blob: e1ddaaf4f2b01db83ebf0c7249f59f6bda2f4455
1 | /* |
2 | * filename: flac_decode.c |
3 | * |
4 | * |
5 | * |
6 | * |
7 | * |
8 | */ |
9 | |
10 | #include <stdio.h> |
11 | #include <stdlib.h> |
12 | #include <limits.h> |
13 | #include "codec.h" |
14 | #ifndef __MW__ |
15 | #include <inttypes.h> |
16 | #else |
17 | #include "types.h" |
18 | #endif |
19 | #include "crc.h" |
20 | #include "avcodec.h" |
21 | #include "get_bits.h" |
22 | #include "golomb.h" |
23 | |
24 | #include "flac.h" |
25 | #include "flacdata.h" |
26 | #include "../../amadec/adec-armdec-mgt.h" |
27 | |
28 | #include <android/log.h> |
29 | |
30 | #define LOG_TAG "FlacDecoder" |
31 | #define audio_codec_print(...) __android_log_print(ANDROID_LOG_INFO,LOG_TAG,__VA_ARGS__) |
32 | |
33 | #define DefaultReadSize 1024*10 //read count from kernel audio buf one time |
34 | #define DefaultOutBufSize 1024*1024 |
35 | |
36 | typedef struct OutData { |
37 | uint8_t * outb; |
38 | int byte; |
39 | int index; |
40 | } OutData; |
41 | |
42 | typedef struct FLACContext { |
43 | FLACSTREAMINFO |
44 | |
45 | AVCodecContext *avctx; ///< parent AVCodecContext |
46 | GetBitContext gb; ///< GetBitContext initialized to start at the current frame |
47 | |
48 | int blocksize; ///< number of samples in the current frame |
49 | int curr_bps; ///< bps for current subframe, adjusted for channel correlation and wasted bits |
50 | int sample_shift; ///< shift required to make output samples 16-bit or 32-bit |
51 | int is32; ///< flag to indicate if output should be 32-bit instead of 16-bit |
52 | int ch_mode; ///< channel decorrelation type in the current frame |
53 | int got_streaminfo; ///< indicates if the STREAMINFO has been read |
54 | |
55 | int32_t *decoded[FLAC_MAX_CHANNELS]; ///< decoded samples |
56 | uint8_t *bitstream; |
57 | unsigned int bitstream_size; |
58 | unsigned int bitstream_index; |
59 | unsigned int allocated_bitstream_size; |
60 | } FLACContext; |
61 | |
62 | static const int sample_size_table[] = |
63 | { 0, 8, 12, 0, 16, 20, 24, 0 }; |
64 | |
65 | static AVCodecContext acodec; |
66 | static FLACContext flactext; |
67 | static OutData outbuffer; |
68 | #define INPUT_BUF_SIZE 32768 |
69 | static int64_t get_utf8(GetBitContext *gb) |
70 | { |
71 | int64_t val; |
72 | GET_UTF8(val, get_bits(gb, 8), return -1;) |
73 | return val; |
74 | } |
75 | |
76 | void *av_malloc(unsigned int size) |
77 | { |
78 | void *ptr = NULL; |
79 | |
80 | ptr = malloc(size); |
81 | return ptr; |
82 | } |
83 | |
84 | void *av_realloc(void *ptr, unsigned int size) |
85 | { |
86 | return realloc(ptr, size); |
87 | } |
88 | |
89 | void av_freep(void **arg) |
90 | { |
91 | if (*arg) { |
92 | free(*arg); |
93 | } |
94 | *arg = NULL; |
95 | } |
96 | void av_free(void *arg) |
97 | { |
98 | if (arg) { |
99 | free(arg); |
100 | } |
101 | } |
102 | void *av_mallocz(unsigned int size) |
103 | { |
104 | void *ptr = malloc(size); |
105 | if (ptr) { |
106 | memset(ptr, 0, size); |
107 | } |
108 | return ptr; |
109 | } |
110 | |
111 | static inline uint32_t bytestream_get_be32(const uint8_t** ptr) |
112 | { |
113 | uint32_t tmp; |
114 | tmp = (*ptr)[3] | ((*ptr)[2] << 8) | ((*ptr)[1] << 16) | ((*ptr)[0] << 24); |
115 | *ptr += 4; |
116 | return tmp; |
117 | } |
118 | |
119 | static inline uint32_t bytestream_get_be24(const uint8_t** ptr) |
120 | { |
121 | uint32_t tmp; |
122 | tmp = (*ptr)[2] | ((*ptr)[1] << 8) | ((*ptr)[0] << 16) ; |
123 | *ptr += 3; |
124 | return tmp; |
125 | } |
126 | |
127 | static inline uint8_t bytestream_get_byte(const uint8_t** ptr) |
128 | { |
129 | uint8_t tmp; |
130 | tmp = **ptr; |
131 | *ptr += 1; |
132 | return tmp; |
133 | } |
134 | |
135 | |
136 | |
137 | void *av_fast_realloc(void *ptr, unsigned int *size, unsigned int min_size) |
138 | { |
139 | if (min_size < *size) { |
140 | return ptr; |
141 | } |
142 | |
143 | *size = FFMAX(17 * min_size / 16 + 32, min_size); |
144 | |
145 | ptr = av_realloc(ptr, *size); |
146 | if (!ptr) { //we could set this to the unmodified min_size but this is safer if the user lost the ptr and uses NULL now |
147 | *size = 0; |
148 | } |
149 | |
150 | return ptr; |
151 | } |
152 | |
153 | static void allocate_buffers(FLACContext *s) |
154 | { |
155 | int i; |
156 | |
157 | if (s->max_framesize == 0 && s->max_blocksize) { |
158 | s->max_framesize = ff_flac_get_max_frame_size(s->max_blocksize, |
159 | s->channels, s->bps); |
160 | } |
161 | |
162 | for (i = 0; i < s->channels; i++) { |
163 | s->decoded[i] = av_realloc(s->decoded[i], |
164 | sizeof(int32_t) * s->max_blocksize); |
165 | } |
166 | } |
167 | |
168 | |
169 | /** |
170 | * Parse the STREAMINFO from an inline header. |
171 | * @param s the flac decoding context |
172 | * @param buf input buffer, starting with the "fLaC" marker |
173 | * @param buf_size buffer size |
174 | * @return non-zero if metadata is invalid |
175 | */ |
176 | static int parse_streaminfo(FLACContext *s, const uint8_t *buf, int buf_size) |
177 | { |
178 | int metadata_type, metadata_size; |
179 | |
180 | if (buf_size < FLAC_STREAMINFO_SIZE + 8) { |
181 | /* need more data */ |
182 | return 0; |
183 | } |
184 | ff_flac_parse_block_header(&buf[4], NULL, &metadata_type, &metadata_size); |
185 | if (metadata_type != FLAC_METADATA_TYPE_STREAMINFO || |
186 | metadata_size != FLAC_STREAMINFO_SIZE) { |
187 | return -1;//AVERROR_INVALIDDATA; |
188 | } |
189 | ff_flac_parse_streaminfo(s->avctx, (FLACStreaminfo *)s, &buf[8]); |
190 | allocate_buffers(s); |
191 | s->got_streaminfo = 1; |
192 | |
193 | return 0; |
194 | } |
195 | /** |
196 | * Determine the size of an inline header. |
197 | * @param buf input buffer, starting with the "fLaC" marker |
198 | * @param buf_size buffer size |
199 | * @return number of bytes in the header, or 0 if more data is needed |
200 | */ |
201 | static int get_metadata_size(const uint8_t *buf, int buf_size) |
202 | { |
203 | int metadata_last, metadata_size; |
204 | const uint8_t *buf_end = buf + buf_size; |
205 | |
206 | buf += 4; |
207 | do { |
208 | ff_flac_parse_block_header(buf, &metadata_last, NULL, &metadata_size); |
209 | buf += 4; |
210 | if (buf + metadata_size > buf_end) { |
211 | /* need more data in order to read the complete header */ |
212 | return 0; |
213 | } |
214 | buf += metadata_size; |
215 | } while (!metadata_last); |
216 | |
217 | return buf_size - (buf_end - buf); |
218 | } |
219 | void ff_flac_parse_streaminfo(AVCodecContext *avctx, struct FLACStreaminfo *s, |
220 | const uint8_t *buffer) |
221 | { |
222 | GetBitContext gb; |
223 | init_get_bits(&gb, buffer, FLAC_STREAMINFO_SIZE * 8); |
224 | |
225 | skip_bits(&gb, 16); /* skip min blocksize */ |
226 | s->max_blocksize = get_bits(&gb, 16); |
227 | if (s->max_blocksize < FLAC_MIN_BLOCKSIZE) { |
228 | audio_codec_print("invalid max blocksize: %d\n", s->max_blocksize); |
229 | s->max_blocksize = 16; |
230 | } |
231 | |
232 | skip_bits(&gb, 24); /* skip min frame size */ |
233 | s->max_framesize = get_bits_long(&gb, 24); |
234 | |
235 | s->samplerate = get_bits_long(&gb, 20); |
236 | s->channels = get_bits(&gb, 3) + 1; |
237 | s->bps = get_bits(&gb, 5) + 1; |
238 | |
239 | audio_codec_print("## METADATA sp=%d, ch=%d, bps=%d,-------------\n", |
240 | s->samplerate, s->channels, s->bps); |
241 | |
242 | avctx->channels = s->channels; |
243 | avctx->sample_rate = s->samplerate; |
244 | avctx->bits_per_raw_sample = s->bps; |
245 | |
246 | s->samples = get_bits_long(&gb, 32) << 4; |
247 | s->samples |= get_bits(&gb, 4); |
248 | |
249 | skip_bits_long(&gb, 64); /* md5 sum */ |
250 | skip_bits_long(&gb, 64); /* md5 sum */ |
251 | |
252 | } |
253 | static void allocate_buffers(FLACContext *s); |
254 | |
255 | int ff_flac_is_extradata_valid(AVCodecContext *avctx, |
256 | enum FLACExtradataFormat *format, |
257 | uint8_t **streaminfo_start) |
258 | { |
259 | if (!avctx->extradata || avctx->extradata_size < FLAC_STREAMINFO_SIZE) { |
260 | audio_codec_print("extradata NULL or too small.\n"); |
261 | return 0; |
262 | } |
263 | if (AV_RL32(avctx->extradata) != MKTAG('f', 'L', 'a', 'C')) { |
264 | /* extradata contains STREAMINFO only */ |
265 | if (avctx->extradata_size != FLAC_STREAMINFO_SIZE) { |
266 | audio_codec_print("extradata contains %d bytes too many.\n", |
267 | FLAC_STREAMINFO_SIZE - avctx->extradata_size); |
268 | } |
269 | *format = FLAC_EXTRADATA_FORMAT_STREAMINFO; |
270 | *streaminfo_start = avctx->extradata; |
271 | } else { |
272 | if (avctx->extradata_size < 8 + FLAC_STREAMINFO_SIZE) { |
273 | audio_codec_print("extradata too small.\n"); |
274 | return 0; |
275 | } |
276 | *format = FLAC_EXTRADATA_FORMAT_FULL_HEADER; |
277 | *streaminfo_start = &avctx->extradata[8]; |
278 | } |
279 | return 1; |
280 | } |
281 | static int decode_residuals(FLACContext *s, int channel, int pred_order) |
282 | { |
283 | int i, tmp, partition, method_type, rice_order; |
284 | int sample = 0, samples; |
285 | |
286 | method_type = get_bits(&s->gb, 2); |
287 | if (method_type > 1) { |
288 | audio_codec_print("illegal residual coding method %d\n", method_type); |
289 | return -1; |
290 | } |
291 | |
292 | rice_order = get_bits(&s->gb, 4); |
293 | |
294 | samples = s->blocksize >> rice_order; |
295 | if (pred_order > samples) { |
296 | audio_codec_print("invalid predictor order: %i > %i\n", pred_order, samples); |
297 | return -1; |
298 | } |
299 | |
300 | sample = |
301 | i = pred_order; |
302 | for (partition = 0; partition < (1 << rice_order); partition++) { |
303 | tmp = get_bits(&s->gb, method_type == 0 ? 4 : 5); |
304 | if (tmp == (method_type == 0 ? 15 : 31)) { |
305 | tmp = get_bits(&s->gb, 5); |
306 | for (; i < samples; i++, sample++) { |
307 | s->decoded[channel][sample] = get_sbits_long(&s->gb, tmp); |
308 | } |
309 | } else { |
310 | for (; i < samples; i++, sample++) { |
311 | s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0); |
312 | } |
313 | } |
314 | i = 0; |
315 | } |
316 | |
317 | return 0; |
318 | } |
319 | |
320 | |
321 | static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order) |
322 | { |
323 | const int blocksize = s->blocksize; |
324 | int32_t *decoded = s->decoded[channel]; |
325 | int av_uninit(a), av_uninit(b), av_uninit(c), av_uninit(d), i; |
326 | |
327 | /* warm up samples */ |
328 | for (i = 0; i < pred_order; i++) { |
329 | decoded[i] = get_sbits_long(&s->gb, s->curr_bps); |
330 | } |
331 | |
332 | if (decode_residuals(s, channel, pred_order) < 0) { |
333 | return -1; |
334 | } |
335 | |
336 | if (pred_order > 0) { |
337 | a = decoded[pred_order - 1]; |
338 | } |
339 | if (pred_order > 1) { |
340 | b = a - decoded[pred_order - 2]; |
341 | } |
342 | if (pred_order > 2) { |
343 | c = b - decoded[pred_order - 2] + decoded[pred_order - 3]; |
344 | } |
345 | if (pred_order > 3) { |
346 | d = c - decoded[pred_order - 2] + 2 * decoded[pred_order - 3] - decoded[pred_order - 4]; |
347 | } |
348 | |
349 | switch (pred_order) { |
350 | case 0: |
351 | break; |
352 | case 1: |
353 | for (i = pred_order; i < blocksize; i++) { |
354 | decoded[i] = a += decoded[i]; |
355 | } |
356 | break; |
357 | case 2: |
358 | for (i = pred_order; i < blocksize; i++) { |
359 | decoded[i] = a += b += decoded[i]; |
360 | } |
361 | break; |
362 | case 3: |
363 | for (i = pred_order; i < blocksize; i++) { |
364 | decoded[i] = a += b += c += decoded[i]; |
365 | } |
366 | break; |
367 | case 4: |
368 | for (i = pred_order; i < blocksize; i++) { |
369 | decoded[i] = a += b += c += d += decoded[i]; |
370 | } |
371 | break; |
372 | default: |
373 | audio_codec_print("illegal pred order %d\n", pred_order); |
374 | return -1; |
375 | } |
376 | |
377 | return 0; |
378 | } |
379 | |
380 | |
381 | static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order) |
382 | { |
383 | int i, j; |
384 | int coeff_prec, qlevel; |
385 | int coeffs[32]; |
386 | int32_t *decoded = s->decoded[channel]; |
387 | |
388 | /* warm up samples */ |
389 | for (i = 0; i < pred_order; i++) { |
390 | decoded[i] = get_sbits_long(&s->gb, s->curr_bps); |
391 | } |
392 | |
393 | coeff_prec = get_bits(&s->gb, 4) + 1; |
394 | if (coeff_prec == 16) { |
395 | audio_codec_print("invalid coeff precision\n"); |
396 | return -1; |
397 | } |
398 | qlevel = get_sbits(&s->gb, 5); |
399 | if (qlevel < 0) { |
400 | audio_codec_print("qlevel %d not supported, maybe buggy stream\n", qlevel); |
401 | return -1; |
402 | } |
403 | |
404 | for (i = 0; i < pred_order; i++) { |
405 | coeffs[i] = get_sbits(&s->gb, coeff_prec); |
406 | } |
407 | |
408 | if (decode_residuals(s, channel, pred_order) < 0) { |
409 | return -1; |
410 | } |
411 | |
412 | if (s->bps > 16) { |
413 | int64_t sum; |
414 | for (i = pred_order; i < s->blocksize; i++) { |
415 | sum = 0; |
416 | for (j = 0; j < pred_order; j++) { |
417 | sum += (int64_t)coeffs[j] * decoded[i - j - 1]; |
418 | } |
419 | decoded[i] += sum >> qlevel; |
420 | } |
421 | } else { |
422 | for (i = pred_order; i < s->blocksize - 1; i += 2) { |
423 | int c; |
424 | int d = decoded[i - pred_order]; |
425 | int s0 = 0, s1 = 0; |
426 | for (j = pred_order - 1; j > 0; j--) { |
427 | c = coeffs[j]; |
428 | s0 += c * d; |
429 | d = decoded[i - j]; |
430 | s1 += c * d; |
431 | } |
432 | c = coeffs[0]; |
433 | s0 += c * d; |
434 | d = decoded[i] += s0 >> qlevel; |
435 | s1 += c * d; |
436 | decoded[i + 1] += s1 >> qlevel; |
437 | } |
438 | if (i < s->blocksize) { |
439 | int sum = 0; |
440 | for (j = 0; j < pred_order; j++) { |
441 | sum += coeffs[j] * decoded[i - j - 1]; |
442 | } |
443 | decoded[i] += sum >> qlevel; |
444 | } |
445 | } |
446 | |
447 | return 0; |
448 | } |
449 | |
450 | void ff_flac_parse_block_header(const uint8_t *block_header, |
451 | int *last, int *type, int *size) |
452 | { |
453 | int tmp = bytestream_get_byte(&block_header); |
454 | if (last) { |
455 | *last = tmp & 0x80; |
456 | } |
457 | if (type) { |
458 | *type = tmp & 0x7F; |
459 | } |
460 | if (size) { |
461 | *size = bytestream_get_be24(&block_header); |
462 | } |
463 | } |
464 | |
465 | static inline int decode_subframe(FLACContext *s, int channel) |
466 | { |
467 | int type, wasted = 0; |
468 | int i, tmp; |
469 | |
470 | s->curr_bps = s->bps; |
471 | if (channel == 0) { |
472 | if (s->ch_mode == FLAC_CHMODE_RIGHT_SIDE) { |
473 | s->curr_bps++; |
474 | } |
475 | } else { |
476 | if (s->ch_mode == FLAC_CHMODE_LEFT_SIDE || s->ch_mode == FLAC_CHMODE_MID_SIDE) { |
477 | s->curr_bps++; |
478 | } |
479 | } |
480 | |
481 | if (get_bits1(&s->gb)) { |
482 | audio_codec_print("invalid subframe padding\n"); |
483 | return -1; |
484 | } |
485 | type = get_bits(&s->gb, 6); |
486 | |
487 | if (get_bits1(&s->gb)) { |
488 | wasted = 1; |
489 | while (!get_bits1(&s->gb)) { |
490 | wasted++; |
491 | } |
492 | s->curr_bps -= wasted; |
493 | } |
494 | if (s->curr_bps > 32) { |
495 | audio_codec_print("decorrelated bit depth > 32"); |
496 | return -1; |
497 | } |
498 | |
499 | //FIXME use av_log2 for types |
500 | if (type == 0) { |
501 | tmp = get_sbits_long(&s->gb, s->curr_bps); |
502 | for (i = 0; i < s->blocksize; i++) { |
503 | s->decoded[channel][i] = tmp; |
504 | } |
505 | } else if (type == 1) { |
506 | for (i = 0; i < s->blocksize; i++) { |
507 | s->decoded[channel][i] = get_sbits_long(&s->gb, s->curr_bps); |
508 | } |
509 | } else if ((type >= 8) && (type <= 12)) { |
510 | if (decode_subframe_fixed(s, channel, type & ~0x8) < 0) { |
511 | return -1; |
512 | } |
513 | } else if (type >= 32) { |
514 | if (decode_subframe_lpc(s, channel, (type & ~0x20) + 1) < 0) { |
515 | return -1; |
516 | } |
517 | } else { |
518 | audio_codec_print("invalid coding type\n"); |
519 | return -1; |
520 | } |
521 | |
522 | if (wasted) { |
523 | int i; |
524 | for (i = 0; i < s->blocksize; i++) { |
525 | s->decoded[channel][i] <<= wasted; |
526 | } |
527 | } |
528 | |
529 | return 0; |
530 | } |
531 | |
532 | static int decode_frame_header(GetBitContext *gb, FLACFrameInfo *fi) |
533 | { |
534 | int bs_code, sr_code, bps_code; |
535 | |
536 | /* frame sync code */ |
537 | //skip_bits(gb, 16); |
538 | if ((get_bits(gb, 15) & 0x7FFF) != 0x7FFC) { |
539 | audio_codec_print("invalid sync code-----------------------\n"); |
540 | return -1; |
541 | } |
542 | skip_bits(gb, 1); |
543 | |
544 | /* block size and sample rate codes */ |
545 | bs_code = get_bits(gb, 4); |
546 | sr_code = get_bits(gb, 4); |
547 | |
548 | /* channels and decorrelation */ |
549 | fi->ch_mode = get_bits(gb, 4); |
550 | if (fi->ch_mode < FLAC_MAX_CHANNELS) { |
551 | fi->channels = fi->ch_mode + 1; |
552 | fi->ch_mode = FLAC_CHMODE_INDEPENDENT; |
553 | } else if (fi->ch_mode <= FLAC_CHMODE_MID_SIDE) { |
554 | fi->channels = 2; |
555 | } else { |
556 | audio_codec_print("invalid channel mode: %d\n", fi->ch_mode); |
557 | return -1; |
558 | } |
559 | |
560 | /* bits per sample */ |
561 | bps_code = get_bits(gb, 3); |
562 | if (bps_code == 3 || bps_code == 7) { |
563 | audio_codec_print("invalid sample size code (%d)\n", bps_code); |
564 | return -1; |
565 | } |
566 | fi->bps = sample_size_table[bps_code]; |
567 | |
568 | /* reserved bit */ |
569 | if (get_bits1(gb)) { |
570 | audio_codec_print("broken stream, invalid padding\n"); |
571 | return -1; |
572 | } |
573 | |
574 | /* sample or frame count */ |
575 | if (get_utf8(gb) < 0) { |
576 | audio_codec_print("utf8 fscked\n"); |
577 | return -1; |
578 | } |
579 | |
580 | /* blocksize */ |
581 | if (bs_code == 0) { |
582 | audio_codec_print("reserved blocksize code: 0\n"); |
583 | return -1; |
584 | } else if (bs_code == 6) { |
585 | fi->blocksize = get_bits(gb, 8) + 1; |
586 | } else if (bs_code == 7) { |
587 | fi->blocksize = get_bits(gb, 16) + 1; |
588 | } else { |
589 | fi->blocksize = ff_flac_blocksize_table[bs_code]; |
590 | } |
591 | |
592 | /* sample rate */ |
593 | if (sr_code < 12) { |
594 | fi->samplerate = ff_flac_sample_rate_table[sr_code]; |
595 | } else if (sr_code == 12) { |
596 | fi->samplerate = get_bits(gb, 8) * 1000; |
597 | } else if (sr_code == 13) { |
598 | fi->samplerate = get_bits(gb, 16); |
599 | } else if (sr_code == 14) { |
600 | fi->samplerate = get_bits(gb, 16) * 10; |
601 | } else { |
602 | audio_codec_print("illegal sample rate code %d\n", sr_code); |
603 | return -1; |
604 | } |
605 | |
606 | /* header CRC-8 check */ |
607 | skip_bits(gb, 8); |
608 | if (av_crc(av_crc_get_table(AV_CRC_8_ATM), 0, gb->buffer, |
609 | get_bits_count(gb) / 8)) { |
610 | audio_codec_print("header crc mismatch\n"); |
611 | return -1; |
612 | } |
613 | |
614 | return 0; |
615 | } |
616 | |
617 | static int decode_frame(FLACContext *s) |
618 | { |
619 | int i; |
620 | GetBitContext *gb = &s->gb; |
621 | FLACFrameInfo fi; |
622 | |
623 | if (decode_frame_header(gb, &fi)) { |
624 | //audio_codec_print("invalid frame header\n"); |
625 | return -1; |
626 | } |
627 | |
628 | if (fi.channels != s->channels) { |
629 | audio_codec_print("switching channel layout mid-stream is not supported\n"); |
630 | return -1; |
631 | } |
632 | s->ch_mode = fi.ch_mode; |
633 | |
634 | if (fi.bps && fi.bps != s->bps) { |
635 | audio_codec_print("switching bps mid-stream is not supported\n"); |
636 | return -1; |
637 | } |
638 | if (0/*s->bps > 16*/) { |
639 | s->avctx->sample_fmt = SAMPLE_FMT_S32; |
640 | s->sample_shift = 32 - s->bps; |
641 | s->is32 = 1; |
642 | } else { |
643 | // s->avctx->sample_fmt = SAMPLE_FMT_S16; |
644 | s->sample_shift = 16 - s->bps; |
645 | s->is32 = 0; |
646 | } |
647 | |
648 | if (fi.blocksize > s->max_blocksize) { |
649 | audio_codec_print("blocksize %d > %d\n", fi.blocksize, s->max_blocksize); |
650 | return -1; |
651 | } |
652 | s->blocksize = fi.blocksize; |
653 | |
654 | if (fi.samplerate == 0) { |
655 | fi.samplerate = s->samplerate; |
656 | } else if (fi.samplerate != s->samplerate) { |
657 | audio_codec_print("sample rate changed from %d to %d\n", s->samplerate, fi.samplerate); |
658 | } |
659 | s->samplerate = s->avctx->sample_rate = fi.samplerate; |
660 | |
661 | |
662 | /* subframes */ |
663 | for (i = 0; i < s->channels; i++) { |
664 | if (decode_subframe(s, i) < 0) { |
665 | return -1; |
666 | } |
667 | } |
668 | |
669 | align_get_bits(gb); |
670 | |
671 | /* frame footer */ |
672 | skip_bits(gb, 16); /* data crc */ |
673 | |
674 | return 0; |
675 | } |
676 | |
677 | static inline int av_clipf1(float a, int amin, int amax) |
678 | { |
679 | if (a < amin) { |
680 | return (int)amin; |
681 | } else if (a > amax) { |
682 | return (int)amax; |
683 | } else { |
684 | return (int)a; |
685 | } |
686 | } |
687 | int audio_dec_decode(audio_decoder_operations_t *adec_ops, char *outbuf, int *outlen, char *inbuf, int inlen) |
688 | { |
689 | AVCodecContext *avctx = &acodec; |
690 | FLACContext *s = &flactext; |
691 | int i, j = 0, input_buf_size = 0, bytes_read = 0; |
692 | int16_t *samples_16 = (int16_t*)outbuf; |
693 | int32_t *samples_32 = (int32_t*)outbuf; |
694 | int output_size = 0; |
695 | unsigned char *buf; |
696 | unsigned buf_size; |
697 | int last_sync_pos = 0; |
698 | |
699 | unsigned int max_framesize = FFMAX(17 * s->max_framesize / 16 + 32, s->max_framesize); |
700 | if (inlen < max_framesize) { |
701 | int inbufindex = 0; |
702 | char *pinbufptr = inbuf; |
703 | char para = *(pinbufptr + 2); |
704 | int frame_num = 0; |
705 | |
706 | inbufindex += 11; |
707 | pinbufptr += 11; |
708 | while (inbufindex < inlen - FLAC_MIN_FRAME_SIZE + 2) { |
709 | if ((AV_RB16(pinbufptr) & 0xFFFF) != 0xFFF8 || AV_RB8(pinbufptr + 2) != para) { |
710 | pinbufptr++; |
711 | inbufindex++; |
712 | } else { |
713 | inlen = inbufindex; |
714 | frame_num++; |
715 | if (frame_num > 1) { |
716 | audio_codec_print("## next frame found, inbufindex=%d, frame_num=%d,-------\n", inbufindex, frame_num); |
717 | goto decodecontinue; |
718 | } |
719 | } |
720 | } |
721 | //audio_codec_print("##inlen=%d, diff=%d,-------------------\n", inlen, pinbufptr-inbuf); |
722 | return -1; |
723 | } |
724 | |
725 | decodecontinue: |
726 | s->bitstream = inbuf; |
727 | s->bitstream_size = inlen; |
728 | s->bitstream_index = 0; |
729 | |
730 | buf = s->bitstream; |
731 | buf_size = s->bitstream_size; |
732 | |
733 | /* check for inline header */ |
734 | if (AV_RB32(buf) == MKBETAG('f', 'L', 'a', 'C')) { |
735 | if (!s->got_streaminfo && parse_streaminfo(s, buf, buf_size)) { |
736 | audio_codec_print("invalid header\n"); |
737 | return -1; |
738 | } |
739 | bytes_read = get_metadata_size(buf, buf_size); |
740 | goto end; |
741 | } |
742 | FIND_SYNC_WORD: |
743 | /* check for frame sync code and resync stream if necessary */ |
744 | if ((AV_RB16(buf) & 0xFFFE) != 0xFFF8) { |
745 | const uint8_t *buf_end = buf + buf_size; |
746 | // av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n"); |
747 | while (buf + 2 < buf_end && (AV_RB16(buf) & 0xFFFE) != 0xFFF8) { |
748 | buf++; |
749 | last_sync_pos++; |
750 | } |
751 | bytes_read = buf_size - (buf_end - buf); |
752 | goto end; // we may not have enough bits left to decode a frame, so try next time |
753 | } |
754 | |
755 | /* decode frame */ |
756 | init_get_bits(&s->gb, buf, buf_size * 8); |
757 | if (decode_frame(s) < 0) { |
758 | audio_codec_print("decode_frame() failed\n"); |
759 | //some times , flac may seek to fake sync word pos, caused decode failed |
760 | //call resync to fix this issue |
761 | buf = s->bitstream + last_sync_pos + 1; |
762 | buf_size = s->bitstream_size - (last_sync_pos + 1); |
763 | last_sync_pos++; |
764 | goto FIND_SYNC_WORD; |
765 | } |
766 | bytes_read = (get_bits_count(&s->gb) + 7) / 8; |
767 | |
768 | /* check if allocated data size is large enough for output */ |
769 | output_size = s->blocksize * (s->channels > 2 ? 2 : s->channels) * (s->is32 ? 4 : 2); |
770 | if (output_size > DefaultOutBufSize) { |
771 | audio_codec_print("output data size is larger than allocated data size\n"); |
772 | goto end; |
773 | } |
774 | |
775 | #define DECORRELATE(left, right)\ |
776 | for (i = 0; i < s->blocksize; i++) {\ |
777 | int a= s->decoded[0][i];\ |
778 | int b= s->decoded[1][i];\ |
779 | if (s->is32) {\ |
780 | *samples_32++ = (left) << s->sample_shift;\ |
781 | *samples_32++ = (right) << s->sample_shift;\ |
782 | } else {\ |
783 | if (s->sample_shift >= 0) \ |
784 | {\ |
785 | *samples_16++ = (left) << s->sample_shift;\ |
786 | *samples_16++ = (right) << s->sample_shift;\ |
787 | }\ |
788 | else\ |
789 | {\ |
790 | *samples_16++ = (left) >> (-1*s->sample_shift);\ |
791 | *samples_16++ = (right) >> (-1*s->sample_shift);\ |
792 | }\ |
793 | }\ |
794 | }\ |
795 | break; |
796 | |
797 | switch (s->ch_mode) { |
798 | case FLAC_CHMODE_INDEPENDENT: |
799 | if (s->channels <= 2) { |
800 | for (j = 0; j < s->blocksize; j++) { |
801 | for (i = 0; i < s->channels; i++) { |
802 | if (s->is32) { |
803 | *samples_32++ = s->decoded[i][j] << s->sample_shift; |
804 | } else { |
805 | if (s->sample_shift >= 0) { |
806 | *samples_16++ = s->decoded[i][j] << s->sample_shift; |
807 | } else { |
808 | *samples_16++ = s->decoded[i][j] >> (-1 * s->sample_shift); |
809 | } |
810 | } |
811 | } |
812 | } |
813 | } else { |
814 | float sum0 = 0, sum1 = 0; |
815 | //downmix here |
816 | //FL + (sum of other surround channels >> 1) |
817 | //FR + (sum of other surround channels >> 1) |
818 | for (j = 0; j < s->blocksize; j++) { |
819 | if (s->is32) { |
820 | sum0 = s->decoded[0][j] << s->sample_shift; |
821 | sum1 = s->decoded[1][j] << s->sample_shift; |
822 | } else { |
823 | if (s->sample_shift >= 0) { |
824 | sum0 = s->decoded[0][j] << s->sample_shift; |
825 | sum1 = s->decoded[1][j] << s->sample_shift; |
826 | } else { |
827 | sum0 = s->decoded[0][j] >> (-1 * s->sample_shift); |
828 | sum1 = s->decoded[1][j] >> (-1 * s->sample_shift); |
829 | } |
830 | } |
831 | for (i = 2; i < s->channels; i++) { |
832 | if (s->is32) { |
833 | sum0 += (s->decoded[i][j] << s->sample_shift) >> 1; |
834 | sum1 += (s->decoded[i][j] << s->sample_shift) >> 1; |
835 | } else { |
836 | if (s->sample_shift >= 0) { |
837 | sum0 += (s->decoded[i][j] << s->sample_shift) >> 1; |
838 | sum1 += (s->decoded[i][j] << s->sample_shift) >> 1; |
839 | } else { |
840 | sum0 += (s->decoded[i][j] >> (-1 * s->sample_shift)) >> 1; |
841 | sum1 += (s->decoded[i][j] >> (-1 * s->sample_shift)) >> 1; |
842 | } |
843 | } |
844 | } |
845 | |
846 | if (s->is32) { |
847 | *samples_32++ = av_clipf1(sum0, -0x80000000, 0x7fffffff); |
848 | *samples_32++ = av_clipf1(sum1, -0x80000000, 0x7fffffff); |
849 | } else { |
850 | *samples_16++ = av_clipf1(sum0, -32768, 32767); |
851 | *samples_16++ = av_clipf1(sum1, -32768, 32767); |
852 | } |
853 | } |
854 | } |
855 | break; |
856 | case FLAC_CHMODE_LEFT_SIDE: |
857 | DECORRELATE(a, a - b) |
858 | case FLAC_CHMODE_RIGHT_SIDE: |
859 | DECORRELATE(a + b, b) |
860 | case FLAC_CHMODE_MID_SIDE: |
861 | DECORRELATE((a -= b >> 1) + b, a) |
862 | } |
863 | |
864 | end: |
865 | if (bytes_read > inlen) { |
866 | // av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", bytes_read - buf_size); |
867 | s->bitstream_size = 0; |
868 | s->bitstream_index = 0; |
869 | return -1; |
870 | } |
871 | |
872 | if (s->bitstream_size) { |
873 | s->bitstream_index += bytes_read; |
874 | s->bitstream_size -= bytes_read; |
875 | } |
876 | |
877 | *outlen = output_size; |
878 | |
879 | return bytes_read; |
880 | } |
881 | |
882 | int audio_dec_init(audio_decoder_operations_t *adec_ops) |
883 | { |
884 | memset(&flactext , 0, sizeof(flactext)); |
885 | memset(&acodec, 0, sizeof(acodec)); |
886 | enum FLACExtradataFormat format; |
887 | uint8_t *streaminfo; |
888 | AVCodecContext *avctx = &acodec; |
889 | FLACContext *s = &flactext; |
890 | s->avctx = &acodec; |
891 | //audio_codec_print("\n\n[%s]BuildDate--%s BuildTime--%s", __FUNCTION__, __DATE__, __TIME__); |
892 | avctx->sample_fmt = SAMPLE_FMT_S16; |
893 | avctx->extradata = adec_ops->extradata; |
894 | avctx->extradata_size = adec_ops->extradata_size; |
895 | |
896 | if (!avctx->extradata_size) { |
897 | return 0; |
898 | } |
899 | if (!ff_flac_is_extradata_valid(avctx, &format, &streaminfo)) { |
900 | return -1; |
901 | } |
902 | |
903 | /* initialize based on the demuxer-supplied streamdata header */ |
904 | ff_flac_parse_streaminfo(avctx, (FLACStreaminfo *)s, streaminfo); |
905 | if (s->bps > 16) { |
906 | avctx->sample_fmt = SAMPLE_FMT_S32; |
907 | } else { |
908 | avctx->sample_fmt = SAMPLE_FMT_S16; |
909 | } |
910 | allocate_buffers(s); |
911 | s->got_streaminfo = 1; |
912 | avctx->channels = (avctx->channels > 2 ? 2 : avctx->channels); |
913 | audio_codec_print("applied flac sr %d,ch num %d\n", avctx->sample_rate, avctx->channels); |
914 | |
915 | adec_ops->nInBufSize = DefaultReadSize; |
916 | adec_ops->nOutBufSize = DefaultOutBufSize; |
917 | audio_codec_print("ape_Init.--------------------------------\n"); |
918 | |
919 | return 0; |
920 | } |
921 | |
922 | int audio_dec_release(audio_decoder_operations_t *adec_ops) |
923 | { |
924 | int i; |
925 | audio_codec_print("audio_dec_release.--------------------------------\n"); |
926 | |
927 | if (outbuffer.outb == NULL) { |
928 | av_freep(&outbuffer.outb); |
929 | } |
930 | |
931 | for (i = 0; i < flactext.channels; i++) { |
932 | av_freep(&flactext.decoded[i]); |
933 | } |
934 | // av_free(flactext.bitstream); |
935 | |
936 | return 0; |
937 | } |
938 | |
939 | int audio_dec_getinfo(audio_decoder_operations_t *adec_ops, void *pAudioInfo) |
940 | { |
941 | return 0; |
942 | } |
943 | |
944 |