fix decoding of multi-channel vorbis streams
[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) {
c55427f8
MH
32
33#ifdef OGGVORBIS_VBR_BY_ESTIMATE
34 /* variable bitrate by estimate */
35
36 return (vorbis_encode_setup_managed(vi, avccontext->channels,
37 avccontext->sample_rate, -1, avccontext->bit_rate, -1) ||
38 vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE_AVG, NULL) ||
39 vorbis_encode_setup_init(vi)) ;
40#else
41 /* constant bitrate */
81e0d0b4
MH
42
43 return vorbis_encode_init(vi, avccontext->channels,
44 avccontext->sample_rate, -1, avccontext->bit_rate, -1) ;
c55427f8 45#endif
81e0d0b4
MH
46}
47
81e0d0b4
MH
48static int oggvorbis_encode_init(AVCodecContext *avccontext) {
49 OggVorbisContext *context = avccontext->priv_data ;
bbb77e7c
MN
50 ogg_packet header, header_comm, header_code;
51 uint8_t *p;
ad2b531d 52 unsigned int offset, len;
81e0d0b4 53
81e0d0b4 54 vorbis_info_init(&context->vi) ;
81e0d0b4 55 if(oggvorbis_init_encoder(&context->vi, avccontext) < 0) {
9b879566 56 av_log(avccontext, AV_LOG_ERROR, "oggvorbis_encode_init: init_encoder failed") ;
81e0d0b4
MH
57 return -1 ;
58 }
81e0d0b4
MH
59 vorbis_analysis_init(&context->vd, &context->vi) ;
60 vorbis_block_init(&context->vd, &context->vb) ;
61
bbb77e7c
MN
62 vorbis_comment_init(&context->vc);
63 vorbis_comment_add_tag(&context->vc, "encoder", LIBAVCODEC_IDENT) ;
64
65 vorbis_analysis_headerout(&context->vd, &context->vc, &header,
66 &header_comm, &header_code);
67
ad2b531d
MR
68 len = header.bytes + header_comm.bytes + header_code.bytes;
69 avccontext->extradata_size= 64 + len + len/255;
70 p = avccontext->extradata= av_mallocz(avccontext->extradata_size);
71 p[0] = 2;
72 offset = 1;
73 offset += av_xiphlacing(&p[offset], header.bytes);
74 offset += av_xiphlacing(&p[offset], header_comm.bytes);
75 memcpy(&p[offset], header.packet, header.bytes);
76 offset += header.bytes;
77 memcpy(&p[offset], header_comm.packet, header_comm.bytes);
78 offset += header_comm.bytes;
79 memcpy(&p[offset], header_code.packet, header_code.bytes);
80 offset += header_code.bytes;
81 avccontext->extradata_size = offset;
82 avccontext->extradata= av_realloc(avccontext->extradata, avccontext->extradata_size);
bbb77e7c
MN
83
84/* vorbis_block_clear(&context->vb);
85 vorbis_dsp_clear(&context->vd);
86 vorbis_info_clear(&context->vi);*/
87 vorbis_comment_clear(&context->vc);
88
81e0d0b4 89 avccontext->frame_size = OGGVORBIS_FRAME_SIZE ;
492cd3a9
MN
90
91 avccontext->coded_frame= avcodec_alloc_frame();
92 avccontext->coded_frame->key_frame= 1;
81e0d0b4
MH
93
94 return 0 ;
95}
96
97
3aca208a
MH
98static int oggvorbis_encode_frame(AVCodecContext *avccontext,
99 unsigned char *packets,
81e0d0b4
MH
100 int buf_size, void *data)
101{
102 OggVorbisContext *context = avccontext->priv_data ;
103 float **buffer ;
104 ogg_packet op ;
0d00cf38 105 signed short *audio = data ;
6f824977 106 int l, samples = data ? OGGVORBIS_FRAME_SIZE : 0;
81e0d0b4
MH
107
108 buffer = vorbis_analysis_buffer(&context->vd, samples) ;
109
110 if(context->vi.channels == 1) {
111 for(l = 0 ; l < samples ; l++)
0d00cf38 112 buffer[0][l]=audio[l]/32768.f;
81e0d0b4
MH
113 } else {
114 for(l = 0 ; l < samples ; l++){
0d00cf38
SS
115 buffer[0][l]=audio[l*2]/32768.f;
116 buffer[1][l]=audio[l*2+1]/32768.f;
81e0d0b4
MH
117 }
118 }
119
120 vorbis_analysis_wrote(&context->vd, samples) ;
121
81e0d0b4
MH
122 while(vorbis_analysis_blockout(&context->vd, &context->vb) == 1) {
123 vorbis_analysis(&context->vb, NULL);
124 vorbis_bitrate_addblock(&context->vb) ;
125
126 while(vorbis_bitrate_flushpacket(&context->vd, &op)) {
6f824977
MN
127 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
128 continue;
6d8f985e
MN
129 memcpy(context->buffer + context->buffer_index, &op, sizeof(ogg_packet));
130 context->buffer_index += sizeof(ogg_packet);
131 memcpy(context->buffer + context->buffer_index, op.packet, op.bytes);
132 context->buffer_index += op.bytes;
133// av_log(avccontext, AV_LOG_DEBUG, "e%d / %d\n", context->buffer_index, op.bytes);
81e0d0b4
MH
134 }
135 }
136
3f4993f1 137 l=0;
6d8f985e 138 if(context->buffer_index){
bbb77e7c 139 ogg_packet *op2= (ogg_packet*)context->buffer;
6d8f985e 140 op2->packet = context->buffer + sizeof(ogg_packet);
3f4993f1 141
6f824977 142 l= op2->bytes;
c0df9d75 143 avccontext->coded_frame->pts= op2->granulepos;
3f4993f1 144
6f824977
MN
145 memcpy(packets, op2->packet, l);
146 context->buffer_index -= l + sizeof(ogg_packet);
147 memcpy(context->buffer, context->buffer + l + sizeof(ogg_packet), context->buffer_index);
6d8f985e 148// av_log(avccontext, AV_LOG_DEBUG, "E%d\n", l);
6d8f985e
MN
149 }
150
3f4993f1 151 return l;
81e0d0b4
MH
152}
153
154
3aca208a 155static int oggvorbis_encode_close(AVCodecContext *avccontext) {
81e0d0b4
MH
156 OggVorbisContext *context = avccontext->priv_data ;
157/* ogg_packet op ; */
158
81e0d0b4
MH
159 vorbis_analysis_wrote(&context->vd, 0) ; /* notify vorbisenc this is EOF */
160
81e0d0b4
MH
161 vorbis_block_clear(&context->vb);
162 vorbis_dsp_clear(&context->vd);
163 vorbis_info_clear(&context->vi);
492cd3a9
MN
164
165 av_freep(&avccontext->coded_frame);
bbb77e7c 166 av_freep(&avccontext->extradata);
81e0d0b4
MH
167
168 return 0 ;
169}
170
171
172AVCodec oggvorbis_encoder = {
173 "vorbis",
174 CODEC_TYPE_AUDIO,
175 CODEC_ID_VORBIS,
176 sizeof(OggVorbisContext),
177 oggvorbis_encode_init,
178 oggvorbis_encode_frame,
6f824977
MN
179 oggvorbis_encode_close,
180 .capabilities= CODEC_CAP_DELAY,
3aca208a
MH
181} ;
182
183
184static int oggvorbis_decode_init(AVCodecContext *avccontext) {
185 OggVorbisContext *context = avccontext->priv_data ;
a03cbe5f 186 uint8_t *p= avccontext->extradata;
ad2b531d
MR
187 int i, hsizes[3];
188 unsigned char *headers[3], *extradata = avccontext->extradata;
189 unsigned int offset;
3aca208a
MH
190
191 vorbis_info_init(&context->vi) ;
192 vorbis_comment_init(&context->vc) ;
a03cbe5f 193
ad2b531d
MR
194 if(! avccontext->extradata_size || ! p) {
195 av_log(avccontext, AV_LOG_ERROR, "vorbis extradata absent\n");
196 return -1;
197 }
198 if(*p != 2) {
199 av_log(avccontext, AV_LOG_ERROR,
200 "vorbis initial header len is wrong: %d\n", *p);
201 return -1;
202 }
203 offset = 1;
204 p++;
205 for(i=0; i<2; i++) {
206 hsizes[i] = 0;
207 while((*p == 0xFF) && (offset < avccontext->extradata_size)) {
208 hsizes[i] += 0xFF;
209 offset++;
210 p++;
211 }
212 if(offset >= avccontext->extradata_size - 1) {
213 av_log(avccontext, AV_LOG_ERROR, "vorbis header sizes damaged\n");
214 return -1;
215 }
216 hsizes[i] += *p;
217 offset++;
218 p++;
219 }
220 hsizes[2] = avccontext->extradata_size - hsizes[0] - hsizes[1] - offset;
221#if 0
222 av_log(avccontext, AV_LOG_DEBUG,
223 "vorbis header sizes: %d, %d, %d, / extradata_len is %d \n",
224 hsizes[0], hsizes[1], hsizes[2], avccontext->extradata_size);
225#endif
226 headers[0] = extradata + offset;
227 headers[1] = extradata + offset + hsizes[0];
228 headers[2] = extradata + offset + hsizes[0] + hsizes[1];
229
a03cbe5f
MN
230 for(i=0; i<3; i++){
231 context->op.b_o_s= i==0;
ad2b531d
MR
232 context->op.bytes = hsizes[i];
233 context->op.packet = headers[i];
234 if(vorbis_synthesis_headerin(&context->vi, &context->vc, &context->op)<0){
a03cbe5f
MN
235 av_log(avccontext, AV_LOG_ERROR, "%d. vorbis header damaged\n", i+1);
236 return -1;
237 }
238 }
239 avccontext->channels = context->vi.channels;
240 avccontext->sample_rate = context->vi.rate;
c0df9d75 241 avccontext->time_base= (AVRational){1, avccontext->sample_rate};
a03cbe5f
MN
242
243 vorbis_synthesis_init(&context->vd, &context->vi);
244 vorbis_block_init(&context->vd, &context->vb);
3aca208a
MH
245
246 return 0 ;
247}
248
249
250static inline int conv(int samples, float **pcm, char *buf, int channels) {
251 int i, j, val ;
252 ogg_int16_t *ptr, *data = (ogg_int16_t*)buf ;
253 float *mono ;
254
255 for(i = 0 ; i < channels ; i++){
256 ptr = &data[i];
257 mono = pcm[i] ;
258
259 for(j = 0 ; j < samples ; j++) {
260
261 val = mono[j] * 32767.f;
262
263 if(val > 32767) val = 32767 ;
264 if(val < -32768) val = -32768 ;
265
266 *ptr = val ;
267 ptr += channels;
268 }
269 }
270
271 return 0 ;
272}
273
274
275static int oggvorbis_decode_frame(AVCodecContext *avccontext,
276 void *data, int *data_size,
0c1a9eda 277 uint8_t *buf, int buf_size)
3aca208a
MH
278{
279 OggVorbisContext *context = avccontext->priv_data ;
3aca208a 280 float **pcm ;
d76f581f 281 ogg_packet *op= &context->op;
8bfed902 282 int samples, total_samples, total_bytes,i;
3aca208a 283
8bfed902
MN
284 if(!buf_size){
285 //FIXME flush
8bfed902
MN
286 return 0;
287 }
288
d76f581f
MN
289 op->packet = buf;
290 op->bytes = buf_size;
3aca208a 291
8bfed902
MN
292// 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);
293
294/* for(i=0; i<op->bytes; i++)
295 av_log(avccontext, AV_LOG_DEBUG, "%02X ", op->packet[i]);
296 av_log(avccontext, AV_LOG_DEBUG, "\n");*/
3aca208a
MH
297
298 if(vorbis_synthesis(&context->vb, op) == 0)
299 vorbis_synthesis_blockin(&context->vd, &context->vb) ;
300
301 total_samples = 0 ;
302 total_bytes = 0 ;
303
304 while((samples = vorbis_synthesis_pcmout(&context->vd, &pcm)) > 0) {
305 conv(samples, pcm, (char*)data + total_bytes, context->vi.channels) ;
306 total_bytes += samples * 2 * context->vi.channels ;
307 total_samples += samples ;
308 vorbis_synthesis_read(&context->vd, samples) ;
309 }
310
311 *data_size = total_bytes ;
312 return buf_size ;
313}
81e0d0b4 314
3aca208a
MH
315
316static int oggvorbis_decode_close(AVCodecContext *avccontext) {
317 OggVorbisContext *context = avccontext->priv_data ;
318
319 vorbis_info_clear(&context->vi) ;
320 vorbis_comment_clear(&context->vc) ;
321
322 return 0 ;
323}
324
325
326AVCodec oggvorbis_decoder = {
327 "vorbis",
328 CODEC_TYPE_AUDIO,
329 CODEC_ID_VORBIS,
330 sizeof(OggVorbisContext),
331 oggvorbis_decode_init,
332 NULL,
333 oggvorbis_decode_close,
334 oggvorbis_decode_frame,
6f824977 335 .capabilities= CODEC_CAP_DELAY,
3aca208a 336} ;