libavcodec/vc1dsp.c

   1 /*
   2  * VC-1 and WMV3 decoder - DSP functions
   3  * Copyright (c) 2006 Konstantin Shishkov
   4  *
   5  * This file is part of FFmpeg.
   6  *
   7  * FFmpeg is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * FFmpeg is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with FFmpeg; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 /**
  23 * @file libavcodec/vc1dsp.c
  24  * VC-1 and WMV3 decoder
  25  *
  26  */
  27
  28 #include "dsputil.h"
  29
  30
  31 /** Apply overlap transform to horizontal edge
  32 */
  33 static void vc1_v_overlap_c(uint8_t* src, int stride)
  34 {
  35     int i;
  36     int a, b, c, d;
  37     int d1, d2;
  38     int rnd = 1;
  39     for(i = 0; i < 8; i++) {
  40         a = src[-2*stride];
  41         b = src[-stride];
  42         c = src[0];
  43         d = src[stride];
  44         d1 = (a - d + 3 + rnd) >> 3;
  45         d2 = (a - d + b - c + 4 - rnd) >> 3;
  46
  47         src[-2*stride] = a - d1;
  48         src[-stride] = av_clip_uint8(b - d2);
  49         src[0] = av_clip_uint8(c + d2);
  50         src[stride] = d + d1;
  51         src++;
  52         rnd = !rnd;
  53     }
  54 }
  55
  56 /** Apply overlap transform to vertical edge
  57 */
  58 static void vc1_h_overlap_c(uint8_t* src, int stride)
  59 {
  60     int i;
  61     int a, b, c, d;
  62     int d1, d2;
  63     int rnd = 1;
  64     for(i = 0; i < 8; i++) {
  65         a = src[-2];
  66         b = src[-1];
  67         c = src[0];
  68         d = src[1];
  69         d1 = (a - d + 3 + rnd) >> 3;
  70         d2 = (a - d + b - c + 4 - rnd) >> 3;
  71
  72         src[-2] = a - d1;
  73         src[-1] = av_clip_uint8(b - d2);
  74         src[0] = av_clip_uint8(c + d2);
  75         src[1] = d + d1;
  76         src += stride;
  77         rnd = !rnd;
  78     }
  79 }
  80
  81 /**
  82  * VC-1 in-loop deblocking filter for one line
  83  * @param src source block type
  84  * @param stride block stride
  85  * @param pq block quantizer
  86  * @return whether other 3 pairs should be filtered or not
  87  * @see 8.6
  88  */
  89 static av_always_inline int vc1_filter_line(uint8_t* src, int stride, int pq){
  90     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
  91
  92     int a0 = (2*(src[-2*stride] - src[ 1*stride]) - 5*(src[-1*stride] - src[ 0*stride]) + 4) >> 3;
  93     int a0_sign = a0 >> 31;        /* Store sign */
  94     a0 = (a0 ^ a0_sign) - a0_sign; /* a0 = FFABS(a0); */
  95     if(a0 < pq){
  96         int a1 = FFABS((2*(src[-4*stride] - src[-1*stride]) - 5*(src[-3*stride] - src[-2*stride]) + 4) >> 3);
  97         int a2 = FFABS((2*(src[ 0*stride] - src[ 3*stride]) - 5*(src[ 1*stride] - src[ 2*stride]) + 4) >> 3);
  98         if(a1 < a0 || a2 < a0){
  99             int clip = src[-1*stride] - src[ 0*stride];
 100             int clip_sign = clip >> 31;
 101             clip = ((clip ^ clip_sign) - clip_sign)>>1;
 102             if(clip){
 103                 int a3 = FFMIN(a1, a2);
 104                 int d = 5 * (a3 - a0);
 105                 int d_sign = (d >> 31);
 106                 d = ((d ^ d_sign) - d_sign) >> 3;
 107                 d_sign ^= a0_sign;
 108
 109                 if( d_sign ^ clip_sign )
 110                     d = 0;
 111                 else{
 112                     d = FFMIN(d, clip);
 113                     d = (d ^ d_sign) - d_sign;          /* Restore sign */
 114                     src[-1*stride] = cm[src[-1*stride] - d];
 115                     src[ 0*stride] = cm[src[ 0*stride] + d];
 116                 }
 117                 return 1;
 118             }
 119         }
 120     }
 121     return 0;
 122 }
 123
 124 /**
 125  * VC-1 in-loop deblocking filter
 126  * @param src source block type
 127  * @param step distance between horizontally adjacent elements
 128  * @param stride distance between vertically adjacent elements
 129  * @param len edge length to filter (4 or 8 pixels)
 130  * @param pq block quantizer
 131  * @see 8.6
 132  */
 133 static inline void vc1_loop_filter(uint8_t* src, int step, int stride, int len, int pq)
 134 {
 135     int i;
 136     int filt3;
 137
 138     for(i = 0; i < len; i += 4){
 139         filt3 = vc1_filter_line(src + 2*step, stride, pq);
 140         if(filt3){
 141             vc1_filter_line(src + 0*step, stride, pq);
 142             vc1_filter_line(src + 1*step, stride, pq);
 143             vc1_filter_line(src + 3*step, stride, pq);
 144         }
 145         src += step * 4;
 146     }
 147 }
 148
 149 static void vc1_v_loop_filter4_c(uint8_t *src, int stride, int pq)
 150 {
 151     vc1_loop_filter(src, 1, stride, 4, pq);
 152 }
 153
 154 static void vc1_h_loop_filter4_c(uint8_t *src, int stride, int pq)
 155 {
 156     vc1_loop_filter(src, stride, 1, 4, pq);
 157 }
 158
 159 static void vc1_v_loop_filter8_c(uint8_t *src, int stride, int pq)
 160 {
 161     vc1_loop_filter(src, 1, stride, 8, pq);
 162 }
 163
 164 static void vc1_h_loop_filter8_c(uint8_t *src, int stride, int pq)
 165 {
 166     vc1_loop_filter(src, stride, 1, 8, pq);
 167 }
 168
 169 static void vc1_v_loop_filter16_c(uint8_t *src, int stride, int pq)
 170 {
 171     vc1_loop_filter(src, 1, stride, 16, pq);
 172 }
 173
 174 static void vc1_h_loop_filter16_c(uint8_t *src, int stride, int pq)
 175 {
 176     vc1_loop_filter(src, stride, 1, 16, pq);
 177 }
 178
 179 /** Do inverse transform on 8x8 block
 180 */
 181 static void vc1_inv_trans_8x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
 182 {
 183     int i;
 184     int dc = block[0];
 185     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
 186     dc = (3 * dc +  1) >> 1;
 187     dc = (3 * dc + 16) >> 5;
 188     for(i = 0; i < 8; i++){
 189         dest[0] = cm[dest[0]+dc];
 190         dest[1] = cm[dest[1]+dc];
 191         dest[2] = cm[dest[2]+dc];
 192         dest[3] = cm[dest[3]+dc];
 193         dest[4] = cm[dest[4]+dc];
 194         dest[5] = cm[dest[5]+dc];
 195         dest[6] = cm[dest[6]+dc];
 196         dest[7] = cm[dest[7]+dc];
 197         dest += linesize;
 198     }
 199 }
 200
 201 static void vc1_inv_trans_8x8_c(DCTELEM block[64])
 202 {
 203     int i;
 204     register int t1,t2,t3,t4,t5,t6,t7,t8;
 205     DCTELEM *src, *dst;
 206
 207     src = block;
 208     dst = block;
 209     for(i = 0; i < 8; i++){
 210         t1 = 12 * (src[0] + src[4]) + 4;
 211         t2 = 12 * (src[0] - src[4]) + 4;
 212         t3 = 16 * src[2] +  6 * src[6];
 213         t4 =  6 * src[2] - 16 * src[6];
 214
 215         t5 = t1 + t3;
 216         t6 = t2 + t4;
 217         t7 = t2 - t4;
 218         t8 = t1 - t3;
 219
 220         t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];
 221         t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];
 222         t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];
 223         t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];
 224
 225         dst[0] = (t5 + t1) >> 3;
 226         dst[1] = (t6 + t2) >> 3;
 227         dst[2] = (t7 + t3) >> 3;
 228         dst[3] = (t8 + t4) >> 3;
 229         dst[4] = (t8 - t4) >> 3;
 230         dst[5] = (t7 - t3) >> 3;
 231         dst[6] = (t6 - t2) >> 3;
 232         dst[7] = (t5 - t1) >> 3;
 233
 234         src += 8;
 235         dst += 8;
 236     }
 237
 238     src = block;
 239     dst = block;
 240     for(i = 0; i < 8; i++){
 241         t1 = 12 * (src[ 0] + src[32]) + 64;
 242         t2 = 12 * (src[ 0] - src[32]) + 64;
 243         t3 = 16 * src[16] +  6 * src[48];
 244         t4 =  6 * src[16] - 16 * src[48];
 245
 246         t5 = t1 + t3;
 247         t6 = t2 + t4;
 248         t7 = t2 - t4;
 249         t8 = t1 - t3;
 250
 251         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
 252         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
 253         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
 254         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
 255
 256         dst[ 0] = (t5 + t1) >> 7;
 257         dst[ 8] = (t6 + t2) >> 7;
 258         dst[16] = (t7 + t3) >> 7;
 259         dst[24] = (t8 + t4) >> 7;
 260         dst[32] = (t8 - t4 + 1) >> 7;
 261         dst[40] = (t7 - t3 + 1) >> 7;
 262         dst[48] = (t6 - t2 + 1) >> 7;
 263         dst[56] = (t5 - t1 + 1) >> 7;
 264
 265         src++;
 266         dst++;
 267     }
 268 }
 269
 270 /** Do inverse transform on 8x4 part of block
 271 */
 272 static void vc1_inv_trans_8x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
 273 {
 274     int i;
 275     int dc = block[0];
 276     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
 277     dc = ( 3 * dc +  1) >> 1;
 278     dc = (17 * dc + 64) >> 7;
 279     for(i = 0; i < 4; i++){
 280         dest[0] = cm[dest[0]+dc];
 281         dest[1] = cm[dest[1]+dc];
 282         dest[2] = cm[dest[2]+dc];
 283         dest[3] = cm[dest[3]+dc];
 284         dest[4] = cm[dest[4]+dc];
 285         dest[5] = cm[dest[5]+dc];
 286         dest[6] = cm[dest[6]+dc];
 287         dest[7] = cm[dest[7]+dc];
 288         dest += linesize;
 289     }
 290 }
 291
 292 static void vc1_inv_trans_8x4_c(uint8_t *dest, int linesize, DCTELEM *block)
 293 {
 294     int i;
 295     register int t1,t2,t3,t4,t5,t6,t7,t8;
 296     DCTELEM *src, *dst;
 297     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
 298
 299     src = block;
 300     dst = block;
 301     for(i = 0; i < 4; i++){
 302         t1 = 12 * (src[0] + src[4]) + 4;
 303         t2 = 12 * (src[0] - src[4]) + 4;
 304         t3 = 16 * src[2] +  6 * src[6];
 305         t4 =  6 * src[2] - 16 * src[6];
 306
 307         t5 = t1 + t3;
 308         t6 = t2 + t4;
 309         t7 = t2 - t4;
 310         t8 = t1 - t3;
 311
 312         t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];
 313         t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];
 314         t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];
 315         t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];
 316
 317         dst[0] = (t5 + t1) >> 3;
 318         dst[1] = (t6 + t2) >> 3;
 319         dst[2] = (t7 + t3) >> 3;
 320         dst[3] = (t8 + t4) >> 3;
 321         dst[4] = (t8 - t4) >> 3;
 322         dst[5] = (t7 - t3) >> 3;
 323         dst[6] = (t6 - t2) >> 3;
 324         dst[7] = (t5 - t1) >> 3;
 325
 326         src += 8;
 327         dst += 8;
 328     }
 329
 330     src = block;
 331     for(i = 0; i < 8; i++){
 332         t1 = 17 * (src[ 0] + src[16]) + 64;
 333         t2 = 17 * (src[ 0] - src[16]) + 64;
 334         t3 = 22 * src[ 8] + 10 * src[24];
 335         t4 = 22 * src[24] - 10 * src[ 8];
 336
 337         dest[0*linesize] = cm[dest[0*linesize] + ((t1 + t3) >> 7)];
 338         dest[1*linesize] = cm[dest[1*linesize] + ((t2 - t4) >> 7)];
 339         dest[2*linesize] = cm[dest[2*linesize] + ((t2 + t4) >> 7)];
 340         dest[3*linesize] = cm[dest[3*linesize] + ((t1 - t3) >> 7)];
 341
 342         src ++;
 343         dest++;
 344     }
 345 }
 346
 347 /** Do inverse transform on 4x8 parts of block
 348 */
 349 static void vc1_inv_trans_4x8_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
 350 {
 351     int i;
 352     int dc = block[0];
 353     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
 354     dc = (17 * dc +  4) >> 3;
 355     dc = (12 * dc + 64) >> 7;
 356     for(i = 0; i < 8; i++){
 357         dest[0] = cm[dest[0]+dc];
 358         dest[1] = cm[dest[1]+dc];
 359         dest[2] = cm[dest[2]+dc];
 360         dest[3] = cm[dest[3]+dc];
 361         dest += linesize;
 362     }
 363 }
 364
 365 static void vc1_inv_trans_4x8_c(uint8_t *dest, int linesize, DCTELEM *block)
 366 {
 367     int i;
 368     register int t1,t2,t3,t4,t5,t6,t7,t8;
 369     DCTELEM *src, *dst;
 370     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
 371
 372     src = block;
 373     dst = block;
 374     for(i = 0; i < 8; i++){
 375         t1 = 17 * (src[0] + src[2]) + 4;
 376         t2 = 17 * (src[0] - src[2]) + 4;
 377         t3 = 22 * src[1] + 10 * src[3];
 378         t4 = 22 * src[3] - 10 * src[1];
 379
 380         dst[0] = (t1 + t3) >> 3;
 381         dst[1] = (t2 - t4) >> 3;
 382         dst[2] = (t2 + t4) >> 3;
 383         dst[3] = (t1 - t3) >> 3;
 384
 385         src += 8;
 386         dst += 8;
 387     }
 388
 389     src = block;
 390     for(i = 0; i < 4; i++){
 391         t1 = 12 * (src[ 0] + src[32]) + 64;
 392         t2 = 12 * (src[ 0] - src[32]) + 64;
 393         t3 = 16 * src[16] +  6 * src[48];
 394         t4 =  6 * src[16] - 16 * src[48];
 395
 396         t5 = t1 + t3;
 397         t6 = t2 + t4;
 398         t7 = t2 - t4;
 399         t8 = t1 - t3;
 400
 401         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
 402         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
 403         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
 404         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
 405
 406         dest[0*linesize] = cm[dest[0*linesize] + ((t5 + t1) >> 7)];
 407         dest[1*linesize] = cm[dest[1*linesize] + ((t6 + t2) >> 7)];
 408         dest[2*linesize] = cm[dest[2*linesize] + ((t7 + t3) >> 7)];
 409         dest[3*linesize] = cm[dest[3*linesize] + ((t8 + t4) >> 7)];
 410         dest[4*linesize] = cm[dest[4*linesize] + ((t8 - t4 + 1) >> 7)];
 411         dest[5*linesize] = cm[dest[5*linesize] + ((t7 - t3 + 1) >> 7)];
 412         dest[6*linesize] = cm[dest[6*linesize] + ((t6 - t2 + 1) >> 7)];
 413         dest[7*linesize] = cm[dest[7*linesize] + ((t5 - t1 + 1) >> 7)];
 414
 415         src ++;
 416         dest++;
 417     }
 418 }
 419
 420 /** Do inverse transform on 4x4 part of block
 421 */
 422 static void vc1_inv_trans_4x4_dc_c(uint8_t *dest, int linesize, DCTELEM *block)
 423 {
 424     int i;
 425     int dc = block[0];
 426     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
 427     dc = (17 * dc +  4) >> 3;
 428     dc = (17 * dc + 64) >> 7;
 429     for(i = 0; i < 4; i++){
 430         dest[0] = cm[dest[0]+dc];
 431         dest[1] = cm[dest[1]+dc];
 432         dest[2] = cm[dest[2]+dc];
 433         dest[3] = cm[dest[3]+dc];
 434         dest += linesize;
 435     }
 436 }
 437
 438 static void vc1_inv_trans_4x4_c(uint8_t *dest, int linesize, DCTELEM *block)
 439 {
 440     int i;
 441     register int t1,t2,t3,t4;
 442     DCTELEM *src, *dst;
 443     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
 444
 445     src = block;
 446     dst = block;
 447     for(i = 0; i < 4; i++){
 448         t1 = 17 * (src[0] + src[2]) + 4;
 449         t2 = 17 * (src[0] - src[2]) + 4;
 450         t3 = 22 * src[1] + 10 * src[3];
 451         t4 = 22 * src[3] - 10 * src[1];
 452
 453         dst[0] = (t1 + t3) >> 3;
 454         dst[1] = (t2 - t4) >> 3;
 455         dst[2] = (t2 + t4) >> 3;
 456         dst[3] = (t1 - t3) >> 3;
 457
 458         src += 8;
 459         dst += 8;
 460     }
 461
 462     src = block;
 463     for(i = 0; i < 4; i++){
 464         t1 = 17 * (src[ 0] + src[16]) + 64;
 465         t2 = 17 * (src[ 0] - src[16]) + 64;
 466         t3 = 22 * src[ 8] + 10 * src[24];
 467         t4 = 22 * src[24] - 10 * src[ 8];
 468
 469         dest[0*linesize] = cm[dest[0*linesize] + ((t1 + t3) >> 7)];
 470         dest[1*linesize] = cm[dest[1*linesize] + ((t2 - t4) >> 7)];
 471         dest[2*linesize] = cm[dest[2*linesize] + ((t2 + t4) >> 7)];
 472         dest[3*linesize] = cm[dest[3*linesize] + ((t1 - t3) >> 7)];
 473
 474         src ++;
 475         dest++;
 476     }
 477 }
 478
 479 /* motion compensation functions */
 480 /** Filter in case of 2 filters */
 481 #define VC1_MSPEL_FILTER_16B(DIR, TYPE)                                 \
 482 static av_always_inline int vc1_mspel_ ## DIR ## _filter_16bits(const TYPE *src, int stride, int mode) \
 483 {                                                                       \
 484     switch(mode){                                                       \
 485     case 0: /* no shift - should not occur */                           \
 486         return 0;                                                       \
 487     case 1: /* 1/4 shift */                                             \
 488         return -4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2]; \
 489     case 2: /* 1/2 shift */                                             \
 490         return -src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2]; \
 491     case 3: /* 3/4 shift */                                             \
 492         return -3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2]; \
 493     }                                                                   \
 494     return 0; /* should not occur */                                    \
 495 }
 496
 497 VC1_MSPEL_FILTER_16B(ver, uint8_t);
 498 VC1_MSPEL_FILTER_16B(hor, int16_t);
 499
 500
 501 /** Filter used to interpolate fractional pel values
 502  */
 503 static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int mode, int r)
 504 {
 505     switch(mode){
 506     case 0: //no shift
 507         return src[0];
 508     case 1: // 1/4 shift
 509         return (-4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2] + 32 - r) >> 6;
 510     case 2: // 1/2 shift
 511         return (-src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2] + 8 - r) >> 4;
 512     case 3: // 3/4 shift
 513         return (-3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2] + 32 - r) >> 6;
 514     }
 515     return 0; //should not occur
 516 }
 517
 518 /** Function used to do motion compensation with bicubic interpolation
 519  */
 520 #define VC1_MSPEL_MC(OP, OPNAME)\
 521 static void OPNAME ## vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)\
 522 {\
 523     int     i, j;\
 524 \
 525     if (vmode) { /* Horizontal filter to apply */\
 526         int r;\
 527 \
 528         if (hmode) { /* Vertical filter to apply, output to tmp */\
 529             static const int shift_value[] = { 0, 5, 1, 5 };\
 530             int              shift = (shift_value[hmode]+shift_value[vmode])>>1;\
 531             int16_t          tmp[11*8], *tptr = tmp;\
 532 \
 533             r = (1<<(shift-1)) + rnd-1;\
 534 \
 535             src -= 1;\
 536             for(j = 0; j < 8; j++) {\
 537                 for(i = 0; i < 11; i++)\
 538                     tptr[i] = (vc1_mspel_ver_filter_16bits(src + i, stride, vmode)+r)>>shift;\
 539                 src += stride;\
 540                 tptr += 11;\
 541             }\
 542 \
 543             r = 64-rnd;\
 544             tptr = tmp+1;\
 545             for(j = 0; j < 8; j++) {\
 546                 for(i = 0; i < 8; i++)\
 547                     OP(dst[i], (vc1_mspel_hor_filter_16bits(tptr + i, 1, hmode)+r)>>7);\
 548                 dst += stride;\
 549                 tptr += 11;\
 550             }\
 551 \
 552             return;\
 553         }\
 554         else { /* No horizontal filter, output 8 lines to dst */\
 555             r = 1-rnd;\
 556 \
 557             for(j = 0; j < 8; j++) {\
 558                 for(i = 0; i < 8; i++)\
 559                     OP(dst[i], vc1_mspel_filter(src + i, stride, vmode, r));\
 560                 src += stride;\
 561                 dst += stride;\
 562             }\
 563             return;\
 564         }\
 565     }\
 566 \
 567     /* Horizontal mode with no vertical mode */\
 568     for(j = 0; j < 8; j++) {\
 569         for(i = 0; i < 8; i++)\
 570             OP(dst[i], vc1_mspel_filter(src + i, 1, hmode, rnd));\
 571         dst += stride;\
 572         src += stride;\
 573     }\
 574 }
 575
 576 #define op_put(a, b) a = av_clip_uint8(b)
 577 #define op_avg(a, b) a = (a + av_clip_uint8(b) + 1) >> 1
 578
 579 VC1_MSPEL_MC(op_put, put_)
 580 VC1_MSPEL_MC(op_avg, avg_)
 581
 582 /* pixel functions - really are entry points to vc1_mspel_mc */
 583
 584 /* this one is defined in dsputil.c */
 585 void ff_put_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd);
 586 void ff_avg_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd);
 587
 588 #define PUT_VC1_MSPEL(a, b)\
 589 static void put_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
 590      put_vc1_mspel_mc(dst, src, stride, a, b, rnd);                         \
 591 }\
 592 static void avg_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
 593      avg_vc1_mspel_mc(dst, src, stride, a, b, rnd);                         \
 594 }
 595
 596 PUT_VC1_MSPEL(1, 0)
 597 PUT_VC1_MSPEL(2, 0)
 598 PUT_VC1_MSPEL(3, 0)
 599
 600 PUT_VC1_MSPEL(0, 1)
 601 PUT_VC1_MSPEL(1, 1)
 602 PUT_VC1_MSPEL(2, 1)
 603 PUT_VC1_MSPEL(3, 1)
 604
 605 PUT_VC1_MSPEL(0, 2)
 606 PUT_VC1_MSPEL(1, 2)
 607 PUT_VC1_MSPEL(2, 2)
 608 PUT_VC1_MSPEL(3, 2)
 609
 610 PUT_VC1_MSPEL(0, 3)
 611 PUT_VC1_MSPEL(1, 3)
 612 PUT_VC1_MSPEL(2, 3)
 613 PUT_VC1_MSPEL(3, 3)
 614
 615 void ff_vc1dsp_init(DSPContext* dsp, AVCodecContext *avctx) {
 616     dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;
 617     dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;
 618     dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;
 619     dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
 620     dsp->vc1_inv_trans_8x8_dc = vc1_inv_trans_8x8_dc_c;
 621     dsp->vc1_inv_trans_4x8_dc = vc1_inv_trans_4x8_dc_c;
 622     dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_c;
 623     dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_c;
 624     dsp->vc1_h_overlap = vc1_h_overlap_c;
 625     dsp->vc1_v_overlap = vc1_v_overlap_c;
 626     dsp->vc1_v_loop_filter4 = vc1_v_loop_filter4_c;
 627     dsp->vc1_h_loop_filter4 = vc1_h_loop_filter4_c;
 628     dsp->vc1_v_loop_filter8 = vc1_v_loop_filter8_c;
 629     dsp->vc1_h_loop_filter8 = vc1_h_loop_filter8_c;
 630     dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_c;
 631     dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_c;
 632
 633     dsp->put_vc1_mspel_pixels_tab[ 0] = ff_put_vc1_mspel_mc00_c;
 634     dsp->put_vc1_mspel_pixels_tab[ 1] = put_vc1_mspel_mc10_c;
 635     dsp->put_vc1_mspel_pixels_tab[ 2] = put_vc1_mspel_mc20_c;
 636     dsp->put_vc1_mspel_pixels_tab[ 3] = put_vc1_mspel_mc30_c;
 637     dsp->put_vc1_mspel_pixels_tab[ 4] = put_vc1_mspel_mc01_c;
 638     dsp->put_vc1_mspel_pixels_tab[ 5] = put_vc1_mspel_mc11_c;
 639     dsp->put_vc1_mspel_pixels_tab[ 6] = put_vc1_mspel_mc21_c;
 640     dsp->put_vc1_mspel_pixels_tab[ 7] = put_vc1_mspel_mc31_c;
 641     dsp->put_vc1_mspel_pixels_tab[ 8] = put_vc1_mspel_mc02_c;
 642     dsp->put_vc1_mspel_pixels_tab[ 9] = put_vc1_mspel_mc12_c;
 643     dsp->put_vc1_mspel_pixels_tab[10] = put_vc1_mspel_mc22_c;
 644     dsp->put_vc1_mspel_pixels_tab[11] = put_vc1_mspel_mc32_c;
 645     dsp->put_vc1_mspel_pixels_tab[12] = put_vc1_mspel_mc03_c;
 646     dsp->put_vc1_mspel_pixels_tab[13] = put_vc1_mspel_mc13_c;
 647     dsp->put_vc1_mspel_pixels_tab[14] = put_vc1_mspel_mc23_c;
 648     dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_c;
 649
 650     dsp->avg_vc1_mspel_pixels_tab[ 0] = ff_avg_vc1_mspel_mc00_c;
 651     dsp->avg_vc1_mspel_pixels_tab[ 1] = avg_vc1_mspel_mc10_c;
 652     dsp->avg_vc1_mspel_pixels_tab[ 2] = avg_vc1_mspel_mc20_c;
 653     dsp->avg_vc1_mspel_pixels_tab[ 3] = avg_vc1_mspel_mc30_c;
 654     dsp->avg_vc1_mspel_pixels_tab[ 4] = avg_vc1_mspel_mc01_c;
 655     dsp->avg_vc1_mspel_pixels_tab[ 5] = avg_vc1_mspel_mc11_c;
 656     dsp->avg_vc1_mspel_pixels_tab[ 6] = avg_vc1_mspel_mc21_c;
 657     dsp->avg_vc1_mspel_pixels_tab[ 7] = avg_vc1_mspel_mc31_c;
 658     dsp->avg_vc1_mspel_pixels_tab[ 8] = avg_vc1_mspel_mc02_c;
 659     dsp->avg_vc1_mspel_pixels_tab[ 9] = avg_vc1_mspel_mc12_c;
 660     dsp->avg_vc1_mspel_pixels_tab[10] = avg_vc1_mspel_mc22_c;
 661     dsp->avg_vc1_mspel_pixels_tab[11] = avg_vc1_mspel_mc32_c;
 662     dsp->avg_vc1_mspel_pixels_tab[12] = avg_vc1_mspel_mc03_c;
 663     dsp->avg_vc1_mspel_pixels_tab[13] = avg_vc1_mspel_mc13_c;
 664     dsp->avg_vc1_mspel_pixels_tab[14] = avg_vc1_mspel_mc23_c;
 665     dsp->avg_vc1_mspel_pixels_tab[15] = avg_vc1_mspel_mc33_c;
 666 }