blob: 807f009f0076e7ef412c35b73f87e353f98c6267 [file] [log] [blame]
/*
* Copyright (C) 2010, Google Inc. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY APPLE INC. AND ITS CONTRIBUTORS ``AS IS'' AND ANY
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
* DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR ITS CONTRIBUTORS BE LIABLE FOR ANY
* DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "config.h"
#if ENABLE(WEB_AUDIO)
#include "AudioNode.h"
#include "AudioContext.h"
#include "AudioNodeInput.h"
#include "AudioNodeOutput.h"
#include "AudioParam.h"
#include "Logging.h"
#include <wtf/Atomics.h>
#include <wtf/IsoMallocInlines.h>
#include <wtf/MainThread.h>
#if DEBUG_AUDIONODE_REFERENCES
#include <stdio.h>
#endif
namespace WebCore {
WTF_MAKE_ISO_ALLOCATED_IMPL(AudioNode);
String convertEnumerationToString(AudioNode::NodeType enumerationValue)
{
static const NeverDestroyed<String> values[] = {
MAKE_STATIC_STRING_IMPL("NodeTypeUnknown"),
MAKE_STATIC_STRING_IMPL("NodeTypeDestination"),
MAKE_STATIC_STRING_IMPL("NodeTypeOscillator"),
MAKE_STATIC_STRING_IMPL("NodeTypeAudioBufferSource"),
MAKE_STATIC_STRING_IMPL("NodeTypeMediaElementAudioSource"),
MAKE_STATIC_STRING_IMPL("NodeTypeMediaStreamAudioDestination"),
MAKE_STATIC_STRING_IMPL("NodeTypeMediaStreamAudioSource"),
MAKE_STATIC_STRING_IMPL("NodeTypeJavaScript"),
MAKE_STATIC_STRING_IMPL("NodeTypeBiquadFilter"),
MAKE_STATIC_STRING_IMPL("NodeTypePanner"),
MAKE_STATIC_STRING_IMPL("NodeTypeConvolver"),
MAKE_STATIC_STRING_IMPL("NodeTypeDelay"),
MAKE_STATIC_STRING_IMPL("NodeTypeGain"),
MAKE_STATIC_STRING_IMPL("NodeTypeChannelSplitter"),
MAKE_STATIC_STRING_IMPL("NodeTypeChannelMerger"),
MAKE_STATIC_STRING_IMPL("NodeTypeAnalyser"),
MAKE_STATIC_STRING_IMPL("NodeTypeDynamicsCompressor"),
MAKE_STATIC_STRING_IMPL("NodeTypeWaveShaper"),
MAKE_STATIC_STRING_IMPL("NodeTypeBasicInspector"),
MAKE_STATIC_STRING_IMPL("NodeTypeEnd"),
};
static_assert(static_cast<size_t>(AudioNode::NodeTypeUnknown) == 0, "AudioNode::NodeTypeUnknown is not 0 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeDestination) == 1, "AudioNode::NodeTypeDestination is not 1 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeOscillator) == 2, "AudioNode::NodeTypeOscillator is not 2 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeAudioBufferSource) == 3, "AudioNode::NodeTypeAudioBufferSource is not 3 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeMediaElementAudioSource) == 4, "AudioNode::NodeTypeMediaElementAudioSource is not 4 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeMediaStreamAudioDestination) == 5, "AudioNode::NodeTypeMediaStreamAudioDestination is not 5 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeMediaStreamAudioSource) == 6, "AudioNode::NodeTypeMediaStreamAudioSource is not 6 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeJavaScript) == 7, "AudioNode::NodeTypeJavaScript is not 7 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeBiquadFilter) == 8, "AudioNode::NodeTypeBiquadFilter is not 8 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypePanner) == 9, "AudioNode::NodeTypePanner is not 9 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeConvolver) == 10, "AudioNode::NodeTypeConvolver is not 10 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeDelay) == 11, "AudioNode::NodeTypeDelay is not 11 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeGain) == 12, "AudioNode::NodeTypeGain is not 12 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeChannelSplitter) == 13, "AudioNode::NodeTypeChannelSplitter is not 13 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeChannelMerger) == 14, "AudioNode::NodeTypeChannelMerger is not 14 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeAnalyser) == 15, "AudioNode::NodeTypeAnalyser is not 15 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeDynamicsCompressor) == 16, "AudioNode::NodeTypeDynamicsCompressor is not 16 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeWaveShaper) == 17, "AudioNode::NodeTypeWaveShaper is not 17 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeBasicInspector) == 18, "AudioNode::NodeTypeBasicInspector is not 18 as expected");
static_assert(static_cast<size_t>(AudioNode::NodeTypeEnd) == 19, "AudioNode::NodeTypeEnd is not 19 as expected");
ASSERT(static_cast<size_t>(enumerationValue) < WTF_ARRAY_LENGTH(values));
return values[static_cast<size_t>(enumerationValue)];
}
AudioNode::AudioNode(AudioContext& context, float sampleRate)
: m_isInitialized(false)
, m_nodeType(NodeTypeUnknown)
, m_context(context)
, m_sampleRate(sampleRate)
, m_lastProcessingTime(-1)
, m_lastNonSilentTime(-1)
, m_normalRefCount(1) // start out with normal refCount == 1 (like WTF::RefCounted class)
, m_connectionRefCount(0)
, m_isMarkedForDeletion(false)
, m_isDisabled(false)
#if !RELEASE_LOG_DISABLED
, m_logger(context.logger())
, m_logIdentifier(context.nextAudioNodeLogIdentifier())
#endif
, m_channelCount(2)
, m_channelCountMode(Max)
, m_channelInterpretation(AudioBus::Speakers)
{
ALWAYS_LOG(LOGIDENTIFIER);
#if DEBUG_AUDIONODE_REFERENCES
if (!s_isNodeCountInitialized) {
s_isNodeCountInitialized = true;
atexit(AudioNode::printNodeCounts);
}
#endif
}
AudioNode::~AudioNode()
{
ALWAYS_LOG(LOGIDENTIFIER);
ASSERT(isMainThread());
#if DEBUG_AUDIONODE_REFERENCES
--s_nodeCount[nodeType()];
fprintf(stderr, "%p: %d: AudioNode::~AudioNode() %d %d\n", this, nodeType(), m_normalRefCount.load(), m_connectionRefCount);
#endif
}
void AudioNode::initialize()
{
m_isInitialized = true;
}
void AudioNode::uninitialize()
{
m_isInitialized = false;
}
void AudioNode::setNodeType(NodeType type)
{
ASSERT(isMainThread());
ALWAYS_LOG(LOGIDENTIFIER, type);
m_nodeType = type;
#if DEBUG_AUDIONODE_REFERENCES
++s_nodeCount[type];
#endif
}
void AudioNode::lazyInitialize()
{
if (!isInitialized())
initialize();
}
void AudioNode::addInput(std::unique_ptr<AudioNodeInput> input)
{
ASSERT(isMainThread());
INFO_LOG(LOGIDENTIFIER, input->node()->nodeType());
m_inputs.append(WTFMove(input));
}
void AudioNode::addOutput(std::unique_ptr<AudioNodeOutput> output)
{
ASSERT(isMainThread());
INFO_LOG(LOGIDENTIFIER, output->node()->nodeType());
m_outputs.append(WTFMove(output));
}
AudioNodeInput* AudioNode::input(unsigned i)
{
if (i < m_inputs.size())
return m_inputs[i].get();
return nullptr;
}
AudioNodeOutput* AudioNode::output(unsigned i)
{
if (i < m_outputs.size())
return m_outputs[i].get();
return nullptr;
}
ExceptionOr<void> AudioNode::connect(AudioNode& destination, unsigned outputIndex, unsigned inputIndex)
{
ASSERT(isMainThread());
AudioContext::AutoLocker locker(context());
ALWAYS_LOG(LOGIDENTIFIER, destination.nodeType(), ", output = ", outputIndex, ", input = ", inputIndex);
// Sanity check input and output indices.
if (outputIndex >= numberOfOutputs())
return Exception { IndexSizeError };
if (inputIndex >= destination.numberOfInputs())
return Exception { IndexSizeError };
if (context() != destination.context())
return Exception { SyntaxError };
auto* input = destination.input(inputIndex);
auto* output = this->output(outputIndex);
input->connect(output);
// Let context know that a connection has been made.
context().incrementConnectionCount();
return { };
}
ExceptionOr<void> AudioNode::connect(AudioParam& param, unsigned outputIndex)
{
AudioContext::AutoLocker locker(context());
ASSERT(isMainThread());
INFO_LOG(LOGIDENTIFIER, param.name(), ", output = ", outputIndex);
if (outputIndex >= numberOfOutputs())
return Exception { IndexSizeError };
if (context() != param.context())
return Exception { SyntaxError };
auto* output = this->output(outputIndex);
param.connect(output);
return { };
}
ExceptionOr<void> AudioNode::disconnect(unsigned outputIndex)
{
ASSERT(isMainThread());
AudioContext::AutoLocker locker(context());
// Sanity check input and output indices.
if (outputIndex >= numberOfOutputs())
return Exception { IndexSizeError };
auto* output = this->output(outputIndex);
INFO_LOG(LOGIDENTIFIER, output->node()->nodeType());
output->disconnectAll();
return { };
}
unsigned AudioNode::channelCount()
{
return m_channelCount;
}
ExceptionOr<void> AudioNode::setChannelCount(unsigned channelCount)
{
ASSERT(isMainThread());
AudioContext::AutoLocker locker(context());
ALWAYS_LOG(LOGIDENTIFIER, channelCount);
if (!(channelCount > 0 && channelCount <= AudioContext::maxNumberOfChannels()))
return Exception { InvalidStateError };
if (m_channelCount == channelCount)
return { };
m_channelCount = channelCount;
if (m_channelCountMode != Max)
updateChannelsForInputs();
return { };
}
String AudioNode::channelCountMode()
{
switch (m_channelCountMode) {
case Max:
return "max"_s;
case ClampedMax:
return "clamped-max"_s;
case Explicit:
return "explicit"_s;
}
ASSERT_NOT_REACHED();
return emptyString();
}
ExceptionOr<void> AudioNode::setChannelCountMode(const String& mode)
{
ASSERT(isMainThread());
AudioContext::AutoLocker locker(context());
ALWAYS_LOG(LOGIDENTIFIER, mode);
ChannelCountMode oldMode = m_channelCountMode;
if (mode == "max")
m_channelCountMode = Max;
else if (mode == "clamped-max")
m_channelCountMode = ClampedMax;
else if (mode == "explicit")
m_channelCountMode = Explicit;
else
return Exception { InvalidStateError };
if (m_channelCountMode != oldMode)
updateChannelsForInputs();
return { };
}
String AudioNode::channelInterpretation()
{
switch (m_channelInterpretation) {
case AudioBus::Speakers:
return "speakers"_s;
case AudioBus::Discrete:
return "discrete"_s;
}
ASSERT_NOT_REACHED();
return emptyString();
}
ExceptionOr<void> AudioNode::setChannelInterpretation(const String& interpretation)
{
ASSERT(isMainThread());
AudioContext::AutoLocker locker(context());
ALWAYS_LOG(LOGIDENTIFIER, interpretation);
if (interpretation == "speakers")
m_channelInterpretation = AudioBus::Speakers;
else if (interpretation == "discrete")
m_channelInterpretation = AudioBus::Discrete;
else
return Exception { InvalidStateError };
return { };
}
void AudioNode::updateChannelsForInputs()
{
for (auto& input : m_inputs)
input->changedOutputs();
}
EventTargetInterface AudioNode::eventTargetInterface() const
{
return AudioNodeEventTargetInterfaceType;
}
ScriptExecutionContext* AudioNode::scriptExecutionContext() const
{
return static_cast<ActiveDOMObject&>(const_cast<AudioNode*>(this)->context()).scriptExecutionContext();
}
void AudioNode::processIfNecessary(size_t framesToProcess)
{
ASSERT(context().isAudioThread());
if (!isInitialized())
return;
// Ensure that we only process once per rendering quantum.
// This handles the "fanout" problem where an output is connected to multiple inputs.
// The first time we're called during this time slice we process, but after that we don't want to re-process,
// instead our output(s) will already have the results cached in their bus;
double currentTime = context().currentTime();
if (m_lastProcessingTime != currentTime) {
m_lastProcessingTime = currentTime; // important to first update this time because of feedback loops in the rendering graph
pullInputs(framesToProcess);
bool silentInputs = inputsAreSilent();
if (!silentInputs)
m_lastNonSilentTime = (context().currentSampleFrame() + framesToProcess) / static_cast<double>(m_sampleRate);
if (silentInputs && propagatesSilence())
silenceOutputs();
else
process(framesToProcess);
}
}
void AudioNode::checkNumberOfChannelsForInput(AudioNodeInput* input)
{
ASSERT(context().isAudioThread() && context().isGraphOwner());
for (auto& savedInput : m_inputs) {
if (input == savedInput.get()) {
input->updateInternalBus();
return;
}
}
ASSERT_NOT_REACHED();
}
bool AudioNode::propagatesSilence() const
{
return m_lastNonSilentTime + latencyTime() + tailTime() < context().currentTime();
}
void AudioNode::pullInputs(size_t framesToProcess)
{
ASSERT(context().isAudioThread());
// Process all of the AudioNodes connected to our inputs.
for (auto& input : m_inputs)
input->pull(0, framesToProcess);
}
bool AudioNode::inputsAreSilent()
{
for (auto& input : m_inputs) {
if (!input->bus()->isSilent())
return false;
}
return true;
}
void AudioNode::silenceOutputs()
{
for (auto& output : m_outputs)
output->bus()->zero();
}
void AudioNode::enableOutputsIfNecessary()
{
if (m_isDisabled && m_connectionRefCount > 0) {
ASSERT(isMainThread());
AudioContext::AutoLocker locker(context());
m_isDisabled = false;
for (auto& output : m_outputs)
output->enable();
}
}
void AudioNode::disableOutputsIfNecessary()
{
// Disable outputs if appropriate. We do this if the number of connections is 0 or 1. The case
// of 0 is from finishDeref() where there are no connections left. The case of 1 is from
// AudioNodeInput::disable() where we want to disable outputs when there's only one connection
// left because we're ready to go away, but can't quite yet.
if (m_connectionRefCount <= 1 && !m_isDisabled) {
// Still may have JavaScript references, but no more "active" connection references, so put all of our outputs in a "dormant" disabled state.
// Garbage collection may take a very long time after this time, so the "dormant" disabled nodes should not bog down the rendering...
// As far as JavaScript is concerned, our outputs must still appear to be connected.
// But internally our outputs should be disabled from the inputs they're connected to.
// disable() can recursively deref connections (and call disable()) down a whole chain of connected nodes.
// FIXME: we special case the convolver and delay since they have a significant tail-time and shouldn't be disconnected simply
// because they no longer have any input connections. This needs to be handled more generally where AudioNodes have
// a tailTime attribute. Then the AudioNode only needs to remain "active" for tailTime seconds after there are no
// longer any active connections.
if (nodeType() != NodeTypeConvolver && nodeType() != NodeTypeDelay) {
m_isDisabled = true;
for (auto& output : m_outputs)
output->disable();
}
}
}
void AudioNode::ref(RefType refType)
{
switch (refType) {
case RefTypeNormal:
++m_normalRefCount;
break;
case RefTypeConnection:
++m_connectionRefCount;
break;
default:
ASSERT_NOT_REACHED();
}
#if DEBUG_AUDIONODE_REFERENCES
fprintf(stderr, "%p: %d: AudioNode::ref(%d) %d %d\n", this, nodeType(), refType, m_normalRefCount, m_connectionRefCount);
#endif
// See the disabling code in finishDeref() below. This handles the case where a node
// is being re-connected after being used at least once and disconnected.
// In this case, we need to re-enable.
if (refType == RefTypeConnection)
enableOutputsIfNecessary();
}
void AudioNode::deref(RefType refType)
{
// The actually work for deref happens completely within the audio context's graph lock.
// In the case of the audio thread, we must use a tryLock to avoid glitches.
bool hasLock = false;
bool mustReleaseLock = false;
if (context().isAudioThread()) {
// Real-time audio thread must not contend lock (to avoid glitches).
hasLock = context().tryLock(mustReleaseLock);
} else {
context().lock(mustReleaseLock);
hasLock = true;
}
if (hasLock) {
// This is where the real deref work happens.
finishDeref(refType);
if (mustReleaseLock)
context().unlock();
} else {
// We were unable to get the lock, so put this in a list to finish up later.
ASSERT(context().isAudioThread());
ASSERT(refType == RefTypeConnection);
context().addDeferredFinishDeref(this);
}
// Once AudioContext::uninitialize() is called there's no more chances for deleteMarkedNodes() to get called, so we call here.
// We can't call in AudioContext::~AudioContext() since it will never be called as long as any AudioNode is alive
// because AudioNodes keep a reference to the context.
if (context().isAudioThreadFinished())
context().deleteMarkedNodes();
}
void AudioNode::finishDeref(RefType refType)
{
ASSERT(context().isGraphOwner());
switch (refType) {
case RefTypeNormal:
ASSERT(m_normalRefCount > 0);
--m_normalRefCount;
break;
case RefTypeConnection:
ASSERT(m_connectionRefCount > 0);
--m_connectionRefCount;
break;
default:
ASSERT_NOT_REACHED();
}
#if DEBUG_AUDIONODE_REFERENCES
fprintf(stderr, "%p: %d: AudioNode::deref(%d) %d %d\n", this, nodeType(), refType, m_normalRefCount, m_connectionRefCount);
#endif
if (!m_connectionRefCount) {
if (!m_normalRefCount) {
if (!m_isMarkedForDeletion) {
// All references are gone - we need to go away.
for (auto& output : m_outputs)
output->disconnectAll(); // This will deref() nodes we're connected to.
// Mark for deletion at end of each render quantum or when context shuts down.
context().markForDeletion(*this);
m_isMarkedForDeletion = true;
}
} else if (refType == RefTypeConnection)
disableOutputsIfNecessary();
}
}
#if DEBUG_AUDIONODE_REFERENCES
bool AudioNode::s_isNodeCountInitialized = false;
int AudioNode::s_nodeCount[NodeTypeEnd];
void AudioNode::printNodeCounts()
{
fprintf(stderr, "\n\n");
fprintf(stderr, "===========================\n");
fprintf(stderr, "AudioNode: reference counts\n");
fprintf(stderr, "===========================\n");
for (unsigned i = 0; i < NodeTypeEnd; ++i)
fprintf(stderr, "%d: %d\n", i, s_nodeCount[i]);
fprintf(stderr, "===========================\n\n\n");
}
#endif // DEBUG_AUDIONODE_REFERENCES
#if !RELEASE_LOG_DISABLED
WTFLogChannel& AudioNode::logChannel() const
{
return LogMedia;
}
#endif
} // namespace WebCore
#endif // ENABLE(WEB_AUDIO)