src/starboard/shared/starboard/player/filter/audio_renderer_impl_internal.cc

// Copyright 2016 Google Inc. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

#include "starboard/shared/starboard/player/filter/audio_renderer_impl_internal.h"

#include <algorithm>

#include "starboard/memory.h"
#include "starboard/shared/starboard/audio_sink/audio_sink_internal.h"
#include "starboard/shared/starboard/media/media_util.h"

namespace starboard {
namespace shared {
namespace starboard {
namespace player {
namespace filter {

namespace {

// This class works only when the input format and output format are the same.
// It allows for a simplified AudioRendererImpl implementation by always using a
// resampler.
class IdentityAudioResampler : public AudioResampler {
 public:
  IdentityAudioResampler() : eos_reached_(false) {}
  scoped_refptr<DecodedAudio> Resample(
      const scoped_refptr<DecodedAudio>& audio_data) SB_OVERRIDE {
    SB_DCHECK(!eos_reached_);

    return audio_data;
  }
  scoped_refptr<DecodedAudio> WriteEndOfStream() SB_OVERRIDE {
    SB_DCHECK(!eos_reached_);
    eos_reached_ = true;
    return new DecodedAudio();
  }

 private:
  bool eos_reached_;
};

// AudioRendererImpl uses AudioTimeStretcher internally to adjust to playback
// rate and AudioTimeStretcher can only process float32 samples.  So we try to
// use kSbMediaAudioSampleTypeFloat32 and only use kSbMediaAudioSampleTypeInt16
// when float32 is not supported.  To use kSbMediaAudioSampleTypeFloat32 will
// cause an extra conversion from float32 to int16 before the samples are sent
// to the audio sink.
SbMediaAudioSampleType GetSinkAudioSampleType() {
  return SbAudioSinkIsAudioSampleTypeSupported(kSbMediaAudioSampleTypeFloat32)
             ? kSbMediaAudioSampleTypeFloat32
             : kSbMediaAudioSampleTypeInt16;
}

}  // namespace

AudioRendererImpl::AudioRendererImpl(scoped_ptr<AudioDecoder> decoder,
                                     const SbMediaAudioHeader& audio_header)
    : eos_state_(kEOSNotReceived),
      channels_(audio_header.number_of_channels),
      sink_sample_type_(GetSinkAudioSampleType()),
      bytes_per_frame_(media::GetBytesPerSample(sink_sample_type_) * channels_),
      playback_rate_(1.0),
      paused_(true),
      seeking_(false),
      seeking_to_pts_(0),
      frame_buffer_(kMaxCachedFrames * bytes_per_frame_),
      frames_sent_to_sink_(0),
      pending_decoder_outputs_(0),
      frames_consumed_by_sink_(0),
      frames_consumed_set_at_(SbTimeGetMonotonicNow()),
      decoder_(decoder.Pass()),
      audio_sink_(kSbAudioSinkInvalid),
      can_accept_more_data_(true),
      process_audio_data_scheduled_(false),
      decoder_needs_full_reset_(false) {
  SB_DCHECK(decoder_ != NULL);

  frame_buffers_[0] = &frame_buffer_[0];

#if defined(NDEBUG)
  const bool kLogFramesConsumed = false;
#else
  const bool kLogFramesConsumed = true;
#endif
  if (kLogFramesConsumed) {
    log_frames_consumed_closure_ =
        Bind(&AudioRendererImpl::LogFramesConsumed, this);
    Schedule(log_frames_consumed_closure_, kSbTimeSecond);
  }

  decoder_->Initialize(Bind(&AudioRendererImpl::OnDecoderOutput, this));

  int source_sample_rate = decoder_->GetSamplesPerSecond();
  int destination_sample_rate =
      SbAudioSinkGetNearestSupportedSampleFrequency(source_sample_rate);
  time_stretcher_.Initialize(channels_, destination_sample_rate);
}

AudioRendererImpl::~AudioRendererImpl() {
  SB_DCHECK(BelongsToCurrentThread());

  if (audio_sink_ != kSbAudioSinkInvalid) {
    SbAudioSinkDestroy(audio_sink_);
  }
}

void AudioRendererImpl::WriteSample(const InputBuffer& input_buffer) {
  SB_DCHECK(BelongsToCurrentThread());
  SB_DCHECK(can_accept_more_data_);

  if (eos_state_.load() >= kEOSWrittenToDecoder) {
    SB_LOG(ERROR) << "Appending audio sample at " << input_buffer.pts()
                  << " after EOS reached.";
    return;
  }

  can_accept_more_data_ = false;

  decoder_->Decode(input_buffer,
                   Bind(&AudioRendererImpl::OnDecoderConsumed, this));
  decoder_needs_full_reset_ = true;
}

void AudioRendererImpl::WriteEndOfStream() {
  SB_DCHECK(BelongsToCurrentThread());
  // TODO: Check |can_accept_more_data_| and make WriteEndOfStream() depend on
  // CanAcceptMoreData() or callback.
  // SB_DCHECK(can_accept_more_data_);
  // can_accept_more_data_ = false;

  if (eos_state_.load() >= kEOSWrittenToDecoder) {
    SB_LOG(ERROR) << "Try to write EOS after EOS is reached";
    return;
  }

  decoder_->WriteEndOfStream();

  eos_state_.store(kEOSWrittenToDecoder);
  decoder_needs_full_reset_ = true;
}

void AudioRendererImpl::Play() {
  SB_DCHECK(BelongsToCurrentThread());

  paused_.store(false);
}

void AudioRendererImpl::Pause() {
  SB_DCHECK(BelongsToCurrentThread());

  paused_.store(true);
}

#if SB_API_VERSION >= 4
void AudioRendererImpl::SetPlaybackRate(double playback_rate) {
  SB_DCHECK(BelongsToCurrentThread());

  playback_rate_ = playback_rate;

  if (audio_sink_) {
    // TODO: Remove SetPlaybackRate() support from audio sink as it only need to
    // support play/pause.
    audio_sink_->SetPlaybackRate(playback_rate_ > 0.0 ? 1.0 : 0.0);
  }
}
#endif  // SB_API_VERSION >= 4

void AudioRendererImpl::Seek(SbMediaTime seek_to_pts) {
  SB_DCHECK(BelongsToCurrentThread());
  SB_DCHECK(seek_to_pts >= 0);

  SbAudioSinkDestroy(audio_sink_);

  // Now the sink is destroyed and the callbacks will no longer be called, so
  // the following modifications are safe without lock.
  audio_sink_ = kSbAudioSinkInvalid;

  if (resampler_) {
    resampler_.reset();
    time_stretcher_.FlushBuffers();
  }

  eos_state_.store(kEOSNotReceived);
  seeking_to_pts_ = std::max<SbMediaTime>(seek_to_pts, 0);
  seeking_.store(true);
  frames_sent_to_sink_.store(0);
  frames_consumed_by_sink_.store(0);
  frames_consumed_by_sink_since_last_get_current_time_.store(0);
  pending_decoder_outputs_ = 0;
  audio_frame_tracker_.Reset();
  frames_consumed_set_at_.store(SbTimeGetMonotonicNow());
  can_accept_more_data_ = true;
  process_audio_data_scheduled_ = false;

  if (decoder_needs_full_reset_) {
    decoder_->Reset();
    decoder_needs_full_reset_ = false;
  }

  CancelPendingJobs();

  if (log_frames_consumed_closure_.is_valid()) {
    Schedule(log_frames_consumed_closure_, kSbTimeSecond);
  }
}

bool AudioRendererImpl::IsEndOfStreamPlayed() const {
  SB_DCHECK(BelongsToCurrentThread());

  return eos_state_.load() >= kEOSSentToSink &&
         frames_sent_to_sink_.load() == frames_consumed_by_sink_.load();
}

bool AudioRendererImpl::CanAcceptMoreData() const {
  SB_DCHECK(BelongsToCurrentThread());

  return eos_state_.load() == kEOSNotReceived && can_accept_more_data_ &&
         !time_stretcher_.IsQueueFull();
}

bool AudioRendererImpl::IsSeekingInProgress() const {
  SB_DCHECK(BelongsToCurrentThread());
  return seeking_.load();
}

SbMediaTime AudioRendererImpl::GetCurrentTime() {
  SB_DCHECK(BelongsToCurrentThread());

  if (seeking_.load()) {
    return seeking_to_pts_;
  }

  audio_frame_tracker_.RecordPlayedFrames(
      frames_consumed_by_sink_since_last_get_current_time_.exchange(0));

  return seeking_to_pts_ +
         audio_frame_tracker_.GetFramePlayedAdjustedToPlaybackRate() *
             kSbMediaTimeSecond / decoder_->GetSamplesPerSecond();
}

void AudioRendererImpl::CreateAudioSinkAndResampler() {
  int source_sample_rate = decoder_->GetSamplesPerSecond();
  SbMediaAudioSampleType source_sample_type = decoder_->GetSampleType();
  SbMediaAudioFrameStorageType source_storage_type = decoder_->GetStorageType();

  int destination_sample_rate =
      SbAudioSinkGetNearestSupportedSampleFrequency(source_sample_rate);

  // AudioTimeStretcher only supports interleaved float32 samples.
  if (source_sample_rate != destination_sample_rate ||
      source_sample_type != kSbMediaAudioSampleTypeFloat32 ||
      source_storage_type != kSbMediaAudioFrameStorageTypeInterleaved) {
    resampler_ = AudioResampler::Create(
        decoder_->GetSampleType(), decoder_->GetStorageType(),
        source_sample_rate, kSbMediaAudioSampleTypeFloat32,
        kSbMediaAudioFrameStorageTypeInterleaved, destination_sample_rate,
        channels_);
    SB_DCHECK(resampler_);
  } else {
    resampler_.reset(new IdentityAudioResampler);
  }

  // TODO: Support planar only audio sink.
  audio_sink_ = SbAudioSinkCreate(
      channels_, destination_sample_rate, sink_sample_type_,
      kSbMediaAudioFrameStorageTypeInterleaved,
      reinterpret_cast<SbAudioSinkFrameBuffers>(frame_buffers_),
      kMaxCachedFrames, &AudioRendererImpl::UpdateSourceStatusFunc,
      &AudioRendererImpl::ConsumeFramesFunc, this);
  SB_DCHECK(SbAudioSinkIsValid(audio_sink_));

#if SB_API_VERSION >= 4
  // TODO: Remove SetPlaybackRate() support from audio sink as it only need to
  // support play/pause.
  audio_sink_->SetPlaybackRate(playback_rate_ > 0.0 ? 1.0 : 0.0);
#endif  // SB_API_VERSION >= 4
}

void AudioRendererImpl::UpdateSourceStatus(int* frames_in_buffer,
                                           int* offset_in_frames,
                                           bool* is_playing,
                                           bool* is_eos_reached) {
  *is_eos_reached = eos_state_.load() >= kEOSSentToSink;

  *is_playing = !paused_.load() && !seeking_.load();

  if (*is_playing) {
    *frames_in_buffer = static_cast<int>(frames_sent_to_sink_.load() -
                                         frames_consumed_by_sink_.load());
    *offset_in_frames = frames_consumed_by_sink_.load() % kMaxCachedFrames;
  } else {
    *frames_in_buffer = *offset_in_frames = 0;
  }
}

void AudioRendererImpl::ConsumeFrames(int frames_consumed) {
  frames_consumed_by_sink_.fetch_add(frames_consumed);
  SB_DCHECK(frames_consumed_by_sink_.load() <= frames_sent_to_sink_.load());
  frames_consumed_by_sink_since_last_get_current_time_.fetch_add(
      frames_consumed);
  frames_consumed_set_at_.store(SbTimeGetMonotonicNow());
}

void AudioRendererImpl::LogFramesConsumed() {
  SbTimeMonotonic time_since =
      SbTimeGetMonotonicNow() - frames_consumed_set_at_.load();
  if (time_since > kSbTimeSecond) {
    SB_DLOG(WARNING) << "|frames_consumed_| has not been updated for "
                     << (time_since / kSbTimeSecond) << "."
                     << ((time_since / (kSbTimeSecond / 10)) % 10)
                     << " seconds";
  }
  Schedule(log_frames_consumed_closure_, kSbTimeSecond);
}

void AudioRendererImpl::OnDecoderConsumed() {
  SB_DCHECK(BelongsToCurrentThread());

  // TODO: Unify EOS and non EOS request once WriteEndOfStream() depends on
  // CanAcceptMoreData().
  if (eos_state_.load() == kEOSNotReceived) {
    SB_DCHECK(!can_accept_more_data_);

    can_accept_more_data_ = true;
  }
}

void AudioRendererImpl::OnDecoderOutput() {
  SB_DCHECK(BelongsToCurrentThread());

  ++pending_decoder_outputs_;

  if (process_audio_data_scheduled_) {
    // A ProcessAudioData() callback has been scheduled and we should let it
    // process the output.
    return;
  }

  process_audio_data_scheduled_ = true;
  ProcessAudioData();
}

void AudioRendererImpl::ProcessAudioData() {
  process_audio_data_scheduled_ = false;

  if (!SbAudioSinkIsValid(audio_sink_)) {
    CreateAudioSinkAndResampler();
  }

  // Loop until no audio is appended, i.e. AppendAudioToFrameBuffer() returns
  // false.
  while (AppendAudioToFrameBuffer()) {
  }

  while (pending_decoder_outputs_ > 0) {
    if (time_stretcher_.IsQueueFull()) {
      // There is no room to do any further processing, schedule the function
      // again for a later time.  The delay time is 1/4 of the buffer size.
      const SbTimeMonotonic delay = kMaxCachedFrames * kSbTimeSecond /
                                    decoder_->GetSamplesPerSecond() / 4;
      process_audio_data_scheduled_ = true;
      Schedule(Bind(&AudioRendererImpl::ProcessAudioData, this), delay);
      return;
    }

    scoped_refptr<DecodedAudio> resampled_audio;
    scoped_refptr<DecodedAudio> decoded_audio = decoder_->Read();

    SB_DCHECK(decoded_audio);
    --pending_decoder_outputs_;

    if (decoded_audio->is_end_of_stream()) {
      SB_DCHECK(eos_state_.load() == kEOSWrittenToDecoder) << eos_state_.load();
      eos_state_.store(kEOSDecoded);
      seeking_.store(false);

      resampled_audio = resampler_->WriteEndOfStream();
    } else {
      // Discard any audio data before the seeking target.
      if (seeking_.load() && decoded_audio->pts() < seeking_to_pts_) {
        continue;
      }

      resampled_audio = resampler_->Resample(decoded_audio);
    }

    if (resampled_audio->size() > 0) {
      time_stretcher_.EnqueueBuffer(resampled_audio);
    }

    // Loop until no audio is appended, i.e. AppendAudioToFrameBuffer() returns
    // false.
    while (AppendAudioToFrameBuffer()) {
    }
  }

  int64_t frames_in_buffer =
      frames_sent_to_sink_.load() - frames_consumed_by_sink_.load();
  if (kMaxCachedFrames - frames_in_buffer < kFrameAppendUnit &&
      eos_state_.load() < kEOSSentToSink) {
    // There are still audio data not appended so schedule a callback later.
    SbTimeMonotonic delay = 0;
    if (kMaxCachedFrames - frames_in_buffer < kMaxCachedFrames / 4) {
      int frames_to_delay = static_cast<int>(
          kMaxCachedFrames / 4 - (kMaxCachedFrames - frames_in_buffer));
      delay = frames_to_delay * kSbTimeSecond / decoder_->GetSamplesPerSecond();
    }
    process_audio_data_scheduled_ = true;
    Schedule(Bind(&AudioRendererImpl::ProcessAudioData, this), delay);
  }
}

bool AudioRendererImpl::AppendAudioToFrameBuffer() {
  SB_DCHECK(BelongsToCurrentThread());

  int frames_in_buffer = static_cast<int>(frames_sent_to_sink_.load() -
                                          frames_consumed_by_sink_.load());

  if (kMaxCachedFrames - frames_in_buffer < kFrameAppendUnit) {
    return false;
  }

  int offset_to_append = frames_sent_to_sink_.load() % kMaxCachedFrames;

  // When |playback_rate_| is 0, try to fill the buffer with playback rate as 1.
  // Otherwise the preroll will never finish.
  float playback_rate_to_fill =
      playback_rate_ == 0.0 ? 1.f : static_cast<float>(playback_rate_);
  scoped_refptr<DecodedAudio> decoded_audio =
      time_stretcher_.Read(kFrameAppendUnit, playback_rate_to_fill);
  SB_DCHECK(decoded_audio);
  if (decoded_audio->frames() == 0 && eos_state_.load() == kEOSDecoded) {
    eos_state_.store(kEOSSentToSink);
  }
  audio_frame_tracker_.AddFrames(decoded_audio->frames(),
                                 playback_rate_to_fill);
  // TODO: Support kSbMediaAudioFrameStorageTypePlanar.
  decoded_audio->SwitchFormatTo(sink_sample_type_,
                                kSbMediaAudioFrameStorageTypeInterleaved);
  const uint8_t* source_buffer = decoded_audio->buffer();
  int frames_to_append = decoded_audio->frames();
  int frames_appended = 0;

  if (frames_to_append > kMaxCachedFrames - offset_to_append) {
    SbMemoryCopy(&frame_buffer_[offset_to_append * bytes_per_frame_],
                 source_buffer,
                 (kMaxCachedFrames - offset_to_append) * bytes_per_frame_);
    source_buffer += (kMaxCachedFrames - offset_to_append) * bytes_per_frame_;
    frames_to_append -= kMaxCachedFrames - offset_to_append;
    frames_appended += kMaxCachedFrames - offset_to_append;
    offset_to_append = 0;
  }

  SbMemoryCopy(&frame_buffer_[offset_to_append * bytes_per_frame_],
               source_buffer, frames_to_append * bytes_per_frame_);
  frames_appended += frames_to_append;

  frames_sent_to_sink_.fetch_add(frames_appended);

  int64_t preroll_frames =
      decoder_->GetSamplesPerSecond() * kPrerollTime / kSbTimeSecond;
  if (seeking_.load() && frames_sent_to_sink_.load() > preroll_frames) {
    seeking_.store(false);
  }

  return frames_appended > 0;
}

// static
void AudioRendererImpl::UpdateSourceStatusFunc(int* frames_in_buffer,
                                               int* offset_in_frames,
                                               bool* is_playing,
                                               bool* is_eos_reached,
                                               void* context) {
  AudioRendererImpl* audio_renderer = static_cast<AudioRendererImpl*>(context);
  SB_DCHECK(audio_renderer);
  SB_DCHECK(frames_in_buffer);
  SB_DCHECK(offset_in_frames);
  SB_DCHECK(is_playing);
  SB_DCHECK(is_eos_reached);

  audio_renderer->UpdateSourceStatus(frames_in_buffer, offset_in_frames,
                                     is_playing, is_eos_reached);
}

// static
void AudioRendererImpl::ConsumeFramesFunc(int frames_consumed, void* context) {
  AudioRendererImpl* audio_renderer = static_cast<AudioRendererImpl*>(context);
  SB_DCHECK(audio_renderer);

  audio_renderer->ConsumeFrames(frames_consumed);
}

}  // namespace filter
}  // namespace player
}  // namespace starboard
}  // namespace shared
}  // namespace starboard