2 * Copyright (C) 2010, Google Inc. All rights reserved.
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
7 * 1. Redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution.
13 * THIS SOFTWARE IS PROVIDED BY APPLE INC. AND ITS CONTRIBUTORS ``AS IS'' AND ANY
14 * EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
15 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
16 * DISCLAIMED. IN NO EVENT SHALL APPLE INC. OR ITS CONTRIBUTORS BE LIABLE FOR ANY
17 * DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
18 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
19 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
20 * ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
21 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
22 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
29 #include "AudioBufferSourceNode.h"
31 #include "AudioContext.h"
32 #include "AudioNodeOutput.h"
33 #include "AudioUtilities.h"
34 #include "FloatConversion.h"
35 #include "ScriptCallStack.h"
36 #include "ScriptExecutionContext.h"
38 #include <wtf/MainThread.h>
39 #include <wtf/MathExtras.h>
45 const double DefaultGrainDuration = 0.020; // 20ms
47 // Arbitrary upper limit on playback rate.
48 // Higher than expected rates can be useful when playing back oversampled buffers
49 // to minimize linear interpolation aliasing.
50 const double MaxRate = 1024;
52 PassRefPtr<AudioBufferSourceNode> AudioBufferSourceNode::create(AudioContext* context, float sampleRate)
54 return adoptRef(new AudioBufferSourceNode(context, sampleRate));
57 AudioBufferSourceNode::AudioBufferSourceNode(AudioContext* context, float sampleRate)
58 : AudioScheduledSourceNode(context, sampleRate)
63 , m_virtualReadIndex(0)
66 , m_grainDuration(DefaultGrainDuration)
70 setNodeType(NodeTypeAudioBufferSource);
72 m_gain = AudioGain::create(context, "gain", 1.0, 0.0, 1.0);
73 m_playbackRate = AudioParam::create(context, "playbackRate", 1.0, 0.0, MaxRate);
75 // Default to mono. A call to setBuffer() will set the number of output channels to that of the buffer.
76 addOutput(adoptPtr(new AudioNodeOutput(this, 1)));
81 AudioBufferSourceNode::~AudioBufferSourceNode()
87 void AudioBufferSourceNode::process(size_t framesToProcess)
89 AudioBus* outputBus = output(0)->bus();
91 if (!isInitialized()) {
96 // The audio thread can't block on this lock, so we call tryLock() instead.
97 MutexTryLocker tryLocker(m_processLock);
98 if (tryLocker.locked()) {
104 // After calling setBuffer() with a buffer having a different number of channels, there can in rare cases be a slight delay
105 // before the output bus is updated to the new number of channels because of use of tryLocks() in the context's updating system.
106 // In this case, if the the buffer has just been changed and we're not quite ready yet, then just output silence.
107 if (numberOfChannels() != buffer()->numberOfChannels()) {
112 size_t quantumFrameOffset;
113 size_t bufferFramesToProcess;
115 updateSchedulingInfo(framesToProcess,
118 bufferFramesToProcess);
120 if (!bufferFramesToProcess) {
125 for (unsigned i = 0; i < outputBus->numberOfChannels(); ++i)
126 m_destinationChannels[i] = outputBus->channel(i)->mutableData();
128 // Render by reading directly from the buffer.
129 if (!renderFromBuffer(outputBus, quantumFrameOffset, bufferFramesToProcess)) {
134 // Apply the gain (in-place) to the output bus.
135 float totalGain = gain()->value() * m_buffer->gain();
136 outputBus->copyWithGainFrom(*outputBus, &m_lastGain, totalGain);
137 outputBus->clearSilentFlag();
139 // Too bad - the tryLock() failed. We must be in the middle of changing buffers and were already outputting silence anyway.
144 // Returns true if we're finished.
145 bool AudioBufferSourceNode::renderSilenceAndFinishIfNotLooping(AudioBus*, unsigned index, size_t framesToProcess)
148 // If we're not looping, then stop playing when we get to the end.
150 if (framesToProcess > 0) {
151 // We're not looping and we've reached the end of the sample data, but we still need to provide more output,
152 // so generate silence for the remaining.
153 for (unsigned i = 0; i < numberOfChannels(); ++i)
154 memset(m_destinationChannels[i] + index, 0, sizeof(float) * framesToProcess);
163 bool AudioBufferSourceNode::renderFromBuffer(AudioBus* bus, unsigned destinationFrameOffset, size_t numberOfFrames)
165 ASSERT(context()->isAudioThread());
167 // Basic sanity checking
170 if (!bus || !buffer())
173 unsigned numberOfChannels = this->numberOfChannels();
174 unsigned busNumberOfChannels = bus->numberOfChannels();
176 bool channelCountGood = numberOfChannels && numberOfChannels == busNumberOfChannels;
177 ASSERT(channelCountGood);
178 if (!channelCountGood)
181 // Sanity check destinationFrameOffset, numberOfFrames.
182 size_t destinationLength = bus->length();
184 bool isLengthGood = destinationLength <= 4096 && numberOfFrames <= 4096;
185 ASSERT(isLengthGood);
189 bool isOffsetGood = destinationFrameOffset <= destinationLength && destinationFrameOffset + numberOfFrames <= destinationLength;
190 ASSERT(isOffsetGood);
194 // Potentially zero out initial frames leading up to the offset.
195 if (destinationFrameOffset) {
196 for (unsigned i = 0; i < numberOfChannels; ++i)
197 memset(m_destinationChannels[i], 0, sizeof(float) * destinationFrameOffset);
200 // Offset the pointers to the correct offset frame.
201 unsigned writeIndex = destinationFrameOffset;
203 size_t bufferLength = buffer()->length();
204 double bufferSampleRate = buffer()->sampleRate();
206 // Avoid converting from time to sample-frames twice by computing
207 // the grain end time first before computing the sample frame.
208 unsigned endFrame = m_isGrain ? AudioUtilities::timeToSampleFrame(m_grainOffset + m_grainDuration, bufferSampleRate) : bufferLength;
210 // This is a HACK to allow for HRTF tail-time - avoids glitch at end.
211 // FIXME: implement tailTime for each AudioNode for a more general solution to this problem.
212 // https://bugs.webkit.org/show_bug.cgi?id=77224
216 // Do some sanity checking.
217 if (endFrame > bufferLength)
218 endFrame = bufferLength;
219 if (m_virtualReadIndex >= endFrame)
220 m_virtualReadIndex = 0; // reset to start
222 // If the .loop attribute is true, then values of m_loopStart == 0 && m_loopEnd == 0 implies
223 // that we should use the entire buffer as the loop, otherwise use the loop values in m_loopStart and m_loopEnd.
224 double virtualEndFrame = endFrame;
225 double virtualDeltaFrames = endFrame;
227 if (loop() && (m_loopStart || m_loopEnd) && m_loopStart >= 0 && m_loopEnd > 0 && m_loopStart < m_loopEnd) {
228 // Convert from seconds to sample-frames.
229 double loopStartFrame = m_loopStart * buffer()->sampleRate();
230 double loopEndFrame = m_loopEnd * buffer()->sampleRate();
232 virtualEndFrame = min(loopEndFrame, virtualEndFrame);
233 virtualDeltaFrames = virtualEndFrame - loopStartFrame;
237 double pitchRate = totalPitchRate();
239 // Sanity check that our playback rate isn't larger than the loop size.
240 if (pitchRate >= virtualDeltaFrames)
244 double virtualReadIndex = m_virtualReadIndex;
246 // Render loop - reading from the source buffer to the destination using linear interpolation.
247 int framesToProcess = numberOfFrames;
249 const float** sourceChannels = m_sourceChannels.get();
250 float** destinationChannels = m_destinationChannels.get();
252 // Optimize for the very common case of playing back with pitchRate == 1.
253 // We can avoid the linear interpolation.
254 if (pitchRate == 1 && virtualReadIndex == floor(virtualReadIndex)
255 && virtualDeltaFrames == floor(virtualDeltaFrames)
256 && virtualEndFrame == floor(virtualEndFrame)) {
257 unsigned readIndex = static_cast<unsigned>(virtualReadIndex);
258 unsigned deltaFrames = static_cast<unsigned>(virtualDeltaFrames);
259 endFrame = static_cast<unsigned>(virtualEndFrame);
260 while (framesToProcess > 0) {
261 int framesToEnd = endFrame - readIndex;
262 int framesThisTime = min(framesToProcess, framesToEnd);
263 framesThisTime = max(0, framesThisTime);
265 for (unsigned i = 0; i < numberOfChannels; ++i)
266 memcpy(destinationChannels[i] + writeIndex, sourceChannels[i] + readIndex, sizeof(float) * framesThisTime);
268 writeIndex += framesThisTime;
269 readIndex += framesThisTime;
270 framesToProcess -= framesThisTime;
273 if (readIndex >= endFrame) {
274 readIndex -= deltaFrames;
275 if (renderSilenceAndFinishIfNotLooping(bus, writeIndex, framesToProcess))
279 virtualReadIndex = readIndex;
281 while (framesToProcess--) {
282 unsigned readIndex = static_cast<unsigned>(virtualReadIndex);
283 double interpolationFactor = virtualReadIndex - readIndex;
285 // For linear interpolation we need the next sample-frame too.
286 unsigned readIndex2 = readIndex + 1;
287 if (readIndex2 >= bufferLength) {
289 // Make sure to wrap around at the end of the buffer.
290 readIndex2 = static_cast<unsigned>(virtualReadIndex + 1 - virtualDeltaFrames);
292 readIndex2 = readIndex;
295 // Final sanity check on buffer access.
296 // FIXME: as an optimization, try to get rid of this inner-loop check and put assertions and guards before the loop.
297 if (readIndex >= bufferLength || readIndex2 >= bufferLength)
300 // Linear interpolation.
301 for (unsigned i = 0; i < numberOfChannels; ++i) {
302 float* destination = destinationChannels[i];
303 const float* source = sourceChannels[i];
305 double sample1 = source[readIndex];
306 double sample2 = source[readIndex2];
307 double sample = (1.0 - interpolationFactor) * sample1 + interpolationFactor * sample2;
309 destination[writeIndex] = narrowPrecisionToFloat(sample);
313 virtualReadIndex += pitchRate;
315 // Wrap-around, retaining sub-sample position since virtualReadIndex is floating-point.
316 if (virtualReadIndex >= virtualEndFrame) {
317 virtualReadIndex -= virtualDeltaFrames;
318 if (renderSilenceAndFinishIfNotLooping(bus, writeIndex, framesToProcess))
324 bus->clearSilentFlag();
326 m_virtualReadIndex = virtualReadIndex;
332 void AudioBufferSourceNode::reset()
334 m_virtualReadIndex = 0;
335 m_lastGain = gain()->value();
338 bool AudioBufferSourceNode::setBuffer(AudioBuffer* buffer)
340 ASSERT(isMainThread());
342 // The context must be locked since changing the buffer can re-configure the number of channels that are output.
343 AudioContext::AutoLocker contextLocker(context());
345 // This synchronizes with process().
346 MutexLocker processLocker(m_processLock);
349 // Do any necesssary re-configuration to the buffer's number of channels.
350 unsigned numberOfChannels = buffer->numberOfChannels();
352 if (numberOfChannels > AudioContext::maxNumberOfChannels())
355 output(0)->setNumberOfChannels(numberOfChannels);
357 m_sourceChannels = adoptArrayPtr(new const float* [numberOfChannels]);
358 m_destinationChannels = adoptArrayPtr(new float* [numberOfChannels]);
360 for (unsigned i = 0; i < numberOfChannels; ++i)
361 m_sourceChannels[i] = buffer->getChannelData(i)->data();
364 m_virtualReadIndex = 0;
370 unsigned AudioBufferSourceNode::numberOfChannels()
372 return output(0)->numberOfChannels();
375 #if ENABLE(TIZEN_WEB_AUDIO)
376 void AudioBufferSourceNode::startGrain(double when, double grainOffset, ExceptionCode& ec)
378 // Duration of 0 has special value, meaning calculate based on the entire buffer's duration.
379 startGrain(when, grainOffset, 0, ec);
382 void AudioBufferSourceNode::startGrain(double when, double grainOffset)
384 // Duration of 0 has special value, meaning calculate based on the entire buffer's duration.
385 startGrain(when, grainOffset, 0);
389 #if ENABLE(TIZEN_WEB_AUDIO)
390 void AudioBufferSourceNode::startGrain(double when, double grainOffset, double grainDuration, ExceptionCode& ec)
392 void AudioBufferSourceNode::startGrain(double when, double grainOffset, double grainDuration)
395 ASSERT(isMainThread());
397 #if ENABLE(TIZEN_WEB_AUDIO)
398 if (m_playbackState != UNSCHEDULED_STATE) {
399 ec = INVALID_STATE_ERR;
403 if (m_playbackState != UNSCHEDULED_STATE)
410 // Do sanity checking of grain parameters versus buffer size.
411 double bufferDuration = buffer()->duration();
413 grainOffset = max(0.0, grainOffset);
414 grainOffset = min(bufferDuration, grainOffset);
415 m_grainOffset = grainOffset;
417 // Handle default/unspecified duration.
418 double maxDuration = bufferDuration - grainOffset;
420 grainDuration = maxDuration;
422 grainDuration = max(0.0, grainDuration);
423 grainDuration = min(maxDuration, grainDuration);
424 m_grainDuration = grainDuration;
429 // We call timeToSampleFrame here since at playbackRate == 1 we don't want to go through linear interpolation
430 // at a sub-sample position since it will degrade the quality.
431 // When aligned to the sample-frame the playback will be identical to the PCM data stored in the buffer.
432 // Since playbackRate == 1 is very common, it's worth considering quality.
433 m_virtualReadIndex = AudioUtilities::timeToSampleFrame(m_grainOffset, buffer()->sampleRate());
435 m_playbackState = SCHEDULED_STATE;
438 #if ENABLE(TIZEN_WEB_AUDIO)
439 #if ENABLE(LEGACY_WEB_AUDIO)
440 void AudioBufferSourceNode::noteGrainOn(double when, double grainOffset, double grainDuration, ExceptionCode& ec)
442 startGrain(when, grainOffset, grainDuration, ec);
446 #if ENABLE(LEGACY_WEB_AUDIO)
447 void AudioBufferSourceNode::noteGrainOn(double when, double grainOffset, double grainDuration)
449 startGrain(when, grainOffset, grainDuration);
454 double AudioBufferSourceNode::totalPitchRate()
456 double dopplerRate = 1.0;
458 dopplerRate = m_pannerNode->dopplerRate();
460 // Incorporate buffer's sample-rate versus AudioContext's sample-rate.
461 // Normally it's not an issue because buffers are loaded at the AudioContext's sample-rate, but we can handle it in any case.
462 double sampleRateFactor = 1.0;
464 sampleRateFactor = buffer()->sampleRate() / sampleRate();
466 double basePitchRate = playbackRate()->value();
468 double totalRate = dopplerRate * sampleRateFactor * basePitchRate;
470 // Sanity check the total rate. It's very important that the resampler not get any bad rate values.
471 totalRate = max(0.0, totalRate);
473 totalRate = 1; // zero rate is considered illegal
474 totalRate = min(MaxRate, totalRate);
476 bool isTotalRateValid = !isnan(totalRate) && !isinf(totalRate);
477 ASSERT(isTotalRateValid);
478 if (!isTotalRateValid)
484 bool AudioBufferSourceNode::looping()
486 static bool firstTime = true;
487 if (firstTime && context() && context()->scriptExecutionContext()) {
488 context()->scriptExecutionContext()->addConsoleMessage(JSMessageSource, LogMessageType, WarningMessageLevel, "AudioBufferSourceNode 'looping' attribute is deprecated. Use 'loop' instead.");
495 void AudioBufferSourceNode::setLooping(bool looping)
497 static bool firstTime = true;
498 if (firstTime && context() && context()->scriptExecutionContext()) {
499 context()->scriptExecutionContext()->addConsoleMessage(JSMessageSource, LogMessageType, WarningMessageLevel, "AudioBufferSourceNode 'looping' attribute is deprecated. Use 'loop' instead.");
503 m_isLooping = looping;
506 bool AudioBufferSourceNode::propagatesSilence() const
508 return !isPlayingOrScheduled() || hasFinished() || !m_buffer;
511 void AudioBufferSourceNode::setPannerNode(PannerNode* pannerNode)
513 if (m_pannerNode != pannerNode && !hasFinished()) {
515 pannerNode->ref(AudioNode::RefTypeConnection);
517 m_pannerNode->deref(AudioNode::RefTypeConnection);
519 m_pannerNode = pannerNode;
523 void AudioBufferSourceNode::clearPannerNode()
526 m_pannerNode->deref(AudioNode::RefTypeConnection);
531 void AudioBufferSourceNode::finish()
534 ASSERT(!m_pannerNode);
535 AudioScheduledSourceNode::finish();
538 } // namespace WebCore
540 #endif // ENABLE(WEB_AUDIO)