Mercurial > libavcodec.hg
annotate wmaprodec.c @ 10553:86c7e3c6de00 libavcodec
Update libtheora wrapper to use the 1.0 API
author | conrad |
---|---|
date | Sun, 22 Nov 2009 21:08:37 +0000 |
parents | 99a26e9d6e1b |
children | 3d011a01a6a0 |
rev | line source |
---|---|
10073 | 1 /* |
2 * Wmapro compatible decoder | |
3 * Copyright (c) 2007 Baptiste Coudurier, Benjamin Larsson, Ulion | |
4 * Copyright (c) 2008 - 2009 Sascha Sommer, Benjamin Larsson | |
5 * | |
6 * This file is part of FFmpeg. | |
7 * | |
8 * FFmpeg is free software; you can redistribute it and/or | |
9 * modify it under the terms of the GNU Lesser General Public | |
10 * License as published by the Free Software Foundation; either | |
11 * version 2.1 of the License, or (at your option) any later version. | |
12 * | |
13 * FFmpeg is distributed in the hope that it will be useful, | |
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
16 * Lesser General Public License for more details. | |
17 * | |
18 * You should have received a copy of the GNU Lesser General Public | |
19 * License along with FFmpeg; if not, write to the Free Software | |
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
21 */ | |
22 | |
23 /** | |
24 * @file libavcodec/wmaprodec.c | |
25 * @brief wmapro decoder implementation | |
26 * Wmapro is an MDCT based codec comparable to wma standard or AAC. | |
27 * The decoding therefore consists of the following steps: | |
28 * - bitstream decoding | |
29 * - reconstruction of per-channel data | |
30 * - rescaling and inverse quantization | |
31 * - IMDCT | |
32 * - windowing and overlapp-add | |
33 * | |
34 * The compressed wmapro bitstream is split into individual packets. | |
35 * Every such packet contains one or more wma frames. | |
36 * The compressed frames may have a variable length and frames may | |
37 * cross packet boundaries. | |
38 * Common to all wmapro frames is the number of samples that are stored in | |
39 * a frame. | |
40 * The number of samples and a few other decode flags are stored | |
41 * as extradata that has to be passed to the decoder. | |
42 * | |
43 * The wmapro frames themselves are again split into a variable number of | |
44 * subframes. Every subframe contains the data for 2^N time domain samples | |
45 * where N varies between 7 and 12. | |
46 * | |
47 * Example wmapro bitstream (in samples): | |
48 * | |
49 * || packet 0 || packet 1 || packet 2 packets | |
50 * --------------------------------------------------- | |
51 * || frame 0 || frame 1 || frame 2 || frames | |
52 * --------------------------------------------------- | |
53 * || | | || | | | || || subframes of channel 0 | |
54 * --------------------------------------------------- | |
55 * || | | || | | | || || subframes of channel 1 | |
56 * --------------------------------------------------- | |
57 * | |
58 * The frame layouts for the individual channels of a wma frame does not need | |
59 * to be the same. | |
60 * | |
61 * However, if the offsets and lengths of several subframes of a frame are the | |
62 * same, the subframes of the channels can be grouped. | |
63 * Every group may then use special coding techniques like M/S stereo coding | |
64 * to improve the compression ratio. These channel transformations do not | |
65 * need to be applied to a whole subframe. Instead, they can also work on | |
66 * individual scale factor bands (see below). | |
67 * The coefficients that carry the audio signal in the frequency domain | |
68 * are transmitted as huffman-coded vectors with 4, 2 and 1 elements. | |
69 * In addition to that, the encoder can switch to a runlevel coding scheme | |
70 * by transmitting subframe_length / 128 zero coefficients. | |
71 * | |
72 * Before the audio signal can be converted to the time domain, the | |
73 * coefficients have to be rescaled and inverse quantized. | |
74 * A subframe is therefore split into several scale factor bands that get | |
75 * scaled individually. | |
76 * Scale factors are submitted for every frame but they might be shared | |
77 * between the subframes of a channel. Scale factors are initially DPCM-coded. | |
78 * Once scale factors are shared, the differences are transmitted as runlevel | |
79 * codes. | |
80 * Every subframe length and offset combination in the frame layout shares a | |
81 * common quantization factor that can be adjusted for every channel by a | |
82 * modifier. | |
83 * After the inverse quantization, the coefficients get processed by an IMDCT. | |
84 * The resulting values are then windowed with a sine window and the first half | |
85 * of the values are added to the second half of the output from the previous | |
86 * subframe in order to reconstruct the output samples. | |
87 */ | |
88 | |
10122 | 89 #include "avcodec.h" |
90 #include "internal.h" | |
91 #include "get_bits.h" | |
92 #include "put_bits.h" | |
93 #include "wmaprodata.h" | |
94 #include "dsputil.h" | |
95 #include "wma.h" | |
96 | |
97 /** current decoder limitations */ | |
98 #define WMAPRO_MAX_CHANNELS 8 ///< max number of handled channels | |
99 #define MAX_SUBFRAMES 32 ///< max number of subframes per channel | |
100 #define MAX_BANDS 29 ///< max number of scale factor bands | |
10353 | 101 #define MAX_FRAMESIZE 32768 ///< maximum compressed frame size |
10122 | 102 |
103 #define WMAPRO_BLOCK_MAX_BITS 12 ///< log2 of max block size | |
104 #define WMAPRO_BLOCK_MAX_SIZE (1 << WMAPRO_BLOCK_MAX_BITS) ///< maximum block size | |
105 #define WMAPRO_BLOCK_SIZES (WMAPRO_BLOCK_MAX_BITS - BLOCK_MIN_BITS + 1) ///< possible block sizes | |
106 | |
107 | |
108 #define VLCBITS 9 | |
109 #define SCALEVLCBITS 8 | |
110 #define VEC4MAXDEPTH ((HUFF_VEC4_MAXBITS+VLCBITS-1)/VLCBITS) | |
111 #define VEC2MAXDEPTH ((HUFF_VEC2_MAXBITS+VLCBITS-1)/VLCBITS) | |
112 #define VEC1MAXDEPTH ((HUFF_VEC1_MAXBITS+VLCBITS-1)/VLCBITS) | |
113 #define SCALEMAXDEPTH ((HUFF_SCALE_MAXBITS+SCALEVLCBITS-1)/SCALEVLCBITS) | |
114 #define SCALERLMAXDEPTH ((HUFF_SCALE_RL_MAXBITS+VLCBITS-1)/VLCBITS) | |
115 | |
116 static VLC sf_vlc; ///< scale factor DPCM vlc | |
117 static VLC sf_rl_vlc; ///< scale factor run length vlc | |
118 static VLC vec4_vlc; ///< 4 coefficients per symbol | |
119 static VLC vec2_vlc; ///< 2 coefficients per symbol | |
120 static VLC vec1_vlc; ///< 1 coefficient per symbol | |
121 static VLC coef_vlc[2]; ///< coefficient run length vlc codes | |
122 static float sin64[33]; ///< sinus table for decorrelation | |
123 | |
124 /** | |
125 * @brief frame specific decoder context for a single channel | |
126 */ | |
127 typedef struct { | |
128 int16_t prev_block_len; ///< length of the previous block | |
129 uint8_t transmit_coefs; | |
130 uint8_t num_subframes; | |
131 uint16_t subframe_len[MAX_SUBFRAMES]; ///< subframe length in samples | |
132 uint16_t subframe_offset[MAX_SUBFRAMES]; ///< subframe positions in the current frame | |
133 uint8_t cur_subframe; ///< current subframe number | |
134 uint16_t decoded_samples; ///< number of already processed samples | |
135 uint8_t grouped; ///< channel is part of a group | |
136 int quant_step; ///< quantization step for the current subframe | |
137 int8_t reuse_sf; ///< share scale factors between subframes | |
138 int8_t scale_factor_step; ///< scaling step for the current subframe | |
139 int max_scale_factor; ///< maximum scale factor for the current subframe | |
10138 | 140 int saved_scale_factors[2][MAX_BANDS]; ///< resampled and (previously) transmitted scale factor values |
141 int8_t scale_factor_idx; ///< index for the transmitted scale factor values (used for resampling) | |
142 int* scale_factors; ///< pointer to the scale factor values used for decoding | |
10122 | 143 uint8_t table_idx; ///< index in sf_offsets for the scale factor reference block |
144 float* coeffs; ///< pointer to the subframe decode buffer | |
145 DECLARE_ALIGNED_16(float, out[WMAPRO_BLOCK_MAX_SIZE + WMAPRO_BLOCK_MAX_SIZE / 2]); ///< output buffer | |
146 } WMAProChannelCtx; | |
147 | |
148 /** | |
149 * @brief channel group for channel transformations | |
150 */ | |
151 typedef struct { | |
152 uint8_t num_channels; ///< number of channels in the group | |
153 int8_t transform; ///< transform on / off | |
154 int8_t transform_band[MAX_BANDS]; ///< controls if the transform is enabled for a certain band | |
155 float decorrelation_matrix[WMAPRO_MAX_CHANNELS*WMAPRO_MAX_CHANNELS]; | |
156 float* channel_data[WMAPRO_MAX_CHANNELS]; ///< transformation coefficients | |
157 } WMAProChannelGrp; | |
158 | |
159 /** | |
160 * @brief main decoder context | |
161 */ | |
162 typedef struct WMAProDecodeCtx { | |
163 /* generic decoder variables */ | |
164 AVCodecContext* avctx; ///< codec context for av_log | |
165 DSPContext dsp; ///< accelerated DSP functions | |
166 uint8_t frame_data[MAX_FRAMESIZE + | |
167 FF_INPUT_BUFFER_PADDING_SIZE];///< compressed frame data | |
168 PutBitContext pb; ///< context for filling the frame_data buffer | |
10199 | 169 FFTContext mdct_ctx[WMAPRO_BLOCK_SIZES]; ///< MDCT context per block size |
10122 | 170 DECLARE_ALIGNED_16(float, tmp[WMAPRO_BLOCK_MAX_SIZE]); ///< IMDCT output buffer |
171 float* windows[WMAPRO_BLOCK_SIZES]; ///< windows for the different block sizes | |
172 | |
173 /* frame size dependent frame information (set during initialization) */ | |
174 uint32_t decode_flags; ///< used compression features | |
175 uint8_t len_prefix; ///< frame is prefixed with its length | |
176 uint8_t dynamic_range_compression; ///< frame contains DRC data | |
177 uint8_t bits_per_sample; ///< integer audio sample size for the unscaled IMDCT output (used to scale to [-1.0, 1.0]) | |
178 uint16_t samples_per_frame; ///< number of samples to output | |
179 uint16_t log2_frame_size; | |
180 int8_t num_channels; ///< number of channels in the stream (same as AVCodecContext.num_channels) | |
181 int8_t lfe_channel; ///< lfe channel index | |
182 uint8_t max_num_subframes; | |
183 uint8_t subframe_len_bits; ///< number of bits used for the subframe length | |
184 uint8_t max_subframe_len_bit; ///< flag indicating that the subframe is of maximum size when the first subframe length bit is 1 | |
185 uint16_t min_samples_per_subframe; | |
186 int8_t num_sfb[WMAPRO_BLOCK_SIZES]; ///< scale factor bands per block size | |
187 int16_t sfb_offsets[WMAPRO_BLOCK_SIZES][MAX_BANDS]; ///< scale factor band offsets (multiples of 4) | |
188 int8_t sf_offsets[WMAPRO_BLOCK_SIZES][WMAPRO_BLOCK_SIZES][MAX_BANDS]; ///< scale factor resample matrix | |
189 int16_t subwoofer_cutoffs[WMAPRO_BLOCK_SIZES]; ///< subwoofer cutoff values | |
190 | |
191 /* packet decode state */ | |
10126 | 192 GetBitContext pgb; ///< bitstream reader context for the packet |
10165 | 193 uint8_t packet_offset; ///< frame offset in the packet |
10122 | 194 uint8_t packet_sequence_number; ///< current packet number |
195 int num_saved_bits; ///< saved number of bits | |
196 int frame_offset; ///< frame offset in the bit reservoir | |
197 int subframe_offset; ///< subframe offset in the bit reservoir | |
198 uint8_t packet_loss; ///< set in case of bitstream error | |
10139
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
199 uint8_t packet_done; ///< set when a packet is fully decoded |
10122 | 200 |
201 /* frame decode state */ | |
202 uint32_t frame_num; ///< current frame number (not used for decoding) | |
203 GetBitContext gb; ///< bitstream reader context | |
204 int buf_bit_size; ///< buffer size in bits | |
205 float* samples; ///< current samplebuffer pointer | |
206 float* samples_end; ///< maximum samplebuffer pointer | |
207 uint8_t drc_gain; ///< gain for the DRC tool | |
208 int8_t skip_frame; ///< skip output step | |
209 int8_t parsed_all_subframes; ///< all subframes decoded? | |
210 | |
211 /* subframe/block decode state */ | |
212 int16_t subframe_len; ///< current subframe length | |
213 int8_t channels_for_cur_subframe; ///< number of channels that contain the subframe | |
214 int8_t channel_indexes_for_cur_subframe[WMAPRO_MAX_CHANNELS]; | |
215 int8_t num_bands; ///< number of scale factor bands | |
216 int16_t* cur_sfb_offsets; ///< sfb offsets for the current block | |
217 uint8_t table_idx; ///< index for the num_sfb, sfb_offsets, sf_offsets and subwoofer_cutoffs tables | |
218 int8_t esc_len; ///< length of escaped coefficients | |
219 | |
220 uint8_t num_chgroups; ///< number of channel groups | |
221 WMAProChannelGrp chgroup[WMAPRO_MAX_CHANNELS]; ///< channel group information | |
222 | |
223 WMAProChannelCtx channel[WMAPRO_MAX_CHANNELS]; ///< per channel data | |
224 } WMAProDecodeCtx; | |
225 | |
226 | |
227 /** | |
228 *@brief helper function to print the most important members of the context | |
229 *@param s context | |
230 */ | |
231 static void av_cold dump_context(WMAProDecodeCtx *s) | |
232 { | |
233 #define PRINT(a, b) av_log(s->avctx, AV_LOG_DEBUG, " %s = %d\n", a, b); | |
234 #define PRINT_HEX(a, b) av_log(s->avctx, AV_LOG_DEBUG, " %s = %x\n", a, b); | |
235 | |
236 PRINT("ed sample bit depth", s->bits_per_sample); | |
237 PRINT_HEX("ed decode flags", s->decode_flags); | |
238 PRINT("samples per frame", s->samples_per_frame); | |
239 PRINT("log2 frame size", s->log2_frame_size); | |
240 PRINT("max num subframes", s->max_num_subframes); | |
241 PRINT("len prefix", s->len_prefix); | |
242 PRINT("num channels", s->num_channels); | |
243 } | |
244 | |
10005 | 245 /** |
246 *@brief Uninitialize the decoder and free all resources. | |
247 *@param avctx codec context | |
248 *@return 0 on success, < 0 otherwise | |
249 */ | |
250 static av_cold int decode_end(AVCodecContext *avctx) | |
251 { | |
10122 | 252 WMAProDecodeCtx *s = avctx->priv_data; |
10005 | 253 int i; |
254 | |
10084 | 255 for (i = 0; i < WMAPRO_BLOCK_SIZES; i++) |
10005 | 256 ff_mdct_end(&s->mdct_ctx[i]); |
257 | |
258 return 0; | |
259 } | |
260 | |
261 /** | |
10122 | 262 *@brief Initialize the decoder. |
263 *@param avctx codec context | |
264 *@return 0 on success, -1 otherwise | |
265 */ | |
266 static av_cold int decode_init(AVCodecContext *avctx) | |
267 { | |
268 WMAProDecodeCtx *s = avctx->priv_data; | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
269 uint8_t *edata_ptr = avctx->extradata; |
10122 | 270 unsigned int channel_mask; |
271 int i; | |
272 int log2_max_num_subframes; | |
273 int num_possible_block_sizes; | |
274 | |
275 s->avctx = avctx; | |
276 dsputil_init(&s->dsp, avctx); | |
277 init_put_bits(&s->pb, s->frame_data, MAX_FRAMESIZE); | |
278 | |
279 avctx->sample_fmt = SAMPLE_FMT_FLT; | |
280 | |
281 if (avctx->extradata_size >= 18) { | |
282 s->decode_flags = AV_RL16(edata_ptr+14); | |
283 channel_mask = AV_RL32(edata_ptr+2); | |
284 s->bits_per_sample = AV_RL16(edata_ptr); | |
285 /** dump the extradata */ | |
286 for (i = 0; i < avctx->extradata_size; i++) | |
287 dprintf(avctx, "[%x] ", avctx->extradata[i]); | |
288 dprintf(avctx, "\n"); | |
289 | |
290 } else { | |
291 av_log_ask_for_sample(avctx, "Unknown extradata size\n"); | |
292 return AVERROR_INVALIDDATA; | |
293 } | |
294 | |
295 /** generic init */ | |
296 s->log2_frame_size = av_log2(avctx->block_align) + 4; | |
297 | |
298 /** frame info */ | |
299 s->skip_frame = 1; /** skip first frame */ | |
300 s->packet_loss = 1; | |
301 s->len_prefix = (s->decode_flags & 0x40); | |
302 | |
303 if (!s->len_prefix) { | |
304 av_log_ask_for_sample(avctx, "no length prefix\n"); | |
305 return AVERROR_INVALIDDATA; | |
306 } | |
307 | |
308 /** get frame len */ | |
309 s->samples_per_frame = 1 << ff_wma_get_frame_len_bits(avctx->sample_rate, | |
310 3, s->decode_flags); | |
311 | |
312 /** init previous block len */ | |
313 for (i = 0; i < avctx->channels; i++) | |
314 s->channel[i].prev_block_len = s->samples_per_frame; | |
315 | |
316 /** subframe info */ | |
317 log2_max_num_subframes = ((s->decode_flags & 0x38) >> 3); | |
318 s->max_num_subframes = 1 << log2_max_num_subframes; | |
319 if (s->max_num_subframes == 16) | |
320 s->max_subframe_len_bit = 1; | |
321 s->subframe_len_bits = av_log2(log2_max_num_subframes) + 1; | |
322 | |
323 num_possible_block_sizes = log2_max_num_subframes + 1; | |
324 s->min_samples_per_subframe = s->samples_per_frame / s->max_num_subframes; | |
325 s->dynamic_range_compression = (s->decode_flags & 0x80); | |
326 | |
327 if (s->max_num_subframes > MAX_SUBFRAMES) { | |
328 av_log(avctx, AV_LOG_ERROR, "invalid number of subframes %i\n", | |
329 s->max_num_subframes); | |
330 return AVERROR_INVALIDDATA; | |
331 } | |
332 | |
333 s->num_channels = avctx->channels; | |
334 | |
335 /** extract lfe channel position */ | |
336 s->lfe_channel = -1; | |
337 | |
338 if (channel_mask & 8) { | |
339 unsigned int mask; | |
340 for (mask = 1; mask < 16; mask <<= 1) { | |
341 if (channel_mask & mask) | |
342 ++s->lfe_channel; | |
343 } | |
344 } | |
345 | |
346 if (s->num_channels < 0 || s->num_channels > WMAPRO_MAX_CHANNELS) { | |
347 av_log_ask_for_sample(avctx, "invalid number of channels\n"); | |
348 return AVERROR_NOTSUPP; | |
349 } | |
350 | |
351 INIT_VLC_STATIC(&sf_vlc, SCALEVLCBITS, HUFF_SCALE_SIZE, | |
352 scale_huffbits, 1, 1, | |
353 scale_huffcodes, 2, 2, 616); | |
354 | |
355 INIT_VLC_STATIC(&sf_rl_vlc, VLCBITS, HUFF_SCALE_RL_SIZE, | |
356 scale_rl_huffbits, 1, 1, | |
357 scale_rl_huffcodes, 4, 4, 1406); | |
358 | |
359 INIT_VLC_STATIC(&coef_vlc[0], VLCBITS, HUFF_COEF0_SIZE, | |
360 coef0_huffbits, 1, 1, | |
361 coef0_huffcodes, 4, 4, 2108); | |
362 | |
363 INIT_VLC_STATIC(&coef_vlc[1], VLCBITS, HUFF_COEF1_SIZE, | |
364 coef1_huffbits, 1, 1, | |
365 coef1_huffcodes, 4, 4, 3912); | |
366 | |
367 INIT_VLC_STATIC(&vec4_vlc, VLCBITS, HUFF_VEC4_SIZE, | |
368 vec4_huffbits, 1, 1, | |
369 vec4_huffcodes, 2, 2, 604); | |
370 | |
371 INIT_VLC_STATIC(&vec2_vlc, VLCBITS, HUFF_VEC2_SIZE, | |
372 vec2_huffbits, 1, 1, | |
373 vec2_huffcodes, 2, 2, 562); | |
374 | |
375 INIT_VLC_STATIC(&vec1_vlc, VLCBITS, HUFF_VEC1_SIZE, | |
376 vec1_huffbits, 1, 1, | |
377 vec1_huffcodes, 2, 2, 562); | |
378 | |
379 /** calculate number of scale factor bands and their offsets | |
380 for every possible block size */ | |
381 for (i = 0; i < num_possible_block_sizes; i++) { | |
382 int subframe_len = s->samples_per_frame >> i; | |
383 int x; | |
384 int band = 1; | |
385 | |
386 s->sfb_offsets[i][0] = 0; | |
387 | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
388 for (x = 0; x < MAX_BANDS-1 && s->sfb_offsets[i][band - 1] < subframe_len; x++) { |
10122 | 389 int offset = (subframe_len * 2 * critical_freq[x]) |
390 / s->avctx->sample_rate + 2; | |
391 offset &= ~3; | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
392 if (offset > s->sfb_offsets[i][band - 1]) |
10122 | 393 s->sfb_offsets[i][band++] = offset; |
394 } | |
395 s->sfb_offsets[i][band - 1] = subframe_len; | |
396 s->num_sfb[i] = band - 1; | |
397 } | |
398 | |
399 | |
400 /** Scale factors can be shared between blocks of different size | |
401 as every block has a different scale factor band layout. | |
402 The matrix sf_offsets is needed to find the correct scale factor. | |
403 */ | |
404 | |
405 for (i = 0; i < num_possible_block_sizes; i++) { | |
406 int b; | |
407 for (b = 0; b < s->num_sfb[i]; b++) { | |
408 int x; | |
409 int offset = ((s->sfb_offsets[i][b] | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
410 + s->sfb_offsets[i][b + 1] - 1) << i) >> 1; |
10122 | 411 for (x = 0; x < num_possible_block_sizes; x++) { |
412 int v = 0; | |
413 while (s->sfb_offsets[x][v + 1] << x < offset) | |
414 ++v; | |
415 s->sf_offsets[i][x][b] = v; | |
416 } | |
417 } | |
418 } | |
419 | |
420 /** init MDCT, FIXME: only init needed sizes */ | |
421 for (i = 0; i < WMAPRO_BLOCK_SIZES; i++) | |
422 ff_mdct_init(&s->mdct_ctx[i], BLOCK_MIN_BITS+1+i, 1, | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
423 1.0 / (1 << (BLOCK_MIN_BITS + i - 1)) |
10122 | 424 / (1 << (s->bits_per_sample - 1))); |
425 | |
426 /** init MDCT windows: simple sinus window */ | |
427 for (i = 0; i < WMAPRO_BLOCK_SIZES; i++) { | |
428 const int n = 1 << (WMAPRO_BLOCK_MAX_BITS - i); | |
10174
89cd870ca180
Add two more sizes to ff_sine_windows[] and also pad it with NULLs so
vitor
parents:
10165
diff
changeset
|
429 const int win_idx = WMAPRO_BLOCK_MAX_BITS - i; |
10122 | 430 ff_sine_window_init(ff_sine_windows[win_idx], n); |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
431 s->windows[WMAPRO_BLOCK_SIZES - i - 1] = ff_sine_windows[win_idx]; |
10122 | 432 } |
433 | |
434 /** calculate subwoofer cutoff values */ | |
435 for (i = 0; i < num_possible_block_sizes; i++) { | |
436 int block_size = s->samples_per_frame >> i; | |
437 int cutoff = (440*block_size + 3 * (s->avctx->sample_rate >> 1) - 1) | |
438 / s->avctx->sample_rate; | |
439 s->subwoofer_cutoffs[i] = av_clip(cutoff, 4, block_size); | |
440 } | |
441 | |
442 /** calculate sine values for the decorrelation matrix */ | |
443 for (i = 0; i < 33; i++) | |
444 sin64[i] = sin(i*M_PI / 64.0); | |
445 | |
446 if (avctx->debug & FF_DEBUG_BITSTREAM) | |
447 dump_context(s); | |
448 | |
449 avctx->channel_layout = channel_mask; | |
450 return 0; | |
451 } | |
452 | |
453 /** | |
454 *@brief Decode the subframe length. | |
455 *@param s context | |
456 *@param offset sample offset in the frame | |
457 *@return decoded subframe length on success, < 0 in case of an error | |
458 */ | |
459 static int decode_subframe_length(WMAProDecodeCtx *s, int offset) | |
460 { | |
461 int frame_len_shift = 0; | |
462 int subframe_len; | |
463 | |
464 /** no need to read from the bitstream when only one length is possible */ | |
465 if (offset == s->samples_per_frame - s->min_samples_per_subframe) | |
466 return s->min_samples_per_subframe; | |
467 | |
468 /** 1 bit indicates if the subframe is of maximum length */ | |
469 if (s->max_subframe_len_bit) { | |
470 if (get_bits1(&s->gb)) | |
471 frame_len_shift = 1 + get_bits(&s->gb, s->subframe_len_bits-1); | |
472 } else | |
473 frame_len_shift = get_bits(&s->gb, s->subframe_len_bits); | |
474 | |
475 subframe_len = s->samples_per_frame >> frame_len_shift; | |
476 | |
477 /** sanity check the length */ | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
478 if (subframe_len < s->min_samples_per_subframe || |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
479 subframe_len > s->samples_per_frame) { |
10122 | 480 av_log(s->avctx, AV_LOG_ERROR, "broken frame: subframe_len %i\n", |
481 subframe_len); | |
482 return AVERROR_INVALIDDATA; | |
483 } | |
484 return subframe_len; | |
485 } | |
486 | |
487 /** | |
488 *@brief Decode how the data in the frame is split into subframes. | |
489 * Every WMA frame contains the encoded data for a fixed number of | |
490 * samples per channel. The data for every channel might be split | |
491 * into several subframes. This function will reconstruct the list of | |
492 * subframes for every channel. | |
493 * | |
494 * If the subframes are not evenly split, the algorithm estimates the | |
495 * channels with the lowest number of total samples. | |
496 * Afterwards, for each of these channels a bit is read from the | |
497 * bitstream that indicates if the channel contains a subframe with the | |
498 * next subframe size that is going to be read from the bitstream or not. | |
499 * If a channel contains such a subframe, the subframe size gets added to | |
500 * the channel's subframe list. | |
501 * The algorithm repeats these steps until the frame is properly divided | |
502 * between the individual channels. | |
503 * | |
504 *@param s context | |
505 *@return 0 on success, < 0 in case of an error | |
506 */ | |
507 static int decode_tilehdr(WMAProDecodeCtx *s) | |
508 { | |
509 uint16_t num_samples[WMAPRO_MAX_CHANNELS]; /** sum of samples for all currently known subframes of a channel */ | |
510 uint8_t contains_subframe[WMAPRO_MAX_CHANNELS]; /** flag indicating if a channel contains the current subframe */ | |
511 int channels_for_cur_subframe = s->num_channels; /** number of channels that contain the current subframe */ | |
512 int fixed_channel_layout = 0; /** flag indicating that all channels use the same subframe offsets and sizes */ | |
513 int min_channel_len = 0; /** smallest sum of samples (channels with this length will be processed first) */ | |
514 int c; | |
515 | |
516 /* Should never consume more than 3073 bits (256 iterations for the | |
517 * while loop when always the minimum amount of 128 samples is substracted | |
518 * from missing samples in the 8 channel case). | |
519 * 1 + BLOCK_MAX_SIZE * MAX_CHANNELS / BLOCK_MIN_SIZE * (MAX_CHANNELS + 4) | |
520 */ | |
521 | |
522 /** reset tiling information */ | |
523 for (c = 0; c < s->num_channels; c++) | |
524 s->channel[c].num_subframes = 0; | |
525 | |
526 memset(num_samples, 0, sizeof(num_samples)); | |
527 | |
528 if (s->max_num_subframes == 1 || get_bits1(&s->gb)) | |
529 fixed_channel_layout = 1; | |
530 | |
531 /** loop until the frame data is split between the subframes */ | |
532 do { | |
533 int subframe_len; | |
534 | |
535 /** check which channels contain the subframe */ | |
536 for (c = 0; c < s->num_channels; c++) { | |
537 if (num_samples[c] == min_channel_len) { | |
538 if (fixed_channel_layout || channels_for_cur_subframe == 1 || | |
539 (min_channel_len == s->samples_per_frame - s->min_samples_per_subframe)) | |
540 contains_subframe[c] = 1; | |
541 else | |
542 contains_subframe[c] = get_bits1(&s->gb); | |
543 } else | |
544 contains_subframe[c] = 0; | |
545 } | |
546 | |
547 /** get subframe length, subframe_len == 0 is not allowed */ | |
548 if ((subframe_len = decode_subframe_length(s, min_channel_len)) <= 0) | |
549 return AVERROR_INVALIDDATA; | |
550 | |
551 /** add subframes to the individual channels and find new min_channel_len */ | |
552 min_channel_len += subframe_len; | |
553 for (c = 0; c < s->num_channels; c++) { | |
554 WMAProChannelCtx* chan = &s->channel[c]; | |
555 | |
556 if (contains_subframe[c]) { | |
557 if (chan->num_subframes >= MAX_SUBFRAMES) { | |
558 av_log(s->avctx, AV_LOG_ERROR, | |
559 "broken frame: num subframes > 31\n"); | |
560 return AVERROR_INVALIDDATA; | |
561 } | |
562 chan->subframe_len[chan->num_subframes] = subframe_len; | |
563 num_samples[c] += subframe_len; | |
564 ++chan->num_subframes; | |
565 if (num_samples[c] > s->samples_per_frame) { | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
566 av_log(s->avctx, AV_LOG_ERROR, "broken frame: " |
10122 | 567 "channel len > samples_per_frame\n"); |
568 return AVERROR_INVALIDDATA; | |
569 } | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
570 } else if (num_samples[c] <= min_channel_len) { |
10122 | 571 if (num_samples[c] < min_channel_len) { |
572 channels_for_cur_subframe = 0; | |
573 min_channel_len = num_samples[c]; | |
574 } | |
575 ++channels_for_cur_subframe; | |
576 } | |
577 } | |
578 } while (min_channel_len < s->samples_per_frame); | |
579 | |
580 for (c = 0; c < s->num_channels; c++) { | |
581 int i; | |
582 int offset = 0; | |
583 for (i = 0; i < s->channel[c].num_subframes; i++) { | |
584 dprintf(s->avctx, "frame[%i] channel[%i] subframe[%i]" | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
585 " len %i\n", s->frame_num, c, i, |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
586 s->channel[c].subframe_len[i]); |
10122 | 587 s->channel[c].subframe_offset[i] = offset; |
588 offset += s->channel[c].subframe_len[i]; | |
589 } | |
590 } | |
591 | |
592 return 0; | |
593 } | |
594 | |
595 /** | |
10005 | 596 *@brief Calculate a decorrelation matrix from the bitstream parameters. |
597 *@param s codec context | |
598 *@param chgroup channel group for which the matrix needs to be calculated | |
599 */ | |
10122 | 600 static void decode_decorrelation_matrix(WMAProDecodeCtx *s, |
601 WMAProChannelGrp *chgroup) | |
10005 | 602 { |
603 int i; | |
604 int offset = 0; | |
605 int8_t rotation_offset[WMAPRO_MAX_CHANNELS * WMAPRO_MAX_CHANNELS]; | |
10122 | 606 memset(chgroup->decorrelation_matrix, 0, s->num_channels * |
607 s->num_channels * sizeof(*chgroup->decorrelation_matrix)); | |
10005 | 608 |
10006 | 609 for (i = 0; i < chgroup->num_channels * (chgroup->num_channels - 1) >> 1; i++) |
10084 | 610 rotation_offset[i] = get_bits(&s->gb, 6); |
10005 | 611 |
10006 | 612 for (i = 0; i < chgroup->num_channels; i++) |
10005 | 613 chgroup->decorrelation_matrix[chgroup->num_channels * i + i] = |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
614 get_bits1(&s->gb) ? 1.0 : -1.0; |
10005 | 615 |
10006 | 616 for (i = 1; i < chgroup->num_channels; i++) { |
10005 | 617 int x; |
10006 | 618 for (x = 0; x < i; x++) { |
10005 | 619 int y; |
10084 | 620 for (y = 0; y < i + 1; y++) { |
10005 | 621 float v1 = chgroup->decorrelation_matrix[x * chgroup->num_channels + y]; |
622 float v2 = chgroup->decorrelation_matrix[i * chgroup->num_channels + y]; | |
623 int n = rotation_offset[offset + x]; | |
624 float sinv; | |
625 float cosv; | |
626 | |
10006 | 627 if (n < 32) { |
10005 | 628 sinv = sin64[n]; |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
629 cosv = sin64[32 - n]; |
10005 | 630 } else { |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
631 sinv = sin64[64 - n]; |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
632 cosv = -sin64[n - 32]; |
10005 | 633 } |
634 | |
635 chgroup->decorrelation_matrix[y + x * chgroup->num_channels] = | |
636 (v1 * sinv) - (v2 * cosv); | |
637 chgroup->decorrelation_matrix[y + i * chgroup->num_channels] = | |
638 (v1 * cosv) + (v2 * sinv); | |
639 } | |
640 } | |
641 offset += i; | |
642 } | |
643 } | |
644 | |
645 /** | |
10122 | 646 *@brief Decode channel transformation parameters |
647 *@param s codec context | |
648 *@return 0 in case of success, < 0 in case of bitstream errors | |
649 */ | |
650 static int decode_channel_transform(WMAProDecodeCtx* s) | |
651 { | |
652 int i; | |
653 /* should never consume more than 1921 bits for the 8 channel case | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
654 * 1 + MAX_CHANNELS * (MAX_CHANNELS + 2 + 3 * MAX_CHANNELS * MAX_CHANNELS |
10122 | 655 * + MAX_CHANNELS + MAX_BANDS + 1) |
656 */ | |
657 | |
658 /** in the one channel case channel transforms are pointless */ | |
659 s->num_chgroups = 0; | |
660 if (s->num_channels > 1) { | |
661 int remaining_channels = s->channels_for_cur_subframe; | |
662 | |
663 if (get_bits1(&s->gb)) { | |
664 av_log_ask_for_sample(s->avctx, | |
665 "unsupported channel transform bit\n"); | |
666 return AVERROR_INVALIDDATA; | |
667 } | |
668 | |
669 for (s->num_chgroups = 0; remaining_channels && | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
670 s->num_chgroups < s->channels_for_cur_subframe; s->num_chgroups++) { |
10122 | 671 WMAProChannelGrp* chgroup = &s->chgroup[s->num_chgroups]; |
672 float** channel_data = chgroup->channel_data; | |
673 chgroup->num_channels = 0; | |
674 chgroup->transform = 0; | |
675 | |
676 /** decode channel mask */ | |
677 if (remaining_channels > 2) { | |
678 for (i = 0; i < s->channels_for_cur_subframe; i++) { | |
679 int channel_idx = s->channel_indexes_for_cur_subframe[i]; | |
680 if (!s->channel[channel_idx].grouped | |
681 && get_bits1(&s->gb)) { | |
682 ++chgroup->num_channels; | |
683 s->channel[channel_idx].grouped = 1; | |
684 *channel_data++ = s->channel[channel_idx].coeffs; | |
685 } | |
686 } | |
687 } else { | |
688 chgroup->num_channels = remaining_channels; | |
689 for (i = 0; i < s->channels_for_cur_subframe; i++) { | |
690 int channel_idx = s->channel_indexes_for_cur_subframe[i]; | |
691 if (!s->channel[channel_idx].grouped) | |
692 *channel_data++ = s->channel[channel_idx].coeffs; | |
693 s->channel[channel_idx].grouped = 1; | |
694 } | |
695 } | |
696 | |
697 /** decode transform type */ | |
698 if (chgroup->num_channels == 2) { | |
699 if (get_bits1(&s->gb)) { | |
700 if (get_bits1(&s->gb)) { | |
701 av_log_ask_for_sample(s->avctx, | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
702 "unsupported channel transform type\n"); |
10122 | 703 } |
704 } else { | |
705 chgroup->transform = 1; | |
706 if (s->num_channels == 2) { | |
707 chgroup->decorrelation_matrix[0] = 1.0; | |
708 chgroup->decorrelation_matrix[1] = -1.0; | |
709 chgroup->decorrelation_matrix[2] = 1.0; | |
710 chgroup->decorrelation_matrix[3] = 1.0; | |
711 } else { | |
712 /** cos(pi/4) */ | |
713 chgroup->decorrelation_matrix[0] = 0.70703125; | |
714 chgroup->decorrelation_matrix[1] = -0.70703125; | |
715 chgroup->decorrelation_matrix[2] = 0.70703125; | |
716 chgroup->decorrelation_matrix[3] = 0.70703125; | |
717 } | |
718 } | |
719 } else if (chgroup->num_channels > 2) { | |
720 if (get_bits1(&s->gb)) { | |
721 chgroup->transform = 1; | |
722 if (get_bits1(&s->gb)) { | |
723 decode_decorrelation_matrix(s, chgroup); | |
724 } else { | |
725 /** FIXME: more than 6 coupled channels not supported */ | |
726 if (chgroup->num_channels > 6) { | |
727 av_log_ask_for_sample(s->avctx, | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
728 "coupled channels > 6\n"); |
10122 | 729 } else { |
730 memcpy(chgroup->decorrelation_matrix, | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
731 default_decorrelation[chgroup->num_channels], |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
732 chgroup->num_channels * chgroup->num_channels * |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
733 sizeof(*chgroup->decorrelation_matrix)); |
10122 | 734 } |
735 } | |
736 } | |
737 } | |
738 | |
739 /** decode transform on / off */ | |
740 if (chgroup->transform) { | |
741 if (!get_bits1(&s->gb)) { | |
742 int i; | |
743 /** transform can be enabled for individual bands */ | |
744 for (i = 0; i < s->num_bands; i++) { | |
745 chgroup->transform_band[i] = get_bits1(&s->gb); | |
746 } | |
747 } else { | |
748 memset(chgroup->transform_band, 1, s->num_bands); | |
749 } | |
750 } | |
751 remaining_channels -= chgroup->num_channels; | |
752 } | |
753 } | |
754 return 0; | |
755 } | |
756 | |
757 /** | |
10097 | 758 *@brief Extract the coefficients from the bitstream. |
759 *@param s codec context | |
760 *@param c current channel number | |
761 *@return 0 on success, < 0 in case of bitstream errors | |
762 */ | |
10122 | 763 static int decode_coeffs(WMAProDecodeCtx *s, int c) |
10097 | 764 { |
10428 | 765 /* Integers 0..15 as single-precision floats. The table saves a |
766 costly int to float conversion, and storing the values as | |
767 integers allows fast sign-flipping. */ | |
768 static const int fval_tab[16] = { | |
769 0x00000000, 0x3f800000, 0x40000000, 0x40400000, | |
770 0x40800000, 0x40a00000, 0x40c00000, 0x40e00000, | |
771 0x41000000, 0x41100000, 0x41200000, 0x41300000, | |
772 0x41400000, 0x41500000, 0x41600000, 0x41700000, | |
773 }; | |
10097 | 774 int vlctable; |
775 VLC* vlc; | |
10122 | 776 WMAProChannelCtx* ci = &s->channel[c]; |
10097 | 777 int rl_mode = 0; |
778 int cur_coeff = 0; | |
779 int num_zeros = 0; | |
780 const uint16_t* run; | |
10314
ab687351bfef
WMA: store level_table as floats, use type punning for sign flip in decode
mru
parents:
10298
diff
changeset
|
781 const float* level; |
10097 | 782 |
783 dprintf(s->avctx, "decode coefficients for channel %i\n", c); | |
784 | |
785 vlctable = get_bits1(&s->gb); | |
786 vlc = &coef_vlc[vlctable]; | |
787 | |
788 if (vlctable) { | |
789 run = coef1_run; | |
790 level = coef1_level; | |
791 } else { | |
792 run = coef0_run; | |
793 level = coef0_level; | |
794 } | |
795 | |
796 /** decode vector coefficients (consumes up to 167 bits per iteration for | |
797 4 vector coded large values) */ | |
798 while (!rl_mode && cur_coeff + 3 < s->subframe_len) { | |
799 int vals[4]; | |
800 int i; | |
801 unsigned int idx; | |
802 | |
803 idx = get_vlc2(&s->gb, vec4_vlc.table, VLCBITS, VEC4MAXDEPTH); | |
804 | |
10101 | 805 if (idx == HUFF_VEC4_SIZE - 1) { |
10097 | 806 for (i = 0; i < 4; i += 2) { |
807 idx = get_vlc2(&s->gb, vec2_vlc.table, VLCBITS, VEC2MAXDEPTH); | |
10101 | 808 if (idx == HUFF_VEC2_SIZE - 1) { |
10428 | 809 int v0, v1; |
810 v0 = get_vlc2(&s->gb, vec1_vlc.table, VLCBITS, VEC1MAXDEPTH); | |
811 if (v0 == HUFF_VEC1_SIZE - 1) | |
812 v0 += ff_wma_get_large_val(&s->gb); | |
813 v1 = get_vlc2(&s->gb, vec1_vlc.table, VLCBITS, VEC1MAXDEPTH); | |
814 if (v1 == HUFF_VEC1_SIZE - 1) | |
815 v1 += ff_wma_get_large_val(&s->gb); | |
816 ((float*)vals)[i ] = v0; | |
817 ((float*)vals)[i+1] = v1; | |
10097 | 818 } else { |
10428 | 819 vals[i] = fval_tab[symbol_to_vec2[idx] >> 4 ]; |
820 vals[i+1] = fval_tab[symbol_to_vec2[idx] & 0xF]; | |
10097 | 821 } |
822 } | |
823 } else { | |
10428 | 824 vals[0] = fval_tab[ symbol_to_vec4[idx] >> 12 ]; |
825 vals[1] = fval_tab[(symbol_to_vec4[idx] >> 8) & 0xF]; | |
826 vals[2] = fval_tab[(symbol_to_vec4[idx] >> 4) & 0xF]; | |
827 vals[3] = fval_tab[ symbol_to_vec4[idx] & 0xF]; | |
10097 | 828 } |
829 | |
830 /** decode sign */ | |
831 for (i = 0; i < 4; i++) { | |
832 if (vals[i]) { | |
833 int sign = get_bits1(&s->gb) - 1; | |
10428 | 834 *(uint32_t*)&ci->coeffs[cur_coeff] = vals[i] ^ sign<<31; |
10097 | 835 num_zeros = 0; |
836 } else { | |
10122 | 837 ci->coeffs[cur_coeff] = 0; |
10097 | 838 /** switch to run level mode when subframe_len / 128 zeros |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
839 were found in a row */ |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
840 rl_mode |= (++num_zeros > s->subframe_len >> 8); |
10097 | 841 } |
842 ++cur_coeff; | |
843 } | |
844 } | |
845 | |
846 /** decode run level coded coefficients */ | |
847 if (rl_mode) { | |
10122 | 848 memset(&ci->coeffs[cur_coeff], 0, |
849 sizeof(*ci->coeffs) * (s->subframe_len - cur_coeff)); | |
10101 | 850 if (ff_wma_run_level_decode(s->avctx, &s->gb, vlc, |
851 level, run, 1, ci->coeffs, | |
852 cur_coeff, s->subframe_len, | |
853 s->subframe_len, s->esc_len, 0)) | |
10097 | 854 return AVERROR_INVALIDDATA; |
855 } | |
856 | |
857 return 0; | |
858 } | |
859 | |
860 /** | |
10122 | 861 *@brief Extract scale factors from the bitstream. |
862 *@param s codec context | |
863 *@return 0 on success, < 0 in case of bitstream errors | |
864 */ | |
865 static int decode_scale_factors(WMAProDecodeCtx* s) | |
866 { | |
867 int i; | |
868 | |
869 /** should never consume more than 5344 bits | |
870 * MAX_CHANNELS * (1 + MAX_BANDS * 23) | |
871 */ | |
872 | |
873 for (i = 0; i < s->channels_for_cur_subframe; i++) { | |
874 int c = s->channel_indexes_for_cur_subframe[i]; | |
875 int* sf; | |
10138 | 876 int* sf_end; |
877 s->channel[c].scale_factors = s->channel[c].saved_scale_factors[!s->channel[c].scale_factor_idx]; | |
878 sf_end = s->channel[c].scale_factors + s->num_bands; | |
10122 | 879 |
880 /** resample scale factors for the new block size | |
881 * as the scale factors might need to be resampled several times | |
882 * before some new values are transmitted, a backup of the last | |
883 * transmitted scale factors is kept in saved_scale_factors | |
884 */ | |
885 if (s->channel[c].reuse_sf) { | |
886 const int8_t* sf_offsets = s->sf_offsets[s->table_idx][s->channel[c].table_idx]; | |
887 int b; | |
888 for (b = 0; b < s->num_bands; b++) | |
889 s->channel[c].scale_factors[b] = | |
10138 | 890 s->channel[c].saved_scale_factors[s->channel[c].scale_factor_idx][*sf_offsets++]; |
10122 | 891 } |
892 | |
893 if (!s->channel[c].cur_subframe || get_bits1(&s->gb)) { | |
894 | |
895 if (!s->channel[c].reuse_sf) { | |
896 int val; | |
897 /** decode DPCM coded scale factors */ | |
898 s->channel[c].scale_factor_step = get_bits(&s->gb, 2) + 1; | |
899 val = 45 / s->channel[c].scale_factor_step; | |
900 for (sf = s->channel[c].scale_factors; sf < sf_end; sf++) { | |
901 val += get_vlc2(&s->gb, sf_vlc.table, SCALEVLCBITS, SCALEMAXDEPTH) - 60; | |
902 *sf = val; | |
903 } | |
904 } else { | |
905 int i; | |
906 /** run level decode differences to the resampled factors */ | |
907 for (i = 0; i < s->num_bands; i++) { | |
908 int idx; | |
909 int skip; | |
910 int val; | |
911 int sign; | |
912 | |
913 idx = get_vlc2(&s->gb, sf_rl_vlc.table, VLCBITS, SCALERLMAXDEPTH); | |
914 | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
915 if (!idx) { |
10122 | 916 uint32_t code = get_bits(&s->gb, 14); |
917 val = code >> 6; | |
918 sign = (code & 1) - 1; | |
919 skip = (code & 0x3f) >> 1; | |
920 } else if (idx == 1) { | |
921 break; | |
922 } else { | |
923 skip = scale_rl_run[idx]; | |
924 val = scale_rl_level[idx]; | |
925 sign = get_bits1(&s->gb)-1; | |
926 } | |
927 | |
928 i += skip; | |
929 if (i >= s->num_bands) { | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
930 av_log(s->avctx, AV_LOG_ERROR, |
10122 | 931 "invalid scale factor coding\n"); |
932 return AVERROR_INVALIDDATA; | |
933 } | |
934 s->channel[c].scale_factors[i] += (val ^ sign) - sign; | |
935 } | |
936 } | |
10138 | 937 /** swap buffers */ |
938 s->channel[c].scale_factor_idx = !s->channel[c].scale_factor_idx; | |
10122 | 939 s->channel[c].table_idx = s->table_idx; |
940 s->channel[c].reuse_sf = 1; | |
941 } | |
942 | |
943 /** calculate new scale factor maximum */ | |
944 s->channel[c].max_scale_factor = s->channel[c].scale_factors[0]; | |
945 for (sf = s->channel[c].scale_factors + 1; sf < sf_end; sf++) { | |
946 s->channel[c].max_scale_factor = | |
947 FFMAX(s->channel[c].max_scale_factor, *sf); | |
948 } | |
949 | |
950 } | |
951 return 0; | |
952 } | |
953 | |
954 /** | |
10005 | 955 *@brief Reconstruct the individual channel data. |
956 *@param s codec context | |
957 */ | |
10122 | 958 static void inverse_channel_transform(WMAProDecodeCtx *s) |
10005 | 959 { |
960 int i; | |
961 | |
10006 | 962 for (i = 0; i < s->num_chgroups; i++) { |
10096
071d6d272a77
merge 2-channel M/S stereo decoding code with the multichannel version
faust3
parents:
10084
diff
changeset
|
963 if (s->chgroup[i].transform) { |
10005 | 964 float data[WMAPRO_MAX_CHANNELS]; |
965 const int num_channels = s->chgroup[i].num_channels; | |
966 float** ch_data = s->chgroup[i].channel_data; | |
967 float** ch_end = ch_data + num_channels; | |
968 const int8_t* tb = s->chgroup[i].transform_band; | |
969 int16_t* sfb; | |
970 | |
971 /** multichannel decorrelation */ | |
10084 | 972 for (sfb = s->cur_sfb_offsets; |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
973 sfb < s->cur_sfb_offsets + s->num_bands; sfb++) { |
10096
071d6d272a77
merge 2-channel M/S stereo decoding code with the multichannel version
faust3
parents:
10084
diff
changeset
|
974 int y; |
10005 | 975 if (*tb++ == 1) { |
976 /** multiply values with the decorrelation_matrix */ | |
10006 | 977 for (y = sfb[0]; y < FFMIN(sfb[1], s->subframe_len); y++) { |
10005 | 978 const float* mat = s->chgroup[i].decorrelation_matrix; |
10006 | 979 const float* data_end = data + num_channels; |
980 float* data_ptr = data; | |
10005 | 981 float** ch; |
982 | |
10084 | 983 for (ch = ch_data; ch < ch_end; ch++) |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
984 *data_ptr++ = (*ch)[y]; |
10005 | 985 |
986 for (ch = ch_data; ch < ch_end; ch++) { | |
987 float sum = 0; | |
988 data_ptr = data; | |
989 while (data_ptr < data_end) | |
990 sum += *data_ptr++ * *mat++; | |
991 | |
992 (*ch)[y] = sum; | |
993 } | |
994 } | |
10096
071d6d272a77
merge 2-channel M/S stereo decoding code with the multichannel version
faust3
parents:
10084
diff
changeset
|
995 } else if (s->num_channels == 2) { |
10298
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
996 int len = FFMIN(sfb[1], s->subframe_len) - sfb[0]; |
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
997 s->dsp.vector_fmul_scalar(ch_data[0] + sfb[0], |
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
998 ch_data[0] + sfb[0], |
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
999 181.0 / 128, len); |
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
1000 s->dsp.vector_fmul_scalar(ch_data[1] + sfb[0], |
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
1001 ch_data[1] + sfb[0], |
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
1002 181.0 / 128, len); |
10005 | 1003 } |
1004 } | |
1005 } | |
1006 } | |
1007 } | |
1008 | |
10122 | 1009 /** |
1010 *@brief Apply sine window and reconstruct the output buffer. | |
1011 *@param s codec context | |
1012 */ | |
1013 static void wmapro_window(WMAProDecodeCtx *s) | |
1014 { | |
1015 int i; | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1016 for (i = 0; i < s->channels_for_cur_subframe; i++) { |
10122 | 1017 int c = s->channel_indexes_for_cur_subframe[i]; |
1018 float* window; | |
1019 int winlen = s->channel[c].prev_block_len; | |
1020 float* start = s->channel[c].coeffs - (winlen >> 1); | |
1021 | |
1022 if (s->subframe_len < winlen) { | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1023 start += (winlen - s->subframe_len) >> 1; |
10122 | 1024 winlen = s->subframe_len; |
1025 } | |
1026 | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1027 window = s->windows[av_log2(winlen) - BLOCK_MIN_BITS]; |
10122 | 1028 |
1029 winlen >>= 1; | |
1030 | |
1031 s->dsp.vector_fmul_window(start, start, start + winlen, | |
1032 window, 0, winlen); | |
1033 | |
1034 s->channel[c].prev_block_len = s->subframe_len; | |
1035 } | |
1036 } | |
1037 | |
1038 /** | |
1039 *@brief Decode a single subframe (block). | |
1040 *@param s codec context | |
1041 *@return 0 on success, < 0 when decoding failed | |
1042 */ | |
1043 static int decode_subframe(WMAProDecodeCtx *s) | |
1044 { | |
1045 int offset = s->samples_per_frame; | |
1046 int subframe_len = s->samples_per_frame; | |
1047 int i; | |
1048 int total_samples = s->samples_per_frame * s->num_channels; | |
1049 int transmit_coeffs = 0; | |
1050 int cur_subwoofer_cutoff; | |
1051 | |
1052 s->subframe_offset = get_bits_count(&s->gb); | |
1053 | |
1054 /** reset channel context and find the next block offset and size | |
1055 == the next block of the channel with the smallest number of | |
1056 decoded samples | |
1057 */ | |
1058 for (i = 0; i < s->num_channels; i++) { | |
1059 s->channel[i].grouped = 0; | |
1060 if (offset > s->channel[i].decoded_samples) { | |
1061 offset = s->channel[i].decoded_samples; | |
1062 subframe_len = | |
1063 s->channel[i].subframe_len[s->channel[i].cur_subframe]; | |
1064 } | |
1065 } | |
1066 | |
1067 dprintf(s->avctx, | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1068 "processing subframe with offset %i len %i\n", offset, subframe_len); |
10122 | 1069 |
1070 /** get a list of all channels that contain the estimated block */ | |
1071 s->channels_for_cur_subframe = 0; | |
1072 for (i = 0; i < s->num_channels; i++) { | |
1073 const int cur_subframe = s->channel[i].cur_subframe; | |
1074 /** substract already processed samples */ | |
1075 total_samples -= s->channel[i].decoded_samples; | |
1076 | |
1077 /** and count if there are multiple subframes that match our profile */ | |
1078 if (offset == s->channel[i].decoded_samples && | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1079 subframe_len == s->channel[i].subframe_len[cur_subframe]) { |
10122 | 1080 total_samples -= s->channel[i].subframe_len[cur_subframe]; |
1081 s->channel[i].decoded_samples += | |
1082 s->channel[i].subframe_len[cur_subframe]; | |
1083 s->channel_indexes_for_cur_subframe[s->channels_for_cur_subframe] = i; | |
1084 ++s->channels_for_cur_subframe; | |
1085 } | |
1086 } | |
1087 | |
1088 /** check if the frame will be complete after processing the | |
1089 estimated block */ | |
1090 if (!total_samples) | |
1091 s->parsed_all_subframes = 1; | |
1092 | |
1093 | |
1094 dprintf(s->avctx, "subframe is part of %i channels\n", | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1095 s->channels_for_cur_subframe); |
10122 | 1096 |
1097 /** calculate number of scale factor bands and their offsets */ | |
1098 s->table_idx = av_log2(s->samples_per_frame/subframe_len); | |
1099 s->num_bands = s->num_sfb[s->table_idx]; | |
1100 s->cur_sfb_offsets = s->sfb_offsets[s->table_idx]; | |
1101 cur_subwoofer_cutoff = s->subwoofer_cutoffs[s->table_idx]; | |
1102 | |
1103 /** configure the decoder for the current subframe */ | |
1104 for (i = 0; i < s->channels_for_cur_subframe; i++) { | |
1105 int c = s->channel_indexes_for_cur_subframe[i]; | |
1106 | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1107 s->channel[c].coeffs = &s->channel[c].out[(s->samples_per_frame >> 1) |
10122 | 1108 + offset]; |
1109 } | |
1110 | |
1111 s->subframe_len = subframe_len; | |
1112 s->esc_len = av_log2(s->subframe_len - 1) + 1; | |
1113 | |
1114 /** skip extended header if any */ | |
1115 if (get_bits1(&s->gb)) { | |
1116 int num_fill_bits; | |
1117 if (!(num_fill_bits = get_bits(&s->gb, 2))) { | |
1118 int len = get_bits(&s->gb, 4); | |
1119 num_fill_bits = get_bits(&s->gb, len) + 1; | |
1120 } | |
1121 | |
1122 if (num_fill_bits >= 0) { | |
1123 if (get_bits_count(&s->gb) + num_fill_bits > s->num_saved_bits) { | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1124 av_log(s->avctx, AV_LOG_ERROR, "invalid number of fill bits\n"); |
10122 | 1125 return AVERROR_INVALIDDATA; |
1126 } | |
1127 | |
1128 skip_bits_long(&s->gb, num_fill_bits); | |
1129 } | |
1130 } | |
1131 | |
1132 /** no idea for what the following bit is used */ | |
1133 if (get_bits1(&s->gb)) { | |
1134 av_log_ask_for_sample(s->avctx, "reserved bit set\n"); | |
1135 return AVERROR_INVALIDDATA; | |
1136 } | |
1137 | |
1138 | |
1139 if (decode_channel_transform(s) < 0) | |
1140 return AVERROR_INVALIDDATA; | |
1141 | |
1142 | |
1143 for (i = 0; i < s->channels_for_cur_subframe; i++) { | |
1144 int c = s->channel_indexes_for_cur_subframe[i]; | |
1145 if ((s->channel[c].transmit_coefs = get_bits1(&s->gb))) | |
1146 transmit_coeffs = 1; | |
1147 } | |
1148 | |
1149 if (transmit_coeffs) { | |
1150 int step; | |
1151 int quant_step = 90 * s->bits_per_sample >> 4; | |
1152 if ((get_bits1(&s->gb))) { | |
1153 /** FIXME: might change run level mode decision */ | |
1154 av_log_ask_for_sample(s->avctx, "unsupported quant step coding\n"); | |
1155 return AVERROR_INVALIDDATA; | |
1156 } | |
1157 /** decode quantization step */ | |
1158 step = get_sbits(&s->gb, 6); | |
1159 quant_step += step; | |
1160 if (step == -32 || step == 31) { | |
1161 const int sign = (step == 31) - 1; | |
1162 int quant = 0; | |
1163 while (get_bits_count(&s->gb) + 5 < s->num_saved_bits && | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1164 (step = get_bits(&s->gb, 5)) == 31) { |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1165 quant += 31; |
10122 | 1166 } |
1167 quant_step += ((quant + step) ^ sign) - sign; | |
1168 } | |
1169 if (quant_step < 0) { | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1170 av_log(s->avctx, AV_LOG_DEBUG, "negative quant step\n"); |
10122 | 1171 } |
1172 | |
1173 /** decode quantization step modifiers for every channel */ | |
1174 | |
1175 if (s->channels_for_cur_subframe == 1) { | |
1176 s->channel[s->channel_indexes_for_cur_subframe[0]].quant_step = quant_step; | |
1177 } else { | |
1178 int modifier_len = get_bits(&s->gb, 3); | |
1179 for (i = 0; i < s->channels_for_cur_subframe; i++) { | |
1180 int c = s->channel_indexes_for_cur_subframe[i]; | |
1181 s->channel[c].quant_step = quant_step; | |
1182 if (get_bits1(&s->gb)) { | |
1183 if (modifier_len) { | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1184 s->channel[c].quant_step += get_bits(&s->gb, modifier_len) + 1; |
10122 | 1185 } else |
1186 ++s->channel[c].quant_step; | |
1187 } | |
1188 } | |
1189 } | |
1190 | |
1191 /** decode scale factors */ | |
1192 if (decode_scale_factors(s) < 0) | |
1193 return AVERROR_INVALIDDATA; | |
1194 } | |
1195 | |
1196 dprintf(s->avctx, "BITSTREAM: subframe header length was %i\n", | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1197 get_bits_count(&s->gb) - s->subframe_offset); |
10122 | 1198 |
1199 /** parse coefficients */ | |
1200 for (i = 0; i < s->channels_for_cur_subframe; i++) { | |
1201 int c = s->channel_indexes_for_cur_subframe[i]; | |
1202 if (s->channel[c].transmit_coefs && | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1203 get_bits_count(&s->gb) < s->num_saved_bits) { |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1204 decode_coeffs(s, c); |
10122 | 1205 } else |
1206 memset(s->channel[c].coeffs, 0, | |
1207 sizeof(*s->channel[c].coeffs) * subframe_len); | |
1208 } | |
1209 | |
1210 dprintf(s->avctx, "BITSTREAM: subframe length was %i\n", | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1211 get_bits_count(&s->gb) - s->subframe_offset); |
10122 | 1212 |
1213 if (transmit_coeffs) { | |
1214 /** reconstruct the per channel data */ | |
1215 inverse_channel_transform(s); | |
1216 for (i = 0; i < s->channels_for_cur_subframe; i++) { | |
1217 int c = s->channel_indexes_for_cur_subframe[i]; | |
1218 const int* sf = s->channel[c].scale_factors; | |
1219 int b; | |
1220 | |
1221 if (c == s->lfe_channel) | |
1222 memset(&s->tmp[cur_subwoofer_cutoff], 0, sizeof(*s->tmp) * | |
1223 (subframe_len - cur_subwoofer_cutoff)); | |
1224 | |
1225 /** inverse quantization and rescaling */ | |
1226 for (b = 0; b < s->num_bands; b++) { | |
1227 const int end = FFMIN(s->cur_sfb_offsets[b+1], s->subframe_len); | |
1228 const int exp = s->channel[c].quant_step - | |
1229 (s->channel[c].max_scale_factor - *sf++) * | |
1230 s->channel[c].scale_factor_step; | |
1231 const float quant = pow(10.0, exp / 20.0); | |
10298
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
1232 int start = s->cur_sfb_offsets[b]; |
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
1233 s->dsp.vector_fmul_scalar(s->tmp + start, |
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
1234 s->channel[c].coeffs + start, |
9699e4499f2d
WMAPRO: use vector_fmul_scalar from dsputil where possible
mru
parents:
10199
diff
changeset
|
1235 quant, end - start); |
10122 | 1236 } |
1237 | |
1238 /** apply imdct (ff_imdct_half == DCTIV with reverse) */ | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1239 ff_imdct_half(&s->mdct_ctx[av_log2(subframe_len) - BLOCK_MIN_BITS], |
10122 | 1240 s->channel[c].coeffs, s->tmp); |
1241 } | |
1242 } | |
1243 | |
1244 /** window and overlapp-add */ | |
1245 wmapro_window(s); | |
1246 | |
1247 /** handled one subframe */ | |
1248 for (i = 0; i < s->channels_for_cur_subframe; i++) { | |
1249 int c = s->channel_indexes_for_cur_subframe[i]; | |
1250 if (s->channel[c].cur_subframe >= s->channel[c].num_subframes) { | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1251 av_log(s->avctx, AV_LOG_ERROR, "broken subframe\n"); |
10122 | 1252 return AVERROR_INVALIDDATA; |
1253 } | |
1254 ++s->channel[c].cur_subframe; | |
1255 } | |
1256 | |
1257 return 0; | |
1258 } | |
1259 | |
1260 /** | |
1261 *@brief Decode one WMA frame. | |
1262 *@param s codec context | |
1263 *@return 0 if the trailer bit indicates that this is the last frame, | |
1264 * 1 if there are additional frames | |
1265 */ | |
1266 static int decode_frame(WMAProDecodeCtx *s) | |
1267 { | |
1268 GetBitContext* gb = &s->gb; | |
1269 int more_frames = 0; | |
1270 int len = 0; | |
1271 int i; | |
1272 | |
1273 /** check for potential output buffer overflow */ | |
1274 if (s->num_channels * s->samples_per_frame > s->samples_end - s->samples) { | |
10127 | 1275 /** return an error if no frame could be decoded at all */ |
10139
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1276 av_log(s->avctx, AV_LOG_ERROR, |
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1277 "not enough space for the output samples\n"); |
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1278 s->packet_loss = 1; |
10122 | 1279 return 0; |
1280 } | |
1281 | |
1282 /** get frame length */ | |
1283 if (s->len_prefix) | |
1284 len = get_bits(gb, s->log2_frame_size); | |
1285 | |
1286 dprintf(s->avctx, "decoding frame with length %x\n", len); | |
1287 | |
1288 /** decode tile information */ | |
1289 if (decode_tilehdr(s)) { | |
1290 s->packet_loss = 1; | |
1291 return 0; | |
1292 } | |
1293 | |
1294 /** read postproc transform */ | |
1295 if (s->num_channels > 1 && get_bits1(gb)) { | |
1296 av_log_ask_for_sample(s->avctx, "Unsupported postproc transform found\n"); | |
1297 s->packet_loss = 1; | |
1298 return 0; | |
1299 } | |
1300 | |
1301 /** read drc info */ | |
1302 if (s->dynamic_range_compression) { | |
1303 s->drc_gain = get_bits(gb, 8); | |
1304 dprintf(s->avctx, "drc_gain %i\n", s->drc_gain); | |
1305 } | |
1306 | |
1307 /** no idea what these are for, might be the number of samples | |
1308 that need to be skipped at the beginning or end of a stream */ | |
1309 if (get_bits1(gb)) { | |
1310 int skip; | |
1311 | |
1312 /** usually true for the first frame */ | |
1313 if (get_bits1(gb)) { | |
1314 skip = get_bits(gb, av_log2(s->samples_per_frame * 2)); | |
1315 dprintf(s->avctx, "start skip: %i\n", skip); | |
1316 } | |
1317 | |
1318 /** sometimes true for the last frame */ | |
1319 if (get_bits1(gb)) { | |
1320 skip = get_bits(gb, av_log2(s->samples_per_frame * 2)); | |
1321 dprintf(s->avctx, "end skip: %i\n", skip); | |
1322 } | |
1323 | |
1324 } | |
1325 | |
1326 dprintf(s->avctx, "BITSTREAM: frame header length was %i\n", | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1327 get_bits_count(gb) - s->frame_offset); |
10122 | 1328 |
1329 /** reset subframe states */ | |
1330 s->parsed_all_subframes = 0; | |
1331 for (i = 0; i < s->num_channels; i++) { | |
1332 s->channel[i].decoded_samples = 0; | |
1333 s->channel[i].cur_subframe = 0; | |
1334 s->channel[i].reuse_sf = 0; | |
1335 } | |
1336 | |
1337 /** decode all subframes */ | |
1338 while (!s->parsed_all_subframes) { | |
1339 if (decode_subframe(s) < 0) { | |
1340 s->packet_loss = 1; | |
1341 return 0; | |
1342 } | |
1343 } | |
1344 | |
1345 /** interleave samples and write them to the output buffer */ | |
1346 for (i = 0; i < s->num_channels; i++) { | |
1347 float* ptr; | |
1348 int incr = s->num_channels; | |
1349 float* iptr = s->channel[i].out; | |
1350 int x; | |
1351 | |
1352 ptr = s->samples + i; | |
1353 | |
1354 for (x = 0; x < s->samples_per_frame; x++) { | |
1355 *ptr = av_clipf(*iptr++, -1.0, 32767.0 / 32768.0); | |
1356 ptr += incr; | |
1357 } | |
1358 | |
1359 /** reuse second half of the IMDCT output for the next frame */ | |
1360 memcpy(&s->channel[i].out[0], | |
1361 &s->channel[i].out[s->samples_per_frame], | |
1362 s->samples_per_frame * sizeof(*s->channel[i].out) >> 1); | |
1363 } | |
1364 | |
1365 if (s->skip_frame) { | |
1366 s->skip_frame = 0; | |
1367 } else | |
1368 s->samples += s->num_channels * s->samples_per_frame; | |
1369 | |
1370 if (len != (get_bits_count(gb) - s->frame_offset) + 2) { | |
1371 /** FIXME: not sure if this is always an error */ | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1372 av_log(s->avctx, AV_LOG_ERROR, "frame[%i] would have to skip %i bits\n", |
10122 | 1373 s->frame_num, len - (get_bits_count(gb) - s->frame_offset) - 1); |
1374 s->packet_loss = 1; | |
1375 return 0; | |
1376 } | |
1377 | |
1378 /** skip the rest of the frame data */ | |
1379 skip_bits_long(gb, len - (get_bits_count(gb) - s->frame_offset) - 1); | |
1380 | |
1381 /** decode trailer bit */ | |
1382 more_frames = get_bits1(gb); | |
1383 | |
1384 ++s->frame_num; | |
1385 return more_frames; | |
1386 } | |
1387 | |
1388 /** | |
1389 *@brief Calculate remaining input buffer length. | |
1390 *@param s codec context | |
1391 *@param gb bitstream reader context | |
1392 *@return remaining size in bits | |
1393 */ | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1394 static int remaining_bits(WMAProDecodeCtx *s, GetBitContext *gb) |
10122 | 1395 { |
1396 return s->buf_bit_size - get_bits_count(gb); | |
1397 } | |
1398 | |
1399 /** | |
1400 *@brief Fill the bit reservoir with a (partial) frame. | |
1401 *@param s codec context | |
1402 *@param gb bitstream reader context | |
1403 *@param len length of the partial frame | |
1404 *@param append decides wether to reset the buffer or not | |
1405 */ | |
1406 static void save_bits(WMAProDecodeCtx *s, GetBitContext* gb, int len, | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1407 int append) |
10122 | 1408 { |
1409 int buflen; | |
1410 | |
1411 /** when the frame data does not need to be concatenated, the input buffer | |
1412 is resetted and additional bits from the previous frame are copyed | |
1413 and skipped later so that a fast byte copy is possible */ | |
1414 | |
1415 if (!append) { | |
1416 s->frame_offset = get_bits_count(gb) & 7; | |
1417 s->num_saved_bits = s->frame_offset; | |
1418 init_put_bits(&s->pb, s->frame_data, MAX_FRAMESIZE); | |
1419 } | |
1420 | |
1421 buflen = (s->num_saved_bits + len + 8) >> 3; | |
1422 | |
1423 if (len <= 0 || buflen > MAX_FRAMESIZE) { | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1424 av_log_ask_for_sample(s->avctx, "input buffer too small\n"); |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1425 s->packet_loss = 1; |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1426 return; |
10122 | 1427 } |
1428 | |
1429 s->num_saved_bits += len; | |
1430 if (!append) { | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1431 ff_copy_bits(&s->pb, gb->buffer + (get_bits_count(gb) >> 3), |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1432 s->num_saved_bits); |
10122 | 1433 } else { |
1434 int align = 8 - (get_bits_count(gb) & 7); | |
1435 align = FFMIN(align, len); | |
1436 put_bits(&s->pb, align, get_bits(gb, align)); | |
1437 len -= align; | |
1438 ff_copy_bits(&s->pb, gb->buffer + (get_bits_count(gb) >> 3), len); | |
1439 } | |
1440 skip_bits_long(gb, len); | |
1441 | |
1442 { | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1443 PutBitContext tmp = s->pb; |
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1444 flush_put_bits(&tmp); |
10122 | 1445 } |
1446 | |
1447 init_get_bits(&s->gb, s->frame_data, s->num_saved_bits); | |
1448 skip_bits(&s->gb, s->frame_offset); | |
1449 } | |
1450 | |
1451 /** | |
1452 *@brief Decode a single WMA packet. | |
1453 *@param avctx codec context | |
1454 *@param data the output buffer | |
1455 *@param data_size number of bytes that were written to the output buffer | |
1456 *@param avpkt input packet | |
1457 *@return number of bytes that were read from the input buffer | |
1458 */ | |
1459 static int decode_packet(AVCodecContext *avctx, | |
10123
e1bd75a1eab2
cosmetics: indentation and other prettyprinting fixes
diego
parents:
10122
diff
changeset
|
1460 void *data, int *data_size, AVPacket* avpkt) |
10122 | 1461 { |
1462 WMAProDecodeCtx *s = avctx->priv_data; | |
10129 | 1463 GetBitContext* gb = &s->pgb; |
1464 const uint8_t* buf = avpkt->data; | |
1465 int buf_size = avpkt->size; | |
10122 | 1466 int num_bits_prev_frame; |
1467 int packet_sequence_number; | |
1468 | |
10129 | 1469 s->samples = data; |
1470 s->samples_end = (float*)((int8_t*)data + *data_size); | |
10127 | 1471 *data_size = 0; |
10122 | 1472 |
10139
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1473 if (s->packet_done || s->packet_loss) { |
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1474 s->packet_done = 0; |
10129 | 1475 s->buf_bit_size = buf_size << 3; |
10122 | 1476 |
10129 | 1477 /** sanity check for the buffer length */ |
1478 if (buf_size < avctx->block_align) | |
1479 return 0; | |
10122 | 1480 |
10129 | 1481 buf_size = avctx->block_align; |
10122 | 1482 |
10129 | 1483 /** parse packet header */ |
1484 init_get_bits(gb, buf, s->buf_bit_size); | |
1485 packet_sequence_number = get_bits(gb, 4); | |
1486 skip_bits(gb, 2); | |
10122 | 1487 |
10129 | 1488 /** get number of bits that need to be added to the previous frame */ |
1489 num_bits_prev_frame = get_bits(gb, s->log2_frame_size); | |
1490 dprintf(avctx, "packet[%d]: nbpf %x\n", avctx->frame_number, | |
1491 num_bits_prev_frame); | |
10122 | 1492 |
10129 | 1493 /** check for packet loss */ |
1494 if (!s->packet_loss && | |
1495 ((s->packet_sequence_number + 1) & 0xF) != packet_sequence_number) { | |
1496 s->packet_loss = 1; | |
1497 av_log(avctx, AV_LOG_ERROR, "Packet loss detected! seq %x vs %x\n", | |
1498 s->packet_sequence_number, packet_sequence_number); | |
1499 } | |
1500 s->packet_sequence_number = packet_sequence_number; | |
10122 | 1501 |
10129 | 1502 if (num_bits_prev_frame > 0) { |
1503 /** append the previous frame data to the remaining data from the | |
1504 previous packet to create a full frame */ | |
1505 save_bits(s, gb, num_bits_prev_frame, 1); | |
1506 dprintf(avctx, "accumulated %x bits of frame data\n", | |
1507 s->num_saved_bits - s->frame_offset); | |
10122 | 1508 |
10129 | 1509 /** decode the cross packet frame if it is valid */ |
1510 if (!s->packet_loss) | |
1511 decode_frame(s); | |
1512 } else if (s->num_saved_bits - s->frame_offset) { | |
1513 dprintf(avctx, "ignoring %x previously saved bits\n", | |
1514 s->num_saved_bits - s->frame_offset); | |
1515 } | |
10122 | 1516 |
10129 | 1517 s->packet_loss = 0; |
10127 | 1518 |
1519 } else { | |
10139
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1520 int frame_size; |
10165 | 1521 s->buf_bit_size = avpkt->size << 3; |
1522 init_get_bits(gb, avpkt->data, s->buf_bit_size); | |
1523 skip_bits(gb, s->packet_offset); | |
10139
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1524 if (remaining_bits(s, gb) > s->log2_frame_size && |
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1525 (frame_size = show_bits(gb, s->log2_frame_size)) && |
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1526 frame_size <= remaining_bits(s, gb)) { |
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1527 save_bits(s, gb, frame_size, 0); |
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1528 s->packet_done = !decode_frame(s); |
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1529 } else |
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1530 s->packet_done = 1; |
10127 | 1531 } |
1532 | |
10139
de9335d4b6b1
do not decode more than one audio frame in a decode packet call
faust3
parents:
10138
diff
changeset
|
1533 if (s->packet_done && !s->packet_loss && |
10127 | 1534 remaining_bits(s, gb) > 0) { |
10122 | 1535 /** save the rest of the data so that it can be decoded |
1536 with the next packet */ | |
10126 | 1537 save_bits(s, gb, remaining_bits(s, gb), 0); |
10122 | 1538 } |
1539 | |
1540 *data_size = (int8_t *)s->samples - (int8_t *)data; | |
10165 | 1541 s->packet_offset = get_bits_count(gb) & 7; |
10122 | 1542 |
10352
8eb51d3eccdf
return AVERROR_INVALIDDATA when the bitstream could not be decoded
faust3
parents:
10314
diff
changeset
|
1543 return (s->packet_loss) ? AVERROR_INVALIDDATA : get_bits_count(gb) >> 3; |
10122 | 1544 } |
1545 | |
1546 /** | |
1547 *@brief Clear decoder buffers (for seeking). | |
1548 *@param avctx codec context | |
1549 */ | |
1550 static void flush(AVCodecContext *avctx) | |
1551 { | |
1552 WMAProDecodeCtx *s = avctx->priv_data; | |
1553 int i; | |
1554 /** reset output buffer as a part of it is used during the windowing of a | |
1555 new frame */ | |
1556 for (i = 0; i < s->num_channels; i++) | |
1557 memset(s->channel[i].out, 0, s->samples_per_frame * | |
1558 sizeof(*s->channel[i].out)); | |
1559 s->packet_loss = 1; | |
1560 } | |
1561 | |
1562 | |
1563 /** | |
1564 *@brief wmapro decoder | |
1565 */ | |
1566 AVCodec wmapro_decoder = { | |
1567 "wmapro", | |
1568 CODEC_TYPE_AUDIO, | |
1569 CODEC_ID_WMAPRO, | |
1570 sizeof(WMAProDecodeCtx), | |
1571 decode_init, | |
1572 NULL, | |
1573 decode_end, | |
1574 decode_packet, | |
10180
a514a601bf26
Add CODEC_CAP_SUBFRAMES for codecs that output multiple subframes
faust3
parents:
10174
diff
changeset
|
1575 .capabilities = CODEC_CAP_SUBFRAMES, |
10122 | 1576 .flush= flush, |
1577 .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 9 Professional"), | |
1578 }; |