Update `early_execute_margin` after process packets, and the test case.
Original change's description:
>Pacer: Reduce TQ wake up and improve packet size estimation
>
>The TQ Pacer schedules delayed task according to target time of
>PacingController. It drains all valid ProcessPackets() in single loop,
>denies retired scheduled tasks, and round up the timeout to 1ms.
>
>This CL also improves packet size estimation in TQ Pacer by removing
>zero initialization, and introduces `include_overhead_` configuration.
>
>Tests:
>1. webrtc_perf_tests: MaybeProcessPackets() calls
> 2075147 -> 2007995
>
>2. module_unittests: MaybeProcessPackets() calls
> 203393 -> 183563
>
>3. peerconnection_unittests: MaybeProcessPackets() calls
> 66713-> 64333
>
>Bug: webrtc:13417, webrtc:13437
>Change-Id: I18eb0a36dbe063c606b1f27014df74a65ebfc486
>Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/242962
>Reviewed-by: Erik Språng <sprang@webrtc.org>
>Reviewed-by: Henrik Boström <hbos@webrtc.org>
>Commit-Queue: Erik Språng <sprang@webrtc.org>
>Cr-Commit-Position: refs/heads/main@{#36179}
Bug: webrtc:13417, webrtc:13437
Change-Id: I79f2554cf02364b67ce7073698611a3ae337a73b
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/256145
Reviewed-by: Erik Språng <sprang@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Owners-Override: Erik Språng <sprang@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36283}
This is a reland of commit 43a69b3f46059b103c35079744dc09a0e2bb2948
Original change's description:
> Experimentally reduce TaskQueuePacedSender's delayed task precision.
>
> This CL reduces the delayed task precision of non-probes in accordance
> with DD go/slacked-task-queue-paced-sender. The precision is only
> deduced if field trial "WebRTC-SlackedTaskQueuePacedSender" is enabled
> though.
>
> Bug: webrtc:13824
> Change-Id: I37e53b24e343f4f08059be08a3cda74f5484cc05
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/255341
> Reviewed-by: Erik Språng <sprang@webrtc.org>
> Commit-Queue: Henrik Boström <hbos@webrtc.org>
> Cr-Commit-Position: refs/heads/main@{#36214}
Bug: webrtc:13824
Change-Id: I86cace6f4f6bf23d51c75b3d18f8d24fff0f5b74
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/255826
Auto-Submit: Mirko Bonadei <mbonadei@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36221}
When congestion window is used, two different mechanisms can currently
update the outstanding data state in the pacer:
* OnPacketSent() withing the pacer itself, when a packet is sent
* UpdateOutstandingData(), when RtpTransportControllerSend either:
a. Receives an OnPacketSent() callback (increase outstanding data)
b. Receives transport feedback (decrease outstanding data)
This creates a lot of calls to UpdateOutstandingData(), more than one
per sent packet. Each requires locking and/or thread jumps. To avoid
that, this CL moves the congestion window state to
RtpTransportController send - and we only post a congested flag down
the the pacer when the state is changed.
The only benefit I can see is of the old way is we prevent sending
new packets immedately when the window is full, rather than in some
edge cases queue extra packets on the network task queue before the
congestion signal is received. That should be rare and benign.
I think this simplified logic, which is easier to read and more
performant, is a better tradeoff.
Bug: webrtc:13417
Change-Id: I326dd88db86dc0d6dc685c61920654ac024e57ef
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/255600
Auto-Submit: Erik Språng <sprang@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36220}
This CL reduces the delayed task precision of non-probes in accordance
with DD go/slacked-task-queue-paced-sender. The precision is only
deduced if field trial "WebRTC-SlackedTaskQueuePacedSender" is enabled
though.
Bug: webrtc:13824
Change-Id: I37e53b24e343f4f08059be08a3cda74f5484cc05
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/255341
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36214}
This CL also removes dependency on the legacy field trial methods.
Bug: webrtc:11926
Change-Id: I53feeee86b92878cf0f2b8ebdce3d101f9e04014
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/255381
Reviewed-by: Jonas Oreland <jonaso@webrtc.org>
Auto-Submit: Erik Språng <sprang@webrtc.org>
Reviewed-by: Björn Terelius <terelius@webrtc.org>
Commit-Queue: Björn Terelius <terelius@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36205}
The TQ Pacer schedules delayed task according to target time of
PacingController. It drains all valid ProcessPackets() in single loop,
denies retired scheduled tasks, and round up the timeout to 1ms.
This CL also improves packet size estimation in TQ Pacer by removing
zero initialization, and introduces `include_overhead_` configuration.
Tests:
1. webrtc_perf_tests: MaybeProcessPackets() calls
2075147 -> 2007995
2. module_unittests: MaybeProcessPackets() calls
203393 -> 183563
3. peerconnection_unittests: MaybeProcessPackets() calls
66713-> 64333
Bug: webrtc:13417, webrtc:13437
Change-Id: I18eb0a36dbe063c606b1f27014df74a65ebfc486
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/242962
Reviewed-by: Erik Språng <sprang@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36179}
This cl/
1) move WebRtcKeyValueConfig from api/transport to api/ directory.
2) add a test/ScopedKeyValueConfig (compare ScopedFieldTrials).
3) removes usage of webrtc::field_trial:: from the pc/ directory.
4) removes a few unused includes of system_wrappers/field_trial.h.
Bug: webrtc:10335
Change-Id: If29c07900dbe791050b0a5ad05332bedfad035f2
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/253903
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Commit-Queue: Jonas Oreland <jonaso@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#36160}
Use cases of TaskQueue or TaskQueueBase that are considered high
precision are updated to make use of PostDelayedHighPrecisionTask
(see go/postdelayedtask-precision-in-webrtc) instead of PostDelayedTask.
The cases here are the ones covered by that document, plus some
testing-only uses. The FrameBuffer2 and DataTracker use cases will
be covered by separate CLs because FrameBuffer2 uses
RepeatingTaskHandle and DataTracker uses dcsctp::Timer.
This protects these use cases against regressions when PostDelayedTask
gets its precision lowered.
This CL also adds TaskQueue::PostDelayedHighPrecisionTask which calls
TaskQueueBase::PostDelayedHighPrecisionTask (same pattern as for
PostDelayedTask).
Bug: webrtc:13604
Change-Id: I7dcab59cbe4d274d27b734ceb4fc06daa12ffd0f
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/248864
Reviewed-by: Erik Språng <sprang@webrtc.org>
Reviewed-by: Tomas Gunnarsson <tommi@webrtc.org>
Commit-Queue: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35781}
The pacer status is never changed unless MaybeProcessPackets() is
called. This CL removes the scheduler, and updates pacer status after
every MaybeProcessPackets().
Bug: webrtc:10809, webrtc:13417
Change-Id: Ib5f18decf44c1596c0a716d799600a72b2332abd
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/239120
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35489}
Holdback window can be specified as absolute time and in terms of packet
send times. Example:
WebRTC-TaskQueuePacer/Enabled,holdback_window:20ms,holdback_packet:3/
If current conditions have us running with 2000kbps pacing rate and
1250byte (10kbit) packets, each packet send time is 5ms.
The holdback window would then be min(20ms, 3*5ms) = 15ms.
The default is like before 1ms and packets no take into account when
TQ pacer is used, parameters have no effect with legacy process thread
pacer.
Bug: webrtc:10809
Change-Id: I800de05107e2d4df461eabaaf1ca04fb4c5de51e
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/233421
Commit-Queue: Erik Språng <sprang@webrtc.org>
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35266}
When new packets are enqueued after a dead-period where media debt is
zero, that time slice should not be used to reduce the debt for the
new packet.
Bug: webrtc:10809
Change-Id: Ifb960548e6aa349b79f37743cbfed78a5c937a13
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/234081
Reviewed-by: Philip Eliasson <philipel@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/main@{#35143}
Since https://webrtc-review.googlesource.com/c/src/+/228433 both audio
and video now only call Get/SetRtpState while not registered to the
packet router.
We can thus remove the lock around packet sequencer and just use a
thread checker.
Bug: webrtc:11340
Change-Id: Ie6865cc96c36208700c31a75747ff4dd992ce68d
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/228435
Commit-Queue: Erik Språng <sprang@webrtc.org>
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#34755}
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}
This removes PacketRouter inheritance from RemoteBitrateObserver and TransportFeedbackSenderInterface.
Call binds methods for sending REMB and transport feedback messages from RemoteCongestionController to PacketRouter.
This is needed until the RTCPTranseiver is used instead of the RTP modules.
Bug: webrtc:12693
Change-Id: I7088de497cd6d1e15c98788ff3e6b0a2c8897ea8
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/215965
Reviewed-by: Harald Alvestrand <hta@webrtc.org>
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Reviewed-by: Danil Chapovalov <danilchap@webrtc.org>
Commit-Queue: Per Kjellander <perkj@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33993}
In highly loaded Media Servers, RoundRobinPacketQueue's usage of
std::map attributes to around 0.3% CPU. In profiling, it has been found
that it's the `streams_` map that is the culprit and std::map::find
being the function that is used most.
By using an unordered map, lookups should be faster. This will be
evaluated and if this commit doesn't show results, it will be reverted.
Bug: webrtc:12689
Change-Id: Ia368f1184130298cfbb9064528828435aa7a5c66
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/216320
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Victor Boivie <boivie@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33848}
Following up on https://webrtc-review.googlesource.com/c/src/+/213000
This CL prevents scheduling work before TaskQueuePacedSender::EnsureStarted(),
making it necessary to function.
Bug: chromium:1152887
Change-Id: I848c9e6d6057a404626ad693b1f4dc7fba797a9c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/214320
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Etienne Pierre-Doray <etiennep@chromium.org>
Cr-Commit-Position: refs/heads/master@{#33695}
This is a reland of 89cb65ed663a9000b9f7c90a78039bd85731e9ae
... and f28aade91dcc2cb8f590dc1379ac7ab5c1981909
... and 2072b87261a6505a88561bdeab3e7405d7038eaa
Reason for revert: Failing DuoGroupsMediaQualityTest due to missing
TaskQueuePacedSender::EnsureStarted() in google3.
Fix: This CL adds the logic behind TaskQueuePacedSender::EnsureStarted,
but initializes with |is_started| = true. Once the caller in google3 is
updated, |is_started| can be switched to false by default.
> Original change's description:
> Reason for revert: crashes due to uninitialized pacing_bitrate_
> crbug.com/1190547
> Apparently pacer() is sometimes being used before EnsureStarted()
> Fix: Instead of delaying first call to SetPacingRates(),
> this CL no-ops MaybeProcessPackets() until EnsureStarted()
> is called for the first time.
> Original change's description:
> > [Battery]: Delay start of TaskQueuePacedSender.
> >
> > To avoid unnecessary repeating tasks, TaskQueuePacedSender is started
> > only upon RtpTransportControllerSend::EnsureStarted().
> >
> > More specifically, the repeating task happens in
> > TaskQueuePacedSender::MaybeProcessPackets() every 500ms, using a self
> > task_queue_.PostDelayedTask().
> >
> > Bug: chromium:1152887
> > Change-Id: I72c96d2c4b491d5edb45a30b210b3797165cbf48
> > Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/208560
> > Commit-Queue: Etienne Pierre-Doray <etiennep@chromium.org>
> > Reviewed-by: Henrik Boström <hbos@webrtc.org>
> > Reviewed-by: Erik Språng <sprang@webrtc.org>
> > Cr-Commit-Position: refs/heads/master@{#33421}
>
> Bug: chromium:1152887
> Change-Id: I9aba4882a64bbee7d97ace9059dea8a24c144f93
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/212880
> Reviewed-by: Erik Språng <sprang@webrtc.org>
> Reviewed-by: Henrik Boström <hbos@webrtc.org>
> Commit-Queue: Etienne Pierre-Doray <etiennep@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#33554}
Bug: chromium:1152887
Change-Id: Ie365562bd83aefdb2757a65e20a4cf3eece678b9
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/213000
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Reviewed-by: Erik Språng <sprang@webrtc.org>
Commit-Queue: Etienne Pierre-Doray <etiennep@chromium.org>
Cr-Commit-Position: refs/heads/master@{#33629}
This reverts commit 2072b87261a6505a88561bdeab3e7405d7038eaa.
Reason for revert: Causing test failure.
Original change's description:
> Reland "[Battery]: Delay start of TaskQueuePacedSender." Take 2
>
> This is a reland of 89cb65ed663a9000b9f7c90a78039bd85731e9ae
> ... and f28aade91dcc2cb8f590dc1379ac7ab5c1981909
>
> Reason for revert: crashes due to uninitialized pacing_bitrate_
> crbug.com/1190547
> Apparently pacer() is sometimes being used before EnsureStarted()
> Fix: Instead of delaying first call to SetPacingRates(),
> this CL no-ops MaybeProcessPackets() until EnsureStarted()
> is called for the first time.
>
> Original change's description:
> > [Battery]: Delay start of TaskQueuePacedSender.
> >
> > To avoid unnecessary repeating tasks, TaskQueuePacedSender is started
> > only upon RtpTransportControllerSend::EnsureStarted().
> >
> > More specifically, the repeating task happens in
> > TaskQueuePacedSender::MaybeProcessPackets() every 500ms, using a self
> > task_queue_.PostDelayedTask().
> >
> > Bug: chromium:1152887
> > Change-Id: I72c96d2c4b491d5edb45a30b210b3797165cbf48
> > Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/208560
> > Commit-Queue: Etienne Pierre-Doray <etiennep@chromium.org>
> > Reviewed-by: Henrik Boström <hbos@webrtc.org>
> > Reviewed-by: Erik Språng <sprang@webrtc.org>
> > Cr-Commit-Position: refs/heads/master@{#33421}
>
> Bug: chromium:1152887
> Change-Id: I9aba4882a64bbee7d97ace9059dea8a24c144f93
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/212880
> Reviewed-by: Erik Språng <sprang@webrtc.org>
> Reviewed-by: Henrik Boström <hbos@webrtc.org>
> Commit-Queue: Etienne Pierre-Doray <etiennep@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#33554}
TBR=hbos@webrtc.org,sprang@webrtc.org,etiennep@chromium.org
Change-Id: I430fd31c7602702c8ec44b9e38e68266abba8854
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: chromium:1152887
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/212965
Reviewed-by: Ying Wang <yinwa@webrtc.org>
Commit-Queue: Ying Wang <yinwa@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#33559}
This is a reland of 89cb65ed663a9000b9f7c90a78039bd85731e9ae
... and f28aade91dcc2cb8f590dc1379ac7ab5c1981909
Reason for revert: crashes due to uninitialized pacing_bitrate_
crbug.com/1190547
Apparently pacer() is sometimes being used before EnsureStarted()
Fix: Instead of delaying first call to SetPacingRates(),
this CL no-ops MaybeProcessPackets() until EnsureStarted()
is called for the first time.
Original change's description:
> [Battery]: Delay start of TaskQueuePacedSender.
>
> To avoid unnecessary repeating tasks, TaskQueuePacedSender is started
> only upon RtpTransportControllerSend::EnsureStarted().
>
> More specifically, the repeating task happens in
> TaskQueuePacedSender::MaybeProcessPackets() every 500ms, using a self
> task_queue_.PostDelayedTask().
>
> Bug: chromium:1152887
> Change-Id: I72c96d2c4b491d5edb45a30b210b3797165cbf48
> Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/208560
> Commit-Queue: Etienne Pierre-Doray <etiennep@chromium.org>
> Reviewed-by: Henrik Boström <hbos@webrtc.org>
> Reviewed-by: Erik Språng <sprang@webrtc.org>
> Cr-Commit-Position: refs/heads/master@{#33421}
Bug: chromium:1152887
Change-Id: I9aba4882a64bbee7d97ace9059dea8a24c144f93
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/212880
Reviewed-by: Erik Språng <sprang@webrtc.org>
Reviewed-by: Henrik Boström <hbos@webrtc.org>
Commit-Queue: Etienne Pierre-Doray <etiennep@chromium.org>
Cr-Commit-Position: refs/heads/master@{#33554}
This rounding triggers a dcheck that crashes debug builds.
Furtunately, in release mode this does not matter as the resulting
value is anyway capped to 0.
Unfortunately, it is almost impossible to write a test for this even
with simulated time as the conditions needed to trigger this condition
includes thread scheduling being slightly off in such a way that an
unscheduled process call preempts a scheduled one at a time when
sending a sufficiently large padding packet becomes possible - and
right after starting a new probe cluster.
We should consider updating this class to make unit testing easier.
Bug: webrtc:10809
Change-Id: I533e6e716bddc106d11e82a9e3edb4e0035fd21c
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/192786
Reviewed-by: Per Kjellander <perkj@webrtc.org>
Commit-Queue: Erik Språng <sprang@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32589}
And use it in a few places that were using RTC_CHECK(false) or FATAL()
to do the exact same job. There should be no change in behavior.
Bug: none
Change-Id: I36d5e6bcf35fd41534e08a8c879fa0811b4f1967
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/191963
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Commit-Queue: Karl Wiberg <kwiberg@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#32567}
"warning: control reaches end of non-void function [-Wreturn-type]"
Reported by gcc (8.3)
In all the reported cases, the end of function is never actually
reached. Add RTC_CHECK(false) to ensure the compiler is aware that
this path is a dead-end.
Bug: webrtc:12008
Change-Id: I7f816fde3d1897ed2774057c7e05da66e1895e60
Reviewed-on: https://webrtc-review.googlesource.com/c/src/+/189784
Reviewed-by: Karl Wiberg <kwiberg@webrtc.org>
Reviewed-by: Mirko Bonadei <mbonadei@webrtc.org>
Commit-Queue: Karl Wiberg <kwiberg@webrtc.org>
Commit-Queue: Fabien VALLÉE <fabien.vallee@netgem.com>
Cr-Commit-Position: refs/heads/master@{#32503}
