[Chromoting] Implement down mixing in AudioPump

remoting/protocol/audio_pump.cc

Index: remoting/protocol/audio_pump.cc
diff --git a/remoting/protocol/audio_pump.cc b/remoting/protocol/audio_pump.cc
index 66eff6ebff4d58946b7746163d2b9a1c32c2ef20..525ea24a490f3668459f83ce46d8def5fc6378fc 100644
--- a/remoting/protocol/audio_pump.cc
+++ b/remoting/protocol/audio_pump.cc
@@ -4,19 +4,86 @@
 
 #include "remoting/protocol/audio_pump.h"
 
+#include <memory>
 #include <utility>
 
 #include "base/bind.h"
 #include "base/location.h"
 #include "base/logging.h"
 #include "base/macros.h"
+#include "base/memory/ptr_util.h"
 #include "base/single_thread_task_runner.h"
 #include "base/threading/thread_task_runner_handle.h"
+#include "media/base/audio_bus.h"
+#include "media/base/audio_sample_types.h"
+#include "media/base/channel_layout.h"
+#include "media/base/channel_mixer.h"
 #include "remoting/codec/audio_encoder.h"
 #include "remoting/proto/audio.pb.h"
 #include "remoting/protocol/audio_source.h"
 #include "remoting/protocol/audio_stub.h"
 
+namespace {
+
+int CalculateFrameCount(const remoting::AudioPacket& packet) {
+  return packet.data(0).size() / packet.channels() / packet.bytes_per_sample();

[chcunningham, 2017/06/08 16:39:45]

I think this calculation could go sideways if you have a channel mask. My read
of the API docs are they wont send you data for channels that aren't in the
mask. 

[Hzj_jie, 2017/06/08 19:07:48]

We do not change the dwChannelMask returned by GetMixFormat(), so all channels
should be returned from Windows. Do you mean this?

+}
+
+std::unique_ptr<media::AudioBus> AudioPacketToAudioBus(
+    const remoting::AudioPacket& packet) {
+  const int frame_count = CalculateFrameCount(packet);
+  DCHECK_GT(frame_count, 0);
+  std::unique_ptr<media::AudioBus> result =
+      media::AudioBus::Create(packet.channels(), frame_count);
+  result->FromInterleaved<media::SignedInt16SampleTypeTraits>(
+      reinterpret_cast<const int16_t*>(packet.data(0).data()), frame_count);
+  return result;
+}
+
+std::unique_ptr<remoting::AudioPacket> AudioBusToAudioPacket(
+    const media::AudioBus& packet) {
+  std::unique_ptr<remoting::AudioPacket> result =
+      base::MakeUnique<remoting::AudioPacket>();
+  result->add_data()->resize(
+      packet.channels() * packet.frames() * sizeof(int16_t));
+  packet.ToInterleaved<media::SignedInt16SampleTypeTraits>(
+      packet.frames(),
+      reinterpret_cast<int16_t*>(&(result->mutable_data(0)->at(0))));
+  result->set_encoding(remoting::AudioPacket::ENCODING_RAW);
+  result->set_channels(
+      static_cast<remoting::AudioPacket::Channels>(packet.channels()));
+  result->set_bytes_per_sample(remoting::AudioPacket::BYTES_PER_SAMPLE_2);
+  return result;
+}
+
+media::ChannelLayout RetrieveLayout(const remoting::AudioPacket& packet) {
+  // This switch should match AudioPacket::Channels enum in audio.proto.
+  switch (packet.channels()) {
+    case remoting::AudioPacket::CHANNELS_INVALID:
+      return media::CHANNEL_LAYOUT_UNSUPPORTED;
+    case remoting::AudioPacket::CHANNELS_MONO:
+      return media::CHANNEL_LAYOUT_MONO;
+    case remoting::AudioPacket::CHANNELS_STEREO:
+      return media::CHANNEL_LAYOUT_STEREO;
+    case remoting::AudioPacket::CHANNELS_SURROUND:
+      return media::CHANNEL_LAYOUT_SURROUND;
+    case remoting::AudioPacket::CHANNELS_4_0:
+      return media::CHANNEL_LAYOUT_4_0;
+    case remoting::AudioPacket::CHANNELS_4_1:
+      return media::CHANNEL_LAYOUT_4_1;
+    case remoting::AudioPacket::CHANNELS_5_1:
+      return media::CHANNEL_LAYOUT_5_1;
+    case remoting::AudioPacket::CHANNELS_6_1:
+      return media::CHANNEL_LAYOUT_6_1;
+    case remoting::AudioPacket::CHANNELS_7_1:
+      return media::CHANNEL_LAYOUT_7_1;
+  }
+  NOTREACHED() << "Invalid AudioPacket::Channels";
+  return media::CHANNEL_LAYOUT_UNSUPPORTED;
+}
+
+}  // namespace
+
 namespace remoting {
 namespace protocol {
 
@@ -36,6 +103,8 @@ class AudioPump::Core {
   void OnPacketSent(int size);
 
  private:
+  std::unique_ptr<AudioPacket> Downmix(std::unique_ptr<AudioPacket> packet);
+
   void EncodeAudioPacket(std::unique_ptr<AudioPacket> packet);
 
   base::ThreadChecker thread_checker_;
@@ -53,6 +122,9 @@ class AudioPump::Core {
   // yet.
   int bytes_pending_;
 
+  std::unique_ptr<media::ChannelMixer> mixer_;
+  media::ChannelLayout mixer_input_layout_ = media::CHANNEL_LAYOUT_NONE;
+
   DISALLOW_COPY_AND_ASSIGN(Core);
 };
 
@@ -98,15 +170,21 @@ void AudioPump::Core::EncodeAudioPacket(std::unique_ptr<AudioPacket> packet) {
 
   int max_buffered_bytes =
       audio_encoder_->GetBitrate() * kMaxBufferedIntervalMs / 1000 / 8;
-  if (!enabled_ || bytes_pending_ > max_buffered_bytes)
+  if (!enabled_ || bytes_pending_ > max_buffered_bytes) {
     return;
+  }
+
+  if (packet->channels() > AudioPacket::CHANNELS_STEREO) {
+    packet = Downmix(std::move(packet));
+  }
 
   std::unique_ptr<AudioPacket> encoded_packet =
       audio_encoder_->Encode(std::move(packet));
 
   // The audio encoder returns a null audio packet if there's no audio to send.
-  if (!encoded_packet)
+  if (!encoded_packet) {
     return;
+  }
 
   int packet_size = encoded_packet->ByteSize();
   bytes_pending_ += packet_size;
@@ -116,6 +194,35 @@ void AudioPump::Core::EncodeAudioPacket(std::unique_ptr<AudioPacket> packet) {
                             base::Passed(&encoded_packet), packet_size));
 }
 
+std::unique_ptr<AudioPacket> AudioPump::Core::Downmix(
+    std::unique_ptr<AudioPacket> packet) {
+  DCHECK(thread_checker_.CalledOnValidThread());
+  DCHECK(packet);
+  DCHECK_EQ(packet->data_size(), 1);
+  DCHECK_EQ(packet->bytes_per_sample(), AudioPacket::BYTES_PER_SAMPLE_2);
+
+  const media::ChannelLayout input_layout = RetrieveLayout(*packet);
+  DCHECK_NE(input_layout, media::CHANNEL_LAYOUT_UNSUPPORTED);
+  DCHECK_NE(input_layout, media::CHANNEL_LAYOUT_MONO);
+  DCHECK_NE(input_layout, media::CHANNEL_LAYOUT_STEREO);
+
+  if (!mixer_ || mixer_input_layout_ != input_layout) {
+    mixer_input_layout_ = input_layout;
+    mixer_ = base::MakeUnique<media::ChannelMixer>(
+        input_layout, media::CHANNEL_LAYOUT_STEREO);
+  }
+
+  std::unique_ptr<media::AudioBus> input = AudioPacketToAudioBus(*packet);
+  DCHECK(input);
+  std::unique_ptr<media::AudioBus> output =
+      media::AudioBus::Create(AudioPacket::CHANNELS_STEREO, input->frames());
+  mixer_->Transform(input.get(), output.get());
+
+  std::unique_ptr<AudioPacket> result = AudioBusToAudioPacket(*output);
+  result->set_sampling_rate(packet->sampling_rate());
+  return result;
+}
+
 AudioPump::AudioPump(
     scoped_refptr<base::SingleThreadTaskRunner> audio_task_runner,
     std::unique_ptr<AudioSource> audio_source,