Update aom to v1.0.0

Update aom to commit id d14c5bb4f336ef1842046089849dee4a301fbbf0.

1 files changed, 552 insertions, 609 deletions

diff --git a/third_party/aom/av1/common/onyxc_int.h b/third_party/aom/av1/common/onyxc_int.h
index 2396ce2f3..fa5f02e52 100644
--- a/third_party/aom/av1/common/onyxc_int.h
+++ b/third_party/aom/av1/common/onyxc_int.h
@@ -12,76 +12,72 @@
 #ifndef AV1_COMMON_ONYXC_INT_H_
 #define AV1_COMMON_ONYXC_INT_H_
 
-#include "./aom_config.h"
-#include "./av1_rtcd.h"
+#include "config/aom_config.h"
+#include "config/av1_rtcd.h"
+
 #include "aom/internal/aom_codec_internal.h"
 #include "aom_util/aom_thread.h"
-#if CONFIG_ANS
-#include "aom_dsp/ans.h"
-#endif
 #include "av1/common/alloccommon.h"
 #include "av1/common/av1_loopfilter.h"
 #include "av1/common/entropy.h"
 #include "av1/common/entropymode.h"
 #include "av1/common/entropymv.h"
+#include "av1/common/enums.h"
 #include "av1/common/frame_buffers.h"
 #include "av1/common/mv.h"
 #include "av1/common/quant_common.h"
-#if CONFIG_LOOP_RESTORATION
 #include "av1/common/restoration.h"
-#endif  // CONFIG_LOOP_RESTORATION
 #include "av1/common/tile_common.h"
+#include "av1/common/timing.h"
 #include "av1/common/odintrin.h"
-#if CONFIG_PVQ
-#include "av1/common/pvq.h"
-#endif
-#if CONFIG_CFL
-#include "av1/common/cfl.h"
-#endif
-#if CONFIG_HASH_ME
-// TODO(youzhou@microsoft.com): Encoder only. Move it out of common
 #include "av1/encoder/hash_motion.h"
-#endif
+#include "aom_dsp/grain_synthesis.h"
+#include "aom_dsp/grain_table.h"
 #ifdef __cplusplus
 extern "C" {
 #endif
 
-#define CDEF_MAX_STRENGTHS 16
+#if defined(__clang__) && defined(__has_warning)
+#if __has_feature(cxx_attributes) && __has_warning("-Wimplicit-fallthrough")
+#define AOM_FALLTHROUGH_INTENDED [[clang::fallthrough]]  // NOLINT
+#endif
+#elif defined(__GNUC__) && __GNUC__ >= 7
+#define AOM_FALLTHROUGH_INTENDED __attribute__((fallthrough))  // NOLINT
+#endif
 
-#define REF_FRAMES_LOG2 3
-#define REF_FRAMES (1 << REF_FRAMES_LOG2)
+#ifndef AOM_FALLTHROUGH_INTENDED
+#define AOM_FALLTHROUGH_INTENDED \
+  do {                           \
+  } while (0)
+#endif
 
-// 4 scratch frames for the new frames to support a maximum of 4 cores decoding
-// in parallel, 3 for scaled references on the encoder.
-// TODO(hkuang): Add ondemand frame buffers instead of hardcoding the number
-// of framebuffers.
-// TODO(jkoleszar): These 3 extra references could probably come from the
-// normal reference pool.
-#define FRAME_BUFFERS (REF_FRAMES + 7)
+#define CDEF_MAX_STRENGTHS 16
 
-#if CONFIG_REFERENCE_BUFFER
 /* Constant values while waiting for the sequence header */
-#define FRAME_ID_NUMBERS_PRESENT_FLAG 1
-#define FRAME_ID_LENGTH_MINUS7 8         // Allows frame id up to 2^15-1
-#define DELTA_FRAME_ID_LENGTH_MINUS2 12  // Allows frame id deltas up to 2^14-1
-#endif                                   // CONFIG_REFERENCE_BUFFER
+#define FRAME_ID_LENGTH 15
+#define DELTA_FRAME_ID_LENGTH 14
 
-#if CONFIG_NO_FRAME_CONTEXT_SIGNALING
 #define FRAME_CONTEXTS (FRAME_BUFFERS + 1)
 // Extra frame context which is always kept at default values
 #define FRAME_CONTEXT_DEFAULTS (FRAME_CONTEXTS - 1)
-#else
+#define PRIMARY_REF_BITS 3
+#define PRIMARY_REF_NONE 7
 
-#if CONFIG_EXT_REFS
-#define FRAME_CONTEXTS_LOG2 3
-#else
-#define FRAME_CONTEXTS_LOG2 2
-#endif
+#define NUM_PING_PONG_BUFFERS 2
 
-#define FRAME_CONTEXTS (1 << FRAME_CONTEXTS_LOG2)
-#endif  // CONFIG_NO_FRAME_CONTEXT_SIGNALING
+#define MAX_NUM_TEMPORAL_LAYERS 8
+#define MAX_NUM_SPATIAL_LAYERS 4
+/* clang-format off */
+// clang-format seems to think this is a pointer dereference and not a
+// multiplication.
+#define MAX_NUM_OPERATING_POINTS \
+  MAX_NUM_TEMPORAL_LAYERS * MAX_NUM_SPATIAL_LAYERS
+/* clang-format on*/
 
-#define NUM_PING_PONG_BUFFERS 2
+// TODO(jingning): Turning this on to set up transform coefficient
+// processing timer.
+#define TXCOEFF_TIMER 0
+#define TXCOEFF_COST_TIMER 0
 
 typedef enum {
   SINGLE_REFERENCE = 0,
@@ -90,20 +86,11 @@ typedef enum {
   REFERENCE_MODES = 3,
 } REFERENCE_MODE;
 
-#if !CONFIG_NO_FRAME_CONTEXT_SIGNALING
-typedef enum {
-  RESET_FRAME_CONTEXT_NONE = 0,
-  RESET_FRAME_CONTEXT_CURRENT = 1,
-  RESET_FRAME_CONTEXT_ALL = 2,
-} RESET_FRAME_CONTEXT_MODE;
-#endif
-
 typedef enum {
   /**
-   * Update frame context to values resulting from forward probability
-   * updates signaled in the frame header
+   * Frame context updates are disabled
    */
-  REFRESH_FRAME_CONTEXT_FORWARD,
+  REFRESH_FRAME_CONTEXT_DISABLED,
   /**
    * Update frame context to values resulting from backward probability
    * updates based on entropy/counts in the decoded frame
@@ -111,57 +98,41 @@ typedef enum {
   REFRESH_FRAME_CONTEXT_BACKWARD,
 } REFRESH_FRAME_CONTEXT_MODE;
 
-#if CONFIG_MFMV
-#define MFMV_STACK_SIZE INTER_REFS_PER_FRAME
-
+#define MFMV_STACK_SIZE 3
 typedef struct {
-  int_mv mfmv[INTER_REFS_PER_FRAME][MFMV_STACK_SIZE];
+  int_mv mfmv0;
+  uint8_t ref_frame_offset;
 } TPL_MV_REF;
-#endif
 
 typedef struct {
-  int_mv mv[2];
-  int_mv pred_mv[2];
-  MV_REFERENCE_FRAME ref_frame[2];
+  int_mv mv;
+  MV_REFERENCE_FRAME ref_frame;
 } MV_REF;
 
 typedef struct {
   int ref_count;
 
-#if CONFIG_FRAME_MARKER
-  int cur_frame_offset;
-  int lst_frame_offset;
-  int alt_frame_offset;
-  int gld_frame_offset;
-#if CONFIG_EXT_REFS
-  int lst2_frame_offset;
-  int lst3_frame_offset;
-  int bwd_frame_offset;
-  int alt2_frame_offset;
-#endif
-#endif  // CONFIG_FRAME_MARKER
+  unsigned int cur_frame_offset;
+  unsigned int ref_frame_offset[INTER_REFS_PER_FRAME];
 
-#if CONFIG_MFMV
-  TPL_MV_REF *tpl_mvs;
-#endif
   MV_REF *mvs;
+  uint8_t *seg_map;
+  struct segmentation seg;
   int mi_rows;
   int mi_cols;
   // Width and height give the size of the buffer (before any upscaling, unlike
   // the sizes that can be derived from the buf structure)
   int width;
   int height;
-#if CONFIG_GLOBAL_MOTION
-  WarpedMotionParams global_motion[TOTAL_REFS_PER_FRAME];
-#endif  // CONFIG_GLOBAL_MOTION
+  WarpedMotionParams global_motion[REF_FRAMES];
+  int showable_frame;  // frame can be used as show existing frame in future
+  int film_grain_params_present;
+  aom_film_grain_t film_grain_params;
   aom_codec_frame_buffer_t raw_frame_buffer;
   YV12_BUFFER_CONFIG buf;
-#if CONFIG_HASH_ME
   hash_table hash_table;
-#endif
-#if CONFIG_TEMPMV_SIGNALING
   uint8_t intra_only;
-#endif
+  FRAME_TYPE frame_type;
   // The Following variables will only be used in frame parallel decode.
 
   // frame_worker_owner indicates which FrameWorker owns this buffer. NULL means
@@ -173,6 +144,12 @@ typedef struct {
   // when the frame is fully decoded.
   int row;
   int col;
+
+  // Inter frame reference frame delta for loop filter
+  int8_t ref_deltas[REF_FRAMES];
+
+  // 0 = ZERO_MV, MV
+  int8_t mode_deltas[MAX_MODE_LF_DELTAS];
 } RefCntBuffer;
 
 typedef struct BufferPool {
@@ -195,28 +172,77 @@ typedef struct BufferPool {
   InternalFrameBufferList int_frame_buffers;
 } BufferPool;
 
-#if CONFIG_LV_MAP
 typedef struct {
-  int base_ctx_table[2 /*row*/][2 /*col*/][2 /*sig_map*/]
+  int base_ctx_table[2 /*row*/][2 /*col*/][3 /*sig_map*/]
                     [BASE_CONTEXT_POSITION_NUM + 1];
 } LV_MAP_CTX_TABLE;
-typedef int BASE_CTX_TABLE[2 /*col*/][2 /*sig_map*/]
+typedef int BASE_CTX_TABLE[2 /*col*/][3 /*sig_map*/]
                           [BASE_CONTEXT_POSITION_NUM + 1];
-#endif
 
-#if CONFIG_REFERENCE_BUFFER
+typedef struct BitstreamLevel {
+  uint8_t major;
+  uint8_t minor;
+} BitstreamLevel;
+
 /* Initial version of sequence header structure */
 typedef struct SequenceHeader {
+  int num_bits_width;
+  int num_bits_height;
+  int max_frame_width;
+  int max_frame_height;
   int frame_id_numbers_present_flag;
-  int frame_id_length_minus7;
-  int delta_frame_id_length_minus2;
+  int frame_id_length;
+  int delta_frame_id_length;
+  BLOCK_SIZE sb_size;  // Size of the superblock used for this frame
+  int mib_size;        // Size of the superblock in units of MI blocks
+  int mib_size_log2;   // Log 2 of above.
+  int order_hint_bits_minus_1;
+  int force_screen_content_tools;  // 0 - force off
+                                   // 1 - force on
+                                   // 2 - adaptive
+  int force_integer_mv;            // 0 - Not to force. MV can be in 1/4 or 1/8
+                                   // 1 - force to integer
+                                   // 2 - adaptive
+  int still_picture;               // Video is a single frame still picture
+  int reduced_still_picture_hdr;   // Use reduced header for still picture
+  int monochrome;                  // Monochorme video
+  int enable_filter_intra;         // enables/disables filterintra
+  int enable_intra_edge_filter;    // enables/disables corner/edge/upsampling
+  int enable_interintra_compound;  // enables/disables interintra_compound
+  int enable_masked_compound;      // enables/disables masked compound
+  int enable_dual_filter;          // 0 - disable dual interpolation filter
+                                   // 1 - enable vert/horiz filter selection
+  int enable_order_hint;           // 0 - disable order hint, and related tools
+                                   // jnt_comp, ref_frame_mvs, frame_sign_bias
+                                   // if 0, enable_jnt_comp and
+                                   // enable_ref_frame_mvs must be set zs 0.
+  int enable_jnt_comp;             // 0 - disable joint compound modes
+                                   // 1 - enable it
+  int enable_ref_frame_mvs;        // 0 - disable ref frame mvs
+                                   // 1 - enable it
+  int enable_warped_motion;        // 0 - disable warped motion for sequence
+                                   // 1 - enable it for the sequence
+  int enable_superres;     // 0 - Disable superres for the sequence, and disable
+                           //     transmitting per-frame superres enabled flag.
+                           // 1 - Enable superres for the sequence, and also
+                           //     enable per-frame flag to denote if superres is
+                           //     enabled for that frame.
+  int enable_cdef;         // To turn on/off CDEF
+  int enable_restoration;  // To turn on/off loop restoration
+  int operating_points_cnt_minus_1;
+  int operating_point_idc[MAX_NUM_OPERATING_POINTS];
+  int display_model_info_present_flag;
+  int decoder_model_info_present_flag;
+  BitstreamLevel level[MAX_NUM_OPERATING_POINTS];
+  uint8_t tier[MAX_NUM_OPERATING_POINTS];  // seq_tier in the spec. One bit: 0
+                                           // or 1.
 } SequenceHeader;
-#endif  // CONFIG_REFERENCE_BUFFER
 
 typedef struct AV1Common {
   struct aom_internal_error_info error;
-  aom_color_space_t color_space;
-  aom_transfer_function_t transfer_function;
+  aom_color_primaries_t color_primaries;
+  aom_transfer_characteristics_t transfer_characteristics;
+  aom_matrix_coefficients_t matrix_coefficients;
   aom_chroma_sample_position_t chroma_sample_position;
   int color_range;
   int width;
@@ -225,6 +251,14 @@ typedef struct AV1Common {
   int render_height;
   int last_width;
   int last_height;
+  int timing_info_present;
+  aom_timing_info_t timing_info;
+  int buffer_removal_delay_present;
+  aom_dec_model_info_t buffer_model;
+  aom_dec_model_op_parameters_t op_params[MAX_NUM_OPERATING_POINTS + 1];
+  aom_op_timing_info_t op_frame_timing[MAX_NUM_OPERATING_POINTS + 1];
+  int tu_presentation_delay_flag;
+  int64_t tu_presentation_delay;
 
   // TODO(jkoleszar): this implies chroma ss right now, but could vary per
   // plane. Revisit as part of the future change to YV12_BUFFER_CONFIG to
@@ -232,10 +266,15 @@ typedef struct AV1Common {
   int subsampling_x;
   int subsampling_y;
 
-#if CONFIG_HIGHBITDEPTH
+  int largest_tile_id;
+  size_t largest_tile_size;
+  int context_update_tile_id;
+
+  // Scale of the current frame with respect to itself.
+  struct scale_factors sf_identity;
+
   // Marks if we need to use 16bit frame buffers (1: yes, 0: no).
   int use_highbitdepth;
-#endif
   YV12_BUFFER_CONFIG *frame_to_show;
   RefCntBuffer *prev_frame;
 
@@ -253,6 +292,10 @@ typedef struct AV1Common {
 
   // Each Inter frame can reference INTER_REFS_PER_FRAME buffers
   RefBuffer frame_refs[INTER_REFS_PER_FRAME];
+  int is_skip_mode_allowed;
+  int skip_mode_flag;
+  int ref_frame_idx_0;
+  int ref_frame_idx_1;
 
   int new_fb_idx;
 
@@ -260,39 +303,26 @@ typedef struct AV1Common {
   FRAME_TYPE frame_type;
 
   int show_frame;
+  int showable_frame;  // frame can be used as show existing frame in future
   int last_show_frame;
   int show_existing_frame;
-#if CONFIG_EXT_REFS
   // Flag for a frame used as a reference - not written to the bitstream
   int is_reference_frame;
-#endif  // CONFIG_EXT_REFS
+  int reset_decoder_state;
 
   // Flag signaling that the frame is encoded using only INTRA modes.
   uint8_t intra_only;
   uint8_t last_intra_only;
-
+  uint8_t disable_cdf_update;
   int allow_high_precision_mv;
-#if CONFIG_AMVR
-  int seq_mv_precision_level;        // 0 the default in AOM, 1 only integer, 2
-                                     // adaptive
-  int cur_frame_mv_precision_level;  // 0 the default in AOM, 1 only integer
-#endif
+  int cur_frame_force_integer_mv;  // 0 the default in AOM, 1 only integer
 
   int allow_screen_content_tools;
-#if CONFIG_INTERINTRA
-  int allow_interintra_compound;
-#endif  // CONFIG_INTERINTRA
-#if CONFIG_WEDGE || CONFIG_COMPOUND_SEGMENT
-  int allow_masked_compound;
-#endif  // CONFIG_WEDGE || CONFIG_COMPOUND_SEGMENT
-
-#if !CONFIG_NO_FRAME_CONTEXT_SIGNALING
-  // Flag signaling which frame contexts should be reset to default values.
-  RESET_FRAME_CONTEXT_MODE reset_frame_context;
-#endif
+  int allow_intrabc;
+  int allow_warped_motion;
 
   // MBs, mb_rows/cols is in 16-pixel units; mi_rows/cols is in
-  // MODE_INFO (8-pixel) units.
+  // MB_MODE_INFO (8-pixel) units.
   int MBs;
   int mb_rows, mi_rows;
   int mb_cols, mi_cols;
@@ -301,119 +331,120 @@ typedef struct AV1Common {
   /* profile settings */
   TX_MODE tx_mode;
 
+#if CONFIG_ENTROPY_STATS
+  int coef_cdf_category;
+#endif
+
   int base_qindex;
   int y_dc_delta_q;
-  int uv_dc_delta_q;
-  int uv_ac_delta_q;
-  int16_t y_dequant[MAX_SEGMENTS][2];
-  int16_t uv_dequant[MAX_SEGMENTS][2];
+  int u_dc_delta_q;
+  int v_dc_delta_q;
+  int u_ac_delta_q;
+  int v_ac_delta_q;
+
+  int separate_uv_delta_q;
+
+  // The dequantizers below are true dequntizers used only in the
+  // dequantization process.  They have the same coefficient
+  // shift/scale as TX.
+  int16_t y_dequant_QTX[MAX_SEGMENTS][2];
+  int16_t u_dequant_QTX[MAX_SEGMENTS][2];
+  int16_t v_dequant_QTX[MAX_SEGMENTS][2];
 
-#if CONFIG_AOM_QM
   // Global quant matrix tables
-  qm_val_t *giqmatrix[NUM_QM_LEVELS][2][2][TX_SIZES_ALL];
-  qm_val_t *gqmatrix[NUM_QM_LEVELS][2][2][TX_SIZES_ALL];
+  const qm_val_t *giqmatrix[NUM_QM_LEVELS][3][TX_SIZES_ALL];
+  const qm_val_t *gqmatrix[NUM_QM_LEVELS][3][TX_SIZES_ALL];
 
   // Local quant matrix tables for each frame
-  qm_val_t *y_iqmatrix[MAX_SEGMENTS][2][TX_SIZES_ALL];
-  qm_val_t *uv_iqmatrix[MAX_SEGMENTS][2][TX_SIZES_ALL];
-  // Encoder
-  qm_val_t *y_qmatrix[MAX_SEGMENTS][2][TX_SIZES_ALL];
-  qm_val_t *uv_qmatrix[MAX_SEGMENTS][2][TX_SIZES_ALL];
+  const qm_val_t *y_iqmatrix[MAX_SEGMENTS][TX_SIZES_ALL];
+  const qm_val_t *u_iqmatrix[MAX_SEGMENTS][TX_SIZES_ALL];
+  const qm_val_t *v_iqmatrix[MAX_SEGMENTS][TX_SIZES_ALL];
 
+  // Encoder
   int using_qmatrix;
+  int qm_y;
+  int qm_u;
+  int qm_v;
   int min_qmlevel;
   int max_qmlevel;
-#endif
-#if CONFIG_NEW_QUANT
-  dequant_val_type_nuq y_dequant_nuq[MAX_SEGMENTS][QUANT_PROFILES][COEF_BANDS];
-  dequant_val_type_nuq uv_dequant_nuq[MAX_SEGMENTS][QUANT_PROFILES][COEF_BANDS];
-#endif
 
-  /* We allocate a MODE_INFO struct for each macroblock, together with
+  /* We allocate a MB_MODE_INFO struct for each macroblock, together with
      an extra row on top and column on the left to simplify prediction. */
   int mi_alloc_size;
-  MODE_INFO *mip; /* Base of allocated array */
-  MODE_INFO *mi;  /* Corresponds to upper left visible macroblock */
+  MB_MODE_INFO *mip; /* Base of allocated array */
+  MB_MODE_INFO *mi;  /* Corresponds to upper left visible macroblock */
 
   // TODO(agrange): Move prev_mi into encoder structure.
   // prev_mip and prev_mi will only be allocated in encoder.
-  MODE_INFO *prev_mip; /* MODE_INFO array 'mip' from last decoded frame */
-  MODE_INFO *prev_mi;  /* 'mi' from last frame (points into prev_mip) */
+  MB_MODE_INFO *prev_mip; /* MB_MODE_INFO array 'mip' from last decoded frame */
+  MB_MODE_INFO *prev_mi;  /* 'mi' from last frame (points into prev_mip) */
 
   // Separate mi functions between encoder and decoder.
   int (*alloc_mi)(struct AV1Common *cm, int mi_size);
   void (*free_mi)(struct AV1Common *cm);
   void (*setup_mi)(struct AV1Common *cm);
 
-  // Grid of pointers to 8x8 MODE_INFO structs.  Any 8x8 not in the visible
+  // Grid of pointers to 8x8 MB_MODE_INFO structs.  Any 8x8 not in the visible
   // area will be NULL.
-  MODE_INFO **mi_grid_base;
-  MODE_INFO **mi_grid_visible;
-  MODE_INFO **prev_mi_grid_base;
-  MODE_INFO **prev_mi_grid_visible;
-
-  // Whether to use previous frame's motion vectors for prediction.
-  int use_prev_frame_mvs;
+  MB_MODE_INFO **mi_grid_base;
+  MB_MODE_INFO **mi_grid_visible;
+  MB_MODE_INFO **prev_mi_grid_base;
+  MB_MODE_INFO **prev_mi_grid_visible;
 
-  // Persistent mb segment id map used in prediction.
-  int seg_map_idx;
-  int prev_seg_map_idx;
+  // Whether to use previous frames' motion vectors for prediction.
+  int allow_ref_frame_mvs;
 
-  uint8_t *seg_map_array[NUM_PING_PONG_BUFFERS];
   uint8_t *last_frame_seg_map;
   uint8_t *current_frame_seg_map;
   int seg_map_alloc_size;
 
   InterpFilter interp_filter;
 
+  int switchable_motion_mode;
+
   loop_filter_info_n lf_info;
-#if CONFIG_FRAME_SUPERRES
   // The denominator of the superres scale; the numerator is fixed.
   uint8_t superres_scale_denominator;
   int superres_upscaled_width;
   int superres_upscaled_height;
-#endif  // CONFIG_FRAME_SUPERRES
-#if CONFIG_LOOP_RESTORATION
   RestorationInfo rst_info[MAX_MB_PLANE];
-  RestorationInternal rst_internal;
-#endif  // CONFIG_LOOP_RESTORATION
+
+  // rst_end_stripe[i] is one more than the index of the bottom stripe
+  // for tile row i.
+  int rst_end_stripe[MAX_TILE_ROWS];
+
+  // Pointer to a scratch buffer used by self-guided restoration
+  int32_t *rst_tmpbuf;
+  RestorationLineBuffers *rlbs;
+
+  // Output of loop restoration
+  YV12_BUFFER_CONFIG rst_frame;
 
   // Flag signaling how frame contexts should be updated at the end of
   // a frame decode
   REFRESH_FRAME_CONTEXT_MODE refresh_frame_context;
 
-  int ref_frame_sign_bias[TOTAL_REFS_PER_FRAME]; /* Two state 0, 1 */
+  int ref_frame_sign_bias[REF_FRAMES]; /* Two state 0, 1 */
 
   struct loopfilter lf;
   struct segmentation seg;
-  int all_lossless;
-  int frame_parallel_decode;  // frame-based threading.
+  int coded_lossless;  // frame is fully lossless at the coded resolution.
+  int all_lossless;    // frame is fully lossless at the upscaled resolution.
 
-#if CONFIG_EXT_TX
   int reduced_tx_set_used;
-#endif  // CONFIG_EXT_TX
 
-// Context probabilities for reference frame prediction
-#if CONFIG_EXT_REFS
+  // Context probabilities for reference frame prediction
   MV_REFERENCE_FRAME comp_fwd_ref[FWD_REFS];
   MV_REFERENCE_FRAME comp_bwd_ref[BWD_REFS];
-#else
-  MV_REFERENCE_FRAME comp_fixed_ref;
-  MV_REFERENCE_FRAME comp_var_ref[COMP_REFS];
-#endif  // CONFIG_EXT_REFS
   REFERENCE_MODE reference_mode;
 
   FRAME_CONTEXT *fc;              /* this frame entropy */
   FRAME_CONTEXT *frame_contexts;  // FRAME_CONTEXTS
-  FRAME_CONTEXT *pre_fc;          // Context referenced in this frame
-#if !CONFIG_NO_FRAME_CONTEXT_SIGNALING
   unsigned int frame_context_idx; /* Context to use/update */
-#endif
-  FRAME_COUNTS counts;
+  int fb_of_context_type[REF_FRAMES];
+  int primary_ref_frame;
 
-#if CONFIG_FRAME_MARKER
   unsigned int frame_offset;
-#endif
 
   unsigned int current_video_frame;
   BITSTREAM_PROFILE profile;
@@ -423,44 +454,27 @@ typedef struct AV1Common {
   aom_bit_depth_t dequant_bit_depth;  // bit_depth of current dequantizer
 
   int error_resilient_mode;
+  int force_primary_ref_none;
 
   int tile_cols, tile_rows;
   int last_tile_cols, last_tile_rows;
 
-#if CONFIG_MAX_TILE
+  int max_tile_width_sb;
   int min_log2_tile_cols;
   int max_log2_tile_cols;
   int max_log2_tile_rows;
   int min_log2_tile_rows;
   int min_log2_tiles;
-  int max_tile_width_sb;
   int max_tile_height_sb;
   int uniform_tile_spacing_flag;
   int log2_tile_cols;                        // only valid for uniform tiles
   int log2_tile_rows;                        // only valid for uniform tiles
   int tile_col_start_sb[MAX_TILE_COLS + 1];  // valid for 0 <= i <= tile_cols
   int tile_row_start_sb[MAX_TILE_ROWS + 1];  // valid for 0 <= i <= tile_rows
-#if CONFIG_DEPENDENT_HORZTILES
-  int tile_row_independent[MAX_TILE_ROWS];  // valid for 0 <= i <  tile_rows
-#endif
-#else
-  int log2_tile_cols, log2_tile_rows;  // Used in non-large_scale_tile_coding.
-  int tile_width, tile_height;         // In MI units
-#endif  // CONFIG_MAX_TILE
+  int tile_width, tile_height;               // In MI units
 
-#if CONFIG_EXT_TILE
   unsigned int large_scale_tile;
   unsigned int single_tile_decoding;
-#endif  // CONFIG_EXT_TILE
-
-#if CONFIG_DEPENDENT_HORZTILES
-  int dependent_horz_tiles;
-  int tile_group_start_row[MAX_TILE_ROWS][MAX_TILE_COLS];
-  int tile_group_start_col[MAX_TILE_ROWS][MAX_TILE_COLS];
-#endif
-#if CONFIG_LOOPFILTERING_ACROSS_TILES
-  int loop_filter_across_tiles_enabled;
-#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
 
   int byte_alignment;
   int skip_loop_filter;
@@ -476,74 +490,65 @@ typedef struct AV1Common {
   // External BufferPool passed from outside.
   BufferPool *buffer_pool;
 
-  PARTITION_CONTEXT *above_seg_context;
-  ENTROPY_CONTEXT *above_context[MAX_MB_PLANE];
-#if CONFIG_VAR_TX
-  TXFM_CONTEXT *above_txfm_context;
-  TXFM_CONTEXT *top_txfm_context[MAX_MB_PLANE];
-  TXFM_CONTEXT left_txfm_context[MAX_MB_PLANE][2 * MAX_MIB_SIZE];
-#endif
-  int above_context_alloc_cols;
-
-  // scratch memory for intraonly/keyframe forward updates from default tables
-  // - this is intentionally not placed in FRAME_CONTEXT since it's reset upon
-  // each keyframe and not used afterwards
-  aom_prob kf_y_prob[INTRA_MODES][INTRA_MODES][INTRA_MODES - 1];
-#if CONFIG_GLOBAL_MOTION
-  WarpedMotionParams global_motion[TOTAL_REFS_PER_FRAME];
-#endif
-
-  BLOCK_SIZE sb_size;  // Size of the superblock used for this frame
-  int mib_size;        // Size of the superblock in units of MI blocks
-  int mib_size_log2;   // Log 2 of above.
-#if CONFIG_CDEF
+  PARTITION_CONTEXT **above_seg_context;
+  ENTROPY_CONTEXT **above_context[MAX_MB_PLANE];
+  TXFM_CONTEXT **above_txfm_context;
+  WarpedMotionParams global_motion[REF_FRAMES];
+  aom_film_grain_table_t *film_grain_table;
+  int film_grain_params_present;
+  aom_film_grain_t film_grain_params;
   int cdef_pri_damping;
   int cdef_sec_damping;
   int nb_cdef_strengths;
   int cdef_strengths[CDEF_MAX_STRENGTHS];
   int cdef_uv_strengths[CDEF_MAX_STRENGTHS];
   int cdef_bits;
-#endif
 
   int delta_q_present_flag;
   // Resolution of delta quant
   int delta_q_res;
-#if CONFIG_EXT_DELTA_Q
   int delta_lf_present_flag;
   // Resolution of delta lf level
   int delta_lf_res;
-#if CONFIG_LOOPFILTER_LEVEL
   // This is a flag for number of deltas of loop filter level
   // 0: use 1 delta, for y_vertical, y_horizontal, u, and v
   // 1: use separate deltas for each filter level
   int delta_lf_multi;
-#endif  // CONFIG_LOOPFILTER_LEVEL
-#endif
   int num_tg;
-#if CONFIG_REFERENCE_BUFFER
   SequenceHeader seq_params;
   int current_frame_id;
   int ref_frame_id[REF_FRAMES];
   int valid_for_referencing[REF_FRAMES];
-  int refresh_mask;
-  int invalid_delta_frame_id_minus1;
-#endif  // CONFIG_REFERENCE_BUFFER
-#if CONFIG_ANS && ANS_MAX_SYMBOLS
-  int ans_window_size_log2;
-#endif
-#if CONFIG_NCOBMC_ADAPT_WEIGHT
-  NCOBMC_KERNELS ncobmc_kernels[ADAPT_OVERLAP_BLOCKS][ALL_NCOBMC_MODES];
-  uint8_t *ncobmcaw_buf[4];
-#endif
-#if CONFIG_LV_MAP
+  int invalid_delta_frame_id_minus_1;
   LV_MAP_CTX_TABLE coeff_ctx_table;
+  TPL_MV_REF *tpl_mvs;
+  int tpl_mvs_mem_size;
+  // TODO(jingning): This can be combined with sign_bias later.
+  int8_t ref_frame_side[REF_FRAMES];
+
+  int is_annexb;
+
+  int frame_refs_short_signaling;
+  int temporal_layer_id;
+  int spatial_layer_id;
+  unsigned int number_temporal_layers;
+  unsigned int number_spatial_layers;
+  int num_allocated_above_context_mi_col;
+  int num_allocated_above_contexts;
+  int num_allocated_above_context_planes;
+
+#if TXCOEFF_TIMER
+  int64_t cum_txcoeff_timer;
+  int64_t txcoeff_timer;
+  int txb_count;
 #endif
-#if CONFIG_LPF_SB
-  int final_lpf_encode;
-#endif
-#if CONFIG_ADAPT_SCAN
-  int use_adapt_scan;
+
+#if TXCOEFF_COST_TIMER
+  int64_t cum_txcoeff_cost_timer;
+  int64_t txcoeff_cost_timer;
+  int64_t txcoeff_cost_count;
 #endif
+  const cfg_options_t *options;
 } AV1_COMMON;
 
 // TODO(hkuang): Don't need to lock the whole pool after implementing atomic
@@ -585,6 +590,17 @@ static INLINE int get_free_fb(AV1_COMMON *cm) {
     if (frame_bufs[i].ref_count == 0) break;
 
   if (i != FRAME_BUFFERS) {
+    if (frame_bufs[i].buf.use_external_refernce_buffers) {
+      // If this frame buffer's y_buffer, u_buffer, and v_buffer point to the
+      // external reference buffers. Restore the buffer pointers to point to the
+      // internally allocated memory.
+      YV12_BUFFER_CONFIG *ybf = &frame_bufs[i].buf;
+      ybf->y_buffer = ybf->store_buf_adr[0];
+      ybf->u_buffer = ybf->store_buf_adr[1];
+      ybf->v_buffer = ybf->store_buf_adr[2];
+      ybf->use_external_refernce_buffers = 0;
+    }
+
     frame_bufs[i].ref_count = 1;
   } else {
     // Reset i to be INVALID_IDX to indicate no free buffer found.
@@ -606,270 +622,236 @@ static INLINE void ref_cnt_fb(RefCntBuffer *bufs, int *idx, int new_idx) {
   bufs[new_idx].ref_count++;
 }
 
-#if CONFIG_TEMPMV_SIGNALING
-// Returns 1 if this frame might use mvs from some previous frame. This
-// function doesn't consider whether prev_frame is actually suitable (see
-// frame_can_use_prev_frame_mvs for that)
-static INLINE int frame_might_use_prev_frame_mvs(const AV1_COMMON *cm) {
-  return !cm->error_resilient_mode && !cm->intra_only;
+static INLINE int frame_is_intra_only(const AV1_COMMON *const cm) {
+  return cm->frame_type == KEY_FRAME || cm->intra_only;
+}
+
+static INLINE int frame_is_sframe(const AV1_COMMON *cm) {
+  return cm->frame_type == S_FRAME;
 }
 
-// Returns 1 if this frame really can use MVs from some previous frame.
-static INLINE int frame_can_use_prev_frame_mvs(const AV1_COMMON *cm) {
-  return (frame_might_use_prev_frame_mvs(cm) && cm->last_show_frame &&
-          cm->prev_frame && !cm->prev_frame->intra_only &&
-          cm->width == cm->prev_frame->width &&
-          cm->height == cm->prev_frame->height);
+static INLINE RefCntBuffer *get_prev_frame(const AV1_COMMON *const cm) {
+  if (cm->primary_ref_frame == PRIMARY_REF_NONE ||
+      cm->frame_refs[cm->primary_ref_frame].idx == INVALID_IDX) {
+    return NULL;
+  } else {
+    return &cm->buffer_pool
+                ->frame_bufs[cm->frame_refs[cm->primary_ref_frame].idx];
+  }
+}
+
+// Returns 1 if this frame might allow mvs from some reference frame.
+static INLINE int frame_might_allow_ref_frame_mvs(const AV1_COMMON *cm) {
+  return !cm->error_resilient_mode && cm->seq_params.enable_ref_frame_mvs &&
+         cm->seq_params.enable_order_hint && !frame_is_intra_only(cm);
+}
+
+// Returns 1 if this frame might use warped_motion
+static INLINE int frame_might_allow_warped_motion(const AV1_COMMON *cm) {
+  return !cm->error_resilient_mode && !frame_is_intra_only(cm) &&
+         cm->seq_params.enable_warped_motion;
 }
-#endif
 
 static INLINE void ensure_mv_buffer(RefCntBuffer *buf, AV1_COMMON *cm) {
-  if (buf->mvs == NULL || buf->mi_rows < cm->mi_rows ||
-      buf->mi_cols < cm->mi_cols) {
+  const int buf_rows = buf->mi_rows;
+  const int buf_cols = buf->mi_cols;
+
+  if (buf->mvs == NULL || buf_rows != cm->mi_rows || buf_cols != cm->mi_cols) {
     aom_free(buf->mvs);
     buf->mi_rows = cm->mi_rows;
     buf->mi_cols = cm->mi_cols;
-#if CONFIG_TMV
     CHECK_MEM_ERROR(cm, buf->mvs,
                     (MV_REF *)aom_calloc(
                         ((cm->mi_rows + 1) >> 1) * ((cm->mi_cols + 1) >> 1),
                         sizeof(*buf->mvs)));
-#else
-    CHECK_MEM_ERROR(
-        cm, buf->mvs,
-        (MV_REF *)aom_calloc(cm->mi_rows * cm->mi_cols, sizeof(*buf->mvs)));
-#endif  // CONFIG_TMV
-
-#if CONFIG_MFMV
-    aom_free(buf->tpl_mvs);
-    CHECK_MEM_ERROR(
-        cm, buf->tpl_mvs,
-        (TPL_MV_REF *)aom_calloc((cm->mi_rows + MAX_MIB_SIZE) * cm->mi_stride,
-                                 sizeof(*buf->tpl_mvs)));
-#endif
+    aom_free(buf->seg_map);
+    CHECK_MEM_ERROR(cm, buf->seg_map,
+                    (uint8_t *)aom_calloc(cm->mi_rows * cm->mi_cols,
+                                          sizeof(*buf->seg_map)));
   }
-}
 
-#if CONFIG_VAR_REFS
-#define LAST_IS_VALID(cm) ((cm)->frame_refs[LAST_FRAME - 1].is_valid)
-#define LAST2_IS_VALID(cm) ((cm)->frame_refs[LAST2_FRAME - 1].is_valid)
-#define LAST3_IS_VALID(cm) ((cm)->frame_refs[LAST3_FRAME - 1].is_valid)
-#define GOLDEN_IS_VALID(cm) ((cm)->frame_refs[GOLDEN_FRAME - 1].is_valid)
-#define BWDREF_IS_VALID(cm) ((cm)->frame_refs[BWDREF_FRAME - 1].is_valid)
-#define ALTREF2_IS_VALID(cm) ((cm)->frame_refs[ALTREF2_FRAME - 1].is_valid)
-#define ALTREF_IS_VALID(cm) ((cm)->frame_refs[ALTREF_FRAME - 1].is_valid)
-
-#define L_OR_L2(cm) (LAST_IS_VALID(cm) || LAST2_IS_VALID(cm))
-#define L_AND_L2(cm) (LAST_IS_VALID(cm) && LAST2_IS_VALID(cm))
-#define L_AND_L3(cm) (LAST_IS_VALID(cm) && LAST3_IS_VALID(cm))
-#define L_AND_G(cm) (LAST_IS_VALID(cm) && GOLDEN_IS_VALID(cm))
-
-#define L3_OR_G(cm) (LAST3_IS_VALID(cm) || GOLDEN_IS_VALID(cm))
-#define L3_AND_G(cm) (LAST3_IS_VALID(cm) && GOLDEN_IS_VALID(cm))
-
-#define BWD_OR_ALT2(cm) (BWDREF_IS_VALID(cm) || ALTREF2_IS_VALID(cm))
-#define BWD_AND_ALT2(cm) (BWDREF_IS_VALID(cm) && ALTREF2_IS_VALID(cm))
-#define BWD_OR_ALT(cm) (BWDREF_IS_VALID(cm) || ALTREF_IS_VALID(cm))
-#define BWD_AND_ALT(cm) (BWDREF_IS_VALID(cm) && ALTREF_IS_VALID(cm))
-#endif  // CONFIG_VAR_REFS
+  const int mem_size =
+      ((cm->mi_rows + MAX_MIB_SIZE) >> 1) * (cm->mi_stride >> 1);
+  int realloc = cm->tpl_mvs == NULL;
+  if (cm->tpl_mvs) realloc |= cm->tpl_mvs_mem_size < mem_size;
+
+  if (realloc) {
+    aom_free(cm->tpl_mvs);
+    CHECK_MEM_ERROR(cm, cm->tpl_mvs,
+                    (TPL_MV_REF *)aom_calloc(mem_size, sizeof(*cm->tpl_mvs)));
+    cm->tpl_mvs_mem_size = mem_size;
+  }
+}
 
 static INLINE int mi_cols_aligned_to_sb(const AV1_COMMON *cm) {
-  return ALIGN_POWER_OF_TWO(cm->mi_cols, cm->mib_size_log2);
+  return ALIGN_POWER_OF_TWO(cm->mi_cols, cm->seq_params.mib_size_log2);
 }
 
 static INLINE int mi_rows_aligned_to_sb(const AV1_COMMON *cm) {
-  return ALIGN_POWER_OF_TWO(cm->mi_rows, cm->mib_size_log2);
+  return ALIGN_POWER_OF_TWO(cm->mi_rows, cm->seq_params.mib_size_log2);
 }
 
-static INLINE int frame_is_intra_only(const AV1_COMMON *const cm) {
-  return cm->frame_type == KEY_FRAME || cm->intra_only;
+void cfl_init(CFL_CTX *cfl, AV1_COMMON *cm);
+
+static INLINE int av1_num_planes(const AV1_COMMON *cm) {
+  return cm->seq_params.monochrome ? 1 : MAX_MB_PLANE;
 }
 
-#if CONFIG_CFL
-#if CONFIG_CHROMA_SUB8X8 && CONFIG_DEBUG
-static INLINE void cfl_clear_sub8x8_val(CFL_CTX *cfl) {
-  memset(cfl->sub8x8_val, 0, sizeof(cfl->sub8x8_val));
+static INLINE void av1_init_above_context(AV1_COMMON *cm, MACROBLOCKD *xd,
+                                          const int tile_row) {
+  const int num_planes = av1_num_planes(cm);
+  for (int i = 0; i < num_planes; ++i) {
+    xd->above_context[i] = cm->above_context[i][tile_row];
+  }
+  xd->above_seg_context = cm->above_seg_context[tile_row];
+  xd->above_txfm_context = cm->above_txfm_context[tile_row];
 }
-#endif  // CONFIG_CHROMA_SUB8X8 && CONFIG_DEBUG
-void cfl_init(CFL_CTX *cfl, AV1_COMMON *cm);
-#endif  // CONFIG_CFL
 
 static INLINE void av1_init_macroblockd(AV1_COMMON *cm, MACROBLOCKD *xd,
-#if CONFIG_PVQ
-                                        tran_low_t *pvq_ref_coeff,
-#endif
-#if CONFIG_CFL
-                                        CFL_CTX *cfl,
-#endif
                                         tran_low_t *dqcoeff) {
-  for (int i = 0; i < MAX_MB_PLANE; ++i) {
+  const int num_planes = av1_num_planes(cm);
+  for (int i = 0; i < num_planes; ++i) {
     xd->plane[i].dqcoeff = dqcoeff;
-#if CONFIG_PVQ
-    xd->plane[i].pvq_ref_coeff = pvq_ref_coeff;
-#endif
-    xd->above_context[i] = cm->above_context[i];
+
     if (xd->plane[i].plane_type == PLANE_TYPE_Y) {
-      memcpy(xd->plane[i].seg_dequant, cm->y_dequant, sizeof(cm->y_dequant));
-#if CONFIG_AOM_QM
+      memcpy(xd->plane[i].seg_dequant_QTX, cm->y_dequant_QTX,
+             sizeof(cm->y_dequant_QTX));
       memcpy(xd->plane[i].seg_iqmatrix, cm->y_iqmatrix, sizeof(cm->y_iqmatrix));
-#endif
 
-#if CONFIG_NEW_QUANT
-      memcpy(xd->plane[i].seg_dequant_nuq, cm->y_dequant_nuq,
-             sizeof(cm->y_dequant_nuq));
-#endif
     } else {
-      memcpy(xd->plane[i].seg_dequant, cm->uv_dequant, sizeof(cm->uv_dequant));
-#if CONFIG_AOM_QM
-      memcpy(xd->plane[i].seg_iqmatrix, cm->uv_iqmatrix,
-             sizeof(cm->uv_iqmatrix));
-#endif
-#if CONFIG_NEW_QUANT
-      memcpy(xd->plane[i].seg_dequant_nuq, cm->uv_dequant_nuq,
-             sizeof(cm->uv_dequant_nuq));
-#endif
+      if (i == AOM_PLANE_U) {
+        memcpy(xd->plane[i].seg_dequant_QTX, cm->u_dequant_QTX,
+               sizeof(cm->u_dequant_QTX));
+        memcpy(xd->plane[i].seg_iqmatrix, cm->u_iqmatrix,
+               sizeof(cm->u_iqmatrix));
+      } else {
+        memcpy(xd->plane[i].seg_dequant_QTX, cm->v_dequant_QTX,
+               sizeof(cm->v_dequant_QTX));
+        memcpy(xd->plane[i].seg_iqmatrix, cm->v_iqmatrix,
+               sizeof(cm->v_iqmatrix));
+      }
     }
   }
-  xd->fc = cm->fc;
-  xd->above_seg_context = cm->above_seg_context;
-#if CONFIG_VAR_TX
-  xd->above_txfm_context = cm->above_txfm_context;
-#endif
-#if CONFIG_CFL
-  cfl_init(cfl, cm);
-  xd->cfl = cfl;
-#endif
   xd->mi_stride = cm->mi_stride;
   xd->error_info = &cm->error;
+  cfl_init(&xd->cfl, cm);
 }
 
-static INLINE void set_skip_context(MACROBLOCKD *xd, int mi_row, int mi_col) {
+static INLINE void set_skip_context(MACROBLOCKD *xd, int mi_row, int mi_col,
+                                    const int num_planes) {
   int i;
   int row_offset = mi_row;
   int col_offset = mi_col;
-  for (i = 0; i < MAX_MB_PLANE; ++i) {
+  for (i = 0; i < num_planes; ++i) {
     struct macroblockd_plane *const pd = &xd->plane[i];
-#if CONFIG_CHROMA_SUB8X8
     // Offset the buffer pointer
-    const BLOCK_SIZE bsize = xd->mi[0]->mbmi.sb_type;
+    const BLOCK_SIZE bsize = xd->mi[0]->sb_type;
     if (pd->subsampling_y && (mi_row & 0x01) && (mi_size_high[bsize] == 1))
       row_offset = mi_row - 1;
     if (pd->subsampling_x && (mi_col & 0x01) && (mi_size_wide[bsize] == 1))
       col_offset = mi_col - 1;
-#endif
-    int above_idx = col_offset << (MI_SIZE_LOG2 - tx_size_wide_log2[0]);
-    int left_idx = (row_offset & MAX_MIB_MASK)
-                   << (MI_SIZE_LOG2 - tx_size_high_log2[0]);
+    int above_idx = col_offset;
+    int left_idx = row_offset & MAX_MIB_MASK;
     pd->above_context = &xd->above_context[i][above_idx >> pd->subsampling_x];
     pd->left_context = &xd->left_context[i][left_idx >> pd->subsampling_y];
   }
 }
 
 static INLINE int calc_mi_size(int len) {
-  // len is in mi units.
-  return len + MAX_MIB_SIZE;
+  // len is in mi units. Align to a multiple of SBs.
+  return ALIGN_POWER_OF_TWO(len, MAX_MIB_SIZE_LOG2);
 }
 
-static INLINE void set_plane_n4(MACROBLOCKD *const xd, int bw, int bh) {
+static INLINE void set_plane_n4(MACROBLOCKD *const xd, int bw, int bh,
+                                const int num_planes) {
   int i;
-  for (i = 0; i < MAX_MB_PLANE; i++) {
-    xd->plane[i].n4_w = (bw << 1) >> xd->plane[i].subsampling_x;
-    xd->plane[i].n4_h = (bh << 1) >> xd->plane[i].subsampling_y;
-
+  for (i = 0; i < num_planes; i++) {
     xd->plane[i].width = (bw * MI_SIZE) >> xd->plane[i].subsampling_x;
     xd->plane[i].height = (bh * MI_SIZE) >> xd->plane[i].subsampling_y;
 
-#if !CONFIG_CHROMA_2X2
     xd->plane[i].width = AOMMAX(xd->plane[i].width, 4);
     xd->plane[i].height = AOMMAX(xd->plane[i].height, 4);
-#endif
   }
 }
 
 static INLINE void set_mi_row_col(MACROBLOCKD *xd, const TileInfo *const tile,
                                   int mi_row, int bh, int mi_col, int bw,
-#if CONFIG_DEPENDENT_HORZTILES
-                                  int dependent_horz_tile_flag,
-#endif  // CONFIG_DEPENDENT_HORZTILES
                                   int mi_rows, int mi_cols) {
   xd->mb_to_top_edge = -((mi_row * MI_SIZE) * 8);
   xd->mb_to_bottom_edge = ((mi_rows - bh - mi_row) * MI_SIZE) * 8;
   xd->mb_to_left_edge = -((mi_col * MI_SIZE) * 8);
   xd->mb_to_right_edge = ((mi_cols - bw - mi_col) * MI_SIZE) * 8;
 
-#if CONFIG_DEPENDENT_HORZTILES
-  if (dependent_horz_tile_flag) {
-    xd->up_available = (mi_row > tile->mi_row_start) || !tile->tg_horz_boundary;
-  } else {
-#endif  // CONFIG_DEPENDENT_HORZTILES
-    // Are edges available for intra prediction?
-    xd->up_available = (mi_row > tile->mi_row_start);
-#if CONFIG_DEPENDENT_HORZTILES
-  }
-#endif  // CONFIG_DEPENDENT_HORZTILES
+  // Are edges available for intra prediction?
+  xd->up_available = (mi_row > tile->mi_row_start);
+
+  const int ss_x = xd->plane[1].subsampling_x;
+  const int ss_y = xd->plane[1].subsampling_y;
 
   xd->left_available = (mi_col > tile->mi_col_start);
-#if CONFIG_CHROMA_SUB8X8
   xd->chroma_up_available = xd->up_available;
   xd->chroma_left_available = xd->left_available;
-  if (xd->plane[1].subsampling_x && bw < mi_size_wide[BLOCK_8X8])
+  if (ss_x && bw < mi_size_wide[BLOCK_8X8])
     xd->chroma_left_available = (mi_col - 1) > tile->mi_col_start;
-  if (xd->plane[1].subsampling_y && bh < mi_size_high[BLOCK_8X8])
+  if (ss_y && bh < mi_size_high[BLOCK_8X8])
     xd->chroma_up_available = (mi_row - 1) > tile->mi_row_start;
-#endif
   if (xd->up_available) {
-    xd->above_mi = xd->mi[-xd->mi_stride];
-    // above_mi may be NULL in encoder's first pass.
-    xd->above_mbmi = xd->above_mi ? &xd->above_mi->mbmi : NULL;
+    xd->above_mbmi = xd->mi[-xd->mi_stride];
   } else {
-    xd->above_mi = NULL;
     xd->above_mbmi = NULL;
   }
 
   if (xd->left_available) {
-    xd->left_mi = xd->mi[-1];
-    // left_mi may be NULL in encoder's first pass.
-    xd->left_mbmi = xd->left_mi ? &xd->left_mi->mbmi : NULL;
+    xd->left_mbmi = xd->mi[-1];
   } else {
-    xd->left_mi = NULL;
     xd->left_mbmi = NULL;
   }
 
+  const int chroma_ref = ((mi_row & 0x01) || !(bh & 0x01) || !ss_y) &&
+                         ((mi_col & 0x01) || !(bw & 0x01) || !ss_x);
+  if (chroma_ref) {
+    // To help calculate the "above" and "left" chroma blocks, note that the
+    // current block may cover multiple luma blocks (eg, if partitioned into
+    // 4x4 luma blocks).
+    // First, find the top-left-most luma block covered by this chroma block
+    MB_MODE_INFO **base_mi =
+        &xd->mi[-(mi_row & ss_y) * xd->mi_stride - (mi_col & ss_x)];
+
+    // Then, we consider the luma region covered by the left or above 4x4 chroma
+    // prediction. We want to point to the chroma reference block in that
+    // region, which is the bottom-right-most mi unit.
+    // This leads to the following offsets:
+    MB_MODE_INFO *chroma_above_mi =
+        xd->chroma_up_available ? base_mi[-xd->mi_stride + ss_x] : NULL;
+    xd->chroma_above_mbmi = chroma_above_mi;
+
+    MB_MODE_INFO *chroma_left_mi =
+        xd->chroma_left_available ? base_mi[ss_y * xd->mi_stride - 1] : NULL;
+    xd->chroma_left_mbmi = chroma_left_mi;
+  }
+
   xd->n8_h = bh;
   xd->n8_w = bw;
   xd->is_sec_rect = 0;
-  if (xd->n8_w < xd->n8_h)
-    if (mi_col & (xd->n8_h - 1)) xd->is_sec_rect = 1;
+  if (xd->n8_w < xd->n8_h) {
+    // Only mark is_sec_rect as 1 for the last block.
+    // For PARTITION_VERT_4, it would be (0, 0, 0, 1);
+    // For other partitions, it would be (0, 1).
+    if (!((mi_col + xd->n8_w) & (xd->n8_h - 1))) xd->is_sec_rect = 1;
+  }
 
   if (xd->n8_w > xd->n8_h)
     if (mi_row & (xd->n8_w - 1)) xd->is_sec_rect = 1;
 }
 
-static INLINE const aom_prob *get_y_mode_probs(const AV1_COMMON *cm,
-                                               const MODE_INFO *mi,
-                                               const MODE_INFO *above_mi,
-                                               const MODE_INFO *left_mi,
-                                               int block) {
-  const PREDICTION_MODE above = av1_above_block_mode(mi, above_mi, block);
-  const PREDICTION_MODE left = av1_left_block_mode(mi, left_mi, block);
-  return cm->kf_y_prob[above][left];
-}
-
 static INLINE aom_cdf_prob *get_y_mode_cdf(FRAME_CONTEXT *tile_ctx,
-                                           const MODE_INFO *mi,
-                                           const MODE_INFO *above_mi,
-                                           const MODE_INFO *left_mi,
-                                           int block) {
-  const PREDICTION_MODE above = av1_above_block_mode(mi, above_mi, block);
-  const PREDICTION_MODE left = av1_left_block_mode(mi, left_mi, block);
-
-#if CONFIG_KF_CTX
-  int above_ctx = intra_mode_context[above];
-  int left_ctx = intra_mode_context[left];
+                                           const MB_MODE_INFO *above_mi,
+                                           const MB_MODE_INFO *left_mi) {
+  const PREDICTION_MODE above = av1_above_block_mode(above_mi);
+  const PREDICTION_MODE left = av1_left_block_mode(left_mi);
+  const int above_ctx = intra_mode_context[above];
+  const int left_ctx = intra_mode_context[left];
   return tile_ctx->kf_y_cdf[above_ctx][left_ctx];
-#else
-  return tile_ctx->kf_y_cdf[above][left];
-#endif
 }
 
 static INLINE void update_partition_context(MACROBLOCKD *xd, int mi_row,
@@ -879,130 +861,117 @@ static INLINE void update_partition_context(MACROBLOCKD *xd, int mi_row,
   PARTITION_CONTEXT *const left_ctx =
       xd->left_seg_context + (mi_row & MAX_MIB_MASK);
 
-#if CONFIG_EXT_PARTITION_TYPES
   const int bw = mi_size_wide[bsize];
   const int bh = mi_size_high[bsize];
   memset(above_ctx, partition_context_lookup[subsize].above, bw);
   memset(left_ctx, partition_context_lookup[subsize].left, bh);
-#else
-  // num_4x4_blocks_wide_lookup[bsize] / 2
-  const int bs = mi_size_wide[bsize];
-
-  // update the partition context at the end notes. set partition bits
-  // of block sizes larger than the current one to be one, and partition
-  // bits of smaller block sizes to be zero.
-  memset(above_ctx, partition_context_lookup[subsize].above, bs);
-  memset(left_ctx, partition_context_lookup[subsize].left, bs);
-#endif  // CONFIG_EXT_PARTITION_TYPES
 }
 
-#if CONFIG_CB4X4
 static INLINE int is_chroma_reference(int mi_row, int mi_col, BLOCK_SIZE bsize,
                                       int subsampling_x, int subsampling_y) {
-#if CONFIG_CHROMA_2X2
-  return 1;
-#endif
-
-#if CONFIG_CHROMA_SUB8X8
   const int bw = mi_size_wide[bsize];
   const int bh = mi_size_high[bsize];
-
   int ref_pos = ((mi_row & 0x01) || !(bh & 0x01) || !subsampling_y) &&
                 ((mi_col & 0x01) || !(bw & 0x01) || !subsampling_x);
-
-  return ref_pos;
-#else
-  int ref_pos = !(((mi_row & 0x01) && subsampling_y) ||
-                  ((mi_col & 0x01) && subsampling_x));
-
-  if (bsize >= BLOCK_8X8) ref_pos = 1;
-
   return ref_pos;
-#endif
-}
-
-#if CONFIG_SUPERTX
-static INLINE int need_handle_chroma_sub8x8(BLOCK_SIZE bsize, int subsampling_x,
-                                            int subsampling_y) {
-  const int bw = mi_size_wide[bsize];
-  const int bh = mi_size_high[bsize];
-
-  if (bsize >= BLOCK_8X8 ||
-      ((!(bh & 0x01) || !subsampling_y) && (!(bw & 0x01) || !subsampling_x)))
-    return 0;
-  else
-    return 1;
 }
-#endif
 
 static INLINE BLOCK_SIZE scale_chroma_bsize(BLOCK_SIZE bsize, int subsampling_x,
                                             int subsampling_y) {
   BLOCK_SIZE bs = bsize;
-
-  if (bs < BLOCK_8X8) {
-    if (subsampling_x == 1 && subsampling_y == 1)
-      bs = BLOCK_8X8;
-    else if (subsampling_x == 1)
-      bs = BLOCK_8X4;
-    else if (subsampling_y == 1)
-      bs = BLOCK_4X8;
+  switch (bsize) {
+    case BLOCK_4X4:
+      if (subsampling_x == 1 && subsampling_y == 1)
+        bs = BLOCK_8X8;
+      else if (subsampling_x == 1)
+        bs = BLOCK_8X4;
+      else if (subsampling_y == 1)
+        bs = BLOCK_4X8;
+      break;
+    case BLOCK_4X8:
+      if (subsampling_x == 1 && subsampling_y == 1)
+        bs = BLOCK_8X8;
+      else if (subsampling_x == 1)
+        bs = BLOCK_8X8;
+      else if (subsampling_y == 1)
+        bs = BLOCK_4X8;
+      break;
+    case BLOCK_8X4:
+      if (subsampling_x == 1 && subsampling_y == 1)
+        bs = BLOCK_8X8;
+      else if (subsampling_x == 1)
+        bs = BLOCK_8X4;
+      else if (subsampling_y == 1)
+        bs = BLOCK_8X8;
+      break;
+    case BLOCK_4X16:
+      if (subsampling_x == 1 && subsampling_y == 1)
+        bs = BLOCK_8X16;
+      else if (subsampling_x == 1)
+        bs = BLOCK_8X16;
+      else if (subsampling_y == 1)
+        bs = BLOCK_4X16;
+      break;
+    case BLOCK_16X4:
+      if (subsampling_x == 1 && subsampling_y == 1)
+        bs = BLOCK_16X8;
+      else if (subsampling_x == 1)
+        bs = BLOCK_16X4;
+      else if (subsampling_y == 1)
+        bs = BLOCK_16X8;
+      break;
+    default: break;
   }
-
   return bs;
 }
-#endif
 
 static INLINE aom_cdf_prob cdf_element_prob(const aom_cdf_prob *cdf,
                                             size_t element) {
   assert(cdf != NULL);
-#if !CONFIG_ANS
   return (element > 0 ? cdf[element - 1] : CDF_PROB_TOP) - cdf[element];
-#else
-  return cdf[element] - (element > 0 ? cdf[element - 1] : 0);
-#endif
 }
 
 static INLINE void partition_gather_horz_alike(aom_cdf_prob *out,
-                                               const aom_cdf_prob *const in) {
+                                               const aom_cdf_prob *const in,
+                                               BLOCK_SIZE bsize) {
+  (void)bsize;
   out[0] = CDF_PROB_TOP;
   out[0] -= cdf_element_prob(in, PARTITION_HORZ);
   out[0] -= cdf_element_prob(in, PARTITION_SPLIT);
-#if CONFIG_EXT_PARTITION_TYPES
   out[0] -= cdf_element_prob(in, PARTITION_HORZ_A);
   out[0] -= cdf_element_prob(in, PARTITION_HORZ_B);
   out[0] -= cdf_element_prob(in, PARTITION_VERT_A);
-#endif
+  if (bsize != BLOCK_128X128) out[0] -= cdf_element_prob(in, PARTITION_HORZ_4);
   out[0] = AOM_ICDF(out[0]);
   out[1] = AOM_ICDF(CDF_PROB_TOP);
 }
 
 static INLINE void partition_gather_vert_alike(aom_cdf_prob *out,
-                                               const aom_cdf_prob *const in) {
+                                               const aom_cdf_prob *const in,
+                                               BLOCK_SIZE bsize) {
+  (void)bsize;
   out[0] = CDF_PROB_TOP;
   out[0] -= cdf_element_prob(in, PARTITION_VERT);
   out[0] -= cdf_element_prob(in, PARTITION_SPLIT);
-#if CONFIG_EXT_PARTITION_TYPES
   out[0] -= cdf_element_prob(in, PARTITION_HORZ_A);
   out[0] -= cdf_element_prob(in, PARTITION_VERT_A);
   out[0] -= cdf_element_prob(in, PARTITION_VERT_B);
-#endif
+  if (bsize != BLOCK_128X128) out[0] -= cdf_element_prob(in, PARTITION_VERT_4);
   out[0] = AOM_ICDF(out[0]);
   out[1] = AOM_ICDF(CDF_PROB_TOP);
 }
 
-#if CONFIG_EXT_PARTITION_TYPES
 static INLINE void update_ext_partition_context(MACROBLOCKD *xd, int mi_row,
                                                 int mi_col, BLOCK_SIZE subsize,
                                                 BLOCK_SIZE bsize,
                                                 PARTITION_TYPE partition) {
   if (bsize >= BLOCK_8X8) {
-#if !CONFIG_EXT_PARTITION_TYPES_AB
     const int hbs = mi_size_wide[bsize] / 2;
-    BLOCK_SIZE bsize2 = get_subsize(bsize, PARTITION_SPLIT);
-#endif
+    BLOCK_SIZE bsize2 = get_partition_subsize(bsize, PARTITION_SPLIT);
     switch (partition) {
       case PARTITION_SPLIT:
         if (bsize != BLOCK_8X8) break;
+        AOM_FALLTHROUGH_INTENDED;
       case PARTITION_NONE:
       case PARTITION_HORZ:
       case PARTITION_VERT:
@@ -1010,30 +979,6 @@ static INLINE void update_ext_partition_context(MACROBLOCKD *xd, int mi_row,
       case PARTITION_VERT_4:
         update_partition_context(xd, mi_row, mi_col, subsize, bsize);
         break;
-#if CONFIG_EXT_PARTITION_TYPES_AB
-      case PARTITION_HORZ_A:
-        update_partition_context(xd, mi_row, mi_col,
-                                 get_subsize(bsize, PARTITION_HORZ_4), subsize);
-        update_partition_context(xd, mi_row + mi_size_high[bsize] / 2, mi_col,
-                                 subsize, subsize);
-        break;
-      case PARTITION_HORZ_B:
-        update_partition_context(xd, mi_row, mi_col, subsize, subsize);
-        update_partition_context(xd, mi_row + mi_size_high[bsize] / 2, mi_col,
-                                 get_subsize(bsize, PARTITION_HORZ_4), subsize);
-        break;
-      case PARTITION_VERT_A:
-        update_partition_context(xd, mi_row, mi_col,
-                                 get_subsize(bsize, PARTITION_VERT_4), subsize);
-        update_partition_context(xd, mi_row, mi_col + mi_size_wide[bsize] / 2,
-                                 subsize, subsize);
-        break;
-      case PARTITION_VERT_B:
-        update_partition_context(xd, mi_row, mi_col, subsize, subsize);
-        update_partition_context(xd, mi_row, mi_col + mi_size_wide[bsize] / 2,
-                                 get_subsize(bsize, PARTITION_VERT_4), subsize);
-        break;
-#else
       case PARTITION_HORZ_A:
         update_partition_context(xd, mi_row, mi_col, bsize2, subsize);
         update_partition_context(xd, mi_row + hbs, mi_col, subsize, subsize);
@@ -1050,41 +995,35 @@ static INLINE void update_ext_partition_context(MACROBLOCKD *xd, int mi_row,
         update_partition_context(xd, mi_row, mi_col, subsize, subsize);
         update_partition_context(xd, mi_row, mi_col + hbs, bsize2, subsize);
         break;
-#endif
       default: assert(0 && "Invalid partition type");
     }
   }
 }
-#endif  // CONFIG_EXT_PARTITION_TYPES
 
 static INLINE int partition_plane_context(const MACROBLOCKD *xd, int mi_row,
-                                          int mi_col,
-#if CONFIG_UNPOISON_PARTITION_CTX
-                                          int has_rows, int has_cols,
-#endif
-                                          BLOCK_SIZE bsize) {
+                                          int mi_col, BLOCK_SIZE bsize) {
   const PARTITION_CONTEXT *above_ctx = xd->above_seg_context + mi_col;
   const PARTITION_CONTEXT *left_ctx =
       xd->left_seg_context + (mi_row & MAX_MIB_MASK);
   // Minimum partition point is 8x8. Offset the bsl accordingly.
-  const int bsl = mi_width_log2_lookup[bsize] - mi_width_log2_lookup[BLOCK_8X8];
+  const int bsl = mi_size_wide_log2[bsize] - mi_size_wide_log2[BLOCK_8X8];
   int above = (*above_ctx >> bsl) & 1, left = (*left_ctx >> bsl) & 1;
 
-  assert(b_width_log2_lookup[bsize] == b_height_log2_lookup[bsize]);
+  assert(mi_size_wide_log2[bsize] == mi_size_high_log2[bsize]);
   assert(bsl >= 0);
 
-#if CONFIG_UNPOISON_PARTITION_CTX
-  if (has_rows && has_cols)
-    return (left * 2 + above) + bsl * PARTITION_PLOFFSET;
-  else if (has_rows && !has_cols)
-    return PARTITION_CONTEXTS_PRIMARY + bsl;
-  else if (!has_rows && has_cols)
-    return PARTITION_CONTEXTS_PRIMARY + PARTITION_BLOCK_SIZES + bsl;
-  else
-    return INVALID_PARTITION_CTX;  // Bogus context, forced SPLIT
-#else
   return (left * 2 + above) + bsl * PARTITION_PLOFFSET;
-#endif
+}
+
+// Return the number of elements in the partition CDF when
+// partitioning the (square) block with luma block size of bsize.
+static INLINE int partition_cdf_length(BLOCK_SIZE bsize) {
+  if (bsize <= BLOCK_8X8)
+    return PARTITION_TYPES;
+  else if (bsize == BLOCK_128X128)
+    return EXT_PARTITION_TYPES - 2;
+  else
+    return EXT_PARTITION_TYPES;
 }
 
 static INLINE int max_block_wide(const MACROBLOCKD *xd, BLOCK_SIZE bsize,
@@ -1107,11 +1046,10 @@ static INLINE int max_block_high(const MACROBLOCKD *xd, BLOCK_SIZE bsize,
   if (xd->mb_to_bottom_edge < 0)
     max_blocks_high += xd->mb_to_bottom_edge >> (3 + pd->subsampling_y);
 
-  // Scale the width in the transform block unit.
-  return max_blocks_high >> tx_size_wide_log2[0];
+  // Scale the height in the transform block unit.
+  return max_blocks_high >> tx_size_high_log2[0];
 }
 
-#if CONFIG_CFL
 static INLINE int max_intra_block_width(const MACROBLOCKD *xd,
                                         BLOCK_SIZE plane_bsize, int plane,
                                         TX_SIZE tx_size) {
@@ -1127,36 +1065,43 @@ static INLINE int max_intra_block_height(const MACROBLOCKD *xd,
                               << tx_size_high_log2[0];
   return ALIGN_POWER_OF_TWO(max_blocks_high, tx_size_high_log2[tx_size]);
 }
-#endif  // CONFIG_CFL
 
 static INLINE void av1_zero_above_context(AV1_COMMON *const cm,
-                                          int mi_col_start, int mi_col_end) {
+  int mi_col_start, int mi_col_end, const int tile_row) {
+  const int num_planes = av1_num_planes(cm);
   const int width = mi_col_end - mi_col_start;
-  const int aligned_width = ALIGN_POWER_OF_TWO(width, cm->mib_size_log2);
+  const int aligned_width =
+    ALIGN_POWER_OF_TWO(width, cm->seq_params.mib_size_log2);
 
-  const int offset_y = mi_col_start << (MI_SIZE_LOG2 - tx_size_wide_log2[0]);
-  const int width_y = aligned_width << (MI_SIZE_LOG2 - tx_size_wide_log2[0]);
+  const int offset_y = mi_col_start;
+  const int width_y = aligned_width;
   const int offset_uv = offset_y >> cm->subsampling_x;
   const int width_uv = width_y >> cm->subsampling_x;
 
-  av1_zero_array(cm->above_context[0] + offset_y, width_y);
-  av1_zero_array(cm->above_context[1] + offset_uv, width_uv);
-  av1_zero_array(cm->above_context[2] + offset_uv, width_uv);
+  av1_zero_array(cm->above_context[0][tile_row] + offset_y, width_y);
+  if (num_planes > 1) {
+    if (cm->above_context[1][tile_row] && cm->above_context[2][tile_row]) {
+      av1_zero_array(cm->above_context[1][tile_row] + offset_uv, width_uv);
+      av1_zero_array(cm->above_context[2][tile_row] + offset_uv, width_uv);
+    } else {
+      aom_internal_error(&cm->error, AOM_CODEC_CORRUPT_FRAME,
+                         "Invalid value of planes");
+    }
+  }
 
-  av1_zero_array(cm->above_seg_context + mi_col_start, aligned_width);
+  av1_zero_array(cm->above_seg_context[tile_row] + mi_col_start, aligned_width);
 
-#if CONFIG_VAR_TX
-  av1_zero_array(cm->above_txfm_context + (mi_col_start << TX_UNIT_WIDE_LOG2),
-                 aligned_width << TX_UNIT_WIDE_LOG2);
-#endif  // CONFIG_VAR_TX
+  memset(cm->above_txfm_context[tile_row] + mi_col_start,
+    tx_size_wide[TX_SIZES_LARGEST],
+    aligned_width * sizeof(TXFM_CONTEXT));
 }
 
 static INLINE void av1_zero_left_context(MACROBLOCKD *const xd) {
   av1_zero(xd->left_context);
   av1_zero(xd->left_seg_context);
-#if CONFIG_VAR_TX
-  av1_zero(xd->left_txfm_context_buffer);
-#endif
+
+  memset(xd->left_txfm_context_buffer, tx_size_high[TX_SIZES_LARGEST],
+         sizeof(xd->left_txfm_context_buffer));
 }
 
 // Disable array-bounds checks as the TX_SIZE enum contains values larger than
@@ -1166,15 +1111,11 @@ static INLINE void av1_zero_left_context(MACROBLOCKD *const xd) {
 #if defined(__GNUC__) && __GNUC__ >= 4
 #pragma GCC diagnostic ignored "-Warray-bounds"
 #endif
-static INLINE TX_SIZE get_min_tx_size(TX_SIZE tx_size) {
-  assert(tx_size < TX_SIZES_ALL);
-  return txsize_sqr_map[tx_size];
-}
+
 #if defined(__GNUC__) && __GNUC__ >= 4
 #pragma GCC diagnostic warning "-Warray-bounds"
 #endif
 
-#if CONFIG_VAR_TX
 static INLINE void set_txfm_ctx(TXFM_CONTEXT *txfm_ctx, uint8_t txs, int len) {
   int i;
   for (i = 0; i < len; ++i) txfm_ctx[i] = txs;
@@ -1190,16 +1131,16 @@ static INLINE void set_txfm_ctxs(TX_SIZE tx_size, int n8_w, int n8_h, int skip,
     bh = n8_h * MI_SIZE;
   }
 
-  set_txfm_ctx(xd->above_txfm_context, bw, n8_w << TX_UNIT_WIDE_LOG2);
-  set_txfm_ctx(xd->left_txfm_context, bh, n8_h << TX_UNIT_HIGH_LOG2);
+  set_txfm_ctx(xd->above_txfm_context, bw, n8_w);
+  set_txfm_ctx(xd->left_txfm_context, bh, n8_h);
 }
 
 static INLINE void txfm_partition_update(TXFM_CONTEXT *above_ctx,
                                          TXFM_CONTEXT *left_ctx,
                                          TX_SIZE tx_size, TX_SIZE txb_size) {
   BLOCK_SIZE bsize = txsize_to_bsize[txb_size];
-  int bh = mi_size_high[bsize] << TX_UNIT_HIGH_LOG2;
-  int bw = mi_size_wide[bsize] << TX_UNIT_WIDE_LOG2;
+  int bh = mi_size_high[bsize];
+  int bw = mi_size_wide[bsize];
   uint8_t txw = tx_size_wide[tx_size];
   uint8_t txh = tx_size_high[tx_size];
   int i;
@@ -1209,16 +1150,8 @@ static INLINE void txfm_partition_update(TXFM_CONTEXT *above_ctx,
 
 static INLINE TX_SIZE get_sqr_tx_size(int tx_dim) {
   switch (tx_dim) {
-#if CONFIG_EXT_PARTITION
     case 128:
-#endif  // CONFIG_EXT_PARTITION
-    case 64:
-#if CONFIG_TX64X64
-      return TX_64X64;
-#else
-      return TX_32X32;
-#endif  // CONFIG_TX64X64
-      break;
+    case 64: return TX_64X64; break;
     case 32: return TX_32X32; break;
     case 16: return TX_16X16; break;
     case 8: return TX_8X8; break;
@@ -1226,6 +1159,45 @@ static INLINE TX_SIZE get_sqr_tx_size(int tx_dim) {
   }
 }
 
+static INLINE TX_SIZE get_tx_size(int width, int height) {
+  if (width == height) {
+    return get_sqr_tx_size(width);
+  }
+  if (width < height) {
+    if (width + width == height) {
+      switch (width) {
+        case 4: return TX_4X8; break;
+        case 8: return TX_8X16; break;
+        case 16: return TX_16X32; break;
+        case 32: return TX_32X64; break;
+      }
+    } else {
+      switch (width) {
+        case 4: return TX_4X16; break;
+        case 8: return TX_8X32; break;
+        case 16: return TX_16X64; break;
+      }
+    }
+  } else {
+    if (height + height == width) {
+      switch (height) {
+        case 4: return TX_8X4; break;
+        case 8: return TX_16X8; break;
+        case 16: return TX_32X16; break;
+        case 32: return TX_64X32; break;
+      }
+    } else {
+      switch (height) {
+        case 4: return TX_16X4; break;
+        case 8: return TX_32X8; break;
+        case 16: return TX_64X16; break;
+      }
+    }
+  }
+  assert(0);
+  return TX_4X4;
+}
+
 static INLINE int txfm_partition_context(TXFM_CONTEXT *above_ctx,
                                          TXFM_CONTEXT *left_ctx,
                                          BLOCK_SIZE bsize, TX_SIZE tx_size) {
@@ -1233,7 +1205,7 @@ static INLINE int txfm_partition_context(TXFM_CONTEXT *above_ctx,
   const uint8_t txh = tx_size_high[tx_size];
   const int above = *above_ctx < txw;
   const int left = *left_ctx < txh;
-  int category = TXFM_PARTITION_CONTEXTS - 1;
+  int category = TXFM_PARTITION_CONTEXTS;
 
   // dummy return, not used by others.
   if (tx_size <= TX_4X4) return 0;
@@ -1242,13 +1214,13 @@ static INLINE int txfm_partition_context(TXFM_CONTEXT *above_ctx,
       get_sqr_tx_size(AOMMAX(block_size_wide[bsize], block_size_high[bsize]));
 
   if (max_tx_size >= TX_8X8) {
-    category = (tx_size != max_tx_size && max_tx_size > TX_8X8) +
-               (TX_SIZES - 1 - max_tx_size) * 2;
+    category =
+        (txsize_sqr_up_map[tx_size] != max_tx_size && max_tx_size > TX_8X8) +
+        (TX_SIZES - 1 - max_tx_size) * 2;
   }
-  if (category == TXFM_PARTITION_CONTEXTS - 1) return category;
+  assert(category != TXFM_PARTITION_CONTEXTS);
   return category * 3 + above + left;
 }
-#endif
 
 // Compute the next partition in the direction of the sb_type stored in the mi
 // array, starting with bsize.
@@ -1258,8 +1230,8 @@ static INLINE PARTITION_TYPE get_partition(const AV1_COMMON *const cm,
   if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols) return PARTITION_INVALID;
 
   const int offset = mi_row * cm->mi_stride + mi_col;
-  MODE_INFO **mi = cm->mi_grid_visible + offset;
-  const BLOCK_SIZE subsize = mi[0]->mbmi.sb_type;
+  MB_MODE_INFO **mi = cm->mi_grid_visible + offset;
+  const BLOCK_SIZE subsize = mi[0]->sb_type;
 
   if (subsize == bsize) return PARTITION_NONE;
 
@@ -1268,25 +1240,14 @@ static INLINE PARTITION_TYPE get_partition(const AV1_COMMON *const cm,
   const int sshigh = mi_size_high[subsize];
   const int sswide = mi_size_wide[subsize];
 
-#if CONFIG_EXT_PARTITION_TYPES
   if (bsize > BLOCK_8X8 && mi_row + bwide / 2 < cm->mi_rows &&
       mi_col + bhigh / 2 < cm->mi_cols) {
     // In this case, the block might be using an extended partition
     // type.
-    const MB_MODE_INFO *const mbmi_right = &mi[bwide / 2]->mbmi;
-    const MB_MODE_INFO *const mbmi_below = &mi[bhigh / 2 * cm->mi_stride]->mbmi;
+    const MB_MODE_INFO *const mbmi_right = mi[bwide / 2];
+    const MB_MODE_INFO *const mbmi_below = mi[bhigh / 2 * cm->mi_stride];
 
     if (sswide == bwide) {
-#if CONFIG_EXT_PARTITION_TYPES_AB
-      // Smaller height but same width. Is PARTITION_HORZ, PARTITION_HORZ_4,
-      // PARTITION_HORZ_A or PARTITION_HORZ_B.
-      if (sshigh * 2 == bhigh)
-        return (mbmi_below->sb_type == subsize) ? PARTITION_HORZ
-                                                : PARTITION_HORZ_B;
-      assert(sshigh * 4 == bhigh);
-      return (mbmi_below->sb_type == subsize) ? PARTITION_HORZ_4
-                                              : PARTITION_HORZ_A;
-#else
       // Smaller height but same width. Is PARTITION_HORZ_4, PARTITION_HORZ or
       // PARTITION_HORZ_B. To distinguish the latter two, check if the lower
       // half was split.
@@ -1297,18 +1258,7 @@ static INLINE PARTITION_TYPE get_partition(const AV1_COMMON *const cm,
         return PARTITION_HORZ;
       else
         return PARTITION_HORZ_B;
-#endif
     } else if (sshigh == bhigh) {
-#if CONFIG_EXT_PARTITION_TYPES_AB
-      // Smaller width but same height. Is PARTITION_VERT, PARTITION_VERT_4,
-      // PARTITION_VERT_A or PARTITION_VERT_B.
-      if (sswide * 2 == bwide)
-        return (mbmi_right->sb_type == subsize) ? PARTITION_VERT
-                                                : PARTITION_VERT_B;
-      assert(sswide * 4 == bwide);
-      return (mbmi_right->sb_type == subsize) ? PARTITION_VERT_4
-                                              : PARTITION_VERT_A;
-#else
       // Smaller width but same height. Is PARTITION_VERT_4, PARTITION_VERT or
       // PARTITION_VERT_B. To distinguish the latter two, check if the right
       // half was split.
@@ -1319,9 +1269,7 @@ static INLINE PARTITION_TYPE get_partition(const AV1_COMMON *const cm,
         return PARTITION_VERT;
       else
         return PARTITION_VERT_B;
-#endif
     } else {
-#if !CONFIG_EXT_PARTITION_TYPES_AB
       // Smaller width and smaller height. Might be PARTITION_SPLIT or could be
       // PARTITION_HORZ_A or PARTITION_VERT_A. If subsize isn't halved in both
       // dimensions, we immediately know this is a split (which will recurse to
@@ -1333,12 +1281,10 @@ static INLINE PARTITION_TYPE get_partition(const AV1_COMMON *const cm,
 
       if (mi_size_wide[mbmi_below->sb_type] == bwide) return PARTITION_HORZ_A;
       if (mi_size_high[mbmi_right->sb_type] == bhigh) return PARTITION_VERT_A;
-#endif
 
       return PARTITION_SPLIT;
     }
   }
-#endif
   const int vert_split = sswide < bwide;
   const int horz_split = sshigh < bhigh;
   const int split_idx = (vert_split << 1) | horz_split;
@@ -1352,49 +1298,46 @@ static INLINE PARTITION_TYPE get_partition(const AV1_COMMON *const cm,
 }
 
 static INLINE void set_use_reference_buffer(AV1_COMMON *const cm, int use) {
-#if CONFIG_REFERENCE_BUFFER
   cm->seq_params.frame_id_numbers_present_flag = use;
-#else
-  (void)cm;
-  (void)use;
-#endif
 }
 
-static INLINE void set_sb_size(AV1_COMMON *const cm, BLOCK_SIZE sb_size) {
-  cm->sb_size = sb_size;
-  cm->mib_size = mi_size_wide[cm->sb_size];
-#if CONFIG_CB4X4
-  cm->mib_size_log2 = b_width_log2_lookup[cm->sb_size];
-#else
-  cm->mib_size_log2 = mi_width_log2_lookup[cm->sb_size];
-#endif
+static INLINE void set_sb_size(SequenceHeader *const seq_params,
+                               BLOCK_SIZE sb_size) {
+  seq_params->sb_size = sb_size;
+  seq_params->mib_size = mi_size_wide[seq_params->sb_size];
+  seq_params->mib_size_log2 = mi_size_wide_log2[seq_params->sb_size];
 }
 
-static INLINE int all_lossless(const AV1_COMMON *cm, const MACROBLOCKD *xd) {
-  int i;
-  int all_lossless = 1;
+// Returns true if the frame is fully lossless at the coded resolution.
+// Note: If super-resolution is used, such a frame will still NOT be lossless at
+// the upscaled resolution.
+static INLINE int is_coded_lossless(const AV1_COMMON *cm,
+                                    const MACROBLOCKD *xd) {
+  int coded_lossless = 1;
   if (cm->seg.enabled) {
-    for (i = 0; i < MAX_SEGMENTS; ++i) {
+    for (int i = 0; i < MAX_SEGMENTS; ++i) {
       if (!xd->lossless[i]) {
-        all_lossless = 0;
+        coded_lossless = 0;
         break;
       }
     }
   } else {
-    all_lossless = xd->lossless[0];
+    coded_lossless = xd->lossless[0];
   }
-  return all_lossless;
+  return coded_lossless;
 }
 
-static INLINE int use_compressed_header(const AV1_COMMON *cm) {
-  (void)cm;
-#if CONFIG_RESTRICT_COMPRESSED_HDR && CONFIG_NEW_MULTISYMBOL
-  return 0;
-#elif CONFIG_RESTRICT_COMPRESSED_HDR
-  return cm->refresh_frame_context == REFRESH_FRAME_CONTEXT_FORWARD;
-#else
-  return 1;
-#endif  // CONFIG_RESTRICT_COMPRESSED_HDR && CONFIG_NEW_MULTISYMBOL
+static INLINE int is_valid_seq_level_idx(uint8_t seq_level_idx) {
+  return seq_level_idx < 24 || seq_level_idx == 31;
+}
+
+static INLINE uint8_t major_minor_to_seq_level_idx(BitstreamLevel bl) {
+  assert(bl.major >= LEVEL_MAJOR_MIN && bl.major <= LEVEL_MAJOR_MAX);
+  // Since bl.minor is unsigned a comparison will return a warning:
+  // comparison is always true due to limited range of data type
+  assert(LEVEL_MINOR_MIN == 0);
+  assert(bl.minor <= LEVEL_MINOR_MAX);
+  return ((bl.major - LEVEL_MAJOR_MIN) << LEVEL_MINOR_BITS) + bl.minor;
 }
 
 #ifdef __cplusplus