Skip to content

Commit

Permalink
Add extended stream data packet
Browse files Browse the repository at this point in the history 
I hate it. But I see no other way.
  • Loading branch information
@cyanreg
cyanreg committed Jul 3, 2024
1 parent ac650d4 commit c86cc06
Show file tree
Hide file tree
Showing 2 changed files with 179 additions and 31 deletions.
181 changes: 164 additions & 17 deletions draft-avtransport-spec.bs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -266,6 +266,10 @@ Below is a table of how they're allocated:
<td>[''0x0100'':''0x01FF'']</td>
<td>[[#stream-data-packets]]</td>
</tr>
<tr>
<td>[''0x0200'':''0x02FF'']</td>
<td>[[#extended-stream-data-packets]]</td>
</tr>
<tr>
<td>''0x0300''</td>
<td>[[#time-synchronization-packets]]</td>
Expand Down Expand Up @@ -1338,7 +1342,7 @@ The data packets indicate the start of a stream packet, which may be fragmented
more [[#stream-data-segment-packets]]. It is laid out as follows:

<figure id="table-StreamData" class="table">
<table class="complex data longlastcol" dfn-for="#stream-data-packet">
<table class="complex data longlastcol" dfn-for="#stream-data-packets">
<tr>
<th>Type</th>
<th>Name</th>
Expand Down Expand Up @@ -1433,6 +1437,155 @@ For information on the layout of the specific codec-specific packet data, consul
the [[#codec-encapsulation]] addenda.


## Extended stream data ## {#extended-stream-data-packets}

Depending upon the codec, using an extended stream data header may be required.
It is laid out in the following way:

<figure id="table-ExtendedStreamData" class="table">
<table class="complex data longlastcol" dfn-for="#extended-stream-data-packet">
<tr>
<th>Type</th>
<th>Name</th>
<th>Data</th>
<th>Description</th>
</tr>
<tr>
<td><code>b(16)</code></td>
<td><dfn noexport>extended_stream_data_descriptor_structure</dfn></td>
<td></td>
<td>
<table>
<tr id="0x0200+0">
<td><code>b(8)</code></td>
<td><dfn noexport>extended_stream_data_descriptor</dfn></td>
<td>0x02</td>
</tr>
<tr id="0x0200+1">
<td><code>b(3)</code></td>
<td>[=frame_type=]</td>
<td>enum {{FrameType}}</td>
</tr>
<tr id="0x0200+2">
<td><code>b(1)</code></td>
<td><dfn noexport>ext_pkt_reserved</dfn></td>
<td>Reserved field. Must be set to 0.</td>
</tr>
<tr id="0x0200+3">
<td><code>b(1)</code></td>
<td>[=pkt_in_fec_group=]</td>
<td>Packet is part of an FEC group and must be retained.</td>
</tr>
<tr id="0x0200+4">
<td><code>b(1)</code></td>
<td>[=field_id=]</td>
<td>As defined in enum {{Interlacing}}.</td>
</tr>
<tr id="0x0200+5">
<td><code>b(2)</code></td>
<td>[=pkt_compression=]</td>
<td>enum {{DataCompression}}</td>
</tr>
</table>
</td>
</tr>
<tr id="0x0200+6">
<td><code>b(16)</code></td>
<td>[=stream_id=]</td>
<td></td>
<td>Indicates the stream ID for which this packet is applicable.</td>
</tr>
<tr id="0x0200+7">
<td><code>u(32)</code></td>
<td>[=global_seq=]</td>
<td></td>
<td>Monotonically incrementing per-packet global sequence number.</td>
</tr>
<tr id="0x0200+8">
<td><code>i(64)</code></td>
<td>[=pts=]</td>
<td></td>
<td>Indicates the presentation timestamp for when this frame should be presented at. To interpret the value, read the [[#timestamps]] section.</td>
</tr>
<tr id="0x0200+9">
<td><code>i(64)</code></td>
<td>[=dts=]</td>
<td></td>
<td>Indicates the timestamp for when to input this frame should be input into the decoder. To interpret the value, read the [[#timestamps]] section.</td>
</tr>
<tr id="0x0200+10">
<td><code>u(32)</code></td>
<td>[=data_length=]</td>
<td></td>
<td>The size of the data in this packet.</td>
</tr>
<tr id="0x0200+11">
<td><code>b(64)</code></td>
<td>[=ldpc_288_224=]</td>
<td>[=LDPC(288, 224)=]</td>
<td>64-bits of LDPC parity data to correct the previous 224 bits of the packet.</td>
</tr>
<tr id="0x0200+12">
<td><code>[=data_length=]*b(8)</code></td>
<td>[=packet_data=]</td>
<td>payload</td>
<td>The packet data itself.</td>
</tr>
<tr id="0x0200+13">
<td><code>i(64)</code></td>
<td>[=duration=]</td>
<td></td>
<td>The duration of this packet in stream timebase unis.</td>
</tr>
<tr id="0x0200+14">
<td><code>u(32)</code></td>
<td><dfn noexport>total_data_length</dfn></td>
<td></td>
<td>The total size of all data needed to receive the payload correctly. Must be either 0 if unknown, or the total size of the payload across all segments.</td>
</tr>
<tr id="0x0200+15">
<td><code>u(32)</code></td>
<td><dfn noexport>side_data_length</dfn></td>
<td></td>
<td>Length of the side data, if any.</td>
</tr>
<tr id="0x0200+16">
<td><code>u(32)</code></td>
<td><dfn noexport>side_data_offset</dfn></td>
<td></td>
<td>The offset within the packet where the side data is located at.</td>
</tr>
<tr id="0x0200+17">
<td><code>u(8)</code></td>
<td><dfn noexport>side_data_type</dfn></td>
<td></td>
<td>The size of the data in this packet.</td>
</tr>
<tr id="0x0200+18">
<td><code>b(48)</code></td>
<td>[=padding=]</td>
<td></td>
<td>Padding, reserved for future use. MUST be ''0x0''.</td>
</tr>
<tr id="0x0200+19">
<td><code>b(64)</code></td>
<td>[=ldpc_288_224=]</td>
<td>[=LDPC(288, 224)=]</td>
<td>64-bits of LDPC parity data to correct the previous 224 bits of the packet.</td>
</tr>
<tr id="0x0200+20">
<td><code>[=data_length=]*b(8)</code></td>
<td>[=packet_data=]</td>
<td>payload</td>
<td>The packet data itself.</td>
</tr>
</table>
</figure>

Extended stream data packets should only be used if there's a need to use them.
Regular stream data packets have a much lower overhead.


### Frame Type Enumeration (enum <dfn enum>FrameType</dfn>) ### {#enum-FrameType}

<div dfn-type="enum-value" dfn-for="#enum-FrameType" id="FrameType">
Expand All @@ -1442,16 +1595,6 @@ the [[#codec-encapsulation]] addenda.
:: Packet data contains a scalable/switch frame, able to be decoded standalone, with acceptable degradation.
: <dfn noexport>FRAME_TYPE_P</dfn> = <i>0x2</i>
:: Packet data contains an <b>inter</b> frame, requiring additional reference frames in order to be validly decoded.
: <dfn noexport>FRAME_TYPE_C1</dfn> = <i>0x3</i>
:: Packet data contains a frame with a type defined in [[#codec-encapsulation]].
: <dfn noexport>FRAME_TYPE_C2</dfn> = <i>0x4</i>
:: Packet data contains a frame with a type defined in [[#codec-encapsulation]].
: <dfn noexport>FRAME_TYPE_C3</dfn> = <i>0x5</i>
:: Packet data contains a frame with a type defined in [[#codec-encapsulation]].
: <dfn noexport>FRAME_TYPE_C4</dfn> = <i>0x6</i>
:: Packet data contains a frame with a type defined in [[#codec-encapsulation]].
: <dfn noexport>FRAME_TYPE_C5</dfn> = <i>0x7</i>
:: Packet data contains a frame with a type defined in [[#codec-encapsulation]].
</div>


Expand Down Expand Up @@ -3865,9 +4008,10 @@ For H264 encapsulation, the [=codec_id=] in
[[#stream-registration-packets]]
must be ''0x48323634'' (''H264'').

The [=packet_data=] must contain the following elements, in order:
- A single <code>b(64)</code> element with the [=dts=], the time, which indicates when a frame should be input into a synchronous 1-in-1-out decoder.
- ''Annex-B'' formatted NAL units, with startcode emulation bits included.
[[#extended-stream-data-packets]] must be used, as H264 requires a [=dts=].

The [=packet_data=] must contain ''Annex-B'' formatted NAL units,
with startcode emulation bits included.

A [[#stream-configuration-packets]] packet may be sent, to speed up stream initialization.
If they are present, they must contain an ''AVCDecoderConfigurationRecord'' structure,
Expand All @@ -3888,9 +4032,10 @@ For H265 encapsulation, the [=codec_id=] in
[[#stream-registration-packets]]
must be ''0x48323635'' (''H265'').

The [=packet_data=] must contain the following elements, in order:
- A single <code>b(64)</code> element with the [=dts=], the time, which indicates when a frame should be input into a synchronous 1-in-1-out decoder.
- ''Annex-B'' formatted NAL units, with startcode emulation bits included.
[[#extended-stream-data-packets]] must be used, as H265 requires a [=dts=].

The [=packet_data=] must contain ''Annex-B'' formatted NAL units,
with startcode emulation bits included.

''Annex-B'' formatted packets must be used, with startcode emulation bits included.

Expand All @@ -3905,6 +4050,8 @@ For Dirac or VC-2 encapsulation, the [=codec_id=] in
[[#stream-registration-packets]]
MUST be ''0x42424344'' (''BBCD'').

[[#extended-stream-data-packets]] must be used, as Dirac packets require a [=dts=].

Dirac streams require no [[#stream-configuration-packets]] packets.

The [=packet_data=] must contain raw sequences,
Expand Down
29 changes: 15 additions & 14 deletions libavtransport/tools/spec2c.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -115,21 +115,22 @@
streamid_field_exceptions = [ "stream_id", "group_id", "user_field", "session_version" ]
codec_tags = { }

# This is able to add arbitrary extra values to any structure during parsing
extra_values = {
"AVTStreamData": [
"dts", {
"fixed": None,
"ldpc": None,
"enum": None,
"size_bits": 64,
"bytestream": 8,
"datatype": "int64_t",
"struct": None,
"array_len": 0,
"string": False,
"payload": False,
}
],
# "AVTStreamData": [
# "dts", {
# "fixed": None,
# "ldpc": None,
# "enum": None,
# "size_bits": 64,
# "bytestream": 8,
# "datatype": "int64_t",
# "struct": None,
# "array_len": 0,
# "string": False,
# "payload": False,
# }
# ],
}

def iter_enum_fields(target_id):
Expand Down

0 comments on commit c86cc06

Please sign in to comment.