A little bit of history about frame dropping:
I did a large number of experiments with frame dropping in old versions
of OBS1, and it's not an easy thing to deal with. I tried just about
everything from standard i-frame delay, to large buffers, to dumping
packets, to super-unnecessarily-complex things that just ended up
causing more problems than they was worth.
When I did my experiments, I found that the most ideal frame drop system
(in terms of reducing the amount of total data that needed to be
dropped) was in the 0.4xx days where I had a 3 second frame-drop buffer
where I could calculate the actual buffer size in bytes, and then
intellgently choose packets in that buffer to trim it down to a specific
size while minimizing the number of p-frames and i-frames dropped, and
preventing the actual impact of dropped frames on the stream. The
downside of it was that it required too much extra latency, and far too
many people complained about it, so it was removed in favor of the
current system.
The current system I just refer to just as 'packet dumping', which when
combined with low keyframe intervals (like most services use these
days), is the next-best method from my experience. Just dump the buffer
when you reach a threshold of buffering (which I prefer to measure with
time rather than in size), then wait for a new i-frame. Simple,
effective, and reduces the risk of consecutive buffering, while still
having fairly low impact on the stream output due to the low keyframe
interval of services.
By the way, audio will not (and should not ever) be dropped, lest you
end up with syncing issues (among other nasty things) specific to server
implementation.
- Implement the RTMP output module. This time around, we just use a
simple FLV muxer, then just write to the stream with RTMP_Write.
Easy and effective.
- Fix the FLV muxer, the muxer now outputs proper FLV packets.
- Output API:
* When using encoders, automatically interleave encoded packets
before sending it to the output.
* Pair encoders and have them automatically wait for the other to
start to ensure sync.
* Change 'obs_output_signal_start_fail' to 'obs_output_signal_stop'
because it was a bit confusing, and doing this makes a lot more
sense for outputs that need to stop suddenly (disconnections/etc).
- Encoder API:
* Remove some unnecessary encoder functions from the actual API and
make them internal. Most of the encoder functions are handled
automatically by outputs anyway, so there's no real need to expose
them and end up inadvertently confusing plugin writers.
* Have audio encoders wait for the video encoder to get a frame, then
start at the exact data point that the first video frame starts to
ensure the most accrate sync of video/audio possible.
* Add a required 'frame_size' callback for audio encoders that
returns the expected number of frames desired to encode with. This
way, the libobs encoder API can handle the circular buffering
internally automatically for the encoder modules, so encoder
writers don't have to do it themselves.
- Fix a few bugs in the serializer interface. It was passing the wrong
variable for the data in a few cases.
- If a source has video, make obs_source_update defer the actual update
callback until the tick function is called to prevent threading
issues.
I was getting cases where the CPU cache was causing issues with the
allocation counter, for the longest time I thought I was doing something
wrong, but when the allocation counter went below 0, I realized it was
because I didn't use atomics for incrementing/decrementing the
allocation counter variable. The allocation counter now always should
have the correct value.
The serializer code is meant to be used as a means of reading/writing
data from any arbitrary type of input/output.
The array output serializer makes it so we can stream data to a dynamic
array on the fly.
- Add dummy GL texture support to allow libobs texture references to be
created for GL without
- Add a texture_getobj function to allow the retrieval of the
context-specific object, such as the D3D texture pointer, or the
OpenGL texture object handle.
- Also cleaned up the export stuff. I realized it was all totally
superfluous. Kind of a dumb moment, but nice to clean it up
regardless.
Also, rename atomic functions to be consistent with the rest of the
platform/threading functions, and move atomic functions to threading*
files rather than platform* files
- Implement OBS encoder interface. It was previously incomplete, but
now is reaching some level of completion, though probably should
still be considered preliminary.
I had originally implemented it so that encoders only have a 'reset'
function to reset their parameters, but I felt that having both a
'start' and 'stop' function would be useful.
Encoders are now assigned to a specific video/audio media output each
rather than implicitely assigned to the main obs video/audio
contexts. This allows separate encoder contexts that aren't
necessarily assigned to the main video/audio context (which is useful
for things such as recording specific sources). Will probably have
to do this for regular obs outputs as well.
When creating an encoder, you must now explicitely state whether that
encoder is an audio or video encoder.
Audio and video can optionally be automatically converted depending
on what the encoder specifies.
When something 'attaches' to an encoder, the first attachment starts
the encoder, and the encoder automatically attaches to the media
output context associated with it. Subsequent attachments won't have
the same effect, they will just start receiving the same encoder data
when the next keyframe plays (along with SEI if any). When detaching
from the encoder, the last detachment will fully stop the encoder and
detach the encoder from the media output context associated with the
encoder.
SEI must actually be exported separately; because new encoder
attachments may not always be at the beginning of the stream, the
first keyframe they get must have that SEI data in it. If the
encoder has SEI data, it needs only add one small function to simply
query that SEI data, and then that data will be handled automatically
by libobs for all subsequent encoder attachments.
- Implement x264 encoder plugin, move x264 files to separate plugin to
separate necessary dependencies.
- Change video/audio frame output structures to not use const
qualifiers to prevent issues with non-const function usage elsewhere.
This was an issue when writing the x264 encoder, as the x264 encoder
expects non-const frame data.
Change stagesurf_map to return a non-const data type to prevent this
as well.
- Change full range parameter of video scaler to be an enum rather than
boolean
...The reason why audio didn't work was because I overwrote the bitrate
values.
As for semaphores, mac doesn't support unnamed semaphores without using
mach semaphores. So, I just implemented a semaphore wrapper for each
OS.
- Add some temporary streaming code using FFmpeg. FFmpeg itself is not
very ideal for streaming; lack of direct control of the sockets and
no framedrop handling means that FFmpeg is definitely not something
you want to use without wrapper code. I'd prefer writing my own
network framework in this particular case just because you give away
so much control of the network interface. Wasted an entire day
trying to go through FFmpeg issues.
There's just no way FFmpeg should be used for real streaming (at
least without being patched or submitting some sort of patch, but I'm
sort of feeling "meh" on that idea)
I had to end up writing multiple threads just to handle both
connecting and writing, because av_interleaved_write_frame blocks
every call, stalling the main encoder thread, and thus also stalling
draw signals.
- Add some temporary user interface for streaming settings. This is
just temporary for the time being. It's in the outputs section of
the basic-mode settings
- Make it so that dynamic arrays do not free all their data when the
size just happens to be reduced to 0. This prevents constant
reallocation when an array keeps going from 1 item to 0 items. Also,
it was bad to become dependent upon that functionality. You must now
always explicitly call "free" on it to ensure the data is free, and
that's how it should be. Implicit functionality can lead to
confusion and maintainability issues.
- Implement windows monitor capture (code is so much cleaner than in
OBS1). Will implement duplication capture later
- Add GDI texture support to d3d11 graphics library
- Fix precision issue with sleep timing, you have to call
timeBeginPeriod otherwise windows sleep will be totally erratic.
- Add WASAPI audio capture for windows, input and output
- Check for null pointer in os_dlopen
- Add exception-safe 'WinHandle' and 'CoTaskMemPtr' helper classes that
will automatically call CloseHandle on handles and call CoTaskMemFree
on certain types of memory returned from windows functions
- Changed the wide <-> MBS/UTF8 conversion functions so that you use
buffers (like these functions are *supposed* to behave), and changed
the ones that allocate to a different naming scheme to be safe
- Split input and output audio captures so that they're different
sources. This allows easier handling and enumeration of audio
devices without having to do some sort of string processing.
This way the user interface code can handle this a bit more easily,
and so that it doesn't confuse users either. This should be done for
all audio capture sources for all operating systems. You don't have
to duplicate any code, you just need to create input/output wrapper
functions to designate the audio as input or output before creation.
- Make it detect soundflower and wavtap devices as mac "output" devices
(even though they're actually input) for the mac output capture, and
make it so that users can select a default output capture and
automatically use soundflower or wavtap.
I'm not entirely happy about having to do this, but because mac is
designed this way, this is really the only way to handle it that
makes it easier for users and UI code to deal with.
Note that soundflower and wavtap are still also designated as input
devices, so will still show up in input device enumeration.
- Remove pragma messages because they were kind polluting the other
compiler messages and just getting in the way. In the future we can
just do a grep for TODO to find them.
- Redo list property again, this time using a safer internal array,
rather than requiring sketchy array inputs. Having functions handle
everything behind the scenes is much safer.
- Remove the reference counter debug log code, as it was included
unintentionally in a commit.
- Signals and dynamic callbacks now require declarations to be made
before being used. What this does is allows us to get information
about the functions dynamically which can be relayed to the user and
plugins for future extended usage (this should have big implications
later for scripting in particular, hopefully).
- Reduced the number of types calldata uses from "everything I could
think of" to simply integer, float, bool, pointer/object, string.
Integer data is now stored as long long. Floats are now stored as
doubles (check em).
- Use a more consistent naming scheme for lexer error/warning macros.
- Fixed a rather nasty bug where switching to an existing scene would
cause it to increment sourceSceneRefs, which would mean that it would
never end up never properly removing the source when the user clicks
removed (stayed in limbo, obs_source_remove never got called)
LOG_ERROR should be used in places where though recoverable (or at least
something that can be handled safely), was unexpected, and may affect
the user/application.
LOG_WARNING should be used in places where it's not entirely unexpected,
is recoverable, and doesn't really affect the user/application.
- Add CoreAudio device input capture for mac audio capturing. The code
should cover just about everything for capturing mac input device
audio. Because of the way mac audio is designed, users may have no
choice but to obtain the open source soundflower software to capture
their mac's desktop audio. It may be necessary for us to distribute
it with the program as well.
- Hide event backend
- Use win32 events for windows
- Allow timed waits for events
- Fix a few warnings
- Changed glMapBuffer to glMapBufferRange to allow invalidation. Using
just glMapBuffer alone was causing some unacceptable stalls.
- Changed dynamic buffers from GL_DYNAMIC_WRITE to GL_STREAM_WRITE
because I had misunderstood the OpenGL specification
- Added _OPENGL and _D3D11 builtin preprocessor macros to effects to
allow special processing if needed
- Added fmod support to shaders (NOTE: D3D and GL do not function
identically with negative numbers when using this. Positive numbers
however function identically)
- Created a planar conversion shader that converts from packed YUV to
planar 420 right on the GPU without any CPU processing. Reduces
required GPU download size to approximately 37.5% of its normal rate
as well. GPU usage down by 10 entire percentage points despite the
extra required pass.
There were a *lot* of warnings, managed to remove most of them.
Also, put warning flags before C_FLAGS and CXX_FLAGS, rather than after,
as -Wall -Wextra was overwriting flags that came before it.
The signature detection code when reading UTF-8 files was causing the
UTF-8 strings read from file to allocate more data than they were
supposed to, causing the last 3 bytes to be garbage
- Fill in the rest of the FFmpeg test output code for testing so it
actually properly outputs data.
- Improve the main video subsystem to be a bit more optimal and
automatically output I420 or NV12 if needed.
- Fix audio subsystem insertation and byte calculation. Now it will
seamlessly insert new audio data in to the audio stream based upon
its timestamp value. (Be extremely cautious when using floating
point calculations for important things like this, and always round
your values and check your values)
- Use 32 byte alignment in case of future optimizations and export a
function to get the current alignment.
- Make os_sleepto_ns return true if slept, false if the time has
already been passed before the call.
- Fix sinewave output so that it actually properly calculates a middle
C sinewave.
- Change the use of row_bytes to linesize (also makes it a bit more
consistent with FFmpeg's naming as well)
- Add planar audio support. FFmpeg and libav use planar audio for many
encoders, so it was somewhat necessary to add support in libobs
itself.
- Improve/adjust FFmpeg test output plugin. The exports were somewhat
messed up (making me rethink how exports should be done). Not yet
functional; it handles video properly, but it still does not handle
audio properly.
- Improve planar video code. The planar video code was not properly
accounting for row sizes for each plane. Specifying row sizes for
each plane has now been added. This will also make it more compatible
with FFmpeg/libav.
- Fixed a bug where callbacks wouldn't create properly in audio-io and
video-io code.
- Implement 'blogva' function to allow for va_list usage with libobs
logging.
- Move over the last of the original settings dialog code to QT. It was
actually a bit easier to write in the QT version. wxWidgets was
definitely not ideal for that because the pages would fully
create/destroy every time.
- [Win32] Fix os_dlopen so that it only appends .dll if not present
- [MacOS] Fix name dialog text edit widget issue (it would be better if
we could just use the list widget for editing labels, will have to
look in to that in the future)
- Tweak the settings UI a bit more and make 30 FPS default
- Add a macro to convert a QString to a UTF-8 const char * string
- Rename build/plugins to build/obs-plugins
- Remove the last of the wxWidgets code
- Added a test audio sinewave test source that should just play a sine
wave of the middle C note. Using unsigned 8 bit mono to test
ffmpeg's audio resampler, seems to work pretty good.
- Fixed a boolean trap in threading.h for the event_init function, it
now uses enum event_type, which can be EVENT_TYPE_MANUAL or
EVENT_TYPE_AUTO, to specify whether the event is automatically reset
or not.
- Changed display names of test sources to something a little less
vague.
- Removed te whole "if timestamp is 0 just use current system time"
when outputting source audio, if you want to use system time you
should just use system time yourself. Using 0 as some sort of
"indicator" like that just makes things confusing, and prevents you
from legitimately using 0 as a timestamp for your audio data.
- Circular buffer code wasn't correctly handling the splitting of
newly placed data segments, the code was untested and turned out to
just be backwards. It now copied the data to the back and front of
the buffer properly.
- Mixing still isn't implemented, but the audio system should be able
to start up, and mix at least once audio line for the time being.
Will have to write some test audio sources to verify things are
working properly, and build the rest of the output functionality.
fixes a deadlock in obs_free_video/obs_video_thread where
video_output_stop would signal the update event before obs_video_thread
enters video_output_wait (the thread calling obs_free_video would
block on pthread_join and obs_video_thread would block on
pthread_cond_wait)
jp9000
Timo R
Zachary Lund
Benjamin Klettbach
Palana
Asgeir Mortensen
Ján Mlynek
John Bradley
Peter SZTANOJEV