aac: SBR support (without PS)
[nihav.git] / nihav-mpeg / src / codecs / aac / mod.rs
CommitLineData
999c25f7
KS
1use nihav_core::formats::*;
2use nihav_core::frame::*;
3use nihav_core::codecs::*;
4use nihav_codec_support::dsp::mdct::IMDCT;
5use nihav_codec_support::dsp::window::*;
6use nihav_core::io::bitreader::*;
7use nihav_core::io::byteio::*;
8use std::str::FromStr;
9
10mod coeff_read;
11use coeff_read::*;
12mod info;
13use info::*;
0e18270d
KS
14mod sbr;
15use sbr::{SBRHeader, SBRCodebooks, SBRState, SBRChannel, SBRDSP, sbr_read_sce, sbr_read_cpe};
27608aa6 16#[allow(clippy::excessive_precision)]
999c25f7
KS
17mod tables;
18use tables::*;
19mod tools;
20use tools::*;
21
22const MAX_WINDOWS: usize = 8;
23const MAX_SFBS: usize = 64;
24
25#[derive(Clone,Copy)]
26pub struct ICSInfo {
27 window_sequence: u8,
28 prev_window_sequence: u8,
29 window_shape: bool,
30 prev_window_shape: bool,
31 scale_factor_grouping: [bool; MAX_WINDOWS],
32 group_start: [usize; MAX_WINDOWS],
33 window_groups: usize,
34 num_windows: usize,
35 max_sfb: usize,
36 predictor_data: Option<LTPData>,
37 long_win: bool,
38}
39
40const ONLY_LONG_SEQUENCE: u8 = 0;
41const LONG_START_SEQUENCE: u8 = 1;
42const EIGHT_SHORT_SEQUENCE: u8 = 2;
43const LONG_STOP_SEQUENCE: u8 = 3;
44
45impl ICSInfo {
46 fn new() -> Self {
47 Self {
48 window_sequence: 0,
49 prev_window_sequence: 0,
50 window_shape: false,
51 prev_window_shape: false,
52 scale_factor_grouping: [false; MAX_WINDOWS],
53 group_start: [0; MAX_WINDOWS],
54 num_windows: 0,
55 window_groups: 0,
56 max_sfb: 0,
57 predictor_data: None,
58 long_win: true,
59 }
60 }
61 fn decode_ics_info(&mut self, br: &mut BitReader) -> DecoderResult<()> {
62 self.prev_window_sequence = self.window_sequence;
63 self.prev_window_shape = self.window_shape;
64 let ics_reserved_bit = br.read(1)?;
65 validate!(ics_reserved_bit == 0);
66 self.window_sequence = br.read(2)? as u8;
67 match self.prev_window_sequence {
68 ONLY_LONG_SEQUENCE | LONG_STOP_SEQUENCE => {
69 if (self.window_sequence != ONLY_LONG_SEQUENCE) &&
70 (self.window_sequence != LONG_START_SEQUENCE) {
71 println!("incorrect previous window");
72 }
73 },
74 LONG_START_SEQUENCE | EIGHT_SHORT_SEQUENCE => {
75 if (self.window_sequence != EIGHT_SHORT_SEQUENCE) &&
76 (self.window_sequence != LONG_STOP_SEQUENCE) {
77 println!("incorrect previous window");
78 }
79 },
80 _ => {},
81 };
82 self.window_shape = br.read_bool()?;
83 self.window_groups = 1;
84 if self.window_sequence == EIGHT_SHORT_SEQUENCE {
85 self.long_win = false;
86 self.num_windows = 8;
87 self.max_sfb = br.read(4)? as usize;
88 for i in 0..MAX_WINDOWS-1 {
89 self.scale_factor_grouping[i] = br.read_bool()?;
90 if !self.scale_factor_grouping[i] {
91 self.group_start[self.window_groups] = i + 1;
92 self.window_groups += 1;
93 }
94 }
95 } else {
96 self.long_win = true;
97 self.num_windows = 1;
98 self.max_sfb = br.read(6)? as usize;
99 self.predictor_data = LTPData::read(br)?;
100 }
101 Ok(())
102 }
103 fn get_group_start(&self, g: usize) -> usize {
104 if g == 0 {
105 0
106 } else if g >= self.window_groups {
107 if self.long_win { 1 } else { 8 }
108 } else {
109 self.group_start[g]
110 }
111 }
112}
113
114#[derive(Clone)]
115struct ICS {
116 global_gain: u8,
117 info: ICSInfo,
118 pulse_data: Option<PulseData>,
119 tns_data: Option<TNSData>,
120 gain_control: Option<GainControlData>,
121 sect_cb: [[u8; MAX_SFBS]; MAX_WINDOWS],
122 sect_len: [[usize; MAX_SFBS]; MAX_WINDOWS],
123 sfb_cb: [[u8; MAX_SFBS]; MAX_WINDOWS],
124 num_sec: [usize; MAX_WINDOWS],
125 scales: [[u8; MAX_SFBS]; MAX_WINDOWS],
126 sbinfo: GASubbandInfo,
127 coeffs: [f32; 1024],
128 delay: [f32; 1024],
129}
130
131const INTENSITY_SCALE_MIN: i16 = -155;
132impl ICS {
133 fn new(sbinfo: GASubbandInfo) -> Self {
134 Self {
135 global_gain: 0,
136 info: ICSInfo::new(),
137 pulse_data: None,
138 tns_data: None,
139 gain_control: None,
140 sect_cb: [[0; MAX_SFBS]; MAX_WINDOWS],
141 sect_len: [[0; MAX_SFBS]; MAX_WINDOWS],
142 sfb_cb: [[0; MAX_SFBS]; MAX_WINDOWS],
143 scales: [[0; MAX_SFBS]; MAX_WINDOWS],
144 num_sec: [0; MAX_WINDOWS],
145 sbinfo,
146 coeffs: [0.0; 1024],
147 delay: [0.0; 1024],
148 }
149 }
150 fn decode_section_data(&mut self, br: &mut BitReader, may_have_intensity: bool) -> DecoderResult<()> {
151 let sect_bits = if self.info.long_win { 5 } else { 3 };
152 let sect_esc_val = (1 << sect_bits) - 1;
153
154 for g in 0..self.info.window_groups {
155 let mut k = 0;
156 let mut l = 0;
157 while k < self.info.max_sfb {
158 self.sect_cb[g][l] = br.read(4)? as u8;
159 self.sect_len[g][l] = 0;
160 validate!(self.sect_cb[g][l] != RESERVED_HCB);
161 if ((self.sect_cb[g][l] == INTENSITY_HCB) || (self.sect_cb[g][l] == INTENSITY_HCB2)) && !may_have_intensity {
162 return Err(DecoderError::InvalidData);
163 }
164 loop {
165 let sect_len_incr = br.read(sect_bits)? as usize;
166 self.sect_len[g][l] += sect_len_incr;
167 if sect_len_incr < sect_esc_val { break; }
168 }
169 validate!(k + self.sect_len[g][l] <= self.info.max_sfb);
170 for _ in 0..self.sect_len[g][l] {
171 self.sfb_cb[g][k] = self.sect_cb[g][l];
172 k += 1;
173 }
174 l += 1;
175 }
176 self.num_sec[g] = l;
177 }
178 Ok(())
179 }
180 fn is_intensity(&self, g: usize, sfb: usize) -> bool {
181 (self.sfb_cb[g][sfb] == INTENSITY_HCB) || (self.sfb_cb[g][sfb] == INTENSITY_HCB2)
182 }
183 fn get_intensity_dir(&self, g: usize, sfb: usize) -> bool {
184 self.sfb_cb[g][sfb] == INTENSITY_HCB
185 }
186 fn decode_scale_factor_data(&mut self, br: &mut BitReader, codebooks: &Codebooks) -> DecoderResult<()> {
187 decode_scale_factor_data(br, &mut self.scales, self.global_gain, &self.info, &self.sfb_cb, codebooks)
188 }
189 fn get_band_start(&self, swb: usize) -> usize {
190 if self.info.long_win {
191 self.sbinfo.long_bands[swb]
192 } else {
193 self.sbinfo.short_bands[swb]
194 }
195 }
196 fn get_num_bands(&self) -> usize {
197 if self.info.long_win {
198 self.sbinfo.long_bands.len() - 1
199 } else {
200 self.sbinfo.short_bands.len() - 1
201 }
202 }
203 fn decode_spectrum(&mut self, br: &mut BitReader, codebooks: &Codebooks) -> DecoderResult<()> {
204 self.coeffs = [0.0; 1024];
205 decode_spectrum(br, &mut self.coeffs, &self.scales, &self.info, &self.sbinfo, &self.sfb_cb, codebooks)
206 }
27608aa6
KS
207 fn iquant(val: f32) -> f32 {
208 if val < 0.0 {
209 -((-val).powf(4.0 / 3.0))
210 } else {
211 val.powf(4.0 / 3.0)
212 }
213 }
999c25f7
KS
214 fn place_pulses(&mut self) {
215 if let Some(ref pdata) = self.pulse_data {
216 if pdata.pulse_start_sfb >= self.sbinfo.long_bands.len() - 1 { return; }
217 let mut k = self.get_band_start(pdata.pulse_start_sfb);
218 let mut band = pdata.pulse_start_sfb;
219 for pno in 0..pdata.number_pulse {
220 k += pdata.pulse_offset[pno] as usize;
221 if k >= 1024 { return; }
222 while self.get_band_start(band + 1) <= k { band += 1; }
223 let scale = get_scale(self.scales[0][band]);
224 let mut base = self.coeffs[k];
225 if base != 0.0 {
226 base = requant(self.coeffs[k], scale);
227 }
228 if base > 0.0 {
229 base += f32::from(pdata.pulse_amp[pno]);
230 } else {
231 base -= f32::from(pdata.pulse_amp[pno]);
232 }
27608aa6 233 self.coeffs[k] = Self::iquant(base) * scale;
999c25f7
KS
234 }
235 }
236 }
237 fn decode_ics(&mut self, br: &mut BitReader, codebooks: &Codebooks, m4atype: M4AType, common_window: bool, may_have_intensity: bool) -> DecoderResult<()> {
238 self.global_gain = br.read(8)? as u8;
239 if !common_window {
240 self.info.decode_ics_info(br)?;
241 }
242 self.decode_section_data(br, may_have_intensity)?;
243 self.decode_scale_factor_data(br, codebooks)?;
244 self.pulse_data = PulseData::read(br)?;
245 validate!(self.pulse_data.is_none() || self.info.long_win);
246 let tns_max_order;
247 if !self.info.long_win {
248 tns_max_order = 7;
249 } else if m4atype == M4AType::LC {
250 tns_max_order = 12;
251 } else {
252 tns_max_order = TNS_MAX_ORDER;
253 }
254 self.tns_data = TNSData::read(br, self.info.long_win, self.info.num_windows, tns_max_order)?;
255 if m4atype == M4AType::SSR {
256 self.gain_control = GainControlData::read(br)?;
257 } else {
258 let gain_control_data_present = br.read_bool()?;
259 validate!(!gain_control_data_present);
260 }
261 self.decode_spectrum(br, codebooks)?;
262 Ok(())
263 }
264 fn synth_channel(&mut self, dsp: &mut DSP, dst: &mut [f32], srate_idx: usize) {
265 self.place_pulses();
266 if let Some(ref tns_data) = self.tns_data {
267 let tns_max_bands = TNSData::get_max_bands(self.info.long_win, srate_idx).min(self.info.max_sfb);
268 for w in 0..self.info.num_windows {
269 let mut bottom = self.get_num_bands();
270 for f in 0..tns_data.n_filt[w] {
271 let top = bottom;
272 bottom = if top >= tns_data.coeffs[w][f].length { top - tns_data.coeffs[w][f].length } else { 0 };
273 let order = tns_data.coeffs[w][f].order;
274 if order == 0 { continue; }
275 let start = w * 128 + self.get_band_start(tns_max_bands.min(bottom));
276 let end = w * 128 + self.get_band_start(tns_max_bands.min(top));
277 tns_data.apply(&mut self.coeffs, w, f, start, end);
278 }
279 }
280 }
281 dsp.synth(&self.coeffs, &mut self.delay, self.info.window_sequence, self.info.window_shape, self.info.prev_window_shape, dst);
282 }
283}
284
285#[derive(Clone)]
286struct ChannelPair {
287 pair: bool,
288 channel: usize,
289 common_window: bool,
290 ms_mask_present: u8,
291 ms_used: [[bool; MAX_SFBS]; MAX_WINDOWS],
292 ics: [ICS; 2],
0e18270d
KS
293 sbr_hdr: SBRHeader,
294 sbr_state: SBRState,
295 sbr_ch: [SBRChannel; 2],
296 do_sbr: bool,
999c25f7
KS
297}
298
299impl ChannelPair {
300 fn new(pair: bool, channel: usize, sbinfo: GASubbandInfo) -> Self {
301 Self {
302 pair, channel,
303 common_window: false,
304 ms_mask_present: 0,
305 ms_used: [[false; MAX_SFBS]; MAX_WINDOWS],
306 ics: [ICS::new(sbinfo), ICS::new(sbinfo)],
0e18270d
KS
307 sbr_hdr: SBRHeader::new(),
308 sbr_state: SBRState::new(),
309 sbr_ch: [SBRChannel::new(), SBRChannel::new()],
310 do_sbr: false,
999c25f7
KS
311 }
312 }
313 fn decode_ga_sce(&mut self, br: &mut BitReader, codebooks: &Codebooks, m4atype: M4AType) -> DecoderResult<()> {
314 self.ics[0].decode_ics(br, codebooks, m4atype, false, false)?;
315 Ok(())
316 }
317 fn decode_ga_cpe(&mut self, br: &mut BitReader, codebooks: &Codebooks, m4atype: M4AType) -> DecoderResult<()> {
318 let common_window = br.read_bool()?;
319 self.common_window = common_window;
320 if common_window {
321 self.ics[0].info.decode_ics_info(br)?;
322 self.ms_mask_present = br.read(2)? as u8;
323 validate!(self.ms_mask_present != 3);
324 if self.ms_mask_present == 1 {
325 for g in 0..self.ics[0].info.window_groups {
326 for sfb in 0..self.ics[0].info.max_sfb {
327 self.ms_used[g][sfb] = br.read_bool()?;
328 }
329 }
330 }
331 self.ics[1].info = self.ics[0].info;
332 }
333 self.ics[0].decode_ics(br, codebooks, m4atype, common_window, true)?;
334 self.ics[1].decode_ics(br, codebooks, m4atype, common_window, true)?;
335 if common_window && self.ms_mask_present != 0 {
336 let mut g = 0;
337 for w in 0..self.ics[0].info.num_windows {
338 if w > 0 && !self.ics[0].info.scale_factor_grouping[w - 1] {
339 g += 1;
340 }
341 for sfb in 0..self.ics[0].info.max_sfb {
342 let start = w * 128 + self.ics[0].get_band_start(sfb);
343 let end = w * 128 + self.ics[0].get_band_start(sfb + 1);
344 if self.ics[0].is_intensity(g, sfb) {
345 let invert = (self.ms_mask_present == 1) && self.ms_used[g][sfb];
346 let dir = self.ics[0].get_intensity_dir(g, sfb) ^ invert;
347 let scale = 0.5f32.powf(0.25 * (f32::from(self.ics[0].scales[g][sfb]) + f32::from(INTENSITY_SCALE_MIN)));
348 if !dir {
349 for i in start..end {
350 self.ics[1].coeffs[i] = scale * self.ics[0].coeffs[i];
351 }
352 } else {
353 for i in start..end {
354 self.ics[1].coeffs[i] = -scale * self.ics[0].coeffs[i];
355 }
356 }
357 } else if (self.ms_mask_present == 2) || self.ms_used[g][sfb] {
358 for i in start..end {
359 let tmp = self.ics[0].coeffs[i] - self.ics[1].coeffs[i];
360 self.ics[0].coeffs[i] += self.ics[1].coeffs[i];
361 self.ics[1].coeffs[i] = tmp;
362 }
363 }
364 }
365 }
366 }
367 Ok(())
368 }
0e18270d
KS
369 fn decode_sbr(&mut self, buf: &[u8], has_crc: bool, cbs: &SBRCodebooks, srate: u32) -> DecoderResult<()> {
370 let mut br = BitReader::new(buf, BitReaderMode::BE);
371 if has_crc {
372 let _bs_sbr_crc_bits = br.read(10)?;
373 }
374 if br.read_bool()? {
375 if let Ok(hdr) = SBRHeader::read(&mut br) {
376 if self.sbr_hdr.differs_from(&hdr) {
377 self.do_sbr = self.sbr_state.init(&hdr, srate).is_ok();
378 self.sbr_ch[0].reset();
379 self.sbr_ch[1].reset();
380 }
381 self.sbr_hdr = hdr;
382 } else {
383 self.do_sbr = false;
384 }
385 }
386 if self.do_sbr {
387 if !self.pair {
388 sbr_read_sce(&mut br, self.sbr_hdr.amp_res, &self.sbr_state, cbs, &mut self.sbr_ch[0])?;
389 } else {
390 sbr_read_cpe(&mut br, self.sbr_hdr.amp_res, &self.sbr_state, cbs, &mut self.sbr_ch)?;
391 }
392 }
393
394 Ok(())
395 }
396 fn synth_audio(&mut self, dsp: &mut DSP, abuf: &mut NABufferType, srate_idx: usize, upsample: bool) {
999c25f7
KS
397 let mut adata = abuf.get_abuf_f32().unwrap();
398 let output = adata.get_data_mut().unwrap();
399 let off0 = abuf.get_offset(self.channel);
400 let off1 = abuf.get_offset(self.channel + 1);
0e18270d
KS
401 if !upsample {
402 self.ics[0].synth_channel(dsp, &mut output[off0..], srate_idx);
403 if self.pair {
404 self.ics[1].synth_channel(dsp, &mut output[off1..], srate_idx);
405 }
406 } else {
407 let mut tmp = [0.0; 1024];
408 let nchannels = if self.pair { 2 } else { 1 };
409 for ch in 0..nchannels {
410 let off = if ch == 0 { off0 } else { off1 };
411 self.ics[ch].synth_channel(dsp, &mut tmp, srate_idx);
412 self.sbr_ch[ch].analysis(&mut dsp.sbr_dsp, &tmp);
413 if self.do_sbr {
414 self.sbr_ch[ch].hf_generate(&self.sbr_state);
415 self.sbr_ch[ch].hf_adjust(&self.sbr_state, &self.sbr_hdr);
416 } else {
417 self.sbr_ch[ch].bypass();
418 }
419 self.sbr_ch[ch].synthesis(&mut dsp.sbr_dsp, &mut output[off..][..2048]);
420 self.sbr_ch[ch].update_frame();
421 }
999c25f7
KS
422 }
423 }
424}
425
426struct DSP {
427 kbd_long_win: [f32; 1024],
428 kbd_short_win: [f32; 128],
429 sine_long_win: [f32; 1024],
430 sine_short_win: [f32; 128],
431 imdct_long: IMDCT,
432 imdct_short: IMDCT,
433 tmp: [f32; 2048],
434 ew_buf: [f32; 1152],
0e18270d 435 sbr_dsp: SBRDSP,
999c25f7
KS
436}
437
438const SHORT_WIN_POINT0: usize = 512 - 64;
439const SHORT_WIN_POINT1: usize = 512 + 64;
440
441impl DSP {
442 fn new() -> Self {
443 let mut kbd_long_win: [f32; 1024] = [0.0; 1024];
444 let mut kbd_short_win: [f32; 128] = [0.0; 128];
445 generate_window(WindowType::KaiserBessel(4.0), 1.0, 1024, true, &mut kbd_long_win);
446 generate_window(WindowType::KaiserBessel(6.0), 1.0, 128, true, &mut kbd_short_win);
447 let mut sine_long_win: [f32; 1024] = [0.0; 1024];
448 let mut sine_short_win: [f32; 128] = [0.0; 128];
449 generate_window(WindowType::Sine, 1.0, 1024, true, &mut sine_long_win);
450 generate_window(WindowType::Sine, 1.0, 128, true, &mut sine_short_win);
451 Self {
452 kbd_long_win, kbd_short_win,
453 sine_long_win, sine_short_win,
454 imdct_long: IMDCT::new(1024 * 2, true),
455 imdct_short: IMDCT::new(128 * 2, true),
456 tmp: [0.0; 2048], ew_buf: [0.0; 1152],
0e18270d 457 sbr_dsp: SBRDSP::new(),
999c25f7
KS
458 }
459 }
460 #[allow(clippy::cognitive_complexity)]
461 fn synth(&mut self, coeffs: &[f32; 1024], delay: &mut [f32; 1024], seq: u8, window_shape: bool, prev_window_shape: bool, dst: &mut [f32]) {
462 let long_win = if window_shape { &self.kbd_long_win } else { &self.sine_long_win };
463 let short_win = if window_shape { &self.kbd_short_win } else { &self.sine_short_win };
464 let left_long_win = if prev_window_shape { &self.kbd_long_win } else { &self.sine_long_win };
465 let left_short_win = if prev_window_shape { &self.kbd_short_win } else { &self.sine_short_win };
466 if seq != EIGHT_SHORT_SEQUENCE {
467 self.imdct_long.imdct(coeffs, &mut self.tmp);
468 } else {
469 for (ain, aout) in coeffs.chunks(128).zip(self.tmp.chunks_mut(256)) {
470 self.imdct_short.imdct(ain, aout);
471 }
472 self.ew_buf = [0.0; 1152];
473 for (w, src) in self.tmp.chunks(256).enumerate() {
474 if w > 0 {
475 for i in 0..128 {
476 self.ew_buf[w * 128 + i] += src[i] * short_win[i];
477 }
478 } else { // to be left-windowed
479 for i in 0..128 {
480 self.ew_buf[i] = src[i];
481 }
482 }
483 for i in 0..128 {
484 self.ew_buf[w * 128 + i + 128] += src[i + 128] * short_win[127 - i];
485 }
486 }
487 }
488 if seq == ONLY_LONG_SEQUENCE { // should be the most common case
489 for i in 0..1024 {
490 dst[i] = delay[i] + self.tmp[i] * left_long_win[i];
491 delay[i] = self.tmp[i + 1024] * long_win[1023 - i];
492 }
493 return;
494 }
495 // output new data
496 match seq {
497 ONLY_LONG_SEQUENCE | LONG_START_SEQUENCE => {
498 for i in 0..1024 {
499 dst[i] = self.tmp[i] * left_long_win[i] + delay[i];
500 }
501 },
502 EIGHT_SHORT_SEQUENCE => {
503 for i in 0..SHORT_WIN_POINT0 {
504 dst[i] = delay[i];
505 }
506 for i in SHORT_WIN_POINT0..SHORT_WIN_POINT1 {
507 let j = i - SHORT_WIN_POINT0;
508 dst[i] = delay[i] + self.ew_buf[j] * left_short_win[j];
509 }
510 for i in SHORT_WIN_POINT1..1024 {
511 let j = i - SHORT_WIN_POINT0;
512 dst[i] = self.ew_buf[j];
513 }
514 },
515 LONG_STOP_SEQUENCE => {
516 for i in 0..SHORT_WIN_POINT0 {
517 dst[i] = delay[i];
518 }
519 for i in SHORT_WIN_POINT0..SHORT_WIN_POINT1 {
520 dst[i] = delay[i] + self.tmp[i] * left_short_win[i - SHORT_WIN_POINT0];
521 }
522 for i in SHORT_WIN_POINT1..1024 {
523 dst[i] = self.tmp[i];
524 }
525 },
526 _ => unreachable!(""),
527 };
528 // save delay
529 match seq {
530 ONLY_LONG_SEQUENCE | LONG_STOP_SEQUENCE => {
531 for i in 0..1024 {
532 delay[i] = self.tmp[i + 1024] * long_win[1023 - i];
533 }
534 },
535 EIGHT_SHORT_SEQUENCE => {
536 for i in 0..SHORT_WIN_POINT1 { // last part is already windowed
537 delay[i] = self.ew_buf[i + 512+64];
538 }
539 for i in SHORT_WIN_POINT1..1024 {
540 delay[i] = 0.0;
541 }
542 },
543 LONG_START_SEQUENCE => {
544 for i in 0..SHORT_WIN_POINT0 {
545 delay[i] = self.tmp[i + 1024];
546 }
547 for i in SHORT_WIN_POINT0..SHORT_WIN_POINT1 {
548 delay[i] = self.tmp[i + 1024] * short_win[127 - (i - SHORT_WIN_POINT0)];
549 }
550 for i in SHORT_WIN_POINT1..1024 {
551 delay[i] = 0.0;
552 }
553 },
554 _ => unreachable!(""),
555 };
556 }
557}
558
559struct AACDecoder {
560 info: NACodecInfoRef,
561 chmap: NAChannelMap,
562 m4ainfo: M4AInfo,
563 pairs: Vec<ChannelPair>,
564 codebooks: Codebooks,
565 dsp: DSP,
566 sbinfo: GASubbandInfo,
0e18270d
KS
567 sbr_cbs: SBRCodebooks,
568 upsample: bool,
999c25f7
KS
569}
570
571impl AACDecoder {
572 fn new() -> Self {
573 AACDecoder {
574 info: NACodecInfo::new_dummy(),
575 chmap: NAChannelMap::new(),
576 m4ainfo: M4AInfo::new(),
577 pairs: Vec::new(),
578 codebooks: Codebooks::new(),
579 dsp: DSP::new(),
580 sbinfo: AAC_SUBBAND_INFO[0],
0e18270d
KS
581 sbr_cbs: SBRCodebooks::new(),
582 upsample: false,
999c25f7
KS
583 }
584 }
585 fn set_pair(&mut self, pair_no: usize, channel: usize, pair: bool) -> DecoderResult<()> {
586 if self.pairs.len() <= pair_no {
587 self.pairs.push(ChannelPair::new(pair, channel, self.sbinfo));
588 } else {
589 validate!(self.pairs[pair_no].channel == channel);
590 validate!(self.pairs[pair_no].pair == pair);
591 }
592 validate!(if pair { channel + 1 } else { channel } < self.m4ainfo.channels);
593 Ok(())
594 }
595 fn decode_ga(&mut self, br: &mut BitReader, abuf: &mut NABufferType) -> DecoderResult<()> {
596 let mut cur_pair = 0;
597 let mut cur_ch = 0;
598 while br.left() > 3 {
599 let id = br.read(3)?;
600 match id {
601 0 => { // ID_SCE
602 let _tag = br.read(4)?;
603 self.set_pair(cur_pair, cur_ch, false)?;
604 self.pairs[cur_pair].decode_ga_sce(br, &self.codebooks, self.m4ainfo.otype)?;
605 cur_pair += 1;
606 cur_ch += 1;
607 },
608 1 => { // ID_CPE
609 let _tag = br.read(4)?;
610 self.set_pair(cur_pair, cur_ch, true)?;
611 self.pairs[cur_pair].decode_ga_cpe(br, &self.codebooks, self.m4ainfo.otype)?;
612 cur_pair += 1;
613 cur_ch += 2;
614 },
615 2 => { // ID_CCE
616 unimplemented!("coupling channel element");
617 },
618 3 => { // ID_LFE
619 let _tag = br.read(4)?;
620 self.set_pair(cur_pair, cur_ch, false)?;
621 self.pairs[cur_pair].decode_ga_sce(br, &self.codebooks, self.m4ainfo.otype)?;
622 cur_pair += 1;
623 cur_ch += 1;
624 },
625 4 => { // ID_DSE
626 let _id = br.read(4)?;
627 let align = br.read_bool()?;
628 let mut count = br.read(8)? as u32;
629 if count == 255 { count += br.read(8)? as u32; }
630 if align { br.align(); }
631 br.skip(count * 8)?; // no SBR payload or such
632 },
633 5 => { // ID_PCE
9c3cd867 634 skimp_through_program_config_element(br)?;
999c25f7
KS
635 },
636 6 => { // ID_FIL
637 let mut count = br.read(4)? as usize;
638 if count == 15 {
639 count += br.read(8)? as usize;
640 count -= 1;
641 }
0e18270d
KS
642 if count > 0 {
643 let extension_type = br.read(4)?;
644 match extension_type {
645 0xD | 0xE => { // SBR data without or with CRC
646 let has_crc = extension_type == 0xE;
647
648 let mut buf = [0; 256 + 16];
649 for el in buf[..count - 1].iter_mut() {
650 *el = br.read(8)? as u8;
651 }
652 buf[count - 1] = br.read(4)? as u8;
653 if cur_pair > 0 {
654 self.pairs[cur_pair - 1].decode_sbr(&buf[..count], has_crc, &self.sbr_cbs, self.m4ainfo.srate * 2)?;
655 }
656 },
657 _ => {
658 for _ in 0..count-1 {
999c25f7 659 br.skip(8)?;
0e18270d
KS
660 }
661 br.skip(4)?;
662 },
663 };
999c25f7
KS
664 }
665 },
666 7 => { // ID_TERM
667 break;
668 },
669 _ => { unreachable!(); },
670 };
671 }
672 let srate_idx = GASubbandInfo::find_idx(self.m4ainfo.srate);
673 for pair in 0..cur_pair {
0e18270d 674 self.pairs[pair].synth_audio(&mut self.dsp, abuf, srate_idx, self.upsample);
999c25f7
KS
675 }
676 Ok(())
677 }
678}
679
680impl NADecoder for AACDecoder {
681 fn init(&mut self, _supp: &mut NADecoderSupport, info: NACodecInfoRef) -> DecoderResult<()> {
682 if let NACodecTypeInfo::Audio(_) = info.get_properties() {
683 let edata = info.get_extradata().unwrap();
684 validate!(edata.len() >= 2);
685
686//print!("edata:"); for s in edata.iter() { print!(" {:02X}", *s);}println!("");
687 if (edata.len() > 12) && (&edata[4..8] == b"esds") {
688 let mut mr = MemoryReader::new_read(edata.as_slice());
689 let mut br = ByteReader::new(&mut mr);
690 let esds_size = br.read_u32be()? as usize;
691 validate!(esds_size <= edata.len());
692 br.read_skip(8)?;
693 let mut info_start = 0;
694 let mut info_size = 0;
695 while br.tell() < (esds_size as u64) {
696 let tag = br.read_byte()?;
697 let mut size = 0;
698 loop {
699 let b = br.read_byte()?;
700 size = (size << 7) | u64::from(b & 0x7F);
701 validate!(br.tell() + size <= (esds_size as u64));
702 if (b & 0x80) == 0 {
703 break;
704 }
705 }
706 match tag {
707 3 => {
708 br.read_u16be()?;
709 let flags = br.read_byte()?;
710 if (flags & 0x80) != 0 {
711 br.read_u16be()?;
712 }
713 if (flags & 0x40) != 0 {
714 let len = br.read_byte()?;
715 br.read_skip(len as usize)?;
716 }
717 if (flags & 0x20) != 0 {
718 br.read_u16be()?;
719 }
720 },
721 4 => {
722 let _otype = br.read_byte()?;
723 let _stype = br.read_byte()?;
724 let _flags = br.read_u24be()?;
725 let _max_br = br.read_u32be()?;
726 let _min_br = br.read_u32be()?;
727 },
728 5 => {
729 info_start = br.tell() as usize;
730 info_size = size as usize;
731 break;
732 },
733 _ => br.read_skip(size as usize)?,
734 }
735 }
736 validate!(info_start > 0 && info_size > 0);
737 self.m4ainfo.read(&edata[info_start..][..info_size])?;
738 } else {
739 self.m4ainfo.read(&edata)?;
740 }
741
742 //println!("{}", self.m4ainfo);
743 if (self.m4ainfo.otype != M4AType::LC) || (self.m4ainfo.channels > 2) || (self.m4ainfo.samples != 1024) {
744 return Err(DecoderError::NotImplemented);
745 }
746 self.sbinfo = GASubbandInfo::find(self.m4ainfo.srate);
747
0e18270d
KS
748 self.upsample = self.m4ainfo.srate < 32000;
749 let (srate, samples) = if !self.upsample {
750 (self.m4ainfo.srate, self.m4ainfo.samples)
751 } else {
752 (self.m4ainfo.srate * 2, self.m4ainfo.samples * 2)
753 };
754
755 let ainfo = NAAudioInfo::new(srate, self.m4ainfo.channels as u8,
756 SND_F32P_FORMAT, samples);
999c25f7
KS
757 self.info = info.replace_info(NACodecTypeInfo::Audio(ainfo));
758
759 if self.m4ainfo.channels >= DEFAULT_CHANNEL_MAP.len() {
760 return Err(DecoderError::NotImplemented);
761 }
762 let chmap_str = DEFAULT_CHANNEL_MAP[self.m4ainfo.channels];
763 if chmap_str.is_empty() { return Err(DecoderError::NotImplemented); }
764 self.chmap = NAChannelMap::from_str(chmap_str).unwrap();
765
766 Ok(())
767 } else {
768 Err(DecoderError::InvalidData)
769 }
770 }
771 fn decode(&mut self, _supp: &mut NADecoderSupport, pkt: &NAPacket) -> DecoderResult<NAFrameRef> {
772 let info = pkt.get_stream().get_info();
773 validate!(info.get_properties().is_audio());
774 let pktbuf = pkt.get_buffer();
775
776 let ainfo = self.info.get_properties().get_audio_info().unwrap();
0e18270d
KS
777 let samples = if !self.upsample { self.m4ainfo.samples } else { self.m4ainfo.samples * 2 };
778 let mut abuf = alloc_audio_buffer(ainfo, samples, self.chmap.clone())?;
999c25f7
KS
779
780 let mut br = BitReader::new(&pktbuf, BitReaderMode::BE);
781 match self.m4ainfo.otype {
782 M4AType::LC => {
783 self.decode_ga(&mut br, &mut abuf)?;
784 },
785 _ => { unimplemented!(""); }
786 }
787
788 let mut frm = NAFrame::new_from_pkt(pkt, self.info.replace_info(NACodecTypeInfo::Audio(ainfo)), abuf);
789 frm.set_keyframe(true);
790 Ok(frm.into_ref())
791 }
792 fn flush(&mut self) {
793 for pair in self.pairs.iter_mut() {
794 pair.ics[0].delay = [0.0; 1024];
795 pair.ics[1].delay = [0.0; 1024];
796 }
797 }
798}
799
800impl NAOptionHandler for AACDecoder {
801 fn get_supported_options(&self) -> &[NAOptionDefinition] { &[] }
802 fn set_options(&mut self, _options: &[NAOption]) { }
803 fn query_option_value(&self, _name: &str) -> Option<NAValue> { None }
804}
805
806pub fn get_decoder() -> Box<dyn NADecoder + Send> {
807 Box::new(AACDecoder::new())
808}
809
810#[cfg(test)]
811mod test {
812 use nihav_core::codecs::RegisteredDecoders;
813 use nihav_core::demuxers::RegisteredDemuxers;
814 use nihav_codec_support::test::dec_video::test_decode_audio;
815 use crate::mpeg_register_all_decoders;
816 use nihav_realmedia::realmedia_register_all_demuxers;
817 #[test]
818 fn test_aac() {
819 let mut dmx_reg = RegisteredDemuxers::new();
820 realmedia_register_all_demuxers(&mut dmx_reg);
821 let mut dec_reg = RegisteredDecoders::new();
822 mpeg_register_all_decoders(&mut dec_reg);
823
824// let file = "assets/RV/rv40_weighted_mc.rmvb";
825 let file = "assets/RV/rv40_weighted_mc_2.rmvb";
826 test_decode_audio("realmedia", file, Some(12000), None/*Some("aac")*/, &dmx_reg, &dec_reg);
827 }
0e18270d
KS
828 #[test]
829 fn test_aac_sbr() {
830 let mut dmx_reg = RegisteredDemuxers::new();
831 nihav_commonfmt::generic_register_all_demuxers(&mut dmx_reg);
832 let mut dec_reg = RegisteredDecoders::new();
833 mpeg_register_all_decoders(&mut dec_reg);
834
835 let file = "assets/MPEG/SBRtestStereoAot29Sig0.mp4";
836 test_decode_audio("mov", file, Some(400), None/*Some("aacsbr")*/, &dmx_reg, &dec_reg);
837 }
999c25f7
KS
838}
839
840const DEFAULT_CHANNEL_MAP: [&str; 9] = [
841 "",
842 "C",
843 "L,R",
844 "C,L,R",
845 "C,L,R,Cs",
846 "C,L,R,Ls,Rs",
847 "C,L,R,Ls,Rs,LFE",
848 "",
849 "C,L,R,Ls,Rs,Lss,Rss,LFE",
850];