H264 HW encode using MediaFoundation

media/gpu/media_foundation_video_encode_accelerator_win.cc

Index: media/gpu/media_foundation_video_encode_accelerator_win.cc
diff --git a/media/gpu/media_foundation_video_encode_accelerator_win.cc b/media/gpu/media_foundation_video_encode_accelerator_win.cc
new file mode 100644
index 0000000000000000000000000000000000000000..542938e16c984b793e4f1763b201b9b95913b2f6
--- /dev/null
+++ b/media/gpu/media_foundation_video_encode_accelerator_win.cc
@@ -0,0 +1,435 @@
+// Copyright 2016 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 "media/gpu/media_foundation_video_encode_accelerator_win.h"
+
+#if !defined(OS_WIN)
+// #error This file should only be built on Windows.

[ananta, 2016/06/28 22:50:01]

Don't think we need this. The _win suffix should be good enough

[emircan, 2016/07/02 00:07:38]

Done.

+#else
+#pragma warning(push)
+#pragma warning(disable : 4800)  // Disable warning for added padding.
+#endif                           // !defined(OS_WIN)
+
+#include <codecapi.h>
+#include <mferror.h>
+#include <mftransform.h>
+
+#include "base/threading/thread_task_runner_handle.h"
+#include "base/win/scoped_co_mem.h"
+#include "base/win/scoped_variant.h"
+#include "media/base/win/mf_helpers.h"
+#include "media/base/win/mf_initializer.h"
+#include "third_party/libyuv/include/libyuv.h"
+
+using base::win::ScopedComPtr;
+using media::mf::MediaBufferScopedPointer;
+
+namespace media {
+
+namespace {
+
+const size_t kMaxFrameRateNumerator = 30;
+const size_t kMaxFrameRateDenominator = 1;
+const size_t kMaxResolutionWidth = 4096;
+const size_t kMaxResolutionHeight = 2160;
+const size_t kNumInputBuffers = 3;
+const size_t kOneSecondInMicroseconds = 1000000;
+const size_t kOutputSampleBufferSizeRatio = 4;
+
+static const wchar_t* const kMediaFoundationVideoEncoderDLLs[] = {
+    L"mf.dll", L"mfplat.dll",
+};
+
+}  // namespace
+
+struct MediaFoundationVideoEncodeAccelerator::BitstreamBufferRef {
+  BitstreamBufferRef(int32_t id,
+                     std::unique_ptr<base::SharedMemory> shm,
+                     size_t size)
+      : id(id), shm(std::move(shm)), size(size) {}
+  const int32_t id;
+  const std::unique_ptr<base::SharedMemory> shm;
+  const size_t size;
+
+ private:
+  DISALLOW_IMPLICIT_CONSTRUCTORS(BitstreamBufferRef);
+};
+
+MediaFoundationVideoEncodeAccelerator::MediaFoundationVideoEncodeAccelerator()
+    : win_version_(base::win::GetVersion()),
+      client_task_runner_(base::ThreadTaskRunnerHandle::Get()),
+      encoder_thread_("MFEncoderThread"),
+      encoder_task_weak_factory_(this) {
+  DVLOG(3) << __FUNCTION__;
+  encoder_weak_ptr_ = encoder_task_weak_factory_.GetWeakPtr();
+}
+
+MediaFoundationVideoEncodeAccelerator::
+    ~MediaFoundationVideoEncodeAccelerator() {
+  DVLOG(3) << __FUNCTION__;
+  DCHECK(thread_checker_.CalledOnValidThread());
+
+  Destroy();
+  DCHECK(!encoder_thread_.IsRunning());
+  DCHECK(!encoder_task_weak_factory_.HasWeakPtrs());
+}
+
+media::VideoEncodeAccelerator::SupportedProfiles
+MediaFoundationVideoEncodeAccelerator::GetSupportedProfiles() {
+  DVLOG(3) << __FUNCTION__;
+  DCHECK(thread_checker_.CalledOnValidThread());
+
+  SupportedProfiles profiles;
+  if (base::win::GetVersion() < base::win::VERSION_WIN8) {
+    DLOG(ERROR) << "Windows versions earlier than 8 are not supported.";
+    return profiles;
+  }
+
+  SupportedProfile profile;
+  profile.profile = media::H264PROFILE_BASELINE;

[ananta, 2016/06/28 22:50:01]

Does this always have to be the baseline profile?. Does this mean that the other
profiles won't be supported going forward?. Please add some notes to that
effect.

[emircan, 2016/07/02 00:07:37]

OpenH264 SW only supports baseline right now. In case of a SW fallback we want
to be consistent with that. I can add additional profile if necessary as well
later.

+  profile.max_framerate_numerator = kMaxFrameRateNumerator;
+  profile.max_framerate_denominator = kMaxFrameRateDenominator;
+  profile.max_resolution = gfx::Size(kMaxResolutionWidth, kMaxResolutionHeight);
+  profiles.push_back(profile);
+  return profiles;
+}
+
+bool MediaFoundationVideoEncodeAccelerator::Initialize(
+    media::VideoPixelFormat format,
+    const gfx::Size& input_visible_size,
+    media::VideoCodecProfile output_profile,
+    uint32_t initial_bitrate,

[ananta, 2016/06/28 22:50:01]

The Initialize function is very big. Please split into smaller functions.

[emircan, 2016/07/02 00:07:37]

Done.

+    Client* client) {
+  DVLOG(3) << __FUNCTION__
+           << ": input_format=" << media::VideoPixelFormatToString(format)
+           << ", input_visible_size=" << input_visible_size.ToString()
+           << ", output_profile=" << output_profile
+           << ", initial_bitrate=" << initial_bitrate;
+  DCHECK(thread_checker_.CalledOnValidThread());
+
+  if (media::PIXEL_FORMAT_I420 != format) {
+    DLOG(ERROR) << "Input format not supported= "
+                << media::VideoPixelFormatToString(format);
+    return false;
+  }

[ananta, 2016/06/28 22:50:01]

newline here.

[emircan, 2016/07/02 00:07:37]

Done.

+  if (media::H264PROFILE_BASELINE != output_profile) {
+    DLOG(ERROR) << "Output profile not supported= " << output_profile;
+    return false;
+  }
+
+  for (const wchar_t* mfdll : kMediaFoundationVideoEncoderDLLs) {
+    HMODULE dll = ::GetModuleHandle(mfdll);
+    if (!dll) {
+      DLOG(ERROR) << mfdll << " is required for encoding";
+      return false;
+    }
+  }
+  media::InitializeMediaFoundation();

[ananta, 2016/06/28 22:50:00]

newline here.

[emircan, 2016/07/02 00:07:37]

Done.

+
+  uint32_t flags = MFT_ENUM_FLAG_HARDWARE | MFT_ENUM_FLAG_SORTANDFILTER;
+  MFT_REGISTER_TYPE_INFO input_info;
+  input_info.guidMajorType = MFMediaType_Video;
+  input_info.guidSubtype = MFVideoFormat_NV12;
+  MFT_REGISTER_TYPE_INFO output_info;
+  output_info.guidMajorType = MFMediaType_Video;
+  output_info.guidSubtype = MFVideoFormat_H264;
+
+  base::win::ScopedCoMem<CLSID> CLSIDs;
+  uint32_t count = 0;
+  HRESULT hr = MFTEnum(MFT_CATEGORY_VIDEO_ENCODER, flags, NULL, &output_info,
+                       NULL, &CLSIDs, &count);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't enumerate hardware encoder", false);
+  RETURN_ON_FAILURE((count > 0), "No HW encoder found", false);
+  DVLOG(3) << "HW encoder(s) found: " << count;
+  hr = encoder_.CreateInstance(CLSIDs[0]);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't activate hardware encoder", false);
+
+  if (!encoder_thread_.Start()) {
+    DLOG(ERROR) << "Failed spawning encoder thread.";
+    return false;
+  }
+  encoder_thread_task_runner_ = encoder_thread_.task_runner();

[ananta, 2016/06/28 22:50:01]

newline

[emircan, 2016/07/02 00:07:38]

Done.

+
+  client_ptr_factory_.reset(new base::WeakPtrFactory<Client>(client));
+  client_ = client_ptr_factory_->GetWeakPtr();
+  input_visible_size_ = input_visible_size;
+  frame_rate_ = kMaxFrameRateNumerator / kMaxFrameRateDenominator;
+  target_bitrate_ = initial_bitrate;
+  bitstream_buffer_size_ = input_visible_size.GetArea();
+
+  u_plane_offset_ =
+      VideoFrame::PlaneSize(PIXEL_FORMAT_I420, VideoFrame::kYPlane,
+                            input_visible_size_)
+          .GetArea();
+  v_plane_offset_ =
+      u_plane_offset_ +
+      VideoFrame::PlaneSize(PIXEL_FORMAT_I420, VideoFrame::kUPlane,
+                            input_visible_size_)
+          .GetArea();
+
+  hr = encoder_->GetStreamLimits(
+      &input_stream_count_min_, &input_stream_count_max_,
+      &output_stream_count_min_, &output_stream_count_max_);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't query stream limits", false);
+  DVLOG(3) << "Stream limits: " << input_stream_count_min_ << ","
+           << input_stream_count_max_ << "," << output_stream_count_min_ << ","
+           << output_stream_count_max_;
+
+  base::win::ScopedComPtr<IMFMediaType> imf_output_media_type;
+  hr = MFCreateMediaType(imf_output_media_type.Receive());
+  hr &= imf_output_media_type->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video);

[ananta, 2016/06/28 22:50:01]

what happens if MFCreateMediaType fails?. Additionally is there a point in
continuing if one of the Set operations fails below?

[emircan, 2016/07/02 00:07:38]

I will add a return there if fails.

+  hr &= imf_output_media_type->SetGUID(MF_MT_SUBTYPE, MFVideoFormat_H264);
+  hr &= imf_output_media_type->SetUINT32(MF_MT_AVG_BITRATE, target_bitrate_);
+  hr &= MFSetAttributeRatio(imf_output_media_type.get(), MF_MT_FRAME_RATE,
+                            frame_rate_, kMaxFrameRateDenominator);
+  hr &= MFSetAttributeSize(imf_output_media_type.get(), MF_MT_FRAME_SIZE,

[ananta, 2016/06/28 22:50:01]

Move this to the next line for consistency

[emircan, 2016/07/02 00:07:37]

Done.

+                           input_visible_size_.width(),
+                           input_visible_size_.height());
+  hr &= imf_output_media_type->SetUINT32(MF_MT_INTERLACE_MODE,
+                                         MFVideoInterlace_Progressive);
+  hr &= imf_output_media_type->SetUINT32(MF_MT_MPEG2_PROFILE,
+                                         eAVEncH264VProfile_Base);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't set output params", false);
+  hr = encoder_->SetOutputType(0, imf_output_media_type.get(), 0);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't set output media type", false);
+
+  base::win::ScopedComPtr<IMFMediaType> imf_input_media_type;
+  hr = MFCreateMediaType(imf_input_media_type.Receive());
+  hr &= imf_input_media_type->SetGUID(MF_MT_MAJOR_TYPE, MFMediaType_Video);

[ananta, 2016/06/28 22:50:01]

Ditto for MFCreateMediaType and the other Set operations here

[emircan, 2016/07/02 00:07:38]

Done.

+  hr &= imf_input_media_type->SetGUID(MF_MT_SUBTYPE, MFVideoFormat_YV12);
+  hr &= MFSetAttributeRatio(imf_input_media_type.get(), MF_MT_FRAME_RATE,
+                            frame_rate_, kMaxFrameRateDenominator);
+  hr &= MFSetAttributeSize(imf_input_media_type.get(), MF_MT_FRAME_SIZE,
+                           input_visible_size_.width(),
+                           input_visible_size_.height());
+  hr &= imf_input_media_type->SetUINT32(MF_MT_INTERLACE_MODE,
+                                        MFVideoInterlace_Progressive);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't set input params", false);
+  hr = encoder_->SetInputType(0, imf_input_media_type.get(), 0);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't set input media type", false);
+
+  hr = encoder_.QueryInterface(IID_ICodecAPI, codec_api_.ReceiveVoid());
+  RETURN_ON_HR_FAILURE(hr, "Couldn't get ICodecAPI", false);
+  VARIANT var;
+  var.vt = VT_UI4;
+  var.ulVal = eAVEncCommonRateControlMode_CBR;
+  hr = codec_api_->SetValue(&CODECAPI_AVEncCommonRateControlMode, &var);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't set CommonRateControlMode", false);
+  var.ulVal = target_bitrate_;
+  hr = codec_api_->SetValue(&CODECAPI_AVEncCommonMeanBitRate, &var);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't set bitrate", false);
+  var.ulVal = eAVEncAdaptiveMode_FrameRate;
+  hr = codec_api_->SetValue(&CODECAPI_AVEncAdaptiveMode, &var);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't set FrameRate", false);
+  var.vt = VT_BOOL;
+  var.boolVal = VARIANT_TRUE;
+  hr = codec_api_->SetValue(&CODECAPI_AVLowLatencyMode, &var);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't set LowLatencyMode", false);
+
+  input_sample_.Attach(mf::CreateEmptySampleWithBuffer(
+      VideoFrame::AllocationSize(PIXEL_FORMAT_I420, input_visible_size), 2));
+  output_sample_.Attach(mf::CreateEmptySampleWithBuffer(
+      bitstream_buffer_size_ * kOutputSampleBufferSizeRatio, 2));
+
+  hr = encoder_->ProcessMessage(MFT_MESSAGE_NOTIFY_BEGIN_STREAMING, NULL);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't set ProcessMessage", false);
+
+  client_task_runner_->PostTask(
+      FROM_HERE,
+      base::Bind(&Client::RequireBitstreamBuffers, client_, kNumInputBuffers,
+                 input_visible_size_, bitstream_buffer_size_));
+  return SUCCEEDED(hr);
+}
+
+void MediaFoundationVideoEncodeAccelerator::Encode(
+    const scoped_refptr<media::VideoFrame>& frame,
+    bool force_keyframe) {
+  DVLOG(3) << __FUNCTION__;
+  DCHECK(thread_checker_.CalledOnValidThread());
+
+  encoder_thread_task_runner_->PostTask(
+      FROM_HERE, base::Bind(&MediaFoundationVideoEncodeAccelerator::EncodeTask,
+                            encoder_weak_ptr_, frame, force_keyframe));
+}
+
+void MediaFoundationVideoEncodeAccelerator::UseOutputBitstreamBuffer(
+    const media::BitstreamBuffer& buffer) {
+  DVLOG(3) << __FUNCTION__ << ": buffer size=" << buffer.size();
+  DCHECK(thread_checker_.CalledOnValidThread());
+
+  if (buffer.size() < bitstream_buffer_size_) {
+    DLOG(ERROR) << "Output BitstreamBuffer isn't big enough: " << buffer.size()
+                << " vs. " << bitstream_buffer_size_;
+    client_->NotifyError(kInvalidArgumentError);
+    return;
+  }
+
+  std::unique_ptr<base::SharedMemory> shm(
+      new base::SharedMemory(buffer.handle(), false));
+  if (!shm->Map(buffer.size())) {
+    DLOG(ERROR) << "Failed mapping shared memory.";
+    client_->NotifyError(kPlatformFailureError);
+    return;
+  }
+
+  std::unique_ptr<BitstreamBufferRef> buffer_ref(
+      new BitstreamBufferRef(buffer.id(), std::move(shm), buffer.size()));
+  encoder_thread_task_runner_->PostTask(
+      FROM_HERE,
+      base::Bind(
+          &MediaFoundationVideoEncodeAccelerator::UseOutputBitstreamBufferTask,
+          encoder_weak_ptr_, base::Passed(&buffer_ref)));
+}
+
+void MediaFoundationVideoEncodeAccelerator::RequestEncodingParametersChange(
+    uint32_t bitrate,
+    uint32_t framerate) {
+  DVLOG(3) << __FUNCTION__ << ": bitrate=" << bitrate
+           << ": framerate=" << framerate;
+  DCHECK(thread_checker_.CalledOnValidThread());
+
+  encoder_thread_task_runner_->PostTask(
+      FROM_HERE, base::Bind(&MediaFoundationVideoEncodeAccelerator::
+                                RequestEncodingParametersChangeTask,
+                            encoder_weak_ptr_, bitrate, framerate));
+}
+
+void MediaFoundationVideoEncodeAccelerator::Destroy() {
+  DVLOG(3) << __FUNCTION__;
+  DCHECK(thread_checker_.CalledOnValidThread());
+
+  // Cancel all callbacks.
+  client_ptr_factory_.reset();
+
+  if (encoder_thread_.IsRunning()) {
+    encoder_thread_task_runner_->PostTask(
+        FROM_HERE,
+        base::Bind(&MediaFoundationVideoEncodeAccelerator::DestroyTask,
+                   encoder_weak_ptr_));
+    encoder_thread_.Stop();
+  } else {
+    DestroyTask();
+  }
+}
+
+void MediaFoundationVideoEncodeAccelerator::EncodeTask(
+    const scoped_refptr<media::VideoFrame>& frame,
+    bool force_keyframe) {
+  DVLOG(3) << __FUNCTION__;
+  DCHECK(encoder_thread_task_runner_->BelongsToCurrentThread());
+
+  base::win::ScopedComPtr<IMFMediaBuffer> input_buffer;
+  input_sample_->GetBufferByIndex(0, input_buffer.Receive());
+  {

[ananta, 2016/06/28 22:50:01]

newline here.

[emircan, 2016/07/02 00:07:37]

Done.

+    MediaBufferScopedPointer scoped_buffer(input_buffer.get());
+    DCHECK(scoped_buffer.get());
+    libyuv::I420Copy(frame->visible_data(media::VideoFrame::kYPlane),
+                     frame->stride(media::VideoFrame::kYPlane),
+                     frame->visible_data(media::VideoFrame::kVPlane),
+                     frame->stride(media::VideoFrame::kVPlane),
+                     frame->visible_data(media::VideoFrame::kUPlane),
+                     frame->stride(media::VideoFrame::kUPlane),
+                     scoped_buffer.get(),
+                     frame->stride(media::VideoFrame::kYPlane),
+                     scoped_buffer.get() + u_plane_offset_,
+                     frame->stride(media::VideoFrame::kUPlane),
+                     scoped_buffer.get() + v_plane_offset_,
+                     frame->stride(media::VideoFrame::kVPlane),
+                     input_visible_size_.width(), input_visible_size_.height());
+  }
+  input_sample_->SetSampleTime(frame->timestamp().InMicroseconds() * 10);

[ananta, 2016/06/28 22:50:00]

newline here.

[emircan, 2016/07/02 00:07:37]

Done.

+  input_sample_->SetSampleDuration(kOneSecondInMicroseconds / frame_rate_);
+  HRESULT hr = encoder_->ProcessInput(0, input_sample_.get(), 0);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't encode", );
+  DVLOG(3) << "Sent for encode " << hr;
+
+  ProcessOutput();

[ananta, 2016/06/28 22:50:01]

Perhaps we need to check if ProcessOutput succeeds.

[emircan, 2016/07/02 00:07:37]

I answered below.

+}
+
+void MediaFoundationVideoEncodeAccelerator::ProcessOutput() {

[ananta, 2016/06/28 22:50:00]

Should this function be returning void?. What happens if the ProcessOutput call
fails below?

[emircan, 2016/07/02 00:07:37]

We are fine if it fails. 
- If it says MF_E_TRANSFORM_NEED_MORE_INPUT, we dont call this until after the
next ProcessInput call. 
- We make recursive calls to ProcessOutput to flush out all the buffers. We stop
when it says MF_E_TRANSFORM_NEED_MORE_INPUT.

+  DVLOG(3) << __FUNCTION__;
+  DCHECK(encoder_thread_task_runner_->BelongsToCurrentThread());
+
+  if (bitstream_buffer_queue_.empty()) {
+    DVLOG(3) << "No bitstream buffers.";
+    return;
+  }
+
+  MFT_OUTPUT_DATA_BUFFER output_data_buffer = {0};
+  output_data_buffer.dwStreamID = 0;
+  output_data_buffer.dwStatus = 0;
+  output_data_buffer.pEvents = NULL;
+  output_data_buffer.pSample = output_sample_.get();
+  DWORD status = 0;
+  HRESULT hr = encoder_->ProcessOutput(0, 1, &output_data_buffer, &status);
+  if (hr == MF_E_TRANSFORM_NEED_MORE_INPUT) {
+    DVLOG(3) << "MF_E_TRANSFORM_NEED_MORE_INPUT";
+    return;
+  }
+  RETURN_ON_HR_FAILURE(hr, "Couldn't get encoded data", );
+  DVLOG(3) << "Got encoded data " << hr;
+
+  std::unique_ptr<MediaFoundationVideoEncodeAccelerator::BitstreamBufferRef>
+      buffer_ref = std::move(bitstream_buffer_queue_.front());
+  bitstream_buffer_queue_.pop_front();
+
+  base::win::ScopedComPtr<IMFMediaBuffer> output_buffer;
+  output_sample_->GetBufferByIndex(0, output_buffer.Receive());
+  DWORD size = 0;

[ananta, 2016/06/28 22:50:01]

What happens if GetBufferByIndex fails

[emircan, 2016/07/02 00:07:37]

Adding hr checks.

+  output_buffer->GetCurrentLength(&size);
+  {
+    MediaBufferScopedPointer scoped_buffer(output_buffer.get());
+    memcpy(buffer_ref->shm->memory(), scoped_buffer.get(), size);
+  }
+
+  const bool keyframe = MFGetAttributeUINT32(
+      output_sample_.get(), MFSampleExtension_CleanPoint, false);
+  DVLOG(3) << "We HAVE encoded data with size:" << size << " keyframe "
+           << keyframe;
+
+  client_task_runner_->PostTask(
+      FROM_HERE,
+      base::Bind(&Client::BitstreamBufferReady, client_, buffer_ref->id, size,
+                 keyframe, base::Time::Now() - base::Time()));
+  ProcessOutput();
+}
+
+void MediaFoundationVideoEncodeAccelerator::UseOutputBitstreamBufferTask(
+    std::unique_ptr<BitstreamBufferRef> buffer_ref) {
+  DVLOG(3) << __FUNCTION__;
+  DCHECK(encoder_thread_task_runner_->BelongsToCurrentThread());
+
+  bitstream_buffer_queue_.push_back(std::move(buffer_ref));
+}
+
+void MediaFoundationVideoEncodeAccelerator::RequestEncodingParametersChangeTask(
+    uint32_t bitrate,
+    uint32_t framerate) {
+  DVLOG(3) << __FUNCTION__;
+  DCHECK(encoder_thread_task_runner_->BelongsToCurrentThread());
+
+  frame_rate_ = framerate > 1 ? framerate : 1;
+  target_bitrate_ = bitrate > 1 ? bitrate : 1;
+
+  VARIANT var;
+  var.vt = VT_UI4;
+  var.ulVal = target_bitrate_;
+  HRESULT hr = codec_api_->SetValue(&CODECAPI_AVEncCommonMeanBitRate, &var);
+  RETURN_ON_HR_FAILURE(hr, "Couldn't set bitrate", );
+
+  base::win::ScopedComPtr<IMFMediaType> imf_output_media_type;
+  hr = MFCreateMediaType(imf_output_media_type.Receive());
+  hr &= imf_output_media_type->SetUINT32(MF_MT_AVG_BITRATE, target_bitrate_);
+  hr &= MFSetAttributeRatio(imf_output_media_type.get(), MF_MT_FRAME_RATE,
+                            framerate, kMaxFrameRateDenominator);
+}
+
+void MediaFoundationVideoEncodeAccelerator::DestroyTask() {
+  DVLOG(3) << __FUNCTION__;
+  DCHECK(encoder_thread_task_runner_->BelongsToCurrentThread());
+
+  encoder_.Release();
+}
+
+}  // namespace content