Details | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
6148 | serge | 1 | /* |
2 | * The simplest mpeg audio layer 2 encoder |
||
3 | * Copyright (c) 2000, 2001 Fabrice Bellard |
||
4 | * |
||
5 | * This file is part of FFmpeg. |
||
6 | * |
||
7 | * FFmpeg is free software; you can redistribute it and/or |
||
8 | * modify it under the terms of the GNU Lesser General Public |
||
9 | * License as published by the Free Software Foundation; either |
||
10 | * version 2.1 of the License, or (at your option) any later version. |
||
11 | * |
||
12 | * FFmpeg is distributed in the hope that it will be useful, |
||
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
||
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
||
15 | * Lesser General Public License for more details. |
||
16 | * |
||
17 | * You should have received a copy of the GNU Lesser General Public |
||
18 | * License along with FFmpeg; if not, write to the Free Software |
||
19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
||
20 | */ |
||
21 | |||
22 | /** |
||
23 | * @file |
||
24 | * The simplest mpeg audio layer 2 encoder. |
||
25 | */ |
||
26 | |||
27 | #include "libavutil/channel_layout.h" |
||
28 | |||
29 | #include "avcodec.h" |
||
30 | #include "internal.h" |
||
31 | #include "put_bits.h" |
||
32 | |||
33 | #define FRAC_BITS 15 /* fractional bits for sb_samples and dct */ |
||
34 | #define WFRAC_BITS 14 /* fractional bits for window */ |
||
35 | |||
36 | #include "mpegaudio.h" |
||
37 | #include "mpegaudiodsp.h" |
||
38 | |||
39 | /* currently, cannot change these constants (need to modify |
||
40 | quantization stage) */ |
||
41 | #define MUL(a,b) (((int64_t)(a) * (int64_t)(b)) >> FRAC_BITS) |
||
42 | |||
43 | #define SAMPLES_BUF_SIZE 4096 |
||
44 | |||
45 | typedef struct MpegAudioContext { |
||
46 | PutBitContext pb; |
||
47 | int nb_channels; |
||
48 | int lsf; /* 1 if mpeg2 low bitrate selected */ |
||
49 | int bitrate_index; /* bit rate */ |
||
50 | int freq_index; |
||
51 | int frame_size; /* frame size, in bits, without padding */ |
||
52 | /* padding computation */ |
||
53 | int frame_frac, frame_frac_incr, do_padding; |
||
54 | short samples_buf[MPA_MAX_CHANNELS][SAMPLES_BUF_SIZE]; /* buffer for filter */ |
||
55 | int samples_offset[MPA_MAX_CHANNELS]; /* offset in samples_buf */ |
||
56 | int sb_samples[MPA_MAX_CHANNELS][3][12][SBLIMIT]; |
||
57 | unsigned char scale_factors[MPA_MAX_CHANNELS][SBLIMIT][3]; /* scale factors */ |
||
58 | /* code to group 3 scale factors */ |
||
59 | unsigned char scale_code[MPA_MAX_CHANNELS][SBLIMIT]; |
||
60 | int sblimit; /* number of used subbands */ |
||
61 | const unsigned char *alloc_table; |
||
62 | } MpegAudioContext; |
||
63 | |||
64 | /* define it to use floats in quantization (I don't like floats !) */ |
||
65 | #define USE_FLOATS |
||
66 | |||
67 | #include "mpegaudiodata.h" |
||
68 | #include "mpegaudiotab.h" |
||
69 | |||
70 | static av_cold int MPA_encode_init(AVCodecContext *avctx) |
||
71 | { |
||
72 | MpegAudioContext *s = avctx->priv_data; |
||
73 | int freq = avctx->sample_rate; |
||
74 | int bitrate = avctx->bit_rate; |
||
75 | int channels = avctx->channels; |
||
76 | int i, v, table; |
||
77 | float a; |
||
78 | |||
79 | if (channels <= 0 || channels > 2){ |
||
80 | av_log(avctx, AV_LOG_ERROR, "encoding %d channel(s) is not allowed in mp2\n", channels); |
||
81 | return AVERROR(EINVAL); |
||
82 | } |
||
83 | bitrate = bitrate / 1000; |
||
84 | s->nb_channels = channels; |
||
85 | avctx->frame_size = MPA_FRAME_SIZE; |
||
86 | avctx->delay = 512 - 32 + 1; |
||
87 | |||
88 | /* encoding freq */ |
||
89 | s->lsf = 0; |
||
90 | for(i=0;i<3;i++) { |
||
91 | if (avpriv_mpa_freq_tab[i] == freq) |
||
92 | break; |
||
93 | if ((avpriv_mpa_freq_tab[i] / 2) == freq) { |
||
94 | s->lsf = 1; |
||
95 | break; |
||
96 | } |
||
97 | } |
||
98 | if (i == 3){ |
||
99 | av_log(avctx, AV_LOG_ERROR, "Sampling rate %d is not allowed in mp2\n", freq); |
||
100 | return AVERROR(EINVAL); |
||
101 | } |
||
102 | s->freq_index = i; |
||
103 | |||
104 | /* encoding bitrate & frequency */ |
||
105 | for(i=0;i<15;i++) { |
||
106 | if (avpriv_mpa_bitrate_tab[s->lsf][1][i] == bitrate) |
||
107 | break; |
||
108 | } |
||
109 | if (i == 15){ |
||
110 | av_log(avctx, AV_LOG_ERROR, "bitrate %d is not allowed in mp2\n", bitrate); |
||
111 | return AVERROR(EINVAL); |
||
112 | } |
||
113 | s->bitrate_index = i; |
||
114 | |||
115 | /* compute total header size & pad bit */ |
||
116 | |||
117 | a = (float)(bitrate * 1000 * MPA_FRAME_SIZE) / (freq * 8.0); |
||
118 | s->frame_size = ((int)a) * 8; |
||
119 | |||
120 | /* frame fractional size to compute padding */ |
||
121 | s->frame_frac = 0; |
||
122 | s->frame_frac_incr = (int)((a - floor(a)) * 65536.0); |
||
123 | |||
124 | /* select the right allocation table */ |
||
125 | table = ff_mpa_l2_select_table(bitrate, s->nb_channels, freq, s->lsf); |
||
126 | |||
127 | /* number of used subbands */ |
||
128 | s->sblimit = ff_mpa_sblimit_table[table]; |
||
129 | s->alloc_table = ff_mpa_alloc_tables[table]; |
||
130 | |||
131 | av_dlog(avctx, "%d kb/s, %d Hz, frame_size=%d bits, table=%d, padincr=%x\n", |
||
132 | bitrate, freq, s->frame_size, table, s->frame_frac_incr); |
||
133 | |||
134 | for(i=0;i |
||
135 | s->samples_offset[i] = 0; |
||
136 | |||
137 | for(i=0;i<257;i++) { |
||
138 | int v; |
||
139 | v = ff_mpa_enwindow[i]; |
||
140 | #if WFRAC_BITS != 16 |
||
141 | v = (v + (1 << (16 - WFRAC_BITS - 1))) >> (16 - WFRAC_BITS); |
||
142 | #endif |
||
143 | filter_bank[i] = v; |
||
144 | if ((i & 63) != 0) |
||
145 | v = -v; |
||
146 | if (i != 0) |
||
147 | filter_bank[512 - i] = v; |
||
148 | } |
||
149 | |||
150 | for(i=0;i<64;i++) { |
||
151 | v = (int)(exp2((3 - i) / 3.0) * (1 << 20)); |
||
152 | if (v <= 0) |
||
153 | v = 1; |
||
154 | scale_factor_table[i] = v; |
||
155 | #ifdef USE_FLOATS |
||
156 | scale_factor_inv_table[i] = exp2(-(3 - i) / 3.0) / (float)(1 << 20); |
||
157 | #else |
||
158 | #define P 15 |
||
159 | scale_factor_shift[i] = 21 - P - (i / 3); |
||
160 | scale_factor_mult[i] = (1 << P) * exp2((i % 3) / 3.0); |
||
161 | #endif |
||
162 | } |
||
163 | for(i=0;i<128;i++) { |
||
164 | v = i - 64; |
||
165 | if (v <= -3) |
||
166 | v = 0; |
||
167 | else if (v < 0) |
||
168 | v = 1; |
||
169 | else if (v == 0) |
||
170 | v = 2; |
||
171 | else if (v < 3) |
||
172 | v = 3; |
||
173 | else |
||
174 | v = 4; |
||
175 | scale_diff_table[i] = v; |
||
176 | } |
||
177 | |||
178 | for(i=0;i<17;i++) { |
||
179 | v = ff_mpa_quant_bits[i]; |
||
180 | if (v < 0) |
||
181 | v = -v; |
||
182 | else |
||
183 | v = v * 3; |
||
184 | total_quant_bits[i] = 12 * v; |
||
185 | } |
||
186 | |||
187 | return 0; |
||
188 | } |
||
189 | |||
190 | /* 32 point floating point IDCT without 1/sqrt(2) coef zero scaling */ |
||
191 | static void idct32(int *out, int *tab) |
||
192 | { |
||
193 | int i, j; |
||
194 | int *t, *t1, xr; |
||
195 | const int *xp = costab32; |
||
196 | |||
197 | for(j=31;j>=3;j-=2) tab[j] += tab[j - 2]; |
||
198 | |||
199 | t = tab + 30; |
||
200 | t1 = tab + 2; |
||
201 | do { |
||
202 | t[0] += t[-4]; |
||
203 | t[1] += t[1 - 4]; |
||
204 | t -= 4; |
||
205 | } while (t != t1); |
||
206 | |||
207 | t = tab + 28; |
||
208 | t1 = tab + 4; |
||
209 | do { |
||
210 | t[0] += t[-8]; |
||
211 | t[1] += t[1-8]; |
||
212 | t[2] += t[2-8]; |
||
213 | t[3] += t[3-8]; |
||
214 | t -= 8; |
||
215 | } while (t != t1); |
||
216 | |||
217 | t = tab; |
||
218 | t1 = tab + 32; |
||
219 | do { |
||
220 | t[ 3] = -t[ 3]; |
||
221 | t[ 6] = -t[ 6]; |
||
222 | |||
223 | t[11] = -t[11]; |
||
224 | t[12] = -t[12]; |
||
225 | t[13] = -t[13]; |
||
226 | t[15] = -t[15]; |
||
227 | t += 16; |
||
228 | } while (t != t1); |
||
229 | |||
230 | |||
231 | t = tab; |
||
232 | t1 = tab + 8; |
||
233 | do { |
||
234 | int x1, x2, x3, x4; |
||
235 | |||
236 | x3 = MUL(t[16], FIX(SQRT2*0.5)); |
||
237 | x4 = t[0] - x3; |
||
238 | x3 = t[0] + x3; |
||
239 | |||
240 | x2 = MUL(-(t[24] + t[8]), FIX(SQRT2*0.5)); |
||
241 | x1 = MUL((t[8] - x2), xp[0]); |
||
242 | x2 = MUL((t[8] + x2), xp[1]); |
||
243 | |||
244 | t[ 0] = x3 + x1; |
||
245 | t[ 8] = x4 - x2; |
||
246 | t[16] = x4 + x2; |
||
247 | t[24] = x3 - x1; |
||
248 | t++; |
||
249 | } while (t != t1); |
||
250 | |||
251 | xp += 2; |
||
252 | t = tab; |
||
253 | t1 = tab + 4; |
||
254 | do { |
||
255 | xr = MUL(t[28],xp[0]); |
||
256 | t[28] = (t[0] - xr); |
||
257 | t[0] = (t[0] + xr); |
||
258 | |||
259 | xr = MUL(t[4],xp[1]); |
||
260 | t[ 4] = (t[24] - xr); |
||
261 | t[24] = (t[24] + xr); |
||
262 | |||
263 | xr = MUL(t[20],xp[2]); |
||
264 | t[20] = (t[8] - xr); |
||
265 | t[ 8] = (t[8] + xr); |
||
266 | |||
267 | xr = MUL(t[12],xp[3]); |
||
268 | t[12] = (t[16] - xr); |
||
269 | t[16] = (t[16] + xr); |
||
270 | t++; |
||
271 | } while (t != t1); |
||
272 | xp += 4; |
||
273 | |||
274 | for (i = 0; i < 4; i++) { |
||
275 | xr = MUL(tab[30-i*4],xp[0]); |
||
276 | tab[30-i*4] = (tab[i*4] - xr); |
||
277 | tab[ i*4] = (tab[i*4] + xr); |
||
278 | |||
279 | xr = MUL(tab[ 2+i*4],xp[1]); |
||
280 | tab[ 2+i*4] = (tab[28-i*4] - xr); |
||
281 | tab[28-i*4] = (tab[28-i*4] + xr); |
||
282 | |||
283 | xr = MUL(tab[31-i*4],xp[0]); |
||
284 | tab[31-i*4] = (tab[1+i*4] - xr); |
||
285 | tab[ 1+i*4] = (tab[1+i*4] + xr); |
||
286 | |||
287 | xr = MUL(tab[ 3+i*4],xp[1]); |
||
288 | tab[ 3+i*4] = (tab[29-i*4] - xr); |
||
289 | tab[29-i*4] = (tab[29-i*4] + xr); |
||
290 | |||
291 | xp += 2; |
||
292 | } |
||
293 | |||
294 | t = tab + 30; |
||
295 | t1 = tab + 1; |
||
296 | do { |
||
297 | xr = MUL(t1[0], *xp); |
||
298 | t1[0] = (t[0] - xr); |
||
299 | t[0] = (t[0] + xr); |
||
300 | t -= 2; |
||
301 | t1 += 2; |
||
302 | xp++; |
||
303 | } while (t >= tab); |
||
304 | |||
305 | for(i=0;i<32;i++) { |
||
306 | out[i] = tab[bitinv32[i]]; |
||
307 | } |
||
308 | } |
||
309 | |||
310 | #define WSHIFT (WFRAC_BITS + 15 - FRAC_BITS) |
||
311 | |||
312 | static void filter(MpegAudioContext *s, int ch, const short *samples, int incr) |
||
313 | { |
||
314 | short *p, *q; |
||
315 | int sum, offset, i, j; |
||
316 | int tmp[64]; |
||
317 | int tmp1[32]; |
||
318 | int *out; |
||
319 | |||
320 | offset = s->samples_offset[ch]; |
||
321 | out = &s->sb_samples[ch][0][0][0]; |
||
322 | for(j=0;j<36;j++) { |
||
323 | /* 32 samples at once */ |
||
324 | for(i=0;i<32;i++) { |
||
325 | s->samples_buf[ch][offset + (31 - i)] = samples[0]; |
||
326 | samples += incr; |
||
327 | } |
||
328 | |||
329 | /* filter */ |
||
330 | p = s->samples_buf[ch] + offset; |
||
331 | q = filter_bank; |
||
332 | /* maxsum = 23169 */ |
||
333 | for(i=0;i<64;i++) { |
||
334 | sum = p[0*64] * q[0*64]; |
||
335 | sum += p[1*64] * q[1*64]; |
||
336 | sum += p[2*64] * q[2*64]; |
||
337 | sum += p[3*64] * q[3*64]; |
||
338 | sum += p[4*64] * q[4*64]; |
||
339 | sum += p[5*64] * q[5*64]; |
||
340 | sum += p[6*64] * q[6*64]; |
||
341 | sum += p[7*64] * q[7*64]; |
||
342 | tmp[i] = sum; |
||
343 | p++; |
||
344 | q++; |
||
345 | } |
||
346 | tmp1[0] = tmp[16] >> WSHIFT; |
||
347 | for( i=1; i<=16; i++ ) tmp1[i] = (tmp[i+16]+tmp[16-i]) >> WSHIFT; |
||
348 | for( i=17; i<=31; i++ ) tmp1[i] = (tmp[i+16]-tmp[80-i]) >> WSHIFT; |
||
349 | |||
350 | idct32(out, tmp1); |
||
351 | |||
352 | /* advance of 32 samples */ |
||
353 | offset -= 32; |
||
354 | out += 32; |
||
355 | /* handle the wrap around */ |
||
356 | if (offset < 0) { |
||
357 | memmove(s->samples_buf[ch] + SAMPLES_BUF_SIZE - (512 - 32), |
||
358 | s->samples_buf[ch], (512 - 32) * 2); |
||
359 | offset = SAMPLES_BUF_SIZE - 512; |
||
360 | } |
||
361 | } |
||
362 | s->samples_offset[ch] = offset; |
||
363 | } |
||
364 | |||
365 | static void compute_scale_factors(unsigned char scale_code[SBLIMIT], |
||
366 | unsigned char scale_factors[SBLIMIT][3], |
||
367 | int sb_samples[3][12][SBLIMIT], |
||
368 | int sblimit) |
||
369 | { |
||
370 | int *p, vmax, v, n, i, j, k, code; |
||
371 | int index, d1, d2; |
||
372 | unsigned char *sf = &scale_factors[0][0]; |
||
373 | |||
374 | for(j=0;j |
||
375 | for(i=0;i<3;i++) { |
||
376 | /* find the max absolute value */ |
||
377 | p = &sb_samples[i][0][j]; |
||
378 | vmax = abs(*p); |
||
379 | for(k=1;k<12;k++) { |
||
380 | p += SBLIMIT; |
||
381 | v = abs(*p); |
||
382 | if (v > vmax) |
||
383 | vmax = v; |
||
384 | } |
||
385 | /* compute the scale factor index using log 2 computations */ |
||
386 | if (vmax > 1) { |
||
387 | n = av_log2(vmax); |
||
388 | /* n is the position of the MSB of vmax. now |
||
389 | use at most 2 compares to find the index */ |
||
390 | index = (21 - n) * 3 - 3; |
||
391 | if (index >= 0) { |
||
392 | while (vmax <= scale_factor_table[index+1]) |
||
393 | index++; |
||
394 | } else { |
||
395 | index = 0; /* very unlikely case of overflow */ |
||
396 | } |
||
397 | } else { |
||
398 | index = 62; /* value 63 is not allowed */ |
||
399 | } |
||
400 | |||
401 | av_dlog(NULL, "%2d:%d in=%x %x %d\n", |
||
402 | j, i, vmax, scale_factor_table[index], index); |
||
403 | /* store the scale factor */ |
||
404 | av_assert2(index >=0 && index <= 63); |
||
405 | sf[i] = index; |
||
406 | } |
||
407 | |||
408 | /* compute the transmission factor : look if the scale factors |
||
409 | are close enough to each other */ |
||
410 | d1 = scale_diff_table[sf[0] - sf[1] + 64]; |
||
411 | d2 = scale_diff_table[sf[1] - sf[2] + 64]; |
||
412 | |||
413 | /* handle the 25 cases */ |
||
414 | switch(d1 * 5 + d2) { |
||
415 | case 0*5+0: |
||
416 | case 0*5+4: |
||
417 | case 3*5+4: |
||
418 | case 4*5+0: |
||
419 | case 4*5+4: |
||
420 | code = 0; |
||
421 | break; |
||
422 | case 0*5+1: |
||
423 | case 0*5+2: |
||
424 | case 4*5+1: |
||
425 | case 4*5+2: |
||
426 | code = 3; |
||
427 | sf[2] = sf[1]; |
||
428 | break; |
||
429 | case 0*5+3: |
||
430 | case 4*5+3: |
||
431 | code = 3; |
||
432 | sf[1] = sf[2]; |
||
433 | break; |
||
434 | case 1*5+0: |
||
435 | case 1*5+4: |
||
436 | case 2*5+4: |
||
437 | code = 1; |
||
438 | sf[1] = sf[0]; |
||
439 | break; |
||
440 | case 1*5+1: |
||
441 | case 1*5+2: |
||
442 | case 2*5+0: |
||
443 | case 2*5+1: |
||
444 | case 2*5+2: |
||
445 | code = 2; |
||
446 | sf[1] = sf[2] = sf[0]; |
||
447 | break; |
||
448 | case 2*5+3: |
||
449 | case 3*5+3: |
||
450 | code = 2; |
||
451 | sf[0] = sf[1] = sf[2]; |
||
452 | break; |
||
453 | case 3*5+0: |
||
454 | case 3*5+1: |
||
455 | case 3*5+2: |
||
456 | code = 2; |
||
457 | sf[0] = sf[2] = sf[1]; |
||
458 | break; |
||
459 | case 1*5+3: |
||
460 | code = 2; |
||
461 | if (sf[0] > sf[2]) |
||
462 | sf[0] = sf[2]; |
||
463 | sf[1] = sf[2] = sf[0]; |
||
464 | break; |
||
465 | default: |
||
466 | av_assert2(0); //cannot happen |
||
467 | code = 0; /* kill warning */ |
||
468 | } |
||
469 | |||
470 | av_dlog(NULL, "%d: %2d %2d %2d %d %d -> %d\n", j, |
||
471 | sf[0], sf[1], sf[2], d1, d2, code); |
||
472 | scale_code[j] = code; |
||
473 | sf += 3; |
||
474 | } |
||
475 | } |
||
476 | |||
477 | /* The most important function : psycho acoustic module. In this |
||
478 | encoder there is basically none, so this is the worst you can do, |
||
479 | but also this is the simpler. */ |
||
480 | static void psycho_acoustic_model(MpegAudioContext *s, short smr[SBLIMIT]) |
||
481 | { |
||
482 | int i; |
||
483 | |||
484 | for(i=0;i |
||
485 | smr[i] = (int)(fixed_smr[i] * 10); |
||
486 | } |
||
487 | } |
||
488 | |||
489 | |||
490 | #define SB_NOTALLOCATED 0 |
||
491 | #define SB_ALLOCATED 1 |
||
492 | #define SB_NOMORE 2 |
||
493 | |||
494 | /* Try to maximize the smr while using a number of bits inferior to |
||
495 | the frame size. I tried to make the code simpler, faster and |
||
496 | smaller than other encoders :-) */ |
||
497 | static void compute_bit_allocation(MpegAudioContext *s, |
||
498 | short smr1[MPA_MAX_CHANNELS][SBLIMIT], |
||
499 | unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT], |
||
500 | int *padding) |
||
501 | { |
||
502 | int i, ch, b, max_smr, max_ch, max_sb, current_frame_size, max_frame_size; |
||
503 | int incr; |
||
504 | short smr[MPA_MAX_CHANNELS][SBLIMIT]; |
||
505 | unsigned char subband_status[MPA_MAX_CHANNELS][SBLIMIT]; |
||
506 | const unsigned char *alloc; |
||
507 | |||
508 | memcpy(smr, smr1, s->nb_channels * sizeof(short) * SBLIMIT); |
||
509 | memset(subband_status, SB_NOTALLOCATED, s->nb_channels * SBLIMIT); |
||
510 | memset(bit_alloc, 0, s->nb_channels * SBLIMIT); |
||
511 | |||
512 | /* compute frame size and padding */ |
||
513 | max_frame_size = s->frame_size; |
||
514 | s->frame_frac += s->frame_frac_incr; |
||
515 | if (s->frame_frac >= 65536) { |
||
516 | s->frame_frac -= 65536; |
||
517 | s->do_padding = 1; |
||
518 | max_frame_size += 8; |
||
519 | } else { |
||
520 | s->do_padding = 0; |
||
521 | } |
||
522 | |||
523 | /* compute the header + bit alloc size */ |
||
524 | current_frame_size = 32; |
||
525 | alloc = s->alloc_table; |
||
526 | for(i=0;i |
||
527 | incr = alloc[0]; |
||
528 | current_frame_size += incr * s->nb_channels; |
||
529 | alloc += 1 << incr; |
||
530 | } |
||
531 | for(;;) { |
||
532 | /* look for the subband with the largest signal to mask ratio */ |
||
533 | max_sb = -1; |
||
534 | max_ch = -1; |
||
535 | max_smr = INT_MIN; |
||
536 | for(ch=0;ch |
||
537 | for(i=0;i |
||
538 | if (smr[ch][i] > max_smr && subband_status[ch][i] != SB_NOMORE) { |
||
539 | max_smr = smr[ch][i]; |
||
540 | max_sb = i; |
||
541 | max_ch = ch; |
||
542 | } |
||
543 | } |
||
544 | } |
||
545 | if (max_sb < 0) |
||
546 | break; |
||
547 | av_dlog(NULL, "current=%d max=%d max_sb=%d max_ch=%d alloc=%d\n", |
||
548 | current_frame_size, max_frame_size, max_sb, max_ch, |
||
549 | bit_alloc[max_ch][max_sb]); |
||
550 | |||
551 | /* find alloc table entry (XXX: not optimal, should use |
||
552 | pointer table) */ |
||
553 | alloc = s->alloc_table; |
||
554 | for(i=0;i |
||
555 | alloc += 1 << alloc[0]; |
||
556 | } |
||
557 | |||
558 | if (subband_status[max_ch][max_sb] == SB_NOTALLOCATED) { |
||
559 | /* nothing was coded for this band: add the necessary bits */ |
||
560 | incr = 2 + nb_scale_factors[s->scale_code[max_ch][max_sb]] * 6; |
||
561 | incr += total_quant_bits[alloc[1]]; |
||
562 | } else { |
||
563 | /* increments bit allocation */ |
||
564 | b = bit_alloc[max_ch][max_sb]; |
||
565 | incr = total_quant_bits[alloc[b + 1]] - |
||
566 | total_quant_bits[alloc[b]]; |
||
567 | } |
||
568 | |||
569 | if (current_frame_size + incr <= max_frame_size) { |
||
570 | /* can increase size */ |
||
571 | b = ++bit_alloc[max_ch][max_sb]; |
||
572 | current_frame_size += incr; |
||
573 | /* decrease smr by the resolution we added */ |
||
574 | smr[max_ch][max_sb] = smr1[max_ch][max_sb] - quant_snr[alloc[b]]; |
||
575 | /* max allocation size reached ? */ |
||
576 | if (b == ((1 << alloc[0]) - 1)) |
||
577 | subband_status[max_ch][max_sb] = SB_NOMORE; |
||
578 | else |
||
579 | subband_status[max_ch][max_sb] = SB_ALLOCATED; |
||
580 | } else { |
||
581 | /* cannot increase the size of this subband */ |
||
582 | subband_status[max_ch][max_sb] = SB_NOMORE; |
||
583 | } |
||
584 | } |
||
585 | *padding = max_frame_size - current_frame_size; |
||
586 | av_assert0(*padding >= 0); |
||
587 | } |
||
588 | |||
589 | /* |
||
590 | * Output the mpeg audio layer 2 frame. Note how the code is small |
||
591 | * compared to other encoders :-) |
||
592 | */ |
||
593 | static void encode_frame(MpegAudioContext *s, |
||
594 | unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT], |
||
595 | int padding) |
||
596 | { |
||
597 | int i, j, k, l, bit_alloc_bits, b, ch; |
||
598 | unsigned char *sf; |
||
599 | int q[3]; |
||
600 | PutBitContext *p = &s->pb; |
||
601 | |||
602 | /* header */ |
||
603 | |||
604 | put_bits(p, 12, 0xfff); |
||
605 | put_bits(p, 1, 1 - s->lsf); /* 1 = mpeg1 ID, 0 = mpeg2 lsf ID */ |
||
606 | put_bits(p, 2, 4-2); /* layer 2 */ |
||
607 | put_bits(p, 1, 1); /* no error protection */ |
||
608 | put_bits(p, 4, s->bitrate_index); |
||
609 | put_bits(p, 2, s->freq_index); |
||
610 | put_bits(p, 1, s->do_padding); /* use padding */ |
||
611 | put_bits(p, 1, 0); /* private_bit */ |
||
612 | put_bits(p, 2, s->nb_channels == 2 ? MPA_STEREO : MPA_MONO); |
||
613 | put_bits(p, 2, 0); /* mode_ext */ |
||
614 | put_bits(p, 1, 0); /* no copyright */ |
||
615 | put_bits(p, 1, 1); /* original */ |
||
616 | put_bits(p, 2, 0); /* no emphasis */ |
||
617 | |||
618 | /* bit allocation */ |
||
619 | j = 0; |
||
620 | for(i=0;i |
||
621 | bit_alloc_bits = s->alloc_table[j]; |
||
622 | for(ch=0;ch |
||
623 | put_bits(p, bit_alloc_bits, bit_alloc[ch][i]); |
||
624 | } |
||
625 | j += 1 << bit_alloc_bits; |
||
626 | } |
||
627 | |||
628 | /* scale codes */ |
||
629 | for(i=0;i |
||
630 | for(ch=0;ch |
||
631 | if (bit_alloc[ch][i]) |
||
632 | put_bits(p, 2, s->scale_code[ch][i]); |
||
633 | } |
||
634 | } |
||
635 | |||
636 | /* scale factors */ |
||
637 | for(i=0;i |
||
638 | for(ch=0;ch |
||
639 | if (bit_alloc[ch][i]) { |
||
640 | sf = &s->scale_factors[ch][i][0]; |
||
641 | switch(s->scale_code[ch][i]) { |
||
642 | case 0: |
||
643 | put_bits(p, 6, sf[0]); |
||
644 | put_bits(p, 6, sf[1]); |
||
645 | put_bits(p, 6, sf[2]); |
||
646 | break; |
||
647 | case 3: |
||
648 | case 1: |
||
649 | put_bits(p, 6, sf[0]); |
||
650 | put_bits(p, 6, sf[2]); |
||
651 | break; |
||
652 | case 2: |
||
653 | put_bits(p, 6, sf[0]); |
||
654 | break; |
||
655 | } |
||
656 | } |
||
657 | } |
||
658 | } |
||
659 | |||
660 | /* quantization & write sub band samples */ |
||
661 | |||
662 | for(k=0;k<3;k++) { |
||
663 | for(l=0;l<12;l+=3) { |
||
664 | j = 0; |
||
665 | for(i=0;i |
||
666 | bit_alloc_bits = s->alloc_table[j]; |
||
667 | for(ch=0;ch |
||
668 | b = bit_alloc[ch][i]; |
||
669 | if (b) { |
||
670 | int qindex, steps, m, sample, bits; |
||
671 | /* we encode 3 sub band samples of the same sub band at a time */ |
||
672 | qindex = s->alloc_table[j+b]; |
||
673 | steps = ff_mpa_quant_steps[qindex]; |
||
674 | for(m=0;m<3;m++) { |
||
675 | sample = s->sb_samples[ch][k][l + m][i]; |
||
676 | /* divide by scale factor */ |
||
677 | #ifdef USE_FLOATS |
||
678 | { |
||
679 | float a; |
||
680 | a = (float)sample * scale_factor_inv_table[s->scale_factors[ch][i][k]]; |
||
681 | q[m] = (int)((a + 1.0) * steps * 0.5); |
||
682 | } |
||
683 | #else |
||
684 | { |
||
685 | int q1, e, shift, mult; |
||
686 | e = s->scale_factors[ch][i][k]; |
||
687 | shift = scale_factor_shift[e]; |
||
688 | mult = scale_factor_mult[e]; |
||
689 | |||
690 | /* normalize to P bits */ |
||
691 | if (shift < 0) |
||
692 | q1 = sample << (-shift); |
||
693 | else |
||
694 | q1 = sample >> shift; |
||
695 | q1 = (q1 * mult) >> P; |
||
696 | q[m] = ((q1 + (1 << P)) * steps) >> (P + 1); |
||
697 | } |
||
698 | #endif |
||
699 | if (q[m] >= steps) |
||
700 | q[m] = steps - 1; |
||
701 | av_assert2(q[m] >= 0 && q[m] < steps); |
||
702 | } |
||
703 | bits = ff_mpa_quant_bits[qindex]; |
||
704 | if (bits < 0) { |
||
705 | /* group the 3 values to save bits */ |
||
706 | put_bits(p, -bits, |
||
707 | q[0] + steps * (q[1] + steps * q[2])); |
||
708 | } else { |
||
709 | put_bits(p, bits, q[0]); |
||
710 | put_bits(p, bits, q[1]); |
||
711 | put_bits(p, bits, q[2]); |
||
712 | } |
||
713 | } |
||
714 | } |
||
715 | /* next subband in alloc table */ |
||
716 | j += 1 << bit_alloc_bits; |
||
717 | } |
||
718 | } |
||
719 | } |
||
720 | |||
721 | /* padding */ |
||
722 | for(i=0;i |
||
723 | put_bits(p, 1, 0); |
||
724 | |||
725 | /* flush */ |
||
726 | flush_put_bits(p); |
||
727 | } |
||
728 | |||
729 | static int MPA_encode_frame(AVCodecContext *avctx, AVPacket *avpkt, |
||
730 | const AVFrame *frame, int *got_packet_ptr) |
||
731 | { |
||
732 | MpegAudioContext *s = avctx->priv_data; |
||
733 | const int16_t *samples = (const int16_t *)frame->data[0]; |
||
734 | short smr[MPA_MAX_CHANNELS][SBLIMIT]; |
||
735 | unsigned char bit_alloc[MPA_MAX_CHANNELS][SBLIMIT]; |
||
736 | int padding, i, ret; |
||
737 | |||
738 | for(i=0;i |
||
739 | filter(s, i, samples + i, s->nb_channels); |
||
740 | } |
||
741 | |||
742 | for(i=0;i |
||
743 | compute_scale_factors(s->scale_code[i], s->scale_factors[i], |
||
744 | s->sb_samples[i], s->sblimit); |
||
745 | } |
||
746 | for(i=0;i |
||
747 | psycho_acoustic_model(s, smr[i]); |
||
748 | } |
||
749 | compute_bit_allocation(s, smr, bit_alloc, &padding); |
||
750 | |||
751 | if ((ret = ff_alloc_packet2(avctx, avpkt, MPA_MAX_CODED_FRAME_SIZE)) < 0) |
||
752 | return ret; |
||
753 | |||
754 | init_put_bits(&s->pb, avpkt->data, avpkt->size); |
||
755 | |||
756 | encode_frame(s, bit_alloc, padding); |
||
757 | |||
758 | if (frame->pts != AV_NOPTS_VALUE) |
||
759 | avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay); |
||
760 | |||
761 | avpkt->size = put_bits_count(&s->pb) / 8; |
||
762 | *got_packet_ptr = 1; |
||
763 | return 0; |
||
764 | } |
||
765 | |||
766 | static const AVCodecDefault mp2_defaults[] = { |
||
767 | { "b", "128k" }, |
||
768 | { NULL }, |
||
769 | }; |
||
770 | |||
771 | AVCodec ff_mp2_encoder = { |
||
772 | .name = "mp2", |
||
773 | .long_name = NULL_IF_CONFIG_SMALL("MP2 (MPEG audio layer 2)"), |
||
774 | .type = AVMEDIA_TYPE_AUDIO, |
||
775 | .id = AV_CODEC_ID_MP2, |
||
776 | .priv_data_size = sizeof(MpegAudioContext), |
||
777 | .init = MPA_encode_init, |
||
778 | .encode2 = MPA_encode_frame, |
||
779 | .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S16, |
||
780 | AV_SAMPLE_FMT_NONE }, |
||
781 | .supported_samplerates = (const int[]){ |
||
782 | 44100, 48000, 32000, 22050, 24000, 16000, 0 |
||
783 | }, |
||
784 | .channel_layouts = (const uint64_t[]){ AV_CH_LAYOUT_MONO, |
||
785 | AV_CH_LAYOUT_STEREO, |
||
786 | |||
787 | .defaults = mp2_defaults, |
||
788 | };> |