Extract common simulcast logic from VP8 wrapper and simulcast adapter

webrtc/modules/video_coding/codecs/vp8/simulcast_encoder_adapter.cc

Index: webrtc/modules/video_coding/codecs/vp8/simulcast_encoder_adapter.cc
diff --git a/webrtc/modules/video_coding/codecs/vp8/simulcast_encoder_adapter.cc b/webrtc/modules/video_coding/codecs/vp8/simulcast_encoder_adapter.cc
index cbd4ec16fda932f84363541babf160af08accd94..f1a629784193aab4ca0c7f2221b7f9e6469b4c4b 100644
--- a/webrtc/modules/video_coding/codecs/vp8/simulcast_encoder_adapter.cc
+++ b/webrtc/modules/video_coding/codecs/vp8/simulcast_encoder_adapter.cc
@@ -26,31 +26,8 @@ const unsigned int kDefaultMaxQp = 56;
 // Max qp for lowest spatial resolution when doing simulcast.
 const unsigned int kLowestResMaxQp = 45;
 
-uint32_t SumStreamTargetBitrate(int streams, const webrtc::VideoCodec& codec) {
-  uint32_t bitrate_sum = 0;
-  for (int i = 0; i < streams; ++i) {
-    bitrate_sum += codec.simulcastStream[i].targetBitrate;
-  }
-  return bitrate_sum;
-}
-
-uint32_t SumStreamMaxBitrate(int streams, const webrtc::VideoCodec& codec) {
-  uint32_t bitrate_sum = 0;
-  for (int i = 0; i < streams; ++i) {
-    bitrate_sum += codec.simulcastStream[i].maxBitrate;
-  }
-  return bitrate_sum;
-}
-
-int NumberOfStreams(const webrtc::VideoCodec& codec) {
-  int streams =
-      codec.numberOfSimulcastStreams < 1 ? 1 : codec.numberOfSimulcastStreams;
-  uint32_t simulcast_max_bitrate = SumStreamMaxBitrate(streams, codec);
-  if (simulcast_max_bitrate == 0) {
-    streams = 1;
-  }
-  return streams;
-}
+static const char* kDefaultImplementationName =
+    "SimulcastEncoderAdapter (uninitialized)";
 
 bool ValidSimulcastResolutions(const webrtc::VideoCodec& codec,
                                int num_streams) {
@@ -92,13 +69,8 @@ int VerifyCodec(const webrtc::VideoCodec* inst) {
   return WEBRTC_VIDEO_CODEC_OK;
 }
 
-// TL1 FrameDropper's max time to drop frames.
-const float kTl1MaxTimeToDropFrames = 20.0f;
-
 struct ScreenshareTemporalLayersFactory : webrtc::TemporalLayersFactory {
-  ScreenshareTemporalLayersFactory()
-      : tl1_frame_dropper_(kTl1MaxTimeToDropFrames) {}
-
+  ScreenshareTemporalLayersFactory() {}
   virtual ~ScreenshareTemporalLayersFactory() {}
 
   virtual webrtc::TemporalLayers* Create(int num_temporal_layers,
@@ -106,9 +78,6 @@ struct ScreenshareTemporalLayersFactory : webrtc::TemporalLayersFactory {
     return new webrtc::ScreenshareLayers(num_temporal_layers, rand(),
                                          webrtc::Clock::GetRealTimeClock());
   }
-
-  mutable webrtc::FrameDropper tl0_frame_dropper_;
-  mutable webrtc::FrameDropper tl1_frame_dropper_;
 };
 
 // An EncodedImageCallback implementation that forwards on calls to a
@@ -133,16 +102,22 @@ class AdapterEncodedImageCallback : public webrtc::EncodedImageCallback {
   const size_t stream_idx_;
 };
 
+webrtc::VideoCodec DefaultCodec() {
+  webrtc::VideoCodec codec;
+  memset(&codec, 0, sizeof(webrtc::VideoCodec));
+  return codec;
+}
+
 }  // namespace
 
 namespace webrtc {
 
 SimulcastEncoderAdapter::SimulcastEncoderAdapter(VideoEncoderFactory* factory)
     : factory_(factory),
-      encoded_complete_callback_(NULL),
-      implementation_name_("SimulcastEncoderAdapter") {
-  memset(&codec_, 0, sizeof(webrtc::VideoCodec));
-}
+      codec_(DefaultCodec()),
+      stream_states_(codec_),
+      encoded_complete_callback_(nullptr),
+      implementation_name_(kDefaultImplementationName) {}
 
 SimulcastEncoderAdapter::~SimulcastEncoderAdapter() {
   Release();
@@ -154,13 +129,11 @@ int SimulcastEncoderAdapter::Release() {
   // resolutions doesn't require reallocation of the first encoder, but only
   // reinitialization, which makes sense. Then Destroy this instance instead in
   // ~SimulcastEncoderAdapter().
-  while (!streaminfos_.empty()) {
-    VideoEncoder* encoder = streaminfos_.back().encoder;
-    EncodedImageCallback* callback = streaminfos_.back().callback;
-    factory_->Destroy(encoder);
-    delete callback;
-    streaminfos_.pop_back();
-  }
+  stream_states_ = SimulcastState(codec_);
+  for (auto& it : encoders_)
+    factory_->Destroy(it.second.first);
+  encoders_.clear();
+
   return WEBRTC_VIDEO_CODEC_OK;
 }
 
@@ -181,7 +154,9 @@ int SimulcastEncoderAdapter::InitEncode(const VideoCodec* inst,
     return ret;
   }
 
-  int number_of_streams = NumberOfStreams(*inst);
+  std::unique_ptr<SimulcastState> simulcast(new SimulcastState(*inst));
+
+  int number_of_streams = simulcast->NumStreams();
   const bool doing_simulcast = (number_of_streams > 1);
 
   if (doing_simulcast && !ValidSimulcastResolutions(*inst, number_of_streams)) {
@@ -198,17 +173,27 @@ int SimulcastEncoderAdapter::InitEncode(const VideoCodec* inst,
 
   std::string implementation_name;
   // Create |number_of_streams| of encoder instances and init them.
-  for (int i = 0; i < number_of_streams; ++i) {
+  stream_states_ = *simulcast;
+  for (const SimulcastState::Stream& stream : stream_states_.Streams()) {
     VideoCodec stream_codec;
-    bool send_stream = true;
     if (!doing_simulcast) {
       stream_codec = codec_;
       stream_codec.numberOfSimulcastStreams = 1;
     } else {
-      bool highest_resolution_stream = (i == (number_of_streams - 1));
-      PopulateStreamCodec(&codec_, i, number_of_streams,
-                          highest_resolution_stream, &stream_codec,
-                          &send_stream);
+      stream_codec = stream.AsCodec();
+      if (stream.idx == 0) {
+        // Settings for lowest spatial resolutions.
+        stream_codec.qpMax = kLowestResMaxQp;
+      }
+      if (stream.idx < codec_.numberOfSimulcastStreams - 1) {
+        // For resolutions below CIF, set the codec |complexity| parameter to
+        // kComplexityHigher, which maps to cpu_used = -4.
+        if (stream_codec.width * stream_codec.height < 352 * 288) {
+          stream_codec.codecSpecific.VP8.complexity = webrtc::kComplexityHigher;
+        }
+        // Turn off denoising for all streams but the highest resolution.
+        stream_codec.codecSpecific.VP8.denoisingOn = false;
+      }
     }
 
     // TODO(ronghuawu): Remove once this is handled in VP8EncoderImpl.
@@ -222,14 +207,16 @@ int SimulcastEncoderAdapter::InitEncode(const VideoCodec* inst,
       Release();
       return ret;
     }
-    EncodedImageCallback* callback = new AdapterEncodedImageCallback(this, i);
+    EncodedImageCallback* callback =
+        new AdapterEncodedImageCallback(this, stream.idx);
     encoder->RegisterEncodeCompleteCallback(callback);
-    streaminfos_.push_back(StreamInfo(encoder, callback, stream_codec.width,
-                                      stream_codec.height, send_stream));
-    if (i != 0)
+    encoders_[stream.idx] = EncoderContext(
+        encoder, std::unique_ptr<EncodedImageCallback>(callback));
+    if (stream.idx != 0)
       implementation_name += ", ";
-    implementation_name += streaminfos_[i].encoder->ImplementationName();
+    implementation_name += encoder->ImplementationName();
   }
+
   if (doing_simulcast) {
     implementation_name_ =
         "SimulcastEncoderAdapter (" + implementation_name + ")";
@@ -250,51 +237,39 @@ int SimulcastEncoderAdapter::Encode(
     return WEBRTC_VIDEO_CODEC_UNINITIALIZED;
   }
 
-  // All active streams should generate a key frame if
-  // a key frame is requested by any stream.
-  bool send_key_frame = false;
-  if (frame_types) {
-    for (size_t i = 0; i < frame_types->size(); ++i) {
-      if (frame_types->at(i) == kVideoFrameKey) {
-        send_key_frame = true;
-        break;
-      }
-    }
-  }
-  for (size_t stream_idx = 0; stream_idx < streaminfos_.size(); ++stream_idx) {
-    if (streaminfos_[stream_idx].key_frame_request &&
-        streaminfos_[stream_idx].send_stream) {
-      send_key_frame = true;
-      break;
-    }
-  }
+  // TODO(sprang): Do we really need to force key frames on all streams?
+  // All active streams should generate a key frame if a key frame is requested
+  // by any stream.
+  bool force_key_frame = frame_types &&
+                         std::find(frame_types->begin(), frame_types->end(),
+                                   kVideoFrameKey) != frame_types->end();
+  int active_streams = stream_states_.NumSendingStreams();
+  for (int i = 0; i < active_streams; ++i)
+    force_key_frame |= stream_states_.GetAndResetKeyFrameRequest(i);
 
   int src_width = input_image.width();
   int src_height = input_image.height();
-  for (size_t stream_idx = 0; stream_idx < streaminfos_.size(); ++stream_idx) {
-    // Don't encode frames in resolutions that we don't intend to send.
-    if (!streaminfos_[stream_idx].send_stream)
+  for (const SimulcastState::Stream& stream : stream_states_.Streams()) {
+    if (!stream.sending)
       continue;
 
     std::vector<FrameType> stream_frame_types;
-    if (send_key_frame) {
+    if (force_key_frame) {
       stream_frame_types.push_back(kVideoFrameKey);
-      streaminfos_[stream_idx].key_frame_request = false;
     } else {
       stream_frame_types.push_back(kVideoFrameDelta);
     }
 
-    int dst_width = streaminfos_[stream_idx].width;
-    int dst_height = streaminfos_[stream_idx].height;
-    // If scaling isn't required, because the input resolution
-    // matches the destination or the input image is empty (e.g.
-    // a keyframe request for encoders with internal camera
-    // sources), pass the image on directly. Otherwise, we'll
-    // scale it to match what the encoder expects (below).
+    int dst_width = stream.config.width;
+    int dst_height = stream.config.height;
+    // If scaling isn't required, because the input resolution matches the
+    // destination or the input image is empty (e.g. a keyframe request for
+    // encoders with internal camera sources), pass the image on directly.
+    // Otherwise, we'll scale it to match what the encoder expects (below).
     if ((dst_width == src_width && dst_height == src_height) ||
         input_image.IsZeroSize()) {
-      streaminfos_[stream_idx].encoder->Encode(input_image, codec_specific_info,
-                                               &stream_frame_types);
+      encoders_[stream.idx].first->Encode(input_image, codec_specific_info,
+                                          &stream_frame_types);
     } else {
       VideoFrame dst_frame;
       // Making sure that destination frame is of sufficient size.
@@ -311,8 +286,8 @@ int SimulcastEncoderAdapter::Encode(
           dst_height, libyuv::kFilterBilinear);
       dst_frame.set_timestamp(input_image.timestamp());
       dst_frame.set_render_time_ms(input_image.render_time_ms());
-      streaminfos_[stream_idx].encoder->Encode(dst_frame, codec_specific_info,
-                                               &stream_frame_types);
+      encoders_[stream.idx].first->Encode(dst_frame, codec_specific_info,
+                                          &stream_frame_types);
     }
   }
 
@@ -327,59 +302,35 @@ int SimulcastEncoderAdapter::RegisterEncodeCompleteCallback(
 
 int SimulcastEncoderAdapter::SetChannelParameters(uint32_t packet_loss,
                                                   int64_t rtt) {
-  for (size_t stream_idx = 0; stream_idx < streaminfos_.size(); ++stream_idx) {
-    streaminfos_[stream_idx].encoder->SetChannelParameters(packet_loss, rtt);
-  }
+  for (const auto& it : encoders_)
+    it.second.first->SetChannelParameters(packet_loss, rtt);
   return WEBRTC_VIDEO_CODEC_OK;
 }
 
 int SimulcastEncoderAdapter::SetRates(uint32_t new_bitrate_kbit,
                                       uint32_t new_framerate) {
-  if (!Initialized()) {
+  if (!Initialized())
     return WEBRTC_VIDEO_CODEC_UNINITIALIZED;
-  }
-  if (new_framerate < 1) {
+
+  if (new_framerate < 1)
     return WEBRTC_VIDEO_CODEC_ERR_PARAMETER;
-  }
-  if (codec_.maxBitrate > 0 && new_bitrate_kbit > codec_.maxBitrate) {
+
+  if (codec_.maxBitrate > 0 && new_bitrate_kbit > codec_.maxBitrate)
     new_bitrate_kbit = codec_.maxBitrate;
-  }
-  if (new_bitrate_kbit < codec_.minBitrate) {
+
+  if (new_bitrate_kbit < codec_.minBitrate)
     new_bitrate_kbit = codec_.minBitrate;
-  }
+
   if (codec_.numberOfSimulcastStreams > 0 &&
       new_bitrate_kbit < codec_.simulcastStream[0].minBitrate) {
     new_bitrate_kbit = codec_.simulcastStream[0].minBitrate;
   }
-  codec_.maxFramerate = new_framerate;
-
-  bool send_stream = true;
-  uint32_t stream_bitrate = 0;
-  for (size_t stream_idx = 0; stream_idx < streaminfos_.size(); ++stream_idx) {
-    stream_bitrate = GetStreamBitrate(stream_idx, streaminfos_.size(),
-                                      new_bitrate_kbit, &send_stream);
-    // Need a key frame if we have not sent this stream before.
-    if (send_stream && !streaminfos_[stream_idx].send_stream) {
-      streaminfos_[stream_idx].key_frame_request = true;
-    }
-    streaminfos_[stream_idx].send_stream = send_stream;
-
-    // TODO(holmer): This is a temporary hack for screensharing, where we
-    // interpret the startBitrate as the encoder target bitrate. This is
-    // to allow for a different max bitrate, so if the codec can't meet
-    // the target we still allow it to overshoot up to the max before dropping
-    // frames. This hack should be improved.
-    if (codec_.targetBitrate > 0 &&
-        (codec_.codecSpecific.VP8.numberOfTemporalLayers == 2 ||
-         codec_.simulcastStream[0].numberOfTemporalLayers == 2)) {
-      stream_bitrate = std::min(codec_.maxBitrate, stream_bitrate);
-      // TODO(ronghuawu): Can't change max bitrate via the VideoEncoder
-      // interface. And VP8EncoderImpl doesn't take negative framerate.
-      // max_bitrate = std::min(codec_.maxBitrate, stream_bitrate);
-      // new_framerate = -1;
-    }
 
-    streaminfos_[stream_idx].encoder->SetRates(stream_bitrate, new_framerate);
+  codec_.maxFramerate = new_framerate;
+  stream_states_.AllocateBitrate(new_bitrate_kbit * 1000);
+  for (const SimulcastState::Stream& stream : stream_states_.Streams()) {
+    encoders_[stream.idx].first->SetRates(stream.allocated_rate_bps / 1000,
+                                          new_framerate);
   }
 
   return WEBRTC_VIDEO_CODEC_OK;
@@ -398,109 +349,22 @@ int32_t SimulcastEncoderAdapter::Encoded(
       encodedImage, &stream_codec_specific, fragmentation);
 }
 
-uint32_t SimulcastEncoderAdapter::GetStreamBitrate(
-    int stream_idx,
-    size_t total_number_of_streams,
-    uint32_t new_bitrate_kbit,
-    bool* send_stream) const {
-  if (total_number_of_streams == 1) {
-    *send_stream = true;
-    return new_bitrate_kbit;
-  }
-
-  // The bitrate needed to start sending this stream is given by the
-  // minimum bitrate allowed for encoding this stream, plus the sum target
-  // rates of all lower streams.
-  uint32_t sum_target_lower_streams =
-      SumStreamTargetBitrate(stream_idx, codec_);
-  uint32_t bitrate_to_send_this_layer =
-      codec_.simulcastStream[stream_idx].minBitrate + sum_target_lower_streams;
-  if (new_bitrate_kbit >= bitrate_to_send_this_layer) {
-    // We have enough bandwidth to send this stream.
-    *send_stream = true;
-    // Bitrate for this stream is the new bitrate (|new_bitrate_kbit|) minus the
-    // sum target rates of the lower streams, and capped to a maximum bitrate.
-    // The maximum cap depends on whether we send the next higher stream.
-    // If we will be sending the next higher stream, |max_rate| is given by
-    // current stream's |targetBitrate|, otherwise it's capped by |maxBitrate|.
-    if (stream_idx < codec_.numberOfSimulcastStreams - 1) {
-      unsigned int max_rate = codec_.simulcastStream[stream_idx].maxBitrate;
-      if (new_bitrate_kbit >=
-          SumStreamTargetBitrate(stream_idx + 1, codec_) +
-              codec_.simulcastStream[stream_idx + 1].minBitrate) {
-        max_rate = codec_.simulcastStream[stream_idx].targetBitrate;
-      }
-      return std::min(new_bitrate_kbit - sum_target_lower_streams, max_rate);
-    } else {
-      // For the highest stream (highest resolution), the |targetBitRate| and
-      // |maxBitrate| are not used. Any excess bitrate (above the targets of
-      // all lower streams) is given to this (highest resolution) stream.
-      return new_bitrate_kbit - sum_target_lower_streams;
-    }
-  } else {
-    // Not enough bitrate for this stream.
-    // Return our max bitrate of |stream_idx| - 1, but we don't send it. We need
-    // to keep this resolution coding in order for the multi-encoder to work.
-    *send_stream = false;
-    return codec_.simulcastStream[stream_idx - 1].maxBitrate;
-  }
-}
-
-void SimulcastEncoderAdapter::PopulateStreamCodec(
-    const webrtc::VideoCodec* inst,
-    int stream_index,
-    size_t total_number_of_streams,
-    bool highest_resolution_stream,
-    webrtc::VideoCodec* stream_codec,
-    bool* send_stream) {
-  *stream_codec = *inst;
-
-  // Stream specific settings.
-  stream_codec->codecSpecific.VP8.numberOfTemporalLayers =
-      inst->simulcastStream[stream_index].numberOfTemporalLayers;
-  stream_codec->numberOfSimulcastStreams = 0;
-  stream_codec->width = inst->simulcastStream[stream_index].width;
-  stream_codec->height = inst->simulcastStream[stream_index].height;
-  stream_codec->maxBitrate = inst->simulcastStream[stream_index].maxBitrate;
-  stream_codec->minBitrate = inst->simulcastStream[stream_index].minBitrate;
-  stream_codec->qpMax = inst->simulcastStream[stream_index].qpMax;
-  // Settings that are based on stream/resolution.
-  if (stream_index == 0) {
-    // Settings for lowest spatial resolutions.
-    stream_codec->qpMax = kLowestResMaxQp;
-  }
-  if (!highest_resolution_stream) {
-    // For resolutions below CIF, set the codec |complexity| parameter to
-    // kComplexityHigher, which maps to cpu_used = -4.
-    int pixels_per_frame = stream_codec->width * stream_codec->height;
-    if (pixels_per_frame < 352 * 288) {
-      stream_codec->codecSpecific.VP8.complexity = webrtc::kComplexityHigher;
-    }
-    // Turn off denoising for all streams but the highest resolution.
-    stream_codec->codecSpecific.VP8.denoisingOn = false;
-  }
-  // TODO(ronghuawu): what to do with targetBitrate.
-
-  int stream_bitrate = GetStreamBitrate(stream_index, total_number_of_streams,
-                                        inst->startBitrate, send_stream);
-  stream_codec->startBitrate = stream_bitrate;
-}
-
 bool SimulcastEncoderAdapter::Initialized() const {
-  return !streaminfos_.empty();
+  return !encoders_.empty();
 }
 
 void SimulcastEncoderAdapter::OnDroppedFrame() {
-  streaminfos_[0].encoder->OnDroppedFrame();
+  for (const auto& it : encoders_)
+    it.second.first->OnDroppedFrame();
 }
 
 bool SimulcastEncoderAdapter::SupportsNativeHandle() const {
   // We should not be calling this method before streaminfos_ are configured.
-  RTC_DCHECK(!streaminfos_.empty());
+  RTC_DCHECK(Initialized());
   // TODO(pbos): Support textures when using more than one encoder.
-  if (streaminfos_.size() != 1)
+  if (encoders_.size() != 1)
     return false;
-  return streaminfos_[0].encoder->SupportsNativeHandle();
+  return encoders_.begin()->second.first->SupportsNativeHandle();
 }
 
 const char* SimulcastEncoderAdapter::ImplementationName() const {