VideoTrackRecorder: Adding inner class H264Encoder

content/renderer/media/video_track_recorder.cc

 // Copyright 2015 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "content/renderer/media/video_track_recorder.h"
 
 #include <utility>
 
 #include "base/bind.h"
 #include "base/logging.h"
 #include "base/macros.h"
 #include "base/sys_info.h"
 #include "base/threading/thread.h"
 #include "base/time/time.h"
 #include "base/trace_event/trace_event.h"
 #include "media/base/video_frame.h"
 #include "media/base/video_util.h"
 #include "ui/gfx/geometry/size.h"
 
+#if BUILDFLAG(RTC_USE_H264)
+#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"
+#endif  // #if BUILDFLAG(RTC_USE_H264)

[hbos_chromium, 2016/04/20 20:28:41]

content_renderer now has a direct dependency on
third_party/openh264.gyp:openh264_encoder (gyp) and third_party/openh264:encoder
(GN). It is indirectly resolved because content_renderer depends on a bunch of
webrtc stuff, some of which depend on the openh264 encoder.

It would be nice if the direct dependency is listed as a dependency if
'rtc_use_h264==1' (buildflag defined in third_party/webrtc/build/common.gypi and
//third_party/webrtc/build/webrtc.gni, see common_features target as an example
of a chromium build file including it). If it's not too much of a hassle,
sometimes GYP is completely borked.

[mcasas, 2016/04/20 23:20:34]

Added a direct dependency in the conditions
section of content.gyp:content_renderer.

+
 extern "C" {
 // VPX_CODEC_DISABLE_COMPAT excludes parts of the libvpx API that provide
 // backwards compatibility for legacy applications using the library.
 #define VPX_CODEC_DISABLE_COMPAT 1
 #include "third_party/libvpx/source/libvpx/vpx/vp8cx.h"
 #include "third_party/libvpx/source/libvpx/vpx/vpx_encoder.h"
 }
 
 using media::VideoFrame;
 using media::VideoFrameMetadata;
 
 namespace content {
 
-// Base class to describe a generic Encoder. This class is used to encapsulate
-// interactions with actual encoders, encoding and delivery of received frames.
-// This class is ref-counted to allow the MediaStreamVideoTrack to hold a
-// reference to it, via the callback that MediaStreamVideoSink passes along.
-// Also, it is quite common that encoders run in a background thread.
+// Base class to describe a generic Encoder, encapsulating all actual encoder
+// (re)configurations, encoding and delivery of received frames. This class is
+// ref-counted to allow the MediaStreamVideoTrack to hold a reference to it (via
+// the callback that MediaStreamVideoSink passes along) and to jump back and
+// forth to an internal encoder thread. Moreover, this class:
+// - is created and destroyed on its parent's thread (usually the main Render
+// thread), |main_task_runner_|.
+// - receives VideoFrames on |origin_task_runner_| and ruyns OnEncodedVideoCB on

[hbos_chromium, 2016/04/20 20:28:42]

nit: ruyns = runs

[mcasas, 2016/04/20 23:20:34]

Done.

+// that thread as well. This task runner is cached on first frame arrival, and
+// is supposed to be the render IO thread (but this is not enforced);
+// - uses an internal |encoding_thread_| for actual encoder interactions, namely
+// configuration, encoding (which might take some time) and destruction.
 class VideoTrackRecorder::Encoder : public base::RefCountedThreadSafe<Encoder> {
  public:
   Encoder(const OnEncodedVideoCB& on_encoded_video_callback,
           int32_t bits_per_second)
-      : paused_(false),
+      : main_task_runner_(base::MessageLoop::current()->task_runner()),
+        encoding_thread_(new base::Thread("EncodingThread")),
+        paused_(false),
         on_encoded_video_callback_(on_encoded_video_callback),
-        bits_per_second_(bits_per_second) {}
+        bits_per_second_(bits_per_second) {
+    DCHECK(!on_encoded_video_callback_.is_null());
+  }
 
-  virtual void StartFrameEncode(const scoped_refptr<VideoFrame>& frame,
-                                base::TimeTicks capture_timestamp) = 0;
+  // Start encoding |frame|, returning via |on_encoded_video_callback_|. This
+  // call will also trigger a ConfigureEncoderOnEncodingThread() upon first
+  // frame arrival or parameter change, and an EncodeOnEncodingThread() to
+  // actually encode the frame.
+  void StartFrameEncode(const scoped_refptr<VideoFrame>& frame,
+                        base::TimeTicks capture_timestamp);
 
-  void set_paused(bool paused) { paused_ = paused; }
+  void SetPaused(bool paused);
 
  protected:
   friend class base::RefCountedThreadSafe<Encoder>;
   virtual ~Encoder() {}
 
-  // While |paused_|, frames are not encoded.
+  virtual void EncodeOnEncodingThread(const scoped_refptr<VideoFrame>& frame,
+                                      base::TimeTicks capture_timestamp) = 0;
+  virtual void ConfigureEncoderOnEncodingThread(const gfx::Size& size) = 0;
+
+  // Used to shutdown properly on the same thread we were created.
+  const scoped_refptr<base::SingleThreadTaskRunner> main_task_runner_;
+
+  // Task runner where frames to encode and reply callbacks must happen.
+  scoped_refptr<base::SingleThreadTaskRunner> origin_task_runner_;
+
+  // Thread for encoding. Active for the lifetime of VpxEncoder.
+  std::unique_ptr<base::Thread> encoding_thread_;
+
+  // While |paused_|, frames are not encoded. Used only from |encoding_thread_|.
   bool paused_;
 
   // This callback should be exercised on IO thread.
   const OnEncodedVideoCB on_encoded_video_callback_;
 
-  // Target bitrate or video encoding. If 0, a standard bitrate is used.
+  // Target bitrate for video encoding. If 0, a standard bitrate is used.
   const int32_t bits_per_second_;
 
   DISALLOW_COPY_AND_ASSIGN(Encoder);
 };
 
+void VideoTrackRecorder::Encoder::StartFrameEncode(
+    const scoped_refptr<VideoFrame>& video_frame,
+    base::TimeTicks capture_timestamp) {
+  // Cache the thread sending frames on first frame arrival.
+  if (!origin_task_runner_.get())
+    origin_task_runner_ = base::MessageLoop::current()->task_runner();
+  DCHECK(origin_task_runner_->BelongsToCurrentThread());
+  if (paused_)
+    return;
+
+  if (!(video_frame->format() == media::PIXEL_FORMAT_I420 ||
+        video_frame->format() == media::PIXEL_FORMAT_YV12 ||
+        video_frame->format() == media::PIXEL_FORMAT_YV12A)) {
+    NOTREACHED();
+    return;
+  }
+  scoped_refptr<media::VideoFrame> frame = video_frame;
+  // Drop alpha channel since we do not support it yet.
+  if (frame->format() == media::PIXEL_FORMAT_YV12A)
+    frame = media::WrapAsI420VideoFrame(video_frame);
+
+  encoding_thread_->task_runner()->PostTask(
+      FROM_HERE, base::Bind(&Encoder::EncodeOnEncodingThread,
+                            this, frame, capture_timestamp));
+}
+
+void VideoTrackRecorder::Encoder::SetPaused(bool paused) {
+  if (!encoding_thread_->task_runner()->BelongsToCurrentThread()) {
+    encoding_thread_->task_runner()->PostTask(
+        FROM_HERE, base::Bind(&Encoder::SetPaused, this, paused));
+    return;
+  }
+  paused_ = paused;
+}
+
 namespace {
 
-const vpx_codec_flags_t kNoFlags = 0;
-
 // Originally from remoting/codec/scoped_vpx_codec.h.
 // TODO(mcasas): Refactor into a common location.
 struct VpxCodecDeleter {
   void operator()(vpx_codec_ctx_t* codec) {
     if (!codec)
       return;
     vpx_codec_err_t ret = vpx_codec_destroy(codec);
     CHECK_EQ(ret, VPX_CODEC_OK);
     delete codec;
   }
 };
-
 typedef std::unique_ptr<vpx_codec_ctx_t, VpxCodecDeleter> ScopedVpxCodecCtxPtr;
 
-void OnFrameEncodeCompleted(
+static void OnFrameEncodeCompleted(
     const VideoTrackRecorder::OnEncodedVideoCB& on_encoded_video_cb,
     const scoped_refptr<VideoFrame>& frame,
     std::unique_ptr<std::string> data,
     base::TimeTicks capture_timestamp,
     bool keyframe) {
   DVLOG(1) << (keyframe ? "" : "non ") << "keyframe "<< data->length() << "B, "
            << capture_timestamp << " ms";
   on_encoded_video_cb.Run(frame, std::move(data), capture_timestamp, keyframe);
 }
 
-// Class encapsulating libvpx interactions, encoding and delivery of received
-// frames. This class:
-// - is created and destroyed on its parent's thread (usually the main Render
-// thread);
-// - receives VideoFrames and Run()s the callbacks on |origin_task_runner_|,
-// which is cached on first frame arrival, and is supposed to be the render IO
-// thread, but this is not enforced;
-// - uses an internal |encoding_thread_| for libvpx interactions, notably for
-// encoding (which might take some time).
+static int GetNumberOfThreadsForEncoding() {
+  // Do not saturate CPU utilization just for encoding. On a lower-end system
+  // with only 1 or 2 cores, use only one thread for encoding. On systems with
+  // more cores, allow half of the cores to be used for encoding.
+  return std::min(8, (base::SysInfo::NumberOfProcessors() + 1) / 2);
+}
+
+// Class encapsulating all libvpx interactions for VP8/VP9 encoding.
 class VpxEncoder final : public VideoTrackRecorder::Encoder {
  public:
   static void ShutdownEncoder(std::unique_ptr<base::Thread> encoding_thread,
                               ScopedVpxCodecCtxPtr encoder);
 
   VpxEncoder(
       bool use_vp9,
       const VideoTrackRecorder::OnEncodedVideoCB& on_encoded_video_callback,
       int32_t bits_per_second);
 
-  void StartFrameEncode(const scoped_refptr<VideoFrame>& frame,
-                        base::TimeTicks capture_timestamp) override;
-
  private:
+  // VideoTrackRecorder::Encoder
   ~VpxEncoder() override;
-
   void EncodeOnEncodingThread(const scoped_refptr<VideoFrame>& frame,
-                              base::TimeTicks capture_timestamp);
-
-  void ConfigureEncoding(const gfx::Size& size);
+                              base::TimeTicks capture_timestamp) override;
+  void ConfigureEncoderOnEncodingThread(const gfx::Size& size) override;
 
   // Returns true if |codec_config_| has been filled in at least once.
   bool IsInitialized() const;
 
   // Estimate the frame duration from |frame| and |last_frame_timestamp_|.
-  base::TimeDelta CalculateFrameDuration(
-      const scoped_refptr<VideoFrame>& frame);
+  base::TimeDelta EstimateFrameDuration(const scoped_refptr<VideoFrame>& frame);
 
   // Force usage of VP9 for encoding, instead of VP8 which is the default.
   const bool use_vp9_;
 
-  // Used to shutdown properly on the same thread we were created.
-  const scoped_refptr<base::SingleThreadTaskRunner> main_task_runner_;
-
-  // Task runner where frames to encode and reply callbacks must happen.
-  scoped_refptr<base::SingleThreadTaskRunner> origin_task_runner_;
-
-  // Thread for encoding. Active for the lifetime of VpxEncoder. All variables
-  // below this are used in this thread.
-  std::unique_ptr<base::Thread> encoding_thread_;
-  // VP8 internal objects: configuration and encoder.
+  // VPx internal objects: configuration and encoder. |encoder_| is a special
+  // scoped pointer to guarantee proper destruction, particularly when
+  // reconfiguring due to parameters change. Only used on |encoding_thread_|.
   vpx_codec_enc_cfg_t codec_config_;
-  // |encoder_| is a special scoped pointer to guarantee proper destruction.
-  // Again, it should only be accessed on |encoding_thread_|.
   ScopedVpxCodecCtxPtr encoder_;
 
   // The |VideoFrame::timestamp()| of the last encoded frame.  This is used to
-  // predict the duration of the next frame.
+  // predict the duration of the next frame. Only used on |encoding_thread_|.
   base::TimeDelta last_frame_timestamp_;
 
   DISALLOW_COPY_AND_ASSIGN(VpxEncoder);
 };
 
+#if BUILDFLAG(RTC_USE_H264)
+
+struct ISVCEncoderDeleter {
+  void operator()(ISVCEncoder* codec) {
+    if (!codec)
+      return;
+    const int uninit_ret = codec->Uninitialize();
+    CHECK_EQ(cmResultSuccess, uninit_ret);
+    WelsDestroySVCEncoder(codec);

[hbos_chromium, 2016/04/20 20:28:41]

This runs when ~H264Encoder runs, which is ref counted? It could run when
VideoTrackRecorder is destroyed. This would be the main renderer thread? Not the
encoding thread? (Thread safety concern.) I don't think this is a problem
though, supposedly if there is encoding still going on the encoder is still
referenced and protected against destruction.

[mcasas, 2016/04/20 23:20:34]

It runs on ShutdownEncoder(), posted from ~H264Encoder,
but also when |encoder_| is destroyed because of a 
change in encoding parameters, in l.549, 
ConfigureEncoderOnEncodingThread(). In both cases this
is only executed on |encoding_thread_|.

FTR, this sequence has been polished already for the 
VPx encoder, where the thread hopping was very carefully 
reviewed by miu@.

[hbos_chromium, 2016/04/21 07:36:09]

Acknowledged.

+  }
+};
+typedef std::unique_ptr<ISVCEncoder, ISVCEncoderDeleter> ScopedISVCEncoderPtr;
+
+// Class encapsulating all openh264 interactions for H264 encoding.
+class H264Encoder final : public VideoTrackRecorder::Encoder {
+ public:
+  static void ShutdownEncoder(std::unique_ptr<base::Thread> encoding_thread,
+                              ScopedISVCEncoderPtr encoder);
+
+  H264Encoder(
+      const VideoTrackRecorder::OnEncodedVideoCB& on_encoded_video_callback,
+      int32_t bits_per_second);
+
+ private:
+  // VideoTrackRecorder::Encoder
+  ~H264Encoder() override;
+  void EncodeOnEncodingThread(const scoped_refptr<VideoFrame>& frame,
+                              base::TimeTicks capture_timestamp) override;
+  void ConfigureEncoderOnEncodingThread(const gfx::Size& size) override;
+
+  // |openh264_encoder_| is a special scoped pointer to guarantee proper
+  // destruction, also when reconfiguring due to parameters change. Only used on
+  // |encoding_thread_|.
+  gfx::Size configured_size_;
+  ScopedISVCEncoderPtr openh264_encoder_;
+
+  // The |VideoFrame::timestamp()| of the first received frame. Only used on
+  // |encoding_thread_|.
+  base::TimeTicks first_frame_timestamp_;
+
+  DISALLOW_COPY_AND_ASSIGN(H264Encoder);
+};
+
+#endif  // #if BUILDFLAG(RTC_USE_H264)
+
 // static
 void VpxEncoder::ShutdownEncoder(std::unique_ptr<base::Thread> encoding_thread,
                                  ScopedVpxCodecCtxPtr encoder) {
   DCHECK(encoding_thread->IsRunning());
   encoding_thread->Stop();
   // Both |encoding_thread| and |encoder| will be destroyed at end-of-scope.
 }
 
 VpxEncoder::VpxEncoder(
     bool use_vp9,
     const VideoTrackRecorder::OnEncodedVideoCB& on_encoded_video_callback,
     int32_t bits_per_second)
     : Encoder(on_encoded_video_callback, bits_per_second),
-      use_vp9_(use_vp9),
-      main_task_runner_(base::MessageLoop::current()->task_runner()),
-      encoding_thread_(new base::Thread("EncodingThread")) {
-  DCHECK(!on_encoded_video_callback_.is_null());
-
+      use_vp9_(use_vp9) {
   codec_config_.g_timebase.den = 0;  // Not initialized.
 
   DCHECK(!encoding_thread_->IsRunning());
   encoding_thread_->Start();
 }
 
 VpxEncoder::~VpxEncoder() {
   main_task_runner_->PostTask(FROM_HERE,
                               base::Bind(&VpxEncoder::ShutdownEncoder,
                                          base::Passed(&encoding_thread_),
                                          base::Passed(&encoder_)));
 }
 
-void VpxEncoder::StartFrameEncode(const scoped_refptr<VideoFrame>& frame,
-                                  base::TimeTicks capture_timestamp) {
-  // Cache the thread sending frames on first frame arrival.
-  if (!origin_task_runner_.get())
-    origin_task_runner_ = base::MessageLoop::current()->task_runner();
-  DCHECK(origin_task_runner_->BelongsToCurrentThread());
-  if (paused_)
-    return;
-  encoding_thread_->task_runner()->PostTask(
-      FROM_HERE, base::Bind(&VpxEncoder::EncodeOnEncodingThread,
-                            this, frame, capture_timestamp));
-}
-
-void VpxEncoder::EncodeOnEncodingThread(
-    const scoped_refptr<VideoFrame>& video_frame,
-    base::TimeTicks capture_timestamp) {
+void VpxEncoder::EncodeOnEncodingThread(const scoped_refptr<VideoFrame>& frame,
+                                        base::TimeTicks capture_timestamp) {
   TRACE_EVENT0("video", "VpxEncoder::EncodeOnEncodingThread");
   DCHECK(encoding_thread_->task_runner()->BelongsToCurrentThread());
 
-  if (!(video_frame->format() == media::PIXEL_FORMAT_I420 ||
-        video_frame->format() == media::PIXEL_FORMAT_YV12 ||
-        video_frame->format() == media::PIXEL_FORMAT_YV12A)) {
-    NOTREACHED();
-    return;
-  }
-  scoped_refptr<media::VideoFrame> frame = video_frame;
-  // Drop alpha channel since we do not support it yet.
-  if (frame->format() == media::PIXEL_FORMAT_YV12A)
-    frame = media::WrapAsI420VideoFrame(video_frame);
-
   const gfx::Size frame_size = frame->visible_rect().size();
   if (!IsInitialized() ||
       gfx::Size(codec_config_.g_w, codec_config_.g_h) != frame_size) {
-    ConfigureEncoding(frame_size);
+    ConfigureEncoderOnEncodingThread(frame_size);
   }
 
   vpx_image_t vpx_image;
   vpx_image_t* const result = vpx_img_wrap(&vpx_image,
                                            VPX_IMG_FMT_I420,
                                            frame_size.width(),
                                            frame_size.height(),
                                            1  /* align */,
                                            frame->data(VideoFrame::kYPlane));
   DCHECK_EQ(result, &vpx_image);
   vpx_image.planes[VPX_PLANE_Y] = frame->visible_data(VideoFrame::kYPlane);
   vpx_image.planes[VPX_PLANE_U] = frame->visible_data(VideoFrame::kUPlane);
   vpx_image.planes[VPX_PLANE_V] = frame->visible_data(VideoFrame::kVPlane);
   vpx_image.stride[VPX_PLANE_Y] = frame->stride(VideoFrame::kYPlane);
   vpx_image.stride[VPX_PLANE_U] = frame->stride(VideoFrame::kUPlane);
   vpx_image.stride[VPX_PLANE_V] = frame->stride(VideoFrame::kVPlane);
 
-  const base::TimeDelta duration = CalculateFrameDuration(frame);
+  const base::TimeDelta duration = EstimateFrameDuration(frame);
   // Encode the frame.  The presentation time stamp argument here is fixed to
   // zero to force the encoder to base its single-frame bandwidth calculations
   // entirely on |predicted_frame_duration|.
   const vpx_codec_err_t ret = vpx_codec_encode(encoder_.get(),
                                                &vpx_image,
                                                0  /* pts */,
                                                duration.InMicroseconds(),
-                                               kNoFlags,
+                                               0 /* flags */,
                                                VPX_DL_REALTIME);
   DCHECK_EQ(ret, VPX_CODEC_OK) << vpx_codec_err_to_string(ret) << ", #"
                                << vpx_codec_error(encoder_.get()) << " -"
                                << vpx_codec_error_detail(encoder_.get());
 
   std::unique_ptr<std::string> data(new std::string);
   bool keyframe = false;
   vpx_codec_iter_t iter = NULL;
   const vpx_codec_cx_pkt_t* pkt = NULL;
   while ((pkt = vpx_codec_get_cx_data(encoder_.get(), &iter)) != NULL) {
     if (pkt->kind != VPX_CODEC_CX_FRAME_PKT)
       continue;
     data->assign(static_cast<char*>(pkt->data.frame.buf), pkt->data.frame.sz);
     keyframe = (pkt->data.frame.flags & VPX_FRAME_IS_KEY) != 0;
     break;
   }
   origin_task_runner_->PostTask(FROM_HERE,
       base::Bind(OnFrameEncodeCompleted,
                  on_encoded_video_callback_,
                  frame,
                  base::Passed(&data),
                  capture_timestamp,
                  keyframe));
 }
 
-void VpxEncoder::ConfigureEncoding(const gfx::Size& size) {
+void VpxEncoder::ConfigureEncoderOnEncodingThread(const gfx::Size& size) {
+  DCHECK(encoding_thread_->task_runner()->BelongsToCurrentThread());
   if (IsInitialized()) {
     // TODO(mcasas) VP8 quirk/optimisation: If the new |size| is strictly less-
     // than-or-equal than the old size, in terms of area, the existing encoder
     // instance could be reused after changing |codec_config_.{g_w,g_h}|.
     DVLOG(1) << "Destroying/Re-Creating encoder for new frame size: "
              << gfx::Size(codec_config_.g_w, codec_config_.g_h).ToString()
              << " --> " << size.ToString() << (use_vp9_ ? " vp9" : " vp8");
     encoder_.reset();
   }
 
@@ skipping 44 matching lines @@
   // max distance out of necessity.
   // Note that due to http://crbug.com/440223, it might be necessary to force a
   // key frame after 10,000frames since decoding fails after 30,000 non-key
   // frames.
   // Forcing a keyframe in regular intervals also allows seeking in the
   // resulting recording with decent performance.
   codec_config_.kf_mode = VPX_KF_AUTO;
   codec_config_.kf_min_dist = 0;
   codec_config_.kf_max_dist = 100;
 
-  // Do not saturate CPU utilization just for encoding. On a lower-end system
-  // with only 1 or 2 cores, use only one thread for encoding. On systems with
-  // more cores, allow half of the cores to be used for encoding.
-  codec_config_.g_threads =
-      std::min(8, (base::SysInfo::NumberOfProcessors() + 1) / 2);
+  codec_config_.g_threads = GetNumberOfThreadsForEncoding();
 
   // Number of frames to consume before producing output.
   codec_config_.g_lag_in_frames = 0;
 
   DCHECK(!encoder_);
   encoder_.reset(new vpx_codec_ctx_t);
   const vpx_codec_err_t ret = vpx_codec_enc_init(encoder_.get(), interface,
-                                                 &codec_config_, kNoFlags);
+                                                 &codec_config_, 0 /* flags */);
   DCHECK_EQ(VPX_CODEC_OK, ret);
 
   if (use_vp9_) {
     // Values of VP8E_SET_CPUUSED greater than 0 will increase encoder speed at
     // the expense of quality up to a maximum value of 8 for VP9, by tuning the
     // target time spent encoding the frame. Go from 8 to 5 (values for real
     // time encoding) depending on the amount of cores available in the system.
     const int kCpuUsed =
         std::max(5, 8 - base::SysInfo::NumberOfProcessors() / 2);
     result = vpx_codec_control(encoder_.get(), VP8E_SET_CPUUSED, kCpuUsed);
     DLOG_IF(WARNING, VPX_CODEC_OK != result) << "VP8E_SET_CPUUSED failed";
   }
 }
 
 bool VpxEncoder::IsInitialized() const {
   DCHECK(encoding_thread_->task_runner()->BelongsToCurrentThread());
   return codec_config_.g_timebase.den != 0;
 }
 
-base::TimeDelta VpxEncoder::CalculateFrameDuration(
+base::TimeDelta VpxEncoder::EstimateFrameDuration(
     const scoped_refptr<VideoFrame>& frame) {
   DCHECK(encoding_thread_->task_runner()->BelongsToCurrentThread());
 
   using base::TimeDelta;
   TimeDelta predicted_frame_duration;
   if (!frame->metadata()->GetTimeDelta(VideoFrameMetadata::FRAME_DURATION,
                                        &predicted_frame_duration) ||
       predicted_frame_duration <= TimeDelta()) {
     // The source of the video frame did not provide the frame duration.  Use
     // the actual amount of time between the current and previous frame as a
     // prediction for the next frame's duration.
     // TODO(mcasas): This duration estimation could lead to artifacts if the
     // cadence of the received stream is compromised (e.g. camera freeze, pause,
     // remote packet loss).  Investigate using GetFrameRate() in this case.
     predicted_frame_duration = frame->timestamp() - last_frame_timestamp_;
   }
   last_frame_timestamp_ = frame->timestamp();
   // Make sure |predicted_frame_duration| is in a safe range of values.
   const TimeDelta kMaxFrameDuration = TimeDelta::FromSecondsD(1.0 / 8);
   const TimeDelta kMinFrameDuration = TimeDelta::FromMilliseconds(1);
   return std::min(kMaxFrameDuration, std::max(predicted_frame_duration,
                                               kMinFrameDuration));
 }
 
+#if BUILDFLAG(RTC_USE_H264)
+
+// static
+void H264Encoder::ShutdownEncoder(std::unique_ptr<base::Thread> encoding_thread,
+                                  ScopedISVCEncoderPtr encoder) {
+  DCHECK(encoding_thread->IsRunning());
+  encoding_thread->Stop();
+  // Both |encoding_thread| and |encoder| will be destroyed at end-of-scope.
+}
+
+H264Encoder::H264Encoder(
+    const VideoTrackRecorder::OnEncodedVideoCB& on_encoded_video_callback,
+    int32_t bits_per_second)
+    : Encoder(on_encoded_video_callback, bits_per_second) {
+  DCHECK(!encoding_thread_->IsRunning());
+  encoding_thread_->Start();
+}
+
+H264Encoder::~H264Encoder() {
+  main_task_runner_->PostTask(FROM_HERE,
+                              base::Bind(&H264Encoder::ShutdownEncoder,
+                                         base::Passed(&encoding_thread_),
+                                         base::Passed(&openh264_encoder_)));
+}
+
+void H264Encoder::EncodeOnEncodingThread(const scoped_refptr<VideoFrame>& frame,
+                                         base::TimeTicks capture_timestamp) {
+  TRACE_EVENT0("video", "H264Encoder::EncodeOnEncodingThread");
+  DCHECK(encoding_thread_->task_runner()->BelongsToCurrentThread());
+
+  const gfx::Size frame_size = frame->visible_rect().size();
+  if (!openh264_encoder_ || configured_size_ != frame_size) {
+    ConfigureEncoderOnEncodingThread(frame_size);
+    first_frame_timestamp_ = capture_timestamp;
+  }
+
+  SSourcePicture picture = {};
+  picture.iPicWidth = frame_size.width();
+  picture.iPicHeight = frame_size.height();
+  picture.iColorFormat = EVideoFormatType::videoFormatI420;
+  picture.uiTimeStamp =
+      (capture_timestamp - first_frame_timestamp_).InMilliseconds();
+  picture.iStride[0] = frame->stride(VideoFrame::kYPlane);
+  picture.iStride[1] = frame->stride(VideoFrame::kUPlane);
+  picture.iStride[2] = frame->stride(VideoFrame::kVPlane);
+  picture.pData[0] = frame->data(VideoFrame::kYPlane);
+  picture.pData[1] = frame->data(VideoFrame::kUPlane);
+  picture.pData[2] = frame->data(VideoFrame::kVPlane);

[hbos_chromium, 2016/04/20 20:28:41]

How does this work with visible_rect() that is not aligned from (0,0)? Or the
strides if (width,height) is not 100% of the image?

[mcasas, 2016/04/20 23:20:34]

Oh you're right.  Corrected.

+
+  SFrameBSInfo info = {};
+  if (openh264_encoder_->EncodeFrame(&picture, &info) != cmResultSuccess) {
+    NOTREACHED() << "OpenH264 encoding failed";
+    return;
+  }
+
+  std::unique_ptr<std::string> data(new std::string);
+  const uint8_t kNALStartCode[4] = {0, 0, 0, 1};
+  for (int layer = 0; layer < info.iLayerNum; ++layer) {
+    const SLayerBSInfo& layerInfo = info.sLayerInfo[layer];
+    // Iterate NAL units making up this layer, noting fragments.
+    size_t layer_len = 0;
+    for (int nal = 0; nal < layerInfo.iNalCount; ++nal) {
+      // The following DCHECKs make sure that the header of each NAl is OK.

[hbos_chromium, 2016/04/20 20:28:41]

nit: NAL or NAL unit

[mcasas, 2016/04/20 23:20:34]

Done.

+      DCHECK_GE(layerInfo.pNalLengthInByte[nal], 4);
+      DCHECK_EQ(kNALStartCode[0], layerInfo.pBsBuf[layer_len+0]);
+      DCHECK_EQ(kNALStartCode[1], layerInfo.pBsBuf[layer_len+1]);
+      DCHECK_EQ(kNALStartCode[2], layerInfo.pBsBuf[layer_len+2]);
+      DCHECK_EQ(kNALStartCode[3], layerInfo.pBsBuf[layer_len+3]);
+
+      layer_len += layerInfo.pNalLengthInByte[nal];
+    }
+    // Copy the entire layer's data (including NAL start codes).
+    data->append(reinterpret_cast<char*>(layerInfo.pBsBuf), layer_len);
+  }

[hbos_chromium, 2016/04/20 20:28:42]

(The NAL start codes should be part of the data and are necessary for decoding.
In the webrtc repo they are stripped before being transmitted and then
reinserted on the other end or something, packetization magic, not sure why.)

[mcasas, 2016/04/20 23:20:34]

Acknowledged.

+
+  const bool is_key_frame = info.eFrameType == videoFrameTypeIDR;
+  origin_task_runner_->PostTask(
+      FROM_HERE,
+      base::Bind(OnFrameEncodeCompleted, on_encoded_video_callback_, frame,
+                 base::Passed(&data), capture_timestamp, is_key_frame));
+}
+
+void H264Encoder::ConfigureEncoderOnEncodingThread(const gfx::Size& size) {
+  DCHECK(encoding_thread_->task_runner()->BelongsToCurrentThread());
+  ISVCEncoder* temp_encoder = nullptr;
+  if (WelsCreateSVCEncoder(&temp_encoder) != 0) {
+    NOTREACHED() << "Failed to create OpenH264 encoder";
+    return;
+  }
+  openh264_encoder_.reset(temp_encoder);

[hbos_chromium, 2016/04/20 20:28:41]

DCHECK that the encoder is not already set (or delete the old one) to make sure
we don't leak an existing encoder instance.

[mcasas, 2016/04/20 23:20:34]

See my comments in ISVCEncoderDeleter, the idea
is that we use a ScopedISVCEncoderPtr with the
appropriate semantics for its destruction. This
is the same pattern used for Vpx encoder in
l.350 (of this patch) using ScopedVpxCodecCtxPtr.

FTR, while writing this code and before adding the
specific destructor, I just used a plan unique_ptr
and openh264 was correctly calling Uninitialize()
and then WelsDestroySVCEncoder(). Go figure :)

[hbos_chromium, 2016/04/21 07:36:09]

Oh right, if there is an old one it is appropriately destroyed. No need to
DCHECK.

+  configured_size_ = size;
+
+#if DCHECK_IS_ON()
+  int trace_level = WELS_LOG_INFO;
+  openh264_encoder_->SetOption(ENCODER_OPTION_TRACE_LEVEL, &trace_level);
+#endif
+
+  SEncParamExt init_params;
+  openh264_encoder_->GetDefaultParams(&init_params);
+  init_params.iUsageType = CAMERA_VIDEO_REAL_TIME;
+
+  DCHECK_EQ(AUTO_REF_PIC_COUNT, init_params.iNumRefFrame);
+  DCHECK(!init_params.bSimulcastAVC);
+
+  init_params.uiIntraPeriod = 100;  // Same as for VpxEncoder.
+  init_params.iPicWidth = size.width();
+  init_params.iPicHeight = size.height();
+
+  DCHECK_EQ(RC_QUALITY_MODE, init_params.iRCMode);
+  DCHECK_EQ(0, init_params.iPaddingFlag);
+  DCHECK_EQ(UNSPECIFIED_BIT_RATE, init_params.iTargetBitrate);
+  DCHECK_EQ(UNSPECIFIED_BIT_RATE, init_params.iMaxBitrate);
+  if (bits_per_second_ > 0) {
+    init_params.iRCMode = RC_BITRATE_MODE;
+    init_params.iTargetBitrate = bits_per_second_;
+  } else {
+    init_params.iRCMode = RC_OFF_MODE;
+  }
+
+  // Threading model: 0 means auto, here we select explicitly.
+  init_params.iMultipleThreadIdc = GetNumberOfThreadsForEncoding();

[hbos_chromium, 2016/04/20 20:28:41]

I had to limit the number of threads to 1 due to unresolved issue:
crbug.com/583348.

[mcasas, 2016/04/20 23:20:34]

Added comment, done.

+
+  // TODO(mcasas): consider reducing complexity if there are few CPUs available.
+  DCHECK_EQ(MEDIUM_COMPLEXITY, init_params.iComplexityMode);
+  DCHECK(!init_params.bEnableDenoise);
+  DCHECK(init_params.bEnableFrameSkip);
+
+  // The base spatial layer 0 is the only one we use.
+  DCHECK_EQ(1, init_params.iSpatialLayerNum);
+  init_params.sSpatialLayers[0].iVideoWidth        = init_params.iPicWidth;
+  init_params.sSpatialLayers[0].iVideoHeight       = init_params.iPicHeight;
+  init_params.sSpatialLayers[0].iSpatialBitrate    = init_params.iTargetBitrate;
+  // Slice num according to number of threads.
+  init_params.sSpatialLayers[0].sSliceCfg.uiSliceMode = SM_AUTO_SLICE;
+
+  if (openh264_encoder_->InitializeExt(&init_params) != cmResultSuccess) {
+    NOTREACHED() << "Failed to initialize OpenH264 encoder";
+    return;
+  }
+
+  int pixel_format = EVideoFormatType::videoFormatI420;
+  openh264_encoder_->SetOption(ENCODER_OPTION_DATAFORMAT, &pixel_format);
+}
+#endif //#if BUILDFLAG(RTC_USE_H264)
+
 }  // anonymous namespace
 
 VideoTrackRecorder::VideoTrackRecorder(
     CodecId codec,
     const blink::WebMediaStreamTrack& track,
     const OnEncodedVideoCB& on_encoded_video_callback,
     int32_t bits_per_second)
-    : track_(track),
-      encoder_(new VpxEncoder(codec == CodecId::VP9,
-                              on_encoded_video_callback,
-                              bits_per_second)) {
+    : track_(track) {
   DCHECK(main_render_thread_checker_.CalledOnValidThread());
   DCHECK(!track_.isNull());
   DCHECK(track_.getExtraData());
 
+  switch (codec) {
+#if BUILDFLAG(RTC_USE_H264)
+    case CodecId::H264:
+      encoder_ = new H264Encoder(on_encoded_video_callback, bits_per_second);
+      break;
+#endif
+    case CodecId::VP8:
+    case CodecId::VP9:
+      encoder_ = new VpxEncoder(codec == CodecId::VP9,
+                                on_encoded_video_callback, bits_per_second);
+      break;
+    default:
+      NOTREACHED() << "Unsupported codec";
+  }
+
   // StartFrameEncode() will be called on Render IO thread.
   MediaStreamVideoSink::ConnectToTrack(
       track_,
       base::Bind(&VideoTrackRecorder::Encoder::StartFrameEncode, encoder_));
 }
 
 VideoTrackRecorder::~VideoTrackRecorder() {
   DCHECK(main_render_thread_checker_.CalledOnValidThread());
   MediaStreamVideoSink::DisconnectFromTrack();
   track_.reset();
 }
 
 void VideoTrackRecorder::Pause() {
   DCHECK(main_render_thread_checker_.CalledOnValidThread());
   DCHECK(encoder_);
-  encoder_->set_paused(true);
+  encoder_->SetPaused(true);
 }
 
 void VideoTrackRecorder::Resume() {
   DCHECK(main_render_thread_checker_.CalledOnValidThread());
   DCHECK(encoder_);
-  encoder_->set_paused(false);
+  encoder_->SetPaused(false);
 }
 
 void VideoTrackRecorder::OnVideoFrameForTesting(
     const scoped_refptr<media::VideoFrame>& frame,
     base::TimeTicks timestamp) {
   encoder_->StartFrameEncode(frame, timestamp);
 }
 
 }  // namespace content