[H264][Simulcast] Implement the simulcast logic for h264 encoder

webrtc/modules/video_coding/codecs/h264/h264_encoder_impl.cc

Index: webrtc/modules/video_coding/codecs/h264/h264_encoder_impl.cc
diff --git a/webrtc/modules/video_coding/codecs/h264/h264_encoder_impl.cc b/webrtc/modules/video_coding/codecs/h264/h264_encoder_impl.cc
index 18eccb25a5d4c1950935b54a060b03ce8006250b..6da71c9bae757a46807b3ef4a5f18775f4dbeb30 100644
--- a/webrtc/modules/video_coding/codecs/h264/h264_encoder_impl.cc
+++ b/webrtc/modules/video_coding/codecs/h264/h264_encoder_impl.cc
@@ -11,8 +11,10 @@
 
 #include "webrtc/modules/video_coding/codecs/h264/h264_encoder_impl.h"
 
+#include <algorithm>
 #include <limits>
 
+#include "webrtc/common_video/libyuv/include/scaler.h"
 #include "third_party/openh264/src/codec/api/svc/codec_api.h"
 #include "third_party/openh264/src/codec/api/svc/codec_app_def.h"
 #include "third_party/openh264/src/codec/api/svc/codec_def.h"
@@ -21,6 +23,7 @@
 #include "webrtc/base/logging.h"
 #include "webrtc/common_video/libyuv/include/webrtc_libyuv.h"
 #include "webrtc/system_wrappers/include/metrics.h"
+#include "webrtc/video_frame.h"
 
 namespace webrtc {
 
@@ -50,6 +53,78 @@ int NumberOfThreads(int width, int height, int number_of_cores) {
   return 1;
 }
 
+std::vector<int> GetStreamBitratesKbps(const VideoCodec& codec,
+                                       int bitrate_to_allocate_kbps) {
+  if (codec.numberOfSimulcastStreams <= 1) {
+    return std::vector<int>(1, bitrate_to_allocate_kbps);
+  }
+
+  std::vector<int> bitrates_kbps(codec.numberOfSimulcastStreams);
+  // Allocate min -> target bitrates as long as we have bitrate to spend.
+  size_t last_active_stream = 0;
+  for (size_t i = 0; i < static_cast<size_t>(codec.numberOfSimulcastStreams) &&
+                     bitrate_to_allocate_kbps >=
+                         static_cast<int>(codec.simulcastStream[i].minBitrate);
+       ++i) {
+    last_active_stream = i;
+    int allocated_bitrate_kbps =
+        std::min(static_cast<int>(codec.simulcastStream[i].targetBitrate),
+                 bitrate_to_allocate_kbps);
+    bitrates_kbps[i] = allocated_bitrate_kbps;
+    bitrate_to_allocate_kbps -= allocated_bitrate_kbps;
+  }
+
+  // Spend additional bits on the highest-quality active layer, up to max
+  // bitrate.
+  // TODO(pbos): Consider spending additional bits on last_active_stream-1 down
+  // to 0 and not just the top layer when we have additional bitrate to spend.
+  int allocated_bitrate_kbps = std::min(
+      static_cast<int>(codec.simulcastStream[last_active_stream].maxBitrate -
+                       bitrates_kbps[last_active_stream]),
+      bitrate_to_allocate_kbps);
+  bitrates_kbps[last_active_stream] += allocated_bitrate_kbps;
+  bitrate_to_allocate_kbps -= allocated_bitrate_kbps;
+
+  // Make sure we can always send something. Suspending below min bitrate is
+  // controlled outside the codec implementation and is not overriden by this.
+  if (bitrates_kbps[0] < static_cast<int>(codec.simulcastStream[0].minBitrate))
+    bitrates_kbps[0] = static_cast<int>(codec.simulcastStream[0].minBitrate);
+
+  return bitrates_kbps;
+}
+
+uint32_t SumStreamMaxBitrate(int streams, const 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 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;
+}
+
+bool ValidSimulcastResolutions(const VideoCodec& codec, int num_streams) {
+  if (codec.width != codec.simulcastStream[num_streams - 1].width ||
+      codec.height != codec.simulcastStream[num_streams - 1].height) {
+    return false;
+  }
+  for (int i = 0; i < num_streams; ++i) {
+    if (codec.width * codec.simulcastStream[i].height !=
+        codec.height * codec.simulcastStream[i].width) {
+      return false;
+    }
+  }
+  return true;
+}
+
 }  // namespace
 
 static FrameType EVideoFrameType_to_FrameType(EVideoFrameType type) {
@@ -148,16 +223,29 @@ static void RtpFragmentize(EncodedImage* encoded_image,
 }
 
 H264EncoderImpl::H264EncoderImpl()
-    : openh264_encoder_(nullptr),
-      encoded_image_callback_(nullptr),
+    : encoded_image_callback_(nullptr),
       has_reported_init_(false),
-      has_reported_error_(false) {
+      has_reported_error_(false),
+      key_frame_request_(kMaxSimulcastStreams, false) {
+  encoded_images_.reserve(kMaxSimulcastStreams);
+  encoded_image_buffers_.reserve(kMaxSimulcastStreams);
+  send_streams_.reserve(kMaxSimulcastStreams);
+  encoders_.reserve(kMaxSimulcastStreams);
+  scaled_input_frames_.reserve(kMaxSimulcastStreams);
 }
 
 H264EncoderImpl::~H264EncoderImpl() {
   Release();
 }
 
+void H264EncoderImpl::SetStreamState(bool send_stream, int stream_idx) {
+  if (send_stream && !send_streams_[stream_idx]) {
+    // Need a key frame if we have not sent this stream before.
+    key_frame_request_[stream_idx] = true;
+  }
+  send_streams_[stream_idx] = send_stream;
+}
+
 int32_t H264EncoderImpl::InitEncode(const VideoCodec* codec_settings,
                                     int32_t number_of_cores,
                                     size_t /*max_payload_size*/) {
@@ -181,23 +269,27 @@ int32_t H264EncoderImpl::InitEncode(const VideoCodec* codec_settings,
     ReportError();
     return release_ret;
   }
-  RTC_DCHECK(!openh264_encoder_);
 
-  // Create encoder.
-  if (WelsCreateSVCEncoder(&openh264_encoder_) != 0) {
-    // Failed to create encoder.
-    LOG(LS_ERROR) << "Failed to create OpenH264 encoder";
-    RTC_DCHECK(!openh264_encoder_);
-    ReportError();
-    return WEBRTC_VIDEO_CODEC_ERROR;
+  int number_of_streams = NumberOfStreams(*codec_settings);
+  bool doing_simulcast = (number_of_streams > 1);
+
+  if (doing_simulcast &&
+      !ValidSimulcastResolutions(*codec_settings, number_of_streams)) {
+    return WEBRTC_VIDEO_CODEC_ERR_PARAMETER;
   }
-  RTC_DCHECK(openh264_encoder_);
-  if (kOpenH264EncoderDetailedLogging) {
-    int trace_level = WELS_LOG_DETAIL;
-    openh264_encoder_->SetOption(ENCODER_OPTION_TRACE_LEVEL,
-                                 &trace_level);
+  // Code expects simulcastStream resolutions to be correct, make sure they are
+  // filled even when there are no simulcast layers.
+  if (codec_settings->numberOfSimulcastStreams == 0) {
+    codec_settings_.simulcastStream[0].width = codec_settings->width;
+    codec_settings_.simulcastStream[0].height = codec_settings->height;
   }
-  // else WELS_LOG_DEFAULT is used by default.
+
+  encoded_images_.resize(number_of_streams);
+  encoded_image_buffers_.resize(number_of_streams);
+  encoders_.resize(number_of_streams);
+  scaled_input_frames_.resize(number_of_streams);
+  key_frame_request_.resize(number_of_streams);
+  std::fill(key_frame_request_.begin(), key_frame_request_.end(), false);
 
   codec_settings_ = *codec_settings;
   if (codec_settings_.targetBitrate == 0)
@@ -208,86 +300,127 @@ int32_t H264EncoderImpl::InitEncode(const VideoCodec* codec_settings,
   // memset(&p, 0, sizeof(SEncParamBase)) used in Initialize, and SEncParamExt
   // which is a superset of SEncParamBase (cleared with GetDefaultParams) used
   // in InitializeExt.
-  SEncParamExt init_params;
-  openh264_encoder_->GetDefaultParams(&init_params);
-  if (codec_settings_.mode == kRealtimeVideo) {
-    init_params.iUsageType = CAMERA_VIDEO_REAL_TIME;
-  } else if (codec_settings_.mode == kScreensharing) {
-    init_params.iUsageType = SCREEN_CONTENT_REAL_TIME;
-  } else {
-    ReportError();
-    return WEBRTC_VIDEO_CODEC_ERR_PARAMETER;
-  }
-  init_params.iPicWidth = codec_settings_.width;
-  init_params.iPicHeight = codec_settings_.height;
-  // |init_params| uses bit/s, |codec_settings_| uses kbit/s.
-  init_params.iTargetBitrate = codec_settings_.targetBitrate * 1000;
-  init_params.iMaxBitrate = codec_settings_.maxBitrate * 1000;
-  // Rate Control mode
-  init_params.iRCMode = RC_BITRATE_MODE;
-  init_params.fMaxFrameRate = static_cast<float>(codec_settings_.maxFramerate);
-
-  // The following parameters are extension parameters (they're in SEncParamExt,
-  // not in SEncParamBase).
-  init_params.bEnableFrameSkip =
-      codec_settings_.codecSpecific.H264.frameDroppingOn;
-  // |uiIntraPeriod|    - multiple of GOP size
-  // |keyFrameInterval| - number of frames
-  init_params.uiIntraPeriod =
-      codec_settings_.codecSpecific.H264.keyFrameInterval;
-  init_params.uiMaxNalSize = 0;
-  // Threading model: use auto.
-  //  0: auto (dynamic imp. internal encoder)
-  //  1: single thread (default value)
-  // >1: number of threads
-  init_params.iMultipleThreadIdc = NumberOfThreads(init_params.iPicWidth,
-                                                   init_params.iPicHeight,
-                                                   number_of_cores);
-  // The base spatial layer 0 is the only one we use.
-  init_params.sSpatialLayers[0].iVideoWidth        = init_params.iPicWidth;
-  init_params.sSpatialLayers[0].iVideoHeight       = init_params.iPicHeight;
-  init_params.sSpatialLayers[0].fFrameRate         = init_params.fMaxFrameRate;
-  init_params.sSpatialLayers[0].iSpatialBitrate    = init_params.iTargetBitrate;
-  init_params.sSpatialLayers[0].iMaxSpatialBitrate = init_params.iMaxBitrate;
-  // Slice num according to number of threads.
-  init_params.sSpatialLayers[0].sSliceCfg.uiSliceMode = SM_AUTO_SLICE;
-
-  // Initialize.
-  if (openh264_encoder_->InitializeExt(&init_params) != 0) {
-    LOG(LS_ERROR) << "Failed to initialize OpenH264 encoder";
-    Release();
-    ReportError();
-    return WEBRTC_VIDEO_CODEC_ERROR;
+
+  for (int i = 0; i < number_of_streams; ++i) {
+    // Create encoder.
+    if (WelsCreateSVCEncoder(&encoders_[i]) != 0) {
+      // Failed to create encoder.
+      LOG(LS_ERROR) << "Failed to create OpenH264 encoder";
+      RTC_DCHECK(!encoders_[i]);
+      ReportError();
+      return WEBRTC_VIDEO_CODEC_ERROR;
+    }
+
+    RTC_DCHECK(encoders_[i]);
+    if (kOpenH264EncoderDetailedLogging) {
+      int trace_level = WELS_LOG_DETAIL;
+      encoders_[i]->SetOption(ENCODER_OPTION_TRACE_LEVEL, &trace_level);
+    }
+    SEncParamExt init_params;
+    memset(&init_params, 0, sizeof(SEncParamExt));
+    encoders_[i]->GetDefaultParams(&init_params);
+    if (codec_settings_.mode == kRealtimeVideo) {
+      init_params.iUsageType = CAMERA_VIDEO_REAL_TIME;
+    } else if (codec_settings_.mode == kScreensharing) {
+      init_params.iUsageType = SCREEN_CONTENT_REAL_TIME;
+    } else {
+      return WEBRTC_VIDEO_CODEC_ERR_PARAMETER;
+    }
+    init_params.iPicWidth = codec_settings_.simulcastStream[i].width;
+    init_params.iPicHeight = codec_settings_.simulcastStream[i].height;
+    // |init_params| uses bit/s, |codec_settings_| uses kbit/s.
+    init_params.iTargetBitrate = codec_settings_.startBitrate * 1000;
+    init_params.iMaxBitrate = codec_settings_.maxBitrate * 1000;
+    // Rate Control mode
+    init_params.iRCMode = RC_BITRATE_MODE;
+    init_params.fMaxFrameRate =
+        static_cast<float>(codec_settings_.maxFramerate);
+
+    // The following parameters are extension parameters
+    // (they're in SEncParamExt, not in SEncParamBase).
+    init_params.bEnableFrameSkip =
+        codec_settings_.codecSpecific.H264.frameDroppingOn;
+    // |uiIntraPeriod|    - multiple of GOP size
+    // |keyFrameInterval| - number of frames
+    init_params.uiIntraPeriod =
+        codec_settings_.codecSpecific.H264.keyFrameInterval;
+    init_params.uiMaxNalSize = 0;
+    init_params.iComplexityMode = ECOMPLEXITY_MODE::LOW_COMPLEXITY;
+
+    // Threading model: use auto.
+    //  0: auto (dynamic imp. internal encoder)
+    //  1: single thread (default value)
+    // >1: number of threads
+    init_params.iMultipleThreadIdc = NumberOfThreads(
+        init_params.iPicWidth, init_params.iPicHeight, number_of_cores);
+    // The base spatial layer 0 is the only one we use.
+    init_params.sSpatialLayers[0].iVideoWidth = init_params.iPicWidth;
+    init_params.sSpatialLayers[0].iVideoHeight = init_params.iPicHeight;
+    init_params.sSpatialLayers[0].fFrameRate = init_params.fMaxFrameRate;
+    init_params.sSpatialLayers[0].iSpatialBitrate = init_params.iTargetBitrate;
+    init_params.sSpatialLayers[0].iMaxSpatialBitrate = init_params.iMaxBitrate;
+
+    // Slice num according to number of threads.
+    init_params.sSpatialLayers[0].sSliceCfg.uiSliceMode = SM_AUTO_SLICE;
+    // Initialize.
+    if (encoders_[i]->InitializeExt(&init_params) != 0) {
+      LOG(LS_ERROR) << "Failed to initialize OpenH264 encoder";
+      Release();
+      ReportError();
+      return WEBRTC_VIDEO_CODEC_ERROR;
+    }
+    int video_format = EVideoFormatType::videoFormatI420;
+    encoders_[i]->SetOption(ENCODER_OPTION_DATAFORMAT, &video_format);
+    // Initialize encoded image. Default buffer size: size of unencoded data.
+    // allocate memory for encoded image
+    if (encoded_images_[i]._buffer != NULL) {
+      delete[] encoded_images_[i]._buffer;
+    }
+    encoded_images_[i]._size =
+        CalcBufferSize(kI420, codec_settings->simulcastStream[i].width,
+            codec_settings->simulcastStream[i].height);
+    encoded_images_[i]._buffer = new uint8_t[encoded_images_[i]._size];
+    encoded_image_buffers_[i].reset(encoded_images_[i]._buffer);
+    encoded_images_[i]._completeFrame = true;
+    encoded_images_[i]._encodedWidth = 0;
+    encoded_images_[i]._encodedHeight = 0;
+    encoded_images_[i]._length = 0;
+
+    // Initialize scaled input frames.
+    scaled_input_frames_[i] = *new VideoFrame();
+    scaled_input_frames_[i].CreateEmptyFrame(
+        codec_settings->simulcastStream[i].width,
+        codec_settings->simulcastStream[i].height,
+        CalculateYStrideSize(codec_settings->simulcastStream[i].width,
+                             codec_settings->simulcastStream[i].height),
+        CalculateUVStrideSize(codec_settings->simulcastStream[i].width,
+                              codec_settings->simulcastStream[i].height),
+        CalculateUVStrideSize(codec_settings->simulcastStream[i].width,
+                              codec_settings->simulcastStream[i].height));
   }
-  int video_format = EVideoFormatType::videoFormatI420;
-  openh264_encoder_->SetOption(ENCODER_OPTION_DATAFORMAT,
-                               &video_format);
-
-  // Initialize encoded image. Default buffer size: size of unencoded data.
-  encoded_image_._size = CalcBufferSize(
-      kI420, codec_settings_.width, codec_settings_.height);
-  encoded_image_._buffer = new uint8_t[encoded_image_._size];
-  encoded_image_buffer_.reset(encoded_image_._buffer);
-  encoded_image_._completeFrame = true;
-  encoded_image_._encodedWidth = 0;
-  encoded_image_._encodedHeight = 0;
-  encoded_image_._length = 0;
   return WEBRTC_VIDEO_CODEC_OK;
 }
 
 int32_t H264EncoderImpl::Release() {
-  if (openh264_encoder_) {
-    int uninit_ret = openh264_encoder_->Uninitialize();
-    if (uninit_ret != 0) {
-      LOG(LS_WARNING) << "OpenH264 encoder's Uninitialize() returned "
-                      << "unsuccessful: " << uninit_ret;
+  while (!encoders_.empty()) {
+    ISVCEncoder* openh264_encoder = encoders_.back();
+    if (openh264_encoder) {
+      int uninit_ret = openh264_encoder->Uninitialize();
+      if (uninit_ret != 0) {
+        LOG(LS_WARNING) << "OpenH264 encoder's Uninitialize() returned "
+                        << "unsuccessful: " << uninit_ret;
+      }
+      WelsDestroySVCEncoder(openh264_encoder);
+      openh264_encoder = nullptr;
+      encoders_.pop_back();
+      EncodedImage encoded_image = encoded_images_.back();
+      if (encoded_image._buffer != nullptr) {
+        encoded_image._buffer = nullptr;
+        encoded_image_buffers_.back().reset();
+      }
+      encoded_images_.pop_back();
+      encoded_image_buffers_.pop_back();
     }
-    WelsDestroySVCEncoder(openh264_encoder_);
-    openh264_encoder_ = nullptr;
-  }
-  if (encoded_image_._buffer != nullptr) {
-    encoded_image_._buffer = nullptr;
-    encoded_image_buffer_.reset();
   }
   return WEBRTC_VIDEO_CODEC_OK;
 }
@@ -302,21 +435,45 @@ int32_t H264EncoderImpl::SetRates(uint32_t bitrate, uint32_t framerate) {
   if (bitrate <= 0 || framerate <= 0) {
     return WEBRTC_VIDEO_CODEC_ERR_PARAMETER;
   }
+  if (codec_settings_.maxBitrate > 0 && bitrate > codec_settings_.maxBitrate) {
+    bitrate = codec_settings_.maxBitrate;
+  }
+  if (bitrate < codec_settings_.minBitrate) {
+    bitrate = codec_settings_.minBitrate;
+  }
+  if (codec_settings_.numberOfSimulcastStreams > 0 &&
+      bitrate < codec_settings_.simulcastStream[0].minBitrate) {
+    bitrate = codec_settings_.simulcastStream[0].minBitrate;
+  }
   codec_settings_.targetBitrate = bitrate;
   codec_settings_.maxFramerate = framerate;
 
-  SBitrateInfo target_bitrate;
-  memset(&target_bitrate, 0, sizeof(SBitrateInfo));
-  target_bitrate.iLayer = SPATIAL_LAYER_ALL,
-  target_bitrate.iBitrate = codec_settings_.targetBitrate * 1000;
-  openh264_encoder_->SetOption(ENCODER_OPTION_BITRATE,
-                               &target_bitrate);
-  float max_framerate = static_cast<float>(codec_settings_.maxFramerate);
-  openh264_encoder_->SetOption(ENCODER_OPTION_FRAME_RATE,
-                               &max_framerate);
+  std::vector<int> stream_bitrates =
+      GetStreamBitratesKbps(codec_settings_, bitrate);
+  for (size_t i = 0; i < encoders_.size(); ++i) {
+    SetStreamState(stream_bitrates[i] > 0, i);
+    if (send_streams_[i]) {
+      SBitrateInfo target_bitrate;
+      memset(&target_bitrate, 0, sizeof(SBitrateInfo));
+      target_bitrate.iLayer = SPATIAL_LAYER_ALL,
+      target_bitrate.iBitrate = stream_bitrates[i] * 1000;  // bps
+      encoders_[i]->SetOption(ENCODER_OPTION_BITRATE, &target_bitrate);
+      float max_framerate = static_cast<float>(framerate);
+      encoders_[i]->SetOption(ENCODER_OPTION_FRAME_RATE, &max_framerate);
+    }
+  }
   return WEBRTC_VIDEO_CODEC_OK;
 }
 
+void H264EncoderImpl::Scale(const VideoFrame& input_frame,
+                            VideoFrame* output_frame) {
+  Scaler scaler;
+  scaler.Set(input_frame.width(), input_frame.height(), output_frame->width(),
+             output_frame->height(), webrtc::kI420, webrtc::kI420,
+             webrtc::kScaleBilinear);
+  scaler.Scale(input_frame, output_frame);
+}
+
 int32_t H264EncoderImpl::Encode(
     const VideoFrame& frame, const CodecSpecificInfo* codec_specific_info,
     const std::vector<FrameType>* frame_types) {
@@ -334,88 +491,111 @@ int32_t H264EncoderImpl::Encode(
     ReportError();
     return WEBRTC_VIDEO_CODEC_UNINITIALIZED;
   }
-  if (frame.width()  != codec_settings_.width ||
-      frame.height() != codec_settings_.height) {
-    LOG(LS_WARNING) << "Encoder initialized for " << codec_settings_.width
-                    << "x" << codec_settings_.height << " but trying to encode "
-                    << frame.width() << "x" << frame.height() << " frame.";
-    ReportError();
-    return WEBRTC_VIDEO_CODEC_ERR_SIZE;
-  }
 
-  bool force_key_frame = false;
-  if (frame_types != nullptr) {
-    // We only support a single stream.
-    RTC_DCHECK_EQ(frame_types->size(), static_cast<size_t>(1));
-    // Skip frame?
-    if ((*frame_types)[0] == kEmptyFrame) {
-      return WEBRTC_VIDEO_CODEC_OK;
+  std::vector<bool> force_key_frame;
+  force_key_frame.resize(encoders_.size());
+  std::fill(force_key_frame.begin(), force_key_frame.end(), false);
+  for (size_t i = 0; i < key_frame_request_.size() && i < send_streams_.size();
+       ++i) {
+    if (key_frame_request_[i] && send_streams_[i]) {
+      force_key_frame[i] = true;
     }
-    // Force key frame?
-    force_key_frame = (*frame_types)[0] == kVideoFrameKey;
   }
-  if (force_key_frame) {
-    // API doc says ForceIntraFrame(false) does nothing, but calling this
-    // function forces a key frame regardless of the |bIDR| argument's value.
-    // (If every frame is a key frame we get lag/delays.)
-    openh264_encoder_->ForceIntraFrame(true);
+  if (frame_types) {
+    for (size_t i = 0; i < frame_types->size(); ++i) {
+      if ((*frame_types)[i] == kVideoFrameKey) {
+        force_key_frame[i] = true;
+      }
+    }
   }
 
-  // EncodeFrame input.
-  SSourcePicture picture;
-  memset(&picture, 0, sizeof(SSourcePicture));
-  picture.iPicWidth = frame.width();
-  picture.iPicHeight = frame.height();
-  picture.iColorFormat = EVideoFormatType::videoFormatI420;
-  picture.uiTimeStamp = frame.ntp_time_ms();
-  picture.iStride[0] = frame.stride(kYPlane);
-  picture.iStride[1] = frame.stride(kUPlane);
-  picture.iStride[2] = frame.stride(kVPlane);
-  picture.pData[0] = const_cast<uint8_t*>(frame.buffer(kYPlane));
-  picture.pData[1] = const_cast<uint8_t*>(frame.buffer(kUPlane));
-  picture.pData[2] = const_cast<uint8_t*>(frame.buffer(kVPlane));
-
-  // EncodeFrame output.
-  SFrameBSInfo info;
-  memset(&info, 0, sizeof(SFrameBSInfo));
-
-  // Encode!
-  int enc_ret = openh264_encoder_->EncodeFrame(&picture, &info);
-  if (enc_ret != 0) {
-    LOG(LS_ERROR) << "OpenH264 frame encoding failed, EncodeFrame returned "
-                  << enc_ret << ".";
-    ReportError();
-    return WEBRTC_VIDEO_CODEC_ERROR;
-  }
+  for (size_t i = 0; i < encoders_.size(); ++i) {
+    if (!send_streams_[i] || (*frame_types)[i] == kEmptyFrame) {
+      continue;
+    }
+    // Scale input to match encode dimensions
+    Scale(frame, &scaled_input_frames_[i]);
+
+    if (scaled_input_frames_[i].width() !=
+            codec_settings_.simulcastStream[i].width ||
+        scaled_input_frames_[i].height() !=
+            codec_settings_.simulcastStream[i].height) {
+      LOG(LS_ERROR) << "Encoder initialized for "
+                    << codec_settings_.simulcastStream[i].width << "x"
+                    << codec_settings_.simulcastStream[i].height
+                    << " but trying to encode "
+                    << scaled_input_frames_[i].width() << "x"
+                    << scaled_input_frames_[i].height() << " frame.";
+      ReportError();
+      return WEBRTC_VIDEO_CODEC_ERR_SIZE;
+    }
+
+    SSourcePicture picture;
+    memset(&picture, 0, sizeof(SSourcePicture));
+    picture.iPicWidth = scaled_input_frames_[i].width();
+    picture.iPicHeight = scaled_input_frames_[i].height();
+    picture.iColorFormat = EVideoFormatType::videoFormatI420;
+    picture.uiTimeStamp = frame.ntp_time_ms();
+    picture.iStride[0] = scaled_input_frames_[i].stride(kYPlane);
+    picture.iStride[1] = scaled_input_frames_[i].stride(kUPlane);
+    picture.iStride[2] = scaled_input_frames_[i].stride(kVPlane);
+    picture.pData[0] =
+        const_cast<uint8_t*>(scaled_input_frames_[i].buffer(kYPlane));
+    picture.pData[1] =
+        const_cast<uint8_t*>(scaled_input_frames_[i].buffer(kUPlane));
+    picture.pData[2] =
+        const_cast<uint8_t*>(scaled_input_frames_[i].buffer(kVPlane));
+    if (force_key_frame[i]) {
+      // API doc says ForceIntraFrame(false) does nothing, but calling this
+      // function forces a key frame regardless of the |bIDR| argument's value.
+      // (If every frame is a key frame we get lag/delays.)
+      encoders_[i]->ForceIntraFrame(true);
+      std::fill(key_frame_request_.begin(), key_frame_request_.end(), false);
+    }
+    // EncodeFrame output.
+    SFrameBSInfo info;
+    memset(&info, 0, sizeof(SFrameBSInfo));
+    int enc_ret = encoders_[i]->EncodeFrame(&picture, &info);
+    if (enc_ret != 0) {
+      LOG(LS_ERROR) << "OpenH264 frame encoding failed, EncodeFrame returned "
+                    << enc_ret << ".";
+      ReportError();
+      return WEBRTC_VIDEO_CODEC_ERROR;
+    }
 
-  encoded_image_._encodedWidth = frame.width();
-  encoded_image_._encodedHeight = frame.height();
-  encoded_image_._timeStamp = frame.timestamp();
-  encoded_image_.ntp_time_ms_ = frame.ntp_time_ms();
-  encoded_image_.capture_time_ms_ = frame.render_time_ms();
-  encoded_image_.rotation_ = frame.rotation();
-  encoded_image_._frameType = EVideoFrameType_to_FrameType(info.eFrameType);
-
-  // Split encoded image up into fragments. This also updates |encoded_image_|.
-  RTPFragmentationHeader frag_header;
-  RtpFragmentize(&encoded_image_, &encoded_image_buffer_, frame, &info,
-                 &frag_header);
-
-  // Encoder can skip frames to save bandwidth in which case
-  // |encoded_image_._length| == 0.
-  if (encoded_image_._length > 0) {
-    // Deliver encoded image.
-    CodecSpecificInfo codec_specific;
-    codec_specific.codecType = kVideoCodecH264;
-    encoded_image_callback_->Encoded(encoded_image_,
-                                     &codec_specific,
-                                     &frag_header);
+    encoded_images_[i]._encodedWidth = codec_settings_.simulcastStream[i].width;
+    encoded_images_[i]._encodedHeight =
+        codec_settings_.simulcastStream[i].height;
+    encoded_images_[i]._timeStamp = frame.timestamp();
+    encoded_images_[i].ntp_time_ms_ = frame.ntp_time_ms();
+    encoded_images_[i].capture_time_ms_ = frame.render_time_ms();
+    encoded_images_[i]._frameType =
+        EVideoFrameType_to_FrameType(info.eFrameType);
+    // Split encoded image up into fragments. This also updates
+    // |encoded_image_|.
+    RTPFragmentationHeader frag_header;
+    RtpFragmentize(&encoded_images_[i], &encoded_image_buffers_[i], frame,
+                   &info, &frag_header);
+    if (encoded_images_[i]._length > 0) {
+      // Deliver encoded image.
+      CodecSpecificInfo codec_specific;
+      CodecSpecificInfoH264* h264Info = &(codec_specific.codecSpecific.H264);
+      h264Info->simulcastIdx = i;
+      codec_specific.codecType = kVideoCodecH264;
+      encoded_image_callback_->Encoded(encoded_images_[i], &codec_specific,
+                                       &frag_header);
+    }
   }
   return WEBRTC_VIDEO_CODEC_OK;
 }
 
 bool H264EncoderImpl::IsInitialized() const {
-  return openh264_encoder_ != nullptr;
+  for (auto openh264_encoder : encoders_) {
+    if (openh264_encoder == nullptr) {
+      return false;
+    }
+  }
+  return true;
 }
 
 void H264EncoderImpl::ReportInit() {