Some of the state that's managed in VideoStreamEncoder, is updated
and accessed on the encoder queue, but a method that's used for testing
only (GetAdaptationResources()), represents a race between PostTask
operations that update state and checking for said state on the worker
thread.
This CL removes Wait() operations related to adaptation resources from
the common path and puts one in the test path instead.
Bug: webrtc:13612
Change-Id: Ie3e018e815e24951bc0634ed70de17eaf336a508
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/249220
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Tomas Gunnarsson <tommi@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35797}
Resource adaptation needs refactoring for async adaptations. For now
the resource adaptation processor can work on the encoder thread, until
it is refactored to support async adaptation.
Bug: webrtc:11867
Change-Id: I9c46da356db19c0fd52748c999ccb216f2ca923b
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/182040
Commit-Queue: Evan Shrubsole <eshr@google.com>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31991}
This is in preperation for eventual multi-stream and multi-mitigation
adaptation. This logic only applied to a single stream and thus is
better fit in the VideoStreamAdapter.
Bug: webrtc:11754
Change-Id: Icc5c7920038c82b574f4b5f7efbc92698691076f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/181585
Commit-Queue: Evan Shrubsole <eshr@google.com>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31944}
This occurs when a resource causes an adaptation down but the current
adaptations can not be adapted any more. Any further adaptation will result in the status kLimitReached,
and so any resource that adapts down should also be most limited.
Bug: webrtc:11695
Change-Id: Idfdf23f482b1b4a132cec49a9be76adc0aec4361
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/181586
Commit-Queue: Evan Shrubsole <eshr@google.com>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31933}
It was not used by any class and all future uses can use the
VideoSourceRestrictionsListener.
Bug: webrtc:11834
Change-Id: I5c71b93cc503f458dce0ccdd78b91b5a1debc56d
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/181062
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Evan Shrubsole <eshr@google.com>
Cr-Commit-Position: refs/heads/master@{#31896}
This is a step needed for multi-stream and new mitigations. It also
cleans up needing to signal adaptation changes in mutiple places
from ResourceAdaptationProcessor.
R=hbos@webrtc.org
Bug: webrtc:11754
Change-Id: Ib185dc9f66fbb4a087eb9e970c68c3f47eafb17f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/178874
Commit-Queue: Evan Shrubsole <eshr@google.com>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31684}
Changes in effective degradation preference is now done directly
on the VideoStreamAdapter. This means that the
DegradationPreferenceListener class can be removed.
Bug: webrtc:11700
Change-Id: I9ce4b627de0279df5ef2e5a38435a6d352fbb135
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/178609
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Evan Shrubsole <eshr@google.com>
Cr-Commit-Position: refs/heads/master@{#31642}
This patch creates a new class which provides the DegradationPreference
thread safe to all classes that need if (BalancedConstraint and
QpScalerResource). It also broadcasts to all listeners when the
preferences are updated, so the ResourceAdaptationProcessor can update
the video if needed.
In future work, this could be used to remove the need for two task
queues for the VideoStreamEncoder resources.
Bug: webrtc:11700
Change-Id: I05480db8b7157b5643f6f86ec9c64850839b3e76
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/177522
Commit-Queue: Evan Shrubsole <eshr@google.com>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31623}
This moves this responsibility from the ResourceAdaptaitonProcessor
to the VideoStreamAdapter. A new interface for listening to adaptation
changes was added, and the ResourceAdaptationProcessor now listens on
the VideoStreamAdapter for those changes.
This means that going forward,
1. We can do adaptations outside of resource limitations, like setting
prior adaptations on a resource like initial frame dropper is designed
to.
2. Adaptations can be on a different queue than the
ResourceAdaptaitonProcessor's queue since updates are pushed through
the listener.
Bug: webrtc:11700
Change-Id: I6de0dec748dba095e702f0b9893c5afa50b51aa9
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/176859
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Evan Shrubsole <eshr@google.com>
Cr-Commit-Position: refs/heads/master@{#31615}
When a resource is removed from the ResourceAdaptationProcessor,
and the resource is the most limited resource, we reset the current
adaptation to the next most limited resource level. In the case that
this resource is the only adapted resource then we reset all adaptations.
Bug: webrtc:11636
Change-Id: I29acc5a3934f42f00db79b3bb2171156b1313937
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/176406
Commit-Queue: Evan Shrubsole <eshr@google.com>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31520}
A more detailed explaination is in the bug, but this changes
the way that adaptation happens when multiple resources are
limited. Only the one that is most limited can trigger an
adaptation up. If multiple resources are most limited both
need to underuse to adapt up.
Some of the changes in this patch to make it all work:
* VideoStreamEncoder unittests that did not reflect this
new behaviour have been changed.
* PeekNextRestrictions returns the adaptation counters as
well as the restrictions.
* Adaptation statstics have changed so that when adapting
up all resources are tagged as triggering the adaptation.
Additionally the statistics for the current adaptation is
now the total number of adaptations per reason, rather then
the number of adaptations due to that reason.
* PreventAdaptUpDueToActiveCounts is removed as most limited
resource is a strong implementation of that.
Bug: webrtc:11553
Change-Id: If1545a201c8e019598edf82657a1befde8b05268
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/176128
Commit-Queue: Evan Shrubsole <eshr@google.com>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31497}
Use the current task queue in the test instead.
R=hbos@webrtc.org
Bug: None
Change-Id: I266796df3de675a808bcf2da08901411fadc86ae
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/176414
Commit-Queue: Evan Shrubsole <eshr@google.com>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31439}
This is a prerequisite to implementing it externally.
Bug: webrtc:11525
Change-Id: I9cb3b4418396485d3eb9f25cafa51cbff6db7817
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/176401
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Stefan Holmer <stefan@webrtc.org>
Reviewed-by: Evan Shrubsole <eshr@google.com>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31418}
ResourceListener::OnResourceUsageStateMeasured() now takes
ResourceUsageState as argument, making Resource::UsageState()
superfluous.
With the existing "fire-and-forget" behavior of always clearing usage
state on reacting to a signal, there is no longer a need to call
ClearUsageState() so this too is removed. (We may want to have a
callback in the future to hint to the Resource that it is a good idea
to clear internal measurement samples, i.e. because the load of the
system is about to change, but we can revisit that when we need it.)
Moving the usage state to the callback has the benefit of getting rid
of the assumption that UsageState() has to return the same value every
time it is called in the same task.
This CL is also the final nail in the coffin for Resource needing to
know about the adaptation task queue: ResourceAdaptationProcessor's
ResourceListener now takes care of posting to the adaptation task
queue. To support this, the processor's SequenceChecker is replaced
by a TaskQueueBase pointer.
Bug: webrtc:11525, webrtc:11618
Change-Id: I2277e71cc3759c85b62465020935603f03792c94
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/176376
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Evan Shrubsole <eshr@google.com>
Cr-Commit-Position: refs/heads/master@{#31416}
IsAdaptationUpAllowed is moved from Resource to AdaptationConstraint.
OnAdaptationApplied is moved from Resource to AdaptationListener.
In a future CL, Resource will be moved to api/, but
AdaptationConstraint and AdaptationListener will stay in call/.
The processor, encode stream and manager are updated to keep track of
both resources, constraints and listeners. Fakes and tests are updated.
After this CL, the manager's inner classes that prevent adaptation
implement AdaptationConstraint instead of Resource.
Bug: webrtc:11525
Change-Id: Ie9cd5b1ba7d8e161951e131ab8f6bd9d5cf765bf
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/176368
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Reviewed-by: Evan Shrubsole <eshr@google.com>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31409}
This CL is in preparation for moving Resource to the api/ folder. It
does not move it, but makes it such that the moving CL can be a pure
move.
In order to do this, we must stop depending on rtc_base/rtc::TaskQueue
in favor of api/webrtc::TaskQueueBase.
There are also other rtc_base/ dependencies that we do not want to
expose to the api/ folder, like critical sections and thread
annotations which are not publically exposed. To get around this, we
make Resource an abstract interface and move all of the base class
functionality into a new non-api/ class: VideoStreamEncoderResource.
The Resource now has Register/UnregisterAdaptationTaskQueue() methods.
By explicitly unregistering, we can ensure validity of the pointer even
if the Resource outlives the PeerConnection. While public interface
methods are only to be called on the adaptation task queue, posting to
the task queue happens off-queue, so a |lock_| is introduced to guard
it.
Bug: webrtc:11525
Change-Id: I50b3a30960cdec9032016c779b47001c01dad32f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/176320
Reviewed-by: Evan Shrubsole <eshr@google.com>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31402}
This CL unblocks future Call-Level Mitigation strategies by moving the
ResourceAdaptationProcessor to a separate task queue. This signifies a
major milestone in the new resource adaptation architecture because
with this CL the threading model is in place and moving the Processor
to the Call and increasing its responsibilities is made possible.
In this CL, we still have one Processor per VideoStreamEncoder and the
VideoStreamEncoder is responsible for the creation and the destruction
of its Processor and that Processor's task queue. But the PostTasks are
in place and the decision-making is executed on a separate queue.
This CL:
- Moves ResourceAdaptationProcessor to an adaptation task queue.
It continues to be entirely single-threaded, but now operates on a
separate task queue.
- Makes Resources thread-safe: Interaction with the Processor, i.e.
OnResourceUsageStateMeasured() and IsAdaptationUpAllowed(), happens
on the adaptation task queue. State updates are pushed from the
encoder task queue with PostTasks.
- QualityScalerResource operates on both task queues; the QP usage
callbacks are invoked asynchronously.
- The VideoStreamEncoderResourceManager operates on the encoder task
queue with the following exceptions:
1) Its resources are accessible on any thread (using a mutex). This
is OK because resources are reference counted and thread safe.
This aids adding and removing resources to the Processor on the
adaptation task queue.
2) |active_counts_| is moved to the adaptation task queue. This makes
it possible for PreventAdaptUpDueToActiveCounts to run
IsAdaptationUpAllowed() on the adaptation task queue.
A side-effect of this is that some stats reporting now happen on
the adaptation task queue, but that is OK because
VideoStreamEncoderObserver is thread-safe.
The Manager is updated to take the new threading model into account:
- OnFrameDroppedDueToSize() posts to the adaptation task queue to
invoke the Processor.
- OnVideoSourceRestrictionsUpdated(), now invoked on the adaptation
task queue, updates |active_counts_| synchronously but posts to the
encoder task queue to update video source restrictions (which it
only uses to calculate target frame rate).
- MaybePerformQualityRampupExperiment() posts to the adaptation task
queue to maybe reset video source restrictions on the Processor.
|quality_rampup_done_| is made std::atomic.
Bug: webrtc:11542, webrtc:11520
Change-Id: I1cfd76e0cd42f006a6d2527f5aa2aeb5266ba6d6
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/174441
Reviewed-by: Evan Shrubsole <eshr@google.com>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31231}
In a future CL, adaptation processing and stream encoder resource
management will happen on different task queues. When this is the case,
asynchronous tasks will be posted in both directions and some resources
will have internal states used on multiple threads.
This CL makes the Resource class reference counted in order to support
posting tasks to a different threads without risk of use-after-free
when a posted task is executed with a delay. This is preferred over
WeakPtr strategies because WeakPtrs are single-threaded and preferred
over raw pointer usage because the reference counted approach enables
more compile-time and run-time assurance. This is also "future proof";
when resources can be injected through public APIs, ownership needs to
be shared between libwebrtc and the application (e.g. Chrome).
To reduce the risk of making mistakes in the future CL, sequence
checkers and task queue DCHECKs are added as well as other DCHECKs to
make sure things have been cleaned up before destruction, e.g:
- Processor gets a sequence checker. It is entirely single-threaded.
- Processor must not have any attached listeners or resources on
destruction.
- Resources must not have any listeners on destruction.
- The Manager, EncodeUsageResource and QualityScalerResource DCHECKs
they are running on the encoder queue.
- TODOs are added illustrating where we want to add PostTasks in the
future CL.
Lastly, upon VideoStreamEncoder::Stop() we delete the
ResourceAdaptationProcessor. Because the Processor is already used in
posted tasks, some if statements are added to ensure the Processor is
not used after destruction.
Bug: webrtc:11542, webrtc:11520
Change-Id: Ibaa8a61d86d87a71f477d1075a117c28d9d2d285
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/174760
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Evan Shrubsole <eshr@google.com>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31217}
The problem is that a resource that signals Underuse would be
able to trigger an adapt up when it was never limited in the past.
This means that an underused resource would be able to negate the
adaptations made for an overused one.
For example, consider a fast CPU on a bad link. The QP for the image
is high but the CPU is underused. Without requiring previous underuse,
everytime the QP would signal overuse and trigger an adpatation down,
the CPU would signal underuse and trigger an adaptation up.
This works today as we want by using the active counts in the
VideoStreamEncoderResourceManager. This change
makes it a normal behaviour independant of active counts.
The problem with active counts is that is only works with 2 resources.
When resources are injectable it no longer works as expected.
Bug: webrtc:11522, webrtc:11523
Change-Id: I140636ce206d74e00a6b6f8558162bb8afffda1c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/174482
Commit-Queue: Evan Shrubsole <eshr@google.com>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31181}
This CL adds miscellaneous unit tests for the
ResourceAdaptationProcessor, the VideoSourceRestrictions comparators and
the VideoStreamInputStateProvider.
Bug: webrtc:11172
Change-Id: If95f69644aaf2b43e3b19d5729bedef0b438c77b
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/174101
Reviewed-by: Evan Shrubsole <eshr@google.com>
Reviewed-by: Ilya Nikolaevskiy <ilnik@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#31147}
The Resource interface (previously a skeleton not used outside of
testing) is updated to inform listeners of changes to resource
usage. Debugging methods are removed (Name, UsageUnitsOfMeasurements,
CurrentUsage). The interface is implemented by
OveruseFrameDetectorResourceAdaptationModule's inner classes
EncodeUsageResource and QualityScalerResource.
The new ResourceUsageListener interface is implemented by
OveruseFrameDetectorResourceAdaptationModule. In order to avoid adding
AdaptationObserverInterface::AdaptReason to the ResourceUsageListener
interface, the module figures out if the reason is "kCpu" or "kQuality"
by looking which Resource object triggered
OnResourceUsageStateMeasured(). These resources no longer need an
explicit reference to OveruseFrameDetectorResourceAdaptationModule and
could potentially be used by a different module.
In this CL, AdaptationObserverInterface::AdaptDown()'s return value is
still needed by QualityScaler. This is mirrored in the return value of
ResourceUsageListener::OnResourceUsageStateMeasured(). A TODO is added
to remove it and a comment explains how the current implementation
seems to break the contract of the method (as was the case prior to
this CL).
Follow-up work include:
- Move EncodeUsageResource and QualityScalerResource to separate files.
- Make resources injectable, allowing fake resources in testing and
removing OnResourceOveruseForTesting() methods.
(Investigate adding the necessary input signals to the Resource
interface or relevant sub-interfaces so that the module does not need
to know which Resource implementation is used.)
- And more! See whiteboard :)
Bug: webrtc:11222
Change-Id: I0a46ace4a2e617874e3ee97e67e3a199fef420a2
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/168180
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Reviewed-by: Evan Shrubsole <eshr@google.com>
Cr-Commit-Position: refs/heads/master@{#30469}
This CL adds Resource, ResourceConsumer, ResourceConsumerConfiguration
and ResourceAdaptationProcessor and implements the algorithm outlined
in
https://docs.google.com/presentation/d/13jyqCWNpIa873iKT6yDuB5Q5ma-c0CvxBpX--0tCclY/edit?usp=sharing.
Simply put, if any resource (such as "CPU") is overusing, the most
expensive consumer (e.g. encoded stream) is adapted one step down.
If all resources are underusing, the least expensive consumer is
adapted one step up.
The current resources, consumers and configurations are all fakes;
this CL has no effect on the current adaptation algorithms used in
practise, but it lays down the foundation for future work in this
area.
Bug: webrtc:11167, webrtc:11168, webrtc:11169
Change-Id: I4054ec7728a52a49e137eee6fa67fa27debd9254
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/161237
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Evan Shrubsole <eshr@google.com>
Reviewed-by: Stefan Holmer <stefan@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#30053}
