Line | Branch | Exec | Source |
---|---|---|---|
1 | /* | ||
2 | * Wmapro compatible decoder | ||
3 | * Copyright (c) 2007 Baptiste Coudurier, Benjamin Larsson, Ulion | ||
4 | * Copyright (c) 2008 - 2011 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 | ||
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 | |||
89 | #include <inttypes.h> | ||
90 | |||
91 | #include "libavutil/audio_fifo.h" | ||
92 | #include "libavutil/mem.h" | ||
93 | #include "libavutil/tx.h" | ||
94 | #include "libavutil/ffmath.h" | ||
95 | #include "libavutil/float_dsp.h" | ||
96 | #include "libavutil/intfloat.h" | ||
97 | #include "libavutil/intreadwrite.h" | ||
98 | #include "libavutil/mem_internal.h" | ||
99 | #include "libavutil/thread.h" | ||
100 | |||
101 | #include "avcodec.h" | ||
102 | #include "codec_internal.h" | ||
103 | #include "decode.h" | ||
104 | #include "get_bits.h" | ||
105 | #include "internal.h" | ||
106 | #include "put_bits.h" | ||
107 | #include "wmaprodata.h" | ||
108 | #include "sinewin.h" | ||
109 | #include "wma.h" | ||
110 | #include "wma_common.h" | ||
111 | |||
112 | /** current decoder limitations */ | ||
113 | #define WMAPRO_MAX_CHANNELS 8 ///< max number of handled channels | ||
114 | #define MAX_SUBFRAMES 32 ///< max number of subframes per channel | ||
115 | #define MAX_BANDS 29 ///< max number of scale factor bands | ||
116 | #define MAX_FRAMESIZE 32768 ///< maximum compressed frame size | ||
117 | #define XMA_MAX_STREAMS 8 | ||
118 | #define XMA_MAX_CHANNELS_STREAM 2 | ||
119 | #define XMA_MAX_CHANNELS (XMA_MAX_STREAMS * XMA_MAX_CHANNELS_STREAM) | ||
120 | |||
121 | #define WMAPRO_BLOCK_MIN_BITS 6 ///< log2 of min block size | ||
122 | #define WMAPRO_BLOCK_MAX_BITS 13 ///< log2 of max block size | ||
123 | #define WMAPRO_BLOCK_MIN_SIZE (1 << WMAPRO_BLOCK_MIN_BITS) ///< minimum block size | ||
124 | #define WMAPRO_BLOCK_MAX_SIZE (1 << WMAPRO_BLOCK_MAX_BITS) ///< maximum block size | ||
125 | #define WMAPRO_BLOCK_SIZES (WMAPRO_BLOCK_MAX_BITS - WMAPRO_BLOCK_MIN_BITS + 1) ///< possible block sizes | ||
126 | |||
127 | |||
128 | #define VLCBITS 9 | ||
129 | #define SCALEVLCBITS 8 | ||
130 | #define VEC4MAXDEPTH ((HUFF_VEC4_MAXBITS+VLCBITS-1)/VLCBITS) | ||
131 | #define VEC2MAXDEPTH ((HUFF_VEC2_MAXBITS+VLCBITS-1)/VLCBITS) | ||
132 | #define VEC1MAXDEPTH ((HUFF_VEC1_MAXBITS+VLCBITS-1)/VLCBITS) | ||
133 | #define SCALEMAXDEPTH ((HUFF_SCALE_MAXBITS+SCALEVLCBITS-1)/SCALEVLCBITS) | ||
134 | #define SCALERLMAXDEPTH ((HUFF_SCALE_RL_MAXBITS+VLCBITS-1)/VLCBITS) | ||
135 | |||
136 | static VLCElem sf_vlc[616]; ///< scale factor DPCM vlc | ||
137 | static VLCElem sf_rl_vlc[1406]; ///< scale factor run length vlc | ||
138 | static VLCElem vec4_vlc[604]; ///< 4 coefficients per symbol | ||
139 | static VLCElem vec2_vlc[562]; ///< 2 coefficients per symbol | ||
140 | static VLCElem vec1_vlc[562]; ///< 1 coefficient per symbol | ||
141 | static const VLCElem *coef_vlc[2]; ///< coefficient run length vlc codes | ||
142 | static float sin64[33]; ///< sine table for decorrelation | ||
143 | |||
144 | /** | ||
145 | * @brief frame specific decoder context for a single channel | ||
146 | */ | ||
147 | typedef struct WMAProChannelCtx { | ||
148 | int16_t prev_block_len; ///< length of the previous block | ||
149 | uint8_t transmit_coefs; | ||
150 | uint8_t num_subframes; | ||
151 | uint16_t subframe_len[MAX_SUBFRAMES]; ///< subframe length in samples | ||
152 | uint16_t subframe_offset[MAX_SUBFRAMES]; ///< subframe positions in the current frame | ||
153 | uint8_t cur_subframe; ///< current subframe number | ||
154 | uint16_t decoded_samples; ///< number of already processed samples | ||
155 | uint8_t grouped; ///< channel is part of a group | ||
156 | int quant_step; ///< quantization step for the current subframe | ||
157 | int8_t reuse_sf; ///< share scale factors between subframes | ||
158 | int8_t scale_factor_step; ///< scaling step for the current subframe | ||
159 | int max_scale_factor; ///< maximum scale factor for the current subframe | ||
160 | int saved_scale_factors[2][MAX_BANDS]; ///< resampled and (previously) transmitted scale factor values | ||
161 | int8_t scale_factor_idx; ///< index for the transmitted scale factor values (used for resampling) | ||
162 | int* scale_factors; ///< pointer to the scale factor values used for decoding | ||
163 | uint8_t table_idx; ///< index in sf_offsets for the scale factor reference block | ||
164 | float* coeffs; ///< pointer to the subframe decode buffer | ||
165 | uint16_t num_vec_coeffs; ///< number of vector coded coefficients | ||
166 | DECLARE_ALIGNED(32, float, out)[WMAPRO_BLOCK_MAX_SIZE + WMAPRO_BLOCK_MAX_SIZE / 2]; ///< output buffer | ||
167 | } WMAProChannelCtx; | ||
168 | |||
169 | /** | ||
170 | * @brief channel group for channel transformations | ||
171 | */ | ||
172 | typedef struct WMAProChannelGrp { | ||
173 | uint8_t num_channels; ///< number of channels in the group | ||
174 | int8_t transform; ///< transform on / off | ||
175 | int8_t transform_band[MAX_BANDS]; ///< controls if the transform is enabled for a certain band | ||
176 | float decorrelation_matrix[WMAPRO_MAX_CHANNELS*WMAPRO_MAX_CHANNELS]; | ||
177 | float* channel_data[WMAPRO_MAX_CHANNELS]; ///< transformation coefficients | ||
178 | } WMAProChannelGrp; | ||
179 | |||
180 | /** | ||
181 | * @brief main decoder context | ||
182 | */ | ||
183 | typedef struct WMAProDecodeCtx { | ||
184 | /* generic decoder variables */ | ||
185 | AVCodecContext* avctx; ///< codec context for av_log | ||
186 | AVFloatDSPContext *fdsp; | ||
187 | uint8_t frame_data[MAX_FRAMESIZE + | ||
188 | AV_INPUT_BUFFER_PADDING_SIZE];///< compressed frame data | ||
189 | PutBitContext pb; ///< context for filling the frame_data buffer | ||
190 | AVTXContext *tx[WMAPRO_BLOCK_SIZES]; ///< MDCT context per block size | ||
191 | av_tx_fn tx_fn[WMAPRO_BLOCK_SIZES]; | ||
192 | DECLARE_ALIGNED(32, float, tmp)[WMAPRO_BLOCK_MAX_SIZE]; ///< IMDCT output buffer | ||
193 | const float* windows[WMAPRO_BLOCK_SIZES]; ///< windows for the different block sizes | ||
194 | |||
195 | /* frame size dependent frame information (set during initialization) */ | ||
196 | uint32_t decode_flags; ///< used compression features | ||
197 | uint8_t len_prefix; ///< frame is prefixed with its length | ||
198 | uint8_t dynamic_range_compression; ///< frame contains DRC data | ||
199 | uint8_t bits_per_sample; ///< integer audio sample size for the unscaled IMDCT output (used to scale to [-1.0, 1.0]) | ||
200 | uint16_t samples_per_frame; ///< number of samples to output | ||
201 | uint16_t trim_start; ///< number of samples to skip at start | ||
202 | uint16_t trim_end; ///< number of samples to skip at end | ||
203 | uint16_t log2_frame_size; | ||
204 | int8_t lfe_channel; ///< lfe channel index | ||
205 | uint8_t max_num_subframes; | ||
206 | uint8_t subframe_len_bits; ///< number of bits used for the subframe length | ||
207 | uint8_t max_subframe_len_bit; ///< flag indicating that the subframe is of maximum size when the first subframe length bit is 1 | ||
208 | uint16_t min_samples_per_subframe; | ||
209 | int8_t num_sfb[WMAPRO_BLOCK_SIZES]; ///< scale factor bands per block size | ||
210 | int16_t sfb_offsets[WMAPRO_BLOCK_SIZES][MAX_BANDS]; ///< scale factor band offsets (multiples of 4) | ||
211 | int8_t sf_offsets[WMAPRO_BLOCK_SIZES][WMAPRO_BLOCK_SIZES][MAX_BANDS]; ///< scale factor resample matrix | ||
212 | int16_t subwoofer_cutoffs[WMAPRO_BLOCK_SIZES]; ///< subwoofer cutoff values | ||
213 | |||
214 | /* packet decode state */ | ||
215 | GetBitContext pgb; ///< bitstream reader context for the packet | ||
216 | int next_packet_start; ///< start offset of the next wma packet in the demuxer packet | ||
217 | uint8_t packet_offset; ///< frame offset in the packet | ||
218 | uint8_t packet_sequence_number; ///< current packet number | ||
219 | int num_saved_bits; ///< saved number of bits | ||
220 | int frame_offset; ///< frame offset in the bit reservoir | ||
221 | int subframe_offset; ///< subframe offset in the bit reservoir | ||
222 | uint8_t packet_loss; ///< set in case of bitstream error | ||
223 | uint8_t packet_done; ///< set when a packet is fully decoded | ||
224 | uint8_t eof_done; ///< set when EOF reached and extra subframe is written (XMA1/2) | ||
225 | |||
226 | /* frame decode state */ | ||
227 | uint32_t frame_num; ///< current frame number (not used for decoding) | ||
228 | GetBitContext gb; ///< bitstream reader context | ||
229 | int buf_bit_size; ///< buffer size in bits | ||
230 | uint8_t drc_gain; ///< gain for the DRC tool | ||
231 | int8_t skip_frame; ///< skip output step | ||
232 | int8_t parsed_all_subframes; ///< all subframes decoded? | ||
233 | uint8_t skip_packets; ///< packets to skip to find next packet in a stream (XMA1/2) | ||
234 | |||
235 | /* subframe/block decode state */ | ||
236 | int16_t subframe_len; ///< current subframe length | ||
237 | int8_t nb_channels; ///< number of channels in stream (XMA1/2) | ||
238 | int8_t channels_for_cur_subframe; ///< number of channels that contain the subframe | ||
239 | int8_t channel_indexes_for_cur_subframe[WMAPRO_MAX_CHANNELS]; | ||
240 | int8_t num_bands; ///< number of scale factor bands | ||
241 | int8_t transmit_num_vec_coeffs; ///< number of vector coded coefficients is part of the bitstream | ||
242 | int16_t* cur_sfb_offsets; ///< sfb offsets for the current block | ||
243 | uint8_t table_idx; ///< index for the num_sfb, sfb_offsets, sf_offsets and subwoofer_cutoffs tables | ||
244 | int8_t esc_len; ///< length of escaped coefficients | ||
245 | |||
246 | uint8_t num_chgroups; ///< number of channel groups | ||
247 | WMAProChannelGrp chgroup[WMAPRO_MAX_CHANNELS]; ///< channel group information | ||
248 | |||
249 | WMAProChannelCtx channel[WMAPRO_MAX_CHANNELS]; ///< per channel data | ||
250 | } WMAProDecodeCtx; | ||
251 | |||
252 | typedef struct XMADecodeCtx { | ||
253 | WMAProDecodeCtx xma[XMA_MAX_STREAMS]; | ||
254 | AVFrame *frames[XMA_MAX_STREAMS]; | ||
255 | int current_stream; | ||
256 | int num_streams; | ||
257 | AVAudioFifo *samples[2][XMA_MAX_STREAMS]; | ||
258 | int start_channel[XMA_MAX_STREAMS]; | ||
259 | int trim_start, trim_end; | ||
260 | int flushed; | ||
261 | } XMADecodeCtx; | ||
262 | |||
263 | /** | ||
264 | *@brief helper function to print the most important members of the context | ||
265 | *@param s context | ||
266 | */ | ||
267 | ✗ | static av_cold void dump_context(WMAProDecodeCtx *s) | |
268 | { | ||
269 | #define PRINT(a, b) av_log(s->avctx, AV_LOG_DEBUG, " %s = %d\n", a, b); | ||
270 | #define PRINT_HEX(a, b) av_log(s->avctx, AV_LOG_DEBUG, " %s = %"PRIx32"\n", a, b); | ||
271 | |||
272 | ✗ | PRINT("ed sample bit depth", s->bits_per_sample); | |
273 | ✗ | PRINT_HEX("ed decode flags", s->decode_flags); | |
274 | ✗ | PRINT("samples per frame", s->samples_per_frame); | |
275 | ✗ | PRINT("log2 frame size", s->log2_frame_size); | |
276 | ✗ | PRINT("max num subframes", s->max_num_subframes); | |
277 | ✗ | PRINT("len prefix", s->len_prefix); | |
278 | ✗ | PRINT("num channels", s->nb_channels); | |
279 | ✗ | } | |
280 | |||
281 | /** | ||
282 | *@brief Uninitialize the decoder and free all resources. | ||
283 | *@param avctx codec context | ||
284 | *@return 0 on success, < 0 otherwise | ||
285 | */ | ||
286 | 8 | static av_cold int decode_end(WMAProDecodeCtx *s) | |
287 | { | ||
288 | int i; | ||
289 | |||
290 | 8 | av_freep(&s->fdsp); | |
291 | |||
292 |
2/2✓ Branch 0 taken 64 times.
✓ Branch 1 taken 8 times.
|
72 | for (i = 0; i < WMAPRO_BLOCK_SIZES; i++) |
293 | 64 | av_tx_uninit(&s->tx[i]); | |
294 | |||
295 | 8 | return 0; | |
296 | } | ||
297 | |||
298 | 8 | static av_cold int wmapro_decode_end(AVCodecContext *avctx) | |
299 | { | ||
300 | 8 | WMAProDecodeCtx *s = avctx->priv_data; | |
301 | |||
302 | 8 | decode_end(s); | |
303 | |||
304 | 8 | return 0; | |
305 | } | ||
306 | |||
307 | 40 | static av_cold int get_rate(AVCodecContext *avctx) | |
308 | { | ||
309 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 40 times.
|
40 | if (avctx->codec_id != AV_CODEC_ID_WMAPRO) { // XXX: is this really only for XMA? |
310 | ✗ | if (avctx->sample_rate > 44100) | |
311 | ✗ | return 48000; | |
312 | ✗ | else if (avctx->sample_rate > 32000) | |
313 | ✗ | return 44100; | |
314 | ✗ | else if (avctx->sample_rate > 24000) | |
315 | ✗ | return 32000; | |
316 | ✗ | return 24000; | |
317 | } | ||
318 | |||
319 | 40 | return avctx->sample_rate; | |
320 | } | ||
321 | |||
322 | 5 | static av_cold void decode_init_static(void) | |
323 | { | ||
324 | static VLCElem vlc_buf[2108 + 3912]; | ||
325 | 5 | VLCInitState state = VLC_INIT_STATE(vlc_buf); | |
326 | |||
327 | 5 | VLC_INIT_STATIC_TABLE_FROM_LENGTHS(sf_vlc, SCALEVLCBITS, HUFF_SCALE_SIZE, | |
328 | &scale_table[0][1], 2, | ||
329 | &scale_table[0][0], 2, 1, -60, 0); | ||
330 | 5 | VLC_INIT_STATIC_TABLE_FROM_LENGTHS(sf_rl_vlc, VLCBITS, HUFF_SCALE_RL_SIZE, | |
331 | &scale_rl_table[0][1], 2, | ||
332 | &scale_rl_table[0][0], 2, 1, 0, 0); | ||
333 | 5 | coef_vlc[0] = | |
334 | 5 | ff_vlc_init_tables_from_lengths(&state, VLCBITS, HUFF_COEF0_SIZE, | |
335 | coef0_lens, 1, | ||
336 | coef0_syms, 2, 2, 0, 0); | ||
337 | 5 | coef_vlc[1] = | |
338 | 5 | ff_vlc_init_tables_from_lengths(&state, VLCBITS, HUFF_COEF1_SIZE, | |
339 | &coef1_table[0][1], 2, | ||
340 | &coef1_table[0][0], 2, 1, 0, 0); | ||
341 | 5 | VLC_INIT_STATIC_TABLE_FROM_LENGTHS(vec4_vlc, VLCBITS, HUFF_VEC4_SIZE, | |
342 | vec4_lens, 1, | ||
343 | vec4_syms, 2, 2, -1, 0); | ||
344 | 5 | VLC_INIT_STATIC_TABLE_FROM_LENGTHS(vec2_vlc, VLCBITS, HUFF_VEC2_SIZE, | |
345 | &vec2_table[0][1], 2, | ||
346 | &vec2_table[0][0], 2, 1, -1, 0); | ||
347 | 5 | VLC_INIT_STATIC_TABLE_FROM_LENGTHS(vec1_vlc, VLCBITS, HUFF_VEC1_SIZE, | |
348 | &vec1_table[0][1], 2, | ||
349 | &vec1_table[0][0], 2, 1, 0, 0); | ||
350 | |||
351 | /** calculate sine values for the decorrelation matrix */ | ||
352 |
2/2✓ Branch 0 taken 165 times.
✓ Branch 1 taken 5 times.
|
170 | for (int i = 0; i < 33; i++) |
353 | 165 | sin64[i] = sin(i * M_PI / 64.0); | |
354 | |||
355 |
2/2✓ Branch 0 taken 40 times.
✓ Branch 1 taken 5 times.
|
45 | for (int i = WMAPRO_BLOCK_MIN_BITS; i <= WMAPRO_BLOCK_MAX_BITS; i++) |
356 | 40 | ff_init_ff_sine_windows(i); | |
357 | 5 | } | |
358 | |||
359 | /** | ||
360 | *@brief Initialize the decoder. | ||
361 | *@param avctx codec context | ||
362 | *@return 0 on success, -1 otherwise | ||
363 | */ | ||
364 | 8 | static av_cold int decode_init(WMAProDecodeCtx *s, AVCodecContext *avctx, int num_stream) | |
365 | { | ||
366 | static AVOnce init_static_once = AV_ONCE_INIT; | ||
367 | 8 | uint8_t *edata_ptr = avctx->extradata; | |
368 | unsigned int channel_mask; | ||
369 | int i, bits; | ||
370 | int log2_max_num_subframes; | ||
371 | int num_possible_block_sizes; | ||
372 | |||
373 | 8 | s->avctx = avctx; | |
374 | |||
375 | 8 | init_put_bits(&s->pb, s->frame_data, MAX_FRAMESIZE); | |
376 | |||
377 | 8 | avctx->sample_fmt = AV_SAMPLE_FMT_FLTP; | |
378 | |||
379 | /** dump the extradata */ | ||
380 | 8 | av_log(avctx, AV_LOG_DEBUG, "extradata:\n"); | |
381 |
2/2✓ Branch 0 taken 144 times.
✓ Branch 1 taken 8 times.
|
152 | for (i = 0; i < avctx->extradata_size; i++) |
382 | 144 | av_log(avctx, AV_LOG_DEBUG, "[%x] ", avctx->extradata[i]); | |
383 | 8 | av_log(avctx, AV_LOG_DEBUG, "\n"); | |
384 | |||
385 |
1/4✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
|
8 | if (avctx->codec_id == AV_CODEC_ID_XMA2 && avctx->extradata_size == 34) { /* XMA2WAVEFORMATEX */ |
386 | ✗ | s->decode_flags = 0x10d6; | |
387 | ✗ | s->bits_per_sample = 16; | |
388 | ✗ | channel_mask = 0; //AV_RL32(edata_ptr+2); /* not always in expected order */ | |
389 | ✗ | if ((num_stream+1) * XMA_MAX_CHANNELS_STREAM > avctx->ch_layout.nb_channels) /* stream config is 2ch + 2ch + ... + 1/2ch */ | |
390 | ✗ | s->nb_channels = 1; | |
391 | else | ||
392 | ✗ | s->nb_channels = 2; | |
393 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
|
8 | } else if (avctx->codec_id == AV_CODEC_ID_XMA2) { /* XMA2WAVEFORMAT */ |
394 | ✗ | s->decode_flags = 0x10d6; | |
395 | ✗ | s->bits_per_sample = 16; | |
396 | ✗ | channel_mask = 0; /* would need to aggregate from all streams */ | |
397 | ✗ | s->nb_channels = edata_ptr[32 + ((edata_ptr[0]==3)?0:8) + 4*num_stream + 0]; /* nth stream config */ | |
398 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
|
8 | } else if (avctx->codec_id == AV_CODEC_ID_XMA1) { /* XMAWAVEFORMAT */ |
399 | ✗ | s->decode_flags = 0x10d6; | |
400 | ✗ | s->bits_per_sample = 16; | |
401 | ✗ | channel_mask = 0; /* would need to aggregate from all streams */ | |
402 | ✗ | s->nb_channels = edata_ptr[8 + 20*num_stream + 17]; /* nth stream config */ | |
403 |
2/4✓ Branch 0 taken 8 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 8 times.
✗ Branch 3 not taken.
|
8 | } else if (avctx->codec_id == AV_CODEC_ID_WMAPRO && avctx->extradata_size >= 18) { |
404 | 8 | s->decode_flags = AV_RL16(edata_ptr+14); | |
405 | 8 | channel_mask = AV_RL32(edata_ptr+2); | |
406 | 8 | s->bits_per_sample = AV_RL16(edata_ptr); | |
407 |
1/2✓ Branch 0 taken 8 times.
✗ Branch 1 not taken.
|
8 | s->nb_channels = channel_mask ? av_popcount(channel_mask) : avctx->ch_layout.nb_channels; |
408 | |||
409 |
2/4✓ Branch 0 taken 8 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 8 times.
|
8 | if (s->bits_per_sample > 32 || s->bits_per_sample < 1) { |
410 | ✗ | avpriv_request_sample(avctx, "bits per sample is %d", s->bits_per_sample); | |
411 | ✗ | return AVERROR_PATCHWELCOME; | |
412 | } | ||
413 | } else { | ||
414 | ✗ | avpriv_request_sample(avctx, "Unknown extradata size"); | |
415 | ✗ | return AVERROR_PATCHWELCOME; | |
416 | } | ||
417 | |||
418 | /** generic init */ | ||
419 | 8 | s->log2_frame_size = av_log2(avctx->block_align) + 4; | |
420 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
|
8 | if (s->log2_frame_size > 25) { |
421 | ✗ | avpriv_request_sample(avctx, "Large block align"); | |
422 | ✗ | return AVERROR_PATCHWELCOME; | |
423 | } | ||
424 | |||
425 | /** frame info */ | ||
426 | 8 | s->skip_frame = 1; /* skip first frame */ | |
427 | |||
428 | 8 | s->packet_loss = 1; | |
429 | 8 | s->len_prefix = (s->decode_flags & 0x40); | |
430 | |||
431 | /** get frame len */ | ||
432 |
1/2✓ Branch 0 taken 8 times.
✗ Branch 1 not taken.
|
8 | if (avctx->codec_id == AV_CODEC_ID_WMAPRO) { |
433 | 8 | bits = ff_wma_get_frame_len_bits(avctx->sample_rate, 3, s->decode_flags); | |
434 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
|
8 | if (bits > WMAPRO_BLOCK_MAX_BITS) { |
435 | ✗ | avpriv_request_sample(avctx, "14-bit block sizes"); | |
436 | ✗ | return AVERROR_PATCHWELCOME; | |
437 | } | ||
438 | 8 | s->samples_per_frame = 1 << bits; | |
439 | } else { | ||
440 | ✗ | s->samples_per_frame = 512; | |
441 | } | ||
442 | |||
443 | /** subframe info */ | ||
444 | 8 | log2_max_num_subframes = ((s->decode_flags & 0x38) >> 3); | |
445 | 8 | s->max_num_subframes = 1 << log2_max_num_subframes; | |
446 |
1/4✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
|
8 | if (s->max_num_subframes == 16 || s->max_num_subframes == 4) |
447 | 8 | s->max_subframe_len_bit = 1; | |
448 | 8 | s->subframe_len_bits = av_log2(log2_max_num_subframes) + 1; | |
449 | |||
450 | 8 | num_possible_block_sizes = log2_max_num_subframes + 1; | |
451 | 8 | s->min_samples_per_subframe = s->samples_per_frame / s->max_num_subframes; | |
452 | 8 | s->dynamic_range_compression = (s->decode_flags & 0x80); | |
453 | |||
454 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
|
8 | if (s->max_num_subframes > MAX_SUBFRAMES) { |
455 | ✗ | av_log(avctx, AV_LOG_ERROR, "invalid number of subframes %"PRId8"\n", | |
456 | ✗ | s->max_num_subframes); | |
457 | ✗ | return AVERROR_INVALIDDATA; | |
458 | } | ||
459 | |||
460 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
|
8 | if (s->min_samples_per_subframe < WMAPRO_BLOCK_MIN_SIZE) { |
461 | ✗ | av_log(avctx, AV_LOG_ERROR, "min_samples_per_subframe of %d too small\n", | |
462 | ✗ | s->min_samples_per_subframe); | |
463 | ✗ | return AVERROR_INVALIDDATA; | |
464 | } | ||
465 | |||
466 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
|
8 | if (s->nb_channels <= 0) { |
467 | ✗ | av_log(avctx, AV_LOG_ERROR, "invalid number of channels %d\n", | |
468 | ✗ | s->nb_channels); | |
469 | ✗ | return AVERROR_INVALIDDATA; | |
470 |
1/4✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
|
8 | } else if (avctx->codec_id != AV_CODEC_ID_WMAPRO && s->nb_channels > XMA_MAX_CHANNELS_STREAM) { |
471 | ✗ | av_log(avctx, AV_LOG_ERROR, "invalid number of channels per XMA stream %d\n", | |
472 | ✗ | s->nb_channels); | |
473 | ✗ | return AVERROR_INVALIDDATA; | |
474 |
2/4✓ Branch 0 taken 8 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 8 times.
|
8 | } else if (s->nb_channels > WMAPRO_MAX_CHANNELS || s->nb_channels > avctx->ch_layout.nb_channels) { |
475 | ✗ | avpriv_request_sample(avctx, | |
476 | "More than %d channels", WMAPRO_MAX_CHANNELS); | ||
477 | ✗ | return AVERROR_PATCHWELCOME; | |
478 | } | ||
479 | |||
480 | /** init previous block len */ | ||
481 |
2/2✓ Branch 0 taken 24 times.
✓ Branch 1 taken 8 times.
|
32 | for (i = 0; i < s->nb_channels; i++) |
482 | 24 | s->channel[i].prev_block_len = s->samples_per_frame; | |
483 | |||
484 | /** extract lfe channel position */ | ||
485 | 8 | s->lfe_channel = -1; | |
486 | |||
487 |
2/2✓ Branch 0 taken 2 times.
✓ Branch 1 taken 6 times.
|
8 | if (channel_mask & 8) { |
488 | unsigned int mask; | ||
489 |
2/2✓ Branch 0 taken 8 times.
✓ Branch 1 taken 2 times.
|
10 | for (mask = 1; mask < 16; mask <<= 1) { |
490 |
1/2✓ Branch 0 taken 8 times.
✗ Branch 1 not taken.
|
8 | if (channel_mask & mask) |
491 | 8 | ++s->lfe_channel; | |
492 | } | ||
493 | } | ||
494 | |||
495 | /** calculate number of scale factor bands and their offsets | ||
496 | for every possible block size */ | ||
497 |
2/2✓ Branch 0 taken 40 times.
✓ Branch 1 taken 8 times.
|
48 | for (i = 0; i < num_possible_block_sizes; i++) { |
498 | 40 | int subframe_len = s->samples_per_frame >> i; | |
499 | int x; | ||
500 | 40 | int band = 1; | |
501 | 40 | int rate = get_rate(avctx); | |
502 | |||
503 | 40 | s->sfb_offsets[i][0] = 0; | |
504 | |||
505 |
2/4✓ Branch 0 taken 1040 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 1040 times.
✗ Branch 3 not taken.
|
1040 | for (x = 0; x < MAX_BANDS-1 && s->sfb_offsets[i][band - 1] < subframe_len; x++) { |
506 | 1040 | int offset = (subframe_len * 2 * critical_freq[x]) / rate + 2; | |
507 | 1040 | offset &= ~3; | |
508 |
2/2✓ Branch 0 taken 860 times.
✓ Branch 1 taken 180 times.
|
1040 | if (offset > s->sfb_offsets[i][band - 1]) |
509 | 860 | s->sfb_offsets[i][band++] = offset; | |
510 | |||
511 |
2/2✓ Branch 0 taken 40 times.
✓ Branch 1 taken 1000 times.
|
1040 | if (offset >= subframe_len) |
512 | 40 | break; | |
513 | } | ||
514 | 40 | s->sfb_offsets[i][band - 1] = subframe_len; | |
515 | 40 | s->num_sfb[i] = band - 1; | |
516 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 40 times.
|
40 | if (s->num_sfb[i] <= 0) { |
517 | ✗ | av_log(avctx, AV_LOG_ERROR, "num_sfb invalid\n"); | |
518 | ✗ | return AVERROR_INVALIDDATA; | |
519 | } | ||
520 | } | ||
521 | |||
522 | |||
523 | /** Scale factors can be shared between blocks of different size | ||
524 | as every block has a different scale factor band layout. | ||
525 | The matrix sf_offsets is needed to find the correct scale factor. | ||
526 | */ | ||
527 | |||
528 |
2/2✓ Branch 0 taken 40 times.
✓ Branch 1 taken 8 times.
|
48 | for (i = 0; i < num_possible_block_sizes; i++) { |
529 | int b; | ||
530 |
2/2✓ Branch 0 taken 860 times.
✓ Branch 1 taken 40 times.
|
900 | for (b = 0; b < s->num_sfb[i]; b++) { |
531 | int x; | ||
532 | 860 | int offset = ((s->sfb_offsets[i][b] | |
533 | 860 | + s->sfb_offsets[i][b + 1] - 1) << i) >> 1; | |
534 |
2/2✓ Branch 0 taken 4300 times.
✓ Branch 1 taken 860 times.
|
5160 | for (x = 0; x < num_possible_block_sizes; x++) { |
535 | 4300 | int v = 0; | |
536 |
2/2✓ Branch 0 taken 43876 times.
✓ Branch 1 taken 4300 times.
|
48176 | while (s->sfb_offsets[x][v + 1] << x < offset) { |
537 | 43876 | v++; | |
538 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 43876 times.
|
43876 | av_assert0(v < MAX_BANDS); |
539 | } | ||
540 | 4300 | s->sf_offsets[i][x][b] = v; | |
541 | } | ||
542 | } | ||
543 | } | ||
544 | |||
545 | 8 | s->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT); | |
546 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
|
8 | if (!s->fdsp) |
547 | ✗ | return AVERROR(ENOMEM); | |
548 | |||
549 | /** init MDCT, FIXME: only init needed sizes */ | ||
550 |
2/2✓ Branch 0 taken 64 times.
✓ Branch 1 taken 8 times.
|
72 | for (i = 0; i < WMAPRO_BLOCK_SIZES; i++) { |
551 | 64 | const float scale = 1.0 / (1 << (WMAPRO_BLOCK_MIN_BITS + i - 1)) | |
552 | 64 | / (1ll << (s->bits_per_sample - 1)); | |
553 | 64 | int err = av_tx_init(&s->tx[i], &s->tx_fn[i], AV_TX_FLOAT_MDCT, 1, | |
554 | 64 | 1 << (WMAPRO_BLOCK_MIN_BITS + i), &scale, 0); | |
555 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 64 times.
|
64 | if (err < 0) |
556 | ✗ | return err; | |
557 | } | ||
558 | |||
559 | /** init MDCT windows: simple sine window */ | ||
560 |
2/2✓ Branch 0 taken 64 times.
✓ Branch 1 taken 8 times.
|
72 | for (i = 0; i < WMAPRO_BLOCK_SIZES; i++) { |
561 | 64 | const int win_idx = WMAPRO_BLOCK_MAX_BITS - i; | |
562 | 64 | s->windows[WMAPRO_BLOCK_SIZES - i - 1] = ff_sine_windows[win_idx]; | |
563 | } | ||
564 | |||
565 | /** calculate subwoofer cutoff values */ | ||
566 |
2/2✓ Branch 0 taken 40 times.
✓ Branch 1 taken 8 times.
|
48 | for (i = 0; i < num_possible_block_sizes; i++) { |
567 | 40 | int block_size = s->samples_per_frame >> i; | |
568 | 40 | int cutoff = (440*block_size + 3LL * (s->avctx->sample_rate >> 1) - 1) | |
569 | 40 | / s->avctx->sample_rate; | |
570 | 40 | s->subwoofer_cutoffs[i] = av_clip(cutoff, 4, block_size); | |
571 | } | ||
572 | |||
573 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
|
8 | if (avctx->debug & FF_DEBUG_BITSTREAM) |
574 | ✗ | dump_context(s); | |
575 | |||
576 |
1/2✓ Branch 0 taken 8 times.
✗ Branch 1 not taken.
|
8 | if (avctx->codec_id == AV_CODEC_ID_WMAPRO) { |
577 |
1/2✓ Branch 0 taken 8 times.
✗ Branch 1 not taken.
|
8 | if (channel_mask) { |
578 | 8 | av_channel_layout_uninit(&avctx->ch_layout); | |
579 | 8 | av_channel_layout_from_mask(&avctx->ch_layout, channel_mask); | |
580 | } else | ||
581 | ✗ | avctx->ch_layout.order = AV_CHANNEL_ORDER_UNSPEC; | |
582 | } | ||
583 | |||
584 | 8 | ff_thread_once(&init_static_once, decode_init_static); | |
585 | |||
586 | 8 | return 0; | |
587 | } | ||
588 | |||
589 | /** | ||
590 | *@brief Initialize the decoder. | ||
591 | *@param avctx codec context | ||
592 | *@return 0 on success, -1 otherwise | ||
593 | */ | ||
594 | 8 | static av_cold int wmapro_decode_init(AVCodecContext *avctx) | |
595 | { | ||
596 | 8 | WMAProDecodeCtx *s = avctx->priv_data; | |
597 | |||
598 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 8 times.
|
8 | if (!avctx->block_align) { |
599 | ✗ | av_log(avctx, AV_LOG_ERROR, "block_align is not set\n"); | |
600 | ✗ | return AVERROR(EINVAL); | |
601 | } | ||
602 | |||
603 | 8 | return decode_init(s, avctx, 0); | |
604 | } | ||
605 | |||
606 | /** | ||
607 | *@brief Decode the subframe length. | ||
608 | *@param s context | ||
609 | *@param offset sample offset in the frame | ||
610 | *@return decoded subframe length on success, < 0 in case of an error | ||
611 | */ | ||
612 | 527 | static int decode_subframe_length(WMAProDecodeCtx *s, int offset) | |
613 | { | ||
614 | 527 | int frame_len_shift = 0; | |
615 | int subframe_len; | ||
616 | |||
617 | /** no need to read from the bitstream when only one length is possible */ | ||
618 |
2/2✓ Branch 0 taken 9 times.
✓ Branch 1 taken 518 times.
|
527 | if (offset == s->samples_per_frame - s->min_samples_per_subframe) |
619 | 9 | return s->min_samples_per_subframe; | |
620 | |||
621 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 518 times.
|
518 | if (get_bits_left(&s->gb) < 1) |
622 | ✗ | return AVERROR_INVALIDDATA; | |
623 | |||
624 | /** 1 bit indicates if the subframe is of maximum length */ | ||
625 |
1/2✓ Branch 0 taken 518 times.
✗ Branch 1 not taken.
|
518 | if (s->max_subframe_len_bit) { |
626 |
2/2✓ Branch 1 taken 279 times.
✓ Branch 2 taken 239 times.
|
518 | if (get_bits1(&s->gb)) |
627 | 279 | frame_len_shift = 1 + get_bits(&s->gb, s->subframe_len_bits-1); | |
628 | } else | ||
629 | ✗ | frame_len_shift = get_bits(&s->gb, s->subframe_len_bits); | |
630 | |||
631 | 518 | subframe_len = s->samples_per_frame >> frame_len_shift; | |
632 | |||
633 | /** sanity check the length */ | ||
634 |
1/2✓ Branch 0 taken 518 times.
✗ Branch 1 not taken.
|
518 | if (subframe_len < s->min_samples_per_subframe || |
635 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 518 times.
|
518 | subframe_len > s->samples_per_frame) { |
636 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "broken frame: subframe_len %i\n", | |
637 | subframe_len); | ||
638 | ✗ | return AVERROR_INVALIDDATA; | |
639 | } | ||
640 | 518 | return subframe_len; | |
641 | } | ||
642 | |||
643 | /** | ||
644 | *@brief Decode how the data in the frame is split into subframes. | ||
645 | * Every WMA frame contains the encoded data for a fixed number of | ||
646 | * samples per channel. The data for every channel might be split | ||
647 | * into several subframes. This function will reconstruct the list of | ||
648 | * subframes for every channel. | ||
649 | * | ||
650 | * If the subframes are not evenly split, the algorithm estimates the | ||
651 | * channels with the lowest number of total samples. | ||
652 | * Afterwards, for each of these channels a bit is read from the | ||
653 | * bitstream that indicates if the channel contains a subframe with the | ||
654 | * next subframe size that is going to be read from the bitstream or not. | ||
655 | * If a channel contains such a subframe, the subframe size gets added to | ||
656 | * the channel's subframe list. | ||
657 | * The algorithm repeats these steps until the frame is properly divided | ||
658 | * between the individual channels. | ||
659 | * | ||
660 | *@param s context | ||
661 | *@return 0 on success, < 0 in case of an error | ||
662 | */ | ||
663 | 257 | static int decode_tilehdr(WMAProDecodeCtx *s) | |
664 | { | ||
665 | 257 | uint16_t num_samples[WMAPRO_MAX_CHANNELS] = { 0 };/**< sum of samples for all currently known subframes of a channel */ | |
666 | uint8_t contains_subframe[WMAPRO_MAX_CHANNELS]; /**< flag indicating if a channel contains the current subframe */ | ||
667 | 257 | int channels_for_cur_subframe = s->nb_channels; /**< number of channels that contain the current subframe */ | |
668 | 257 | int fixed_channel_layout = 0; /**< flag indicating that all channels use the same subframe offsets and sizes */ | |
669 | 257 | int min_channel_len = 0; /**< smallest sum of samples (channels with this length will be processed first) */ | |
670 | int c; | ||
671 | |||
672 | /* Should never consume more than 3073 bits (256 iterations for the | ||
673 | * while loop when always the minimum amount of 128 samples is subtracted | ||
674 | * from missing samples in the 8 channel case). | ||
675 | * 1 + BLOCK_MAX_SIZE * MAX_CHANNELS / BLOCK_MIN_SIZE * (MAX_CHANNELS + 4) | ||
676 | */ | ||
677 | |||
678 | /** reset tiling information */ | ||
679 |
2/2✓ Branch 0 taken 926 times.
✓ Branch 1 taken 257 times.
|
1183 | for (c = 0; c < s->nb_channels; c++) |
680 | 926 | s->channel[c].num_subframes = 0; | |
681 | |||
682 |
3/4✓ Branch 0 taken 257 times.
✗ Branch 1 not taken.
✓ Branch 3 taken 225 times.
✓ Branch 4 taken 32 times.
|
257 | if (s->max_num_subframes == 1 || get_bits1(&s->gb)) |
683 | 225 | fixed_channel_layout = 1; | |
684 | |||
685 | /** loop until the frame data is split between the subframes */ | ||
686 | do { | ||
687 | int subframe_len; | ||
688 | |||
689 | /** check which channels contain the subframe */ | ||
690 |
2/2✓ Branch 0 taken 2286 times.
✓ Branch 1 taken 527 times.
|
2813 | for (c = 0; c < s->nb_channels; c++) { |
691 |
2/2✓ Branch 0 taken 1277 times.
✓ Branch 1 taken 1009 times.
|
2286 | if (num_samples[c] == min_channel_len) { |
692 |
4/4✓ Branch 0 taken 413 times.
✓ Branch 1 taken 864 times.
✓ Branch 2 taken 224 times.
✓ Branch 3 taken 189 times.
|
1277 | if (fixed_channel_layout || channels_for_cur_subframe == 1 || |
693 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 224 times.
|
224 | (min_channel_len == s->samples_per_frame - s->min_samples_per_subframe)) |
694 | 1053 | contains_subframe[c] = 1; | |
695 | else | ||
696 | 224 | contains_subframe[c] = get_bits1(&s->gb); | |
697 | } else | ||
698 | 1009 | contains_subframe[c] = 0; | |
699 | } | ||
700 | |||
701 | /** get subframe length, subframe_len == 0 is not allowed */ | ||
702 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 527 times.
|
527 | if ((subframe_len = decode_subframe_length(s, min_channel_len)) <= 0) |
703 | ✗ | return AVERROR_INVALIDDATA; | |
704 | |||
705 | /** add subframes to the individual channels and find new min_channel_len */ | ||
706 | 527 | min_channel_len += subframe_len; | |
707 |
2/2✓ Branch 0 taken 2286 times.
✓ Branch 1 taken 527 times.
|
2813 | for (c = 0; c < s->nb_channels; c++) { |
708 | 2286 | WMAProChannelCtx* chan = &s->channel[c]; | |
709 | |||
710 |
2/2✓ Branch 0 taken 1234 times.
✓ Branch 1 taken 1052 times.
|
2286 | if (contains_subframe[c]) { |
711 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1234 times.
|
1234 | if (chan->num_subframes >= MAX_SUBFRAMES) { |
712 | ✗ | av_log(s->avctx, AV_LOG_ERROR, | |
713 | "broken frame: num subframes > 31\n"); | ||
714 | ✗ | return AVERROR_INVALIDDATA; | |
715 | } | ||
716 | 1234 | chan->subframe_len[chan->num_subframes] = subframe_len; | |
717 | 1234 | num_samples[c] += subframe_len; | |
718 | 1234 | ++chan->num_subframes; | |
719 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1234 times.
|
1234 | if (num_samples[c] > s->samples_per_frame) { |
720 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "broken frame: " | |
721 | "channel len > samples_per_frame\n"); | ||
722 | ✗ | return AVERROR_INVALIDDATA; | |
723 | } | ||
724 |
2/2✓ Branch 0 taken 211 times.
✓ Branch 1 taken 841 times.
|
1052 | } else if (num_samples[c] <= min_channel_len) { |
725 |
2/2✓ Branch 0 taken 43 times.
✓ Branch 1 taken 168 times.
|
211 | if (num_samples[c] < min_channel_len) { |
726 | 43 | channels_for_cur_subframe = 0; | |
727 | 43 | min_channel_len = num_samples[c]; | |
728 | } | ||
729 | 211 | ++channels_for_cur_subframe; | |
730 | } | ||
731 | } | ||
732 |
2/2✓ Branch 0 taken 270 times.
✓ Branch 1 taken 257 times.
|
527 | } while (min_channel_len < s->samples_per_frame); |
733 | |||
734 |
2/2✓ Branch 0 taken 926 times.
✓ Branch 1 taken 257 times.
|
1183 | for (c = 0; c < s->nb_channels; c++) { |
735 | int i; | ||
736 | 926 | int offset = 0; | |
737 |
2/2✓ Branch 0 taken 1234 times.
✓ Branch 1 taken 926 times.
|
2160 | for (i = 0; i < s->channel[c].num_subframes; i++) { |
738 | ff_dlog(s->avctx, "frame[%"PRIu32"] channel[%i] subframe[%i]" | ||
739 | " len %i\n", s->frame_num, c, i, | ||
740 | s->channel[c].subframe_len[i]); | ||
741 | 1234 | s->channel[c].subframe_offset[i] = offset; | |
742 | 1234 | offset += s->channel[c].subframe_len[i]; | |
743 | } | ||
744 | } | ||
745 | |||
746 | 257 | return 0; | |
747 | } | ||
748 | |||
749 | /** | ||
750 | *@brief Calculate a decorrelation matrix from the bitstream parameters. | ||
751 | *@param s codec context | ||
752 | *@param chgroup channel group for which the matrix needs to be calculated | ||
753 | */ | ||
754 | 79 | static void decode_decorrelation_matrix(WMAProDecodeCtx *s, | |
755 | WMAProChannelGrp *chgroup) | ||
756 | { | ||
757 | int i; | ||
758 | 79 | int offset = 0; | |
759 | int8_t rotation_offset[WMAPRO_MAX_CHANNELS * WMAPRO_MAX_CHANNELS]; | ||
760 | 79 | memset(chgroup->decorrelation_matrix, 0, s->nb_channels * | |
761 | 79 | s->nb_channels * sizeof(*chgroup->decorrelation_matrix)); | |
762 | |||
763 |
2/2✓ Branch 0 taken 237 times.
✓ Branch 1 taken 79 times.
|
316 | for (i = 0; i < chgroup->num_channels * (chgroup->num_channels - 1) >> 1; i++) |
764 | 237 | rotation_offset[i] = get_bits(&s->gb, 6); | |
765 | |||
766 |
2/2✓ Branch 0 taken 237 times.
✓ Branch 1 taken 79 times.
|
316 | for (i = 0; i < chgroup->num_channels; i++) |
767 | 237 | chgroup->decorrelation_matrix[chgroup->num_channels * i + i] = | |
768 |
1/2✓ Branch 1 taken 237 times.
✗ Branch 2 not taken.
|
237 | get_bits1(&s->gb) ? 1.0 : -1.0; |
769 | |||
770 |
2/2✓ Branch 0 taken 158 times.
✓ Branch 1 taken 79 times.
|
237 | for (i = 1; i < chgroup->num_channels; i++) { |
771 | int x; | ||
772 |
2/2✓ Branch 0 taken 237 times.
✓ Branch 1 taken 158 times.
|
395 | for (x = 0; x < i; x++) { |
773 | int y; | ||
774 |
2/2✓ Branch 0 taken 632 times.
✓ Branch 1 taken 237 times.
|
869 | for (y = 0; y < i + 1; y++) { |
775 | 632 | float v1 = chgroup->decorrelation_matrix[x * chgroup->num_channels + y]; | |
776 | 632 | float v2 = chgroup->decorrelation_matrix[i * chgroup->num_channels + y]; | |
777 | 632 | int n = rotation_offset[offset + x]; | |
778 | float sinv; | ||
779 | float cosv; | ||
780 | |||
781 |
2/2✓ Branch 0 taken 442 times.
✓ Branch 1 taken 190 times.
|
632 | if (n < 32) { |
782 | 442 | sinv = sin64[n]; | |
783 | 442 | cosv = sin64[32 - n]; | |
784 | } else { | ||
785 | 190 | sinv = sin64[64 - n]; | |
786 | 190 | cosv = -sin64[n - 32]; | |
787 | } | ||
788 | |||
789 | 632 | chgroup->decorrelation_matrix[y + x * chgroup->num_channels] = | |
790 | 632 | (v1 * sinv) - (v2 * cosv); | |
791 | 632 | chgroup->decorrelation_matrix[y + i * chgroup->num_channels] = | |
792 | 632 | (v1 * cosv) + (v2 * sinv); | |
793 | } | ||
794 | } | ||
795 | 158 | offset += i; | |
796 | } | ||
797 | 79 | } | |
798 | |||
799 | /** | ||
800 | *@brief Decode channel transformation parameters | ||
801 | *@param s codec context | ||
802 | *@return >= 0 in case of success, < 0 in case of bitstream errors | ||
803 | */ | ||
804 | 527 | static int decode_channel_transform(WMAProDecodeCtx* s) | |
805 | { | ||
806 | int i; | ||
807 | /* should never consume more than 1921 bits for the 8 channel case | ||
808 | * 1 + MAX_CHANNELS * (MAX_CHANNELS + 2 + 3 * MAX_CHANNELS * MAX_CHANNELS | ||
809 | * + MAX_CHANNELS + MAX_BANDS + 1) | ||
810 | */ | ||
811 | |||
812 | /** in the one channel case channel transforms are pointless */ | ||
813 | 527 | s->num_chgroups = 0; | |
814 |
1/2✓ Branch 0 taken 527 times.
✗ Branch 1 not taken.
|
527 | if (s->nb_channels > 1) { |
815 | 527 | int remaining_channels = s->channels_for_cur_subframe; | |
816 | |||
817 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 527 times.
|
527 | if (get_bits1(&s->gb)) { |
818 | ✗ | avpriv_request_sample(s->avctx, | |
819 | "Channel transform bit"); | ||
820 | ✗ | return AVERROR_PATCHWELCOME; | |
821 | } | ||
822 | |||
823 |
2/2✓ Branch 0 taken 629 times.
✓ Branch 1 taken 527 times.
|
1156 | for (s->num_chgroups = 0; remaining_channels && |
824 |
1/2✓ Branch 0 taken 629 times.
✗ Branch 1 not taken.
|
629 | s->num_chgroups < s->channels_for_cur_subframe; s->num_chgroups++) { |
825 | 629 | WMAProChannelGrp* chgroup = &s->chgroup[s->num_chgroups]; | |
826 | 629 | float** channel_data = chgroup->channel_data; | |
827 | 629 | chgroup->num_channels = 0; | |
828 | 629 | chgroup->transform = 0; | |
829 | |||
830 | /** decode channel mask */ | ||
831 |
2/2✓ Branch 0 taken 182 times.
✓ Branch 1 taken 447 times.
|
629 | if (remaining_channels > 2) { |
832 |
2/2✓ Branch 0 taken 1035 times.
✓ Branch 1 taken 182 times.
|
1217 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
833 | 1035 | int channel_idx = s->channel_indexes_for_cur_subframe[i]; | |
834 |
2/2✓ Branch 0 taken 819 times.
✓ Branch 1 taken 216 times.
|
1035 | if (!s->channel[channel_idx].grouped |
835 |
2/2✓ Branch 1 taken 536 times.
✓ Branch 2 taken 283 times.
|
819 | && get_bits1(&s->gb)) { |
836 | 536 | ++chgroup->num_channels; | |
837 | 536 | s->channel[channel_idx].grouped = 1; | |
838 | 536 | *channel_data++ = s->channel[channel_idx].coeffs; | |
839 | } | ||
840 | } | ||
841 | } else { | ||
842 | 447 | chgroup->num_channels = remaining_channels; | |
843 |
2/2✓ Branch 0 taken 775 times.
✓ Branch 1 taken 447 times.
|
1222 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
844 | 775 | int channel_idx = s->channel_indexes_for_cur_subframe[i]; | |
845 |
2/2✓ Branch 0 taken 698 times.
✓ Branch 1 taken 77 times.
|
775 | if (!s->channel[channel_idx].grouped) |
846 | 698 | *channel_data++ = s->channel[channel_idx].coeffs; | |
847 | 775 | s->channel[channel_idx].grouped = 1; | |
848 | } | ||
849 | } | ||
850 | |||
851 | /** decode transform type */ | ||
852 |
2/2✓ Branch 0 taken 276 times.
✓ Branch 1 taken 353 times.
|
629 | if (chgroup->num_channels == 2) { |
853 |
2/2✓ Branch 1 taken 38 times.
✓ Branch 2 taken 238 times.
|
276 | if (get_bits1(&s->gb)) { |
854 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 38 times.
|
38 | if (get_bits1(&s->gb)) { |
855 | ✗ | avpriv_request_sample(s->avctx, | |
856 | "Unknown channel transform type"); | ||
857 | ✗ | return AVERROR_PATCHWELCOME; | |
858 | } | ||
859 | } else { | ||
860 | 238 | chgroup->transform = 1; | |
861 |
2/2✓ Branch 0 taken 215 times.
✓ Branch 1 taken 23 times.
|
238 | if (s->nb_channels == 2) { |
862 | 215 | chgroup->decorrelation_matrix[0] = 1.0; | |
863 | 215 | chgroup->decorrelation_matrix[1] = -1.0; | |
864 | 215 | chgroup->decorrelation_matrix[2] = 1.0; | |
865 | 215 | chgroup->decorrelation_matrix[3] = 1.0; | |
866 | } else { | ||
867 | /** cos(pi/4) */ | ||
868 | 23 | chgroup->decorrelation_matrix[0] = 0.70703125; | |
869 | 23 | chgroup->decorrelation_matrix[1] = -0.70703125; | |
870 | 23 | chgroup->decorrelation_matrix[2] = 0.70703125; | |
871 | 23 | chgroup->decorrelation_matrix[3] = 0.70703125; | |
872 | } | ||
873 | } | ||
874 |
2/2✓ Branch 0 taken 157 times.
✓ Branch 1 taken 196 times.
|
353 | } else if (chgroup->num_channels > 2) { |
875 |
2/2✓ Branch 1 taken 79 times.
✓ Branch 2 taken 78 times.
|
157 | if (get_bits1(&s->gb)) { |
876 | 79 | chgroup->transform = 1; | |
877 |
1/2✓ Branch 1 taken 79 times.
✗ Branch 2 not taken.
|
79 | if (get_bits1(&s->gb)) { |
878 | 79 | decode_decorrelation_matrix(s, chgroup); | |
879 | } else { | ||
880 | /** FIXME: more than 6 coupled channels not supported */ | ||
881 | ✗ | if (chgroup->num_channels > 6) { | |
882 | ✗ | avpriv_request_sample(s->avctx, | |
883 | "Coupled channels > 6"); | ||
884 | } else { | ||
885 | ✗ | memcpy(chgroup->decorrelation_matrix, | |
886 | ✗ | default_decorrelation[chgroup->num_channels], | |
887 | ✗ | chgroup->num_channels * chgroup->num_channels * | |
888 | sizeof(*chgroup->decorrelation_matrix)); | ||
889 | } | ||
890 | } | ||
891 | } | ||
892 | } | ||
893 | |||
894 | /** decode transform on / off */ | ||
895 |
2/2✓ Branch 0 taken 317 times.
✓ Branch 1 taken 312 times.
|
629 | if (chgroup->transform) { |
896 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 317 times.
|
317 | if (!get_bits1(&s->gb)) { |
897 | int i; | ||
898 | /** transform can be enabled for individual bands */ | ||
899 | ✗ | for (i = 0; i < s->num_bands; i++) { | |
900 | ✗ | chgroup->transform_band[i] = get_bits1(&s->gb); | |
901 | } | ||
902 | } else { | ||
903 | 317 | memset(chgroup->transform_band, 1, s->num_bands); | |
904 | } | ||
905 | } | ||
906 | 629 | remaining_channels -= chgroup->num_channels; | |
907 | } | ||
908 | } | ||
909 | 527 | return 0; | |
910 | } | ||
911 | |||
912 | /** | ||
913 | *@brief Extract the coefficients from the bitstream. | ||
914 | *@param s codec context | ||
915 | *@param c current channel number | ||
916 | *@return 0 on success, < 0 in case of bitstream errors | ||
917 | */ | ||
918 | 1038 | static int decode_coeffs(WMAProDecodeCtx *s, int c) | |
919 | { | ||
920 | /* Integers 0..15 as single-precision floats. The table saves a | ||
921 | costly int to float conversion, and storing the values as | ||
922 | integers allows fast sign-flipping. */ | ||
923 | static const uint32_t fval_tab[16] = { | ||
924 | 0x00000000, 0x3f800000, 0x40000000, 0x40400000, | ||
925 | 0x40800000, 0x40a00000, 0x40c00000, 0x40e00000, | ||
926 | 0x41000000, 0x41100000, 0x41200000, 0x41300000, | ||
927 | 0x41400000, 0x41500000, 0x41600000, 0x41700000, | ||
928 | }; | ||
929 | int vlctable; | ||
930 | const VLCElem *vlc; | ||
931 | 1038 | WMAProChannelCtx* ci = &s->channel[c]; | |
932 | 1038 | int rl_mode = 0; | |
933 | 1038 | int cur_coeff = 0; | |
934 | 1038 | int num_zeros = 0; | |
935 | const uint16_t* run; | ||
936 | const float* level; | ||
937 | |||
938 | ff_dlog(s->avctx, "decode coefficients for channel %i\n", c); | ||
939 | |||
940 | 1038 | vlctable = get_bits1(&s->gb); | |
941 | 1038 | vlc = coef_vlc[vlctable]; | |
942 | |||
943 |
2/2✓ Branch 0 taken 321 times.
✓ Branch 1 taken 717 times.
|
1038 | if (vlctable) { |
944 | 321 | run = coef1_run; | |
945 | 321 | level = coef1_level; | |
946 | } else { | ||
947 | 717 | run = coef0_run; | |
948 | 717 | level = coef0_level; | |
949 | } | ||
950 | |||
951 | /** decode vector coefficients (consumes up to 167 bits per iteration for | ||
952 | 4 vector coded large values) */ | ||
953 |
3/4✓ Branch 0 taken 39251 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 38213 times.
✓ Branch 3 taken 1038 times.
|
39251 | while ((s->transmit_num_vec_coeffs || !rl_mode) && |
954 |
1/2✓ Branch 0 taken 38213 times.
✗ Branch 1 not taken.
|
38213 | (cur_coeff + 3 < ci->num_vec_coeffs)) { |
955 | uint32_t vals[4]; | ||
956 | int i; | ||
957 | unsigned int idx; | ||
958 | |||
959 | 38213 | idx = get_vlc2(&s->gb, vec4_vlc, VLCBITS, VEC4MAXDEPTH); | |
960 | |||
961 |
2/2✓ Branch 0 taken 21810 times.
✓ Branch 1 taken 16403 times.
|
38213 | if ((int)idx < 0) { |
962 |
2/2✓ Branch 0 taken 43620 times.
✓ Branch 1 taken 21810 times.
|
65430 | for (i = 0; i < 4; i += 2) { |
963 | 43620 | idx = get_vlc2(&s->gb, vec2_vlc, VLCBITS, VEC2MAXDEPTH); | |
964 |
2/2✓ Branch 0 taken 20544 times.
✓ Branch 1 taken 23076 times.
|
43620 | if ((int)idx < 0) { |
965 | uint32_t v0, v1; | ||
966 | 20544 | v0 = get_vlc2(&s->gb, vec1_vlc, VLCBITS, VEC1MAXDEPTH); | |
967 |
2/2✓ Branch 0 taken 1608 times.
✓ Branch 1 taken 18936 times.
|
20544 | if (v0 == HUFF_VEC1_SIZE - 1) |
968 | 1608 | v0 += ff_wma_get_large_val(&s->gb); | |
969 | 20544 | v1 = get_vlc2(&s->gb, vec1_vlc, VLCBITS, VEC1MAXDEPTH); | |
970 |
2/2✓ Branch 0 taken 1665 times.
✓ Branch 1 taken 18879 times.
|
20544 | if (v1 == HUFF_VEC1_SIZE - 1) |
971 | 1665 | v1 += ff_wma_get_large_val(&s->gb); | |
972 | 20544 | vals[i ] = av_float2int(v0); | |
973 | 20544 | vals[i+1] = av_float2int(v1); | |
974 | } else { | ||
975 | 23076 | vals[i] = fval_tab[idx >> 4 ]; | |
976 | 23076 | vals[i+1] = fval_tab[idx & 0xF]; | |
977 | } | ||
978 | } | ||
979 | } else { | ||
980 | 16403 | vals[0] = fval_tab[ idx >> 12 ]; | |
981 | 16403 | vals[1] = fval_tab[(idx >> 8) & 0xF]; | |
982 | 16403 | vals[2] = fval_tab[(idx >> 4) & 0xF]; | |
983 | 16403 | vals[3] = fval_tab[ idx & 0xF]; | |
984 | } | ||
985 | |||
986 | /** decode sign */ | ||
987 |
2/2✓ Branch 0 taken 152852 times.
✓ Branch 1 taken 38213 times.
|
191065 | for (i = 0; i < 4; i++) { |
988 |
2/2✓ Branch 0 taken 111800 times.
✓ Branch 1 taken 41052 times.
|
152852 | if (vals[i]) { |
989 | 111800 | uint32_t sign = get_bits1(&s->gb) - 1; | |
990 | 111800 | AV_WN32A(&ci->coeffs[cur_coeff], vals[i] ^ sign << 31); | |
991 | 111800 | num_zeros = 0; | |
992 | } else { | ||
993 | 41052 | ci->coeffs[cur_coeff] = 0; | |
994 | /** switch to run level mode when subframe_len / 128 zeros | ||
995 | were found in a row */ | ||
996 | 41052 | rl_mode |= (++num_zeros > s->subframe_len >> 8); | |
997 | } | ||
998 | 152852 | ++cur_coeff; | |
999 | } | ||
1000 | } | ||
1001 | |||
1002 | /** decode run level coded coefficients */ | ||
1003 |
1/2✓ Branch 0 taken 1038 times.
✗ Branch 1 not taken.
|
1038 | if (cur_coeff < s->subframe_len) { |
1004 | int ret; | ||
1005 | |||
1006 | 1038 | memset(&ci->coeffs[cur_coeff], 0, | |
1007 | 1038 | sizeof(*ci->coeffs) * (s->subframe_len - cur_coeff)); | |
1008 | 1038 | ret = ff_wma_run_level_decode(s->avctx, &s->gb, vlc, | |
1009 | 1038 | level, run, 1, ci->coeffs, | |
1010 | 1038 | cur_coeff, s->subframe_len, | |
1011 | 1038 | s->subframe_len, s->esc_len, 0); | |
1012 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1038 times.
|
1038 | if (ret < 0) |
1013 | ✗ | return ret; | |
1014 | } | ||
1015 | |||
1016 | 1038 | return 0; | |
1017 | } | ||
1018 | |||
1019 | /** | ||
1020 | *@brief Extract scale factors from the bitstream. | ||
1021 | *@param s codec context | ||
1022 | *@return 0 on success, < 0 in case of bitstream errors | ||
1023 | */ | ||
1024 | 521 | static int decode_scale_factors(WMAProDecodeCtx* s) | |
1025 | { | ||
1026 | int i; | ||
1027 | |||
1028 | /** should never consume more than 5344 bits | ||
1029 | * MAX_CHANNELS * (1 + MAX_BANDS * 23) | ||
1030 | */ | ||
1031 | |||
1032 |
2/2✓ Branch 0 taken 1224 times.
✓ Branch 1 taken 521 times.
|
1745 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
1033 | 1224 | int c = s->channel_indexes_for_cur_subframe[i]; | |
1034 | int* sf; | ||
1035 | int* sf_end; | ||
1036 | 1224 | s->channel[c].scale_factors = s->channel[c].saved_scale_factors[!s->channel[c].scale_factor_idx]; | |
1037 | 1224 | sf_end = s->channel[c].scale_factors + s->num_bands; | |
1038 | |||
1039 | /** resample scale factors for the new block size | ||
1040 | * as the scale factors might need to be resampled several times | ||
1041 | * before some new values are transmitted, a backup of the last | ||
1042 | * transmitted scale factors is kept in saved_scale_factors | ||
1043 | */ | ||
1044 |
2/2✓ Branch 0 taken 298 times.
✓ Branch 1 taken 926 times.
|
1224 | if (s->channel[c].reuse_sf) { |
1045 | 298 | const int8_t* sf_offsets = s->sf_offsets[s->table_idx][s->channel[c].table_idx]; | |
1046 | int b; | ||
1047 |
2/2✓ Branch 0 taken 5173 times.
✓ Branch 1 taken 298 times.
|
5471 | for (b = 0; b < s->num_bands; b++) |
1048 | 5173 | s->channel[c].scale_factors[b] = | |
1049 | 5173 | s->channel[c].saved_scale_factors[s->channel[c].scale_factor_idx][*sf_offsets++]; | |
1050 | } | ||
1051 | |||
1052 |
4/4✓ Branch 0 taken 301 times.
✓ Branch 1 taken 923 times.
✓ Branch 3 taken 283 times.
✓ Branch 4 taken 18 times.
|
1224 | if (!s->channel[c].cur_subframe || get_bits1(&s->gb)) { |
1053 | |||
1054 |
2/2✓ Branch 0 taken 926 times.
✓ Branch 1 taken 280 times.
|
1206 | if (!s->channel[c].reuse_sf) { |
1055 | int val; | ||
1056 | /** decode DPCM coded scale factors */ | ||
1057 | 926 | s->channel[c].scale_factor_step = get_bits(&s->gb, 2) + 1; | |
1058 | 926 | val = 45 / s->channel[c].scale_factor_step; | |
1059 |
2/2✓ Branch 0 taken 23831 times.
✓ Branch 1 taken 926 times.
|
24757 | for (sf = s->channel[c].scale_factors; sf < sf_end; sf++) { |
1060 | 23831 | val += get_vlc2(&s->gb, sf_vlc, SCALEVLCBITS, SCALEMAXDEPTH); | |
1061 | 23831 | *sf = val; | |
1062 | } | ||
1063 | } else { | ||
1064 | int i; | ||
1065 | /** run level decode differences to the resampled factors */ | ||
1066 |
1/2✓ Branch 0 taken 1904 times.
✗ Branch 1 not taken.
|
1904 | for (i = 0; i < s->num_bands; i++) { |
1067 | int idx; | ||
1068 | int skip; | ||
1069 | int val; | ||
1070 | int sign; | ||
1071 | |||
1072 | 1904 | idx = get_vlc2(&s->gb, sf_rl_vlc, VLCBITS, SCALERLMAXDEPTH); | |
1073 | |||
1074 |
2/2✓ Branch 0 taken 41 times.
✓ Branch 1 taken 1863 times.
|
1904 | if (!idx) { |
1075 | 41 | uint32_t code = get_bits(&s->gb, 14); | |
1076 | 41 | val = code >> 6; | |
1077 | 41 | sign = (code & 1) - 1; | |
1078 | 41 | skip = (code & 0x3f) >> 1; | |
1079 |
2/2✓ Branch 0 taken 280 times.
✓ Branch 1 taken 1583 times.
|
1863 | } else if (idx == 1) { |
1080 | 280 | break; | |
1081 | } else { | ||
1082 | 1583 | skip = scale_rl_run[idx]; | |
1083 | 1583 | val = scale_rl_level[idx]; | |
1084 | 1583 | sign = get_bits1(&s->gb)-1; | |
1085 | } | ||
1086 | |||
1087 | 1624 | i += skip; | |
1088 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1624 times.
|
1624 | if (i >= s->num_bands) { |
1089 | ✗ | av_log(s->avctx, AV_LOG_ERROR, | |
1090 | "invalid scale factor coding\n"); | ||
1091 | ✗ | return AVERROR_INVALIDDATA; | |
1092 | } | ||
1093 | 1624 | s->channel[c].scale_factors[i] += (val ^ sign) - sign; | |
1094 | } | ||
1095 | } | ||
1096 | /** swap buffers */ | ||
1097 | 1206 | s->channel[c].scale_factor_idx = !s->channel[c].scale_factor_idx; | |
1098 | 1206 | s->channel[c].table_idx = s->table_idx; | |
1099 | 1206 | s->channel[c].reuse_sf = 1; | |
1100 | } | ||
1101 | |||
1102 | /** calculate new scale factor maximum */ | ||
1103 | 1224 | s->channel[c].max_scale_factor = s->channel[c].scale_factors[0]; | |
1104 |
2/2✓ Branch 0 taken 27780 times.
✓ Branch 1 taken 1224 times.
|
29004 | for (sf = s->channel[c].scale_factors + 1; sf < sf_end; sf++) { |
1105 | 27780 | s->channel[c].max_scale_factor = | |
1106 | 27780 | FFMAX(s->channel[c].max_scale_factor, *sf); | |
1107 | } | ||
1108 | |||
1109 | } | ||
1110 | 521 | return 0; | |
1111 | } | ||
1112 | |||
1113 | /** | ||
1114 | *@brief Reconstruct the individual channel data. | ||
1115 | *@param s codec context | ||
1116 | */ | ||
1117 | 521 | static void inverse_channel_transform(WMAProDecodeCtx *s) | |
1118 | { | ||
1119 | int i; | ||
1120 | |||
1121 |
2/2✓ Branch 0 taken 623 times.
✓ Branch 1 taken 521 times.
|
1144 | for (i = 0; i < s->num_chgroups; i++) { |
1122 |
2/2✓ Branch 0 taken 317 times.
✓ Branch 1 taken 306 times.
|
623 | if (s->chgroup[i].transform) { |
1123 | float data[WMAPRO_MAX_CHANNELS]; | ||
1124 | 317 | const int num_channels = s->chgroup[i].num_channels; | |
1125 | 317 | float** ch_data = s->chgroup[i].channel_data; | |
1126 | 317 | float** ch_end = ch_data + num_channels; | |
1127 | 317 | const int8_t* tb = s->chgroup[i].transform_band; | |
1128 | int16_t* sfb; | ||
1129 | |||
1130 | /** multichannel decorrelation */ | ||
1131 | 317 | for (sfb = s->cur_sfb_offsets; | |
1132 |
2/2✓ Branch 0 taken 7696 times.
✓ Branch 1 taken 317 times.
|
8013 | sfb < s->cur_sfb_offsets + s->num_bands; sfb++) { |
1133 | int y; | ||
1134 |
1/2✓ Branch 0 taken 7696 times.
✗ Branch 1 not taken.
|
7696 | if (*tb++ == 1) { |
1135 | /** multiply values with the decorrelation_matrix */ | ||
1136 |
2/2✓ Branch 0 taken 522368 times.
✓ Branch 1 taken 7696 times.
|
530064 | for (y = sfb[0]; y < FFMIN(sfb[1], s->subframe_len); y++) { |
1137 | 522368 | const float* mat = s->chgroup[i].decorrelation_matrix; | |
1138 | 522368 | const float* data_end = data + num_channels; | |
1139 | 522368 | float* data_ptr = data; | |
1140 | float** ch; | ||
1141 | |||
1142 |
2/2✓ Branch 0 taken 1206528 times.
✓ Branch 1 taken 522368 times.
|
1728896 | for (ch = ch_data; ch < ch_end; ch++) |
1143 | 1206528 | *data_ptr++ = (*ch)[y]; | |
1144 | |||
1145 |
2/2✓ Branch 0 taken 1206528 times.
✓ Branch 1 taken 522368 times.
|
1728896 | for (ch = ch_data; ch < ch_end; ch++) { |
1146 | 1206528 | float sum = 0; | |
1147 | 1206528 | data_ptr = data; | |
1148 |
2/2✓ Branch 0 taken 2898432 times.
✓ Branch 1 taken 1206528 times.
|
4104960 | while (data_ptr < data_end) |
1149 | 2898432 | sum += *data_ptr++ * *mat++; | |
1150 | |||
1151 | 1206528 | (*ch)[y] = sum; | |
1152 | } | ||
1153 | } | ||
1154 | ✗ | } else if (s->nb_channels == 2) { | |
1155 | ✗ | int len = FFMIN(sfb[1], s->subframe_len) - sfb[0]; | |
1156 | ✗ | s->fdsp->vector_fmul_scalar(ch_data[0] + sfb[0], | |
1157 | ✗ | ch_data[0] + sfb[0], | |
1158 | 181.0 / 128, len); | ||
1159 | ✗ | s->fdsp->vector_fmul_scalar(ch_data[1] + sfb[0], | |
1160 | ✗ | ch_data[1] + sfb[0], | |
1161 | 181.0 / 128, len); | ||
1162 | } | ||
1163 | } | ||
1164 | } | ||
1165 | } | ||
1166 | 521 | } | |
1167 | |||
1168 | /** | ||
1169 | *@brief Apply sine window and reconstruct the output buffer. | ||
1170 | *@param s codec context | ||
1171 | */ | ||
1172 | 527 | static void wmapro_window(WMAProDecodeCtx *s) | |
1173 | { | ||
1174 | int i; | ||
1175 |
2/2✓ Branch 0 taken 1234 times.
✓ Branch 1 taken 527 times.
|
1761 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
1176 | 1234 | int c = s->channel_indexes_for_cur_subframe[i]; | |
1177 | const float* window; | ||
1178 | 1234 | int winlen = s->channel[c].prev_block_len; | |
1179 | 1234 | float* start = s->channel[c].coeffs - (winlen >> 1); | |
1180 | |||
1181 |
2/2✓ Branch 0 taken 142 times.
✓ Branch 1 taken 1092 times.
|
1234 | if (s->subframe_len < winlen) { |
1182 | 142 | start += (winlen - s->subframe_len) >> 1; | |
1183 | 142 | winlen = s->subframe_len; | |
1184 | } | ||
1185 | |||
1186 | 1234 | window = s->windows[av_log2(winlen) - WMAPRO_BLOCK_MIN_BITS]; | |
1187 | |||
1188 | 1234 | winlen >>= 1; | |
1189 | |||
1190 | 1234 | s->fdsp->vector_fmul_window(start, start, start + winlen, | |
1191 | window, winlen); | ||
1192 | |||
1193 | 1234 | s->channel[c].prev_block_len = s->subframe_len; | |
1194 | } | ||
1195 | 527 | } | |
1196 | |||
1197 | /** | ||
1198 | *@brief Decode a single subframe (block). | ||
1199 | *@param s codec context | ||
1200 | *@return 0 on success, < 0 when decoding failed | ||
1201 | */ | ||
1202 | 527 | static int decode_subframe(WMAProDecodeCtx *s) | |
1203 | { | ||
1204 | 527 | int offset = s->samples_per_frame; | |
1205 | 527 | int subframe_len = s->samples_per_frame; | |
1206 | int i; | ||
1207 | 527 | int total_samples = s->samples_per_frame * s->nb_channels; | |
1208 | 527 | int transmit_coeffs = 0; | |
1209 | int cur_subwoofer_cutoff; | ||
1210 | |||
1211 | 527 | s->subframe_offset = get_bits_count(&s->gb); | |
1212 | |||
1213 | /** reset channel context and find the next block offset and size | ||
1214 | == the next block of the channel with the smallest number of | ||
1215 | decoded samples | ||
1216 | */ | ||
1217 |
2/2✓ Branch 0 taken 2286 times.
✓ Branch 1 taken 527 times.
|
2813 | for (i = 0; i < s->nb_channels; i++) { |
1218 | 2286 | s->channel[i].grouped = 0; | |
1219 |
2/2✓ Branch 0 taken 539 times.
✓ Branch 1 taken 1747 times.
|
2286 | if (offset > s->channel[i].decoded_samples) { |
1220 | 539 | offset = s->channel[i].decoded_samples; | |
1221 | 539 | subframe_len = | |
1222 | 539 | s->channel[i].subframe_len[s->channel[i].cur_subframe]; | |
1223 | } | ||
1224 | } | ||
1225 | |||
1226 | ff_dlog(s->avctx, | ||
1227 | "processing subframe with offset %i len %i\n", offset, subframe_len); | ||
1228 | |||
1229 | /** get a list of all channels that contain the estimated block */ | ||
1230 | 527 | s->channels_for_cur_subframe = 0; | |
1231 |
2/2✓ Branch 0 taken 2286 times.
✓ Branch 1 taken 527 times.
|
2813 | for (i = 0; i < s->nb_channels; i++) { |
1232 | 2286 | const int cur_subframe = s->channel[i].cur_subframe; | |
1233 | /** subtract already processed samples */ | ||
1234 | 2286 | total_samples -= s->channel[i].decoded_samples; | |
1235 | |||
1236 | /** and count if there are multiple subframes that match our profile */ | ||
1237 |
2/2✓ Branch 0 taken 1277 times.
✓ Branch 1 taken 1009 times.
|
2286 | if (offset == s->channel[i].decoded_samples && |
1238 |
2/2✓ Branch 0 taken 1234 times.
✓ Branch 1 taken 43 times.
|
1277 | subframe_len == s->channel[i].subframe_len[cur_subframe]) { |
1239 | 1234 | total_samples -= s->channel[i].subframe_len[cur_subframe]; | |
1240 | 1234 | s->channel[i].decoded_samples += | |
1241 | 1234 | s->channel[i].subframe_len[cur_subframe]; | |
1242 | 1234 | s->channel_indexes_for_cur_subframe[s->channels_for_cur_subframe] = i; | |
1243 | 1234 | ++s->channels_for_cur_subframe; | |
1244 | } | ||
1245 | } | ||
1246 | |||
1247 | /** check if the frame will be complete after processing the | ||
1248 | estimated block */ | ||
1249 |
2/2✓ Branch 0 taken 257 times.
✓ Branch 1 taken 270 times.
|
527 | if (!total_samples) |
1250 | 257 | s->parsed_all_subframes = 1; | |
1251 | |||
1252 | |||
1253 | ff_dlog(s->avctx, "subframe is part of %i channels\n", | ||
1254 | s->channels_for_cur_subframe); | ||
1255 | |||
1256 | /** calculate number of scale factor bands and their offsets */ | ||
1257 | 527 | s->table_idx = av_log2(s->samples_per_frame/subframe_len); | |
1258 | 527 | s->num_bands = s->num_sfb[s->table_idx]; | |
1259 | 527 | s->cur_sfb_offsets = s->sfb_offsets[s->table_idx]; | |
1260 | 527 | cur_subwoofer_cutoff = s->subwoofer_cutoffs[s->table_idx]; | |
1261 | |||
1262 | /** configure the decoder for the current subframe */ | ||
1263 | 527 | offset += s->samples_per_frame >> 1; | |
1264 | |||
1265 |
2/2✓ Branch 0 taken 1234 times.
✓ Branch 1 taken 527 times.
|
1761 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
1266 | 1234 | int c = s->channel_indexes_for_cur_subframe[i]; | |
1267 | |||
1268 | 1234 | s->channel[c].coeffs = &s->channel[c].out[offset]; | |
1269 | } | ||
1270 | |||
1271 | 527 | s->subframe_len = subframe_len; | |
1272 | 527 | s->esc_len = av_log2(s->subframe_len - 1) + 1; | |
1273 | |||
1274 | /** skip extended header if any */ | ||
1275 |
2/2✓ Branch 1 taken 158 times.
✓ Branch 2 taken 369 times.
|
527 | if (get_bits1(&s->gb)) { |
1276 | int num_fill_bits; | ||
1277 |
1/2✓ Branch 1 taken 158 times.
✗ Branch 2 not taken.
|
158 | if (!(num_fill_bits = get_bits(&s->gb, 2))) { |
1278 | 158 | int len = get_bits(&s->gb, 4); | |
1279 | 158 | num_fill_bits = get_bitsz(&s->gb, len) + 1; | |
1280 | } | ||
1281 | |||
1282 |
1/2✓ Branch 0 taken 158 times.
✗ Branch 1 not taken.
|
158 | if (num_fill_bits >= 0) { |
1283 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 158 times.
|
158 | if (get_bits_count(&s->gb) + num_fill_bits > s->num_saved_bits) { |
1284 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "invalid number of fill bits\n"); | |
1285 | ✗ | return AVERROR_INVALIDDATA; | |
1286 | } | ||
1287 | |||
1288 | 158 | skip_bits_long(&s->gb, num_fill_bits); | |
1289 | } | ||
1290 | } | ||
1291 | |||
1292 | /** no idea for what the following bit is used */ | ||
1293 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 527 times.
|
527 | if (get_bits1(&s->gb)) { |
1294 | ✗ | avpriv_request_sample(s->avctx, "Reserved bit"); | |
1295 | ✗ | return AVERROR_PATCHWELCOME; | |
1296 | } | ||
1297 | |||
1298 | |||
1299 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 527 times.
|
527 | if (decode_channel_transform(s) < 0) |
1300 | ✗ | return AVERROR_INVALIDDATA; | |
1301 | |||
1302 | |||
1303 |
2/2✓ Branch 0 taken 1234 times.
✓ Branch 1 taken 527 times.
|
1761 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
1304 | 1234 | int c = s->channel_indexes_for_cur_subframe[i]; | |
1305 |
2/2✓ Branch 1 taken 1038 times.
✓ Branch 2 taken 196 times.
|
1234 | if ((s->channel[c].transmit_coefs = get_bits1(&s->gb))) |
1306 | 1038 | transmit_coeffs = 1; | |
1307 | } | ||
1308 | |||
1309 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 527 times.
|
527 | av_assert0(s->subframe_len <= WMAPRO_BLOCK_MAX_SIZE); |
1310 |
2/2✓ Branch 0 taken 521 times.
✓ Branch 1 taken 6 times.
|
527 | if (transmit_coeffs) { |
1311 | int step; | ||
1312 | 521 | int quant_step = 90 * s->bits_per_sample >> 4; | |
1313 | |||
1314 | /** decode number of vector coded coefficients */ | ||
1315 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 521 times.
|
521 | if ((s->transmit_num_vec_coeffs = get_bits1(&s->gb))) { |
1316 | ✗ | int num_bits = av_log2((s->subframe_len + 3)/4) + 1; | |
1317 | ✗ | for (i = 0; i < s->channels_for_cur_subframe; i++) { | |
1318 | ✗ | int c = s->channel_indexes_for_cur_subframe[i]; | |
1319 | ✗ | int num_vec_coeffs = get_bits(&s->gb, num_bits) << 2; | |
1320 | ✗ | if (num_vec_coeffs > s->subframe_len) { | |
1321 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "num_vec_coeffs %d is too large\n", num_vec_coeffs); | |
1322 | ✗ | return AVERROR_INVALIDDATA; | |
1323 | } | ||
1324 | ✗ | av_assert0(num_vec_coeffs + offset <= FF_ARRAY_ELEMS(s->channel[c].out)); | |
1325 | ✗ | s->channel[c].num_vec_coeffs = num_vec_coeffs; | |
1326 | } | ||
1327 | } else { | ||
1328 |
2/2✓ Branch 0 taken 1224 times.
✓ Branch 1 taken 521 times.
|
1745 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
1329 | 1224 | int c = s->channel_indexes_for_cur_subframe[i]; | |
1330 | 1224 | s->channel[c].num_vec_coeffs = s->subframe_len; | |
1331 | } | ||
1332 | } | ||
1333 | /** decode quantization step */ | ||
1334 | 521 | step = get_sbits(&s->gb, 6); | |
1335 | 521 | quant_step += step; | |
1336 |
3/4✓ Branch 0 taken 498 times.
✓ Branch 1 taken 23 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 498 times.
|
521 | if (step == -32 || step == 31) { |
1337 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 23 times.
|
23 | const int sign = (step == 31) - 1; |
1338 | 23 | int quant = 0; | |
1339 |
1/2✓ Branch 1 taken 23 times.
✗ Branch 2 not taken.
|
23 | while (get_bits_count(&s->gb) + 5 < s->num_saved_bits && |
1340 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 23 times.
|
23 | (step = get_bits(&s->gb, 5)) == 31) { |
1341 | ✗ | quant += 31; | |
1342 | } | ||
1343 | 23 | quant_step += ((quant + step) ^ sign) - sign; | |
1344 | } | ||
1345 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 521 times.
|
521 | if (quant_step < 0) { |
1346 | ✗ | av_log(s->avctx, AV_LOG_DEBUG, "negative quant step\n"); | |
1347 | } | ||
1348 | |||
1349 | /** decode quantization step modifiers for every channel */ | ||
1350 | |||
1351 |
2/2✓ Branch 0 taken 194 times.
✓ Branch 1 taken 327 times.
|
521 | if (s->channels_for_cur_subframe == 1) { |
1352 | 194 | s->channel[s->channel_indexes_for_cur_subframe[0]].quant_step = quant_step; | |
1353 | } else { | ||
1354 | 327 | int modifier_len = get_bits(&s->gb, 3); | |
1355 |
2/2✓ Branch 0 taken 1030 times.
✓ Branch 1 taken 327 times.
|
1357 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
1356 | 1030 | int c = s->channel_indexes_for_cur_subframe[i]; | |
1357 | 1030 | s->channel[c].quant_step = quant_step; | |
1358 |
2/2✓ Branch 1 taken 511 times.
✓ Branch 2 taken 519 times.
|
1030 | if (get_bits1(&s->gb)) { |
1359 |
2/2✓ Branch 0 taken 458 times.
✓ Branch 1 taken 53 times.
|
511 | if (modifier_len) { |
1360 | 458 | s->channel[c].quant_step += get_bits(&s->gb, modifier_len) + 1; | |
1361 | } else | ||
1362 | 53 | ++s->channel[c].quant_step; | |
1363 | } | ||
1364 | } | ||
1365 | } | ||
1366 | |||
1367 | /** decode scale factors */ | ||
1368 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 521 times.
|
521 | if (decode_scale_factors(s) < 0) |
1369 | ✗ | return AVERROR_INVALIDDATA; | |
1370 | } | ||
1371 | |||
1372 | ff_dlog(s->avctx, "BITSTREAM: subframe header length was %i\n", | ||
1373 | get_bits_count(&s->gb) - s->subframe_offset); | ||
1374 | |||
1375 | /** parse coefficients */ | ||
1376 |
2/2✓ Branch 0 taken 1234 times.
✓ Branch 1 taken 527 times.
|
1761 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
1377 | 1234 | int c = s->channel_indexes_for_cur_subframe[i]; | |
1378 |
2/2✓ Branch 0 taken 1038 times.
✓ Branch 1 taken 196 times.
|
1234 | if (s->channel[c].transmit_coefs && |
1379 |
1/2✓ Branch 1 taken 1038 times.
✗ Branch 2 not taken.
|
1038 | get_bits_count(&s->gb) < s->num_saved_bits) { |
1380 | 1038 | decode_coeffs(s, c); | |
1381 | } else | ||
1382 | 196 | memset(s->channel[c].coeffs, 0, | |
1383 | sizeof(*s->channel[c].coeffs) * subframe_len); | ||
1384 | } | ||
1385 | |||
1386 | ff_dlog(s->avctx, "BITSTREAM: subframe length was %i\n", | ||
1387 | get_bits_count(&s->gb) - s->subframe_offset); | ||
1388 | |||
1389 |
2/2✓ Branch 0 taken 521 times.
✓ Branch 1 taken 6 times.
|
527 | if (transmit_coeffs) { |
1390 | 521 | AVTXContext *tx = s->tx[av_log2(subframe_len) - WMAPRO_BLOCK_MIN_BITS]; | |
1391 | 521 | av_tx_fn tx_fn = s->tx_fn[av_log2(subframe_len) - WMAPRO_BLOCK_MIN_BITS]; | |
1392 | /** reconstruct the per channel data */ | ||
1393 | 521 | inverse_channel_transform(s); | |
1394 |
2/2✓ Branch 0 taken 1224 times.
✓ Branch 1 taken 521 times.
|
1745 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
1395 | 1224 | int c = s->channel_indexes_for_cur_subframe[i]; | |
1396 | 1224 | const int* sf = s->channel[c].scale_factors; | |
1397 | int b; | ||
1398 | |||
1399 |
2/2✓ Branch 0 taken 103 times.
✓ Branch 1 taken 1121 times.
|
1224 | if (c == s->lfe_channel) |
1400 | 103 | memset(&s->tmp[cur_subwoofer_cutoff], 0, sizeof(*s->tmp) * | |
1401 | 103 | (subframe_len - cur_subwoofer_cutoff)); | |
1402 | |||
1403 | /** inverse quantization and rescaling */ | ||
1404 |
2/2✓ Branch 0 taken 29004 times.
✓ Branch 1 taken 1224 times.
|
30228 | for (b = 0; b < s->num_bands; b++) { |
1405 | 29004 | const int end = FFMIN(s->cur_sfb_offsets[b+1], s->subframe_len); | |
1406 | 29004 | const int exp = s->channel[c].quant_step - | |
1407 | 29004 | (s->channel[c].max_scale_factor - *sf++) * | |
1408 | 29004 | s->channel[c].scale_factor_step; | |
1409 | 29004 | const float quant = ff_exp10(exp / 20.0); | |
1410 | 29004 | int start = s->cur_sfb_offsets[b]; | |
1411 | 29004 | s->fdsp->vector_fmul_scalar(s->tmp + start, | |
1412 | 29004 | s->channel[c].coeffs + start, | |
1413 | quant, end - start); | ||
1414 | } | ||
1415 | |||
1416 | /** apply imdct (imdct_half == DCTIV with reverse) */ | ||
1417 | 1224 | tx_fn(tx, s->channel[c].coeffs, s->tmp, sizeof(float)); | |
1418 | } | ||
1419 | } | ||
1420 | |||
1421 | /** window and overlapp-add */ | ||
1422 | 527 | wmapro_window(s); | |
1423 | |||
1424 | /** handled one subframe */ | ||
1425 |
2/2✓ Branch 0 taken 1234 times.
✓ Branch 1 taken 527 times.
|
1761 | for (i = 0; i < s->channels_for_cur_subframe; i++) { |
1426 | 1234 | int c = s->channel_indexes_for_cur_subframe[i]; | |
1427 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1234 times.
|
1234 | if (s->channel[c].cur_subframe >= s->channel[c].num_subframes) { |
1428 | ✗ | av_log(s->avctx, AV_LOG_ERROR, "broken subframe\n"); | |
1429 | ✗ | return AVERROR_INVALIDDATA; | |
1430 | } | ||
1431 | 1234 | ++s->channel[c].cur_subframe; | |
1432 | } | ||
1433 | |||
1434 | 527 | return 0; | |
1435 | } | ||
1436 | |||
1437 | /** | ||
1438 | *@brief Decode one WMA frame. | ||
1439 | *@param s codec context | ||
1440 | *@return 0 if the trailer bit indicates that this is the last frame, | ||
1441 | * 1 if there are additional frames | ||
1442 | */ | ||
1443 | 257 | static int decode_frame(WMAProDecodeCtx *s, AVFrame *frame, int *got_frame_ptr) | |
1444 | { | ||
1445 | 257 | GetBitContext* gb = &s->gb; | |
1446 | 257 | int more_frames = 0; | |
1447 | 257 | int len = 0; | |
1448 | int i; | ||
1449 | |||
1450 | /** get frame length */ | ||
1451 |
1/2✓ Branch 0 taken 257 times.
✗ Branch 1 not taken.
|
257 | if (s->len_prefix) |
1452 | 257 | len = get_bits(gb, s->log2_frame_size); | |
1453 | |||
1454 | ff_dlog(s->avctx, "decoding frame with length %x\n", len); | ||
1455 | |||
1456 | /** decode tile information */ | ||
1457 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 257 times.
|
257 | if (decode_tilehdr(s)) { |
1458 | ✗ | s->packet_loss = 1; | |
1459 | ✗ | return 0; | |
1460 | } | ||
1461 | |||
1462 | /** read postproc transform */ | ||
1463 |
2/4✓ Branch 0 taken 257 times.
✗ Branch 1 not taken.
✗ Branch 3 not taken.
✓ Branch 4 taken 257 times.
|
257 | if (s->nb_channels > 1 && get_bits1(gb)) { |
1464 | ✗ | if (get_bits1(gb)) { | |
1465 | ✗ | for (i = 0; i < s->nb_channels * s->nb_channels; i++) | |
1466 | ✗ | skip_bits(gb, 4); | |
1467 | } | ||
1468 | } | ||
1469 | |||
1470 | /** read drc info */ | ||
1471 |
1/2✓ Branch 0 taken 257 times.
✗ Branch 1 not taken.
|
257 | if (s->dynamic_range_compression) { |
1472 | 257 | s->drc_gain = get_bits(gb, 8); | |
1473 | ff_dlog(s->avctx, "drc_gain %i\n", s->drc_gain); | ||
1474 | } | ||
1475 | |||
1476 |
2/2✓ Branch 1 taken 4 times.
✓ Branch 2 taken 253 times.
|
257 | if (get_bits1(gb)) { |
1477 |
2/2✓ Branch 1 taken 3 times.
✓ Branch 2 taken 1 times.
|
4 | if (get_bits1(gb)) |
1478 | 3 | s->trim_start = get_bits(gb, av_log2(s->samples_per_frame * 2)); | |
1479 | |||
1480 |
2/2✓ Branch 1 taken 1 times.
✓ Branch 2 taken 3 times.
|
4 | if (get_bits1(gb)) |
1481 | 1 | s->trim_end = get_bits(gb, av_log2(s->samples_per_frame * 2)); | |
1482 | } else { | ||
1483 | 253 | s->trim_start = s->trim_end = 0; | |
1484 | } | ||
1485 | |||
1486 | ff_dlog(s->avctx, "BITSTREAM: frame header length was %i\n", | ||
1487 | get_bits_count(gb) - s->frame_offset); | ||
1488 | |||
1489 | /** reset subframe states */ | ||
1490 | 257 | s->parsed_all_subframes = 0; | |
1491 |
2/2✓ Branch 0 taken 926 times.
✓ Branch 1 taken 257 times.
|
1183 | for (i = 0; i < s->nb_channels; i++) { |
1492 | 926 | s->channel[i].decoded_samples = 0; | |
1493 | 926 | s->channel[i].cur_subframe = 0; | |
1494 | 926 | s->channel[i].reuse_sf = 0; | |
1495 | } | ||
1496 | |||
1497 | /** decode all subframes */ | ||
1498 |
2/2✓ Branch 0 taken 527 times.
✓ Branch 1 taken 257 times.
|
784 | while (!s->parsed_all_subframes) { |
1499 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 527 times.
|
527 | if (decode_subframe(s) < 0) { |
1500 | ✗ | s->packet_loss = 1; | |
1501 | ✗ | return 0; | |
1502 | } | ||
1503 | } | ||
1504 | |||
1505 | /** copy samples to the output buffer */ | ||
1506 |
2/2✓ Branch 0 taken 926 times.
✓ Branch 1 taken 257 times.
|
1183 | for (i = 0; i < s->nb_channels; i++) |
1507 | 926 | memcpy(frame->extended_data[i], s->channel[i].out, | |
1508 | 926 | s->samples_per_frame * sizeof(*s->channel[i].out)); | |
1509 | |||
1510 |
2/2✓ Branch 0 taken 926 times.
✓ Branch 1 taken 257 times.
|
1183 | for (i = 0; i < s->nb_channels; i++) { |
1511 | /** reuse second half of the IMDCT output for the next frame */ | ||
1512 | 926 | memcpy(&s->channel[i].out[0], | |
1513 | 926 | &s->channel[i].out[s->samples_per_frame], | |
1514 | 926 | s->samples_per_frame * sizeof(*s->channel[i].out) >> 1); | |
1515 | } | ||
1516 | |||
1517 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 254 times.
|
257 | if (s->skip_frame) { |
1518 | 3 | s->skip_frame = 0; | |
1519 | 3 | *got_frame_ptr = 0; | |
1520 | 3 | av_frame_unref(frame); | |
1521 | } else { | ||
1522 | 254 | *got_frame_ptr = 1; | |
1523 | } | ||
1524 | |||
1525 |
1/2✓ Branch 0 taken 257 times.
✗ Branch 1 not taken.
|
257 | if (s->len_prefix) { |
1526 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 257 times.
|
257 | if (len != (get_bits_count(gb) - s->frame_offset) + 2) { |
1527 | /** FIXME: not sure if this is always an error */ | ||
1528 | ✗ | av_log(s->avctx, AV_LOG_ERROR, | |
1529 | "frame[%"PRIu32"] would have to skip %i bits\n", | ||
1530 | s->frame_num, | ||
1531 | ✗ | len - (get_bits_count(gb) - s->frame_offset) - 1); | |
1532 | ✗ | s->packet_loss = 1; | |
1533 | ✗ | return 0; | |
1534 | } | ||
1535 | |||
1536 | /** skip the rest of the frame data */ | ||
1537 | 257 | skip_bits_long(gb, len - (get_bits_count(gb) - s->frame_offset) - 1); | |
1538 | } else { | ||
1539 | ✗ | while (get_bits_count(gb) < s->num_saved_bits && get_bits1(gb) == 0) { | |
1540 | } | ||
1541 | } | ||
1542 | |||
1543 | /** decode trailer bit */ | ||
1544 | 257 | more_frames = get_bits1(gb); | |
1545 | |||
1546 | 257 | ++s->frame_num; | |
1547 | 257 | return more_frames; | |
1548 | } | ||
1549 | |||
1550 | /** | ||
1551 | *@brief Calculate remaining input buffer length. | ||
1552 | *@param s codec context | ||
1553 | *@param gb bitstream reader context | ||
1554 | *@return remaining size in bits | ||
1555 | */ | ||
1556 | 849 | static int remaining_bits(WMAProDecodeCtx *s, GetBitContext *gb) | |
1557 | { | ||
1558 | 849 | return s->buf_bit_size - get_bits_count(gb); | |
1559 | } | ||
1560 | |||
1561 | /** | ||
1562 | *@brief Fill the bit reservoir with a (partial) frame. | ||
1563 | *@param s codec context | ||
1564 | *@param gb bitstream reader context | ||
1565 | *@param len length of the partial frame | ||
1566 | *@param append decides whether to reset the buffer or not | ||
1567 | */ | ||
1568 | 281 | static void save_bits(WMAProDecodeCtx *s, GetBitContext* gb, int len, | |
1569 | int append) | ||
1570 | { | ||
1571 | int buflen; | ||
1572 | |||
1573 | /** when the frame data does not need to be concatenated, the input buffer | ||
1574 | is reset and additional bits from the previous frame are copied | ||
1575 | and skipped later so that a fast byte copy is possible */ | ||
1576 | |||
1577 |
2/2✓ Branch 0 taken 260 times.
✓ Branch 1 taken 21 times.
|
281 | if (!append) { |
1578 | 260 | s->frame_offset = get_bits_count(gb) & 7; | |
1579 | 260 | s->num_saved_bits = s->frame_offset; | |
1580 | 260 | init_put_bits(&s->pb, s->frame_data, MAX_FRAMESIZE); | |
1581 | 260 | buflen = (s->num_saved_bits + len + 7) >> 3; | |
1582 | } else | ||
1583 | 21 | buflen = (put_bits_count(&s->pb) + len + 7) >> 3; | |
1584 | |||
1585 |
2/4✓ Branch 0 taken 281 times.
✗ Branch 1 not taken.
✗ Branch 2 not taken.
✓ Branch 3 taken 281 times.
|
281 | if (len <= 0 || buflen > MAX_FRAMESIZE) { |
1586 | ✗ | avpriv_request_sample(s->avctx, "Too small input buffer"); | |
1587 | ✗ | s->packet_loss = 1; | |
1588 | ✗ | return; | |
1589 | } | ||
1590 | |||
1591 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 281 times.
|
281 | av_assert0(len <= put_bits_left(&s->pb)); |
1592 | |||
1593 | 281 | s->num_saved_bits += len; | |
1594 |
2/2✓ Branch 0 taken 260 times.
✓ Branch 1 taken 21 times.
|
281 | if (!append) { |
1595 | 260 | ff_copy_bits(&s->pb, gb->buffer + (get_bits_count(gb) >> 3), | |
1596 | s->num_saved_bits); | ||
1597 | } else { | ||
1598 | 21 | int align = 8 - (get_bits_count(gb) & 7); | |
1599 | 21 | align = FFMIN(align, len); | |
1600 | 21 | put_bits(&s->pb, align, get_bits(gb, align)); | |
1601 | 21 | len -= align; | |
1602 | 21 | ff_copy_bits(&s->pb, gb->buffer + (get_bits_count(gb) >> 3), len); | |
1603 | } | ||
1604 | 281 | skip_bits_long(gb, len); | |
1605 | |||
1606 | { | ||
1607 | 281 | PutBitContext tmp = s->pb; | |
1608 | 281 | flush_put_bits(&tmp); | |
1609 | } | ||
1610 | |||
1611 | 281 | init_get_bits(&s->gb, s->frame_data, s->num_saved_bits); | |
1612 | 281 | skip_bits(&s->gb, s->frame_offset); | |
1613 | } | ||
1614 | |||
1615 | 285 | static int decode_packet(AVCodecContext *avctx, WMAProDecodeCtx *s, | |
1616 | AVFrame *frame, int *got_frame_ptr, AVPacket *avpkt) | ||
1617 | { | ||
1618 | 285 | GetBitContext* gb = &s->pgb; | |
1619 | 285 | const uint8_t* buf = avpkt->data; | |
1620 | 285 | int buf_size = avpkt->size; | |
1621 | int num_bits_prev_frame; | ||
1622 | int packet_sequence_number; | ||
1623 | int ret; | ||
1624 | |||
1625 | 285 | *got_frame_ptr = 0; | |
1626 | |||
1627 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 285 times.
|
285 | if (!buf_size) { |
1628 | int i; | ||
1629 | |||
1630 | /** Must output remaining samples after stream end. WMAPRO 5.1 created | ||
1631 | * by XWMA encoder don't though (maybe only 1/2ch streams need it). */ | ||
1632 | ✗ | s->packet_done = 0; | |
1633 | ✗ | if (s->eof_done) | |
1634 | ✗ | return 0; | |
1635 | |||
1636 | /** clean output buffer and copy last IMDCT samples */ | ||
1637 | ✗ | for (i = 0; i < s->nb_channels; i++) { | |
1638 | ✗ | memset(frame->extended_data[i], 0, | |
1639 | ✗ | s->samples_per_frame * sizeof(*s->channel[i].out)); | |
1640 | |||
1641 | ✗ | memcpy(frame->extended_data[i], s->channel[i].out, | |
1642 | ✗ | s->samples_per_frame * sizeof(*s->channel[i].out) >> 1); | |
1643 | } | ||
1644 | |||
1645 | ✗ | s->eof_done = 1; | |
1646 | ✗ | s->packet_done = 1; | |
1647 | ✗ | *got_frame_ptr = 1; | |
1648 | ✗ | return 0; | |
1649 | } | ||
1650 |
4/4✓ Branch 0 taken 262 times.
✓ Branch 1 taken 23 times.
✓ Branch 2 taken 3 times.
✓ Branch 3 taken 259 times.
|
285 | else if (s->packet_done || s->packet_loss) { |
1651 | 26 | s->packet_done = 0; | |
1652 | |||
1653 | /** sanity check for the buffer length */ | ||
1654 |
3/4✓ Branch 0 taken 26 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 2 times.
✓ Branch 3 taken 24 times.
|
26 | if (avctx->codec_id == AV_CODEC_ID_WMAPRO && buf_size < avctx->block_align) { |
1655 | 2 | av_log(avctx, AV_LOG_ERROR, "Input packet too small (%d < %d)\n", | |
1656 | buf_size, avctx->block_align); | ||
1657 | 2 | s->packet_loss = 1; | |
1658 | 2 | return AVERROR_INVALIDDATA; | |
1659 | } | ||
1660 | |||
1661 |
1/2✓ Branch 0 taken 24 times.
✗ Branch 1 not taken.
|
24 | if (avctx->codec_id == AV_CODEC_ID_WMAPRO) { |
1662 | 24 | s->next_packet_start = buf_size - avctx->block_align; | |
1663 | 24 | buf_size = avctx->block_align; | |
1664 | } else { | ||
1665 | ✗ | s->next_packet_start = buf_size - FFMIN(buf_size, avctx->block_align); | |
1666 | ✗ | buf_size = FFMIN(buf_size, avctx->block_align); | |
1667 | } | ||
1668 | 24 | s->buf_bit_size = buf_size << 3; | |
1669 | |||
1670 | /** parse packet header */ | ||
1671 | 24 | ret = init_get_bits8(gb, buf, buf_size); | |
1672 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 24 times.
|
24 | if (ret < 0) |
1673 | ✗ | return ret; | |
1674 |
1/2✓ Branch 0 taken 24 times.
✗ Branch 1 not taken.
|
24 | if (avctx->codec_id != AV_CODEC_ID_XMA2) { |
1675 | 24 | packet_sequence_number = get_bits(gb, 4); | |
1676 | 24 | skip_bits(gb, 2); | |
1677 | } else { | ||
1678 | ✗ | int num_frames = get_bits(gb, 6); | |
1679 | ff_dlog(avctx, "packet[%"PRId64"]: number of frames %d\n", avctx->frame_num, num_frames); | ||
1680 | ✗ | packet_sequence_number = 0; | |
1681 | } | ||
1682 | |||
1683 | /** get number of bits that need to be added to the previous frame */ | ||
1684 | 24 | num_bits_prev_frame = get_bits(gb, s->log2_frame_size); | |
1685 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 24 times.
|
24 | if (avctx->codec_id != AV_CODEC_ID_WMAPRO) { |
1686 | ✗ | skip_bits(gb, 3); | |
1687 | ✗ | s->skip_packets = get_bits(gb, 8); | |
1688 | ff_dlog(avctx, "packet[%"PRId64"]: skip packets %d\n", avctx->frame_num, s->skip_packets); | ||
1689 | } | ||
1690 | |||
1691 | ff_dlog(avctx, "packet[%"PRId64"]: nbpf %x\n", avctx->frame_num, | ||
1692 | num_bits_prev_frame); | ||
1693 | |||
1694 | /** check for packet loss */ | ||
1695 |
3/4✓ Branch 0 taken 24 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 21 times.
✓ Branch 3 taken 3 times.
|
24 | if (avctx->codec_id == AV_CODEC_ID_WMAPRO && !s->packet_loss && |
1696 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 21 times.
|
21 | ((s->packet_sequence_number + 1) & 0xF) != packet_sequence_number) { |
1697 | ✗ | s->packet_loss = 1; | |
1698 | ✗ | av_log(avctx, AV_LOG_ERROR, | |
1699 | "Packet loss detected! seq %"PRIx8" vs %x\n", | ||
1700 | ✗ | s->packet_sequence_number, packet_sequence_number); | |
1701 | } | ||
1702 | 24 | s->packet_sequence_number = packet_sequence_number; | |
1703 | |||
1704 |
2/2✓ Branch 0 taken 21 times.
✓ Branch 1 taken 3 times.
|
24 | if (num_bits_prev_frame > 0) { |
1705 | 21 | int remaining_packet_bits = s->buf_bit_size - get_bits_count(gb); | |
1706 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 21 times.
|
21 | if (num_bits_prev_frame >= remaining_packet_bits) { |
1707 | ✗ | num_bits_prev_frame = remaining_packet_bits; | |
1708 | ✗ | s->packet_done = 1; | |
1709 | } | ||
1710 | |||
1711 | /** append the previous frame data to the remaining data from the | ||
1712 | previous packet to create a full frame */ | ||
1713 | 21 | save_bits(s, gb, num_bits_prev_frame, 1); | |
1714 | ff_dlog(avctx, "accumulated %x bits of frame data\n", | ||
1715 | s->num_saved_bits - s->frame_offset); | ||
1716 | |||
1717 | /** decode the cross packet frame if it is valid */ | ||
1718 |
1/2✓ Branch 0 taken 21 times.
✗ Branch 1 not taken.
|
21 | if (!s->packet_loss) |
1719 | 21 | decode_frame(s, frame, got_frame_ptr); | |
1720 | 3 | } else if (s->num_saved_bits - s->frame_offset) { | |
1721 | ff_dlog(avctx, "ignoring %x previously saved bits\n", | ||
1722 | s->num_saved_bits - s->frame_offset); | ||
1723 | } | ||
1724 | |||
1725 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 21 times.
|
24 | if (s->packet_loss) { |
1726 | /** reset number of saved bits so that the decoder | ||
1727 | does not start to decode incomplete frames in the | ||
1728 | s->len_prefix == 0 case */ | ||
1729 | 3 | s->num_saved_bits = 0; | |
1730 | 3 | s->packet_loss = 0; | |
1731 | } | ||
1732 | } else { | ||
1733 | int frame_size; | ||
1734 | |||
1735 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 259 times.
|
259 | if (avpkt->size < s->next_packet_start) { |
1736 | ✗ | s->packet_loss = 1; | |
1737 | ✗ | return AVERROR_INVALIDDATA; | |
1738 | } | ||
1739 | |||
1740 | 259 | s->buf_bit_size = (avpkt->size - s->next_packet_start) << 3; | |
1741 | 259 | ret = init_get_bits8(gb, avpkt->data, avpkt->size - s->next_packet_start); | |
1742 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 259 times.
|
259 | if (ret < 0) |
1743 | ✗ | return ret; | |
1744 | 259 | skip_bits(gb, s->packet_offset); | |
1745 |
2/4✓ Branch 0 taken 259 times.
✗ Branch 1 not taken.
✓ Branch 3 taken 259 times.
✗ Branch 4 not taken.
|
259 | if (s->len_prefix && remaining_bits(s, gb) > s->log2_frame_size && |
1746 |
3/4✓ Branch 1 taken 259 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 236 times.
✓ Branch 4 taken 23 times.
|
518 | (frame_size = show_bits(gb, s->log2_frame_size)) && |
1747 | 259 | frame_size <= remaining_bits(s, gb)) { | |
1748 | 236 | save_bits(s, gb, frame_size, 0); | |
1749 |
1/2✓ Branch 0 taken 236 times.
✗ Branch 1 not taken.
|
236 | if (!s->packet_loss) |
1750 | 236 | s->packet_done = !decode_frame(s, frame, got_frame_ptr); | |
1751 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 23 times.
|
23 | } else if (!s->len_prefix |
1752 | ✗ | && s->num_saved_bits > get_bits_count(&s->gb)) { | |
1753 | /** when the frames do not have a length prefix, we don't know | ||
1754 | the compressed length of the individual frames | ||
1755 | however, we know what part of a new packet belongs to the | ||
1756 | previous frame | ||
1757 | therefore we save the incoming packet first, then we append | ||
1758 | the "previous frame" data from the next packet so that | ||
1759 | we get a buffer that only contains full frames */ | ||
1760 | ✗ | s->packet_done = !decode_frame(s, frame, got_frame_ptr); | |
1761 | } else { | ||
1762 | 23 | s->packet_done = 1; | |
1763 | } | ||
1764 | } | ||
1765 | |||
1766 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 283 times.
|
283 | if (remaining_bits(s, gb) < 0) { |
1767 | ✗ | av_log(avctx, AV_LOG_ERROR, "Overread %d\n", -remaining_bits(s, gb)); | |
1768 | ✗ | s->packet_loss = 1; | |
1769 | } | ||
1770 | |||
1771 |
4/6✓ Branch 0 taken 24 times.
✓ Branch 1 taken 259 times.
✓ Branch 2 taken 24 times.
✗ Branch 3 not taken.
✓ Branch 4 taken 24 times.
✗ Branch 5 not taken.
|
307 | if (s->packet_done && !s->packet_loss && |
1772 | 24 | remaining_bits(s, gb) > 0) { | |
1773 | /** save the rest of the data so that it can be decoded | ||
1774 | with the next packet */ | ||
1775 | 24 | save_bits(s, gb, remaining_bits(s, gb), 0); | |
1776 | } | ||
1777 | |||
1778 | 283 | s->packet_offset = get_bits_count(gb) & 7; | |
1779 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 283 times.
|
283 | if (s->packet_loss) |
1780 | ✗ | return AVERROR_INVALIDDATA; | |
1781 | |||
1782 |
3/4✓ Branch 0 taken 3 times.
✓ Branch 1 taken 280 times.
✓ Branch 2 taken 3 times.
✗ Branch 3 not taken.
|
283 | if (s->trim_start && avctx->codec_id == AV_CODEC_ID_WMAPRO) { |
1783 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 3 times.
|
3 | if (s->trim_start < frame->nb_samples) { |
1784 | ✗ | for (int ch = 0; ch < frame->ch_layout.nb_channels; ch++) | |
1785 | ✗ | frame->extended_data[ch] += s->trim_start * 4; | |
1786 | |||
1787 | ✗ | frame->nb_samples -= s->trim_start; | |
1788 | } else { | ||
1789 | 3 | *got_frame_ptr = 0; | |
1790 | } | ||
1791 | |||
1792 | 3 | s->trim_start = 0; | |
1793 | } | ||
1794 | |||
1795 |
3/4✓ Branch 0 taken 1 times.
✓ Branch 1 taken 282 times.
✓ Branch 2 taken 1 times.
✗ Branch 3 not taken.
|
283 | if (s->trim_end && avctx->codec_id == AV_CODEC_ID_WMAPRO) { |
1796 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | if (s->trim_end < frame->nb_samples) { |
1797 | 1 | frame->nb_samples -= s->trim_end; | |
1798 | } else { | ||
1799 | ✗ | *got_frame_ptr = 0; | |
1800 | } | ||
1801 | |||
1802 | 1 | s->trim_end = 0; | |
1803 | } | ||
1804 | |||
1805 | 283 | return get_bits_count(gb) >> 3; | |
1806 | } | ||
1807 | |||
1808 | /** | ||
1809 | *@brief Decode a single WMA packet. | ||
1810 | *@param avctx codec context | ||
1811 | *@param data the output buffer | ||
1812 | *@param avpkt input packet | ||
1813 | *@return number of bytes that were read from the input buffer | ||
1814 | */ | ||
1815 | 285 | static int wmapro_decode_packet(AVCodecContext *avctx, AVFrame *frame, | |
1816 | int *got_frame_ptr, AVPacket *avpkt) | ||
1817 | { | ||
1818 | 285 | WMAProDecodeCtx *s = avctx->priv_data; | |
1819 | int ret; | ||
1820 | |||
1821 | /* get output buffer */ | ||
1822 | 285 | frame->nb_samples = s->samples_per_frame; | |
1823 |
1/2✗ Branch 1 not taken.
✓ Branch 2 taken 285 times.
|
285 | if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) { |
1824 | ✗ | s->packet_loss = 1; | |
1825 | ✗ | return 0; | |
1826 | } | ||
1827 | |||
1828 | 285 | return decode_packet(avctx, s, frame, got_frame_ptr, avpkt); | |
1829 | } | ||
1830 | |||
1831 | ✗ | static int xma_decode_packet(AVCodecContext *avctx, AVFrame *frame, | |
1832 | int *got_frame_ptr, AVPacket *avpkt) | ||
1833 | { | ||
1834 | ✗ | XMADecodeCtx *s = avctx->priv_data; | |
1835 | ✗ | int got_stream_frame_ptr = 0; | |
1836 | ✗ | int i, ret = 0, eof = 0; | |
1837 | |||
1838 | ✗ | if (!s->frames[s->current_stream]->data[0]) { | |
1839 | ✗ | avctx->internal->skip_samples = 64; | |
1840 | ✗ | s->frames[s->current_stream]->nb_samples = 512; | |
1841 | ✗ | if ((ret = ff_get_buffer(avctx, s->frames[s->current_stream], 0)) < 0) | |
1842 | ✗ | return ret; | |
1843 | ✗ | } else if (s->frames[s->current_stream]->nb_samples != 512) { | |
1844 | ✗ | avctx->internal->skip_samples = 64; | |
1845 | ✗ | av_frame_unref(s->frames[s->current_stream]); | |
1846 | ✗ | s->frames[s->current_stream]->nb_samples = 512; | |
1847 | ✗ | if ((ret = ff_get_buffer(avctx, s->frames[s->current_stream], 0)) < 0) | |
1848 | ✗ | return ret; | |
1849 | } | ||
1850 | /* decode current stream packet */ | ||
1851 | ✗ | if (!s->xma[s->current_stream].eof_done) { | |
1852 | ✗ | ret = decode_packet(avctx, &s->xma[s->current_stream], s->frames[s->current_stream], | |
1853 | &got_stream_frame_ptr, avpkt); | ||
1854 | } | ||
1855 | |||
1856 | ✗ | if (!avpkt->size) { | |
1857 | ✗ | eof = 1; | |
1858 | |||
1859 | ✗ | for (i = 0; i < s->num_streams; i++) { | |
1860 | ✗ | if (!s->xma[i].eof_done && s->frames[i]->data[0]) { | |
1861 | ✗ | ret = decode_packet(avctx, &s->xma[i], s->frames[i], | |
1862 | &got_stream_frame_ptr, avpkt); | ||
1863 | } | ||
1864 | |||
1865 | ✗ | eof &= s->xma[i].eof_done; | |
1866 | } | ||
1867 | } | ||
1868 | |||
1869 | ✗ | if (s->xma[0].trim_start) | |
1870 | ✗ | s->trim_start = s->xma[0].trim_start; | |
1871 | ✗ | if (s->xma[0].trim_end) | |
1872 | ✗ | s->trim_end = s->xma[0].trim_end; | |
1873 | |||
1874 | /* copy stream samples (1/2ch) to sample buffer (Nch) */ | ||
1875 | ✗ | if (got_stream_frame_ptr) { | |
1876 | ✗ | const int nb_samples = s->frames[s->current_stream]->nb_samples; | |
1877 | ✗ | void *left[1] = { s->frames[s->current_stream]->extended_data[0] }; | |
1878 | ✗ | void *right[1] = { s->frames[s->current_stream]->extended_data[1] }; | |
1879 | |||
1880 | ✗ | av_audio_fifo_write(s->samples[0][s->current_stream], left, nb_samples); | |
1881 | ✗ | if (s->xma[s->current_stream].nb_channels > 1) | |
1882 | ✗ | av_audio_fifo_write(s->samples[1][s->current_stream], right, nb_samples); | |
1883 | ✗ | } else if (ret < 0) { | |
1884 | ✗ | s->current_stream = 0; | |
1885 | ✗ | return ret; | |
1886 | } | ||
1887 | |||
1888 | /* find next XMA packet's owner stream, and update. | ||
1889 | * XMA streams find their packets following packet_skips | ||
1890 | * (at start there is one packet per stream, then interleave non-linearly). */ | ||
1891 | ✗ | if (s->xma[s->current_stream].packet_done || | |
1892 | ✗ | s->xma[s->current_stream].packet_loss) { | |
1893 | ✗ | int nb_samples = INT_MAX; | |
1894 | |||
1895 | /* select stream with 0 skip_packets (= uses next packet) */ | ||
1896 | ✗ | if (s->xma[s->current_stream].skip_packets != 0) { | |
1897 | int min[2]; | ||
1898 | |||
1899 | ✗ | min[0] = s->xma[0].skip_packets; | |
1900 | ✗ | min[1] = i = 0; | |
1901 | |||
1902 | ✗ | for (i = 1; i < s->num_streams; i++) { | |
1903 | ✗ | if (s->xma[i].skip_packets < min[0]) { | |
1904 | ✗ | min[0] = s->xma[i].skip_packets; | |
1905 | ✗ | min[1] = i; | |
1906 | } | ||
1907 | } | ||
1908 | |||
1909 | ✗ | s->current_stream = min[1]; | |
1910 | } | ||
1911 | |||
1912 | /* all other streams skip next packet */ | ||
1913 | ✗ | for (i = 0; i < s->num_streams; i++) { | |
1914 | ✗ | s->xma[i].skip_packets = FFMAX(0, s->xma[i].skip_packets - 1); | |
1915 | ✗ | nb_samples = FFMIN(nb_samples, av_audio_fifo_size(s->samples[0][i])); | |
1916 | } | ||
1917 | |||
1918 | ✗ | if (!eof && avpkt->size) | |
1919 | ✗ | nb_samples -= FFMIN(nb_samples, 4096); | |
1920 | |||
1921 | /* copy samples from buffer to output if possible */ | ||
1922 | ✗ | if ((nb_samples > 0 || eof || !avpkt->size) && !s->flushed) { | |
1923 | int bret; | ||
1924 | |||
1925 | ✗ | if (eof) { | |
1926 | ✗ | nb_samples -= av_clip(s->trim_end + s->trim_start - 128 - 64, 0, nb_samples); | |
1927 | ✗ | s->flushed = 1; | |
1928 | } | ||
1929 | |||
1930 | ✗ | frame->nb_samples = nb_samples; | |
1931 | ✗ | if ((bret = ff_get_buffer(avctx, frame, 0)) < 0) | |
1932 | ✗ | return bret; | |
1933 | |||
1934 | ✗ | for (i = 0; i < s->num_streams; i++) { | |
1935 | ✗ | const int start_ch = s->start_channel[i]; | |
1936 | ✗ | void *left[1] = { frame->extended_data[start_ch + 0] }; | |
1937 | |||
1938 | ✗ | av_audio_fifo_read(s->samples[0][i], left, nb_samples); | |
1939 | ✗ | if (s->xma[i].nb_channels > 1) { | |
1940 | ✗ | void *right[1] = { frame->extended_data[start_ch + 1] }; | |
1941 | ✗ | av_audio_fifo_read(s->samples[1][i], right, nb_samples); | |
1942 | } | ||
1943 | } | ||
1944 | |||
1945 | ✗ | *got_frame_ptr = nb_samples > 0; | |
1946 | } | ||
1947 | } | ||
1948 | |||
1949 | ✗ | return ret; | |
1950 | } | ||
1951 | |||
1952 | ✗ | static av_cold int xma_decode_init(AVCodecContext *avctx) | |
1953 | { | ||
1954 | ✗ | XMADecodeCtx *s = avctx->priv_data; | |
1955 | ✗ | int i, ret, start_channels = 0; | |
1956 | |||
1957 | ✗ | avctx->block_align = 2048; | |
1958 | |||
1959 | ✗ | if (avctx->ch_layout.nb_channels <= 0 || avctx->extradata_size == 0) | |
1960 | ✗ | return AVERROR_INVALIDDATA; | |
1961 | |||
1962 | /* get stream config */ | ||
1963 | ✗ | if (avctx->codec_id == AV_CODEC_ID_XMA2 && avctx->extradata_size == 34) { /* XMA2WAVEFORMATEX */ | |
1964 | ✗ | unsigned int channel_mask = AV_RL32(avctx->extradata + 2); | |
1965 | ✗ | if (channel_mask) { | |
1966 | ✗ | av_channel_layout_uninit(&avctx->ch_layout); | |
1967 | ✗ | av_channel_layout_from_mask(&avctx->ch_layout, channel_mask); | |
1968 | } else | ||
1969 | ✗ | avctx->ch_layout.order = AV_CHANNEL_ORDER_UNSPEC; | |
1970 | ✗ | s->num_streams = AV_RL16(avctx->extradata); | |
1971 | ✗ | } else if (avctx->codec_id == AV_CODEC_ID_XMA2 && avctx->extradata_size >= 2) { /* XMA2WAVEFORMAT */ | |
1972 | ✗ | s->num_streams = avctx->extradata[1]; | |
1973 | ✗ | if (avctx->extradata_size != (32 + ((avctx->extradata[0]==3)?0:8) + 4*s->num_streams)) { | |
1974 | ✗ | av_log(avctx, AV_LOG_ERROR, "Incorrect XMA2 extradata size\n"); | |
1975 | ✗ | s->num_streams = 0; | |
1976 | ✗ | return AVERROR(EINVAL); | |
1977 | } | ||
1978 | ✗ | } else if (avctx->codec_id == AV_CODEC_ID_XMA1 && avctx->extradata_size >= 4) { /* XMAWAVEFORMAT */ | |
1979 | ✗ | s->num_streams = avctx->extradata[4]; | |
1980 | ✗ | if (avctx->extradata_size != (8 + 20*s->num_streams)) { | |
1981 | ✗ | av_log(avctx, AV_LOG_ERROR, "Incorrect XMA1 extradata size\n"); | |
1982 | ✗ | s->num_streams = 0; | |
1983 | ✗ | return AVERROR(EINVAL); | |
1984 | } | ||
1985 | } else { | ||
1986 | ✗ | av_log(avctx, AV_LOG_ERROR, "Incorrect XMA config\n"); | |
1987 | ✗ | return AVERROR(EINVAL); | |
1988 | } | ||
1989 | |||
1990 | /* encoder supports up to 64 streams / 64*2 channels (would have to alloc arrays) */ | ||
1991 | ✗ | if (avctx->ch_layout.nb_channels > XMA_MAX_CHANNELS || s->num_streams > XMA_MAX_STREAMS || | |
1992 | ✗ | s->num_streams <= 0 | |
1993 | ) { | ||
1994 | ✗ | avpriv_request_sample(avctx, "More than %d channels in %d streams", XMA_MAX_CHANNELS, s->num_streams); | |
1995 | ✗ | s->num_streams = 0; | |
1996 | ✗ | return AVERROR_PATCHWELCOME; | |
1997 | } | ||
1998 | |||
1999 | /* init all streams (several streams of 1/2ch make Nch files) */ | ||
2000 | ✗ | for (i = 0; i < s->num_streams; i++) { | |
2001 | ✗ | ret = decode_init(&s->xma[i], avctx, i); | |
2002 | ✗ | if (ret < 0) | |
2003 | ✗ | return ret; | |
2004 | ✗ | s->frames[i] = av_frame_alloc(); | |
2005 | ✗ | if (!s->frames[i]) | |
2006 | ✗ | return AVERROR(ENOMEM); | |
2007 | |||
2008 | ✗ | s->start_channel[i] = start_channels; | |
2009 | ✗ | start_channels += s->xma[i].nb_channels; | |
2010 | } | ||
2011 | ✗ | if (start_channels != avctx->ch_layout.nb_channels) | |
2012 | ✗ | return AVERROR_INVALIDDATA; | |
2013 | |||
2014 | ✗ | for (int i = 0; i < XMA_MAX_STREAMS; i++) { | |
2015 | ✗ | s->samples[0][i] = av_audio_fifo_alloc(avctx->sample_fmt, 1, 64 * 512); | |
2016 | ✗ | s->samples[1][i] = av_audio_fifo_alloc(avctx->sample_fmt, 1, 64 * 512); | |
2017 | ✗ | if (!s->samples[0][i] || !s->samples[1][i]) | |
2018 | ✗ | return AVERROR(ENOMEM); | |
2019 | } | ||
2020 | |||
2021 | ✗ | return 0; | |
2022 | } | ||
2023 | |||
2024 | ✗ | static av_cold int xma_decode_end(AVCodecContext *avctx) | |
2025 | { | ||
2026 | ✗ | XMADecodeCtx *s = avctx->priv_data; | |
2027 | int i; | ||
2028 | |||
2029 | ✗ | for (i = 0; i < s->num_streams; i++) { | |
2030 | ✗ | decode_end(&s->xma[i]); | |
2031 | ✗ | av_frame_free(&s->frames[i]); | |
2032 | } | ||
2033 | ✗ | s->num_streams = 0; | |
2034 | |||
2035 | ✗ | for (i = 0; i < XMA_MAX_STREAMS; i++) { | |
2036 | ✗ | av_audio_fifo_free(s->samples[0][i]); | |
2037 | ✗ | av_audio_fifo_free(s->samples[1][i]); | |
2038 | } | ||
2039 | |||
2040 | ✗ | return 0; | |
2041 | } | ||
2042 | |||
2043 | ✗ | static void flush(WMAProDecodeCtx *s) | |
2044 | { | ||
2045 | int i; | ||
2046 | /** reset output buffer as a part of it is used during the windowing of a | ||
2047 | new frame */ | ||
2048 | ✗ | for (i = 0; i < s->nb_channels; i++) | |
2049 | ✗ | memset(s->channel[i].out, 0, s->samples_per_frame * | |
2050 | sizeof(*s->channel[i].out)); | ||
2051 | ✗ | s->packet_loss = 1; | |
2052 | ✗ | s->skip_packets = 0; | |
2053 | ✗ | s->eof_done = 0; | |
2054 | ✗ | s->skip_frame = 1; | |
2055 | ✗ | } | |
2056 | |||
2057 | /** | ||
2058 | *@brief Clear decoder buffers (for seeking). | ||
2059 | *@param avctx codec context | ||
2060 | */ | ||
2061 | ✗ | static void wmapro_flush(AVCodecContext *avctx) | |
2062 | { | ||
2063 | ✗ | WMAProDecodeCtx *s = avctx->priv_data; | |
2064 | |||
2065 | ✗ | flush(s); | |
2066 | ✗ | } | |
2067 | |||
2068 | ✗ | static void xma_flush(AVCodecContext *avctx) | |
2069 | { | ||
2070 | ✗ | XMADecodeCtx *s = avctx->priv_data; | |
2071 | int i; | ||
2072 | |||
2073 | ✗ | for (i = 0; i < XMA_MAX_STREAMS; i++) { | |
2074 | ✗ | av_audio_fifo_reset(s->samples[0][i]); | |
2075 | ✗ | av_audio_fifo_reset(s->samples[1][i]); | |
2076 | } | ||
2077 | |||
2078 | ✗ | for (i = 0; i < s->num_streams; i++) | |
2079 | ✗ | flush(&s->xma[i]); | |
2080 | |||
2081 | ✗ | s->current_stream = 0; | |
2082 | ✗ | s->flushed = 0; | |
2083 | ✗ | } | |
2084 | |||
2085 | /** | ||
2086 | *@brief wmapro decoder | ||
2087 | */ | ||
2088 | const FFCodec ff_wmapro_decoder = { | ||
2089 | .p.name = "wmapro", | ||
2090 | CODEC_LONG_NAME("Windows Media Audio 9 Professional"), | ||
2091 | .p.type = AVMEDIA_TYPE_AUDIO, | ||
2092 | .p.id = AV_CODEC_ID_WMAPRO, | ||
2093 | .priv_data_size = sizeof(WMAProDecodeCtx), | ||
2094 | .init = wmapro_decode_init, | ||
2095 | .close = wmapro_decode_end, | ||
2096 | FF_CODEC_DECODE_CB(wmapro_decode_packet), | ||
2097 | .p.capabilities = AV_CODEC_CAP_DR1, | ||
2098 | .flush = wmapro_flush, | ||
2099 | CODEC_SAMPLEFMTS(AV_SAMPLE_FMT_FLTP), | ||
2100 | .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, | ||
2101 | }; | ||
2102 | |||
2103 | const FFCodec ff_xma1_decoder = { | ||
2104 | .p.name = "xma1", | ||
2105 | CODEC_LONG_NAME("Xbox Media Audio 1"), | ||
2106 | .p.type = AVMEDIA_TYPE_AUDIO, | ||
2107 | .p.id = AV_CODEC_ID_XMA1, | ||
2108 | .priv_data_size = sizeof(XMADecodeCtx), | ||
2109 | .init = xma_decode_init, | ||
2110 | .close = xma_decode_end, | ||
2111 | FF_CODEC_DECODE_CB(xma_decode_packet), | ||
2112 | .flush = xma_flush, | ||
2113 | .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY, | ||
2114 | CODEC_SAMPLEFMTS(AV_SAMPLE_FMT_FLTP), | ||
2115 | .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, | ||
2116 | }; | ||
2117 | |||
2118 | const FFCodec ff_xma2_decoder = { | ||
2119 | .p.name = "xma2", | ||
2120 | CODEC_LONG_NAME("Xbox Media Audio 2"), | ||
2121 | .p.type = AVMEDIA_TYPE_AUDIO, | ||
2122 | .p.id = AV_CODEC_ID_XMA2, | ||
2123 | .priv_data_size = sizeof(XMADecodeCtx), | ||
2124 | .init = xma_decode_init, | ||
2125 | .close = xma_decode_end, | ||
2126 | FF_CODEC_DECODE_CB(xma_decode_packet), | ||
2127 | .flush = xma_flush, | ||
2128 | .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY, | ||
2129 | CODEC_SAMPLEFMTS(AV_SAMPLE_FMT_FLTP), | ||
2130 | .caps_internal = FF_CODEC_CAP_INIT_CLEANUP, | ||
2131 | }; | ||
2132 |