-
Notifications
You must be signed in to change notification settings - Fork 5
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
TWCC manager relies on buffers meta for recovery percentage #13
Open
MishaBaranov
wants to merge
391
commits into
main
Choose a base branch
from
misha/twcc_protection
base: main
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Multiple SIMD-ified vpx plugins will be built if available. These plugins will be put in a separate plugin directory which can be linked to if configuring with --enable-simd-symlink=avx etc.
Hard code limit to 1500 ms between each generated key frame. This should ideally be a property and set to a value derived from RTT. This avoids generating bursts of keyframes when receiving many PLIs/NACKs in close succession. 1500ms is quite large, but with 500ms and lots of packetloss, the frequent keyframe generation will more then double the bitrate from the libvpx encoder (configured at 1mbps, producing over 2mbps), potentially making an already bad situation much worse. By adding another second we only see about 1,4mbps, which is more acceptable.
Adds property that can be used to disable direct rendering.
When buffer is pushed downstream, we should not hold the buffer mapped with write access. Doing so would often lead to unneccesary memcpy later. For instance, gst_buffer_make_writable() in gst_video_decoder_finish_frame() will cause a memcpy because of _memory_get_exclusive_reference().
Change so that if the intersection of upstream and template caps does not contain a specific annex it will be not added to the caps. Previous behavior was that it was set explictily to false. This fixes an issue if an encoder (avenc_h263p) is linked late with the payloader it could end up as not negotiated.
These can spam a lot if it actually happens.
The caps for Opus is generated via a utility function that creates a new GstCaps instance after validating the codec arguments. There are possible ways to solve this, we chose the one requiring the least number of modifications. It would be also possible to simply set the caps with the specified arguments; however, this will required to move all the validation to qtdemux or require a change to gstreamer codec-utils.
…t to upstream default of 4096.
...it crashes our libvpx...
Basically, all frames are "i-frames" here.
Can't have a mix of GstValueArray and GValueArray here.
As this could end up blocking the audio listener thread (gst_osx_audio_device_change_cb) when probing devices caps in gst_core_audio_open() and gst_core_audio_probe_caps()
This way we avoid having potential races between querying "is-default" on devices and changing the system defaults. We now update the system default on every device change to get new connected devices added with the correct default state.
MishaBaranov
force-pushed
the
misha/twcc_protection
branch
from
May 16, 2024 14:06
137c427
to
8bf0b24
Compare
This condition is more common than I expected, for example, in the case we disconnect a bluetooth headset set as system-default. The AudioDeviceID is allowed to be different each time you re-plug the same device, so we might not have it in the internal list by the time it was removed and reported as changed as system-default.
This is basically to ensure better testability and reduce racyness by ensuring each operation completes for each input buffer. Also add some more logging and refactor a bit around locking.
If VTDecompressionSessionDecodeFrame returns kVTInvalidSessionErr or kVTVideoEncoderMalfunctionErr we need to reset the decoder session and request an intra frame. This change allows the decoder to continue decoding after the application switches between foreground and background modes.
MishaBaranov
force-pushed
the
misha/twcc_protection
branch
from
May 29, 2024 16:23
8bf0b24
to
766fb6e
Compare
MishaBaranov
force-pushed
the
misha/twcc_protection
branch
3 times, most recently
from
July 8, 2024 12:56
9a21777
to
633ffe8
Compare
Redundancy Block book keeping moved to stats collecting calls. Sender thread just turns SentPacket ring buffer as well as TWCC feed back parser just passes list of updated packets into stats context.
Set seqnums array to NULL for those callers who ignore return value
MishaBaranov
force-pushed
the
misha/twcc_protection
branch
from
July 8, 2024 14:00
633ffe8
to
52faddc
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This introduces Buffers Meta API for attaching info of which ssrc and seqnums any particular buffer cares redundant content about.
So far it just substitutes present mechanism for recovery percentage computation for RTX only.