Parse VP9 compressed header

media/filters/vp9_parser.h

 // 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.
 //
 // This file contains an implementation of a VP9 bitstream parser. The main
 // purpose of this parser is to support hardware decode acceleration. Some
 // accelerators, e.g. libva which implements VA-API, require the caller
 // (chrome) to feed them parsed VP9 frame header.
 //
 // See media::VP9Decoder for example usage.
 //
 #ifndef MEDIA_FILTERS_VP9_PARSER_H_
 #define MEDIA_FILTERS_VP9_PARSER_H_
 
 #include <stddef.h>
 #include <stdint.h>
 #include <sys/types.h>
 
 #include <deque>
 
+#include "base/callback.h"
 #include "base/macros.h"
+#include "base/memory/weak_ptr.h"
 #include "media/base/media_export.h"
-#include "media/filters/vp9_raw_bits_reader.h"
 
 namespace media {
 
 const int kVp9MaxProfile = 4;
 const int kVp9NumRefFramesLog2 = 3;
 const size_t kVp9NumRefFrames = 1 << kVp9NumRefFramesLog2;
 const uint8_t kVp9MaxProb = 255;
 const size_t kVp9NumRefsPerFrame = 3;
+const int kVp9NumFrameContext = 4;

[Pawel Osciak, 2016/08/04 10:20:19]

size_t ?

s/Context/Contexts/ ?

[kcwu, 2016/08/05 11:38:47]

Done.

+
+typedef uint8_t Vp9Prob;

[Pawel Osciak, 2016/08/04 10:20:19]

Nit: perhaps "using" syntax?
Perhaps move to into the context class?

[kcwu, 2016/08/05 11:38:47]

It's not only used in context class.

 
 enum class Vp9ColorSpace {
   UNKNOWN = 0,
   BT_601 = 1,
   BT_709 = 2,
   SMPTE_170 = 3,
   SMPTE_240 = 4,
   BT_2020 = 5,
   RESERVED = 6,
   SRGB = 7,
 };
 
-enum Vp9InterpFilter {
+enum Vp9InterpolationFilter {
   EIGHTTAP = 0,
   EIGHTTAP_SMOOTH = 1,
   EIGHTTAP_SHARP = 2,
   BILINEAR = 3,
-  SWICHABLE = 4,
+  SWITCHABLE = 4,
 };
 
-struct MEDIA_EXPORT Vp9Segmentation {
+enum Vp9FrameType {

[Pawel Osciak, 2016/08/04 10:20:19]

Perhaps RefType? We already have FrameType in header, so it'd be a bit
confusing, and spec uses FrameType for keyframe/non-keyframe.

[kcwu, 2016/08/05 11:38:48]

Done.

+  VP9_FRAME_INTRA = 0,
+  VP9_FRAME_LAST = 1,
+  VP9_FRAME_GOLDEN = 2,
+  VP9_FRAME_ALTREF = 3,
+  VP9_FRAME_MAX = 4,
+};
+
+enum Vp9TxMode {

[Pawel Osciak, 2016/08/04 10:20:19]

Perhaps move to Vp9Compressed header?

[kcwu, 2016/08/05 11:38:48]

Done.

+  ONLY_4X4 = 0,
+  ALLOW_8X8 = 1,
+  ALLOW_16X16 = 2,
+  ALLOW_32X32 = 3,
+  TX_MODE_SELECT = 4,
+  TX_MODES = 5,
+};
+
+enum Vp9TxSize {

[Pawel Osciak, 2016/08/04 10:20:19]

Perhaps we should move this to ReadCoefProbs()?

[kcwu, 2016/08/05 11:38:47]

Done.

+  TX_4X4 = 0,
+  TX_8X8 = 1,
+  TX_16X16 = 2,
+  TX_32X32 = 3,
+  TX_SIZES = 4,
+};
+
+enum Vp9ReferenceMode {
+  SINGLE_REFERENCE = 0,
+  COMPOUND_REFERENCE = 1,
+  REFERENCE_MODE_SELECT = 2,
+};
+
+struct MEDIA_EXPORT Vp9SegmentationParams {
   static const size_t kNumSegments = 8;
   static const size_t kNumTreeProbs = kNumSegments - 1;
   static const size_t kNumPredictionProbs = 3;
   enum SegmentLevelFeature {
     SEG_LVL_ALT_Q = 0,
     SEG_LVL_ALT_LF = 1,
     SEG_LVL_REF_FRAME = 2,
     SEG_LVL_SKIP = 3,
     SEG_LVL_MAX
   };
 
   bool enabled;
 
   bool update_map;
   uint8_t tree_probs[kNumTreeProbs];
   bool temporal_update;
   uint8_t pred_probs[kNumPredictionProbs];
 
   bool update_data;
-  bool abs_delta;
+  bool abs_or_delta_update;
   bool feature_enabled[kNumSegments][SEG_LVL_MAX];
   int16_t feature_data[kNumSegments][SEG_LVL_MAX];
 
   int16_t y_dequant[kNumSegments][2];
   int16_t uv_dequant[kNumSegments][2];
 
   bool FeatureEnabled(size_t seg_id, SegmentLevelFeature feature) const {
     return feature_enabled[seg_id][feature];
   }
 
   int16_t FeatureData(size_t seg_id, SegmentLevelFeature feature) const {
     return feature_data[seg_id][feature];
   }
 };
 
-struct MEDIA_EXPORT Vp9LoopFilter {
-  enum Vp9FrameType {
-    VP9_FRAME_INTRA = 0,
-    VP9_FRAME_LAST = 1,
-    VP9_FRAME_GOLDEN = 2,
-    VP9_FRAME_ALTREF = 3,
-    VP9_FRAME_MAX = 4,
-  };
-
+struct MEDIA_EXPORT Vp9LoopFilterParams {
   static const size_t kNumModeDeltas = 2;
 
-  uint8_t filter_level;
-  uint8_t sharpness_level;
+  uint8_t level;
+  uint8_t sharpness;
 
-  bool mode_ref_delta_enabled;
-  bool mode_ref_delta_update;
+  bool delta_enabled;
+  bool delta_update;
   bool update_ref_deltas[VP9_FRAME_MAX];
   int8_t ref_deltas[VP9_FRAME_MAX];
   bool update_mode_deltas[kNumModeDeltas];
   int8_t mode_deltas[kNumModeDeltas];
 
-  uint8_t lvl[Vp9Segmentation::kNumSegments][VP9_FRAME_MAX][kNumModeDeltas];
+  // calculated from above fields

[Pawel Osciak, 2016/08/04 10:20:19]

Nit: to match the style, please capitalize first letters and add dots at end of
sentences.

[kcwu, 2016/08/05 11:38:48]

Done.

+  uint8_t lvl[Vp9SegmentationParams::kNumSegments][VP9_FRAME_MAX]
+             [kNumModeDeltas];
 };
 
 // Members of Vp9FrameHeader will be 0-initialized by Vp9Parser::ParseNextFrame.
 struct MEDIA_EXPORT Vp9QuantizationParams {
   bool IsLossless() const {
-    return base_qindex == 0 && y_dc_delta == 0 && uv_dc_delta == 0 &&
-           uv_ac_delta == 0;
+    return base_q_idx == 0 && delta_q_y_dc == 0 && delta_q_uv_dc == 0 &&
+           delta_q_uv_ac == 0;
   }
 
-  uint8_t base_qindex;
-  int8_t y_dc_delta;
-  int8_t uv_dc_delta;
-  int8_t uv_ac_delta;
+  uint8_t base_q_idx;
+  int8_t delta_q_y_dc;
+  int8_t delta_q_uv_dc;
+  int8_t delta_q_uv_ac;
+};
+
+// Entropy context for frame parsing
+struct MEDIA_EXPORT Vp9FrameContext {
+  Vp9Prob tx_probs_8x8[2][1];
+  Vp9Prob tx_probs_16x16[2][2];
+  Vp9Prob tx_probs_32x32[2][3];
+
+  Vp9Prob coef_probs[4][2][2][6][6][3];
+  Vp9Prob skip_prob[3];
+  Vp9Prob inter_mode_probs[7][3];
+  Vp9Prob interp_filter_probs[4][2];
+  Vp9Prob is_inter_prob[4];
+
+  Vp9Prob comp_mode_prob[5];
+  Vp9Prob single_ref_prob[5][2];
+  Vp9Prob comp_ref_prob[5];
+
+  Vp9Prob y_mode_probs[4][9];
+  Vp9Prob uv_mode_probs[10][9];
+  Vp9Prob partition_probs[16][3];
+
+  Vp9Prob mv_joint_probs[3];
+  Vp9Prob mv_sign_prob[2];
+  Vp9Prob mv_class_probs[2][10];
+  Vp9Prob mv_class0_bit_prob[2];
+  Vp9Prob mv_bits_prob[2][10];
+  Vp9Prob mv_class0_fr_probs[2][2][3];
+  Vp9Prob mv_fr_probs[2][3];
+  Vp9Prob mv_class0_hp_prob[2];
+  Vp9Prob mv_hp_prob[2];
+};
+
+struct MEDIA_EXPORT Vp9CompressedHeader {
+  Vp9TxMode tx_mode;
+  Vp9ReferenceMode reference_mode;
 };
 
 // VP9 frame header.
 struct MEDIA_EXPORT Vp9FrameHeader {
   enum FrameType {
     KEYFRAME = 0,
     INTERFRAME = 1,
   };
 
   bool IsKeyframe() const;
-  bool RefreshFlag(size_t i) const { return !!(refresh_flags & (1u << i)); }
+  bool IsIntra() const;
+  bool RefreshFlag(size_t i) const {
+    return !!(refresh_frame_flags & (1u << i));
+  }
 
   uint8_t profile;
 
   bool show_existing_frame;
-  uint8_t frame_to_show;
+  uint8_t frame_to_show_map_idx;
 
   FrameType frame_type;
 
   bool show_frame;
   bool error_resilient_mode;
 
   uint8_t bit_depth;
   Vp9ColorSpace color_space;
-  bool yuv_range;
+  bool color_range;
   uint8_t subsampling_x;
   uint8_t subsampling_y;
 
-  // The range of width and height is 1..2^16.
-  uint32_t width;
-  uint32_t height;
-  uint32_t display_width;
-  uint32_t display_height;
+  // The range of frame_width and frame_height is 1..2^16.
+  uint32_t frame_width;
+  uint32_t frame_height;
+  uint32_t render_width;
+  uint32_t render_height;
 
   bool intra_only;
-  uint8_t reset_context;
-  uint8_t refresh_flags;
-  uint8_t frame_refs[kVp9NumRefsPerFrame];
-  bool ref_sign_biases[kVp9NumRefsPerFrame];
+  uint8_t reset_frame_context;
+  uint8_t refresh_frame_flags;
+  uint8_t ref_frame_idx[kVp9NumRefsPerFrame];
+  bool ref_frame_sign_bias[Vp9FrameType::VP9_FRAME_MAX];
   bool allow_high_precision_mv;
-  Vp9InterpFilter interp_filter;
+  Vp9InterpolationFilter interpolation_filter;
 
   bool refresh_frame_context;
   bool frame_parallel_decoding_mode;
   uint8_t frame_context_idx;
+  // |frame_context_idx| may change later, backup the original value parsed
+  // from the stream.
+  uint8_t frame_context_idx_backup;

[Pawel Osciak, 2016/08/04 10:20:19]

Perhaps frame_context_idx and frame_context_idx_to_save_probs?

We should probably add a comment saying "frame_context_idx_to_save_probs is to
be used by save_probs() only, and frame_context_idx otherwise".

[kcwu, 2016/08/05 11:38:48]

Done.
However, it seem look odd for vaapi_vda.

 
   Vp9QuantizationParams quant_params;
 
-  uint8_t log2_tile_cols;
-  uint8_t log2_tile_rows;
+  uint8_t tile_cols_log2;
+  uint8_t tile_rows_log2;
 
   // Pointer to the beginning of frame data. It is a responsibility of the
   // client of the Vp9Parser to maintain validity of this data while it is
   // being used outside of that class.
   const uint8_t* data;
 
   // Size of |data| in bytes.
   size_t frame_size;
 
   // Size of compressed header in bytes.
-  size_t first_partition_size;
+  size_t header_size_in_bytes;
 
   // Size of uncompressed header in bytes.
   size_t uncompressed_header_size;
+
+  Vp9CompressedHeader compressed_header;
+  // Initial frame entropy context after load_probs2(frame_context_idx).
+  Vp9FrameContext initial_frame_context;
+  // Current frame entropy context after header parsing.
+  Vp9FrameContext frame_context;
+};
+
+class Vp9FrameContextManager {
+ public:
+  using ContextRefreshCallback = base::Callback<void(const Vp9FrameContext&)>;

[Pawel Osciak, 2016/08/04 10:20:19]

Please document.

[kcwu, 2016/08/05 11:38:48]

Done.

+
+  static bool IsValidFrameContext(const Vp9FrameContext& context);
+
+  Vp9FrameContextManager();
+  ~Vp9FrameContextManager();
+  bool initialized() const { return initialized_; }
+  bool need_update() const { return need_update_; }

[Pawel Osciak, 2016/08/04 10:20:19]

s/need/needs/

[kcwu, 2016/08/05 11:38:48]

Done.

+  const Vp9FrameContext& frame_context() const;
+
+  void Reset();

[Pawel Osciak, 2016/08/04 10:20:19]

Please document methods.

[kcwu, 2016/08/05 11:38:48]

Done.

+  ContextRefreshCallback SetNeedUpdate();
+  void Update(const Vp9FrameContext& frame_context);
+
+ private:
+  bool initialized_;
+  bool need_update_;
+  Vp9FrameContext frame_context_;
+
+  base::WeakPtrFactory<Vp9FrameContextManager> weak_ptr_factory_;
 };
 
 // A parser for VP9 bitstream.
 class MEDIA_EXPORT Vp9Parser {
  public:
   // ParseNextFrame() return values. See documentation for ParseNextFrame().
   enum Result {
     kOk,
     kInvalidStream,
     kEOStream,
+    kAwaitingRefresh,
   };
 
-  Vp9Parser();
+  // The parsing context to keep track of references.
+  struct ReferenceSlot {
+    bool initialized;
+    uint32_t frame_width;
+    uint32_t frame_height;
+    uint8_t subsampling_x;
+    uint8_t subsampling_y;
+    uint8_t bit_depth;
+
+    // More fields for consistency checking.

[Pawel Osciak, 2016/08/04 10:20:19]

I think color_space we also use to copy to frame_hdr?

[kcwu, 2016/08/05 11:38:48]

Yes, but it is not mentioned in spec 8.10. I list them separately according to
8.10. Perhaps the spec forgot to mention it?

[Pawel Osciak, 2016/08/08 08:13:36]

Acknowledged.

+    uint8_t profile;
+    Vp9ColorSpace color_space;
+  };
+
+  // The parsing context that persists across frames.
+  struct Context {
+    void Reset();
+
+    // Segmentation and loop filter state.
+    Vp9SegmentationParams segmentation;
+    Vp9LoopFilterParams loop_filter;
+
+    // Frame references.
+    ReferenceSlot ref_slots[kVp9NumRefFrames];
+
+    Vp9FrameContextManager frame_context_managers[kVp9NumFrameContext];
+  };
+
+  explicit Vp9Parser(bool parsing_compressed_header);

[Pawel Osciak, 2016/08/04 10:20:19]

Please document, maybe referring to ParseNextFrame() doc.

[kcwu, 2016/08/05 11:38:47]

Done.

   ~Vp9Parser();
 
   // Set a new stream buffer to read from, starting at |stream| and of size
   // |stream_size| in bytes. |stream| must point to the beginning of a single
   // frame or a single superframe, is owned by caller and must remain valid
   // until the next call to SetStream().
   void SetStream(const uint8_t* stream, off_t stream_size);
 
   // Parse the next frame in the current stream buffer, filling |fhdr| with
   // the parsed frame header and updating current segmentation and loop filter
-  // state. Return kOk if a frame has successfully been parsed, kEOStream if
-  // there is no more data in the current stream buffer, or kInvalidStream
-  // on error.
-  Result ParseNextFrame(Vp9FrameHeader* fhdr);
+  // state. If |parsing_compressed_header|, this function also fills

[Pawel Osciak, 2016/08/04 10:20:19]

s/parsing_compressed_header/parsing_compressed_header_/

[kcwu, 2016/08/05 11:38:47]

Done.

+  // |context_refresh_cb|, which is used to update frame context. If
+  // |*context_refresh_cb| is null, no callback is necessary.
+  // Return kOk if a frame has successfully been parsed,
+  //        kEOStream if there is no more data in the current stream buffer,
+  //        kAwaitingRefresh if this frame awaiting frame context update, or
+  //        kInvalidStream on error.
+  Result ParseNextFrame(
+      Vp9FrameHeader* fhdr,
+      base::Callback<void(const Vp9FrameContext&)>* context_refresh_cb);
 
   // Return current segmentation state.
-  const Vp9Segmentation& GetSegmentation() const { return segmentation_; }
+  const Vp9SegmentationParams& GetSegmentation() const {
+    return context_.segmentation;
+  }
 
   // Return current loop filter state.
-  const Vp9LoopFilter& GetLoopFilter() const { return loop_filter_; }
+  const Vp9LoopFilterParams& GetLoopFilter() const {
+    return context_.loop_filter;
+  }
 
   // Clear parser state and return to an initialized state.
   void Reset();
 
  private:
-  // The parsing context to keep track of references.
-  struct ReferenceSlot {
-    uint32_t width;
-    uint32_t height;
-  };
+  class UncompressedHeaderParser;
+  class CompressedHeaderParser;
 
   bool ParseSuperframe();
-  uint8_t ReadProfile();
-  bool VerifySyncCode();
-  bool ReadBitDepthColorSpaceSampling(Vp9FrameHeader* fhdr);
-  void ReadFrameSize(Vp9FrameHeader* fhdr);
-  bool ReadFrameSizeFromRefs(Vp9FrameHeader* fhdr);
-  void ReadDisplayFrameSize(Vp9FrameHeader* fhdr);
-  Vp9InterpFilter ReadInterpFilter();
-  void ReadLoopFilter();
-  void ReadQuantization(Vp9QuantizationParams* quants);
-  void ReadSegmentationMap();
-  void ReadSegmentationData();
-  void ReadSegmentation();
-  void ReadTiles(Vp9FrameHeader* fhdr);
-  bool ParseUncompressedHeader(const uint8_t* stream,
-                               off_t frame_size,
-                               Vp9FrameHeader* fhdr);
+
+  size_t GetQIndex(const Vp9QuantizationParams& quant, size_t segid) const;
+  void SetupSegmentationDequant(const Vp9FrameHeader* fhdr);
+  void SetupLoopFilter();
   void UpdateSlots(const Vp9FrameHeader* fhdr);
 
-  void ResetLoopfilter();
-  void SetupPastIndependence();
-  size_t GetQIndex(const Vp9QuantizationParams& quant, size_t segid) const;
-  void SetupSegmentationDequant(const Vp9QuantizationParams& quant);
-  void SetupLoopFilter();
-
   // Current address in the bitstream buffer.
   const uint8_t* stream_;
 
   // Remaining bytes in stream_.
   off_t bytes_left_;
 
   // Stores start pointer and size of each frame within the current superframe.
   struct FrameInfo {
     FrameInfo(const uint8_t* ptr, off_t size);
 
     // Starting address of the frame.
     const uint8_t* ptr;
 
     // Size of the frame in bytes.
     off_t size;
   };
 
+  bool parsing_compressed_header_;
+
   // FrameInfo for the remaining frames in the current superframe to be parsed.
   std::deque<FrameInfo> frames_;
 
-  // Raw bits decoder for uncompressed frame header.
-  Vp9RawBitsReader reader_;
-
-  // Segmentation and loop filter state that persists across frames.
-  Vp9Segmentation segmentation_;
-  Vp9LoopFilter loop_filter_;
-
-  // The parsing context to keep track of references.
-  ReferenceSlot ref_slots_[kVp9NumRefFrames];
+  Context context_;
 
   DISALLOW_COPY_AND_ASSIGN(Vp9Parser);
 };
 
 }  // namespace media
 
 #endif  // MEDIA_FILTERS_VP9_PARSER_H_