blob: 7a2ab3ef95d6309363b6db58141bc767cd427252
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 | ff_mdct_end(&ctx->mdct_ctx); |
76 | ff_mdct_end(&ctx->ipqf_dct_ctx); |
77 | |
78 | return 0; |
79 | } |
80 | |
81 | static av_cold int set_channel_params(ATRAC3PContext *ctx, |
82 | AVCodecContext *avctx) |
83 | { |
84 | memset(ctx->channel_blocks, 0, sizeof(ctx->channel_blocks)); |
85 | |
86 | switch (avctx->channels) { |
87 | case 1: |
88 | if (avctx->channel_layout != AV_CH_FRONT_LEFT) |
89 | avctx->channel_layout = AV_CH_LAYOUT_MONO; |
90 | |
91 | ctx->num_channel_blocks = 1; |
92 | ctx->channel_blocks[0] = CH_UNIT_MONO; |
93 | break; |
94 | case 2: |
95 | avctx->channel_layout = AV_CH_LAYOUT_STEREO; |
96 | ctx->num_channel_blocks = 1; |
97 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
98 | break; |
99 | case 3: |
100 | avctx->channel_layout = AV_CH_LAYOUT_SURROUND; |
101 | ctx->num_channel_blocks = 2; |
102 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
103 | ctx->channel_blocks[1] = CH_UNIT_MONO; |
104 | break; |
105 | case 4: |
106 | avctx->channel_layout = AV_CH_LAYOUT_4POINT0; |
107 | ctx->num_channel_blocks = 3; |
108 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
109 | ctx->channel_blocks[1] = CH_UNIT_MONO; |
110 | ctx->channel_blocks[2] = CH_UNIT_MONO; |
111 | break; |
112 | case 6: |
113 | avctx->channel_layout = AV_CH_LAYOUT_5POINT1_BACK; |
114 | ctx->num_channel_blocks = 4; |
115 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
116 | ctx->channel_blocks[1] = CH_UNIT_MONO; |
117 | ctx->channel_blocks[2] = CH_UNIT_STEREO; |
118 | ctx->channel_blocks[3] = CH_UNIT_MONO; |
119 | break; |
120 | case 7: |
121 | avctx->channel_layout = AV_CH_LAYOUT_6POINT1_BACK; |
122 | ctx->num_channel_blocks = 5; |
123 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
124 | ctx->channel_blocks[1] = CH_UNIT_MONO; |
125 | ctx->channel_blocks[2] = CH_UNIT_STEREO; |
126 | ctx->channel_blocks[3] = CH_UNIT_MONO; |
127 | ctx->channel_blocks[4] = CH_UNIT_MONO; |
128 | break; |
129 | case 8: |
130 | avctx->channel_layout = AV_CH_LAYOUT_7POINT1; |
131 | ctx->num_channel_blocks = 5; |
132 | ctx->channel_blocks[0] = CH_UNIT_STEREO; |
133 | ctx->channel_blocks[1] = CH_UNIT_MONO; |
134 | ctx->channel_blocks[2] = CH_UNIT_STEREO; |
135 | ctx->channel_blocks[3] = CH_UNIT_STEREO; |
136 | ctx->channel_blocks[4] = CH_UNIT_MONO; |
137 | break; |
138 | default: |
139 | av_log(avctx, AV_LOG_ERROR, |
140 | "Unsupported channel count: %d!\n", avctx->channels); |
141 | return AVERROR_INVALIDDATA; |
142 | } |
143 | |
144 | return 0; |
145 | } |
146 | |
147 | static av_cold int atrac3p_decode_init(AVCodecContext *avctx) |
148 | { |
149 | ATRAC3PContext *ctx = avctx->priv_data; |
150 | int i, ch, ret; |
151 | |
152 | if (!avctx->block_align) { |
153 | av_log(avctx, AV_LOG_ERROR, "block_align is not set\n"); |
154 | return AVERROR(EINVAL); |
155 | } |
156 | |
157 | ff_atrac3p_init_vlcs(); |
158 | |
159 | /* initialize IPQF */ |
160 | ff_mdct_init(&ctx->ipqf_dct_ctx, 5, 1, 32.0 / 32768.0); |
161 | |
162 | ff_atrac3p_init_imdct(avctx, &ctx->mdct_ctx); |
163 | |
164 | ff_atrac_init_gain_compensation(&ctx->gainc_ctx, 6, 2); |
165 | |
166 | ff_atrac3p_init_wave_synth(); |
167 | |
168 | if ((ret = set_channel_params(ctx, avctx)) < 0) |
169 | return ret; |
170 | |
171 | ctx->my_channel_layout = avctx->channel_layout; |
172 | |
173 | ctx->ch_units = av_mallocz_array(ctx->num_channel_blocks, sizeof(*ctx->ch_units)); |
174 | ctx->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT); |
175 | |
176 | if (!ctx->ch_units || !ctx->fdsp) { |
177 | atrac3p_decode_close(avctx); |
178 | return AVERROR(ENOMEM); |
179 | } |
180 | |
181 | for (i = 0; i < ctx->num_channel_blocks; i++) { |
182 | for (ch = 0; ch < 2; ch++) { |
183 | ctx->ch_units[i].channels[ch].ch_num = ch; |
184 | ctx->ch_units[i].channels[ch].wnd_shape = &ctx->ch_units[i].channels[ch].wnd_shape_hist[0][0]; |
185 | ctx->ch_units[i].channels[ch].wnd_shape_prev = &ctx->ch_units[i].channels[ch].wnd_shape_hist[1][0]; |
186 | ctx->ch_units[i].channels[ch].gain_data = &ctx->ch_units[i].channels[ch].gain_data_hist[0][0]; |
187 | ctx->ch_units[i].channels[ch].gain_data_prev = &ctx->ch_units[i].channels[ch].gain_data_hist[1][0]; |
188 | ctx->ch_units[i].channels[ch].tones_info = &ctx->ch_units[i].channels[ch].tones_info_hist[0][0]; |
189 | ctx->ch_units[i].channels[ch].tones_info_prev = &ctx->ch_units[i].channels[ch].tones_info_hist[1][0]; |
190 | } |
191 | |
192 | ctx->ch_units[i].waves_info = &ctx->ch_units[i].wave_synth_hist[0]; |
193 | ctx->ch_units[i].waves_info_prev = &ctx->ch_units[i].wave_synth_hist[1]; |
194 | } |
195 | |
196 | avctx->sample_fmt = AV_SAMPLE_FMT_FLTP; |
197 | |
198 | return 0; |
199 | } |
200 | |
201 | static void decode_residual_spectrum(Atrac3pChanUnitCtx *ctx, |
202 | float out[2][ATRAC3P_FRAME_SAMPLES], |
203 | int num_channels, |
204 | AVCodecContext *avctx) |
205 | { |
206 | int i, sb, ch, qu, nspeclines, RNG_index; |
207 | float *dst, q; |
208 | int16_t *src; |
209 | /* calculate RNG table index for each subband */ |
210 | int sb_RNG_index[ATRAC3P_SUBBANDS] = { 0 }; |
211 | |
212 | if (ctx->mute_flag) { |
213 | for (ch = 0; ch < num_channels; ch++) |
214 | memset(out[ch], 0, ATRAC3P_FRAME_SAMPLES * sizeof(*out[ch])); |
215 | return; |
216 | } |
217 | |
218 | for (qu = 0, RNG_index = 0; qu < ctx->used_quant_units; qu++) |
219 | RNG_index += ctx->channels[0].qu_sf_idx[qu] + |
220 | ctx->channels[1].qu_sf_idx[qu]; |
221 | |
222 | for (sb = 0; sb < ctx->num_coded_subbands; sb++, RNG_index += 128) |
223 | sb_RNG_index[sb] = RNG_index & 0x3FC; |
224 | |
225 | /* inverse quant and power compensation */ |
226 | for (ch = 0; ch < num_channels; ch++) { |
227 | /* clear channel's residual spectrum */ |
228 | memset(out[ch], 0, ATRAC3P_FRAME_SAMPLES * sizeof(*out[ch])); |
229 | |
230 | for (qu = 0; qu < ctx->used_quant_units; qu++) { |
231 | src = &ctx->channels[ch].spectrum[ff_atrac3p_qu_to_spec_pos[qu]]; |
232 | dst = &out[ch][ff_atrac3p_qu_to_spec_pos[qu]]; |
233 | nspeclines = ff_atrac3p_qu_to_spec_pos[qu + 1] - |
234 | ff_atrac3p_qu_to_spec_pos[qu]; |
235 | |
236 | if (ctx->channels[ch].qu_wordlen[qu] > 0) { |
237 | q = ff_atrac3p_sf_tab[ctx->channels[ch].qu_sf_idx[qu]] * |
238 | ff_atrac3p_mant_tab[ctx->channels[ch].qu_wordlen[qu]]; |
239 | for (i = 0; i < nspeclines; i++) |
240 | dst[i] = src[i] * q; |
241 | } |
242 | } |
243 | |
244 | for (sb = 0; sb < ctx->num_coded_subbands; sb++) |
245 | ff_atrac3p_power_compensation(ctx, ch, &out[ch][0], |
246 | sb_RNG_index[sb], sb); |
247 | } |
248 | |
249 | if (ctx->unit_type == CH_UNIT_STEREO) { |
250 | for (sb = 0; sb < ctx->num_coded_subbands; sb++) { |
251 | if (ctx->swap_channels[sb]) { |
252 | for (i = 0; i < ATRAC3P_SUBBAND_SAMPLES; i++) |
253 | FFSWAP(float, out[0][sb * ATRAC3P_SUBBAND_SAMPLES + i], |
254 | out[1][sb * ATRAC3P_SUBBAND_SAMPLES + i]); |
255 | } |
256 | |
257 | /* flip coefficients' sign if requested */ |
258 | if (ctx->negate_coeffs[sb]) |
259 | for (i = 0; i < ATRAC3P_SUBBAND_SAMPLES; i++) |
260 | out[1][sb * ATRAC3P_SUBBAND_SAMPLES + i] = -(out[1][sb * ATRAC3P_SUBBAND_SAMPLES + i]); |
261 | } |
262 | } |
263 | } |
264 | |
265 | static void reconstruct_frame(ATRAC3PContext *ctx, Atrac3pChanUnitCtx *ch_unit, |
266 | int num_channels, AVCodecContext *avctx) |
267 | { |
268 | int ch, sb; |
269 | |
270 | for (ch = 0; ch < num_channels; ch++) { |
271 | for (sb = 0; sb < ch_unit->num_subbands; sb++) { |
272 | /* inverse transform and windowing */ |
273 | ff_atrac3p_imdct(ctx->fdsp, &ctx->mdct_ctx, |
274 | &ctx->samples[ch][sb * ATRAC3P_SUBBAND_SAMPLES], |
275 | &ctx->mdct_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES], |
276 | (ch_unit->channels[ch].wnd_shape_prev[sb] << 1) + |
277 | ch_unit->channels[ch].wnd_shape[sb], sb); |
278 | |
279 | /* gain compensation and overlapping */ |
280 | ff_atrac_gain_compensation(&ctx->gainc_ctx, |
281 | &ctx->mdct_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES], |
282 | &ch_unit->prev_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES], |
283 | &ch_unit->channels[ch].gain_data_prev[sb], |
284 | &ch_unit->channels[ch].gain_data[sb], |
285 | ATRAC3P_SUBBAND_SAMPLES, |
286 | &ctx->time_buf[ch][sb * ATRAC3P_SUBBAND_SAMPLES]); |
287 | } |
288 | |
289 | /* zero unused subbands in both output and overlapping buffers */ |
290 | memset(&ch_unit->prev_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES], |
291 | 0, |
292 | (ATRAC3P_SUBBANDS - ch_unit->num_subbands) * |
293 | ATRAC3P_SUBBAND_SAMPLES * |
294 | sizeof(ch_unit->prev_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES])); |
295 | memset(&ctx->time_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES], |
296 | 0, |
297 | (ATRAC3P_SUBBANDS - ch_unit->num_subbands) * |
298 | ATRAC3P_SUBBAND_SAMPLES * |
299 | sizeof(ctx->time_buf[ch][ch_unit->num_subbands * ATRAC3P_SUBBAND_SAMPLES])); |
300 | |
301 | /* resynthesize and add tonal signal */ |
302 | if (ch_unit->waves_info->tones_present || |
303 | ch_unit->waves_info_prev->tones_present) { |
304 | for (sb = 0; sb < ch_unit->num_subbands; sb++) |
305 | if (ch_unit->channels[ch].tones_info[sb].num_wavs || |
306 | ch_unit->channels[ch].tones_info_prev[sb].num_wavs) { |
307 | ff_atrac3p_generate_tones(ch_unit, ctx->fdsp, ch, sb, |
308 | &ctx->time_buf[ch][sb * 128]); |
309 | } |
310 | } |
311 | |
312 | /* subband synthesis and acoustic signal output */ |
313 | ff_atrac3p_ipqf(&ctx->ipqf_dct_ctx, &ch_unit->ipqf_ctx[ch], |
314 | &ctx->time_buf[ch][0], &ctx->outp_buf[ch][0]); |
315 | } |
316 | |
317 | /* swap window shape and gain control buffers. */ |
318 | for (ch = 0; ch < num_channels; ch++) { |
319 | FFSWAP(uint8_t *, ch_unit->channels[ch].wnd_shape, |
320 | ch_unit->channels[ch].wnd_shape_prev); |
321 | FFSWAP(AtracGainInfo *, ch_unit->channels[ch].gain_data, |
322 | ch_unit->channels[ch].gain_data_prev); |
323 | FFSWAP(Atrac3pWavesData *, ch_unit->channels[ch].tones_info, |
324 | ch_unit->channels[ch].tones_info_prev); |
325 | } |
326 | |
327 | FFSWAP(Atrac3pWaveSynthParams *, ch_unit->waves_info, ch_unit->waves_info_prev); |
328 | } |
329 | |
330 | static int atrac3p_decode_frame(AVCodecContext *avctx, void *data, |
331 | int *got_frame_ptr, AVPacket *avpkt) |
332 | { |
333 | ATRAC3PContext *ctx = avctx->priv_data; |
334 | AVFrame *frame = data; |
335 | int i, ret, ch_unit_id, ch_block = 0, out_ch_index = 0, channels_to_process; |
336 | float **samples_p = (float **)frame->extended_data; |
337 | |
338 | frame->nb_samples = ATRAC3P_FRAME_SAMPLES; |
339 | if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) |
340 | return ret; |
341 | |
342 | if ((ret = init_get_bits8(&ctx->gb, avpkt->data, avpkt->size)) < 0) |
343 | return ret; |
344 | |
345 | if (get_bits1(&ctx->gb)) { |
346 | av_log(avctx, AV_LOG_ERROR, "Invalid start bit!\n"); |
347 | return AVERROR_INVALIDDATA; |
348 | } |
349 | |
350 | while (get_bits_left(&ctx->gb) >= 2 && |
351 | (ch_unit_id = get_bits(&ctx->gb, 2)) != CH_UNIT_TERMINATOR) { |
352 | if (ch_unit_id == CH_UNIT_EXTENSION) { |
353 | avpriv_report_missing_feature(avctx, "Channel unit extension"); |
354 | return AVERROR_PATCHWELCOME; |
355 | } |
356 | if (ch_block >= ctx->num_channel_blocks || |
357 | ctx->channel_blocks[ch_block] != ch_unit_id) { |
358 | av_log(avctx, AV_LOG_ERROR, |
359 | "Frame data doesn't match channel configuration!\n"); |
360 | return AVERROR_INVALIDDATA; |
361 | } |
362 | |
363 | ctx->ch_units[ch_block].unit_type = ch_unit_id; |
364 | channels_to_process = ch_unit_id + 1; |
365 | |
366 | if ((ret = ff_atrac3p_decode_channel_unit(&ctx->gb, |
367 | &ctx->ch_units[ch_block], |
368 | channels_to_process, |
369 | avctx)) < 0) |
370 | return ret; |
371 | |
372 | decode_residual_spectrum(&ctx->ch_units[ch_block], ctx->samples, |
373 | channels_to_process, avctx); |
374 | reconstruct_frame(ctx, &ctx->ch_units[ch_block], |
375 | channels_to_process, avctx); |
376 | |
377 | for (i = 0; i < channels_to_process; i++) |
378 | memcpy(samples_p[out_ch_index + i], ctx->outp_buf[i], |
379 | ATRAC3P_FRAME_SAMPLES * sizeof(**samples_p)); |
380 | |
381 | ch_block++; |
382 | out_ch_index += channels_to_process; |
383 | } |
384 | |
385 | *got_frame_ptr = 1; |
386 | |
387 | return avctx->codec_id == AV_CODEC_ID_ATRAC3P ? FFMIN(avctx->block_align, avpkt->size) : avpkt->size; |
388 | } |
389 | |
390 | AVCodec ff_atrac3p_decoder = { |
391 | .name = "atrac3plus", |
392 | .long_name = NULL_IF_CONFIG_SMALL("ATRAC3+ (Adaptive TRansform Acoustic Coding 3+)"), |
393 | .type = AVMEDIA_TYPE_AUDIO, |
394 | .id = AV_CODEC_ID_ATRAC3P, |
395 | .capabilities = AV_CODEC_CAP_DR1, |
396 | .priv_data_size = sizeof(ATRAC3PContext), |
397 | .init = atrac3p_decode_init, |
398 | .close = atrac3p_decode_close, |
399 | .decode = atrac3p_decode_frame, |
400 | }; |
401 | |
402 | AVCodec ff_atrac3pal_decoder = { |
403 | .name = "atrac3plusal", |
404 | .long_name = NULL_IF_CONFIG_SMALL("ATRAC3+ AL (Adaptive TRansform Acoustic Coding 3+ Advanced Lossless)"), |
405 | .type = AVMEDIA_TYPE_AUDIO, |
406 | .id = AV_CODEC_ID_ATRAC3PAL, |
407 | .capabilities = AV_CODEC_CAP_DR1, |
408 | .priv_data_size = sizeof(ATRAC3PContext), |
409 | .init = atrac3p_decode_init, |
410 | .close = atrac3p_decode_close, |
411 | .decode = atrac3p_decode_frame, |
412 | }; |
413 |