| 1 | /* |
| 2 | * ATRAC3+ compatible decoder |
| 3 | * |
| 4 | * Copyright (c) 2010-2013 Maxim Poliakovski |
| 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 | * Sony ATRAC3+ compatible decoder. |
| 26 | * |
| 27 | * Container formats used to store its data: |
| 28 | * RIFF WAV (.at3) and Sony OpenMG (.oma, .aa3). |
| 29 | * |
| 30 | * Technical description of this codec can be found here: |
| 31 | * http://wiki.multimedia.cx/index.php?title=ATRAC3plus |
| 32 | * |
| 33 | * Kudos to Benjamin Larsson and Michael Karcher |
| 34 | * for their precious technical help! |
| 35 | */ |
| 36 | |
| 37 | #include <stdint.h> |
| 38 | #include <string.h> |
| 39 | |
| 40 | #include "libavutil/channel_layout.h" |
| 41 | #include "libavutil/float_dsp.h" |
| 42 | #include "avcodec.h" |
| 43 | #include "get_bits.h" |
| 44 | #include "internal.h" |
| 45 | #include "atrac.h" |
| 46 | #include "atrac3plus.h" |
| 47 | |
| 48 | typedef struct ATRAC3PContext { |
| 49 | GetBitContext gb; |
| 50 | AVFloatDSPContext *fdsp; |
| 51 | |
| 52 | DECLARE_ALIGNED(32, float, samples)[2][ATRAC3P_FRAME_SAMPLES]; ///< quantized MDCT spectrum |
| 53 | DECLARE_ALIGNED(32, float, mdct_buf)[2][ATRAC3P_FRAME_SAMPLES]; ///< output of the IMDCT |
| 54 | DECLARE_ALIGNED(32, float, time_buf)[2][ATRAC3P_FRAME_SAMPLES]; ///< output of the gain compensation |
| 55 | DECLARE_ALIGNED(32, float, outp_buf)[2][ATRAC3P_FRAME_SAMPLES]; |
| 56 | |
| 57 | AtracGCContext gainc_ctx; ///< gain compensation context |
| 58 | FFTContext mdct_ctx; |
| 59 | FFTContext ipqf_dct_ctx; ///< IDCT context used by IPQF |
| 60 | |
| 61 | Atrac3pChanUnitCtx *ch_units; ///< global channel units |
| 62 | |
| 63 | int num_channel_blocks; ///< number of channel blocks |
| 64 | uint8_t channel_blocks[5]; ///< channel configuration descriptor |
| 65 | uint64_t my_channel_layout; ///< current channel layout |
| 66 | } ATRAC3PContext; |
| 67 | |
| 68 | static av_cold int atrac3p_decode_close(AVCodecContext *avctx) |
| 69 | { |
| 70 | ATRAC3PContext *ctx = avctx->priv_data; |
| 71 | |
| 72 | av_freep(&ctx->ch_units); |
| 73 | av_freep(&ctx->fdsp); |
| 74 | |
| 75 | return 0; |
| 76 | } |
| 77 | |
| 78 | static av_cold int set_channel_params(ATRAC3PContext *ctx, |
| 79 | AVCodecContext *avctx) |
| 80 | { |
| 81 | memset(ctx->channel_blocks, 0, sizeof(ctx->channel_blocks)); |
| 82 | |
| 83 | switch (avctx->channels) { |
| 84 | case 1: |
| 85 | if (avctx->channel_layout != AV_CH_FRONT_LEFT) |
| 86 | avctx->channel_layout = AV_CH_LAYOUT_MONO; |
| 87 | |
| 88 | ctx->num_channel_blocks = 1; |
| 89 | ctx->channel_blocks[0] = CH_UNIT_MONO; |
| 90 | break; |
| 91 | case 2: |
| 92 | avctx->channel_layout = AV_CH_LAYOUT_STEREO; |
| 93 | ctx->num_channel_blocks = 1; |
| 94 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
| 95 | break; |
| 96 | case 3: |
| 97 | avctx->channel_layout = AV_CH_LAYOUT_SURROUND; |
| 98 | ctx->num_channel_blocks = 2; |
| 99 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
| 100 | ctx->channel_blocks[1] = CH_UNIT_MONO; |
| 101 | break; |
| 102 | case 4: |
| 103 | avctx->channel_layout = AV_CH_LAYOUT_4POINT0; |
| 104 | ctx->num_channel_blocks = 3; |
| 105 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
| 106 | ctx->channel_blocks[1] = CH_UNIT_MONO; |
| 107 | ctx->channel_blocks[2] = CH_UNIT_MONO; |
| 108 | break; |
| 109 | case 6: |
| 110 | avctx->channel_layout = AV_CH_LAYOUT_5POINT1_BACK; |
| 111 | ctx->num_channel_blocks = 4; |
| 112 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
| 113 | ctx->channel_blocks[1] = CH_UNIT_MONO; |
| 114 | ctx->channel_blocks[2] = CH_UNIT_STEREO; |
| 115 | ctx->channel_blocks[3] = CH_UNIT_MONO; |
| 116 | break; |
| 117 | case 7: |
| 118 | avctx->channel_layout = AV_CH_LAYOUT_6POINT1_BACK; |
| 119 | ctx->num_channel_blocks = 5; |
| 120 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
| 121 | ctx->channel_blocks[1] = CH_UNIT_MONO; |
| 122 | ctx->channel_blocks[2] = CH_UNIT_STEREO; |
| 123 | ctx->channel_blocks[3] = CH_UNIT_MONO; |
| 124 | ctx->channel_blocks[4] = CH_UNIT_MONO; |
| 125 | break; |
| 126 | case 8: |
| 127 | avctx->channel_layout = AV_CH_LAYOUT_7POINT1; |
| 128 | ctx->num_channel_blocks = 5; |
| 129 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
| 130 | ctx->channel_blocks[1] = CH_UNIT_MONO; |
| 131 | ctx->channel_blocks[2] = CH_UNIT_STEREO; |
| 132 | ctx->channel_blocks[3] = CH_UNIT_STEREO; |
| 133 | ctx->channel_blocks[4] = CH_UNIT_MONO; |
| 134 | break; |
| 135 | default: |
| 136 | av_log(avctx, AV_LOG_ERROR, |
| 137 | "Unsupported channel count: %d!\n", avctx->channels); |
| 138 | return AVERROR_INVALIDDATA; |
| 139 | } |
| 140 | |
| 141 | return 0; |
| 142 | } |
| 143 | |
| 144 | static av_cold int atrac3p_decode_init(AVCodecContext *avctx) |
| 145 | { |
| 146 | ATRAC3PContext *ctx = avctx->priv_data; |
| 147 | int i, ch, ret; |
| 148 | |
| 149 | if (!avctx->block_align) { |
| 150 | av_log(avctx, AV_LOG_ERROR, "block_align is not set\n"); |
| 151 | return AVERROR(EINVAL); |
| 152 | } |
| 153 | |
| 154 | ff_atrac3p_init_vlcs(); |
| 155 | |
| 156 | /* initialize IPQF */ |
| 157 | ff_mdct_init(&ctx->ipqf_dct_ctx, 5, 1, 32.0 / 32768.0); |
| 158 | |
| 159 | ff_atrac3p_init_imdct(avctx, &ctx->mdct_ctx); |
| 160 | |
| 161 | ff_atrac_init_gain_compensation(&ctx->gainc_ctx, 6, 2); |
| 162 | |
| 163 | ff_atrac3p_init_wave_synth(); |
| 164 | |
| 165 | if ((ret = set_channel_params(ctx, avctx)) < 0) |
| 166 | return ret; |
| 167 | |
| 168 | ctx->my_channel_layout = avctx->channel_layout; |
| 169 | |
| 170 | ctx->ch_units = av_mallocz_array(ctx->num_channel_blocks, sizeof(*ctx->ch_units)); |
| 171 | ctx->fdsp = avpriv_float_dsp_alloc(avctx->flags & CODEC_FLAG_BITEXACT); |
| 172 | |
| 173 | if (!ctx->ch_units || !ctx->fdsp) { |
| 174 | atrac3p_decode_close(avctx); |
| 175 | return AVERROR(ENOMEM); |
| 176 | } |
| 177 | |
| 178 | for (i = 0; i < ctx->num_channel_blocks; i++) { |
| 179 | for (ch = 0; ch < 2; ch++) { |
| 180 | ctx->ch_units[i].channels[ch].ch_num = ch; |
| 181 | ctx->ch_units[i].channels[ch].wnd_shape = &ctx->ch_units[i].channels[ch].wnd_shape_hist[0][0]; |
| 182 | ctx->ch_units[i].channels[ch].wnd_shape_prev = &ctx->ch_units[i].channels[ch].wnd_shape_hist[1][0]; |
| 183 | ctx->ch_units[i].channels[ch].gain_data = &ctx->ch_units[i].channels[ch].gain_data_hist[0][0]; |
| 184 | ctx->ch_units[i].channels[ch].gain_data_prev = &ctx->ch_units[i].channels[ch].gain_data_hist[1][0]; |
| 185 | ctx->ch_units[i].channels[ch].tones_info = &ctx->ch_units[i].channels[ch].tones_info_hist[0][0]; |
| 186 | ctx->ch_units[i].channels[ch].tones_info_prev = &ctx->ch_units[i].channels[ch].tones_info_hist[1][0]; |
| 187 | } |
| 188 | |
| 189 | ctx->ch_units[i].waves_info = &ctx->ch_units[i].wave_synth_hist[0]; |
| 190 | ctx->ch_units[i].waves_info_prev = &ctx->ch_units[i].wave_synth_hist[1]; |
| 191 | } |
| 192 | |
| 193 | avctx->sample_fmt = AV_SAMPLE_FMT_FLTP; |
| 194 | |
| 195 | return 0; |
| 196 | } |
| 197 | |
| 198 | static void decode_residual_spectrum(Atrac3pChanUnitCtx *ctx, |
| 199 | float out[2][ATRAC3P_FRAME_SAMPLES], |
| 200 | int num_channels, |
| 201 | AVCodecContext *avctx) |
| 202 | { |
| 203 | int i, sb, ch, qu, nspeclines, RNG_index; |
| 204 | float *dst, q; |
| 205 | int16_t *src; |
| 206 | /* calculate RNG table index for each subband */ |
| 207 | int sb_RNG_index[ATRAC3P_SUBBANDS] = { 0 }; |
| 208 | |
| 209 | if (ctx->mute_flag) { |
| 210 | for (ch = 0; ch < num_channels; ch++) |
| 211 | memset(out[ch], 0, ATRAC3P_FRAME_SAMPLES * sizeof(*out[ch])); |
| 212 | return; |
| 213 | } |
| 214 | |
| 215 | for (qu = 0, RNG_index = 0; qu < ctx->used_quant_units; qu++) |
| 216 | RNG_index += ctx->channels[0].qu_sf_idx[qu] + |
| 217 | ctx->channels[1].qu_sf_idx[qu]; |
| 218 | |
| 219 | for (sb = 0; sb < ctx->num_coded_subbands; sb++, RNG_index += 128) |
| 220 | sb_RNG_index[sb] = RNG_index & 0x3FC; |
| 221 | |
| 222 | /* inverse quant and power compensation */ |
| 223 | for (ch = 0; ch < num_channels; ch++) { |
| 224 | /* clear channel's residual spectrum */ |
| 225 | memset(out[ch], 0, ATRAC3P_FRAME_SAMPLES * sizeof(*out[ch])); |
| 226 | |
| 227 | for (qu = 0; qu < ctx->used_quant_units; qu++) { |
| 228 | src = &ctx->channels[ch].spectrum[ff_atrac3p_qu_to_spec_pos[qu]]; |
| 229 | dst = &out[ch][ff_atrac3p_qu_to_spec_pos[qu]]; |
| 230 | nspeclines = ff_atrac3p_qu_to_spec_pos[qu + 1] - |
| 231 | ff_atrac3p_qu_to_spec_pos[qu]; |
| 232 | |
| 233 | if (ctx->channels[ch].qu_wordlen[qu] > 0) { |
| 234 | q = ff_atrac3p_sf_tab[ctx->channels[ch].qu_sf_idx[qu]] * |
| 235 | ff_atrac3p_mant_tab[ctx->channels[ch].qu_wordlen[qu]]; |
| 236 | for (i = 0; i < nspeclines; i++) |
| 237 | dst[i] = src[i] * q; |
| 238 | } |
| 239 | } |
| 240 | |
| 241 | for (sb = 0; sb < ctx->num_coded_subbands; sb++) |
| 242 | ff_atrac3p_power_compensation(ctx, ch, &out[ch][0], |
| 243 | sb_RNG_index[sb], sb); |
| 244 | } |
| 245 | |
| 246 | if (ctx->unit_type == CH_UNIT_STEREO) { |
| 247 | for (sb = 0; sb < ctx->num_coded_subbands; sb++) { |
| 248 | if (ctx->swap_channels[sb]) { |
| 249 | for (i = 0; i < ATRAC3P_SUBBAND_SAMPLES; i++) |
| 250 | FFSWAP(float, out[0][sb * ATRAC3P_SUBBAND_SAMPLES + i], |
| 251 | out[1][sb * ATRAC3P_SUBBAND_SAMPLES + i]); |
| 252 | } |
| 253 | |
| 254 | /* flip coefficients' sign if requested */ |
| 255 | if (ctx->negate_coeffs[sb]) |
| 256 | for (i = 0; i < ATRAC3P_SUBBAND_SAMPLES; i++) |
| 257 | out[1][sb * ATRAC3P_SUBBAND_SAMPLES + i] = -(out[1][sb * ATRAC3P_SUBBAND_SAMPLES + i]); |
| 258 | } |
| 259 | } |
| 260 | } |
| 261 | |
| 262 | static void reconstruct_frame(ATRAC3PContext *ctx, Atrac3pChanUnitCtx *ch_unit, |
| 263 | int num_channels, AVCodecContext *avctx) |
| 264 | { |
| 265 | int ch, sb; |
| 266 | |
| 267 | for (ch = 0; ch < num_channels; ch++) { |
| 268 | for (sb = 0; sb < ch_unit->num_subbands; sb++) { |
| 269 | /* inverse transform and windowing */ |
| 270 | ff_atrac3p_imdct(ctx->fdsp, &ctx->mdct_ctx, |
| 271 | &ctx->samples[ch][sb * ATRAC3P_SUBBAND_SAMPLES], |
| 272 | &ctx->mdct_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES], |
| 273 | (ch_unit->channels[ch].wnd_shape_prev[sb] << 1) + |
| 274 | ch_unit->channels[ch].wnd_shape[sb], sb); |
| 275 | |
| 276 | /* gain compensation and overlapping */ |
| 277 | ff_atrac_gain_compensation(&ctx->gainc_ctx, |
| 278 | &ctx->mdct_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES], |
| 279 | &ch_unit->prev_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES], |
| 280 | &ch_unit->channels[ch].gain_data_prev[sb], |
| 281 | &ch_unit->channels[ch].gain_data[sb], |
| 282 | ATRAC3P_SUBBAND_SAMPLES, |
| 283 | &ctx->time_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES]); |
| 284 | } |
| 285 | |
| 286 | /* zero unused subbands in both output and overlapping buffers */ |
| 287 | memset(&ch_unit->prev_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES], |
| 288 | 0, |
| 289 | (ATRAC3P_SUBBANDS - ch_unit->num_subbands) * |
| 290 | ATRAC3P_SUBBAND_SAMPLES * |
| 291 | sizeof(ch_unit->prev_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES])); |
| 292 | memset(&ctx->time_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES], |
| 293 | 0, |
| 294 | (ATRAC3P_SUBBANDS - ch_unit->num_subbands) * |
| 295 | ATRAC3P_SUBBAND_SAMPLES * |
| 296 | sizeof(ctx->time_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES])); |
| 297 | |
| 298 | /* resynthesize and add tonal signal */ |
| 299 | if (ch_unit->waves_info->tones_present || |
| 300 | ch_unit->waves_info_prev->tones_present) { |
| 301 | for (sb = 0; sb < ch_unit->num_subbands; sb++) |
| 302 | if (ch_unit->channels[ch].tones_info[sb].num_wavs || |
| 303 | ch_unit->channels[ch].tones_info_prev[sb].num_wavs) { |
| 304 | ff_atrac3p_generate_tones(ch_unit, ctx->fdsp, ch, sb, |
| 305 | &ctx->time_buf[ch][sb * 128]); |
| 306 | } |
| 307 | } |
| 308 | |
| 309 | /* subband synthesis and acoustic signal output */ |
| 310 | ff_atrac3p_ipqf(&ctx->ipqf_dct_ctx, &ch_unit->ipqf_ctx[ch], |
| 311 | &ctx->time_buf[ch][0], &ctx->outp_buf[ch][0]); |
| 312 | } |
| 313 | |
| 314 | /* swap window shape and gain control buffers. */ |
| 315 | for (ch = 0; ch < num_channels; ch++) { |
| 316 | FFSWAP(uint8_t *, ch_unit->channels[ch].wnd_shape, |
| 317 | ch_unit->channels[ch].wnd_shape_prev); |
| 318 | FFSWAP(AtracGainInfo *, ch_unit->channels[ch].gain_data, |
| 319 | ch_unit->channels[ch].gain_data_prev); |
| 320 | FFSWAP(Atrac3pWavesData *, ch_unit->channels[ch].tones_info, |
| 321 | ch_unit->channels[ch].tones_info_prev); |
| 322 | } |
| 323 | |
| 324 | FFSWAP(Atrac3pWaveSynthParams *, ch_unit->waves_info, ch_unit->waves_info_prev); |
| 325 | } |
| 326 | |
| 327 | static int atrac3p_decode_frame(AVCodecContext *avctx, void *data, |
| 328 | int *got_frame_ptr, AVPacket *avpkt) |
| 329 | { |
| 330 | ATRAC3PContext *ctx = avctx->priv_data; |
| 331 | AVFrame *frame = data; |
| 332 | int i, ret, ch_unit_id, ch_block = 0, out_ch_index = 0, channels_to_process; |
| 333 | float **samples_p = (float **)frame->extended_data; |
| 334 | |
| 335 | frame->nb_samples = ATRAC3P_FRAME_SAMPLES; |
| 336 | if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) { |
| 337 | av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n"); |
| 338 | return ret; |
| 339 | } |
| 340 | |
| 341 | if ((ret = init_get_bits8(&ctx->gb, avpkt->data, avpkt->size)) < 0) |
| 342 | return ret; |
| 343 | |
| 344 | if (get_bits1(&ctx->gb)) { |
| 345 | av_log(avctx, AV_LOG_ERROR, "Invalid start bit!\n"); |
| 346 | return AVERROR_INVALIDDATA; |
| 347 | } |
| 348 | |
| 349 | while (get_bits_left(&ctx->gb) >= 2 && |
| 350 | (ch_unit_id = get_bits(&ctx->gb, 2)) != CH_UNIT_TERMINATOR) { |
| 351 | if (ch_unit_id == CH_UNIT_EXTENSION) { |
| 352 | avpriv_report_missing_feature(avctx, "Channel unit extension"); |
| 353 | return AVERROR_PATCHWELCOME; |
| 354 | } |
| 355 | if (ch_block >= ctx->num_channel_blocks || |
| 356 | ctx->channel_blocks[ch_block] != ch_unit_id) { |
| 357 | av_log(avctx, AV_LOG_ERROR, |
| 358 | "Frame data doesn't match channel configuration!\n"); |
| 359 | return AVERROR_INVALIDDATA; |
| 360 | } |
| 361 | |
| 362 | ctx->ch_units[ch_block].unit_type = ch_unit_id; |
| 363 | channels_to_process = ch_unit_id + 1; |
| 364 | |
| 365 | if ((ret = ff_atrac3p_decode_channel_unit(&ctx->gb, |
| 366 | &ctx->ch_units[ch_block], |
| 367 | channels_to_process, |
| 368 | avctx)) < 0) |
| 369 | return ret; |
| 370 | |
| 371 | decode_residual_spectrum(&ctx->ch_units[ch_block], ctx->samples, |
| 372 | channels_to_process, avctx); |
| 373 | reconstruct_frame(ctx, &ctx->ch_units[ch_block], |
| 374 | channels_to_process, avctx); |
| 375 | |
| 376 | for (i = 0; i < channels_to_process; i++) |
| 377 | memcpy(samples_p[out_ch_index + i], ctx->outp_buf[i], |
| 378 | ATRAC3P_FRAME_SAMPLES * sizeof(**samples_p)); |
| 379 | |
| 380 | ch_block++; |
| 381 | out_ch_index += channels_to_process; |
| 382 | } |
| 383 | |
| 384 | *got_frame_ptr = 1; |
| 385 | |
| 386 | return avctx->block_align; |
| 387 | } |
| 388 | |
| 389 | AVCodec ff_atrac3p_decoder = { |
| 390 | .name = "atrac3plus", |
| 391 | .long_name = NULL_IF_CONFIG_SMALL("ATRAC3+ (Adaptive TRansform Acoustic Coding 3+)"), |
| 392 | .type = AVMEDIA_TYPE_AUDIO, |
| 393 | .id = AV_CODEC_ID_ATRAC3P, |
| 394 | .priv_data_size = sizeof(ATRAC3PContext), |
| 395 | .init = atrac3p_decode_init, |
| 396 | .close = atrac3p_decode_close, |
| 397 | .decode = atrac3p_decode_frame, |
| 398 | }; |