Denormal numbers (see [1]) may origin in APM when the input is zeroed
after a non-zero signal. In extreme cases, instructions involving
denormal operands may run as much as 100 times slower, which seems to
be the case (to some extent) of crbug.com/1227566.
This CL adds a class that disables denormals only via hardware on x86
and on ARM. The class is used in APM and it is an adaption of [2].
Tested: appr.tc call on Chromium (Win, Mac)
[1] https://en.wikipedia.org/wiki/Denormal_number
[2] https://source.chromium.org/chromium/chromium/src/+/main:third_party/blink/renderer/platform/audio/denormal_disabler.h
Fixed: chromium:1227566
Change-Id: I0ed2eab55dc597529f09f93c26c7a01de051fdbe
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/227768
Reviewed-by: Magnus Flodman <mflodman@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Commit-Queue: Alessio Bazzica <alessiob@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34701}
Acts as a compile time annotation, with corresponding run-time check
only when DCHECKs are enabled, and built using absl or pthreads mutexes.
Bug: None
Change-Id: Ie044c1ea1e576df71d634301f7df9d75cdf10b1b
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/226328
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Reviewed-by: Ivo Creusen <ivoc@webrtc.org>
Commit-Queue: Niels Moller <nisse@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34555}
As part of go/coil update code search links to not point to the
"master" branch.
Bug: chromium:1226942
Change-Id: I0ae9e84ecc660f789a69fe0b226f93bbc39a8a66
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/226081
Commit-Queue: Tony Herre <toprice@chromium.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34531}
Add a histogram WebRTC.Audio.Agc.InputClippingRate and logging of
max clipping rate in AgcManagerDirect.
Bug: webrtc:12774
Change-Id: I4a72119b65ad032fc50672e2a8fb4a4d55e1ff24
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/225264
Commit-Queue: Hanna Silen <silen@webrtc.org>
Reviewed-by: Minyue Li <minyue@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34450}
Integrate ClippingPredictorEvaluator in AgcManagerDirect adding the
possibility to run the predictor without affecting the analog gain
adjustment process.
The evaluator is used to compute precision, recall and F1 score.
F1 score and the measured clipping prediction intervals are logged as
`WebRTC.Audio.Agc.ClippingPredictor.F1Score` and `.PredictionInterval`
histograms respectively.
Bug: webrtc:12774
Change-Id: I708dcda9321f92d5bd17ec4c36ebce1165ead57f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/221921
Commit-Queue: Alessio Bazzica <alessiob@webrtc.org>
Reviewed-by: Hanna Silen <silen@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34327}
The dominant nearend detector uses the residual echo spectrum for
determining whether in nearend state. The residual echo spectrum in
computed using the ERLE. To reduce the risk of echo leaks in the
suppressor, the ERLE is capped. While minimizing echo leaks, the
capping of the ERLE can affect the dominant nearend classification
negatively as the residual echo spectrum is often over estimated.
This change enables the dominant nearend detector to use a residual
echo spectrum computed with a virtually non-capped ERLE. This ERLE
is only used for dominant nearend detection and leads to increased
transparency.
The feature is currently disabled by default and can be enabled
with the field trial "WebRTC-Aec3UseUnboundedEchoSpectrum".
Bug: webrtc:12870
Change-Id: Icb675c6f5d42ab9286e623b5fb38424d5c9cbee4
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/221920
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34270}
This change enables the use of two different adaptation speeds of the
matched filter of the delay estimator of AEC3.
One speed is used when no delay has been found, and one is used after a
reliable delay has been found. The purpose is to use a slower adaptation
speed to reduce the risk of divergence during double-talk without
slowing down the search for the initial delay.
The CL prepares for experimentation by adding field trials for
controlling the two adaptation speeds.
Bug: webrtc:12775
Change-Id: I817a1ab5ded0f78d20de45edcf04c708290173fc
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/219083
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34055}
This change adds the field trial "WebRTC-TransientSuppressorForcedOff"
that can be used to disable the transient suppressor (removal of
keyboard typing sounds). The field trial can be enabled by users via
command-line or via experimentation.
Bug: chromium:1186705
Change-Id: I7272df6a20fbbee24a7ba0904502c76bd775d275
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/219282
Reviewed-by: Per Åhgren <peah@webrtc.org>
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34038}
Update `AudioProcessing::Config::ToString()` to also dump the config
from `AnalogGainController` which is missing.
Bug: webrtc:7494
Change-Id: Iea5dab1f6abb9ec8581ce690a2a119f202b4d1e1
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/219082
Reviewed-by: Per Åhgren <peah@webrtc.org>
Commit-Queue: Alessio Bazzica <alessiob@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34025}
This reverts commit 793bac569f.
Reason for revert: rare compilation error fixed
Original change's description:
> Revert "Refactor the PlatformThread API."
>
> This reverts commit c89fdd716c.
>
> Reason for revert: Causes rare compilation error on win-libfuzzer-asan trybot.
> See https://ci.chromium.org/p/chromium/builders/try/win-libfuzzer-asan-rel/713745?
>
> Original change's description:
> > Refactor the PlatformThread API.
> >
> > PlatformThread's API is using old style function pointers, causes
> > casting, is unintuitive and forces artificial call sequences, and
> > is additionally possible to misuse in release mode.
> >
> > Fix this by an API face lift:
> > 1. The class is turned into a handle, which can be empty.
> > 2. The only way of getting a non-empty PlatformThread is by calling
> > SpawnJoinable or SpawnDetached, clearly conveying the semantics to the
> > code reader.
> > 3. Handles can be Finalized, which works differently for joinable and
> > detached threads:
> > a) Handles for detached threads are simply closed where applicable.
> > b) Joinable threads are joined before handles are closed.
> > 4. The destructor finalizes handles. No explicit call is needed.
> >
> > Fixed: webrtc:12727
> > Change-Id: Id00a0464edf4fc9e552b6a1fbb5d2e1280e88811
> > Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215075
> > Commit-Queue: Markus Handell <handellm@webrtc.org>
> > Reviewed-by: Harald Alvestrand <hta@webrtc.org>
> > Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
> > Reviewed-by: Tommi <tommi@webrtc.org>
> > Cr-Commit-Position: refs/heads/master@{#33923}
>
> # Not skipping CQ checks because original CL landed > 1 day ago.
>
> TBR=handellm@webrtc.org
>
> Bug: webrtc:12727
> Change-Id: Ic0146be8866f6dd3ad9c364fb8646650b8e07419
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/217583
> Reviewed-by: Guido Urdaneta <guidou@webrtc.org>
> Reviewed-by: Markus Handell <handellm@webrtc.org>
> Commit-Queue: Guido Urdaneta <guidou@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#33936}
# Not skipping CQ checks because this is a reland.
Bug: webrtc:12727
Change-Id: Ifd6f44eac72fed84474277a1be03eb84d2f4376e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/217881
Commit-Queue: Mirko Bonadei <mbonadei@webrtc.org>
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Reviewed-by: Markus Handell <handellm@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33950}
This reverts commit c89fdd716c.
Reason for revert: Causes rare compilation error on win-libfuzzer-asan trybot.
See https://ci.chromium.org/p/chromium/builders/try/win-libfuzzer-asan-rel/713745?
Original change's description:
> Refactor the PlatformThread API.
>
> PlatformThread's API is using old style function pointers, causes
> casting, is unintuitive and forces artificial call sequences, and
> is additionally possible to misuse in release mode.
>
> Fix this by an API face lift:
> 1. The class is turned into a handle, which can be empty.
> 2. The only way of getting a non-empty PlatformThread is by calling
> SpawnJoinable or SpawnDetached, clearly conveying the semantics to the
> code reader.
> 3. Handles can be Finalized, which works differently for joinable and
> detached threads:
> a) Handles for detached threads are simply closed where applicable.
> b) Joinable threads are joined before handles are closed.
> 4. The destructor finalizes handles. No explicit call is needed.
>
> Fixed: webrtc:12727
> Change-Id: Id00a0464edf4fc9e552b6a1fbb5d2e1280e88811
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215075
> Commit-Queue: Markus Handell <handellm@webrtc.org>
> Reviewed-by: Harald Alvestrand <hta@webrtc.org>
> Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
> Reviewed-by: Tommi <tommi@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#33923}
# Not skipping CQ checks because original CL landed > 1 day ago.
TBR=handellm@webrtc.org
Bug: webrtc:12727
Change-Id: Ic0146be8866f6dd3ad9c364fb8646650b8e07419
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/217583
Reviewed-by: Guido Urdaneta <guidou@webrtc.org>
Reviewed-by: Markus Handell <handellm@webrtc.org>
Commit-Queue: Guido Urdaneta <guidou@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33936}
PlatformThread's API is using old style function pointers, causes
casting, is unintuitive and forces artificial call sequences, and
is additionally possible to misuse in release mode.
Fix this by an API face lift:
1. The class is turned into a handle, which can be empty.
2. The only way of getting a non-empty PlatformThread is by calling
SpawnJoinable or SpawnDetached, clearly conveying the semantics to the
code reader.
3. Handles can be Finalized, which works differently for joinable and
detached threads:
a) Handles for detached threads are simply closed where applicable.
b) Joinable threads are joined before handles are closed.
4. The destructor finalizes handles. No explicit call is needed.
Fixed: webrtc:12727
Change-Id: Id00a0464edf4fc9e552b6a1fbb5d2e1280e88811
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215075
Commit-Queue: Markus Handell <handellm@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Reviewed-by: Tommi <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33923}
Add the option to run the adaptive digital controller of AGC2 without
side-effects - i.e., no gain applied.
Tested: adapation verified during a video call in chromium
Bug: webrtc:7494
Change-Id: I4776f6012907d76a17a3bca89991da97dc38657f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215964
Commit-Queue: Alessio Bazzica <alessiob@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33875}
This is essentially replacing `new rtc::RefCountedObject` with
`rtc::make_ref_counted` in many files. In a couple of places I
made minor tweaks to make things compile such as adding parenthesis
when they were missing.
Bug: webrtc:12701
Change-Id: I3828dbf3ee0eb0232f3a47067474484ac2f4aed2
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215973
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Commit-Queue: Tommi <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33852}
The change introduces support for detachable PlatformThreads, for which
the Stop() call doesn't wait until the thread has finished executing.
The change also introduces rtc::ThreadAttributes that carries priority
and detachability thread attributes. It additionally refactors all
known use to use the new semantics.
Bug: b:181572711, webrtc:12659
Change-Id: Id96e87c2a0dafabc8047767d241fd5da4505d14c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/214704
Reviewed-by: Tommi <tommi@webrtc.org>
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Commit-Queue: Markus Handell <handellm@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33796}
The ERLE is used to estimate residual echo for echo suppression. The
ERLE is reduced during far-end offset to avoid echo leakage. When there
is a strong near-end present this can cause unnecessary transparency loss.
This change adds an ERLE estimation that does not compensate for onsets and
uses it for residual echo estimation when the suppressor considers the near-end to be dominant.
Bug: webrtc:12686
Change-Id: Ida78eeacf1f95c6e62403f86ba3f2ff055898a84
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215323
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33786}
- Bug fix: the desired initial gain quickly dropped to 0 dB hence
starting a call with a too low level
- New tuning to make AGC2 more robust to VAD mistakes
- Smarter max gain increase speed: to deal with an increased threshold
of adjacent speech frames, the gain applier temporarily allows a
faster gain increase to deal with a longer time spent waiting for
enough speech frames in a row to be observed
- Saturation protector isolated from `AdaptiveModeLevelEstimator` to
simplify the unit tests for the latter (non bit-exact change)
- AGC2 adaptive digital config: unnecessary params deprecated
- Code readability improvements
- Data dumps clean-up and better naming
Bug: webrtc:7494
Change-Id: I4e36059bdf2566cc2a7e1a7e95b7430ba9ae9844
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215140
Commit-Queue: Alessio Bazzica <alessiob@webrtc.org>
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33736}
The current noise level estimator has a bug due to which the estimated
level decays to the lower bound in a few seconds when speech is observed.
Instead of fixing the current implementation, which is based on a
stationarity classifier, an alternative, lightweight, noise floor
estimator has been added and tuned for AGC2.
Tested on several AEC dumps including HW mute, music and fast talking.
Bug: webrtc:7494
Change-Id: Iae4cff9fc955a716878f830957e893cd5bc59446
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/214133
Commit-Queue: Alessio Bazzica <alessiob@webrtc.org>
Reviewed-by: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33733}
Erle Uncertainty changes the residual echo computation during saturated
echo. However, the case of saturated echo is already handled by the
residual echo estimator causing the ErleUncertainty to be a no-op.
The change has been tested for bit-exactness.
Bug: webrtc:8671
Change-Id: I779ba67f99f29d4475a0465d05da03d42d50e075
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215072
Reviewed-by: Jesus de Vicente Pena <devicentepena@webrtc.org>
Commit-Queue: Gustaf Ullberg <gustaf@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33719}
Done in preparation for the child CL which adds an alternative
implementation.
Bug: webrtc:7494
Change-Id: I4963376afc917eae434a0d0ccee18f21880eefe0
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/214125
Reviewed-by: Jakob Ivarsson <jakobi@webrtc.org>
Commit-Queue: Alessio Bazzica <alessiob@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33646}
Code style improvements done in preparation for a bug fix (TODO added)
which requires changes in the unit tests.
Note that one expected value in the unit tests has been adjusted since
the white noise generator is now instanced in each separate test and
therefore, even if the seed remained the same, the generated sequences
differ.
Bug: webrtc:7494
Change-Id: I497513b84f50b5c66cf6241a09946ce853eb1cd2
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/214122
Commit-Queue: Alessio Bazzica <alessiob@webrtc.org>
Reviewed-by: Ivo Creusen <ivoc@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33636}
This is a reland of aa6adffba3
What was changed in the reland is that the merging of the bands is
excluded from the code that is not run when the output is not used.
I.e., the merging is always done.
This is important to have since some clients may apply muting before APM,
and still flag to APM that the signal is muted. If the merging is not
always done, those clients will get nonzero output from APM during muting.
Original change's description:
> Reduce complexity in the APM pipeline when the output is not used
>
> This CL selectively turns off parts of the audio processing when
> the output of APM is not used. The parts turned off are such that
> don't need to continuously need to be trained, but rather can be
> temporarily deactivated.
>
> The purpose of this CL is to allow CPU to be reduced when the
> client is muted.
>
> The CL will be follow by additional CLs, adding similar functionality
> in the echo canceller and the noiser suppressor
>
> Bug: b/177830919
> Change-Id: I72d24505197a53872562c0955f3e7b670c43df6b
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/209703
> Commit-Queue: Per Åhgren <peah@webrtc.org>
> Reviewed-by: Sam Zackrisson <saza@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#33431}
Bug: b/177830919
Change-Id: Ib74dd1cefa173d45101e26c4f2b931860abc6d08
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/211760
Reviewed-by: Sam Zackrisson <saza@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33478}
This CL adds functionality in the noise suppressor that allows the
computational complexity to be reduced when the output of APM is not used.
Bug: b/177830919
Change-Id: I849351ba9559fae770e4667d78e38abde5230eed
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/211342
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33477}
This CL adds functionality in AEC3 that allows the computational
complexity to be reduced when the output of APM is not used.
Bug: b/177830919
Change-Id: I08121364bf966f34311f54ffa5affbfd8b4db1e2
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/211341
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Commit-Queue: Per Åhgren <peah@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33476}
This CL adds functionality that allows adjusting the audio levels
internally in APM. The main purpose of the functionality is to allow
APM to optionally be moved to an integration that does not provide an
analog gain to control, and the implementation of this has been
tailored specifically to meet the requirements for that.
More specifically, this CL does
-Add a new variant of the pre-amplifier gain that is intended to replace
the pre-amplifier gain (but at the moment can coexist with that). The
main differences with the pre-amplifier gain is that an attenuating
gain is allowed, the gain is applied jointly with any emulated analog
gain, and that its packaging fits better with the post gain.
-Add an emulation of an analog microphone gain. The emulation is
designed to match the analog mic gain functionality in Chrome OS (which
is digital) but should be usable also on other platforms.
-Add a post-gain which is applied after all processing has been applied.
The purpose of this gain is for it to work well with the integration
in ChromeOS, and be used to compensate for the offset that there is
applied on some USB audio devices.
Bug: b/177830918
Change-Id: I0f312996e4088c9bd242a713a703eaaeb17f188a
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/209707
Commit-Queue: Per Åhgren <peah@webrtc.org>
Reviewed-by: Gustaf Ullberg <gustaf@webrtc.org>
Reviewed-by: Alessio Bazzica <alessiob@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33466}
