP(dst[1], (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4 ]));\
        OP(dst[2], (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5 ]));\
        OP(dst[3], (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6 ]));\
        OP(dst[4], (src[4]+src[5])*20 - (src[3 ]+src[6])*5 + (src[2 ]+src[7 ]));\
        OP(dst[5], (src[5]+src[6])*20 - (src[4 ]+src[7])*5 + (src[3 ]+src[8 ]));\
        OP(dst[6], (src[6]+src[7])*20 - (src[5 ]+src[8])*5 + (src[4 ]+src[9 ]));\
        OP(dst[7], (src[7]+src[8])*20 - (src[6 ]+src[9])*5 + (src[5 ]+src[10]));\
        dst+=dstStride;\
        src+=srcStride;\
    }\
}\
\
static void OPNAME ## h264_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    const int w=8;\
2389
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421
    int i;\
    for(i=0; i<w; i++)\
    {\
        const int srcB= src[-2*srcStride];\
        const int srcA= src[-1*srcStride];\
        const int src0= src[0 *srcStride];\
        const int src1= src[1 *srcStride];\
        const int src2= src[2 *srcStride];\
        const int src3= src[3 *srcStride];\
        const int src4= src[4 *srcStride];\
        const int src5= src[5 *srcStride];\
        const int src6= src[6 *srcStride];\
        const int src7= src[7 *srcStride];\
        const int src8= src[8 *srcStride];\
        const int src9= src[9 *srcStride];\
        const int src10=src[10*srcStride];\
        OP(dst[0*dstStride], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\
        OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\
        OP(dst[2*dstStride], (src2+src3)*20 - (src1+src4)*5 + (src0+src5));\
        OP(dst[3*dstStride], (src3+src4)*20 - (src2+src5)*5 + (src1+src6));\
        OP(dst[4*dstStride], (src4+src5)*20 - (src3+src6)*5 + (src2+src7));\
        OP(dst[5*dstStride], (src5+src6)*20 - (src4+src7)*5 + (src3+src8));\
        OP(dst[6*dstStride], (src6+src7)*20 - (src5+src8)*5 + (src4+src9));\
        OP(dst[7*dstStride], (src7+src8)*20 - (src6+src9)*5 + (src5+src10));\
        dst++;\
        src++;\
    }\
}\
\
static void OPNAME ## h264_qpel8_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
    const int h=8;\
    const int w=8;\
2422
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638
    int i;\
    src -= 2*srcStride;\
    for(i=0; i<h+5; i++)\
    {\
        tmp[0]= (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3 ]);\
        tmp[1]= (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4 ]);\
        tmp[2]= (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5 ]);\
        tmp[3]= (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6 ]);\
        tmp[4]= (src[4]+src[5])*20 - (src[3 ]+src[6])*5 + (src[2 ]+src[7 ]);\
        tmp[5]= (src[5]+src[6])*20 - (src[4 ]+src[7])*5 + (src[3 ]+src[8 ]);\
        tmp[6]= (src[6]+src[7])*20 - (src[5 ]+src[8])*5 + (src[4 ]+src[9 ]);\
        tmp[7]= (src[7]+src[8])*20 - (src[6 ]+src[9])*5 + (src[5 ]+src[10]);\
        tmp+=tmpStride;\
        src+=srcStride;\
    }\
    tmp -= tmpStride*(h+5-2);\
    for(i=0; i<w; i++)\
    {\
        const int tmpB= tmp[-2*tmpStride];\
        const int tmpA= tmp[-1*tmpStride];\
        const int tmp0= tmp[0 *tmpStride];\
        const int tmp1= tmp[1 *tmpStride];\
        const int tmp2= tmp[2 *tmpStride];\
        const int tmp3= tmp[3 *tmpStride];\
        const int tmp4= tmp[4 *tmpStride];\
        const int tmp5= tmp[5 *tmpStride];\
        const int tmp6= tmp[6 *tmpStride];\
        const int tmp7= tmp[7 *tmpStride];\
        const int tmp8= tmp[8 *tmpStride];\
        const int tmp9= tmp[9 *tmpStride];\
        const int tmp10=tmp[10*tmpStride];\
        OP2(dst[0*dstStride], (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));\
        OP2(dst[1*dstStride], (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));\
        OP2(dst[2*dstStride], (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));\
        OP2(dst[3*dstStride], (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));\
        OP2(dst[4*dstStride], (tmp4+tmp5)*20 - (tmp3+tmp6)*5 + (tmp2+tmp7));\
        OP2(dst[5*dstStride], (tmp5+tmp6)*20 - (tmp4+tmp7)*5 + (tmp3+tmp8));\
        OP2(dst[6*dstStride], (tmp6+tmp7)*20 - (tmp5+tmp8)*5 + (tmp4+tmp9));\
        OP2(dst[7*dstStride], (tmp7+tmp8)*20 - (tmp6+tmp9)*5 + (tmp5+tmp10));\
        dst++;\
        tmp++;\
    }\
}\
\
static void OPNAME ## h264_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    OPNAME ## h264_qpel8_v_lowpass(dst  , src  , dstStride, srcStride);\
    OPNAME ## h264_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride);\
    src += 8*srcStride;\
    dst += 8*dstStride;\
    OPNAME ## h264_qpel8_v_lowpass(dst  , src  , dstStride, srcStride);\
    OPNAME ## h264_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride);\
}\
\
static void OPNAME ## h264_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
    OPNAME ## h264_qpel8_h_lowpass(dst  , src  , dstStride, srcStride);\
    OPNAME ## h264_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride);\
    src += 8*srcStride;\
    dst += 8*dstStride;\
    OPNAME ## h264_qpel8_h_lowpass(dst  , src  , dstStride, srcStride);\
    OPNAME ## h264_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride);\
}\
\
static void OPNAME ## h264_qpel16_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\
    OPNAME ## h264_qpel8_hv_lowpass(dst  , tmp  , src  , dstStride, tmpStride, srcStride);\
    OPNAME ## h264_qpel8_hv_lowpass(dst+8, tmp+8, src+8, dstStride, tmpStride, srcStride);\
    src += 8*srcStride;\
    dst += 8*dstStride;\
    OPNAME ## h264_qpel8_hv_lowpass(dst  , tmp  , src  , dstStride, tmpStride, srcStride);\
    OPNAME ## h264_qpel8_hv_lowpass(dst+8, tmp+8, src+8, dstStride, tmpStride, srcStride);\
}\

#define H264_MC(OPNAME, SIZE) \
static void OPNAME ## h264_qpel ## SIZE ## _mc00_c (uint8_t *dst, uint8_t *src, int stride){\
    OPNAME ## pixels ## SIZE ## _c(dst, src, stride, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t half[SIZE*SIZE];\
    put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2(dst, src, half, stride, stride, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\
    OPNAME ## h264_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t half[SIZE*SIZE];\
    put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2(dst, src+1, half, stride, stride, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[SIZE*(SIZE+5)];\
    uint8_t * const full_mid= full + SIZE*2;\
    uint8_t half[SIZE*SIZE];\
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
    put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\
    OPNAME ## pixels ## SIZE ## _l2(dst, full_mid, half, stride, SIZE, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[SIZE*(SIZE+5)];\
    uint8_t * const full_mid= full + SIZE*2;\
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
    OPNAME ## h264_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[SIZE*(SIZE+5)];\
    uint8_t * const full_mid= full + SIZE*2;\
    uint8_t half[SIZE*SIZE];\
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
    put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\
    OPNAME ## pixels ## SIZE ## _l2(dst, full_mid+SIZE, half, stride, SIZE, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[SIZE*(SIZE+5)];\
    uint8_t * const full_mid= full + SIZE*2;\
    uint8_t halfH[SIZE*SIZE];\
    uint8_t halfV[SIZE*SIZE];\
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
    OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[SIZE*(SIZE+5)];\
    uint8_t * const full_mid= full + SIZE*2;\
    uint8_t halfH[SIZE*SIZE];\
    uint8_t halfV[SIZE*SIZE];\
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\
    copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
    OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[SIZE*(SIZE+5)];\
    uint8_t * const full_mid= full + SIZE*2;\
    uint8_t halfH[SIZE*SIZE];\
    uint8_t halfV[SIZE*SIZE];\
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
    OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc33_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[SIZE*(SIZE+5)];\
    uint8_t * const full_mid= full + SIZE*2;\
    uint8_t halfH[SIZE*SIZE];\
    uint8_t halfV[SIZE*SIZE];\
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\
    copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
    OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
    int16_t tmp[SIZE*(SIZE+5)];\
    OPNAME ## h264_qpel ## SIZE ## _hv_lowpass(dst, tmp, src, stride, SIZE, stride);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
    int16_t tmp[SIZE*(SIZE+5)];\
    uint8_t halfH[SIZE*SIZE];\
    uint8_t halfHV[SIZE*SIZE];\
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\
    put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
    int16_t tmp[SIZE*(SIZE+5)];\
    uint8_t halfH[SIZE*SIZE];\
    uint8_t halfHV[SIZE*SIZE];\
    put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\
    put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[SIZE*(SIZE+5)];\
    uint8_t * const full_mid= full + SIZE*2;\
    int16_t tmp[SIZE*(SIZE+5)];\
    uint8_t halfV[SIZE*SIZE];\
    uint8_t halfHV[SIZE*SIZE];\
    copy_block ## SIZE (full, src - stride*2, SIZE,  stride, SIZE + 5);\
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
    put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\
}\
\
static void OPNAME ## h264_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
    uint8_t full[SIZE*(SIZE+5)];\
    uint8_t * const full_mid= full + SIZE*2;\
    int16_t tmp[SIZE*(SIZE+5)];\
    uint8_t halfV[SIZE*SIZE];\
    uint8_t halfHV[SIZE*SIZE];\
    copy_block ## SIZE (full, src - stride*2 + 1, SIZE,  stride, SIZE + 5);\
    put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\
    put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\
    OPNAME ## pixels ## SIZE ## _l2(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\
}\

#define op_avg(a, b)  a = (((a)+cm[((b) + 16)>>5]+1)>>1)
//#define op_avg2(a, b) a = (((a)*w1+cm[((b) + 16)>>5]*w2 + o + 64)>>7)
#define op_put(a, b)  a = cm[((b) + 16)>>5]
#define op2_avg(a, b)  a = (((a)+cm[((b) + 512)>>10]+1)>>1)
#define op2_put(a, b)  a = cm[((b) + 512)>>10]

H264_LOWPASS(put_       , op_put, op2_put)
H264_LOWPASS(avg_       , op_avg, op2_avg)
2639
H264_MC(put_, 2)
2640 2641 2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652
H264_MC(put_, 4)
H264_MC(put_, 8)
H264_MC(put_, 16)
H264_MC(avg_, 4)
H264_MC(avg_, 8)
H264_MC(avg_, 16)

#undef op_avg
#undef op_put
#undef op2_avg
#undef op2_put
#endif

2653 2654
#define op_scale1(x)  block[x] = av_clip_uint8( (block[x]*weight + offset) >> log2_denom )
#define op_scale2(x)  dst[x] = av_clip_uint8( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1))
2655 2656
#define H264_WEIGHT(W,H) \
static void weight_h264_pixels ## W ## x ## H ## _c(uint8_t *block, int stride, int log2_denom, int weight, int offset){ \
Loren Merritt's avatar
Loren Merritt committed
2657
    int y; \
2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681
    offset <<= log2_denom; \
    if(log2_denom) offset += 1<<(log2_denom-1); \
    for(y=0; y<H; y++, block += stride){ \
        op_scale1(0); \
        op_scale1(1); \
        if(W==2) continue; \
        op_scale1(2); \
        op_scale1(3); \
        if(W==4) continue; \
        op_scale1(4); \
        op_scale1(5); \
        op_scale1(6); \
        op_scale1(7); \
        if(W==8) continue; \
        op_scale1(8); \
        op_scale1(9); \
        op_scale1(10); \
        op_scale1(11); \
        op_scale1(12); \
        op_scale1(13); \
        op_scale1(14); \
        op_scale1(15); \
    } \
} \
Loren Merritt's avatar
Loren Merritt committed
2682 2683 2684
static void biweight_h264_pixels ## W ## x ## H ## _c(uint8_t *dst, uint8_t *src, int stride, int log2_denom, int weightd, int weights, int offset){ \
    int y; \
    offset = ((offset + 1) | 1) << log2_denom; \
2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722
    for(y=0; y<H; y++, dst += stride, src += stride){ \
        op_scale2(0); \
        op_scale2(1); \
        if(W==2) continue; \
        op_scale2(2); \
        op_scale2(3); \
        if(W==4) continue; \
        op_scale2(4); \
        op_scale2(5); \
        op_scale2(6); \
        op_scale2(7); \
        if(W==8) continue; \
        op_scale2(8); \
        op_scale2(9); \
        op_scale2(10); \
        op_scale2(11); \
        op_scale2(12); \
        op_scale2(13); \
        op_scale2(14); \
        op_scale2(15); \
    } \
}

H264_WEIGHT(16,16)
H264_WEIGHT(16,8)
H264_WEIGHT(8,16)
H264_WEIGHT(8,8)
H264_WEIGHT(8,4)
H264_WEIGHT(4,8)
H264_WEIGHT(4,4)
H264_WEIGHT(4,2)
H264_WEIGHT(2,4)
H264_WEIGHT(2,2)

#undef op_scale1
#undef op_scale2
#undef H264_WEIGHT

Michael Niedermayer's avatar
Michael Niedermayer committed
2723
static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){
2724
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
Michael Niedermayer's avatar
Michael Niedermayer committed
2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736
    int i;

    for(i=0; i<h; i++){
        dst[0]= cm[(9*(src[0] + src[1]) - (src[-1] + src[2]) + 8)>>4];
        dst[1]= cm[(9*(src[1] + src[2]) - (src[ 0] + src[3]) + 8)>>4];
        dst[2]= cm[(9*(src[2] + src[3]) - (src[ 1] + src[4]) + 8)>>4];
        dst[3]= cm[(9*(src[3] + src[4]) - (src[ 2] + src[5]) + 8)>>4];
        dst[4]= cm[(9*(src[4] + src[5]) - (src[ 3] + src[6]) + 8)>>4];
        dst[5]= cm[(9*(src[5] + src[6]) - (src[ 4] + src[7]) + 8)>>4];
        dst[6]= cm[(9*(src[6] + src[7]) - (src[ 5] + src[8]) + 8)>>4];
        dst[7]= cm[(9*(src[7] + src[8]) - (src[ 6] + src[9]) + 8)>>4];
        dst+=dstStride;
2737
        src+=srcStride;
Michael Niedermayer's avatar
Michael Niedermayer committed
2738 2739 2740
    }
}

2741
#if CONFIG_CAVS_DECODER
2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756
/* AVS specific */
void ff_cavsdsp_init(DSPContext* c, AVCodecContext *avctx);

void ff_put_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride) {
    put_pixels8_c(dst, src, stride, 8);
}
void ff_avg_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride) {
    avg_pixels8_c(dst, src, stride, 8);
}
void ff_put_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride) {
    put_pixels16_c(dst, src, stride, 16);
}
void ff_avg_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride) {
    avg_pixels16_c(dst, src, stride, 16);
}
2757
#endif /* CONFIG_CAVS_DECODER */
2758

2759
void ff_mlp_init(DSPContext* c, AVCodecContext *avctx);
2760

2761
#if CONFIG_VC1_DECODER
2762 2763 2764 2765
/* VC-1 specific */
void ff_vc1dsp_init(DSPContext* c, AVCodecContext *avctx);

void ff_put_vc1_mspel_mc00_c(uint8_t *dst, uint8_t *src, int stride, int rnd) {
2766 2767
    put_pixels8_c(dst, src, stride, 8);
}
2768 2769 2770
void ff_avg_vc1_mspel_mc00_c(uint8_t *dst, uint8_t *src, int stride, int rnd) {
    avg_pixels8_c(dst, src, stride, 8);
}
2771
#endif /* CONFIG_VC1_DECODER */
2772

2773 2774
void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx);

Panagiotis Issaris's avatar
Panagiotis Issaris committed
2775
/* H264 specific */
2776
void ff_h264dspenc_init(DSPContext* c, AVCodecContext *avctx);
Panagiotis Issaris's avatar
Panagiotis Issaris committed
2777

2778
#if CONFIG_RV30_DECODER
Kostya Shishkov's avatar
Kostya Shishkov committed
2779 2780 2781
void ff_rv30dsp_init(DSPContext* c, AVCodecContext *avctx);
#endif /* CONFIG_RV30_DECODER */

2782
#if CONFIG_RV40_DECODER
Kostya Shishkov's avatar
Kostya Shishkov committed
2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798
static void put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride){
    put_pixels16_xy2_c(dst, src, stride, 16);
}
static void avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride){
    avg_pixels16_xy2_c(dst, src, stride, 16);
}
static void put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride){
    put_pixels8_xy2_c(dst, src, stride, 8);
}
static void avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride){
    avg_pixels8_xy2_c(dst, src, stride, 8);
}

void ff_rv40dsp_init(DSPContext* c, AVCodecContext *avctx);
#endif /* CONFIG_RV40_DECODER */

Michael Niedermayer's avatar
Michael Niedermayer committed
2799
static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){
2800
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
Michael Niedermayer's avatar
Michael Niedermayer committed
2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875
    int i;

    for(i=0; i<w; i++){
        const int src_1= src[ -srcStride];
        const int src0 = src[0          ];
        const int src1 = src[  srcStride];
        const int src2 = src[2*srcStride];
        const int src3 = src[3*srcStride];
        const int src4 = src[4*srcStride];
        const int src5 = src[5*srcStride];
        const int src6 = src[6*srcStride];
        const int src7 = src[7*srcStride];
        const int src8 = src[8*srcStride];
        const int src9 = src[9*srcStride];
        dst[0*dstStride]= cm[(9*(src0 + src1) - (src_1 + src2) + 8)>>4];
        dst[1*dstStride]= cm[(9*(src1 + src2) - (src0  + src3) + 8)>>4];
        dst[2*dstStride]= cm[(9*(src2 + src3) - (src1  + src4) + 8)>>4];
        dst[3*dstStride]= cm[(9*(src3 + src4) - (src2  + src5) + 8)>>4];
        dst[4*dstStride]= cm[(9*(src4 + src5) - (src3  + src6) + 8)>>4];
        dst[5*dstStride]= cm[(9*(src5 + src6) - (src4  + src7) + 8)>>4];
        dst[6*dstStride]= cm[(9*(src6 + src7) - (src5  + src8) + 8)>>4];
        dst[7*dstStride]= cm[(9*(src7 + src8) - (src6  + src9) + 8)>>4];
        src++;
        dst++;
    }
}

static void put_mspel8_mc00_c (uint8_t *dst, uint8_t *src, int stride){
    put_pixels8_c(dst, src, stride, 8);
}

static void put_mspel8_mc10_c(uint8_t *dst, uint8_t *src, int stride){
    uint8_t half[64];
    wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
    put_pixels8_l2(dst, src, half, stride, stride, 8, 8);
}

static void put_mspel8_mc20_c(uint8_t *dst, uint8_t *src, int stride){
    wmv2_mspel8_h_lowpass(dst, src, stride, stride, 8);
}

static void put_mspel8_mc30_c(uint8_t *dst, uint8_t *src, int stride){
    uint8_t half[64];
    wmv2_mspel8_h_lowpass(half, src, 8, stride, 8);
    put_pixels8_l2(dst, src+1, half, stride, stride, 8, 8);
}

static void put_mspel8_mc02_c(uint8_t *dst, uint8_t *src, int stride){
    wmv2_mspel8_v_lowpass(dst, src, stride, stride, 8);
}

static void put_mspel8_mc12_c(uint8_t *dst, uint8_t *src, int stride){
    uint8_t halfH[88];
    uint8_t halfV[64];
    uint8_t halfHV[64];
    wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
    wmv2_mspel8_v_lowpass(halfV, src, 8, stride, 8);
    wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
    put_pixels8_l2(dst, halfV, halfHV, stride, 8, 8, 8);
}
static void put_mspel8_mc32_c(uint8_t *dst, uint8_t *src, int stride){
    uint8_t halfH[88];
    uint8_t halfV[64];
    uint8_t halfHV[64];
    wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
    wmv2_mspel8_v_lowpass(halfV, src+1, 8, stride, 8);
    wmv2_mspel8_v_lowpass(halfHV, halfH+8, 8, 8, 8);
    put_pixels8_l2(dst, halfV, halfHV, stride, 8, 8, 8);
}
static void put_mspel8_mc22_c(uint8_t *dst, uint8_t *src, int stride){
    uint8_t halfH[88];
    wmv2_mspel8_h_lowpass(halfH, src-stride, 8, stride, 11);
    wmv2_mspel8_v_lowpass(dst, halfH+8, stride, 8, 8);
}

Michael Niedermayer's avatar
Michael Niedermayer committed
2876
static void h263_v_loop_filter_c(uint8_t *src, int stride, int qscale){
2877
    if(CONFIG_ANY_H263) {
Michael Niedermayer's avatar
Michael Niedermayer committed
2878 2879
    int x;
    const int strength= ff_h263_loop_filter_strength[qscale];
2880

Michael Niedermayer's avatar
Michael Niedermayer committed
2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893
    for(x=0; x<8; x++){
        int d1, d2, ad1;
        int p0= src[x-2*stride];
        int p1= src[x-1*stride];
        int p2= src[x+0*stride];
        int p3= src[x+1*stride];
        int d = (p0 - p3 + 4*(p2 - p1)) / 8;

        if     (d<-2*strength) d1= 0;
        else if(d<-  strength) d1=-2*strength - d;
        else if(d<   strength) d1= d;
        else if(d< 2*strength) d1= 2*strength - d;
        else                   d1= 0;
2894

Michael Niedermayer's avatar
Michael Niedermayer committed
2895 2896 2897 2898
        p1 += d1;
        p2 -= d1;
        if(p1&256) p1= ~(p1>>31);
        if(p2&256) p2= ~(p2>>31);
2899

Michael Niedermayer's avatar
Michael Niedermayer committed
2900 2901 2902
        src[x-1*stride] = p1;
        src[x+0*stride] = p2;

2903
        ad1= FFABS(d1)>>1;
2904

2905
        d2= av_clip((p0-p3)/4, -ad1, ad1);
2906

Michael Niedermayer's avatar
Michael Niedermayer committed
2907 2908 2909
        src[x-2*stride] = p0 - d2;
        src[x+  stride] = p3 + d2;
    }
2910
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
2911 2912 2913
}

static void h263_h_loop_filter_c(uint8_t *src, int stride, int qscale){
2914
    if(CONFIG_ANY_H263) {
Michael Niedermayer's avatar
Michael Niedermayer committed
2915 2916
    int y;
    const int strength= ff_h263_loop_filter_strength[qscale];
2917

Michael Niedermayer's avatar
Michael Niedermayer committed
2918 2919 2920 2921 2922 2923 2924 2925 2926 2927 2928 2929 2930
    for(y=0; y<8; y++){
        int d1, d2, ad1;
        int p0= src[y*stride-2];
        int p1= src[y*stride-1];
        int p2= src[y*stride+0];
        int p3= src[y*stride+1];
        int d = (p0 - p3 + 4*(p2 - p1)) / 8;

        if     (d<-2*strength) d1= 0;
        else if(d<-  strength) d1=-2*strength - d;
        else if(d<   strength) d1= d;
        else if(d< 2*strength) d1= 2*strength - d;
        else                   d1= 0;
2931

Michael Niedermayer's avatar
Michael Niedermayer committed
2932 2933 2934 2935
        p1 += d1;
        p2 -= d1;
        if(p1&256) p1= ~(p1>>31);
        if(p2&256) p2= ~(p2>>31);
2936

Michael Niedermayer's avatar
Michael Niedermayer committed
2937 2938 2939
        src[y*stride-1] = p1;
        src[y*stride+0] = p2;

2940
        ad1= FFABS(d1)>>1;
2941

2942
        d2= av_clip((p0-p3)/4, -ad1, ad1);
2943

Michael Niedermayer's avatar
Michael Niedermayer committed
2944 2945 2946
        src[y*stride-2] = p0 - d2;
        src[y*stride+1] = p3 + d2;
    }
2947
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
2948
}
Michael Niedermayer's avatar
Michael Niedermayer committed
2949

2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962
static void h261_loop_filter_c(uint8_t *src, int stride){
    int x,y,xy,yz;
    int temp[64];

    for(x=0; x<8; x++){
        temp[x      ] = 4*src[x           ];
        temp[x + 7*8] = 4*src[x + 7*stride];
    }
    for(y=1; y<7; y++){
        for(x=0; x<8; x++){
            xy = y * stride + x;
            yz = y * 8 + x;
            temp[yz] = src[xy - stride] + 2*src[xy] + src[xy + stride];
2963 2964
        }
    }
2965

2966 2967 2968 2969 2970 2971 2972
    for(y=0; y<8; y++){
        src[  y*stride] = (temp[  y*8] + 2)>>2;
        src[7+y*stride] = (temp[7+y*8] + 2)>>2;
        for(x=1; x<7; x++){
            xy = y * stride + x;
            yz = y * 8 + x;
            src[xy] = (temp[yz-1] + 2*temp[yz] + temp[yz+1] + 8)>>4;
2973 2974 2975 2976
        }
    }
}

2977
static inline void h264_loop_filter_luma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991
{
    int i, d;
    for( i = 0; i < 4; i++ ) {
        if( tc0[i] < 0 ) {
            pix += 4*ystride;
            continue;
        }
        for( d = 0; d < 4; d++ ) {
            const int p0 = pix[-1*xstride];
            const int p1 = pix[-2*xstride];
            const int p2 = pix[-3*xstride];
            const int q0 = pix[0];
            const int q1 = pix[1*xstride];
            const int q2 = pix[2*xstride];
2992

2993 2994 2995
            if( FFABS( p0 - q0 ) < alpha &&
                FFABS( p1 - p0 ) < beta &&
                FFABS( q1 - q0 ) < beta ) {
2996

2997 2998
                int tc = tc0[i];
                int i_delta;
2999

3000
                if( FFABS( p2 - p0 ) < beta ) {
3001
                    pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc0[i], tc0[i] );
3002 3003
                    tc++;
                }
3004
                if( FFABS( q2 - q0 ) < beta ) {
3005
                    pix[   xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc0[i], tc0[i] );
3006 3007
                    tc++;
                }
3008

3009 3010 3011
                i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
                pix[-xstride] = av_clip_uint8( p0 + i_delta );    /* p0' */
                pix[0]        = av_clip_uint8( q0 - i_delta );    /* q0' */
3012 3013 3014 3015 3016
            }
            pix += ystride;
        }
    }
}
3017
static void h264_v_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
3018 3019 3020
{
    h264_loop_filter_luma_c(pix, stride, 1, alpha, beta, tc0);
}
3021
static void h264_h_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
3022 3023 3024 3025
{
    h264_loop_filter_luma_c(pix, 1, stride, alpha, beta, tc0);
}

3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082
static inline void h264_loop_filter_luma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
{
    int d;
    for( d = 0; d < 16; d++ ) {
        const int p2 = pix[-3*xstride];
        const int p1 = pix[-2*xstride];
        const int p0 = pix[-1*xstride];

        const int q0 = pix[ 0*xstride];
        const int q1 = pix[ 1*xstride];
        const int q2 = pix[ 2*xstride];

        if( FFABS( p0 - q0 ) < alpha &&
            FFABS( p1 - p0 ) < beta &&
            FFABS( q1 - q0 ) < beta ) {

            if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
                if( FFABS( p2 - p0 ) < beta)
                {
                    const int p3 = pix[-4*xstride];
                    /* p0', p1', p2' */
                    pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
                    pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
                    pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
                } else {
                    /* p0' */
                    pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                }
                if( FFABS( q2 - q0 ) < beta)
                {
                    const int q3 = pix[3*xstride];
                    /* q0', q1', q2' */
                    pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
                    pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
                    pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
                } else {
                    /* q0' */
                    pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
                }
            }else{
                /* p0', q0' */
                pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
            }
        }
        pix += ystride;
    }
}
static void h264_v_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
{
    h264_loop_filter_luma_intra_c(pix, stride, 1, alpha, beta);
}
static void h264_h_loop_filter_luma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
{
    h264_loop_filter_luma_intra_c(pix, 1, stride, alpha, beta);
}

3083
static inline void h264_loop_filter_chroma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0)
3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097
{
    int i, d;
    for( i = 0; i < 4; i++ ) {
        const int tc = tc0[i];
        if( tc <= 0 ) {
            pix += 2*ystride;
            continue;
        }
        for( d = 0; d < 2; d++ ) {
            const int p0 = pix[-1*xstride];
            const int p1 = pix[-2*xstride];
            const int q0 = pix[0];
            const int q1 = pix[1*xstride];

3098 3099 3100
            if( FFABS( p0 - q0 ) < alpha &&
                FFABS( p1 - p0 ) < beta &&
                FFABS( q1 - q0 ) < beta ) {
3101

3102
                int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
3103

3104 3105
                pix[-xstride] = av_clip_uint8( p0 + delta );    /* p0' */
                pix[0]        = av_clip_uint8( q0 - delta );    /* q0' */
3106 3107 3108 3109 3110
            }
            pix += ystride;
        }
    }
}
3111
static void h264_v_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
3112 3113 3114
{
    h264_loop_filter_chroma_c(pix, stride, 1, alpha, beta, tc0);
}
3115
static void h264_h_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0)
3116 3117 3118 3119
{
    h264_loop_filter_chroma_c(pix, 1, stride, alpha, beta, tc0);
}

3120 3121 3122 3123 3124 3125 3126 3127 3128
static inline void h264_loop_filter_chroma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta)
{
    int d;
    for( d = 0; d < 8; d++ ) {
        const int p0 = pix[-1*xstride];
        const int p1 = pix[-2*xstride];
        const int q0 = pix[0];
        const int q1 = pix[1*xstride];

3129 3130 3131
        if( FFABS( p0 - q0 ) < alpha &&
            FFABS( p1 - p0 ) < beta &&
            FFABS( q1 - q0 ) < beta ) {
3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147

            pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;   /* p0' */
            pix[0]        = ( 2*q1 + q0 + p1 + 2 ) >> 2;   /* q0' */
        }
        pix += ystride;
    }
}
static void h264_v_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
{
    h264_loop_filter_chroma_intra_c(pix, stride, 1, alpha, beta);
}
static void h264_h_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta)
{
    h264_loop_filter_chroma_intra_c(pix, 1, stride, alpha, beta);
}

3148
static inline int pix_abs16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
Fabrice Bellard's avatar
Fabrice Bellard committed
3149 3150 3151 3152
{
    int s, i;

    s = 0;
3153
    for(i=0;i<h;i++) {
Fabrice Bellard's avatar
Fabrice Bellard committed
3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175
        s += abs(pix1[0] - pix2[0]);
        s += abs(pix1[1] - pix2[1]);
        s += abs(pix1[2] - pix2[2]);
        s += abs(pix1[3] - pix2[3]);
        s += abs(pix1[4] - pix2[4]);
        s += abs(pix1[5] - pix2[5]);
        s += abs(pix1[6] - pix2[6]);
        s += abs(pix1[7] - pix2[7]);
        s += abs(pix1[8] - pix2[8]);
        s += abs(pix1[9] - pix2[9]);
        s += abs(pix1[10] - pix2[10]);
        s += abs(pix1[11] - pix2[11]);
        s += abs(pix1[12] - pix2[12]);
        s += abs(pix1[13] - pix2[13]);
        s += abs(pix1[14] - pix2[14]);
        s += abs(pix1[15] - pix2[15]);
        pix1 += line_size;
        pix2 += line_size;
    }
    return s;
}

3176
static int pix_abs16_x2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
Fabrice Bellard's avatar
Fabrice Bellard committed
3177 3178 3179 3180
{
    int s, i;

    s = 0;
3181
    for(i=0;i<h;i++) {
Fabrice Bellard's avatar
Fabrice Bellard committed
3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203
        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));
        s += abs(pix1[5] - avg2(pix2[5], pix2[6]));
        s += abs(pix1[6] - avg2(pix2[6], pix2[7]));
        s += abs(pix1[7] - avg2(pix2[7], pix2[8]));
        s += abs(pix1[8] - avg2(pix2[8], pix2[9]));
        s += abs(pix1[9] - avg2(pix2[9], pix2[10]));
        s += abs(pix1[10] - avg2(pix2[10], pix2[11]));
        s += abs(pix1[11] - avg2(pix2[11], pix2[12]));
        s += abs(pix1[12] - avg2(pix2[12], pix2[13]));
        s += abs(pix1[13] - avg2(pix2[13], pix2[14]));
        s += abs(pix1[14] - avg2(pix2[14], pix2[15]));
        s += abs(pix1[15] - avg2(pix2[15], pix2[16]));
        pix1 += line_size;
        pix2 += line_size;
    }
    return s;
}

3204
static int pix_abs16_y2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
Fabrice Bellard's avatar
Fabrice Bellard committed
3205 3206
{
    int s, i;
3207
    uint8_t *pix3 = pix2 + line_size;
Fabrice Bellard's avatar
Fabrice Bellard committed
3208 3209

    s = 0;
3210
    for(i=0;i<h;i++) {
Fabrice Bellard's avatar
Fabrice Bellard committed
3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233
        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));
        s += abs(pix1[5] - avg2(pix2[5], pix3[5]));
        s += abs(pix1[6] - avg2(pix2[6], pix3[6]));
        s += abs(pix1[7] - avg2(pix2[7], pix3[7]));
        s += abs(pix1[8] - avg2(pix2[8], pix3[8]));
        s += abs(pix1[9] - avg2(pix2[9], pix3[9]));
        s += abs(pix1[10] - avg2(pix2[10], pix3[10]));
        s += abs(pix1[11] - avg2(pix2[11], pix3[11]));
        s += abs(pix1[12] - avg2(pix2[12], pix3[12]));
        s += abs(pix1[13] - avg2(pix2[13], pix3[13]));
        s += abs(pix1[14] - avg2(pix2[14], pix3[14]));
        s += abs(pix1[15] - avg2(pix2[15], pix3[15]));
        pix1 += line_size;
        pix2 += line_size;
        pix3 += line_size;
    }
    return s;
}

3234
static int pix_abs16_xy2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
Fabrice Bellard's avatar
Fabrice Bellard committed
3235 3236
{
    int s, i;
3237
    uint8_t *pix3 = pix2 + line_size;
Fabrice Bellard's avatar
Fabrice Bellard committed
3238 3239

    s = 0;
3240
    for(i=0;i<h;i++) {
Fabrice Bellard's avatar
Fabrice Bellard committed
3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263
        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));
        s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));
        s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));
        s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));
        s += abs(pix1[8] - avg4(pix2[8], pix2[9], pix3[8], pix3[9]));
        s += abs(pix1[9] - avg4(pix2[9], pix2[10], pix3[9], pix3[10]));
        s += abs(pix1[10] - avg4(pix2[10], pix2[11], pix3[10], pix3[11]));
        s += abs(pix1[11] - avg4(pix2[11], pix2[12], pix3[11], pix3[12]));
        s += abs(pix1[12] - avg4(pix2[12], pix2[13], pix3[12], pix3[13]));
        s += abs(pix1[13] - avg4(pix2[13], pix2[14], pix3[13], pix3[14]));
        s += abs(pix1[14] - avg4(pix2[14], pix2[15], pix3[14], pix3[15]));
        s += abs(pix1[15] - avg4(pix2[15], pix2[16], pix3[15], pix3[16]));
        pix1 += line_size;
        pix2 += line_size;
        pix3 += line_size;
    }
    return s;
}

3264
static inline int pix_abs8_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
3265 3266 3267 3268
{
    int s, i;

    s = 0;
3269
    for(i=0;i<h;i++) {
3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283
        s += abs(pix1[0] - pix2[0]);
        s += abs(pix1[1] - pix2[1]);
        s += abs(pix1[2] - pix2[2]);
        s += abs(pix1[3] - pix2[3]);
        s += abs(pix1[4] - pix2[4]);
        s += abs(pix1[5] - pix2[5]);
        s += abs(pix1[6] - pix2[6]);
        s += abs(pix1[7] - pix2[7]);
        pix1 += line_size;
        pix2 += line_size;
    }
    return s;
}

3284
static int pix_abs8_x2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
3285 3286 3287 3288
{
    int s, i;

    s = 0;
3289
    for(i=0;i<h;i++) {
3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303
        s += abs(pix1[0] - avg2(pix2[0], pix2[1]));
        s += abs(pix1[1] - avg2(pix2[1], pix2[2]));
        s += abs(pix1[2] - avg2(pix2[2], pix2[3]));
        s += abs(pix1[3] - avg2(pix2[3], pix2[4]));
        s += abs(pix1[4] - avg2(pix2[4], pix2[5]));
        s += abs(pix1[5] - avg2(pix2[5], pix2[6]));
        s += abs(pix1[6] - avg2(pix2[6], pix2[7]));
        s += abs(pix1[7] - avg2(pix2[7], pix2[8]));
        pix1 += line_size;
        pix2 += line_size;
    }
    return s;
}

3304
static int pix_abs8_y2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
3305 3306
{
    int s, i;
3307
    uint8_t *pix3 = pix2 + line_size;
3308 3309

    s = 0;
3310
    for(i=0;i<h;i++) {
3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325
        s += abs(pix1[0] - avg2(pix2[0], pix3[0]));
        s += abs(pix1[1] - avg2(pix2[1], pix3[1]));
        s += abs(pix1[2] - avg2(pix2[2], pix3[2]));
        s += abs(pix1[3] - avg2(pix2[3], pix3[3]));
        s += abs(pix1[4] - avg2(pix2[4], pix3[4]));
        s += abs(pix1[5] - avg2(pix2[5], pix3[5]));
        s += abs(pix1[6] - avg2(pix2[6], pix3[6]));
        s += abs(pix1[7] - avg2(pix2[7], pix3[7]));
        pix1 += line_size;
        pix2 += line_size;
        pix3 += line_size;
    }
    return s;
}

3326
static int pix_abs8_xy2_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h)
3327 3328
{
    int s, i;
3329
    uint8_t *pix3 = pix2 + line_size;
3330 3331

    s = 0;
3332
    for(i=0;i<h;i++) {
3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347
        s += abs(pix1[0] - avg4(pix2[0], pix2[1], pix3[0], pix3[1]));
        s += abs(pix1[1] - avg4(pix2[1], pix2[2], pix3[1], pix3[2]));
        s += abs(pix1[2] - avg4(pix2[2], pix2[3], pix3[2], pix3[3]));
        s += abs(pix1[3] - avg4(pix2[3], pix2[4], pix3[3], pix3[4]));
        s += abs(pix1[4] - avg4(pix2[4], pix2[5], pix3[4], pix3[5]));
        s += abs(pix1[5] - avg4(pix2[5], pix2[6], pix3[5], pix3[6]));
        s += abs(pix1[6] - avg4(pix2[6], pix2[7], pix3[6], pix3[7]));
        s += abs(pix1[7] - avg4(pix2[7], pix2[8], pix3[7], pix3[8]));
        pix1 += line_size;
        pix2 += line_size;
        pix3 += line_size;
    }
    return s;
}

3348 3349
static int nsse16_c(void *v, uint8_t *s1, uint8_t *s2, int stride, int h){
    MpegEncContext *c = v;
3350 3351 3352
    int score1=0;
    int score2=0;
    int x,y;
Michael Niedermayer's avatar
Michael Niedermayer committed
3353

3354 3355 3356 3357 3358 3359
    for(y=0; y<h; y++){
        for(x=0; x<16; x++){
            score1+= (s1[x  ] - s2[x ])*(s1[x  ] - s2[x ]);
        }
        if(y+1<h){
            for(x=0; x<15; x++){
3360
                score2+= FFABS(  s1[x  ] - s1[x  +stride]
3361
                             - s1[x+1] + s1[x+1+stride])
3362
                        -FFABS(  s2[x  ] - s2[x  +stride]
3363 3364 3365 3366 3367 3368
                             - s2[x+1] + s2[x+1+stride]);
            }
        }
        s1+= stride;
        s2+= stride;
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
3369

3370 3371
    if(c) return score1 + FFABS(score2)*c->avctx->nsse_weight;
    else  return score1 + FFABS(score2)*8;
3372 3373
}

3374 3375
static int nsse8_c(void *v, uint8_t *s1, uint8_t *s2, int stride, int h){
    MpegEncContext *c = v;
3376 3377 3378
    int score1=0;
    int score2=0;
    int x,y;
3379

3380 3381 3382 3383 3384 3385
    for(y=0; y<h; y++){
        for(x=0; x<8; x++){
            score1+= (s1[x  ] - s2[x ])*(s1[x  ] - s2[x ]);
        }
        if(y+1<h){
            for(x=0; x<7; x++){
3386
                score2+= FFABS(  s1[x  ] - s1[x  +stride]
3387
                             - s1[x+1] + s1[x+1+stride])
3388
                        -FFABS(  s2[x  ] - s2[x  +stride]
3389 3390 3391 3392 3393 3394
                             - s2[x+1] + s2[x+1+stride]);
            }
        }
        s1+= stride;
        s2+= stride;
    }
3395

3396 3397
    if(c) return score1 + FFABS(score2)*c->avctx->nsse_weight;
    else  return score1 + FFABS(score2)*8;
3398 3399
}

3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419
static int try_8x8basis_c(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale){
    int i;
    unsigned int sum=0;

    for(i=0; i<8*8; i++){
        int b= rem[i] + ((basis[i]*scale + (1<<(BASIS_SHIFT - RECON_SHIFT-1)))>>(BASIS_SHIFT - RECON_SHIFT));
        int w= weight[i];
        b>>= RECON_SHIFT;
        assert(-512<b && b<512);

        sum += (w*b)*(w*b)>>4;
    }
    return sum>>2;
}

static void add_8x8basis_c(int16_t rem[64], int16_t basis[64], int scale){
    int i;

    for(i=0; i<8*8; i++){
        rem[i] += (basis[i]*scale + (1<<(BASIS_SHIFT - RECON_SHIFT-1)))>>(BASIS_SHIFT - RECON_SHIFT);
3420
    }
3421 3422
}

3423 3424
/**
 * permutes an 8x8 block.
3425
 * @param block the block which will be permuted according to the given permutation vector
3426 3427
 * @param permutation the permutation vector
 * @param last the last non zero coefficient in scantable order, used to speed the permutation up
3428
 * @param scantable the used scantable, this is only used to speed the permutation up, the block is not
3429
 *                  (inverse) permutated to scantable order!
3430
 */
3431
void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last)
3432
{
3433
    int i;
3434
    DCTELEM temp[64];
3435

3436
    if(last<=0) return;
Diego Biurrun's avatar
Diego Biurrun committed
3437
    //if(permutation[1]==1) return; //FIXME it is ok but not clean and might fail for some permutations
3438

3439 3440 3441 3442 3443
    for(i=0; i<=last; i++){
        const int j= scantable[i];
        temp[j]= block[j];
        block[j]=0;
    }
3444

3445 3446 3447 3448 3449
    for(i=0; i<=last; i++){
        const int j= scantable[i];
        const int perm_j= permutation[j];
        block[perm_j]= temp[j];
    }
3450
}
3451

3452 3453 3454 3455 3456 3457
static int zero_cmp(void *s, uint8_t *a, uint8_t *b, int stride, int h){
    return 0;
}

void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type){
    int i;
3458

3459
    memset(cmp, 0, sizeof(void*)*6);
3460

3461
    for(i=0; i<6; i++){
3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474
        switch(type&0xFF){
        case FF_CMP_SAD:
            cmp[i]= c->sad[i];
            break;
        case FF_CMP_SATD:
            cmp[i]= c->hadamard8_diff[i];
            break;
        case FF_CMP_SSE:
            cmp[i]= c->sse[i];
            break;
        case FF_CMP_DCT:
            cmp[i]= c->dct_sad[i];
            break;
3475 3476 3477
        case FF_CMP_DCT264:
            cmp[i]= c->dct264_sad[i];
            break;
3478 3479 3480
        case FF_CMP_DCTMAX:
            cmp[i]= c->dct_max[i];
            break;
3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498
        case FF_CMP_PSNR:
            cmp[i]= c->quant_psnr[i];
            break;
        case FF_CMP_BIT:
            cmp[i]= c->bit[i];
            break;
        case FF_CMP_RD:
            cmp[i]= c->rd[i];
            break;
        case FF_CMP_VSAD:
            cmp[i]= c->vsad[i];
            break;
        case FF_CMP_VSSE:
            cmp[i]= c->vsse[i];
            break;
        case FF_CMP_ZERO:
            cmp[i]= zero_cmp;
            break;
3499 3500 3501
        case FF_CMP_NSSE:
            cmp[i]= c->nsse[i];
            break;
3502
#if CONFIG_SNOW_ENCODER
3503 3504 3505 3506 3507 3508
        case FF_CMP_W53:
            cmp[i]= c->w53[i];
            break;
        case FF_CMP_W97:
            cmp[i]= c->w97[i];
            break;
3509
#endif
3510 3511 3512 3513 3514 3515
        default:
            av_log(NULL, AV_LOG_ERROR,"internal error in cmp function selection\n");
        }
    }
}

Loren Merritt's avatar
Loren Merritt committed
3516 3517 3518 3519 3520
static void clear_block_c(DCTELEM *block)
{
    memset(block, 0, sizeof(DCTELEM)*64);
}

3521 3522 3523
/**
 * memset(blocks, 0, sizeof(DCTELEM)*6*64)
 */
3524
static void clear_blocks_c(DCTELEM *blocks)
3525 3526 3527 3528
{
    memset(blocks, 0, sizeof(DCTELEM)*6*64);
}

Michael Niedermayer's avatar
Michael Niedermayer committed
3529
static void add_bytes_c(uint8_t *dst, uint8_t *src, int w){
3530 3531 3532 3533 3534
    long i;
    for(i=0; i<=w-sizeof(long); i+=sizeof(long)){
        long a = *(long*)(src+i);
        long b = *(long*)(dst+i);
        *(long*)(dst+i) = ((a&pb_7f) + (b&pb_7f)) ^ ((a^b)&pb_80);
Michael Niedermayer's avatar
Michael Niedermayer committed
3535 3536 3537 3538 3539
    }
    for(; i<w; i++)
        dst[i+0] += src[i+0];
}

3540
static void add_bytes_l2_c(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){
3541
    long i;
3542 3543 3544
    for(i=0; i<=w-sizeof(long); i+=sizeof(long)){
        long a = *(long*)(src1+i);
        long b = *(long*)(src2+i);
3545
        *(long*)(dst+i) = ((a&pb_7f) + (b&pb_7f)) ^ ((a^b)&pb_80);
3546 3547 3548 3549 3550
    }
    for(; i<w; i++)
        dst[i] = src1[i]+src2[i];
}

Michael Niedermayer's avatar
Michael Niedermayer committed
3551
static void diff_bytes_c(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){
3552
    long i;
3553
#if !HAVE_FAST_UNALIGNED
3554
    if((long)src2 & (sizeof(long)-1)){
Loren Merritt's avatar
Loren Merritt committed
3555 3556 3557 3558 3559 3560 3561 3562 3563 3564
        for(i=0; i+7<w; i+=8){
            dst[i+0] = src1[i+0]-src2[i+0];
            dst[i+1] = src1[i+1]-src2[i+1];
            dst[i+2] = src1[i+2]-src2[i+2];
            dst[i+3] = src1[i+3]-src2[i+3];
            dst[i+4] = src1[i+4]-src2[i+4];
            dst[i+5] = src1[i+5]-src2[i+5];
            dst[i+6] = src1[i+6]-src2[i+6];
            dst[i+7] = src1[i+7]-src2[i+7];
        }
3565 3566 3567 3568 3569 3570 3571
    }else
#endif
    for(i=0; i<=w-sizeof(long); i+=sizeof(long)){
        long a = *(long*)(src1+i);
        long b = *(long*)(src2+i);
        *(long*)(dst+i) = ((a|pb_80) - (b&pb_7f)) ^ ((a^b^pb_80)&pb_80);
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
3572 3573 3574 3575
    for(; i<w; i++)
        dst[i+0] = src1[i+0]-src2[i+0];
}

3576
static void add_hfyu_median_prediction_c(uint8_t *dst, const uint8_t *src1, const uint8_t *diff, int w, int *left, int *left_top){
3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592
    int i;
    uint8_t l, lt;

    l= *left;
    lt= *left_top;

    for(i=0; i<w; i++){
        l= mid_pred(l, src1[i], (l + src1[i] - lt)&0xFF) + diff[i];
        lt= src1[i];
        dst[i]= l;
    }

    *left= l;
    *left_top= lt;
}

3593
static void sub_hfyu_median_prediction_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top){
3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604
    int i;
    uint8_t l, lt;

    l= *left;
    lt= *left_top;

    for(i=0; i<w; i++){
        const int pred= mid_pred(l, src1[i], (l + src1[i] - lt)&0xFF);
        lt= src1[i];
        l= src2[i];
        dst[i]= l - pred;
3605
    }
3606 3607 3608 3609 3610

    *left= l;
    *left_top= lt;
}

3611
static int add_hfyu_left_prediction_c(uint8_t *dst, const uint8_t *src, int w, int acc){
3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636 3637 3638
    int i;

    for(i=0; i<w-1; i++){
        acc+= src[i];
        dst[i]= acc;
        i++;
        acc+= src[i];
        dst[i]= acc;
    }

    for(; i<w; i++){
        acc+= src[i];
        dst[i]= acc;
    }

    return acc;
}

#if HAVE_BIGENDIAN
#define B 3
#define G 2
#define R 1
#else
#define B 0
#define G 1
#define R 2
#endif
3639
static void add_hfyu_left_prediction_bgr32_c(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue){
3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663
    int i;
    int r,g,b;
    r= *red;
    g= *green;
    b= *blue;

    for(i=0; i<w; i++){
        b+= src[4*i+B];
        g+= src[4*i+G];
        r+= src[4*i+R];

        dst[4*i+B]= b;
        dst[4*i+G]= g;
        dst[4*i+R]= r;
    }

    *red= r;
    *green= g;
    *blue= b;
}
#undef B
#undef G
#undef R

Michael Niedermayer's avatar
Michael Niedermayer committed
3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676
#define BUTTERFLY2(o1,o2,i1,i2) \
o1= (i1)+(i2);\
o2= (i1)-(i2);

#define BUTTERFLY1(x,y) \
{\
    int a,b;\
    a= x;\
    b= y;\
    x= a+b;\
    y= a-b;\
}

3677
#define BUTTERFLYA(x,y) (FFABS((x)+(y)) + FFABS((x)-(y)))
Michael Niedermayer's avatar
Michael Niedermayer committed
3678

3679
static int hadamard8_diff8x8_c(/*MpegEncContext*/ void *s, uint8_t *dst, uint8_t *src, int stride, int h){
Michael Niedermayer's avatar
Michael Niedermayer committed
3680 3681 3682
    int i;
    int temp[64];
    int sum=0;
3683

3684
    assert(h==8);
Michael Niedermayer's avatar
Michael Niedermayer committed
3685 3686 3687 3688 3689 3690 3691

    for(i=0; i<8; i++){
        //FIXME try pointer walks
        BUTTERFLY2(temp[8*i+0], temp[8*i+1], src[stride*i+0]-dst[stride*i+0],src[stride*i+1]-dst[stride*i+1]);
        BUTTERFLY2(temp[8*i+2], temp[8*i+3], src[stride*i+2]-dst[stride*i+2],src[stride*i+3]-dst[stride*i+3]);
        BUTTERFLY2(temp[8*i+4], temp[8*i+5], src[stride*i+4]-dst[stride*i+4],src[stride*i+5]-dst[stride*i+5]);
        BUTTERFLY2(temp[8*i+6], temp[8*i+7], src[stride*i+6]-dst[stride*i+6],src[stride*i+7]-dst[stride*i+7]);
3692

Michael Niedermayer's avatar
Michael Niedermayer committed
3693 3694 3695 3696
        BUTTERFLY1(temp[8*i+0], temp[8*i+2]);
        BUTTERFLY1(temp[8*i+1], temp[8*i+3]);
        BUTTERFLY1(temp[8*i+4], temp[8*i+6]);
        BUTTERFLY1(temp[8*i+5], temp[8*i+7]);
3697

Michael Niedermayer's avatar
Michael Niedermayer committed
3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708
        BUTTERFLY1(temp[8*i+0], temp[8*i+4]);
        BUTTERFLY1(temp[8*i+1], temp[8*i+5]);
        BUTTERFLY1(temp[8*i+2], temp[8*i+6]);
        BUTTERFLY1(temp[8*i+3], temp[8*i+7]);
    }

    for(i=0; i<8; i++){
        BUTTERFLY1(temp[8*0+i], temp[8*1+i]);
        BUTTERFLY1(temp[8*2+i], temp[8*3+i]);
        BUTTERFLY1(temp[8*4+i], temp[8*5+i]);
        BUTTERFLY1(temp[8*6+i], temp[8*7+i]);
3709

Michael Niedermayer's avatar
Michael Niedermayer committed
3710 3711 3712 3713 3714
        BUTTERFLY1(temp[8*0+i], temp[8*2+i]);
        BUTTERFLY1(temp[8*1+i], temp[8*3+i]);
        BUTTERFLY1(temp[8*4+i], temp[8*6+i]);
        BUTTERFLY1(temp[8*5+i], temp[8*7+i]);

3715
        sum +=
Michael Niedermayer's avatar
Michael Niedermayer committed
3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730
             BUTTERFLYA(temp[8*0+i], temp[8*4+i])
            +BUTTERFLYA(temp[8*1+i], temp[8*5+i])
            +BUTTERFLYA(temp[8*2+i], temp[8*6+i])
            +BUTTERFLYA(temp[8*3+i], temp[8*7+i]);
    }
#if 0
static int maxi=0;
if(sum>maxi){
    maxi=sum;
    printf("MAX:%d\n", maxi);
}
#endif
    return sum;
}

3731
static int hadamard8_intra8x8_c(/*MpegEncContext*/ void *s, uint8_t *src, uint8_t *dummy, int stride, int h){
Michael Niedermayer's avatar
Michael Niedermayer committed
3732 3733 3734
    int i;
    int temp[64];
    int sum=0;
3735

3736
    assert(h==8);
3737

Michael Niedermayer's avatar
Michael Niedermayer committed
3738 3739
    for(i=0; i<8; i++){
        //FIXME try pointer walks
3740 3741 3742 3743
        BUTTERFLY2(temp[8*i+0], temp[8*i+1], src[stride*i+0],src[stride*i+1]);
        BUTTERFLY2(temp[8*i+2], temp[8*i+3], src[stride*i+2],src[stride*i+3]);
        BUTTERFLY2(temp[8*i+4], temp[8*i+5], src[stride*i+4],src[stride*i+5]);
        BUTTERFLY2(temp[8*i+6], temp[8*i+7], src[stride*i+6],src[stride*i+7]);
3744

Michael Niedermayer's avatar
Michael Niedermayer committed
3745 3746 3747 3748
        BUTTERFLY1(temp[8*i+0], temp[8*i+2]);
        BUTTERFLY1(temp[8*i+1], temp[8*i+3]);
        BUTTERFLY1(temp[8*i+4], temp[8*i+6]);
        BUTTERFLY1(temp[8*i+5], temp[8*i+7]);
3749

Michael Niedermayer's avatar
Michael Niedermayer committed
3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760
        BUTTERFLY1(temp[8*i+0], temp[8*i+4]);
        BUTTERFLY1(temp[8*i+1], temp[8*i+5]);
        BUTTERFLY1(temp[8*i+2], temp[8*i+6]);
        BUTTERFLY1(temp[8*i+3], temp[8*i+7]);
    }

    for(i=0; i<8; i++){
        BUTTERFLY1(temp[8*0+i], temp[8*1+i]);
        BUTTERFLY1(temp[8*2+i], temp[8*3+i]);
        BUTTERFLY1(temp[8*4+i], temp[8*5+i]);
        BUTTERFLY1(temp[8*6+i], temp[8*7+i]);
3761

Michael Niedermayer's avatar
Michael Niedermayer committed
3762 3763 3764 3765
        BUTTERFLY1(temp[8*0+i], temp[8*2+i]);
        BUTTERFLY1(temp[8*1+i], temp[8*3+i]);
        BUTTERFLY1(temp[8*4+i], temp[8*6+i]);
        BUTTERFLY1(temp[8*5+i], temp[8*7+i]);
3766 3767

        sum +=
Michael Niedermayer's avatar
Michael Niedermayer committed
3768 3769 3770 3771 3772
             BUTTERFLYA(temp[8*0+i], temp[8*4+i])
            +BUTTERFLYA(temp[8*1+i], temp[8*5+i])
            +BUTTERFLYA(temp[8*2+i], temp[8*6+i])
            +BUTTERFLYA(temp[8*3+i], temp[8*7+i]);
    }
3773

3774
    sum -= FFABS(temp[8*0] + temp[8*4]); // -mean
3775

Michael Niedermayer's avatar
Michael Niedermayer committed
3776 3777 3778
    return sum;
}

3779
static int dct_sad8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
Michael Niedermayer's avatar
Michael Niedermayer committed
3780
    MpegEncContext * const s= (MpegEncContext *)c;
3781
    DECLARE_ALIGNED_16(uint64_t, aligned_temp[sizeof(DCTELEM)*64/8]);
Michael Niedermayer's avatar
Michael Niedermayer committed
3782
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
3783

3784
    assert(h==8);
Michael Niedermayer's avatar
Michael Niedermayer committed
3785 3786

    s->dsp.diff_pixels(temp, src1, src2, stride);
3787
    s->dsp.fdct(temp);
3788
    return s->dsp.sum_abs_dctelem(temp);
Michael Niedermayer's avatar
Michael Niedermayer committed
3789 3790
}

3791
#if CONFIG_GPL
3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820
#define DCT8_1D {\
    const int s07 = SRC(0) + SRC(7);\
    const int s16 = SRC(1) + SRC(6);\
    const int s25 = SRC(2) + SRC(5);\
    const int s34 = SRC(3) + SRC(4);\
    const int a0 = s07 + s34;\
    const int a1 = s16 + s25;\
    const int a2 = s07 - s34;\
    const int a3 = s16 - s25;\
    const int d07 = SRC(0) - SRC(7);\
    const int d16 = SRC(1) - SRC(6);\
    const int d25 = SRC(2) - SRC(5);\
    const int d34 = SRC(3) - SRC(4);\
    const int a4 = d16 + d25 + (d07 + (d07>>1));\
    const int a5 = d07 - d34 - (d25 + (d25>>1));\
    const int a6 = d07 + d34 - (d16 + (d16>>1));\
    const int a7 = d16 - d25 + (d34 + (d34>>1));\
    DST(0,  a0 + a1     ) ;\
    DST(1,  a4 + (a7>>2)) ;\
    DST(2,  a2 + (a3>>1)) ;\
    DST(3,  a5 + (a6>>2)) ;\
    DST(4,  a0 - a1     ) ;\
    DST(5,  a6 - (a5>>2)) ;\
    DST(6, (a2>>1) - a3 ) ;\
    DST(7, (a4>>2) - a7 ) ;\
}

static int dct264_sad8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
    MpegEncContext * const s= (MpegEncContext *)c;
Måns Rullgård's avatar
Måns Rullgård committed
3821
    DCTELEM dct[8][8];
3822 3823 3824
    int i;
    int sum=0;

Måns Rullgård's avatar
Måns Rullgård committed
3825
    s->dsp.diff_pixels(dct[0], src1, src2, stride);
3826 3827 3828 3829 3830 3831 3832 3833 3834

#define SRC(x) dct[i][x]
#define DST(x,v) dct[i][x]= v
    for( i = 0; i < 8; i++ )
        DCT8_1D
#undef SRC
#undef DST

#define SRC(x) dct[x][i]
3835
#define DST(x,v) sum += FFABS(v)
3836 3837 3838 3839 3840 3841 3842 3843
    for( i = 0; i < 8; i++ )
        DCT8_1D
#undef SRC
#undef DST
    return sum;
}
#endif

3844 3845
static int dct_max8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
    MpegEncContext * const s= (MpegEncContext *)c;
3846
    DECLARE_ALIGNED_16(uint64_t, aligned_temp[sizeof(DCTELEM)*64/8]);
3847 3848
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
    int sum=0, i;
3849

3850 3851 3852 3853 3854 3855
    assert(h==8);

    s->dsp.diff_pixels(temp, src1, src2, stride);
    s->dsp.fdct(temp);

    for(i=0; i<64; i++)
3856
        sum= FFMAX(sum, FFABS(temp[i]));
3857

3858 3859 3860
    return sum;
}

3861
static int quant_psnr8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
Michael Niedermayer's avatar
Michael Niedermayer committed
3862
    MpegEncContext * const s= (MpegEncContext *)c;
3863
    DECLARE_ALIGNED_16(uint64_t, aligned_temp[sizeof(DCTELEM)*64*2/8]);
Michael Niedermayer's avatar
Michael Niedermayer committed
3864 3865
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
    DCTELEM * const bak = ((DCTELEM*)aligned_temp)+64;
Michael Niedermayer's avatar
Michael Niedermayer committed
3866 3867
    int sum=0, i;

3868
    assert(h==8);
Michael Niedermayer's avatar
Michael Niedermayer committed
3869
    s->mb_intra=0;
3870

Michael Niedermayer's avatar
Michael Niedermayer committed
3871
    s->dsp.diff_pixels(temp, src1, src2, stride);
3872

Michael Niedermayer's avatar
Michael Niedermayer committed
3873
    memcpy(bak, temp, 64*sizeof(DCTELEM));
3874

3875
    s->block_last_index[0/*FIXME*/]= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);
3876
    s->dct_unquantize_inter(s, temp, 0, s->qscale);
3877
    ff_simple_idct(temp); //FIXME
3878

Michael Niedermayer's avatar
Michael Niedermayer committed
3879 3880
    for(i=0; i<64; i++)
        sum+= (temp[i]-bak[i])*(temp[i]-bak[i]);
3881

Michael Niedermayer's avatar
Michael Niedermayer committed
3882 3883 3884
    return sum;
}

3885
static int rd8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
3886
    MpegEncContext * const s= (MpegEncContext *)c;
3887
    const uint8_t *scantable= s->intra_scantable.permutated;
3888 3889 3890
    DECLARE_ALIGNED_16(uint64_t, aligned_temp[sizeof(DCTELEM)*64/8]);
    DECLARE_ALIGNED_16(uint64_t, aligned_src1[8]);
    DECLARE_ALIGNED_16(uint64_t, aligned_src2[8]);
Michael Niedermayer's avatar
Michael Niedermayer committed
3891
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
3892 3893
    uint8_t * const lsrc1 = (uint8_t*)aligned_src1;
    uint8_t * const lsrc2 = (uint8_t*)aligned_src2;
3894
    int i, last, run, bits, level, distortion, start_i;
3895 3896 3897
    const int esc_length= s->ac_esc_length;
    uint8_t * length;
    uint8_t * last_length;
3898

3899 3900
    assert(h==8);

3901 3902
    copy_block8(lsrc1, src1, 8, stride, 8);
    copy_block8(lsrc2, src2, 8, stride, 8);
3903

3904
    s->dsp.diff_pixels(temp, lsrc1, lsrc2, 8);
3905 3906 3907 3908

    s->block_last_index[0/*FIXME*/]= last= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);

    bits=0;
3909

3910
    if (s->mb_intra) {
3911
        start_i = 1;
3912 3913
        length     = s->intra_ac_vlc_length;
        last_length= s->intra_ac_vlc_last_length;
3914
        bits+= s->luma_dc_vlc_length[temp[0] + 256]; //FIXME chroma
3915 3916 3917 3918 3919
    } else {
        start_i = 0;
        length     = s->inter_ac_vlc_length;
        last_length= s->inter_ac_vlc_last_length;
    }
3920

3921
    if(last>=start_i){
3922 3923 3924 3925
        run=0;
        for(i=start_i; i<last; i++){
            int j= scantable[i];
            level= temp[j];
3926

3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937
            if(level){
                level+=64;
                if((level&(~127)) == 0){
                    bits+= length[UNI_AC_ENC_INDEX(run, level)];
                }else
                    bits+= esc_length;
                run=0;
            }else
                run++;
        }
        i= scantable[last];
3938

3939
        level= temp[i] + 64;
Michael Niedermayer's avatar
Michael Niedermayer committed
3940 3941

        assert(level - 64);
3942

3943 3944 3945 3946
        if((level&(~127)) == 0){
            bits+= last_length[UNI_AC_ENC_INDEX(run, level)];
        }else
            bits+= esc_length;
3947

3948 3949 3950
    }

    if(last>=0){
3951 3952 3953 3954
        if(s->mb_intra)
            s->dct_unquantize_intra(s, temp, 0, s->qscale);
        else
            s->dct_unquantize_inter(s, temp, 0, s->qscale);
3955
    }
3956

3957
    s->dsp.idct_add(lsrc2, 8, temp);
3958

3959
    distortion= s->dsp.sse[1](NULL, lsrc2, lsrc1, 8, 8);
3960

3961
    return distortion + ((bits*s->qscale*s->qscale*109 + 64)>>7);
3962 3963
}

3964
static int bit8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
3965
    MpegEncContext * const s= (MpegEncContext *)c;
3966
    const uint8_t *scantable= s->intra_scantable.permutated;
3967
    DECLARE_ALIGNED_16(uint64_t, aligned_temp[sizeof(DCTELEM)*64/8]);
Michael Niedermayer's avatar
Michael Niedermayer committed
3968
    DCTELEM * const temp= (DCTELEM*)aligned_temp;
3969 3970 3971 3972
    int i, last, run, bits, level, start_i;
    const int esc_length= s->ac_esc_length;
    uint8_t * length;
    uint8_t * last_length;
3973 3974

    assert(h==8);
3975

3976
    s->dsp.diff_pixels(temp, src1, src2, stride);
3977

3978 3979 3980
    s->block_last_index[0/*FIXME*/]= last= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i);

    bits=0;
3981

3982
    if (s->mb_intra) {
3983
        start_i = 1;
3984 3985
        length     = s->intra_ac_vlc_length;
        last_length= s->intra_ac_vlc_last_length;
3986
        bits+= s->luma_dc_vlc_length[temp[0] + 256]; //FIXME chroma
3987 3988 3989 3990 3991
    } else {
        start_i = 0;
        length     = s->inter_ac_vlc_length;
        last_length= s->inter_ac_vlc_last_length;
    }
3992

3993
    if(last>=start_i){
3994 3995 3996 3997
        run=0;
        for(i=start_i; i<last; i++){
            int j= scantable[i];
            level= temp[j];
3998

3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009
            if(level){
                level+=64;
                if((level&(~127)) == 0){
                    bits+= length[UNI_AC_ENC_INDEX(run, level)];
                }else
                    bits+= esc_length;
                run=0;
            }else
                run++;
        }
        i= scantable[last];
4010

4011
        level= temp[i] + 64;
4012

4013
        assert(level - 64);
4014

4015 4016 4017 4018 4019 4020 4021 4022 4023
        if((level&(~127)) == 0){
            bits+= last_length[UNI_AC_ENC_INDEX(run, level)];
        }else
            bits+= esc_length;
    }

    return bits;
}

4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040
#define VSAD_INTRA(size) \
static int vsad_intra##size##_c(/*MpegEncContext*/ void *c, uint8_t *s, uint8_t *dummy, int stride, int h){ \
    int score=0;                                                                                            \
    int x,y;                                                                                                \
                                                                                                            \
    for(y=1; y<h; y++){                                                                                     \
        for(x=0; x<size; x+=4){                                                                             \
            score+= FFABS(s[x  ] - s[x  +stride]) + FFABS(s[x+1] - s[x+1+stride])                           \
                   +FFABS(s[x+2] - s[x+2+stride]) + FFABS(s[x+3] - s[x+3+stride]);                          \
        }                                                                                                   \
        s+= stride;                                                                                         \
    }                                                                                                       \
                                                                                                            \
    return score;                                                                                           \
}
VSAD_INTRA(8)
VSAD_INTRA(16)
4041 4042 4043 4044

static int vsad16_c(/*MpegEncContext*/ void *c, uint8_t *s1, uint8_t *s2, int stride, int h){
    int score=0;
    int x,y;
4045

4046 4047
    for(y=1; y<h; y++){
        for(x=0; x<16; x++){
4048
            score+= FFABS(s1[x  ] - s2[x ] - s1[x  +stride] + s2[x +stride]);
4049 4050 4051 4052
        }
        s1+= stride;
        s2+= stride;
    }
4053

4054 4055 4056 4057
    return score;
}

#define SQ(a) ((a)*(a))
4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074
#define VSSE_INTRA(size) \
static int vsse_intra##size##_c(/*MpegEncContext*/ void *c, uint8_t *s, uint8_t *dummy, int stride, int h){ \
    int score=0;                                                                                            \
    int x,y;                                                                                                \
                                                                                                            \
    for(y=1; y<h; y++){                                                                                     \
        for(x=0; x<size; x+=4){                                                                               \
            score+= SQ(s[x  ] - s[x  +stride]) + SQ(s[x+1] - s[x+1+stride])                                 \
                   +SQ(s[x+2] - s[x+2+stride]) + SQ(s[x+3] - s[x+3+stride]);                                \
        }                                                                                                   \
        s+= stride;                                                                                         \
    }                                                                                                       \
                                                                                                            \
    return score;                                                                                           \
}
VSSE_INTRA(8)
VSSE_INTRA(16)
4075 4076 4077 4078

static int vsse16_c(/*MpegEncContext*/ void *c, uint8_t *s1, uint8_t *s2, int stride, int h){
    int score=0;
    int x,y;
4079

4080 4081 4082 4083 4084 4085 4086
    for(y=1; y<h; y++){
        for(x=0; x<16; x++){
            score+= SQ(s1[x  ] - s2[x ] - s1[x  +stride] + s2[x +stride]);
        }
        s1+= stride;
        s2+= stride;
    }
4087

4088 4089 4090
    return score;
}

4091 4092
static int ssd_int8_vs_int16_c(const int8_t *pix1, const int16_t *pix2,
                               int size){
4093 4094 4095 4096 4097 4098 4099
    int score=0;
    int i;
    for(i=0; i<size; i++)
        score += (pix1[i]-pix2[i])*(pix1[i]-pix2[i]);
    return score;
}

4100 4101 4102
WRAPPER8_16_SQ(hadamard8_diff8x8_c, hadamard8_diff16_c)
WRAPPER8_16_SQ(hadamard8_intra8x8_c, hadamard8_intra16_c)
WRAPPER8_16_SQ(dct_sad8x8_c, dct_sad16_c)
4103
#if CONFIG_GPL
4104
WRAPPER8_16_SQ(dct264_sad8x8_c, dct264_sad16_c)
Mike Melanson's avatar
Mike Melanson committed
4105
#endif
4106 4107 4108 4109
WRAPPER8_16_SQ(dct_max8x8_c, dct_max16_c)
WRAPPER8_16_SQ(quant_psnr8x8_c, quant_psnr16_c)
WRAPPER8_16_SQ(rd8x8_c, rd16_c)
WRAPPER8_16_SQ(bit8x8_c, bit16_c)
Michael Niedermayer's avatar
Michael Niedermayer committed
4110

4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123
static void vector_fmul_c(float *dst, const float *src, int len){
    int i;
    for(i=0; i<len; i++)
        dst[i] *= src[i];
}

static void vector_fmul_reverse_c(float *dst, const float *src0, const float *src1, int len){
    int i;
    src1 += len-1;
    for(i=0; i<len; i++)
        dst[i] = src0[i] * src1[-i];
}

4124
static void vector_fmul_add_c(float *dst, const float *src0, const float *src1, const float *src2, int len){
4125 4126
    int i;
    for(i=0; i<len; i++)
4127
        dst[i] = src0[i] * src1[i] + src2[i];
4128 4129
}

4130
void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len){
Loren Merritt's avatar
Loren Merritt committed
4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142
    int i,j;
    dst += len;
    win += len;
    src0+= len;
    for(i=-len, j=len-1; i<0; i++, j--) {
        float s0 = src0[i];
        float s1 = src1[j];
        float wi = win[i];
        float wj = win[j];
        dst[i] = s0*wj - s1*wi + add_bias;
        dst[j] = s0*wi + s1*wj + add_bias;
    }
4143 4144
}

4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218
static void vector_fmul_scalar_c(float *dst, const float *src, float mul,
                                 int len)
{
    int i;
    for (i = 0; i < len; i++)
        dst[i] = src[i] * mul;
}

static void vector_fmul_sv_scalar_2_c(float *dst, const float *src,
                                      const float **sv, float mul, int len)
{
    int i;
    for (i = 0; i < len; i += 2, sv++) {
        dst[i  ] = src[i  ] * sv[0][0] * mul;
        dst[i+1] = src[i+1] * sv[0][1] * mul;
    }
}

static void vector_fmul_sv_scalar_4_c(float *dst, const float *src,
                                      const float **sv, float mul, int len)
{
    int i;
    for (i = 0; i < len; i += 4, sv++) {
        dst[i  ] = src[i  ] * sv[0][0] * mul;
        dst[i+1] = src[i+1] * sv[0][1] * mul;
        dst[i+2] = src[i+2] * sv[0][2] * mul;
        dst[i+3] = src[i+3] * sv[0][3] * mul;
    }
}

static void sv_fmul_scalar_2_c(float *dst, const float **sv, float mul,
                               int len)
{
    int i;
    for (i = 0; i < len; i += 2, sv++) {
        dst[i  ] = sv[0][0] * mul;
        dst[i+1] = sv[0][1] * mul;
    }
}

static void sv_fmul_scalar_4_c(float *dst, const float **sv, float mul,
                               int len)
{
    int i;
    for (i = 0; i < len; i += 4, sv++) {
        dst[i  ] = sv[0][0] * mul;
        dst[i+1] = sv[0][1] * mul;
        dst[i+2] = sv[0][2] * mul;
        dst[i+3] = sv[0][3] * mul;
    }
}

static void butterflies_float_c(float *restrict v1, float *restrict v2,
                                int len)
{
    int i;
    for (i = 0; i < len; i++) {
        float t = v1[i] - v2[i];
        v1[i] += v2[i];
        v2[i] = t;
    }
}

static float scalarproduct_float_c(const float *v1, const float *v2, int len)
{
    float p = 0.0;
    int i;

    for (i = 0; i < len; i++)
        p += v1[i] * v2[i];

    return p;
}

Loren Merritt's avatar
Loren Merritt committed
4219 4220 4221 4222 4223 4224
static void int32_to_float_fmul_scalar_c(float *dst, const int *src, float mul, int len){
    int i;
    for(i=0; i<len; i++)
        dst[i] = src[i] * mul;
}

4225 4226 4227 4228 4229 4230 4231 4232 4233
static inline uint32_t clipf_c_one(uint32_t a, uint32_t mini,
                   uint32_t maxi, uint32_t maxisign)
{

    if(a > mini) return mini;
    else if((a^(1<<31)) > maxisign) return maxi;
    else return a;
}

4234
static void vector_clipf_c_opposite_sign(float *dst, const float *src, float *min, float *max, int len){
4235 4236 4237 4238 4239
    int i;
    uint32_t mini = *(uint32_t*)min;
    uint32_t maxi = *(uint32_t*)max;
    uint32_t maxisign = maxi ^ (1<<31);
    uint32_t *dsti = (uint32_t*)dst;
4240
    const uint32_t *srci = (const uint32_t*)src;
4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251
    for(i=0; i<len; i+=8) {
        dsti[i + 0] = clipf_c_one(srci[i + 0], mini, maxi, maxisign);
        dsti[i + 1] = clipf_c_one(srci[i + 1], mini, maxi, maxisign);
        dsti[i + 2] = clipf_c_one(srci[i + 2], mini, maxi, maxisign);
        dsti[i + 3] = clipf_c_one(srci[i + 3], mini, maxi, maxisign);
        dsti[i + 4] = clipf_c_one(srci[i + 4], mini, maxi, maxisign);
        dsti[i + 5] = clipf_c_one(srci[i + 5], mini, maxi, maxisign);
        dsti[i + 6] = clipf_c_one(srci[i + 6], mini, maxi, maxisign);
        dsti[i + 7] = clipf_c_one(srci[i + 7], mini, maxi, maxisign);
    }
}
4252
static void vector_clipf_c(float *dst, const float *src, float min, float max, int len){
4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269
    int i;
    if(min < 0 && max > 0) {
        vector_clipf_c_opposite_sign(dst, src, &min, &max, len);
    } else {
        for(i=0; i < len; i+=8) {
            dst[i    ] = av_clipf(src[i    ], min, max);
            dst[i + 1] = av_clipf(src[i + 1], min, max);
            dst[i + 2] = av_clipf(src[i + 2], min, max);
            dst[i + 3] = av_clipf(src[i + 3], min, max);
            dst[i + 4] = av_clipf(src[i + 4], min, max);
            dst[i + 5] = av_clipf(src[i + 5], min, max);
            dst[i + 6] = av_clipf(src[i + 6], min, max);
            dst[i + 7] = av_clipf(src[i + 7], min, max);
        }
    }
}

4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280
static av_always_inline int float_to_int16_one(const float *src){
    int_fast32_t tmp = *(const int32_t*)src;
    if(tmp & 0xf0000){
        tmp = (0x43c0ffff - tmp)>>31;
        // is this faster on some gcc/cpu combinations?
//      if(tmp > 0x43c0ffff) tmp = 0xFFFF;
//      else                 tmp = 0;
    }
    return tmp - 0x8000;
}

Michael Niedermayer's avatar
Michael Niedermayer committed
4281
void ff_float_to_int16_c(int16_t *dst, const float *src, long len){
4282
    int i;
4283 4284 4285 4286
    for(i=0; i<len; i++)
        dst[i] = float_to_int16_one(src+i);
}

4287
void ff_float_to_int16_interleave_c(int16_t *dst, const float **src, long len, int channels){
4288 4289 4290
    int i,j,c;
    if(channels==2){
        for(i=0; i<len; i++){
4291 4292
            dst[2*i]   = float_to_int16_one(src[0]+i);
            dst[2*i+1] = float_to_int16_one(src[1]+i);
4293
        }
4294
    }else{
4295
        for(c=0; c<channels; c++)
4296
            for(i=0, j=c; i<len; i++, j+=channels)
4297
                dst[j] = float_to_int16_one(src[c]+i);
4298 4299 4300
    }
}

4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318 4319 4320 4321 4322
static void add_int16_c(int16_t * v1, int16_t * v2, int order)
{
    while (order--)
       *v1++ += *v2++;
}

static void sub_int16_c(int16_t * v1, int16_t * v2, int order)
{
    while (order--)
        *v1++ -= *v2++;
}

static int32_t scalarproduct_int16_c(int16_t * v1, int16_t * v2, int order, int shift)
{
    int res = 0;

    while (order--)
        res += (*v1++ * *v2++) >> shift;

    return res;
}

4323 4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394
#define W0 2048
#define W1 2841 /* 2048*sqrt (2)*cos (1*pi/16) */
#define W2 2676 /* 2048*sqrt (2)*cos (2*pi/16) */
#define W3 2408 /* 2048*sqrt (2)*cos (3*pi/16) */
#define W4 2048 /* 2048*sqrt (2)*cos (4*pi/16) */
#define W5 1609 /* 2048*sqrt (2)*cos (5*pi/16) */
#define W6 1108 /* 2048*sqrt (2)*cos (6*pi/16) */
#define W7 565  /* 2048*sqrt (2)*cos (7*pi/16) */

static void wmv2_idct_row(short * b)
{
    int s1,s2;
    int a0,a1,a2,a3,a4,a5,a6,a7;
    /*step 1*/
    a1 = W1*b[1]+W7*b[7];
    a7 = W7*b[1]-W1*b[7];
    a5 = W5*b[5]+W3*b[3];
    a3 = W3*b[5]-W5*b[3];
    a2 = W2*b[2]+W6*b[6];
    a6 = W6*b[2]-W2*b[6];
    a0 = W0*b[0]+W0*b[4];
    a4 = W0*b[0]-W0*b[4];
    /*step 2*/
    s1 = (181*(a1-a5+a7-a3)+128)>>8;//1,3,5,7,
    s2 = (181*(a1-a5-a7+a3)+128)>>8;
    /*step 3*/
    b[0] = (a0+a2+a1+a5 + (1<<7))>>8;
    b[1] = (a4+a6 +s1   + (1<<7))>>8;
    b[2] = (a4-a6 +s2   + (1<<7))>>8;
    b[3] = (a0-a2+a7+a3 + (1<<7))>>8;
    b[4] = (a0-a2-a7-a3 + (1<<7))>>8;
    b[5] = (a4-a6 -s2   + (1<<7))>>8;
    b[6] = (a4+a6 -s1   + (1<<7))>>8;
    b[7] = (a0+a2-a1-a5 + (1<<7))>>8;
}
static void wmv2_idct_col(short * b)
{
    int s1,s2;
    int a0,a1,a2,a3,a4,a5,a6,a7;
    /*step 1, with extended precision*/
    a1 = (W1*b[8*1]+W7*b[8*7] + 4)>>3;
    a7 = (W7*b[8*1]-W1*b[8*7] + 4)>>3;
    a5 = (W5*b[8*5]+W3*b[8*3] + 4)>>3;
    a3 = (W3*b[8*5]-W5*b[8*3] + 4)>>3;
    a2 = (W2*b[8*2]+W6*b[8*6] + 4)>>3;
    a6 = (W6*b[8*2]-W2*b[8*6] + 4)>>3;
    a0 = (W0*b[8*0]+W0*b[8*4]    )>>3;
    a4 = (W0*b[8*0]-W0*b[8*4]    )>>3;
    /*step 2*/
    s1 = (181*(a1-a5+a7-a3)+128)>>8;
    s2 = (181*(a1-a5-a7+a3)+128)>>8;
    /*step 3*/
    b[8*0] = (a0+a2+a1+a5 + (1<<13))>>14;
    b[8*1] = (a4+a6 +s1   + (1<<13))>>14;
    b[8*2] = (a4-a6 +s2   + (1<<13))>>14;
    b[8*3] = (a0-a2+a7+a3 + (1<<13))>>14;

    b[8*4] = (a0-a2-a7-a3 + (1<<13))>>14;
    b[8*5] = (a4-a6 -s2   + (1<<13))>>14;
    b[8*6] = (a4+a6 -s1   + (1<<13))>>14;
    b[8*7] = (a0+a2-a1-a5 + (1<<13))>>14;
}
void ff_wmv2_idct_c(short * block){
    int i;

    for(i=0;i<64;i+=8){
        wmv2_idct_row(block+i);
    }
    for(i=0;i<8;i++){
        wmv2_idct_col(block+i);
    }
}
4395 4396
/* XXX: those functions should be suppressed ASAP when all IDCTs are
 converted */
4397 4398 4399 4400 4401 4402 4403 4404 4405 4406
static void ff_wmv2_idct_put_c(uint8_t *dest, int line_size, DCTELEM *block)
{
    ff_wmv2_idct_c(block);
    put_pixels_clamped_c(block, dest, line_size);
}
static void ff_wmv2_idct_add_c(uint8_t *dest, int line_size, DCTELEM *block)
{
    ff_wmv2_idct_c(block);
    add_pixels_clamped_c(block, dest, line_size);
}
4407 4408 4409 4410 4411 4412 4413 4414 4415 4416 4417
static void ff_jref_idct_put(uint8_t *dest, int line_size, DCTELEM *block)
{
    j_rev_dct (block);
    put_pixels_clamped_c(block, dest, line_size);
}
static void ff_jref_idct_add(uint8_t *dest, int line_size, DCTELEM *block)
{
    j_rev_dct (block);
    add_pixels_clamped_c(block, dest, line_size);
}

4418 4419 4420 4421 4422 4423 4424 4425 4426 4427 4428
static void ff_jref_idct4_put(uint8_t *dest, int line_size, DCTELEM *block)
{
    j_rev_dct4 (block);
    put_pixels_clamped4_c(block, dest, line_size);
}
static void ff_jref_idct4_add(uint8_t *dest, int line_size, DCTELEM *block)
{
    j_rev_dct4 (block);
    add_pixels_clamped4_c(block, dest, line_size);
}

4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439
static void ff_jref_idct2_put(uint8_t *dest, int line_size, DCTELEM *block)
{
    j_rev_dct2 (block);
    put_pixels_clamped2_c(block, dest, line_size);
}
static void ff_jref_idct2_add(uint8_t *dest, int line_size, DCTELEM *block)
{
    j_rev_dct2 (block);
    add_pixels_clamped2_c(block, dest, line_size);
}

4440 4441
static void ff_jref_idct1_put(uint8_t *dest, int line_size, DCTELEM *block)
{
4442
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
4443 4444 4445 4446 4447

    dest[0] = cm[(block[0] + 4)>>3];
}
static void ff_jref_idct1_add(uint8_t *dest, int line_size, DCTELEM *block)
{
4448
    uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
4449 4450 4451 4452

    dest[0] = cm[dest[0] + ((block[0] + 4)>>3)];
}

4453
static void just_return(void *mem av_unused, int stride av_unused, int h av_unused) { return; }
4454

Fabrice Bellard's avatar
Fabrice Bellard committed
4455
/* init static data */
Måns Rullgård's avatar
Måns Rullgård committed
4456
void dsputil_static_init(void)
4457
{
Michael Niedermayer's avatar
Michael Niedermayer committed
4458
    int i;
4459

4460
    for(i=0;i<256;i++) ff_cropTbl[i + MAX_NEG_CROP] = i;
Fabrice Bellard's avatar
Fabrice Bellard committed
4461
    for(i=0;i<MAX_NEG_CROP;i++) {
4462 4463
        ff_cropTbl[i] = 0;
        ff_cropTbl[i + MAX_NEG_CROP + 256] = 255;
Fabrice Bellard's avatar
Fabrice Bellard committed
4464
    }
4465

Fabrice Bellard's avatar
Fabrice Bellard committed
4466
    for(i=0;i<512;i++) {
4467
        ff_squareTbl[i] = (i - 256) * (i - 256);
Fabrice Bellard's avatar
Fabrice Bellard committed
4468
    }
4469

Måns Rullgård's avatar
Måns Rullgård committed
4470
    for(i=0; i<64; i++) inv_zigzag_direct16[ff_zigzag_direct[i]]= i+1;
Fabrice Bellard's avatar
Fabrice Bellard committed
4471
}
4472

4473 4474 4475 4476
int ff_check_alignment(void){
    static int did_fail=0;
    DECLARE_ALIGNED_16(int, aligned);

4477
    if((intptr_t)&aligned & 15){
4478
        if(!did_fail){
4479
#if HAVE_MMX || HAVE_ALTIVEC
4480
            av_log(NULL, AV_LOG_ERROR,
4481 4482
                "Compiler did not align stack variables. Libavcodec has been miscompiled\n"
                "and may be very slow or crash. This is not a bug in libavcodec,\n"
4483 4484
                "but in the compiler. You may try recompiling using gcc >= 4.2.\n"
                "Do not report crashes to FFmpeg developers.\n");
4485 4486 4487 4488 4489 4490 4491
#endif
            did_fail=1;
        }
        return -1;
    }
    return 0;
}
4492

Fabrice Bellard's avatar
Fabrice Bellard committed
4493 4494 4495
void dsputil_init(DSPContext* c, AVCodecContext *avctx)
{
    int i;
Fabrice Bellard's avatar
Fabrice Bellard committed
4496

4497 4498
    ff_check_alignment();

4499
#if CONFIG_ENCODERS
4500
    if(avctx->dct_algo==FF_DCT_FASTINT) {
4501
        c->fdct = fdct_ifast;
4502
        c->fdct248 = fdct_ifast248;
4503
    }
4504
    else if(avctx->dct_algo==FF_DCT_FAAN) {
4505
        c->fdct = ff_faandct;
4506
        c->fdct248 = ff_faandct248;
4507
    }
4508
    else {
4509
        c->fdct = ff_jpeg_fdct_islow; //slow/accurate/default
4510
        c->fdct248 = ff_fdct248_islow;
4511
    }
4512 4513
#endif //CONFIG_ENCODERS

4514
    if(avctx->lowres==1){
4515
        if(avctx->idct_algo==FF_IDCT_INT || avctx->idct_algo==FF_IDCT_AUTO || !CONFIG_H264_DECODER){
4516 4517 4518 4519 4520 4521
            c->idct_put= ff_jref_idct4_put;
            c->idct_add= ff_jref_idct4_add;
        }else{
            c->idct_put= ff_h264_lowres_idct_put_c;
            c->idct_add= ff_h264_lowres_idct_add_c;
        }
4522
        c->idct    = j_rev_dct4;
4523
        c->idct_permutation_type= FF_NO_IDCT_PERM;
4524 4525 4526 4527 4528
    }else if(avctx->lowres==2){
        c->idct_put= ff_jref_idct2_put;
        c->idct_add= ff_jref_idct2_add;
        c->idct    = j_rev_dct2;
        c->idct_permutation_type= FF_NO_IDCT_PERM;
4529 4530 4531 4532 4533
    }else if(avctx->lowres==3){
        c->idct_put= ff_jref_idct1_put;
        c->idct_add= ff_jref_idct1_add;
        c->idct    = j_rev_dct1;
        c->idct_permutation_type= FF_NO_IDCT_PERM;
4534 4535 4536 4537 4538 4539
    }else{
        if(avctx->idct_algo==FF_IDCT_INT){
            c->idct_put= ff_jref_idct_put;
            c->idct_add= ff_jref_idct_add;
            c->idct    = j_rev_dct;
            c->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM;
4540
        }else if((CONFIG_VP3_DECODER || CONFIG_VP5_DECODER || CONFIG_VP6_DECODER ) &&
4541
                avctx->idct_algo==FF_IDCT_VP3){
4542 4543 4544 4545
            c->idct_put= ff_vp3_idct_put_c;
            c->idct_add= ff_vp3_idct_add_c;
            c->idct    = ff_vp3_idct_c;
            c->idct_permutation_type= FF_NO_IDCT_PERM;
4546 4547 4548 4549 4550
        }else if(avctx->idct_algo==FF_IDCT_WMV2){
            c->idct_put= ff_wmv2_idct_put_c;
            c->idct_add= ff_wmv2_idct_add_c;
            c->idct    = ff_wmv2_idct_c;
            c->idct_permutation_type= FF_NO_IDCT_PERM;
4551 4552 4553 4554 4555
        }else if(avctx->idct_algo==FF_IDCT_FAAN){
            c->idct_put= ff_faanidct_put;
            c->idct_add= ff_faanidct_add;
            c->idct    = ff_faanidct;
            c->idct_permutation_type= FF_NO_IDCT_PERM;
4556
        }else if(CONFIG_EATGQ_DECODER && avctx->idct_algo==FF_IDCT_EA) {
4557 4558
            c->idct_put= ff_ea_idct_put_c;
            c->idct_permutation_type= FF_NO_IDCT_PERM;
4559
        }else{ //accurate/default
4560 4561 4562
            c->idct_put= ff_simple_idct_put;
            c->idct_add= ff_simple_idct_add;
            c->idct    = ff_simple_idct;
4563 4564
            c->idct_permutation_type= FF_NO_IDCT_PERM;
        }
4565 4566
    }

4567
    if (CONFIG_H264_DECODER) {
Panagiotis Issaris's avatar
Panagiotis Issaris committed
4568 4569 4570 4571
        c->h264_idct_add= ff_h264_idct_add_c;
        c->h264_idct8_add= ff_h264_idct8_add_c;
        c->h264_idct_dc_add= ff_h264_idct_dc_add_c;
        c->h264_idct8_dc_add= ff_h264_idct8_dc_add_c;
4572 4573 4574 4575
        c->h264_idct_add16     = ff_h264_idct_add16_c;
        c->h264_idct8_add4     = ff_h264_idct8_add4_c;
        c->h264_idct_add8      = ff_h264_idct_add8_c;
        c->h264_idct_add16intra= ff_h264_idct_add16intra_c;
4576
    }
4577

4578 4579 4580
    c->get_pixels = get_pixels_c;
    c->diff_pixels = diff_pixels_c;
    c->put_pixels_clamped = put_pixels_clamped_c;
4581
    c->put_signed_pixels_clamped = put_signed_pixels_clamped_c;
4582
    c->add_pixels_clamped = add_pixels_clamped_c;
Loren Merritt's avatar
Loren Merritt committed
4583 4584
    c->add_pixels8 = add_pixels8_c;
    c->add_pixels4 = add_pixels4_c;
4585
    c->sum_abs_dctelem = sum_abs_dctelem_c;
4586
    c->gmc1 = gmc1_c;
4587
    c->gmc = ff_gmc_c;
Loren Merritt's avatar
Loren Merritt committed
4588
    c->clear_block = clear_block_c;
4589 4590 4591 4592
    c->clear_blocks = clear_blocks_c;
    c->pix_sum = pix_sum_c;
    c->pix_norm1 = pix_norm1_c;

4593
    /* TODO [0] 16  [1] 8 */
4594 4595 4596 4597 4598 4599 4600 4601
    c->pix_abs[0][0] = pix_abs16_c;
    c->pix_abs[0][1] = pix_abs16_x2_c;
    c->pix_abs[0][2] = pix_abs16_y2_c;
    c->pix_abs[0][3] = pix_abs16_xy2_c;
    c->pix_abs[1][0] = pix_abs8_c;
    c->pix_abs[1][1] = pix_abs8_x2_c;
    c->pix_abs[1][2] = pix_abs8_y2_c;
    c->pix_abs[1][3] = pix_abs8_xy2_c;
4602

4603 4604 4605 4606 4607 4608 4609 4610 4611 4612
#define dspfunc(PFX, IDX, NUM) \
    c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## NUM ## _c;     \
    c->PFX ## _pixels_tab[IDX][1] = PFX ## _pixels ## NUM ## _x2_c;  \
    c->PFX ## _pixels_tab[IDX][2] = PFX ## _pixels ## NUM ## _y2_c;  \
    c->PFX ## _pixels_tab[IDX][3] = PFX ## _pixels ## NUM ## _xy2_c

    dspfunc(put, 0, 16);
    dspfunc(put_no_rnd, 0, 16);
    dspfunc(put, 1, 8);
    dspfunc(put_no_rnd, 1, 8);
4613 4614
    dspfunc(put, 2, 4);
    dspfunc(put, 3, 2);
4615 4616 4617 4618 4619

    dspfunc(avg, 0, 16);
    dspfunc(avg_no_rnd, 0, 16);
    dspfunc(avg, 1, 8);
    dspfunc(avg_no_rnd, 1, 8);
4620 4621
    dspfunc(avg, 2, 4);
    dspfunc(avg, 3, 2);
4622 4623
#undef dspfunc

4624 4625 4626
    c->put_no_rnd_pixels_l2[0]= put_no_rnd_pixels16_l2_c;
    c->put_no_rnd_pixels_l2[1]= put_no_rnd_pixels8_l2_c;

4627 4628 4629 4630 4631 4632 4633 4634 4635 4636
    c->put_tpel_pixels_tab[ 0] = put_tpel_pixels_mc00_c;
    c->put_tpel_pixels_tab[ 1] = put_tpel_pixels_mc10_c;
    c->put_tpel_pixels_tab[ 2] = put_tpel_pixels_mc20_c;
    c->put_tpel_pixels_tab[ 4] = put_tpel_pixels_mc01_c;
    c->put_tpel_pixels_tab[ 5] = put_tpel_pixels_mc11_c;
    c->put_tpel_pixels_tab[ 6] = put_tpel_pixels_mc21_c;
    c->put_tpel_pixels_tab[ 8] = put_tpel_pixels_mc02_c;
    c->put_tpel_pixels_tab[ 9] = put_tpel_pixels_mc12_c;
    c->put_tpel_pixels_tab[10] = put_tpel_pixels_mc22_c;

4637 4638 4639 4640 4641 4642 4643 4644 4645 4646
    c->avg_tpel_pixels_tab[ 0] = avg_tpel_pixels_mc00_c;
    c->avg_tpel_pixels_tab[ 1] = avg_tpel_pixels_mc10_c;
    c->avg_tpel_pixels_tab[ 2] = avg_tpel_pixels_mc20_c;
    c->avg_tpel_pixels_tab[ 4] = avg_tpel_pixels_mc01_c;
    c->avg_tpel_pixels_tab[ 5] = avg_tpel_pixels_mc11_c;
    c->avg_tpel_pixels_tab[ 6] = avg_tpel_pixels_mc21_c;
    c->avg_tpel_pixels_tab[ 8] = avg_tpel_pixels_mc02_c;
    c->avg_tpel_pixels_tab[ 9] = avg_tpel_pixels_mc12_c;
    c->avg_tpel_pixels_tab[10] = avg_tpel_pixels_mc22_c;

4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675
#define dspfunc(PFX, IDX, NUM) \
    c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_c; \
    c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_c; \
    c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_c; \
    c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_c; \
    c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_c; \
    c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_c; \
    c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_c; \
    c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_c; \
    c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_c; \
    c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_c; \
    c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_c; \
    c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_c; \
    c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_c; \
    c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_c; \
    c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_c; \
    c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_c

    dspfunc(put_qpel, 0, 16);
    dspfunc(put_no_rnd_qpel, 0, 16);

    dspfunc(avg_qpel, 0, 16);
    /* dspfunc(avg_no_rnd_qpel, 0, 16); */

    dspfunc(put_qpel, 1, 8);
    dspfunc(put_no_rnd_qpel, 1, 8);

    dspfunc(avg_qpel, 1, 8);
    /* dspfunc(avg_no_rnd_qpel, 1, 8); */
4676 4677 4678 4679

    dspfunc(put_h264_qpel, 0, 16);
    dspfunc(put_h264_qpel, 1, 8);
    dspfunc(put_h264_qpel, 2, 4);
4680
    dspfunc(put_h264_qpel, 3, 2);
4681 4682 4683 4684
    dspfunc(avg_h264_qpel, 0, 16);
    dspfunc(avg_h264_qpel, 1, 8);
    dspfunc(avg_h264_qpel, 2, 4);

4685
#undef dspfunc
4686 4687 4688 4689 4690 4691
    c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_c;
    c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_c;
    c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_c;
    c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_c;
    c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_c;
    c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_c;
4692
    c->put_no_rnd_vc1_chroma_pixels_tab[0]= put_no_rnd_vc1_chroma_mc8_c;
4693
    c->avg_no_rnd_vc1_chroma_pixels_tab[0]= avg_no_rnd_vc1_chroma_mc8_c;
4694

4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715
    c->weight_h264_pixels_tab[0]= weight_h264_pixels16x16_c;
    c->weight_h264_pixels_tab[1]= weight_h264_pixels16x8_c;
    c->weight_h264_pixels_tab[2]= weight_h264_pixels8x16_c;
    c->weight_h264_pixels_tab[3]= weight_h264_pixels8x8_c;
    c->weight_h264_pixels_tab[4]= weight_h264_pixels8x4_c;
    c->weight_h264_pixels_tab[5]= weight_h264_pixels4x8_c;
    c->weight_h264_pixels_tab[6]= weight_h264_pixels4x4_c;
    c->weight_h264_pixels_tab[7]= weight_h264_pixels4x2_c;
    c->weight_h264_pixels_tab[8]= weight_h264_pixels2x4_c;
    c->weight_h264_pixels_tab[9]= weight_h264_pixels2x2_c;
    c->biweight_h264_pixels_tab[0]= biweight_h264_pixels16x16_c;
    c->biweight_h264_pixels_tab[1]= biweight_h264_pixels16x8_c;
    c->biweight_h264_pixels_tab[2]= biweight_h264_pixels8x16_c;
    c->biweight_h264_pixels_tab[3]= biweight_h264_pixels8x8_c;
    c->biweight_h264_pixels_tab[4]= biweight_h264_pixels8x4_c;
    c->biweight_h264_pixels_tab[5]= biweight_h264_pixels4x8_c;
    c->biweight_h264_pixels_tab[6]= biweight_h264_pixels4x4_c;
    c->biweight_h264_pixels_tab[7]= biweight_h264_pixels4x2_c;
    c->biweight_h264_pixels_tab[8]= biweight_h264_pixels2x4_c;
    c->biweight_h264_pixels_tab[9]= biweight_h264_pixels2x2_c;

4716 4717
    c->draw_edges = draw_edges_c;

4718
#if CONFIG_CAVS_DECODER
4719
    ff_cavsdsp_init(c,avctx);
4720
#endif
4721 4722 4723 4724

#if CONFIG_MLP_DECODER || CONFIG_TRUEHD_DECODER
    ff_mlp_init(c, avctx);
#endif
4725
#if CONFIG_VC1_DECODER
4726 4727
    ff_vc1dsp_init(c,avctx);
#endif
4728
#if CONFIG_WMV2_DECODER || CONFIG_VC1_DECODER
4729 4730
    ff_intrax8dsp_init(c,avctx);
#endif
4731
#if CONFIG_RV30_DECODER
Kostya Shishkov's avatar
Kostya Shishkov committed
4732 4733
    ff_rv30dsp_init(c,avctx);
#endif
4734
#if CONFIG_RV40_DECODER
Kostya Shishkov's avatar
Kostya Shishkov committed
4735 4736 4737 4738 4739 4740
    ff_rv40dsp_init(c,avctx);
    c->put_rv40_qpel_pixels_tab[0][15] = put_rv40_qpel16_mc33_c;
    c->avg_rv40_qpel_pixels_tab[0][15] = avg_rv40_qpel16_mc33_c;
    c->put_rv40_qpel_pixels_tab[1][15] = put_rv40_qpel8_mc33_c;
    c->avg_rv40_qpel_pixels_tab[1][15] = avg_rv40_qpel8_mc33_c;
#endif
4741

Michael Niedermayer's avatar
Michael Niedermayer committed
4742 4743 4744 4745 4746 4747 4748 4749
    c->put_mspel_pixels_tab[0]= put_mspel8_mc00_c;
    c->put_mspel_pixels_tab[1]= put_mspel8_mc10_c;
    c->put_mspel_pixels_tab[2]= put_mspel8_mc20_c;
    c->put_mspel_pixels_tab[3]= put_mspel8_mc30_c;
    c->put_mspel_pixels_tab[4]= put_mspel8_mc02_c;
    c->put_mspel_pixels_tab[5]= put_mspel8_mc12_c;
    c->put_mspel_pixels_tab[6]= put_mspel8_mc22_c;
    c->put_mspel_pixels_tab[7]= put_mspel8_mc32_c;
4750

4751 4752 4753
#define SET_CMP_FUNC(name) \
    c->name[0]= name ## 16_c;\
    c->name[1]= name ## 8x8_c;
4754

4755
    SET_CMP_FUNC(hadamard8_diff)
4756
    c->hadamard8_diff[4]= hadamard8_intra16_c;
4757
    c->hadamard8_diff[5]= hadamard8_intra8x8_c;
4758
    SET_CMP_FUNC(dct_sad)
4759
    SET_CMP_FUNC(dct_max)
4760
#if CONFIG_GPL
4761
    SET_CMP_FUNC(dct264_sad)
Mike Melanson's avatar
Mike Melanson committed
4762
#endif
4763 4764 4765 4766
    c->sad[0]= pix_abs16_c;
    c->sad[1]= pix_abs8_c;
    c->sse[0]= sse16_c;
    c->sse[1]= sse8_c;
4767
    c->sse[2]= sse4_c;
4768 4769 4770
    SET_CMP_FUNC(quant_psnr)
    SET_CMP_FUNC(rd)
    SET_CMP_FUNC(bit)
4771 4772
    c->vsad[0]= vsad16_c;
    c->vsad[4]= vsad_intra16_c;
4773
    c->vsad[5]= vsad_intra8_c;
4774 4775
    c->vsse[0]= vsse16_c;
    c->vsse[4]= vsse_intra16_c;
4776
    c->vsse[5]= vsse_intra8_c;
4777 4778
    c->nsse[0]= nsse16_c;
    c->nsse[1]= nsse8_c;
4779
#if CONFIG_SNOW_ENCODER
4780 4781 4782 4783
    c->w53[0]= w53_16_c;
    c->w53[1]= w53_8_c;
    c->w97[0]= w97_16_c;
    c->w97[1]= w97_8_c;
4784
#endif
4785

4786 4787
    c->ssd_int8_vs_int16 = ssd_int8_vs_int16_c;

Michael Niedermayer's avatar
Michael Niedermayer committed
4788
    c->add_bytes= add_bytes_c;
4789
    c->add_bytes_l2= add_bytes_l2_c;
Michael Niedermayer's avatar
Michael Niedermayer committed
4790
    c->diff_bytes= diff_bytes_c;
4791
    c->add_hfyu_median_prediction= add_hfyu_median_prediction_c;
4792
    c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_c;
4793 4794
    c->add_hfyu_left_prediction  = add_hfyu_left_prediction_c;
    c->add_hfyu_left_prediction_bgr32 = add_hfyu_left_prediction_bgr32_c;
Michael Niedermayer's avatar
Michael Niedermayer committed
4795
    c->bswap_buf= bswap_buf;
4796
#if CONFIG_PNG_DECODER
4797 4798
    c->add_png_paeth_prediction= ff_add_png_paeth_prediction;
#endif
4799 4800 4801

    c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_c;
    c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_c;
4802 4803
    c->h264_v_loop_filter_luma_intra= h264_v_loop_filter_luma_intra_c;
    c->h264_h_loop_filter_luma_intra= h264_h_loop_filter_luma_intra_c;
4804 4805
    c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_c;
    c->h264_h_loop_filter_chroma= h264_h_loop_filter_chroma_c;
4806 4807
    c->h264_v_loop_filter_chroma_intra= h264_v_loop_filter_chroma_intra_c;
    c->h264_h_loop_filter_chroma_intra= h264_h_loop_filter_chroma_intra_c;
4808
    c->h264_loop_filter_strength= NULL;
4809

4810
    if (CONFIG_ANY_H263) {
Aurelien Jacobs's avatar
Aurelien Jacobs committed
4811 4812
        c->h263_h_loop_filter= h263_h_loop_filter_c;
        c->h263_v_loop_filter= h263_v_loop_filter_c;
4813
    }
4814

4815
    if (CONFIG_VP3_DECODER) {
4816 4817 4818
        c->vp3_h_loop_filter= ff_vp3_h_loop_filter_c;
        c->vp3_v_loop_filter= ff_vp3_v_loop_filter_c;
    }
4819 4820 4821
    if (CONFIG_VP6_DECODER) {
        c->vp6_filter_diag4= ff_vp6_filter_diag4_c;
    }
4822

4823
    c->h261_loop_filter= h261_loop_filter_c;
4824

4825 4826
    c->try_8x8basis= try_8x8basis_c;
    c->add_8x8basis= add_8x8basis_c;
Michael Niedermayer's avatar
Michael Niedermayer committed
4827

4828
#if CONFIG_SNOW_DECODER
4829 4830 4831
    c->vertical_compose97i = ff_snow_vertical_compose97i;
    c->horizontal_compose97i = ff_snow_horizontal_compose97i;
    c->inner_add_yblock = ff_snow_inner_add_yblock;
4832
#endif
4833

4834
#if CONFIG_VORBIS_DECODER
4835
    c->vorbis_inverse_coupling = vorbis_inverse_coupling;
4836
#endif
4837
#if CONFIG_AC3_DECODER
Loren Merritt's avatar
Loren Merritt committed
4838 4839
    c->ac3_downmix = ff_ac3_downmix_c;
#endif
4840
#if CONFIG_LPC
4841
    c->lpc_compute_autocorr = ff_lpc_compute_autocorr;
4842
#endif
4843 4844
    c->vector_fmul = vector_fmul_c;
    c->vector_fmul_reverse = vector_fmul_reverse_c;
4845
    c->vector_fmul_add = vector_fmul_add_c;
4846
    c->vector_fmul_window = ff_vector_fmul_window_c;
Loren Merritt's avatar
Loren Merritt committed
4847
    c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_c;
4848
    c->vector_clipf = vector_clipf_c;
4849
    c->float_to_int16 = ff_float_to_int16_c;
4850
    c->float_to_int16_interleave = ff_float_to_int16_interleave_c;
4851 4852 4853
    c->add_int16 = add_int16_c;
    c->sub_int16 = sub_int16_c;
    c->scalarproduct_int16 = scalarproduct_int16_c;
4854 4855 4856 4857 4858 4859 4860 4861 4862
    c->scalarproduct_float = scalarproduct_float_c;
    c->butterflies_float = butterflies_float_c;
    c->vector_fmul_scalar = vector_fmul_scalar_c;

    c->vector_fmul_sv_scalar[0] = vector_fmul_sv_scalar_2_c;
    c->vector_fmul_sv_scalar[1] = vector_fmul_sv_scalar_4_c;

    c->sv_fmul_scalar[0] = sv_fmul_scalar_2_c;
    c->sv_fmul_scalar[1] = sv_fmul_scalar_4_c;
4863

4864 4865 4866 4867 4868
    c->shrink[0]= ff_img_copy_plane;
    c->shrink[1]= ff_shrink22;
    c->shrink[2]= ff_shrink44;
    c->shrink[3]= ff_shrink88;

4869 4870
    c->prefetch= just_return;

4871 4872 4873
    memset(c->put_2tap_qpel_pixels_tab, 0, sizeof(c->put_2tap_qpel_pixels_tab));
    memset(c->avg_2tap_qpel_pixels_tab, 0, sizeof(c->avg_2tap_qpel_pixels_tab));

4874 4875 4876 4877 4878 4879 4880 4881 4882
    if (HAVE_MMX)        dsputil_init_mmx   (c, avctx);
    if (ARCH_ARM)        dsputil_init_arm   (c, avctx);
    if (CONFIG_MLIB)     dsputil_init_mlib  (c, avctx);
    if (HAVE_VIS)        dsputil_init_vis   (c, avctx);
    if (ARCH_ALPHA)      dsputil_init_alpha (c, avctx);
    if (ARCH_PPC)        dsputil_init_ppc   (c, avctx);
    if (HAVE_MMI)        dsputil_init_mmi   (c, avctx);
    if (ARCH_SH4)        dsputil_init_sh4   (c, avctx);
    if (ARCH_BFIN)       dsputil_init_bfin  (c, avctx);
4883

4884 4885 4886 4887 4888 4889 4890
    for(i=0; i<64; i++){
        if(!c->put_2tap_qpel_pixels_tab[0][i])
            c->put_2tap_qpel_pixels_tab[0][i]= c->put_h264_qpel_pixels_tab[0][i];
        if(!c->avg_2tap_qpel_pixels_tab[0][i])
            c->avg_2tap_qpel_pixels_tab[0][i]= c->avg_h264_qpel_pixels_tab[0][i];
    }

4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907
    switch(c->idct_permutation_type){
    case FF_NO_IDCT_PERM:
        for(i=0; i<64; i++)
            c->idct_permutation[i]= i;
        break;
    case FF_LIBMPEG2_IDCT_PERM:
        for(i=0; i<64; i++)
            c->idct_permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2);
        break;
    case FF_SIMPLE_IDCT_PERM:
        for(i=0; i<64; i++)
            c->idct_permutation[i]= simple_mmx_permutation[i];
        break;
    case FF_TRANSPOSE_IDCT_PERM:
        for(i=0; i<64; i++)
            c->idct_permutation[i]= ((i&7)<<3) | (i>>3);
        break;
4908 4909 4910 4911
    case FF_PARTTRANS_IDCT_PERM:
        for(i=0; i<64; i++)
            c->idct_permutation[i]= (i&0x24) | ((i&3)<<3) | ((i>>3)&3);
        break;
4912 4913 4914 4915
    case FF_SSE2_IDCT_PERM:
        for(i=0; i<64; i++)
            c->idct_permutation[i]= (i&0x38) | idct_sse2_row_perm[i&7];
        break;
4916
    default:
4917
        av_log(avctx, AV_LOG_ERROR, "Internal error, IDCT permutation not set\n");
4918
    }
4919
}
4920