Disable B-frames decoding until they produce correct picture
[libav.git] / libavcodec / oggvorbis.c
CommitLineData
983e3246
MN
1/**
2 * @file oggvorbis.c
3 * Ogg Vorbis codec support via libvorbisenc.
4 * @author Mark Hills <mark@pogo.org.uk>
81e0d0b4
MH
5 */
6
81e0d0b4
MH
7#include <vorbis/vorbisenc.h>
8
9#include "avcodec.h"
81e0d0b4 10
3f4993f1
MN
11#undef NDEBUG
12#include <assert.h>
13
6d8f985e 14#define OGGVORBIS_FRAME_SIZE 64
81e0d0b4 15
6d8f985e 16#define BUFFER_SIZE (1024*64)
81e0d0b4
MH
17
18typedef struct OggVorbisContext {
19 vorbis_info vi ;
20 vorbis_dsp_state vd ;
21 vorbis_block vb ;
6d8f985e
MN
22 uint8_t buffer[BUFFER_SIZE];
23 int buffer_index;
3aca208a
MH
24
25 /* decoder */
26 vorbis_comment vc ;
d76f581f 27 ogg_packet op;
81e0d0b4
MH
28} OggVorbisContext ;
29
30
bbb77e7c 31static int oggvorbis_init_encoder(vorbis_info *vi, AVCodecContext *avccontext) {
13c71451 32 double cfreq;
c55427f8 33
1ff8f2b3 34 if(avccontext->flags & CODEC_FLAG_QSCALE) {
13c71451
JR
35 /* variable bitrate */
36 if(vorbis_encode_setup_vbr(vi, avccontext->channels,
1ff8f2b3 37 avccontext->sample_rate,
13c71451
JR
38 avccontext->global_quality / (float)FF_QP2LAMBDA))
39 return -1;
40 } else {
41 /* constant bitrate */
42 if(vorbis_encode_setup_managed(vi, avccontext->channels,
43 avccontext->sample_rate, -1, avccontext->bit_rate, -1))
44 return -1;
45
c55427f8 46#ifdef OGGVORBIS_VBR_BY_ESTIMATE
13c71451
JR
47 /* variable bitrate by estimate */
48 if(vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE_AVG, NULL))
49 return -1;
50#endif
51 }
c55427f8 52
13c71451
JR
53 /* cutoff frequency */
54 if(avccontext->cutoff > 0) {
55 cfreq = avccontext->cutoff / 1000.0;
56 if(vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq))
57 return -1;
58 }
81e0d0b4 59
13c71451 60 return vorbis_encode_setup_init(vi);
81e0d0b4
MH
61}
62
81e0d0b4
MH
63static int oggvorbis_encode_init(AVCodecContext *avccontext) {
64 OggVorbisContext *context = avccontext->priv_data ;
bbb77e7c
MN
65 ogg_packet header, header_comm, header_code;
66 uint8_t *p;
ad2b531d 67 unsigned int offset, len;
81e0d0b4 68
81e0d0b4 69 vorbis_info_init(&context->vi) ;
81e0d0b4 70 if(oggvorbis_init_encoder(&context->vi, avccontext) < 0) {
bb270c08
DB
71 av_log(avccontext, AV_LOG_ERROR, "oggvorbis_encode_init: init_encoder failed") ;
72 return -1 ;
81e0d0b4 73 }
81e0d0b4
MH
74 vorbis_analysis_init(&context->vd, &context->vi) ;
75 vorbis_block_init(&context->vd, &context->vb) ;
76
bbb77e7c
MN
77 vorbis_comment_init(&context->vc);
78 vorbis_comment_add_tag(&context->vc, "encoder", LIBAVCODEC_IDENT) ;
79
80 vorbis_analysis_headerout(&context->vd, &context->vc, &header,
81 &header_comm, &header_code);
115329f1 82
ad2b531d
MR
83 len = header.bytes + header_comm.bytes + header_code.bytes;
84 avccontext->extradata_size= 64 + len + len/255;
85 p = avccontext->extradata= av_mallocz(avccontext->extradata_size);
86 p[0] = 2;
87 offset = 1;
88 offset += av_xiphlacing(&p[offset], header.bytes);
89 offset += av_xiphlacing(&p[offset], header_comm.bytes);
90 memcpy(&p[offset], header.packet, header.bytes);
91 offset += header.bytes;
92 memcpy(&p[offset], header_comm.packet, header_comm.bytes);
93 offset += header_comm.bytes;
94 memcpy(&p[offset], header_code.packet, header_code.bytes);
95 offset += header_code.bytes;
96 avccontext->extradata_size = offset;
97 avccontext->extradata= av_realloc(avccontext->extradata, avccontext->extradata_size);
115329f1 98
bbb77e7c
MN
99/* vorbis_block_clear(&context->vb);
100 vorbis_dsp_clear(&context->vd);
101 vorbis_info_clear(&context->vi);*/
102 vorbis_comment_clear(&context->vc);
115329f1 103
81e0d0b4 104 avccontext->frame_size = OGGVORBIS_FRAME_SIZE ;
115329f1 105
492cd3a9
MN
106 avccontext->coded_frame= avcodec_alloc_frame();
107 avccontext->coded_frame->key_frame= 1;
115329f1 108
81e0d0b4
MH
109 return 0 ;
110}
111
112
3aca208a 113static int oggvorbis_encode_frame(AVCodecContext *avccontext,
bb270c08
DB
114 unsigned char *packets,
115 int buf_size, void *data)
81e0d0b4
MH
116{
117 OggVorbisContext *context = avccontext->priv_data ;
118 float **buffer ;
119 ogg_packet op ;
0d00cf38 120 signed short *audio = data ;
6f824977 121 int l, samples = data ? OGGVORBIS_FRAME_SIZE : 0;
81e0d0b4
MH
122
123 buffer = vorbis_analysis_buffer(&context->vd, samples) ;
124
125 if(context->vi.channels == 1) {
bb270c08
DB
126 for(l = 0 ; l < samples ; l++)
127 buffer[0][l]=audio[l]/32768.f;
81e0d0b4 128 } else {
bb270c08
DB
129 for(l = 0 ; l < samples ; l++){
130 buffer[0][l]=audio[l*2]/32768.f;
131 buffer[1][l]=audio[l*2+1]/32768.f;
132 }
81e0d0b4 133 }
115329f1
DB
134
135 vorbis_analysis_wrote(&context->vd, samples) ;
81e0d0b4 136
81e0d0b4 137 while(vorbis_analysis_blockout(&context->vd, &context->vb) == 1) {
bb270c08
DB
138 vorbis_analysis(&context->vb, NULL);
139 vorbis_bitrate_addblock(&context->vb) ;
81e0d0b4 140
bb270c08 141 while(vorbis_bitrate_flushpacket(&context->vd, &op)) {
6f824977
MN
142 if(op.bytes==1) //id love to say this is a hack, bad sadly its not, appearently the end of stream decission is in libogg
143 continue;
6d8f985e
MN
144 memcpy(context->buffer + context->buffer_index, &op, sizeof(ogg_packet));
145 context->buffer_index += sizeof(ogg_packet);
146 memcpy(context->buffer + context->buffer_index, op.packet, op.bytes);
147 context->buffer_index += op.bytes;
148// av_log(avccontext, AV_LOG_DEBUG, "e%d / %d\n", context->buffer_index, op.bytes);
bb270c08 149 }
81e0d0b4
MH
150 }
151
3f4993f1 152 l=0;
6d8f985e 153 if(context->buffer_index){
bbb77e7c 154 ogg_packet *op2= (ogg_packet*)context->buffer;
6d8f985e 155 op2->packet = context->buffer + sizeof(ogg_packet);
3f4993f1 156
6f824977 157 l= op2->bytes;
42661adf 158 avccontext->coded_frame->pts= av_rescale_q(op2->granulepos, (AVRational){1, avccontext->sample_rate}, avccontext->time_base);
3723f01c 159 //FIXME we should reorder the user supplied pts and not assume that they are spaced by 1/sample_rate
3f4993f1 160
6f824977
MN
161 memcpy(packets, op2->packet, l);
162 context->buffer_index -= l + sizeof(ogg_packet);
163 memcpy(context->buffer, context->buffer + l + sizeof(ogg_packet), context->buffer_index);
6d8f985e 164// av_log(avccontext, AV_LOG_DEBUG, "E%d\n", l);
6d8f985e
MN
165 }
166
3f4993f1 167 return l;
81e0d0b4
MH
168}
169
170
3aca208a 171static int oggvorbis_encode_close(AVCodecContext *avccontext) {
81e0d0b4
MH
172 OggVorbisContext *context = avccontext->priv_data ;
173/* ogg_packet op ; */
115329f1 174
81e0d0b4
MH
175 vorbis_analysis_wrote(&context->vd, 0) ; /* notify vorbisenc this is EOF */
176
81e0d0b4
MH
177 vorbis_block_clear(&context->vb);
178 vorbis_dsp_clear(&context->vd);
179 vorbis_info_clear(&context->vi);
492cd3a9
MN
180
181 av_freep(&avccontext->coded_frame);
bbb77e7c 182 av_freep(&avccontext->extradata);
115329f1 183
81e0d0b4
MH
184 return 0 ;
185}
186
187
188AVCodec oggvorbis_encoder = {
189 "vorbis",
190 CODEC_TYPE_AUDIO,
191 CODEC_ID_VORBIS,
192 sizeof(OggVorbisContext),
193 oggvorbis_encode_init,
194 oggvorbis_encode_frame,
6f824977
MN
195 oggvorbis_encode_close,
196 .capabilities= CODEC_CAP_DELAY,
3aca208a
MH
197} ;
198
3aca208a
MH
199static int oggvorbis_decode_init(AVCodecContext *avccontext) {
200 OggVorbisContext *context = avccontext->priv_data ;
a03cbe5f 201 uint8_t *p= avccontext->extradata;
ad2b531d
MR
202 int i, hsizes[3];
203 unsigned char *headers[3], *extradata = avccontext->extradata;
3aca208a
MH
204
205 vorbis_info_init(&context->vi) ;
206 vorbis_comment_init(&context->vc) ;
a03cbe5f 207
ad2b531d
MR
208 if(! avccontext->extradata_size || ! p) {
209 av_log(avccontext, AV_LOG_ERROR, "vorbis extradata absent\n");
210 return -1;
211 }
2c578058
MR
212
213 if(p[0] == 0 && p[1] == 30) {
214 for(i = 0; i < 3; i++){
215 hsizes[i] = *p++ << 8;
216 hsizes[i] += *p++;
217 headers[i] = p;
218 p += hsizes[i];
219 }
220 } else if(*p == 2) {
221 unsigned int offset = 1;
222 p++;
223 for(i=0; i<2; i++) {
224 hsizes[i] = 0;
225 while((*p == 0xFF) && (offset < avccontext->extradata_size)) {
226 hsizes[i] += 0xFF;
227 offset++;
228 p++;
229 }
230 if(offset >= avccontext->extradata_size - 1) {
231 av_log(avccontext, AV_LOG_ERROR,
232 "vorbis header sizes damaged\n");
233 return -1;
234 }
235 hsizes[i] += *p;
ad2b531d
MR
236 offset++;
237 p++;
238 }
2c578058 239 hsizes[2] = avccontext->extradata_size - hsizes[0]-hsizes[1]-offset;
ad2b531d 240#if 0
2c578058
MR
241 av_log(avccontext, AV_LOG_DEBUG,
242 "vorbis header sizes: %d, %d, %d, / extradata_len is %d \n",
243 hsizes[0], hsizes[1], hsizes[2], avccontext->extradata_size);
ad2b531d 244#endif
2c578058
MR
245 headers[0] = extradata + offset;
246 headers[1] = extradata + offset + hsizes[0];
247 headers[2] = extradata + offset + hsizes[0] + hsizes[1];
248 } else {
249 av_log(avccontext, AV_LOG_ERROR,
250 "vorbis initial header len is wrong: %d\n", *p);
251 return -1;
252 }
ad2b531d 253
a03cbe5f
MN
254 for(i=0; i<3; i++){
255 context->op.b_o_s= i==0;
ad2b531d
MR
256 context->op.bytes = hsizes[i];
257 context->op.packet = headers[i];
258 if(vorbis_synthesis_headerin(&context->vi, &context->vc, &context->op)<0){
a03cbe5f
MN
259 av_log(avccontext, AV_LOG_ERROR, "%d. vorbis header damaged\n", i+1);
260 return -1;
261 }
262 }
2c578058 263
a03cbe5f
MN
264 avccontext->channels = context->vi.channels;
265 avccontext->sample_rate = context->vi.rate;
c0df9d75 266 avccontext->time_base= (AVRational){1, avccontext->sample_rate};
a03cbe5f
MN
267
268 vorbis_synthesis_init(&context->vd, &context->vi);
115329f1 269 vorbis_block_init(&context->vd, &context->vb);
3aca208a
MH
270
271 return 0 ;
272}
273
274
275static inline int conv(int samples, float **pcm, char *buf, int channels) {
276 int i, j, val ;
277 ogg_int16_t *ptr, *data = (ogg_int16_t*)buf ;
278 float *mono ;
115329f1 279
3aca208a 280 for(i = 0 ; i < channels ; i++){
bb270c08
DB
281 ptr = &data[i];
282 mono = pcm[i] ;
115329f1 283
bb270c08 284 for(j = 0 ; j < samples ; j++) {
115329f1 285
bb270c08 286 val = mono[j] * 32767.f;
115329f1 287
bb270c08
DB
288 if(val > 32767) val = 32767 ;
289 if(val < -32768) val = -32768 ;
115329f1 290
bb270c08
DB
291 *ptr = val ;
292 ptr += channels;
293 }
3aca208a 294 }
115329f1 295
3aca208a
MH
296 return 0 ;
297}
115329f1
DB
298
299
3aca208a
MH
300static int oggvorbis_decode_frame(AVCodecContext *avccontext,
301 void *data, int *data_size,
0c1a9eda 302 uint8_t *buf, int buf_size)
3aca208a
MH
303{
304 OggVorbisContext *context = avccontext->priv_data ;
3aca208a 305 float **pcm ;
115329f1 306 ogg_packet *op= &context->op;
62ab0b78 307 int samples, total_samples, total_bytes;
115329f1 308
8bfed902
MN
309 if(!buf_size){
310 //FIXME flush
8bfed902
MN
311 return 0;
312 }
115329f1 313
d76f581f
MN
314 op->packet = buf;
315 op->bytes = buf_size;
3aca208a 316
8bfed902 317// av_log(avccontext, AV_LOG_DEBUG, "%d %d %d %lld %lld %d %d\n", op->bytes, op->b_o_s, op->e_o_s, op->granulepos, op->packetno, buf_size, context->vi.rate);
115329f1 318
8bfed902
MN
319/* for(i=0; i<op->bytes; i++)
320 av_log(avccontext, AV_LOG_DEBUG, "%02X ", op->packet[i]);
321 av_log(avccontext, AV_LOG_DEBUG, "\n");*/
3aca208a
MH
322
323 if(vorbis_synthesis(&context->vb, op) == 0)
bb270c08 324 vorbis_synthesis_blockin(&context->vd, &context->vb) ;
115329f1 325
3aca208a
MH
326 total_samples = 0 ;
327 total_bytes = 0 ;
328
329 while((samples = vorbis_synthesis_pcmout(&context->vd, &pcm)) > 0) {
bb270c08
DB
330 conv(samples, pcm, (char*)data + total_bytes, context->vi.channels) ;
331 total_bytes += samples * 2 * context->vi.channels ;
332 total_samples += samples ;
3aca208a
MH
333 vorbis_synthesis_read(&context->vd, samples) ;
334 }
335
115329f1 336 *data_size = total_bytes ;
3aca208a
MH
337 return buf_size ;
338}
81e0d0b4 339
3aca208a
MH
340
341static int oggvorbis_decode_close(AVCodecContext *avccontext) {
342 OggVorbisContext *context = avccontext->priv_data ;
115329f1 343
3aca208a
MH
344 vorbis_info_clear(&context->vi) ;
345 vorbis_comment_clear(&context->vc) ;
346
347 return 0 ;
348}
349
350
351AVCodec oggvorbis_decoder = {
352 "vorbis",
353 CODEC_TYPE_AUDIO,
354 CODEC_ID_VORBIS,
355 sizeof(OggVorbisContext),
356 oggvorbis_decode_init,
357 NULL,
358 oggvorbis_decode_close,
359 oggvorbis_decode_frame,
6f824977 360 .capabilities= CODEC_CAP_DELAY,
3aca208a 361} ;