Minor audio handling APIs adjustments

Summary:
This diff isolates minor changes to audio handling related APIs or how we use them, without changing functionality.
This diff reduces the size of upcoming diffs that contain actual functional changes that need closer attention.

Reviewed By: kiminoue7

Differential Revision: D53675816

fbshipit-source-id: ba1178b1bba510a9442c8517257dc4dcda04c713

sample_apps/SamplePlaybackApp.cpp

@@ -110,7 +110,7 @@ class AudioStreamPlayer : public RecordFormatStreamPlayer {
     // the audio data was not read yet. Use your own buffers.
     utils::AudioBlock audio;
     if (XR_VERIFY(audio.readBlock(record.reader, cb))) {
-      const auto* audioData = reinterpret_cast<const uint16_t*>(audio.rdata());
+      const auto* audioData = audio.data<uint16_t>();
       // use audio data. In this sample code, we verify that the data matches the expected pattern
       for (size_t k = 0; k < kAudioBlockSize; k++) {
         XR_CHECK(audioData[k] == static_cast<int16_t>(audioBlockIndex * kAudioBlockSize + k));

tools/vrsplayer/AudioPlayer.cpp

@@ -16,8 +16,6 @@
 
 #include "AudioPlayer.h"
 
-#include <vector>
-
 #include <portaudio.h>
 
 #define DEFAULT_LOG_CHANNEL "AudioPlayer"
@@ -74,7 +72,7 @@ bool AudioPlayer::onAudioRead(const CurrentRecord& record, size_t blkIdx, const
             audio.asString());
       } else if (
           VideoTime::getPlaybackSpeed() <= 1 && sampleFormat_ == audio.getSampleFormat() &&
-          audio.getSampleCount() >= channelCount_) {
+          audio.getChannelCount() >= channelCount_ && audioBlock.getSampleCount() > 0) {
         playbackQueue_.sendJob(std::move(audioBlock));
       }
     }
@@ -191,7 +189,7 @@ void AudioPlayer::playbackThread() {
         Pa_WriteStream(paStream_, src, frameBatchSize);
       } else {
         // either we play fewer channels than provided, or frames are padded, we need to compact
-        uint8_t* dst = const_cast<uint8_t*>(reinterpret_cast<const uint8_t*>(block.rdata()));
+        uint8_t* dst = block.data<uint8_t>();
         uint32_t sample = 0;
         while (dst + paFrameStride > src && sample < frameBatchSize) {
           memmove(dst, src, paFrameStride); // more expensive, but safe in case of overlap

vrs/RecordFormat.cpp

@@ -914,6 +914,9 @@ ContentBlock::ContentBlock(
 ContentBlock::ContentBlock(AudioFormat audioFormat, uint8_t channelCount)
     : contentType_{ContentType::AUDIO}, audioSpec_{audioFormat, channelCount} {}
 
+ContentBlock::ContentBlock(const AudioContentBlockSpec& audioSpec, size_t size)
+    : contentType_{ContentType::AUDIO}, size_{size}, audioSpec_{audioSpec} {}
+
 ContentBlock::ContentBlock(
     AudioFormat audioFormat,
     AudioSampleFormat sampleFormat,

vrs/RecordFormat.h

@@ -429,6 +429,10 @@ class AudioContentBlockSpec {
   uint32_t getSampleCount() const {
     return sampleFrameCount_;
   }
+  /// Set the number of audio sample frames in the content block.
+  void setSampleCount(uint32_t sampleCount) {
+    sampleFrameCount_ = sampleCount;
+  }
   /// Tell if the audio sample format is fully defined.
   /// For instance, PCM audio data when we have enough details: sample format & channel count.
   bool isSampleBlockFormatDefined() const {
@@ -520,6 +524,7 @@ class ContentBlock {
   /// Very generic audio block description.
   // NOLINTNEXTLINE(hicpp-explicit-conversions)
   ContentBlock(AudioFormat audioFormat, uint8_t channelCount = 0);
+  ContentBlock(const AudioContentBlockSpec& audioSpec, size_t size);
 
   /// Audio block description.
   ContentBlock(

vrs/utils/AudioBlock.cpp

@@ -16,8 +16,6 @@
 
 #include "AudioBlock.h"
 
-#include <cstring>
-
 #define DEFAULT_LOG_CHANNEL "AudioBlock"
 #include <logging/Checks.h>
 #include <logging/Log.h>
@@ -46,14 +44,22 @@ AudioBlock::AudioBlock(const AudioContentBlockSpec& spec, vector<uint8_t>&& fram
   THROTTLED_VERIFY(nullptr, size == ContentBlock::kSizeUnknown || size == audioBytes_.size());
 }
 
-void AudioBlock::init(const AudioContentBlockSpec& spec) {
-  audioSpec_ = spec;
+AudioBlock::AudioBlock(const AudioContentBlockSpec& spec) : audioSpec_{spec} {
+  allocateBytes();
+}
+
+void AudioBlock::allocateBytes() {
   size_t size = audioSpec_.getBlockSize();
   if (size != ContentBlock::kSizeUnknown) {
     audioBytes_.resize(size);
   }
 }
 
+void AudioBlock::init(const AudioContentBlockSpec& spec) {
+  audioSpec_ = spec;
+  allocateBytes();
+}
+
 void AudioBlock::init(const AudioContentBlockSpec& spec, vector<uint8_t>&& frameBytes) {
   audioSpec_ = spec;
   audioBytes_ = std::move(frameBytes);

vrs/utils/AudioBlock.h

@@ -32,7 +32,7 @@ class AudioBlock {
   AudioBlock(AudioBlock&& other) noexcept = default;
   AudioBlock(const AudioContentBlockSpec& spec, vector<uint8_t>&& frameBytes);
   explicit AudioBlock(const AudioContentBlockSpec& spec);
-  explicit AudioBlock(
+  AudioBlock(
       AudioFormat audioFormat,
       AudioSampleFormat sampleFormat,
       uint8_t channelCount = 0,
@@ -84,6 +84,9 @@ class AudioBlock {
   uint32_t getSampleCount() const {
     return audioSpec_.getSampleCount();
   }
+  void setSampleCount(uint32_t sampleCount) {
+    audioSpec_.setSampleCount(sampleCount);
+  }
 
   vector<uint8_t>& getBuffer() {
     return audioBytes_;
@@ -94,6 +97,14 @@ class AudioBlock {
   uint8_t* wdata() {
     return audioBytes_.data();
   }
+  template <class T>
+  const T* data(size_t byte_offset = 0) const {
+    return reinterpret_cast<const T*>(audioBytes_.data() + byte_offset);
+  }
+  template <class T>
+  T* data(size_t byte_offset = 0) {
+    return reinterpret_cast<T*>(audioBytes_.data() + byte_offset);
+  }
   size_t size() const {
     return audioBytes_.size();
   }
@@ -109,6 +120,8 @@ class AudioBlock {
   bool readBlock(RecordReader* reader, const ContentBlock& cb);
 
  private:
+  void allocateBytes();
+
   AudioContentBlockSpec audioSpec_;
   vector<uint8_t> audioBytes_;
 };

vrs/utils/AudioExtractor.cpp

@@ -78,7 +78,7 @@ int AudioExtractor::createWavFile(
     format = 7;
   }
 
-  uint32_t bytesPerSample = (audioBlock.getBitsPerSample() + 7) / 8;
+  uint32_t bytesPerSample = audioBlock.getBytesPerSample();
   writeHeader(fileHeader.data() + 20, htole16(format)); // audio format, write in little endian
   writeHeader(
       fileHeader.data() + 22,
@@ -111,8 +111,7 @@ int AudioExtractor::writeWavAudioData(
     const AudioContentBlockSpec& audioBlock,
     const std::vector<uint8_t>& audio) {
   uint32_t srcOffset = 0;
-  uint32_t bytesPerSampleBlock =
-      (audioBlock.getBitsPerSample() + 7) / 8 * audioBlock.getChannelCount();
+  uint32_t bytesPerSampleBlock = audioBlock.getBytesPerSample() * audioBlock.getChannelCount();
   uint32_t srcStride = audioBlock.getSampleFrameStride();
   uint32_t totalSamples = audioBlock.getSampleCount();
   for (uint32_t i = 0; i < totalSamples; ++i) {