Implement runtime audio post-processing pipeline. See go/cast_audio.json

chromecast/media/cma/backend/alsa/filter_group.cc

Index: chromecast/media/cma/backend/alsa/filter_group.cc
diff --git a/chromecast/media/cma/backend/alsa/filter_group.cc b/chromecast/media/cma/backend/alsa/filter_group.cc
index 1fa4df6952b435096a288b02e31cad404cfc3bf0..b054ca04ad482cfdbf5ec919943fdb7cbcac9862 100644
--- a/chromecast/media/cma/backend/alsa/filter_group.cc
+++ b/chromecast/media/cma/backend/alsa/filter_group.cc
@@ -3,40 +3,30 @@
 // found in the LICENSE file.
 
 #include "chromecast/media/cma/backend/alsa/filter_group.h"
+
+#include "base/memory/ptr_util.h"
+#include "chromecast/media/cma/backend/alsa/post_processing_pipeline.h"
 #include "media/base/audio_bus.h"
 
 namespace chromecast {
 namespace media {
 
-namespace {
-
-// How many seconds of silence should be passed to the filters to flush them.
-const float kSilenceSecondsToFilter = 1.0f;
-const int kNumOutputChannels = 2;
-
-}  // namespace
-
 FilterGroup::FilterGroup(const std::unordered_set<std::string>& input_types,
-                         AudioFilterFactory::FilterType filter_type,
-                         AudioContentType content_type)
+                         AudioContentType content_type,
+                         int channels,
+                         const base::ListValue* filter_list)
     : input_types_(input_types),
       content_type_(content_type),
+      channels_(channels),
       output_samples_per_second_(0),
-      sample_format_(::media::SampleFormat::kUnknownSampleFormat),
-      audio_filter_(AudioFilterFactory::MakeAudioFilter(filter_type)),
-      silence_frames_filtered_(0) {}
+      post_processing_pipeline_(
+          base::MakeUnique<PostProcessingPipeline>(filter_list, channels_)) {}
 
 FilterGroup::~FilterGroup() = default;
 
-void FilterGroup::Initialize(int output_samples_per_second,
-                             ::media::SampleFormat format) {
+void FilterGroup::Initialize(int output_samples_per_second) {
   output_samples_per_second_ = output_samples_per_second;
-  sample_format_ = format;
-  if (audio_filter_) {
-    audio_filter_->SetSampleRateAndFormat(output_samples_per_second_,
-                                          sample_format_);
-  }
-  silence_frames_filtered_ = 0;
+  post_processing_pipeline_->SetSampleRate(output_samples_per_second);
 }
 
 bool FilterGroup::CanProcessInput(StreamMixerAlsa::InputQueue* input) {
@@ -53,17 +43,8 @@ std::vector<uint8_t>* FilterGroup::GetInterleaved() {
 
 bool FilterGroup::MixAndFilter(int chunk_size) {
   DCHECK_NE(output_samples_per_second_, 0);
-  DCHECK_NE(sample_format_, ::media::SampleFormat::kUnknownSampleFormat);
-  if (active_inputs_.empty()) {
-    int silence_frames_to_filter =
-        output_samples_per_second_ * kSilenceSecondsToFilter;
-    if (audio_filter_ && silence_frames_filtered_ < silence_frames_to_filter) {
-      silence_frames_filtered_ += chunk_size;
-    } else {
-      return false;  // Output will be silence, no need to mix.
-    }
-  } else {
-    silence_frames_filtered_ = 0;
+  if (active_inputs_.empty() && !post_processing_pipeline_->IsRinging()) {
+    return false;  // Output will be silence, no need to mix.
   }
 
   ResizeBuffersIfNecessary(chunk_size);
@@ -71,7 +52,7 @@ bool FilterGroup::MixAndFilter(int chunk_size) {
   mixed_->ZeroFramesPartial(0, chunk_size);
   for (StreamMixerAlsa::InputQueue* input : active_inputs_) {
     input->GetResampledData(temp_.get(), chunk_size);
-    for (int c = 0; c < kNumOutputChannels; ++c) {
+    for (int c = 0; c < channels_; ++c) {
       input->VolumeScaleAccumulate(c, temp_->channel(c), chunk_size,
                                    mixed_->channel(c));
     }
@@ -79,30 +60,29 @@ bool FilterGroup::MixAndFilter(int chunk_size) {
 
   mixed_->ToInterleaved(chunk_size, BytesPerOutputFormatSample(),
                         interleaved_.data());
-  if (audio_filter_) {
-    audio_filter_->ProcessInterleaved(interleaved_.data(), chunk_size, volume_);
-  }
+  post_processing_pipeline_->ProcessFrames(interleaved_.data(), chunk_size,

[kmackay, 2017/03/24 22:15:54]

Don't we need to use the return value of ProcessFrames() to accurately estimate
the rendering delay?

[bshaya, 2017/03/24 22:55:49]

Added TODO, will add in separate CL.

+                                           volume_, active_inputs_.empty());
 
   return true;
 }
 
 void FilterGroup::ClearInterleaved(int chunk_size) {
   ResizeBuffersIfNecessary(chunk_size);
-  memset(interleaved_.data(), 0, static_cast<size_t>(chunk_size) *
-                                     kNumOutputChannels *
-                                     BytesPerOutputFormatSample());
+  memset(interleaved_.data(), 0,
+         static_cast<size_t>(chunk_size) * channels_ *
+             BytesPerOutputFormatSample());
 }
 
 void FilterGroup::ResizeBuffersIfNecessary(int chunk_size) {
   if (!mixed_ || mixed_->frames() < chunk_size) {
-    mixed_ = ::media::AudioBus::Create(kNumOutputChannels, chunk_size);
+    mixed_ = ::media::AudioBus::Create(channels_, chunk_size);
   }
   if (!temp_ || temp_->frames() < chunk_size) {
-    temp_ = ::media::AudioBus::Create(kNumOutputChannels, chunk_size);
+    temp_ = ::media::AudioBus::Create(channels_, chunk_size);
   }
 
-  size_t interleaved_size = static_cast<size_t>(chunk_size) *
-                            kNumOutputChannels * BytesPerOutputFormatSample();
+  size_t interleaved_size = static_cast<size_t>(chunk_size) * channels_ *
+                            BytesPerOutputFormatSample();
 
   if (interleaved_.size() < interleaved_size) {
     interleaved_.resize(interleaved_size);
@@ -110,7 +90,7 @@ void FilterGroup::ResizeBuffersIfNecessary(int chunk_size) {
 }
 
 int FilterGroup::BytesPerOutputFormatSample() {
-  return ::media::SampleFormatToBytesPerChannel(sample_format_);
+  return sizeof(int32_t);
 }
 
 void FilterGroup::ClearActiveInputs() {