This is a reland of 72e6cb0b3f548900fd3b548b4b6966e3f5ee854f
Was not the cause of perf alert, relanding.
TBR=ilnik@webrtc.org
Original change's description:
> Fixes dynamic mode pacing issues.
>
> This CL fixes a few issues in the (default-disabled) dynamic pacing
> mode:
> * Slight update to sleep timing to avoid short spin loops
> * Removed support for early execution as that lead to time-travel
> contradictions that were difficult to solve.
> * Makes sure we schedule a process call when a packet is due to be
> drained even if the queue is empty, so that padding will start at
> the correct time.
> * While paused or empty, sleep relative last send time if we send
> padding while silent - otherwise just relative to last process
> time.
> * If target send time shifts so far back that packet should have
> been sent prior to the last process, make sure we don't let the
> buffer level remain.
> * Update the PacedSender test to _actually_ use dynamic processing
> when the param says so.
>
> Bug: webrtc:10809
> Change-Id: Iebfde9769647d2390fd192a40bbe2d5bf1f6cc62
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160407
> Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
> Commit-Queue: Erik Språng <sprang@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#29911}
Bug: webrtc:10809
Change-Id: Ie7b307e574c2057bb05af87b6718a132d639a416
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160786
Reviewed-by: Erik Språng <sprang@webrtc.org>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29928}
This CL removes the experimental status of the multi-channel processing
in APM, and accordingly updates the variable naming.
It also splits the activation of multi-channel processing to be separate
for render and capture.
Bug: webrtc:10859
Change-Id: I0e5d04dcb94b6637c33d97146231b8ddddbaea39
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160707
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29926}
This CL corrects the re-initialization behavior of the analog
AGC to work correctly when the AGC is reinitialized.
Bug: webrtc:11131
Change-Id: Ie455ba3db1aa3936cbcbb2fab023528124853284
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160650
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29924}
This reverts commit 72e6cb0b3f548900fd3b548b4b6966e3f5ee854f.
Reason for revert: Speculative revert due to perf change
Original change's description:
> Fixes dynamic mode pacing issues.
>
> This CL fixes a few issues in the (default-disabled) dynamic pacing
> mode:
> * Slight update to sleep timing to avoid short spin loops
> * Removed support for early execution as that lead to time-travel
> contradictions that were difficult to solve.
> * Makes sure we schedule a process call when a packet is due to be
> drained even if the queue is empty, so that padding will start at
> the correct time.
> * While paused or empty, sleep relative last send time if we send
> padding while silent - otherwise just relative to last process
> time.
> * If target send time shifts so far back that packet should have
> been sent prior to the last process, make sure we don't let the
> buffer level remain.
> * Update the PacedSender test to _actually_ use dynamic processing
> when the param says so.
>
> Bug: webrtc:10809
> Change-Id: Iebfde9769647d2390fd192a40bbe2d5bf1f6cc62
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160407
> Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
> Commit-Queue: Erik Språng <sprang@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#29911}
TBR=ilnik@webrtc.org,sprang@webrtc.org
Change-Id: I5d1532d2e041e60a7f1bfeb8185f7760c9789711
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: webrtc:10809
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160701
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29920}
The NetEqFactory is currently expected to wrap the AudioDecoderFactory,
but this turns out not to be a good idea. Instead, it makes more sense
to pass the AudioDecoderFactory through the CreateNetEq method.
Bug: webrtc:11005
Change-Id: I8027ff6593f40c92072e7e88157631dcf329a984
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160644
Commit-Queue: Ivo Creusen <ivoc@webrtc.org>
Reviewed-by: Karl Wiberg <kwiberg@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29918}
This CL adds a DCHECK for the deprecated 8 kHz rate in APM.
It also updates the agc fuzzer code to properly do band-split on
the signals, and not send 8 kHz signals into the AGC.
Bug: chromium:1028092,chromium:1028172
Change-Id: I1e7c8d721834310e94b0e21efea07f75da837cab
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160600
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29914}
This CL fixes a few issues in the (default-disabled) dynamic pacing
mode:
* Slight update to sleep timing to avoid short spin loops
* Removed support for early execution as that lead to time-travel
contradictions that were difficult to solve.
* Makes sure we schedule a process call when a packet is due to be
drained even if the queue is empty, so that padding will start at
the correct time.
* While paused or empty, sleep relative last send time if we send
padding while silent - otherwise just relative to last process
time.
* If target send time shifts so far back that packet should have
been sent prior to the last process, make sure we don't let the
buffer level remain.
* Update the PacedSender test to _actually_ use dynamic processing
when the param says so.
Bug: webrtc:10809
Change-Id: Iebfde9769647d2390fd192a40bbe2d5bf1f6cc62
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160407
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29911}
In the unlikely event that the decoded audio is really short, the
downsampling would read outside of the decoded audio vector. This CL
fixes that, and adds a unit test that verifies the fix (when running
with ASan).
Bug: chromium:1016506
Change-Id: Ifb8071ce0550111cd66e7f7c1bed7f17b33f93c5
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160304
Commit-Queue: Henrik Lundin <henrik.lundin@webrtc.org>
Reviewed-by: Karl Wiberg <kwiberg@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29898}
Beyond making the digital AGC1 code properly support
multichannel, this CL also
-Removes deprecated debug logging code.
-Converts the gain application to be fully in floating point
which
--Is less computationally complex.
--Does not quantize the samples to 16 bit before applying the
gains.
Bug: webrtc:10859
Change-Id: I6020ba8ae7e311dfc93a72783a2bb68d935f90c5
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/159861
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29886}
AECM only supports up to two capture channels, this CL extends it to arbitrary channel counts.
Bug: webrtc:10859
Change-Id: Id56ca633cd9de706fa1254bfa8153de88de0ef70
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160340
Reviewed-by: Per Åhgren <peah@webrtc.org>
Commit-Queue: Sam Zackrisson <saza@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29880}
This CL adds proper multi-channel support to the analog AGC.
Beyond that, it prepares adding multi-channel support to the digital
AGC by removing the tight dependency between the analog and digital
AGC codes.
Bug: webrtc:10859
Change-Id: I4414ccbc3db5dbb5ae069fdf426cbd038375ca7b
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/159480
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29878}
Summary of changes/improvements and fixes:
Changes container for list of devices from std::vector to std:deque to
allow fast insertion and deletion at both its beginning and its end. This
approach makes it easier to first build a list of all available devices
and then check the size of the list. If size > 0 => two more devices are
added at the front (Default and Default Communication). The old solution
contained a risk of adding invalid Default and Default Communication
devices in cases where not physical device could be found.
Adds usage of |device_index_| in CoreAudioBase to ensure that the selected
device is unique. The previous version used only an ID but that ID is not
unique when e.g. only one device exists since it can have up to three
different roles.
Improves logging and comments.
No-Try: True
Tbr: thaloun@chromium.org
Bug: webrtc:11107
Change-Id: I9a09f7716ed8d8858dcc6a5354b038fc06496166
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160050
Commit-Queue: Henrik Andreassson <henrika@webrtc.org>
Reviewed-by: Henrik Andreassson <henrika@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29874}
Fix signed/unsigned mismatch.
Protect against NumberOfEnumeratedDevices and Get[In|Out]putDeviceNames returning inconsistent results.
It's possible for an device to be counted but getting its name fails, in which case the utility function returns true but would continue from its loop filling the AudioDeviceNames vector, leading to a smaller output than the later code expects.
Bug: b/144382120
Change-Id: Iab008c28f03023c830011d229b1f1c7e3e7bb5ee
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160226
Commit-Queue: Henrik Andreassson <henrika@webrtc.org>
Reviewed-by: Henrik Andreassson <henrika@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29871}
Usage of this class has now been simplified so that we can do some
cleanup:
* Removes dead code: Push() with 9 args, CancelPop()
* Replaces BeginPop()/CancelPop() with a single Pop() method
* Makes QueuePacket a private class
* Replaces rtp_packets_ with direct ownership from QueuePacket
Bug: webrtc:10809
Change-Id: Iea131ee87d5d920360c71fb180b2af0ea4fc6c7f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160007
Commit-Queue: Erik Språng <sprang@webrtc.org>
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29869}
This is a reland of acdc22d7845c5dde7c23366110e54e5d26127c85
Original change's description:
> Prepares PacingController for simplified packet queue.
>
> This CL removes references to RoundRobinPacketQueue::QueuedPacket,
> other than the method to release an RtpPacketToSend. It also moves
> both the BeginPop() and FinalizePop() to within a single helper
> method.
>
> A follow-up cleanup of the packet queue will stop exposing the
> QueuedPacket struct and replaces the the pop-methods with a single
> new one that just returns an RtpPacketToSend.
>
> Bug: webrtc:10809
> Change-Id: I5208a93e12e6b56714d483cc12d2a37225ea8e5e
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/159889
> Commit-Queue: Erik Språng <sprang@webrtc.org>
> Reviewed-by: Philip Eliasson <philipel@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#29820}
TBR=philipel@webrtc.org
Bug: webrtc:10809
Change-Id: Id8196d9348d7fa69a5e410367b8a88e6039ef1b8
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160205
Commit-Queue: Erik Språng <sprang@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29867}
Change #159711 adds the option to filter out small packets on the
input to the delay-based BWE. This change adds similar functionality
to BitrateEstimator by reducing the weight of small observations.
Bug: webrtc:10932
Change-Id: I0a673a067f7ef86769cabd30443e60e9de70053c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160009
Commit-Queue: Christoffer Rodbro <crodbro@webrtc.org>
Reviewed-by: Björn Terelius <terelius@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29865}
ADM2 for Windows is based on the CoreAudioUtil class in Chrome.
CoreAudioUtil in Chrome does not use a special string to identify
the Default Communication device but instead a combination of a
string (Default) and a role parameter [1].
When CoreAudioUtil was ported to WebRTC, I accidentally added an
invalid usage of a unique string to identify the default comm device
and it can lead to errors since there are then two different ways to
identify this device. It will also complicate life when we want to
merge changes from Chrome into WebRTC.
This CL removes usage of AudioDeviceName::kDefaultCommunicationsDeviceId
in WebRTC to reduce the risk of errors.
[1] https://cs.chromium.org/chromium/src/media/audio/win/core_audio_util_win.cc?q=core_audio_ut&sq=package:chromium&g=0&l=464
Excluding flaky bot win_x86_msvc_dbg and using Tbr.
Tbr: thaloun@chromium.org
No-Try: True
Bug: webrtc:11107
Change-Id: Ie6687adbe9c3940a217456e4025967f71d86214c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160047
Commit-Queue: Henrik Andreassson <henrika@webrtc.org>
Reviewed-by: Henrik Andreassson <henrika@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29848}
A bug currently causes the packet queue to not get any trials enabled
unless an injected key value map is used.
Bug: None
Change-Id: I5c21aa296e8a202a63e81a57c5d13297ad7333bd
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160012
Reviewed-by: Christoffer Rodbro <crodbro@webrtc.org>
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29845}
Keeps behavior for old periodic processing.
Rounding sleep time reduced chance for small bursts of busy-looping when
time approaches 0.
Also fixes a DCHECK which may trigger if there are rounding errors in
the timing.
Bug: webrtc:10809
Change-Id: Iba8450f906fd6ab3b1da97e04507b16ac6bbde3f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160000
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29841}
It's possible for an device to be counted but getting its name fails, in which case the utility function returns true but would continue from its loop filling the AudioDeviceNames vector, leading to a smaller output than the later code expects.
No-Try: True
Bug: b/144729866
Change-Id: If902cada4ef2911bc24fbec0f169da75ff6e6a83
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160020
Commit-Queue: Henrik Andreassson <henrika@webrtc.org>
Reviewed-by: Henrik Andreassson <henrika@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29840}
On top of adding unittests for the remixing, the CL
moves the code tested to a separate file in order
to allow it to be tested.
Bug: webrtc:11007
Change-Id: I531736517bbcc715b3c1bf3a4256c42208c5b778
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/155740
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Karl Wiberg <kwiberg@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29839}
This CL refactors the analog AGC functionality. In particular it:
-Breaks then tight dependency between the analog AGC and the digital
AGC implementation.
-Removes the complicated callback interface for reporting the analog
level and replaces it with an int.
Bug: webrtc:10859
Change-Id: I3572d60ab98edebbcffa25af64cc74c66f9868fc
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/159039
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29838}
This reverts commit acdc22d7845c5dde7c23366110e54e5d26127c85.
Reason for revert: Field trials are not enabled in the same way, will reland after that is fixed.
Original change's description:
> Prepares PacingController for simplified packet queue.
>
> This CL removes references to RoundRobinPacketQueue::QueuedPacket,
> other than the method to release an RtpPacketToSend. It also moves
> both the BeginPop() and FinalizePop() to within a single helper
> method.
>
> A follow-up cleanup of the packet queue will stop exposing the
> QueuedPacket struct and replaces the the pop-methods with a single
> new one that just returns an RtpPacketToSend.
>
> Bug: webrtc:10809
> Change-Id: I5208a93e12e6b56714d483cc12d2a37225ea8e5e
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/159889
> Commit-Queue: Erik Språng <sprang@webrtc.org>
> Reviewed-by: Philip Eliasson <philipel@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#29820}
TBR=sprang@webrtc.org,philipel@webrtc.org
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: webrtc:10809
Change-Id: I02fccbfbba6b9670b0ce2008e067df3aa9d3c5f2
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/160010
Reviewed-by: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29836}
The reported audio interruption metrics are too high. If GetAudio
calls start before the first packets are arriving, and the sample rate
of the encoded audio is different from the one used to initialize
NetEq (default 16 kHz), the initial silent period of GetAudio calls
will be reported as an interruption.
Modifying a unit test to trigger the bug, and make sure it won't come
back.
Bug: webrtc:11094, b/144567257
Change-Id: Id540422cb7f35d3bef68b9e7c03c6e7c8bdb8d97
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/159980
Commit-Queue: Henrik Lundin <henrik.lundin@webrtc.org>
Reviewed-by: Karl Wiberg <kwiberg@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29831}
New tests are:
- AudioDeviceTest.StartStopPlayoutWithRealDevice
- AudioDeviceTest.StartStopRecordingWithRealDevice
(the comments below only affects ADM2 on Windows):
When adding these tests it was found that we could hit the same known issue
as in https://bugs.chromium.org/p/chromium/issues/detail?id=803056 and the
same solution as in Chrome was therefore ported from Chrome to WebRTC.
Hence, this change also adds support for core_audio_utility::WaveFormatWrapper
to support devices that can return a format where only the WAVEFORMATEX parts is
initialized. The old version would only DCHECK for these devices and that could
lead to an unpredictable behavior.
Tbr: minyue
Bug: webrtc:11093
Change-Id: Icb238c5475100f251ce4e55e39a03653da04dbda
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/159982
Reviewed-by: Henrik Andreassson <henrika@webrtc.org>
Commit-Queue: Henrik Andreassson <henrika@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29824}
Some of the field trial default values are changed as well.
Now available bitrate estimation will be decreasing when RTT is more than 3 seconds.
Unless different parameters for the field trial are specified.
Bug: None
Change-Id: Icd1923fc2e2e7766a7f645016c5432a52537145f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/158840
Reviewed-by: Sebastian Jansson <srte@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Reviewed-by: Konrad Hofbauer <hofbauer@webrtc.org>
Reviewed-by: Christoffer Rodbro <crodbro@webrtc.org>
Commit-Queue: Nikita Zetilov <zetilovn@google.com>
Cr-Commit-Position: refs/heads/master@{#29823}
This CL removes references to RoundRobinPacketQueue::QueuedPacket,
other than the method to release an RtpPacketToSend. It also moves
both the BeginPop() and FinalizePop() to within a single helper
method.
A follow-up cleanup of the packet queue will stop exposing the
QueuedPacket struct and replaces the the pop-methods with a single
new one that just returns an RtpPacketToSend.
Bug: webrtc:10809
Change-Id: I5208a93e12e6b56714d483cc12d2a37225ea8e5e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/159889
Commit-Queue: Erik Språng <sprang@webrtc.org>
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#29820}
