Add an IVF parser

The IVF parser would be shared among VP8 and VP9 unittests.

BUG=509500

Review URL: https://codereview.chromium.org/1269473002

Cr-Commit-Position: refs/heads/master@{#342341}

7 files changed, 270 insertions, 67 deletions

diff --git a/content/common/gpu/media/video_encode_accelerator_unittest.cc b/content/common/gpu/media/video_encode_accelerator_unittest.cc
index 0e1e863..caf41c8 100644
--- a/content/common/gpu/media/video_encode_accelerator_unittest.cc
+++ b/content/common/gpu/media/video_encode_accelerator_unittest.cc
@@ -17,7 +17,6 @@
 #include "base/strings/string_number_conversions.h"
 #include "base/strings/string_split.h"
 #include "base/strings/stringprintf.h"
-#include "base/sys_byteorder.h"
 #include "base/threading/thread.h"
 #include "base/threading/thread_checker.h"
 #include "base/time/time.h"
@@ -27,6 +26,7 @@
 #include "media/base/bitstream_buffer.h"
 #include "media/base/test_data_util.h"
 #include "media/filters/h264_parser.h"
+#include "media/filters/ivf_parser.h"
 #include "media/video/fake_video_encode_accelerator.h"
 #include "media/video/video_encode_accelerator.h"
 #include "testing/gtest/include/gtest/gtest.h"
@@ -114,26 +114,6 @@ bool g_fake_encoder = false;
 class VideoEncodeAcceleratorTestEnvironment;
 VideoEncodeAcceleratorTestEnvironment* g_env;
 
-struct IvfFileHeader {
-  char signature[4];     // signature: 'DKIF'
-  uint16_t version;      // version (should be 0)
-  uint16_t header_size;  // size of header in bytes
-  uint32_t fourcc;       // codec FourCC (e.g., 'VP80')
-  uint16_t width;        // width in pixels
-  uint16_t height;       // height in pixels
-  uint32_t framerate;    // frame rate per seconds
-  uint32_t timescale;    // time scale. For example, if framerate is 30 and
-                         // timescale is 2, the unit of IvfFrameHeader.timestamp
-                         // is 2/30 seconds.
-  uint32_t num_frames;   // number of frames in file
-  uint32_t unused;       // unused
-} __attribute__((packed));
-
-struct IvfFrameHeader {
-  uint32_t frame_size;  // Size of frame in bytes (not including the header)
-  uint64_t timestamp;   // 64-bit presentation timestamp
-} __attribute__((packed));
-
 // The number of frames to be encoded. This variable is set by the switch
 // "--num_frames_to_encode". Ignored if 0.
 int g_num_frames_to_encode = 0;
@@ -1285,23 +1265,21 @@ void VEAClient::VerifyStreamProperties() {
 }
 
 void VEAClient::WriteIvfFileHeader() {
-  IvfFileHeader header;
+  media::IvfFileHeader header = {};
 
-  memset(&header, 0, sizeof(header));
-  header.signature[0] = 'D';
-  header.signature[1] = 'K';
-  header.signature[2] = 'I';
-  header.signature[3] = 'F';
+  memcpy(header.signature, media::kIvfHeaderSignature,
+         sizeof(header.signature));
   header.version = 0;
-  header.header_size = base::ByteSwapToLE16(sizeof(header));
-  header.fourcc = base::ByteSwapToLE32(0x30385056);  // VP80
-  header.width = base::ByteSwapToLE16(
-      base::checked_cast<uint16_t>(test_stream_->visible_size.width()));
-  header.height = base::ByteSwapToLE16(
-      base::checked_cast<uint16_t>(test_stream_->visible_size.height()));
-  header.framerate = base::ByteSwapToLE32(requested_framerate_);
-  header.timescale = base::ByteSwapToLE32(1);
-  header.num_frames = base::ByteSwapToLE32(num_frames_to_encode_);
+  header.header_size = sizeof(header);
+  header.fourcc = 0x30385056;  // VP80
+  header.width =
+      base::checked_cast<uint16_t>(test_stream_->visible_size.width());
+  header.height =
+      base::checked_cast<uint16_t>(test_stream_->visible_size.height());
+  header.timebase_denum = requested_framerate_;
+  header.timebase_num = 1;
+  header.num_frames = num_frames_to_encode_;
+  header.ByteSwap();
 
   EXPECT_TRUE(base::AppendToFile(
       base::FilePath::FromUTF8Unsafe(test_stream_->out_filename),
@@ -1309,11 +1287,11 @@ void VEAClient::WriteIvfFileHeader() {
 }
 
 void VEAClient::WriteIvfFrameHeader(int frame_index, size_t frame_size) {
-  IvfFrameHeader header;
+  media::IvfFrameHeader header = {};
 
-  memset(&header, 0, sizeof(header));
-  header.frame_size = base::ByteSwapToLE32(frame_size);
-  header.timestamp = base::ByteSwapToLE64(frame_index);
+  header.frame_size = frame_size;
+  header.timestamp = frame_index;
+  header.ByteSwap();
   EXPECT_TRUE(base::AppendToFile(
       base::FilePath::FromUTF8Unsafe(test_stream_->out_filename),
       reinterpret_cast<char*>(&header), sizeof(header)));
diff --git a/media/BUILD.gn b/media/BUILD.gn
index 744dab9..6861395 100644
--- a/media/BUILD.gn
+++ b/media/BUILD.gn
@@ -200,6 +200,8 @@ component("media") {
     "filters/h264_bit_reader.h",
     "filters/h264_parser.cc",
     "filters/h264_parser.h",
+    "filters/ivf_parser.cc",
+    "filters/ivf_parser.h",
     "filters/jpeg_parser.cc",
     "filters/jpeg_parser.h",
     "filters/source_buffer_platform.cc",
@@ -559,6 +561,7 @@ test("media_unittests") {
     "filters/frame_processor_unittest.cc",
     "filters/h264_bit_reader_unittest.cc",
     "filters/h264_parser_unittest.cc",
+    "filters/ivf_parser_unittest.cc",
     "filters/jpeg_parser_unittest.cc",
     "filters/source_buffer_stream_unittest.cc",
     "filters/video_cadence_estimator_unittest.cc",
diff --git a/media/filters/ivf_parser.cc b/media/filters/ivf_parser.cc
new file mode 100644
index 0000000..b6160fb
--- /dev/null
+++ b/media/filters/ivf_parser.cc
@@ -0,0 +1,89 @@
+// 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 "base/logging.h"
+#include "base/sys_byteorder.h"
+#include "media/filters/ivf_parser.h"
+
+namespace media {
+
+void IvfFileHeader::ByteSwap() {
+  version = base::ByteSwapToLE16(version);
+  header_size = base::ByteSwapToLE16(header_size);
+  fourcc = base::ByteSwapToLE32(fourcc);
+  width = base::ByteSwapToLE16(width);
+  height = base::ByteSwapToLE16(height);
+  timebase_denum = base::ByteSwapToLE32(timebase_denum);
+  timebase_num = base::ByteSwapToLE32(timebase_num);
+  num_frames = base::ByteSwapToLE32(num_frames);
+  unused = base::ByteSwapToLE32(unused);
+}
+
+void IvfFrameHeader::ByteSwap() {
+  frame_size = base::ByteSwapToLE32(frame_size);
+  timestamp = base::ByteSwapToLE64(timestamp);
+}
+
+IvfParser::IvfParser() : ptr_(nullptr), end_(nullptr) {}
+
+bool IvfParser::Initialize(const uint8_t* stream,
+                           size_t size,
+                           IvfFileHeader* file_header) {
+  DCHECK(stream);
+  DCHECK(file_header);
+  ptr_ = stream;
+  end_ = stream + size;
+
+  if (size < sizeof(IvfFileHeader)) {
+    DLOG(ERROR) << "EOF before file header";
+    return false;
+  }
+
+  memcpy(file_header, ptr_, sizeof(IvfFileHeader));
+  file_header->ByteSwap();
+
+  if (memcmp(file_header->signature, kIvfHeaderSignature,
+             sizeof(file_header->signature)) != 0) {
+    DLOG(ERROR) << "IVF signature mismatch";
+    return false;
+  }
+  DLOG_IF(WARNING, file_header->version != 0)
+      << "IVF version unknown: " << file_header->version
+      << ", the parser may not be able to parse correctly";
+  if (file_header->header_size != sizeof(IvfFileHeader)) {
+    DLOG(ERROR) << "IVF file header size mismatch";
+    return false;
+  }
+
+  ptr_ += sizeof(IvfFileHeader);
+
+  return true;
+}
+
+bool IvfParser::ParseNextFrame(IvfFrameHeader* frame_header,
+                               const uint8_t** payload) {
+  DCHECK(ptr_);
+  DCHECK(payload);
+
+  if (end_ < ptr_ + sizeof(IvfFrameHeader)) {
+    DLOG_IF(ERROR, ptr_ != end_) << "Incomplete frame header";
+    return false;
+  }
+
+  memcpy(frame_header, ptr_, sizeof(IvfFrameHeader));
+  frame_header->ByteSwap();
+  ptr_ += sizeof(IvfFrameHeader);
+
+  if (end_ < ptr_ + frame_header->frame_size) {
+    DLOG(ERROR) << "Not enough frame data";
+    return false;
+  }
+
+  *payload = ptr_;
+  ptr_ += frame_header->frame_size;
+
+  return true;
+}
+
+}  // namespace media
diff --git a/media/filters/ivf_parser.h b/media/filters/ivf_parser.h
new file mode 100644
index 0000000..fe234aa
--- /dev/null
+++ b/media/filters/ivf_parser.h
@@ -0,0 +1,86 @@
+// 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.
+
+#ifndef MEDIA_FILTERS_IVF_PARSER_H_
+#define MEDIA_FILTERS_IVF_PARSER_H_
+
+#include <stddef.h>
+#include <stdint.h>
+
+#include "base/macros.h"
+#include "media/base/media_export.h"
+
+namespace media {
+
+const char kIvfHeaderSignature[] = "DKIF";
+
+#pragma pack(push, 1)
+struct MEDIA_EXPORT IvfFileHeader {
+  // Byte swap interger fields between native and (on disk) little endian.
+  void ByteSwap();
+
+  char signature[4];        // signature: 'DKIF'
+  uint16_t version;         // version (should be 0)
+  uint16_t header_size;     // size of header in bytes
+  uint32_t fourcc;          // codec FourCC (e.g., 'VP80')
+  uint16_t width;           // width in pixels
+  uint16_t height;          // height in pixels
+  uint32_t timebase_denum;  // timebase denumerator
+  uint32_t timebase_num;    // timebase numerator. For example, if
+                            // timebase_denum is 30 and timebase_num is 2, the
+                            // unit of IvfFrameHeader.timestamp is 2/30
+                            // seconds.
+  uint32_t num_frames;      // number of frames in file
+  uint32_t unused;          // unused
+};
+static_assert(
+    sizeof(IvfFileHeader) == 32,
+    "sizeof(IvfFileHeader) must be fixed since it will be used with file IO");
+
+struct MEDIA_EXPORT IvfFrameHeader {
+  // Byte swap interger fields between native and (on disk) little endian.
+  void ByteSwap();
+
+  uint32_t frame_size;  // Size of frame in bytes (not including the header)
+  uint64_t timestamp;   // 64-bit presentation timestamp in unit timebase,
+                        // which is defined in IvfFileHeader.
+};
+static_assert(
+    sizeof(IvfFrameHeader) == 12,
+    "sizeof(IvfFrameHeader) must be fixed since it will be used with file IO");
+#pragma pack(pop)
+
+// IVF is a simple container format for video frame. It is used by libvpx to
+// transport VP8 and VP9 bitstream.
+class MEDIA_EXPORT IvfParser {
+ public:
+  IvfParser();
+
+  // Initializes the parser for IVF |stream| with size |size| and parses the
+  // file header. Returns true on success.
+  bool Initialize(const uint8_t* stream,
+                  size_t size,
+                  IvfFileHeader* file_header);
+
+  // Parses the next frame. Returns true if the next frame is parsed without
+  // error. |frame_header| will be filled with the frame header and |payload|
+  // will point to frame payload (inside the |stream| buffer given to
+  // Initialize.)
+  bool ParseNextFrame(IvfFrameHeader* frame_header, const uint8_t** payload);
+
+ private:
+  bool ParseFileHeader(IvfFileHeader* file_header);
+
+  // Current reading position of input stream.
+  const uint8_t* ptr_;
+
+  // The end position of input stream.
+  const uint8_t* end_;
+
+  DISALLOW_COPY_AND_ASSIGN(IvfParser);
+};
+
+}  // namespace media
+
+#endif  // MEDIA_FILTERS_IVF_PARSER_H_
diff --git a/media/filters/ivf_parser_unittest.cc b/media/filters/ivf_parser_unittest.cc
new file mode 100644
index 0000000..2d89dc2
--- /dev/null
+++ b/media/filters/ivf_parser_unittest.cc
@@ -0,0 +1,56 @@
+// 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 "base/files/memory_mapped_file.h"
+#include "media/base/test_data_util.h"
+#include "media/filters/ivf_parser.h"
+#include "testing/gtest/include/gtest/gtest.h"
+
+namespace media {
+
+TEST(IvfParserTest, StreamFileParsing) {
+  base::FilePath file_path = GetTestDataFilePath("test-25fps.vp8");
+
+  base::MemoryMappedFile stream;
+  ASSERT_TRUE(stream.Initialize(file_path)) << "Couldn't open stream file: "
+                                            << file_path.MaybeAsASCII();
+
+  IvfParser parser;
+  IvfFileHeader file_header = {};
+
+  EXPECT_TRUE(parser.Initialize(stream.data(), stream.length(), &file_header));
+
+  // Check file header fields.
+  EXPECT_EQ(0, memcmp(file_header.signature, kIvfHeaderSignature,
+                      sizeof(file_header.signature)));
+  EXPECT_EQ(0, file_header.version);
+  EXPECT_EQ(sizeof(IvfFileHeader), file_header.header_size);
+  EXPECT_EQ(0x30385056u, file_header.fourcc);  // VP80
+  EXPECT_EQ(320u, file_header.width);
+  EXPECT_EQ(240u, file_header.height);
+  EXPECT_EQ(50u, file_header.timebase_denum);
+  EXPECT_EQ(2u, file_header.timebase_num);
+  EXPECT_EQ(250u, file_header.num_frames);
+
+  IvfFrameHeader frame_header;
+  size_t num_parsed_frames = 0;
+  const uint8_t* payload = nullptr;
+  while (parser.ParseNextFrame(&frame_header, &payload)) {
+    ++num_parsed_frames;
+    EXPECT_TRUE(payload != nullptr);
+
+    // Only check the first frame.
+    if (num_parsed_frames == 1u) {
+      EXPECT_EQ(14788u, frame_header.frame_size);
+      EXPECT_EQ(0u, frame_header.timestamp);
+      EXPECT_EQ(
+          static_cast<ptrdiff_t>(sizeof(file_header) + sizeof(frame_header)),
+          payload - stream.data());
+    }
+  }
+
+  EXPECT_EQ(file_header.num_frames, num_parsed_frames);
+}
+
+}  // namespace media
diff --git a/media/filters/vp8_parser_unittest.cc b/media/filters/vp8_parser_unittest.cc
index 39a2a80..3087ed0 100644
--- a/media/filters/vp8_parser_unittest.cc
+++ b/media/filters/vp8_parser_unittest.cc
@@ -2,12 +2,10 @@
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
-#include "base/command_line.h"
 #include "base/files/memory_mapped_file.h"
 #include "base/logging.h"
-#include "base/strings/string_number_conversions.h"
-#include "base/sys_byteorder.h"
 #include "media/base/test_data_util.h"
+#include "media/filters/ivf_parser.h"
 #include "media/filters/vp8_parser.h"
 #include "testing/gtest/include/gtest/gtest.h"
 
@@ -15,45 +13,35 @@ namespace media {
 
 TEST(Vp8ParserTest, StreamFileParsing) {
   base::FilePath file_path = GetTestDataFilePath("test-25fps.vp8");
-  // Number of frames in the test stream to be parsed.
-  const int num_frames = 250;
-
   base::MemoryMappedFile stream;
   ASSERT_TRUE(stream.Initialize(file_path))
       << "Couldn't open stream file: " << file_path.MaybeAsASCII();
 
-  Vp8Parser parser;
+  IvfParser ivf_parser;
+  IvfFileHeader ivf_file_header = {};
+  ASSERT_TRUE(
+      ivf_parser.Initialize(stream.data(), stream.length(), &ivf_file_header));
+  ASSERT_EQ(ivf_file_header.fourcc, 0x30385056u);  // VP80
+
+  Vp8Parser vp8_parser;
+  IvfFrameHeader ivf_frame_header = {};
+  size_t num_parsed_frames = 0;
 
   // Parse until the end of stream/unsupported stream/error in stream is found.
-  int num_parsed_frames = 0;
-  const uint8_t* stream_ptr = stream.data();
-  size_t bytes_left = stream.length();
-  // Skip IVF file header.
-  const size_t kIvfStreamHeaderLen = 32;
-  CHECK_GE(bytes_left, kIvfStreamHeaderLen);
-  stream_ptr += kIvfStreamHeaderLen;
-  bytes_left -= kIvfStreamHeaderLen;
-
-  const size_t kIvfFrameHeaderLen = 12;
-  while (bytes_left > kIvfFrameHeaderLen) {
+  const uint8_t* payload = nullptr;
+  while (ivf_parser.ParseNextFrame(&ivf_frame_header, &payload)) {
     Vp8FrameHeader fhdr;
-    uint32_t frame_size =
-        base::ByteSwapToLE32(*reinterpret_cast<const uint32_t*>(stream_ptr));
-    // Skip IVF frame header.
-    stream_ptr += kIvfFrameHeaderLen;
-    bytes_left -= kIvfFrameHeaderLen;
 
-    ASSERT_TRUE(parser.ParseFrame(stream_ptr, frame_size, &fhdr));
+    ASSERT_TRUE(
+        vp8_parser.ParseFrame(payload, ivf_frame_header.frame_size, &fhdr));
 
-    stream_ptr += frame_size;
-    bytes_left -= frame_size;
     ++num_parsed_frames;
   }
 
   DVLOG(1) << "Number of successfully parsed frames before EOS: "
            << num_parsed_frames;
 
-  EXPECT_EQ(num_frames, num_parsed_frames);
+  EXPECT_EQ(ivf_file_header.num_frames, num_parsed_frames);
 }
 
 }  // namespace media
diff --git a/media/media.gyp b/media/media.gyp
index 667c39d..a6bfe01 100644
--- a/media/media.gyp
+++ b/media/media.gyp
@@ -539,6 +539,8 @@
         'filters/h264_bit_reader.h',
         'filters/h264_parser.cc',
         'filters/h264_parser.h',
+        'filters/ivf_parser.cc',
+        'filters/ivf_parser.h',
         'filters/in_memory_url_protocol.cc',
         'filters/in_memory_url_protocol.h',
         'filters/jpeg_parser.cc',
@@ -1254,6 +1256,7 @@
         'filters/h264_bit_reader_unittest.cc',
         'filters/h264_parser_unittest.cc',
         'filters/in_memory_url_protocol_unittest.cc',
+        'filters/ivf_parser_unittest.cc',
         'filters/jpeg_parser_unittest.cc',
         'filters/source_buffer_stream_unittest.cc',
         'filters/video_cadence_estimator_unittest.cc',