AAudio: Add sample rate conversion to MMAP
This CL adds sample rate conversion to the MMAP path in AAudio.
AAudio no longer needs to use AudioFlinger's sample rate converter for
MMAP streams. This effectively reduces the latency from 100ms to 30ms
on speaker, 3.5mm, and USB if MMAP is enabled and sample rate conversion
is needed.
A sample rate converter is added to the AAudio flowgraph. Since input
and output sample rates are different for the flowgraph, a variable
number of frames must be used. 8 frames are passed through the flowgraph
at any given point of times and the output of the flowgraph pulls as
much data as there exists in the flowgraph.
Burst sizes, capacity, latency, frames read, and frames written are all
changed reflect the sample rate of the app.
Flowgraph was also added to input MMAP streams.
Bug: 219533889
Test: OboeTester Test Input/Output with a variety of sample rates and formats
Test: atest AAudioTests
Change-Id: I9b6a2694921ea558fb91a231f956f638d3342837
diff --git a/media/libaaudio/src/client/AudioStreamInternalPlay.cpp b/media/libaaudio/src/client/AudioStreamInternalPlay.cpp
index 89dd8ff..99aa910 100644
--- a/media/libaaudio/src/client/AudioStreamInternalPlay.cpp
+++ b/media/libaaudio/src/client/AudioStreamInternalPlay.cpp
@@ -48,14 +48,18 @@
aaudio_result_t AudioStreamInternalPlay::open(const AudioStreamBuilder &builder) {
aaudio_result_t result = AudioStreamInternal::open(builder);
+ const bool useVolumeRamps = (getSharingMode() == AAUDIO_SHARING_MODE_EXCLUSIVE);
if (result == AAUDIO_OK) {
result = mFlowGraph.configure(getFormat(),
getSamplesPerFrame(),
+ getSampleRate(),
getDeviceFormat(),
getDeviceChannelCount(),
+ getDeviceSampleRate(),
getRequireMonoBlend(),
+ useVolumeRamps,
getAudioBalance(),
- (getSharingMode() == AAUDIO_SHARING_MODE_EXCLUSIVE));
+ aaudio::resampler::MultiChannelResampler::Quality::Medium);
if (result != AAUDIO_OK) {
safeReleaseClose();
@@ -186,7 +190,7 @@
// Sleep if there is too much data in the buffer.
// Calculate an ideal time to wake up.
if (wakeTimePtr != nullptr
- && (mAudioEndpoint->getFullFramesAvailable() >= getBufferSize())) {
+ && (mAudioEndpoint->getFullFramesAvailable() >= getDeviceBufferSize())) {
// By default wake up a few milliseconds from now. // TODO review
int64_t wakeTime = currentNanoTime + (1 * AAUDIO_NANOS_PER_MILLISECOND);
aaudio_stream_state_t state = getState();
@@ -206,12 +210,12 @@
// If the appBufferSize is smaller than the endpointBufferSize then
// we will have room to write data beyond the appBufferSize.
// That is a technique used to reduce glitches without adding latency.
- const int32_t appBufferSize = getBufferSize();
+ const int64_t appBufferSize = getDeviceBufferSize();
// The endpoint buffer size is set to the maximum that can be written.
// If we use it then we must carve out some room to write data when we wake up.
- const int32_t endBufferSize = mAudioEndpoint->getBufferSizeInFrames()
- - getFramesPerBurst();
- const int32_t bestBufferSize = std::min(appBufferSize, endBufferSize);
+ const int64_t endBufferSize = mAudioEndpoint->getBufferSizeInFrames()
+ - getDeviceFramesPerBurst();
+ const int64_t bestBufferSize = std::min(appBufferSize, endBufferSize);
int64_t targetReadPosition = mAudioEndpoint->getDataWriteCounter() - bestBufferSize;
wakeTime = mClockModel.convertPositionToTime(targetReadPosition);
}
@@ -232,37 +236,78 @@
int32_t numFrames) {
WrappingBuffer wrappingBuffer;
uint8_t *byteBuffer = (uint8_t *) buffer;
- int32_t framesLeft = numFrames;
+ int32_t framesLeftInByteBuffer = numFrames;
mAudioEndpoint->getEmptyFramesAvailable(&wrappingBuffer);
// Write data in one or two parts.
int partIndex = 0;
- while (framesLeft > 0 && partIndex < WrappingBuffer::SIZE) {
- int32_t framesToWrite = framesLeft;
- int32_t framesAvailable = wrappingBuffer.numFrames[partIndex];
- if (framesAvailable > 0) {
- if (framesToWrite > framesAvailable) {
- framesToWrite = framesAvailable;
- }
+ int framesWrittenToAudioEndpoint = 0;
+ while (framesLeftInByteBuffer > 0 && partIndex < WrappingBuffer::SIZE) {
+ int32_t framesAvailableInWrappingBuffer = wrappingBuffer.numFrames[partIndex];
+ uint8_t *currentWrappingBuffer = (uint8_t *) wrappingBuffer.data[partIndex];
- int32_t numBytes = getBytesPerFrame() * framesToWrite;
+ if (framesAvailableInWrappingBuffer > 0) {
+ // Pull data from the flowgraph in case there is residual data.
+ const int32_t framesActuallyWrittenToWrappingBuffer = mFlowGraph.pull(
+ (void*) currentWrappingBuffer,
+ framesAvailableInWrappingBuffer);
- mFlowGraph.process((void *)byteBuffer,
- wrappingBuffer.data[partIndex],
- framesToWrite);
+ const int32_t numBytesActuallyWrittenToWrappingBuffer =
+ framesActuallyWrittenToWrappingBuffer * getBytesPerDeviceFrame();
+ currentWrappingBuffer += numBytesActuallyWrittenToWrappingBuffer;
+ framesAvailableInWrappingBuffer -= framesActuallyWrittenToWrappingBuffer;
+ framesWrittenToAudioEndpoint += framesActuallyWrittenToWrappingBuffer;
+ }
- byteBuffer += numBytes;
- framesLeft -= framesToWrite;
- } else {
- break;
+ // Put data from byteBuffer into the flowgraph one buffer (8 frames) at a time.
+ // Continuously pull as much data as possible from the flowgraph into the wrapping buffer.
+ // The return value of mFlowGraph.process is the number of frames actually pulled.
+ while (framesAvailableInWrappingBuffer > 0 && framesLeftInByteBuffer > 0) {
+ const int32_t framesToWriteFromByteBuffer = std::min(flowgraph::kDefaultBufferSize,
+ framesLeftInByteBuffer);
+
+ const int32_t numBytesToWriteFromByteBuffer = getBytesPerFrame() *
+ framesToWriteFromByteBuffer;
+
+ //ALOGD("%s() framesLeftInByteBuffer %d, framesAvailableInWrappingBuffer %d"
+ // "framesToWriteFromByteBuffer %d, numBytesToWriteFromByteBuffer %d"
+ // , __func__, framesLeftInByteBuffer, framesAvailableInWrappingBuffer,
+ // framesToWriteFromByteBuffer, numBytesToWriteFromByteBuffer);
+
+ const int32_t framesActuallyWrittenToWrappingBuffer = mFlowGraph.process(
+ (void *)byteBuffer,
+ framesToWriteFromByteBuffer,
+ (void *)currentWrappingBuffer,
+ framesAvailableInWrappingBuffer);
+
+ byteBuffer += numBytesToWriteFromByteBuffer;
+ framesLeftInByteBuffer -= framesToWriteFromByteBuffer;
+ const int32_t numBytesActuallyWrittenToWrappingBuffer =
+ framesActuallyWrittenToWrappingBuffer * getBytesPerDeviceFrame();
+ currentWrappingBuffer += numBytesActuallyWrittenToWrappingBuffer;
+ framesAvailableInWrappingBuffer -= framesActuallyWrittenToWrappingBuffer;
+ framesWrittenToAudioEndpoint += framesActuallyWrittenToWrappingBuffer;
+
+ //ALOGD("%s() numBytesActuallyWrittenToWrappingBuffer %d, framesLeftInByteBuffer %d"
+ // "framesActuallyWrittenToWrappingBuffer %d, numBytesToWriteFromByteBuffer %d"
+ // "framesWrittenToAudioEndpoint %d"
+ // , __func__, numBytesActuallyWrittenToWrappingBuffer, framesLeftInByteBuffer,
+ // framesActuallyWrittenToWrappingBuffer, numBytesToWriteFromByteBuffer,
+ // framesWrittenToAudioEndpoint);
}
partIndex++;
}
- int32_t framesWritten = numFrames - framesLeft;
- mAudioEndpoint->advanceWriteIndex(framesWritten);
+ //ALOGD("%s() framesWrittenToAudioEndpoint %d, numFrames %d"
+ // "framesLeftInByteBuffer %d"
+ // , __func__, framesWrittenToAudioEndpoint, numFrames,
+ // framesLeftInByteBuffer);
- return framesWritten;
+ // The audio endpoint should reference the number of frames written to the wrapping buffer.
+ mAudioEndpoint->advanceWriteIndex(framesWrittenToAudioEndpoint);
+
+ // The internal code should use the number of frames read from the app.
+ return numFrames - framesLeftInByteBuffer;
}
int64_t AudioStreamInternalPlay::getFramesRead() {
@@ -284,7 +329,6 @@
return mLastFramesWritten;
}
-
// Render audio in the application callback and then write the data to the stream.
void *AudioStreamInternalPlay::callbackLoop() {
ALOGD("%s() entering >>>>>>>>>>>>>>>", __func__);