[cros] Implement WebM encoder/muxer for animated avatar capture.

media/webm/chromeos/webm_encoder.cc

+// Copyright (c) 2012 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/webm/chromeos/webm_encoder.h"
+
+#include "base/bind.h"
+#include "base/file_util.h"
+#include "base/logging.h"
+#include "libyuv/convert.h"

[scherkus (not reviewing), 2012/07/19 21:46:03]

shouldn't this be third_party/libyuv/...?

[Ivan Korotkov, 2012/07/20 11:10:36]

libyuv.gyp adds third_party/libyuv/include/ to -D, so this is ok:
https://cs.corp.google.com/#chrome/src/third_party/libjingle/source/talk/sess...
Or it could be #include "third_party/libyuv/include/libyuv/..."

+#include "libyuv/video_common.h"
+#include "third_party/skia/include/core/SkBitmap.h"
+
+extern "C" {
+// Getting the right degree of C compatibility has been a constant struggle.
+// - Stroustrup, C++ Report, 12(7), July/August 2000.
+#define private priv
+#include "third_party/libvpx/source/libvpx/libmkv/EbmlIDs.h"

[scherkus (not reviewing), 2012/07/19 21:46:03]

what's going on here?

is there a particular header that causes grief?

[Ivan Korotkov, 2012/07/20 11:10:36]

One of functions in EbmlWriter.h has argument named "private" :)

[scherkus (not reviewing), 2012/07/20 19:20:51]

Are we planning on fixing it upstream?

[Ivan Korotkov, 2012/07/23 18:57:52]

Not sure. This code not used anywhere but in vpxenc console tool and isn't part
of the libvpx interface. I've just exported it to Chrome because there is no
other WebM encoder present in tree.

+#include "third_party/libvpx/source/libvpx/libmkv/EbmlWriter.h"
+#undef private
+}
+
+namespace {
+
+// Number of encoder threads to use.
+const int kNumEncoderThreads = 2;

[scherkus (not reviewing), 2012/07/19 21:46:03]

the rest of the media codebase uses static over anonymous namespaces here +
function below

[Ivan Korotkov, 2012/07/20 11:10:36]

Done.

+
+// Need a fixed size serializer for the track ID. libmkv provides a 64 bit
+// one, but not a 32 bit one.
+void Ebml_SerializeUnsigned32(EbmlGlobal *ebml,

[Nikita (slow), 2012/07/19 08:37:12]

nit: EbmlGlobal* ebml

[scherkus (not reviewing), 2012/07/19 21:46:03]

pointers go with types

EbmlGlobal* embl

[Ivan Korotkov, 2012/07/20 09:09:29]

Done.

+                              unsigned long class_id,
+                              uint64_t value) {
+  unsigned char size_serialized = 4 | 0x80;

[scherkus (not reviewing), 2012/07/19 21:46:03]

prefer uint8 over unsigned keyword

[Ivan Korotkov, 2012/07/20 11:10:36]

Done.

+  Ebml_WriteID(ebml, class_id);
+  Ebml_Serialize(ebml, &size_serialized, sizeof(size_serialized), 1);
+  Ebml_Serialize(ebml, &value, sizeof(value), 4);
+}
+
+}  // namespace
+
+namespace media {
+
+namespace chromeos {
+
+WebmEncoder::WebmEncoder(const FilePath& output_path,
+                         unsigned bitrate,
+                         bool realtime)
+    : bitrate_(bitrate),
+      deadline_(realtime ? VPX_DL_REALTIME : VPX_DL_GOOD_QUALITY),
+      output_path_(output_path) {
+  ebml_writer_.write_cb = base::Bind(

[scherkus (not reviewing), 2012/07/19 21:46:03]

what happens if I use two of these classes at the same time?

[Ivan Korotkov, 2012/07/20 11:10:36]

What's the problem with this? Each instance would pass its own ebml_writer_
context to EbmlXxx functions, so its own callbacks will be called.

+      &WebmEncoder::EbmlWrite, base::Unretained(this));
+  ebml_writer_.serialize_cb = base::Bind(
+      &WebmEncoder::EbmlSerialize, base::Unretained(this));
+}
+
+void WebmEncoder::EncodeFromSprite(const SkBitmap& sprite,
+                                   int fps_n,
+                                   int fps_d) {
+  DCHECK(!sprite.isNull() && !sprite.empty());

[scherkus (not reviewing), 2012/07/19 21:46:03]

nit: separate the DCHECKs

[Ivan Korotkov, 2012/07/23 18:57:52]

Done.

+
+  width_ = sprite.width();
+  height_ = sprite.width();
+  fps_.num = fps_n;
+  fps_.den = fps_d;
+
+  size_t y_size = width_ * height_;
+  size_t uv_size = ((width_ + 1) / 2) * ((height_ + 1) / 2);
+  size_t frame_size = y_size + 2 * uv_size;
+  size_t frame_count = sprite.height() / width_;  // Sprite is tiled vertically.
+
+  yuv_frame_.reset(new uint8[frame_size]);
+
+  vpx_image_t image;
+  vpx_img_wrap(&image, VPX_IMG_FMT_I420, width_, height_, 1, yuv_frame_.get());
+
+  vpx_codec_ctx_t codec;
+  const vpx_codec_iface_t* codec_iface = vpx_codec_vp8_cx();
+  DCHECK(codec_iface);
+  vpx_codec_err_t ret = vpx_codec_enc_config_default(codec_iface, &config_, 0);
+  DCHECK_EQ(VPX_CODEC_OK, ret);
+
+  config_.rc_target_bitrate = bitrate_;
+  config_.g_w = width_;
+  config_.g_h = height_;
+  config_.g_pass = VPX_RC_ONE_PASS;

[Nikita (slow), 2012/07/19 08:37:12]

nit: align comments if possible

[Ivan Korotkov, 2012/07/20 09:09:29]

Done.

+  config_.g_profile = 0;  // Default profile.
+  config_.g_threads = kNumEncoderThreads;
+  config_.rc_min_quantizer = 0;
+  config_.rc_max_quantizer = 63;  // Maximum possible range.
+  config_.g_timebase.num = fps_.den;
+  config_.g_timebase.den = fps_.num;
+  config_.kf_mode = VPX_KF_AUTO;  // Auto key frames.
+
+  ret = vpx_codec_enc_init(&codec, codec_iface, &config_, 0);
+  DCHECK_EQ(VPX_CODEC_OK, ret);

[Nikita (slow), 2012/07/19 08:37:12]

Pass error to upper level if that fails

[Ivan Korotkov, 2012/07/20 09:09:29]

Done.

+
+  SkAutoLockPixels lock_sprite(sprite);
+
+  const uint8* src = reinterpret_cast<const uint8*>(sprite.getAddr32(0, 0));
+  size_t src_frame_size = sprite.getSize();
+  int crop_y = 0;
+
+  WriteWebmHeader();
+
+  for (size_t frame = 0; frame < frame_count; ++frame) {
+    int res = libyuv::ConvertToI420(
+        src, src_frame_size,
+        image.planes[VPX_PLANE_Y], image.stride[VPX_PLANE_Y],
+        image.planes[VPX_PLANE_U], image.stride[VPX_PLANE_U],
+        image.planes[VPX_PLANE_V], image.stride[VPX_PLANE_V],
+        0, crop_y,                // src origin
+        width_, sprite.height(),  // src size
+        width_, height_,          // dest size
+        libyuv::kRotate0,
+        libyuv::FOURCC_ARGB);
+    DCHECK_EQ(0, res);

[Nikita (slow), 2012/07/19 08:37:12]

Pass error to upper level if that fails?

[Ivan Korotkov, 2012/07/20 09:09:29]

Done.

+    crop_y += height_;
+
+    ret = vpx_codec_encode(&codec, &image, frame, 1, 0, deadline_);
+    DCHECK_EQ(VPX_CODEC_OK, ret);

[Nikita (slow), 2012/07/19 08:37:12]

Same here.

[Ivan Korotkov, 2012/07/20 09:09:29]

Done.

+
+    vpx_codec_iter_t iter = NULL;
+    const vpx_codec_cx_pkt_t* packet;
+    while ((packet = vpx_codec_get_cx_data(&codec, &iter))) {
+      if (packet->kind == VPX_CODEC_CX_FRAME_PKT)
+        WriteWebmBlock(packet);
+    }
+  }
+
+  vpx_codec_destroy(&codec);
+
+  WriteWebmFooter();
+}
+
+void WebmEncoder::WriteWebmHeader() {
+  output_ = file_util::OpenFile(output_path_, "wb");
+  DCHECK(output_);
+
+  // Global header.
+  StartSubElement(EBML); {

[Nikita (slow), 2012/07/19 08:37:12]

How about

StartSubElement(EBML);
{
  ...
}
EndSubElement();  

formatting?

[Ivan Korotkov, 2012/07/20 09:09:29]

IMO that looks against Chrome's style. This one follows if/else and do/while
look.

+    Ebml_SerializeUnsigned(&ebml_writer_, EBMLVersion, 1);
+    Ebml_SerializeUnsigned(&ebml_writer_, EBMLReadVersion, 1);
+    Ebml_SerializeUnsigned(&ebml_writer_, EBMLMaxIDLength, 4);
+    Ebml_SerializeUnsigned(&ebml_writer_, EBMLMaxSizeLength, 8);
+    Ebml_SerializeString(&ebml_writer_, DocType, "webm");
+    Ebml_SerializeUnsigned(&ebml_writer_, DocTypeVersion, 2);
+    Ebml_SerializeUnsigned(&ebml_writer_, DocTypeReadVersion, 2);
+  } EndSubElement();  // EBML
+  // Single segment with a video track.

[Nikita (slow), 2012/07/19 08:37:12]

nit: insert extra line

[Ivan Korotkov, 2012/07/20 09:09:29]

Done.

+  StartSubElement(Segment); {

[Nikita (slow), 2012/07/19 08:37:12]

nit: fix formatting as suggested

+    StartSubElement(Info); {
+      // All timecodes in the segment will be expressed in milliseconds.
+      Ebml_SerializeUnsigned(&ebml_writer_, TimecodeScale, 1000000);
+    } EndSubElement();  // Info
+    StartSubElement(Tracks); {

[Nikita (slow), 2012/07/19 08:37:12]

nit: fix formatting as suggested

+      StartSubElement(TrackEntry); {
+        Ebml_SerializeUnsigned(&ebml_writer_, TrackNumber, 1);
+        Ebml_SerializeUnsigned32(&ebml_writer_, TrackUID, 1);
+        Ebml_SerializeUnsigned(&ebml_writer_, TrackType, 1);  // Video
+        Ebml_SerializeString(&ebml_writer_, CodecID, "V_VP8");
+        StartSubElement(Video); {
+          Ebml_SerializeUnsigned(&ebml_writer_, PixelWidth, width_);
+          Ebml_SerializeUnsigned(&ebml_writer_, PixelHeight, height_);
+          Ebml_SerializeUnsigned(&ebml_writer_, StereoMode, 0);  // Mono
+          float fps = static_cast<float>(fps_.num) / fps_.den;
+          Ebml_SerializeFloat(&ebml_writer_, FrameRate, fps);
+        } EndSubElement();  // Video
+      } EndSubElement();  // TrackEntry
+    } EndSubElement();  // Tracks
+    StartSubElement(Cluster); {
+      Ebml_SerializeUnsigned(&ebml_writer_, Timecode, 0);
+    }  // Cluster left open.
+  }  // Segment left open.
+}
+
+void WebmEncoder::WriteWebmBlock(const vpx_codec_cx_pkt_t* packet) {
+  bool is_keyframe = packet->data.frame.flags & VPX_FRAME_IS_KEY;
+  int64_t pts_ms = 1000 * packet->data.frame.pts * fps_.den / fps_.num;
+
+  DVLOG(1) << "Video packet @" << pts_ms << " ms "
+           << packet->data.frame.sz << " bytes "
+           << (is_keyframe ? "K" : "");
+
+  Ebml_WriteID(&ebml_writer_, SimpleBlock);
+
+  unsigned long block_length = (packet->data.frame.sz + 4) | 0x10000000;

[scherkus (not reviewing), 2012/07/19 21:46:03]

what's 0x10000000?

[Ivan Korotkov, 2012/07/20 11:10:36]

I guess these higher bit masks (0x10000000, 0x80, etc) denote the length of the
following field in EBML or something like that. The code comes from libvpx.

+  EbmlSerializeHelper(&block_length, 4);
+
+  unsigned char track_number = 1 | 0x80;

[scherkus (not reviewing), 2012/07/19 21:46:03]

what's 0x80?

+  EbmlSerializeHelper(&track_number, 1);
+
+  EbmlSerializeHelper(&pts_ms, 2);
+
+  unsigned char flags = 0;
+  if (is_keyframe)
+    flags |= 0x80;

[scherkus (not reviewing), 2012/07/19 21:46:03]

are there no constants for flags?

[Ivan Korotkov, 2012/07/20 11:10:36]

libvpx/webrtc use these as is, so I assume not.
What's interesting is that EBML doesn't define anything for bit 7 ("unused"), I
guess using it as keyframe flag is some sort of WebM extension.

+  if (packet->data.frame.flags & VPX_FRAME_IS_INVISIBLE)
+    flags |= 0x08;
+  EbmlSerializeHelper(&flags, 1);
+
+  EbmlWrite(packet->data.frame.buf, packet->data.frame.sz);
+}
+
+void WebmEncoder::WriteWebmFooter() {
+  EndSubElement();  // Cluster
+  EndSubElement();  // Segment
+  DCHECK(ebml_sub_elements_.empty());
+  file_util::CloseFile(output_);
+}
+
+void WebmEncoder::StartSubElement(unsigned long class_id) {
+  Ebml_WriteID(&ebml_writer_, class_id);
+  ebml_sub_elements_.push(ftell(output_));
+  uint64_t unknown_len = 0x01FFFFFFFFFFFFFFLLU;

[scherkus (not reviewing), 2012/07/19 21:46:03]

no constant?

[Ivan Korotkov, 2012/07/20 11:10:36]

Same here. I think it's "-1" EBML-encoded. Made it static const.

+  Ebml_Serialize(&ebml_writer_, &unknown_len, sizeof(unknown_len), 8);
+}
+
+void WebmEncoder::EndSubElement() {
+  DCHECK(!ebml_sub_elements_.empty());
+
+  long int end_pos = ftell(output_);
+  long int start_pos = ebml_sub_elements_.top();
+  ebml_sub_elements_.pop();
+
+  uint64_t size = (end_pos - start_pos - 8) | 0x0100000000000000ULL;
+  // Seek to the beginning of the sub-element and patch in the calculated size.
+  fseek(output_, start_pos, SEEK_SET);

[Nikita (slow), 2012/07/19 08:37:12]

Check return value?

[Ivan Korotkov, 2012/07/20 09:09:29]

Done.

+  EbmlSerializeHelper(&size, 8);
+
+  // Restore write position.
+  fseek(output_, end_pos, SEEK_SET);

[Nikita (slow), 2012/07/19 08:37:12]

Check return value?

[Ivan Korotkov, 2012/07/20 09:09:29]

Done.

+}
+
+void WebmEncoder::EbmlWrite(const void* buffer,
+                            unsigned long len) {
+  unsigned long ret = fwrite(buffer, 1, len, output_);
+  CHECK_EQ(len, ret);

[Nikita (slow), 2012/07/19 08:37:12]

Not sure that it makes sense to have a CHECK here. How about if encoding
operation fails let caller know about it and handle on that level.

[Ivan Korotkov, 2012/07/20 09:09:29]

It's difficult to do with EbmlXxx interface. Maybe it makes sense to simply
LOG(ERROR) in this case, because it's pretty rare (disk full, for example).

+}
+
+template <class T>
+void WebmEncoder::EbmlSerializeHelper(const T* buffer, unsigned long len) {
+  for (unsigned long i = len; i-- > 0; ) {

[scherkus (not reviewing), 2012/07/19 21:46:03]

nit: let's try to keep conditions as they are and put the iterator modifying
stuff in the third clause

for (i = len - 1; i >= 0; --i) {
  // ...
}

[Ivan Korotkov, 2012/07/20 11:10:36]

Ok, I guess we can use a signed iterator since len is anyways <= 8.

+    char c = *buffer >> (i * CHAR_BIT);
+    EbmlWrite(&c, 1);
+  }
+}
+
+void WebmEncoder::EbmlSerialize(const void* buffer,
+                                int buffer_size,
+                                unsigned long len) {
+  switch (buffer_size) {
+    case 1:
+      return EbmlSerializeHelper(static_cast<const int8_t*>(buffer), len);
+    case 2:
+      return EbmlSerializeHelper(static_cast<const int16_t*>(buffer), len);
+    case 4:
+      return EbmlSerializeHelper(static_cast<const int32_t*>(buffer), len);
+    case 8:
+      return EbmlSerializeHelper(static_cast<const int64_t*>(buffer), len);
+    default:
+      NOTREACHED();

[Nikita (slow), 2012/07/19 08:37:12]

nit: Add some error message?

[Ivan Korotkov, 2012/07/20 09:09:29]

Done.

+  }
+}
+
+}  // namespace chromeos
+
+}  // namespace media