| 1 | /* |
| 2 | * RV40 decoder |
| 3 | * Copyright (c) 2007 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 |
| 24 | * RV40 decoder |
| 25 | */ |
| 26 | |
| 27 | #include "libavutil/imgutils.h" |
| 28 | |
| 29 | #include "avcodec.h" |
| 30 | #include "mpegutils.h" |
| 31 | #include "mpegvideo.h" |
| 32 | #include "golomb.h" |
| 33 | |
| 34 | #include "rv34.h" |
| 35 | #include "rv40vlc2.h" |
| 36 | #include "rv40data.h" |
| 37 | |
| 38 | static VLC aic_top_vlc; |
| 39 | static VLC aic_mode1_vlc[AIC_MODE1_NUM], aic_mode2_vlc[AIC_MODE2_NUM]; |
| 40 | static VLC ptype_vlc[NUM_PTYPE_VLCS], btype_vlc[NUM_BTYPE_VLCS]; |
| 41 | |
| 42 | static const int16_t mode2_offs[] = { |
| 43 | 0, 614, 1222, 1794, 2410, 3014, 3586, 4202, 4792, 5382, 5966, 6542, |
| 44 | 7138, 7716, 8292, 8864, 9444, 10030, 10642, 11212, 11814 |
| 45 | }; |
| 46 | |
| 47 | /** |
| 48 | * Initialize all tables. |
| 49 | */ |
| 50 | static av_cold void rv40_init_tables(void) |
| 51 | { |
| 52 | int i; |
| 53 | static VLC_TYPE aic_table[1 << AIC_TOP_BITS][2]; |
| 54 | static VLC_TYPE aic_mode1_table[AIC_MODE1_NUM << AIC_MODE1_BITS][2]; |
| 55 | static VLC_TYPE aic_mode2_table[11814][2]; |
| 56 | static VLC_TYPE ptype_table[NUM_PTYPE_VLCS << PTYPE_VLC_BITS][2]; |
| 57 | static VLC_TYPE btype_table[NUM_BTYPE_VLCS << BTYPE_VLC_BITS][2]; |
| 58 | |
| 59 | aic_top_vlc.table = aic_table; |
| 60 | aic_top_vlc.table_allocated = 1 << AIC_TOP_BITS; |
| 61 | init_vlc(&aic_top_vlc, AIC_TOP_BITS, AIC_TOP_SIZE, |
| 62 | rv40_aic_top_vlc_bits, 1, 1, |
| 63 | rv40_aic_top_vlc_codes, 1, 1, INIT_VLC_USE_NEW_STATIC); |
| 64 | for(i = 0; i < AIC_MODE1_NUM; i++){ |
| 65 | // Every tenth VLC table is empty |
| 66 | if((i % 10) == 9) continue; |
| 67 | aic_mode1_vlc[i].table = &aic_mode1_table[i << AIC_MODE1_BITS]; |
| 68 | aic_mode1_vlc[i].table_allocated = 1 << AIC_MODE1_BITS; |
| 69 | init_vlc(&aic_mode1_vlc[i], AIC_MODE1_BITS, AIC_MODE1_SIZE, |
| 70 | aic_mode1_vlc_bits[i], 1, 1, |
| 71 | aic_mode1_vlc_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC); |
| 72 | } |
| 73 | for(i = 0; i < AIC_MODE2_NUM; i++){ |
| 74 | aic_mode2_vlc[i].table = &aic_mode2_table[mode2_offs[i]]; |
| 75 | aic_mode2_vlc[i].table_allocated = mode2_offs[i + 1] - mode2_offs[i]; |
| 76 | init_vlc(&aic_mode2_vlc[i], AIC_MODE2_BITS, AIC_MODE2_SIZE, |
| 77 | aic_mode2_vlc_bits[i], 1, 1, |
| 78 | aic_mode2_vlc_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC); |
| 79 | } |
| 80 | for(i = 0; i < NUM_PTYPE_VLCS; i++){ |
| 81 | ptype_vlc[i].table = &ptype_table[i << PTYPE_VLC_BITS]; |
| 82 | ptype_vlc[i].table_allocated = 1 << PTYPE_VLC_BITS; |
| 83 | ff_init_vlc_sparse(&ptype_vlc[i], PTYPE_VLC_BITS, PTYPE_VLC_SIZE, |
| 84 | ptype_vlc_bits[i], 1, 1, |
| 85 | ptype_vlc_codes[i], 1, 1, |
| 86 | ptype_vlc_syms, 1, 1, INIT_VLC_USE_NEW_STATIC); |
| 87 | } |
| 88 | for(i = 0; i < NUM_BTYPE_VLCS; i++){ |
| 89 | btype_vlc[i].table = &btype_table[i << BTYPE_VLC_BITS]; |
| 90 | btype_vlc[i].table_allocated = 1 << BTYPE_VLC_BITS; |
| 91 | ff_init_vlc_sparse(&btype_vlc[i], BTYPE_VLC_BITS, BTYPE_VLC_SIZE, |
| 92 | btype_vlc_bits[i], 1, 1, |
| 93 | btype_vlc_codes[i], 1, 1, |
| 94 | btype_vlc_syms, 1, 1, INIT_VLC_USE_NEW_STATIC); |
| 95 | } |
| 96 | } |
| 97 | |
| 98 | /** |
| 99 | * Get stored dimension from bitstream. |
| 100 | * |
| 101 | * If the width/height is the standard one then it's coded as a 3-bit index. |
| 102 | * Otherwise it is coded as escaped 8-bit portions. |
| 103 | */ |
| 104 | static int get_dimension(GetBitContext *gb, const int *dim) |
| 105 | { |
| 106 | int t = get_bits(gb, 3); |
| 107 | int val = dim[t]; |
| 108 | if(val < 0) |
| 109 | val = dim[get_bits1(gb) - val]; |
| 110 | if(!val){ |
| 111 | do{ |
| 112 | t = get_bits(gb, 8); |
| 113 | val += t << 2; |
| 114 | }while(t == 0xFF); |
| 115 | } |
| 116 | return val; |
| 117 | } |
| 118 | |
| 119 | /** |
| 120 | * Get encoded picture size - usually this is called from rv40_parse_slice_header. |
| 121 | */ |
| 122 | static void rv40_parse_picture_size(GetBitContext *gb, int *w, int *h) |
| 123 | { |
| 124 | *w = get_dimension(gb, rv40_standard_widths); |
| 125 | *h = get_dimension(gb, rv40_standard_heights); |
| 126 | } |
| 127 | |
| 128 | static int rv40_parse_slice_header(RV34DecContext *r, GetBitContext *gb, SliceInfo *si) |
| 129 | { |
| 130 | int mb_bits; |
| 131 | int w = r->s.width, h = r->s.height; |
| 132 | int mb_size; |
| 133 | |
| 134 | memset(si, 0, sizeof(SliceInfo)); |
| 135 | if(get_bits1(gb)) |
| 136 | return -1; |
| 137 | si->type = get_bits(gb, 2); |
| 138 | if(si->type == 1) si->type = 0; |
| 139 | si->quant = get_bits(gb, 5); |
| 140 | if(get_bits(gb, 2)) |
| 141 | return -1; |
| 142 | si->vlc_set = get_bits(gb, 2); |
| 143 | skip_bits1(gb); |
| 144 | si->pts = get_bits(gb, 13); |
| 145 | if(!si->type || !get_bits1(gb)) |
| 146 | rv40_parse_picture_size(gb, &w, &h); |
| 147 | if(av_image_check_size(w, h, 0, r->s.avctx) < 0) |
| 148 | return -1; |
| 149 | si->width = w; |
| 150 | si->height = h; |
| 151 | mb_size = ((w + 15) >> 4) * ((h + 15) >> 4); |
| 152 | mb_bits = ff_rv34_get_start_offset(gb, mb_size); |
| 153 | si->start = get_bits(gb, mb_bits); |
| 154 | |
| 155 | return 0; |
| 156 | } |
| 157 | |
| 158 | /** |
| 159 | * Decode 4x4 intra types array. |
| 160 | */ |
| 161 | static int rv40_decode_intra_types(RV34DecContext *r, GetBitContext *gb, int8_t *dst) |
| 162 | { |
| 163 | MpegEncContext *s = &r->s; |
| 164 | int i, j, k, v; |
| 165 | int A, B, C; |
| 166 | int pattern; |
| 167 | int8_t *ptr; |
| 168 | |
| 169 | for(i = 0; i < 4; i++, dst += r->intra_types_stride){ |
| 170 | if(!i && s->first_slice_line){ |
| 171 | pattern = get_vlc2(gb, aic_top_vlc.table, AIC_TOP_BITS, 1); |
| 172 | dst[0] = (pattern >> 2) & 2; |
| 173 | dst[1] = (pattern >> 1) & 2; |
| 174 | dst[2] = pattern & 2; |
| 175 | dst[3] = (pattern << 1) & 2; |
| 176 | continue; |
| 177 | } |
| 178 | ptr = dst; |
| 179 | for(j = 0; j < 4; j++){ |
| 180 | /* Coefficients are read using VLC chosen by the prediction pattern |
| 181 | * The first one (used for retrieving a pair of coefficients) is |
| 182 | * constructed from the top, top right and left coefficients |
| 183 | * The second one (used for retrieving only one coefficient) is |
| 184 | * top + 10 * left. |
| 185 | */ |
| 186 | A = ptr[-r->intra_types_stride + 1]; // it won't be used for the last coefficient in a row |
| 187 | B = ptr[-r->intra_types_stride]; |
| 188 | C = ptr[-1]; |
| 189 | pattern = A + (B << 4) + (C << 8); |
| 190 | for(k = 0; k < MODE2_PATTERNS_NUM; k++) |
| 191 | if(pattern == rv40_aic_table_index[k]) |
| 192 | break; |
| 193 | if(j < 3 && k < MODE2_PATTERNS_NUM){ //pattern is found, decoding 2 coefficients |
| 194 | v = get_vlc2(gb, aic_mode2_vlc[k].table, AIC_MODE2_BITS, 2); |
| 195 | *ptr++ = v/9; |
| 196 | *ptr++ = v%9; |
| 197 | j++; |
| 198 | }else{ |
| 199 | if(B != -1 && C != -1) |
| 200 | v = get_vlc2(gb, aic_mode1_vlc[B + C*10].table, AIC_MODE1_BITS, 1); |
| 201 | else{ // tricky decoding |
| 202 | v = 0; |
| 203 | switch(C){ |
| 204 | case -1: // code 0 -> 1, 1 -> 0 |
| 205 | if(B < 2) |
| 206 | v = get_bits1(gb) ^ 1; |
| 207 | break; |
| 208 | case 0: |
| 209 | case 2: // code 0 -> 2, 1 -> 0 |
| 210 | v = (get_bits1(gb) ^ 1) << 1; |
| 211 | break; |
| 212 | } |
| 213 | } |
| 214 | *ptr++ = v; |
| 215 | } |
| 216 | } |
| 217 | } |
| 218 | return 0; |
| 219 | } |
| 220 | |
| 221 | /** |
| 222 | * Decode macroblock information. |
| 223 | */ |
| 224 | static int rv40_decode_mb_info(RV34DecContext *r) |
| 225 | { |
| 226 | MpegEncContext *s = &r->s; |
| 227 | GetBitContext *gb = &s->gb; |
| 228 | int q, i; |
| 229 | int prev_type = 0; |
| 230 | int mb_pos = s->mb_x + s->mb_y * s->mb_stride; |
| 231 | |
| 232 | if(!r->s.mb_skip_run) { |
| 233 | r->s.mb_skip_run = svq3_get_ue_golomb(gb) + 1; |
| 234 | if(r->s.mb_skip_run > (unsigned)s->mb_num) |
| 235 | return -1; |
| 236 | } |
| 237 | |
| 238 | if(--r->s.mb_skip_run) |
| 239 | return RV34_MB_SKIP; |
| 240 | |
| 241 | if(r->avail_cache[6-4]){ |
| 242 | int blocks[RV34_MB_TYPES] = {0}; |
| 243 | int count = 0; |
| 244 | if(r->avail_cache[6-1]) |
| 245 | blocks[r->mb_type[mb_pos - 1]]++; |
| 246 | blocks[r->mb_type[mb_pos - s->mb_stride]]++; |
| 247 | if(r->avail_cache[6-2]) |
| 248 | blocks[r->mb_type[mb_pos - s->mb_stride + 1]]++; |
| 249 | if(r->avail_cache[6-5]) |
| 250 | blocks[r->mb_type[mb_pos - s->mb_stride - 1]]++; |
| 251 | for(i = 0; i < RV34_MB_TYPES; i++){ |
| 252 | if(blocks[i] > count){ |
| 253 | count = blocks[i]; |
| 254 | prev_type = i; |
| 255 | if(count>1) |
| 256 | break; |
| 257 | } |
| 258 | } |
| 259 | } else if (r->avail_cache[6-1]) |
| 260 | prev_type = r->mb_type[mb_pos - 1]; |
| 261 | |
| 262 | if(s->pict_type == AV_PICTURE_TYPE_P){ |
| 263 | prev_type = block_num_to_ptype_vlc_num[prev_type]; |
| 264 | q = get_vlc2(gb, ptype_vlc[prev_type].table, PTYPE_VLC_BITS, 1); |
| 265 | if(q < PBTYPE_ESCAPE) |
| 266 | return q; |
| 267 | q = get_vlc2(gb, ptype_vlc[prev_type].table, PTYPE_VLC_BITS, 1); |
| 268 | av_log(s->avctx, AV_LOG_ERROR, "Dquant for P-frame\n"); |
| 269 | }else{ |
| 270 | prev_type = block_num_to_btype_vlc_num[prev_type]; |
| 271 | q = get_vlc2(gb, btype_vlc[prev_type].table, BTYPE_VLC_BITS, 1); |
| 272 | if(q < PBTYPE_ESCAPE) |
| 273 | return q; |
| 274 | q = get_vlc2(gb, btype_vlc[prev_type].table, BTYPE_VLC_BITS, 1); |
| 275 | av_log(s->avctx, AV_LOG_ERROR, "Dquant for B-frame\n"); |
| 276 | } |
| 277 | return 0; |
| 278 | } |
| 279 | |
| 280 | enum RV40BlockPos{ |
| 281 | POS_CUR, |
| 282 | POS_TOP, |
| 283 | POS_LEFT, |
| 284 | POS_BOTTOM, |
| 285 | }; |
| 286 | |
| 287 | #define MASK_CUR 0x0001 |
| 288 | #define MASK_RIGHT 0x0008 |
| 289 | #define MASK_BOTTOM 0x0010 |
| 290 | #define MASK_TOP 0x1000 |
| 291 | #define MASK_Y_TOP_ROW 0x000F |
| 292 | #define MASK_Y_LAST_ROW 0xF000 |
| 293 | #define MASK_Y_LEFT_COL 0x1111 |
| 294 | #define MASK_Y_RIGHT_COL 0x8888 |
| 295 | #define MASK_C_TOP_ROW 0x0003 |
| 296 | #define MASK_C_LAST_ROW 0x000C |
| 297 | #define MASK_C_LEFT_COL 0x0005 |
| 298 | #define MASK_C_RIGHT_COL 0x000A |
| 299 | |
| 300 | static const int neighbour_offs_x[4] = { 0, 0, -1, 0 }; |
| 301 | static const int neighbour_offs_y[4] = { 0, -1, 0, 1 }; |
| 302 | |
| 303 | static void rv40_adaptive_loop_filter(RV34DSPContext *rdsp, |
| 304 | uint8_t *src, int stride, int dmode, |
| 305 | int lim_q1, int lim_p1, |
| 306 | int alpha, int beta, int beta2, |
| 307 | int chroma, int edge, int dir) |
| 308 | { |
| 309 | int filter_p1, filter_q1; |
| 310 | int strong; |
| 311 | int lims; |
| 312 | |
| 313 | strong = rdsp->rv40_loop_filter_strength[dir](src, stride, beta, beta2, |
| 314 | edge, &filter_p1, &filter_q1); |
| 315 | |
| 316 | lims = filter_p1 + filter_q1 + ((lim_q1 + lim_p1) >> 1) + 1; |
| 317 | |
| 318 | if (strong) { |
| 319 | rdsp->rv40_strong_loop_filter[dir](src, stride, alpha, |
| 320 | lims, dmode, chroma); |
| 321 | } else if (filter_p1 & filter_q1) { |
| 322 | rdsp->rv40_weak_loop_filter[dir](src, stride, 1, 1, alpha, beta, |
| 323 | lims, lim_q1, lim_p1); |
| 324 | } else if (filter_p1 | filter_q1) { |
| 325 | rdsp->rv40_weak_loop_filter[dir](src, stride, filter_p1, filter_q1, |
| 326 | alpha, beta, lims >> 1, lim_q1 >> 1, |
| 327 | lim_p1 >> 1); |
| 328 | } |
| 329 | } |
| 330 | |
| 331 | /** |
| 332 | * RV40 loop filtering function |
| 333 | */ |
| 334 | static void rv40_loop_filter(RV34DecContext *r, int row) |
| 335 | { |
| 336 | MpegEncContext *s = &r->s; |
| 337 | int mb_pos, mb_x; |
| 338 | int i, j, k; |
| 339 | uint8_t *Y, *C; |
| 340 | int alpha, beta, betaY, betaC; |
| 341 | int q; |
| 342 | int mbtype[4]; ///< current macroblock and its neighbours types |
| 343 | /** |
| 344 | * flags indicating that macroblock can be filtered with strong filter |
| 345 | * it is set only for intra coded MB and MB with DCs coded separately |
| 346 | */ |
| 347 | int mb_strong[4]; |
| 348 | int clip[4]; ///< MB filter clipping value calculated from filtering strength |
| 349 | /** |
| 350 | * coded block patterns for luma part of current macroblock and its neighbours |
| 351 | * Format: |
| 352 | * LSB corresponds to the top left block, |
| 353 | * each nibble represents one row of subblocks. |
| 354 | */ |
| 355 | int cbp[4]; |
| 356 | /** |
| 357 | * coded block patterns for chroma part of current macroblock and its neighbours |
| 358 | * Format is the same as for luma with two subblocks in a row. |
| 359 | */ |
| 360 | int uvcbp[4][2]; |
| 361 | /** |
| 362 | * This mask represents the pattern of luma subblocks that should be filtered |
| 363 | * in addition to the coded ones because they lie at the edge of |
| 364 | * 8x8 block with different enough motion vectors |
| 365 | */ |
| 366 | unsigned mvmasks[4]; |
| 367 | |
| 368 | mb_pos = row * s->mb_stride; |
| 369 | for(mb_x = 0; mb_x < s->mb_width; mb_x++, mb_pos++){ |
| 370 | int mbtype = s->current_picture_ptr->mb_type[mb_pos]; |
| 371 | if(IS_INTRA(mbtype) || IS_SEPARATE_DC(mbtype)) |
| 372 | r->cbp_luma [mb_pos] = r->deblock_coefs[mb_pos] = 0xFFFF; |
| 373 | if(IS_INTRA(mbtype)) |
| 374 | r->cbp_chroma[mb_pos] = 0xFF; |
| 375 | } |
| 376 | mb_pos = row * s->mb_stride; |
| 377 | for(mb_x = 0; mb_x < s->mb_width; mb_x++, mb_pos++){ |
| 378 | int y_h_deblock, y_v_deblock; |
| 379 | int c_v_deblock[2], c_h_deblock[2]; |
| 380 | int clip_left; |
| 381 | int avail[4]; |
| 382 | unsigned y_to_deblock; |
| 383 | int c_to_deblock[2]; |
| 384 | |
| 385 | q = s->current_picture_ptr->qscale_table[mb_pos]; |
| 386 | alpha = rv40_alpha_tab[q]; |
| 387 | beta = rv40_beta_tab [q]; |
| 388 | betaY = betaC = beta * 3; |
| 389 | if(s->width * s->height <= 176*144) |
| 390 | betaY += beta; |
| 391 | |
| 392 | avail[0] = 1; |
| 393 | avail[1] = row; |
| 394 | avail[2] = mb_x; |
| 395 | avail[3] = row < s->mb_height - 1; |
| 396 | for(i = 0; i < 4; i++){ |
| 397 | if(avail[i]){ |
| 398 | int pos = mb_pos + neighbour_offs_x[i] + neighbour_offs_y[i]*s->mb_stride; |
| 399 | mvmasks[i] = r->deblock_coefs[pos]; |
| 400 | mbtype [i] = s->current_picture_ptr->mb_type[pos]; |
| 401 | cbp [i] = r->cbp_luma[pos]; |
| 402 | uvcbp[i][0] = r->cbp_chroma[pos] & 0xF; |
| 403 | uvcbp[i][1] = r->cbp_chroma[pos] >> 4; |
| 404 | }else{ |
| 405 | mvmasks[i] = 0; |
| 406 | mbtype [i] = mbtype[0]; |
| 407 | cbp [i] = 0; |
| 408 | uvcbp[i][0] = uvcbp[i][1] = 0; |
| 409 | } |
| 410 | mb_strong[i] = IS_INTRA(mbtype[i]) || IS_SEPARATE_DC(mbtype[i]); |
| 411 | clip[i] = rv40_filter_clip_tbl[mb_strong[i] + 1][q]; |
| 412 | } |
| 413 | y_to_deblock = mvmasks[POS_CUR] |
| 414 | | (mvmasks[POS_BOTTOM] << 16); |
| 415 | /* This pattern contains bits signalling that horizontal edges of |
| 416 | * the current block can be filtered. |
| 417 | * That happens when either of adjacent subblocks is coded or lies on |
| 418 | * the edge of 8x8 blocks with motion vectors differing by more than |
| 419 | * 3/4 pel in any component (any edge orientation for some reason). |
| 420 | */ |
| 421 | y_h_deblock = y_to_deblock |
| 422 | | ((cbp[POS_CUR] << 4) & ~MASK_Y_TOP_ROW) |
| 423 | | ((cbp[POS_TOP] & MASK_Y_LAST_ROW) >> 12); |
| 424 | /* This pattern contains bits signalling that vertical edges of |
| 425 | * the current block can be filtered. |
| 426 | * That happens when either of adjacent subblocks is coded or lies on |
| 427 | * the edge of 8x8 blocks with motion vectors differing by more than |
| 428 | * 3/4 pel in any component (any edge orientation for some reason). |
| 429 | */ |
| 430 | y_v_deblock = y_to_deblock |
| 431 | | ((cbp[POS_CUR] << 1) & ~MASK_Y_LEFT_COL) |
| 432 | | ((cbp[POS_LEFT] & MASK_Y_RIGHT_COL) >> 3); |
| 433 | if(!mb_x) |
| 434 | y_v_deblock &= ~MASK_Y_LEFT_COL; |
| 435 | if(!row) |
| 436 | y_h_deblock &= ~MASK_Y_TOP_ROW; |
| 437 | if(row == s->mb_height - 1 || (mb_strong[POS_CUR] | mb_strong[POS_BOTTOM])) |
| 438 | y_h_deblock &= ~(MASK_Y_TOP_ROW << 16); |
| 439 | /* Calculating chroma patterns is similar and easier since there is |
| 440 | * no motion vector pattern for them. |
| 441 | */ |
| 442 | for(i = 0; i < 2; i++){ |
| 443 | c_to_deblock[i] = (uvcbp[POS_BOTTOM][i] << 4) | uvcbp[POS_CUR][i]; |
| 444 | c_v_deblock[i] = c_to_deblock[i] |
| 445 | | ((uvcbp[POS_CUR] [i] << 1) & ~MASK_C_LEFT_COL) |
| 446 | | ((uvcbp[POS_LEFT][i] & MASK_C_RIGHT_COL) >> 1); |
| 447 | c_h_deblock[i] = c_to_deblock[i] |
| 448 | | ((uvcbp[POS_TOP][i] & MASK_C_LAST_ROW) >> 2) |
| 449 | | (uvcbp[POS_CUR][i] << 2); |
| 450 | if(!mb_x) |
| 451 | c_v_deblock[i] &= ~MASK_C_LEFT_COL; |
| 452 | if(!row) |
| 453 | c_h_deblock[i] &= ~MASK_C_TOP_ROW; |
| 454 | if(row == s->mb_height - 1 || (mb_strong[POS_CUR] | mb_strong[POS_BOTTOM])) |
| 455 | c_h_deblock[i] &= ~(MASK_C_TOP_ROW << 4); |
| 456 | } |
| 457 | |
| 458 | for(j = 0; j < 16; j += 4){ |
| 459 | Y = s->current_picture_ptr->f->data[0] + mb_x*16 + (row*16 + j) * s->linesize; |
| 460 | for(i = 0; i < 4; i++, Y += 4){ |
| 461 | int ij = i + j; |
| 462 | int clip_cur = y_to_deblock & (MASK_CUR << ij) ? clip[POS_CUR] : 0; |
| 463 | int dither = j ? ij : i*4; |
| 464 | |
| 465 | // if bottom block is coded then we can filter its top edge |
| 466 | // (or bottom edge of this block, which is the same) |
| 467 | if(y_h_deblock & (MASK_BOTTOM << ij)){ |
| 468 | rv40_adaptive_loop_filter(&r->rdsp, Y+4*s->linesize, |
| 469 | s->linesize, dither, |
| 470 | y_to_deblock & (MASK_BOTTOM << ij) ? clip[POS_CUR] : 0, |
| 471 | clip_cur, alpha, beta, betaY, |
| 472 | 0, 0, 0); |
| 473 | } |
| 474 | // filter left block edge in ordinary mode (with low filtering strength) |
| 475 | if(y_v_deblock & (MASK_CUR << ij) && (i || !(mb_strong[POS_CUR] | mb_strong[POS_LEFT]))){ |
| 476 | if(!i) |
| 477 | clip_left = mvmasks[POS_LEFT] & (MASK_RIGHT << j) ? clip[POS_LEFT] : 0; |
| 478 | else |
| 479 | clip_left = y_to_deblock & (MASK_CUR << (ij-1)) ? clip[POS_CUR] : 0; |
| 480 | rv40_adaptive_loop_filter(&r->rdsp, Y, s->linesize, dither, |
| 481 | clip_cur, |
| 482 | clip_left, |
| 483 | alpha, beta, betaY, 0, 0, 1); |
| 484 | } |
| 485 | // filter top edge of the current macroblock when filtering strength is high |
| 486 | if(!j && y_h_deblock & (MASK_CUR << i) && (mb_strong[POS_CUR] | mb_strong[POS_TOP])){ |
| 487 | rv40_adaptive_loop_filter(&r->rdsp, Y, s->linesize, dither, |
| 488 | clip_cur, |
| 489 | mvmasks[POS_TOP] & (MASK_TOP << i) ? clip[POS_TOP] : 0, |
| 490 | alpha, beta, betaY, 0, 1, 0); |
| 491 | } |
| 492 | // filter left block edge in edge mode (with high filtering strength) |
| 493 | if(y_v_deblock & (MASK_CUR << ij) && !i && (mb_strong[POS_CUR] | mb_strong[POS_LEFT])){ |
| 494 | clip_left = mvmasks[POS_LEFT] & (MASK_RIGHT << j) ? clip[POS_LEFT] : 0; |
| 495 | rv40_adaptive_loop_filter(&r->rdsp, Y, s->linesize, dither, |
| 496 | clip_cur, |
| 497 | clip_left, |
| 498 | alpha, beta, betaY, 0, 1, 1); |
| 499 | } |
| 500 | } |
| 501 | } |
| 502 | for(k = 0; k < 2; k++){ |
| 503 | for(j = 0; j < 2; j++){ |
| 504 | C = s->current_picture_ptr->f->data[k + 1] + mb_x*8 + (row*8 + j*4) * s->uvlinesize; |
| 505 | for(i = 0; i < 2; i++, C += 4){ |
| 506 | int ij = i + j*2; |
| 507 | int clip_cur = c_to_deblock[k] & (MASK_CUR << ij) ? clip[POS_CUR] : 0; |
| 508 | if(c_h_deblock[k] & (MASK_CUR << (ij+2))){ |
| 509 | int clip_bot = c_to_deblock[k] & (MASK_CUR << (ij+2)) ? clip[POS_CUR] : 0; |
| 510 | rv40_adaptive_loop_filter(&r->rdsp, C+4*s->uvlinesize, s->uvlinesize, i*8, |
| 511 | clip_bot, |
| 512 | clip_cur, |
| 513 | alpha, beta, betaC, 1, 0, 0); |
| 514 | } |
| 515 | if((c_v_deblock[k] & (MASK_CUR << ij)) && (i || !(mb_strong[POS_CUR] | mb_strong[POS_LEFT]))){ |
| 516 | if(!i) |
| 517 | clip_left = uvcbp[POS_LEFT][k] & (MASK_CUR << (2*j+1)) ? clip[POS_LEFT] : 0; |
| 518 | else |
| 519 | clip_left = c_to_deblock[k] & (MASK_CUR << (ij-1)) ? clip[POS_CUR] : 0; |
| 520 | rv40_adaptive_loop_filter(&r->rdsp, C, s->uvlinesize, j*8, |
| 521 | clip_cur, |
| 522 | clip_left, |
| 523 | alpha, beta, betaC, 1, 0, 1); |
| 524 | } |
| 525 | if(!j && c_h_deblock[k] & (MASK_CUR << ij) && (mb_strong[POS_CUR] | mb_strong[POS_TOP])){ |
| 526 | int clip_top = uvcbp[POS_TOP][k] & (MASK_CUR << (ij+2)) ? clip[POS_TOP] : 0; |
| 527 | rv40_adaptive_loop_filter(&r->rdsp, C, s->uvlinesize, i*8, |
| 528 | clip_cur, |
| 529 | clip_top, |
| 530 | alpha, beta, betaC, 1, 1, 0); |
| 531 | } |
| 532 | if(c_v_deblock[k] & (MASK_CUR << ij) && !i && (mb_strong[POS_CUR] | mb_strong[POS_LEFT])){ |
| 533 | clip_left = uvcbp[POS_LEFT][k] & (MASK_CUR << (2*j+1)) ? clip[POS_LEFT] : 0; |
| 534 | rv40_adaptive_loop_filter(&r->rdsp, C, s->uvlinesize, j*8, |
| 535 | clip_cur, |
| 536 | clip_left, |
| 537 | alpha, beta, betaC, 1, 1, 1); |
| 538 | } |
| 539 | } |
| 540 | } |
| 541 | } |
| 542 | } |
| 543 | } |
| 544 | |
| 545 | /** |
| 546 | * Initialize decoder. |
| 547 | */ |
| 548 | static av_cold int rv40_decode_init(AVCodecContext *avctx) |
| 549 | { |
| 550 | RV34DecContext *r = avctx->priv_data; |
| 551 | int ret; |
| 552 | |
| 553 | r->rv30 = 0; |
| 554 | if ((ret = ff_rv34_decode_init(avctx)) < 0) |
| 555 | return ret; |
| 556 | if(!aic_top_vlc.bits) |
| 557 | rv40_init_tables(); |
| 558 | r->parse_slice_header = rv40_parse_slice_header; |
| 559 | r->decode_intra_types = rv40_decode_intra_types; |
| 560 | r->decode_mb_info = rv40_decode_mb_info; |
| 561 | r->loop_filter = rv40_loop_filter; |
| 562 | r->luma_dc_quant_i = rv40_luma_dc_quant[0]; |
| 563 | r->luma_dc_quant_p = rv40_luma_dc_quant[1]; |
| 564 | return 0; |
| 565 | } |
| 566 | |
| 567 | AVCodec ff_rv40_decoder = { |
| 568 | .name = "rv40", |
| 569 | .long_name = NULL_IF_CONFIG_SMALL("RealVideo 4.0"), |
| 570 | .type = AVMEDIA_TYPE_VIDEO, |
| 571 | .id = AV_CODEC_ID_RV40, |
| 572 | .priv_data_size = sizeof(RV34DecContext), |
| 573 | .init = rv40_decode_init, |
| 574 | .close = ff_rv34_decode_end, |
| 575 | .decode = ff_rv34_decode_frame, |
| 576 | .capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY | |
| 577 | CODEC_CAP_FRAME_THREADS, |
| 578 | .flush = ff_mpeg_flush, |
| 579 | .pix_fmts = (const enum AVPixelFormat[]) { |
| 580 | AV_PIX_FMT_YUV420P, |
| 581 | AV_PIX_FMT_NONE |
| 582 | }, |
| 583 | .init_thread_copy = ONLY_IF_THREADS_ENABLED(ff_rv34_decode_init_thread_copy), |
| 584 | .update_thread_context = ONLY_IF_THREADS_ENABLED(ff_rv34_decode_update_thread_context), |
| 585 | }; |