void
InterleaveAndConvertBuffer(const void** aSourceChannels,
AudioSampleFormat aSourceFormat,
int32_t aLength, float aVolume,
int32_t aChannels,
AudioDataValue* aOutput)
{
switch (aSourceFormat) {
case AUDIO_FORMAT_FLOAT32:
InterleaveAndConvertBuffer(reinterpret_cast<const float**>(aSourceChannels),
aLength,
aVolume,
aChannels,
aOutput);
break;
case AUDIO_FORMAT_S16:
InterleaveAndConvertBuffer(reinterpret_cast<const int16_t**>(aSourceChannels),
aLength,
aVolume,
aChannels,
aOutput);
break;
case AUDIO_FORMAT_SILENCE:
// nothing to do here.
break;
}
}
static void
InterleaveAndConvertBuffer(const void* aSource, nsAudioStream::SampleFormat aSourceFormat,
int32_t aSourceLength,
int32_t aOffset, int32_t aLength,
float aVolume,
int32_t aChannels,
void* aOutput, nsAudioStream::SampleFormat aOutputFormat)
{
switch (aSourceFormat) {
case nsAudioStream::FORMAT_FLOAT32:
InterleaveAndConvertBuffer(static_cast<const float*>(aSource) + aOffset, aSourceLength,
aLength,
aVolume,
aChannels,
aOutput, aOutputFormat);
break;
case nsAudioStream::FORMAT_S16_LE:
InterleaveAndConvertBuffer(static_cast<const int16_t*>(aSource) + aOffset, aSourceLength,
aLength,
aVolume,
aChannels,
aOutput, aOutputFormat);
break;
case nsAudioStream::FORMAT_U8:
InterleaveAndConvertBuffer(static_cast<const uint8_t*>(aSource) + aOffset, aSourceLength,
aLength,
aVolume,
aChannels,
aOutput, aOutputFormat);
break;
}
}
void
AudioSegment::WriteTo(nsAudioStream* aOutput)
{
NS_ASSERTION(mChannels == aOutput->GetChannels(), "Wrong number of channels");
nsAutoTArray<uint8_t,STATIC_AUDIO_BUFFER_BYTES> buf;
uint32_t frameSize = GetSampleSize(aOutput->GetFormat())*mChannels;
for (ChunkIterator ci(*this); !ci.IsEnded(); ci.Next()) {
AudioChunk& c = *ci;
if (frameSize*c.mDuration > PR_UINT32_MAX) {
NS_ERROR("Buffer overflow");
return;
}
buf.SetLength(int32_t(frameSize*c.mDuration));
if (c.mBuffer) {
InterleaveAndConvertBuffer(c.mBuffer->Data(), c.mBufferFormat, c.mBufferLength,
c.mOffset, int32_t(c.mDuration),
c.mVolume,
aOutput->GetChannels(),
buf.Elements(), aOutput->GetFormat());
} else {
// Assumes that a bit pattern of zeroes == 0.0f
memset(buf.Elements(), 0, buf.Length());
}
aOutput->Write(buf.Elements(), int32_t(c.mDuration));
}
}
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
AudioTrackEncoder::InterleaveTrackData(AudioChunk& aChunk,
int32_t aDuration,
uint32_t aOutputChannels,
AudioDataValue* aOutput)
{
if (aChunk.mChannelData.Length() < aOutputChannels) {
// Up-mix. This might make the mChannelData have more than aChannels.
AudioChannelsUpMix(&aChunk.mChannelData, aOutputChannels, gZeroChannel);
}
if (aChunk.mChannelData.Length() > aOutputChannels) {
DownmixAndInterleave(aChunk.mChannelData, aChunk.mBufferFormat, aDuration,
aChunk.mVolume, mChannels, aOutput);
} else {
InterleaveAndConvertBuffer(aChunk.mChannelData.Elements(),
aChunk.mBufferFormat, aDuration, aChunk.mVolume,
mChannels, aOutput);
}
}
void TestInterleaveAndConvert() {
size_t arraySize = 1024;
size_t maxChannels = 8; // 7.1
for (uint32_t channels = 1; channels < maxChannels; channels++) {
const SrcT* const* src = GetPlanarChannelArray<SrcT>(channels, arraySize);
DstT* dst = new DstT[channels * arraySize];
InterleaveAndConvertBuffer(src, arraySize, 1.0, channels, dst);
uint32_t channelIndex = 0;
for (size_t i = 0; i < arraySize * channels; i++) {
ASSERT_TRUE(FuzzyEqual(
dst[i], FloatToAudioSample<DstT>(1. / (channelIndex + 1))));
channelIndex++;
channelIndex %= channels;
}
DeletePlanarChannelsArray(src, channels);
delete[] dst;
}
}
void
AudioSegment::WriteTo(uint64_t aID, AudioStream* aOutput, AudioMixer* aMixer)
{
uint32_t outputChannels = aOutput->GetChannels();
nsAutoTArray<AudioDataValue,AUDIO_PROCESSING_FRAMES*GUESS_AUDIO_CHANNELS> buf;
nsAutoTArray<const void*,GUESS_AUDIO_CHANNELS> channelData;
// Offset in the buffer that will end up sent to the AudioStream, in samples.
uint32_t offset = 0;
if (!GetDuration()) {
return;
}
uint32_t outBufferLength = GetDuration() * outputChannels;
buf.SetLength(outBufferLength);
for (ChunkIterator ci(*this); !ci.IsEnded(); ci.Next()) {
AudioChunk& c = *ci;
uint32_t frames = c.mDuration;
// If we have written data in the past, or we have real (non-silent) data
// to write, we can proceed. Otherwise, it means we just started the
// AudioStream, and we don't have real data to write to it (just silence).
// To avoid overbuffering in the AudioStream, we simply drop the silence,
// here. The stream will underrun and output silence anyways.
if (c.mBuffer || aOutput->GetWritten()) {
if (c.mBuffer && c.mBufferFormat != AUDIO_FORMAT_SILENCE) {
channelData.SetLength(c.mChannelData.Length());
for (uint32_t i = 0; i < channelData.Length(); ++i) {
channelData[i] = c.mChannelData[i];
}
if (channelData.Length() < outputChannels) {
// Up-mix. Note that this might actually make channelData have more
// than outputChannels temporarily.
AudioChannelsUpMix(&channelData, outputChannels, gZeroChannel);
}
if (channelData.Length() > outputChannels) {
// Down-mix.
DownmixAndInterleave(channelData, c.mBufferFormat, frames,
c.mVolume, outputChannels, buf.Elements() + offset);
} else {
InterleaveAndConvertBuffer(channelData.Elements(), c.mBufferFormat,
frames, c.mVolume,
outputChannels,
buf.Elements() + offset);
}
} else {
// Assumes that a bit pattern of zeroes == 0.0f
memset(buf.Elements() + offset, 0, outputChannels * frames * sizeof(AudioDataValue));
}
offset += frames * outputChannels;
}
if (!c.mTimeStamp.IsNull()) {
TimeStamp now = TimeStamp::Now();
// would be more efficient to c.mTimeStamp to ms on create time then pass here
LogTime(AsyncLatencyLogger::AudioMediaStreamTrack, aID,
(now - c.mTimeStamp).ToMilliseconds(), c.mTimeStamp);
}
}
aOutput->Write(buf.Elements(), offset / outputChannels, &(mChunks[mChunks.Length() - 1].mTimeStamp));
if (aMixer) {
aMixer->Mix(buf.Elements(), outputChannels, GetDuration(), aOutput->GetRate());
}
aOutput->Start();
}
