Mercurial > libavformat.hg
comparison nutenc.c @ 3046:c12367fe98ad libavformat
Header elision muxing support.
This only removes 2 bytes from MP3 and MP2 currently.
Up to 4 could be removed from MP3/MP2 though this might need a 2pass muxer.
Primitive code to remove headers from MPEG-1/2/4 is there too but for the
single file I tried it on (the one in the regression tests), it was a loss
because all video frames were >4096 byte, so that it is disabled ATM.
author | michael |
---|---|
date | Fri, 15 Feb 2008 02:57:19 +0000 |
parents | eac2648449a5 |
children | 3c374f48768d |
comparison
equal
deleted
inserted
replaced
3045:f4c7d2d8acbb | 3046:c12367fe98ad |
---|---|
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
20 */ | 20 */ |
21 | 21 |
22 #include "nut.h" | 22 #include "nut.h" |
23 #include "tree.h" | 23 #include "tree.h" |
24 #include "mpegaudiodata.h" | |
25 | |
26 static int find_expected_header(AVCodecContext *c, int size, int key_frame, uint8_t out[64]){ | |
27 int sample_rate= c->sample_rate; | |
28 | |
29 if(size>4096) | |
30 return 0; | |
31 | |
32 AV_WB24(out, 1); | |
33 | |
34 if(c->codec_id == CODEC_ID_MPEG4){ | |
35 if(key_frame){ | |
36 return 3; | |
37 }else{ | |
38 out[3]= 0xB6; | |
39 return 4; | |
40 } | |
41 }else if(c->codec_id == CODEC_ID_MPEG1VIDEO || c->codec_id == CODEC_ID_MPEG2VIDEO){ | |
42 return 3; | |
43 }else if(c->codec_id == CODEC_ID_H264){ | |
44 return 3; | |
45 }else if(c->codec_id == CODEC_ID_MP3 || c->codec_id == CODEC_ID_MP2){ | |
46 int lsf, mpeg25, sample_rate_index, bitrate_index, frame_size; | |
47 int layer= c->codec_id == CODEC_ID_MP3 ? 3 : 2; | |
48 unsigned int header= 0xFFF00000; | |
49 | |
50 lsf = sample_rate < (24000+32000)/2; | |
51 mpeg25 = sample_rate < (12000+16000)/2; | |
52 sample_rate <<= lsf + mpeg25; | |
53 if (sample_rate < (32000 + 44100)/2) sample_rate_index=2; | |
54 else if(sample_rate < (44100 + 48000)/2) sample_rate_index=0; | |
55 else sample_rate_index=1; | |
56 | |
57 sample_rate= ff_mpa_freq_tab[sample_rate_index] >> (lsf + mpeg25); | |
58 | |
59 for(bitrate_index=2; bitrate_index<30; bitrate_index++){ | |
60 frame_size = ff_mpa_bitrate_tab[lsf][layer-1][bitrate_index>>1]; | |
61 frame_size = (frame_size * 144000) / (sample_rate << lsf) + (bitrate_index&1); | |
62 | |
63 if(frame_size == size) | |
64 break; | |
65 } | |
66 | |
67 header |= (!lsf)<<19; | |
68 header |= (4-layer)<<17; | |
69 header |= 1<<16; //no crc | |
70 AV_WB32(out, header); | |
71 if(size <= 0) | |
72 return 2; //we guess theres no crc, if there is one the user clearly doesnt care about overhead | |
73 if(bitrate_index == 30) | |
74 return -1; //something is wrong ... | |
75 | |
76 header |= (bitrate_index>>1)<<12; | |
77 header |= sample_rate_index<<10; | |
78 header |= (bitrate_index&1)<<9; | |
79 | |
80 return 2; //FIXME actually put the needed ones in build_elision_headers() | |
81 return 3; //we guess that the private bit isnt set | |
82 //FIXME the above asumtations should be checked, if these turn out false too often something should be done | |
83 } | |
84 return 0; | |
85 } | |
86 | |
87 static int find_header_idx(AVFormatContext *s, AVCodecContext *c, int size, int frame_type){ | |
88 NUTContext *nut = s->priv_data; | |
89 uint8_t out[64]; | |
90 int i; | |
91 int len= find_expected_header(c, size, frame_type, out); | |
92 | |
93 //av_log(NULL, AV_LOG_ERROR, "expected_h len=%d size=%d codec_id=%d\n", len, size, c->codec_id); | |
94 | |
95 for(i=1; i<nut->header_count; i++){ | |
96 if( len == nut->header_len[i] | |
97 && !memcmp(out, nut->header[i], len)){ | |
98 // av_log(NULL, AV_LOG_ERROR, "found %d\n", i); | |
99 return i; | |
100 } | |
101 } | |
102 // av_log(NULL, AV_LOG_ERROR, "nothing found\n"); | |
103 return 0; | |
104 } | |
105 | |
106 static void build_elision_headers(AVFormatContext *s){ | |
107 NUTContext *nut = s->priv_data; | |
108 int i; | |
109 //FIXME this is lame | |
110 //FIXME write a 2pass mode to find the maximal headers | |
111 const static uint8_t headers[][5]={ | |
112 {3, 0x00, 0x00, 0x01}, | |
113 {4, 0x00, 0x00, 0x01, 0xB6}, | |
114 {2, 0xFF, 0xFA}, //mp3+crc | |
115 {2, 0xFF, 0xFB}, //mp3 | |
116 {2, 0xFF, 0xFC}, //mp2+crc | |
117 {2, 0xFF, 0xFD}, //mp2 | |
118 }; | |
119 | |
120 nut->header_count= 7; | |
121 for(i=1; i<nut->header_count; i++){ | |
122 nut->header_len[i]= headers[i-1][0]; | |
123 nut->header [i]= &headers[i-1][1]; | |
124 } | |
125 } | |
24 | 126 |
25 static void build_frame_code(AVFormatContext *s){ | 127 static void build_frame_code(AVFormatContext *s){ |
26 NUTContext *nut = s->priv_data; | 128 NUTContext *nut = s->priv_data; |
27 int key_frame, index, pred, stream_id; | 129 int key_frame, index, pred, stream_id; |
28 int start=1; | 130 int start=1; |
61 FrameCode *ft= &nut->frame_code[start2]; | 163 FrameCode *ft= &nut->frame_code[start2]; |
62 ft->flags= FLAG_KEY*key_frame; | 164 ft->flags= FLAG_KEY*key_frame; |
63 ft->flags|= FLAG_SIZE_MSB | FLAG_CODED_PTS; | 165 ft->flags|= FLAG_SIZE_MSB | FLAG_CODED_PTS; |
64 ft->stream_id= stream_id; | 166 ft->stream_id= stream_id; |
65 ft->size_mul=1; | 167 ft->size_mul=1; |
168 if(is_audio) | |
169 ft->header_idx= find_header_idx(s, codec, -1, key_frame); | |
66 start2++; | 170 start2++; |
67 } | 171 } |
68 } | 172 } |
69 | 173 |
70 key_frame= intra_only; | 174 key_frame= intra_only; |
78 ft->flags= FLAG_KEY*key_frame; | 182 ft->flags= FLAG_KEY*key_frame; |
79 ft->stream_id= stream_id; | 183 ft->stream_id= stream_id; |
80 ft->size_mul=frame_bytes + 2; | 184 ft->size_mul=frame_bytes + 2; |
81 ft->size_lsb=frame_bytes + pred; | 185 ft->size_lsb=frame_bytes + pred; |
82 ft->pts_delta=pts; | 186 ft->pts_delta=pts; |
187 ft->header_idx= find_header_idx(s, codec, frame_bytes + pred, key_frame); | |
83 start2++; | 188 start2++; |
84 } | 189 } |
85 } | 190 } |
86 }else{ | 191 }else{ |
87 FrameCode *ft= &nut->frame_code[start2]; | 192 FrameCode *ft= &nut->frame_code[start2]; |
121 ft->stream_id= stream_id; | 226 ft->stream_id= stream_id; |
122 //FIXME use single byte size and pred from last | 227 //FIXME use single byte size and pred from last |
123 ft->size_mul= end3-start3; | 228 ft->size_mul= end3-start3; |
124 ft->size_lsb= index - start3; | 229 ft->size_lsb= index - start3; |
125 ft->pts_delta= pred_table[pred]; | 230 ft->pts_delta= pred_table[pred]; |
231 if(is_audio) | |
232 ft->header_idx= find_header_idx(s, codec, -1, key_frame); | |
126 } | 233 } |
127 } | 234 } |
128 } | 235 } |
129 memmove(&nut->frame_code['N'+1], &nut->frame_code['N'], sizeof(FrameCode)*(255-'N')); | 236 memmove(&nut->frame_code['N'+1], &nut->frame_code['N'], sizeof(FrameCode)*(255-'N')); |
130 nut->frame_code[ 0].flags= | 237 nut->frame_code[ 0].flags= |
210 | 317 |
211 av_free(dyn_buf); | 318 av_free(dyn_buf); |
212 } | 319 } |
213 | 320 |
214 static void write_mainheader(NUTContext *nut, ByteIOContext *bc){ | 321 static void write_mainheader(NUTContext *nut, ByteIOContext *bc){ |
215 int i, j, tmp_pts, tmp_flags, tmp_stream, tmp_mul, tmp_size, tmp_fields; | 322 int i, j, tmp_pts, tmp_flags, tmp_stream, tmp_mul, tmp_size, tmp_fields, tmp_head_idx; |
323 int64_t tmp_match; | |
216 | 324 |
217 put_v(bc, 3); /* version */ | 325 put_v(bc, 3); /* version */ |
218 put_v(bc, nut->avf->nb_streams); | 326 put_v(bc, nut->avf->nb_streams); |
219 put_v(bc, nut->max_distance); | 327 put_v(bc, nut->max_distance); |
220 put_v(bc, nut->time_base_count); | 328 put_v(bc, nut->time_base_count); |
225 } | 333 } |
226 | 334 |
227 tmp_pts=0; | 335 tmp_pts=0; |
228 tmp_mul=1; | 336 tmp_mul=1; |
229 tmp_stream=0; | 337 tmp_stream=0; |
338 tmp_match= 1-(1LL<<62); | |
339 tmp_head_idx= 0; | |
230 for(i=0; i<256;){ | 340 for(i=0; i<256;){ |
231 tmp_fields=0; | 341 tmp_fields=0; |
232 tmp_size=0; | 342 tmp_size=0; |
233 // tmp_res=0; | 343 // tmp_res=0; |
234 if(tmp_pts != nut->frame_code[i].pts_delta) tmp_fields=1; | 344 if(tmp_pts != nut->frame_code[i].pts_delta) tmp_fields=1; |
235 if(tmp_mul != nut->frame_code[i].size_mul ) tmp_fields=2; | 345 if(tmp_mul != nut->frame_code[i].size_mul ) tmp_fields=2; |
236 if(tmp_stream != nut->frame_code[i].stream_id) tmp_fields=3; | 346 if(tmp_stream != nut->frame_code[i].stream_id) tmp_fields=3; |
237 if(tmp_size != nut->frame_code[i].size_lsb ) tmp_fields=4; | 347 if(tmp_size != nut->frame_code[i].size_lsb ) tmp_fields=4; |
238 // if(tmp_res != nut->frame_code[i].res ) tmp_fields=5; | 348 // if(tmp_res != nut->frame_code[i].res ) tmp_fields=5; |
349 if(tmp_head_idx!=nut->frame_code[i].header_idx)tmp_fields=8; | |
239 | 350 |
240 tmp_pts = nut->frame_code[i].pts_delta; | 351 tmp_pts = nut->frame_code[i].pts_delta; |
241 tmp_flags = nut->frame_code[i].flags; | 352 tmp_flags = nut->frame_code[i].flags; |
242 tmp_stream= nut->frame_code[i].stream_id; | 353 tmp_stream= nut->frame_code[i].stream_id; |
243 tmp_mul = nut->frame_code[i].size_mul; | 354 tmp_mul = nut->frame_code[i].size_mul; |
244 tmp_size = nut->frame_code[i].size_lsb; | 355 tmp_size = nut->frame_code[i].size_lsb; |
245 // tmp_res = nut->frame_code[i].res; | 356 // tmp_res = nut->frame_code[i].res; |
357 tmp_head_idx= nut->frame_code[i].header_idx; | |
246 | 358 |
247 for(j=0; i<256; j++,i++){ | 359 for(j=0; i<256; j++,i++){ |
248 if(i == 'N'){ | 360 if(i == 'N'){ |
249 j--; | 361 j--; |
250 continue; | 362 continue; |
253 if(nut->frame_code[i].flags != tmp_flags ) break; | 365 if(nut->frame_code[i].flags != tmp_flags ) break; |
254 if(nut->frame_code[i].stream_id != tmp_stream) break; | 366 if(nut->frame_code[i].stream_id != tmp_stream) break; |
255 if(nut->frame_code[i].size_mul != tmp_mul ) break; | 367 if(nut->frame_code[i].size_mul != tmp_mul ) break; |
256 if(nut->frame_code[i].size_lsb != tmp_size+j) break; | 368 if(nut->frame_code[i].size_lsb != tmp_size+j) break; |
257 // if(nut->frame_code[i].res != tmp_res ) break; | 369 // if(nut->frame_code[i].res != tmp_res ) break; |
370 if(nut->frame_code[i].header_idx!= tmp_head_idx) break; | |
258 } | 371 } |
259 if(j != tmp_mul - tmp_size) tmp_fields=6; | 372 if(j != tmp_mul - tmp_size) tmp_fields=6; |
260 | 373 |
261 put_v(bc, tmp_flags); | 374 put_v(bc, tmp_flags); |
262 put_v(bc, tmp_fields); | 375 put_v(bc, tmp_fields); |
264 if(tmp_fields>1) put_v(bc, tmp_mul); | 377 if(tmp_fields>1) put_v(bc, tmp_mul); |
265 if(tmp_fields>2) put_v(bc, tmp_stream); | 378 if(tmp_fields>2) put_v(bc, tmp_stream); |
266 if(tmp_fields>3) put_v(bc, tmp_size); | 379 if(tmp_fields>3) put_v(bc, tmp_size); |
267 if(tmp_fields>4) put_v(bc, 0 /*tmp_res*/); | 380 if(tmp_fields>4) put_v(bc, 0 /*tmp_res*/); |
268 if(tmp_fields>5) put_v(bc, j); | 381 if(tmp_fields>5) put_v(bc, j); |
382 if(tmp_fields>6) put_v(bc, tmp_match); | |
383 if(tmp_fields>7) put_v(bc, tmp_head_idx); | |
384 } | |
385 put_v(bc, nut->header_count-1); | |
386 for(i=1; i<nut->header_count; i++){ | |
387 put_v(bc, nut->header_len[i]); | |
388 put_buffer(bc, nut->header[i], nut->header_len[i]); | |
269 } | 389 } |
270 } | 390 } |
271 | 391 |
272 static int write_streamheader(NUTContext *nut, ByteIOContext *bc, AVCodecContext *codec, int i){ | 392 static int write_streamheader(NUTContext *nut, ByteIOContext *bc, AVCodecContext *codec, int i){ |
273 put_v(bc, i); | 393 put_v(bc, i); |
417 nut->stream[i].msb_pts_shift = 14; | 537 nut->stream[i].msb_pts_shift = 14; |
418 nut->stream[i].max_pts_distance= FFMAX(1/av_q2d(time_base), 1); | 538 nut->stream[i].max_pts_distance= FFMAX(1/av_q2d(time_base), 1); |
419 } | 539 } |
420 | 540 |
421 nut->max_distance = MAX_DISTANCE; | 541 nut->max_distance = MAX_DISTANCE; |
542 build_elision_headers(s); | |
422 build_frame_code(s); | 543 build_frame_code(s); |
423 assert(nut->frame_code['N'].flags == FLAG_INVALID); | 544 assert(nut->frame_code['N'].flags == FLAG_INVALID); |
424 | 545 |
425 put_buffer(bc, ID_STRING, strlen(ID_STRING)); | 546 put_buffer(bc, ID_STRING, strlen(ID_STRING)); |
426 put_byte(bc, 0); | 547 put_byte(bc, 0); |
442 if(pkt->size / fc->size_mul ) flags |= FLAG_SIZE_MSB; | 563 if(pkt->size / fc->size_mul ) flags |= FLAG_SIZE_MSB; |
443 if(pkt->pts - nus->last_pts != fc->pts_delta) flags |= FLAG_CODED_PTS; | 564 if(pkt->pts - nus->last_pts != fc->pts_delta) flags |= FLAG_CODED_PTS; |
444 if(pkt->size > 2*nut->max_distance ) flags |= FLAG_CHECKSUM; | 565 if(pkt->size > 2*nut->max_distance ) flags |= FLAG_CHECKSUM; |
445 if(FFABS(pkt->pts - nus->last_pts) | 566 if(FFABS(pkt->pts - nus->last_pts) |
446 > nus->max_pts_distance) flags |= FLAG_CHECKSUM; | 567 > nus->max_pts_distance) flags |= FLAG_CHECKSUM; |
568 if( pkt->size < nut->header_len[fc->header_idx] | |
569 || (pkt->size > 4096 && fc->header_idx) | |
570 || memcmp(pkt->data, nut->header[fc->header_idx], nut->header_len[fc->header_idx])) | |
571 flags |= FLAG_HEADER_IDX; | |
447 | 572 |
448 return flags | (fc->flags & FLAG_CODED); | 573 return flags | (fc->flags & FLAG_CODED); |
574 } | |
575 | |
576 static int find_best_header_idx(NUTContext *nut, AVPacket *pkt){ | |
577 int i; | |
578 int best_i = 0; | |
579 int best_len= 0; | |
580 | |
581 if(pkt->size > 4096) | |
582 return 0; | |
583 | |
584 for(i=1; i<nut->header_count; i++){ | |
585 if( pkt->size >= nut->header_len[i] | |
586 && nut->header_len[i] > best_len | |
587 && !memcmp(pkt->data, nut->header[i], nut->header_len[i])){ | |
588 best_i= i; | |
589 best_len= nut->header_len[i]; | |
590 } | |
591 } | |
592 return best_i; | |
449 } | 593 } |
450 | 594 |
451 static int write_packet(AVFormatContext *s, AVPacket *pkt){ | 595 static int write_packet(AVFormatContext *s, AVPacket *pkt){ |
452 NUTContext *nut = s->priv_data; | 596 NUTContext *nut = s->priv_data; |
453 StreamContext *nus= &nut->stream[pkt->stream_index]; | 597 StreamContext *nus= &nut->stream[pkt->stream_index]; |
454 ByteIOContext *bc = s->pb, *dyn_bc; | 598 ByteIOContext *bc = s->pb, *dyn_bc; |
455 FrameCode *fc; | 599 FrameCode *fc; |
456 int64_t coded_pts; | 600 int64_t coded_pts; |
457 int best_length, frame_code, flags, needed_flags, i; | 601 int best_length, frame_code, flags, needed_flags, i, header_idx, best_header_idx; |
458 int key_frame = !!(pkt->flags & PKT_FLAG_KEY); | 602 int key_frame = !!(pkt->flags & PKT_FLAG_KEY); |
459 int store_sp=0; | 603 int store_sp=0; |
460 int ret; | 604 int ret; |
461 | 605 |
462 if(1LL<<(20+3*nut->header_count) <= url_ftell(bc)) | 606 if(1LL<<(20+3*nut->header_count) <= url_ftell(bc)) |
501 | 645 |
502 coded_pts = pkt->pts & ((1<<nus->msb_pts_shift)-1); | 646 coded_pts = pkt->pts & ((1<<nus->msb_pts_shift)-1); |
503 if(ff_lsb2full(nus, coded_pts) != pkt->pts) | 647 if(ff_lsb2full(nus, coded_pts) != pkt->pts) |
504 coded_pts= pkt->pts + (1<<nus->msb_pts_shift); | 648 coded_pts= pkt->pts + (1<<nus->msb_pts_shift); |
505 | 649 |
650 best_header_idx= find_best_header_idx(nut, pkt); | |
651 | |
506 best_length=INT_MAX; | 652 best_length=INT_MAX; |
507 frame_code= -1; | 653 frame_code= -1; |
508 for(i=0; i<256; i++){ | 654 for(i=0; i<256; i++){ |
509 int length= 0; | 655 int length= 0; |
510 FrameCode *fc= &nut->frame_code[i]; | 656 FrameCode *fc= &nut->frame_code[i]; |
537 length+=4; | 683 length+=4; |
538 | 684 |
539 if(flags & FLAG_CODED_PTS) | 685 if(flags & FLAG_CODED_PTS) |
540 length += get_length(coded_pts); | 686 length += get_length(coded_pts); |
541 | 687 |
688 if( (flags & FLAG_CODED) | |
689 && nut->header_len[best_header_idx] > nut->header_len[fc->header_idx]+1){ | |
690 flags |= FLAG_HEADER_IDX; | |
691 } | |
692 | |
693 if(flags & FLAG_HEADER_IDX){ | |
694 length += 1 - nut->header_len[best_header_idx]; | |
695 }else{ | |
696 length -= nut->header_len[fc->header_idx]; | |
697 } | |
698 | |
542 length*=4; | 699 length*=4; |
543 length+= !(flags & FLAG_CODED_PTS); | 700 length+= !(flags & FLAG_CODED_PTS); |
544 length+= !(flags & FLAG_CHECKSUM); | 701 length+= !(flags & FLAG_CHECKSUM); |
545 | 702 |
546 if(length < best_length){ | 703 if(length < best_length){ |
550 } | 707 } |
551 assert(frame_code != -1); | 708 assert(frame_code != -1); |
552 fc= &nut->frame_code[frame_code]; | 709 fc= &nut->frame_code[frame_code]; |
553 flags= fc->flags; | 710 flags= fc->flags; |
554 needed_flags= get_needed_flags(nut, nus, fc, pkt); | 711 needed_flags= get_needed_flags(nut, nus, fc, pkt); |
712 header_idx= fc->header_idx; | |
555 | 713 |
556 init_checksum(bc, ff_crc04C11DB7_update, 0); | 714 init_checksum(bc, ff_crc04C11DB7_update, 0); |
557 put_byte(bc, frame_code); | 715 put_byte(bc, frame_code); |
558 if(flags & FLAG_CODED){ | 716 if(flags & FLAG_CODED){ |
559 put_v(bc, (flags^needed_flags) & ~(FLAG_CODED)); | 717 put_v(bc, (flags^needed_flags) & ~(FLAG_CODED)); |
560 flags = needed_flags; | 718 flags = needed_flags; |
561 } | 719 } |
562 if(flags & FLAG_STREAM_ID) put_v(bc, pkt->stream_index); | 720 if(flags & FLAG_STREAM_ID) put_v(bc, pkt->stream_index); |
563 if(flags & FLAG_CODED_PTS) put_v(bc, coded_pts); | 721 if(flags & FLAG_CODED_PTS) put_v(bc, coded_pts); |
564 if(flags & FLAG_SIZE_MSB) put_v(bc, pkt->size / fc->size_mul); | 722 if(flags & FLAG_SIZE_MSB) put_v(bc, pkt->size / fc->size_mul); |
723 if(flags & FLAG_HEADER_IDX) put_v(bc, header_idx= best_header_idx); | |
565 | 724 |
566 if(flags & FLAG_CHECKSUM) put_le32(bc, get_checksum(bc)); | 725 if(flags & FLAG_CHECKSUM) put_le32(bc, get_checksum(bc)); |
567 else get_checksum(bc); | 726 else get_checksum(bc); |
568 | 727 |
569 put_buffer(bc, pkt->data, pkt->size); | 728 put_buffer(bc, pkt->data + nut->header_len[header_idx], pkt->size - nut->header_len[header_idx]); |
570 nus->last_flags= flags; | 729 nus->last_flags= flags; |
571 | 730 |
572 //FIXME just store one per syncpoint | 731 //FIXME just store one per syncpoint |
573 if(flags & FLAG_KEY) | 732 if(flags & FLAG_KEY) |
574 av_add_index_entry( | 733 av_add_index_entry( |