| 1 | /* |
| 2 | * Copyright (C) 2006 Aurelien Jacobs <aurel@gnuage.org> |
| 3 | * |
| 4 | * This file is part of FFmpeg. |
| 5 | * |
| 6 | * FFmpeg is free software; you can redistribute it and/or |
| 7 | * modify it under the terms of the GNU Lesser General Public |
| 8 | * License as published by the Free Software Foundation; either |
| 9 | * version 2.1 of the License, or (at your option) any later version. |
| 10 | * |
| 11 | * FFmpeg is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 14 | * Lesser General Public License for more details. |
| 15 | * |
| 16 | * You should have received a copy of the GNU Lesser General Public |
| 17 | * License along with FFmpeg; if not, write to the Free Software |
| 18 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 19 | */ |
| 20 | |
| 21 | /** |
| 22 | * @file |
| 23 | * VP5 and VP6 compatible video decoder (common features) |
| 24 | */ |
| 25 | |
| 26 | #ifndef AVCODEC_VP56_H |
| 27 | #define AVCODEC_VP56_H |
| 28 | |
| 29 | #include "get_bits.h" |
| 30 | #include "hpeldsp.h" |
| 31 | #include "bytestream.h" |
| 32 | #include "h264chroma.h" |
| 33 | #include "videodsp.h" |
| 34 | #include "vp3dsp.h" |
| 35 | #include "vp56dsp.h" |
| 36 | |
| 37 | typedef struct vp56_context VP56Context; |
| 38 | |
| 39 | typedef enum { |
| 40 | VP56_FRAME_NONE =-1, |
| 41 | VP56_FRAME_CURRENT = 0, |
| 42 | VP56_FRAME_PREVIOUS = 1, |
| 43 | VP56_FRAME_GOLDEN = 2, |
| 44 | VP56_FRAME_GOLDEN2 = 3, |
| 45 | } VP56Frame; |
| 46 | |
| 47 | typedef enum { |
| 48 | VP56_MB_INTER_NOVEC_PF = 0, /**< Inter MB, no vector, from previous frame */ |
| 49 | VP56_MB_INTRA = 1, /**< Intra MB */ |
| 50 | VP56_MB_INTER_DELTA_PF = 2, /**< Inter MB, above/left vector + delta, from previous frame */ |
| 51 | VP56_MB_INTER_V1_PF = 3, /**< Inter MB, first vector, from previous frame */ |
| 52 | VP56_MB_INTER_V2_PF = 4, /**< Inter MB, second vector, from previous frame */ |
| 53 | VP56_MB_INTER_NOVEC_GF = 5, /**< Inter MB, no vector, from golden frame */ |
| 54 | VP56_MB_INTER_DELTA_GF = 6, /**< Inter MB, above/left vector + delta, from golden frame */ |
| 55 | VP56_MB_INTER_4V = 7, /**< Inter MB, 4 vectors, from previous frame */ |
| 56 | VP56_MB_INTER_V1_GF = 8, /**< Inter MB, first vector, from golden frame */ |
| 57 | VP56_MB_INTER_V2_GF = 9, /**< Inter MB, second vector, from golden frame */ |
| 58 | } VP56mb; |
| 59 | |
| 60 | typedef struct VP56Tree { |
| 61 | int8_t val; |
| 62 | int8_t prob_idx; |
| 63 | } VP56Tree; |
| 64 | |
| 65 | typedef struct VP56mv { |
| 66 | DECLARE_ALIGNED(4, int16_t, x); |
| 67 | int16_t y; |
| 68 | } VP56mv; |
| 69 | |
| 70 | #define VP56_SIZE_CHANGE 1 |
| 71 | |
| 72 | typedef void (*VP56ParseVectorAdjustment)(VP56Context *s, |
| 73 | VP56mv *vect); |
| 74 | typedef void (*VP56Filter)(VP56Context *s, uint8_t *dst, uint8_t *src, |
| 75 | int offset1, int offset2, int stride, |
| 76 | VP56mv mv, int mask, int select, int luma); |
| 77 | typedef void (*VP56ParseCoeff)(VP56Context *s); |
| 78 | typedef void (*VP56DefaultModelsInit)(VP56Context *s); |
| 79 | typedef void (*VP56ParseVectorModels)(VP56Context *s); |
| 80 | typedef int (*VP56ParseCoeffModels)(VP56Context *s); |
| 81 | typedef int (*VP56ParseHeader)(VP56Context *s, const uint8_t *buf, |
| 82 | int buf_size); |
| 83 | |
| 84 | typedef struct VP56RangeCoder { |
| 85 | int high; |
| 86 | int bits; /* stored negated (i.e. negative "bits" is a positive number of |
| 87 | bits left) in order to eliminate a negate in cache refilling */ |
| 88 | const uint8_t *buffer; |
| 89 | const uint8_t *end; |
| 90 | unsigned int code_word; |
| 91 | } VP56RangeCoder; |
| 92 | |
| 93 | typedef struct VP56RefDc { |
| 94 | uint8_t not_null_dc; |
| 95 | VP56Frame ref_frame; |
| 96 | int16_t dc_coeff; |
| 97 | } VP56RefDc; |
| 98 | |
| 99 | typedef struct VP56Macroblock { |
| 100 | uint8_t type; |
| 101 | VP56mv mv; |
| 102 | } VP56Macroblock; |
| 103 | |
| 104 | typedef struct VP56Model { |
| 105 | uint8_t coeff_reorder[64]; /* used in vp6 only */ |
| 106 | uint8_t coeff_index_to_pos[64]; /* used in vp6 only */ |
| 107 | uint8_t vector_sig[2]; /* delta sign */ |
| 108 | uint8_t vector_dct[2]; /* delta coding types */ |
| 109 | uint8_t vector_pdi[2][2]; /* predefined delta init */ |
| 110 | uint8_t vector_pdv[2][7]; /* predefined delta values */ |
| 111 | uint8_t vector_fdv[2][8]; /* 8 bit delta value definition */ |
| 112 | uint8_t coeff_dccv[2][11]; /* DC coeff value */ |
| 113 | uint8_t coeff_ract[2][3][6][11]; /* Run/AC coding type and AC coeff value */ |
| 114 | uint8_t coeff_acct[2][3][3][6][5];/* vp5 only AC coding type for coding group < 3 */ |
| 115 | uint8_t coeff_dcct[2][36][5]; /* DC coeff coding type */ |
| 116 | uint8_t coeff_runv[2][14]; /* run value (vp6 only) */ |
| 117 | uint8_t mb_type[3][10][10]; /* model for decoding MB type */ |
| 118 | uint8_t mb_types_stats[3][10][2];/* contextual, next MB type stats */ |
| 119 | } VP56Model; |
| 120 | |
| 121 | struct vp56_context { |
| 122 | AVCodecContext *avctx; |
| 123 | H264ChromaContext h264chroma; |
| 124 | HpelDSPContext hdsp; |
| 125 | VideoDSPContext vdsp; |
| 126 | VP3DSPContext vp3dsp; |
| 127 | VP56DSPContext vp56dsp; |
| 128 | uint8_t idct_scantable[64]; |
| 129 | AVFrame *frames[4]; |
| 130 | uint8_t *edge_emu_buffer_alloc; |
| 131 | uint8_t *edge_emu_buffer; |
| 132 | VP56RangeCoder c; |
| 133 | VP56RangeCoder cc; |
| 134 | VP56RangeCoder *ccp; |
| 135 | int sub_version; |
| 136 | |
| 137 | /* frame info */ |
| 138 | int golden_frame; |
| 139 | int plane_width[4]; |
| 140 | int plane_height[4]; |
| 141 | int mb_width; /* number of horizontal MB */ |
| 142 | int mb_height; /* number of vertical MB */ |
| 143 | int block_offset[6]; |
| 144 | |
| 145 | int quantizer; |
| 146 | uint16_t dequant_dc; |
| 147 | uint16_t dequant_ac; |
| 148 | |
| 149 | /* DC predictors management */ |
| 150 | VP56RefDc *above_blocks; |
| 151 | VP56RefDc left_block[4]; |
| 152 | int above_block_idx[6]; |
| 153 | int16_t prev_dc[3][3]; /* [plan][ref_frame] */ |
| 154 | |
| 155 | /* blocks / macroblock */ |
| 156 | VP56mb mb_type; |
| 157 | VP56Macroblock *macroblocks; |
| 158 | DECLARE_ALIGNED(16, int16_t, block_coeff)[6][64]; |
| 159 | |
| 160 | /* motion vectors */ |
| 161 | VP56mv mv[6]; /* vectors for each block in MB */ |
| 162 | VP56mv vector_candidate[2]; |
| 163 | int vector_candidate_pos; |
| 164 | |
| 165 | /* filtering hints */ |
| 166 | int filter_header; /* used in vp6 only */ |
| 167 | int deblock_filtering; |
| 168 | int filter_selection; |
| 169 | int filter_mode; |
| 170 | int max_vector_length; |
| 171 | int sample_variance_threshold; |
| 172 | |
| 173 | uint8_t coeff_ctx[4][64]; /* used in vp5 only */ |
| 174 | uint8_t coeff_ctx_last[4]; /* used in vp5 only */ |
| 175 | |
| 176 | int has_alpha; |
| 177 | |
| 178 | /* upside-down flipping hints */ |
| 179 | int flip; /* are we flipping ? */ |
| 180 | int frbi; /* first row block index in MB */ |
| 181 | int srbi; /* second row block index in MB */ |
| 182 | int stride[4]; /* stride for each plan */ |
| 183 | |
| 184 | const uint8_t *vp56_coord_div; |
| 185 | VP56ParseVectorAdjustment parse_vector_adjustment; |
| 186 | VP56Filter filter; |
| 187 | VP56ParseCoeff parse_coeff; |
| 188 | VP56DefaultModelsInit default_models_init; |
| 189 | VP56ParseVectorModels parse_vector_models; |
| 190 | VP56ParseCoeffModels parse_coeff_models; |
| 191 | VP56ParseHeader parse_header; |
| 192 | |
| 193 | /* for "slice" parallelism between YUV and A */ |
| 194 | VP56Context *alpha_context; |
| 195 | |
| 196 | VP56Model *modelp; |
| 197 | VP56Model model; |
| 198 | |
| 199 | /* huffman decoding */ |
| 200 | int use_huffman; |
| 201 | GetBitContext gb; |
| 202 | VLC dccv_vlc[2]; |
| 203 | VLC runv_vlc[2]; |
| 204 | VLC ract_vlc[2][3][6]; |
| 205 | unsigned int nb_null[2][2]; /* number of consecutive NULL DC/AC */ |
| 206 | }; |
| 207 | |
| 208 | |
| 209 | int ff_vp56_init(AVCodecContext *avctx, int flip, int has_alpha); |
| 210 | int ff_vp56_init_context(AVCodecContext *avctx, VP56Context *s, |
| 211 | int flip, int has_alpha); |
| 212 | int ff_vp56_free(AVCodecContext *avctx); |
| 213 | int ff_vp56_free_context(VP56Context *s); |
| 214 | void ff_vp56_init_dequant(VP56Context *s, int quantizer); |
| 215 | int ff_vp56_decode_frame(AVCodecContext *avctx, void *data, int *got_frame, |
| 216 | AVPacket *avpkt); |
| 217 | |
| 218 | |
| 219 | /** |
| 220 | * vp56 specific range coder implementation |
| 221 | */ |
| 222 | |
| 223 | extern const uint8_t ff_vp56_norm_shift[256]; |
| 224 | void ff_vp56_init_range_decoder(VP56RangeCoder *c, const uint8_t *buf, int buf_size); |
| 225 | |
| 226 | static av_always_inline unsigned int vp56_rac_renorm(VP56RangeCoder *c) |
| 227 | { |
| 228 | int shift = ff_vp56_norm_shift[c->high]; |
| 229 | int bits = c->bits; |
| 230 | unsigned int code_word = c->code_word; |
| 231 | |
| 232 | c->high <<= shift; |
| 233 | code_word <<= shift; |
| 234 | bits += shift; |
| 235 | if(bits >= 0 && c->buffer < c->end) { |
| 236 | code_word |= bytestream_get_be16(&c->buffer) << bits; |
| 237 | bits -= 16; |
| 238 | } |
| 239 | c->bits = bits; |
| 240 | return code_word; |
| 241 | } |
| 242 | |
| 243 | #if ARCH_ARM |
| 244 | #include "arm/vp56_arith.h" |
| 245 | #elif ARCH_X86 |
| 246 | #include "x86/vp56_arith.h" |
| 247 | #endif |
| 248 | |
| 249 | #ifndef vp56_rac_get_prob |
| 250 | #define vp56_rac_get_prob vp56_rac_get_prob |
| 251 | static av_always_inline int vp56_rac_get_prob(VP56RangeCoder *c, uint8_t prob) |
| 252 | { |
| 253 | unsigned int code_word = vp56_rac_renorm(c); |
| 254 | unsigned int low = 1 + (((c->high - 1) * prob) >> 8); |
| 255 | unsigned int low_shift = low << 16; |
| 256 | int bit = code_word >= low_shift; |
| 257 | |
| 258 | c->high = bit ? c->high - low : low; |
| 259 | c->code_word = bit ? code_word - low_shift : code_word; |
| 260 | |
| 261 | return bit; |
| 262 | } |
| 263 | #endif |
| 264 | |
| 265 | #ifndef vp56_rac_get_prob_branchy |
| 266 | // branchy variant, to be used where there's a branch based on the bit decoded |
| 267 | static av_always_inline int vp56_rac_get_prob_branchy(VP56RangeCoder *c, int prob) |
| 268 | { |
| 269 | unsigned long code_word = vp56_rac_renorm(c); |
| 270 | unsigned low = 1 + (((c->high - 1) * prob) >> 8); |
| 271 | unsigned low_shift = low << 16; |
| 272 | |
| 273 | if (code_word >= low_shift) { |
| 274 | c->high -= low; |
| 275 | c->code_word = code_word - low_shift; |
| 276 | return 1; |
| 277 | } |
| 278 | |
| 279 | c->high = low; |
| 280 | c->code_word = code_word; |
| 281 | return 0; |
| 282 | } |
| 283 | #endif |
| 284 | |
| 285 | static av_always_inline int vp56_rac_get(VP56RangeCoder *c) |
| 286 | { |
| 287 | unsigned int code_word = vp56_rac_renorm(c); |
| 288 | /* equiprobable */ |
| 289 | int low = (c->high + 1) >> 1; |
| 290 | unsigned int low_shift = low << 16; |
| 291 | int bit = code_word >= low_shift; |
| 292 | if (bit) { |
| 293 | c->high -= low; |
| 294 | code_word -= low_shift; |
| 295 | } else { |
| 296 | c->high = low; |
| 297 | } |
| 298 | |
| 299 | c->code_word = code_word; |
| 300 | return bit; |
| 301 | } |
| 302 | |
| 303 | // rounding is different than vp56_rac_get, is vp56_rac_get wrong? |
| 304 | static av_always_inline int vp8_rac_get(VP56RangeCoder *c) |
| 305 | { |
| 306 | return vp56_rac_get_prob(c, 128); |
| 307 | } |
| 308 | |
| 309 | static int vp56_rac_gets(VP56RangeCoder *c, int bits) |
| 310 | { |
| 311 | int value = 0; |
| 312 | |
| 313 | while (bits--) { |
| 314 | value = (value << 1) | vp56_rac_get(c); |
| 315 | } |
| 316 | |
| 317 | return value; |
| 318 | } |
| 319 | |
| 320 | static int vp8_rac_get_uint(VP56RangeCoder *c, int bits) |
| 321 | { |
| 322 | int value = 0; |
| 323 | |
| 324 | while (bits--) { |
| 325 | value = (value << 1) | vp8_rac_get(c); |
| 326 | } |
| 327 | |
| 328 | return value; |
| 329 | } |
| 330 | |
| 331 | // fixme: add 1 bit to all the calls to this? |
| 332 | static av_unused int vp8_rac_get_sint(VP56RangeCoder *c, int bits) |
| 333 | { |
| 334 | int v; |
| 335 | |
| 336 | if (!vp8_rac_get(c)) |
| 337 | return 0; |
| 338 | |
| 339 | v = vp8_rac_get_uint(c, bits); |
| 340 | |
| 341 | if (vp8_rac_get(c)) |
| 342 | v = -v; |
| 343 | |
| 344 | return v; |
| 345 | } |
| 346 | |
| 347 | // P(7) |
| 348 | static av_unused int vp56_rac_gets_nn(VP56RangeCoder *c, int bits) |
| 349 | { |
| 350 | int v = vp56_rac_gets(c, 7) << 1; |
| 351 | return v + !v; |
| 352 | } |
| 353 | |
| 354 | static av_unused int vp8_rac_get_nn(VP56RangeCoder *c) |
| 355 | { |
| 356 | int v = vp8_rac_get_uint(c, 7) << 1; |
| 357 | return v + !v; |
| 358 | } |
| 359 | |
| 360 | static av_always_inline |
| 361 | int vp56_rac_get_tree(VP56RangeCoder *c, |
| 362 | const VP56Tree *tree, |
| 363 | const uint8_t *probs) |
| 364 | { |
| 365 | while (tree->val > 0) { |
| 366 | if (vp56_rac_get_prob_branchy(c, probs[tree->prob_idx])) |
| 367 | tree += tree->val; |
| 368 | else |
| 369 | tree++; |
| 370 | } |
| 371 | return -tree->val; |
| 372 | } |
| 373 | |
| 374 | // how probabilities are associated with decisions is different I think |
| 375 | // well, the new scheme fits in the old but this way has one fewer branches per decision |
| 376 | static av_always_inline int vp8_rac_get_tree(VP56RangeCoder *c, const int8_t (*tree)[2], |
| 377 | const uint8_t *probs) |
| 378 | { |
| 379 | int i = 0; |
| 380 | |
| 381 | do { |
| 382 | i = tree[i][vp56_rac_get_prob(c, probs[i])]; |
| 383 | } while (i > 0); |
| 384 | |
| 385 | return -i; |
| 386 | } |
| 387 | |
| 388 | // DCTextra |
| 389 | static av_always_inline int vp8_rac_get_coeff(VP56RangeCoder *c, const uint8_t *prob) |
| 390 | { |
| 391 | int v = 0; |
| 392 | |
| 393 | do { |
| 394 | v = (v<<1) + vp56_rac_get_prob(c, *prob++); |
| 395 | } while (*prob); |
| 396 | |
| 397 | return v; |
| 398 | } |
| 399 | |
| 400 | #endif /* AVCODEC_VP56_H */ |