Source code
Revision control
Copy as Markdown
// Copyright (C) 2016 The Qt Company Ltd.
// SPDX-License-Identifier: LicenseRef-Qt-Commercial OR LGPL-3.0-only OR GPL-2.0-only OR GPL-3.0-only
#include "qaudiosink.h"
#include <QtMultimedia/private/qtmm_alignment_support_p.h>
#include <QtMultimedia/private/qaudiosystem_p.h>
#include <QtMultimedia/private/qaudiohelpers_p.h>
#include <QtMultimedia/private/qaudiosystem_p.h>
#include <QtMultimedia/private/qplatformaudiodevices_p.h>
#include <QtMultimedia/private/qplatformmediaintegration_p.h>
#include <QtMultimedia/qaudio.h>
#include <QtMultimedia/qaudiodevice.h>
QT_BEGIN_NAMESPACE
/*!
\class QAudioSink
\brief The QAudioSink class provides an interface for sending audio data to
an audio output device.
\inmodule QtMultimedia
\ingroup multimedia
\ingroup multimedia_audio
You can construct an audio output with the system's
default audio output device. It is also possible to
create QAudioSink with a specific QAudioDevice. When
you create the audio output, you should also send in
the QAudioFormat to be used for the playback (see
the QAudioFormat class description for details).
QAudioSink can be used in two different modes:
\list
\li Using a QIODevice from an application thread
\li Using a callback-based interface from the audio thread
\endlist
\section1 QIODevice interface
Starting to play an audio stream is simply a matter of calling
start() with a QIODevice. QAudioSink will then fetch the data it
needs from the io device. So playing back an audio file is as
simple as:
\snippet multimedia-snippets/audio.cpp Audio output class members
\snippet multimedia-snippets/audio.cpp Audio output setup
The file will start playing assuming that the audio system and
output device support it. If you run out of luck, check what's
up with the error() function.
After the file has finished playing, we need to stop the device:
\snippet multimedia-snippets/audio.cpp Audio output stop
At any given time, the QAudioSink will be in one of four states:
active, suspended, stopped, or idle. These states are described
by the QtAudio::State enum.
\section2 Threading model and buffering
The QIODevice interface is designed to be used from the application thread.
A wait-free ringbuffer is used to communicate to the audio thread. The size
of this ringbuffer can be configured with setBufferSize() and defaults to
250ms. The state of this buffer can be queried with bytesFree(). If the
ringbuffer runs out of data, the audio thread will send silence to the audio
device and the state will change to QtAudio::IdleState and resume to QtAudio::ActiveState
when more data is available from the QIODevice.
\section1 Callback interface
The preferred way to achieve low audio latency is to use the callback-based interface.
It allows you to write audio data directly to the audio device without having to go through
a QIODevice. This is done by calling start() with a callback function that will be called
from the audio thread. This callback function will be called with a QSpan<SampleType> whenever
the audio backend requires data.
\snippet multimedia-snippets/audio.cpp Audio callback output class members
\snippet multimedia-snippets/audio.cpp Audio callback output setup sine
Unlike the QIODevice-based interface, the QAudioSink can only be in the states active,
suspendend and stopped. The setBufferSize() API is not available when using the callback,
the size of the callback argument is determined by the audio backend.
\qtmmaudiocallbacksupportednote
\qtmmaudiocallbacknote
\section1 State and error handling
State changes are reported through the stateChanged() signal. You
can use this signal to, for instance, update the GUI of the
application; the mundane example here being changing the state of
a \c { play/pause } button. You request a state change directly
with suspend(), stop(), reset(), resume(), and start().
The QAudioSink will enter the \l{QtAudio::}{StoppedState} when an error is encountered.
The \l{QtAudio::Error}{error type} can be retrieved with the error() function. Please see the
QtAudio::Error enum for a description of the possible errors that are reported. Calling stop()
or reset() will reset the error state to \l{QtAudio::Error}{NoError}.
You can check for errors by connecting to the stateChanged()
signal:
\snippet multimedia-snippets/audio.cpp Audio output state changed
\sa QAudioSource, QAudioDevice
*/
/*!
Construct a new audio output and attach it to \a parent.
The default audio output device is used with the output
\a format parameters. If \a format is default-initialized,
the format will be set to the preferred format of the audio device.
*/
QAudioSink::QAudioSink(const QAudioFormat &format, QObject *parent)
: QAudioSink({}, format, parent)
{
}
/*!
Construct a new audio output and attach it to \a parent.
The device referenced by \a audioDevice is used with the output
\a format parameters. If \a format is default-initialized,
the format will be set to the preferred format of \a audioDevice.
*/
QAudioSink::QAudioSink(const QAudioDevice &audioDevice, const QAudioFormat &format, QObject *parent):
QObject(parent)
{
d = QPlatformMediaIntegration::instance()->audioDevices()->audioOutputDevice(format,
audioDevice, this);
if (d)
connect(d, &QPlatformAudioSink::stateChanged, this, &QAudioSink::stateChanged);
else
qWarning("No audio device detected");
}
/*!
\fn bool QAudioSink::isNull() const
Returns \c true is the QAudioSink instance is \c null, otherwise returns
\c false.
*/
/*!
Destroys this audio output.
This will release any system resources used and free any buffers.
*/
QAudioSink::~QAudioSink()
{
delete d;
}
/*!
Returns the QAudioFormat being used.
*/
QAudioFormat QAudioSink::format() const
{
return d ? d->format() : QAudioFormat();
}
static bool validateFormatAtStart(QPlatformAudioSink *d)
{
if (!d->format().isValid()) {
qWarning() << "QAudioSink::start: QAudioFormat not valid";
d->setError(QAudio::OpenError);
return false;
}
if (!d->isFormatSupported(d->format())) {
qWarning() << "QAudioSink::start: QAudioFormat not supported by QAudioDevice";
d->setError(QAudio::OpenError);
return false;
}
return true;
};
/*!
Starts transferring audio data from the \a device to the system's audio output.
The \a device must have been opened in the \l{QIODevice::ReadOnly}{ReadOnly} or
\l{QIODevice::ReadWrite}{ReadWrite} modes.
If the QAudioSink is able to successfully output audio data, state() returns
QtAudio::ActiveState, error() returns QtAudio::NoError
and the stateChanged() signal is emitted.
If a problem occurs during this process, error() returns QtAudio::OpenError,
state() returns QtAudio::StoppedState and the stateChanged() signal is emitted.
\sa QIODevice, {QAudioSink#Callback interface}{QIODevice interface}
*/
void QAudioSink::start(QIODevice* device)
{
if (!d)
return;
d->setError(QAudio::NoError);
if (!device->isReadable()) {
qWarning() << "QAudioSink::start: QIODevice is not readable";
d->setError(QAudio::OpenError);
return;
}
if (!validateFormatAtStart(d))
return;
d->elapsedTime.start();
d->start(device);
}
/*!
Returns a pointer to the internal QIODevice being used to transfer data to
the system's audio output. The device will already be open and
\l{QIODevice::write()}{write()} can write data directly to it.
\note The pointer will become invalid after the stream is stopped or
if you start another stream.
If the QAudioSink is able to access the system's audio device, state() returns
QtAudio::IdleState, error() returns QtAudio::NoError
and the stateChanged() signal is emitted.
If a problem occurs during this process, error() returns QtAudio::OpenError,
state() returns QtAudio::StoppedState and the stateChanged() signal is emitted.
\sa QIODevice, {QAudioSink#Callback interface}{QIODevice interface}
*/
QIODevice* QAudioSink::start()
{
if (!d)
return nullptr;
d->setError(QAudio::NoError);
if (!validateFormatAtStart(d))
return nullptr;
d->elapsedTime.start();
return d->start();
}
/*!
\fn template <typename Callback, QtAudio::if_audio_sink_callback<Callback> = true> void QAudioSink::start(Callback &&)
Starts the QAudioSink with a callback function that will be called on a soft-realtime audio
thread. The callback is a callable that takes a QSpan<SampleType> as an argument, SampleType has
to match the QAudioFormat::SampleFormat of the QAudioSink's format. The span needs to be filled
with interleaved audio data.
If the QAudioSink is able to successfully start, error() returns QtAudio::NoError.
If a problem occurs during this process, error() returns QtAudio::OpenError, state() returns
QtAudio::StoppedState and the stateChanged() signal is emitted.
\qtmmaudiocallbacksupportednote
\qtmmaudiocallbacknote
\sa {QAudioSink#Callback interface}{Callback interface}
\since 6.11
*/
template <typename T>
void QAudioSink::startImpl(T &&callback)
{
if (!d)
return;
if (!d->hasCallbackAPI()) {
qWarning() << "QAudioSink::start: Callback API not supported on this platform";
d->setError(QAudio::OpenError);
return;
}
using namespace QtMultimediaPrivate;
if (!validateAudioCallback(callback, format())) {
d->setError(QAudio::OpenError);
return;
}
if (!validateFormatAtStart(d))
return;
d->elapsedTime.start();
d->start(std::forward<T>(callback));
}
void QAudioSink::startABIImpl(QtAudioPrivate::AudioSinkCallback &&callback)
{
return QAudioSink::startImpl(QtMultimediaPrivate::asAudioSinkCallback(std::move(callback)));
}
/*!
Stops the audio output, detaching from the system resource.
Sets error() to QtAudio::NoError, state() to QtAudio::StoppedState and
emit stateChanged() signal.
\note On Linux, and Darwin, this operation synchronously drains the
underlying audio buffer, which may cause delays accordingly to the
buffer payload. To reset all the buffers immediately, use the method
\l reset instead.
\sa reset()
*/
void QAudioSink::stop()
{
if (d)
d->stop();
}
/*!
Immediately halts audio output and discards any audio data currently in the buffers. All pending
audio data pushed to QIODevice is ignored.
\sa stop()
*/
void QAudioSink::reset()
{
if (d)
d->reset();
}
/*!
Stops processing audio data, preserving buffered audio data.
Sets state() to QtAudio::SuspendedState and emits stateChanged() signal.
*/
void QAudioSink::suspend()
{
if (d)
d->suspend();
}
/*!
Resumes processing audio data after a suspend().
Sets state() to the state the sink had when suspend() was called. This function does nothing if
the audio sink's state is not QtAudio::SuspendedState.
*/
void QAudioSink::resume()
{
if (d)
d->resume();
}
/*!
Returns the number of free bytes available in the audio buffer.
\note The returned value is only valid while in QtAudio::ActiveState or QtAudio::IdleState
state, otherwise returns zero.
\sa framesFree
*/
qsizetype QAudioSink::bytesFree() const
{
return d ? d->bytesFree() : 0;
}
/*!
Returns the number of free frames available in the audio buffer.
\note The returned value is only valid while in QtAudio::ActiveState or QtAudio::IdleState
state, otherwise returns zero.
\sa bytesFree
\since 6.10
*/
qsizetype QAudioSink::framesFree() const
{
return d ? d->format().framesForBytes(bytesFree()) : 0;
}
/*!
Sets the audio buffer size to \a value in bytes.
\note This function can be called anytime before start(). Calls to this
are ignored after start(). It should not be assumed that the buffer size
set is the actual buffer size used - call bufferSize() anytime after start()
to return the actual buffer size being used.
The bufferFrameCount() and bufferSize() properties denote the ringbuffer size
when using the QIODevice APIs. The native period size is controlled via setNativePeriodFrameCount().
\sa setBufferFrameCount
*/
void QAudioSink::setBufferSize(qsizetype value)
{
if (d)
d->setBufferSize(value);
}
/*!
Returns the audio buffer size in bytes.
If called before \l start(), returns platform default value.
If called before \c start() but \l setBufferSize() or \l setBufferFrameCount() was called prior, returns
value set by \c setBufferSize() or \c setBufferFrameCount(). If called after \c start(), returns the actual
buffer size being used. This may not be what was set previously by
\c setBufferSize() or \c setBufferFrameCount().
\sa bufferFrameCount
*/
qsizetype QAudioSink::bufferSize() const
{
return d ? d->bufferSize() : 0;
}
/*!
Sets the audio buffer size to \a value in frame count.
\note This function can be called anytime before start(). Calls to this
are ignored after start(). It should not be assumed that the buffer size
set is the actual buffer size used - call bufferFrameCount() anytime after
start() to return the actual buffer size being used.
The bufferFrameCount() and bufferSize() properties denote the ringbuffer size
when using the QIODevice APIs. The native period size is controlled via setNativePeriodFrameCount().
\sa setBufferSize
\since 6.10
*/
void QAudioSink::setBufferFrameCount(qsizetype value)
{
if (d)
setBufferSize(d->format().bytesForFrames(value));
}
/*!
Returns the audio buffer size in frames.
If called before \l start(), returns platform default value.
If called before \c start() but \l setBufferSize() or \l setBufferFrameCount() was called prior, returns
value set by \c setBufferSize() or \c setBufferFrameCount(). If called after \c start(), returns the actual
buffer size being used. This may not be what was set previously by
\c setBufferSize() or \c setBufferFrameCount().
\sa bufferSize
\since 6.10
*/
qsizetype QAudioSink::bufferFrameCount() const
{
return d ? d->format().framesForBytes(bufferSize()) : 0;
}
/*!
Sets the native period frame count to \a frameCount.
Tunes the audio buffer size used by the operating system and hardware. By default (when not set),
the system uses a platform-dependent value, typically 1024 frames (~23ms at 44100 Hz).
Use smaller values (64-256) to reduce latency, or larger values (2048-4096) to reduce the risk of
buffer underruns (dropouts).
Valid values are powers of 2 between 32 and 4096 (inclusive), or -1 to unset. Can only be called
when the audio sink is stopped.
Maps to platform-specific settings: \c{kAudioDevicePropertyBufferFrameSize} (macOS),
\c{IAudioClient::bufferDuration} (Windows), \c{PW_KEY_NODE_FORCE_QUANTUM} (PipeWire/Linux),
and \c{AAudioStreamBuilder_setBufferCapacityInFrames} (Android).
\note This is separate from the ringbuffer configured with setBufferSize() or setBufferFrameCount().
\since 6.12
\sa nativePeriodFrameCount
*/
void QAudioSink::setNativePeriodFrameCount(qsizetype frameCount)
{
if (!d)
return;
// Allow -1 as special case (unset)
if (frameCount != -1) {
if (state() != QAudio::StoppedState) {
qWarning("QAudioSink::setNativePeriodFrameCount: can only be set when stopped");
return;
}
if (frameCount < 32 || frameCount > 4096
|| !QtMultimediaPrivate::isPowerOfTwo(frameCount)) {
qWarning("QAudioSink::setNativePeriodFrameCount: invalid frame count %lld "
"(must be -1 or a power of 2 between 32 and 4096)",
qint64(frameCount));
return;
}
}
if (frameCount != -1)
d->setNativePeriodFrames(QtMultimediaPrivate::NativePeriodFrames(frameCount));
else
d->setNativePeriodFrames(std::nullopt);
}
/*!
Returns the native period frame count.
Returns -1 if not set, otherwise returns the value previously set with
setNativePeriodFrameCount().
\since 6.12
\sa setNativePeriodFrameCount
*/
qsizetype QAudioSink::nativePeriodFrameCount() const
{
std::optional hwbf = d ? d->nativePeriodFrames() : std::nullopt;
return hwbf ? qsizetype(qToUnderlying(*hwbf)) : qsizetype(-1);
}
/*!
Returns the amount of audio data processed since start()
was called (in microseconds).
*/
qint64 QAudioSink::processedUSecs() const
{
return d ? d->processedUSecs() : 0;
}
/*!
Returns the microseconds since start() was called, including time in Idle and
Suspend states.
*/
qint64 QAudioSink::elapsedUSecs() const
{
return state() == QAudio::StoppedState ? 0 : d->elapsedTime.nsecsElapsed()/1000;
}
/*!
Returns the error state.
*/
QtAudio::Error QAudioSink::error() const
{
return d ? d->error() : QAudio::OpenError;
}
/*!
Returns the state of audio processing.
*/
QtAudio::State QAudioSink::state() const
{
return d ? d->state() : QAudio::StoppedState;
}
/*!
Sets the output volume to \a volume.
The volume is scaled linearly from \c 0.0 (silence) to \c 1.0 (full volume).
Values outside this range will be clamped.
The default volume is \c 1.0.
\note Adjustments to the volume will change the volume of this audio stream,
not the global volume.
UI volume controls should usually be scaled non-linearly. For example, using
a logarithmic scale will produce linear changes in perceived loudness, which
is what a user would normally expect from a volume control. See
QtAudio::convertVolume() for more details.
*/
void QAudioSink::setVolume(qreal volume)
{
if (!d)
return;
std::optional<float> newVolume = QAudioHelperInternal::sanitizeVolume(volume, this->volume());
if (newVolume)
d->setVolume(*newVolume);
}
/*!
Returns the volume between 0.0 and 1.0 inclusive.
*/
qreal QAudioSink::volume() const
{
return d ? d->volume() : 1.0;
}
/*!
\fn QAudioSink::stateChanged(QtAudio::State state)
This signal is emitted when the device \a state has changed.
This is the current state of the audio output.
\note The QtAudio namespace was named QAudio up to and including Qt 6.6.
String-based connections to this signal have to use \c{QAudio::State} as
the parameter type: \c{connect(source, SIGNAL(stateChanged(QAudio::State)), ...);}
*/
QT_END_NAMESPACE
#include "moc_qaudiosink.cpp"