media/cast/audio_sender/audio_encoder.cc

   1 // Copyright 2013 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "media/cast/audio_sender/audio_encoder.h"
   6
   7 #include <algorithm>
   8
   9 #include "base/bind.h"
  10 #include "base/bind_helpers.h"
  11 #include "base/logging.h"
  12 #include "base/message_loop/message_loop.h"
  13 #include "base/sys_byteorder.h"
  14 #include "base/time/time.h"
  15 #include "media/base/audio_bus.h"
  16 #include "media/cast/cast_defines.h"
  17 #include "media/cast/cast_environment.h"
  18 #include "third_party/opus/src/include/opus.h"
  19
  20 namespace media {
  21 namespace cast {
  22
  23 void LogAudioEncodedEvent(CastEnvironment* const cast_environment,
  24                           const base::TimeTicks& recorded_time) {
  25   // TODO(mikhal): Resolve timestamp calculation for audio.
  26   base::TimeTicks now = cast_environment->Clock()->NowTicks();
  27
  28   cast_environment->Logging()->InsertFrameEvent(now, kAudioFrameEncoded,
  29       GetVideoRtpTimestamp(recorded_time), kFrameIdUnknown);
  30 }
  31
  32 // Base class that handles the common problem of feeding one or more AudioBus'
  33 // data into a 10 ms buffer and then, once the buffer is full, encoding the
  34 // signal and emitting an EncodedAudioFrame via the FrameEncodedCallback.
  35 //
  36 // Subclasses complete the implementation by handling the actual encoding
  37 // details.
  38 class AudioEncoder::ImplBase {
  39  public:
  40   ImplBase(CastEnvironment* cast_environment,
  41            transport::AudioCodec codec, int num_channels, int sampling_rate,
  42            const FrameEncodedCallback& callback)
  43       : cast_environment_(cast_environment),
  44         codec_(codec), num_channels_(num_channels),
  45         samples_per_10ms_(sampling_rate / 100),
  46         callback_(callback),
  47         buffer_fill_end_(0),
  48         frame_id_(0) {
  49     CHECK_GT(num_channels_, 0);
  50     CHECK_GT(samples_per_10ms_, 0);
  51     CHECK_EQ(sampling_rate % 100, 0);
  52     CHECK_LE(samples_per_10ms_ * num_channels_,
  53              transport::EncodedAudioFrame::kMaxNumberOfSamples);
  54   }
  55
  56   virtual ~ImplBase() {}
  57
  58   void EncodeAudio(const AudioBus* audio_bus,
  59                    const base::TimeTicks& recorded_time,
  60                    const base::Closure& done_callback) {
  61     int src_pos = 0;
  62     while (audio_bus && src_pos < audio_bus->frames()) {
  63       const int num_samples_to_xfer =
  64           std::min(samples_per_10ms_ - buffer_fill_end_,
  65                    audio_bus->frames() - src_pos);
  66       DCHECK_EQ(audio_bus->channels(), num_channels_);
  67       TransferSamplesIntoBuffer(
  68           audio_bus, src_pos, buffer_fill_end_, num_samples_to_xfer);
  69       src_pos += num_samples_to_xfer;
  70       buffer_fill_end_ += num_samples_to_xfer;
  71
  72       if (src_pos == audio_bus->frames()) {
  73         cast_environment_->PostTask(CastEnvironment::MAIN, FROM_HERE,
  74                                     done_callback);
  75         // Note: |audio_bus| is invalid once done_callback is invoked.
  76         audio_bus = NULL;
  77       }
  78
  79       if (buffer_fill_end_ == samples_per_10ms_) {
  80         scoped_ptr<transport::EncodedAudioFrame> audio_frame(
  81             new transport::EncodedAudioFrame());
  82         audio_frame->codec = codec_;
  83         audio_frame->frame_id = frame_id_++;
  84         audio_frame->samples = samples_per_10ms_;
  85         if (EncodeFromFilledBuffer(&audio_frame->data)) {
  86           // Compute an offset to determine the recorded time for the first
  87           // audio sample in the buffer.
  88           const base::TimeDelta buffer_time_offset =
  89               (buffer_fill_end_ - src_pos) *
  90               base::TimeDelta::FromMilliseconds(10) / samples_per_10ms_;
  91           // TODO(miu): Consider batching EncodedAudioFrames so we only post a
  92           // at most one task for each call to this method.
  93           cast_environment_->PostTask(
  94               CastEnvironment::MAIN, FROM_HERE,
  95               base::Bind(callback_, base::Passed(&audio_frame),
  96                          recorded_time - buffer_time_offset));
  97         }
  98         buffer_fill_end_ = 0;
  99       }
 100     }
 101   }
 102
 103  protected:
 104   virtual void TransferSamplesIntoBuffer(const AudioBus* audio_bus,
 105                                          int source_offset,
 106                                          int buffer_fill_offset,
 107                                          int num_samples) = 0;
 108   virtual bool EncodeFromFilledBuffer(std::string* out) = 0;
 109
 110   CastEnvironment* const cast_environment_;
 111   const transport::AudioCodec codec_;
 112   const int num_channels_;
 113   const int samples_per_10ms_;
 114   const FrameEncodedCallback callback_;
 115
 116  private:
 117   // In the case where a call to EncodeAudio() cannot completely fill the
 118   // buffer, this points to the position at which to populate data in a later
 119   // call.
 120   int buffer_fill_end_;
 121
 122   // A counter used to label EncodedAudioFrames.
 123   uint32 frame_id_;
 124
 125  private:
 126   DISALLOW_COPY_AND_ASSIGN(ImplBase);
 127 };
 128
 129 class AudioEncoder::OpusImpl : public AudioEncoder::ImplBase {
 130  public:
 131   OpusImpl(CastEnvironment* cast_environment,
 132            int num_channels, int sampling_rate, int bitrate,
 133            const FrameEncodedCallback& callback)
 134       : ImplBase(cast_environment, transport::kOpus, num_channels,
 135                  sampling_rate, callback),
 136         encoder_memory_(new uint8[opus_encoder_get_size(num_channels)]),
 137         opus_encoder_(reinterpret_cast<OpusEncoder*>(encoder_memory_.get())),
 138         buffer_(new float[num_channels * samples_per_10ms_]) {
 139     CHECK_EQ(opus_encoder_init(opus_encoder_, sampling_rate, num_channels,
 140                                OPUS_APPLICATION_AUDIO),
 141              OPUS_OK);
 142     if (bitrate <= 0) {
 143       // Note: As of 2013-10-31, the encoder in "auto bitrate" mode would use a
 144       // variable bitrate up to 102kbps for 2-channel, 48 kHz audio and a 10 ms
 145       // frame size.  The opus library authors may, of course, adjust this in
 146       // later versions.
 147       bitrate = OPUS_AUTO;
 148     }
 149     CHECK_EQ(opus_encoder_ctl(opus_encoder_, OPUS_SET_BITRATE(bitrate)),
 150              OPUS_OK);
 151   }
 152
 153   virtual ~OpusImpl() {}
 154
 155  private:
 156   virtual void TransferSamplesIntoBuffer(const AudioBus* audio_bus,
 157                                          int source_offset,
 158                                          int buffer_fill_offset,
 159                                          int num_samples) OVERRIDE {
 160     // Opus requires channel-interleaved samples in a single array.
 161     for (int ch = 0; ch < audio_bus->channels(); ++ch) {
 162       const float* src = audio_bus->channel(ch) + source_offset;
 163       const float* const src_end = src + num_samples;
 164       float* dest = buffer_.get() + buffer_fill_offset * num_channels_ + ch;
 165       for (; src < src_end; ++src, dest += num_channels_)
 166         *dest = *src;
 167     }
 168   }
 169
 170   virtual bool EncodeFromFilledBuffer(std::string* out) OVERRIDE {
 171     out->resize(kOpusMaxPayloadSize);
 172     const opus_int32 result = opus_encode_float(
 173         opus_encoder_, buffer_.get(), samples_per_10ms_,
 174         reinterpret_cast<uint8*>(&out->at(0)), kOpusMaxPayloadSize);
 175     if (result > 1) {
 176       out->resize(result);
 177       return true;
 178     } else if (result < 0) {
 179       LOG(ERROR) << "Error code from opus_encode_float(): " << result;
 180       return false;
 181     } else {
 182       // Do nothing: The documentation says that a return value of zero or
 183       // one byte means the packet does not need to be transmitted.
 184       return false;
 185     }
 186   }
 187
 188   const scoped_ptr<uint8[]> encoder_memory_;
 189   OpusEncoder* const opus_encoder_;
 190   const scoped_ptr<float[]> buffer_;
 191
 192   // This is the recommended value, according to documentation in
 193   // third_party/opus/src/include/opus.h, so that the Opus encoder does not
 194   // degrade the audio due to memory constraints.
 195   //
 196   // Note: Whereas other RTP implementations do not, the cast library is
 197   // perfectly capable of transporting larger than MTU-sized audio frames.
 198   static const int kOpusMaxPayloadSize = 4000;
 199
 200   DISALLOW_COPY_AND_ASSIGN(OpusImpl);
 201 };
 202
 203 class AudioEncoder::Pcm16Impl : public AudioEncoder::ImplBase {
 204  public:
 205   Pcm16Impl(CastEnvironment* cast_environment,
 206             int num_channels, int sampling_rate,
 207             const FrameEncodedCallback& callback)
 208       : ImplBase(cast_environment, transport::kPcm16, num_channels,
 209                  sampling_rate, callback),
 210         buffer_(new int16[num_channels * samples_per_10ms_]) {}
 211
 212   virtual ~Pcm16Impl() {}
 213
 214  private:
 215   virtual void TransferSamplesIntoBuffer(const AudioBus* audio_bus,
 216                                          int source_offset,
 217                                          int buffer_fill_offset,
 218                                          int num_samples) OVERRIDE {
 219     audio_bus->ToInterleavedPartial(
 220         source_offset, num_samples, sizeof(int16),
 221         buffer_.get() + buffer_fill_offset * num_channels_);
 222   }
 223
 224   virtual bool EncodeFromFilledBuffer(std::string* out) OVERRIDE {
 225     // Output 16-bit PCM integers in big-endian byte order.
 226     out->resize(num_channels_ * samples_per_10ms_ * sizeof(int16));
 227     const int16* src = buffer_.get();
 228     const int16* const src_end = src + num_channels_ * samples_per_10ms_;
 229     uint16* dest = reinterpret_cast<uint16*>(&out->at(0));
 230     for (; src < src_end; ++src, ++dest)
 231       *dest = base::HostToNet16(*src);
 232     return true;
 233   }
 234
 235  private:
 236   const scoped_ptr<int16[]> buffer_;
 237
 238   DISALLOW_COPY_AND_ASSIGN(Pcm16Impl);
 239 };
 240
 241 AudioEncoder::AudioEncoder(
 242     const scoped_refptr<CastEnvironment>& cast_environment,
 243     const AudioSenderConfig& audio_config,
 244     const FrameEncodedCallback& frame_encoded_callback)
 245     : cast_environment_(cast_environment) {
 246   // Note: It doesn't matter which thread constructs AudioEncoder, just so long
 247   // as all calls to InsertAudio() are by the same thread.
 248   insert_thread_checker_.DetachFromThread();
 249
 250   switch (audio_config.codec) {
 251     case transport::kOpus:
 252       impl_.reset(new OpusImpl(
 253           cast_environment, audio_config.channels, audio_config.frequency,
 254           audio_config.bitrate, frame_encoded_callback));
 255       break;
 256     case transport::kPcm16:
 257       impl_.reset(new Pcm16Impl(
 258           cast_environment, audio_config.channels, audio_config.frequency,
 259           frame_encoded_callback));
 260       break;
 261     default:
 262       NOTREACHED() << "Unsupported or unspecified codec for audio encoder";
 263       break;
 264   }
 265 }
 266
 267 AudioEncoder::~AudioEncoder() {}
 268
 269 void AudioEncoder::InsertAudio(
 270     const AudioBus* audio_bus,
 271     const base::TimeTicks& recorded_time,
 272     const base::Closure& done_callback) {
 273   DCHECK(insert_thread_checker_.CalledOnValidThread());
 274   if (!impl_) {
 275     NOTREACHED();
 276     cast_environment_->PostTask(CastEnvironment::MAIN, FROM_HERE,
 277                                 done_callback);
 278     return;
 279   }
 280   cast_environment_->PostTask(CastEnvironment::AUDIO_ENCODER, FROM_HERE,
 281       base::Bind(&AudioEncoder::EncodeAudio, this, audio_bus, recorded_time,
 282                  done_callback));
 283 }
 284
 285 void AudioEncoder::EncodeAudio(
 286     const AudioBus* audio_bus,
 287     const base::TimeTicks& recorded_time,
 288     const base::Closure& done_callback) {
 289   DCHECK(cast_environment_->CurrentlyOn(CastEnvironment::AUDIO_ENCODER));
 290   impl_->EncodeAudio(audio_bus, recorded_time, done_callback);
 291   cast_environment_->PostTask(CastEnvironment::MAIN, FROM_HERE,
 292       base::Bind(LogAudioEncodedEvent, cast_environment_, recorded_time));
 293 }
 294
 295 }  // namespace cast
 296 }  // namespace media