Commit Graph

97 Commits

Author SHA1 Message Date
Anton Khirnov
8996945d45 fftools/ffmpeg_enc: set AV_PKT_FLAG_TRUSTED on encoded packets
This allows using WRAPPED_AVFRAME encoders with loopback decoders in
order to connect multiple filtergraphs together.

Clear the flag in muxers, since lavf does not need it for anything and
it would change the results of framecrc FATE tests.
2024-03-13 08:01:28 +01:00
Anton Khirnov
a9193f7b7d fftools/ffmpeg: add loopback decoding
This allows to send an encoder's output back to decoding and feed the
result into a complex filtergraph.
2024-03-13 08:01:27 +01:00
Anton Khirnov
3febc84e5e fftools/ffmpeg_enc: merge do_{audio,video}_out into frame_encode()
These functions used to be much longer, but now they are only a couple
lines each, some of them duplicated between audio and video.
2024-03-13 08:01:15 +01:00
Anton Khirnov
2f5c570dd6 fftools/ffmpeg_enc: drop unnecessary parameter from forced_kf_apply()
Encoder timebase is equal to the frame timebase, so does not need to be
passed separately.

Also, rename in_picture to frame, which is shorter and more accurate -
it always contains a frame, never a field.
2024-03-13 08:01:15 +01:00
Anton Khirnov
2ee9362419 fftools/ffmpeg: remove unncessary casts for *_thread() return values
These functions used to be passed directly to pthread_create(), which
required them to return void*. This is no longer the case, so they can
return a plain int.
2024-03-13 08:01:15 +01:00
Anton Khirnov
a2fc86378a fftools/ffmpeg: declare loop indices inside loops 2024-02-24 11:51:41 +01:00
王昊然
344ec7a3f7 fftools/ffmpeg: use correct unit prefix symbol for binary scaled byte values
Fix bug #10501

Signed-off-by: WHR <msl0000023508@gmail.com>
2024-02-03 12:50:42 +01:00
Anton Khirnov
b43d4a0692 fftools/ffmpeg_dec: export subtitle_header in Decoder
This way the encoder does not need to access the decoder AVCodecContext,
which will allow to make it private in future commits.
2024-01-30 09:52:00 +01:00
Zhao Zhili
f72d781339 fftools/ffmpeg_enc: assert necessary frame fields before create encoder
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>
2023-12-21 19:30:25 +08:00
Anton Khirnov
2ad0b8e0ea fftools/ffmpeg: use a mutex for enc_stats_write()
It may be called concurrently from different threads to write into the
same file.
2023-12-18 08:50:02 +01:00
Anton Khirnov
02a4393647 fftools/ffmpeg: print keyframe information with -stats_* 2023-12-18 08:50:02 +01:00
Anton Khirnov
4224895a87 fftools/ffmpeg: replace OutputStream.file_index by a pointer
Reduces the need to use the output_files global array.
2023-12-14 20:16:53 +01:00
Anton Khirnov
9d7000b1be fftools/ffmpeg: attach wallclock timing information to packets and frames
Will become useful in following commits.
2023-12-14 08:11:05 +01:00
Anton Khirnov
d119ae2fd8 fftools/ffmpeg: convert to a threaded architecture
Change the main loop and every component (demuxers, decoders, filters,
encoders, muxers) to use the previously added transcode scheduler. Every
instance of every such component was already running in a separate
thread, but now they can actually run in parallel.

Changes the results of ffmpeg-fix_sub_duration_heartbeat - tested by
JEEB to be more correct and deterministic.
2023-12-12 08:24:18 +01:00
Anton Khirnov
9b8cc36ce0 fftools/ffmpeg: add thread-aware transcode scheduling infrastructure
See the comment block at the top of fftools/ffmpeg_sched.h for more
details on what this scheduler is for.

This commit adds the scheduling code itself, along with minimal
integration with the rest of the program:
* allocating and freeing the scheduler
* passing it throughout the call stack in order to register the
  individual components (demuxers/decoders/filtergraphs/encoders/muxers)
  with the scheduler

The scheduler is not actually used as of this commit, so it should not
result in any change in behavior. That will change in future commits.
2023-12-12 08:24:18 +01:00
Anton Khirnov
ee2a8cbfd1 fftools/ffmpeg_enc: move encoding to a separate thread
As for the analogous decoding change, this is only a preparatory step to
a fully threaded architecture and does not yet make encoding truly
parallel. The main thread will currently submit a frame and wait until
it has been fully processed by the encoder before moving on. That will
change in future commits after filters are moved to threads and a
thread-aware scheduler is added.

This code suffers from a known issue -  if an encoder with a sync queue
receives EOF it will terminate after processing everything it currently
has, even though the sync queue might still be triggered by other
threads. That will be fixed in following commits.
2023-12-12 08:24:18 +01:00
Anton Khirnov
33f058f2ec fftools/ffmpeg_enc: constify the frame passed to enc_open() 2023-10-10 12:41:31 +02:00
Anton Khirnov
9196be2fb1 fftools/ffmpeg_enc: move fps conversion code to ffmpeg_filter
Its function is analogous to that of the fps filter, so filtering is a
more appropriate place for this.

The main practical reason for this move is that it places the encoding
sync queue right at the boundary between filters and encoders. This will
be important when switching to threaded scheduling, as the sync queue
involves multiple streams and will thus need to do nontrivial
inter-thread synchronization.

In addition to framerate conversion, the closely-related
* encoder timebase selection
* applying the start_time offset
are also moved to filtering.
2023-10-10 12:41:31 +02:00
Anton Khirnov
f0f6d6d0e1 fftools/ffmpeg_enc: move framerate conversion state into a separate struct
Makes it more clear what state is specific to framerate conversion,
which will be useful in the following commit.
2023-10-10 12:41:31 +02:00
Anton Khirnov
61777c62d7 fftools/ffmpeg: stop accessing OutputStream.last_dropped in print_report()
That field is used by the framerate code to track whether any output has
been generated for the last input frame(*). Its use in the last
invocation of print_report() is meant to account for the very last
filtered frame being dropped in the number of dropped frames printed in
the log. However, that is a highly inappropriate place to do so, as it
makes assumptions about vsync logic in completely unrelated code. Move
the increment to encoder flush instead.

(*) the name is misleading, as the input frame has not yet been dropped
    and may still be output in the future
2023-10-10 12:41:31 +02:00
Anton Khirnov
d2c416fdf1 fftools/ffmpeg_enc: merge -force_key_frames source/source_no_drop
Always use the functionality of the latter, which makes more sense as it
avoids losing keyframes due to vsync code dropping frames.

Deprecate the 'source_no_drop' value, as it is now redundant.
2023-10-10 12:41:31 +02:00
Anton Khirnov
735b082231 fftools/ffmpeg_enc: unbreak -force_key_frames source_no_drop
Unlike the 'source' mode, which preserves source keyframe-marking as-is,
the 'source_no_drop' mode attempts to keep track of keyframes dropped by
framerate conversion and mark the next output frame as key in such
cases. However,
* c75be06148 broke this functionality entirely, and made it equivalent
  to 'source'
* even before it would only work when the frame immediately following
  the dropped keyframe is preserved and not dropped as well

Also, drop a redundant check for 'frame' in setting dropped_keyframe, as
it is redundant with the check on the above line.
2023-10-10 12:41:31 +02:00
Anton Khirnov
99567ee870 fftools/ffmpeg_enc: simplify adjust_frame_pts_to_encoder_tb() signature
It does not need an OutputFile and an OutputStream, only the target
timebase and the timestamp offset.
2023-10-10 12:41:31 +02:00
Anton Khirnov
096d88dcc6 fftools/ffmpeg_enc: move remaining vsync-related code to video_sync_process() 2023-10-10 12:41:31 +02:00
Anton Khirnov
a9b9e771b3 fftools/ffmpeg_enc: move handling video frame duration to video_sync_process()
That is a more appropriate place for this.
2023-10-10 12:41:31 +02:00
James Almer
3fd37b5268 fftools/ffmpeg: stop using AVStream.side_data
Signed-off-by: James Almer <jamrial@gmail.com>
2023-10-06 10:11:00 -03:00
Anton Khirnov
8d12762b42 fftools/ffmpeg: move derivation of frame duration from filter framerate
From ffmpeg_enc to ffmpeg_filter, which is a more appropriate
place for it.
2023-10-03 16:57:02 +02:00
Anton Khirnov
9d4ca76c08 fftools/ffmpeg_enc: do not round frame durations prematurely
Changes the results of fate-idroq-video-encode and fate-lavf* tests,
where different frames now get duplicated by framerate conversion code.
2023-10-03 16:57:02 +02:00
Anton Khirnov
5d58a35f98 fftools/ffmpeg: deprecate the -top option
It is badly named (should have been -top_field_first, or at least -tff),
underdocumented and underspecified, and (most importantly) entirely
redundant with the setfield filter.
2023-09-18 17:16:06 +02:00
Anton Khirnov
43a0004b5c fftools/ffmpeg_enc: apply -top to individual encoded frames
Fixes #9339.
2023-09-18 17:15:53 +02:00
Anton Khirnov
74b643a51e fftools/ffmpeg_enc: refactor setting encoding field_order
Merge three blocks with slightly inconsistent checks into one, treating
encoder input as interlaced when either:
* at least one of ilme/ildct flags is set
* the first frame in the stream is marked as interlaced
* the user specified the -top option

Stop modifying the frame passed to enc_open().
2023-09-18 17:15:53 +02:00
Anton Khirnov
d65d0f4df1 fftools/ffmpeg_enc: reindent after previous commit 2023-08-30 11:53:50 +02:00
Anton Khirnov
411ada649f fftools/ffmpeg_enc: only use fallback framerate when encoding CFR
When no output video framerate is specified by the user with -r or can
be inferred from the filtergraph, encoder setup will arbitrarily decide
that the framerate is 25fps. However, making up any framerate value for
VFR encoding is at best unnecessary.

Changes the results of the sub2video tests, where the input timebase is
now used instead of 1/25.
2023-08-30 11:53:50 +02:00
Anton Khirnov
8b6b2518fa fftools/ffmpeg_enc: reindent after previous commit 2023-08-30 11:53:50 +02:00
Anton Khirnov
d1a2cd1e13 fftools/ffmpeg_enc: factor out setting encoder timebase
Mainly this fixes handling special values of -enc_time_base ('demux' or
'filter') for audio. It also prints a warning if -enc_time_base is
specified for subtitles, instead of ignoring it silently (current
subtitle encoding API only works with AV_TIME_BASE_Q).
2023-08-30 11:53:50 +02:00
Anton Khirnov
8ecbb1f9af fftools/ffmpeg_mux: stop rescaling timestamps in of_streamcopy()
This function converts packet timestamps from the input stream timebase
to OutputStream.mux_timebase, which may or may not be equal to the
actual output AVStream timebase (and even when it is, this may not
always be the optimal choice due to bitstream filtering).

Just keep the timestamps in input stream timebase, they will be rescaled
as needed before bitstream filtering and/or sending the packet to the
muxer.

Move the av_rescale_delta() call for audio (needed to preserve accuracy
with coarse demuxer timebases) to write_packet.

Drop now-unused OutputStream.mux_timebase.
2023-08-30 11:53:46 +02:00
Anton Khirnov
2f155b18a1 fftools/ffmpeg: return an error from assert_avoptions() instead of aborting
Rename it to check_avoptions().
2023-07-20 20:47:46 +02:00
Anton Khirnov
e0f4259689 fftools/ffmpeg_mux: return errors from of_output_packet() instead of aborting 2023-07-20 20:30:13 +02:00
Anton Khirnov
2b4afe8169 fftools/ffmpeg_enc: return errors from encode_frame() instead of aborting 2023-07-20 20:30:13 +02:00
Anton Khirnov
43bcf631d0 fftools/ffmpeg_enc: return errors from enc_flush() instead of aborting 2023-07-20 20:30:13 +02:00
Anton Khirnov
dde3de0170 fftools/ffmpeg_enc: return errors from do_*_out() instead of aborting 2023-07-20 20:30:13 +02:00
Anton Khirnov
aa1173feee fftools/ffmpeg_enc: return errors from enc_open() instead of aborting 2023-07-20 20:30:13 +02:00
Anton Khirnov
80a64800ea fftools/ffmpeg_enc: return errors from enc_frame() instead of aborting 2023-07-20 20:30:13 +02:00
Andreas Rheinhardt
463a472426 fftools/ffmpeg_enc: Reindentate after the last commit
Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2023-07-17 16:18:12 +02:00
Andreas Rheinhardt
bfe15b81b3 fftools/ffmpeg_enc: Remove always-true checks
frame is always != NULL for audio and video here
(this is checked by an assert and the frame is already dereferenced
before it reaches this code here).
Fixes Coverity issue #1538858.

Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
2023-07-17 16:14:41 +02:00
Anton Khirnov
dff3a283cd fftools/ffmpeg: rework -enc_time_base handling
Read the timebase from FrameData rather than the input stream. This
should fix #10393 and generally be more reliable.

Replace the use of '-1' to indicate demuxing timebase with the string
'demux'. Also allow to request filter timebase with
'-enc_time_base filter'.
2023-07-15 11:02:11 +02:00
Anton Khirnov
4d06742b93 fftools/ffmpeg: add more structure to FrameData
It now contains data from multiple sources, so group those items that
always come from the decoder. Also, initialize them to invalid values,
so that frames that did not originate from a decoder can be
distinguished.
2023-07-15 11:02:11 +02:00
Anton Khirnov
9a97f9cf3a fftools/ffmpeg_enc: initialize audio/video encoders from frame parameters
This is possible now that enc_open() is always called with a non-NULL
frame for audio/video.

Previously the code would directly reach into the buffersink, which is a
layering violation.
2023-07-15 11:02:11 +02:00
Anton Khirnov
0a6751a78a fftools/ffmpeg: rework initializing encoders with no frames
When no frames were passed from a filtergraph to an encoder, but the
filtergraph is configured (i.e. has output parameters), encoder flush
code will use those parameters to initialize the encoder in a last-ditch
effort to produce some useful output.

Rework this process so that it is triggered by the filtergraph, which
now sends a dummy frame with parameters, but no data, to the encoder,
rather than the encoder reaching backwards into the filter.

This approach is more in line with the natural data flow from filters to
encoders and will allow to reduce encoder-filter interactions in
following commits.

This code is tested by fate-adpcm-ima-cunning-trunc-t2-track1, which (as
confirmed by Zane) is supposed to produce empty output.
2023-07-15 11:02:11 +02:00
Anton Khirnov
c328bff4da fftools/ffmpeg_enc: return errors from enc_subtitle() instead of aborting 2023-07-15 11:02:11 +02:00