2012-09-01 00:59:37 +04:00
|
|
|
/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
|
|
|
|
/* vim:set ts=2 sw=2 sts=2 et cindent: */
|
|
|
|
/* This Source Code Form is subject to the terms of the Mozilla Public
|
|
|
|
* License, v. 2.0. If a copy of the MPL was not distributed with this
|
|
|
|
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
|
|
|
|
|
2012-10-31 01:39:38 +04:00
|
|
|
#ifndef AudioContext_h_
|
|
|
|
#define AudioContext_h_
|
2012-09-19 03:07:33 +04:00
|
|
|
|
2017-08-09 05:55:43 +03:00
|
|
|
#include "mozilla/dom/OfflineAudioContextBinding.h"
|
2013-02-02 02:13:23 +04:00
|
|
|
#include "MediaBufferDecoder.h"
|
2013-07-24 11:31:06 +04:00
|
|
|
#include "mozilla/Attributes.h"
|
2014-04-01 10:13:50 +04:00
|
|
|
#include "mozilla/DOMEventTargetHelper.h"
|
|
|
|
#include "mozilla/MemoryReporting.h"
|
2013-07-24 11:31:06 +04:00
|
|
|
#include "mozilla/dom/TypedArray.h"
|
2018-03-01 20:00:12 +03:00
|
|
|
#include "mozilla/RelativeTimeline.h"
|
2017-01-04 00:52:03 +03:00
|
|
|
#include "mozilla/UniquePtr.h"
|
2013-07-24 11:31:06 +04:00
|
|
|
#include "nsCOMPtr.h"
|
|
|
|
#include "nsCycleCollectionParticipant.h"
|
|
|
|
#include "nsHashKeys.h"
|
2013-04-25 07:32:41 +04:00
|
|
|
#include "nsTHashtable.h"
|
2013-08-28 06:59:14 +04:00
|
|
|
#include "js/TypeDecls.h"
|
2014-01-04 22:15:41 +04:00
|
|
|
#include "nsIMemoryReporter.h"
|
2012-09-01 00:59:37 +04:00
|
|
|
|
2013-03-15 05:01:02 +04:00
|
|
|
// X11 has a #define for CurrentTime. Unbelievable :-(.
|
2014-10-25 21:24:36 +04:00
|
|
|
// See dom/media/DOMMediaStream.h for more fun!
|
2013-03-15 05:01:02 +04:00
|
|
|
#ifdef CurrentTime
|
|
|
|
#undef CurrentTime
|
|
|
|
#endif
|
|
|
|
|
2015-06-15 19:16:16 +03:00
|
|
|
namespace WebCore {
|
|
|
|
class PeriodicWave;
|
2015-07-13 18:25:42 +03:00
|
|
|
} // namespace WebCore
|
2015-06-15 19:16:16 +03:00
|
|
|
|
2016-01-30 20:05:36 +03:00
|
|
|
class nsPIDOMWindowInner;
|
2012-09-01 00:59:37 +04:00
|
|
|
|
|
|
|
namespace mozilla {
|
|
|
|
|
2013-08-15 23:44:14 +04:00
|
|
|
class DOMMediaStream;
|
2012-09-01 00:59:37 +04:00
|
|
|
class ErrorResult;
|
2013-08-15 23:44:14 +04:00
|
|
|
class MediaStream;
|
|
|
|
class MediaStreamGraph;
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
class AudioNodeStream;
|
2012-09-01 00:59:37 +04:00
|
|
|
|
2012-09-08 02:13:26 +04:00
|
|
|
namespace dom {
|
|
|
|
|
2017-01-12 04:23:37 +03:00
|
|
|
enum class AudioContextState : uint8_t;
|
2013-04-01 07:41:14 +04:00
|
|
|
class AnalyserNode;
|
2012-09-22 02:42:14 +04:00
|
|
|
class AudioBuffer;
|
2012-11-06 04:26:03 +04:00
|
|
|
class AudioBufferSourceNode;
|
|
|
|
class AudioDestinationNode;
|
|
|
|
class AudioListener;
|
2013-09-17 03:53:40 +04:00
|
|
|
class AudioNode;
|
2012-11-08 05:59:14 +04:00
|
|
|
class BiquadFilterNode;
|
2013-05-05 19:49:37 +04:00
|
|
|
class ChannelMergerNode;
|
2013-05-05 19:49:13 +04:00
|
|
|
class ChannelSplitterNode;
|
2016-10-13 17:40:29 +03:00
|
|
|
class ConstantSourceNode;
|
2013-06-11 00:07:55 +04:00
|
|
|
class ConvolverNode;
|
2012-11-01 04:26:03 +04:00
|
|
|
class DelayNode;
|
2012-11-07 05:01:11 +04:00
|
|
|
class DynamicsCompressorNode;
|
2012-11-06 04:26:03 +04:00
|
|
|
class GainNode;
|
2016-06-07 12:50:16 +03:00
|
|
|
class GlobalObject;
|
2013-07-25 07:01:49 +04:00
|
|
|
class HTMLMediaElement;
|
2016-06-07 12:50:16 +03:00
|
|
|
class IIRFilterNode;
|
2013-07-25 07:01:49 +04:00
|
|
|
class MediaElementAudioSourceNode;
|
2013-05-21 23:17:47 +04:00
|
|
|
class MediaStreamAudioDestinationNode;
|
2013-07-24 15:29:39 +04:00
|
|
|
class MediaStreamAudioSourceNode;
|
2013-08-19 22:53:00 +04:00
|
|
|
class OscillatorNode;
|
2012-11-06 06:14:13 +04:00
|
|
|
class PannerNode;
|
2013-04-14 05:37:04 +04:00
|
|
|
class ScriptProcessorNode;
|
2014-11-19 20:15:13 +03:00
|
|
|
class StereoPannerNode;
|
2013-05-14 08:12:30 +04:00
|
|
|
class WaveShaperNode;
|
2018-06-27 12:31:02 +03:00
|
|
|
class Worklet;
|
2013-06-20 02:24:26 +04:00
|
|
|
class PeriodicWave;
|
2016-04-26 13:54:38 +03:00
|
|
|
struct PeriodicWaveConstraints;
|
2014-10-23 14:07:48 +04:00
|
|
|
class Promise;
|
2017-01-12 04:23:37 +03:00
|
|
|
enum class OscillatorType : uint8_t;
|
2015-06-15 19:16:16 +03:00
|
|
|
|
|
|
|
// This is addrefed by the OscillatorNodeEngine on the main thread
|
|
|
|
// and then used from the MSG thread.
|
|
|
|
// It can be released either from the graph thread or the main thread.
|
|
|
|
class BasicWaveFormCache
|
|
|
|
{
|
|
|
|
public:
|
2015-07-06 15:27:18 +03:00
|
|
|
explicit BasicWaveFormCache(uint32_t aSampleRate);
|
2015-06-15 19:16:16 +03:00
|
|
|
NS_INLINE_DECL_THREADSAFE_REFCOUNTING(BasicWaveFormCache)
|
|
|
|
WebCore::PeriodicWave* GetBasicWaveForm(OscillatorType aType);
|
|
|
|
private:
|
|
|
|
~BasicWaveFormCache();
|
2015-10-18 08:24:48 +03:00
|
|
|
RefPtr<WebCore::PeriodicWave> mSawtooth;
|
|
|
|
RefPtr<WebCore::PeriodicWave> mSquare;
|
|
|
|
RefPtr<WebCore::PeriodicWave> mTriangle;
|
2015-06-15 19:16:16 +03:00
|
|
|
uint32_t mSampleRate;
|
|
|
|
};
|
|
|
|
|
2012-09-19 03:07:33 +04:00
|
|
|
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
/* This runnable allows the MSG to notify the main thread when audio is actually
|
|
|
|
* flowing */
|
2016-04-26 03:23:21 +03:00
|
|
|
class StateChangeTask final : public Runnable
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
{
|
|
|
|
public:
|
|
|
|
/* This constructor should be used when this event is sent from the main
|
|
|
|
* thread. */
|
|
|
|
StateChangeTask(AudioContext* aAudioContext, void* aPromise, AudioContextState aNewState);
|
|
|
|
|
|
|
|
/* This constructor should be used when this event is sent from the audio
|
|
|
|
* thread. */
|
|
|
|
StateChangeTask(AudioNodeStream* aStream, void* aPromise, AudioContextState aNewState);
|
|
|
|
|
|
|
|
NS_IMETHOD Run() override;
|
|
|
|
|
|
|
|
private:
|
2015-10-18 08:24:48 +03:00
|
|
|
RefPtr<AudioContext> mAudioContext;
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
void* mPromise;
|
2015-10-18 08:24:48 +03:00
|
|
|
RefPtr<AudioNodeStream> mAudioNodeStream;
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
AudioContextState mNewState;
|
|
|
|
};
|
|
|
|
|
2015-09-24 23:49:03 +03:00
|
|
|
enum class AudioContextOperation { Suspend, Resume, Close };
|
2018-04-03 20:02:15 +03:00
|
|
|
struct AudioContextOptions;
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
|
2015-03-21 19:28:04 +03:00
|
|
|
class AudioContext final : public DOMEventTargetHelper,
|
2018-03-01 20:00:12 +03:00
|
|
|
public nsIMemoryReporter,
|
|
|
|
public RelativeTimeline
|
2012-09-01 00:59:37 +04:00
|
|
|
{
|
2016-01-30 20:05:36 +03:00
|
|
|
AudioContext(nsPIDOMWindowInner* aParentWindow,
|
2013-05-17 03:30:42 +04:00
|
|
|
bool aIsOffline,
|
|
|
|
uint32_t aNumberOfChannels = 0,
|
|
|
|
uint32_t aLength = 0,
|
|
|
|
float aSampleRate = 0.0f);
|
|
|
|
~AudioContext();
|
2012-09-01 00:59:37 +04:00
|
|
|
|
2015-10-05 05:35:00 +03:00
|
|
|
nsresult Init();
|
2015-07-04 07:13:03 +03:00
|
|
|
|
2012-09-01 00:59:37 +04:00
|
|
|
public:
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
typedef uint64_t AudioContextId;
|
|
|
|
|
2013-04-25 08:28:39 +04:00
|
|
|
NS_DECL_ISUPPORTS_INHERITED
|
|
|
|
NS_DECL_CYCLE_COLLECTION_CLASS_INHERITED(AudioContext,
|
2014-04-01 10:13:50 +04:00
|
|
|
DOMEventTargetHelper)
|
2014-01-04 22:15:41 +04:00
|
|
|
MOZ_DEFINE_MALLOC_SIZE_OF(MallocSizeOf)
|
2012-09-01 00:59:37 +04:00
|
|
|
|
2016-01-30 20:05:36 +03:00
|
|
|
nsPIDOMWindowInner* GetParentObject() const
|
2012-09-01 00:59:37 +04:00
|
|
|
{
|
2013-04-25 08:28:39 +04:00
|
|
|
return GetOwner();
|
2012-09-01 00:59:37 +04:00
|
|
|
}
|
|
|
|
|
2017-03-15 19:36:40 +03:00
|
|
|
virtual void DisconnectFromOwner() override;
|
2018-04-16 16:08:52 +03:00
|
|
|
virtual void BindToOwner(nsIGlobalObject* aNew) override;
|
2017-03-15 19:36:40 +03:00
|
|
|
|
2013-09-10 09:05:22 +04:00
|
|
|
void Shutdown(); // idempotent
|
2013-03-22 06:59:33 +04:00
|
|
|
|
2016-01-18 06:22:51 +03:00
|
|
|
JSObject* WrapObject(JSContext* aCx, JS::Handle<JSObject*> aGivenProto) override;
|
2012-09-01 00:59:37 +04:00
|
|
|
|
2014-04-01 10:13:50 +04:00
|
|
|
using DOMEventTargetHelper::DispatchTrustedEvent;
|
2013-05-17 03:30:57 +04:00
|
|
|
|
2013-05-17 03:30:41 +04:00
|
|
|
// Constructor for regular AudioContext
|
2012-09-01 00:59:37 +04:00
|
|
|
static already_AddRefed<AudioContext>
|
2018-04-03 20:02:15 +03:00
|
|
|
Constructor(const GlobalObject& aGlobal,
|
|
|
|
const AudioContextOptions& aOptions,
|
|
|
|
ErrorResult& aRv);
|
2012-09-01 00:59:37 +04:00
|
|
|
|
2017-08-09 05:55:43 +03:00
|
|
|
// Constructor for offline AudioContext with options object
|
|
|
|
static already_AddRefed<AudioContext>
|
|
|
|
Constructor(const GlobalObject& aGlobal,
|
|
|
|
const OfflineAudioContextOptions& aOptions,
|
|
|
|
ErrorResult& aRv);
|
|
|
|
|
2013-05-17 03:30:41 +04:00
|
|
|
// Constructor for offline AudioContext
|
|
|
|
static already_AddRefed<AudioContext>
|
|
|
|
Constructor(const GlobalObject& aGlobal,
|
|
|
|
uint32_t aNumberOfChannels,
|
|
|
|
uint32_t aLength,
|
|
|
|
float aSampleRate,
|
|
|
|
ErrorResult& aRv);
|
|
|
|
|
|
|
|
// AudioContext methods
|
|
|
|
|
2012-09-19 03:07:33 +04:00
|
|
|
AudioDestinationNode* Destination() const
|
|
|
|
{
|
|
|
|
return mDestination;
|
|
|
|
}
|
|
|
|
|
2012-11-30 07:31:39 +04:00
|
|
|
float SampleRate() const
|
|
|
|
{
|
2013-05-24 21:09:29 +04:00
|
|
|
return mSampleRate;
|
2012-11-30 07:31:39 +04:00
|
|
|
}
|
|
|
|
|
2015-09-16 07:15:21 +03:00
|
|
|
bool ShouldSuspendNewStream() const { return mSuspendCalled; }
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
|
2018-03-01 20:00:12 +03:00
|
|
|
double CurrentTime();
|
2013-03-15 05:01:02 +04:00
|
|
|
|
2012-11-06 04:26:03 +04:00
|
|
|
AudioListener* Listener();
|
|
|
|
|
2015-09-03 08:35:19 +03:00
|
|
|
AudioContextState State() const { return mAudioContextState; }
|
2018-06-27 12:31:02 +03:00
|
|
|
|
|
|
|
Worklet* GetAudioWorklet(ErrorResult& aRv);
|
|
|
|
|
2017-07-24 12:17:14 +03:00
|
|
|
bool IsRunning() const;
|
2015-09-03 08:35:19 +03:00
|
|
|
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
// Those three methods return a promise to content, that is resolved when an
|
|
|
|
// (possibly long) operation is completed on the MSG (and possibly other)
|
|
|
|
// thread(s). To avoid having to match the calls and asychronous result when
|
|
|
|
// the operation is completed, we keep a reference to the promises on the main
|
|
|
|
// thread, and then send the promises pointers down the MSG thread, as a void*
|
|
|
|
// (to make it very clear that the pointer is to merely be treated as an ID).
|
|
|
|
// When back on the main thread, we can resolve or reject the promise, by
|
|
|
|
// casting it back to a `Promise*` while asserting we're back on the main
|
|
|
|
// thread and removing the reference we added.
|
|
|
|
already_AddRefed<Promise> Suspend(ErrorResult& aRv);
|
|
|
|
already_AddRefed<Promise> Resume(ErrorResult& aRv);
|
|
|
|
already_AddRefed<Promise> Close(ErrorResult& aRv);
|
|
|
|
IMPL_EVENT_HANDLER(statechange)
|
|
|
|
|
|
|
|
already_AddRefed<AudioBufferSourceNode> CreateBufferSource(ErrorResult& aRv);
|
2012-09-19 03:07:33 +04:00
|
|
|
|
2016-10-13 17:40:29 +03:00
|
|
|
already_AddRefed<ConstantSourceNode> CreateConstantSource(ErrorResult& aRv);
|
|
|
|
|
2012-09-22 02:42:14 +04:00
|
|
|
already_AddRefed<AudioBuffer>
|
2016-03-12 00:43:31 +03:00
|
|
|
CreateBuffer(uint32_t aNumberOfChannels, uint32_t aLength, float aSampleRate,
|
2012-09-22 02:42:14 +04:00
|
|
|
ErrorResult& aRv);
|
|
|
|
|
2013-05-21 23:17:47 +04:00
|
|
|
already_AddRefed<MediaStreamAudioDestinationNode>
|
2013-07-18 13:57:38 +04:00
|
|
|
CreateMediaStreamDestination(ErrorResult& aRv);
|
2013-05-21 23:17:47 +04:00
|
|
|
|
2013-04-14 05:37:04 +04:00
|
|
|
already_AddRefed<ScriptProcessorNode>
|
|
|
|
CreateScriptProcessor(uint32_t aBufferSize,
|
|
|
|
uint32_t aNumberOfInputChannels,
|
|
|
|
uint32_t aNumberOfOutputChannels,
|
|
|
|
ErrorResult& aRv);
|
|
|
|
|
2014-11-19 20:15:13 +03:00
|
|
|
already_AddRefed<StereoPannerNode>
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
CreateStereoPanner(ErrorResult& aRv);
|
2014-11-19 20:15:13 +03:00
|
|
|
|
2013-04-01 07:41:14 +04:00
|
|
|
already_AddRefed<AnalyserNode>
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
CreateAnalyser(ErrorResult& aRv);
|
2013-04-01 07:41:14 +04:00
|
|
|
|
2012-10-31 23:09:32 +04:00
|
|
|
already_AddRefed<GainNode>
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
CreateGain(ErrorResult& aRv);
|
2012-10-31 23:09:32 +04:00
|
|
|
|
2013-05-14 08:12:30 +04:00
|
|
|
already_AddRefed<WaveShaperNode>
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
CreateWaveShaper(ErrorResult& aRv);
|
2013-05-14 08:12:30 +04:00
|
|
|
|
2013-07-25 07:01:49 +04:00
|
|
|
already_AddRefed<MediaElementAudioSourceNode>
|
|
|
|
CreateMediaElementSource(HTMLMediaElement& aMediaElement, ErrorResult& aRv);
|
2013-07-24 15:29:39 +04:00
|
|
|
already_AddRefed<MediaStreamAudioSourceNode>
|
2013-07-25 06:07:34 +04:00
|
|
|
CreateMediaStreamSource(DOMMediaStream& aMediaStream, ErrorResult& aRv);
|
2013-07-24 15:29:39 +04:00
|
|
|
|
2012-11-01 04:26:03 +04:00
|
|
|
already_AddRefed<DelayNode>
|
2012-11-20 00:52:29 +04:00
|
|
|
CreateDelay(double aMaxDelayTime, ErrorResult& aRv);
|
2012-11-01 04:26:03 +04:00
|
|
|
|
2012-11-06 06:14:13 +04:00
|
|
|
already_AddRefed<PannerNode>
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
CreatePanner(ErrorResult& aRv);
|
2012-11-06 06:14:13 +04:00
|
|
|
|
2013-06-11 00:07:55 +04:00
|
|
|
already_AddRefed<ConvolverNode>
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
CreateConvolver(ErrorResult& aRv);
|
2013-06-11 00:07:55 +04:00
|
|
|
|
2013-05-05 19:49:13 +04:00
|
|
|
already_AddRefed<ChannelSplitterNode>
|
|
|
|
CreateChannelSplitter(uint32_t aNumberOfOutputs, ErrorResult& aRv);
|
|
|
|
|
2013-05-05 19:49:37 +04:00
|
|
|
already_AddRefed<ChannelMergerNode>
|
|
|
|
CreateChannelMerger(uint32_t aNumberOfInputs, ErrorResult& aRv);
|
|
|
|
|
2012-11-07 05:01:11 +04:00
|
|
|
already_AddRefed<DynamicsCompressorNode>
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
CreateDynamicsCompressor(ErrorResult& aRv);
|
2012-11-07 05:01:11 +04:00
|
|
|
|
2012-11-08 05:59:14 +04:00
|
|
|
already_AddRefed<BiquadFilterNode>
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
CreateBiquadFilter(ErrorResult& aRv);
|
2012-11-08 05:59:14 +04:00
|
|
|
|
2016-06-07 12:50:16 +03:00
|
|
|
already_AddRefed<IIRFilterNode>
|
2017-01-03 06:54:37 +03:00
|
|
|
CreateIIRFilter(const Sequence<double>& aFeedforward,
|
|
|
|
const Sequence<double>& aFeedback,
|
2016-06-07 12:50:16 +03:00
|
|
|
mozilla::ErrorResult& aRv);
|
|
|
|
|
2013-08-19 22:53:00 +04:00
|
|
|
already_AddRefed<OscillatorNode>
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
CreateOscillator(ErrorResult& aRv);
|
2013-08-19 22:53:00 +04:00
|
|
|
|
2013-06-20 02:24:26 +04:00
|
|
|
already_AddRefed<PeriodicWave>
|
|
|
|
CreatePeriodicWave(const Float32Array& aRealData, const Float32Array& aImagData,
|
2016-04-26 13:54:38 +03:00
|
|
|
const PeriodicWaveConstraints& aConstraints,
|
2013-06-20 02:24:26 +04:00
|
|
|
ErrorResult& aRv);
|
2013-05-28 15:19:07 +04:00
|
|
|
|
2014-10-23 14:07:48 +04:00
|
|
|
already_AddRefed<Promise>
|
|
|
|
DecodeAudioData(const ArrayBuffer& aBuffer,
|
|
|
|
const Optional<OwningNonNull<DecodeSuccessCallback> >& aSuccessCallback,
|
2015-01-05 15:43:00 +03:00
|
|
|
const Optional<OwningNonNull<DecodeErrorCallback> >& aFailureCallback,
|
|
|
|
ErrorResult& aRv);
|
2013-02-02 02:13:23 +04:00
|
|
|
|
2013-05-17 03:30:41 +04:00
|
|
|
// OfflineAudioContext methods
|
2014-11-19 13:38:39 +03:00
|
|
|
already_AddRefed<Promise> StartRendering(ErrorResult& aRv);
|
2013-05-17 03:30:41 +04:00
|
|
|
IMPL_EVENT_HANDLER(complete)
|
2016-04-25 12:45:59 +03:00
|
|
|
unsigned long Length();
|
2013-05-17 03:30:41 +04:00
|
|
|
|
2013-05-17 03:31:08 +04:00
|
|
|
bool IsOffline() const { return mIsOffline; }
|
2013-02-05 03:07:25 +04:00
|
|
|
|
|
|
|
MediaStreamGraph* Graph() const;
|
|
|
|
MediaStream* DestinationStream() const;
|
2013-09-17 03:53:40 +04:00
|
|
|
|
|
|
|
// Nodes register here if they will produce sound even if they have silent
|
|
|
|
// or no input connections. The AudioContext will keep registered nodes
|
|
|
|
// alive until the context is collected. This takes care of "playing"
|
|
|
|
// references and "tail-time" references.
|
|
|
|
void RegisterActiveNode(AudioNode* aNode);
|
|
|
|
// Nodes unregister when they have finished producing sound for the
|
|
|
|
// foreseeable future.
|
|
|
|
// Do NOT call UnregisterActiveNode from an AudioNode destructor.
|
|
|
|
// If the destructor is called, then the Node has already been unregistered.
|
|
|
|
// The destructor may be called during hashtable enumeration, during which
|
|
|
|
// unregistering would not be safe.
|
|
|
|
void UnregisterActiveNode(AudioNode* aNode);
|
|
|
|
|
2013-06-10 21:32:28 +04:00
|
|
|
uint32_t MaxChannelCount() const;
|
|
|
|
|
2016-05-06 22:08:58 +03:00
|
|
|
uint32_t ActiveNodeCount() const;
|
|
|
|
|
2013-07-04 04:44:32 +04:00
|
|
|
void Mute() const;
|
|
|
|
void Unmute() const;
|
|
|
|
|
2014-05-09 22:33:17 +04:00
|
|
|
JSObject* GetGlobalJSObject() const;
|
|
|
|
|
2015-09-03 08:12:25 +03:00
|
|
|
void RegisterNode(AudioNode* aNode);
|
|
|
|
void UnregisterNode(AudioNode* aNode);
|
2014-01-15 15:08:20 +04:00
|
|
|
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
void OnStateChanged(void* aPromise, AudioContextState aNewState);
|
|
|
|
|
2015-06-15 19:16:16 +03:00
|
|
|
BasicWaveFormCache* GetBasicWaveFormCache();
|
|
|
|
|
2016-12-15 21:24:41 +03:00
|
|
|
bool CheckClosed(ErrorResult& aRv);
|
|
|
|
|
2017-04-04 14:47:47 +03:00
|
|
|
void Dispatch(already_AddRefed<nsIRunnable>&& aRunnable);
|
|
|
|
|
2013-02-02 02:13:23 +04:00
|
|
|
private:
|
2016-03-11 10:54:31 +03:00
|
|
|
void DisconnectFromWindow();
|
2013-02-02 02:13:23 +04:00
|
|
|
void RemoveFromDecodeQueue(WebAudioDecodeJob* aDecodeJob);
|
2013-08-02 20:07:17 +04:00
|
|
|
void ShutdownDecoder();
|
2013-02-02 02:13:23 +04:00
|
|
|
|
2014-01-04 22:15:41 +04:00
|
|
|
size_t SizeOfIncludingThis(mozilla::MallocSizeOf aMallocSizeOf) const;
|
2016-08-24 08:23:45 +03:00
|
|
|
NS_DECL_NSIMEMORYREPORTER
|
2014-01-04 22:15:41 +04:00
|
|
|
|
2013-02-02 02:13:23 +04:00
|
|
|
friend struct ::mozilla::WebAudioDecodeJob;
|
|
|
|
|
2015-09-16 07:15:21 +03:00
|
|
|
nsTArray<MediaStream*> GetAllStreams() const;
|
|
|
|
|
2012-09-01 00:59:37 +04:00
|
|
|
private:
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
// Each AudioContext has an id, that is passed down the MediaStreams that
|
|
|
|
// back the AudioNodes, so we can easily compute the set of all the
|
|
|
|
// MediaStreams for a given context, on the MediasStreamGraph side.
|
|
|
|
const AudioContextId mId;
|
2013-05-24 21:09:29 +04:00
|
|
|
// Note that it's important for mSampleRate to be initialized before
|
|
|
|
// mDestination, as mDestination's constructor needs to access it!
|
|
|
|
const float mSampleRate;
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
AudioContextState mAudioContextState;
|
2015-10-18 08:24:48 +03:00
|
|
|
RefPtr<AudioDestinationNode> mDestination;
|
|
|
|
RefPtr<AudioListener> mListener;
|
2018-06-27 12:31:02 +03:00
|
|
|
RefPtr<Worklet> mWorklet;
|
2017-01-04 00:52:03 +03:00
|
|
|
nsTArray<UniquePtr<WebAudioDecodeJob> > mDecodeJobs;
|
2017-12-13 23:05:35 +03:00
|
|
|
// This array is used to keep the suspend/close promises alive until
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
// they are resolved, so we can safely pass them accross threads.
|
2015-10-18 08:24:48 +03:00
|
|
|
nsTArray<RefPtr<Promise>> mPromiseGripArray;
|
2017-12-13 23:05:35 +03:00
|
|
|
// This array is used to onlly keep the resume promises alive until they are
|
|
|
|
// resolved, so we can safely pass them accross threads. If the audio context
|
|
|
|
// is not allowed to play, the promise would be pending in this array and be
|
|
|
|
// resolved until audio context has been allowed and user call resume() again.
|
|
|
|
nsTArray<RefPtr<Promise>> mPendingResumePromises;
|
2013-09-17 03:53:40 +04:00
|
|
|
// See RegisterActiveNode. These will keep the AudioContext alive while it
|
|
|
|
// is rendering and the window remains alive.
|
|
|
|
nsTHashtable<nsRefPtrHashKey<AudioNode> > mActiveNodes;
|
2015-09-03 08:12:25 +03:00
|
|
|
// Raw (non-owning) references to all AudioNodes for this AudioContext.
|
|
|
|
nsTHashtable<nsPtrHashKey<AudioNode> > mAllNodes;
|
2015-06-15 19:16:16 +03:00
|
|
|
// Cache to avoid recomputing basic waveforms all the time.
|
2015-10-18 08:24:48 +03:00
|
|
|
RefPtr<BasicWaveFormCache> mBasicWaveFormCache;
|
2013-06-10 21:32:28 +04:00
|
|
|
// Number of channels passed in the OfflineAudioContext ctor.
|
|
|
|
uint32_t mNumberOfChannels;
|
2013-05-17 03:30:41 +04:00
|
|
|
bool mIsOffline;
|
2013-09-16 09:14:45 +04:00
|
|
|
bool mIsStarted;
|
2013-09-18 05:10:30 +04:00
|
|
|
bool mIsShutDown;
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
// Close has been called, reject suspend and resume call.
|
|
|
|
bool mCloseCalled;
|
2015-09-16 07:15:21 +03:00
|
|
|
// Suspend has been called with no following resume.
|
|
|
|
bool mSuspendCalled;
|
2017-03-15 19:36:40 +03:00
|
|
|
bool mIsDisconnecting;
|
2012-09-01 00:59:37 +04:00
|
|
|
};
|
|
|
|
|
Bug 1094764 - Implement AudioContext.suspend and friends. r=roc,ehsan
- Relevant spec text:
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-suspend-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-resume-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-close-Promise
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-state
- http://webaudio.github.io/web-audio-api/#widl-AudioContext-onstatechange
- In a couple words, the behavior we want:
- Closed context cannot have new nodes created, but can do decodeAudioData,
and create buffers, and such.
- OfflineAudioContexts don't support those methods, transitions happen at
startRendering and at the end of processing. onstatechange is used to make
this observable.
- (regular) AudioContexts support those methods. The promises and
onstatechange should be resolved/called when the operation has actually
completed on the rendering thread. Once a context has been closed, it
cannot transition back to "running". An AudioContext switches to "running"
when the audio callback start running, this allow authors to know how long
the audio stack takes to start running.
- MediaStreams that feed in/go out of a suspended graph should respectively
not buffer at the graph input, and output silence
- suspended context should not be doing much on the CPU, and we should try
to pause audio streams if we can (this behaviour is the main reason we need
this in the first place, for saving battery on mobile, and CPU on all
platforms)
- Now, the implementation:
- AudioNodeStreams are now tagged with a context id, to be able to operate
on all the streams of a given AudioContext on the Graph thread without
having to go and lock everytime to touch the AudioContext. This happens in
the AudioNodeStream ctor. IDs are of course constant for the lifetime of the
node.
- When an AudioContext goes into suspended mode, streams for this
AudioContext are moved out of the mStreams array to a second array,
mSuspendedStreams. Streams in mSuspendedStream are not ordered, and are not
processed.
- The MSG will automatically switch to a SystemClockDriver when it finds
that there are no more AudioNodeStream/Stream with an audio track. This is
how pausing the audio subsystem and saving battery works. Subsequently, when
the MSG finds that there are only streams in mSuspendedStreams, it will go
to sleep (block on a monitor), so we save CPU, but it does not shut itself
down. This is mostly not a new behaviour (this is what the MSG does since
the refactoring), but is important to note.
- Promises are gripped (addref-ed) on the main thread, and then shepherd
down other threads and to the GraphDriver, if needed (sometimes we can
resolve them right away). They move between threads as void* to prevent
calling methods on them, as they are not thread safe. Then, the driver
executes the operation, and when it's done (initializing and closing audio
streams can take some time), we send the promise back to the main thread,
and resolve it, casting back to Promise* after asserting we're back on the
main thread. This way, we can send them back on the main thread once an
operation has complete (suspending an audio stream, starting it again on
resume(), etc.), without having to do bookkeeping between suspend calls and
their result. Promises are not thread safe, so we can't move them around
AddRef-ed.
- The stream destruction logic now takes into account that a stream can be
destroyed while not being in mStreams.
- A graph can now switch GraphDriver twice or more per iteration, for
example if an author goes suspend()/resume()/suspend() in the same script.
- Some operation have to be done on suspended stream, so we now use double
for-loop around mSuspendedStreams and mStreams in some places in
MediaStreamGraph.cpp.
- A tricky part was making sure everything worked at AudioContext
boundaries. TrackUnionStream that have one of their input stream suspended
append null ticks instead.
- The graph ordering algorithm had to be altered to not include suspended
streams.
- There are some edge cases (adding a stream on a suspended graph, calling
suspend/resume when a graph has just been close()d).
2015-02-27 20:22:05 +03:00
|
|
|
static const dom::AudioContext::AudioContextId NO_AUDIO_CONTEXT = 0;
|
|
|
|
|
2015-07-13 18:25:42 +03:00
|
|
|
} // namespace dom
|
|
|
|
} // namespace mozilla
|
2012-09-01 00:59:37 +04:00
|
|
|
|
2012-10-31 01:39:38 +04:00
|
|
|
#endif
|
|
|
|
|