It used input frame resolution before this change which caused unnecessary resolution adaptations when resolution scaling is used.
Found that initial frame dropping was always enabled for AV1 SVC. After fixing DropDueToSize the AV1 SVC tests [1] started to fail ("number of encoded temporal layers is less than expected") on bots. The tests encode 1850x1110 in L3T3 for 5s using the default 300kbps start bitrate. Before the fix the initial frame dropping kicked in and reduced the resolution to a level that let encoder to generate all temporal layers. After the fix the resolution stayed at 1850x1110 and encoder dropped all T1 and T2 layer frames. Mitigated this by increasing test duration from 5 to 10s. This gives enough time for BWE to ramp up and for encoder to generate (stop dropping) all temporal layers.
[1] https://source.chromium.org/chromium/chromium/src/+/main:third_party/webrtc/pc/test/svc_e2e_tests.cc;l=460;bpv=1
Bug: chromium:1466809
Change-Id: I16802689e234f8fc16f891f024d5f644985de01c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/315142
Reviewed-by: Florent Castelli <orphis@webrtc.org>
Commit-Queue: Sergey Silkin <ssilkin@webrtc.org>
Reviewed-by: Åsa Persson <asapersson@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40536}
This CL adds [[deprecated]] to the old signatures, and uses the new
signatures throughout.
Bug: webrtc:14870
Change-Id: Ic9a8198ac0a2f954e1b2e7d05a55dbe04342f958
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/314962
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Commit-Queue: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40517}
EncoderStreamFactory has two code paths for creating a stream: the
"simulcast path" and the "default path". Only the former cares about
encoding paramter's maxBitrate. The latter assumes that
`encoder_config.max_bitrate_bps` already encompasses the maxBitrate of
the first encoding, but this is not always the case.
As of M113, when scalability mode is specified, {active,inactive} does
not count as simulcast stream but as a default stream represented by
encoding[0].
The problem is that `encoder_config.max_bitrate_bps` only includes
`encodings[0].max_bitrate_bps` when `encodings.size() == 1` which isn't
the case here.
This CL fixes the problem by making the "create default stream" code
path look at the first encoding's maxBitrate and remove existing
assumptions that `encoder_config.max_bitrate_bps` encompasses
`encodings[0].max_bitrate_bps`. This is a step in the right direction
since we're trying to remove all special cases and have encodings map
1:1 with SSRCs, so the "max bps of entire stream" should indeed be a
separate limit than the per-encoding limits and it was confusing that
sometimes it included and sometimes it excluded encoding[0]'s limit.
This issue did not happen in {inactive,active} since that code path
counts as "simulcast stream", so "default stream" is only ever
applicable for index 0.
TESTED=Simulcast Playground, see https://crbug.com/1455962.
Bug: chromium:1455962
Change-Id: I7c44925b780623b5979751e8959e972293648a3d
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/313282
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40482}
It was discovered that if libvpx reported a scalability mode in getStats
(e.g. L3T3_KEY) and we then changed encoder implementation to an
RTCVideoEncoder (such as MediaFoundationVideoEncodeAccelerator),
getStats continued to report the old scalability mode value.
This CL makes sure to clear the scalability mode on encoder
implementation change or if the `codec_info` is missing.
We should update MediaFoundation to report L1T1 as well, but in the
meantime we should clear any old scalability modes values when the
implementation changes (if the scalability mode is not known it is
better to report nothing than to report an old misleading value).
Bug: chromium:1426440
Change-Id: I1b5f324c4d29a00a6c73404cbee0faa2ae9cd843
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/312900
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40467}
This extension is documented to carry one bit: Screenshare.
It's been used for carrying simulcast layers and experiment IDs.
This CL removes that usage.
Bug: webrtc:15383
Change-Id: I048b283cde59bf1f607d8abdd53ced07a7add6f8
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/312420
Commit-Queue: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40457}
Without this, 'Sender' frames inserted into the writer of an encoded
transform have an invalid receive time (0), which breaks all later
heuristics which build on the receive time, eg the VCMTiming estimators
used for controlling the playback delay.
Bug: chromium:1463451
Change-Id: I413c884e08986148d4a854cd275212b21d093ceb
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/311544
Reviewed-by: Guido Urdaneta <guidou@webrtc.org>
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Reviewed-by: Palak Agarwal <agpalak@google.com>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Tony Herre <herre@google.com>
Cr-Commit-Position: refs/heads/main@{#40416}
allowing for better correlation with MaybeEncodeVideoFrame
which also logs the ntp timestamp.
BUG=None
Change-Id: I00fc99e69cd703f6da3f25043361d68b3cb3f3fa
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/311542
Commit-Queue: Philipp Hancke <phancke@microsoft.com>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40415}
It is partial reland, which adds call to Start() to all relevant places,
but doesn't actually switches frame generator to not produce frames from
the moment it was created.
Bug: b/272350185
Change-Id: I6e3bd7af6f5cd8d9baff79c2aada7b2ddfae1c8d
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/310782
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Florent Castelli <orphis@webrtc.org>
Commit-Queue: Artem Titov <titovartem@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40379}
which means this will not show up in getStats inbound-rtp/outbound-rtp
until the encoder/decoder is known. This has implications in particular
for inbound-rtp where the value is currently "unknown" until video
frames have been received.
This is safe to change as the previous change to gate
decoderImplementation behind getUserMedia access already broke
the assumption that the field is always string.
BUG=webrtc:14906
Change-Id: Ie6040ada3656e80f792c0c32c1b86ad1d6609d3c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/293600
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Philipp Hancke <phancke@microsoft.com>
Cr-Commit-Position: refs/heads/main@{#40334}
There are now multiple ways to configure VP9 L1Tx:
- Legacy API: configure legacy SVC and disable encodings, this gets
interpreted as disabling spatial layers (non-standard API hack).
- Standard API: configure scalability_mode. This can be done either
with a single encoding or multiple encodings. As long as only one
encoding is active we get a single L1Tx ssrc, same as legacy API.
Due to a bug, the ApplySpatialLayerBitrateLimits() logic which tweaks
bitrates was only applied in the legacy API code path, not the standard
API code path, despite both code paths configuring L1Tx.
The issue is that IsSimulcastOrMultipleSpatialLayers() was checking if
`number_of_streams == 1`. This is true in legacy code path but not
standard code path. The fix is to look at
`numberOfSimulcastStreams == 1` instead, which is set to the correct
value regardless of code path used.
This CL adds comments documenting the difference between
`number_of_streams` and `numberOfSimulcastStreams` to reduce the risk
of more mistakes like this in the future.
Bug: chromium:1455039, b:279161263
Change-Id: I69789b68cc5d45ef1b3becd310687c8dec8e7c87
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/308722
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40287}
Most of the usage of the H264Encoder::Create(codec) method passes a
simple codec with just the H264 codec name. This simplified the call
sites in many places and removes references to the codec types.
Bug: webrtc:15214
Change-Id: I4039c0be4ce6e3147c14c7853df4635f344b7d70
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/307222
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Florent Castelli <orphis@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40214}
This completes the split-channel work for the Video side.
Note: For ease of review, the implementations in the .cc
file have not been sorted between sender and receiver. This
can be done in a later purely-editorial CL.
Bug: webrtc:13931
Change-Id: I36cf015d5facb1eed368070cb204a8763ac19a9c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/307180
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40207}
Prior to this CL, the video `jitterBufferDelay` stat was the accumulated current delay, which is a smoothened version of the target delay. This is not correct according to the spec [1]. Rather, the stat should be the accumulated time spent in the jitter buffer, for all emitted frames. This CL fixes this spec compliance problem.
Expect changes to test metrics and product monitoring as this CL rolls out.
[1]: https://www.w3.org/TR/webrtc-stats/#dom-rtcinboundrtpstreamstats-jitterbufferdelay
Tested:
1. Go to https://jsfiddle.net/jib1/0L6duga2/show
2. Apply 2.0 seconds of video delay.
3. Notice that "Video jitter buffer delay" is slightly less than 1990ms. (2000ms playoutdelayhint - 10ms render delay - Xms decode delay).
Bug: webrtc:15085
Change-Id: I42805faafd7dd3bcdcf3ad08e751e08d6de38906
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/304521
Reviewed-by: Åsa Persson <asapersson@webrtc.org>
Commit-Queue: Rasmus Brandt <brandtr@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40138}
Finally remove bogus code after a year of no feedback on the matter.
Bug: webrtc:14138
Change-Id: I8083c9e1986e3779c9023a7d8935b717f63f0d86
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/306180
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Markus Handell <handellm@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40115}
This CL removes histograms which are no longer of use.
Bug: chromium:1255737
Change-Id: I7eb7e2cfbb03126257b51bfaa30d764b37dd9bd0
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/306200
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Markus Handell <handellm@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40114}
Reasons:
* Old code that has not been updated or tuned in several years.
* Nobody seems to intentionally use it.
* The application can do this itself by looking at GetStats.
Bug: None
Change-Id: Ib34bbebcf5885cf41ba05036506b500db7eb6b69
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/306160
Commit-Queue: Rasmus Brandt <brandtr@webrtc.org>
Reviewed-by: Stefan Holmer <stefan@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40104}
The VCMReceiveStatisticsCallback interface is both implemented (by ReceiveStatisticsProxy) and called (by VideoStreamBufferController) in `video/`, so there's no reason it should be declared in `modules/video_coding`. I also took the opportunity to update the name.
No functional changes are intended by this change, but following CLs will make some changes.
Bug: webrtc:15085
Change-Id: Ib8da30ca56675e4f638d0b9778c329b9c1138acf
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/304662
Reviewed-by: Åsa Persson <asapersson@webrtc.org>
Commit-Queue: Rasmus Brandt <brandtr@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40034}
RtpRtcpInterface::RTT follows discouraged style of using return values,
uses raw integers to represent time delta,
and returns values that no code uses (min, max, average RTT)
added LastRtt function addresses all these stylistic issues.
Bug: webrtc:13757
Change-Id: Iaf947dd1b7139026f2beb991e69634c606c6b608
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/304520
Commit-Queue: Danil Chapovalov <danilchap@webrtc.org>
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Reviewed-by: Jakob Ivarsson <jakobi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40028}
There is a lot of different delays tracked by this class. Renaming this
member clarifies what delay it actually tracks.
Bug: webrtc:15085
Change-Id: I8b038ecf84ca262efdc9f69c0f9a2a9eaad81d37
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/304402
Commit-Queue: Rasmus Brandt <brandtr@webrtc.org>
Reviewed-by: Åsa Persson <asapersson@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40007}
guarding the change in receive stats behind the
WebRTC-Stats-RtxReceiveStats
field trial which acts as a killswitch.
BUG=webrtc:15096
Change-Id: I475a2ce4fe4bddd454aa6477f8818384696c007b
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/304160
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Philipp Hancke <phancke@microsoft.com>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#40005}
This reduces dependency on the struct RTCPReportBlock and would allow to
delete it in favor of class ReportBlockData
Bug: None
Change-Id: Ia46a2516e26453724eed2e499f475f65df6cd3fa
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/304163
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Danil Chapovalov <danilchap@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#39990}
With repeated CreateSynchronizedFrameScheduler/Stop calls, the
DecodeSynchronizer can register & keep multiple callbacks in
the metronome. Fix this to only keep at most one callback
installed.
Fixed: chromium:1434747
Change-Id: I61f67a871339dbcc7560e9d545a5217f361a9b87
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/303840
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Markus Handell <handellm@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#39964}
* retransmittedBytesReceived
* retransmittedPacketsReceived
added to the specification in
https://github.com/w3c/webrtc-stats/pull/735
BUG=webrtc:15096
Change-Id: I6770e5d8d09ac1c2693c918fd943b0ab257ec7ba
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/295260
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Philipp Hancke <phancke@microsoft.com>
Cr-Commit-Position: refs/heads/main@{#39959}
Cleanup and remove usage of MaybeWorkerThread from VideoSendStream.
VideoSendStream is now created and lives on the worker thread.
Bug: webrtc:14502
Change-Id: I81ccf6b9fc6e8889db81b09bd4a75a3831a003e2
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/300842
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Per Kjellander <perkj@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#39814}
The EncoderStreamFactory triggers different code paths depending on
`number_of_streams`: one for simulcast and one for non-simulcast.
The non-simulcast path is desired for both normal streams and SVC
streams.
The simulcast path gives sensible max bitrates for 4:2:1 scenarios, but
when encodings like {active,inactive,inactive} is specified in order to
do standard SVC, the max bps of the first encoding is so low that an
SVC stream will never send more than its first spatial layer (even when
scaleResolutionDownBy is 1).
Because of this, standard SVC is broken. This CL fixes this problem by
using the CreateDefaultVideoStreams() code path instead, which is the
same one that legacy SVC uses. With this fix, legacy and standard SVC
produce the same behavior regarding bitrate.
An added benefit of this is that numberOfSimulcastStreams == 1 in the
standard SVC path as well.
{active,inactive,inactive} tests are updated to verify the full
resolution is achieved after ramp-up. I've also confirmed that this
fixes the bug in Canary, see https://crbug.com/1428098#c2.
Bug: chromium:1428098, webrtc:15041, webrtc:15034
Change-Id: Ia1eb4ff59c4e2a56af833f7ac907a66bca8ea054
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/299147
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#39697}
Various "if streams == 1" cases are updated to "if
IsSinglecastOrAllNonFirstLayersInactive()" in order not to cause subtle
differences between VP9 {active} and VP9 {active,inactive,inactive}.
This CL also affects a line that conditionally sets
`simulcastStream[0].active = codec_active` so it seemed fitting to
improve the test coverage of "if all streams are inactive, don't send".
Bug: webrtc:15028
Change-Id: I8872dc8be0f2dfc1d8914bdba5e6433f9ba8cbfd
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/298881
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#39656}
