2 * WMA compatible encoder
3 * Copyright (c) 2007 Michael Niedermayer
5 * This file is part of FFmpeg.
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.
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.
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
29 static int encode_init(AVCodecContext * avctx){
30 WMADecodeContext *s = avctx->priv_data;
31 int i, flags1, flags2;
34 if(avctx->channels > MAX_CHANNELS)
37 /* extract flag infos */
40 if (avctx->codec->id == CODEC_ID_WMAV1) {
41 extradata= av_malloc(4);
42 avctx->extradata_size= 4;
43 extradata[0] = flags1;
44 extradata[1] = flags1>>8;
45 extradata[2] = flags2;
46 extradata[3] = flags2>>8;
47 } else if (avctx->codec->id == CODEC_ID_WMAV2) {
48 extradata= av_mallocz(10);
49 avctx->extradata_size= 10;
50 extradata[0] = flags1;
51 extradata[1] = flags1>>8;
52 extradata[2] = flags1>>16;
53 extradata[3] = flags1>>24;
54 extradata[4] = flags2;
55 extradata[5] = flags2>>8;
58 avctx->extradata= extradata;
59 s->use_exp_vlc = flags2 & 0x0001;
60 s->use_bit_reservoir = flags2 & 0x0002;
61 s->use_variable_block_len = flags2 & 0x0004;
63 ff_wma_init(avctx, flags2);
66 for(i = 0; i < s->nb_block_sizes; i++)
67 ff_mdct_init(&s->mdct_ctx[i], s->frame_len_bits - i + 1, 0);
70 s->block_align= avctx->bit_rate*(int64_t)s->frame_len / (avctx->sample_rate*8);
71 //av_log(NULL, AV_LOG_ERROR, "%d %d %d %d\n", s->block_align, avctx->bit_rate, s->frame_len, avctx->sample_rate);
72 avctx->frame_size= s->frame_len;
78 static void apply_window_and_mdct(AVCodecContext * avctx, signed short * audio, int len) {
79 WMADecodeContext *s = avctx->priv_data;
80 int window_index= s->frame_len_bits - s->block_len_bits;
82 const float * win = s->windows[window_index];
83 int window_len = 1 << s->block_len_bits;
84 float n = window_len/2;
86 for (channel = 0; channel < avctx->channels; channel++) {
87 memcpy(s->output, s->frame_out[channel], sizeof(float)*window_len);
89 for (i = 0; i < len; i++, j += avctx->channels){
90 s->output[i+window_len] = audio[j] / n * win[i];
91 s->frame_out[channel][i] = audio[j] / n * win[window_len - i - 1];
93 ff_mdct_calc(&s->mdct_ctx[window_index], s->coefs[channel], s->output, s->mdct_tmp);
97 //FIXME use for decoding too
98 static void init_exp(WMADecodeContext *s, int ch, int *exp_param){
101 float v, *q, max_scale, *q_end;
103 ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
104 q = s->exponents[ch];
105 q_end = q + s->block_len;
108 /* XXX: use a table */
109 v = pow(10, *exp_param++ * (1.0 / 16.0));
110 max_scale= FFMAX(max_scale, v);
116 s->max_exponent[ch] = max_scale;
119 static void encode_exp_vlc(WMADecodeContext *s, int ch, const int *exp_param){
124 ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
125 q = s->exponents[ch];
126 q_end = q + s->block_len;
127 if (s->version == 1) {
128 last_exp= *exp_param++;
129 assert(last_exp-10 >= 0 && last_exp-10 < 32);
130 put_bits(&s->pb, 5, last_exp - 10);
135 int exp = *exp_param++;
136 int code = exp - last_exp + 60;
137 assert(code >= 0 && code < 120);
138 put_bits(&s->pb, ff_wma_scale_huffbits[code], ff_wma_scale_huffcodes[code]);
139 /* XXX: use a table */
145 static int encode_block(WMADecodeContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], int total_gain){
146 int v, bsize, ch, coef_nb_bits, parse_exponents;
148 int nb_coefs[MAX_CHANNELS];
149 static const int fixed_exp[25]={20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20};
151 //FIXME remove duplication relative to decoder
152 if (s->use_variable_block_len) {
153 assert(0); //FIXME not implemented
155 /* fixed block len */
156 s->next_block_len_bits = s->frame_len_bits;
157 s->prev_block_len_bits = s->frame_len_bits;
158 s->block_len_bits = s->frame_len_bits;
161 s->block_len = 1 << s->block_len_bits;
162 // assert((s->block_pos + s->block_len) <= s->frame_len);
163 bsize = s->frame_len_bits - s->block_len_bits;
166 v = s->coefs_end[bsize] - s->coefs_start;
167 for(ch = 0; ch < s->nb_channels; ch++)
170 int n4 = s->block_len / 2;
171 mdct_norm = 1.0 / (float)n4;
172 if (s->version == 1) {
173 mdct_norm *= sqrt(n4);
177 if (s->nb_channels == 2) {
178 put_bits(&s->pb, 1, s->ms_stereo= 1);
181 for(ch = 0; ch < s->nb_channels; ch++) {
182 if (s->channel_coded[ch]= 1) { //FIXME
183 init_exp(s, ch, fixed_exp);
187 for(ch = 0; ch < s->nb_channels; ch++) {
188 if (s->channel_coded[ch]) {
190 float *coefs, *exponents, mult;
193 coefs1 = s->coefs1[ch];
194 exponents = s->exponents[ch];
195 mult = pow(10, total_gain * 0.05) / s->max_exponent[ch];
197 coefs = src_coefs[ch];
198 if (s->use_noise_coding && 0) {
199 assert(0); //FIXME not implemented
201 coefs += s->coefs_start;
203 for(i = 0;i < n; i++){
204 double t= *coefs++ / (exponents[i] * mult);
205 if(t<-32768 || t>32767)
208 coefs1[i] = lrint(t);
215 for(ch = 0; ch < s->nb_channels; ch++) {
216 int a = s->channel_coded[ch];
217 put_bits(&s->pb, 1, a);
224 for(v= total_gain-1; v>=127; v-= 127)
225 put_bits(&s->pb, 7, 127);
226 put_bits(&s->pb, 7, v);
228 coef_nb_bits= ff_wma_total_gain_to_bits(total_gain);
230 if (s->use_noise_coding) {
231 for(ch = 0; ch < s->nb_channels; ch++) {
232 if (s->channel_coded[ch]) {
234 n = s->exponent_high_sizes[bsize];
236 put_bits(&s->pb, 1, s->high_band_coded[ch][i]= 0);
238 nb_coefs[ch] -= s->exponent_high_bands[bsize][i];
245 if (s->block_len_bits != s->frame_len_bits) {
246 put_bits(&s->pb, 1, parse_exponents);
249 if (parse_exponents) {
250 for(ch = 0; ch < s->nb_channels; ch++) {
251 if (s->channel_coded[ch]) {
252 if (s->use_exp_vlc) {
253 encode_exp_vlc(s, ch, fixed_exp);
255 assert(0); //FIXME not implemented
256 // encode_exp_lsp(s, ch);
261 assert(0); //FIXME not implemented
264 for(ch = 0; ch < s->nb_channels; ch++) {
265 if (s->channel_coded[ch]) {
268 tindex = (ch == 1 && s->ms_stereo);
269 ptr = &s->coefs1[ch][0];
270 eptr = ptr + nb_coefs[ch];
273 for(;ptr < eptr; ptr++){
276 int abs_level= FFABS(level);
278 if(abs_level <= s->coef_vlcs[tindex]->max_level){
279 if(run < s->coef_vlcs[tindex]->levels[abs_level-1])
280 code= run + s->int_table[tindex][abs_level-1];
283 assert(code < s->coef_vlcs[tindex]->n);
284 put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[code], s->coef_vlcs[tindex]->huffcodes[code]);
287 if(1<<coef_nb_bits <= abs_level)
290 put_bits(&s->pb, coef_nb_bits, abs_level);
291 put_bits(&s->pb, s->frame_len_bits, run);
293 put_bits(&s->pb, 1, level < 0); //FIXME the sign is fliped somewhere
300 put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[1], s->coef_vlcs[tindex]->huffcodes[1]);
302 if (s->version == 1 && s->nb_channels >= 2) {
303 align_put_bits(&s->pb);
309 static int encode_frame(WMADecodeContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], uint8_t *buf, int buf_size, int total_gain){
310 init_put_bits(&s->pb, buf, buf_size);
312 if (s->use_bit_reservoir) {
313 assert(0);//FIXME not implemented
315 if(encode_block(s, src_coefs, total_gain) < 0)
319 align_put_bits(&s->pb);
321 return put_bits_count(&s->pb)/8 - s->block_align;
324 static int encode_superframe(AVCodecContext *avctx,
325 unsigned char *buf, int buf_size, void *data){
326 WMADecodeContext *s = avctx->priv_data;
327 short *samples = data;
328 int i, total_gain, best;
330 s->block_len_bits= s->frame_len_bits; //required by non variable block len
331 s->block_len = 1 << s->block_len_bits;
333 apply_window_and_mdct(avctx, samples, avctx->frame_size);
339 for(i = 0; i < s->block_len; i++) {
340 a = s->coefs[0][i]*0.5;
341 b = s->coefs[1][i]*0.5;
342 s->coefs[0][i] = a + b;
343 s->coefs[1][i] = a - b;
350 int error= encode_frame(s, s->coefs, buf, buf_size, total_gain-i);
356 best= encode_frame(s, s->coefs, buf, buf_size, total_gain);
358 int scoreL= encode_frame(s, s->coefs, buf, buf_size, total_gain-i);
359 int scoreR= encode_frame(s, s->coefs, buf, buf_size, total_gain+i);
360 av_log(NULL, AV_LOG_ERROR, "%d %d %d (%d)\n", scoreL, best, scoreR, total_gain);
361 if(scoreL < FFMIN(best, scoreR)){
364 }else if(scoreR < best){
371 encode_frame(s, s->coefs, buf, buf_size, total_gain);
372 assert((put_bits_count(&s->pb) & 7) == 0);
373 i= s->block_align - (put_bits_count(&s->pb)+7)/8;
376 put_bits(&s->pb, 8, 'N');
378 flush_put_bits(&s->pb);
379 return pbBufPtr(&s->pb) - s->pb.buf;
382 AVCodec wmav1_encoder =
387 sizeof(WMADecodeContext),
393 AVCodec wmav2_encoder =
398 sizeof(WMADecodeContext),