API tools faq
paste
Advertisement
Guest User

AAC fix for myth 0.24

a guest
Oct 5th, 2011
172
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 69.44 KB | None | 0 0
raw download clone embed print report
  1. diff --git a/mythtv/configure b/mythtv/configure
  2. index e17edb0..659e8d0 100755
  3. --- a/mythtv/configure
  4. +++ b/mythtv/configure
  5. @@ -1473,7 +1473,7 @@ rdft_select="fft"
  6. # decoders / encoders / hardware accelerators
  7. aac_decoder_select="mdct rdft"
  8. aac_encoder_select="mdct"
  9. -aac_latm_decoder_select="aac_decoder"
  10. +aac_latm_decoder_select="aac_decoder aac_latm_parser"
  11. ac3_decoder_select="mdct ac3_parser"
  12. alac_encoder_select="lpc"
  13. amrnb_decoder_select="lsp"
  14. diff --git a/mythtv/external/FFmpeg/libavcodec/Makefile b/mythtv/external/FFmpeg/libavcodec/Makefile
  15. index 0d6ab83..15a9493 100644
  16. --- a/mythtv/external/FFmpeg/libavcodec/Makefile
  17. +++ b/mythtv/external/FFmpeg/libavcodec/Makefile
  18. @@ -61,7 +61,6 @@ OBJS-$(CONFIG_AAC_ENCODER) += aacenc.o aaccoder.o \
  19. aacpsy.o aactab.o \
  20. psymodel.o iirfilter.o \
  21. mpeg4audio.o
  22. -OBJS-$(CONFIG_AAC_LATM_DECODER) += aaclatmdec.o
  23. OBJS-$(CONFIG_AASC_DECODER) += aasc.o msrledec.o
  24. OBJS-$(CONFIG_AC3_DECODER) += ac3dec.o ac3dec_data.o ac3.o
  25. OBJS-$(CONFIG_AC3_ENCODER) += ac3enc.o ac3tab.o ac3.o
  26. diff --git a/mythtv/external/FFmpeg/libavcodec/aac.h b/mythtv/external/FFmpeg/libavcodec/aac.h
  27. index 94f578f..da6603a 100644
  28. --- a/mythtv/external/FFmpeg/libavcodec/aac.h
  29. +++ b/mythtv/external/FFmpeg/libavcodec/aac.h
  30. @@ -42,6 +42,7 @@
  31. #define MAX_ELEM_ID 16
  32.  
  33. #define TNS_MAX_ORDER 20
  34. +#define MAX_LTP_LONG_SFB 40
  35.  
  36. enum RawDataBlockType {
  37. TYPE_SCE,
  38. @@ -128,6 +129,17 @@ typedef struct {
  39. #define SCALE_MAX_POS 255 ///< scalefactor index maximum value
  40. #define SCALE_MAX_DIFF 60 ///< maximum scalefactor difference allowed by standard
  41. #define SCALE_DIFF_ZERO 60 ///< codebook index corresponding to zero scalefactor indices difference
  42. +#define POW_SF2_ZERO 200 ///< ff_aac_pow2sf_tab index corresponding to pow(2, 0);
  43. +
  44. +/**
  45. + * Long Term Prediction
  46. + */
  47. +typedef struct {
  48. + int8_t present;
  49. + int16_t lag;
  50. + float coef;
  51. + int8_t used[MAX_LTP_LONG_SFB];
  52. +} LongTermPrediction;
  53.  
  54. /**
  55. * Individual Channel Stream
  56. @@ -138,6 +150,7 @@ typedef struct {
  57. uint8_t use_kb_window[2]; ///< If set, use Kaiser-Bessel window, otherwise use a sinus window.
  58. int num_window_groups;
  59. uint8_t group_len[8];
  60. + LongTermPrediction ltp;
  61. const uint16_t *swb_offset; ///< table of offsets to the lowest spectral coefficient of a scalefactor band, sfb, for a particular window
  62. const uint8_t *swb_sizes; ///< table of scalefactor band sizes for a particular window
  63. int num_swb; ///< number of scalefactor window bands
  64. @@ -205,14 +218,15 @@ typedef struct {
  65. IndividualChannelStream ics;
  66. TemporalNoiseShaping tns;
  67. Pulse pulse;
  68. - enum BandType band_type[128]; ///< band types
  69. - int band_type_run_end[120]; ///< band type run end points
  70. - float sf[120]; ///< scalefactors
  71. - int sf_idx[128]; ///< scalefactor indices (used by encoder)
  72. - uint8_t zeroes[128]; ///< band is not coded (used by encoder)
  73. - DECLARE_ALIGNED(16, float, coeffs)[1024]; ///< coefficients for IMDCT
  74. - DECLARE_ALIGNED(16, float, saved)[1024]; ///< overlap
  75. - DECLARE_ALIGNED(16, float, ret)[2048]; ///< PCM output
  76. + enum BandType band_type[128]; ///< band types
  77. + int band_type_run_end[120]; ///< band type run end points
  78. + float sf[120]; ///< scalefactors
  79. + int sf_idx[128]; ///< scalefactor indices (used by encoder)
  80. + uint8_t zeroes[128]; ///< band is not coded (used by encoder)
  81. + DECLARE_ALIGNED(16, float, coeffs)[1024]; ///< coefficients for IMDCT
  82. + DECLARE_ALIGNED(16, float, saved)[1024]; ///< overlap
  83. + DECLARE_ALIGNED(16, float, ret)[2048]; ///< PCM output
  84. + DECLARE_ALIGNED(16, int16_t, ltp_state)[3072]; ///< time signal for LTP
  85. PredictorState predictor_state[MAX_PREDICTORS];
  86. } SingleChannelElement;
  87.  
  88. @@ -251,7 +265,6 @@ typedef struct {
  89. */
  90. ChannelElement *che[4][MAX_ELEM_ID];
  91. ChannelElement *tag_che_map[4][MAX_ELEM_ID];
  92. - uint8_t tags_seen_this_frame[4][MAX_ELEM_ID];
  93. int tags_mapped;
  94. /** @} */
  95.  
  96. @@ -259,7 +272,7 @@ typedef struct {
  97. * @defgroup temporary aligned temporary buffers (We do not want to have these on the stack.)
  98. * @{
  99. */
  100. - DECLARE_ALIGNED(16, float, buf_mdct)[1024];
  101. + DECLARE_ALIGNED(16, float, buf_mdct)[2048];
  102. /** @} */
  103.  
  104. /**
  105. @@ -268,7 +281,9 @@ typedef struct {
  106. */
  107. FFTContext mdct;
  108. FFTContext mdct_small;
  109. + FFTContext mdct_ltp;
  110. DSPContext dsp;
  111. +// FmtConvertContext fmt_conv;
  112. int random_state;
  113. /** @} */
  114.  
  115. @@ -277,7 +292,6 @@ typedef struct {
  116. * @{
  117. */
  118. float *output_data[MAX_CHANNELS]; ///< Points to each element's 'ret' buffer (PCM output).
  119. - float add_bias; ///< offset for dsp.float_to_int16
  120. float sf_scale; ///< Pre-scale for correct IMDCT and dsp.float_to_int16.
  121. int sf_offset; ///< offset into pow2sf_tab as appropriate for dsp.float_to_int16
  122. /** @} */
  123. diff --git a/mythtv/external/FFmpeg/libavcodec/aacdec.c b/mythtv/external/FFmpeg/libavcodec/aacdec.c
  124. index 62aab34..c983141 100644
  125. --- a/mythtv/external/FFmpeg/libavcodec/aacdec.c
  126. +++ b/mythtv/external/FFmpeg/libavcodec/aacdec.c
  127. @@ -3,6 +3,10 @@
  128. * Copyright (c) 2005-2006 Oded Shimon ( ods15 ods15 dyndns org )
  129. * Copyright (c) 2006-2007 Maxim Gavrilov ( maxim.gavrilov gmail com )
  130. *
  131. + * AAC LATM decoder
  132. + * Copyright (c) 2008-2010 Paul Kendall <paul@kcbbs.gen.nz>
  133. + * Copyright (c) 2010 Janne Grunau <janne-ffmpeg@jannau.net>
  134. + *
  135. * This file is part of FFmpeg.
  136. *
  137. * FFmpeg is free software; you can redistribute it and/or
  138. @@ -38,7 +42,7 @@
  139. * Y filterbank - standard
  140. * N (code in SoC repo) filterbank - Scalable Sample Rate
  141. * Y Temporal Noise Shaping
  142. - * N (code in SoC repo) Long Term Prediction
  143. + * Y Long Term Prediction
  144. * Y intensity stereo
  145. * Y channel coupling
  146. * Y frequency domain prediction
  147. @@ -113,28 +117,11 @@ static const char overread_err[] = "Input buffer exhausted before END element fo
  148.  
  149. static ChannelElement *get_che(AACContext *ac, int type, int elem_id)
  150. {
  151. - /* Some buggy encoders appear to set all elem_ids to zero and rely on
  152. - channels always occurring in the same order. This is expressly forbidden
  153. - by the spec but we will try to work around it.
  154. - */
  155. - int err_printed = 0;
  156. - while (ac->tags_seen_this_frame[type][elem_id] && elem_id < MAX_ELEM_ID) {
  157. - if (ac->output_configured < OC_LOCKED && !err_printed) {
  158. - av_log(ac->avctx, AV_LOG_WARNING, "Duplicate channel tag found, attempting to remap.\n");
  159. - err_printed = 1;
  160. - }
  161. - elem_id++;
  162. - }
  163. - if (elem_id == MAX_ELEM_ID)
  164. - return NULL;
  165. - ac->tags_seen_this_frame[type][elem_id] = 1;
  166. -
  167. - if (ac->tag_che_map[type][elem_id]) {
  168. + // For PCE based channel configurations map the channels solely based on tags.
  169. + if (!ac->m4ac.chan_config) {
  170. return ac->tag_che_map[type][elem_id];
  171. }
  172. - if (ac->tags_mapped >= tags_per_config[ac->m4ac.chan_config]) {
  173. - return NULL;
  174. - }
  175. + // For indexed channel configurations map the channels solely based on position.
  176. switch (ac->m4ac.chan_config) {
  177. case 7:
  178. if (ac->tags_mapped == 3 && type == TYPE_CPE) {
  179. @@ -190,9 +177,8 @@ static ChannelElement *get_che(AACContext *ac, int type, int elem_id)
  180. * @return Returns error status. 0 - OK, !0 - error
  181. */
  182. static av_cold int che_configure(AACContext *ac,
  183. - enum ChannelPosition che_pos[4][MAX_ELEM_ID],
  184. - int type, int id,
  185. - int *channels)
  186. + enum ChannelPosition che_pos[4][MAX_ELEM_ID],
  187. + int type, int id, int *channels)
  188. {
  189. if (che_pos[type][id]) {
  190. if (!ac->che[type][id] && !(ac->che[type][id] = av_mallocz(sizeof(ChannelElement))))
  191. @@ -222,9 +208,9 @@ static av_cold int che_configure(AACContext *ac,
  192. * @return Returns error status. 0 - OK, !0 - error
  193. */
  194. static av_cold int output_configure(AACContext *ac,
  195. - enum ChannelPosition che_pos[4][MAX_ELEM_ID],
  196. - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
  197. - int channel_config, enum OCStatus oc_type)
  198. + enum ChannelPosition che_pos[4][MAX_ELEM_ID],
  199. + enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
  200. + int channel_config, enum OCStatus oc_type)
  201. {
  202. AVCodecContext *avctx = ac->avctx;
  203. int i, type, channels = 0, ret;
  204. @@ -241,8 +227,7 @@ static av_cold int output_configure(AACContext *ac,
  205. return ret;
  206. }
  207.  
  208. - memset(ac->tag_che_map, 0, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0]));
  209. - ac->tags_mapped = 0;
  210. + memset(ac->tag_che_map, 0, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0]));
  211.  
  212. avctx->channel_layout = aac_channel_layout[channel_config - 1];
  213. } else {
  214. @@ -263,9 +248,6 @@ static av_cold int output_configure(AACContext *ac,
  215. }
  216.  
  217. memcpy(ac->tag_che_map, ac->che, 4 * MAX_ELEM_ID * sizeof(ac->che[0][0]));
  218. - ac->tags_mapped = 4 * MAX_ELEM_ID;
  219. -
  220. - avctx->channel_layout = 0;
  221. }
  222.  
  223. avctx->channels = channels;
  224. @@ -300,7 +282,8 @@ static void decode_channel_map(enum ChannelPosition *cpe_map,
  225. *
  226. * @return Returns error status. 0 - OK, !0 - error
  227. */
  228. -static int decode_pce(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
  229. +static int decode_pce(AVCodecContext *avctx, MPEG4AudioConfig *m4ac,
  230. + enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
  231. GetBitContext *gb)
  232. {
  233. int num_front, num_side, num_back, num_lfe, num_assoc_data, num_cc, sampling_index;
  234. @@ -309,8 +292,8 @@ static int decode_pce(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_EL
  235. skip_bits(gb, 2); // object_type
  236.  
  237. sampling_index = get_bits(gb, 4);
  238. - if (ac->m4ac.sampling_index != sampling_index)
  239. - av_log(ac->avctx, AV_LOG_WARNING, "Sample rate index in program config element does not match the sample rate index configured by the container.\n");
  240. + if (m4ac->sampling_index != sampling_index)
  241. + av_log(avctx, AV_LOG_WARNING, "Sample rate index in program config element does not match the sample rate index configured by the container.\n");
  242.  
  243. num_front = get_bits(gb, 4);
  244. num_side = get_bits(gb, 4);
  245. @@ -327,6 +310,10 @@ static int decode_pce(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_EL
  246. if (get_bits1(gb))
  247. skip_bits(gb, 3); // mixdown_coeff_index and pseudo_surround
  248.  
  249. + if (get_bits_left(gb) < 4 * (num_front + num_side + num_back + num_lfe + num_assoc_data + num_cc)) {
  250. + av_log(avctx, AV_LOG_ERROR, overread_err);
  251. + return -1;
  252. + }
  253. decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_FRONT, gb, num_front);
  254. decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_SIDE, gb, num_side );
  255. decode_channel_map(new_che_pos[TYPE_CPE], new_che_pos[TYPE_SCE], AAC_CHANNEL_BACK, gb, num_back );
  256. @@ -341,7 +328,7 @@ static int decode_pce(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_EL
  257. /* comment field, first byte is length */
  258. comment_len = get_bits(gb, 8) * 8;
  259. if (get_bits_left(gb) < comment_len) {
  260. - av_log(ac->avctx, AV_LOG_ERROR, overread_err);
  261. + av_log(avctx, AV_LOG_ERROR, overread_err);
  262. return -1;
  263. }
  264. skip_bits_long(gb, comment_len);
  265. @@ -356,12 +343,12 @@ static int decode_pce(AACContext *ac, enum ChannelPosition new_che_pos[4][MAX_EL
  266. *
  267. * @return Returns error status. 0 - OK, !0 - error
  268. */
  269. -static av_cold int set_default_channel_config(AACContext *ac,
  270. - enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
  271. - int channel_config)
  272. +static av_cold int set_default_channel_config(AVCodecContext *avctx,
  273. + enum ChannelPosition new_che_pos[4][MAX_ELEM_ID],
  274. + int channel_config)
  275. {
  276. if (channel_config < 1 || channel_config > 7) {
  277. - av_log(ac->avctx, AV_LOG_ERROR, "invalid default channel configuration (%d)\n",
  278. + av_log(avctx, AV_LOG_ERROR, "invalid default channel configuration (%d)\n",
  279. channel_config);
  280. return -1;
  281. }
  282. @@ -397,16 +384,21 @@ static av_cold int set_default_channel_config(AACContext *ac,
  283. /**
  284. * Decode GA "General Audio" specific configuration; reference: table 4.1.
  285. *
  286. + * @param ac pointer to AACContext, may be null
  287. + * @param avctx pointer to AVCCodecContext, used for logging
  288. + *
  289. * @return Returns error status. 0 - OK, !0 - error
  290. */
  291. -static int decode_ga_specific_config(AACContext *ac, GetBitContext *gb,
  292. +static int decode_ga_specific_config(AACContext *ac, AVCodecContext *avctx,
  293. + GetBitContext *gb,
  294. + MPEG4AudioConfig *m4ac,
  295. int channel_config)
  296. {
  297. enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
  298. int extension_flag, ret;
  299.  
  300. if (get_bits1(gb)) { // frameLengthFlag
  301. - av_log_missing_feature(ac->avctx, "960/120 MDCT window is", 1);
  302. + av_log_missing_feature(avctx, "960/120 MDCT window is", 1);
  303. return -1;
  304. }
  305.  
  306. @@ -414,24 +406,24 @@ static int decode_ga_specific_config(AACContext *ac, GetBitContext *gb,
  307. skip_bits(gb, 14); // coreCoderDelay
  308. extension_flag = get_bits1(gb);
  309.  
  310. - if (ac->m4ac.object_type == AOT_AAC_SCALABLE ||
  311. - ac->m4ac.object_type == AOT_ER_AAC_SCALABLE)
  312. + if (m4ac->object_type == AOT_AAC_SCALABLE ||
  313. + m4ac->object_type == AOT_ER_AAC_SCALABLE)
  314. skip_bits(gb, 3); // layerNr
  315.  
  316. memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
  317. if (channel_config == 0) {
  318. skip_bits(gb, 4); // element_instance_tag
  319. - if ((ret = decode_pce(ac, new_che_pos, gb)))
  320. + if ((ret = decode_pce(avctx, m4ac, new_che_pos, gb)))
  321. return ret;
  322. } else {
  323. - if ((ret = set_default_channel_config(ac, new_che_pos, channel_config)))
  324. + if ((ret = set_default_channel_config(avctx, new_che_pos, channel_config)))
  325. return ret;
  326. }
  327. - if ((ret = output_configure(ac, ac->che_pos, new_che_pos, channel_config, OC_GLOBAL_HDR)))
  328. + if (ac && (ret = output_configure(ac, ac->che_pos, new_che_pos, channel_config, OC_GLOBAL_HDR)))
  329. return ret;
  330.  
  331. if (extension_flag) {
  332. - switch (ac->m4ac.object_type) {
  333. + switch (m4ac->object_type) {
  334. case AOT_ER_BSAC:
  335. skip_bits(gb, 5); // numOfSubFrame
  336. skip_bits(gb, 11); // layer_length
  337. @@ -454,42 +446,58 @@ static int decode_ga_specific_config(AACContext *ac, GetBitContext *gb,
  338. /**
  339. * Decode audio specific configuration; reference: table 1.13.
  340. *
  341. + * @param ac pointer to AACContext, may be null
  342. + * @param avctx pointer to AVCCodecContext, used for logging
  343. + * @param m4ac pointer to MPEG4AudioConfig, used for parsing
  344. * @param data pointer to AVCodecContext extradata
  345. * @param data_size size of AVCCodecContext extradata
  346. *
  347. - * @return Returns error status. 0 - OK, !0 - error
  348. + * @return Returns error status or number of consumed bits. <0 - error
  349. */
  350. -static int decode_audio_specific_config(AACContext *ac, void *data,
  351. - int data_size)
  352. +static int decode_audio_specific_config(AACContext *ac,
  353. + AVCodecContext *avctx,
  354. + MPEG4AudioConfig *m4ac,
  355. + const uint8_t *data, int data_size, int asclen)
  356. {
  357. GetBitContext gb;
  358. int i;
  359.  
  360. + av_log(avctx, AV_LOG_DEBUG, "extradata size %d\n", avctx->extradata_size);
  361. + for (i = 0; i < avctx->extradata_size; i++)
  362. + av_log(avctx, AV_LOG_DEBUG, "%02x ", avctx->extradata[i]);
  363. + av_log(avctx, AV_LOG_DEBUG, "\n");
  364. +
  365. init_get_bits(&gb, data, data_size * 8);
  366.  
  367. - if ((i = ff_mpeg4audio_get_config(&ac->m4ac, data, data_size)) < 0)
  368. + if ((i = ff_mpeg4audio_get_config(m4ac, data, asclen/8)) < 0)
  369. return -1;
  370. - if (ac->m4ac.sampling_index > 12) {
  371. - av_log(ac->avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
  372. + if (m4ac->sampling_index > 12) {
  373. + av_log(avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", m4ac->sampling_index);
  374. return -1;
  375. }
  376. - if (ac->m4ac.sbr == 1 && ac->m4ac.ps == -1)
  377. - ac->m4ac.ps = 1;
  378. + if (m4ac->sbr == 1 && m4ac->ps == -1)
  379. + m4ac->ps = 1;
  380.  
  381. skip_bits_long(&gb, i);
  382.  
  383. - switch (ac->m4ac.object_type) {
  384. + switch (m4ac->object_type) {
  385. case AOT_AAC_MAIN:
  386. case AOT_AAC_LC:
  387. - if (decode_ga_specific_config(ac, &gb, ac->m4ac.chan_config))
  388. + case AOT_AAC_LTP:
  389. + if (decode_ga_specific_config(ac, avctx, &gb, m4ac, m4ac->chan_config))
  390. return -1;
  391. break;
  392. default:
  393. - av_log(ac->avctx, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n",
  394. - ac->m4ac.sbr == 1? "SBR+" : "", ac->m4ac.object_type);
  395. + av_log(avctx, AV_LOG_ERROR, "Audio object type %s%d is not supported.\n",
  396. + m4ac->sbr == 1? "SBR+" : "", m4ac->object_type);
  397. return -1;
  398. }
  399. - return 0;
  400. +
  401. + av_log(avctx, AV_LOG_DEBUG, "AOT %d chan config %d sampling index %d (%d) SBR %d PS %d\n",
  402. + m4ac->object_type, m4ac->chan_config, m4ac->sampling_index,
  403. + m4ac->sample_rate, m4ac->sbr, m4ac->ps);
  404. +
  405. + return get_bits_count(&gb);
  406. }
  407.  
  408. /**
  409. @@ -521,6 +529,22 @@ static void reset_all_predictors(PredictorState *ps)
  410. reset_predict_state(&ps[i]);
  411. }
  412.  
  413. +static int sample_rate_idx (int rate)
  414. +{
  415. + if (92017 <= rate) return 0;
  416. + else if (75132 <= rate) return 1;
  417. + else if (55426 <= rate) return 2;
  418. + else if (46009 <= rate) return 3;
  419. + else if (37566 <= rate) return 4;
  420. + else if (27713 <= rate) return 5;
  421. + else if (23004 <= rate) return 6;
  422. + else if (18783 <= rate) return 7;
  423. + else if (13856 <= rate) return 8;
  424. + else if (11502 <= rate) return 9;
  425. + else if (9391 <= rate) return 10;
  426. + else return 11;
  427. +}
  428. +
  429. static void reset_predictor_group(PredictorState *ps, int group_num)
  430. {
  431. int i;
  432. @@ -542,8 +566,31 @@ static av_cold int aac_decode_init(AVCodecContext *avctx)
  433. ac->m4ac.sample_rate = avctx->sample_rate;
  434.  
  435. if (avctx->extradata_size > 0) {
  436. - if (decode_audio_specific_config(ac, avctx->extradata, avctx->extradata_size))
  437. + if (decode_audio_specific_config(ac, ac->avctx, &ac->m4ac,
  438. + avctx->extradata,
  439. + avctx->extradata_size, 8*avctx->extradata_size) < 0)
  440. return -1;
  441. + } else {
  442. + int sr, i;
  443. + enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
  444. +
  445. + sr = sample_rate_idx(avctx->sample_rate);
  446. + ac->m4ac.sampling_index = sr;
  447. + ac->m4ac.channels = avctx->channels;
  448. + ac->m4ac.sbr = -1;
  449. + ac->m4ac.ps = -1;
  450. +
  451. + for (i = 0; i < FF_ARRAY_ELEMS(ff_mpeg4audio_channels); i++)
  452. + if (ff_mpeg4audio_channels[i] == avctx->channels)
  453. + break;
  454. + if (i == FF_ARRAY_ELEMS(ff_mpeg4audio_channels)) {
  455. + i = 0;
  456. + }
  457. + ac->m4ac.chan_config = i;
  458. +
  459. + if (ac->m4ac.chan_config) {
  460. + set_default_channel_config(avctx, new_che_pos, ac->m4ac.chan_config);
  461. + }
  462. }
  463.  
  464. avctx->sample_fmt = SAMPLE_FMT_S16;
  465. @@ -563,23 +610,10 @@ static av_cold int aac_decode_init(AVCodecContext *avctx)
  466. ff_aac_sbr_init();
  467.  
  468. dsputil_init(&ac->dsp, avctx);
  469. +// ff_fmt_convert_init(&ac->fmt_conv, avctx);
  470.  
  471. ac->random_state = 0x1f2e3d4c;
  472.  
  473. - // -1024 - Compensate wrong IMDCT method.
  474. - // 32768 - Required to scale values to the correct range for the bias method
  475. - // for float to int16 conversion.
  476. -
  477. - if (ac->dsp.float_to_int16_interleave == ff_float_to_int16_interleave_c) {
  478. - ac->add_bias = 385.0f;
  479. - ac->sf_scale = 1. / (-1024. * 32768.);
  480. - ac->sf_offset = 0;
  481. - } else {
  482. - ac->add_bias = 0.0f;
  483. - ac->sf_scale = 1. / -1024.;
  484. - ac->sf_offset = 60;
  485. - }
  486. -
  487. ff_aac_tableinit();
  488.  
  489. INIT_VLC_STATIC(&vlc_scalefactors,7,FF_ARRAY_ELEMS(ff_aac_scalefactor_code),
  490. @@ -587,8 +621,9 @@ static av_cold int aac_decode_init(AVCodecContext *avctx)
  491. ff_aac_scalefactor_code, sizeof(ff_aac_scalefactor_code[0]), sizeof(ff_aac_scalefactor_code[0]),
  492. 352);
  493.  
  494. - ff_mdct_init(&ac->mdct, 11, 1, 1.0);
  495. - ff_mdct_init(&ac->mdct_small, 8, 1, 1.0);
  496. + ff_mdct_init(&ac->mdct, 11, 1, 1.0/1024.0);
  497. + ff_mdct_init(&ac->mdct_small, 8, 1, 1.0/128.0);
  498. + ff_mdct_init(&ac->mdct_ltp, 11, 0, -2.0);
  499. // window initialization
  500. ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
  501. ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
  502. @@ -638,6 +673,20 @@ static int decode_prediction(AACContext *ac, IndividualChannelStream *ics,
  503. }
  504.  
  505. /**
  506. + * Decode Long Term Prediction data; reference: table 4.xx.
  507. + */
  508. +static void decode_ltp(AACContext *ac, LongTermPrediction *ltp,
  509. + GetBitContext *gb, uint8_t max_sfb)
  510. +{
  511. + int sfb;
  512. +
  513. + ltp->lag = get_bits(gb, 11);
  514. + ltp->coef = ltp_coef[get_bits(gb, 3)];
  515. + for (sfb = 0; sfb < FFMIN(max_sfb, MAX_LTP_LONG_SFB); sfb++)
  516. + ltp->used[sfb] = get_bits1(gb);
  517. +}
  518. +
  519. +/**
  520. * Decode Individual Channel Stream info; reference: table 4.6.
  521. *
  522. * @param common_window Channels have independent [0], or shared [1], Individual Channel Stream information.
  523. @@ -691,9 +740,8 @@ static int decode_ics_info(AACContext *ac, IndividualChannelStream *ics,
  524. memset(ics, 0, sizeof(IndividualChannelStream));
  525. return -1;
  526. } else {
  527. - av_log_missing_feature(ac->avctx, "Predictor bit set but LTP is", 1);
  528. - memset(ics, 0, sizeof(IndividualChannelStream));
  529. - return -1;
  530. + if ((ics->ltp.present = get_bits(gb, 1)))
  531. + decode_ltp(ac, &ics->ltp, gb, ics->max_sfb);
  532. }
  533. }
  534. }
  535. @@ -771,9 +819,9 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb,
  536. enum BandType band_type[120],
  537. int band_type_run_end[120])
  538. {
  539. - const int sf_offset = ac->sf_offset + (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE ? 12 : 0);
  540. int g, i, idx = 0;
  541. - int offset[3] = { global_gain, global_gain - 90, 100 };
  542. + int offset[3] = { global_gain, global_gain - 90, 0 };
  543. + int clipped_offset;
  544. int noise_flag = 1;
  545. static const char *sf_str[3] = { "Global gain", "Noise gain", "Intensity stereo position" };
  546. for (g = 0; g < ics->num_window_groups; g++) {
  547. @@ -785,12 +833,15 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb,
  548. } else if ((band_type[idx] == INTENSITY_BT) || (band_type[idx] == INTENSITY_BT2)) {
  549. for (; i < run_end; i++, idx++) {
  550. offset[2] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
  551. - if (offset[2] > 255U) {
  552. - av_log(ac->avctx, AV_LOG_ERROR,
  553. - "%s (%d) out of range.\n", sf_str[2], offset[2]);
  554. - return -1;
  555. + clipped_offset = av_clip(offset[2], -155, 100);
  556. + if (offset[2] != clipped_offset) {
  557. +/* av_log_ask_for_sample(ac->avctx, "Intensity stereo "
  558. + "position clipped (%d -> %d).\nIf you heard an "
  559. + "audible artifact, there may be a bug in the "
  560. + "decoder. ", offset[2], clipped_offset);
  561. +*/
  562. }
  563. - sf[idx] = ff_aac_pow2sf_tab[-offset[2] + 300];
  564. + sf[idx] = ff_aac_pow2sf_tab[-clipped_offset + POW_SF2_ZERO];
  565. }
  566. } else if (band_type[idx] == NOISE_BT) {
  567. for (; i < run_end; i++, idx++) {
  568. @@ -798,12 +849,15 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb,
  569. offset[1] += get_bits(gb, 9) - 256;
  570. else
  571. offset[1] += get_vlc2(gb, vlc_scalefactors.table, 7, 3) - 60;
  572. - if (offset[1] > 255U) {
  573. - av_log(ac->avctx, AV_LOG_ERROR,
  574. - "%s (%d) out of range.\n", sf_str[1], offset[1]);
  575. - return -1;
  576. + clipped_offset = av_clip(offset[1], -100, 155);
  577. + if (offset[1] != clipped_offset) {
  578. +/* av_log_ask_for_sample(ac->avctx, "Noise gain clipped "
  579. + "(%d -> %d).\nIf you heard an audible "
  580. + "artifact, there may be a bug in the decoder. ",
  581. + offset[1], clipped_offset);
  582. +*/
  583. }
  584. - sf[idx] = -ff_aac_pow2sf_tab[offset[1] + sf_offset + 100];
  585. + sf[idx] = -ff_aac_pow2sf_tab[clipped_offset + POW_SF2_ZERO];
  586. }
  587. } else {
  588. for (; i < run_end; i++, idx++) {
  589. @@ -813,7 +867,7 @@ static int decode_scalefactors(AACContext *ac, float sf[120], GetBitContext *gb,
  590. "%s (%d) out of range.\n", sf_str[0], offset[0]);
  591. return -1;
  592. }
  593. - sf[idx] = -ff_aac_pow2sf_tab[ offset[0] + sf_offset];
  594. + sf[idx] = -ff_aac_pow2sf_tab[offset[0] - 100 + POW_SF2_ZERO];
  595. }
  596. }
  597. }
  598. @@ -954,19 +1008,19 @@ static inline float *VMUL4S(float *dst, const float *v, unsigned idx,
  599. union float754 s = { .f = *scale };
  600. union float754 t;
  601.  
  602. - t.i = s.i ^ (sign & 1<<31);
  603. + t.i = s.i ^ (sign & 1U<<31);
  604. *dst++ = v[idx & 3] * t.f;
  605.  
  606. sign <<= nz & 1; nz >>= 1;
  607. - t.i = s.i ^ (sign & 1<<31);
  608. + t.i = s.i ^ (sign & 1U<<31);
  609. *dst++ = v[idx>>2 & 3] * t.f;
  610.  
  611. sign <<= nz & 1; nz >>= 1;
  612. - t.i = s.i ^ (sign & 1<<31);
  613. + t.i = s.i ^ (sign & 1U<<31);
  614. *dst++ = v[idx>>4 & 3] * t.f;
  615.  
  616. sign <<= nz & 1; nz >>= 1;
  617. - t.i = s.i ^ (sign & 1<<31);
  618. + t.i = s.i ^ (sign & 1U<<31);
  619. *dst++ = v[idx>>6 & 3] * t.f;
  620.  
  621. return dst;
  622. @@ -1063,9 +1117,6 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024],
  623.  
  624. UPDATE_CACHE(re, gb);
  625. GET_VLC(code, re, gb, vlc_tab, 8, 2);
  626. -#if MIN_CACHE_BITS < 20
  627. - UPDATE_CACHE(re, gb);
  628. -#endif
  629. cb_idx = cb_vector_idx[code];
  630. nnz = cb_idx >> 8 & 15;
  631. bits = SHOW_UBITS(re, gb, nnz) << (32-nnz);
  632. @@ -1158,20 +1209,15 @@ static int decode_spectrum_and_dequant(AACContext *ac, float coef[1024],
  633. return -1;
  634. }
  635.  
  636. -#if MIN_CACHE_BITS < 21
  637. - LAST_SKIP_BITS(re, gb, b + 1);
  638. - UPDATE_CACHE(re, gb);
  639. -#else
  640. SKIP_BITS(re, gb, b + 1);
  641. -#endif
  642. b += 4;
  643. n = (1 << b) + SHOW_UBITS(re, gb, b);
  644. LAST_SKIP_BITS(re, gb, b);
  645. - *icf++ = cbrt_tab[n] | (bits & 1<<31);
  646. + *icf++ = cbrt_tab[n] | (bits & 1U<<31);
  647. bits <<= 1;
  648. } else {
  649. unsigned v = ((const uint32_t*)vq)[cb_idx & 15];
  650. - *icf++ = (bits & 1<<31) | v;
  651. + *icf++ = (bits & 1U<<31) | v;
  652. bits <<= !!v;
  653. }
  654. cb_idx >>= 4;
  655. @@ -1232,8 +1278,7 @@ static av_always_inline float flt16_trunc(float pf)
  656. }
  657.  
  658. static av_always_inline void predict(PredictorState *ps, float *coef,
  659. - float sf_scale, float inv_sf_scale,
  660. - int output_enable)
  661. + int output_enable)
  662. {
  663. const float a = 0.953125; // 61.0 / 64
  664. const float alpha = 0.90625; // 29.0 / 32
  665. @@ -1249,9 +1294,9 @@ static av_always_inline void predict(PredictorState *ps, float *coef,
  666.  
  667. pv = flt16_round(k1 * r0 + k2 * r1);
  668. if (output_enable)
  669. - *coef += pv * sf_scale;
  670. + *coef += pv;
  671.  
  672. - e0 = *coef * inv_sf_scale;
  673. + e0 = *coef;
  674. e1 = e0 - k1 * r0;
  675.  
  676. ps->cor1 = flt16_trunc(alpha * cor1 + r1 * e1);
  677. @@ -1269,7 +1314,6 @@ static av_always_inline void predict(PredictorState *ps, float *coef,
  678. static void apply_prediction(AACContext *ac, SingleChannelElement *sce)
  679. {
  680. int sfb, k;
  681. - float sf_scale = ac->sf_scale, inv_sf_scale = 1 / ac->sf_scale;
  682.  
  683. if (!sce->ics.predictor_initialized) {
  684. reset_all_predictors(sce->predictor_state);
  685. @@ -1280,7 +1324,6 @@ static void apply_prediction(AACContext *ac, SingleChannelElement *sce)
  686. for (sfb = 0; sfb < ff_aac_pred_sfb_max[ac->m4ac.sampling_index]; sfb++) {
  687. for (k = sce->ics.swb_offset[sfb]; k < sce->ics.swb_offset[sfb + 1]; k++) {
  688. predict(&sce->predictor_state[k], &sce->coeffs[k],
  689. - sf_scale, inv_sf_scale,
  690. sce->ics.predictor_present && sce->ics.prediction_used[sfb]);
  691. }
  692. }
  693. @@ -1386,13 +1429,13 @@ static void apply_mid_side_stereo(AACContext *ac, ChannelElement *cpe)
  694. * [1] mask is decoded from bitstream; [2] mask is all 1s;
  695. * [3] reserved for scalable AAC
  696. */
  697. -static void apply_intensity_stereo(ChannelElement *cpe, int ms_present)
  698. +static void apply_intensity_stereo(AACContext *ac, ChannelElement *cpe, int ms_present)
  699. {
  700. const IndividualChannelStream *ics = &cpe->ch[1].ics;
  701. SingleChannelElement *sce1 = &cpe->ch[1];
  702. float *coef0 = cpe->ch[0].coeffs, *coef1 = cpe->ch[1].coeffs;
  703. const uint16_t *offsets = ics->swb_offset;
  704. - int g, group, i, k, idx = 0;
  705. + int g, group, i, idx = 0;
  706. int c;
  707. float scale;
  708. for (g = 0; g < ics->num_window_groups; g++) {
  709. @@ -1405,8 +1448,10 @@ static void apply_intensity_stereo(ChannelElement *cpe, int ms_present)
  710. c *= 1 - 2 * cpe->ms_mask[idx];
  711. scale = c * sce1->sf[idx];
  712. for (group = 0; group < ics->group_len[g]; group++)
  713. - for (k = offsets[i]; k < offsets[i + 1]; k++)
  714. - coef1[group * 128 + k] = scale * coef0[group * 128 + k];
  715. + ac->dsp.vector_fmul_scalar(coef1 + group * 128 + offsets[i],
  716. + coef0 + group * 128 + offsets[i],
  717. + scale,
  718. + offsets[i + 1] - offsets[i]);
  719. }
  720. } else {
  721. int bt_run_end = sce1->band_type_run_end[idx];
  722. @@ -1435,6 +1480,9 @@ static int decode_cpe(AACContext *ac, GetBitContext *gb, ChannelElement *cpe)
  723. i = cpe->ch[1].ics.use_kb_window[0];
  724. cpe->ch[1].ics = cpe->ch[0].ics;
  725. cpe->ch[1].ics.use_kb_window[1] = i;
  726. + if (cpe->ch[1].ics.predictor_present && (ac->m4ac.object_type != AOT_AAC_MAIN))
  727. + if ((cpe->ch[1].ics.ltp.present = get_bits(gb, 1)))
  728. + decode_ltp(ac, &cpe->ch[1].ics.ltp, gb, cpe->ch[1].ics.max_sfb);
  729. ms_present = get_bits(gb, 2);
  730. if (ms_present == 3) {
  731. av_log(ac->avctx, AV_LOG_ERROR, "ms_present = 3 is reserved.\n");
  732. @@ -1456,7 +1504,7 @@ static int decode_cpe(AACContext *ac, GetBitContext *gb, ChannelElement *cpe)
  733. }
  734. }
  735.  
  736. - apply_intensity_stereo(cpe, ms_present);
  737. + apply_intensity_stereo(ac, cpe, ms_present);
  738. return 0;
  739. }
  740.  
  741. @@ -1674,6 +1722,7 @@ static void apply_tns(float coef[1024], TemporalNoiseShaping *tns,
  742. int w, filt, m, i;
  743. int bottom, top, order, start, end, size, inc;
  744. float lpc[TNS_MAX_ORDER];
  745. + float tmp[TNS_MAX_ORDER];
  746.  
  747. for (w = 0; w < ics->num_windows; w++) {
  748. bottom = ics->num_swb;
  749. @@ -1699,18 +1748,129 @@ static void apply_tns(float coef[1024], TemporalNoiseShaping *tns,
  750. }
  751. start += w * 128;
  752.  
  753. - // ar filter
  754. - for (m = 0; m < size; m++, start += inc)
  755. - for (i = 1; i <= FFMIN(m, order); i++)
  756. - coef[start] -= coef[start - i * inc] * lpc[i - 1];
  757. + if (decode) {
  758. + // ar filter
  759. + for (m = 0; m < size; m++, start += inc)
  760. + for (i = 1; i <= FFMIN(m, order); i++)
  761. + coef[start] -= coef[start - i * inc] * lpc[i - 1];
  762. + } else {
  763. + // ma filter
  764. + for (m = 0; m < size; m++, start += inc) {
  765. + tmp[0] = coef[start];
  766. + for (i = 1; i <= FFMIN(m, order); i++)
  767. + coef[start] += tmp[i] * lpc[i - 1];
  768. + for (i = order; i > 0; i--)
  769. + tmp[i] = tmp[i - 1];
  770. + }
  771. + }
  772. }
  773. }
  774. }
  775.  
  776. +static void vector_fmul(float *dst, const float *src0, const float *src1, int len){
  777. + int i;
  778. + for(i=0; i<len; i++)
  779. + dst[i] = src0[i] * src1[i];
  780. +}
  781. +
  782. +/**
  783. + * Apply windowing and MDCT to obtain the spectral
  784. + * coefficient from the predicted sample by LTP.
  785. + */
  786. +static void windowing_and_mdct_ltp(AACContext *ac, float *out,
  787. + float *in, IndividualChannelStream *ics)
  788. +{
  789. + const float *lwindow = ics->use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
  790. + const float *swindow = ics->use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
  791. + const float *lwindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_long_1024 : ff_sine_1024;
  792. + const float *swindow_prev = ics->use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
  793. +
  794. + if (ics->window_sequence[0] != LONG_STOP_SEQUENCE) {
  795. + vector_fmul(in, in, lwindow_prev, 1024);
  796. + } else {
  797. + memset(in, 0, 448 * sizeof(float));
  798. + vector_fmul(in + 448, in + 448, swindow_prev, 128);
  799. + }
  800. + if (ics->window_sequence[0] != LONG_START_SEQUENCE) {
  801. + ac->dsp.vector_fmul_reverse(in + 1024, in + 1024, lwindow, 1024);
  802. + } else {
  803. + ac->dsp.vector_fmul_reverse(in + 1024 + 448, in + 1024 + 448, swindow, 128);
  804. + memset(in + 1024 + 576, 0, 448 * sizeof(float));
  805. + }
  806. + ac->mdct_ltp.mdct_calc(&ac->mdct_ltp, out, in);
  807. +}
  808. +
  809. +/**
  810. + * Apply the long term prediction
  811. + */
  812. +static void apply_ltp(AACContext *ac, SingleChannelElement *sce)
  813. +{
  814. + const LongTermPrediction *ltp = &sce->ics.ltp;
  815. + const uint16_t *offsets = sce->ics.swb_offset;
  816. + int i, sfb;
  817. +
  818. + if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
  819. + float *predTime = sce->ret;
  820. + float *predFreq = ac->buf_mdct;
  821. + int16_t num_samples = 2048;
  822. +
  823. + if (ltp->lag < 1024)
  824. + num_samples = ltp->lag + 1024;
  825. + for (i = 0; i < num_samples; i++)
  826. + predTime[i] = sce->ltp_state[i + 2048 - ltp->lag] * ltp->coef;
  827. + memset(&predTime[i], 0, (2048 - i) * sizeof(float));
  828. +
  829. + windowing_and_mdct_ltp(ac, predFreq, predTime, &sce->ics);
  830. +
  831. + if (sce->tns.present)
  832. + apply_tns(predFreq, &sce->tns, &sce->ics, 0);
  833. +
  834. + for (sfb = 0; sfb < FFMIN(sce->ics.max_sfb, MAX_LTP_LONG_SFB); sfb++)
  835. + if (ltp->used[sfb])
  836. + for (i = offsets[sfb]; i < offsets[sfb + 1]; i++)
  837. + sce->coeffs[i] += predFreq[i];
  838. + }
  839. +}
  840. +
  841. +/**
  842. + * Update the LTP buffer for next frame
  843. + */
  844. +static void update_ltp(AACContext *ac, SingleChannelElement *sce)
  845. +{
  846. + IndividualChannelStream *ics = &sce->ics;
  847. + float *saved = sce->saved;
  848. + float *saved_ltp = sce->coeffs;
  849. + const float *lwindow = ics->use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
  850. + const float *swindow = ics->use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
  851. + int i;
  852. +
  853. + if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
  854. + memcpy(saved_ltp, saved, 512 * sizeof(float));
  855. + memset(saved_ltp + 576, 0, 448 * sizeof(float));
  856. + ac->dsp.vector_fmul_reverse(saved_ltp + 448, ac->buf_mdct + 960, &swindow[64], 64);
  857. + for (i = 0; i < 64; i++)
  858. + saved_ltp[i + 512] = ac->buf_mdct[1023 - i] * swindow[63 - i];
  859. + } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
  860. + memcpy(saved_ltp, ac->buf_mdct + 512, 448 * sizeof(float));
  861. + memset(saved_ltp + 576, 0, 448 * sizeof(float));
  862. + ac->dsp.vector_fmul_reverse(saved_ltp + 448, ac->buf_mdct + 960, &swindow[64], 64);
  863. + for (i = 0; i < 64; i++)
  864. + saved_ltp[i + 512] = ac->buf_mdct[1023 - i] * swindow[63 - i];
  865. + } else { // LONG_STOP or ONLY_LONG
  866. + ac->dsp.vector_fmul_reverse(saved_ltp, ac->buf_mdct + 512, &lwindow[512], 512);
  867. + for (i = 0; i < 512; i++)
  868. + saved_ltp[i + 512] = ac->buf_mdct[1023 - i] * lwindow[511 - i];
  869. + }
  870. +
  871. + memcpy(sce->ltp_state, sce->ltp_state+1024, 1024 * sizeof(*sce->ltp_state));
  872. + memcpy(sce->ltp_state+1024, sce->ret, 1024 * sizeof(*sce->ltp_state));
  873. + memcpy(sce->ltp_state+2048, saved_ltp, 1024 * sizeof(*sce->ltp_state));
  874. +}
  875. +
  876. /**
  877. * Conduct IMDCT and windowing.
  878. */
  879. -static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce, float bias)
  880. +static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce)
  881. {
  882. IndividualChannelStream *ics = &sce->ics;
  883. float *in = sce->coeffs;
  884. @@ -1726,9 +1886,9 @@ static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce, float
  885. // imdct
  886. if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
  887. for (i = 0; i < 1024; i += 128)
  888. - ff_imdct_half(&ac->mdct_small, buf + i, in + i);
  889. + ac->mdct_small.imdct_half(&ac->mdct_small, buf + i, in + i);
  890. } else
  891. - ff_imdct_half(&ac->mdct, buf, in);
  892. + ac->mdct.imdct_half(&ac->mdct, buf, in);
  893.  
  894. /* window overlapping
  895. * NOTE: To simplify the overlapping code, all 'meaningless' short to long
  896. @@ -1738,32 +1898,29 @@ static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce, float
  897. */
  898. if ((ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE) &&
  899. (ics->window_sequence[0] == ONLY_LONG_SEQUENCE || ics->window_sequence[0] == LONG_START_SEQUENCE)) {
  900. - ac->dsp.vector_fmul_window( out, saved, buf, lwindow_prev, bias, 512);
  901. + ac->dsp.vector_fmul_window( out, saved, buf, lwindow_prev, 0.0f, 512);
  902. } else {
  903. - for (i = 0; i < 448; i++)
  904. - out[i] = saved[i] + bias;
  905. + memcpy( out, saved, 448 * sizeof(float));
  906.  
  907. if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
  908. - ac->dsp.vector_fmul_window(out + 448 + 0*128, saved + 448, buf + 0*128, swindow_prev, bias, 64);
  909. - ac->dsp.vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow, bias, 64);
  910. - ac->dsp.vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow, bias, 64);
  911. - ac->dsp.vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow, bias, 64);
  912. - ac->dsp.vector_fmul_window(temp, buf + 3*128 + 64, buf + 4*128, swindow, bias, 64);
  913. + ac->dsp.vector_fmul_window(out + 448 + 0*128, saved + 448, buf + 0*128, swindow_prev, 0.0f, 64);
  914. + ac->dsp.vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow, 0.0f, 64);
  915. + ac->dsp.vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow, 0.0f, 64);
  916. + ac->dsp.vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow, 0.0f, 64);
  917. + ac->dsp.vector_fmul_window(temp, buf + 3*128 + 64, buf + 4*128, swindow, 0.0f, 64);
  918. memcpy( out + 448 + 4*128, temp, 64 * sizeof(float));
  919. } else {
  920. - ac->dsp.vector_fmul_window(out + 448, saved + 448, buf, swindow_prev, bias, 64);
  921. - for (i = 576; i < 1024; i++)
  922. - out[i] = buf[i-512] + bias;
  923. + ac->dsp.vector_fmul_window(out + 448, saved + 448, buf, swindow_prev, 0.0f, 64);
  924. + memcpy( out + 576, buf + 64, 448 * sizeof(float));
  925. }
  926. }
  927.  
  928. // buffer update
  929. if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
  930. - for (i = 0; i < 64; i++)
  931. - saved[i] = temp[64 + i] - bias;
  932. - ac->dsp.vector_fmul_window(saved + 64, buf + 4*128 + 64, buf + 5*128, swindow, 0, 64);
  933. - ac->dsp.vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 0, 64);
  934. - ac->dsp.vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, swindow, 0, 64);
  935. + memcpy( saved, temp + 64, 64 * sizeof(float));
  936. + ac->dsp.vector_fmul_window(saved + 64, buf + 4*128 + 64, buf + 5*128, swindow, 0.0f, 64);
  937. + ac->dsp.vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 0.0f, 64);
  938. + ac->dsp.vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, swindow, 0.0f, 64);
  939. memcpy( saved + 448, buf + 7*128 + 64, 64 * sizeof(float));
  940. } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
  941. memcpy( saved, buf + 512, 448 * sizeof(float));
  942. @@ -1820,13 +1977,12 @@ static void apply_independent_coupling(AACContext *ac,
  943. {
  944. int i;
  945. const float gain = cce->coup.gain[index][0];
  946. - const float bias = ac->add_bias;
  947. const float *src = cce->ch[0].ret;
  948. float *dest = target->ret;
  949. const int len = 1024 << (ac->m4ac.sbr == 1);
  950.  
  951. for (i = 0; i < len; i++)
  952. - dest[i] += gain * (src[i] - bias);
  953. + dest[i] += gain * src[i];
  954. }
  955.  
  956. /**
  957. @@ -1870,13 +2026,20 @@ static void apply_channel_coupling(AACContext *ac, ChannelElement *cc,
  958. static void spectral_to_sample(AACContext *ac)
  959. {
  960. int i, type;
  961. - float imdct_bias = (ac->m4ac.sbr <= 0) ? ac->add_bias : 0.0f;
  962. for (type = 3; type >= 0; type--) {
  963. for (i = 0; i < MAX_ELEM_ID; i++) {
  964. ChannelElement *che = ac->che[type][i];
  965. if (che) {
  966. if (type <= TYPE_CPE)
  967. apply_channel_coupling(ac, che, type, i, BEFORE_TNS, apply_dependent_coupling);
  968. + if (ac->m4ac.object_type == AOT_AAC_LTP) {
  969. + if (che->ch[0].ics.predictor_present) {
  970. + if (che->ch[0].ics.ltp.present)
  971. + apply_ltp(ac, &che->ch[0]);
  972. + if (che->ch[1].ics.ltp.present && type == TYPE_CPE)
  973. + apply_ltp(ac, &che->ch[1]);
  974. + }
  975. + }
  976. if (che->ch[0].tns.present)
  977. apply_tns(che->ch[0].coeffs, &che->ch[0].tns, &che->ch[0].ics, 1);
  978. if (che->ch[1].tns.present)
  979. @@ -1884,9 +2047,13 @@ static void spectral_to_sample(AACContext *ac)
  980. if (type <= TYPE_CPE)
  981. apply_channel_coupling(ac, che, type, i, BETWEEN_TNS_AND_IMDCT, apply_dependent_coupling);
  982. if (type != TYPE_CCE || che->coup.coupling_point == AFTER_IMDCT) {
  983. - imdct_and_windowing(ac, &che->ch[0], imdct_bias);
  984. + imdct_and_windowing(ac, &che->ch[0]);
  985. + if (ac->m4ac.object_type == AOT_AAC_LTP)
  986. + update_ltp(ac, &che->ch[0]);
  987. if (type == TYPE_CPE) {
  988. - imdct_and_windowing(ac, &che->ch[1], imdct_bias);
  989. + imdct_and_windowing(ac, &che->ch[1]);
  990. + if (ac->m4ac.object_type == AOT_AAC_LTP)
  991. + update_ltp(ac, &che->ch[1]);
  992. }
  993. if (ac->m4ac.sbr > 0) {
  994. ff_sbr_apply(ac, &che->sbr, type, che->ch[0].ret, che->ch[1].ret);
  995. @@ -1906,24 +2073,25 @@ static int parse_adts_frame_header(AACContext *ac, GetBitContext *gb)
  996.  
  997. size = ff_aac_parse_header(gb, &hdr_info);
  998. if (size > 0) {
  999. - if (ac->output_configured != OC_LOCKED && hdr_info.chan_config) {
  1000. + if (hdr_info.chan_config) {
  1001. enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
  1002. memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
  1003. ac->m4ac.chan_config = hdr_info.chan_config;
  1004. - if (set_default_channel_config(ac, new_che_pos, hdr_info.chan_config))
  1005. + if (set_default_channel_config(ac->avctx, new_che_pos, hdr_info.chan_config))
  1006. return -7;
  1007. if (output_configure(ac, ac->che_pos, new_che_pos, hdr_info.chan_config, OC_TRIAL_FRAME))
  1008. return -7;
  1009. } else if (ac->output_configured != OC_LOCKED) {
  1010. + ac->m4ac.chan_config = 0;
  1011. ac->output_configured = OC_NONE;
  1012. }
  1013. if (ac->output_configured != OC_LOCKED) {
  1014. ac->m4ac.sbr = -1;
  1015. ac->m4ac.ps = -1;
  1016. + ac->m4ac.sample_rate = hdr_info.sample_rate;
  1017. + ac->m4ac.sampling_index = hdr_info.sampling_index;
  1018. + ac->m4ac.object_type = hdr_info.object_type;
  1019. }
  1020. - ac->m4ac.sample_rate = hdr_info.sample_rate;
  1021. - ac->m4ac.sampling_index = hdr_info.sampling_index;
  1022. - ac->m4ac.object_type = hdr_info.object_type;
  1023. if (!ac->avctx->sample_rate)
  1024. ac->avctx->sample_rate = hdr_info.sample_rate;
  1025. if (hdr_info.num_aac_frames == 1) {
  1026. @@ -1937,24 +2105,17 @@ static int parse_adts_frame_header(AACContext *ac, GetBitContext *gb)
  1027. return size;
  1028. }
  1029.  
  1030. -static int aac_decode_frame(AVCodecContext *avctx, void *data,
  1031. - int *data_size, AVPacket *avpkt)
  1032. +static int aac_decode_frame_int(AVCodecContext *avctx, void *data,
  1033. + int *data_size, GetBitContext *gb)
  1034. {
  1035. - const uint8_t *buf = avpkt->data;
  1036. - int buf_size = avpkt->size;
  1037. AACContext *ac = avctx->priv_data;
  1038. ChannelElement *che = NULL, *che_prev = NULL;
  1039. - GetBitContext gb;
  1040. enum RawDataBlockType elem_type, elem_type_prev = TYPE_END;
  1041. int err, elem_id, data_size_tmp;
  1042. - int buf_consumed;
  1043. - int samples = 0, multiplier;
  1044. - int buf_offset;
  1045. + int samples = 0, multiplier, audio_found = 0;
  1046.  
  1047. - init_get_bits(&gb, buf, buf_size * 8);
  1048. -
  1049. - if (show_bits(&gb, 12) == 0xfff) {
  1050. - if (parse_adts_frame_header(ac, &gb) < 0) {
  1051. + if (show_bits(gb, 12) == 0xfff) {
  1052. + if (parse_adts_frame_header(ac, gb) < 0) {
  1053. av_log(avctx, AV_LOG_ERROR, "Error decoding AAC frame header.\n");
  1054. return -1;
  1055. }
  1056. @@ -1964,10 +2125,10 @@ static int aac_decode_frame(AVCodecContext *avctx, void *data,
  1057. }
  1058. }
  1059.  
  1060. - memset(ac->tags_seen_this_frame, 0, sizeof(ac->tags_seen_this_frame));
  1061. + ac->tags_mapped = 0;
  1062. // parse
  1063. - while ((elem_type = get_bits(&gb, 3)) != TYPE_END) {
  1064. - elem_id = get_bits(&gb, 4);
  1065. + while ((elem_type = get_bits(gb, 3)) != TYPE_END) {
  1066. + elem_id = get_bits(gb, 4);
  1067.  
  1068. if (elem_type < TYPE_DSE) {
  1069. if (!(che=get_che(ac, elem_type, elem_id))) {
  1070. @@ -1981,29 +2142,32 @@ static int aac_decode_frame(AVCodecContext *avctx, void *data,
  1071. switch (elem_type) {
  1072.  
  1073. case TYPE_SCE:
  1074. - err = decode_ics(ac, &che->ch[0], &gb, 0, 0);
  1075. + err = decode_ics(ac, &che->ch[0], gb, 0, 0);
  1076. + audio_found = 1;
  1077. break;
  1078.  
  1079. case TYPE_CPE:
  1080. - err = decode_cpe(ac, &gb, che);
  1081. + err = decode_cpe(ac, gb, che);
  1082. + audio_found = 1;
  1083. break;
  1084.  
  1085. case TYPE_CCE:
  1086. - err = decode_cce(ac, &gb, che);
  1087. + err = decode_cce(ac, gb, che);
  1088. break;
  1089.  
  1090. case TYPE_LFE:
  1091. - err = decode_ics(ac, &che->ch[0], &gb, 0, 0);
  1092. + err = decode_ics(ac, &che->ch[0], gb, 0, 0);
  1093. + audio_found = 1;
  1094. break;
  1095.  
  1096. case TYPE_DSE:
  1097. - err = skip_data_stream_element(ac, &gb);
  1098. + err = skip_data_stream_element(ac, gb);
  1099. break;
  1100.  
  1101. case TYPE_PCE: {
  1102. enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
  1103. memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
  1104. - if ((err = decode_pce(ac, new_che_pos, &gb)))
  1105. + if ((err = decode_pce(avctx, &ac->m4ac, new_che_pos, gb)))
  1106. break;
  1107. if (ac->output_configured > OC_TRIAL_PCE)
  1108. av_log(avctx, AV_LOG_ERROR,
  1109. @@ -2015,13 +2179,13 @@ static int aac_decode_frame(AVCodecContext *avctx, void *data,
  1110.  
  1111. case TYPE_FIL:
  1112. if (elem_id == 15)
  1113. - elem_id += get_bits(&gb, 8) - 1;
  1114. - if (get_bits_left(&gb) < 8 * elem_id) {
  1115. + elem_id += get_bits(gb, 8) - 1;
  1116. + if (get_bits_left(gb) < 8 * elem_id) {
  1117. av_log(avctx, AV_LOG_ERROR, overread_err);
  1118. return -1;
  1119. }
  1120. while (elem_id > 0)
  1121. - elem_id -= decode_extension_payload(ac, &gb, elem_id, che_prev, elem_type_prev);
  1122. + elem_id -= decode_extension_payload(ac, gb, elem_id, che_prev, elem_type_prev);
  1123. err = 0; /* FIXME */
  1124. break;
  1125.  
  1126. @@ -2036,7 +2200,7 @@ static int aac_decode_frame(AVCodecContext *avctx, void *data,
  1127. if (err)
  1128. return err;
  1129.  
  1130. - if (get_bits_left(&gb) < 3) {
  1131. + if (get_bits_left(gb) < 3) {
  1132. av_log(avctx, AV_LOG_ERROR, overread_err);
  1133. return -1;
  1134. }
  1135. @@ -2063,9 +2227,30 @@ static int aac_decode_frame(AVCodecContext *avctx, void *data,
  1136. if (samples)
  1137. ac->dsp.float_to_int16_interleave(data, (const float **)ac->output_data, samples, avctx->channels);
  1138.  
  1139. - if (ac->output_configured)
  1140. +// if (samples)
  1141. +// ac->fmt_conv.float_to_int16_interleave(data, (const float **)ac->output_data, samples, avctx->channels);
  1142. +
  1143. + if (ac->output_configured && audio_found)
  1144. ac->output_configured = OC_LOCKED;
  1145.  
  1146. + return 0;
  1147. +}
  1148. +
  1149. +static int aac_decode_frame(AVCodecContext *avctx, void *data,
  1150. + int *data_size, AVPacket *avpkt)
  1151. +{
  1152. + const uint8_t *buf = avpkt->data;
  1153. + int buf_size = avpkt->size;
  1154. + GetBitContext gb;
  1155. + int buf_consumed;
  1156. + int buf_offset;
  1157. + int err;
  1158. +
  1159. + init_get_bits(&gb, buf, buf_size * 8);
  1160. +
  1161. + if ((err = aac_decode_frame_int(avctx, data, data_size, &gb)) < 0)
  1162. + return err;
  1163. +
  1164. buf_consumed = (get_bits_count(&gb) + 7) >> 3;
  1165. for (buf_offset = buf_consumed; buf_offset < buf_size; buf_offset++)
  1166. if (buf[buf_offset])
  1167. @@ -2089,9 +2274,270 @@ static av_cold int aac_decode_close(AVCodecContext *avctx)
  1168.  
  1169. ff_mdct_end(&ac->mdct);
  1170. ff_mdct_end(&ac->mdct_small);
  1171. + ff_mdct_end(&ac->mdct_ltp);
  1172. return 0;
  1173. }
  1174.  
  1175. +
  1176. +#define LOAS_SYNC_WORD 0x2b7 ///< 11 bits LOAS sync word
  1177. +
  1178. +struct LATMContext {
  1179. + AACContext aac_ctx; ///< containing AACContext
  1180. + int initialized; ///< initilized after a valid extradata was seen
  1181. +
  1182. + // parser data
  1183. + int audio_mux_version_A; ///< LATM syntax version
  1184. + int frame_length_type; ///< 0/1 variable/fixed frame length
  1185. + int frame_length; ///< frame length for fixed frame length
  1186. +};
  1187. +
  1188. +static inline uint32_t latm_get_value(GetBitContext *b)
  1189. +{
  1190. + int length = get_bits(b, 2);
  1191. +
  1192. + return get_bits_long(b, (length+1)*8);
  1193. +}
  1194. +
  1195. +static int latm_decode_audio_specific_config(struct LATMContext *latmctx,
  1196. + GetBitContext *gb, int asclen)
  1197. +{
  1198. + AVCodecContext *avctx = latmctx->aac_ctx.avctx;
  1199. + MPEG4AudioConfig m4ac;
  1200. + AACContext *ac= &latmctx->aac_ctx;
  1201. + int config_start_bit = get_bits_count(gb);
  1202. + int bits_consumed, esize;
  1203. +
  1204. + if (config_start_bit % 8) {
  1205. + av_log_missing_feature(latmctx->aac_ctx.avctx, "audio specific "
  1206. + "config not byte aligned.\n", 1);
  1207. + return AVERROR_INVALIDDATA;
  1208. + } else {
  1209. + bits_consumed =
  1210. + decode_audio_specific_config(ac, avctx, &m4ac,
  1211. + gb->buffer + (config_start_bit / 8),
  1212. + get_bits_left(gb) / 8, asclen);
  1213. +
  1214. + if (bits_consumed < 0)
  1215. + return AVERROR_INVALIDDATA;
  1216. + ac->m4ac= m4ac;
  1217. +
  1218. + esize = (bits_consumed+7) / 8;
  1219. +
  1220. + if (avctx->extradata_size <= esize) {
  1221. + av_free(avctx->extradata);
  1222. + avctx->extradata = av_malloc(esize + FF_INPUT_BUFFER_PADDING_SIZE);
  1223. + if (!avctx->extradata)
  1224. + return AVERROR(ENOMEM);
  1225. + }
  1226. +
  1227. + avctx->extradata_size = esize;
  1228. + memcpy(avctx->extradata, gb->buffer + (config_start_bit/8), esize);
  1229. + memset(avctx->extradata+esize, 0, FF_INPUT_BUFFER_PADDING_SIZE);
  1230. +
  1231. + skip_bits_long(gb, bits_consumed);
  1232. + }
  1233. +
  1234. + return bits_consumed;
  1235. +}
  1236. +
  1237. +static int read_stream_mux_config(struct LATMContext *latmctx,
  1238. + GetBitContext *gb)
  1239. +{
  1240. + int ret, audio_mux_version = get_bits(gb, 1);
  1241. +
  1242. + latmctx->audio_mux_version_A = 0;
  1243. + if (audio_mux_version)
  1244. + latmctx->audio_mux_version_A = get_bits(gb, 1);
  1245. +
  1246. + if (!latmctx->audio_mux_version_A) {
  1247. +
  1248. + if (audio_mux_version)
  1249. + latm_get_value(gb); // taraFullness
  1250. +
  1251. + skip_bits(gb, 1); // allStreamSameTimeFraming
  1252. + skip_bits(gb, 6); // numSubFrames
  1253. + // numPrograms
  1254. + if (get_bits(gb, 4)) { // numPrograms
  1255. + av_log_missing_feature(latmctx->aac_ctx.avctx,
  1256. + "multiple programs are not supported\n", 1);
  1257. + return AVERROR_PATCHWELCOME;
  1258. + }
  1259. +
  1260. + // for each program (which there is only on in DVB)
  1261. +
  1262. + // for each layer (which there is only on in DVB)
  1263. + if (get_bits(gb, 3)) { // numLayer
  1264. + av_log_missing_feature(latmctx->aac_ctx.avctx,
  1265. + "multiple layers are not supported\n", 1);
  1266. + return AVERROR_PATCHWELCOME;
  1267. + }
  1268. +
  1269. + // for all but first stream: use_same_config = get_bits(gb, 1);
  1270. + if (!audio_mux_version) {
  1271. + if ((ret = latm_decode_audio_specific_config(latmctx, gb, 0)) < 0)
  1272. + return ret;
  1273. + } else {
  1274. + int ascLen = latm_get_value(gb);
  1275. + if ((ret = latm_decode_audio_specific_config(latmctx, gb, ascLen)) < 0)
  1276. + return ret;
  1277. + ascLen -= ret;
  1278. + skip_bits_long(gb, ascLen);
  1279. + }
  1280. +
  1281. + latmctx->frame_length_type = get_bits(gb, 3);
  1282. + switch (latmctx->frame_length_type) {
  1283. + case 0:
  1284. + skip_bits(gb, 8); // latmBufferFullness
  1285. + break;
  1286. + case 1:
  1287. + latmctx->frame_length = get_bits(gb, 9);
  1288. + break;
  1289. + case 3:
  1290. + case 4:
  1291. + case 5:
  1292. + skip_bits(gb, 6); // CELP frame length table index
  1293. + break;
  1294. + case 6:
  1295. + case 7:
  1296. + skip_bits(gb, 1); // HVXC frame length table index
  1297. + break;
  1298. + }
  1299. +
  1300. + if (get_bits(gb, 1)) { // other data
  1301. + if (audio_mux_version) {
  1302. + latm_get_value(gb); // other_data_bits
  1303. + } else {
  1304. + int esc;
  1305. + do {
  1306. + esc = get_bits(gb, 1);
  1307. + skip_bits(gb, 8);
  1308. + } while (esc);
  1309. + }
  1310. + }
  1311. +
  1312. + if (get_bits(gb, 1)) // crc present
  1313. + skip_bits(gb, 8); // config_crc
  1314. + }
  1315. +
  1316. + return 0;
  1317. +}
  1318. +
  1319. +static int read_payload_length_info(struct LATMContext *ctx, GetBitContext *gb)
  1320. +{
  1321. + uint8_t tmp;
  1322. +
  1323. + if (ctx->frame_length_type == 0) {
  1324. + int mux_slot_length = 0;
  1325. + do {
  1326. + tmp = get_bits(gb, 8);
  1327. + mux_slot_length += tmp;
  1328. + } while (tmp == 255);
  1329. + return mux_slot_length;
  1330. + } else if (ctx->frame_length_type == 1) {
  1331. + return ctx->frame_length;
  1332. + } else if (ctx->frame_length_type == 3 ||
  1333. + ctx->frame_length_type == 5 ||
  1334. + ctx->frame_length_type == 7) {
  1335. + skip_bits(gb, 2); // mux_slot_length_coded
  1336. + }
  1337. + return 0;
  1338. +}
  1339. +
  1340. +static int read_audio_mux_element(struct LATMContext *latmctx,
  1341. + GetBitContext *gb)
  1342. +{
  1343. + int err;
  1344. + uint8_t use_same_mux = get_bits(gb, 1);
  1345. + if (!use_same_mux) {
  1346. + if ((err = read_stream_mux_config(latmctx, gb)) < 0)
  1347. + return err;
  1348. + } else if (!latmctx->aac_ctx.avctx->extradata) {
  1349. + av_log(latmctx->aac_ctx.avctx, AV_LOG_DEBUG,
  1350. + "no decoder config found\n");
  1351. + return AVERROR(EAGAIN);
  1352. + }
  1353. + if (latmctx->audio_mux_version_A == 0) {
  1354. + int mux_slot_length_bytes = read_payload_length_info(latmctx, gb);
  1355. + if (mux_slot_length_bytes * 8 > get_bits_left(gb)) {
  1356. + av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR, "incomplete frame\n");
  1357. + return AVERROR_INVALIDDATA;
  1358. + } else if (mux_slot_length_bytes * 8 + 256 < get_bits_left(gb)) {
  1359. + av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR,
  1360. + "frame length mismatch %d << %d\n",
  1361. + mux_slot_length_bytes * 8, get_bits_left(gb));
  1362. + return AVERROR_INVALIDDATA;
  1363. + }
  1364. + }
  1365. + return 0;
  1366. +}
  1367. +
  1368. +
  1369. +static int latm_decode_frame(AVCodecContext *avctx, void *out, int *out_size,
  1370. + AVPacket *avpkt)
  1371. +{
  1372. + struct LATMContext *latmctx = avctx->priv_data;
  1373. + int muxlength, err;
  1374. + GetBitContext gb;
  1375. +
  1376. + if (avpkt->size == 0)
  1377. + return 0;
  1378. +
  1379. + init_get_bits(&gb, avpkt->data, avpkt->size * 8);
  1380. +
  1381. + // check for LOAS sync word
  1382. + if (get_bits(&gb, 11) != LOAS_SYNC_WORD)
  1383. + return AVERROR_INVALIDDATA;
  1384. +
  1385. + muxlength = get_bits(&gb, 13) + 3;
  1386. + // not enough data, the parser should have sorted this
  1387. + if (muxlength > avpkt->size)
  1388. + return AVERROR_INVALIDDATA;
  1389. +
  1390. + if ((err = read_audio_mux_element(latmctx, &gb)) < 0)
  1391. + return err;
  1392. +
  1393. + if (!latmctx->initialized) {
  1394. + if (!avctx->extradata) {
  1395. + *out_size = 0;
  1396. + return avpkt->size;
  1397. + } else {
  1398. + aac_decode_close(avctx);
  1399. + if ((err = aac_decode_init(avctx)) < 0)
  1400. + return err;
  1401. + latmctx->initialized = 1;
  1402. + }
  1403. + }
  1404. +
  1405. + if (show_bits(&gb, 12) == 0xfff) {
  1406. + av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR,
  1407. + "ADTS header detected, probably as result of configuration "
  1408. + "misparsing\n");
  1409. + return AVERROR_INVALIDDATA;
  1410. + }
  1411. +
  1412. + if ((err = aac_decode_frame_int(avctx, out, out_size, &gb)) < 0)
  1413. + return err;
  1414. +
  1415. + return muxlength;
  1416. +}
  1417. +
  1418. +av_cold static int latm_decode_init(AVCodecContext *avctx)
  1419. +{
  1420. + struct LATMContext *latmctx = avctx->priv_data;
  1421. + int ret;
  1422. +
  1423. + ret = aac_decode_init(avctx);
  1424. +
  1425. + if (avctx->extradata_size > 0) {
  1426. + latmctx->initialized = !ret;
  1427. + } else {
  1428. + latmctx->initialized = 0;
  1429. + }
  1430. +
  1431. + return ret;
  1432. +}
  1433. +
  1434. +
  1435. AVCodec aac_decoder = {
  1436. "aac",
  1437. AVMEDIA_TYPE_AUDIO,
  1438. @@ -2107,3 +2553,23 @@ AVCodec aac_decoder = {
  1439. },
  1440. .channel_layouts = aac_channel_layout,
  1441. };
  1442. +
  1443. +/*
  1444. + Note: This decoder filter is intended to decode LATM streams transferred
  1445. + in MPEG transport streams which only contain one program.
  1446. + To do a more complex LATM demuxing a separate LATM demuxer should be used.
  1447. +*/
  1448. +AVCodec aac_latm_decoder = {
  1449. + .name = "aac_latm",
  1450. + .type = AVMEDIA_TYPE_AUDIO,
  1451. + .id = CODEC_ID_AAC_LATM,
  1452. + .priv_data_size = sizeof(struct LATMContext),
  1453. + .init = latm_decode_init,
  1454. + .close = aac_decode_close,
  1455. + .decode = latm_decode_frame,
  1456. + .long_name = NULL_IF_CONFIG_SMALL("AAC LATM (Advanced Audio Codec LATM syntax)"),
  1457. + .sample_fmts = (const enum SampleFormat[]) {
  1458. + SAMPLE_FMT_S16,SAMPLE_FMT_NONE
  1459. + },
  1460. + .channel_layouts = aac_channel_layout,
  1461. +};
  1462. diff --git a/mythtv/external/FFmpeg/libavcodec/aacdectab.h b/mythtv/external/FFmpeg/libavcodec/aacdectab.h
  1463. index b74f100..2759c29 100644
  1464. --- a/mythtv/external/FFmpeg/libavcodec/aacdectab.h
  1465. +++ b/mythtv/external/FFmpeg/libavcodec/aacdectab.h
  1466. @@ -34,6 +34,14 @@
  1467.  
  1468. #include <stdint.h>
  1469.  
  1470. +/* @name ltp_coef
  1471. + * Table of the LTP coefficient (multiplied by 2)
  1472. + */
  1473. +static const float ltp_coef[8] = {
  1474. + 1.141658, 1.393232, 1.626008, 1.822608,
  1475. + 1.969800, 2.135788, 2.2389202, 2.739066,
  1476. +};
  1477. +
  1478. /* @name tns_tmp2_map
  1479. * Tables of the tmp2[] arrays of LPC coefficients used for TNS.
  1480. * The suffix _M_N[] indicate the values of coef_compress and coef_res
  1481. diff --git a/mythtv/external/FFmpeg/libavcodec/aacsbr.c b/mythtv/external/FFmpeg/libavcodec/aacsbr.c
  1482. index 050305a..ec7ad3c 100644
  1483. --- a/mythtv/external/FFmpeg/libavcodec/aacsbr.c
  1484. +++ b/mythtv/external/FFmpeg/libavcodec/aacsbr.c
  1485. @@ -32,9 +32,11 @@
  1486. #include "aacsbrdata.h"
  1487. #include "fft.h"
  1488. #include "aacps.h"
  1489. +#include "libavutil/libm.h"
  1490.  
  1491. #include <stdint.h>
  1492. #include <float.h>
  1493. +#include <math.h>
  1494.  
  1495. #define ENVELOPE_ADJUSTMENT_OFFSET 2
  1496. #define NOISE_FLOOR_OFFSET 6.0f
  1497. @@ -127,11 +129,13 @@ av_cold void ff_aac_sbr_init(void)
  1498.  
  1499. av_cold void ff_aac_sbr_ctx_init(SpectralBandReplication *sbr)
  1500. {
  1501. + if(sbr->mdct.mdct_bits)
  1502. + return;
  1503. sbr->kx[0] = sbr->kx[1] = 32; //Typo in spec, kx' inits to 32
  1504. sbr->data[0].e_a[1] = sbr->data[1].e_a[1] = -1;
  1505. sbr->data[0].synthesis_filterbank_samples_offset = SBR_SYNTHESIS_BUF_SIZE - (1280 - 128);
  1506. sbr->data[1].synthesis_filterbank_samples_offset = SBR_SYNTHESIS_BUF_SIZE - (1280 - 128);
  1507. - ff_mdct_init(&sbr->mdct, 7, 1, 1.0/64);
  1508. + ff_mdct_init(&sbr->mdct, 7, 1, 1.0 / 64.0);
  1509. ff_mdct_init(&sbr->mdct_ana, 7, 1, -2.0);
  1510. ff_ps_ctx_init(&sbr->ps);
  1511. }
  1512. @@ -1134,16 +1138,12 @@ static void sbr_dequant(SpectralBandReplication *sbr, int id_aac)
  1513. * @param W array of complex-valued samples split into subbands
  1514. */
  1515. static void sbr_qmf_analysis(DSPContext *dsp, FFTContext *mdct, const float *in, float *x,
  1516. - float z[320], float W[2][32][32][2],
  1517. - float scale)
  1518. + float z[320], float W[2][32][32][2])
  1519. {
  1520. int i, k;
  1521. memcpy(W[0], W[1], sizeof(W[0]));
  1522. memcpy(x , x+1024, (320-32)*sizeof(x[0]));
  1523. - if (scale != 1.0f)
  1524. - dsp->vector_fmul_scalar(x+288, in, scale, 1024);
  1525. - else
  1526. - memcpy(x+288, in, 1024*sizeof(*x));
  1527. + memcpy(x+288, in, 1024*sizeof(x[0]));
  1528. for (i = 0; i < 32; i++) { // numTimeSlots*RATE = 16*2 as 960 sample frames
  1529. // are not supported
  1530. dsp->vector_fmul_reverse(z, sbr_qmf_window_ds, x, 320);
  1531. @@ -1159,7 +1159,7 @@ static void sbr_qmf_analysis(DSPContext *dsp, FFTContext *mdct, const float *in,
  1532. }
  1533. z[64+63] = z[32];
  1534.  
  1535. - ff_imdct_half(mdct, z, z+64);
  1536. + mdct->imdct_half(mdct, z, z+64);
  1537. for (k = 0; k < 32; k++) {
  1538. W[1][i][k][0] = -z[63-k];
  1539. W[1][i][k][1] = z[k];
  1540. @@ -1175,12 +1175,10 @@ static void sbr_qmf_analysis(DSPContext *dsp, FFTContext *mdct, const float *in,
  1541. static void sbr_qmf_synthesis(DSPContext *dsp, FFTContext *mdct,
  1542. float *out, float X[2][38][64],
  1543. float mdct_buf[2][64],
  1544. - float *v0, int *v_off, const unsigned int div,
  1545. - float bias, float scale)
  1546. + float *v0, int *v_off, const unsigned int div)
  1547. {
  1548. int i, n;
  1549. const float *sbr_qmf_window = div ? sbr_qmf_window_ds : sbr_qmf_window_us;
  1550. - int scale_and_bias = scale != 1.0f || bias != 0.0f;
  1551. float *v;
  1552. for (i = 0; i < 32; i++) {
  1553. if (*v_off == 0) {
  1554. @@ -1196,7 +1194,7 @@ static void sbr_qmf_synthesis(DSPContext *dsp, FFTContext *mdct,
  1555. X[0][i][ n] = -X[0][i][n];
  1556. X[0][i][32+n] = X[1][i][31-n];
  1557. }
  1558. - ff_imdct_half(mdct, mdct_buf[0], X[0][i]);
  1559. + mdct->imdct_half(mdct, mdct_buf[0], X[0][i]);
  1560. for (n = 0; n < 32; n++) {
  1561. v[ n] = mdct_buf[0][63 - 2*n];
  1562. v[63 - n] = -mdct_buf[0][62 - 2*n];
  1563. @@ -1205,8 +1203,8 @@ static void sbr_qmf_synthesis(DSPContext *dsp, FFTContext *mdct,
  1564. for (n = 1; n < 64; n+=2) {
  1565. X[1][i][n] = -X[1][i][n];
  1566. }
  1567. - ff_imdct_half(mdct, mdct_buf[0], X[0][i]);
  1568. - ff_imdct_half(mdct, mdct_buf[1], X[1][i]);
  1569. + mdct->imdct_half(mdct, mdct_buf[0], X[0][i]);
  1570. + mdct->imdct_half(mdct, mdct_buf[1], X[1][i]);
  1571. for (n = 0; n < 64; n++) {
  1572. v[ n] = -mdct_buf[0][63 - n] + mdct_buf[1][ n ];
  1573. v[127 - n] = mdct_buf[0][63 - n] + mdct_buf[1][ n ];
  1574. @@ -1222,9 +1220,6 @@ static void sbr_qmf_synthesis(DSPContext *dsp, FFTContext *mdct,
  1575. dsp->vector_fmul_add(out, v + ( 960 >> div), sbr_qmf_window + (448 >> div), out , 64 >> div);
  1576. dsp->vector_fmul_add(out, v + (1024 >> div), sbr_qmf_window + (512 >> div), out , 64 >> div);
  1577. dsp->vector_fmul_add(out, v + (1216 >> div), sbr_qmf_window + (576 >> div), out , 64 >> div);
  1578. - if (scale_and_bias)
  1579. - for (n = 0; n < 64 >> div; n++)
  1580. - out[n] = out[n] * scale + bias;
  1581. out += 64 >> div;
  1582. }
  1583. }
  1584. @@ -1459,6 +1454,7 @@ static void sbr_mapping(AACContext *ac, SpectralBandReplication *sbr,
  1585. uint16_t *table = ch_data->bs_freq_res[e + 1] ? sbr->f_tablehigh : sbr->f_tablelow;
  1586. int k;
  1587.  
  1588. + //av_assert0(sbr->kx[1] <= table[0]);
  1589. for (i = 0; i < ilim; i++)
  1590. for (m = table[i]; m < table[i + 1]; m++)
  1591. sbr->e_origmapped[e][m - sbr->kx[1]] = ch_data->env_facs[e+1][i];
  1592. @@ -1727,7 +1723,7 @@ void ff_sbr_apply(AACContext *ac, SpectralBandReplication *sbr, int id_aac,
  1593. /* decode channel */
  1594. sbr_qmf_analysis(&ac->dsp, &sbr->mdct_ana, ch ? R : L, sbr->data[ch].analysis_filterbank_samples,
  1595. (float*)sbr->qmf_filter_scratch,
  1596. - sbr->data[ch].W, 1/(-1024 * ac->sf_scale));
  1597. + sbr->data[ch].W);
  1598. sbr_lf_gen(ac, sbr, sbr->X_low, sbr->data[ch].W);
  1599. if (sbr->start) {
  1600. sbr_hf_inverse_filter(sbr->alpha0, sbr->alpha1, sbr->X_low, sbr->k[0]);
  1601. @@ -1760,12 +1756,10 @@ void ff_sbr_apply(AACContext *ac, SpectralBandReplication *sbr, int id_aac,
  1602. sbr_qmf_synthesis(&ac->dsp, &sbr->mdct, L, sbr->X[0], sbr->qmf_filter_scratch,
  1603. sbr->data[0].synthesis_filterbank_samples,
  1604. &sbr->data[0].synthesis_filterbank_samples_offset,
  1605. - downsampled,
  1606. - ac->add_bias, -1024 * ac->sf_scale);
  1607. + downsampled);
  1608. if (nch == 2)
  1609. sbr_qmf_synthesis(&ac->dsp, &sbr->mdct, R, sbr->X[1], sbr->qmf_filter_scratch,
  1610. sbr->data[1].synthesis_filterbank_samples,
  1611. &sbr->data[1].synthesis_filterbank_samples_offset,
  1612. - downsampled,
  1613. - ac->add_bias, -1024 * ac->sf_scale);
  1614. + downsampled);
  1615. }
  1616.  
  1617.  
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement
Public Pastes
Advertisement
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!