2bbb0994cc9157e882d189b301b8833ae325013d
[ffmpeg.git] / libavcodec / wma.c
1 /*
2  * WMA compatible codec
3  * Copyright (c) 2002-2007 The FFmpeg Project.
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
31 static void init_coef_vlc(VLC *vlc,
32                           uint16_t **prun_table, uint16_t **plevel_table, uint16_t **pint_table,
33                           const CoefVLCTable *vlc_table)
34 {
35     int n = vlc_table->n;
36     const uint8_t *table_bits = vlc_table->huffbits;
37     const uint32_t *table_codes = vlc_table->huffcodes;
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
44     run_table = av_malloc(n * sizeof(uint16_t));
45     level_table = av_malloc(n * sizeof(uint16_t));
46     int_table = av_malloc(n * sizeof(uint16_t));
47     i = 2;
48     level = 1;
49     k = 0;
50     while (i < n) {
51         int_table[k]= i;
52         l = levels_table[k++];
53         for(j=0;j<l;j++) {
54             run_table[i] = j;
55             level_table[i] = level;
56             i++;
57         }
58         level++;
59     }
60     *prun_table = run_table;
61     *plevel_table = level_table;
62     *pint_table= int_table;
63 }
64
65 int ff_wma_init(AVCodecContext * avctx, int flags2)
66 {
67     WMACodecContext *s = avctx->priv_data;
68     int i;
69     float bps1, high_freq;
70     volatile float bps;
71     int sample_rate1;
72     int coef_vlc_table;
73
74     if(   avctx->sample_rate<=0 || avctx->sample_rate>50000
75        || avctx->channels<=0 || avctx->channels>8
76        || avctx->bit_rate<=0)
77         return -1;
78
79     s->sample_rate = avctx->sample_rate;
80     s->nb_channels = avctx->channels;
81     s->bit_rate = avctx->bit_rate;
82     s->block_align = avctx->block_align;
83
84     dsputil_init(&s->dsp, avctx);
85
86     if (avctx->codec->id == CODEC_ID_WMAV1) {
87         s->version = 1;
88     } else {
89         s->version = 2;
90     }
91
92     /* compute MDCT block size */
93     if (s->sample_rate <= 16000) {
94         s->frame_len_bits = 9;
95     } else if (s->sample_rate <= 22050 ||
96                (s->sample_rate <= 32000 && s->version == 1)) {
97         s->frame_len_bits = 10;
98     } else {
99         s->frame_len_bits = 11;
100     }
101     s->frame_len = 1 << s->frame_len_bits;
102     if (s->use_variable_block_len) {
103         int nb_max, nb;
104         nb = ((flags2 >> 3) & 3) + 1;
105         if ((s->bit_rate / s->nb_channels) >= 32000)
106             nb += 2;
107         nb_max = s->frame_len_bits - BLOCK_MIN_BITS;
108         if (nb > nb_max)
109             nb = nb_max;
110         s->nb_block_sizes = nb + 1;
111     } else {
112         s->nb_block_sizes = 1;
113     }
114
115     /* init rate dependent parameters */
116     s->use_noise_coding = 1;
117     high_freq = s->sample_rate * 0.5;
118
119     /* if version 2, then the rates are normalized */
120     sample_rate1 = s->sample_rate;
121     if (s->version == 2) {
122         if (sample_rate1 >= 44100)
123             sample_rate1 = 44100;
124         else if (sample_rate1 >= 22050)
125             sample_rate1 = 22050;
126         else if (sample_rate1 >= 16000)
127             sample_rate1 = 16000;
128         else if (sample_rate1 >= 11025)
129             sample_rate1 = 11025;
130         else if (sample_rate1 >= 8000)
131             sample_rate1 = 8000;
132     }
133
134     bps = (float)s->bit_rate / (float)(s->nb_channels * s->sample_rate);
135     s->byte_offset_bits = av_log2((int)(bps * s->frame_len / 8.0 + 0.5)) + 2;
136
137     /* compute high frequency value and choose if noise coding should
138        be activated */
139     bps1 = bps;
140     if (s->nb_channels == 2)
141         bps1 = bps * 1.6;
142     if (sample_rate1 == 44100) {
143         if (bps1 >= 0.61)
144             s->use_noise_coding = 0;
145         else
146             high_freq = high_freq * 0.4;
147     } else if (sample_rate1 == 22050) {
148         if (bps1 >= 1.16)
149             s->use_noise_coding = 0;
150         else if (bps1 >= 0.72)
151             high_freq = high_freq * 0.7;
152         else
153             high_freq = high_freq * 0.6;
154     } else if (sample_rate1 == 16000) {
155         if (bps > 0.5)
156             high_freq = high_freq * 0.5;
157         else
158             high_freq = high_freq * 0.3;
159     } else if (sample_rate1 == 11025) {
160         high_freq = high_freq * 0.7;
161     } else if (sample_rate1 == 8000) {
162         if (bps <= 0.625) {
163             high_freq = high_freq * 0.5;
164         } else if (bps > 0.75) {
165             s->use_noise_coding = 0;
166         } else {
167             high_freq = high_freq * 0.65;
168         }
169     } else {
170         if (bps >= 0.8) {
171             high_freq = high_freq * 0.75;
172         } else if (bps >= 0.6) {
173             high_freq = high_freq * 0.6;
174         } else {
175             high_freq = high_freq * 0.5;
176         }
177     }
178     dprintf(s->avctx, "flags2=0x%x\n", flags2);
179     dprintf(s->avctx, "version=%d channels=%d sample_rate=%d bitrate=%d block_align=%d\n",
180            s->version, s->nb_channels, s->sample_rate, s->bit_rate,
181            s->block_align);
182     dprintf(s->avctx, "bps=%f bps1=%f high_freq=%f bitoffset=%d\n",
183            bps, bps1, high_freq, s->byte_offset_bits);
184     dprintf(s->avctx, "use_noise_coding=%d use_exp_vlc=%d nb_block_sizes=%d\n",
185            s->use_noise_coding, s->use_exp_vlc, s->nb_block_sizes);
186
187     /* compute the scale factor band sizes for each MDCT block size */
188     {
189         int a, b, pos, lpos, k, block_len, i, j, n;
190         const uint8_t *table;
191
192         if (s->version == 1) {
193             s->coefs_start = 3;
194         } else {
195             s->coefs_start = 0;
196         }
197         for(k = 0; k < s->nb_block_sizes; k++) {
198             block_len = s->frame_len >> k;
199
200             if (s->version == 1) {
201                 lpos = 0;
202                 for(i=0;i<25;i++) {
203                     a = wma_critical_freqs[i];
204                     b = s->sample_rate;
205                     pos = ((block_len * 2 * a)  + (b >> 1)) / b;
206                     if (pos > block_len)
207                         pos = block_len;
208                     s->exponent_bands[0][i] = pos - lpos;
209                     if (pos >= block_len) {
210                         i++;
211                         break;
212                     }
213                     lpos = pos;
214                 }
215                 s->exponent_sizes[0] = i;
216             } else {
217                 /* hardcoded tables */
218                 table = NULL;
219                 a = s->frame_len_bits - BLOCK_MIN_BITS - k;
220                 if (a < 3) {
221                     if (s->sample_rate >= 44100)
222                         table = exponent_band_44100[a];
223                     else if (s->sample_rate >= 32000)
224                         table = exponent_band_32000[a];
225                     else if (s->sample_rate >= 22050)
226                         table = exponent_band_22050[a];
227                 }
228                 if (table) {
229                     n = *table++;
230                     for(i=0;i<n;i++)
231                         s->exponent_bands[k][i] = table[i];
232                     s->exponent_sizes[k] = n;
233                 } else {
234                     j = 0;
235                     lpos = 0;
236                     for(i=0;i<25;i++) {
237                         a = wma_critical_freqs[i];
238                         b = s->sample_rate;
239                         pos = ((block_len * 2 * a)  + (b << 1)) / (4 * b);
240                         pos <<= 2;
241                         if (pos > block_len)
242                             pos = block_len;
243                         if (pos > lpos)
244                             s->exponent_bands[k][j++] = pos - lpos;
245                         if (pos >= block_len)
246                             break;
247                         lpos = pos;
248                     }
249                     s->exponent_sizes[k] = j;
250                 }
251             }
252
253             /* max number of coefs */
254             s->coefs_end[k] = (s->frame_len - ((s->frame_len * 9) / 100)) >> k;
255             /* high freq computation */
256             s->high_band_start[k] = (int)((block_len * 2 * high_freq) /
257                                           s->sample_rate + 0.5);
258             n = s->exponent_sizes[k];
259             j = 0;
260             pos = 0;
261             for(i=0;i<n;i++) {
262                 int start, end;
263                 start = pos;
264                 pos += s->exponent_bands[k][i];
265                 end = pos;
266                 if (start < s->high_band_start[k])
267                     start = s->high_band_start[k];
268                 if (end > s->coefs_end[k])
269                     end = s->coefs_end[k];
270                 if (end > start)
271                     s->exponent_high_bands[k][j++] = end - start;
272             }
273             s->exponent_high_sizes[k] = j;
274 #if 0
275             tprintf(s->avctx, "%5d: coefs_end=%d high_band_start=%d nb_high_bands=%d: ",
276                   s->frame_len >> k,
277                   s->coefs_end[k],
278                   s->high_band_start[k],
279                   s->exponent_high_sizes[k]);
280             for(j=0;j<s->exponent_high_sizes[k];j++)
281                 tprintf(s->avctx, " %d", s->exponent_high_bands[k][j]);
282             tprintf(s->avctx, "\n");
283 #endif
284         }
285     }
286
287 #ifdef TRACE
288     {
289         int i, j;
290         for(i = 0; i < s->nb_block_sizes; i++) {
291             tprintf(s->avctx, "%5d: n=%2d:",
292                    s->frame_len >> i,
293                    s->exponent_sizes[i]);
294             for(j=0;j<s->exponent_sizes[i];j++)
295                 tprintf(s->avctx, " %d", s->exponent_bands[i][j]);
296             tprintf(s->avctx, "\n");
297         }
298     }
299 #endif
300
301     /* init MDCT windows : simple sinus window */
302     for(i = 0; i < s->nb_block_sizes; i++) {
303         int n;
304         n = 1 << (s->frame_len_bits - i);
305         ff_sine_window_init(ff_sine_windows[s->frame_len_bits - i - 7], n);
306         s->windows[i] = ff_sine_windows[s->frame_len_bits - i - 7];
307     }
308
309     s->reset_block_lengths = 1;
310
311     if (s->use_noise_coding) {
312
313         /* init the noise generator */
314         if (s->use_exp_vlc)
315             s->noise_mult = 0.02;
316         else
317             s->noise_mult = 0.04;
318
319 #ifdef TRACE
320         for(i=0;i<NOISE_TAB_SIZE;i++)
321             s->noise_table[i] = 1.0 * s->noise_mult;
322 #else
323         {
324             unsigned int seed;
325             float norm;
326             seed = 1;
327             norm = (1.0 / (float)(1LL << 31)) * sqrt(3) * s->noise_mult;
328             for(i=0;i<NOISE_TAB_SIZE;i++) {
329                 seed = seed * 314159 + 1;
330                 s->noise_table[i] = (float)((int)seed) * norm;
331             }
332         }
333 #endif
334     }
335
336     /* choose the VLC tables for the coefficients */
337     coef_vlc_table = 2;
338     if (s->sample_rate >= 32000) {
339         if (bps1 < 0.72)
340             coef_vlc_table = 0;
341         else if (bps1 < 1.16)
342             coef_vlc_table = 1;
343     }
344     s->coef_vlcs[0]= &coef_vlcs[coef_vlc_table * 2    ];
345     s->coef_vlcs[1]= &coef_vlcs[coef_vlc_table * 2 + 1];
346     init_coef_vlc(&s->coef_vlc[0], &s->run_table[0], &s->level_table[0], &s->int_table[0],
347                   s->coef_vlcs[0]);
348     init_coef_vlc(&s->coef_vlc[1], &s->run_table[1], &s->level_table[1], &s->int_table[1],
349                   s->coef_vlcs[1]);
350
351     return 0;
352 }
353
354 int ff_wma_total_gain_to_bits(int total_gain){
355          if (total_gain < 15) return 13;
356     else if (total_gain < 32) return 12;
357     else if (total_gain < 40) return 11;
358     else if (total_gain < 45) return 10;
359     else                      return  9;
360 }
361
362 int ff_wma_end(AVCodecContext *avctx)
363 {
364     WMACodecContext *s = avctx->priv_data;
365     int i;
366
367     for(i = 0; i < s->nb_block_sizes; i++)
368         ff_mdct_end(&s->mdct_ctx[i]);
369
370     if (s->use_exp_vlc) {
371         free_vlc(&s->exp_vlc);
372     }
373     if (s->use_noise_coding) {
374         free_vlc(&s->hgain_vlc);
375     }
376     for(i = 0;i < 2; i++) {
377         free_vlc(&s->coef_vlc[i]);
378         av_free(s->run_table[i]);
379         av_free(s->level_table[i]);
380         av_free(s->int_table[i]);
381     }
382
383     return 0;
384 }