| Line | Branch | Exec | Source | 
|---|---|---|---|
| 1 | /* | ||
| 2 | * ATRAC9 decoder | ||
| 3 | * Copyright (c) 2018 Rostislav Pehlivanov <atomnuker@gmail.com> | ||
| 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 | #include "libavutil/channel_layout.h" | ||
| 23 | #include "libavutil/mem.h" | ||
| 24 | #include "libavutil/thread.h" | ||
| 25 | |||
| 26 | #include "codec_internal.h" | ||
| 27 | #include "decode.h" | ||
| 28 | #include "get_bits.h" | ||
| 29 | #include "atrac9tab.h" | ||
| 30 | #include "libavutil/tx.h" | ||
| 31 | #include "libavutil/lfg.h" | ||
| 32 | #include "libavutil/float_dsp.h" | ||
| 33 | #include "libavutil/mem_internal.h" | ||
| 34 | |||
| 35 | #define ATRAC9_SF_VLC_BITS 8 | ||
| 36 | #define ATRAC9_COEFF_VLC_BITS 9 | ||
| 37 | |||
| 38 | typedef struct ATRAC9ChannelData { | ||
| 39 | int band_ext; | ||
| 40 | int q_unit_cnt; | ||
| 41 | int band_ext_data[4]; | ||
| 42 | int32_t scalefactors[31]; | ||
| 43 | int32_t scalefactors_prev[31]; | ||
| 44 | |||
| 45 | int precision_coarse[30]; | ||
| 46 | int precision_fine[30]; | ||
| 47 | int precision_mask[30]; | ||
| 48 | |||
| 49 | int codebookset[30]; | ||
| 50 | |||
| 51 | int32_t q_coeffs_coarse[256]; | ||
| 52 | int32_t q_coeffs_fine[256]; | ||
| 53 | |||
| 54 | DECLARE_ALIGNED(32, float, coeffs )[256]; | ||
| 55 | DECLARE_ALIGNED(32, float, prev_win)[128]; | ||
| 56 | } ATRAC9ChannelData; | ||
| 57 | |||
| 58 | typedef struct ATRAC9BlockData { | ||
| 59 | ATRAC9ChannelData channel[2]; | ||
| 60 | |||
| 61 | /* Base */ | ||
| 62 | int band_count; | ||
| 63 | int q_unit_cnt; | ||
| 64 | int q_unit_cnt_prev; | ||
| 65 | |||
| 66 | /* Stereo block only */ | ||
| 67 | int stereo_q_unit; | ||
| 68 | |||
| 69 | /* Band extension only */ | ||
| 70 | int has_band_ext; | ||
| 71 | int has_band_ext_data; | ||
| 72 | int band_ext_q_unit; | ||
| 73 | |||
| 74 | /* Gradient */ | ||
| 75 | int grad_mode; | ||
| 76 | int grad_boundary; | ||
| 77 | int gradient[31]; | ||
| 78 | |||
| 79 | /* Stereo */ | ||
| 80 | int cpe_base_channel; | ||
| 81 | int is_signs[30]; | ||
| 82 | |||
| 83 | int reusable; | ||
| 84 | |||
| 85 | } ATRAC9BlockData; | ||
| 86 | |||
| 87 | typedef struct ATRAC9Context { | ||
| 88 | AVCodecContext *avctx; | ||
| 89 | AVFloatDSPContext *fdsp; | ||
| 90 | AVTXContext *tx; | ||
| 91 | av_tx_fn tx_fn; | ||
| 92 | ATRAC9BlockData block[5]; | ||
| 93 | AVLFG lfg; | ||
| 94 | |||
| 95 | /* Set on init */ | ||
| 96 | int frame_log2; | ||
| 97 | int avg_frame_size; | ||
| 98 | int frame_count; | ||
| 99 | int samplerate_idx; | ||
| 100 | const ATRAC9BlockConfig *block_config; | ||
| 101 | |||
| 102 | /* Generated on init */ | ||
| 103 | uint8_t alloc_curve[48][48]; | ||
| 104 | DECLARE_ALIGNED(32, float, imdct_win)[256]; | ||
| 105 | |||
| 106 | DECLARE_ALIGNED(32, float, temp)[2048]; | ||
| 107 | } ATRAC9Context; | ||
| 108 | |||
| 109 | static const VLCElem *sf_vlc[2][8]; /* Signed/unsigned, length */ | ||
| 110 | static const VLCElem *coeff_vlc[2][8][4]; /* Cookbook, precision, cookbook index */ | ||
| 111 | |||
| 112 | ✗ | static inline int parse_gradient(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 113 | GetBitContext *gb) | ||
| 114 | { | ||
| 115 | int grad_range[2]; | ||
| 116 | int grad_value[2]; | ||
| 117 | int values, sign, base; | ||
| 118 | uint8_t *curve; | ||
| 119 | float scale; | ||
| 120 | |||
| 121 | ✗ | b->grad_mode = get_bits(gb, 2); | |
| 122 | ✗ | if (b->grad_mode) { | |
| 123 | ✗ | grad_range[0] = get_bits(gb, 5); | |
| 124 | ✗ | grad_range[1] = 31; | |
| 125 | ✗ | grad_value[0] = get_bits(gb, 5); | |
| 126 | ✗ | grad_value[1] = 31; | |
| 127 | } else { | ||
| 128 | ✗ | grad_range[0] = get_bits(gb, 6); | |
| 129 | ✗ | grad_range[1] = get_bits(gb, 6) + 1; | |
| 130 | ✗ | grad_value[0] = get_bits(gb, 5); | |
| 131 | ✗ | grad_value[1] = get_bits(gb, 5); | |
| 132 | } | ||
| 133 | ✗ | b->grad_boundary = get_bits(gb, 4); | |
| 134 | |||
| 135 | ✗ | if (grad_range[0] >= grad_range[1] || grad_range[1] > 31) | |
| 136 | ✗ | return AVERROR_INVALIDDATA; | |
| 137 | |||
| 138 | ✗ | if (b->grad_boundary > b->q_unit_cnt) | |
| 139 | ✗ | return AVERROR_INVALIDDATA; | |
| 140 | |||
| 141 | ✗ | values = grad_value[1] - grad_value[0]; | |
| 142 | ✗ | sign = 1 - 2*(values < 0); | |
| 143 | ✗ | base = grad_value[0] + sign; | |
| 144 | ✗ | scale = (FFABS(values) - 1) / 31.0f; | |
| 145 | ✗ | curve = s->alloc_curve[grad_range[1] - grad_range[0] - 1]; | |
| 146 | |||
| 147 | ✗ | for (int i = 0; i <= b->q_unit_cnt; i++) | |
| 148 | ✗ | b->gradient[i] = grad_value[i >= grad_range[0]]; | |
| 149 | |||
| 150 | ✗ | for (int i = grad_range[0]; i < grad_range[1]; i++) | |
| 151 | ✗ | b->gradient[i] = base + sign*((int)(scale*curve[i - grad_range[0]])); | |
| 152 | |||
| 153 | ✗ | return 0; | |
| 154 | } | ||
| 155 | |||
| 156 | ✗ | static inline void calc_precision(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 157 | ATRAC9ChannelData *c) | ||
| 158 | { | ||
| 159 | ✗ | memset(c->precision_mask, 0, sizeof(c->precision_mask)); | |
| 160 | ✗ | for (int i = 1; i < b->q_unit_cnt; i++) { | |
| 161 | ✗ | const int delta = FFABS(c->scalefactors[i] - c->scalefactors[i - 1]) - 1; | |
| 162 | ✗ | if (delta > 0) { | |
| 163 | ✗ | const int neg = c->scalefactors[i - 1] > c->scalefactors[i]; | |
| 164 | ✗ | c->precision_mask[i - neg] += FFMIN(delta, 5); | |
| 165 | } | ||
| 166 | } | ||
| 167 | |||
| 168 | ✗ | if (b->grad_mode) { | |
| 169 | ✗ | for (int i = 0; i < b->q_unit_cnt; i++) { | |
| 170 | ✗ | c->precision_coarse[i] = c->scalefactors[i]; | |
| 171 | ✗ | c->precision_coarse[i] += c->precision_mask[i] - b->gradient[i]; | |
| 172 | ✗ | if (c->precision_coarse[i] < 0) | |
| 173 | ✗ | continue; | |
| 174 | ✗ | switch (b->grad_mode) { | |
| 175 | ✗ | case 1: | |
| 176 | ✗ | c->precision_coarse[i] >>= 1; | |
| 177 | ✗ | break; | |
| 178 | ✗ | case 2: | |
| 179 | ✗ | c->precision_coarse[i] = (3 * c->precision_coarse[i]) >> 3; | |
| 180 | ✗ | break; | |
| 181 | ✗ | case 3: | |
| 182 | ✗ | c->precision_coarse[i] >>= 2; | |
| 183 | ✗ | break; | |
| 184 | } | ||
| 185 | } | ||
| 186 | } else { | ||
| 187 | ✗ | for (int i = 0; i < b->q_unit_cnt; i++) | |
| 188 | ✗ | c->precision_coarse[i] = c->scalefactors[i] - b->gradient[i]; | |
| 189 | } | ||
| 190 | |||
| 191 | |||
| 192 | ✗ | for (int i = 0; i < b->q_unit_cnt; i++) | |
| 193 | ✗ | c->precision_coarse[i] = FFMAX(c->precision_coarse[i], 1); | |
| 194 | |||
| 195 | ✗ | for (int i = 0; i < b->grad_boundary; i++) | |
| 196 | ✗ | c->precision_coarse[i]++; | |
| 197 | |||
| 198 | ✗ | for (int i = 0; i < b->q_unit_cnt; i++) { | |
| 199 | ✗ | c->precision_fine[i] = 0; | |
| 200 | ✗ | if (c->precision_coarse[i] > 15) { | |
| 201 | ✗ | c->precision_fine[i] = FFMIN(c->precision_coarse[i], 30) - 15; | |
| 202 | ✗ | c->precision_coarse[i] = 15; | |
| 203 | } | ||
| 204 | } | ||
| 205 | ✗ | } | |
| 206 | |||
| 207 | ✗ | static inline int parse_band_ext(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 208 | GetBitContext *gb, int stereo) | ||
| 209 | { | ||
| 210 | ✗ | int ext_band = 0; | |
| 211 | |||
| 212 | ✗ | if (b->has_band_ext) { | |
| 213 | ✗ | if (b->q_unit_cnt < 13 || b->q_unit_cnt > 20) | |
| 214 | ✗ | return AVERROR_INVALIDDATA; | |
| 215 | ✗ | ext_band = at9_tab_band_ext_group[b->q_unit_cnt - 13][2]; | |
| 216 | ✗ | if (stereo) { | |
| 217 | ✗ | b->channel[1].band_ext = get_bits(gb, 2); | |
| 218 | ✗ | b->channel[1].band_ext = ext_band > 2 ? b->channel[1].band_ext : 4; | |
| 219 | } else { | ||
| 220 | ✗ | skip_bits1(gb); | |
| 221 | } | ||
| 222 | } | ||
| 223 | |||
| 224 | ✗ | b->has_band_ext_data = get_bits1(gb); | |
| 225 | ✗ | if (!b->has_band_ext_data) | |
| 226 | ✗ | return 0; | |
| 227 | |||
| 228 | ✗ | if (!b->has_band_ext) { | |
| 229 | ✗ | skip_bits(gb, 2); | |
| 230 | ✗ | skip_bits_long(gb, get_bits(gb, 5)); | |
| 231 | ✗ | return 0; | |
| 232 | } | ||
| 233 | |||
| 234 | ✗ | b->channel[0].band_ext = get_bits(gb, 2); | |
| 235 | ✗ | b->channel[0].band_ext = ext_band > 2 ? b->channel[0].band_ext : 4; | |
| 236 | |||
| 237 | ✗ | if (!get_bits(gb, 5)) { | |
| 238 | ✗ | for (int i = 0; i <= stereo; i++) { | |
| 239 | ✗ | ATRAC9ChannelData *c = &b->channel[i]; | |
| 240 | ✗ | const int count = at9_tab_band_ext_cnt[c->band_ext][ext_band]; | |
| 241 | ✗ | for (int j = 0; j < count; j++) { | |
| 242 | ✗ | int len = at9_tab_band_ext_lengths[c->band_ext][ext_band][j]; | |
| 243 | ✗ | c->band_ext_data[j] = av_clip_uintp2_c(c->band_ext_data[j], len); | |
| 244 | } | ||
| 245 | } | ||
| 246 | |||
| 247 | ✗ | return 0; | |
| 248 | } | ||
| 249 | |||
| 250 | ✗ | for (int i = 0; i <= stereo; i++) { | |
| 251 | ✗ | ATRAC9ChannelData *c = &b->channel[i]; | |
| 252 | ✗ | const int count = at9_tab_band_ext_cnt[c->band_ext][ext_band]; | |
| 253 | ✗ | for (int j = 0; j < count; j++) { | |
| 254 | ✗ | int len = at9_tab_band_ext_lengths[c->band_ext][ext_band][j]; | |
| 255 | ✗ | c->band_ext_data[j] = get_bits(gb, len); | |
| 256 | } | ||
| 257 | } | ||
| 258 | |||
| 259 | ✗ | return 0; | |
| 260 | } | ||
| 261 | |||
| 262 | ✗ | static inline int read_scalefactors(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 263 | ATRAC9ChannelData *c, GetBitContext *gb, | ||
| 264 | int channel_idx, int first_in_pkt) | ||
| 265 | { | ||
| 266 | static const uint8_t mode_map[2][4] = { { 0, 1, 2, 3 }, { 0, 2, 3, 4 } }; | ||
| 267 | ✗ | const int mode = mode_map[channel_idx][get_bits(gb, 2)]; | |
| 268 | |||
| 269 | ✗ | memset(c->scalefactors, 0, sizeof(c->scalefactors)); | |
| 270 | |||
| 271 | ✗ | if (first_in_pkt && (mode == 4 || ((mode == 3) && !channel_idx))) { | |
| 272 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "Invalid scalefactor coding mode!\n"); | |
| 273 | ✗ | return AVERROR_INVALIDDATA; | |
| 274 | } | ||
| 275 | |||
| 276 | ✗ | switch (mode) { | |
| 277 | ✗ | case 0: { /* VLC delta offset */ | |
| 278 | ✗ | const uint8_t *sf_weights = at9_tab_sf_weights[get_bits(gb, 3)]; | |
| 279 | ✗ | const int base = get_bits(gb, 5); | |
| 280 | ✗ | const int len = get_bits(gb, 2) + 3; | |
| 281 | ✗ | const VLCElem *tab = sf_vlc[0][len]; | |
| 282 | |||
| 283 | ✗ | c->scalefactors[0] = get_bits(gb, len); | |
| 284 | |||
| 285 | ✗ | for (int i = 1; i < b->band_ext_q_unit; i++) { | |
| 286 | ✗ | int val = c->scalefactors[i - 1] + get_vlc2(gb, tab, | |
| 287 | ATRAC9_SF_VLC_BITS, 1); | ||
| 288 | ✗ | c->scalefactors[i] = av_zero_extend(val, len); | |
| 289 | } | ||
| 290 | |||
| 291 | ✗ | for (int i = 0; i < b->band_ext_q_unit; i++) | |
| 292 | ✗ | c->scalefactors[i] += base - sf_weights[i]; | |
| 293 | |||
| 294 | ✗ | break; | |
| 295 | } | ||
| 296 | ✗ | case 1: { /* CLC offset */ | |
| 297 | ✗ | const int len = get_bits(gb, 2) + 2; | |
| 298 | ✗ | const int base = len < 5 ? get_bits(gb, 5) : 0; | |
| 299 | ✗ | for (int i = 0; i < b->band_ext_q_unit; i++) | |
| 300 | ✗ | c->scalefactors[i] = base + get_bits(gb, len); | |
| 301 | ✗ | break; | |
| 302 | } | ||
| 303 | ✗ | case 2: | |
| 304 | case 4: { /* VLC dist to baseline */ | ||
| 305 | ✗ | const int *baseline = mode == 4 ? c->scalefactors_prev : | |
| 306 | channel_idx ? b->channel[0].scalefactors : | ||
| 307 | c->scalefactors_prev; | ||
| 308 | ✗ | const int baseline_len = mode == 4 ? b->q_unit_cnt_prev : | |
| 309 | ✗ | channel_idx ? b->band_ext_q_unit : | |
| 310 | b->q_unit_cnt_prev; | ||
| 311 | |||
| 312 | ✗ | const int len = get_bits(gb, 2) + 2; | |
| 313 | ✗ | const int unit_cnt = FFMIN(b->band_ext_q_unit, baseline_len); | |
| 314 | ✗ | const VLCElem *tab = sf_vlc[1][len]; | |
| 315 | |||
| 316 | ✗ | for (int i = 0; i < unit_cnt; i++) { | |
| 317 | ✗ | int dist = get_vlc2(gb, tab, ATRAC9_SF_VLC_BITS, 1); | |
| 318 | ✗ | c->scalefactors[i] = baseline[i] + dist; | |
| 319 | } | ||
| 320 | |||
| 321 | ✗ | for (int i = unit_cnt; i < b->band_ext_q_unit; i++) | |
| 322 | ✗ | c->scalefactors[i] = get_bits(gb, 5); | |
| 323 | |||
| 324 | ✗ | break; | |
| 325 | } | ||
| 326 | ✗ | case 3: { /* VLC offset with baseline */ | |
| 327 | ✗ | const int *baseline = channel_idx ? b->channel[0].scalefactors : | |
| 328 | c->scalefactors_prev; | ||
| 329 | ✗ | const int baseline_len = channel_idx ? b->band_ext_q_unit : | |
| 330 | b->q_unit_cnt_prev; | ||
| 331 | |||
| 332 | ✗ | const int base = get_bits(gb, 5) - (1 << (5 - 1)); | |
| 333 | ✗ | const int len = get_bits(gb, 2) + 1; | |
| 334 | ✗ | const int unit_cnt = FFMIN(b->band_ext_q_unit, baseline_len); | |
| 335 | ✗ | const VLCElem *tab = sf_vlc[0][len]; | |
| 336 | |||
| 337 | ✗ | c->scalefactors[0] = get_bits(gb, len); | |
| 338 | |||
| 339 | ✗ | for (int i = 1; i < unit_cnt; i++) { | |
| 340 | ✗ | int val = c->scalefactors[i - 1] + get_vlc2(gb, tab, | |
| 341 | ATRAC9_SF_VLC_BITS, 1); | ||
| 342 | ✗ | c->scalefactors[i] = av_zero_extend(val, len); | |
| 343 | } | ||
| 344 | |||
| 345 | ✗ | for (int i = 0; i < unit_cnt; i++) | |
| 346 | ✗ | c->scalefactors[i] += base + baseline[i]; | |
| 347 | |||
| 348 | ✗ | for (int i = unit_cnt; i < b->band_ext_q_unit; i++) | |
| 349 | ✗ | c->scalefactors[i] = get_bits(gb, 5); | |
| 350 | ✗ | break; | |
| 351 | } | ||
| 352 | } | ||
| 353 | |||
| 354 | ✗ | for (int i = 0; i < b->band_ext_q_unit; i++) | |
| 355 | ✗ | if (c->scalefactors[i] < 0 || c->scalefactors[i] > 31) | |
| 356 | ✗ | return AVERROR_INVALIDDATA; | |
| 357 | |||
| 358 | ✗ | memcpy(c->scalefactors_prev, c->scalefactors, sizeof(c->scalefactors)); | |
| 359 | |||
| 360 | ✗ | return 0; | |
| 361 | } | ||
| 362 | |||
| 363 | ✗ | static inline void calc_codebook_idx(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 364 | ATRAC9ChannelData *c) | ||
| 365 | { | ||
| 366 | ✗ | int avg = 0; | |
| 367 | ✗ | const int last_sf = c->scalefactors[c->q_unit_cnt]; | |
| 368 | |||
| 369 | ✗ | memset(c->codebookset, 0, sizeof(c->codebookset)); | |
| 370 | |||
| 371 | ✗ | if (c->q_unit_cnt <= 1) | |
| 372 | ✗ | return; | |
| 373 | ✗ | if (s->samplerate_idx > 7) | |
| 374 | ✗ | return; | |
| 375 | |||
| 376 | ✗ | c->scalefactors[c->q_unit_cnt] = c->scalefactors[c->q_unit_cnt - 1]; | |
| 377 | |||
| 378 | ✗ | if (c->q_unit_cnt > 12) { | |
| 379 | ✗ | for (int i = 0; i < 12; i++) | |
| 380 | ✗ | avg += c->scalefactors[i]; | |
| 381 | ✗ | avg = (avg + 6) / 12; | |
| 382 | } | ||
| 383 | |||
| 384 | ✗ | for (int i = 8; i < c->q_unit_cnt; i++) { | |
| 385 | ✗ | const int prev = c->scalefactors[i - 1]; | |
| 386 | ✗ | const int cur = c->scalefactors[i ]; | |
| 387 | ✗ | const int next = c->scalefactors[i + 1]; | |
| 388 | ✗ | const int min = FFMIN(prev, next); | |
| 389 | ✗ | if ((cur - min >= 3 || 2*cur - prev - next >= 3)) | |
| 390 | ✗ | c->codebookset[i] = 1; | |
| 391 | } | ||
| 392 | |||
| 393 | |||
| 394 | ✗ | for (int i = 12; i < c->q_unit_cnt; i++) { | |
| 395 | ✗ | const int cur = c->scalefactors[i]; | |
| 396 | ✗ | const int cnd = at9_q_unit_to_coeff_cnt[i] == 16; | |
| 397 | ✗ | const int min = FFMIN(c->scalefactors[i + 1], c->scalefactors[i - 1]); | |
| 398 | ✗ | if (c->codebookset[i]) | |
| 399 | ✗ | continue; | |
| 400 | |||
| 401 | ✗ | c->codebookset[i] = (((cur - min) >= 2) && (cur >= (avg - cnd))); | |
| 402 | } | ||
| 403 | |||
| 404 | ✗ | c->scalefactors[c->q_unit_cnt] = last_sf; | |
| 405 | } | ||
| 406 | |||
| 407 | ✗ | static inline void read_coeffs_coarse(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 408 | ATRAC9ChannelData *c, GetBitContext *gb) | ||
| 409 | { | ||
| 410 | ✗ | const int max_prec = s->samplerate_idx > 7 ? 1 : 7; | |
| 411 | |||
| 412 | ✗ | memset(c->q_coeffs_coarse, 0, sizeof(c->q_coeffs_coarse)); | |
| 413 | |||
| 414 | ✗ | for (int i = 0; i < c->q_unit_cnt; i++) { | |
| 415 | ✗ | int *coeffs = &c->q_coeffs_coarse[at9_q_unit_to_coeff_idx[i]]; | |
| 416 | ✗ | const int bands = at9_q_unit_to_coeff_cnt[i]; | |
| 417 | ✗ | const int prec = c->precision_coarse[i] + 1; | |
| 418 | |||
| 419 | ✗ | if (prec <= max_prec) { | |
| 420 | ✗ | const int cb = c->codebookset[i]; | |
| 421 | ✗ | const int cbi = at9_q_unit_to_codebookidx[i]; | |
| 422 | ✗ | const VLCElem *tab = coeff_vlc[cb][prec][cbi]; | |
| 423 | ✗ | const HuffmanCodebook *huff = &at9_huffman_coeffs[cb][prec][cbi]; | |
| 424 | ✗ | const int groups = bands >> huff->value_cnt_pow; | |
| 425 | |||
| 426 | ✗ | for (int j = 0; j < groups; j++) { | |
| 427 | ✗ | uint16_t val = get_vlc2(gb, tab, ATRAC9_COEFF_VLC_BITS, 2); | |
| 428 | |||
| 429 | ✗ | for (int k = 0; k < huff->value_cnt; k++) { | |
| 430 | ✗ | coeffs[k] = sign_extend(val, huff->value_bits); | |
| 431 | ✗ | val >>= huff->value_bits; | |
| 432 | } | ||
| 433 | |||
| 434 | ✗ | coeffs += huff->value_cnt; | |
| 435 | } | ||
| 436 | } else { | ||
| 437 | ✗ | for (int j = 0; j < bands; j++) | |
| 438 | ✗ | coeffs[j] = sign_extend(get_bits(gb, prec), prec); | |
| 439 | } | ||
| 440 | } | ||
| 441 | ✗ | } | |
| 442 | |||
| 443 | ✗ | static inline void read_coeffs_fine(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 444 | ATRAC9ChannelData *c, GetBitContext *gb) | ||
| 445 | { | ||
| 446 | ✗ | memset(c->q_coeffs_fine, 0, sizeof(c->q_coeffs_fine)); | |
| 447 | |||
| 448 | ✗ | for (int i = 0; i < c->q_unit_cnt; i++) { | |
| 449 | ✗ | const int start = at9_q_unit_to_coeff_idx[i + 0]; | |
| 450 | ✗ | const int end = at9_q_unit_to_coeff_idx[i + 1]; | |
| 451 | ✗ | const int len = c->precision_fine[i] + 1; | |
| 452 | |||
| 453 | ✗ | if (c->precision_fine[i] <= 0) | |
| 454 | ✗ | continue; | |
| 455 | |||
| 456 | ✗ | for (int j = start; j < end; j++) | |
| 457 | ✗ | c->q_coeffs_fine[j] = sign_extend(get_bits(gb, len), len); | |
| 458 | } | ||
| 459 | ✗ | } | |
| 460 | |||
| 461 | ✗ | static inline void dequantize(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 462 | ATRAC9ChannelData *c) | ||
| 463 | { | ||
| 464 | ✗ | memset(c->coeffs, 0, sizeof(c->coeffs)); | |
| 465 | |||
| 466 | ✗ | for (int i = 0; i < c->q_unit_cnt; i++) { | |
| 467 | ✗ | const int start = at9_q_unit_to_coeff_idx[i + 0]; | |
| 468 | ✗ | const int end = at9_q_unit_to_coeff_idx[i + 1]; | |
| 469 | |||
| 470 | ✗ | const float coarse_c = at9_quant_step_coarse[c->precision_coarse[i]]; | |
| 471 | ✗ | const float fine_c = at9_quant_step_fine[c->precision_fine[i]]; | |
| 472 | |||
| 473 | ✗ | for (int j = start; j < end; j++) { | |
| 474 | ✗ | const float vc = c->q_coeffs_coarse[j] * coarse_c; | |
| 475 | ✗ | const float vf = c->q_coeffs_fine[j] * fine_c; | |
| 476 | ✗ | c->coeffs[j] = vc + vf; | |
| 477 | } | ||
| 478 | } | ||
| 479 | ✗ | } | |
| 480 | |||
| 481 | ✗ | static inline void apply_intensity_stereo(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 482 | const int stereo) | ||
| 483 | { | ||
| 484 | ✗ | float *src = b->channel[ b->cpe_base_channel].coeffs; | |
| 485 | ✗ | float *dst = b->channel[!b->cpe_base_channel].coeffs; | |
| 486 | |||
| 487 | ✗ | if (!stereo) | |
| 488 | ✗ | return; | |
| 489 | |||
| 490 | ✗ | if (b->q_unit_cnt <= b->stereo_q_unit) | |
| 491 | ✗ | return; | |
| 492 | |||
| 493 | ✗ | for (int i = b->stereo_q_unit; i < b->q_unit_cnt; i++) { | |
| 494 | ✗ | const int sign = b->is_signs[i]; | |
| 495 | ✗ | const int start = at9_q_unit_to_coeff_idx[i + 0]; | |
| 496 | ✗ | const int end = at9_q_unit_to_coeff_idx[i + 1]; | |
| 497 | ✗ | for (int j = start; j < end; j++) | |
| 498 | ✗ | dst[j] = sign*src[j]; | |
| 499 | } | ||
| 500 | } | ||
| 501 | |||
| 502 | ✗ | static inline void apply_scalefactors(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 503 | const int stereo) | ||
| 504 | { | ||
| 505 | ✗ | for (int i = 0; i <= stereo; i++) { | |
| 506 | ✗ | float *coeffs = b->channel[i].coeffs; | |
| 507 | ✗ | for (int j = 0; j < b->q_unit_cnt; j++) { | |
| 508 | ✗ | const int start = at9_q_unit_to_coeff_idx[j + 0]; | |
| 509 | ✗ | const int end = at9_q_unit_to_coeff_idx[j + 1]; | |
| 510 | ✗ | const int scalefactor = b->channel[i].scalefactors[j]; | |
| 511 | ✗ | const float scale = at9_scalefactor_c[scalefactor]; | |
| 512 | ✗ | for (int k = start; k < end; k++) | |
| 513 | ✗ | coeffs[k] *= scale; | |
| 514 | } | ||
| 515 | } | ||
| 516 | ✗ | } | |
| 517 | |||
| 518 | ✗ | static inline void fill_with_noise(ATRAC9Context *s, ATRAC9ChannelData *c, | |
| 519 | int start, int count) | ||
| 520 | { | ||
| 521 | ✗ | float maxval = 0.0f; | |
| 522 | ✗ | for (int i = 0; i < count; i += 2) { | |
| 523 | double tmp[2]; | ||
| 524 | ✗ | av_bmg_get(&s->lfg, tmp); | |
| 525 | ✗ | c->coeffs[start + i + 0] = tmp[0]; | |
| 526 | ✗ | c->coeffs[start + i + 1] = tmp[1]; | |
| 527 | ✗ | maxval = FFMAX(FFMAX(FFABS(tmp[0]), FFABS(tmp[1])), maxval); | |
| 528 | } | ||
| 529 | /* Normalize */ | ||
| 530 | ✗ | for (int i = 0; i < count; i++) | |
| 531 | ✗ | c->coeffs[start + i] /= maxval; | |
| 532 | ✗ | } | |
| 533 | |||
| 534 | ✗ | static inline void scale_band_ext_coeffs(ATRAC9ChannelData *c, float sf[6], | |
| 535 | const int s_unit, const int e_unit) | ||
| 536 | { | ||
| 537 | ✗ | for (int i = s_unit; i < e_unit; i++) { | |
| 538 | ✗ | const int start = at9_q_unit_to_coeff_idx[i + 0]; | |
| 539 | ✗ | const int end = at9_q_unit_to_coeff_idx[i + 1]; | |
| 540 | ✗ | for (int j = start; j < end; j++) | |
| 541 | ✗ | c->coeffs[j] *= sf[i - s_unit]; | |
| 542 | } | ||
| 543 | ✗ | } | |
| 544 | |||
| 545 | ✗ | static inline void apply_band_extension(ATRAC9Context *s, ATRAC9BlockData *b, | |
| 546 | const int stereo) | ||
| 547 | { | ||
| 548 | ✗ | const int g_units[4] = { /* A, B, C, total units */ | |
| 549 | ✗ | b->q_unit_cnt, | |
| 550 | ✗ | at9_tab_band_ext_group[b->q_unit_cnt - 13][0], | |
| 551 | ✗ | at9_tab_band_ext_group[b->q_unit_cnt - 13][1], | |
| 552 | ✗ | FFMAX(g_units[2], 22), | |
| 553 | }; | ||
| 554 | |||
| 555 | ✗ | const int g_bins[4] = { /* A, B, C, total bins */ | |
| 556 | ✗ | at9_q_unit_to_coeff_idx[g_units[0]], | |
| 557 | ✗ | at9_q_unit_to_coeff_idx[g_units[1]], | |
| 558 | ✗ | at9_q_unit_to_coeff_idx[g_units[2]], | |
| 559 | ✗ | at9_q_unit_to_coeff_idx[g_units[3]], | |
| 560 | }; | ||
| 561 | |||
| 562 | ✗ | for (int ch = 0; ch <= stereo; ch++) { | |
| 563 | ✗ | ATRAC9ChannelData *c = &b->channel[ch]; | |
| 564 | |||
| 565 | /* Mirror the spectrum */ | ||
| 566 | ✗ | for (int i = 0; i < 3; i++) | |
| 567 | ✗ | for (int j = 0; j < (g_bins[i + 1] - g_bins[i + 0]); j++) | |
| 568 | ✗ | c->coeffs[g_bins[i] + j] = c->coeffs[g_bins[i] - j - 1]; | |
| 569 | |||
| 570 | ✗ | switch (c->band_ext) { | |
| 571 | ✗ | case 0: { | |
| 572 | ✗ | float sf[6] = { 0.0f }; | |
| 573 | ✗ | const int l = g_units[3] - g_units[0] - 1; | |
| 574 | ✗ | const int n_start = at9_q_unit_to_coeff_idx[g_units[3] - 1]; | |
| 575 | ✗ | const int n_cnt = at9_q_unit_to_coeff_cnt[g_units[3] - 1]; | |
| 576 | ✗ | switch (at9_tab_band_ext_group[b->q_unit_cnt - 13][2]) { | |
| 577 | ✗ | case 3: | |
| 578 | ✗ | sf[0] = at9_band_ext_scales_m0[0][0][c->band_ext_data[0]]; | |
| 579 | ✗ | sf[1] = at9_band_ext_scales_m0[0][1][c->band_ext_data[0]]; | |
| 580 | ✗ | sf[2] = at9_band_ext_scales_m0[0][2][c->band_ext_data[1]]; | |
| 581 | ✗ | sf[3] = at9_band_ext_scales_m0[0][3][c->band_ext_data[2]]; | |
| 582 | ✗ | sf[4] = at9_band_ext_scales_m0[0][4][c->band_ext_data[3]]; | |
| 583 | ✗ | break; | |
| 584 | ✗ | case 4: | |
| 585 | ✗ | sf[0] = at9_band_ext_scales_m0[1][0][c->band_ext_data[0]]; | |
| 586 | ✗ | sf[1] = at9_band_ext_scales_m0[1][1][c->band_ext_data[0]]; | |
| 587 | ✗ | sf[2] = at9_band_ext_scales_m0[1][2][c->band_ext_data[1]]; | |
| 588 | ✗ | sf[3] = at9_band_ext_scales_m0[1][3][c->band_ext_data[2]]; | |
| 589 | ✗ | sf[4] = at9_band_ext_scales_m0[1][4][c->band_ext_data[3]]; | |
| 590 | ✗ | break; | |
| 591 | ✗ | case 5: | |
| 592 | ✗ | sf[0] = at9_band_ext_scales_m0[2][0][c->band_ext_data[0]]; | |
| 593 | ✗ | sf[1] = at9_band_ext_scales_m0[2][1][c->band_ext_data[1]]; | |
| 594 | ✗ | sf[2] = at9_band_ext_scales_m0[2][2][c->band_ext_data[1]]; | |
| 595 | ✗ | break; | |
| 596 | } | ||
| 597 | |||
| 598 | ✗ | sf[l] = at9_scalefactor_c[c->scalefactors[g_units[0]]]; | |
| 599 | |||
| 600 | ✗ | fill_with_noise(s, c, n_start, n_cnt); | |
| 601 | ✗ | scale_band_ext_coeffs(c, sf, g_units[0], g_units[3]); | |
| 602 | ✗ | break; | |
| 603 | } | ||
| 604 | ✗ | case 1: { | |
| 605 | float sf[6]; | ||
| 606 | ✗ | for (int i = g_units[0]; i < g_units[3]; i++) | |
| 607 | ✗ | sf[i - g_units[0]] = at9_scalefactor_c[c->scalefactors[i]]; | |
| 608 | |||
| 609 | ✗ | fill_with_noise(s, c, g_bins[0], g_bins[3] - g_bins[0]); | |
| 610 | ✗ | scale_band_ext_coeffs(c, sf, g_units[0], g_units[3]); | |
| 611 | ✗ | break; | |
| 612 | } | ||
| 613 | ✗ | case 2: { | |
| 614 | ✗ | const float g_sf[2] = { | |
| 615 | ✗ | at9_band_ext_scales_m2[c->band_ext_data[0]], | |
| 616 | ✗ | at9_band_ext_scales_m2[c->band_ext_data[1]], | |
| 617 | }; | ||
| 618 | |||
| 619 | ✗ | for (int i = 0; i < 2; i++) | |
| 620 | ✗ | for (int j = g_bins[i + 0]; j < g_bins[i + 1]; j++) | |
| 621 | ✗ | c->coeffs[j] *= g_sf[i]; | |
| 622 | ✗ | break; | |
| 623 | } | ||
| 624 | ✗ | case 3: { | |
| 625 | ✗ | float scale = at9_band_ext_scales_m3[c->band_ext_data[0]][0]; | |
| 626 | ✗ | float rate = at9_band_ext_scales_m3[c->band_ext_data[1]][1]; | |
| 627 | ✗ | rate = pow(2, rate); | |
| 628 | ✗ | for (int i = g_bins[0]; i < g_bins[3]; i++) { | |
| 629 | ✗ | scale *= rate; | |
| 630 | ✗ | c->coeffs[i] *= scale; | |
| 631 | } | ||
| 632 | ✗ | break; | |
| 633 | } | ||
| 634 | ✗ | case 4: { | |
| 635 | ✗ | const float m = at9_band_ext_scales_m4[c->band_ext_data[0]]; | |
| 636 | ✗ | const float g_sf[3] = { 0.7079468f*m, 0.5011902f*m, 0.3548279f*m }; | |
| 637 | |||
| 638 | ✗ | for (int i = 0; i < 3; i++) | |
| 639 | ✗ | for (int j = g_bins[i + 0]; j < g_bins[i + 1]; j++) | |
| 640 | ✗ | c->coeffs[j] *= g_sf[i]; | |
| 641 | ✗ | break; | |
| 642 | } | ||
| 643 | } | ||
| 644 | } | ||
| 645 | ✗ | } | |
| 646 | |||
| 647 | ✗ | static int atrac9_decode_block(ATRAC9Context *s, GetBitContext *gb, | |
| 648 | ATRAC9BlockData *b, AVFrame *frame, | ||
| 649 | int frame_idx, int block_idx) | ||
| 650 | { | ||
| 651 | ✗ | const int first_in_pkt = !get_bits1(gb); | |
| 652 | ✗ | const int reuse_params = get_bits1(gb); | |
| 653 | ✗ | const int stereo = s->block_config->type[block_idx] == ATRAC9_BLOCK_TYPE_CPE; | |
| 654 | |||
| 655 | ✗ | if (s->block_config->type[block_idx] == ATRAC9_BLOCK_TYPE_LFE) { | |
| 656 | ✗ | ATRAC9ChannelData *c = &b->channel[0]; | |
| 657 | ✗ | const int precision = reuse_params ? 8 : 4; | |
| 658 | ✗ | c->q_unit_cnt = b->q_unit_cnt = 2; | |
| 659 | |||
| 660 | ✗ | memset(c->scalefactors, 0, sizeof(c->scalefactors)); | |
| 661 | ✗ | memset(c->q_coeffs_fine, 0, sizeof(c->q_coeffs_fine)); | |
| 662 | ✗ | memset(c->q_coeffs_coarse, 0, sizeof(c->q_coeffs_coarse)); | |
| 663 | |||
| 664 | ✗ | for (int i = 0; i < b->q_unit_cnt; i++) { | |
| 665 | ✗ | c->scalefactors[i] = get_bits(gb, 5); | |
| 666 | ✗ | c->precision_coarse[i] = precision; | |
| 667 | ✗ | c->precision_fine[i] = 0; | |
| 668 | } | ||
| 669 | |||
| 670 | ✗ | for (int i = 0; i < c->q_unit_cnt; i++) { | |
| 671 | ✗ | const int start = at9_q_unit_to_coeff_idx[i + 0]; | |
| 672 | ✗ | const int end = at9_q_unit_to_coeff_idx[i + 1]; | |
| 673 | ✗ | for (int j = start; j < end; j++) | |
| 674 | ✗ | c->q_coeffs_coarse[j] = get_bits(gb, c->precision_coarse[i] + 1); | |
| 675 | } | ||
| 676 | |||
| 677 | ✗ | dequantize (s, b, c); | |
| 678 | ✗ | apply_scalefactors(s, b, 0); | |
| 679 | |||
| 680 | ✗ | goto imdct; | |
| 681 | } | ||
| 682 | |||
| 683 | ✗ | if (first_in_pkt && reuse_params) { | |
| 684 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "Invalid block flags!\n"); | |
| 685 | ✗ | return AVERROR_INVALIDDATA; | |
| 686 | } | ||
| 687 | |||
| 688 | /* Band parameters */ | ||
| 689 | ✗ | if (!reuse_params) { | |
| 690 | int stereo_band, ext_band; | ||
| 691 | ✗ | const int min_band_count = s->samplerate_idx > 7 ? 1 : 3; | |
| 692 | ✗ | b->reusable = 0; | |
| 693 | ✗ | b->band_count = get_bits(gb, 4) + min_band_count; | |
| 694 | ✗ | b->q_unit_cnt = at9_tab_band_q_unit_map[b->band_count]; | |
| 695 | |||
| 696 | ✗ | b->band_ext_q_unit = b->stereo_q_unit = b->q_unit_cnt; | |
| 697 | |||
| 698 | ✗ | if (b->band_count > at9_tab_sri_max_bands[s->samplerate_idx]) { | |
| 699 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "Invalid band count %i!\n", | |
| 700 | b->band_count); | ||
| 701 | ✗ | return AVERROR_INVALIDDATA; | |
| 702 | } | ||
| 703 | |||
| 704 | ✗ | if (stereo) { | |
| 705 | ✗ | stereo_band = get_bits(gb, 4) + min_band_count; | |
| 706 | ✗ | if (stereo_band > b->band_count) { | |
| 707 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "Invalid stereo band %i!\n", | |
| 708 | stereo_band); | ||
| 709 | ✗ | return AVERROR_INVALIDDATA; | |
| 710 | } | ||
| 711 | ✗ | b->stereo_q_unit = at9_tab_band_q_unit_map[stereo_band]; | |
| 712 | } | ||
| 713 | |||
| 714 | ✗ | b->has_band_ext = get_bits1(gb); | |
| 715 | ✗ | if (b->has_band_ext) { | |
| 716 | ✗ | ext_band = get_bits(gb, 4) + min_band_count; | |
| 717 | ✗ | if (ext_band < b->band_count) { | |
| 718 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "Invalid extension band %i!\n", | |
| 719 | ext_band); | ||
| 720 | ✗ | return AVERROR_INVALIDDATA; | |
| 721 | } | ||
| 722 | ✗ | b->band_ext_q_unit = at9_tab_band_q_unit_map[ext_band]; | |
| 723 | } | ||
| 724 | ✗ | b->reusable = 1; | |
| 725 | } | ||
| 726 | ✗ | if (!b->reusable) { | |
| 727 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "invalid block reused!\n"); | |
| 728 | ✗ | return AVERROR_INVALIDDATA; | |
| 729 | } | ||
| 730 | |||
| 731 | /* Calculate bit alloc gradient */ | ||
| 732 | ✗ | if (parse_gradient(s, b, gb)) | |
| 733 | ✗ | return AVERROR_INVALIDDATA; | |
| 734 | |||
| 735 | /* IS data */ | ||
| 736 | ✗ | b->cpe_base_channel = 0; | |
| 737 | ✗ | if (stereo) { | |
| 738 | ✗ | b->cpe_base_channel = get_bits1(gb); | |
| 739 | ✗ | if (get_bits1(gb)) { | |
| 740 | ✗ | for (int i = b->stereo_q_unit; i < b->q_unit_cnt; i++) | |
| 741 | ✗ | b->is_signs[i] = 1 - 2*get_bits1(gb); | |
| 742 | } else { | ||
| 743 | ✗ | for (int i = 0; i < FF_ARRAY_ELEMS(b->is_signs); i++) | |
| 744 | ✗ | b->is_signs[i] = 1; | |
| 745 | } | ||
| 746 | } | ||
| 747 | |||
| 748 | /* Band extension */ | ||
| 749 | ✗ | if (parse_band_ext(s, b, gb, stereo)) | |
| 750 | ✗ | return AVERROR_INVALIDDATA; | |
| 751 | |||
| 752 | /* Scalefactors */ | ||
| 753 | ✗ | for (int i = 0; i <= stereo; i++) { | |
| 754 | ✗ | ATRAC9ChannelData *c = &b->channel[i]; | |
| 755 | ✗ | c->q_unit_cnt = i == b->cpe_base_channel ? b->q_unit_cnt : | |
| 756 | b->stereo_q_unit; | ||
| 757 | ✗ | if (read_scalefactors(s, b, c, gb, i, first_in_pkt)) | |
| 758 | ✗ | return AVERROR_INVALIDDATA; | |
| 759 | |||
| 760 | ✗ | calc_precision (s, b, c); | |
| 761 | ✗ | calc_codebook_idx (s, b, c); | |
| 762 | ✗ | read_coeffs_coarse(s, b, c, gb); | |
| 763 | ✗ | read_coeffs_fine (s, b, c, gb); | |
| 764 | ✗ | dequantize (s, b, c); | |
| 765 | } | ||
| 766 | |||
| 767 | ✗ | b->q_unit_cnt_prev = b->has_band_ext ? b->band_ext_q_unit : b->q_unit_cnt; | |
| 768 | |||
| 769 | ✗ | apply_intensity_stereo(s, b, stereo); | |
| 770 | ✗ | apply_scalefactors (s, b, stereo); | |
| 771 | |||
| 772 | ✗ | if (b->has_band_ext && b->has_band_ext_data) | |
| 773 | ✗ | apply_band_extension (s, b, stereo); | |
| 774 | |||
| 775 | ✗ | imdct: | |
| 776 | ✗ | for (int i = 0; i <= stereo; i++) { | |
| 777 | ✗ | ATRAC9ChannelData *c = &b->channel[i]; | |
| 778 | ✗ | const int dst_idx = s->block_config->plane_map[block_idx][i]; | |
| 779 | ✗ | const int wsize = 1 << s->frame_log2; | |
| 780 | ✗ | const ptrdiff_t offset = wsize*frame_idx*sizeof(float); | |
| 781 | ✗ | float *dst = (float *)(frame->extended_data[dst_idx] + offset); | |
| 782 | |||
| 783 | ✗ | s->tx_fn(s->tx, s->temp, c->coeffs, sizeof(float)); | |
| 784 | ✗ | s->fdsp->vector_fmul_window(dst, c->prev_win, s->temp, | |
| 785 | ✗ | s->imdct_win, wsize >> 1); | |
| 786 | ✗ | memcpy(c->prev_win, s->temp + (wsize >> 1), sizeof(float)*wsize >> 1); | |
| 787 | } | ||
| 788 | |||
| 789 | ✗ | return 0; | |
| 790 | } | ||
| 791 | |||
| 792 | ✗ | static int atrac9_decode_frame(AVCodecContext *avctx, AVFrame *frame, | |
| 793 | int *got_frame_ptr, AVPacket *avpkt) | ||
| 794 | { | ||
| 795 | int ret; | ||
| 796 | GetBitContext gb; | ||
| 797 | ✗ | ATRAC9Context *s = avctx->priv_data; | |
| 798 | ✗ | const int frames = FFMIN(avpkt->size / s->avg_frame_size, s->frame_count); | |
| 799 | |||
| 800 | ✗ | frame->nb_samples = (1 << s->frame_log2) * frames; | |
| 801 | ✗ | ret = ff_get_buffer(avctx, frame, 0); | |
| 802 | ✗ | if (ret < 0) | |
| 803 | ✗ | return ret; | |
| 804 | |||
| 805 | ✗ | ret = init_get_bits8(&gb, avpkt->data, avpkt->size); | |
| 806 | ✗ | if (ret < 0) | |
| 807 | ✗ | return ret; | |
| 808 | |||
| 809 | ✗ | for (int i = 0; i < frames; i++) { | |
| 810 | ✗ | for (int j = 0; j < s->block_config->count; j++) { | |
| 811 | ✗ | ret = atrac9_decode_block(s, &gb, &s->block[j], frame, i, j); | |
| 812 | ✗ | if (ret) | |
| 813 | ✗ | return ret; | |
| 814 | ✗ | align_get_bits(&gb); | |
| 815 | } | ||
| 816 | } | ||
| 817 | |||
| 818 | ✗ | *got_frame_ptr = 1; | |
| 819 | |||
| 820 | ✗ | return avctx->block_align; | |
| 821 | } | ||
| 822 | |||
| 823 | ✗ | static av_cold void atrac9_decode_flush(AVCodecContext *avctx) | |
| 824 | { | ||
| 825 | ✗ | ATRAC9Context *s = avctx->priv_data; | |
| 826 | |||
| 827 | ✗ | for (int j = 0; j < s->block_config->count; j++) { | |
| 828 | ✗ | ATRAC9BlockData *b = &s->block[j]; | |
| 829 | ✗ | const int stereo = s->block_config->type[j] == ATRAC9_BLOCK_TYPE_CPE; | |
| 830 | ✗ | for (int i = 0; i <= stereo; i++) { | |
| 831 | ✗ | ATRAC9ChannelData *c = &b->channel[i]; | |
| 832 | ✗ | memset(c->prev_win, 0, sizeof(c->prev_win)); | |
| 833 | } | ||
| 834 | } | ||
| 835 | ✗ | } | |
| 836 | |||
| 837 | ✗ | static av_cold int atrac9_decode_close(AVCodecContext *avctx) | |
| 838 | { | ||
| 839 | ✗ | ATRAC9Context *s = avctx->priv_data; | |
| 840 | |||
| 841 | ✗ | av_tx_uninit(&s->tx); | |
| 842 | ✗ | av_freep(&s->fdsp); | |
| 843 | |||
| 844 | ✗ | return 0; | |
| 845 | } | ||
| 846 | |||
| 847 | ✗ | static av_cold const VLCElem *atrac9_init_vlc(VLCInitState *state, | |
| 848 | int nb_bits, int nb_codes, | ||
| 849 | const uint8_t (**tab)[2], int offset) | ||
| 850 | { | ||
| 851 | ✗ | const uint8_t (*table)[2] = *tab; | |
| 852 | |||
| 853 | ✗ | *tab += nb_codes; | |
| 854 | ✗ | return ff_vlc_init_tables_from_lengths(state, nb_bits, nb_codes, | |
| 855 | ✗ | &table[0][1], 2, &table[0][0], 2, 1, | |
| 856 | offset, 0); | ||
| 857 | } | ||
| 858 | |||
| 859 | ✗ | static av_cold void atrac9_init_static(void) | |
| 860 | { | ||
| 861 | static VLCElem vlc_buf[24812]; | ||
| 862 | ✗ | VLCInitState state = VLC_INIT_STATE(vlc_buf); | |
| 863 | const uint8_t (*tab)[2]; | ||
| 864 | |||
| 865 | /* Unsigned scalefactor VLCs */ | ||
| 866 | ✗ | tab = at9_sfb_a_tab; | |
| 867 | ✗ | for (int i = 1; i < 7; i++) { | |
| 868 | ✗ | const HuffmanCodebook *hf = &at9_huffman_sf_unsigned[i]; | |
| 869 | |||
| 870 | ✗ | sf_vlc[0][i] = atrac9_init_vlc(&state, ATRAC9_SF_VLC_BITS, | |
| 871 | ✗ | hf->size, &tab, 0); | |
| 872 | } | ||
| 873 | |||
| 874 | /* Signed scalefactor VLCs */ | ||
| 875 | ✗ | tab = at9_sfb_b_tab; | |
| 876 | ✗ | for (int i = 2; i < 6; i++) { | |
| 877 | ✗ | const HuffmanCodebook *hf = &at9_huffman_sf_signed[i]; | |
| 878 | |||
| 879 | /* The symbols are signed integers in the range -16..15; | ||
| 880 | * the values in the source table are offset by 16 to make | ||
| 881 | * them fit into an uint8_t; the -16 reverses this shift. */ | ||
| 882 | ✗ | sf_vlc[1][i] = atrac9_init_vlc(&state, ATRAC9_SF_VLC_BITS, | |
| 883 | ✗ | hf->size, &tab, -16); | |
| 884 | } | ||
| 885 | |||
| 886 | /* Coefficient VLCs */ | ||
| 887 | ✗ | tab = at9_coeffs_tab; | |
| 888 | ✗ | for (int i = 0; i < 2; i++) { | |
| 889 | ✗ | for (int j = 2; j < 8; j++) { | |
| 890 | ✗ | for (int k = i; k < 4; k++) { | |
| 891 | ✗ | const HuffmanCodebook *hf = &at9_huffman_coeffs[i][j][k]; | |
| 892 | ✗ | coeff_vlc[i][j][k] = atrac9_init_vlc(&state, ATRAC9_COEFF_VLC_BITS, | |
| 893 | ✗ | hf->size, &tab, 0); | |
| 894 | } | ||
| 895 | } | ||
| 896 | } | ||
| 897 | ✗ | } | |
| 898 | |||
| 899 | ✗ | static av_cold int atrac9_decode_init(AVCodecContext *avctx) | |
| 900 | { | ||
| 901 | float scale; | ||
| 902 | static AVOnce static_table_init = AV_ONCE_INIT; | ||
| 903 | GetBitContext gb; | ||
| 904 | ✗ | ATRAC9Context *s = avctx->priv_data; | |
| 905 | int err, version, block_config_idx, superframe_idx, alloc_c_len; | ||
| 906 | |||
| 907 | ✗ | s->avctx = avctx; | |
| 908 | |||
| 909 | ✗ | av_lfg_init(&s->lfg, 0xFBADF00D); | |
| 910 | |||
| 911 | ✗ | if (avctx->block_align <= 0) { | |
| 912 | ✗ | av_log(avctx, AV_LOG_ERROR, "Invalid block align\n"); | |
| 913 | ✗ | return AVERROR_INVALIDDATA; | |
| 914 | } | ||
| 915 | |||
| 916 | ✗ | if (avctx->extradata_size != 12) { | |
| 917 | ✗ | av_log(avctx, AV_LOG_ERROR, "Invalid extradata length!\n"); | |
| 918 | ✗ | return AVERROR_INVALIDDATA; | |
| 919 | } | ||
| 920 | |||
| 921 | ✗ | version = AV_RL32(avctx->extradata); | |
| 922 | ✗ | if (version > 2) { | |
| 923 | ✗ | av_log(avctx, AV_LOG_ERROR, "Unsupported version (%i)!\n", version); | |
| 924 | ✗ | return AVERROR_INVALIDDATA; | |
| 925 | } | ||
| 926 | |||
| 927 | ✗ | err = init_get_bits8(&gb, avctx->extradata + 4, avctx->extradata_size); | |
| 928 | ✗ | if (err < 0) | |
| 929 | ✗ | return err; | |
| 930 | |||
| 931 | ✗ | if (get_bits(&gb, 8) != 0xFE) { | |
| 932 | ✗ | av_log(avctx, AV_LOG_ERROR, "Incorrect magic byte!\n"); | |
| 933 | ✗ | return AVERROR_INVALIDDATA; | |
| 934 | } | ||
| 935 | |||
| 936 | ✗ | s->samplerate_idx = get_bits(&gb, 4); | |
| 937 | ✗ | avctx->sample_rate = at9_tab_samplerates[s->samplerate_idx]; | |
| 938 | |||
| 939 | ✗ | block_config_idx = get_bits(&gb, 3); | |
| 940 | ✗ | if (block_config_idx > 5) { | |
| 941 | ✗ | av_log(avctx, AV_LOG_ERROR, "Incorrect block config!\n"); | |
| 942 | ✗ | return AVERROR_INVALIDDATA; | |
| 943 | } | ||
| 944 | ✗ | s->block_config = &at9_block_layout[block_config_idx]; | |
| 945 | |||
| 946 | ✗ | av_channel_layout_uninit(&avctx->ch_layout); | |
| 947 | ✗ | avctx->ch_layout = s->block_config->channel_layout; | |
| 948 | ✗ | avctx->sample_fmt = AV_SAMPLE_FMT_FLTP; | |
| 949 | |||
| 950 | ✗ | if (get_bits1(&gb)) { | |
| 951 | ✗ | av_log(avctx, AV_LOG_ERROR, "Incorrect verification bit!\n"); | |
| 952 | ✗ | return AVERROR_INVALIDDATA; | |
| 953 | } | ||
| 954 | |||
| 955 | /* Average frame size in bytes */ | ||
| 956 | ✗ | s->avg_frame_size = get_bits(&gb, 11) + 1; | |
| 957 | |||
| 958 | ✗ | superframe_idx = get_bits(&gb, 2); | |
| 959 | ✗ | if (superframe_idx & 1) { | |
| 960 | ✗ | av_log(avctx, AV_LOG_ERROR, "Invalid superframe index!\n"); | |
| 961 | ✗ | return AVERROR_INVALIDDATA; | |
| 962 | } | ||
| 963 | |||
| 964 | ✗ | s->frame_count = 1 << superframe_idx; | |
| 965 | ✗ | s->frame_log2 = at9_tab_sri_frame_log2[s->samplerate_idx]; | |
| 966 | |||
| 967 | ✗ | scale = 1.0f / 32768.0; | |
| 968 | ✗ | err = av_tx_init(&s->tx, &s->tx_fn, AV_TX_FLOAT_MDCT, 1, | |
| 969 | ✗ | 1 << s->frame_log2, &scale, 0); | |
| 970 | ✗ | if (err < 0) | |
| 971 | ✗ | return err; | |
| 972 | |||
| 973 | ✗ | s->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT); | |
| 974 | ✗ | if (!s->fdsp) | |
| 975 | ✗ | return AVERROR(ENOMEM); | |
| 976 | |||
| 977 | /* iMDCT window */ | ||
| 978 | ✗ | for (int i = 0; i < (1 << s->frame_log2); i++) { | |
| 979 | ✗ | const int len = 1 << s->frame_log2; | |
| 980 | ✗ | const float sidx = ( i + 0.5f) / len; | |
| 981 | ✗ | const float eidx = (len - i - 0.5f) / len; | |
| 982 | ✗ | const float s_c = sinf(sidx*M_PI - M_PI_2)*0.5f + 0.5f; | |
| 983 | ✗ | const float e_c = sinf(eidx*M_PI - M_PI_2)*0.5f + 0.5f; | |
| 984 | ✗ | s->imdct_win[i] = s_c / ((s_c * s_c) + (e_c * e_c)); | |
| 985 | } | ||
| 986 | |||
| 987 | /* Allocation curve */ | ||
| 988 | ✗ | alloc_c_len = FF_ARRAY_ELEMS(at9_tab_b_dist); | |
| 989 | ✗ | for (int i = 1; i <= alloc_c_len; i++) | |
| 990 | ✗ | for (int j = 0; j < i; j++) | |
| 991 | ✗ | s->alloc_curve[i - 1][j] = at9_tab_b_dist[(j * alloc_c_len) / i]; | |
| 992 | |||
| 993 | ✗ | ff_thread_once(&static_table_init, atrac9_init_static); | |
| 994 | |||
| 995 | ✗ | return 0; | |
| 996 | } | ||
| 997 | |||
| 998 | const FFCodec ff_atrac9_decoder = { | ||
| 999 | .p.name = "atrac9", | ||
| 1000 | CODEC_LONG_NAME("ATRAC9 (Adaptive TRansform Acoustic Coding 9)"), | ||
| 1001 | .p.type = AVMEDIA_TYPE_AUDIO, | ||
| 1002 | .p.id = AV_CODEC_ID_ATRAC9, | ||
| 1003 | .priv_data_size = sizeof(ATRAC9Context), | ||
| 1004 | .init = atrac9_decode_init, | ||
| 1005 | .close = atrac9_decode_close, | ||
| 1006 | FF_CODEC_DECODE_CB(atrac9_decode_frame), | ||
| 1007 | .flush = atrac9_decode_flush, | ||
| 1008 | .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, | ||
| 1009 | .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_CHANNEL_CONF, | ||
| 1010 | }; | ||
| 1011 |