| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | /* | ||
| 2 | * ATRAC3 compatible decoder | ||
| 3 | * Copyright (c) 2006-2008 Maxim Poliakovski | ||
| 4 | * Copyright (c) 2006-2008 Benjamin Larsson | ||
| 5 | * | ||
| 6 | * This file is part of FFmpeg. | ||
| 7 | * | ||
| 8 | * FFmpeg is free software; you can redistribute it and/or | ||
| 9 | * modify it under the terms of the GNU Lesser General Public | ||
| 10 | * License as published by the Free Software Foundation; either | ||
| 11 | * version 2.1 of the License, or (at your option) any later version. | ||
| 12 | * | ||
| 13 | * FFmpeg is distributed in the hope that it will be useful, | ||
| 14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
| 16 | * Lesser General Public License for more details. | ||
| 17 | * | ||
| 18 | * You should have received a copy of the GNU Lesser General Public | ||
| 19 | * License along with FFmpeg; if not, write to the Free Software | ||
| 20 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
| 21 | */ | ||
| 22 | |||
| 23 | /** | ||
| 24 | * @file | ||
| 25 | * ATRAC3 compatible decoder. | ||
| 26 | * This decoder handles Sony's ATRAC3 data. | ||
| 27 | * | ||
| 28 | * Container formats used to store ATRAC3 data: | ||
| 29 | * RealMedia (.rm), RIFF WAV (.wav, .at3), Sony OpenMG (.oma, .aa3). | ||
| 30 | * | ||
| 31 | * To use this decoder, a calling application must supply the extradata | ||
| 32 | * bytes provided in the containers above. | ||
| 33 | */ | ||
| 34 | |||
| 35 | #include <math.h> | ||
| 36 | #include <stddef.h> | ||
| 37 | |||
| 38 | #include "libavutil/attributes.h" | ||
| 39 | #include "libavutil/float_dsp.h" | ||
| 40 | #include "libavutil/libm.h" | ||
| 41 | #include "libavutil/mem.h" | ||
| 42 | #include "libavutil/mem_internal.h" | ||
| 43 | #include "libavutil/thread.h" | ||
| 44 | #include "libavutil/tx.h" | ||
| 45 | |||
| 46 | #include "avcodec.h" | ||
| 47 | #include "bytestream.h" | ||
| 48 | #include "codec_internal.h" | ||
| 49 | #include "decode.h" | ||
| 50 | #include "get_bits.h" | ||
| 51 | |||
| 52 | #include "atrac.h" | ||
| 53 | #include "atrac3data.h" | ||
| 54 | |||
| 55 | #define MIN_CHANNELS 1 | ||
| 56 | #define MAX_CHANNELS 8 | ||
| 57 | #define MAX_JS_PAIRS 8 / 2 | ||
| 58 | |||
| 59 | #define JOINT_STEREO 0x12 | ||
| 60 | #define SINGLE 0x2 | ||
| 61 | |||
| 62 | #define SAMPLES_PER_FRAME 1024 | ||
| 63 | #define MDCT_SIZE 512 | ||
| 64 | |||
| 65 | #define ATRAC3_VLC_BITS 8 | ||
| 66 | |||
| 67 | typedef struct GainBlock { | ||
| 68 | AtracGainInfo g_block[4]; | ||
| 69 | } GainBlock; | ||
| 70 | |||
| 71 | typedef struct TonalComponent { | ||
| 72 | int pos; | ||
| 73 | int num_coefs; | ||
| 74 | float coef[8]; | ||
| 75 | } TonalComponent; | ||
| 76 | |||
| 77 | typedef struct ChannelUnit { | ||
| 78 | int bands_coded; | ||
| 79 | int num_components; | ||
| 80 | float prev_frame[SAMPLES_PER_FRAME]; | ||
| 81 | int gc_blk_switch; | ||
| 82 | TonalComponent components[64]; | ||
| 83 | GainBlock gain_block[2]; | ||
| 84 | |||
| 85 | DECLARE_ALIGNED(32, float, spectrum)[SAMPLES_PER_FRAME]; | ||
| 86 | DECLARE_ALIGNED(32, float, imdct_buf)[SAMPLES_PER_FRAME]; | ||
| 87 | |||
| 88 | float delay_buf1[46]; ///<qmf delay buffers | ||
| 89 | float delay_buf2[46]; | ||
| 90 | float delay_buf3[46]; | ||
| 91 | } ChannelUnit; | ||
| 92 | |||
| 93 | typedef struct ATRAC3Context { | ||
| 94 | GetBitContext gb; | ||
| 95 | //@{ | ||
| 96 | /** stream data */ | ||
| 97 | int coding_mode; | ||
| 98 | |||
| 99 | ChannelUnit *units; | ||
| 100 | //@} | ||
| 101 | //@{ | ||
| 102 | /** joint-stereo related variables */ | ||
| 103 | int matrix_coeff_index_prev[MAX_JS_PAIRS][4]; | ||
| 104 | int matrix_coeff_index_now[MAX_JS_PAIRS][4]; | ||
| 105 | int matrix_coeff_index_next[MAX_JS_PAIRS][4]; | ||
| 106 | int weighting_delay[MAX_JS_PAIRS][6]; | ||
| 107 | //@} | ||
| 108 | //@{ | ||
| 109 | /** data buffers */ | ||
| 110 | uint8_t *decoded_bytes_buffer; | ||
| 111 | float temp_buf[1070]; | ||
| 112 | //@} | ||
| 113 | //@{ | ||
| 114 | /** extradata */ | ||
| 115 | int scrambled_stream; | ||
| 116 | //@} | ||
| 117 | |||
| 118 | AtracGCContext gainc_ctx; | ||
| 119 | AVTXContext *mdct_ctx; | ||
| 120 | av_tx_fn mdct_fn; | ||
| 121 | void (*vector_fmul)(float *dst, const float *src0, const float *src1, | ||
| 122 | int len); | ||
| 123 | } ATRAC3Context; | ||
| 124 | |||
| 125 | static DECLARE_ALIGNED(32, float, mdct_window)[MDCT_SIZE]; | ||
| 126 | static VLCElem atrac3_vlc_table[7 * 1 << ATRAC3_VLC_BITS]; | ||
| 127 | static VLC spectral_coeff_tab[7]; | ||
| 128 | |||
| 129 | /** | ||
| 130 | * Regular 512 points IMDCT without overlapping, with the exception of the | ||
| 131 | * swapping of odd bands caused by the reverse spectra of the QMF. | ||
| 132 | * | ||
| 133 | * @param odd_band 1 if the band is an odd band | ||
| 134 | */ | ||
| 135 | 2804 | static void imlt(ATRAC3Context *q, float *input, float *output, int odd_band) | |
| 136 | { | ||
| 137 | int i; | ||
| 138 | |||
| 139 |
2/2✓ Branch 0 taken 848 times.
✓ Branch 1 taken 1956 times.
|
2804 | if (odd_band) { |
| 140 | /** | ||
| 141 | * Reverse the odd bands before IMDCT, this is an effect of the QMF | ||
| 142 | * transform or it gives better compression to do it this way. | ||
| 143 | * FIXME: It should be possible to handle this in imdct_calc | ||
| 144 | * for that to happen a modification of the prerotation step of | ||
| 145 | * all SIMD code and C code is needed. | ||
| 146 | * Or fix the functions before so they generate a pre reversed spectrum. | ||
| 147 | */ | ||
| 148 |
2/2✓ Branch 0 taken 108544 times.
✓ Branch 1 taken 848 times.
|
109392 | for (i = 0; i < 128; i++) |
| 149 | 108544 | FFSWAP(float, input[i], input[255 - i]); | |
| 150 | } | ||
| 151 | |||
| 152 | 2804 | q->mdct_fn(q->mdct_ctx, output, input, sizeof(float)); | |
| 153 | |||
| 154 | /* Perform windowing on the output. */ | ||
| 155 | 2804 | q->vector_fmul(output, output, mdct_window, MDCT_SIZE); | |
| 156 | 2804 | } | |
| 157 | |||
| 158 | /* | ||
| 159 | * indata descrambling, only used for data coming from the rm container | ||
| 160 | */ | ||
| 161 | ✗ | static int decode_bytes(const uint8_t *input, uint8_t *out, int bytes) | |
| 162 | { | ||
| 163 | int i, off; | ||
| 164 | uint32_t c; | ||
| 165 | const uint32_t *buf; | ||
| 166 | ✗ | uint32_t *output = (uint32_t *)out; | |
| 167 | |||
| 168 | ✗ | off = (intptr_t)input & 3; | |
| 169 | ✗ | buf = (const uint32_t *)(input - off); | |
| 170 | ✗ | if (off) | |
| 171 | ✗ | c = av_be2ne32((0x537F6103U >> (off * 8)) | (0x537F6103U << (32 - (off * 8)))); | |
| 172 | else | ||
| 173 | ✗ | c = av_be2ne32(0x537F6103U); | |
| 174 | ✗ | bytes += 3 + off; | |
| 175 | ✗ | for (i = 0; i < bytes / 4; i++) | |
| 176 | ✗ | output[i] = c ^ buf[i]; | |
| 177 | |||
| 178 | ✗ | if (off) | |
| 179 | ✗ | avpriv_request_sample(NULL, "Offset of %d", off); | |
| 180 | |||
| 181 | ✗ | return off; | |
| 182 | } | ||
| 183 | |||
| 184 | 6 | static av_cold void init_imdct_window(void) | |
| 185 | { | ||
| 186 | int i, j; | ||
| 187 | |||
| 188 | /* generate the mdct window, for details see | ||
| 189 | * http://wiki.multimedia.cx/index.php?title=RealAudio_atrc#Windows */ | ||
| 190 |
2/2✓ Branch 0 taken 768 times.
✓ Branch 1 taken 6 times.
|
774 | for (i = 0, j = 255; i < 128; i++, j--) { |
| 191 | 768 | float wi = sin(((i + 0.5) / 256.0 - 0.5) * M_PI) + 1.0; | |
| 192 | 768 | float wj = sin(((j + 0.5) / 256.0 - 0.5) * M_PI) + 1.0; | |
| 193 | 768 | float w = 0.5 * (wi * wi + wj * wj); | |
| 194 | 768 | mdct_window[i] = mdct_window[511 - i] = wi / w; | |
| 195 | 768 | mdct_window[j] = mdct_window[511 - j] = wj / w; | |
| 196 | } | ||
| 197 | 6 | } | |
| 198 | |||
| 199 | 9 | static av_cold int atrac3_decode_close(AVCodecContext *avctx) | |
| 200 | { | ||
| 201 | 9 | ATRAC3Context *q = avctx->priv_data; | |
| 202 | |||
| 203 | 9 | av_freep(&q->units); | |
| 204 | 9 | av_freep(&q->decoded_bytes_buffer); | |
| 205 | |||
| 206 | 9 | av_tx_uninit(&q->mdct_ctx); | |
| 207 | |||
| 208 | 9 | return 0; | |
| 209 | } | ||
| 210 | |||
| 211 | /** | ||
| 212 | * Mantissa decoding | ||
| 213 | * | ||
| 214 | * @param selector which table the output values are coded with | ||
| 215 | * @param coding_flag constant length coding or variable length coding | ||
| 216 | * @param mantissas mantissa output table | ||
| 217 | * @param num_codes number of values to get | ||
| 218 | */ | ||
| 219 | 28209 | static void read_quant_spectral_coeffs(GetBitContext *gb, int selector, | |
| 220 | int coding_flag, int *mantissas, | ||
| 221 | int num_codes) | ||
| 222 | { | ||
| 223 | int i, code, huff_symb; | ||
| 224 | |||
| 225 |
2/2✓ Branch 0 taken 15517 times.
✓ Branch 1 taken 12692 times.
|
28209 | if (selector == 1) |
| 226 | 15517 | num_codes /= 2; | |
| 227 | |||
| 228 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 28209 times.
|
28209 | if (coding_flag != 0) { |
| 229 | /* constant length coding (CLC) */ | ||
| 230 | ✗ | int num_bits = clc_length_tab[selector]; | |
| 231 | |||
| 232 | ✗ | if (selector > 1) { | |
| 233 | ✗ | for (i = 0; i < num_codes; i++) { | |
| 234 | ✗ | if (num_bits) | |
| 235 | ✗ | code = get_sbits(gb, num_bits); | |
| 236 | else | ||
| 237 | ✗ | code = 0; | |
| 238 | ✗ | mantissas[i] = code; | |
| 239 | } | ||
| 240 | } else { | ||
| 241 | ✗ | for (i = 0; i < num_codes; i++) { | |
| 242 | ✗ | if (num_bits) | |
| 243 | ✗ | code = get_bits(gb, num_bits); // num_bits is always 4 in this case | |
| 244 | else | ||
| 245 | ✗ | code = 0; | |
| 246 | ✗ | mantissas[i * 2 ] = mantissa_clc_tab[code >> 2]; | |
| 247 | ✗ | mantissas[i * 2 + 1] = mantissa_clc_tab[code & 3]; | |
| 248 | } | ||
| 249 | } | ||
| 250 | } else { | ||
| 251 | /* variable length coding (VLC) */ | ||
| 252 |
2/2✓ Branch 0 taken 12692 times.
✓ Branch 1 taken 15517 times.
|
28209 | if (selector != 1) { |
| 253 |
2/2✓ Branch 0 taken 179824 times.
✓ Branch 1 taken 12692 times.
|
192516 | for (i = 0; i < num_codes; i++) { |
| 254 | 179824 | mantissas[i] = get_vlc2(gb, spectral_coeff_tab[selector-1].table, | |
| 255 | ATRAC3_VLC_BITS, 1); | ||
| 256 | } | ||
| 257 | } else { | ||
| 258 |
2/2✓ Branch 0 taken 234888 times.
✓ Branch 1 taken 15517 times.
|
250405 | for (i = 0; i < num_codes; i++) { |
| 259 | 234888 | huff_symb = get_vlc2(gb, spectral_coeff_tab[selector - 1].table, | |
| 260 | ATRAC3_VLC_BITS, 1); | ||
| 261 | 234888 | mantissas[i * 2 ] = mantissa_vlc_tab[huff_symb * 2 ]; | |
| 262 | 234888 | mantissas[i * 2 + 1] = mantissa_vlc_tab[huff_symb * 2 + 1]; | |
| 263 | } | ||
| 264 | } | ||
| 265 | } | ||
| 266 | 28209 | } | |
| 267 | |||
| 268 | /** | ||
| 269 | * Restore the quantized band spectrum coefficients | ||
| 270 | * | ||
| 271 | * @return subband count, fix for broken specification/files | ||
| 272 | */ | ||
| 273 | 1108 | static int decode_spectrum(GetBitContext *gb, float *output) | |
| 274 | { | ||
| 275 | int num_subbands, coding_mode, i, j, first, last, subband_size; | ||
| 276 | int subband_vlc_index[32], sf_index[32]; | ||
| 277 | int mantissas[128]; | ||
| 278 | float scale_factor; | ||
| 279 | |||
| 280 | 1108 | num_subbands = get_bits(gb, 5); // number of coded subbands | |
| 281 | 1108 | coding_mode = get_bits1(gb); // coding Mode: 0 - VLC/ 1-CLC | |
| 282 | |||
| 283 | /* get the VLC selector table for the subbands, 0 means not coded */ | ||
| 284 |
2/2✓ Branch 0 taken 28209 times.
✓ Branch 1 taken 1108 times.
|
29317 | for (i = 0; i <= num_subbands; i++) |
| 285 | 28209 | subband_vlc_index[i] = get_bits(gb, 3); | |
| 286 | |||
| 287 | /* read the scale factor indexes from the stream */ | ||
| 288 |
2/2✓ Branch 0 taken 28209 times.
✓ Branch 1 taken 1108 times.
|
29317 | for (i = 0; i <= num_subbands; i++) { |
| 289 |
1/2✓ Branch 0 taken 28209 times.
✗ Branch 1 not taken.
|
28209 | if (subband_vlc_index[i] != 0) |
| 290 | 28209 | sf_index[i] = get_bits(gb, 6); | |
| 291 | } | ||
| 292 | |||
| 293 |
2/2✓ Branch 0 taken 28209 times.
✓ Branch 1 taken 1108 times.
|
29317 | for (i = 0; i <= num_subbands; i++) { |
| 294 | 28209 | first = subband_tab[i ]; | |
| 295 | 28209 | last = subband_tab[i + 1]; | |
| 296 | |||
| 297 | 28209 | subband_size = last - first; | |
| 298 | |||
| 299 |
1/2✓ Branch 0 taken 28209 times.
✗ Branch 1 not taken.
|
28209 | if (subband_vlc_index[i] != 0) { |
| 300 | /* decode spectral coefficients for this subband */ | ||
| 301 | /* TODO: This can be done faster is several blocks share the | ||
| 302 | * same VLC selector (subband_vlc_index) */ | ||
| 303 | 28209 | read_quant_spectral_coeffs(gb, subband_vlc_index[i], coding_mode, | |
| 304 | mantissas, subband_size); | ||
| 305 | |||
| 306 | /* decode the scale factor for this subband */ | ||
| 307 | 28209 | scale_factor = ff_atrac_sf_table[sf_index[i]] * | |
| 308 | 28209 | inv_max_quant[subband_vlc_index[i]]; | |
| 309 | |||
| 310 | /* inverse quantize the coefficients */ | ||
| 311 |
2/2✓ Branch 0 taken 649600 times.
✓ Branch 1 taken 28209 times.
|
677809 | for (j = 0; first < last; first++, j++) |
| 312 | 649600 | output[first] = mantissas[j] * scale_factor; | |
| 313 | } else { | ||
| 314 | /* this subband was not coded, so zero the entire subband */ | ||
| 315 | ✗ | memset(output + first, 0, subband_size * sizeof(*output)); | |
| 316 | } | ||
| 317 | } | ||
| 318 | |||
| 319 | /* clear the subbands that were not coded */ | ||
| 320 | 1108 | first = subband_tab[i]; | |
| 321 | 1108 | memset(output + first, 0, (SAMPLES_PER_FRAME - first) * sizeof(*output)); | |
| 322 | 1108 | return num_subbands; | |
| 323 | } | ||
| 324 | |||
| 325 | /** | ||
| 326 | * Restore the quantized tonal components | ||
| 327 | * | ||
| 328 | * @param components tonal components | ||
| 329 | * @param num_bands number of coded bands | ||
| 330 | */ | ||
| 331 | 1108 | static int decode_tonal_components(GetBitContext *gb, | |
| 332 | TonalComponent *components, int num_bands) | ||
| 333 | { | ||
| 334 | int i, b, c, m; | ||
| 335 | int nb_components, coding_mode_selector, coding_mode; | ||
| 336 | int band_flags[4], mantissa[8]; | ||
| 337 | 1108 | int component_count = 0; | |
| 338 | |||
| 339 | 1108 | nb_components = get_bits(gb, 5); | |
| 340 | |||
| 341 | /* no tonal components */ | ||
| 342 |
1/2✓ Branch 0 taken 1108 times.
✗ Branch 1 not taken.
|
1108 | if (nb_components == 0) |
| 343 | 1108 | return 0; | |
| 344 | |||
| 345 | ✗ | coding_mode_selector = get_bits(gb, 2); | |
| 346 | ✗ | if (coding_mode_selector == 2) | |
| 347 | ✗ | return AVERROR_INVALIDDATA; | |
| 348 | |||
| 349 | ✗ | coding_mode = coding_mode_selector & 1; | |
| 350 | |||
| 351 | ✗ | for (i = 0; i < nb_components; i++) { | |
| 352 | int coded_values_per_component, quant_step_index; | ||
| 353 | |||
| 354 | ✗ | for (b = 0; b <= num_bands; b++) | |
| 355 | ✗ | band_flags[b] = get_bits1(gb); | |
| 356 | |||
| 357 | ✗ | coded_values_per_component = get_bits(gb, 3); | |
| 358 | |||
| 359 | ✗ | quant_step_index = get_bits(gb, 3); | |
| 360 | ✗ | if (quant_step_index <= 1) | |
| 361 | ✗ | return AVERROR_INVALIDDATA; | |
| 362 | |||
| 363 | ✗ | if (coding_mode_selector == 3) | |
| 364 | ✗ | coding_mode = get_bits1(gb); | |
| 365 | |||
| 366 | ✗ | for (b = 0; b < (num_bands + 1) * 4; b++) { | |
| 367 | int coded_components; | ||
| 368 | |||
| 369 | ✗ | if (band_flags[b >> 2] == 0) | |
| 370 | ✗ | continue; | |
| 371 | |||
| 372 | ✗ | coded_components = get_bits(gb, 3); | |
| 373 | |||
| 374 | ✗ | for (c = 0; c < coded_components; c++) { | |
| 375 | ✗ | TonalComponent *cmp = &components[component_count]; | |
| 376 | int sf_index, coded_values, max_coded_values; | ||
| 377 | float scale_factor; | ||
| 378 | |||
| 379 | ✗ | sf_index = get_bits(gb, 6); | |
| 380 | ✗ | if (component_count >= 64) | |
| 381 | ✗ | return AVERROR_INVALIDDATA; | |
| 382 | |||
| 383 | ✗ | cmp->pos = b * 64 + get_bits(gb, 6); | |
| 384 | |||
| 385 | ✗ | max_coded_values = SAMPLES_PER_FRAME - cmp->pos; | |
| 386 | ✗ | coded_values = coded_values_per_component + 1; | |
| 387 | ✗ | coded_values = FFMIN(max_coded_values, coded_values); | |
| 388 | |||
| 389 | ✗ | scale_factor = ff_atrac_sf_table[sf_index] * | |
| 390 | ✗ | inv_max_quant[quant_step_index]; | |
| 391 | |||
| 392 | ✗ | read_quant_spectral_coeffs(gb, quant_step_index, coding_mode, | |
| 393 | mantissa, coded_values); | ||
| 394 | |||
| 395 | ✗ | cmp->num_coefs = coded_values; | |
| 396 | |||
| 397 | /* inverse quant */ | ||
| 398 | ✗ | for (m = 0; m < coded_values; m++) | |
| 399 | ✗ | cmp->coef[m] = mantissa[m] * scale_factor; | |
| 400 | |||
| 401 | ✗ | component_count++; | |
| 402 | } | ||
| 403 | } | ||
| 404 | } | ||
| 405 | |||
| 406 | ✗ | return component_count; | |
| 407 | } | ||
| 408 | |||
| 409 | /** | ||
| 410 | * Decode gain parameters for the coded bands | ||
| 411 | * | ||
| 412 | * @param block the gainblock for the current band | ||
| 413 | * @param num_bands amount of coded bands | ||
| 414 | */ | ||
| 415 | 1108 | static int decode_gain_control(GetBitContext *gb, GainBlock *block, | |
| 416 | int num_bands) | ||
| 417 | { | ||
| 418 | int b, j; | ||
| 419 | int *level, *loc; | ||
| 420 | |||
| 421 | 1108 | AtracGainInfo *gain = block->g_block; | |
| 422 | |||
| 423 |
2/2✓ Branch 0 taken 2804 times.
✓ Branch 1 taken 1108 times.
|
3912 | for (b = 0; b <= num_bands; b++) { |
| 424 | 2804 | gain[b].num_points = get_bits(gb, 3); | |
| 425 | 2804 | level = gain[b].lev_code; | |
| 426 | 2804 | loc = gain[b].loc_code; | |
| 427 | |||
| 428 |
2/2✓ Branch 0 taken 874 times.
✓ Branch 1 taken 2804 times.
|
3678 | for (j = 0; j < gain[b].num_points; j++) { |
| 429 | 874 | level[j] = get_bits(gb, 4); | |
| 430 | 874 | loc[j] = get_bits(gb, 5); | |
| 431 |
3/4✓ Branch 0 taken 295 times.
✓ Branch 1 taken 579 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 295 times.
|
874 | if (j && loc[j] <= loc[j - 1]) |
| 432 | ✗ | return AVERROR_INVALIDDATA; | |
| 433 | } | ||
| 434 | } | ||
| 435 | |||
| 436 | /* Clear the unused blocks. */ | ||
| 437 |
2/2✓ Branch 0 taken 1628 times.
✓ Branch 1 taken 1108 times.
|
2736 | for (; b < 4 ; b++) |
| 438 | 1628 | gain[b].num_points = 0; | |
| 439 | |||
| 440 | 1108 | return 0; | |
| 441 | } | ||
| 442 | |||
| 443 | /** | ||
| 444 | * Combine the tonal band spectrum and regular band spectrum | ||
| 445 | * | ||
| 446 | * @param spectrum output spectrum buffer | ||
| 447 | * @param num_components number of tonal components | ||
| 448 | * @param components tonal components for this band | ||
| 449 | * @return position of the last tonal coefficient | ||
| 450 | */ | ||
| 451 | 1108 | static int add_tonal_components(float *spectrum, int num_components, | |
| 452 | TonalComponent *components) | ||
| 453 | { | ||
| 454 | 1108 | int i, j, last_pos = -1; | |
| 455 | float *input, *output; | ||
| 456 | |||
| 457 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1108 times.
|
1108 | for (i = 0; i < num_components; i++) { |
| 458 | ✗ | last_pos = FFMAX(components[i].pos + components[i].num_coefs, last_pos); | |
| 459 | ✗ | input = components[i].coef; | |
| 460 | ✗ | output = &spectrum[components[i].pos]; | |
| 461 | |||
| 462 | ✗ | for (j = 0; j < components[i].num_coefs; j++) | |
| 463 | ✗ | output[j] += input[j]; | |
| 464 | } | ||
| 465 | |||
| 466 | 1108 | return last_pos; | |
| 467 | } | ||
| 468 | |||
| 469 | #define INTERPOLATE(old, new, nsample) \ | ||
| 470 | ((old) + (nsample) * 0.125 * ((new) - (old))) | ||
| 471 | |||
| 472 | 260 | static void reverse_matrixing(float *su1, float *su2, int *prev_code, | |
| 473 | int *curr_code) | ||
| 474 | { | ||
| 475 | int i, nsample, band; | ||
| 476 | float mc1_l, mc1_r, mc2_l, mc2_r; | ||
| 477 | |||
| 478 |
2/2✓ Branch 0 taken 1040 times.
✓ Branch 1 taken 260 times.
|
1300 | for (i = 0, band = 0; band < 4 * 256; band += 256, i++) { |
| 479 | 1040 | int s1 = prev_code[i]; | |
| 480 | 1040 | int s2 = curr_code[i]; | |
| 481 | 1040 | nsample = band; | |
| 482 | |||
| 483 |
2/2✓ Branch 0 taken 38 times.
✓ Branch 1 taken 1002 times.
|
1040 | if (s1 != s2) { |
| 484 | /* Selector value changed, interpolation needed. */ | ||
| 485 | 38 | mc1_l = matrix_coeffs[s1 * 2 ]; | |
| 486 | 38 | mc1_r = matrix_coeffs[s1 * 2 + 1]; | |
| 487 | 38 | mc2_l = matrix_coeffs[s2 * 2 ]; | |
| 488 | 38 | mc2_r = matrix_coeffs[s2 * 2 + 1]; | |
| 489 | |||
| 490 | /* Interpolation is done over the first eight samples. */ | ||
| 491 |
2/2✓ Branch 0 taken 304 times.
✓ Branch 1 taken 38 times.
|
342 | for (; nsample < band + 8; nsample++) { |
| 492 | 304 | float c1 = su1[nsample]; | |
| 493 | 304 | float c2 = su2[nsample]; | |
| 494 | 304 | c2 = c1 * INTERPOLATE(mc1_l, mc2_l, nsample - band) + | |
| 495 | 304 | c2 * INTERPOLATE(mc1_r, mc2_r, nsample - band); | |
| 496 | 304 | su1[nsample] = c2; | |
| 497 | 304 | su2[nsample] = c1 * 2.0 - c2; | |
| 498 | } | ||
| 499 | } | ||
| 500 | |||
| 501 | /* Apply the matrix without interpolation. */ | ||
| 502 |
2/4✓ Branch 0 taken 45 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 995 times.
✗ Branch 3 not taken.
|
1040 | switch (s2) { |
| 503 | 45 | case 0: /* M/S decoding */ | |
| 504 |
2/2✓ Branch 0 taken 11368 times.
✓ Branch 1 taken 45 times.
|
11413 | for (; nsample < band + 256; nsample++) { |
| 505 | 11368 | float c1 = su1[nsample]; | |
| 506 | 11368 | float c2 = su2[nsample]; | |
| 507 | 11368 | su1[nsample] = c2 * 2.0; | |
| 508 | 11368 | su2[nsample] = (c1 - c2) * 2.0; | |
| 509 | } | ||
| 510 | 45 | break; | |
| 511 | ✗ | case 1: | |
| 512 | ✗ | for (; nsample < band + 256; nsample++) { | |
| 513 | ✗ | float c1 = su1[nsample]; | |
| 514 | ✗ | float c2 = su2[nsample]; | |
| 515 | ✗ | su1[nsample] = (c1 + c2) * 2.0; | |
| 516 | ✗ | su2[nsample] = c2 * -2.0; | |
| 517 | } | ||
| 518 | ✗ | break; | |
| 519 | 995 | case 2: | |
| 520 | case 3: | ||
| 521 |
2/2✓ Branch 0 taken 254568 times.
✓ Branch 1 taken 995 times.
|
255563 | for (; nsample < band + 256; nsample++) { |
| 522 | 254568 | float c1 = su1[nsample]; | |
| 523 | 254568 | float c2 = su2[nsample]; | |
| 524 | 254568 | su1[nsample] = c1 + c2; | |
| 525 | 254568 | su2[nsample] = c1 - c2; | |
| 526 | } | ||
| 527 | 995 | break; | |
| 528 | ✗ | default: | |
| 529 | ✗ | av_unreachable("curr_code/matrix_coeff_index_* values are stored in two bits"); | |
| 530 | } | ||
| 531 | } | ||
| 532 | 260 | } | |
| 533 | |||
| 534 | 492 | static void get_channel_weights(int index, int flag, float ch[2]) | |
| 535 | { | ||
| 536 |
2/2✓ Branch 0 taken 53 times.
✓ Branch 1 taken 439 times.
|
492 | if (index == 7) { |
| 537 | 53 | ch[0] = 1.0; | |
| 538 | 53 | ch[1] = 1.0; | |
| 539 | } else { | ||
| 540 | 439 | ch[0] = (index & 7) / 7.0; | |
| 541 | 439 | ch[1] = sqrt(2 - ch[0] * ch[0]); | |
| 542 |
2/2✓ Branch 0 taken 115 times.
✓ Branch 1 taken 324 times.
|
439 | if (flag) |
| 543 | 115 | FFSWAP(float, ch[0], ch[1]); | |
| 544 | } | ||
| 545 | 492 | } | |
| 546 | |||
| 547 | 260 | static void channel_weighting(float *su1, float *su2, int *p3) | |
| 548 | { | ||
| 549 | int band, nsample; | ||
| 550 | /* w[x][y] y=0 is left y=1 is right */ | ||
| 551 | float w[2][2]; | ||
| 552 | |||
| 553 |
4/4✓ Branch 0 taken 41 times.
✓ Branch 1 taken 219 times.
✓ Branch 2 taken 27 times.
✓ Branch 3 taken 14 times.
|
260 | if (p3[1] != 7 || p3[3] != 7) { |
| 554 | 246 | get_channel_weights(p3[1], p3[0], w[0]); | |
| 555 | 246 | get_channel_weights(p3[3], p3[2], w[1]); | |
| 556 | |||
| 557 |
2/2✓ Branch 0 taken 738 times.
✓ Branch 1 taken 246 times.
|
984 | for (band = 256; band < 4 * 256; band += 256) { |
| 558 |
2/2✓ Branch 0 taken 5904 times.
✓ Branch 1 taken 738 times.
|
6642 | for (nsample = band; nsample < band + 8; nsample++) { |
| 559 | 5904 | su1[nsample] *= INTERPOLATE(w[0][0], w[0][1], nsample - band); | |
| 560 | 5904 | su2[nsample] *= INTERPOLATE(w[1][0], w[1][1], nsample - band); | |
| 561 | } | ||
| 562 |
2/2✓ Branch 0 taken 183024 times.
✓ Branch 1 taken 738 times.
|
183762 | for(; nsample < band + 256; nsample++) { |
| 563 | 183024 | su1[nsample] *= w[1][0]; | |
| 564 | 183024 | su2[nsample] *= w[1][1]; | |
| 565 | } | ||
| 566 | } | ||
| 567 | } | ||
| 568 | 260 | } | |
| 569 | |||
| 570 | /** | ||
| 571 | * Decode a Sound Unit | ||
| 572 | * | ||
| 573 | * @param snd the channel unit to be used | ||
| 574 | * @param output the decoded samples before IQMF in float representation | ||
| 575 | * @param channel_num channel number | ||
| 576 | * @param coding_mode the coding mode (JOINT_STEREO or single channels) | ||
| 577 | */ | ||
| 578 | 1108 | static int decode_channel_sound_unit(ATRAC3Context *q, GetBitContext *gb, | |
| 579 | ChannelUnit *snd, float *output, | ||
| 580 | int channel_num, int coding_mode) | ||
| 581 | { | ||
| 582 | int band, ret, num_subbands, last_tonal, num_bands; | ||
| 583 | 1108 | GainBlock *gain1 = &snd->gain_block[ snd->gc_blk_switch]; | |
| 584 | 1108 | GainBlock *gain2 = &snd->gain_block[1 - snd->gc_blk_switch]; | |
| 585 | |||
| 586 |
4/4✓ Branch 0 taken 520 times.
✓ Branch 1 taken 588 times.
✓ Branch 2 taken 260 times.
✓ Branch 3 taken 260 times.
|
1108 | if (coding_mode == JOINT_STEREO && (channel_num % 2) == 1) { |
| 587 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 260 times.
|
260 | if (get_bits(gb, 2) != 3) { |
| 588 | ✗ | av_log(NULL,AV_LOG_ERROR,"JS mono Sound Unit id != 3.\n"); | |
| 589 | ✗ | return AVERROR_INVALIDDATA; | |
| 590 | } | ||
| 591 | } else { | ||
| 592 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 848 times.
|
848 | if (get_bits(gb, 6) != 0x28) { |
| 593 | ✗ | av_log(NULL,AV_LOG_ERROR,"Sound Unit id != 0x28.\n"); | |
| 594 | ✗ | return AVERROR_INVALIDDATA; | |
| 595 | } | ||
| 596 | } | ||
| 597 | |||
| 598 | /* number of coded QMF bands */ | ||
| 599 | 1108 | snd->bands_coded = get_bits(gb, 2); | |
| 600 | |||
| 601 | 1108 | ret = decode_gain_control(gb, gain2, snd->bands_coded); | |
| 602 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1108 times.
|
1108 | if (ret) |
| 603 | ✗ | return ret; | |
| 604 | |||
| 605 | 1108 | snd->num_components = decode_tonal_components(gb, snd->components, | |
| 606 | snd->bands_coded); | ||
| 607 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1108 times.
|
1108 | if (snd->num_components < 0) |
| 608 | ✗ | return snd->num_components; | |
| 609 | |||
| 610 | 1108 | num_subbands = decode_spectrum(gb, snd->spectrum); | |
| 611 | |||
| 612 | /* Merge the decoded spectrum and tonal components. */ | ||
| 613 | 1108 | last_tonal = add_tonal_components(snd->spectrum, snd->num_components, | |
| 614 | 1108 | snd->components); | |
| 615 | |||
| 616 | |||
| 617 | /* calculate number of used MLT/QMF bands according to the amount of coded | ||
| 618 | spectral lines */ | ||
| 619 | 1108 | num_bands = (subband_tab[num_subbands + 1] - 1) >> 8; | |
| 620 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1108 times.
|
1108 | if (last_tonal >= 0) |
| 621 | ✗ | num_bands = FFMAX((last_tonal + 256) >> 8, num_bands); | |
| 622 | |||
| 623 | |||
| 624 | /* Reconstruct time domain samples. */ | ||
| 625 |
2/2✓ Branch 0 taken 4432 times.
✓ Branch 1 taken 1108 times.
|
5540 | for (band = 0; band < 4; band++) { |
| 626 | /* Perform the IMDCT step without overlapping. */ | ||
| 627 |
2/2✓ Branch 0 taken 2804 times.
✓ Branch 1 taken 1628 times.
|
4432 | if (band <= num_bands) |
| 628 | 2804 | imlt(q, &snd->spectrum[band * 256], snd->imdct_buf, band & 1); | |
| 629 | else | ||
| 630 | 1628 | memset(snd->imdct_buf, 0, 512 * sizeof(*snd->imdct_buf)); | |
| 631 | |||
| 632 | /* gain compensation and overlapping */ | ||
| 633 | 4432 | ff_atrac_gain_compensation(&q->gainc_ctx, snd->imdct_buf, | |
| 634 | 4432 | &snd->prev_frame[band * 256], | |
| 635 | &gain1->g_block[band], &gain2->g_block[band], | ||
| 636 | 4432 | 256, &output[band * 256]); | |
| 637 | } | ||
| 638 | |||
| 639 | /* Swap the gain control buffers for the next frame. */ | ||
| 640 | 1108 | snd->gc_blk_switch ^= 1; | |
| 641 | |||
| 642 | 1108 | return 0; | |
| 643 | } | ||
| 644 | |||
| 645 | 554 | static int decode_frame(AVCodecContext *avctx, const uint8_t *databuf, | |
| 646 | float **out_samples) | ||
| 647 | { | ||
| 648 | 554 | ATRAC3Context *q = avctx->priv_data; | |
| 649 | int ret, i, ch; | ||
| 650 | uint8_t *ptr1; | ||
| 651 | 554 | int channels = avctx->ch_layout.nb_channels; | |
| 652 | |||
| 653 |
2/2✓ Branch 0 taken 260 times.
✓ Branch 1 taken 294 times.
|
554 | if (q->coding_mode == JOINT_STEREO) { |
| 654 | /* channel coupling mode */ | ||
| 655 | |||
| 656 | /* Decode sound unit pairs (channels are expected to be even). | ||
| 657 | * Multichannel joint stereo interleaves pairs (6ch: 2ch + 2ch + 2ch) */ | ||
| 658 | const uint8_t *js_databuf; | ||
| 659 | int js_pair, js_block_align; | ||
| 660 | |||
| 661 | 260 | js_block_align = (avctx->block_align / channels) * 2; /* block pair */ | |
| 662 | |||
| 663 |
2/2✓ Branch 0 taken 260 times.
✓ Branch 1 taken 260 times.
|
520 | for (ch = 0; ch < channels; ch = ch + 2) { |
| 664 | 260 | js_pair = ch/2; | |
| 665 | 260 | js_databuf = databuf + js_pair * js_block_align; /* align to current pair */ | |
| 666 | |||
| 667 | /* Set the bitstream reader at the start of first channel sound unit. */ | ||
| 668 | 260 | init_get_bits(&q->gb, | |
| 669 | js_databuf, js_block_align * 8); | ||
| 670 | |||
| 671 | /* decode Sound Unit 1 */ | ||
| 672 | 260 | ret = decode_channel_sound_unit(q, &q->gb, &q->units[ch], | |
| 673 | 260 | out_samples[ch], ch, JOINT_STEREO); | |
| 674 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 260 times.
|
260 | if (ret != 0) |
| 675 | ✗ | return ret; | |
| 676 | |||
| 677 | /* Framedata of the su2 in the joint-stereo mode is encoded in | ||
| 678 | * reverse byte order so we need to swap it first. */ | ||
| 679 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 260 times.
|
260 | if (js_databuf == q->decoded_bytes_buffer) { |
| 680 | ✗ | uint8_t *ptr2 = q->decoded_bytes_buffer + js_block_align - 1; | |
| 681 | ✗ | ptr1 = q->decoded_bytes_buffer; | |
| 682 | ✗ | for (i = 0; i < js_block_align / 2; i++, ptr1++, ptr2--) | |
| 683 | ✗ | FFSWAP(uint8_t, *ptr1, *ptr2); | |
| 684 | } else { | ||
| 685 | 260 | const uint8_t *ptr2 = js_databuf + js_block_align - 1; | |
| 686 |
2/2✓ Branch 0 taken 49920 times.
✓ Branch 1 taken 260 times.
|
50180 | for (i = 0; i < js_block_align; i++) |
| 687 | 49920 | q->decoded_bytes_buffer[i] = *ptr2--; | |
| 688 | } | ||
| 689 | |||
| 690 | /* Skip the sync codes (0xF8). */ | ||
| 691 | 260 | ptr1 = q->decoded_bytes_buffer; | |
| 692 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 260 times.
|
260 | for (i = 4; *ptr1 == 0xF8; i++, ptr1++) { |
| 693 | ✗ | if (i >= js_block_align) | |
| 694 | ✗ | return AVERROR_INVALIDDATA; | |
| 695 | } | ||
| 696 | |||
| 697 | |||
| 698 | /* set the bitstream reader at the start of the second Sound Unit */ | ||
| 699 | 260 | ret = init_get_bits8(&q->gb, | |
| 700 | 260 | ptr1, q->decoded_bytes_buffer + js_block_align - ptr1); | |
| 701 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 260 times.
|
260 | if (ret < 0) |
| 702 | ✗ | return ret; | |
| 703 | |||
| 704 | /* Fill the Weighting coeffs delay buffer */ | ||
| 705 | 260 | memmove(q->weighting_delay[js_pair], &q->weighting_delay[js_pair][2], | |
| 706 | 4 * sizeof(*q->weighting_delay[js_pair])); | ||
| 707 | 260 | q->weighting_delay[js_pair][4] = get_bits1(&q->gb); | |
| 708 | 260 | q->weighting_delay[js_pair][5] = get_bits(&q->gb, 3); | |
| 709 | |||
| 710 |
2/2✓ Branch 0 taken 1040 times.
✓ Branch 1 taken 260 times.
|
1300 | for (i = 0; i < 4; i++) { |
| 711 | 1040 | q->matrix_coeff_index_prev[js_pair][i] = q->matrix_coeff_index_now[js_pair][i]; | |
| 712 | 1040 | q->matrix_coeff_index_now[js_pair][i] = q->matrix_coeff_index_next[js_pair][i]; | |
| 713 | 1040 | q->matrix_coeff_index_next[js_pair][i] = get_bits(&q->gb, 2); | |
| 714 | } | ||
| 715 | |||
| 716 | /* Decode Sound Unit 2. */ | ||
| 717 | 260 | ret = decode_channel_sound_unit(q, &q->gb, &q->units[ch+1], | |
| 718 | 260 | out_samples[ch+1], ch+1, JOINT_STEREO); | |
| 719 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 260 times.
|
260 | if (ret != 0) |
| 720 | ✗ | return ret; | |
| 721 | |||
| 722 | /* Reconstruct the channel coefficients. */ | ||
| 723 | 260 | reverse_matrixing(out_samples[ch], out_samples[ch+1], | |
| 724 | 260 | q->matrix_coeff_index_prev[js_pair], | |
| 725 | 260 | q->matrix_coeff_index_now[js_pair]); | |
| 726 | |||
| 727 | 260 | channel_weighting(out_samples[ch], out_samples[ch+1], q->weighting_delay[js_pair]); | |
| 728 | } | ||
| 729 | } else { | ||
| 730 | /* single channels */ | ||
| 731 | /* Decode the channel sound units. */ | ||
| 732 |
2/2✓ Branch 0 taken 588 times.
✓ Branch 1 taken 294 times.
|
882 | for (i = 0; i < channels; i++) { |
| 733 | /* Set the bitstream reader at the start of a channel sound unit. */ | ||
| 734 | 588 | init_get_bits(&q->gb, | |
| 735 | 588 | databuf + i * avctx->block_align / channels, | |
| 736 | 588 | avctx->block_align * 8 / channels); | |
| 737 | |||
| 738 | 588 | ret = decode_channel_sound_unit(q, &q->gb, &q->units[i], | |
| 739 | 588 | out_samples[i], i, q->coding_mode); | |
| 740 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 588 times.
|
588 | if (ret != 0) |
| 741 | ✗ | return ret; | |
| 742 | } | ||
| 743 | } | ||
| 744 | |||
| 745 | /* Apply the iQMF synthesis filter. */ | ||
| 746 |
2/2✓ Branch 0 taken 1108 times.
✓ Branch 1 taken 554 times.
|
1662 | for (i = 0; i < channels; i++) { |
| 747 | 1108 | float *p1 = out_samples[i]; | |
| 748 | 1108 | float *p2 = p1 + 256; | |
| 749 | 1108 | float *p3 = p2 + 256; | |
| 750 | 1108 | float *p4 = p3 + 256; | |
| 751 | 1108 | ff_atrac_iqmf(p1, p2, 256, p1, q->units[i].delay_buf1, q->temp_buf); | |
| 752 | 1108 | ff_atrac_iqmf(p4, p3, 256, p3, q->units[i].delay_buf2, q->temp_buf); | |
| 753 | 1108 | ff_atrac_iqmf(p1, p3, 512, p1, q->units[i].delay_buf3, q->temp_buf); | |
| 754 | } | ||
| 755 | |||
| 756 | 554 | return 0; | |
| 757 | } | ||
| 758 | |||
| 759 | ✗ | static int al_decode_frame(AVCodecContext *avctx, const uint8_t *databuf, | |
| 760 | int size, float **out_samples) | ||
| 761 | { | ||
| 762 | ✗ | ATRAC3Context *q = avctx->priv_data; | |
| 763 | ✗ | int channels = avctx->ch_layout.nb_channels; | |
| 764 | int ret, i; | ||
| 765 | |||
| 766 | /* Set the bitstream reader at the start of a channel sound unit. */ | ||
| 767 | ✗ | init_get_bits(&q->gb, databuf, size * 8); | |
| 768 | /* single channels */ | ||
| 769 | /* Decode the channel sound units. */ | ||
| 770 | ✗ | for (i = 0; i < channels; i++) { | |
| 771 | ✗ | ret = decode_channel_sound_unit(q, &q->gb, &q->units[i], | |
| 772 | ✗ | out_samples[i], i, q->coding_mode); | |
| 773 | ✗ | if (ret != 0) | |
| 774 | ✗ | return ret; | |
| 775 | ✗ | while (i < channels && get_bits_left(&q->gb) > 6 && show_bits(&q->gb, 6) != 0x28) { | |
| 776 | ✗ | skip_bits(&q->gb, 1); | |
| 777 | } | ||
| 778 | } | ||
| 779 | |||
| 780 | /* Apply the iQMF synthesis filter. */ | ||
| 781 | ✗ | for (i = 0; i < channels; i++) { | |
| 782 | ✗ | float *p1 = out_samples[i]; | |
| 783 | ✗ | float *p2 = p1 + 256; | |
| 784 | ✗ | float *p3 = p2 + 256; | |
| 785 | ✗ | float *p4 = p3 + 256; | |
| 786 | ✗ | ff_atrac_iqmf(p1, p2, 256, p1, q->units[i].delay_buf1, q->temp_buf); | |
| 787 | ✗ | ff_atrac_iqmf(p4, p3, 256, p3, q->units[i].delay_buf2, q->temp_buf); | |
| 788 | ✗ | ff_atrac_iqmf(p1, p3, 512, p1, q->units[i].delay_buf3, q->temp_buf); | |
| 789 | } | ||
| 790 | |||
| 791 | ✗ | return 0; | |
| 792 | } | ||
| 793 | |||
| 794 | 557 | static int atrac3_decode_frame(AVCodecContext *avctx, AVFrame *frame, | |
| 795 | int *got_frame_ptr, AVPacket *avpkt) | ||
| 796 | { | ||
| 797 | 557 | const uint8_t *buf = avpkt->data; | |
| 798 | 557 | int buf_size = avpkt->size; | |
| 799 | 557 | ATRAC3Context *q = avctx->priv_data; | |
| 800 | int ret; | ||
| 801 | const uint8_t *databuf; | ||
| 802 | |||
| 803 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 554 times.
|
557 | if (buf_size < avctx->block_align) { |
| 804 | 3 | av_log(avctx, AV_LOG_ERROR, | |
| 805 | "Frame too small (%d bytes). Truncated file?\n", buf_size); | ||
| 806 | 3 | return AVERROR_INVALIDDATA; | |
| 807 | } | ||
| 808 | |||
| 809 | /* get output buffer */ | ||
| 810 | 554 | frame->nb_samples = SAMPLES_PER_FRAME; | |
| 811 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 554 times.
|
554 | if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) |
| 812 | ✗ | return ret; | |
| 813 | |||
| 814 | /* Check if we need to descramble and what buffer to pass on. */ | ||
| 815 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 554 times.
|
554 | if (q->scrambled_stream) { |
| 816 | ✗ | decode_bytes(buf, q->decoded_bytes_buffer, avctx->block_align); | |
| 817 | ✗ | databuf = q->decoded_bytes_buffer; | |
| 818 | } else { | ||
| 819 | 554 | databuf = buf; | |
| 820 | } | ||
| 821 | |||
| 822 | 554 | ret = decode_frame(avctx, databuf, (float **)frame->extended_data); | |
| 823 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 554 times.
|
554 | if (ret) { |
| 824 | ✗ | av_log(avctx, AV_LOG_ERROR, "Frame decoding error!\n"); | |
| 825 | ✗ | return ret; | |
| 826 | } | ||
| 827 | |||
| 828 | 554 | *got_frame_ptr = 1; | |
| 829 | |||
| 830 | 554 | return avctx->block_align; | |
| 831 | } | ||
| 832 | |||
| 833 | ✗ | static int atrac3al_decode_frame(AVCodecContext *avctx, AVFrame *frame, | |
| 834 | int *got_frame_ptr, AVPacket *avpkt) | ||
| 835 | { | ||
| 836 | int ret; | ||
| 837 | |||
| 838 | ✗ | frame->nb_samples = SAMPLES_PER_FRAME; | |
| 839 | ✗ | if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) | |
| 840 | ✗ | return ret; | |
| 841 | |||
| 842 | ✗ | ret = al_decode_frame(avctx, avpkt->data, avpkt->size, | |
| 843 | ✗ | (float **)frame->extended_data); | |
| 844 | ✗ | if (ret) { | |
| 845 | ✗ | av_log(avctx, AV_LOG_ERROR, "Frame decoding error!\n"); | |
| 846 | ✗ | return ret; | |
| 847 | } | ||
| 848 | |||
| 849 | ✗ | *got_frame_ptr = 1; | |
| 850 | |||
| 851 | ✗ | return avpkt->size; | |
| 852 | } | ||
| 853 | |||
| 854 | 6 | static av_cold void atrac3_init_static_data(void) | |
| 855 | { | ||
| 856 | 6 | VLCElem *table = atrac3_vlc_table; | |
| 857 | 6 | const uint8_t (*hufftabs)[2] = atrac3_hufftabs; | |
| 858 | int i; | ||
| 859 | |||
| 860 | 6 | init_imdct_window(); | |
| 861 | 6 | ff_atrac_generate_tables(); | |
| 862 | |||
| 863 | /* Initialize the VLC tables. */ | ||
| 864 |
2/2✓ Branch 0 taken 42 times.
✓ Branch 1 taken 6 times.
|
48 | for (i = 0; i < 7; i++) { |
| 865 | 42 | spectral_coeff_tab[i].table = table; | |
| 866 | 42 | spectral_coeff_tab[i].table_allocated = 256; | |
| 867 | 42 | ff_vlc_init_from_lengths(&spectral_coeff_tab[i], ATRAC3_VLC_BITS, huff_tab_sizes[i], | |
| 868 | 42 | &hufftabs[0][1], 2, | |
| 869 | 42 | &hufftabs[0][0], 2, 1, | |
| 870 | -31, VLC_INIT_USE_STATIC, NULL); | ||
| 871 | 42 | hufftabs += huff_tab_sizes[i]; | |
| 872 | 42 | table += 256; | |
| 873 | } | ||
| 874 | 6 | } | |
| 875 | |||
| 876 | 9 | static av_cold int atrac3_decode_init(AVCodecContext *avctx) | |
| 877 | { | ||
| 878 | static AVOnce init_static_once = AV_ONCE_INIT; | ||
| 879 | int i, js_pair, ret; | ||
| 880 | int version, delay, samples_per_frame, frame_factor; | ||
| 881 | 9 | const uint8_t *edata_ptr = avctx->extradata; | |
| 882 | 9 | ATRAC3Context *q = avctx->priv_data; | |
| 883 | AVFloatDSPContext *fdsp; | ||
| 884 | 9 | float scale = 1.0 / 32768; | |
| 885 | 9 | int channels = avctx->ch_layout.nb_channels; | |
| 886 | |||
| 887 |
2/4✓ Branch 0 taken 9 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 9 times.
|
9 | if (channels < MIN_CHANNELS || channels > MAX_CHANNELS) { |
| 888 | ✗ | av_log(avctx, AV_LOG_ERROR, "Channel configuration error!\n"); | |
| 889 | ✗ | return AVERROR(EINVAL); | |
| 890 | } | ||
| 891 | |||
| 892 | /* Take care of the codec-specific extradata. */ | ||
| 893 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 9 times.
|
9 | if (avctx->codec_id == AV_CODEC_ID_ATRAC3AL) { |
| 894 | ✗ | version = 4; | |
| 895 | ✗ | samples_per_frame = SAMPLES_PER_FRAME * channels; | |
| 896 | ✗ | delay = 0x88E; | |
| 897 | ✗ | q->coding_mode = SINGLE; | |
| 898 |
1/2✓ Branch 0 taken 9 times.
✗ Branch 1 not taken.
|
9 | } else if (avctx->extradata_size == 14) { |
| 899 | /* Parse the extradata, WAV format */ | ||
| 900 | 9 | av_log(avctx, AV_LOG_DEBUG, "[0-1] %d\n", | |
| 901 | bytestream_get_le16(&edata_ptr)); // Unknown value always 1 | ||
| 902 | 9 | edata_ptr += 4; // samples per channel | |
| 903 | 9 | q->coding_mode = bytestream_get_le16(&edata_ptr); | |
| 904 | 9 | av_log(avctx, AV_LOG_DEBUG,"[8-9] %d\n", | |
| 905 | bytestream_get_le16(&edata_ptr)); //Dupe of coding mode | ||
| 906 | 9 | frame_factor = bytestream_get_le16(&edata_ptr); // Unknown always 1 | |
| 907 | 9 | av_log(avctx, AV_LOG_DEBUG,"[12-13] %d\n", | |
| 908 | bytestream_get_le16(&edata_ptr)); // Unknown always 0 | ||
| 909 | |||
| 910 | /* setup */ | ||
| 911 | 9 | samples_per_frame = SAMPLES_PER_FRAME * channels; | |
| 912 | 9 | version = 4; | |
| 913 | 9 | delay = 0x88E; | |
| 914 |
2/2✓ Branch 0 taken 2 times.
✓ Branch 1 taken 7 times.
|
9 | q->coding_mode = q->coding_mode ? JOINT_STEREO : SINGLE; |
| 915 | 9 | q->scrambled_stream = 0; | |
| 916 | |||
| 917 |
2/2✓ Branch 0 taken 7 times.
✓ Branch 1 taken 2 times.
|
9 | if (avctx->block_align != 96 * channels * frame_factor && |
| 918 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 2 times.
|
7 | avctx->block_align != 152 * channels * frame_factor && |
| 919 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 5 times.
|
5 | avctx->block_align != 192 * channels * frame_factor) { |
| 920 | ✗ | av_log(avctx, AV_LOG_ERROR, "Unknown frame/channel/frame_factor " | |
| 921 | "configuration %d/%d/%d\n", avctx->block_align, | ||
| 922 | channels, frame_factor); | ||
| 923 | ✗ | return AVERROR_INVALIDDATA; | |
| 924 | } | ||
| 925 | ✗ | } else if (avctx->extradata_size == 12 || avctx->extradata_size == 10) { | |
| 926 | /* Parse the extradata, RM format. */ | ||
| 927 | ✗ | version = bytestream_get_be32(&edata_ptr); | |
| 928 | ✗ | samples_per_frame = bytestream_get_be16(&edata_ptr); | |
| 929 | ✗ | delay = bytestream_get_be16(&edata_ptr); | |
| 930 | ✗ | q->coding_mode = bytestream_get_be16(&edata_ptr); | |
| 931 | ✗ | q->scrambled_stream = 1; | |
| 932 | |||
| 933 | } else { | ||
| 934 | ✗ | av_log(avctx, AV_LOG_ERROR, "Unknown extradata size %d.\n", | |
| 935 | avctx->extradata_size); | ||
| 936 | ✗ | return AVERROR(EINVAL); | |
| 937 | } | ||
| 938 | |||
| 939 | /* Check the extradata */ | ||
| 940 | |||
| 941 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 9 times.
|
9 | if (version != 4) { |
| 942 | ✗ | av_log(avctx, AV_LOG_ERROR, "Version %d != 4.\n", version); | |
| 943 | ✗ | return AVERROR_INVALIDDATA; | |
| 944 | } | ||
| 945 | |||
| 946 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 9 times.
|
9 | if (samples_per_frame != SAMPLES_PER_FRAME * channels) { |
| 947 | ✗ | av_log(avctx, AV_LOG_ERROR, "Unknown amount of samples per frame %d.\n", | |
| 948 | samples_per_frame); | ||
| 949 | ✗ | return AVERROR_INVALIDDATA; | |
| 950 | } | ||
| 951 | |||
| 952 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 9 times.
|
9 | if (delay != 0x88E) { |
| 953 | ✗ | av_log(avctx, AV_LOG_ERROR, "Unknown amount of delay %x != 0x88E.\n", | |
| 954 | delay); | ||
| 955 | ✗ | return AVERROR_INVALIDDATA; | |
| 956 | } | ||
| 957 | |||
| 958 |
2/2✓ Branch 0 taken 7 times.
✓ Branch 1 taken 2 times.
|
9 | if (q->coding_mode == SINGLE) |
| 959 | 7 | av_log(avctx, AV_LOG_DEBUG, "Single channels detected.\n"); | |
| 960 |
1/2✓ Branch 0 taken 2 times.
✗ Branch 1 not taken.
|
2 | else if (q->coding_mode == JOINT_STEREO) { |
| 961 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 2 times.
|
2 | if (channels % 2 == 1) { /* Joint stereo channels must be even */ |
| 962 | ✗ | av_log(avctx, AV_LOG_ERROR, "Invalid joint stereo channel configuration.\n"); | |
| 963 | ✗ | return AVERROR_INVALIDDATA; | |
| 964 | } | ||
| 965 | 2 | av_log(avctx, AV_LOG_DEBUG, "Joint stereo detected.\n"); | |
| 966 | } else { | ||
| 967 | ✗ | av_log(avctx, AV_LOG_ERROR, "Unknown channel coding mode %x!\n", | |
| 968 | q->coding_mode); | ||
| 969 | ✗ | return AVERROR_INVALIDDATA; | |
| 970 | } | ||
| 971 | |||
| 972 |
2/4✓ Branch 0 taken 9 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 9 times.
|
9 | if (avctx->block_align > 4096 || avctx->block_align <= 0) |
| 973 | ✗ | return AVERROR(EINVAL); | |
| 974 | |||
| 975 | 9 | q->decoded_bytes_buffer = av_mallocz(FFALIGN(avctx->block_align, 4) + | |
| 976 | AV_INPUT_BUFFER_PADDING_SIZE); | ||
| 977 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 9 times.
|
9 | if (!q->decoded_bytes_buffer) |
| 978 | ✗ | return AVERROR(ENOMEM); | |
| 979 | |||
| 980 | 9 | avctx->sample_fmt = AV_SAMPLE_FMT_FLTP; | |
| 981 | |||
| 982 | /* initialize the MDCT transform */ | ||
| 983 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 9 times.
|
9 | if ((ret = av_tx_init(&q->mdct_ctx, &q->mdct_fn, AV_TX_FLOAT_MDCT, 1, 256, |
| 984 | &scale, AV_TX_FULL_IMDCT)) < 0) { | ||
| 985 | ✗ | av_log(avctx, AV_LOG_ERROR, "Error initializing MDCT\n"); | |
| 986 | ✗ | return ret; | |
| 987 | } | ||
| 988 | |||
| 989 | /* init the joint-stereo decoding data */ | ||
| 990 |
2/2✓ Branch 0 taken 36 times.
✓ Branch 1 taken 9 times.
|
45 | for (js_pair = 0; js_pair < MAX_JS_PAIRS; js_pair++) { |
| 991 | 36 | q->weighting_delay[js_pair][0] = 0; | |
| 992 | 36 | q->weighting_delay[js_pair][1] = 7; | |
| 993 | 36 | q->weighting_delay[js_pair][2] = 0; | |
| 994 | 36 | q->weighting_delay[js_pair][3] = 7; | |
| 995 | 36 | q->weighting_delay[js_pair][4] = 0; | |
| 996 | 36 | q->weighting_delay[js_pair][5] = 7; | |
| 997 | |||
| 998 |
2/2✓ Branch 0 taken 144 times.
✓ Branch 1 taken 36 times.
|
180 | for (i = 0; i < 4; i++) { |
| 999 | 144 | q->matrix_coeff_index_prev[js_pair][i] = 3; | |
| 1000 | 144 | q->matrix_coeff_index_now[js_pair][i] = 3; | |
| 1001 | 144 | q->matrix_coeff_index_next[js_pair][i] = 3; | |
| 1002 | } | ||
| 1003 | } | ||
| 1004 | |||
| 1005 | 9 | ff_atrac_init_gain_compensation(&q->gainc_ctx, 4, 3); | |
| 1006 | 9 | fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT); | |
| 1007 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 9 times.
|
9 | if (!fdsp) |
| 1008 | ✗ | return AVERROR(ENOMEM); | |
| 1009 | 9 | q->vector_fmul = fdsp->vector_fmul; | |
| 1010 | 9 | av_free(fdsp); | |
| 1011 | |||
| 1012 | 9 | q->units = av_calloc(channels, sizeof(*q->units)); | |
| 1013 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 9 times.
|
9 | if (!q->units) |
| 1014 | ✗ | return AVERROR(ENOMEM); | |
| 1015 | |||
| 1016 | 9 | ff_thread_once(&init_static_once, atrac3_init_static_data); | |
| 1017 | |||
| 1018 | 9 | return 0; | |
| 1019 | } | ||
| 1020 | |||
| 1021 | const FFCodec ff_atrac3_decoder = { | ||
| 1022 | .p.name = "atrac3", | ||
| 1023 | CODEC_LONG_NAME("ATRAC3 (Adaptive TRansform Acoustic Coding 3)"), | ||
| 1024 | .p.type = AVMEDIA_TYPE_AUDIO, | ||
| 1025 | .p.id = AV_CODEC_ID_ATRAC3, | ||
| 1026 | .priv_data_size = sizeof(ATRAC3Context), | ||
| 1027 | .init = atrac3_decode_init, | ||
| 1028 | .close = atrac3_decode_close, | ||
| 1029 | FF_CODEC_DECODE_CB(atrac3_decode_frame), | ||
| 1030 | .p.capabilities = AV_CODEC_CAP_DR1, | ||
| 1031 | CODEC_SAMPLEFMTS(AV_SAMPLE_FMT_FLTP), | ||
| 1032 | .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, | ||
| 1033 | }; | ||
| 1034 | |||
| 1035 | const FFCodec ff_atrac3al_decoder = { | ||
| 1036 | .p.name = "atrac3al", | ||
| 1037 | CODEC_LONG_NAME("ATRAC3 AL (Adaptive TRansform Acoustic Coding 3 Advanced Lossless)"), | ||
| 1038 | .p.type = AVMEDIA_TYPE_AUDIO, | ||
| 1039 | .p.id = AV_CODEC_ID_ATRAC3AL, | ||
| 1040 | .priv_data_size = sizeof(ATRAC3Context), | ||
| 1041 | .init = atrac3_decode_init, | ||
| 1042 | .close = atrac3_decode_close, | ||
| 1043 | FF_CODEC_DECODE_CB(atrac3al_decode_frame), | ||
| 1044 | .p.capabilities = AV_CODEC_CAP_DR1, | ||
| 1045 | CODEC_SAMPLEFMTS(AV_SAMPLE_FMT_FLTP), | ||
| 1046 | .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, | ||
| 1047 | }; | ||
| 1048 |