ffmpeg / libavcodec / ac3enc.c @ b6675279
History | View | Annotate | Download (77.3 KB)
1 |
/*
|
---|---|
2 |
* The simplest AC-3 encoder
|
3 |
* Copyright (c) 2000 Fabrice Bellard
|
4 |
* Copyright (c) 2006-2010 Justin Ruggles <justin.ruggles@gmail.com>
|
5 |
* Copyright (c) 2006-2010 Prakash Punnoor <prakash@punnoor.de>
|
6 |
*
|
7 |
* This file is part of Libav.
|
8 |
*
|
9 |
* Libav is free software; you can redistribute it and/or
|
10 |
* modify it under the terms of the GNU Lesser General Public
|
11 |
* License as published by the Free Software Foundation; either
|
12 |
* version 2.1 of the License, or (at your option) any later version.
|
13 |
*
|
14 |
* Libav is distributed in the hope that it will be useful,
|
15 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
16 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
17 |
* Lesser General Public License for more details.
|
18 |
*
|
19 |
* You should have received a copy of the GNU Lesser General Public
|
20 |
* License along with Libav; if not, write to the Free Software
|
21 |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
22 |
*/
|
23 |
|
24 |
/**
|
25 |
* @file
|
26 |
* The simplest AC-3 encoder.
|
27 |
*/
|
28 |
|
29 |
//#define DEBUG
|
30 |
//#define ASSERT_LEVEL 2
|
31 |
|
32 |
#include <stdint.h> |
33 |
|
34 |
#include "libavutil/audioconvert.h" |
35 |
#include "libavutil/avassert.h" |
36 |
#include "libavutil/avstring.h" |
37 |
#include "libavutil/crc.h" |
38 |
#include "libavutil/opt.h" |
39 |
#include "avcodec.h" |
40 |
#include "put_bits.h" |
41 |
#include "dsputil.h" |
42 |
#include "ac3dsp.h" |
43 |
#include "ac3.h" |
44 |
#include "audioconvert.h" |
45 |
#include "fft.h" |
46 |
|
47 |
|
48 |
#ifndef CONFIG_AC3ENC_FLOAT
|
49 |
#define CONFIG_AC3ENC_FLOAT 0 |
50 |
#endif
|
51 |
|
52 |
|
53 |
/** Maximum number of exponent groups. +1 for separate DC exponent. */
|
54 |
#define AC3_MAX_EXP_GROUPS 85 |
55 |
|
56 |
#if CONFIG_AC3ENC_FLOAT
|
57 |
#define MAC_COEF(d,a,b) ((d)+=(a)*(b))
|
58 |
typedef float SampleType; |
59 |
typedef float CoefType; |
60 |
typedef float CoefSumType; |
61 |
#else
|
62 |
#define MAC_COEF(d,a,b) MAC64(d,a,b)
|
63 |
typedef int16_t SampleType;
|
64 |
typedef int32_t CoefType;
|
65 |
typedef int64_t CoefSumType;
|
66 |
#endif
|
67 |
|
68 |
typedef struct AC3MDCTContext { |
69 |
const SampleType *window; ///< MDCT window function |
70 |
FFTContext fft; ///< FFT context for MDCT calculation
|
71 |
} AC3MDCTContext; |
72 |
|
73 |
/**
|
74 |
* Encoding Options used by AVOption.
|
75 |
*/
|
76 |
typedef struct AC3EncOptions { |
77 |
/* AC-3 metadata options*/
|
78 |
int dialogue_level;
|
79 |
int bitstream_mode;
|
80 |
float center_mix_level;
|
81 |
float surround_mix_level;
|
82 |
int dolby_surround_mode;
|
83 |
int audio_production_info;
|
84 |
int mixing_level;
|
85 |
int room_type;
|
86 |
int copyright;
|
87 |
int original;
|
88 |
int extended_bsi_1;
|
89 |
int preferred_stereo_downmix;
|
90 |
float ltrt_center_mix_level;
|
91 |
float ltrt_surround_mix_level;
|
92 |
float loro_center_mix_level;
|
93 |
float loro_surround_mix_level;
|
94 |
int extended_bsi_2;
|
95 |
int dolby_surround_ex_mode;
|
96 |
int dolby_headphone_mode;
|
97 |
int ad_converter_type;
|
98 |
|
99 |
/* other encoding options */
|
100 |
int allow_per_frame_metadata;
|
101 |
int stereo_rematrixing;
|
102 |
} AC3EncOptions; |
103 |
|
104 |
/**
|
105 |
* Data for a single audio block.
|
106 |
*/
|
107 |
typedef struct AC3Block { |
108 |
uint8_t **bap; ///< bit allocation pointers (bap)
|
109 |
CoefType **mdct_coef; ///< MDCT coefficients
|
110 |
int32_t **fixed_coef; ///< fixed-point MDCT coefficients
|
111 |
uint8_t **exp; ///< original exponents
|
112 |
uint8_t **grouped_exp; ///< grouped exponents
|
113 |
int16_t **psd; ///< psd per frequency bin
|
114 |
int16_t **band_psd; ///< psd per critical band
|
115 |
int16_t **mask; ///< masking curve
|
116 |
uint16_t **qmant; ///< quantized mantissas
|
117 |
uint8_t coeff_shift[AC3_MAX_CHANNELS]; ///< fixed-point coefficient shift values
|
118 |
uint8_t new_rematrixing_strategy; ///< send new rematrixing flags in this block
|
119 |
uint8_t rematrixing_flags[4]; ///< rematrixing flags |
120 |
struct AC3Block *exp_ref_block[AC3_MAX_CHANNELS]; ///< reference blocks for EXP_REUSE |
121 |
} AC3Block; |
122 |
|
123 |
/**
|
124 |
* AC-3 encoder private context.
|
125 |
*/
|
126 |
typedef struct AC3EncodeContext { |
127 |
AVClass *av_class; ///< AVClass used for AVOption
|
128 |
AC3EncOptions options; ///< encoding options
|
129 |
PutBitContext pb; ///< bitstream writer context
|
130 |
DSPContext dsp; |
131 |
AC3DSPContext ac3dsp; ///< AC-3 optimized functions
|
132 |
AC3MDCTContext mdct; ///< MDCT context
|
133 |
|
134 |
AC3Block blocks[AC3_MAX_BLOCKS]; ///< per-block info
|
135 |
|
136 |
int bitstream_id; ///< bitstream id (bsid) |
137 |
int bitstream_mode; ///< bitstream mode (bsmod) |
138 |
|
139 |
int bit_rate; ///< target bit rate, in bits-per-second |
140 |
int sample_rate; ///< sampling frequency, in Hz |
141 |
|
142 |
int frame_size_min; ///< minimum frame size in case rounding is necessary |
143 |
int frame_size; ///< current frame size in bytes |
144 |
int frame_size_code; ///< frame size code (frmsizecod) |
145 |
uint16_t crc_inv[2];
|
146 |
int bits_written; ///< bit count (used to avg. bitrate) |
147 |
int samples_written; ///< sample count (used to avg. bitrate) |
148 |
|
149 |
int fbw_channels; ///< number of full-bandwidth channels (nfchans) |
150 |
int channels; ///< total number of channels (nchans) |
151 |
int lfe_on; ///< indicates if there is an LFE channel (lfeon) |
152 |
int lfe_channel; ///< channel index of the LFE channel |
153 |
int has_center; ///< indicates if there is a center channel |
154 |
int has_surround; ///< indicates if there are one or more surround channels |
155 |
int channel_mode; ///< channel mode (acmod) |
156 |
const uint8_t *channel_map; ///< channel map used to reorder channels |
157 |
|
158 |
int center_mix_level; ///< center mix level code |
159 |
int surround_mix_level; ///< surround mix level code |
160 |
int ltrt_center_mix_level; ///< Lt/Rt center mix level code |
161 |
int ltrt_surround_mix_level; ///< Lt/Rt surround mix level code |
162 |
int loro_center_mix_level; ///< Lo/Ro center mix level code |
163 |
int loro_surround_mix_level; ///< Lo/Ro surround mix level code |
164 |
|
165 |
int cutoff; ///< user-specified cutoff frequency, in Hz |
166 |
int bandwidth_code; ///< bandwidth code (0 to 60) (chbwcod) |
167 |
int nb_coefs[AC3_MAX_CHANNELS];
|
168 |
|
169 |
int rematrixing_enabled; ///< stereo rematrixing enabled |
170 |
int num_rematrixing_bands; ///< number of rematrixing bands |
171 |
|
172 |
/* bitrate allocation control */
|
173 |
int slow_gain_code; ///< slow gain code (sgaincod) |
174 |
int slow_decay_code; ///< slow decay code (sdcycod) |
175 |
int fast_decay_code; ///< fast decay code (fdcycod) |
176 |
int db_per_bit_code; ///< dB/bit code (dbpbcod) |
177 |
int floor_code; ///< floor code (floorcod) |
178 |
AC3BitAllocParameters bit_alloc; ///< bit allocation parameters
|
179 |
int coarse_snr_offset; ///< coarse SNR offsets (csnroffst) |
180 |
int fast_gain_code[AC3_MAX_CHANNELS]; ///< fast gain codes (signal-to-mask ratio) (fgaincod) |
181 |
int fine_snr_offset[AC3_MAX_CHANNELS]; ///< fine SNR offsets (fsnroffst) |
182 |
int frame_bits_fixed; ///< number of non-coefficient bits for fixed parameters |
183 |
int frame_bits; ///< all frame bits except exponents and mantissas |
184 |
int exponent_bits; ///< number of bits used for exponents |
185 |
|
186 |
SampleType **planar_samples; |
187 |
uint8_t *bap_buffer; |
188 |
uint8_t *bap1_buffer; |
189 |
CoefType *mdct_coef_buffer; |
190 |
int32_t *fixed_coef_buffer; |
191 |
uint8_t *exp_buffer; |
192 |
uint8_t *grouped_exp_buffer; |
193 |
int16_t *psd_buffer; |
194 |
int16_t *band_psd_buffer; |
195 |
int16_t *mask_buffer; |
196 |
uint16_t *qmant_buffer; |
197 |
|
198 |
uint8_t exp_strategy[AC3_MAX_CHANNELS][AC3_MAX_BLOCKS]; ///< exponent strategies
|
199 |
|
200 |
DECLARE_ALIGNED(32, SampleType, windowed_samples)[AC3_WINDOW_SIZE];
|
201 |
} AC3EncodeContext; |
202 |
|
203 |
typedef struct AC3Mant { |
204 |
uint16_t *qmant1_ptr, *qmant2_ptr, *qmant4_ptr; ///< mantissa pointers for bap=1,2,4
|
205 |
int mant1_cnt, mant2_cnt, mant4_cnt; ///< mantissa counts for bap=1,2,4 |
206 |
} AC3Mant; |
207 |
|
208 |
#define CMIXLEV_NUM_OPTIONS 3 |
209 |
static const float cmixlev_options[CMIXLEV_NUM_OPTIONS] = { |
210 |
LEVEL_MINUS_3DB, LEVEL_MINUS_4POINT5DB, LEVEL_MINUS_6DB |
211 |
}; |
212 |
|
213 |
#define SURMIXLEV_NUM_OPTIONS 3 |
214 |
static const float surmixlev_options[SURMIXLEV_NUM_OPTIONS] = { |
215 |
LEVEL_MINUS_3DB, LEVEL_MINUS_6DB, LEVEL_ZERO |
216 |
}; |
217 |
|
218 |
#define EXTMIXLEV_NUM_OPTIONS 8 |
219 |
static const float extmixlev_options[EXTMIXLEV_NUM_OPTIONS] = { |
220 |
LEVEL_PLUS_3DB, LEVEL_PLUS_1POINT5DB, LEVEL_ONE, LEVEL_MINUS_4POINT5DB, |
221 |
LEVEL_MINUS_3DB, LEVEL_MINUS_4POINT5DB, LEVEL_MINUS_6DB, LEVEL_ZERO |
222 |
}; |
223 |
|
224 |
|
225 |
#define OFFSET(param) offsetof(AC3EncodeContext, options.param)
|
226 |
#define AC3ENC_PARAM (AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM)
|
227 |
|
228 |
static const AVOption options[] = { |
229 |
/* Metadata Options */
|
230 |
{"per_frame_metadata", "Allow Changing Metadata Per-Frame", OFFSET(allow_per_frame_metadata), FF_OPT_TYPE_INT, {.dbl = 0 }, 0, 1, AC3ENC_PARAM}, |
231 |
/* downmix levels */
|
232 |
{"center_mixlev", "Center Mix Level", OFFSET(center_mix_level), FF_OPT_TYPE_FLOAT, {.dbl = LEVEL_MINUS_4POINT5DB }, 0.0, 1.0, AC3ENC_PARAM}, |
233 |
{"surround_mixlev", "Surround Mix Level", OFFSET(surround_mix_level), FF_OPT_TYPE_FLOAT, {.dbl = LEVEL_MINUS_6DB }, 0.0, 1.0, AC3ENC_PARAM}, |
234 |
/* audio production information */
|
235 |
{"mixing_level", "Mixing Level", OFFSET(mixing_level), FF_OPT_TYPE_INT, {.dbl = -1 }, -1, 111, AC3ENC_PARAM}, |
236 |
{"room_type", "Room Type", OFFSET(room_type), FF_OPT_TYPE_INT, {.dbl = -1 }, -1, 2, AC3ENC_PARAM, "room_type"}, |
237 |
{"notindicated", "Not Indicated (default)", 0, FF_OPT_TYPE_CONST, {.dbl = 0 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "room_type"}, |
238 |
{"large", "Large Room", 0, FF_OPT_TYPE_CONST, {.dbl = 1 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "room_type"}, |
239 |
{"small", "Small Room", 0, FF_OPT_TYPE_CONST, {.dbl = 2 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "room_type"}, |
240 |
/* other metadata options */
|
241 |
{"copyright", "Copyright Bit", OFFSET(copyright), FF_OPT_TYPE_INT, {.dbl = 0 }, 0, 1, AC3ENC_PARAM}, |
242 |
{"dialnorm", "Dialogue Level (dB)", OFFSET(dialogue_level), FF_OPT_TYPE_INT, {.dbl = -31 }, -31, -1, AC3ENC_PARAM}, |
243 |
{"dsur_mode", "Dolby Surround Mode", OFFSET(dolby_surround_mode), FF_OPT_TYPE_INT, {.dbl = 0 }, 0, 2, AC3ENC_PARAM, "dsur_mode"}, |
244 |
{"notindicated", "Not Indicated (default)", 0, FF_OPT_TYPE_CONST, {.dbl = 0 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsur_mode"}, |
245 |
{"on", "Dolby Surround Encoded", 0, FF_OPT_TYPE_CONST, {.dbl = 1 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsur_mode"}, |
246 |
{"off", "Not Dolby Surround Encoded", 0, FF_OPT_TYPE_CONST, {.dbl = 2 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsur_mode"}, |
247 |
{"original", "Original Bit Stream", OFFSET(original), FF_OPT_TYPE_INT, {.dbl = 1 }, 0, 1, AC3ENC_PARAM}, |
248 |
/* extended bitstream information */
|
249 |
{"dmix_mode", "Preferred Stereo Downmix Mode", OFFSET(preferred_stereo_downmix), FF_OPT_TYPE_INT, {.dbl = -1 }, -1, 2, AC3ENC_PARAM, "dmix_mode"}, |
250 |
{"notindicated", "Not Indicated (default)", 0, FF_OPT_TYPE_CONST, {.dbl = 0 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dmix_mode"}, |
251 |
{"ltrt", "Lt/Rt Downmix Preferred", 0, FF_OPT_TYPE_CONST, {.dbl = 1 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dmix_mode"}, |
252 |
{"loro", "Lo/Ro Downmix Preferred", 0, FF_OPT_TYPE_CONST, {.dbl = 2 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dmix_mode"}, |
253 |
{"ltrt_cmixlev", "Lt/Rt Center Mix Level", OFFSET(ltrt_center_mix_level), FF_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, AC3ENC_PARAM}, |
254 |
{"ltrt_surmixlev", "Lt/Rt Surround Mix Level", OFFSET(ltrt_surround_mix_level), FF_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, AC3ENC_PARAM}, |
255 |
{"loro_cmixlev", "Lo/Ro Center Mix Level", OFFSET(loro_center_mix_level), FF_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, AC3ENC_PARAM}, |
256 |
{"loro_surmixlev", "Lo/Ro Surround Mix Level", OFFSET(loro_surround_mix_level), FF_OPT_TYPE_FLOAT, {.dbl = -1.0 }, -1.0, 2.0, AC3ENC_PARAM}, |
257 |
{"dsurex_mode", "Dolby Surround EX Mode", OFFSET(dolby_surround_ex_mode), FF_OPT_TYPE_INT, {.dbl = -1 }, -1, 2, AC3ENC_PARAM, "dsurex_mode"}, |
258 |
{"notindicated", "Not Indicated (default)", 0, FF_OPT_TYPE_CONST, {.dbl = 0 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsurex_mode"}, |
259 |
{"on", "Dolby Surround EX Encoded", 0, FF_OPT_TYPE_CONST, {.dbl = 1 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsurex_mode"}, |
260 |
{"off", "Not Dolby Surround EX Encoded", 0, FF_OPT_TYPE_CONST, {.dbl = 2 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dsurex_mode"}, |
261 |
{"dheadphone_mode", "Dolby Headphone Mode", OFFSET(dolby_headphone_mode), FF_OPT_TYPE_INT, {.dbl = -1 }, -1, 2, AC3ENC_PARAM, "dheadphone_mode"}, |
262 |
{"notindicated", "Not Indicated (default)", 0, FF_OPT_TYPE_CONST, {.dbl = 0 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dheadphone_mode"}, |
263 |
{"on", "Dolby Headphone Encoded", 0, FF_OPT_TYPE_CONST, {.dbl = 1 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dheadphone_mode"}, |
264 |
{"off", "Not Dolby Headphone Encoded", 0, FF_OPT_TYPE_CONST, {.dbl = 2 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "dheadphone_mode"}, |
265 |
{"ad_conv_type", "A/D Converter Type", OFFSET(ad_converter_type), FF_OPT_TYPE_INT, {.dbl = -1 }, -1, 1, AC3ENC_PARAM, "ad_conv_type"}, |
266 |
{"standard", "Standard (default)", 0, FF_OPT_TYPE_CONST, {.dbl = 0 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "ad_conv_type"}, |
267 |
{"hdcd", "HDCD", 0, FF_OPT_TYPE_CONST, {.dbl = 1 }, INT_MIN, INT_MAX, AC3ENC_PARAM, "ad_conv_type"}, |
268 |
/* Other Encoding Options */
|
269 |
{"stereo_rematrixing", "Stereo Rematrixing", OFFSET(stereo_rematrixing), FF_OPT_TYPE_INT, {.dbl = 1 }, 0, 1, AC3ENC_PARAM}, |
270 |
{NULL}
|
271 |
}; |
272 |
|
273 |
#if CONFIG_AC3ENC_FLOAT
|
274 |
static AVClass ac3enc_class = { "AC-3 Encoder", av_default_item_name, |
275 |
options, LIBAVUTIL_VERSION_INT }; |
276 |
#else
|
277 |
static AVClass ac3enc_class = { "Fixed-Point AC-3 Encoder", av_default_item_name, |
278 |
options, LIBAVUTIL_VERSION_INT }; |
279 |
#endif
|
280 |
|
281 |
|
282 |
/* prototypes for functions in ac3enc_fixed.c and ac3enc_float.c */
|
283 |
|
284 |
static av_cold void mdct_end(AC3MDCTContext *mdct); |
285 |
|
286 |
static av_cold int mdct_init(AVCodecContext *avctx, AC3MDCTContext *mdct, |
287 |
int nbits);
|
288 |
|
289 |
static void apply_window(DSPContext *dsp, SampleType *output, const SampleType *input, |
290 |
const SampleType *window, unsigned int len); |
291 |
|
292 |
static int normalize_samples(AC3EncodeContext *s); |
293 |
|
294 |
static void scale_coefficients(AC3EncodeContext *s); |
295 |
|
296 |
|
297 |
/**
|
298 |
* LUT for number of exponent groups.
|
299 |
* exponent_group_tab[exponent strategy-1][number of coefficients]
|
300 |
*/
|
301 |
static uint8_t exponent_group_tab[3][256]; |
302 |
|
303 |
|
304 |
/**
|
305 |
* List of supported channel layouts.
|
306 |
*/
|
307 |
static const int64_t ac3_channel_layouts[] = { |
308 |
AV_CH_LAYOUT_MONO, |
309 |
AV_CH_LAYOUT_STEREO, |
310 |
AV_CH_LAYOUT_2_1, |
311 |
AV_CH_LAYOUT_SURROUND, |
312 |
AV_CH_LAYOUT_2_2, |
313 |
AV_CH_LAYOUT_QUAD, |
314 |
AV_CH_LAYOUT_4POINT0, |
315 |
AV_CH_LAYOUT_5POINT0, |
316 |
AV_CH_LAYOUT_5POINT0_BACK, |
317 |
(AV_CH_LAYOUT_MONO | AV_CH_LOW_FREQUENCY), |
318 |
(AV_CH_LAYOUT_STEREO | AV_CH_LOW_FREQUENCY), |
319 |
(AV_CH_LAYOUT_2_1 | AV_CH_LOW_FREQUENCY), |
320 |
(AV_CH_LAYOUT_SURROUND | AV_CH_LOW_FREQUENCY), |
321 |
(AV_CH_LAYOUT_2_2 | AV_CH_LOW_FREQUENCY), |
322 |
(AV_CH_LAYOUT_QUAD | AV_CH_LOW_FREQUENCY), |
323 |
(AV_CH_LAYOUT_4POINT0 | AV_CH_LOW_FREQUENCY), |
324 |
AV_CH_LAYOUT_5POINT1, |
325 |
AV_CH_LAYOUT_5POINT1_BACK, |
326 |
0
|
327 |
}; |
328 |
|
329 |
|
330 |
/**
|
331 |
* LUT to select the bandwidth code based on the bit rate, sample rate, and
|
332 |
* number of full-bandwidth channels.
|
333 |
* bandwidth_tab[fbw_channels-1][sample rate code][bit rate code]
|
334 |
*/
|
335 |
static const uint8_t ac3_bandwidth_tab[5][3][19] = { |
336 |
// 32 40 48 56 64 80 96 112 128 160 192 224 256 320 384 448 512 576 640
|
337 |
|
338 |
{ { 0, 0, 0, 12, 16, 32, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48, 48 }, |
339 |
{ 0, 0, 0, 16, 20, 36, 56, 56, 56, 56, 56, 56, 56, 56, 56, 56, 56, 56, 56 }, |
340 |
{ 0, 0, 0, 32, 40, 60, 60, 60, 60, 60, 60, 60, 60, 60, 60, 60, 60, 60, 60 } }, |
341 |
|
342 |
{ { 0, 0, 0, 0, 0, 0, 0, 20, 24, 32, 48, 48, 48, 48, 48, 48, 48, 48, 48 }, |
343 |
{ 0, 0, 0, 0, 0, 0, 4, 24, 28, 36, 56, 56, 56, 56, 56, 56, 56, 56, 56 }, |
344 |
{ 0, 0, 0, 0, 0, 0, 20, 44, 52, 60, 60, 60, 60, 60, 60, 60, 60, 60, 60 } }, |
345 |
|
346 |
{ { 0, 0, 0, 0, 0, 0, 0, 0, 0, 16, 24, 32, 40, 48, 48, 48, 48, 48, 48 }, |
347 |
{ 0, 0, 0, 0, 0, 0, 0, 0, 4, 20, 28, 36, 44, 56, 56, 56, 56, 56, 56 }, |
348 |
{ 0, 0, 0, 0, 0, 0, 0, 0, 20, 40, 48, 60, 60, 60, 60, 60, 60, 60, 60 } }, |
349 |
|
350 |
{ { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 12, 24, 32, 48, 48, 48, 48, 48, 48 }, |
351 |
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 16, 28, 36, 56, 56, 56, 56, 56, 56 }, |
352 |
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 32, 48, 60, 60, 60, 60, 60, 60, 60 } }, |
353 |
|
354 |
{ { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 8, 20, 32, 40, 48, 48, 48, 48 }, |
355 |
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 12, 24, 36, 44, 56, 56, 56, 56 }, |
356 |
{ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 28, 44, 60, 60, 60, 60, 60, 60 } } |
357 |
}; |
358 |
|
359 |
|
360 |
/**
|
361 |
* Adjust the frame size to make the average bit rate match the target bit rate.
|
362 |
* This is only needed for 11025, 22050, and 44100 sample rates.
|
363 |
*/
|
364 |
static void adjust_frame_size(AC3EncodeContext *s) |
365 |
{ |
366 |
while (s->bits_written >= s->bit_rate && s->samples_written >= s->sample_rate) {
|
367 |
s->bits_written -= s->bit_rate; |
368 |
s->samples_written -= s->sample_rate; |
369 |
} |
370 |
s->frame_size = s->frame_size_min + |
371 |
2 * (s->bits_written * s->sample_rate < s->samples_written * s->bit_rate);
|
372 |
s->bits_written += s->frame_size * 8;
|
373 |
s->samples_written += AC3_FRAME_SIZE; |
374 |
} |
375 |
|
376 |
|
377 |
/**
|
378 |
* Deinterleave input samples.
|
379 |
* Channels are reordered from Libav's default order to AC-3 order.
|
380 |
*/
|
381 |
static void deinterleave_input_samples(AC3EncodeContext *s, |
382 |
const SampleType *samples)
|
383 |
{ |
384 |
int ch, i;
|
385 |
|
386 |
/* deinterleave and remap input samples */
|
387 |
for (ch = 0; ch < s->channels; ch++) { |
388 |
const SampleType *sptr;
|
389 |
int sinc;
|
390 |
|
391 |
/* copy last 256 samples of previous frame to the start of the current frame */
|
392 |
memcpy(&s->planar_samples[ch][0], &s->planar_samples[ch][AC3_FRAME_SIZE],
|
393 |
AC3_BLOCK_SIZE * sizeof(s->planar_samples[0][0])); |
394 |
|
395 |
/* deinterleave */
|
396 |
sinc = s->channels; |
397 |
sptr = samples + s->channel_map[ch]; |
398 |
for (i = AC3_BLOCK_SIZE; i < AC3_FRAME_SIZE+AC3_BLOCK_SIZE; i++) {
|
399 |
s->planar_samples[ch][i] = *sptr; |
400 |
sptr += sinc; |
401 |
} |
402 |
} |
403 |
} |
404 |
|
405 |
|
406 |
/**
|
407 |
* Apply the MDCT to input samples to generate frequency coefficients.
|
408 |
* This applies the KBD window and normalizes the input to reduce precision
|
409 |
* loss due to fixed-point calculations.
|
410 |
*/
|
411 |
static void apply_mdct(AC3EncodeContext *s) |
412 |
{ |
413 |
int blk, ch;
|
414 |
|
415 |
for (ch = 0; ch < s->channels; ch++) { |
416 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
417 |
AC3Block *block = &s->blocks[blk]; |
418 |
const SampleType *input_samples = &s->planar_samples[ch][blk * AC3_BLOCK_SIZE];
|
419 |
|
420 |
apply_window(&s->dsp, s->windowed_samples, input_samples, s->mdct.window, AC3_WINDOW_SIZE); |
421 |
|
422 |
block->coeff_shift[ch] = normalize_samples(s); |
423 |
|
424 |
s->mdct.fft.mdct_calcw(&s->mdct.fft, block->mdct_coef[ch], |
425 |
s->windowed_samples); |
426 |
} |
427 |
} |
428 |
} |
429 |
|
430 |
|
431 |
/**
|
432 |
* Determine rematrixing flags for each block and band.
|
433 |
*/
|
434 |
static void compute_rematrixing_strategy(AC3EncodeContext *s) |
435 |
{ |
436 |
int nb_coefs;
|
437 |
int blk, bnd, i;
|
438 |
AC3Block *block, *block0; |
439 |
|
440 |
if (s->channel_mode != AC3_CHMODE_STEREO)
|
441 |
return;
|
442 |
|
443 |
s->num_rematrixing_bands = 4;
|
444 |
|
445 |
nb_coefs = FFMIN(s->nb_coefs[0], s->nb_coefs[1]); |
446 |
|
447 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
448 |
block = &s->blocks[blk]; |
449 |
block->new_rematrixing_strategy = !blk; |
450 |
if (!s->rematrixing_enabled)
|
451 |
continue;
|
452 |
for (bnd = 0; bnd < s->num_rematrixing_bands; bnd++) { |
453 |
/* calculate calculate sum of squared coeffs for one band in one block */
|
454 |
int start = ff_ac3_rematrix_band_tab[bnd];
|
455 |
int end = FFMIN(nb_coefs, ff_ac3_rematrix_band_tab[bnd+1]); |
456 |
CoefSumType sum[4] = {0,}; |
457 |
for (i = start; i < end; i++) {
|
458 |
CoefType lt = block->mdct_coef[0][i];
|
459 |
CoefType rt = block->mdct_coef[1][i];
|
460 |
CoefType md = lt + rt; |
461 |
CoefType sd = lt - rt; |
462 |
MAC_COEF(sum[0], lt, lt);
|
463 |
MAC_COEF(sum[1], rt, rt);
|
464 |
MAC_COEF(sum[2], md, md);
|
465 |
MAC_COEF(sum[3], sd, sd);
|
466 |
} |
467 |
|
468 |
/* compare sums to determine if rematrixing will be used for this band */
|
469 |
if (FFMIN(sum[2], sum[3]) < FFMIN(sum[0], sum[1])) |
470 |
block->rematrixing_flags[bnd] = 1;
|
471 |
else
|
472 |
block->rematrixing_flags[bnd] = 0;
|
473 |
|
474 |
/* determine if new rematrixing flags will be sent */
|
475 |
if (blk &&
|
476 |
block->rematrixing_flags[bnd] != block0->rematrixing_flags[bnd]) { |
477 |
block->new_rematrixing_strategy = 1;
|
478 |
} |
479 |
} |
480 |
block0 = block; |
481 |
} |
482 |
} |
483 |
|
484 |
|
485 |
/**
|
486 |
* Apply stereo rematrixing to coefficients based on rematrixing flags.
|
487 |
*/
|
488 |
static void apply_rematrixing(AC3EncodeContext *s) |
489 |
{ |
490 |
int nb_coefs;
|
491 |
int blk, bnd, i;
|
492 |
int start, end;
|
493 |
uint8_t *flags; |
494 |
|
495 |
if (!s->rematrixing_enabled)
|
496 |
return;
|
497 |
|
498 |
nb_coefs = FFMIN(s->nb_coefs[0], s->nb_coefs[1]); |
499 |
|
500 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
501 |
AC3Block *block = &s->blocks[blk]; |
502 |
if (block->new_rematrixing_strategy)
|
503 |
flags = block->rematrixing_flags; |
504 |
for (bnd = 0; bnd < s->num_rematrixing_bands; bnd++) { |
505 |
if (flags[bnd]) {
|
506 |
start = ff_ac3_rematrix_band_tab[bnd]; |
507 |
end = FFMIN(nb_coefs, ff_ac3_rematrix_band_tab[bnd+1]);
|
508 |
for (i = start; i < end; i++) {
|
509 |
int32_t lt = block->fixed_coef[0][i];
|
510 |
int32_t rt = block->fixed_coef[1][i];
|
511 |
block->fixed_coef[0][i] = (lt + rt) >> 1; |
512 |
block->fixed_coef[1][i] = (lt - rt) >> 1; |
513 |
} |
514 |
} |
515 |
} |
516 |
} |
517 |
} |
518 |
|
519 |
|
520 |
/**
|
521 |
* Initialize exponent tables.
|
522 |
*/
|
523 |
static av_cold void exponent_init(AC3EncodeContext *s) |
524 |
{ |
525 |
int expstr, i, grpsize;
|
526 |
|
527 |
for (expstr = EXP_D15-1; expstr <= EXP_D45-1; expstr++) { |
528 |
grpsize = 3 << expstr;
|
529 |
for (i = 73; i < 256; i++) { |
530 |
exponent_group_tab[expstr][i] = (i + grpsize - 4) / grpsize;
|
531 |
} |
532 |
} |
533 |
/* LFE */
|
534 |
exponent_group_tab[0][7] = 2; |
535 |
} |
536 |
|
537 |
|
538 |
/**
|
539 |
* Extract exponents from the MDCT coefficients.
|
540 |
* This takes into account the normalization that was done to the input samples
|
541 |
* by adjusting the exponents by the exponent shift values.
|
542 |
*/
|
543 |
static void extract_exponents(AC3EncodeContext *s) |
544 |
{ |
545 |
int blk, ch;
|
546 |
|
547 |
for (ch = 0; ch < s->channels; ch++) { |
548 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
549 |
AC3Block *block = &s->blocks[blk]; |
550 |
s->ac3dsp.extract_exponents(block->exp[ch], block->fixed_coef[ch], |
551 |
AC3_MAX_COEFS); |
552 |
} |
553 |
} |
554 |
} |
555 |
|
556 |
|
557 |
/**
|
558 |
* Exponent Difference Threshold.
|
559 |
* New exponents are sent if their SAD exceed this number.
|
560 |
*/
|
561 |
#define EXP_DIFF_THRESHOLD 500 |
562 |
|
563 |
|
564 |
/**
|
565 |
* Calculate exponent strategies for all channels.
|
566 |
* Array arrangement is reversed to simplify the per-channel calculation.
|
567 |
*/
|
568 |
static void compute_exp_strategy(AC3EncodeContext *s) |
569 |
{ |
570 |
int ch, blk, blk1;
|
571 |
|
572 |
for (ch = 0; ch < s->fbw_channels; ch++) { |
573 |
uint8_t *exp_strategy = s->exp_strategy[ch]; |
574 |
uint8_t *exp = s->blocks[0].exp[ch];
|
575 |
int exp_diff;
|
576 |
|
577 |
/* estimate if the exponent variation & decide if they should be
|
578 |
reused in the next frame */
|
579 |
exp_strategy[0] = EXP_NEW;
|
580 |
exp += AC3_MAX_COEFS; |
581 |
for (blk = 1; blk < AC3_MAX_BLOCKS; blk++) { |
582 |
exp_diff = s->dsp.sad[0](NULL, exp, exp - AC3_MAX_COEFS, 16, 16); |
583 |
if (exp_diff > EXP_DIFF_THRESHOLD)
|
584 |
exp_strategy[blk] = EXP_NEW; |
585 |
else
|
586 |
exp_strategy[blk] = EXP_REUSE; |
587 |
exp += AC3_MAX_COEFS; |
588 |
} |
589 |
|
590 |
/* now select the encoding strategy type : if exponents are often
|
591 |
recoded, we use a coarse encoding */
|
592 |
blk = 0;
|
593 |
while (blk < AC3_MAX_BLOCKS) {
|
594 |
blk1 = blk + 1;
|
595 |
while (blk1 < AC3_MAX_BLOCKS && exp_strategy[blk1] == EXP_REUSE)
|
596 |
blk1++; |
597 |
switch (blk1 - blk) {
|
598 |
case 1: exp_strategy[blk] = EXP_D45; break; |
599 |
case 2: |
600 |
case 3: exp_strategy[blk] = EXP_D25; break; |
601 |
default: exp_strategy[blk] = EXP_D15; break; |
602 |
} |
603 |
blk = blk1; |
604 |
} |
605 |
} |
606 |
if (s->lfe_on) {
|
607 |
ch = s->lfe_channel; |
608 |
s->exp_strategy[ch][0] = EXP_D15;
|
609 |
for (blk = 1; blk < AC3_MAX_BLOCKS; blk++) |
610 |
s->exp_strategy[ch][blk] = EXP_REUSE; |
611 |
} |
612 |
} |
613 |
|
614 |
|
615 |
/**
|
616 |
* Update the exponents so that they are the ones the decoder will decode.
|
617 |
*/
|
618 |
static void encode_exponents_blk_ch(uint8_t *exp, int nb_exps, int exp_strategy) |
619 |
{ |
620 |
int nb_groups, i, k;
|
621 |
|
622 |
nb_groups = exponent_group_tab[exp_strategy-1][nb_exps] * 3; |
623 |
|
624 |
/* for each group, compute the minimum exponent */
|
625 |
switch(exp_strategy) {
|
626 |
case EXP_D25:
|
627 |
for (i = 1, k = 1; i <= nb_groups; i++) { |
628 |
uint8_t exp_min = exp[k]; |
629 |
if (exp[k+1] < exp_min) |
630 |
exp_min = exp[k+1];
|
631 |
exp[i] = exp_min; |
632 |
k += 2;
|
633 |
} |
634 |
break;
|
635 |
case EXP_D45:
|
636 |
for (i = 1, k = 1; i <= nb_groups; i++) { |
637 |
uint8_t exp_min = exp[k]; |
638 |
if (exp[k+1] < exp_min) |
639 |
exp_min = exp[k+1];
|
640 |
if (exp[k+2] < exp_min) |
641 |
exp_min = exp[k+2];
|
642 |
if (exp[k+3] < exp_min) |
643 |
exp_min = exp[k+3];
|
644 |
exp[i] = exp_min; |
645 |
k += 4;
|
646 |
} |
647 |
break;
|
648 |
} |
649 |
|
650 |
/* constraint for DC exponent */
|
651 |
if (exp[0] > 15) |
652 |
exp[0] = 15; |
653 |
|
654 |
/* decrease the delta between each groups to within 2 so that they can be
|
655 |
differentially encoded */
|
656 |
for (i = 1; i <= nb_groups; i++) |
657 |
exp[i] = FFMIN(exp[i], exp[i-1] + 2); |
658 |
i--; |
659 |
while (--i >= 0) |
660 |
exp[i] = FFMIN(exp[i], exp[i+1] + 2); |
661 |
|
662 |
/* now we have the exponent values the decoder will see */
|
663 |
switch (exp_strategy) {
|
664 |
case EXP_D25:
|
665 |
for (i = nb_groups, k = nb_groups * 2; i > 0; i--) { |
666 |
uint8_t exp1 = exp[i]; |
667 |
exp[k--] = exp1; |
668 |
exp[k--] = exp1; |
669 |
} |
670 |
break;
|
671 |
case EXP_D45:
|
672 |
for (i = nb_groups, k = nb_groups * 4; i > 0; i--) { |
673 |
exp[k] = exp[k-1] = exp[k-2] = exp[k-3] = exp[i]; |
674 |
k -= 4;
|
675 |
} |
676 |
break;
|
677 |
} |
678 |
} |
679 |
|
680 |
|
681 |
/**
|
682 |
* Encode exponents from original extracted form to what the decoder will see.
|
683 |
* This copies and groups exponents based on exponent strategy and reduces
|
684 |
* deltas between adjacent exponent groups so that they can be differentially
|
685 |
* encoded.
|
686 |
*/
|
687 |
static void encode_exponents(AC3EncodeContext *s) |
688 |
{ |
689 |
int blk, blk1, ch;
|
690 |
uint8_t *exp, *exp_strategy; |
691 |
int nb_coefs, num_reuse_blocks;
|
692 |
|
693 |
for (ch = 0; ch < s->channels; ch++) { |
694 |
exp = s->blocks[0].exp[ch];
|
695 |
exp_strategy = s->exp_strategy[ch]; |
696 |
nb_coefs = s->nb_coefs[ch]; |
697 |
|
698 |
blk = 0;
|
699 |
while (blk < AC3_MAX_BLOCKS) {
|
700 |
blk1 = blk + 1;
|
701 |
|
702 |
/* count the number of EXP_REUSE blocks after the current block
|
703 |
and set exponent reference block pointers */
|
704 |
s->blocks[blk].exp_ref_block[ch] = &s->blocks[blk]; |
705 |
while (blk1 < AC3_MAX_BLOCKS && exp_strategy[blk1] == EXP_REUSE) {
|
706 |
s->blocks[blk1].exp_ref_block[ch] = &s->blocks[blk]; |
707 |
blk1++; |
708 |
} |
709 |
num_reuse_blocks = blk1 - blk - 1;
|
710 |
|
711 |
/* for the EXP_REUSE case we select the min of the exponents */
|
712 |
s->ac3dsp.ac3_exponent_min(exp, num_reuse_blocks, nb_coefs); |
713 |
|
714 |
encode_exponents_blk_ch(exp, nb_coefs, exp_strategy[blk]); |
715 |
|
716 |
exp += AC3_MAX_COEFS * (num_reuse_blocks + 1);
|
717 |
blk = blk1; |
718 |
} |
719 |
} |
720 |
} |
721 |
|
722 |
|
723 |
/**
|
724 |
* Group exponents.
|
725 |
* 3 delta-encoded exponents are in each 7-bit group. The number of groups
|
726 |
* varies depending on exponent strategy and bandwidth.
|
727 |
*/
|
728 |
static void group_exponents(AC3EncodeContext *s) |
729 |
{ |
730 |
int blk, ch, i;
|
731 |
int group_size, nb_groups, bit_count;
|
732 |
uint8_t *p; |
733 |
int delta0, delta1, delta2;
|
734 |
int exp0, exp1;
|
735 |
|
736 |
bit_count = 0;
|
737 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
738 |
AC3Block *block = &s->blocks[blk]; |
739 |
for (ch = 0; ch < s->channels; ch++) { |
740 |
int exp_strategy = s->exp_strategy[ch][blk];
|
741 |
if (exp_strategy == EXP_REUSE)
|
742 |
continue;
|
743 |
group_size = exp_strategy + (exp_strategy == EXP_D45); |
744 |
nb_groups = exponent_group_tab[exp_strategy-1][s->nb_coefs[ch]];
|
745 |
bit_count += 4 + (nb_groups * 7); |
746 |
p = block->exp[ch]; |
747 |
|
748 |
/* DC exponent */
|
749 |
exp1 = *p++; |
750 |
block->grouped_exp[ch][0] = exp1;
|
751 |
|
752 |
/* remaining exponents are delta encoded */
|
753 |
for (i = 1; i <= nb_groups; i++) { |
754 |
/* merge three delta in one code */
|
755 |
exp0 = exp1; |
756 |
exp1 = p[0];
|
757 |
p += group_size; |
758 |
delta0 = exp1 - exp0 + 2;
|
759 |
av_assert2(delta0 >= 0 && delta0 <= 4); |
760 |
|
761 |
exp0 = exp1; |
762 |
exp1 = p[0];
|
763 |
p += group_size; |
764 |
delta1 = exp1 - exp0 + 2;
|
765 |
av_assert2(delta1 >= 0 && delta1 <= 4); |
766 |
|
767 |
exp0 = exp1; |
768 |
exp1 = p[0];
|
769 |
p += group_size; |
770 |
delta2 = exp1 - exp0 + 2;
|
771 |
av_assert2(delta2 >= 0 && delta2 <= 4); |
772 |
|
773 |
block->grouped_exp[ch][i] = ((delta0 * 5 + delta1) * 5) + delta2; |
774 |
} |
775 |
} |
776 |
} |
777 |
|
778 |
s->exponent_bits = bit_count; |
779 |
} |
780 |
|
781 |
|
782 |
/**
|
783 |
* Calculate final exponents from the supplied MDCT coefficients and exponent shift.
|
784 |
* Extract exponents from MDCT coefficients, calculate exponent strategies,
|
785 |
* and encode final exponents.
|
786 |
*/
|
787 |
static void process_exponents(AC3EncodeContext *s) |
788 |
{ |
789 |
extract_exponents(s); |
790 |
|
791 |
compute_exp_strategy(s); |
792 |
|
793 |
encode_exponents(s); |
794 |
|
795 |
group_exponents(s); |
796 |
|
797 |
emms_c(); |
798 |
} |
799 |
|
800 |
|
801 |
/**
|
802 |
* Count frame bits that are based solely on fixed parameters.
|
803 |
* This only has to be run once when the encoder is initialized.
|
804 |
*/
|
805 |
static void count_frame_bits_fixed(AC3EncodeContext *s) |
806 |
{ |
807 |
static const int frame_bits_inc[8] = { 0, 0, 2, 2, 2, 4, 2, 4 }; |
808 |
int blk;
|
809 |
int frame_bits;
|
810 |
|
811 |
/* assumptions:
|
812 |
* no dynamic range codes
|
813 |
* no channel coupling
|
814 |
* bit allocation parameters do not change between blocks
|
815 |
* SNR offsets do not change between blocks
|
816 |
* no delta bit allocation
|
817 |
* no skipped data
|
818 |
* no auxilliary data
|
819 |
*/
|
820 |
|
821 |
/* header size */
|
822 |
frame_bits = 65;
|
823 |
frame_bits += frame_bits_inc[s->channel_mode]; |
824 |
|
825 |
/* audio blocks */
|
826 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
827 |
frame_bits += s->fbw_channels * 2 + 2; /* blksw * c, dithflag * c, dynrnge, cplstre */ |
828 |
if (s->channel_mode == AC3_CHMODE_STEREO) {
|
829 |
frame_bits++; /* rematstr */
|
830 |
} |
831 |
frame_bits += 2 * s->fbw_channels; /* chexpstr[2] * c */ |
832 |
if (s->lfe_on)
|
833 |
frame_bits++; /* lfeexpstr */
|
834 |
frame_bits++; /* baie */
|
835 |
frame_bits++; /* snr */
|
836 |
frame_bits += 2; /* delta / skip */ |
837 |
} |
838 |
frame_bits++; /* cplinu for block 0 */
|
839 |
/* bit alloc info */
|
840 |
/* sdcycod[2], fdcycod[2], sgaincod[2], dbpbcod[2], floorcod[3] */
|
841 |
/* csnroffset[6] */
|
842 |
/* (fsnoffset[4] + fgaincod[4]) * c */
|
843 |
frame_bits += 2*4 + 3 + 6 + s->channels * (4 + 3); |
844 |
|
845 |
/* auxdatae, crcrsv */
|
846 |
frame_bits += 2;
|
847 |
|
848 |
/* CRC */
|
849 |
frame_bits += 16;
|
850 |
|
851 |
s->frame_bits_fixed = frame_bits; |
852 |
} |
853 |
|
854 |
|
855 |
/**
|
856 |
* Initialize bit allocation.
|
857 |
* Set default parameter codes and calculate parameter values.
|
858 |
*/
|
859 |
static void bit_alloc_init(AC3EncodeContext *s) |
860 |
{ |
861 |
int ch;
|
862 |
|
863 |
/* init default parameters */
|
864 |
s->slow_decay_code = 2;
|
865 |
s->fast_decay_code = 1;
|
866 |
s->slow_gain_code = 1;
|
867 |
s->db_per_bit_code = 3;
|
868 |
s->floor_code = 7;
|
869 |
for (ch = 0; ch < s->channels; ch++) |
870 |
s->fast_gain_code[ch] = 4;
|
871 |
|
872 |
/* initial snr offset */
|
873 |
s->coarse_snr_offset = 40;
|
874 |
|
875 |
/* compute real values */
|
876 |
/* currently none of these values change during encoding, so we can just
|
877 |
set them once at initialization */
|
878 |
s->bit_alloc.slow_decay = ff_ac3_slow_decay_tab[s->slow_decay_code] >> s->bit_alloc.sr_shift; |
879 |
s->bit_alloc.fast_decay = ff_ac3_fast_decay_tab[s->fast_decay_code] >> s->bit_alloc.sr_shift; |
880 |
s->bit_alloc.slow_gain = ff_ac3_slow_gain_tab[s->slow_gain_code]; |
881 |
s->bit_alloc.db_per_bit = ff_ac3_db_per_bit_tab[s->db_per_bit_code]; |
882 |
s->bit_alloc.floor = ff_ac3_floor_tab[s->floor_code]; |
883 |
|
884 |
count_frame_bits_fixed(s); |
885 |
} |
886 |
|
887 |
|
888 |
/**
|
889 |
* Count the bits used to encode the frame, minus exponents and mantissas.
|
890 |
* Bits based on fixed parameters have already been counted, so now we just
|
891 |
* have to add the bits based on parameters that change during encoding.
|
892 |
*/
|
893 |
static void count_frame_bits(AC3EncodeContext *s) |
894 |
{ |
895 |
AC3EncOptions *opt = &s->options; |
896 |
int blk, ch;
|
897 |
int frame_bits = 0; |
898 |
|
899 |
if (opt->audio_production_info)
|
900 |
frame_bits += 7;
|
901 |
if (s->bitstream_id == 6) { |
902 |
if (opt->extended_bsi_1)
|
903 |
frame_bits += 14;
|
904 |
if (opt->extended_bsi_2)
|
905 |
frame_bits += 14;
|
906 |
} |
907 |
|
908 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
909 |
/* stereo rematrixing */
|
910 |
if (s->channel_mode == AC3_CHMODE_STEREO &&
|
911 |
s->blocks[blk].new_rematrixing_strategy) { |
912 |
frame_bits += s->num_rematrixing_bands; |
913 |
} |
914 |
|
915 |
for (ch = 0; ch < s->fbw_channels; ch++) { |
916 |
if (s->exp_strategy[ch][blk] != EXP_REUSE)
|
917 |
frame_bits += 6 + 2; /* chbwcod[6], gainrng[2] */ |
918 |
} |
919 |
} |
920 |
s->frame_bits = s->frame_bits_fixed + frame_bits; |
921 |
} |
922 |
|
923 |
|
924 |
/**
|
925 |
* Finalize the mantissa bit count by adding in the grouped mantissas.
|
926 |
*/
|
927 |
static int compute_mantissa_size_final(int mant_cnt[5]) |
928 |
{ |
929 |
// bap=1 : 3 mantissas in 5 bits
|
930 |
int bits = (mant_cnt[1] / 3) * 5; |
931 |
// bap=2 : 3 mantissas in 7 bits
|
932 |
// bap=4 : 2 mantissas in 7 bits
|
933 |
bits += ((mant_cnt[2] / 3) + (mant_cnt[4] >> 1)) * 7; |
934 |
// bap=3 : each mantissa is 3 bits
|
935 |
bits += mant_cnt[3] * 3; |
936 |
return bits;
|
937 |
} |
938 |
|
939 |
|
940 |
/**
|
941 |
* Calculate masking curve based on the final exponents.
|
942 |
* Also calculate the power spectral densities to use in future calculations.
|
943 |
*/
|
944 |
static void bit_alloc_masking(AC3EncodeContext *s) |
945 |
{ |
946 |
int blk, ch;
|
947 |
|
948 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
949 |
AC3Block *block = &s->blocks[blk]; |
950 |
for (ch = 0; ch < s->channels; ch++) { |
951 |
/* We only need psd and mask for calculating bap.
|
952 |
Since we currently do not calculate bap when exponent
|
953 |
strategy is EXP_REUSE we do not need to calculate psd or mask. */
|
954 |
if (s->exp_strategy[ch][blk] != EXP_REUSE) {
|
955 |
ff_ac3_bit_alloc_calc_psd(block->exp[ch], 0,
|
956 |
s->nb_coefs[ch], |
957 |
block->psd[ch], block->band_psd[ch]); |
958 |
ff_ac3_bit_alloc_calc_mask(&s->bit_alloc, block->band_psd[ch], |
959 |
0, s->nb_coefs[ch],
|
960 |
ff_ac3_fast_gain_tab[s->fast_gain_code[ch]], |
961 |
ch == s->lfe_channel, |
962 |
DBA_NONE, 0, NULL, NULL, NULL, |
963 |
block->mask[ch]); |
964 |
} |
965 |
} |
966 |
} |
967 |
} |
968 |
|
969 |
|
970 |
/**
|
971 |
* Ensure that bap for each block and channel point to the current bap_buffer.
|
972 |
* They may have been switched during the bit allocation search.
|
973 |
*/
|
974 |
static void reset_block_bap(AC3EncodeContext *s) |
975 |
{ |
976 |
int blk, ch;
|
977 |
if (s->blocks[0].bap[0] == s->bap_buffer) |
978 |
return;
|
979 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
980 |
for (ch = 0; ch < s->channels; ch++) { |
981 |
s->blocks[blk].bap[ch] = &s->bap_buffer[AC3_MAX_COEFS * (blk * s->channels + ch)]; |
982 |
} |
983 |
} |
984 |
} |
985 |
|
986 |
|
987 |
/**
|
988 |
* Run the bit allocation with a given SNR offset.
|
989 |
* This calculates the bit allocation pointers that will be used to determine
|
990 |
* the quantization of each mantissa.
|
991 |
* @return the number of bits needed for mantissas if the given SNR offset is
|
992 |
* is used.
|
993 |
*/
|
994 |
static int bit_alloc(AC3EncodeContext *s, int snr_offset) |
995 |
{ |
996 |
int blk, ch;
|
997 |
int mantissa_bits;
|
998 |
int mant_cnt[5]; |
999 |
|
1000 |
snr_offset = (snr_offset - 240) << 2; |
1001 |
|
1002 |
reset_block_bap(s); |
1003 |
mantissa_bits = 0;
|
1004 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
1005 |
AC3Block *block = &s->blocks[blk]; |
1006 |
AC3Block *ref_block; |
1007 |
// initialize grouped mantissa counts. these are set so that they are
|
1008 |
// padded to the next whole group size when bits are counted in
|
1009 |
// compute_mantissa_size_final
|
1010 |
mant_cnt[0] = mant_cnt[3] = 0; |
1011 |
mant_cnt[1] = mant_cnt[2] = 2; |
1012 |
mant_cnt[4] = 1; |
1013 |
for (ch = 0; ch < s->channels; ch++) { |
1014 |
/* Currently the only bit allocation parameters which vary across
|
1015 |
blocks within a frame are the exponent values. We can take
|
1016 |
advantage of that by reusing the bit allocation pointers
|
1017 |
whenever we reuse exponents. */
|
1018 |
ref_block = block->exp_ref_block[ch]; |
1019 |
if (s->exp_strategy[ch][blk] != EXP_REUSE) {
|
1020 |
s->ac3dsp.bit_alloc_calc_bap(ref_block->mask[ch], |
1021 |
ref_block->psd[ch], 0,
|
1022 |
s->nb_coefs[ch], snr_offset, |
1023 |
s->bit_alloc.floor, ff_ac3_bap_tab, |
1024 |
ref_block->bap[ch]); |
1025 |
} |
1026 |
mantissa_bits += s->ac3dsp.compute_mantissa_size(mant_cnt, |
1027 |
ref_block->bap[ch], |
1028 |
s->nb_coefs[ch]); |
1029 |
} |
1030 |
mantissa_bits += compute_mantissa_size_final(mant_cnt); |
1031 |
} |
1032 |
return mantissa_bits;
|
1033 |
} |
1034 |
|
1035 |
|
1036 |
/**
|
1037 |
* Constant bitrate bit allocation search.
|
1038 |
* Find the largest SNR offset that will allow data to fit in the frame.
|
1039 |
*/
|
1040 |
static int cbr_bit_allocation(AC3EncodeContext *s) |
1041 |
{ |
1042 |
int ch;
|
1043 |
int bits_left;
|
1044 |
int snr_offset, snr_incr;
|
1045 |
|
1046 |
bits_left = 8 * s->frame_size - (s->frame_bits + s->exponent_bits);
|
1047 |
if (bits_left < 0) |
1048 |
return AVERROR(EINVAL);
|
1049 |
|
1050 |
snr_offset = s->coarse_snr_offset << 4;
|
1051 |
|
1052 |
/* if previous frame SNR offset was 1023, check if current frame can also
|
1053 |
use SNR offset of 1023. if so, skip the search. */
|
1054 |
if ((snr_offset | s->fine_snr_offset[0]) == 1023) { |
1055 |
if (bit_alloc(s, 1023) <= bits_left) |
1056 |
return 0; |
1057 |
} |
1058 |
|
1059 |
while (snr_offset >= 0 && |
1060 |
bit_alloc(s, snr_offset) > bits_left) { |
1061 |
snr_offset -= 64;
|
1062 |
} |
1063 |
if (snr_offset < 0) |
1064 |
return AVERROR(EINVAL);
|
1065 |
|
1066 |
FFSWAP(uint8_t *, s->bap_buffer, s->bap1_buffer); |
1067 |
for (snr_incr = 64; snr_incr > 0; snr_incr >>= 2) { |
1068 |
while (snr_offset + snr_incr <= 1023 && |
1069 |
bit_alloc(s, snr_offset + snr_incr) <= bits_left) { |
1070 |
snr_offset += snr_incr; |
1071 |
FFSWAP(uint8_t *, s->bap_buffer, s->bap1_buffer); |
1072 |
} |
1073 |
} |
1074 |
FFSWAP(uint8_t *, s->bap_buffer, s->bap1_buffer); |
1075 |
reset_block_bap(s); |
1076 |
|
1077 |
s->coarse_snr_offset = snr_offset >> 4;
|
1078 |
for (ch = 0; ch < s->channels; ch++) |
1079 |
s->fine_snr_offset[ch] = snr_offset & 0xF;
|
1080 |
|
1081 |
return 0; |
1082 |
} |
1083 |
|
1084 |
|
1085 |
/**
|
1086 |
* Downgrade exponent strategies to reduce the bits used by the exponents.
|
1087 |
* This is a fallback for when bit allocation fails with the normal exponent
|
1088 |
* strategies. Each time this function is run it only downgrades the
|
1089 |
* strategy in 1 channel of 1 block.
|
1090 |
* @return non-zero if downgrade was unsuccessful
|
1091 |
*/
|
1092 |
static int downgrade_exponents(AC3EncodeContext *s) |
1093 |
{ |
1094 |
int ch, blk;
|
1095 |
|
1096 |
for (ch = 0; ch < s->fbw_channels; ch++) { |
1097 |
for (blk = AC3_MAX_BLOCKS-1; blk >= 0; blk--) { |
1098 |
if (s->exp_strategy[ch][blk] == EXP_D15) {
|
1099 |
s->exp_strategy[ch][blk] = EXP_D25; |
1100 |
return 0; |
1101 |
} |
1102 |
} |
1103 |
} |
1104 |
for (ch = 0; ch < s->fbw_channels; ch++) { |
1105 |
for (blk = AC3_MAX_BLOCKS-1; blk >= 0; blk--) { |
1106 |
if (s->exp_strategy[ch][blk] == EXP_D25) {
|
1107 |
s->exp_strategy[ch][blk] = EXP_D45; |
1108 |
return 0; |
1109 |
} |
1110 |
} |
1111 |
} |
1112 |
for (ch = 0; ch < s->fbw_channels; ch++) { |
1113 |
/* block 0 cannot reuse exponents, so only downgrade D45 to REUSE if
|
1114 |
the block number > 0 */
|
1115 |
for (blk = AC3_MAX_BLOCKS-1; blk > 0; blk--) { |
1116 |
if (s->exp_strategy[ch][blk] > EXP_REUSE) {
|
1117 |
s->exp_strategy[ch][blk] = EXP_REUSE; |
1118 |
return 0; |
1119 |
} |
1120 |
} |
1121 |
} |
1122 |
return -1; |
1123 |
} |
1124 |
|
1125 |
|
1126 |
/**
|
1127 |
* Perform bit allocation search.
|
1128 |
* Finds the SNR offset value that maximizes quality and fits in the specified
|
1129 |
* frame size. Output is the SNR offset and a set of bit allocation pointers
|
1130 |
* used to quantize the mantissas.
|
1131 |
*/
|
1132 |
static int compute_bit_allocation(AC3EncodeContext *s) |
1133 |
{ |
1134 |
int ret;
|
1135 |
|
1136 |
count_frame_bits(s); |
1137 |
|
1138 |
bit_alloc_masking(s); |
1139 |
|
1140 |
ret = cbr_bit_allocation(s); |
1141 |
while (ret) {
|
1142 |
/* fallback 1: downgrade exponents */
|
1143 |
if (!downgrade_exponents(s)) {
|
1144 |
extract_exponents(s); |
1145 |
encode_exponents(s); |
1146 |
group_exponents(s); |
1147 |
ret = compute_bit_allocation(s); |
1148 |
continue;
|
1149 |
} |
1150 |
|
1151 |
/* fallbacks were not enough... */
|
1152 |
break;
|
1153 |
} |
1154 |
|
1155 |
return ret;
|
1156 |
} |
1157 |
|
1158 |
|
1159 |
/**
|
1160 |
* Symmetric quantization on 'levels' levels.
|
1161 |
*/
|
1162 |
static inline int sym_quant(int c, int e, int levels) |
1163 |
{ |
1164 |
int v = (((levels * c) >> (24 - e)) + levels) >> 1; |
1165 |
av_assert2(v >= 0 && v < levels);
|
1166 |
return v;
|
1167 |
} |
1168 |
|
1169 |
|
1170 |
/**
|
1171 |
* Asymmetric quantization on 2^qbits levels.
|
1172 |
*/
|
1173 |
static inline int asym_quant(int c, int e, int qbits) |
1174 |
{ |
1175 |
int lshift, m, v;
|
1176 |
|
1177 |
lshift = e + qbits - 24;
|
1178 |
if (lshift >= 0) |
1179 |
v = c << lshift; |
1180 |
else
|
1181 |
v = c >> (-lshift); |
1182 |
/* rounding */
|
1183 |
v = (v + 1) >> 1; |
1184 |
m = (1 << (qbits-1)); |
1185 |
if (v >= m)
|
1186 |
v = m - 1;
|
1187 |
av_assert2(v >= -m); |
1188 |
return v & ((1 << qbits)-1); |
1189 |
} |
1190 |
|
1191 |
|
1192 |
/**
|
1193 |
* Quantize a set of mantissas for a single channel in a single block.
|
1194 |
*/
|
1195 |
static void quantize_mantissas_blk_ch(AC3Mant *s, int32_t *fixed_coef, |
1196 |
uint8_t *exp, |
1197 |
uint8_t *bap, uint16_t *qmant, int n)
|
1198 |
{ |
1199 |
int i;
|
1200 |
|
1201 |
for (i = 0; i < n; i++) { |
1202 |
int v;
|
1203 |
int c = fixed_coef[i];
|
1204 |
int e = exp[i];
|
1205 |
int b = bap[i];
|
1206 |
switch (b) {
|
1207 |
case 0: |
1208 |
v = 0;
|
1209 |
break;
|
1210 |
case 1: |
1211 |
v = sym_quant(c, e, 3);
|
1212 |
switch (s->mant1_cnt) {
|
1213 |
case 0: |
1214 |
s->qmant1_ptr = &qmant[i]; |
1215 |
v = 9 * v;
|
1216 |
s->mant1_cnt = 1;
|
1217 |
break;
|
1218 |
case 1: |
1219 |
*s->qmant1_ptr += 3 * v;
|
1220 |
s->mant1_cnt = 2;
|
1221 |
v = 128;
|
1222 |
break;
|
1223 |
default:
|
1224 |
*s->qmant1_ptr += v; |
1225 |
s->mant1_cnt = 0;
|
1226 |
v = 128;
|
1227 |
break;
|
1228 |
} |
1229 |
break;
|
1230 |
case 2: |
1231 |
v = sym_quant(c, e, 5);
|
1232 |
switch (s->mant2_cnt) {
|
1233 |
case 0: |
1234 |
s->qmant2_ptr = &qmant[i]; |
1235 |
v = 25 * v;
|
1236 |
s->mant2_cnt = 1;
|
1237 |
break;
|
1238 |
case 1: |
1239 |
*s->qmant2_ptr += 5 * v;
|
1240 |
s->mant2_cnt = 2;
|
1241 |
v = 128;
|
1242 |
break;
|
1243 |
default:
|
1244 |
*s->qmant2_ptr += v; |
1245 |
s->mant2_cnt = 0;
|
1246 |
v = 128;
|
1247 |
break;
|
1248 |
} |
1249 |
break;
|
1250 |
case 3: |
1251 |
v = sym_quant(c, e, 7);
|
1252 |
break;
|
1253 |
case 4: |
1254 |
v = sym_quant(c, e, 11);
|
1255 |
switch (s->mant4_cnt) {
|
1256 |
case 0: |
1257 |
s->qmant4_ptr = &qmant[i]; |
1258 |
v = 11 * v;
|
1259 |
s->mant4_cnt = 1;
|
1260 |
break;
|
1261 |
default:
|
1262 |
*s->qmant4_ptr += v; |
1263 |
s->mant4_cnt = 0;
|
1264 |
v = 128;
|
1265 |
break;
|
1266 |
} |
1267 |
break;
|
1268 |
case 5: |
1269 |
v = sym_quant(c, e, 15);
|
1270 |
break;
|
1271 |
case 14: |
1272 |
v = asym_quant(c, e, 14);
|
1273 |
break;
|
1274 |
case 15: |
1275 |
v = asym_quant(c, e, 16);
|
1276 |
break;
|
1277 |
default:
|
1278 |
v = asym_quant(c, e, b - 1);
|
1279 |
break;
|
1280 |
} |
1281 |
qmant[i] = v; |
1282 |
} |
1283 |
} |
1284 |
|
1285 |
|
1286 |
/**
|
1287 |
* Quantize mantissas using coefficients, exponents, and bit allocation pointers.
|
1288 |
*/
|
1289 |
static void quantize_mantissas(AC3EncodeContext *s) |
1290 |
{ |
1291 |
int blk, ch;
|
1292 |
|
1293 |
|
1294 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
1295 |
AC3Block *block = &s->blocks[blk]; |
1296 |
AC3Block *ref_block; |
1297 |
AC3Mant m = { 0 };
|
1298 |
|
1299 |
for (ch = 0; ch < s->channels; ch++) { |
1300 |
ref_block = block->exp_ref_block[ch]; |
1301 |
quantize_mantissas_blk_ch(&m, block->fixed_coef[ch], |
1302 |
ref_block->exp[ch], ref_block->bap[ch], |
1303 |
block->qmant[ch], s->nb_coefs[ch]); |
1304 |
} |
1305 |
} |
1306 |
} |
1307 |
|
1308 |
|
1309 |
/**
|
1310 |
* Write the AC-3 frame header to the output bitstream.
|
1311 |
*/
|
1312 |
static void output_frame_header(AC3EncodeContext *s) |
1313 |
{ |
1314 |
AC3EncOptions *opt = &s->options; |
1315 |
|
1316 |
put_bits(&s->pb, 16, 0x0b77); /* frame header */ |
1317 |
put_bits(&s->pb, 16, 0); /* crc1: will be filled later */ |
1318 |
put_bits(&s->pb, 2, s->bit_alloc.sr_code);
|
1319 |
put_bits(&s->pb, 6, s->frame_size_code + (s->frame_size - s->frame_size_min) / 2); |
1320 |
put_bits(&s->pb, 5, s->bitstream_id);
|
1321 |
put_bits(&s->pb, 3, s->bitstream_mode);
|
1322 |
put_bits(&s->pb, 3, s->channel_mode);
|
1323 |
if ((s->channel_mode & 0x01) && s->channel_mode != AC3_CHMODE_MONO) |
1324 |
put_bits(&s->pb, 2, s->center_mix_level);
|
1325 |
if (s->channel_mode & 0x04) |
1326 |
put_bits(&s->pb, 2, s->surround_mix_level);
|
1327 |
if (s->channel_mode == AC3_CHMODE_STEREO)
|
1328 |
put_bits(&s->pb, 2, opt->dolby_surround_mode);
|
1329 |
put_bits(&s->pb, 1, s->lfe_on); /* LFE */ |
1330 |
put_bits(&s->pb, 5, -opt->dialogue_level);
|
1331 |
put_bits(&s->pb, 1, 0); /* no compression control word */ |
1332 |
put_bits(&s->pb, 1, 0); /* no lang code */ |
1333 |
put_bits(&s->pb, 1, opt->audio_production_info);
|
1334 |
if (opt->audio_production_info) {
|
1335 |
put_bits(&s->pb, 5, opt->mixing_level - 80); |
1336 |
put_bits(&s->pb, 2, opt->room_type);
|
1337 |
} |
1338 |
put_bits(&s->pb, 1, opt->copyright);
|
1339 |
put_bits(&s->pb, 1, opt->original);
|
1340 |
if (s->bitstream_id == 6) { |
1341 |
/* alternate bit stream syntax */
|
1342 |
put_bits(&s->pb, 1, opt->extended_bsi_1);
|
1343 |
if (opt->extended_bsi_1) {
|
1344 |
put_bits(&s->pb, 2, opt->preferred_stereo_downmix);
|
1345 |
put_bits(&s->pb, 3, s->ltrt_center_mix_level);
|
1346 |
put_bits(&s->pb, 3, s->ltrt_surround_mix_level);
|
1347 |
put_bits(&s->pb, 3, s->loro_center_mix_level);
|
1348 |
put_bits(&s->pb, 3, s->loro_surround_mix_level);
|
1349 |
} |
1350 |
put_bits(&s->pb, 1, opt->extended_bsi_2);
|
1351 |
if (opt->extended_bsi_2) {
|
1352 |
put_bits(&s->pb, 2, opt->dolby_surround_ex_mode);
|
1353 |
put_bits(&s->pb, 2, opt->dolby_headphone_mode);
|
1354 |
put_bits(&s->pb, 1, opt->ad_converter_type);
|
1355 |
put_bits(&s->pb, 9, 0); /* xbsi2 and encinfo : reserved */ |
1356 |
} |
1357 |
} else {
|
1358 |
put_bits(&s->pb, 1, 0); /* no time code 1 */ |
1359 |
put_bits(&s->pb, 1, 0); /* no time code 2 */ |
1360 |
} |
1361 |
put_bits(&s->pb, 1, 0); /* no additional bit stream info */ |
1362 |
} |
1363 |
|
1364 |
|
1365 |
/**
|
1366 |
* Write one audio block to the output bitstream.
|
1367 |
*/
|
1368 |
static void output_audio_block(AC3EncodeContext *s, int blk) |
1369 |
{ |
1370 |
int ch, i, baie, rbnd;
|
1371 |
AC3Block *block = &s->blocks[blk]; |
1372 |
|
1373 |
/* block switching */
|
1374 |
for (ch = 0; ch < s->fbw_channels; ch++) |
1375 |
put_bits(&s->pb, 1, 0); |
1376 |
|
1377 |
/* dither flags */
|
1378 |
for (ch = 0; ch < s->fbw_channels; ch++) |
1379 |
put_bits(&s->pb, 1, 1); |
1380 |
|
1381 |
/* dynamic range codes */
|
1382 |
put_bits(&s->pb, 1, 0); |
1383 |
|
1384 |
/* channel coupling */
|
1385 |
if (!blk) {
|
1386 |
put_bits(&s->pb, 1, 1); /* coupling strategy present */ |
1387 |
put_bits(&s->pb, 1, 0); /* no coupling strategy */ |
1388 |
} else {
|
1389 |
put_bits(&s->pb, 1, 0); /* no new coupling strategy */ |
1390 |
} |
1391 |
|
1392 |
/* stereo rematrixing */
|
1393 |
if (s->channel_mode == AC3_CHMODE_STEREO) {
|
1394 |
put_bits(&s->pb, 1, block->new_rematrixing_strategy);
|
1395 |
if (block->new_rematrixing_strategy) {
|
1396 |
/* rematrixing flags */
|
1397 |
for (rbnd = 0; rbnd < s->num_rematrixing_bands; rbnd++) |
1398 |
put_bits(&s->pb, 1, block->rematrixing_flags[rbnd]);
|
1399 |
} |
1400 |
} |
1401 |
|
1402 |
/* exponent strategy */
|
1403 |
for (ch = 0; ch < s->fbw_channels; ch++) |
1404 |
put_bits(&s->pb, 2, s->exp_strategy[ch][blk]);
|
1405 |
if (s->lfe_on)
|
1406 |
put_bits(&s->pb, 1, s->exp_strategy[s->lfe_channel][blk]);
|
1407 |
|
1408 |
/* bandwidth */
|
1409 |
for (ch = 0; ch < s->fbw_channels; ch++) { |
1410 |
if (s->exp_strategy[ch][blk] != EXP_REUSE)
|
1411 |
put_bits(&s->pb, 6, s->bandwidth_code);
|
1412 |
} |
1413 |
|
1414 |
/* exponents */
|
1415 |
for (ch = 0; ch < s->channels; ch++) { |
1416 |
int nb_groups;
|
1417 |
|
1418 |
if (s->exp_strategy[ch][blk] == EXP_REUSE)
|
1419 |
continue;
|
1420 |
|
1421 |
/* DC exponent */
|
1422 |
put_bits(&s->pb, 4, block->grouped_exp[ch][0]); |
1423 |
|
1424 |
/* exponent groups */
|
1425 |
nb_groups = exponent_group_tab[s->exp_strategy[ch][blk]-1][s->nb_coefs[ch]];
|
1426 |
for (i = 1; i <= nb_groups; i++) |
1427 |
put_bits(&s->pb, 7, block->grouped_exp[ch][i]);
|
1428 |
|
1429 |
/* gain range info */
|
1430 |
if (ch != s->lfe_channel)
|
1431 |
put_bits(&s->pb, 2, 0); |
1432 |
} |
1433 |
|
1434 |
/* bit allocation info */
|
1435 |
baie = (blk == 0);
|
1436 |
put_bits(&s->pb, 1, baie);
|
1437 |
if (baie) {
|
1438 |
put_bits(&s->pb, 2, s->slow_decay_code);
|
1439 |
put_bits(&s->pb, 2, s->fast_decay_code);
|
1440 |
put_bits(&s->pb, 2, s->slow_gain_code);
|
1441 |
put_bits(&s->pb, 2, s->db_per_bit_code);
|
1442 |
put_bits(&s->pb, 3, s->floor_code);
|
1443 |
} |
1444 |
|
1445 |
/* snr offset */
|
1446 |
put_bits(&s->pb, 1, baie);
|
1447 |
if (baie) {
|
1448 |
put_bits(&s->pb, 6, s->coarse_snr_offset);
|
1449 |
for (ch = 0; ch < s->channels; ch++) { |
1450 |
put_bits(&s->pb, 4, s->fine_snr_offset[ch]);
|
1451 |
put_bits(&s->pb, 3, s->fast_gain_code[ch]);
|
1452 |
} |
1453 |
} |
1454 |
|
1455 |
put_bits(&s->pb, 1, 0); /* no delta bit allocation */ |
1456 |
put_bits(&s->pb, 1, 0); /* no data to skip */ |
1457 |
|
1458 |
/* mantissas */
|
1459 |
for (ch = 0; ch < s->channels; ch++) { |
1460 |
int b, q;
|
1461 |
AC3Block *ref_block = block->exp_ref_block[ch]; |
1462 |
for (i = 0; i < s->nb_coefs[ch]; i++) { |
1463 |
q = block->qmant[ch][i]; |
1464 |
b = ref_block->bap[ch][i]; |
1465 |
switch (b) {
|
1466 |
case 0: break; |
1467 |
case 1: if (q != 128) put_bits(&s->pb, 5, q); break; |
1468 |
case 2: if (q != 128) put_bits(&s->pb, 7, q); break; |
1469 |
case 3: put_bits(&s->pb, 3, q); break; |
1470 |
case 4: if (q != 128) put_bits(&s->pb, 7, q); break; |
1471 |
case 14: put_bits(&s->pb, 14, q); break; |
1472 |
case 15: put_bits(&s->pb, 16, q); break; |
1473 |
default: put_bits(&s->pb, b-1, q); break; |
1474 |
} |
1475 |
} |
1476 |
} |
1477 |
} |
1478 |
|
1479 |
|
1480 |
/** CRC-16 Polynomial */
|
1481 |
#define CRC16_POLY ((1 << 0) | (1 << 2) | (1 << 15) | (1 << 16)) |
1482 |
|
1483 |
|
1484 |
static unsigned int mul_poly(unsigned int a, unsigned int b, unsigned int poly) |
1485 |
{ |
1486 |
unsigned int c; |
1487 |
|
1488 |
c = 0;
|
1489 |
while (a) {
|
1490 |
if (a & 1) |
1491 |
c ^= b; |
1492 |
a = a >> 1;
|
1493 |
b = b << 1;
|
1494 |
if (b & (1 << 16)) |
1495 |
b ^= poly; |
1496 |
} |
1497 |
return c;
|
1498 |
} |
1499 |
|
1500 |
|
1501 |
static unsigned int pow_poly(unsigned int a, unsigned int n, unsigned int poly) |
1502 |
{ |
1503 |
unsigned int r; |
1504 |
r = 1;
|
1505 |
while (n) {
|
1506 |
if (n & 1) |
1507 |
r = mul_poly(r, a, poly); |
1508 |
a = mul_poly(a, a, poly); |
1509 |
n >>= 1;
|
1510 |
} |
1511 |
return r;
|
1512 |
} |
1513 |
|
1514 |
|
1515 |
/**
|
1516 |
* Fill the end of the frame with 0's and compute the two CRCs.
|
1517 |
*/
|
1518 |
static void output_frame_end(AC3EncodeContext *s) |
1519 |
{ |
1520 |
const AVCRC *crc_ctx = av_crc_get_table(AV_CRC_16_ANSI);
|
1521 |
int frame_size_58, pad_bytes, crc1, crc2_partial, crc2, crc_inv;
|
1522 |
uint8_t *frame; |
1523 |
|
1524 |
frame_size_58 = ((s->frame_size >> 2) + (s->frame_size >> 4)) << 1; |
1525 |
|
1526 |
/* pad the remainder of the frame with zeros */
|
1527 |
av_assert2(s->frame_size * 8 - put_bits_count(&s->pb) >= 18); |
1528 |
flush_put_bits(&s->pb); |
1529 |
frame = s->pb.buf; |
1530 |
pad_bytes = s->frame_size - (put_bits_ptr(&s->pb) - frame) - 2;
|
1531 |
av_assert2(pad_bytes >= 0);
|
1532 |
if (pad_bytes > 0) |
1533 |
memset(put_bits_ptr(&s->pb), 0, pad_bytes);
|
1534 |
|
1535 |
/* compute crc1 */
|
1536 |
/* this is not so easy because it is at the beginning of the data... */
|
1537 |
crc1 = av_bswap16(av_crc(crc_ctx, 0, frame + 4, frame_size_58 - 4)); |
1538 |
crc_inv = s->crc_inv[s->frame_size > s->frame_size_min]; |
1539 |
crc1 = mul_poly(crc_inv, crc1, CRC16_POLY); |
1540 |
AV_WB16(frame + 2, crc1);
|
1541 |
|
1542 |
/* compute crc2 */
|
1543 |
crc2_partial = av_crc(crc_ctx, 0, frame + frame_size_58,
|
1544 |
s->frame_size - frame_size_58 - 3);
|
1545 |
crc2 = av_crc(crc_ctx, crc2_partial, frame + s->frame_size - 3, 1); |
1546 |
/* ensure crc2 does not match sync word by flipping crcrsv bit if needed */
|
1547 |
if (crc2 == 0x770B) { |
1548 |
frame[s->frame_size - 3] ^= 0x1; |
1549 |
crc2 = av_crc(crc_ctx, crc2_partial, frame + s->frame_size - 3, 1); |
1550 |
} |
1551 |
crc2 = av_bswap16(crc2); |
1552 |
AV_WB16(frame + s->frame_size - 2, crc2);
|
1553 |
} |
1554 |
|
1555 |
|
1556 |
/**
|
1557 |
* Write the frame to the output bitstream.
|
1558 |
*/
|
1559 |
static void output_frame(AC3EncodeContext *s, unsigned char *frame) |
1560 |
{ |
1561 |
int blk;
|
1562 |
|
1563 |
init_put_bits(&s->pb, frame, AC3_MAX_CODED_FRAME_SIZE); |
1564 |
|
1565 |
output_frame_header(s); |
1566 |
|
1567 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) |
1568 |
output_audio_block(s, blk); |
1569 |
|
1570 |
output_frame_end(s); |
1571 |
} |
1572 |
|
1573 |
|
1574 |
static void dprint_options(AVCodecContext *avctx) |
1575 |
{ |
1576 |
#ifdef DEBUG
|
1577 |
AC3EncodeContext *s = avctx->priv_data; |
1578 |
AC3EncOptions *opt = &s->options; |
1579 |
char strbuf[32]; |
1580 |
|
1581 |
switch (s->bitstream_id) {
|
1582 |
case 6: av_strlcpy(strbuf, "AC-3 (alt syntax)", 32); break; |
1583 |
case 8: av_strlcpy(strbuf, "AC-3 (standard)", 32); break; |
1584 |
case 9: av_strlcpy(strbuf, "AC-3 (dnet half-rate)", 32); break; |
1585 |
case 10: av_strlcpy(strbuf, "AC-3 (dnet quater-rate", 32); break; |
1586 |
default: snprintf(strbuf, 32, "ERROR"); |
1587 |
} |
1588 |
av_dlog(avctx, "bitstream_id: %s (%d)\n", strbuf, s->bitstream_id);
|
1589 |
av_dlog(avctx, "sample_fmt: %s\n", av_get_sample_fmt_name(avctx->sample_fmt));
|
1590 |
av_get_channel_layout_string(strbuf, 32, s->channels, avctx->channel_layout);
|
1591 |
av_dlog(avctx, "channel_layout: %s\n", strbuf);
|
1592 |
av_dlog(avctx, "sample_rate: %d\n", s->sample_rate);
|
1593 |
av_dlog(avctx, "bit_rate: %d\n", s->bit_rate);
|
1594 |
if (s->cutoff)
|
1595 |
av_dlog(avctx, "cutoff: %d\n", s->cutoff);
|
1596 |
|
1597 |
av_dlog(avctx, "per_frame_metadata: %s\n",
|
1598 |
opt->allow_per_frame_metadata?"on":"off"); |
1599 |
if (s->has_center)
|
1600 |
av_dlog(avctx, "center_mixlev: %0.3f (%d)\n", opt->center_mix_level,
|
1601 |
s->center_mix_level); |
1602 |
else
|
1603 |
av_dlog(avctx, "center_mixlev: {not written}\n");
|
1604 |
if (s->has_surround)
|
1605 |
av_dlog(avctx, "surround_mixlev: %0.3f (%d)\n", opt->surround_mix_level,
|
1606 |
s->surround_mix_level); |
1607 |
else
|
1608 |
av_dlog(avctx, "surround_mixlev: {not written}\n");
|
1609 |
if (opt->audio_production_info) {
|
1610 |
av_dlog(avctx, "mixing_level: %ddB\n", opt->mixing_level);
|
1611 |
switch (opt->room_type) {
|
1612 |
case 0: av_strlcpy(strbuf, "notindicated", 32); break; |
1613 |
case 1: av_strlcpy(strbuf, "large", 32); break; |
1614 |
case 2: av_strlcpy(strbuf, "small", 32); break; |
1615 |
default: snprintf(strbuf, 32, "ERROR (%d)", opt->room_type); |
1616 |
} |
1617 |
av_dlog(avctx, "room_type: %s\n", strbuf);
|
1618 |
} else {
|
1619 |
av_dlog(avctx, "mixing_level: {not written}\n");
|
1620 |
av_dlog(avctx, "room_type: {not written}\n");
|
1621 |
} |
1622 |
av_dlog(avctx, "copyright: %s\n", opt->copyright?"on":"off"); |
1623 |
av_dlog(avctx, "dialnorm: %ddB\n", opt->dialogue_level);
|
1624 |
if (s->channel_mode == AC3_CHMODE_STEREO) {
|
1625 |
switch (opt->dolby_surround_mode) {
|
1626 |
case 0: av_strlcpy(strbuf, "notindicated", 32); break; |
1627 |
case 1: av_strlcpy(strbuf, "on", 32); break; |
1628 |
case 2: av_strlcpy(strbuf, "off", 32); break; |
1629 |
default: snprintf(strbuf, 32, "ERROR (%d)", opt->dolby_surround_mode); |
1630 |
} |
1631 |
av_dlog(avctx, "dsur_mode: %s\n", strbuf);
|
1632 |
} else {
|
1633 |
av_dlog(avctx, "dsur_mode: {not written}\n");
|
1634 |
} |
1635 |
av_dlog(avctx, "original: %s\n", opt->original?"on":"off"); |
1636 |
|
1637 |
if (s->bitstream_id == 6) { |
1638 |
if (opt->extended_bsi_1) {
|
1639 |
switch (opt->preferred_stereo_downmix) {
|
1640 |
case 0: av_strlcpy(strbuf, "notindicated", 32); break; |
1641 |
case 1: av_strlcpy(strbuf, "ltrt", 32); break; |
1642 |
case 2: av_strlcpy(strbuf, "loro", 32); break; |
1643 |
default: snprintf(strbuf, 32, "ERROR (%d)", opt->preferred_stereo_downmix); |
1644 |
} |
1645 |
av_dlog(avctx, "dmix_mode: %s\n", strbuf);
|
1646 |
av_dlog(avctx, "ltrt_cmixlev: %0.3f (%d)\n",
|
1647 |
opt->ltrt_center_mix_level, s->ltrt_center_mix_level); |
1648 |
av_dlog(avctx, "ltrt_surmixlev: %0.3f (%d)\n",
|
1649 |
opt->ltrt_surround_mix_level, s->ltrt_surround_mix_level); |
1650 |
av_dlog(avctx, "loro_cmixlev: %0.3f (%d)\n",
|
1651 |
opt->loro_center_mix_level, s->loro_center_mix_level); |
1652 |
av_dlog(avctx, "loro_surmixlev: %0.3f (%d)\n",
|
1653 |
opt->loro_surround_mix_level, s->loro_surround_mix_level); |
1654 |
} else {
|
1655 |
av_dlog(avctx, "extended bitstream info 1: {not written}\n");
|
1656 |
} |
1657 |
if (opt->extended_bsi_2) {
|
1658 |
switch (opt->dolby_surround_ex_mode) {
|
1659 |
case 0: av_strlcpy(strbuf, "notindicated", 32); break; |
1660 |
case 1: av_strlcpy(strbuf, "on", 32); break; |
1661 |
case 2: av_strlcpy(strbuf, "off", 32); break; |
1662 |
default: snprintf(strbuf, 32, "ERROR (%d)", opt->dolby_surround_ex_mode); |
1663 |
} |
1664 |
av_dlog(avctx, "dsurex_mode: %s\n", strbuf);
|
1665 |
switch (opt->dolby_headphone_mode) {
|
1666 |
case 0: av_strlcpy(strbuf, "notindicated", 32); break; |
1667 |
case 1: av_strlcpy(strbuf, "on", 32); break; |
1668 |
case 2: av_strlcpy(strbuf, "off", 32); break; |
1669 |
default: snprintf(strbuf, 32, "ERROR (%d)", opt->dolby_headphone_mode); |
1670 |
} |
1671 |
av_dlog(avctx, "dheadphone_mode: %s\n", strbuf);
|
1672 |
|
1673 |
switch (opt->ad_converter_type) {
|
1674 |
case 0: av_strlcpy(strbuf, "standard", 32); break; |
1675 |
case 1: av_strlcpy(strbuf, "hdcd", 32); break; |
1676 |
default: snprintf(strbuf, 32, "ERROR (%d)", opt->ad_converter_type); |
1677 |
} |
1678 |
av_dlog(avctx, "ad_conv_type: %s\n", strbuf);
|
1679 |
} else {
|
1680 |
av_dlog(avctx, "extended bitstream info 2: {not written}\n");
|
1681 |
} |
1682 |
} |
1683 |
#endif
|
1684 |
} |
1685 |
|
1686 |
|
1687 |
#define FLT_OPTION_THRESHOLD 0.01 |
1688 |
|
1689 |
static int validate_float_option(float v, const float *v_list, int v_list_size) |
1690 |
{ |
1691 |
int i;
|
1692 |
|
1693 |
for (i = 0; i < v_list_size; i++) { |
1694 |
if (v < (v_list[i] + FLT_OPTION_THRESHOLD) &&
|
1695 |
v > (v_list[i] - FLT_OPTION_THRESHOLD)) |
1696 |
break;
|
1697 |
} |
1698 |
if (i == v_list_size)
|
1699 |
return -1; |
1700 |
|
1701 |
return i;
|
1702 |
} |
1703 |
|
1704 |
|
1705 |
static void validate_mix_level(void *log_ctx, const char *opt_name, |
1706 |
float *opt_param, const float *list, |
1707 |
int list_size, int default_value, int min_value, |
1708 |
int *ctx_param)
|
1709 |
{ |
1710 |
int mixlev = validate_float_option(*opt_param, list, list_size);
|
1711 |
if (mixlev < min_value) {
|
1712 |
mixlev = default_value; |
1713 |
if (*opt_param >= 0.0) { |
1714 |
av_log(log_ctx, AV_LOG_WARNING, "requested %s is not valid. using "
|
1715 |
"default value: %0.3f\n", opt_name, list[mixlev]);
|
1716 |
} |
1717 |
} |
1718 |
*opt_param = list[mixlev]; |
1719 |
*ctx_param = mixlev; |
1720 |
} |
1721 |
|
1722 |
|
1723 |
/**
|
1724 |
* Validate metadata options as set by AVOption system.
|
1725 |
* These values can optionally be changed per-frame.
|
1726 |
*/
|
1727 |
static int validate_metadata(AVCodecContext *avctx) |
1728 |
{ |
1729 |
AC3EncodeContext *s = avctx->priv_data; |
1730 |
AC3EncOptions *opt = &s->options; |
1731 |
|
1732 |
/* validate mixing levels */
|
1733 |
if (s->has_center) {
|
1734 |
validate_mix_level(avctx, "center_mix_level", &opt->center_mix_level,
|
1735 |
cmixlev_options, CMIXLEV_NUM_OPTIONS, 1, 0, |
1736 |
&s->center_mix_level); |
1737 |
} |
1738 |
if (s->has_surround) {
|
1739 |
validate_mix_level(avctx, "surround_mix_level", &opt->surround_mix_level,
|
1740 |
surmixlev_options, SURMIXLEV_NUM_OPTIONS, 1, 0, |
1741 |
&s->surround_mix_level); |
1742 |
} |
1743 |
|
1744 |
/* set audio production info flag */
|
1745 |
if (opt->mixing_level >= 0 || opt->room_type >= 0) { |
1746 |
if (opt->mixing_level < 0) { |
1747 |
av_log(avctx, AV_LOG_ERROR, "mixing_level must be set if "
|
1748 |
"room_type is set\n");
|
1749 |
return AVERROR(EINVAL);
|
1750 |
} |
1751 |
if (opt->mixing_level < 80) { |
1752 |
av_log(avctx, AV_LOG_ERROR, "invalid mixing level. must be between "
|
1753 |
"80dB and 111dB\n");
|
1754 |
return AVERROR(EINVAL);
|
1755 |
} |
1756 |
/* default room type */
|
1757 |
if (opt->room_type < 0) |
1758 |
opt->room_type = 0;
|
1759 |
opt->audio_production_info = 1;
|
1760 |
} else {
|
1761 |
opt->audio_production_info = 0;
|
1762 |
} |
1763 |
|
1764 |
/* set extended bsi 1 flag */
|
1765 |
if ((s->has_center || s->has_surround) &&
|
1766 |
(opt->preferred_stereo_downmix >= 0 ||
|
1767 |
opt->ltrt_center_mix_level >= 0 ||
|
1768 |
opt->ltrt_surround_mix_level >= 0 ||
|
1769 |
opt->loro_center_mix_level >= 0 ||
|
1770 |
opt->loro_surround_mix_level >= 0)) {
|
1771 |
/* default preferred stereo downmix */
|
1772 |
if (opt->preferred_stereo_downmix < 0) |
1773 |
opt->preferred_stereo_downmix = 0;
|
1774 |
/* validate Lt/Rt center mix level */
|
1775 |
validate_mix_level(avctx, "ltrt_center_mix_level",
|
1776 |
&opt->ltrt_center_mix_level, extmixlev_options, |
1777 |
EXTMIXLEV_NUM_OPTIONS, 5, 0, |
1778 |
&s->ltrt_center_mix_level); |
1779 |
/* validate Lt/Rt surround mix level */
|
1780 |
validate_mix_level(avctx, "ltrt_surround_mix_level",
|
1781 |
&opt->ltrt_surround_mix_level, extmixlev_options, |
1782 |
EXTMIXLEV_NUM_OPTIONS, 6, 3, |
1783 |
&s->ltrt_surround_mix_level); |
1784 |
/* validate Lo/Ro center mix level */
|
1785 |
validate_mix_level(avctx, "loro_center_mix_level",
|
1786 |
&opt->loro_center_mix_level, extmixlev_options, |
1787 |
EXTMIXLEV_NUM_OPTIONS, 5, 0, |
1788 |
&s->loro_center_mix_level); |
1789 |
/* validate Lo/Ro surround mix level */
|
1790 |
validate_mix_level(avctx, "loro_surround_mix_level",
|
1791 |
&opt->loro_surround_mix_level, extmixlev_options, |
1792 |
EXTMIXLEV_NUM_OPTIONS, 6, 3, |
1793 |
&s->loro_surround_mix_level); |
1794 |
opt->extended_bsi_1 = 1;
|
1795 |
} else {
|
1796 |
opt->extended_bsi_1 = 0;
|
1797 |
} |
1798 |
|
1799 |
/* set extended bsi 2 flag */
|
1800 |
if (opt->dolby_surround_ex_mode >= 0 || |
1801 |
opt->dolby_headphone_mode >= 0 ||
|
1802 |
opt->ad_converter_type >= 0) {
|
1803 |
/* default dolby surround ex mode */
|
1804 |
if (opt->dolby_surround_ex_mode < 0) |
1805 |
opt->dolby_surround_ex_mode = 0;
|
1806 |
/* default dolby headphone mode */
|
1807 |
if (opt->dolby_headphone_mode < 0) |
1808 |
opt->dolby_headphone_mode = 0;
|
1809 |
/* default A/D converter type */
|
1810 |
if (opt->ad_converter_type < 0) |
1811 |
opt->ad_converter_type = 0;
|
1812 |
opt->extended_bsi_2 = 1;
|
1813 |
} else {
|
1814 |
opt->extended_bsi_2 = 0;
|
1815 |
} |
1816 |
|
1817 |
/* set bitstream id for alternate bitstream syntax */
|
1818 |
if (opt->extended_bsi_1 || opt->extended_bsi_2) {
|
1819 |
if (s->bitstream_id > 8 && s->bitstream_id < 11) { |
1820 |
static int warn_once = 1; |
1821 |
if (warn_once) {
|
1822 |
av_log(avctx, AV_LOG_WARNING, "alternate bitstream syntax is "
|
1823 |
"not compatible with reduced samplerates. writing of "
|
1824 |
"extended bitstream information will be disabled.\n");
|
1825 |
warn_once = 0;
|
1826 |
} |
1827 |
} else {
|
1828 |
s->bitstream_id = 6;
|
1829 |
} |
1830 |
} |
1831 |
|
1832 |
return 0; |
1833 |
} |
1834 |
|
1835 |
|
1836 |
/**
|
1837 |
* Encode a single AC-3 frame.
|
1838 |
*/
|
1839 |
static int ac3_encode_frame(AVCodecContext *avctx, unsigned char *frame, |
1840 |
int buf_size, void *data) |
1841 |
{ |
1842 |
AC3EncodeContext *s = avctx->priv_data; |
1843 |
const SampleType *samples = data;
|
1844 |
int ret;
|
1845 |
|
1846 |
if (s->options.allow_per_frame_metadata) {
|
1847 |
ret = validate_metadata(avctx); |
1848 |
if (ret)
|
1849 |
return ret;
|
1850 |
} |
1851 |
|
1852 |
if (s->bit_alloc.sr_code == 1) |
1853 |
adjust_frame_size(s); |
1854 |
|
1855 |
deinterleave_input_samples(s, samples); |
1856 |
|
1857 |
apply_mdct(s); |
1858 |
|
1859 |
scale_coefficients(s); |
1860 |
|
1861 |
compute_rematrixing_strategy(s); |
1862 |
|
1863 |
apply_rematrixing(s); |
1864 |
|
1865 |
process_exponents(s); |
1866 |
|
1867 |
ret = compute_bit_allocation(s); |
1868 |
if (ret) {
|
1869 |
av_log(avctx, AV_LOG_ERROR, "Bit allocation failed. Try increasing the bitrate.\n");
|
1870 |
return ret;
|
1871 |
} |
1872 |
|
1873 |
quantize_mantissas(s); |
1874 |
|
1875 |
output_frame(s, frame); |
1876 |
|
1877 |
return s->frame_size;
|
1878 |
} |
1879 |
|
1880 |
|
1881 |
/**
|
1882 |
* Finalize encoding and free any memory allocated by the encoder.
|
1883 |
*/
|
1884 |
static av_cold int ac3_encode_close(AVCodecContext *avctx) |
1885 |
{ |
1886 |
int blk, ch;
|
1887 |
AC3EncodeContext *s = avctx->priv_data; |
1888 |
|
1889 |
for (ch = 0; ch < s->channels; ch++) |
1890 |
av_freep(&s->planar_samples[ch]); |
1891 |
av_freep(&s->planar_samples); |
1892 |
av_freep(&s->bap_buffer); |
1893 |
av_freep(&s->bap1_buffer); |
1894 |
av_freep(&s->mdct_coef_buffer); |
1895 |
av_freep(&s->fixed_coef_buffer); |
1896 |
av_freep(&s->exp_buffer); |
1897 |
av_freep(&s->grouped_exp_buffer); |
1898 |
av_freep(&s->psd_buffer); |
1899 |
av_freep(&s->band_psd_buffer); |
1900 |
av_freep(&s->mask_buffer); |
1901 |
av_freep(&s->qmant_buffer); |
1902 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
1903 |
AC3Block *block = &s->blocks[blk]; |
1904 |
av_freep(&block->bap); |
1905 |
av_freep(&block->mdct_coef); |
1906 |
av_freep(&block->fixed_coef); |
1907 |
av_freep(&block->exp); |
1908 |
av_freep(&block->grouped_exp); |
1909 |
av_freep(&block->psd); |
1910 |
av_freep(&block->band_psd); |
1911 |
av_freep(&block->mask); |
1912 |
av_freep(&block->qmant); |
1913 |
} |
1914 |
|
1915 |
mdct_end(&s->mdct); |
1916 |
|
1917 |
av_freep(&avctx->coded_frame); |
1918 |
return 0; |
1919 |
} |
1920 |
|
1921 |
|
1922 |
/**
|
1923 |
* Set channel information during initialization.
|
1924 |
*/
|
1925 |
static av_cold int set_channel_info(AC3EncodeContext *s, int channels, |
1926 |
int64_t *channel_layout) |
1927 |
{ |
1928 |
int ch_layout;
|
1929 |
|
1930 |
if (channels < 1 || channels > AC3_MAX_CHANNELS) |
1931 |
return AVERROR(EINVAL);
|
1932 |
if ((uint64_t)*channel_layout > 0x7FF) |
1933 |
return AVERROR(EINVAL);
|
1934 |
ch_layout = *channel_layout; |
1935 |
if (!ch_layout)
|
1936 |
ch_layout = avcodec_guess_channel_layout(channels, CODEC_ID_AC3, NULL);
|
1937 |
|
1938 |
s->lfe_on = !!(ch_layout & AV_CH_LOW_FREQUENCY); |
1939 |
s->channels = channels; |
1940 |
s->fbw_channels = channels - s->lfe_on; |
1941 |
s->lfe_channel = s->lfe_on ? s->fbw_channels : -1;
|
1942 |
if (s->lfe_on)
|
1943 |
ch_layout -= AV_CH_LOW_FREQUENCY; |
1944 |
|
1945 |
switch (ch_layout) {
|
1946 |
case AV_CH_LAYOUT_MONO: s->channel_mode = AC3_CHMODE_MONO; break; |
1947 |
case AV_CH_LAYOUT_STEREO: s->channel_mode = AC3_CHMODE_STEREO; break; |
1948 |
case AV_CH_LAYOUT_SURROUND: s->channel_mode = AC3_CHMODE_3F; break; |
1949 |
case AV_CH_LAYOUT_2_1: s->channel_mode = AC3_CHMODE_2F1R; break; |
1950 |
case AV_CH_LAYOUT_4POINT0: s->channel_mode = AC3_CHMODE_3F1R; break; |
1951 |
case AV_CH_LAYOUT_QUAD:
|
1952 |
case AV_CH_LAYOUT_2_2: s->channel_mode = AC3_CHMODE_2F2R; break; |
1953 |
case AV_CH_LAYOUT_5POINT0:
|
1954 |
case AV_CH_LAYOUT_5POINT0_BACK: s->channel_mode = AC3_CHMODE_3F2R; break; |
1955 |
default:
|
1956 |
return AVERROR(EINVAL);
|
1957 |
} |
1958 |
s->has_center = (s->channel_mode & 0x01) && s->channel_mode != AC3_CHMODE_MONO;
|
1959 |
s->has_surround = s->channel_mode & 0x04;
|
1960 |
|
1961 |
s->channel_map = ff_ac3_enc_channel_map[s->channel_mode][s->lfe_on]; |
1962 |
*channel_layout = ch_layout; |
1963 |
if (s->lfe_on)
|
1964 |
*channel_layout |= AV_CH_LOW_FREQUENCY; |
1965 |
|
1966 |
return 0; |
1967 |
} |
1968 |
|
1969 |
|
1970 |
static av_cold int validate_options(AVCodecContext *avctx, AC3EncodeContext *s) |
1971 |
{ |
1972 |
int i, ret;
|
1973 |
|
1974 |
/* validate channel layout */
|
1975 |
if (!avctx->channel_layout) {
|
1976 |
av_log(avctx, AV_LOG_WARNING, "No channel layout specified. The "
|
1977 |
"encoder will guess the layout, but it "
|
1978 |
"might be incorrect.\n");
|
1979 |
} |
1980 |
ret = set_channel_info(s, avctx->channels, &avctx->channel_layout); |
1981 |
if (ret) {
|
1982 |
av_log(avctx, AV_LOG_ERROR, "invalid channel layout\n");
|
1983 |
return ret;
|
1984 |
} |
1985 |
|
1986 |
/* validate sample rate */
|
1987 |
for (i = 0; i < 9; i++) { |
1988 |
if ((ff_ac3_sample_rate_tab[i / 3] >> (i % 3)) == avctx->sample_rate) |
1989 |
break;
|
1990 |
} |
1991 |
if (i == 9) { |
1992 |
av_log(avctx, AV_LOG_ERROR, "invalid sample rate\n");
|
1993 |
return AVERROR(EINVAL);
|
1994 |
} |
1995 |
s->sample_rate = avctx->sample_rate; |
1996 |
s->bit_alloc.sr_shift = i % 3;
|
1997 |
s->bit_alloc.sr_code = i / 3;
|
1998 |
s->bitstream_id = 8 + s->bit_alloc.sr_shift;
|
1999 |
|
2000 |
/* validate bit rate */
|
2001 |
for (i = 0; i < 19; i++) { |
2002 |
if ((ff_ac3_bitrate_tab[i] >> s->bit_alloc.sr_shift)*1000 == avctx->bit_rate) |
2003 |
break;
|
2004 |
} |
2005 |
if (i == 19) { |
2006 |
av_log(avctx, AV_LOG_ERROR, "invalid bit rate\n");
|
2007 |
return AVERROR(EINVAL);
|
2008 |
} |
2009 |
s->bit_rate = avctx->bit_rate; |
2010 |
s->frame_size_code = i << 1;
|
2011 |
|
2012 |
/* validate cutoff */
|
2013 |
if (avctx->cutoff < 0) { |
2014 |
av_log(avctx, AV_LOG_ERROR, "invalid cutoff frequency\n");
|
2015 |
return AVERROR(EINVAL);
|
2016 |
} |
2017 |
s->cutoff = avctx->cutoff; |
2018 |
if (s->cutoff > (s->sample_rate >> 1)) |
2019 |
s->cutoff = s->sample_rate >> 1;
|
2020 |
|
2021 |
/* validate audio service type / channels combination */
|
2022 |
if ((avctx->audio_service_type == AV_AUDIO_SERVICE_TYPE_KARAOKE &&
|
2023 |
avctx->channels == 1) ||
|
2024 |
((avctx->audio_service_type == AV_AUDIO_SERVICE_TYPE_COMMENTARY || |
2025 |
avctx->audio_service_type == AV_AUDIO_SERVICE_TYPE_EMERGENCY || |
2026 |
avctx->audio_service_type == AV_AUDIO_SERVICE_TYPE_VOICE_OVER) |
2027 |
&& avctx->channels > 1)) {
|
2028 |
av_log(avctx, AV_LOG_ERROR, "invalid audio service type for the "
|
2029 |
"specified number of channels\n");
|
2030 |
return AVERROR(EINVAL);
|
2031 |
} |
2032 |
|
2033 |
ret = validate_metadata(avctx); |
2034 |
if (ret)
|
2035 |
return ret;
|
2036 |
|
2037 |
s->rematrixing_enabled = s->options.stereo_rematrixing && |
2038 |
(s->channel_mode == AC3_CHMODE_STEREO); |
2039 |
|
2040 |
return 0; |
2041 |
} |
2042 |
|
2043 |
|
2044 |
/**
|
2045 |
* Set bandwidth for all channels.
|
2046 |
* The user can optionally supply a cutoff frequency. Otherwise an appropriate
|
2047 |
* default value will be used.
|
2048 |
*/
|
2049 |
static av_cold void set_bandwidth(AC3EncodeContext *s) |
2050 |
{ |
2051 |
int ch;
|
2052 |
|
2053 |
if (s->cutoff) {
|
2054 |
/* calculate bandwidth based on user-specified cutoff frequency */
|
2055 |
int fbw_coeffs;
|
2056 |
fbw_coeffs = s->cutoff * 2 * AC3_MAX_COEFS / s->sample_rate;
|
2057 |
s->bandwidth_code = av_clip((fbw_coeffs - 73) / 3, 0, 60); |
2058 |
} else {
|
2059 |
/* use default bandwidth setting */
|
2060 |
s->bandwidth_code = ac3_bandwidth_tab[s->fbw_channels-1][s->bit_alloc.sr_code][s->frame_size_code/2]; |
2061 |
} |
2062 |
|
2063 |
/* set number of coefficients for each channel */
|
2064 |
for (ch = 0; ch < s->fbw_channels; ch++) { |
2065 |
s->nb_coefs[ch] = s->bandwidth_code * 3 + 73; |
2066 |
} |
2067 |
if (s->lfe_on)
|
2068 |
s->nb_coefs[s->lfe_channel] = 7; /* LFE channel always has 7 coefs */ |
2069 |
} |
2070 |
|
2071 |
|
2072 |
static av_cold int allocate_buffers(AVCodecContext *avctx) |
2073 |
{ |
2074 |
int blk, ch;
|
2075 |
AC3EncodeContext *s = avctx->priv_data; |
2076 |
|
2077 |
FF_ALLOC_OR_GOTO(avctx, s->planar_samples, s->channels * sizeof(*s->planar_samples),
|
2078 |
alloc_fail); |
2079 |
for (ch = 0; ch < s->channels; ch++) { |
2080 |
FF_ALLOCZ_OR_GOTO(avctx, s->planar_samples[ch], |
2081 |
(AC3_FRAME_SIZE+AC3_BLOCK_SIZE) * sizeof(**s->planar_samples),
|
2082 |
alloc_fail); |
2083 |
} |
2084 |
FF_ALLOC_OR_GOTO(avctx, s->bap_buffer, AC3_MAX_BLOCKS * s->channels * |
2085 |
AC3_MAX_COEFS * sizeof(*s->bap_buffer), alloc_fail);
|
2086 |
FF_ALLOC_OR_GOTO(avctx, s->bap1_buffer, AC3_MAX_BLOCKS * s->channels * |
2087 |
AC3_MAX_COEFS * sizeof(*s->bap1_buffer), alloc_fail);
|
2088 |
FF_ALLOC_OR_GOTO(avctx, s->mdct_coef_buffer, AC3_MAX_BLOCKS * s->channels * |
2089 |
AC3_MAX_COEFS * sizeof(*s->mdct_coef_buffer), alloc_fail);
|
2090 |
FF_ALLOC_OR_GOTO(avctx, s->exp_buffer, AC3_MAX_BLOCKS * s->channels * |
2091 |
AC3_MAX_COEFS * sizeof(*s->exp_buffer), alloc_fail);
|
2092 |
FF_ALLOC_OR_GOTO(avctx, s->grouped_exp_buffer, AC3_MAX_BLOCKS * s->channels * |
2093 |
128 * sizeof(*s->grouped_exp_buffer), alloc_fail); |
2094 |
FF_ALLOC_OR_GOTO(avctx, s->psd_buffer, AC3_MAX_BLOCKS * s->channels * |
2095 |
AC3_MAX_COEFS * sizeof(*s->psd_buffer), alloc_fail);
|
2096 |
FF_ALLOC_OR_GOTO(avctx, s->band_psd_buffer, AC3_MAX_BLOCKS * s->channels * |
2097 |
64 * sizeof(*s->band_psd_buffer), alloc_fail); |
2098 |
FF_ALLOC_OR_GOTO(avctx, s->mask_buffer, AC3_MAX_BLOCKS * s->channels * |
2099 |
64 * sizeof(*s->mask_buffer), alloc_fail); |
2100 |
FF_ALLOC_OR_GOTO(avctx, s->qmant_buffer, AC3_MAX_BLOCKS * s->channels * |
2101 |
AC3_MAX_COEFS * sizeof(*s->qmant_buffer), alloc_fail);
|
2102 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
2103 |
AC3Block *block = &s->blocks[blk]; |
2104 |
FF_ALLOC_OR_GOTO(avctx, block->bap, s->channels * sizeof(*block->bap),
|
2105 |
alloc_fail); |
2106 |
FF_ALLOCZ_OR_GOTO(avctx, block->mdct_coef, s->channels * sizeof(*block->mdct_coef),
|
2107 |
alloc_fail); |
2108 |
FF_ALLOCZ_OR_GOTO(avctx, block->exp, s->channels * sizeof(*block->exp),
|
2109 |
alloc_fail); |
2110 |
FF_ALLOCZ_OR_GOTO(avctx, block->grouped_exp, s->channels * sizeof(*block->grouped_exp),
|
2111 |
alloc_fail); |
2112 |
FF_ALLOCZ_OR_GOTO(avctx, block->psd, s->channels * sizeof(*block->psd),
|
2113 |
alloc_fail); |
2114 |
FF_ALLOCZ_OR_GOTO(avctx, block->band_psd, s->channels * sizeof(*block->band_psd),
|
2115 |
alloc_fail); |
2116 |
FF_ALLOCZ_OR_GOTO(avctx, block->mask, s->channels * sizeof(*block->mask),
|
2117 |
alloc_fail); |
2118 |
FF_ALLOCZ_OR_GOTO(avctx, block->qmant, s->channels * sizeof(*block->qmant),
|
2119 |
alloc_fail); |
2120 |
|
2121 |
for (ch = 0; ch < s->channels; ch++) { |
2122 |
/* arrangement: block, channel, coeff */
|
2123 |
block->bap[ch] = &s->bap_buffer [AC3_MAX_COEFS * (blk * s->channels + ch)]; |
2124 |
block->mdct_coef[ch] = &s->mdct_coef_buffer [AC3_MAX_COEFS * (blk * s->channels + ch)]; |
2125 |
block->grouped_exp[ch] = &s->grouped_exp_buffer[128 * (blk * s->channels + ch)];
|
2126 |
block->psd[ch] = &s->psd_buffer [AC3_MAX_COEFS * (blk * s->channels + ch)]; |
2127 |
block->band_psd[ch] = &s->band_psd_buffer [64 * (blk * s->channels + ch)];
|
2128 |
block->mask[ch] = &s->mask_buffer [64 * (blk * s->channels + ch)];
|
2129 |
block->qmant[ch] = &s->qmant_buffer [AC3_MAX_COEFS * (blk * s->channels + ch)]; |
2130 |
|
2131 |
/* arrangement: channel, block, coeff */
|
2132 |
block->exp[ch] = &s->exp_buffer [AC3_MAX_COEFS * (AC3_MAX_BLOCKS * ch + blk)]; |
2133 |
} |
2134 |
} |
2135 |
|
2136 |
if (CONFIG_AC3ENC_FLOAT) {
|
2137 |
FF_ALLOC_OR_GOTO(avctx, s->fixed_coef_buffer, AC3_MAX_BLOCKS * s->channels * |
2138 |
AC3_MAX_COEFS * sizeof(*s->fixed_coef_buffer), alloc_fail);
|
2139 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
2140 |
AC3Block *block = &s->blocks[blk]; |
2141 |
FF_ALLOCZ_OR_GOTO(avctx, block->fixed_coef, s->channels * |
2142 |
sizeof(*block->fixed_coef), alloc_fail);
|
2143 |
for (ch = 0; ch < s->channels; ch++) |
2144 |
block->fixed_coef[ch] = &s->fixed_coef_buffer[AC3_MAX_COEFS * (blk * s->channels + ch)]; |
2145 |
} |
2146 |
} else {
|
2147 |
for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) { |
2148 |
AC3Block *block = &s->blocks[blk]; |
2149 |
FF_ALLOCZ_OR_GOTO(avctx, block->fixed_coef, s->channels * |
2150 |
sizeof(*block->fixed_coef), alloc_fail);
|
2151 |
for (ch = 0; ch < s->channels; ch++) |
2152 |
block->fixed_coef[ch] = (int32_t *)block->mdct_coef[ch]; |
2153 |
} |
2154 |
} |
2155 |
|
2156 |
return 0; |
2157 |
alloc_fail:
|
2158 |
return AVERROR(ENOMEM);
|
2159 |
} |
2160 |
|
2161 |
|
2162 |
/**
|
2163 |
* Initialize the encoder.
|
2164 |
*/
|
2165 |
static av_cold int ac3_encode_init(AVCodecContext *avctx) |
2166 |
{ |
2167 |
AC3EncodeContext *s = avctx->priv_data; |
2168 |
int ret, frame_size_58;
|
2169 |
|
2170 |
avctx->frame_size = AC3_FRAME_SIZE; |
2171 |
|
2172 |
ff_ac3_common_init(); |
2173 |
|
2174 |
ret = validate_options(avctx, s); |
2175 |
if (ret)
|
2176 |
return ret;
|
2177 |
|
2178 |
s->bitstream_mode = avctx->audio_service_type; |
2179 |
if (s->bitstream_mode == AV_AUDIO_SERVICE_TYPE_KARAOKE)
|
2180 |
s->bitstream_mode = 0x7;
|
2181 |
|
2182 |
s->frame_size_min = 2 * ff_ac3_frame_size_tab[s->frame_size_code][s->bit_alloc.sr_code];
|
2183 |
s->bits_written = 0;
|
2184 |
s->samples_written = 0;
|
2185 |
s->frame_size = s->frame_size_min; |
2186 |
|
2187 |
/* calculate crc_inv for both possible frame sizes */
|
2188 |
frame_size_58 = (( s->frame_size >> 2) + ( s->frame_size >> 4)) << 1; |
2189 |
s->crc_inv[0] = pow_poly((CRC16_POLY >> 1), (8 * frame_size_58) - 16, CRC16_POLY); |
2190 |
if (s->bit_alloc.sr_code == 1) { |
2191 |
frame_size_58 = (((s->frame_size+2) >> 2) + ((s->frame_size+2) >> 4)) << 1; |
2192 |
s->crc_inv[1] = pow_poly((CRC16_POLY >> 1), (8 * frame_size_58) - 16, CRC16_POLY); |
2193 |
} |
2194 |
|
2195 |
set_bandwidth(s); |
2196 |
|
2197 |
exponent_init(s); |
2198 |
|
2199 |
bit_alloc_init(s); |
2200 |
|
2201 |
ret = mdct_init(avctx, &s->mdct, 9);
|
2202 |
if (ret)
|
2203 |
goto init_fail;
|
2204 |
|
2205 |
ret = allocate_buffers(avctx); |
2206 |
if (ret)
|
2207 |
goto init_fail;
|
2208 |
|
2209 |
avctx->coded_frame= avcodec_alloc_frame(); |
2210 |
|
2211 |
dsputil_init(&s->dsp, avctx); |
2212 |
ff_ac3dsp_init(&s->ac3dsp, avctx->flags & CODEC_FLAG_BITEXACT); |
2213 |
|
2214 |
dprint_options(avctx); |
2215 |
|
2216 |
return 0; |
2217 |
init_fail:
|
2218 |
ac3_encode_close(avctx); |
2219 |
return ret;
|
2220 |
} |