AOMedia AV1 Codec
encoder.h
Go to the documentation of this file.
1/*
2 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
3 *
4 * This source code is subject to the terms of the BSD 2 Clause License and
5 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
6 * was not distributed with this source code in the LICENSE file, you can
7 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
8 * Media Patent License 1.0 was not distributed with this source code in the
9 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
10 */
11
15#ifndef AOM_AV1_ENCODER_ENCODER_H_
16#define AOM_AV1_ENCODER_ENCODER_H_
17
18#include <stdbool.h>
19#include <stdio.h>
20
21#include "config/aom_config.h"
22
23#include "aom/aomcx.h"
24
25#include "av1/common/alloccommon.h"
26#include "av1/common/av1_common_int.h"
27#include "av1/common/blockd.h"
28#include "av1/common/entropymode.h"
29#include "av1/common/enums.h"
30#include "av1/common/resize.h"
31#include "av1/common/thread_common.h"
32#include "av1/common/timing.h"
33#include "av1/encoder/aq_cyclicrefresh.h"
34#include "av1/encoder/av1_quantize.h"
35#include "av1/encoder/block.h"
36#include "av1/encoder/context_tree.h"
37#include "av1/encoder/encodemb.h"
38#include "av1/encoder/firstpass.h"
39#include "av1/encoder/global_motion.h"
40#include "av1/encoder/level.h"
42#include "av1/encoder/mcomp.h"
43#include "av1/encoder/pickcdef.h"
44#include "av1/encoder/ratectrl.h"
45#include "av1/encoder/rd.h"
47#include "av1/encoder/svc_layercontext.h"
48#include "av1/encoder/temporal_filter.h"
49#include "av1/encoder/tokenize.h"
50#include "av1/encoder/tpl_model.h"
51#include "av1/encoder/av1_noise_estimate.h"
52
53#if CONFIG_INTERNAL_STATS
54#include "aom_dsp/ssim.h"
55#endif
56#include "aom_dsp/variance.h"
57#if CONFIG_DENOISE
58#include "aom_dsp/noise_model.h"
59#endif
60#if CONFIG_TUNE_VMAF
61#include "av1/encoder/tune_vmaf.h"
62#endif
63#if CONFIG_AV1_TEMPORAL_DENOISING
64#include "av1/encoder/av1_temporal_denoiser.h"
65#endif
66#if CONFIG_TUNE_BUTTERAUGLI
67#include "av1/encoder/tune_butteraugli.h"
68#endif
69
70#include "aom/internal/aom_codec_internal.h"
71#include "aom_util/aom_thread.h"
72
73#ifdef __cplusplus
74extern "C" {
75#endif
76
77// TODO(yunqing, any): Added suppression tag to quiet Doxygen warnings. Need to
78// adjust it while we work on documentation.
80// Number of frames required to test for scene cut detection
81#define SCENE_CUT_KEY_TEST_INTERVAL 16
82
83// Rational number with an int64 numerator
84// This structure holds a fractional value
85typedef struct aom_rational64 {
86 int64_t num; // fraction numerator
87 int den; // fraction denominator
88} aom_rational64_t; // alias for struct aom_rational
89
90enum {
91 NORMAL = 0,
92 FOURFIVE = 1,
93 THREEFIVE = 2,
94 THREEFOUR = 3,
95 ONEFOUR = 4,
96 ONEEIGHT = 5,
97 ONETWO = 6
98} UENUM1BYTE(AOM_SCALING);
99
100enum {
101 // Good Quality Fast Encoding. The encoder balances quality with the amount of
102 // time it takes to encode the output. Speed setting controls how fast.
103 GOOD,
104 // Realtime Fast Encoding. Will force some restrictions on bitrate
105 // constraints.
106 REALTIME,
107 // All intra mode. All the frames are coded as intra frames.
108 ALLINTRA
109} UENUM1BYTE(MODE);
110
111enum {
112 FRAMEFLAGS_KEY = 1 << 0,
113 FRAMEFLAGS_GOLDEN = 1 << 1,
114 FRAMEFLAGS_BWDREF = 1 << 2,
115 // TODO(zoeliu): To determine whether a frame flag is needed for ALTREF2_FRAME
116 FRAMEFLAGS_ALTREF = 1 << 3,
117 FRAMEFLAGS_INTRAONLY = 1 << 4,
118 FRAMEFLAGS_SWITCH = 1 << 5,
119 FRAMEFLAGS_ERROR_RESILIENT = 1 << 6,
120} UENUM1BYTE(FRAMETYPE_FLAGS);
121
122enum {
123 NO_AQ = 0,
124 VARIANCE_AQ = 1,
125 COMPLEXITY_AQ = 2,
126 CYCLIC_REFRESH_AQ = 3,
127 AQ_MODE_COUNT // This should always be the last member of the enum
128} UENUM1BYTE(AQ_MODE);
129enum {
130 NO_DELTA_Q = 0,
131 DELTA_Q_OBJECTIVE = 1, // Modulation to improve objective quality
132 DELTA_Q_PERCEPTUAL = 2, // Modulation to improve perceptual quality
133 DELTA_Q_MODE_COUNT // This should always be the last member of the enum
134} UENUM1BYTE(DELTAQ_MODE);
135
136enum {
137 RESIZE_NONE = 0, // No frame resizing allowed.
138 RESIZE_FIXED = 1, // All frames are coded at the specified scale.
139 RESIZE_RANDOM = 2, // All frames are coded at a random scale.
140 RESIZE_DYNAMIC = 3, // Frames coded at lower scale based on rate control.
141 RESIZE_MODES
142} UENUM1BYTE(RESIZE_MODE);
143
144enum {
145 SS_CFG_SRC = 0,
146 SS_CFG_LOOKAHEAD = 1,
147 SS_CFG_FPF = 2,
148 SS_CFG_TOTAL = 3
149} UENUM1BYTE(SS_CFG_OFFSET);
150
151enum {
152 DISABLE_SCENECUT, // For LAP, lag_in_frames < 19
153 ENABLE_SCENECUT_MODE_1, // For LAP, lag_in_frames >=19 and < 33
154 ENABLE_SCENECUT_MODE_2 // For twopass and LAP - lag_in_frames >=33
155} UENUM1BYTE(SCENECUT_MODE);
156
157#define MAX_VBR_CORPUS_COMPLEXITY 10000
158
161typedef enum {
162 MOD_FP, // First pass
163 MOD_TF, // Temporal filtering
164 MOD_TPL, // TPL
165 MOD_GME, // Global motion estimation
166 MOD_ENC, // Encode stage
167 MOD_LPF, // Deblocking loop filter
168 MOD_CDEF_SEARCH, // CDEF search
169 MOD_LR, // Loop restoration filtering
170 NUM_MT_MODULES
171} MULTI_THREADED_MODULES;
172
178typedef enum {
184
188typedef struct {
192 RESIZE_MODE resize_mode;
203} ResizeCfg;
204
208typedef struct {
232
236typedef struct {
268
272typedef struct {
306
310typedef struct {
337
341typedef struct {
373
377typedef struct {
382
387
392
398
405
410
415
420
426
430typedef struct {
432 // BUFFERING PARAMETERS
450
455
474 unsigned int gf_cbr_boost_pct;
479 unsigned int min_cr;
533
535typedef struct {
536 // Indicates the number of frames lag before encoding is started.
537 int lag_in_frames;
538 // Indicates the minimum gf/arf interval to be used.
539 int min_gf_interval;
540 // Indicates the maximum gf/arf interval to be used.
541 int max_gf_interval;
542 // Indicates the minimum height for GF group pyramid structure to be used.
543 int gf_min_pyr_height;
544 // Indicates the maximum height for GF group pyramid structure to be used.
545 int gf_max_pyr_height;
546 // Indicates if automatic set and use of altref frames should be enabled.
547 bool enable_auto_arf;
548 // Indicates if automatic set and use of (b)ackward (r)ef (f)rames should be
549 // enabled.
550 bool enable_auto_brf;
551} GFConfig;
552
553typedef struct {
554 // Indicates the number of tile groups.
555 unsigned int num_tile_groups;
556 // Indicates the MTU size for a tile group. If mtu is non-zero,
557 // num_tile_groups is set to DEFAULT_MAX_NUM_TG.
558 unsigned int mtu;
559 // Indicates the number of tile columns in log2.
560 int tile_columns;
561 // Indicates the number of tile rows in log2.
562 int tile_rows;
563 // Indicates the number of widths in the tile_widths[] array.
564 int tile_width_count;
565 // Indicates the number of heights in the tile_heights[] array.
566 int tile_height_count;
567 // Indicates the tile widths, and may be empty.
568 int tile_widths[MAX_TILE_COLS];
569 // Indicates the tile heights, and may be empty.
570 int tile_heights[MAX_TILE_ROWS];
571 // Indicates if large scale tile coding should be used.
572 bool enable_large_scale_tile;
573 // Indicates if single tile decoding mode should be enabled.
574 bool enable_single_tile_decoding;
575 // Indicates if EXT_TILE_DEBUG should be enabled.
576 bool enable_ext_tile_debug;
577} TileConfig;
578
579typedef struct {
580 // Indicates the width of the input frame.
581 int width;
582 // Indicates the height of the input frame.
583 int height;
584 // If forced_max_frame_width is non-zero then it is used to force the maximum
585 // frame width written in write_sequence_header().
586 int forced_max_frame_width;
587 // If forced_max_frame_width is non-zero then it is used to force the maximum
588 // frame height written in write_sequence_header().
589 int forced_max_frame_height;
590 // Indicates the frame width after applying both super-resolution and resize
591 // to the coded frame.
592 int render_width;
593 // Indicates the frame height after applying both super-resolution and resize
594 // to the coded frame.
595 int render_height;
596} FrameDimensionCfg;
597
598typedef struct {
599 // Indicates if warped motion should be enabled.
600 bool enable_warped_motion;
601 // Indicates if warped motion should be evaluated or not.
602 bool allow_warped_motion;
603 // Indicates if OBMC motion should be enabled.
604 bool enable_obmc;
605} MotionModeCfg;
606
607typedef struct {
608 // Timing info for each frame.
609 aom_timing_info_t timing_info;
610 // Indicates the number of time units of a decoding clock.
611 uint32_t num_units_in_decoding_tick;
612 // Indicates if decoder model information is present in the coded sequence
613 // header.
614 bool decoder_model_info_present_flag;
615 // Indicates if display model information is present in the coded sequence
616 // header.
617 bool display_model_info_present_flag;
618 // Indicates if timing info for each frame is present.
619 bool timing_info_present;
620} DecoderModelCfg;
621
622typedef struct {
623 // Indicates the update frequency for coeff costs.
624 COST_UPDATE_TYPE coeff;
625 // Indicates the update frequency for mode costs.
626 COST_UPDATE_TYPE mode;
627 // Indicates the update frequency for mv costs.
629 // Indicates the update frequency for dv costs.
631} CostUpdateFreq;
632
633typedef struct {
634 // Indicates the maximum number of reference frames allowed per frame.
635 unsigned int max_reference_frames;
636 // Indicates if the reduced set of references should be enabled.
637 bool enable_reduced_reference_set;
638 // Indicates if one-sided compound should be enabled.
639 bool enable_onesided_comp;
640} RefFrameCfg;
641
642typedef struct {
643 // Indicates the color space that should be used.
644 aom_color_primaries_t color_primaries;
645 // Indicates the characteristics of transfer function to be used.
646 aom_transfer_characteristics_t transfer_characteristics;
647 // Indicates the matrix coefficients to be used for the transfer function.
648 aom_matrix_coefficients_t matrix_coefficients;
649 // Indicates the chroma 4:2:0 sample position info.
650 aom_chroma_sample_position_t chroma_sample_position;
651 // Indicates if a limited color range or full color range should be used.
652 aom_color_range_t color_range;
653} ColorCfg;
654
655typedef struct {
656 // Indicates if extreme motion vector unit test should be enabled or not.
657 unsigned int motion_vector_unit_test;
658 // Indicates if superblock multipass unit test should be enabled or not.
659 unsigned int sb_multipass_unit_test;
660} UnitTestCfg;
661
662typedef struct {
663 // Indicates the file path to the VMAF model.
664 const char *vmaf_model_path;
665 // Indicates the path to the film grain parameters.
666 const char *film_grain_table_filename;
667 // Indicates the visual tuning metric.
668 aom_tune_metric tuning;
669 // Indicates if the current content is screen or default type.
670 aom_tune_content content;
671 // Indicates the film grain parameters.
672 int film_grain_test_vector;
673} TuneCfg;
674
675typedef struct {
676 // Indicates the framerate of the input video.
677 double init_framerate;
678 // Indicates the bit-depth of the input video.
679 unsigned int input_bit_depth;
680 // Indicates the maximum number of frames to be encoded.
681 unsigned int limit;
682 // Indicates the chrome subsampling x value.
683 unsigned int chroma_subsampling_x;
684 // Indicates the chrome subsampling y value.
685 unsigned int chroma_subsampling_y;
686} InputCfg;
687
688typedef struct {
689 // List of QP offsets for: keyframe, ALTREF, and 3 levels of internal ARFs.
690 // If any of these values are negative, fixed offsets are disabled.
691 // Uses internal q range.
692 double fixed_qp_offsets[FIXED_QP_OFFSET_COUNT];
693 // If true, encoder will use fixed QP offsets, that are either:
694 // - Given by the user, and stored in 'fixed_qp_offsets' array, OR
695 // - Picked automatically from cq_level.
696 int use_fixed_qp_offsets;
697 // Indicates the minimum flatness of the quantization matrix.
698 int qm_minlevel;
699 // Indicates the maximum flatness of the quantization matrix.
700 int qm_maxlevel;
701 // Indicates if adaptive quantize_b should be enabled.
702 int quant_b_adapt;
703 // Indicates the Adaptive Quantization mode to be used.
704 AQ_MODE aq_mode;
705 // Indicates the delta q mode to be used.
706 DELTAQ_MODE deltaq_mode;
707 // Indicates if delta quantization should be enabled in chroma planes.
708 bool enable_chroma_deltaq;
709 // Indicates if encoding with quantization matrices should be enabled.
710 bool using_qm;
711} QuantizationCfg;
712
717typedef struct {
722
731
736
741
749
754
760} AlgoCfg;
763typedef struct {
764 // Indicates the codec bit-depth.
765 aom_bit_depth_t bit_depth;
766 // Indicates the superblock size that should be used by the encoder.
767 aom_superblock_size_t superblock_size;
768 // Indicates if loopfilter modulation should be enabled.
769 bool enable_deltalf_mode;
770 // Indicates if CDEF should be enabled.
771 bool enable_cdef;
772 // Indicates if loop restoration filter should be enabled.
773 bool enable_restoration;
774 // When enabled, video mode should be used even for single frame input.
775 bool force_video_mode;
776 // Indicates if the error resiliency features should be enabled.
777 bool error_resilient_mode;
778 // Indicates if frame parallel decoding feature should be enabled.
779 bool frame_parallel_decoding_mode;
780 // Indicates if the input should be encoded as monochrome.
781 bool enable_monochrome;
782 // When enabled, the encoder will use a full header even for still pictures.
783 // When disabled, a reduced header is used for still pictures.
784 bool full_still_picture_hdr;
785 // Indicates if dual interpolation filters should be enabled.
786 bool enable_dual_filter;
787 // Indicates if frame order hint should be enabled or not.
788 bool enable_order_hint;
789 // Indicates if ref_frame_mvs should be enabled at the sequence level.
790 bool ref_frame_mvs_present;
791 // Indicates if ref_frame_mvs should be enabled at the frame level.
792 bool enable_ref_frame_mvs;
793 // Indicates if interintra compound mode is enabled.
794 bool enable_interintra_comp;
795 // Indicates if global motion should be enabled.
796 bool enable_global_motion;
797 // Indicates if palette should be enabled.
798 bool enable_palette;
799} ToolCfg;
800
805typedef struct AV1EncoderConfig {
807 // Configuration related to the input video.
808 InputCfg input_cfg;
809
810 // Configuration related to frame-dimensions.
811 FrameDimensionCfg frm_dim_cfg;
812
818
823
830 // Configuration related to Quantization.
831 QuantizationCfg q_cfg;
832
833 // Internal frame size scaling.
834 ResizeCfg resize_cfg;
835
836 // Frame Super-Resolution size scaling.
837 SuperResCfg superres_cfg;
838
847 // Configuration related to encoder toolsets.
848 ToolCfg tool_cfg;
849
850 // Configuration related to Group of frames.
851 GFConfig gf_cfg;
852
853 // Tile related configuration parameters.
854 TileConfig tile_cfg;
855
856 // Configuration related to Tune.
857 TuneCfg tune_cfg;
858
859 // Configuration related to color.
860 ColorCfg color_cfg;
861
862 // Configuration related to decoder model.
863 DecoderModelCfg dec_model_cfg;
864
865 // Configuration related to reference frames.
866 RefFrameCfg ref_frm_cfg;
867
868 // Configuration related to unit tests.
869 UnitTestCfg unit_test_cfg;
870
871 // Flags related to motion mode.
872 MotionModeCfg motion_mode_cfg;
873
874 // Flags related to intra mode search.
875 IntraModeCfg intra_mode_cfg;
876
877 // Flags related to transform size/type.
878 TxfmSizeTypeCfg txfm_cfg;
879
880 // Flags related to compound type.
881 CompoundTypeCfg comp_type_cfg;
882
883 // Partition related information.
884 PartitionCfg part_cfg;
885
886 // Configuration related to frequency of cost update.
887 CostUpdateFreq cost_upd_freq;
888
889#if CONFIG_DENOISE
890 // Indicates the noise level.
891 float noise_level;
892 // Indicates the the denoisers block size.
893 int noise_block_size;
894 // Indicates whether to apply denoising to the frame to be encoded
895 int enable_dnl_denoising;
896#endif
897
898#if CONFIG_AV1_TEMPORAL_DENOISING
899 // Noise sensitivity.
900 int noise_sensitivity;
901#endif
902 // Bit mask to specify which tier each of the 32 possible operating points
903 // conforms to.
904 unsigned int tier_mask;
905
906 // Indicates the number of pixels off the edge of a reference frame we're
907 // allowed to go when forming an inter prediction.
908 int border_in_pixels;
909
910 // Indicates the maximum number of threads that may be used by the encoder.
911 int max_threads;
912
913 // Indicates the speed preset to be used.
914 int speed;
915
916 // Indicates the target sequence level index for each operating point(OP).
917 AV1_LEVEL target_seq_level_idx[MAX_NUM_OPERATING_POINTS];
918
919 // Indicates the bitstream profile to be used.
920 BITSTREAM_PROFILE profile;
921
933 // Indicates if the encoding is GOOD or REALTIME.
934 MODE mode;
935
936 // Indicates if row-based multi-threading should be enabled or not.
937 bool row_mt;
938
939 // Indicates if 16bit frame buffers are to be used i.e., the content is >
940 // 8-bit.
941 bool use_highbitdepth;
942
943 // Indicates the bitstream syntax mode. 0 indicates bitstream is saved as
944 // Section 5 bitstream, while 1 indicates the bitstream is saved in Annex - B
945 // format.
946 bool save_as_annexb;
947
950
952static INLINE int is_lossless_requested(const RateControlCfg *const rc_cfg) {
953 return rc_cfg->best_allowed_q == 0 && rc_cfg->worst_allowed_q == 0;
954}
960typedef struct {
966 int obmc_probs[FRAME_UPDATE_TYPES][BLOCK_SIZES_ALL];
967
973 int warped_probs[FRAME_UPDATE_TYPES];
974
981 int tx_type_probs[FRAME_UPDATE_TYPES][TX_SIZES_ALL][TX_TYPES];
982
989 int switchable_interp_probs[FRAME_UPDATE_TYPES][SWITCHABLE_FILTER_CONTEXTS]
990 [SWITCHABLE_FILTERS];
992
995typedef struct FRAME_COUNTS {
996// Note: This structure should only contain 'unsigned int' fields, or
997// aggregates built solely from 'unsigned int' fields/elements
998#if CONFIG_ENTROPY_STATS
999 unsigned int kf_y_mode[KF_MODE_CONTEXTS][KF_MODE_CONTEXTS][INTRA_MODES];
1000 unsigned int angle_delta[DIRECTIONAL_MODES][2 * MAX_ANGLE_DELTA + 1];
1001 unsigned int y_mode[BLOCK_SIZE_GROUPS][INTRA_MODES];
1002 unsigned int uv_mode[CFL_ALLOWED_TYPES][INTRA_MODES][UV_INTRA_MODES];
1003 unsigned int cfl_sign[CFL_JOINT_SIGNS];
1004 unsigned int cfl_alpha[CFL_ALPHA_CONTEXTS][CFL_ALPHABET_SIZE];
1005 unsigned int palette_y_mode[PALATTE_BSIZE_CTXS][PALETTE_Y_MODE_CONTEXTS][2];
1006 unsigned int palette_uv_mode[PALETTE_UV_MODE_CONTEXTS][2];
1007 unsigned int palette_y_size[PALATTE_BSIZE_CTXS][PALETTE_SIZES];
1008 unsigned int palette_uv_size[PALATTE_BSIZE_CTXS][PALETTE_SIZES];
1009 unsigned int palette_y_color_index[PALETTE_SIZES]
1010 [PALETTE_COLOR_INDEX_CONTEXTS]
1011 [PALETTE_COLORS];
1012 unsigned int palette_uv_color_index[PALETTE_SIZES]
1013 [PALETTE_COLOR_INDEX_CONTEXTS]
1014 [PALETTE_COLORS];
1015 unsigned int partition[PARTITION_CONTEXTS][EXT_PARTITION_TYPES];
1016 unsigned int txb_skip[TOKEN_CDF_Q_CTXS][TX_SIZES][TXB_SKIP_CONTEXTS][2];
1017 unsigned int eob_extra[TOKEN_CDF_Q_CTXS][TX_SIZES][PLANE_TYPES]
1018 [EOB_COEF_CONTEXTS][2];
1019 unsigned int dc_sign[PLANE_TYPES][DC_SIGN_CONTEXTS][2];
1020 unsigned int coeff_lps[TX_SIZES][PLANE_TYPES][BR_CDF_SIZE - 1][LEVEL_CONTEXTS]
1021 [2];
1022 unsigned int eob_flag[TX_SIZES][PLANE_TYPES][EOB_COEF_CONTEXTS][2];
1023 unsigned int eob_multi16[TOKEN_CDF_Q_CTXS][PLANE_TYPES][2][5];
1024 unsigned int eob_multi32[TOKEN_CDF_Q_CTXS][PLANE_TYPES][2][6];
1025 unsigned int eob_multi64[TOKEN_CDF_Q_CTXS][PLANE_TYPES][2][7];
1026 unsigned int eob_multi128[TOKEN_CDF_Q_CTXS][PLANE_TYPES][2][8];
1027 unsigned int eob_multi256[TOKEN_CDF_Q_CTXS][PLANE_TYPES][2][9];
1028 unsigned int eob_multi512[TOKEN_CDF_Q_CTXS][PLANE_TYPES][2][10];
1029 unsigned int eob_multi1024[TOKEN_CDF_Q_CTXS][PLANE_TYPES][2][11];
1030 unsigned int coeff_lps_multi[TOKEN_CDF_Q_CTXS][TX_SIZES][PLANE_TYPES]
1031 [LEVEL_CONTEXTS][BR_CDF_SIZE];
1032 unsigned int coeff_base_multi[TOKEN_CDF_Q_CTXS][TX_SIZES][PLANE_TYPES]
1033 [SIG_COEF_CONTEXTS][NUM_BASE_LEVELS + 2];
1034 unsigned int coeff_base_eob_multi[TOKEN_CDF_Q_CTXS][TX_SIZES][PLANE_TYPES]
1035 [SIG_COEF_CONTEXTS_EOB][NUM_BASE_LEVELS + 1];
1036 unsigned int newmv_mode[NEWMV_MODE_CONTEXTS][2];
1037 unsigned int zeromv_mode[GLOBALMV_MODE_CONTEXTS][2];
1038 unsigned int refmv_mode[REFMV_MODE_CONTEXTS][2];
1039 unsigned int drl_mode[DRL_MODE_CONTEXTS][2];
1040 unsigned int inter_compound_mode[INTER_MODE_CONTEXTS][INTER_COMPOUND_MODES];
1041 unsigned int wedge_idx[BLOCK_SIZES_ALL][16];
1042 unsigned int interintra[BLOCK_SIZE_GROUPS][2];
1043 unsigned int interintra_mode[BLOCK_SIZE_GROUPS][INTERINTRA_MODES];
1044 unsigned int wedge_interintra[BLOCK_SIZES_ALL][2];
1045 unsigned int compound_type[BLOCK_SIZES_ALL][MASKED_COMPOUND_TYPES];
1046 unsigned int motion_mode[BLOCK_SIZES_ALL][MOTION_MODES];
1047 unsigned int obmc[BLOCK_SIZES_ALL][2];
1048 unsigned int intra_inter[INTRA_INTER_CONTEXTS][2];
1049 unsigned int comp_inter[COMP_INTER_CONTEXTS][2];
1050 unsigned int comp_ref_type[COMP_REF_TYPE_CONTEXTS][2];
1051 unsigned int uni_comp_ref[UNI_COMP_REF_CONTEXTS][UNIDIR_COMP_REFS - 1][2];
1052 unsigned int single_ref[REF_CONTEXTS][SINGLE_REFS - 1][2];
1053 unsigned int comp_ref[REF_CONTEXTS][FWD_REFS - 1][2];
1054 unsigned int comp_bwdref[REF_CONTEXTS][BWD_REFS - 1][2];
1055 unsigned int intrabc[2];
1056
1057 unsigned int txfm_partition[TXFM_PARTITION_CONTEXTS][2];
1058 unsigned int intra_tx_size[MAX_TX_CATS][TX_SIZE_CONTEXTS][MAX_TX_DEPTH + 1];
1059 unsigned int skip_mode[SKIP_MODE_CONTEXTS][2];
1060 unsigned int skip_txfm[SKIP_CONTEXTS][2];
1061 unsigned int compound_index[COMP_INDEX_CONTEXTS][2];
1062 unsigned int comp_group_idx[COMP_GROUP_IDX_CONTEXTS][2];
1063 unsigned int delta_q[DELTA_Q_PROBS][2];
1064 unsigned int delta_lf_multi[FRAME_LF_COUNT][DELTA_LF_PROBS][2];
1065 unsigned int delta_lf[DELTA_LF_PROBS][2];
1066
1067 unsigned int inter_ext_tx[EXT_TX_SETS_INTER][EXT_TX_SIZES][TX_TYPES];
1068 unsigned int intra_ext_tx[EXT_TX_SETS_INTRA][EXT_TX_SIZES][INTRA_MODES]
1069 [TX_TYPES];
1070 unsigned int filter_intra_mode[FILTER_INTRA_MODES];
1071 unsigned int filter_intra[BLOCK_SIZES_ALL][2];
1072 unsigned int switchable_restore[RESTORE_SWITCHABLE_TYPES];
1073 unsigned int wiener_restore[2];
1074 unsigned int sgrproj_restore[2];
1075#endif // CONFIG_ENTROPY_STATS
1076
1077 unsigned int switchable_interp[SWITCHABLE_FILTER_CONTEXTS]
1078 [SWITCHABLE_FILTERS];
1079} FRAME_COUNTS;
1080
1081#define INTER_MODE_RD_DATA_OVERALL_SIZE 6400
1082
1083typedef struct {
1084 int ready;
1085 double a;
1086 double b;
1087 double dist_mean;
1088 double ld_mean;
1089 double sse_mean;
1090 double sse_sse_mean;
1091 double sse_ld_mean;
1092 int num;
1093 double dist_sum;
1094 double ld_sum;
1095 double sse_sum;
1096 double sse_sse_sum;
1097 double sse_ld_sum;
1098} InterModeRdModel;
1099
1100typedef struct {
1101 int idx;
1102 int64_t rd;
1103} RdIdxPair;
1104// TODO(angiebird): This is an estimated size. We still need to figure what is
1105// the maximum number of modes.
1106#define MAX_INTER_MODES 1024
1107// TODO(any): rename this struct to something else. There is already another
1108// struct called inter_mode_info, which makes this terribly confusing.
1116typedef struct inter_modes_info {
1121 int num;
1125 MB_MODE_INFO mbmi_arr[MAX_INTER_MODES];
1129 int mode_rate_arr[MAX_INTER_MODES];
1133 int64_t sse_arr[MAX_INTER_MODES];
1137 int64_t est_rd_arr[MAX_INTER_MODES];
1141 RdIdxPair rd_idx_pair_arr[MAX_INTER_MODES];
1145 RD_STATS rd_cost_arr[MAX_INTER_MODES];
1149 RD_STATS rd_cost_y_arr[MAX_INTER_MODES];
1153 RD_STATS rd_cost_uv_arr[MAX_INTER_MODES];
1155
1157typedef struct {
1158 // TODO(kyslov): consider changing to 64bit
1159
1160 // This struct is used for computing variance in choose_partitioning(), where
1161 // the max number of samples within a superblock is 32x32 (with 4x4 avg).
1162 // With 8bit bitdepth, uint32_t is enough for sum_square_error (2^8 * 2^8 * 32
1163 // * 32 = 2^26). For high bitdepth we need to consider changing this to 64 bit
1164 uint32_t sum_square_error;
1165 int32_t sum_error;
1166 int log2_count;
1167 int variance;
1168} VPartVar;
1169
1170typedef struct {
1171 VPartVar none;
1172 VPartVar horz[2];
1173 VPartVar vert[2];
1174} VPVariance;
1175
1176typedef struct {
1177 VPVariance part_variances;
1178 VPartVar split[4];
1179} VP4x4;
1180
1181typedef struct {
1182 VPVariance part_variances;
1183 VP4x4 split[4];
1184} VP8x8;
1185
1186typedef struct {
1187 VPVariance part_variances;
1188 VP8x8 split[4];
1189} VP16x16;
1190
1191typedef struct {
1192 VPVariance part_variances;
1193 VP16x16 split[4];
1194} VP32x32;
1195
1196typedef struct {
1197 VPVariance part_variances;
1198 VP32x32 split[4];
1199} VP64x64;
1200
1201typedef struct {
1202 VPVariance part_variances;
1203 VP64x64 *split;
1204} VP128x128;
1205
1211typedef struct {
1220 int64_t thresholds[5];
1221
1228
1232typedef struct {
1233#if CONFIG_MULTITHREAD
1238 pthread_mutex_t *mutex_;
1239 pthread_cond_t *cond_;
1241#endif // CONFIG_MULTITHREAD
1247 int *num_finished_cols;
1257 int rows;
1267
1270// TODO(jingning) All spatially adaptive variables should go to TileDataEnc.
1271typedef struct TileDataEnc {
1272 TileInfo tile_info;
1273 DECLARE_ALIGNED(16, FRAME_CONTEXT, tctx);
1274 FRAME_CONTEXT *row_ctx;
1275 uint8_t allow_update_cdf;
1276 InterModeRdModel inter_mode_rd_models[BLOCK_SIZES_ALL];
1277 AV1EncRowMultiThreadSync row_mt_sync;
1278 MV firstpass_top_mv;
1279} TileDataEnc;
1280
1281typedef struct RD_COUNTS {
1282 int64_t comp_pred_diff[REFERENCE_MODES];
1283 int compound_ref_used_flag;
1284 int skip_mode_used_flag;
1285 int tx_type_used[TX_SIZES_ALL][TX_TYPES];
1286 int obmc_used[BLOCK_SIZES_ALL][2];
1287 int warped_used[2];
1288} RD_COUNTS;
1289
1290typedef struct ThreadData {
1291 MACROBLOCK mb;
1292 RD_COUNTS rd_counts;
1293 FRAME_COUNTS *counts;
1294 PC_TREE_SHARED_BUFFERS shared_coeff_buf;
1295 SIMPLE_MOTION_DATA_TREE *sms_tree;
1296 SIMPLE_MOTION_DATA_TREE *sms_root;
1298 uint32_t *hash_value_buffer[2][2];
1299 OBMCBuffer obmc_buffer;
1300 PALETTE_BUFFER *palette_buffer;
1301 CompoundTypeRdBuffers comp_rd_buffer;
1302 CONV_BUF_TYPE *tmp_conv_dst;
1303 uint8_t *tmp_pred_bufs[2];
1304 int intrabc_used;
1305 int deltaq_used;
1306 FRAME_CONTEXT *tctx;
1307 VP64x64 *vt64x64;
1308 int32_t num_64x64_blocks;
1309 PICK_MODE_CONTEXT *firstpass_ctx;
1310 TemporalFilterData tf_data;
1311 TplTxfmStats tpl_txfm_stats;
1312} ThreadData;
1313
1314struct EncWorkerData;
1315
1321typedef struct {
1344
1348 int thread_id_to_tile_id[MAX_NUM_THREADS];
1349
1350#if CONFIG_MULTITHREAD
1354 pthread_mutex_t *mutex_;
1355#endif
1356
1364 void (*sync_read_ptr)(AV1EncRowMultiThreadSync *const, int, int);
1368 void (*sync_write_ptr)(AV1EncRowMultiThreadSync *const, int, int, int);
1371
1375typedef struct MultiThreadInfo {
1380
1384 int num_mod_workers[NUM_MT_MODULES];
1385
1391
1397
1401 AVxWorker *workers;
1402
1407 struct EncWorkerData *tile_thr_data;
1408
1414
1419
1423 AV1TplRowMultiThreadInfo tpl_row_mt;
1424
1428 AV1LfSync lf_row_sync;
1429
1433 AV1LrSync lr_row_sync;
1434
1438 AV1GlobalMotionSync gm_sync;
1439
1443 AV1TemporalFilterSync tf_sync;
1444
1448 AV1CdefSync cdef_sync;
1450
1453typedef struct ActiveMap {
1454 int enabled;
1455 int update;
1456 unsigned char *map;
1457} ActiveMap;
1458
1464typedef struct {
1469 double cs_rate_array[32];
1479
1482#if CONFIG_INTERNAL_STATS
1483// types of stats
1484enum {
1485 STAT_Y,
1486 STAT_U,
1487 STAT_V,
1488 STAT_ALL,
1489 NUM_STAT_TYPES // This should always be the last member of the enum
1490} UENUM1BYTE(StatType);
1491
1492typedef struct IMAGE_STAT {
1493 double stat[NUM_STAT_TYPES];
1494 double worst;
1495} ImageStat;
1496#endif // CONFIG_INTERNAL_STATS
1497
1498typedef struct {
1499 int ref_count;
1501} EncRefCntBuffer;
1502
1510typedef struct {
1525
1528#if CONFIG_COLLECT_PARTITION_STATS
1529typedef struct FramePartitionTimingStats {
1530 int partition_decisions[6][EXT_PARTITION_TYPES];
1531 int partition_attempts[6][EXT_PARTITION_TYPES];
1532 int64_t partition_times[6][EXT_PARTITION_TYPES];
1533
1534 int partition_redo;
1535} FramePartitionTimingStats;
1536#endif // CONFIG_COLLECT_PARTITION_STATS
1537
1538#if CONFIG_COLLECT_COMPONENT_TIMING
1539#include "aom_ports/aom_timer.h"
1540// Adjust the following to add new components.
1541enum {
1542 av1_encode_strategy_time,
1543 av1_get_second_pass_params_time,
1544 denoise_and_encode_time,
1545 apply_filtering_time,
1546 av1_tpl_setup_stats_time,
1547 encode_frame_to_data_rate_time,
1548 encode_with_recode_loop_time,
1549 loop_filter_time,
1550 cdef_time,
1551 loop_restoration_time,
1552 av1_pack_bitstream_final_time,
1553 av1_encode_frame_time,
1554 av1_compute_global_motion_time,
1555 av1_setup_motion_field_time,
1556 encode_sb_row_time,
1557
1558 rd_pick_partition_time,
1559 av1_prune_partitions_time,
1560 none_partition_search_time,
1561 split_partition_search_time,
1562 rectangular_partition_search_time,
1563 ab_partitions_search_time,
1564 rd_pick_4partition_time,
1565 encode_sb_time,
1566
1567 rd_pick_sb_modes_time,
1568 av1_rd_pick_intra_mode_sb_time,
1569 av1_rd_pick_inter_mode_sb_time,
1570 handle_inter_mode_time,
1571 evaluate_motion_mode_for_winner_candidates_time,
1572 handle_intra_mode_time,
1573 do_tx_search_time,
1574 av1_search_palette_mode_time,
1575 handle_newmv_time,
1576 compound_type_rd_time,
1577 interpolation_filter_search_time,
1578 motion_mode_rd_time,
1579 kTimingComponents,
1580} UENUM1BYTE(TIMING_COMPONENT);
1581
1582static INLINE char const *get_component_name(int index) {
1583 switch (index) {
1584 case av1_encode_strategy_time: return "av1_encode_strategy_time";
1585 case av1_get_second_pass_params_time:
1586 return "av1_get_second_pass_params_time";
1587 case denoise_and_encode_time: return "denoise_and_encode_time";
1588 case apply_filtering_time: return "apply_filtering_time";
1589 case av1_tpl_setup_stats_time: return "av1_tpl_setup_stats_time";
1590 case encode_frame_to_data_rate_time:
1591 return "encode_frame_to_data_rate_time";
1592 case encode_with_recode_loop_time: return "encode_with_recode_loop_time";
1593 case loop_filter_time: return "loop_filter_time";
1594 case cdef_time: return "cdef_time";
1595 case loop_restoration_time: return "loop_restoration_time";
1596 case av1_pack_bitstream_final_time: return "av1_pack_bitstream_final_time";
1597 case av1_encode_frame_time: return "av1_encode_frame_time";
1598 case av1_compute_global_motion_time:
1599 return "av1_compute_global_motion_time";
1600 case av1_setup_motion_field_time: return "av1_setup_motion_field_time";
1601 case encode_sb_row_time: return "encode_sb_row_time";
1602
1603 case rd_pick_partition_time: return "rd_pick_partition_time";
1604 case av1_prune_partitions_time: return "av1_prune_partitions_time";
1605 case none_partition_search_time: return "none_partition_search_time";
1606 case split_partition_search_time: return "split_partition_search_time";
1607 case rectangular_partition_search_time:
1608 return "rectangular_partition_search_time";
1609 case ab_partitions_search_time: return "ab_partitions_search_time";
1610 case rd_pick_4partition_time: return "rd_pick_4partition_time";
1611 case encode_sb_time: return "encode_sb_time";
1612
1613 case rd_pick_sb_modes_time: return "rd_pick_sb_modes_time";
1614 case av1_rd_pick_intra_mode_sb_time:
1615 return "av1_rd_pick_intra_mode_sb_time";
1616 case av1_rd_pick_inter_mode_sb_time:
1617 return "av1_rd_pick_inter_mode_sb_time";
1618 case handle_inter_mode_time: return "handle_inter_mode_time";
1619 case evaluate_motion_mode_for_winner_candidates_time:
1620 return "evaluate_motion_mode_for_winner_candidates_time";
1621 case handle_intra_mode_time: return "handle_intra_mode_time";
1622 case do_tx_search_time: return "do_tx_search_time";
1623 case av1_search_palette_mode_time: return "av1_search_palette_mode_time";
1624 case handle_newmv_time: return "handle_newmv_time";
1625 case compound_type_rd_time: return "compound_type_rd_time";
1626 case interpolation_filter_search_time:
1627 return "interpolation_filter_search_time";
1628 case motion_mode_rd_time: return "motion_mode_rd_time";
1629 default: assert(0);
1630 }
1631 return "error";
1632}
1633#endif
1634
1635// The maximum number of internal ARFs except ALTREF_FRAME
1636#define MAX_INTERNAL_ARFS (REF_FRAMES - BWDREF_FRAME - 1)
1637
1643typedef struct {
1648
1654 YV12_BUFFER_CONFIG *ref_buf[REF_FRAMES];
1655
1659 unsigned char *src_buffer;
1660
1666 int num_ref_frames[MAX_DIRECTIONS];
1667
1674 FrameDistPair reference_frames[MAX_DIRECTIONS][REF_FRAMES - 1];
1675
1688
1694 int src_corners[2 * MAX_CORNERS];
1696
1706typedef struct {
1707 int width;
1710
1714typedef struct {
1725
1729typedef struct {
1749 fractional_mv_step_fp *find_fractional_mv_step;
1756 search_site_config search_site_cfg[SS_CFG_TOTAL][NUM_DISTINCT_SEARCH_METHODS];
1758
1767typedef struct {
1772
1780typedef struct {
1781 int width;
1784
1788typedef struct {
1792 int ref_relative_dist[INTER_REFS_PER_FRAME];
1802
1818typedef struct {
1826 unsigned int coeff_opt_thresholds[MODE_EVAL_TYPES][2];
1827
1832 TX_SIZE_SEARCH_METHOD tx_size_search_methods[MODE_EVAL_TYPES];
1833
1840 unsigned int use_transform_domain_distortion[MODE_EVAL_TYPES];
1841
1847 unsigned int tx_domain_dist_threshold[MODE_EVAL_TYPES];
1848
1854 unsigned int skip_txfm_level[MODE_EVAL_TYPES];
1855
1861 unsigned int predict_dc_level[MODE_EVAL_TYPES];
1863
1871typedef struct {
1882
1886typedef struct {
1891
1896
1901
1907
1912
1917
1922
1929
1932typedef struct {
1933 int arf_stack[FRAME_BUFFERS];
1934 int arf_stack_size;
1935 int lst_stack[FRAME_BUFFERS];
1936 int lst_stack_size;
1937 int gld_stack[FRAME_BUFFERS];
1938 int gld_stack_size;
1939} RefBufferStack;
1940
1941typedef struct {
1942 // Some misc info
1943 int high_prec;
1944 int q;
1945 int order;
1946
1947 // MV counters
1948 int inter_count;
1949 int intra_count;
1950 int default_mvs;
1951 int mv_joint_count[4];
1952 int last_bit_zero;
1953 int last_bit_nonzero;
1954
1955 // Keep track of the rates
1956 int total_mv_rate;
1957 int hp_total_mv_rate;
1958 int lp_total_mv_rate;
1959
1960 // Texture info
1961 int horz_text;
1962 int vert_text;
1963 int diag_text;
1964
1965 // Whether the current struct contains valid data
1966 int valid;
1967} MV_STATS;
1968
1969typedef struct {
1970 struct loopfilter lf;
1971 CdefInfo cdef_info;
1972 YV12_BUFFER_CONFIG copy_buffer;
1973 RATE_CONTROL rc;
1974 MV_STATS mv_stats;
1975} CODING_CONTEXT;
1976
1977typedef struct {
1978 int frame_width;
1979 int frame_height;
1980 int mi_rows;
1981 int mi_cols;
1982 int mb_rows;
1983 int mb_cols;
1984 int num_mbs;
1985 aom_bit_depth_t bit_depth;
1986 int subsampling_x;
1987 int subsampling_y;
1988} FRAME_INFO;
1989
1993typedef struct {
1994 int show_frame_count;
1995} FRAME_INDEX_SET;
1996
2002typedef struct {
2008 uint8_t *map;
2016
2020typedef struct {
2033} TimeStamps;
2034
2039typedef struct {
2043 tran_low_t *tcoeff;
2047 uint16_t *eobs;
2051 uint8_t *entropy_ctx;
2053
2057typedef struct AV1_PRIMARY {
2061 struct AV1_COMP *cpi;
2062
2067
2071 struct lookahead_ctx *lookahead;
2072
2079
2084 struct aom_codec_pkt_list *output_pkt_list;
2085
2091
2095typedef struct AV1_COMP {
2100
2105 EncQuantDequantParams enc_quant_dequant_params;
2106
2110 ThreadData td;
2111
2115 FRAME_COUNTS counts;
2116
2121
2128
2134
2139
2144
2149
2154 TRELLIS_OPT_TYPE optimize_seg_arr[MAX_SEGMENTS];
2155
2162
2171
2177
2182
2187
2192
2198
2203
2208
2213
2218
2223 RefCntBuffer *scaled_ref_buf[INTER_REFS_PER_FRAME];
2224
2228 RefCntBuffer *last_show_frame_buf;
2229
2234
2241 int fb_of_context_type[REF_FRAMES];
2242
2247
2253
2259
2264
2268 RD_OPT rd;
2269
2274 CODING_CONTEXT coding_context;
2275
2280
2285
2290
2295
2300
2305
2310
2315
2320
2326
2331
2340 ActiveMap active_map;
2341
2347 aom_variance_fn_ptr_t fn_ptr[BLOCK_SIZES_ALL];
2348
2353
2358
2362 unsigned char gf_frame_index;
2363
2367 GF_STATE gf_state;
2368
2372 RefBufferStack ref_buffer_stack;
2373
2379
2384
2385#if CONFIG_INTERNAL_STATS
2387 uint64_t time_receive_data;
2388 uint64_t time_compress_data;
2389
2390 unsigned int mode_chosen_counts[MAX_MODES];
2391
2392 int count[2];
2393 uint64_t total_sq_error[2];
2394 uint64_t total_samples[2];
2395 ImageStat psnr[2];
2396
2397 double total_blockiness;
2398 double worst_blockiness;
2399
2400 int bytes;
2401 double summed_quality;
2402 double summed_weights;
2403 double summed_quality_hbd;
2404 double summed_weights_hbd;
2405 unsigned int tot_recode_hits;
2406 double worst_ssim;
2407 double worst_ssim_hbd;
2408
2409 ImageStat fastssim;
2410 ImageStat psnrhvs;
2411
2412 int b_calculate_blockiness;
2413 int b_calculate_consistency;
2414
2415 double total_inconsistency;
2416 double worst_consistency;
2417 Ssimv *ssim_vars;
2418 Metrics metrics;
2420#endif
2421
2426
2427#if CONFIG_SPEED_STATS
2431 unsigned int tx_search_count;
2432#endif // CONFIG_SPEED_STATS
2433
2439
2443 FRAME_INFO frame_info;
2444
2448 FRAME_INDEX_SET frame_index_set;
2449
2454
2462
2467
2472 TileDataEnc *tile_data;
2477
2481 TokenInfo token_info;
2482
2487
2492
2497
2502
2509
2514
2519
2523 AV1LrStruct lr_ctxt;
2524
2528 aom_film_grain_table_t *film_grain_table;
2529
2530#if CONFIG_DENOISE
2535 struct aom_denoise_and_model_t *denoise_and_model;
2536#endif
2537
2542
2551
2559
2560#if CONFIG_COLLECT_PARTITION_STATS
2564 FramePartitionTimingStats partition_stats;
2565#endif // CONFIG_COLLECT_PARTITION_STATS
2566
2567#if CONFIG_COLLECT_COMPONENT_TIMING
2571 uint64_t component_time[kTimingComponents];
2572 struct aom_usec_timer component_timer[kTimingComponents];
2576 uint64_t frame_component_time[kTimingComponents];
2577#endif
2578
2582 AV1LevelParams level_params;
2583
2588
2593
2614
2615#if CONFIG_TUNE_VMAF
2619 TuneVMAFInfo vmaf_info;
2620#endif
2621
2622#if CONFIG_TUNE_BUTTERAUGLI
2626 TuneButteraugliInfo butteraugli_info;
2627#endif
2628
2637
2645 COMPRESSOR_STAGE compressor_stage;
2646
2651 MV_STATS mv_stats;
2652
2658
2663
2670
2674 FirstPassData firstpass_data;
2675
2679 NOISE_ESTIMATE noise_estimate;
2680
2681#if CONFIG_AV1_TEMPORAL_DENOISING
2685 AV1_DENOISER denoiser;
2686#endif
2687
2693
2698
2702 BLOCK_SIZE fp_block_size;
2703
2710
2714typedef struct EncodeFrameInput {
2716 YV12_BUFFER_CONFIG *source;
2717 YV12_BUFFER_CONFIG *last_source;
2718 int64_t ts_duration;
2721
2726typedef struct EncodeFrameParams {
2734 FRAME_TYPE frame_type;
2735
2737 int primary_ref_frame;
2738 int order_offset;
2739
2745
2747 int refresh_frame_flags;
2748
2749 int show_existing_frame;
2750 int existing_fb_idx_to_show;
2751
2757
2761 int remapped_ref_idx[REF_FRAMES];
2762
2768
2774
2777// EncodeFrameResults contains information about the result of encoding a
2778// single frame
2779typedef struct {
2780 size_t size; // Size of resulting bitstream
2781} EncodeFrameResults;
2782
2783// Must not be called more than once.
2784void av1_initialize_enc(void);
2785
2786struct AV1_COMP *av1_create_compressor(AV1_PRIMARY *ppi, AV1EncoderConfig *oxcf,
2787 BufferPool *const pool,
2788 FIRSTPASS_STATS *frame_stats_buf,
2789 COMPRESSOR_STAGE stage,
2790 int num_lap_buffers,
2791 int lap_lag_in_frames,
2792 STATS_BUFFER_CTX *stats_buf_context);
2793
2794struct AV1_PRIMARY *av1_create_primary_compressor(
2795 struct aom_codec_pkt_list *pkt_list_head);
2796
2797void av1_remove_compressor(AV1_COMP *cpi);
2798
2799void av1_remove_primary_compressor(AV1_PRIMARY *ppi);
2800
2801void av1_change_config(AV1_COMP *cpi, const AV1EncoderConfig *oxcf);
2802
2803void av1_check_initial_width(AV1_COMP *cpi, int use_highbitdepth,
2804 int subsampling_x, int subsampling_y);
2805
2806void av1_init_seq_coding_tools(SequenceHeader *seq, AV1_COMMON *cm,
2807 const AV1EncoderConfig *oxcf, int use_svc);
2808
2828 YV12_BUFFER_CONFIG *sd, int64_t time_stamp,
2829 int64_t end_time_stamp);
2830
2855int av1_get_compressed_data(AV1_COMP *cpi, unsigned int *frame_flags,
2856 size_t *size, uint8_t *dest, int64_t *time_stamp,
2857 int64_t *time_end, int flush,
2858 const aom_rational64_t *timebase);
2859
2866int av1_encode(AV1_COMP *const cpi, uint8_t *const dest,
2867 const EncodeFrameInput *const frame_input,
2868 const EncodeFrameParams *const frame_params,
2869 EncodeFrameResults *const frame_results);
2870
2872int av1_get_preview_raw_frame(AV1_COMP *cpi, YV12_BUFFER_CONFIG *dest);
2873
2874int av1_get_last_show_frame(AV1_COMP *cpi, YV12_BUFFER_CONFIG *frame);
2875
2876aom_codec_err_t av1_copy_new_frame_enc(AV1_COMMON *cm,
2877 YV12_BUFFER_CONFIG *new_frame,
2878 YV12_BUFFER_CONFIG *sd);
2879
2880int av1_use_as_reference(int *ext_ref_frame_flags, int ref_frame_flags);
2881
2882int av1_copy_reference_enc(AV1_COMP *cpi, int idx, YV12_BUFFER_CONFIG *sd);
2883
2884int av1_set_reference_enc(AV1_COMP *cpi, int idx, YV12_BUFFER_CONFIG *sd);
2885
2886int av1_set_size_literal(AV1_COMP *cpi, int width, int height);
2887
2888void av1_set_frame_size(AV1_COMP *cpi, int width, int height);
2889
2890int av1_set_active_map(AV1_COMP *cpi, unsigned char *map, int rows, int cols);
2891
2892int av1_get_active_map(AV1_COMP *cpi, unsigned char *map, int rows, int cols);
2893
2894int av1_set_internal_size(AV1EncoderConfig *const oxcf,
2895 ResizePendingParams *resize_pending_params,
2896 AOM_SCALING horiz_mode, AOM_SCALING vert_mode);
2897
2898int av1_get_quantizer(struct AV1_COMP *cpi);
2899
2900int av1_convert_sect5obus_to_annexb(uint8_t *buffer, size_t *input_size);
2901
2902// Set screen content options.
2903// This function estimates whether to use screen content tools, by counting
2904// the portion of blocks that have few luma colors.
2905// Modifies:
2906// cpi->commom.features.allow_screen_content_tools
2907// cpi->common.features.allow_intrabc
2908// cpi->use_screen_content_tools
2909// cpi->is_screen_content_type
2910// However, the estimation is not accurate and may misclassify videos.
2911// A slower but more accurate approach that determines whether to use screen
2912// content tools is employed later. See av1_determine_sc_tools_with_encoding().
2913void av1_set_screen_content_options(struct AV1_COMP *cpi,
2914 FeatureFlags *features);
2915
2916void av1_update_frame_size(AV1_COMP *cpi);
2917
2918// TODO(jingning): Move these functions as primitive members for the new cpi
2919// class.
2920static INLINE void stack_push(int *stack, int *stack_size, int item) {
2921 for (int i = *stack_size - 1; i >= 0; --i) stack[i + 1] = stack[i];
2922 stack[0] = item;
2923 ++*stack_size;
2924}
2925
2926static INLINE int stack_pop(int *stack, int *stack_size) {
2927 if (*stack_size <= 0) return -1;
2928
2929 int item = stack[0];
2930 for (int i = 0; i < *stack_size; ++i) stack[i] = stack[i + 1];
2931 --*stack_size;
2932
2933 return item;
2934}
2935
2936static INLINE int stack_pop_end(int *stack, int *stack_size) {
2937 int item = stack[*stack_size - 1];
2938 stack[*stack_size - 1] = -1;
2939 --*stack_size;
2940
2941 return item;
2942}
2943
2944static INLINE void stack_reset(int *stack, int *stack_size) {
2945 for (int i = 0; i < *stack_size; ++i) stack[i] = INVALID_IDX;
2946 *stack_size = 0;
2947}
2948
2949// av1 uses 10,000,000 ticks/second as time stamp
2950#define TICKS_PER_SEC 10000000LL
2951
2952static INLINE int64_t
2953timebase_units_to_ticks(const aom_rational64_t *timestamp_ratio, int64_t n) {
2954 return n * timestamp_ratio->num / timestamp_ratio->den;
2955}
2956
2957static INLINE int64_t
2958ticks_to_timebase_units(const aom_rational64_t *timestamp_ratio, int64_t n) {
2959 int64_t round = timestamp_ratio->num / 2;
2960 if (round > 0) --round;
2961 return (n * timestamp_ratio->den + round) / timestamp_ratio->num;
2962}
2963
2964static INLINE int frame_is_kf_gf_arf(const AV1_COMP *cpi) {
2965 const GF_GROUP *const gf_group = &cpi->gf_group;
2966 const FRAME_UPDATE_TYPE update_type =
2967 gf_group->update_type[cpi->gf_frame_index];
2968
2969 return frame_is_intra_only(&cpi->common) || update_type == ARF_UPDATE ||
2970 update_type == GF_UPDATE;
2971}
2972
2973// TODO(huisu@google.com, youzhou@microsoft.com): enable hash-me for HBD.
2974static INLINE int av1_use_hash_me(const AV1_COMP *const cpi) {
2977 frame_is_intra_only(&cpi->common));
2978}
2979
2980static INLINE const YV12_BUFFER_CONFIG *get_ref_frame_yv12_buf(
2981 const AV1_COMMON *const cm, MV_REFERENCE_FRAME ref_frame) {
2982 const RefCntBuffer *const buf = get_ref_frame_buf(cm, ref_frame);
2983 return buf != NULL ? &buf->buf : NULL;
2984}
2985
2986static INLINE void alloc_frame_mvs(AV1_COMMON *const cm, RefCntBuffer *buf) {
2987 assert(buf != NULL);
2988 ensure_mv_buffer(buf, cm);
2989 buf->width = cm->width;
2990 buf->height = cm->height;
2991}
2992
2993// Get the allocated token size for a tile. It does the same calculation as in
2994// the frame token allocation.
2995static INLINE unsigned int allocated_tokens(TileInfo tile, int sb_size_log2,
2996 int num_planes) {
2997 int tile_mb_rows = (tile.mi_row_end - tile.mi_row_start + 2) >> 2;
2998 int tile_mb_cols = (tile.mi_col_end - tile.mi_col_start + 2) >> 2;
2999
3000 return get_token_alloc(tile_mb_rows, tile_mb_cols, sb_size_log2, num_planes);
3001}
3002
3003static INLINE void get_start_tok(AV1_COMP *cpi, int tile_row, int tile_col,
3004 int mi_row, TokenExtra **tok, int sb_size_log2,
3005 int num_planes) {
3006 AV1_COMMON *const cm = &cpi->common;
3007 const int tile_cols = cm->tiles.cols;
3008 TileDataEnc *this_tile = &cpi->tile_data[tile_row * tile_cols + tile_col];
3009 const TileInfo *const tile_info = &this_tile->tile_info;
3010
3011 const int tile_mb_cols =
3012 (tile_info->mi_col_end - tile_info->mi_col_start + 2) >> 2;
3013 const int tile_mb_row = (mi_row - tile_info->mi_row_start + 2) >> 2;
3014
3015 *tok = cpi->token_info.tile_tok[tile_row][tile_col] +
3016 get_token_alloc(tile_mb_row, tile_mb_cols, sb_size_log2, num_planes);
3017}
3018
3019void av1_apply_encoding_flags(AV1_COMP *cpi, aom_enc_frame_flags_t flags);
3020
3021#define ALT_MIN_LAG 3
3022static INLINE int is_altref_enabled(int lag_in_frames, bool enable_auto_arf) {
3023 return lag_in_frames >= ALT_MIN_LAG && enable_auto_arf;
3024}
3025
3026// Check if statistics generation stage
3027static INLINE int is_stat_generation_stage(const AV1_COMP *const cpi) {
3028 assert(IMPLIES(cpi->compressor_stage == LAP_STAGE,
3029 cpi->oxcf.pass == 0 && cpi->lap_enabled));
3030 return (cpi->oxcf.pass == 1 || (cpi->compressor_stage == LAP_STAGE));
3031}
3032// Check if statistics consumption stage
3033static INLINE int is_stat_consumption_stage_twopass(const AV1_COMP *const cpi) {
3034 return (cpi->oxcf.pass == 2);
3035}
3036
3037// Check if statistics consumption stage
3038static INLINE int is_stat_consumption_stage(const AV1_COMP *const cpi) {
3039 return (is_stat_consumption_stage_twopass(cpi) ||
3040 (cpi->oxcf.pass == 0 && (cpi->compressor_stage == ENCODE_STAGE) &&
3041 cpi->lap_enabled));
3042}
3043
3053static INLINE int has_no_stats_stage(const AV1_COMP *const cpi) {
3054 assert(IMPLIES(!cpi->lap_enabled, cpi->compressor_stage == ENCODE_STAGE));
3055 return (cpi->oxcf.pass == 0 && !cpi->lap_enabled);
3056}
3059// Function return size of frame stats buffer
3060static INLINE int get_stats_buf_size(int num_lap_buffer, int num_lag_buffer) {
3061 /* if lookahead is enabled return num_lap_buffers else num_lag_buffers */
3062 return (num_lap_buffer > 0 ? num_lap_buffer + 1 : num_lag_buffer);
3063}
3064
3065// TODO(zoeliu): To set up cpi->oxcf.gf_cfg.enable_auto_brf
3066
3067static INLINE void set_ref_ptrs(const AV1_COMMON *cm, MACROBLOCKD *xd,
3068 MV_REFERENCE_FRAME ref0,
3069 MV_REFERENCE_FRAME ref1) {
3071 get_ref_scale_factors_const(cm, ref0 >= LAST_FRAME ? ref0 : 1);
3073 get_ref_scale_factors_const(cm, ref1 >= LAST_FRAME ? ref1 : 1);
3074}
3075
3076static INLINE int get_chessboard_index(int frame_index) {
3077 return frame_index & 0x1;
3078}
3079
3080static INLINE const int *cond_cost_list_const(const struct AV1_COMP *cpi,
3081 const int *cost_list) {
3082 const int use_cost_list = cpi->sf.mv_sf.subpel_search_method != SUBPEL_TREE &&
3083 cpi->sf.mv_sf.use_fullpel_costlist;
3084 return use_cost_list ? cost_list : NULL;
3085}
3086
3087static INLINE int *cond_cost_list(const struct AV1_COMP *cpi, int *cost_list) {
3088 const int use_cost_list = cpi->sf.mv_sf.subpel_search_method != SUBPEL_TREE &&
3089 cpi->sf.mv_sf.use_fullpel_costlist;
3090 return use_cost_list ? cost_list : NULL;
3091}
3092
3093// Compression ratio of current frame.
3094double av1_get_compression_ratio(const AV1_COMMON *const cm,
3095 size_t encoded_frame_size);
3096
3097void av1_new_framerate(AV1_COMP *cpi, double framerate);
3098
3099void av1_setup_frame_size(AV1_COMP *cpi);
3100
3101#define LAYER_IDS_TO_IDX(sl, tl, num_tl) ((sl) * (num_tl) + (tl))
3102
3103// Returns 1 if a frame is scaled and 0 otherwise.
3104static INLINE int av1_resize_scaled(const AV1_COMMON *cm) {
3105 return !(cm->superres_upscaled_width == cm->render_width &&
3107}
3108
3109static INLINE int av1_frame_scaled(const AV1_COMMON *cm) {
3110 return !av1_superres_scaled(cm) && av1_resize_scaled(cm);
3111}
3112
3113// Don't allow a show_existing_frame to coincide with an error resilient
3114// frame. An exception can be made for a forward keyframe since it has no
3115// previous dependencies.
3116static INLINE int encode_show_existing_frame(const AV1_COMMON *cm) {
3118 cm->current_frame.frame_type == KEY_FRAME);
3119}
3120
3121// Get index into the 'cpi->mbmi_ext_info.frame_base' array for the given
3122// 'mi_row' and 'mi_col'.
3123static INLINE int get_mi_ext_idx(const int mi_row, const int mi_col,
3124 const BLOCK_SIZE mi_alloc_bsize,
3125 const int mbmi_ext_stride) {
3126 const int mi_ext_size_1d = mi_size_wide[mi_alloc_bsize];
3127 const int mi_ext_row = mi_row / mi_ext_size_1d;
3128 const int mi_ext_col = mi_col / mi_ext_size_1d;
3129 return mi_ext_row * mbmi_ext_stride + mi_ext_col;
3130}
3131
3132// Lighter version of set_offsets that only sets the mode info
3133// pointers.
3134static INLINE void set_mode_info_offsets(
3135 const CommonModeInfoParams *const mi_params,
3136 const MBMIExtFrameBufferInfo *const mbmi_ext_info, MACROBLOCK *const x,
3137 MACROBLOCKD *const xd, int mi_row, int mi_col) {
3138 set_mi_offsets(mi_params, xd, mi_row, mi_col);
3139 const int ext_idx = get_mi_ext_idx(mi_row, mi_col, mi_params->mi_alloc_bsize,
3140 mbmi_ext_info->stride);
3141 x->mbmi_ext_frame = mbmi_ext_info->frame_base + ext_idx;
3142}
3143
3144// Check to see if the given partition size is allowed for a specified number
3145// of mi block rows and columns remaining in the image.
3146// If not then return the largest allowed partition size
3147static INLINE BLOCK_SIZE find_partition_size(BLOCK_SIZE bsize, int rows_left,
3148 int cols_left, int *bh, int *bw) {
3149 int int_size = (int)bsize;
3150 if (rows_left <= 0 || cols_left <= 0) {
3151 return AOMMIN(bsize, BLOCK_8X8);
3152 } else {
3153 for (; int_size > 0; int_size -= 3) {
3154 *bh = mi_size_high[int_size];
3155 *bw = mi_size_wide[int_size];
3156 if ((*bh <= rows_left) && (*bw <= cols_left)) {
3157 break;
3158 }
3159 }
3160 }
3161 return (BLOCK_SIZE)int_size;
3162}
3163
3164static const uint8_t av1_ref_frame_flag_list[REF_FRAMES] = { 0,
3165 AOM_LAST_FLAG,
3166 AOM_LAST2_FLAG,
3167 AOM_LAST3_FLAG,
3168 AOM_GOLD_FLAG,
3169 AOM_BWD_FLAG,
3170 AOM_ALT2_FLAG,
3171 AOM_ALT_FLAG };
3172
3173// When more than 'max_allowed_refs' are available, we reduce the number of
3174// reference frames one at a time based on this order.
3175static const MV_REFERENCE_FRAME disable_order[] = {
3176 LAST3_FRAME,
3177 LAST2_FRAME,
3178 ALTREF2_FRAME,
3179 GOLDEN_FRAME,
3180};
3181
3182static const MV_REFERENCE_FRAME
3183 ref_frame_priority_order[INTER_REFS_PER_FRAME] = {
3184 LAST_FRAME, ALTREF_FRAME, BWDREF_FRAME, GOLDEN_FRAME,
3185 ALTREF2_FRAME, LAST2_FRAME, LAST3_FRAME,
3186 };
3187
3188static INLINE int get_ref_frame_flags(const SPEED_FEATURES *const sf,
3189 const YV12_BUFFER_CONFIG **ref_frames,
3190 const int ext_ref_frame_flags) {
3191 // cpi->ext_flags.ref_frame_flags allows certain reference types to be
3192 // disabled by the external interface. These are set by
3193 // av1_apply_encoding_flags(). Start with what the external interface allows,
3194 // then suppress any reference types which we have found to be duplicates.
3195 int flags = ext_ref_frame_flags;
3196
3197 for (int i = 1; i < INTER_REFS_PER_FRAME; ++i) {
3198 const YV12_BUFFER_CONFIG *const this_ref = ref_frames[i];
3199 // If this_ref has appeared before, mark the corresponding ref frame as
3200 // invalid. For nonrd mode, only disable GOLDEN_FRAME if it's the same
3201 // as LAST_FRAME or ALTREF_FRAME (if ALTREF is being used in nonrd).
3202 int index = (sf->rt_sf.use_nonrd_pick_mode &&
3203 ref_frame_priority_order[i] == GOLDEN_FRAME)
3204 ? (1 + sf->rt_sf.use_nonrd_altref_frame)
3205 : i;
3206 for (int j = 0; j < index; ++j) {
3207 if (this_ref == ref_frames[j]) {
3208 flags &= ~(1 << (ref_frame_priority_order[i] - 1));
3209 break;
3210 }
3211 }
3212 }
3213 return flags;
3214}
3215
3216// Returns a Sequence Header OBU stored in an aom_fixed_buf_t, or NULL upon
3217// failure. When a non-NULL aom_fixed_buf_t pointer is returned by this
3218// function, the memory must be freed by the caller. Both the buf member of the
3219// aom_fixed_buf_t, and the aom_fixed_buf_t pointer itself must be freed. Memory
3220// returned must be freed via call to free().
3221//
3222// Note: The OBU returned is in Low Overhead Bitstream Format. Specifically,
3223// the obu_has_size_field bit is set, and the buffer contains the obu_size
3224// field.
3225aom_fixed_buf_t *av1_get_global_headers(AV1_COMP *cpi);
3226
3227#define MAX_GFUBOOST_FACTOR 10.0
3228#define MIN_GFUBOOST_FACTOR 4.0
3229
3230static INLINE int is_frame_tpl_eligible(const GF_GROUP *const gf_group,
3231 uint8_t index) {
3232 const FRAME_UPDATE_TYPE update_type = gf_group->update_type[index];
3233 return update_type == ARF_UPDATE || update_type == GF_UPDATE ||
3234 update_type == KF_UPDATE;
3235}
3236
3237static INLINE int is_frame_eligible_for_ref_pruning(const GF_GROUP *gf_group,
3238 int selective_ref_frame,
3239 int prune_ref_frames,
3240 int gf_index) {
3241 return (selective_ref_frame > 0) && (prune_ref_frames > 0) &&
3242 !is_frame_tpl_eligible(gf_group, gf_index);
3243}
3244
3245// Get update type of the current frame.
3246static INLINE FRAME_UPDATE_TYPE get_frame_update_type(const GF_GROUP *gf_group,
3247 int gf_frame_index) {
3248 return gf_group->update_type[gf_frame_index];
3249}
3250
3251static INLINE int av1_pixels_to_mi(int pixels) {
3252 return ALIGN_POWER_OF_TWO(pixels, 3) >> MI_SIZE_LOG2;
3253}
3254
3255static AOM_INLINE int is_psnr_calc_enabled(const AV1_COMP *cpi) {
3256 const AV1_COMMON *const cm = &cpi->common;
3257
3258 return cpi->b_calculate_psnr && !is_stat_generation_stage(cpi) &&
3259 cm->show_frame;
3260}
3261
3262#if CONFIG_AV1_TEMPORAL_DENOISING
3263static INLINE int denoise_svc(const struct AV1_COMP *const cpi) {
3264 return (!cpi->use_svc || (cpi->use_svc && cpi->svc.spatial_layer_id >=
3265 cpi->svc.first_layer_denoise));
3266}
3267#endif
3268
3269#if CONFIG_COLLECT_PARTITION_STATS == 2
3270static INLINE void av1_print_fr_partition_timing_stats(
3271 const FramePartitionTimingStats *part_stats, const char *filename) {
3272 FILE *f = fopen(filename, "w");
3273 if (!f) {
3274 return;
3275 }
3276
3277 fprintf(f, "bsize,redo,");
3278 for (int part = 0; part < EXT_PARTITION_TYPES; part++) {
3279 fprintf(f, "decision_%d,", part);
3280 }
3281 for (int part = 0; part < EXT_PARTITION_TYPES; part++) {
3282 fprintf(f, "attempt_%d,", part);
3283 }
3284 for (int part = 0; part < EXT_PARTITION_TYPES; part++) {
3285 fprintf(f, "time_%d,", part);
3286 }
3287 fprintf(f, "\n");
3288
3289 static const int bsizes[6] = { 128, 64, 32, 16, 8, 4 };
3290
3291 for (int bsize_idx = 0; bsize_idx < 6; bsize_idx++) {
3292 fprintf(f, "%d,%d,", bsizes[bsize_idx], part_stats->partition_redo);
3293 for (int part = 0; part < EXT_PARTITION_TYPES; part++) {
3294 fprintf(f, "%d,", part_stats->partition_decisions[bsize_idx][part]);
3295 }
3296 for (int part = 0; part < EXT_PARTITION_TYPES; part++) {
3297 fprintf(f, "%d,", part_stats->partition_attempts[bsize_idx][part]);
3298 }
3299 for (int part = 0; part < EXT_PARTITION_TYPES; part++) {
3300 fprintf(f, "%ld,", part_stats->partition_times[bsize_idx][part]);
3301 }
3302 fprintf(f, "\n");
3303 }
3304 fclose(f);
3305}
3306#endif // CONFIG_COLLECT_PARTITION_STATS == 2
3307
3308#if CONFIG_COLLECT_PARTITION_STATS
3309static INLINE int av1_get_bsize_idx_for_part_stats(BLOCK_SIZE bsize) {
3310 assert(bsize == BLOCK_128X128 || bsize == BLOCK_64X64 ||
3311 bsize == BLOCK_32X32 || bsize == BLOCK_16X16 || bsize == BLOCK_8X8 ||
3312 bsize == BLOCK_4X4);
3313 switch (bsize) {
3314 case BLOCK_128X128: return 0;
3315 case BLOCK_64X64: return 1;
3316 case BLOCK_32X32: return 2;
3317 case BLOCK_16X16: return 3;
3318 case BLOCK_8X8: return 4;
3319 case BLOCK_4X4: return 5;
3320 default: assert(0 && "Invalid bsize for partition_stats."); return -1;
3321 }
3322}
3323#endif // CONFIG_COLLECT_PARTITION_STATS
3324
3325#if CONFIG_COLLECT_COMPONENT_TIMING
3326static INLINE void start_timing(AV1_COMP *cpi, int component) {
3327 aom_usec_timer_start(&cpi->component_timer[component]);
3328}
3329static INLINE void end_timing(AV1_COMP *cpi, int component) {
3330 aom_usec_timer_mark(&cpi->component_timer[component]);
3331 cpi->frame_component_time[component] +=
3332 aom_usec_timer_elapsed(&cpi->component_timer[component]);
3333}
3334static INLINE char const *get_frame_type_enum(int type) {
3335 switch (type) {
3336 case 0: return "KEY_FRAME";
3337 case 1: return "INTER_FRAME";
3338 case 2: return "INTRA_ONLY_FRAME";
3339 case 3: return "S_FRAME";
3340 default: assert(0);
3341 }
3342 return "error";
3343}
3344#endif
3345
3348#ifdef __cplusplus
3349} // extern "C"
3350#endif
3351
3352#endif // AOM_AV1_ENCODER_ENCODER_H_
#define FIXED_QP_OFFSET_COUNT
Number of fixed QP offsets.
Definition: aom_encoder.h:876
enum aom_chroma_sample_position aom_chroma_sample_position_t
List of chroma sample positions.
enum aom_transfer_characteristics aom_transfer_characteristics_t
List of supported transfer functions.
enum aom_color_range aom_color_range_t
List of supported color range.
enum aom_color_primaries aom_color_primaries_t
List of supported color primaries.
enum aom_matrix_coefficients aom_matrix_coefficients_t
List of supported matrix coefficients.
Provides definitions for using AOM or AV1 encoder algorithm within the aom Codec Interface.
struct AV1_COMP AV1_COMP
Top level encoder structure.
COST_UPDATE_TYPE
This enum controls how often the entropy costs should be updated.
Definition: encoder.h:178
@ COST_UPD_SBROW
Definition: encoder.h:180
@ COST_UPD_TILE
Definition: encoder.h:181
@ COST_UPD_OFF
Definition: encoder.h:182
@ COST_UPD_SB
Definition: encoder.h:179
struct EncodeFrameParams EncodeFrameParams
contains per-frame encoding parameters decided upon by av1_encode_strategy() and passed down to av1_e...
struct EncodeFrameInput EncodeFrameInput
Input frames and last input frame.
struct MultiThreadInfo MultiThreadInfo
Encoder parameters related to multi-threading.
struct AV1_PRIMARY AV1_PRIMARY
Top level primary encoder structure.
struct inter_modes_info InterModesInfo
Struct used to hold inter mode data for fast tx search.
struct AV1EncoderConfig AV1EncoderConfig
Main encoder configuration data structure.
@ RESTORE_SWITCHABLE_TYPES
Definition: enums.h:655
aom_tune_content
Definition: aomcx.h:1400
aom_tune_metric
Model tuning parameters.
Definition: aomcx.h:1419
enum aom_bit_depth aom_bit_depth_t
Bit depth for codecThis enumeration determines the bit depth of the codec.
enum aom_superblock_size aom_superblock_size_t
Superblock size selection.
aom_codec_err_t
Algorithm return codes.
Definition: aom_codec.h:155
aom_superres_mode
Frame super-resolution mode.
Definition: aom_encoder.h:187
aom_rc_mode
Rate control mode.
Definition: aom_encoder.h:165
aom_enc_pass
Multi-pass Encoding Pass.
Definition: aom_encoder.h:158
long aom_enc_frame_flags_t
Encoded Frame Flags.
Definition: aom_encoder.h:357
int av1_receive_raw_frame(AV1_COMP *cpi, aom_enc_frame_flags_t frame_flags, YV12_BUFFER_CONFIG *sd, int64_t time_stamp, int64_t end_time_stamp)
Obtain the raw frame data.
Definition: encoder.c:3478
int av1_encode(AV1_COMP *const cpi, uint8_t *const dest, const EncodeFrameInput *const frame_input, const EncodeFrameParams *const frame_params, EncodeFrameResults *const frame_results)
Run 1-pass/2-pass encoding.
Definition: encoder.c:3389
int av1_get_compressed_data(AV1_COMP *cpi, unsigned int *frame_flags, size_t *size, uint8_t *dest, int64_t *time_stamp, int64_t *time_end, int flush, const aom_rational64_t *timebase)
Encode a frame.
Definition: encoder.c:3683
static int has_no_stats_stage(const AV1_COMP *const cpi)
Check if the current stage has statistics.
Definition: encoder.h:3053
Describes look ahead buffer operations.
Top level common structure used by both encoder and decoder.
Definition: av1_common_int.h:725
int superres_upscaled_width
Definition: av1_common_int.h:774
int superres_upscaled_height
Definition: av1_common_int.h:775
int width
Definition: av1_common_int.h:750
CurrentFrame current_frame
Definition: av1_common_int.h:729
int show_existing_frame
Definition: av1_common_int.h:879
FeatureFlags features
Definition: av1_common_int.h:884
int show_frame
Definition: av1_common_int.h:864
CommonTileParams tiles
Definition: av1_common_int.h:971
int height
Definition: av1_common_int.h:751
int render_width
Definition: av1_common_int.h:761
int render_height
Definition: av1_common_int.h:762
Encoder data related to row-based multi-threading.
Definition: encoder.h:1321
pthread_mutex_t * mutex_
Definition: encoder.h:1354
int allocated_tile_cols
Definition: encoder.h:1329
int allocated_cols
Definition: encoder.h:1343
int allocated_tile_rows
Definition: encoder.h:1325
int allocated_rows
Definition: encoder.h:1336
Encoder parameters for synchronization of row based multi-threading.
Definition: encoder.h:1232
int rows
Definition: encoder.h:1257
int sync_range
Definition: encoder.h:1253
int next_mi_row
Definition: encoder.h:1261
pthread_mutex_t * mutex_
Definition: encoder.h:1238
pthread_cond_t * cond_
Definition: encoder.h:1239
int num_threads_working
Definition: encoder.h:1265
Main encoder configuration data structure.
Definition: encoder.h:805
RateControlCfg rc_cfg
Definition: encoder.h:827
KeyFrameCfg kf_cfg
Definition: encoder.h:822
enum aom_enc_pass pass
Definition: encoder.h:930
AlgoCfg algo_cfg
Definition: encoder.h:817
aom_fixed_buf_t twopass_stats_in
Definition: encoder.h:844
Top level encoder structure.
Definition: encoder.h:2095
uint8_t * consec_zero_mv
Definition: encoder.h:2692
struct aom_denoise_and_model_t * denoise_and_model
Definition: encoder.h:2535
int vaq_refresh
Definition: encoder.h:2486
FRAME_TYPE last_frame_type
Definition: encoder.h:2657
YV12_BUFFER_CONFIG * unscaled_source
Definition: encoder.h:2176
CYCLIC_REFRESH * cyclic_refresh
Definition: encoder.h:2335
int partition_search_skippable_frame
Definition: encoder.h:2212
aom_variance_fn_ptr_t fn_ptr[BLOCK_SIZES_ALL]
Definition: encoder.h:2347
InitialDimensions initial_dimensions
Definition: encoder.h:2453
RATE_CONTROL rc
Definition: encoder.h:2294
int deltaq_used
Definition: encoder.h:2587
ActiveMap active_map
Definition: encoder.h:2340
TokenInfo token_info
Definition: encoder.h:2481
int64_t ambient_err
Definition: encoder.h:2263
aom_film_grain_table_t * film_grain_table
Definition: encoder.h:2528
int b_calculate_psnr
Definition: encoder.h:2425
int lap_enabled
Definition: encoder.h:2641
SPEED_FEATURES sf
Definition: encoder.h:2314
TRELLIS_OPT_TYPE optimize_seg_arr[8]
Definition: encoder.h:2154
double * ssim_rdmult_scaling_factors
Definition: encoder.h:2613
RD_OPT rd
Definition: encoder.h:2268
int num_tg
Definition: encoder.h:2662
FrameProbInfo frame_probs
Definition: encoder.h:2496
WinnerModeParams winner_mode_params
Definition: encoder.h:2284
double * tpl_sb_rdmult_scaling_factors
Definition: encoder.h:2607
ExternalFlags ext_flags
Definition: encoder.h:2246
EncSegmentationInfo enc_seg
Definition: encoder.h:2330
MotionVectorSearchParams mv_search_params
Definition: encoder.h:2319
int use_screen_content_tools
Definition: encoder.h:2550
CODING_CONTEXT coding_context
Definition: encoder.h:2274
TemporalFilterCtx tf_ctx
Definition: encoder.h:2207
ForceIntegerMVInfo force_intpel_info
Definition: encoder.h:2217
GlobalMotionInfo gm_info
Definition: encoder.h:2279
CoeffBufferPool coeff_buffer_pool
Definition: encoder.h:2133
FRAME_INDEX_SET frame_index_set
Definition: encoder.h:2448
int ref_frame_flags
Definition: encoder.h:2304
RefCntBuffer * scaled_ref_buf[INTER_REFS_PER_FRAME]
Definition: encoder.h:2223
unsigned char gf_frame_index
Definition: encoder.h:2362
AV1EncoderConfig oxcf
Definition: encoder.h:2143
AV1_COMMON common
Definition: encoder.h:2138
int show_existing_alt_ref
Definition: encoder.h:2383
AV1LrStruct lr_ctxt
Definition: encoder.h:2523
TplParams tpl_data
Definition: encoder.h:2202
FRAME_COUNTS counts
Definition: encoder.h:2115
GF_GROUP gf_group
Definition: encoder.h:2357
COMPRESSOR_STAGE compressor_stage
Definition: encoder.h:2645
int intrabc_used
Definition: encoder.h:2513
int no_show_fwd_kf
Definition: encoder.h:2148
int prune_ref_frame_mask
Definition: encoder.h:2518
YV12_BUFFER_CONFIG * unscaled_last_source
Definition: encoder.h:2186
int all_one_sided_refs
Definition: encoder.h:2325
MultiThreadInfo mt_info
Definition: encoder.h:2501
VarBasedPartitionInfo vbp_info
Definition: encoder.h:2491
YV12_BUFFER_CONFIG * last_source
Definition: encoder.h:2170
int existing_fb_idx_to_show
Definition: encoder.h:2508
YV12_BUFFER_CONFIG alt_ref_buffer
Definition: encoder.h:2378
YV12_BUFFER_CONFIG * unfiltered_source
Definition: encoder.h:2197
GF_STATE gf_state
Definition: encoder.h:2367
FRAME_INFO frame_info
Definition: encoder.h:2443
int droppable
Definition: encoder.h:2438
RefCntBuffer * last_show_frame_buf
Definition: encoder.h:2228
aom_superres_mode superres_mode
Definition: encoder.h:2669
MBMIExtFrameBufferInfo mbmi_ext_info
Definition: encoder.h:2120
AV1_PRIMARY * ppi
Definition: encoder.h:2099
int use_svc
Definition: encoder.h:2632
double * tpl_rdmult_scaling_factors
Definition: encoder.h:2602
RefBufferStack ref_buffer_stack
Definition: encoder.h:2372
int sb_counter
Definition: encoder.h:2708
TileDataEnc * tile_data
Definition: encoder.h:2472
int is_screen_content_type
Definition: encoder.h:2558
InterpSearchFlags interp_search_flags
Definition: encoder.h:2541
AV1LevelParams level_params
Definition: encoder.h:2582
TimeStamps time_stamps
Definition: encoder.h:2289
int frames_left
Definition: encoder.h:2697
FirstPassData firstpass_data
Definition: encoder.h:2674
double framerate
Definition: encoder.h:2299
TWO_PASS twopass
Definition: encoder.h:2352
int speed
Definition: encoder.h:2309
BLOCK_SIZE fp_block_size
Definition: encoder.h:2702
int fb_of_context_type[REF_FRAMES]
Definition: encoder.h:2241
YV12_BUFFER_CONFIG trial_frame_rst
Definition: encoder.h:2258
ThreadData td
Definition: encoder.h:2110
ResizePendingParams resize_pending_params
Definition: encoder.h:2466
YV12_BUFFER_CONFIG scaled_source
Definition: encoder.h:2181
YV12_BUFFER_CONFIG last_frame_uf
Definition: encoder.h:2252
EncQuantDequantParams enc_quant_dequant_params
Definition: encoder.h:2105
RefFrameDistanceInfo ref_frame_dist_info
Definition: encoder.h:2592
int initial_mbs
Definition: encoder.h:2461
YV12_BUFFER_CONFIG scaled_last_source
Definition: encoder.h:2191
MV_STATS mv_stats
Definition: encoder.h:2651
YV12_BUFFER_CONFIG * source
Definition: encoder.h:2161
int allocated_tiles
Definition: encoder.h:2476
RefreshFrameFlagsInfo refresh_frame
Definition: encoder.h:2233
SVC svc
Definition: encoder.h:2636
CB_COEFF_BUFFER * coeff_buffer_base
Definition: encoder.h:2127
NOISE_ESTIMATE noise_estimate
Definition: encoder.h:2679
Top level primary encoder structure.
Definition: encoder.h:2057
struct aom_codec_pkt_list * output_pkt_list
Definition: encoder.h:2084
struct AV1_COMP * cpi
Definition: encoder.h:2061
int internal_altref_allowed
Definition: encoder.h:2089
struct AV1_COMP * cpi_lap
Definition: encoder.h:2066
struct lookahead_ctx * lookahead
Definition: encoder.h:2071
int seq_params_locked
Definition: encoder.h:2078
Algorithm configuration parameters.
Definition: encoder.h:717
int disable_trellis_quant
Definition: encoder.h:730
int sharpness
Definition: encoder.h:721
int arnr_max_frames
Definition: encoder.h:735
bool enable_tpl_model
Definition: encoder.h:753
uint8_t cdf_update_mode
Definition: encoder.h:748
bool enable_overlay
Definition: encoder.h:759
int arnr_strength
Definition: encoder.h:740
Stores the transforms coefficients for the whole superblock.
Definition: block.h:180
The stucture of CYCLIC_REFRESH.
Definition: aq_cyclicrefresh.h:36
Parameters related to CDEF.
Definition: av1_common_int.h:196
Definition: encoder.h:2039
uint8_t * entropy_ctx
Definition: encoder.h:2051
tran_low_t * tcoeff
Definition: encoder.h:2043
uint16_t * eobs
Definition: encoder.h:2047
Params related to MB_MODE_INFO arrays and related info.
Definition: av1_common_int.h:479
BLOCK_SIZE mi_alloc_bsize
Definition: av1_common_int.h:528
int cols
Definition: av1_common_int.h:406
Encoder flags for compound prediction modes.
Definition: encoder.h:310
bool enable_masked_comp
Definition: encoder.h:319
bool enable_diff_wtd_comp
Definition: encoder.h:327
bool enable_smooth_interintra
Definition: encoder.h:323
bool enable_interintra_wedge
Definition: encoder.h:335
bool enable_interinter_wedge
Definition: encoder.h:331
bool enable_dist_wtd_comp
Definition: encoder.h:314
Contains buffers used by av1_compound_type_rd()
Definition: block.h:372
Segmentation related information for the current frame.
Definition: encoder.h:2002
uint8_t * map
Definition: encoder.h:2008
bool has_lossless_segment
Definition: encoder.h:2014
Input frames and last input frame.
Definition: encoder.h:2714
contains per-frame encoding parameters decided upon by av1_encode_strategy() and passed down to av1_e...
Definition: encoder.h:2726
int error_resilient_mode
Definition: encoder.h:2730
int remapped_ref_idx[REF_FRAMES]
Definition: encoder.h:2761
int ref_frame_flags
Definition: encoder.h:2756
int speed
Definition: encoder.h:2772
FRAME_TYPE frame_type
Definition: encoder.h:2734
int show_frame
Definition: encoder.h:2744
RefreshFrameFlagsInfo refresh_frame
Definition: encoder.h:2767
Frame refresh flags set by the external interface.
Definition: encoder.h:1871
bool golden_frame
Definition: encoder.h:1873
bool bwd_ref_frame
Definition: encoder.h:1874
bool update_pending
Definition: encoder.h:1880
bool last_frame
Definition: encoder.h:1872
bool alt_ref_frame
Definition: encoder.h:1876
bool alt2_ref_frame
Definition: encoder.h:1875
Flags signalled by the external interface at frame level.
Definition: encoder.h:1886
bool use_primary_ref_none
Definition: encoder.h:1927
bool use_ref_frame_mvs
Definition: encoder.h:1911
ExtRefreshFrameFlagsInfo refresh_frame
Definition: encoder.h:1895
int ref_frame_flags
Definition: encoder.h:1890
bool use_error_resilient
Definition: encoder.h:1916
bool use_s_frame
Definition: encoder.h:1921
bool refresh_frame_context
Definition: encoder.h:1900
bool refresh_frame_context_pending
Definition: encoder.h:1906
The stucture of acummulated frame stats in the first pass.
Definition: firstpass.h:36
Frame level features.
Definition: av1_common_int.h:336
bool allow_screen_content_tools
Definition: av1_common_int.h:353
bool allow_intrabc
Definition: av1_common_int.h:354
bool error_resilient_mode
Definition: av1_common_int.h:378
Encoder info used for decision on forcing integer motion vectors.
Definition: encoder.h:1464
int rate_size
Definition: encoder.h:1477
int rate_index
Definition: encoder.h:1473
Encoder-side probabilities for pruning of various AV1 tools.
Definition: encoder.h:960
Data related to the current GF/ARF group and the individual frames within the group.
Definition: firstpass.h:171
Parameters related to global motion search.
Definition: encoder.h:1643
bool search_done
Definition: encoder.h:1647
int segment_map_h
Definition: encoder.h:1681
int num_src_corners
Definition: encoder.h:1687
unsigned char * src_buffer
Definition: encoder.h:1659
int segment_map_w
Definition: encoder.h:1680
Initial frame dimensions.
Definition: encoder.h:1706
int height
Definition: encoder.h:1708
int width
Definition: encoder.h:1707
Flags related to interpolation filter search.
Definition: encoder.h:1714
int default_interp_skip_flags
Definition: encoder.h:1719
uint16_t interp_filter_search_mask
Definition: encoder.h:1723
Encoder flags for intra prediction.
Definition: encoder.h:236
bool enable_diagonal_intra
Definition: encoder.h:261
bool enable_smooth_intra
Definition: encoder.h:249
bool enable_filter_intra
Definition: encoder.h:245
bool enable_paeth_intra
Definition: encoder.h:253
bool enable_intra_edge_filter
Definition: encoder.h:240
bool enable_cfl_intra
Definition: encoder.h:257
bool enable_angle_delta
Definition: encoder.h:266
Encoder config related to the coding of key frames.
Definition: encoder.h:377
int key_freq_max
Definition: encoder.h:386
int sframe_mode
Definition: encoder.h:404
bool auto_key
Definition: encoder.h:409
bool enable_intrabc
Definition: encoder.h:424
int sframe_dist
Definition: encoder.h:397
bool enable_sframe
Definition: encoder.h:419
int enable_keyframe_filtering
Definition: encoder.h:391
int key_freq_min
Definition: encoder.h:381
bool fwd_kf_enabled
Definition: encoder.h:414
Buffer to store mode information at mi_alloc_bsize (4x4 or 8x8) level.
Definition: encoder.h:1510
int alloc_size
Definition: encoder.h:1519
int stride
Definition: encoder.h:1523
MB_MODE_INFO_EXT_FRAME * frame_base
Definition: encoder.h:1515
Stores best extended mode information at frame level.
Definition: block.h:216
Stores the prediction/txfm mode of the current coding block.
Definition: blockd.h:216
Parameters for motion vector search process.
Definition: encoder.h:1729
int max_mv_magnitude
Definition: encoder.h:1735
fractional_mv_step_fp * find_fractional_mv_step
Definition: encoder.h:1749
int mv_step_param
Definition: encoder.h:1740
Encoder parameters related to multi-threading.
Definition: encoder.h:1375
AV1LrSync lr_row_sync
Definition: encoder.h:1433
struct EncWorkerData * tile_thr_data
Definition: encoder.h:1407
AV1TplRowMultiThreadInfo tpl_row_mt
Definition: encoder.h:1423
int enc_mt_buf_init_done
Definition: encoder.h:1396
AV1EncRowMultiThreadInfo enc_row_mt
Definition: encoder.h:1418
AV1LfSync lf_row_sync
Definition: encoder.h:1428
AV1CdefSync cdef_sync
Definition: encoder.h:1448
int num_mod_workers[NUM_MT_MODULES]
Definition: encoder.h:1384
int num_workers
Definition: encoder.h:1379
AVxWorker * workers
Definition: encoder.h:1401
int fp_mt_buf_init_done
Definition: encoder.h:1390
bool row_mt_enabled
Definition: encoder.h:1413
AV1TemporalFilterSync tf_sync
Definition: encoder.h:1443
AV1GlobalMotionSync gm_sync
Definition: encoder.h:1438
Contains buffers used to speed up rdopt for obmc.
Definition: block.h:335
Contains color maps used in palette mode.
Definition: block.h:360
Encoder config for coding block partitioning.
Definition: encoder.h:208
bool enable_rect_partitions
Definition: encoder.h:212
bool enable_1to4_partitions
Definition: encoder.h:220
BLOCK_SIZE max_partition_size
Definition: encoder.h:230
bool enable_ab_partitions
Definition: encoder.h:216
BLOCK_SIZE min_partition_size
Definition: encoder.h:225
Rate Control parameters and status.
Definition: ratectrl.h:118
Encoder rate control configuration parameters.
Definition: encoder.h:430
int worst_allowed_q
Definition: encoder.h:500
int over_shoot_pct
Definition: encoder.h:495
unsigned int max_intra_bitrate_pct
Definition: encoder.h:465
int drop_frames_water_mark
Definition: encoder.h:483
int vbrmax_section
Definition: encoder.h:531
int64_t maximum_buffer_size_ms
Definition: encoder.h:449
unsigned int vbr_corpus_complexity_lap
Definition: encoder.h:460
unsigned int min_cr
Definition: encoder.h:479
int vbrbias
Definition: encoder.h:521
unsigned int gf_cbr_boost_pct
Definition: encoder.h:474
int vbrmin_section
Definition: encoder.h:526
enum aom_rc_mode mode
Definition: encoder.h:514
unsigned int max_inter_bitrate_pct
Definition: encoder.h:470
int64_t starting_buffer_level_ms
Definition: encoder.h:439
int best_allowed_q
Definition: encoder.h:505
int under_shoot_pct
Definition: encoder.h:489
int64_t target_bandwidth
Definition: encoder.h:454
int64_t optimal_buffer_level_ms
Definition: encoder.h:444
int cq_level
Definition: encoder.h:509
Refrence frame distance related variables.
Definition: encoder.h:1788
int8_t nearest_past_ref
Definition: encoder.h:1796
int8_t nearest_future_ref
Definition: encoder.h:1800
Refresh frame flags for different type of frames.
Definition: encoder.h:1767
bool alt_ref_frame
Definition: encoder.h:1770
bool bwd_ref_frame
Definition: encoder.h:1769
bool golden_frame
Definition: encoder.h:1768
Encoder config related to resize.
Definition: encoder.h:188
uint8_t resize_scale_denominator
Definition: encoder.h:197
uint8_t resize_kf_scale_denominator
Definition: encoder.h:202
RESIZE_MODE resize_mode
Definition: encoder.h:192
Desired dimensions for an externally triggered resize.
Definition: encoder.h:1780
int width
Definition: encoder.h:1781
int height
Definition: encoder.h:1782
Top level speed vs quality trade off data struture.
Definition: speed_features.h:1120
MV_SPEED_FEATURES mv_sf
Definition: speed_features.h:1144
REAL_TIME_SPEED_FEATURES rt_sf
Definition: speed_features.h:1184
The stucture of SVC.
Definition: svc_layercontext.h:91
Encoder config related to frame super-resolution.
Definition: encoder.h:341
uint8_t superres_kf_scale_denominator
Definition: encoder.h:363
aom_superres_mode superres_mode
Definition: encoder.h:367
int superres_kf_qthresh
Definition: encoder.h:351
bool enable_superres
Definition: encoder.h:371
uint8_t superres_scale_denominator
Definition: encoder.h:357
int superres_qthresh
Definition: encoder.h:346
Two pass status and control data.
Definition: firstpass.h:214
Parameters related to temporal filtering.
Definition: temporal_filter.h:75
Frame time stamps.
Definition: encoder.h:2020
int64_t prev_ts_start
Definition: encoder.h:2024
int64_t first_ts_start
Definition: encoder.h:2032
int64_t prev_ts_end
Definition: encoder.h:2028
Params related to temporal dependency model.
Definition: tpl_model.h:129
Encoder flags for transform sizes and types.
Definition: encoder.h:272
bool enable_tx64
Definition: encoder.h:276
bool use_inter_dct_only
Definition: encoder.h:299
bool enable_flip_idtx
Definition: encoder.h:280
bool use_intra_default_tx_only
Definition: encoder.h:304
bool use_intra_dct_only
Definition: encoder.h:294
bool enable_rect_tx
Definition: encoder.h:284
bool reduced_tx_type_set
Definition: encoder.h:289
Thresholds for variance based partitioning.
Definition: encoder.h:1211
int64_t threshold_minmax
Definition: encoder.h:1226
Parameters used for winner mode processing.
Definition: encoder.h:1818
Generic fixed size buffer structure.
Definition: aom_encoder.h:76
Struct used to hold inter mode data for fast tx search.
Definition: encoder.h:1116
RD_STATS rd_cost_arr[MAX_INTER_MODES]
Definition: encoder.h:1145
int64_t est_rd_arr[MAX_INTER_MODES]
Definition: encoder.h:1137
int64_t sse_arr[MAX_INTER_MODES]
Definition: encoder.h:1133
RD_STATS rd_cost_y_arr[MAX_INTER_MODES]
Definition: encoder.h:1149
RD_STATS rd_cost_uv_arr[MAX_INTER_MODES]
Definition: encoder.h:1153
MB_MODE_INFO mbmi_arr[MAX_INTER_MODES]
Definition: encoder.h:1125
RdIdxPair rd_idx_pair_arr[MAX_INTER_MODES]
Definition: encoder.h:1141
int mode_rate_arr[MAX_INTER_MODES]
Definition: encoder.h:1129
int num
Definition: encoder.h:1121
Encoder's parameters related to the current coding block.
Definition: block.h:846
MB_MODE_INFO_EXT_FRAME * mbmi_ext_frame
Finalized mbmi_ext for the whole frame.
Definition: block.h:878
Variables related to current coding block.
Definition: blockd.h:568
const struct scale_factors * block_ref_scale_factors[2]
Definition: blockd.h:685
YV12 frame buffer data structure.
Definition: yv12config.h:38