Home | History | Annotate | Download | only in encoder
      1 /*
      2  *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
      3  *
      4  *  Use of this source code is governed by a BSD-style license
      5  *  that can be found in the LICENSE file in the root of the source
      6  *  tree. An additional intellectual property rights grant can be found
      7  *  in the file PATENTS.  All contributing project authors may
      8  *  be found in the AUTHORS file in the root of the source tree.
      9  */
     10 
     11 #ifndef VP9_ENCODER_VP9_SPEED_FEATURES_H_
     12 #define VP9_ENCODER_VP9_SPEED_FEATURES_H_
     13 
     14 #include "vp9/common/vp9_enums.h"
     15 
     16 #ifdef __cplusplus
     17 extern "C" {
     18 #endif
     19 
     20 typedef enum {
     21   DIAMOND = 0,
     22   NSTEP = 1,
     23   HEX = 2,
     24   BIGDIA = 3,
     25   SQUARE = 4,
     26   FAST_HEX = 5,
     27   FAST_DIAMOND = 6
     28 } SEARCH_METHODS;
     29 
     30 typedef enum {
     31   // No recode.
     32   DISALLOW_RECODE = 0,
     33   // Allow recode for KF and exceeding maximum frame bandwidth.
     34   ALLOW_RECODE_KFMAXBW = 1,
     35   // Allow recode only for KF/ARF/GF frames.
     36   ALLOW_RECODE_KFARFGF = 2,
     37   // Allow recode for all frames based on bitrate constraints.
     38   ALLOW_RECODE = 3,
     39 } RECODE_LOOP_TYPE;
     40 
     41 typedef enum {
     42   SUBPEL_TREE = 0,
     43   // Other methods to come
     44 } SUBPEL_SEARCH_METHODS;
     45 
     46 typedef enum {
     47   LAST_FRAME_PARTITION_OFF = 0,
     48   LAST_FRAME_PARTITION_LOW_MOTION = 1,
     49   LAST_FRAME_PARTITION_ALL = 2
     50 } LAST_FRAME_PARTITION_METHOD;
     51 
     52 typedef enum {
     53   USE_FULL_RD = 0,
     54   USE_LARGESTINTRA,
     55   USE_LARGESTINTRA_MODELINTER,
     56   USE_LARGESTALL
     57 } TX_SIZE_SEARCH_METHOD;
     58 
     59 typedef enum {
     60   NOT_IN_USE = 0,
     61   RELAXED_NEIGHBORING_MIN_MAX = 1,
     62   STRICT_NEIGHBORING_MIN_MAX = 2
     63 } AUTO_MIN_MAX_MODE;
     64 
     65 typedef enum {
     66   // Try the full image with different values.
     67   LPF_PICK_FROM_FULL_IMAGE,
     68   // Try a small portion of the image with different values.
     69   LPF_PICK_FROM_SUBIMAGE,
     70   // Estimate the level based on quantizer and frame type
     71   LPF_PICK_FROM_Q,
     72 } LPF_PICK_METHOD;
     73 
     74 typedef enum {
     75   // Terminate search early based on distortion so far compared to
     76   // qp step, distortion in the neighborhood of the frame, etc.
     77   FLAG_EARLY_TERMINATE = 1 << 0,
     78 
     79   // Skips comp inter modes if the best so far is an intra mode.
     80   FLAG_SKIP_COMP_BESTINTRA = 1 << 1,
     81 
     82   // Skips comp inter modes if the best single intermode so far does
     83   // not have the same reference as one of the two references being
     84   // tested.
     85   FLAG_SKIP_COMP_REFMISMATCH = 1 << 2,
     86 
     87   // Skips oblique intra modes if the best so far is an inter mode.
     88   FLAG_SKIP_INTRA_BESTINTER = 1 << 3,
     89 
     90   // Skips oblique intra modes  at angles 27, 63, 117, 153 if the best
     91   // intra so far is not one of the neighboring directions.
     92   FLAG_SKIP_INTRA_DIRMISMATCH = 1 << 4,
     93 
     94   // Skips intra modes other than DC_PRED if the source variance is small
     95   FLAG_SKIP_INTRA_LOWVAR = 1 << 5,
     96 } MODE_SEARCH_SKIP_LOGIC;
     97 
     98 typedef enum {
     99   // Search partitions using RD/NONRD criterion
    100   SEARCH_PARTITION = 0,
    101 
    102   // Always use a fixed size partition
    103   FIXED_PARTITION = 1,
    104 
    105   // Use a fixed size partition in every 64X64 SB, where the size is
    106   // determined based on source variance
    107   VAR_BASED_FIXED_PARTITION = 2,
    108 
    109   REFERENCE_PARTITION = 3,
    110 
    111   // Use an arbitrary partitioning scheme based on source variance within
    112   // a 64X64 SB
    113   VAR_BASED_PARTITION,
    114 
    115   // Use non-fixed partitions based on source variance
    116   SOURCE_VAR_BASED_PARTITION
    117 } PARTITION_SEARCH_TYPE;
    118 
    119 typedef enum {
    120   // Does a dry run to see if any of the contexts need to be updated or not,
    121   // before the final run.
    122   TWO_LOOP = 0,
    123 
    124   // No dry run conducted.
    125   ONE_LOOP = 1,
    126 
    127   // No dry run, also only half the coef contexts and bands are updated.
    128   // The rest are not updated at all.
    129   ONE_LOOP_REDUCED = 2
    130 } FAST_COEFF_UPDATE;
    131 
    132 typedef struct {
    133   // Frame level coding parameter update
    134   int frame_parameter_update;
    135 
    136   // Motion search method (Diamond, NSTEP, Hex, Big Diamond, Square, etc).
    137   SEARCH_METHODS search_method;
    138 
    139   RECODE_LOOP_TYPE recode_loop;
    140 
    141   // Subpel_search_method can only be subpel_tree which does a subpixel
    142   // logarithmic search that keeps stepping at 1/2 pixel units until
    143   // you stop getting a gain, and then goes on to 1/4 and repeats
    144   // the same process. Along the way it skips many diagonals.
    145   SUBPEL_SEARCH_METHODS subpel_search_method;
    146 
    147   // Maximum number of steps in logarithmic subpel search before giving up.
    148   int subpel_iters_per_step;
    149 
    150   // Control when to stop subpel search
    151   int subpel_force_stop;
    152 
    153   // This parameter controls the number of steps we'll do in a diamond
    154   // search.
    155   int max_step_search_steps;
    156 
    157   // This parameter controls which step in the n-step process we start at.
    158   // It's changed adaptively based on circumstances.
    159   int reduce_first_step_size;
    160 
    161   // If this is set to 1, we limit the motion search range to 2 times the
    162   // largest motion vector found in the last frame.
    163   int auto_mv_step_size;
    164 
    165   // Trellis (dynamic programming) optimization of quantized values (+1, 0).
    166   int optimize_coefficients;
    167 
    168   // Always set to 0. If on it enables 0 cost background transmission
    169   // (except for the initial transmission of the segmentation). The feature is
    170   // disabled because the addition of very large block sizes make the
    171   // backgrounds very to cheap to encode, and the segmentation we have
    172   // adds overhead.
    173   int static_segmentation;
    174 
    175   // If 1 we iterate finding a best reference for 2 ref frames together - via
    176   // a log search that iterates 4 times (check around mv for last for best
    177   // error of combined predictor then check around mv for alt). If 0 we
    178   // we just use the best motion vector found for each frame by itself.
    179   int comp_inter_joint_search_thresh;
    180 
    181   // This variable is used to cap the maximum number of times we skip testing a
    182   // mode to be evaluated. A high value means we will be faster.
    183   int adaptive_rd_thresh;
    184 
    185   // Enables skipping the reconstruction step (idct, recon) in the
    186   // intermediate steps assuming the last frame didn't have too many intra
    187   // blocks and the q is less than a threshold.
    188   int skip_encode_sb;
    189   int skip_encode_frame;
    190   // Speed feature to allow or disallow skipping of recode at block
    191   // level within a frame.
    192   int allow_skip_recode;
    193 
    194   // This variable allows us to reuse the last frames partition choices
    195   // (64x64 v 32x32 etc) for this frame. It can be set to only use the last
    196   // frame as a starting point in low motion scenes or always use it. If set
    197   // we use last partitioning_redo frequency to determine how often to redo
    198   // the partitioning from scratch. Adjust_partitioning_from_last_frame
    199   // enables us to adjust up or down one partitioning from the last frames
    200   // partitioning.
    201   LAST_FRAME_PARTITION_METHOD use_lastframe_partitioning;
    202 
    203   // Determine which method we use to determine transform size. We can choose
    204   // between options like full rd, largest for prediction size, largest
    205   // for intra and model coefs for the rest.
    206   TX_SIZE_SEARCH_METHOD tx_size_search_method;
    207 
    208   // Low precision 32x32 fdct keeps everything in 16 bits and thus is less
    209   // precise but significantly faster than the non lp version.
    210   int use_lp32x32fdct;
    211 
    212   // TODO(JBB): remove this as its no longer used.
    213 
    214   // After looking at the first set of modes (set by index here), skip
    215   // checking modes for reference frames that don't match the reference frame
    216   // of the best so far.
    217   int mode_skip_start;
    218 
    219   // TODO(JBB): Remove this.
    220   int reference_masking;
    221 
    222   PARTITION_SEARCH_TYPE partition_search_type;
    223 
    224   // Used if partition_search_type = FIXED_SIZE_PARTITION
    225   BLOCK_SIZE always_this_block_size;
    226 
    227   // Skip rectangular partition test when partition type none gives better
    228   // rd than partition type split.
    229   int less_rectangular_check;
    230 
    231   // Disable testing non square partitions. (eg 16x32)
    232   int use_square_partition_only;
    233 
    234   // Sets min and max partition sizes for this 64x64 region based on the
    235   // same 64x64 in last encoded frame, and the left and above neighbor.
    236   AUTO_MIN_MAX_MODE auto_min_max_partition_size;
    237 
    238   // Min and max partition size we enable (block_size) as per auto
    239   // min max, but also used by adjust partitioning, and pick_partitioning.
    240   BLOCK_SIZE min_partition_size;
    241   BLOCK_SIZE max_partition_size;
    242 
    243   // Whether or not we allow partitions one smaller or one greater than the last
    244   // frame's partitioning. Only used if use_lastframe_partitioning is set.
    245   int adjust_partitioning_from_last_frame;
    246 
    247   // How frequently we re do the partitioning from scratch. Only used if
    248   // use_lastframe_partitioning is set.
    249   int last_partitioning_redo_frequency;
    250 
    251   // This enables constrained copy partitioning, which, given an input block
    252   // size bsize, will copy previous partition for partitions less than bsize,
    253   // otherwise bsize partition is used. bsize is currently set to 16x16.
    254   // Used for the case where motion is detected in superblock.
    255   int constrain_copy_partition;
    256 
    257   // Disables sub 8x8 blocksizes in different scenarios: Choices are to disable
    258   // it always, to allow it for only Last frame and Intra, disable it for all
    259   // inter modes or to enable it always.
    260   int disable_split_mask;
    261 
    262   // TODO(jingning): combine the related motion search speed features
    263   // This allows us to use motion search at other sizes as a starting
    264   // point for this motion search and limits the search range around it.
    265   int adaptive_motion_search;
    266 
    267   // Allows sub 8x8 modes to use the prediction filter that was determined
    268   // best for 8x8 mode. If set to 0 we always re check all the filters for
    269   // sizes less than 8x8, 1 means we check all filter modes if no 8x8 filter
    270   // was selected, and 2 means we use 8 tap if no 8x8 filter mode was selected.
    271   int adaptive_pred_interp_filter;
    272 
    273   // Search through variable block partition types in non-RD mode decision
    274   // encoding process for RTC.
    275   int partition_check;
    276 
    277   // Use finer quantizer in every other few frames that run variable block
    278   // partition type search.
    279   int force_frame_boost;
    280 
    281   // Maximally allowed base quantization index fluctuation.
    282   int max_delta_qindex;
    283 
    284   // Implements various heuristics to skip searching modes
    285   // The heuristics selected are based on  flags
    286   // defined in the MODE_SEARCH_SKIP_HEURISTICS enum
    287   unsigned int mode_search_skip_flags;
    288 
    289   // A source variance threshold below which the split mode is disabled
    290   unsigned int disable_split_var_thresh;
    291 
    292   // A source variance threshold below which filter search is disabled
    293   // Choose a very large value (UINT_MAX) to use 8-tap always
    294   unsigned int disable_filter_search_var_thresh;
    295 
    296   // These bit masks allow you to enable or disable intra modes for each
    297   // transform size separately.
    298   int intra_y_mode_mask[TX_SIZES];
    299   int intra_uv_mode_mask[TX_SIZES];
    300 
    301   // This variable enables an early break out of mode testing if the model for
    302   // rd built from the prediction signal indicates a value that's much
    303   // higher than the best rd we've seen so far.
    304   int use_rd_breakout;
    305 
    306   // This enables us to use an estimate for intra rd based on dc mode rather
    307   // than choosing an actual uv mode in the stage of encoding before the actual
    308   // final encode.
    309   int use_uv_intra_rd_estimate;
    310 
    311   // This feature controls how the loop filter level is determined.
    312   LPF_PICK_METHOD lpf_pick;
    313 
    314   // This feature limits the number of coefficients updates we actually do
    315   // by only looking at counts from 1/2 the bands.
    316   FAST_COEFF_UPDATE use_fast_coef_updates;
    317 
    318   // This flag controls the use of non-RD mode decision.
    319   int use_nonrd_pick_mode;
    320 
    321   // This variable sets the encode_breakout threshold. Currently, it is only
    322   // enabled in real time mode.
    323   int encode_breakout_thresh;
    324 
    325   // A binary mask indicating if NEARESTMV, NEARMV, ZEROMV, NEWMV
    326   // modes are disabled in order from LSB to MSB for each BLOCK_SIZE.
    327   int disable_inter_mode_mask[BLOCK_SIZES];
    328 
    329   // This feature controls whether we do the expensive context update and
    330   // calculation in the rd coefficient costing loop.
    331   int use_fast_coef_costing;
    332 
    333   // This feature controls the tolerence vs target used in deciding whether to
    334   // recode a frame. It has no meaning if recode is disabled.
    335   int recode_tolerance;
    336 
    337   // This variable controls the maximum block size where intra blocks can be
    338   // used in inter frames.
    339   // TODO(aconverse): Fold this into one of the other many mode skips
    340   BLOCK_SIZE max_intra_bsize;
    341 
    342   // The frequency that we check if SOURCE_VAR_BASED_PARTITION or
    343   // FIXED_PARTITION search type should be used.
    344   int search_type_check_frequency;
    345 
    346   // The threshold used in SOURCE_VAR_BASED_PARTITION search type.
    347   int source_var_thresh;
    348 } SPEED_FEATURES;
    349 
    350 struct VP9_COMP;
    351 
    352 void vp9_set_speed_features(struct VP9_COMP *cpi);
    353 
    354 #ifdef __cplusplus
    355 }  // extern "C"
    356 #endif
    357 
    358 #endif  // VP9_ENCODER_VP9_SPEED_FEATURES_H_
    359 
    360