void
AudioBuffer::MixToMono(JSContext* aJSContext)
{
if (mJSChannels.Length() == 1) {
// The buffer is already mono
return;
}
// Prepare the input channels
nsAutoTArray<const void*, GUESS_AUDIO_CHANNELS> channels;
channels.SetLength(mJSChannels.Length());
for (uint32_t i = 0; i < mJSChannels.Length(); ++i) {
channels[i] = JS_GetFloat32ArrayData(mJSChannels[i]);
}
// Prepare the output channels
float* downmixBuffer = new float[mLength];
// Perform the down-mix
AudioChannelsDownMix(channels, &downmixBuffer, 1, mLength);
// Truncate the shared channels and copy the downmixed data over
mJSChannels.SetLength(1);
SetRawChannelContents(aJSContext, 0, downmixBuffer);
delete[] downmixBuffer;
}
void TestDownmixStereo() {
const size_t arraySize = 1024;
nsTArray<const T*> inputptr;
nsTArray<T*> input;
T** output;
output = new T*[1];
output[0] = new T[arraySize];
input.SetLength(2);
inputptr.SetLength(2);
for (size_t channel = 0; channel < input.Length(); channel++) {
input[channel] = new T[arraySize];
for (size_t i = 0; i < arraySize; i++) {
input[channel][i] = channel == 0 ? GetLowValue<T>() : GetHighValue<T>();
}
inputptr[channel] = input[channel];
}
AudioChannelsDownMix(inputptr, output, 1, arraySize);
for (size_t i = 0; i < arraySize; i++) {
ASSERT_TRUE(output[0][i] == GetSilentValue<T>());
ASSERT_TRUE(output[0][i] == GetSilentValue<T>());
}
delete[] output[0];
delete[] output;
}
void
DownmixAndInterleave(const nsTArray<const void*>& aChannelData,
AudioSampleFormat aSourceFormat, int32_t aDuration,
float aVolume, uint32_t aOutputChannels,
AudioDataValue* aOutput)
{
nsAutoTArray<const void*,GUESS_AUDIO_CHANNELS> channelData;
nsAutoTArray<float,AUDIO_PROCESSING_FRAMES*GUESS_AUDIO_CHANNELS> downmixConversionBuffer;
nsAutoTArray<float,AUDIO_PROCESSING_FRAMES*GUESS_AUDIO_CHANNELS> downmixOutputBuffer;
channelData.SetLength(aChannelData.Length());
if (aSourceFormat != AUDIO_FORMAT_FLOAT32) {
NS_ASSERTION(aSourceFormat == AUDIO_FORMAT_S16, "unknown format");
downmixConversionBuffer.SetLength(aDuration*aChannelData.Length());
for (uint32_t i = 0; i < aChannelData.Length(); ++i) {
float* conversionBuf = downmixConversionBuffer.Elements() + (i*aDuration);
const int16_t* sourceBuf = static_cast<const int16_t*>(aChannelData[i]);
for (uint32_t j = 0; j < (uint32_t)aDuration; ++j) {
conversionBuf[j] = AudioSampleToFloat(sourceBuf[j]);
}
channelData[i] = conversionBuf;
}
} else {
for (uint32_t i = 0; i < aChannelData.Length(); ++i) {
channelData[i] = aChannelData[i];
}
}
downmixOutputBuffer.SetLength(aDuration*aOutputChannels);
nsAutoTArray<float*,GUESS_AUDIO_CHANNELS> outputChannelBuffers;
nsAutoTArray<const void*,GUESS_AUDIO_CHANNELS> outputChannelData;
outputChannelBuffers.SetLength(aOutputChannels);
outputChannelData.SetLength(aOutputChannels);
for (uint32_t i = 0; i < (uint32_t)aOutputChannels; ++i) {
outputChannelData[i] = outputChannelBuffers[i] =
downmixOutputBuffer.Elements() + aDuration*i;
}
if (channelData.Length() > aOutputChannels) {
AudioChannelsDownMix(channelData, outputChannelBuffers.Elements(),
aOutputChannels, aDuration);
}
InterleaveAndConvertBuffer(outputChannelData.Elements(), AUDIO_FORMAT_FLOAT32,
aDuration, aVolume, aOutputChannels, aOutput);
}
void
AudioNodeStream::UpMixDownMixChunk(const AudioChunk* aChunk,
uint32_t aOutputChannelCount,
nsTArray<const void*>& aOutputChannels,
nsTArray<float>& aDownmixBuffer)
{
static const float silenceChannel[WEBAUDIO_BLOCK_SIZE] = {0.f};
aOutputChannels.AppendElements(aChunk->mChannelData);
if (aOutputChannels.Length() < aOutputChannelCount) {
if (mChannelInterpretation == ChannelInterpretation::Speakers) {
AudioChannelsUpMix(&aOutputChannels, aOutputChannelCount, nullptr);
NS_ASSERTION(aOutputChannelCount == aOutputChannels.Length(),
"We called GetAudioChannelsSuperset to avoid this");
} else {
// Fill up the remaining aOutputChannels by zeros
for (uint32_t j = aOutputChannels.Length(); j < aOutputChannelCount; ++j) {
aOutputChannels.AppendElement(silenceChannel);
}
}
} else if (aOutputChannels.Length() > aOutputChannelCount) {
if (mChannelInterpretation == ChannelInterpretation::Speakers) {
nsAutoTArray<float*,GUESS_AUDIO_CHANNELS> outputChannels;
outputChannels.SetLength(aOutputChannelCount);
aDownmixBuffer.SetLength(aOutputChannelCount * WEBAUDIO_BLOCK_SIZE);
for (uint32_t j = 0; j < aOutputChannelCount; ++j) {
outputChannels[j] = &aDownmixBuffer[j * WEBAUDIO_BLOCK_SIZE];
}
AudioChannelsDownMix(aOutputChannels, outputChannels.Elements(),
aOutputChannelCount, WEBAUDIO_BLOCK_SIZE);
aOutputChannels.SetLength(aOutputChannelCount);
for (uint32_t j = 0; j < aOutputChannels.Length(); ++j) {
aOutputChannels[j] = outputChannels[j];
}
} else {
// Drop the remaining aOutputChannels
aOutputChannels.RemoveElementsAt(aOutputChannelCount,
aOutputChannels.Length() - aOutputChannelCount);
}
}
}
void
AudioNodeStream::ObtainInputBlock(AudioChunk& aTmpChunk, uint32_t aPortIndex)
{
uint32_t inputCount = mInputs.Length();
uint32_t outputChannelCount = 1;
nsAutoTArray<AudioChunk*,250> inputChunks;
for (uint32_t i = 0; i < inputCount; ++i) {
if (aPortIndex != mInputs[i]->InputNumber()) {
// This input is connected to a different port
continue;
}
MediaStream* s = mInputs[i]->GetSource();
AudioNodeStream* a = static_cast<AudioNodeStream*>(s);
MOZ_ASSERT(a == s->AsAudioNodeStream());
if (a->IsFinishedOnGraphThread() ||
a->IsAudioParamStream()) {
continue;
}
AudioChunk* chunk = &a->mLastChunks[mInputs[i]->OutputNumber()];
MOZ_ASSERT(chunk);
if (chunk->IsNull()) {
continue;
}
inputChunks.AppendElement(chunk);
outputChannelCount =
GetAudioChannelsSuperset(outputChannelCount, chunk->mChannelData.Length());
}
switch (mChannelCountMode) {
case ChannelCountMode::Explicit:
// Disregard the output channel count that we've calculated, and just use
// mNumberOfInputChannels.
outputChannelCount = mNumberOfInputChannels;
break;
case ChannelCountMode::Clamped_max:
// Clamp the computed output channel count to mNumberOfInputChannels.
outputChannelCount = std::min(outputChannelCount, mNumberOfInputChannels);
break;
case ChannelCountMode::Max:
// Nothing to do here, just shut up the compiler warning.
break;
}
uint32_t inputChunkCount = inputChunks.Length();
if (inputChunkCount == 0 ||
(inputChunkCount == 1 && inputChunks[0]->mChannelData.Length() == 0)) {
aTmpChunk.SetNull(WEBAUDIO_BLOCK_SIZE);
return;
}
if (inputChunkCount == 1 &&
inputChunks[0]->mChannelData.Length() == outputChannelCount) {
aTmpChunk = *inputChunks[0];
return;
}
AllocateAudioBlock(outputChannelCount, &aTmpChunk);
float silenceChannel[WEBAUDIO_BLOCK_SIZE] = {0.f};
// The static storage here should be 1KB, so it's fine
nsAutoTArray<float, GUESS_AUDIO_CHANNELS*WEBAUDIO_BLOCK_SIZE> downmixBuffer;
for (uint32_t i = 0; i < inputChunkCount; ++i) {
AudioChunk* chunk = inputChunks[i];
nsAutoTArray<const void*,GUESS_AUDIO_CHANNELS> channels;
channels.AppendElements(chunk->mChannelData);
if (channels.Length() < outputChannelCount) {
if (mChannelInterpretation == ChannelInterpretation::Speakers) {
AudioChannelsUpMix(&channels, outputChannelCount, nullptr);
NS_ASSERTION(outputChannelCount == channels.Length(),
"We called GetAudioChannelsSuperset to avoid this");
} else {
// Fill up the remaining channels by zeros
for (uint32_t j = channels.Length(); j < outputChannelCount; ++j) {
channels.AppendElement(silenceChannel);
}
}
} else if (channels.Length() > outputChannelCount) {
if (mChannelInterpretation == ChannelInterpretation::Speakers) {
nsAutoTArray<float*,GUESS_AUDIO_CHANNELS> outputChannels;
outputChannels.SetLength(outputChannelCount);
downmixBuffer.SetLength(outputChannelCount * WEBAUDIO_BLOCK_SIZE);
for (uint32_t j = 0; j < outputChannelCount; ++j) {
outputChannels[j] = &downmixBuffer[j * WEBAUDIO_BLOCK_SIZE];
}
AudioChannelsDownMix(channels, outputChannels.Elements(),
outputChannelCount, WEBAUDIO_BLOCK_SIZE);
channels.SetLength(outputChannelCount);
for (uint32_t j = 0; j < channels.Length(); ++j) {
channels[j] = outputChannels[j];
}
} else {
// Drop the remaining channels
channels.RemoveElementsAt(outputChannelCount,
channels.Length() - outputChannelCount);
}
}
for (uint32_t c = 0; c < channels.Length(); ++c) {
const float* inputData = static_cast<const float*>(channels[c]);
float* outputData = static_cast<float*>(const_cast<void*>(aTmpChunk.mChannelData[c]));
if (inputData) {
if (i == 0) {
AudioBlockCopyChannelWithScale(inputData, chunk->mVolume, outputData);
} else {
AudioBlockAddChannelWithScale(inputData, chunk->mVolume, outputData);
}
} else {
if (i == 0) {
memset(outputData, 0, WEBAUDIO_BLOCK_SIZE*sizeof(float));
}
}
}
}
}
void
AudioNodeExternalInputStream::TrackMapEntry::ResampleInputData(AudioSegment* aSegment)
{
AudioSegment::ChunkIterator ci(*aSegment);
while (!ci.IsEnded()) {
const AudioChunk& chunk = *ci;
nsAutoTArray<const void*,2> channels;
if (chunk.GetDuration() > UINT32_MAX) {
// This will cause us to OOM or overflow below. So let's just bail.
NS_ERROR("Chunk duration out of bounds");
return;
}
uint32_t duration = uint32_t(chunk.GetDuration());
if (chunk.IsNull()) {
nsAutoTArray<AudioDataValue,1024> silence;
silence.SetLength(duration);
PodZero(silence.Elements(), silence.Length());
channels.SetLength(mResamplerChannelCount);
for (uint32_t i = 0; i < channels.Length(); ++i) {
channels[i] = silence.Elements();
}
ResampleChannels(channels, duration, AUDIO_OUTPUT_FORMAT, 0.0f);
} else if (chunk.mChannelData.Length() == mResamplerChannelCount) {
// Common case, since mResamplerChannelCount is set to the first chunk's
// number of channels.
channels.AppendElements(chunk.mChannelData);
ResampleChannels(channels, duration, chunk.mBufferFormat, chunk.mVolume);
} else {
// Uncommon case. Since downmixing requires channels to be floats,
// convert everything to floats now.
uint32_t upChannels = GetAudioChannelsSuperset(chunk.mChannelData.Length(), mResamplerChannelCount);
nsTArray<float> buffer;
if (chunk.mBufferFormat == AUDIO_FORMAT_FLOAT32) {
channels.AppendElements(chunk.mChannelData);
} else {
NS_ASSERTION(chunk.mBufferFormat == AUDIO_FORMAT_S16, "Unknown format");
if (duration > UINT32_MAX/chunk.mChannelData.Length()) {
NS_ERROR("Chunk duration out of bounds");
return;
}
buffer.SetLength(chunk.mChannelData.Length()*duration);
for (uint32_t i = 0; i < chunk.mChannelData.Length(); ++i) {
const int16_t* samples = static_cast<const int16_t*>(chunk.mChannelData[i]);
float* converted = &buffer[i*duration];
for (uint32_t j = 0; j < duration; ++j) {
converted[j] = AudioSampleToFloat(samples[j]);
}
channels.AppendElement(converted);
}
}
nsTArray<float> zeroes;
if (channels.Length() < upChannels) {
zeroes.SetLength(duration);
PodZero(zeroes.Elements(), zeroes.Length());
AudioChannelsUpMix(&channels, upChannels, zeroes.Elements());
}
if (channels.Length() == mResamplerChannelCount) {
ResampleChannels(channels, duration, AUDIO_FORMAT_FLOAT32, chunk.mVolume);
} else {
nsTArray<float> output;
if (duration > UINT32_MAX/mResamplerChannelCount) {
NS_ERROR("Chunk duration out of bounds");
return;
}
output.SetLength(duration*mResamplerChannelCount);
nsAutoTArray<float*,2> outputPtrs;
nsAutoTArray<const void*,2> outputPtrsConst;
for (uint32_t i = 0; i < mResamplerChannelCount; ++i) {
outputPtrs.AppendElement(output.Elements() + i*duration);
outputPtrsConst.AppendElement(outputPtrs[i]);
}
AudioChannelsDownMix(channels, outputPtrs.Elements(), outputPtrs.Length(), duration);
ResampleChannels(outputPtrsConst, duration, AUDIO_FORMAT_FLOAT32, chunk.mVolume);
}
}
ci.Next();
}
}