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 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] = b - d2;
  49         src[0] = 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] = b - d2;
  74         src[0] = c + d2;
  75         src[1] = d + d1;
  76         src += stride;
  77         rnd = !rnd;
  78     }
  79 }
  80
  81
  82 /** Do inverse transform on 8x8 block
  83 */
  84 static void vc1_inv_trans_8x8_c(DCTELEM block[64])
  85 {
  86     int i;
  87     register int t1,t2,t3,t4,t5,t6,t7,t8;
  88     DCTELEM *src, *dst;
  89
  90     src = block;
  91     dst = block;
  92     for(i = 0; i < 8; i++){
  93         t1 = 12 * (src[0] + src[4]);
  94         t2 = 12 * (src[0] - src[4]);
  95         t3 = 16 * src[2] +  6 * src[6];
  96         t4 =  6 * src[2] - 16 * src[6];
  97
  98         t5 = t1 + t3;
  99         t6 = t2 + t4;
 100         t7 = t2 - t4;
 101         t8 = t1 - t3;
 102
 103         t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];
 104         t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];
 105         t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];
 106         t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];
 107
 108         dst[0] = (t5 + t1 + 4) >> 3;
 109         dst[1] = (t6 + t2 + 4) >> 3;
 110         dst[2] = (t7 + t3 + 4) >> 3;
 111         dst[3] = (t8 + t4 + 4) >> 3;
 112         dst[4] = (t8 - t4 + 4) >> 3;
 113         dst[5] = (t7 - t3 + 4) >> 3;
 114         dst[6] = (t6 - t2 + 4) >> 3;
 115         dst[7] = (t5 - t1 + 4) >> 3;
 116
 117         src += 8;
 118         dst += 8;
 119     }
 120
 121     src = block;
 122     dst = block;
 123     for(i = 0; i < 8; i++){
 124         t1 = 12 * (src[ 0] + src[32]);
 125         t2 = 12 * (src[ 0] - src[32]);
 126         t3 = 16 * src[16] +  6 * src[48];
 127         t4 =  6 * src[16] - 16 * src[48];
 128
 129         t5 = t1 + t3;
 130         t6 = t2 + t4;
 131         t7 = t2 - t4;
 132         t8 = t1 - t3;
 133
 134         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
 135         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
 136         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
 137         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
 138
 139         dst[ 0] = (t5 + t1 + 64) >> 7;
 140         dst[ 8] = (t6 + t2 + 64) >> 7;
 141         dst[16] = (t7 + t3 + 64) >> 7;
 142         dst[24] = (t8 + t4 + 64) >> 7;
 143         dst[32] = (t8 - t4 + 64 + 1) >> 7;
 144         dst[40] = (t7 - t3 + 64 + 1) >> 7;
 145         dst[48] = (t6 - t2 + 64 + 1) >> 7;
 146         dst[56] = (t5 - t1 + 64 + 1) >> 7;
 147
 148         src++;
 149         dst++;
 150     }
 151 }
 152
 153 /** Do inverse transform on 8x4 part of block
 154 */
 155 static void vc1_inv_trans_8x4_c(DCTELEM block[64], int n)
 156 {
 157     int i;
 158     register int t1,t2,t3,t4,t5,t6,t7,t8;
 159     DCTELEM *src, *dst;
 160     int off;
 161
 162     off = n * 32;
 163     src = block + off;
 164     dst = block + off;
 165     for(i = 0; i < 4; i++){
 166         t1 = 12 * (src[0] + src[4]);
 167         t2 = 12 * (src[0] - src[4]);
 168         t3 = 16 * src[2] +  6 * src[6];
 169         t4 =  6 * src[2] - 16 * src[6];
 170
 171         t5 = t1 + t3;
 172         t6 = t2 + t4;
 173         t7 = t2 - t4;
 174         t8 = t1 - t3;
 175
 176         t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];
 177         t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];
 178         t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];
 179         t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];
 180
 181         dst[0] = (t5 + t1 + 4) >> 3;
 182         dst[1] = (t6 + t2 + 4) >> 3;
 183         dst[2] = (t7 + t3 + 4) >> 3;
 184         dst[3] = (t8 + t4 + 4) >> 3;
 185         dst[4] = (t8 - t4 + 4) >> 3;
 186         dst[5] = (t7 - t3 + 4) >> 3;
 187         dst[6] = (t6 - t2 + 4) >> 3;
 188         dst[7] = (t5 - t1 + 4) >> 3;
 189
 190         src += 8;
 191         dst += 8;
 192     }
 193
 194     src = block + off;
 195     dst = block + off;
 196     for(i = 0; i < 8; i++){
 197         t1 = 17 * (src[ 0] + src[16]);
 198         t2 = 17 * (src[ 0] - src[16]);
 199         t3 = 22 * src[ 8];
 200         t4 = 22 * src[24];
 201         t5 = 10 * src[ 8];
 202         t6 = 10 * src[24];
 203
 204         dst[ 0] = (t1 + t3 + t6 + 64) >> 7;
 205         dst[ 8] = (t2 - t4 + t5 + 64) >> 7;
 206         dst[16] = (t2 + t4 - t5 + 64) >> 7;
 207         dst[24] = (t1 - t3 - t6 + 64) >> 7;
 208
 209         src ++;
 210         dst ++;
 211     }
 212 }
 213
 214 /** Do inverse transform on 4x8 parts of block
 215 */
 216 static void vc1_inv_trans_4x8_c(DCTELEM block[64], int n)
 217 {
 218     int i;
 219     register int t1,t2,t3,t4,t5,t6,t7,t8;
 220     DCTELEM *src, *dst;
 221     int off;
 222
 223     off = n * 4;
 224     src = block + off;
 225     dst = block + off;
 226     for(i = 0; i < 8; i++){
 227         t1 = 17 * (src[0] + src[2]);
 228         t2 = 17 * (src[0] - src[2]);
 229         t3 = 22 * src[1];
 230         t4 = 22 * src[3];
 231         t5 = 10 * src[1];
 232         t6 = 10 * src[3];
 233
 234         dst[0] = (t1 + t3 + t6 + 4) >> 3;
 235         dst[1] = (t2 - t4 + t5 + 4) >> 3;
 236         dst[2] = (t2 + t4 - t5 + 4) >> 3;
 237         dst[3] = (t1 - t3 - t6 + 4) >> 3;
 238
 239         src += 8;
 240         dst += 8;
 241     }
 242
 243     src = block + off;
 244     dst = block + off;
 245     for(i = 0; i < 4; i++){
 246         t1 = 12 * (src[ 0] + src[32]);
 247         t2 = 12 * (src[ 0] - src[32]);
 248         t3 = 16 * src[16] +  6 * src[48];
 249         t4 =  6 * src[16] - 16 * src[48];
 250
 251         t5 = t1 + t3;
 252         t6 = t2 + t4;
 253         t7 = t2 - t4;
 254         t8 = t1 - t3;
 255
 256         t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];
 257         t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];
 258         t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];
 259         t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];
 260
 261         dst[ 0] = (t5 + t1 + 64) >> 7;
 262         dst[ 8] = (t6 + t2 + 64) >> 7;
 263         dst[16] = (t7 + t3 + 64) >> 7;
 264         dst[24] = (t8 + t4 + 64) >> 7;
 265         dst[32] = (t8 - t4 + 64 + 1) >> 7;
 266         dst[40] = (t7 - t3 + 64 + 1) >> 7;
 267         dst[48] = (t6 - t2 + 64 + 1) >> 7;
 268         dst[56] = (t5 - t1 + 64 + 1) >> 7;
 269
 270         src++;
 271         dst++;
 272     }
 273 }
 274
 275 /** Do inverse transform on 4x4 part of block
 276 */
 277 static void vc1_inv_trans_4x4_c(DCTELEM block[64], int n)
 278 {
 279     int i;
 280     register int t1,t2,t3,t4,t5,t6;
 281     DCTELEM *src, *dst;
 282     int off;
 283
 284     off = (n&1) * 4 + (n&2) * 16;
 285     src = block + off;
 286     dst = block + off;
 287     for(i = 0; i < 4; i++){
 288         t1 = 17 * (src[0] + src[2]);
 289         t2 = 17 * (src[0] - src[2]);
 290         t3 = 22 * src[1];
 291         t4 = 22 * src[3];
 292         t5 = 10 * src[1];
 293         t6 = 10 * src[3];
 294
 295         dst[0] = (t1 + t3 + t6 + 4) >> 3;
 296         dst[1] = (t2 - t4 + t5 + 4) >> 3;
 297         dst[2] = (t2 + t4 - t5 + 4) >> 3;
 298         dst[3] = (t1 - t3 - t6 + 4) >> 3;
 299
 300         src += 8;
 301         dst += 8;
 302     }
 303
 304     src = block + off;
 305     dst = block + off;
 306     for(i = 0; i < 4; i++){
 307         t1 = 17 * (src[ 0] + src[16]);
 308         t2 = 17 * (src[ 0] - src[16]);
 309         t3 = 22 * src[ 8];
 310         t4 = 22 * src[24];
 311         t5 = 10 * src[ 8];
 312         t6 = 10 * src[24];
 313
 314         dst[ 0] = (t1 + t3 + t6 + 64) >> 7;
 315         dst[ 8] = (t2 - t4 + t5 + 64) >> 7;
 316         dst[16] = (t2 + t4 - t5 + 64) >> 7;
 317         dst[24] = (t1 - t3 - t6 + 64) >> 7;
 318
 319         src ++;
 320         dst ++;
 321     }
 322 }
 323
 324 /* motion compensation functions */
 325
 326 /** Filter used to interpolate fractional pel values
 327  */
 328 static av_always_inline int vc1_mspel_filter(const uint8_t *src, int stride, int mode, int r)
 329 {
 330     switch(mode){
 331     case 0: //no shift
 332         return src[0];
 333     case 1: // 1/4 shift
 334         return (-4*src[-stride] + 53*src[0] + 18*src[stride] - 3*src[stride*2] + 32 - r) >> 6;
 335     case 2: // 1/2 shift
 336         return (-src[-stride] + 9*src[0] + 9*src[stride] - src[stride*2] + 8 - r) >> 4;
 337     case 3: // 3/4 shift
 338         return (-3*src[-stride] + 18*src[0] + 53*src[stride] - 4*src[stride*2] + 32 - r) >> 6;
 339     }
 340     return 0; //should not occur
 341 }
 342
 343 /** Function used to do motion compensation with bicubic interpolation
 344  */
 345 static void vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride, int hmode, int vmode, int rnd)
 346 {
 347     int i, j;
 348     uint8_t tmp[8*11], *tptr;
 349     int r;
 350
 351     r = rnd;
 352     src -= stride;
 353     tptr = tmp;
 354     for(j = 0; j < 11; j++) {
 355         for(i = 0; i < 8; i++)
 356             tptr[i] = av_clip_uint8(vc1_mspel_filter(src + i, 1, hmode, r));
 357         src += stride;
 358         tptr += 8;
 359     }
 360     r = 1 - rnd;
 361
 362     tptr = tmp + 8;
 363     for(j = 0; j < 8; j++) {
 364         for(i = 0; i < 8; i++)
 365             dst[i] = av_clip_uint8(vc1_mspel_filter(tptr + i, 8, vmode, r));
 366         dst += stride;
 367         tptr += 8;
 368     }
 369 }
 370
 371 /* pixel functions - really are entry points to vc1_mspel_mc */
 372
 373 /* this one is defined in dsputil.c */
 374 void ff_put_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd);
 375
 376 #define PUT_VC1_MSPEL(a, b)\
 377 static void put_vc1_mspel_mc ## a ## b ##_c(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
 378      vc1_mspel_mc(dst, src, stride, a, b, rnd);                         \
 379 }
 380
 381 PUT_VC1_MSPEL(1, 0)
 382 PUT_VC1_MSPEL(2, 0)
 383 PUT_VC1_MSPEL(3, 0)
 384
 385 PUT_VC1_MSPEL(0, 1)
 386 PUT_VC1_MSPEL(1, 1)
 387 PUT_VC1_MSPEL(2, 1)
 388 PUT_VC1_MSPEL(3, 1)
 389
 390 PUT_VC1_MSPEL(0, 2)
 391 PUT_VC1_MSPEL(1, 2)
 392 PUT_VC1_MSPEL(2, 2)
 393 PUT_VC1_MSPEL(3, 2)
 394
 395 PUT_VC1_MSPEL(0, 3)
 396 PUT_VC1_MSPEL(1, 3)
 397 PUT_VC1_MSPEL(2, 3)
 398 PUT_VC1_MSPEL(3, 3)
 399
 400 void ff_vc1dsp_init(DSPContext* dsp, AVCodecContext *avctx) {
 401     dsp->vc1_inv_trans_8x8 = vc1_inv_trans_8x8_c;
 402     dsp->vc1_inv_trans_4x8 = vc1_inv_trans_4x8_c;
 403     dsp->vc1_inv_trans_8x4 = vc1_inv_trans_8x4_c;
 404     dsp->vc1_inv_trans_4x4 = vc1_inv_trans_4x4_c;
 405     dsp->vc1_h_overlap = vc1_h_overlap_c;
 406     dsp->vc1_v_overlap = vc1_v_overlap_c;
 407
 408     dsp->put_vc1_mspel_pixels_tab[ 0] = ff_put_vc1_mspel_mc00_c;
 409     dsp->put_vc1_mspel_pixels_tab[ 1] = put_vc1_mspel_mc10_c;
 410     dsp->put_vc1_mspel_pixels_tab[ 2] = put_vc1_mspel_mc20_c;
 411     dsp->put_vc1_mspel_pixels_tab[ 3] = put_vc1_mspel_mc30_c;
 412     dsp->put_vc1_mspel_pixels_tab[ 4] = put_vc1_mspel_mc01_c;
 413     dsp->put_vc1_mspel_pixels_tab[ 5] = put_vc1_mspel_mc11_c;
 414     dsp->put_vc1_mspel_pixels_tab[ 6] = put_vc1_mspel_mc21_c;
 415     dsp->put_vc1_mspel_pixels_tab[ 7] = put_vc1_mspel_mc31_c;
 416     dsp->put_vc1_mspel_pixels_tab[ 8] = put_vc1_mspel_mc02_c;
 417     dsp->put_vc1_mspel_pixels_tab[ 9] = put_vc1_mspel_mc12_c;
 418     dsp->put_vc1_mspel_pixels_tab[10] = put_vc1_mspel_mc22_c;
 419     dsp->put_vc1_mspel_pixels_tab[11] = put_vc1_mspel_mc32_c;
 420     dsp->put_vc1_mspel_pixels_tab[12] = put_vc1_mspel_mc03_c;
 421     dsp->put_vc1_mspel_pixels_tab[13] = put_vc1_mspel_mc13_c;
 422     dsp->put_vc1_mspel_pixels_tab[14] = put_vc1_mspel_mc23_c;
 423     dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_c;
 424 }