Fix placment of the DPCM Xan comment
[libav.git] / libavcodec / wma.c
CommitLineData
59686073
MN
1/*
2 * WMA compatible codec
406792e7 3 * Copyright (c) 2002-2007 The FFmpeg Project
59686073
MN
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#include "avcodec.h"
23#include "wma.h"
24#include "wmadata.h"
25
26#undef NDEBUG
27#include <assert.h>
28
29/* XXX: use same run/length optimization as mpeg decoders */
30//FIXME maybe split decode / encode or pass flag
ae1374ce
DB
31static void init_coef_vlc(VLC *vlc, uint16_t **prun_table,
32 uint16_t **plevel_table, uint16_t **pint_table,
59686073
MN
33 const CoefVLCTable *vlc_table)
34{
35 int n = vlc_table->n;
ae1374ce
DB
36 const uint8_t *table_bits = vlc_table->huffbits;
37 const uint32_t *table_codes = vlc_table->huffcodes;
59686073
MN
38 const uint16_t *levels_table = vlc_table->levels;
39 uint16_t *run_table, *level_table, *int_table;
40 int i, l, j, k, level;
41
42 init_vlc(vlc, VLCBITS, n, table_bits, 1, 1, table_codes, 4, 4, 0);
43
ae1374ce 44 run_table = av_malloc(n * sizeof(uint16_t));
59686073 45 level_table = av_malloc(n * sizeof(uint16_t));
ae1374ce 46 int_table = av_malloc(n * sizeof(uint16_t));
59686073
MN
47 i = 2;
48 level = 1;
49 k = 0;
50 while (i < n) {
ae1374ce 51 int_table[k] = i;
59686073 52 l = levels_table[k++];
ae1374ce
DB
53 for (j = 0; j < l; j++) {
54 run_table[i] = j;
59686073
MN
55 level_table[i] = level;
56 i++;
57 }
58 level++;
59 }
ae1374ce 60 *prun_table = run_table;
59686073 61 *plevel_table = level_table;
ae1374ce 62 *pint_table = int_table;
59686073
MN
63}
64
cee4bb89
SS
65/**
66 *@brief Get the samples per frame for this stream.
67 *@param sample_rate output sample_rate
68 *@param version wma version
69 *@param decode_flags codec compression features
70 *@return log2 of the number of output samples per frame
71 */
72int av_cold ff_wma_get_frame_len_bits(int sample_rate, int version,
73 unsigned int decode_flags)
74{
75
76 int frame_len_bits;
77
78 if (sample_rate <= 16000)
79 frame_len_bits = 9;
80 else if (sample_rate <= 22050 ||
81 (sample_rate <= 32000 && version == 1))
82 frame_len_bits = 10;
83 else
84 frame_len_bits = 11;
85
86 return frame_len_bits;
87}
88
ae1374ce 89int ff_wma_init(AVCodecContext *avctx, int flags2)
59686073 90{
911b9faf 91 WMACodecContext *s = avctx->priv_data;
59686073 92 int i;
59686073
MN
93 float bps1, high_freq;
94 volatile float bps;
95 int sample_rate1;
96 int coef_vlc_table;
97
ae1374ce
DB
98 if ( avctx->sample_rate <= 0 || avctx->sample_rate > 50000
99 || avctx->channels <= 0 || avctx->channels > 8
100 || avctx->bit_rate <= 0)
47b777ce
MN
101 return -1;
102
59686073
MN
103 s->sample_rate = avctx->sample_rate;
104 s->nb_channels = avctx->channels;
ae1374ce 105 s->bit_rate = avctx->bit_rate;
59686073
MN
106 s->block_align = avctx->block_align;
107
108 dsputil_init(&s->dsp, avctx);
109
110 if (avctx->codec->id == CODEC_ID_WMAV1) {
111 s->version = 1;
112 } else {
113 s->version = 2;
114 }
115
116 /* compute MDCT block size */
cee4bb89
SS
117 s->frame_len_bits = ff_wma_get_frame_len_bits(s->sample_rate, s->version, 0);
118
59686073
MN
119 s->frame_len = 1 << s->frame_len_bits;
120 if (s->use_variable_block_len) {
121 int nb_max, nb;
122 nb = ((flags2 >> 3) & 3) + 1;
123 if ((s->bit_rate / s->nb_channels) >= 32000)
124 nb += 2;
125 nb_max = s->frame_len_bits - BLOCK_MIN_BITS;
126 if (nb > nb_max)
127 nb = nb_max;
128 s->nb_block_sizes = nb + 1;
129 } else {
130 s->nb_block_sizes = 1;
131 }
132
eafcac6a 133 /* init rate dependent parameters */
59686073
MN
134 s->use_noise_coding = 1;
135 high_freq = s->sample_rate * 0.5;
136
137 /* if version 2, then the rates are normalized */
138 sample_rate1 = s->sample_rate;
139 if (s->version == 2) {
140 if (sample_rate1 >= 44100)
141 sample_rate1 = 44100;
142 else if (sample_rate1 >= 22050)
143 sample_rate1 = 22050;
144 else if (sample_rate1 >= 16000)
145 sample_rate1 = 16000;
146 else if (sample_rate1 >= 11025)
147 sample_rate1 = 11025;
148 else if (sample_rate1 >= 8000)
149 sample_rate1 = 8000;
150 }
151
152 bps = (float)s->bit_rate / (float)(s->nb_channels * s->sample_rate);
153 s->byte_offset_bits = av_log2((int)(bps * s->frame_len / 8.0 + 0.5)) + 2;
154
155 /* compute high frequency value and choose if noise coding should
156 be activated */
157 bps1 = bps;
158 if (s->nb_channels == 2)
159 bps1 = bps * 1.6;
160 if (sample_rate1 == 44100) {
161 if (bps1 >= 0.61)
162 s->use_noise_coding = 0;
163 else
164 high_freq = high_freq * 0.4;
165 } else if (sample_rate1 == 22050) {
166 if (bps1 >= 1.16)
167 s->use_noise_coding = 0;
168 else if (bps1 >= 0.72)
169 high_freq = high_freq * 0.7;
170 else
171 high_freq = high_freq * 0.6;
172 } else if (sample_rate1 == 16000) {
173 if (bps > 0.5)
174 high_freq = high_freq * 0.5;
175 else
176 high_freq = high_freq * 0.3;
177 } else if (sample_rate1 == 11025) {
178 high_freq = high_freq * 0.7;
179 } else if (sample_rate1 == 8000) {
180 if (bps <= 0.625) {
181 high_freq = high_freq * 0.5;
182 } else if (bps > 0.75) {
183 s->use_noise_coding = 0;
184 } else {
185 high_freq = high_freq * 0.65;
186 }
187 } else {
188 if (bps >= 0.8) {
189 high_freq = high_freq * 0.75;
190 } else if (bps >= 0.6) {
191 high_freq = high_freq * 0.6;
192 } else {
193 high_freq = high_freq * 0.5;
194 }
195 }
318c5e05
MB
196 dprintf(s->avctx, "flags2=0x%x\n", flags2);
197 dprintf(s->avctx, "version=%d channels=%d sample_rate=%d bitrate=%d block_align=%d\n",
ae1374ce
DB
198 s->version, s->nb_channels, s->sample_rate, s->bit_rate,
199 s->block_align);
318c5e05 200 dprintf(s->avctx, "bps=%f bps1=%f high_freq=%f bitoffset=%d\n",
ae1374ce 201 bps, bps1, high_freq, s->byte_offset_bits);
318c5e05 202 dprintf(s->avctx, "use_noise_coding=%d use_exp_vlc=%d nb_block_sizes=%d\n",
ae1374ce 203 s->use_noise_coding, s->use_exp_vlc, s->nb_block_sizes);
59686073
MN
204
205 /* compute the scale factor band sizes for each MDCT block size */
206 {
207 int a, b, pos, lpos, k, block_len, i, j, n;
208 const uint8_t *table;
209
210 if (s->version == 1) {
211 s->coefs_start = 3;
212 } else {
213 s->coefs_start = 0;
214 }
ae1374ce 215 for (k = 0; k < s->nb_block_sizes; k++) {
59686073
MN
216 block_len = s->frame_len >> k;
217
218 if (s->version == 1) {
219 lpos = 0;
ae1374ce 220 for (i = 0; i < 25; i++) {
59686073
MN
221 a = wma_critical_freqs[i];
222 b = s->sample_rate;
ae1374ce 223 pos = ((block_len * 2 * a) + (b >> 1)) / b;
59686073
MN
224 if (pos > block_len)
225 pos = block_len;
226 s->exponent_bands[0][i] = pos - lpos;
227 if (pos >= block_len) {
228 i++;
229 break;
230 }
231 lpos = pos;
232 }
233 s->exponent_sizes[0] = i;
234 } else {
235 /* hardcoded tables */
236 table = NULL;
237 a = s->frame_len_bits - BLOCK_MIN_BITS - k;
238 if (a < 3) {
239 if (s->sample_rate >= 44100)
240 table = exponent_band_44100[a];
241 else if (s->sample_rate >= 32000)
242 table = exponent_band_32000[a];
243 else if (s->sample_rate >= 22050)
244 table = exponent_band_22050[a];
245 }
246 if (table) {
247 n = *table++;
ae1374ce 248 for (i = 0; i < n; i++)
59686073
MN
249 s->exponent_bands[k][i] = table[i];
250 s->exponent_sizes[k] = n;
251 } else {
252 j = 0;
253 lpos = 0;
ae1374ce 254 for (i = 0; i < 25; i++) {
59686073
MN
255 a = wma_critical_freqs[i];
256 b = s->sample_rate;
ae1374ce 257 pos = ((block_len * 2 * a) + (b << 1)) / (4 * b);
59686073
MN
258 pos <<= 2;
259 if (pos > block_len)
260 pos = block_len;
261 if (pos > lpos)
262 s->exponent_bands[k][j++] = pos - lpos;
263 if (pos >= block_len)
264 break;
265 lpos = pos;
266 }
267 s->exponent_sizes[k] = j;
268 }
269 }
270
271 /* max number of coefs */
272 s->coefs_end[k] = (s->frame_len - ((s->frame_len * 9) / 100)) >> k;
273 /* high freq computation */
274 s->high_band_start[k] = (int)((block_len * 2 * high_freq) /
275 s->sample_rate + 0.5);
276 n = s->exponent_sizes[k];
277 j = 0;
278 pos = 0;
ae1374ce 279 for (i = 0; i < n; i++) {
59686073
MN
280 int start, end;
281 start = pos;
282 pos += s->exponent_bands[k][i];
283 end = pos;
284 if (start < s->high_band_start[k])
285 start = s->high_band_start[k];
286 if (end > s->coefs_end[k])
287 end = s->coefs_end[k];
288 if (end > start)
289 s->exponent_high_bands[k][j++] = end - start;
290 }
291 s->exponent_high_sizes[k] = j;
292#if 0
a9c9a240 293 tprintf(s->avctx, "%5d: coefs_end=%d high_band_start=%d nb_high_bands=%d: ",
ae1374ce
DB
294 s->frame_len >> k,
295 s->coefs_end[k],
296 s->high_band_start[k],
297 s->exponent_high_sizes[k]);
298 for (j = 0; j < s->exponent_high_sizes[k]; j++)
a9c9a240
MB
299 tprintf(s->avctx, " %d", s->exponent_high_bands[k][j]);
300 tprintf(s->avctx, "\n");
59686073
MN
301#endif
302 }
303 }
304
305#ifdef TRACE
306 {
307 int i, j;
ae1374ce 308 for (i = 0; i < s->nb_block_sizes; i++) {
a9c9a240 309 tprintf(s->avctx, "%5d: n=%2d:",
ae1374ce
DB
310 s->frame_len >> i,
311 s->exponent_sizes[i]);
312 for (j = 0; j < s->exponent_sizes[i]; j++)
a9c9a240
MB
313 tprintf(s->avctx, " %d", s->exponent_bands[i][j]);
314 tprintf(s->avctx, "\n");
59686073
MN
315 }
316 }
317#endif
318
319 /* init MDCT windows : simple sinus window */
ae1374ce 320 for (i = 0; i < s->nb_block_sizes; i++) {
ece6b83c 321 int n;
59686073 322 n = 1 << (s->frame_len_bits - i);
cbcbf439
RS
323 ff_sine_window_init(ff_sine_windows[s->frame_len_bits - i - 7], n);
324 s->windows[i] = ff_sine_windows[s->frame_len_bits - i - 7];
59686073
MN
325 }
326
327 s->reset_block_lengths = 1;
328
329 if (s->use_noise_coding) {
330
331 /* init the noise generator */
332 if (s->use_exp_vlc)
333 s->noise_mult = 0.02;
334 else
335 s->noise_mult = 0.04;
336
337#ifdef TRACE
ae1374ce 338 for (i = 0; i < NOISE_TAB_SIZE; i++)
59686073
MN
339 s->noise_table[i] = 1.0 * s->noise_mult;
340#else
341 {
342 unsigned int seed;
343 float norm;
344 seed = 1;
345 norm = (1.0 / (float)(1LL << 31)) * sqrt(3) * s->noise_mult;
ae1374ce 346 for (i = 0; i < NOISE_TAB_SIZE; i++) {
59686073
MN
347 seed = seed * 314159 + 1;
348 s->noise_table[i] = (float)((int)seed) * norm;
349 }
350 }
351#endif
352 }
353
354 /* choose the VLC tables for the coefficients */
355 coef_vlc_table = 2;
356 if (s->sample_rate >= 32000) {
357 if (bps1 < 0.72)
358 coef_vlc_table = 0;
359 else if (bps1 < 1.16)
360 coef_vlc_table = 1;
361 }
362 s->coef_vlcs[0]= &coef_vlcs[coef_vlc_table * 2 ];
363 s->coef_vlcs[1]= &coef_vlcs[coef_vlc_table * 2 + 1];
364 init_coef_vlc(&s->coef_vlc[0], &s->run_table[0], &s->level_table[0], &s->int_table[0],
365 s->coef_vlcs[0]);
366 init_coef_vlc(&s->coef_vlc[1], &s->run_table[1], &s->level_table[1], &s->int_table[1],
367 s->coef_vlcs[1]);
368
369 return 0;
370}
371
ae1374ce
DB
372int ff_wma_total_gain_to_bits(int total_gain)
373{
59686073
MN
374 if (total_gain < 15) return 13;
375 else if (total_gain < 32) return 12;
376 else if (total_gain < 40) return 11;
377 else if (total_gain < 45) return 10;
378 else return 9;
379}
380
381int ff_wma_end(AVCodecContext *avctx)
382{
911b9faf 383 WMACodecContext *s = avctx->priv_data;
59686073
MN
384 int i;
385
ae1374ce 386 for (i = 0; i < s->nb_block_sizes; i++)
59686073 387 ff_mdct_end(&s->mdct_ctx[i]);
59686073
MN
388
389 if (s->use_exp_vlc) {
390 free_vlc(&s->exp_vlc);
391 }
392 if (s->use_noise_coding) {
393 free_vlc(&s->hgain_vlc);
394 }
ae1374ce 395 for (i = 0; i < 2; i++) {
59686073
MN
396 free_vlc(&s->coef_vlc[i]);
397 av_free(s->run_table[i]);
398 av_free(s->level_table[i]);
2c79288d 399 av_free(s->int_table[i]);
59686073
MN
400 }
401
402 return 0;
403}