CVS change done by Michael Niedermayer CVS
Update of /cvsroot/mplayer/main/DOCS/tech
In directory mail:/var2/tmp/cvs-serv11727
Modified Files:
mpcf.txt
Log Message:
removing subpackets (everyone hates them ...)
removing shuffle_type (meaningless without subpackets)
making timestamp_lsb a simple unsigned value (someone proposed that a long time ago, dunno dont remember who, IIRC it was rejected as it more often required the timestamp_msb to be coded but by defining lsb relative to the last lsb we need very few msbs in the error free case and for damaged files its also pretty difficult to trash the timestamp, for example for a fixed fps stream with 7bit lsb_timestamps we need to loose >64 frames in a row to end up with a wrong timestamp)
cleanup (filesize and such where only partially removed)
frame_code byte, with the meaning of each value stored in the main header
the frame_code contains the keyframe_flag, packet_type and can contain the timestamp_delta, stream_id and the data_size or part if it
Index: mpcf.txt
===================================================================
RCS file: /cvsroot/mplayer/main/DOCS/tech/mpcf.txt,v
retrieving revision 1.36
retrieving revision 1.37
diff -u -r1.36 -r1.37
--- mpcf.txt 18 Feb 2004 13:33:57 -0000 1.36
+++ mpcf.txt 30 Mar 2004 01:05:51 -0000 1.37
@@ -1,4 +1,4 @@
- NUT Open Container Format DRAFT 20030906
+ NUT Open Container Format DRAFT 20040330
----------------------------------------
@@ -19,11 +19,10 @@
~0.2% overhead, for normal bitrates
index is <10kb per hour (1 keyframe every 3sec)
a usual header for a file is about 100bytes (audio + video headers together)
- a packet header is about ~8 bytes
+ a packet header is about ~1-8 bytes
Error resistant
seeking / playback without an index
headers & index can be repeated
- audio packet reshuffle
checksums to allow quick redownloading of damaged parts
damaged files can be played back with minimal data lost and fast
resyncing times
@@ -68,8 +67,8 @@
Bitstream syntax:
packet header
- forward ptr v
backward ptr v
+ forward ptr v
align_byte
while(not byte aligned)
@@ -88,6 +87,15 @@
packet header
version v
stream_count v
+ checksum_threshold v
+ for(i=0; i<256; i++){
+ flags[i] v
+ if(flags&64){
+ stream_id[i] v
+ lsb_size[i] v
+ data_size_mul[i] v
+ }
+ }
reserved_bytes
checksum u(32)
@@ -102,7 +110,8 @@
time_base_nom v
time_base_denom v
msb_timestamp_shift v
- shuffle_type v
+ inital_timestamp_predictor v(3)
+ initial_data_size_predictor v(2)
fixed_fps u(1)
index_flag u(1)
reserved u(6)
@@ -130,44 +139,31 @@
reserved_bytes
checksum u(32)
+
frame
- if(keyframe){
- keyframe_startcode f(64)
+ if(frame_type == 2){
+ frame_type2_startcode f(64)
}
- zero_bit f(1)
- priority u(2)
- checksum_flag u(1)
- msb_timestamp_flag u(1)
- subpacket_type u(2)
- reserved u(1)
- packet header
- stream_id v
- if(msb_timestamp_flag)
+ frame_code f(8)
+ if(flags[frame_code]&1){
+ packet header
+ }
+ if(stream_id[frame_code]==stream_count){
+ stream_id v
+ }
+ if(frame_type == 2){
msb_timestamp v
- lsb_timestamp s
- if(sub_packet_type==01)
- sub_packet[0]
- else{
- subpacket_count v
- if(sub_packet_type&01)
- for(i=0; ; i++)
- keyframe_run[i] v
- for(i=0; ; i++){
- timestamp_diff[i] v
- if(timestamp_diff[i] & 1)
- timestamp_diff_run[i] v
- }
- if(sub_packet_type&10){
- for(i=0; i<subpacket_count-1; i++)
- subpacket_size_diff[i] s
- }
- for(i=0; i<subpacket_count; i++)
- subpacket[i]
- }
}
- if(checksum_flag)
+ if((flags[frame_code]&12) == 12){
+ lsb_timestamp v
+ }
+ if(flags[frame_code]&2){
+ data_size_msb v
+ }
+ data
+ if(checksum_threshold < frame_type)
frame_checksum u(32)
-
+
Index:
index_startcode f(64)
packet header
@@ -230,13 +226,7 @@
version
0 for now
-
-file_size
- size in bytes. 0 invalidates the field
-
-length_in_msec
- length of the file in milli seconds (0 invalidates the field)
-
+
stream_id
Note: streams with a lower relative class MUST have a lower relative id
so a stream with class 0 MUST allways have a id which is lower then any
@@ -305,56 +295,76 @@
codec_specific_data
private global data for a codec (could be huffman tables or ...)
-
-msb_timestamp_flag
- indicates that the msb_timestamp is coded
- MUST be 1 for keyframes
-
-subpacket_type
- subpacket_count subpacket_size keyframe_flag
- 00 >1 constant constant
- 01 1 NA NA
- 10 >1 variable constant
- 11 >1 variable variable
- the only legal subpacket_type for video streams is 01, so video streams
- MUST NOT contain multiple subpackets per packet
- Note, if there are multiple subpackets then the timestamp of the packet
- is the timestamp of the first subpacket
- Note, if multiple subpackets are stored in one frame then the resulting
- framesize SHOULD be < 16kbyte and not more then 0.5 sec of data SHOULD
- be put in a single packet
-
-subpacket_count
- the number of subpackets, if not pressent then 1
-
-keyframe_run[i]
- the number of subpackets with an identical keyframe_flag
- Note, the value of first flag is stored in the packet header
- is equal to subpacket count if not coded
-
-timestamp_diff[i]
- the difference from the last subpacket timestamp to the current one in
- time_base precission
- the lowwest bit is used to indicate if timestamp_diff_run[i] is coded
-
-timestamp_diff_run[i]
- the number of subpackets which have the same timestamp_diff
- if not coded than 1
- 0 is forbidden
-
-subpacket_size_diff
- the difference between the predicted size of this subpacket and the
- actual size
- the predicted size is 64 for the first subpacket in a packet
- otherwise it is MAX(64, last_subpacket_size)
- the size of the last subpacket is not coded and is simply the space left
- Note a subpacket MUST be in exactly one packet, it cannot be split
-
+
+frame_code
+ the meaning of this byte is stored in the main header
+ the value 78 ('N') is forbidden to ensure that the byte is always
+ different from the first byte of any startcode
+
+flags[frame_code]
+ the bits of the flags from MSB to LSB are CKKTTDP
+ P is 1 for type 1 and 2 packets, 0 for type 0 packets
+ TT is the timestamp_code 00,01,10 use the last timestamp + the first,
+ second and third last unique timestamp difference, so if the
+ timestamp differences, are +3,+1,+2,+2,+1 then last diff is
+ +1, second is +2 and third is +3
+ if TT is 11, then the timestamp is calculated by
+ mask = (1<<msb_timestamp_shift)-1;
+ delta= last_timestamp - mask/2
+ timestamp= ((timestamp_lsb-delta)&mask) + delta
+ TT must be 11 if packet_type is not 0
+ the last timestamp differences are reset to the default values
+ from the stream header if a packet of type not 0 in encountered
+ if D is 1 then the data_size_msb is coded otherwise data_size_msb is 0
+ KK is the keyframe_type
+ 00-> no keyframe,
+ 01-> keyframe,
+ 10-> equal to last of the same stream,
+ 11-> opposite from last of the same stream
+ KK must be 00 or 01 if the packet_type is not 0
+ if C is 1 then stream_id, data_size_mul and data_size_lsb are not
+ stored, but predicted from the last ones
+ the value 1000001 (65) is used to mark illegal frame_code bytes, at
+ least flags[78] must be 65
+
+frame_type
+ 0 is indicated by (flags[frame_code]&1)==0
+ 1 is indicated by (flags[frame_code]&1)==1 && !startcode
+ 2 is indicated by (flags[frame_code]&1)==1 && startcode
+ there SHOULD not be more then 0.5 seconds or 16kbyte of type 0 frames
+ wihout a intervening frame of different frame_type
+
+stream_id[frame_code]
+ if its not coded in the main_header then its equal to the last one
+ from the main header
+ must be <250
+
+data_size_mul[frame_code]
+ if its not coded in the main_header then its equal to the last one
+ from the main header
+ must be <250
+
+data_size_lsb[frame_code]
+ if its not coded in the main_header then its equal to the last one
+ from the main header + 1
+ must be <250
+
+data_size
+ if(data_size_lsb == data_size_mul)
+ data_size= last;
+ else if(data_size_lsb == data_size_mul+1)
+ data_size= next last;
+ else if(data_size_lsb < data_size_mul)
+ data_size= data_size_lsb + data_size_msb*data_size_mul;
+ else reserved
+ last and next last are reset to the values stored in the stream header
+ if an frame with type > 0 is encountered
+
msb_timestamp
most significant bits of the timestamp, SHOULD be 0 for the first frame
lsb_timestamp
- difference from the msb_timestamp in time_base precission
+ least significant bits of the timestamp in time_base precission
Example: IBBP display order
keyframe msb_timestamp=0 lsb_timestamp=0 -> timestamp=0
frame lsb_timestamp=3 -> timestamp=3
@@ -362,7 +372,7 @@
frame lsb_timestamp=2 -> timestamp=2
...
keyframe msb_timestamp=1 lsb_timestamp=1 -> timestamp=257
- frame lsb_timestamp=-1-> timestamp=255
+ frame lsb_timestamp=255->timestamp=255
frame lsb_timestamp=0 -> timestamp=256
frame lsb_timestamp=4 -> timestamp=260
frame lsb_timestamp=2 -> timestamp=258
@@ -391,29 +401,13 @@
MUST be 0, its there to distinguish non keyframes from other packets,
Note: all packets have a 64-bit startcode except non-keyframes to reduce
their size, and all startcodes start with a 1 bit
-
-priority
- if 0 then the frame isnt used as reference (b frame) and can be droped
- MUST be > 0 for keyframes
-
-shuffle_type
- audio is often encoded in small subpackets, and to increase the
- error robustness these can be shuffled
- 0 -> no shuffle
- 1-16 -> interleave packets by 2^n
checksum
crc32 checksum using the generator polynomial 0x104c11db7 (same as ogg)
-checksum_flag
- indicates that the frame_checksum is coded
- must be 1 for the last non keyframe before a keyframe
-
frame_checksum
identical to checksum, but instead of covering just the current
- packet, it covers all frames of the same stream id since the last
- frame_checksum
- this field is only coded if checksum_flag=1
+ packet, it covers all frames since the last frame_checksum
index_timestamp
value in time_base precission, relative to the last index_timestamp
@@ -586,17 +580,6 @@
return 0;
}
-static inline int put_s(BufferContext *bc, uint64_t val){
- if(val<=0) return put_v(bc, -2*val );
- else return put_v(bc, 2*val-1);
-}
-
-static inline int64_t get_s(BufferContext *bc){
- int64_t v= get_v(bc) + 1;
- if(v&1) return -(v>>1);
- else return (v>>1);
-}
-
Example stream
