mirror of https://git.ffmpeg.org/ffmpeg.git
VP8: optimize VP8Context struct ordering
Shaves at least 3KB off code size on x86, should improve cache utilization. This would probably be useful to do for other decoders/encoders as well.
This commit is contained in:
parent
3efbe13739
commit
1eeca88691
|
@ -474,7 +474,7 @@ void decode_mvs(VP8Context *s, VP8Macroblock *mb, int mb_x, int mb_y)
|
||||||
enum { EDGE_TOP, EDGE_LEFT, EDGE_TOPLEFT };
|
enum { EDGE_TOP, EDGE_LEFT, EDGE_TOPLEFT };
|
||||||
int idx = CNT_ZERO;
|
int idx = CNT_ZERO;
|
||||||
int cur_sign_bias = s->sign_bias[mb->ref_frame];
|
int cur_sign_bias = s->sign_bias[mb->ref_frame];
|
||||||
int *sign_bias = s->sign_bias;
|
int8_t *sign_bias = s->sign_bias;
|
||||||
VP56mv near_mv[4];
|
VP56mv near_mv[4];
|
||||||
uint8_t cnt[4] = { 0 };
|
uint8_t cnt[4] = { 0 };
|
||||||
VP56RangeCoder *c = &s->c;
|
VP56RangeCoder *c = &s->c;
|
||||||
|
|
178
libavcodec/vp8.h
178
libavcodec/vp8.h
|
@ -85,49 +85,88 @@ typedef struct {
|
||||||
|
|
||||||
typedef struct {
|
typedef struct {
|
||||||
AVCodecContext *avctx;
|
AVCodecContext *avctx;
|
||||||
DSPContext dsp;
|
|
||||||
VP8DSPContext vp8dsp;
|
|
||||||
H264PredContext hpc;
|
|
||||||
vp8_mc_func put_pixels_tab[3][3][3];
|
|
||||||
AVFrame frames[4];
|
|
||||||
AVFrame *framep[4];
|
AVFrame *framep[4];
|
||||||
uint8_t *edge_emu_buffer;
|
uint8_t *edge_emu_buffer;
|
||||||
VP56RangeCoder c; ///< header context, includes mb modes and motion vectors
|
|
||||||
int profile;
|
|
||||||
|
|
||||||
int mb_width; /* number of horizontal MB */
|
uint16_t mb_width; /* number of horizontal MB */
|
||||||
int mb_height; /* number of vertical MB */
|
uint16_t mb_height; /* number of vertical MB */
|
||||||
int linesize;
|
int linesize;
|
||||||
int uvlinesize;
|
int uvlinesize;
|
||||||
|
|
||||||
int keyframe;
|
uint8_t keyframe;
|
||||||
int invisible;
|
uint8_t deblock_filter;
|
||||||
int update_last; ///< update VP56_FRAME_PREVIOUS with the current one
|
uint8_t mbskip_enabled;
|
||||||
int update_golden; ///< VP56_FRAME_NONE if not updated, or which frame to copy if so
|
uint8_t segment; ///< segment of the current macroblock
|
||||||
int update_altref;
|
uint8_t chroma_pred_mode; ///< 8x8c pred mode of the current macroblock
|
||||||
int deblock_filter;
|
uint8_t profile;
|
||||||
|
VP56mv mv_min;
|
||||||
|
VP56mv mv_max;
|
||||||
|
|
||||||
|
int8_t sign_bias[4]; ///< one state [0, 1] per ref frame type
|
||||||
|
int ref_count[3];
|
||||||
|
|
||||||
/**
|
/**
|
||||||
* If this flag is not set, all the probability updates
|
* Base parameters for segmentation, i.e. per-macroblock parameters.
|
||||||
* are discarded after this frame is decoded.
|
* These must be kept unchanged even if segmentation is not used for
|
||||||
|
* a frame, since the values persist between interframes.
|
||||||
*/
|
*/
|
||||||
int update_probabilities;
|
struct {
|
||||||
|
uint8_t enabled;
|
||||||
|
uint8_t absolute_vals;
|
||||||
|
uint8_t update_map;
|
||||||
|
int8_t base_quant[4];
|
||||||
|
int8_t filter_level[4]; ///< base loop filter level
|
||||||
|
} segmentation;
|
||||||
|
|
||||||
/**
|
struct {
|
||||||
* All coefficients are contained in separate arith coding contexts.
|
uint8_t simple;
|
||||||
* There can be 1, 2, 4, or 8 of these after the header context.
|
uint8_t level;
|
||||||
*/
|
uint8_t sharpness;
|
||||||
int num_coeff_partitions;
|
} filter;
|
||||||
VP56RangeCoder coeff_partition[8];
|
|
||||||
|
|
||||||
VP8Macroblock *macroblocks;
|
VP8Macroblock *macroblocks;
|
||||||
VP8Macroblock *macroblocks_base;
|
|
||||||
VP8FilterStrength *filter_strength;
|
VP8FilterStrength *filter_strength;
|
||||||
|
|
||||||
uint8_t *intra4x4_pred_mode_top;
|
uint8_t *intra4x4_pred_mode_top;
|
||||||
uint8_t intra4x4_pred_mode_left[4];
|
uint8_t intra4x4_pred_mode_left[4];
|
||||||
uint8_t *segmentation_map;
|
uint8_t *segmentation_map;
|
||||||
|
|
||||||
|
/**
|
||||||
|
* Macroblocks can have one of 4 different quants in a frame when
|
||||||
|
* segmentation is enabled.
|
||||||
|
* If segmentation is disabled, only the first segment's values are used.
|
||||||
|
*/
|
||||||
|
struct {
|
||||||
|
// [0] - DC qmul [1] - AC qmul
|
||||||
|
int16_t luma_qmul[2];
|
||||||
|
int16_t luma_dc_qmul[2]; ///< luma dc-only block quant
|
||||||
|
int16_t chroma_qmul[2];
|
||||||
|
} qmat[4];
|
||||||
|
|
||||||
|
struct {
|
||||||
|
uint8_t enabled; ///< whether each mb can have a different strength based on mode/ref
|
||||||
|
|
||||||
|
/**
|
||||||
|
* filter strength adjustment for the following macroblock modes:
|
||||||
|
* [0-3] - i16x16 (always zero)
|
||||||
|
* [4] - i4x4
|
||||||
|
* [5] - zero mv
|
||||||
|
* [6] - inter modes except for zero or split mv
|
||||||
|
* [7] - split mv
|
||||||
|
* i16x16 modes never have any adjustment
|
||||||
|
*/
|
||||||
|
int8_t mode[VP8_MVMODE_SPLIT+1];
|
||||||
|
|
||||||
|
/**
|
||||||
|
* filter strength adjustment for macroblocks that reference:
|
||||||
|
* [0] - intra / VP56_FRAME_CURRENT
|
||||||
|
* [1] - VP56_FRAME_PREVIOUS
|
||||||
|
* [2] - VP56_FRAME_GOLDEN
|
||||||
|
* [3] - altref / VP56_FRAME_GOLDEN2
|
||||||
|
*/
|
||||||
|
int8_t ref[4];
|
||||||
|
} lf_delta;
|
||||||
|
|
||||||
/**
|
/**
|
||||||
* Cache of the top row needed for intra prediction
|
* Cache of the top row needed for intra prediction
|
||||||
* 16 for luma, 8 for each chroma plane
|
* 16 for luma, 8 for each chroma plane
|
||||||
|
@ -151,74 +190,11 @@ typedef struct {
|
||||||
* 2+-> full transform
|
* 2+-> full transform
|
||||||
*/
|
*/
|
||||||
DECLARE_ALIGNED(16, uint8_t, non_zero_count_cache)[6][4];
|
DECLARE_ALIGNED(16, uint8_t, non_zero_count_cache)[6][4];
|
||||||
|
VP56RangeCoder c; ///< header context, includes mb modes and motion vectors
|
||||||
DECLARE_ALIGNED(16, DCTELEM, block)[6][4][16];
|
DECLARE_ALIGNED(16, DCTELEM, block)[6][4][16];
|
||||||
DECLARE_ALIGNED(16, DCTELEM, block_dc)[16];
|
DECLARE_ALIGNED(16, DCTELEM, block_dc)[16];
|
||||||
uint8_t intra4x4_pred_mode_mb[16];
|
uint8_t intra4x4_pred_mode_mb[16];
|
||||||
|
|
||||||
int chroma_pred_mode; ///< 8x8c pred mode of the current macroblock
|
|
||||||
int segment; ///< segment of the current macroblock
|
|
||||||
VP56mv mv_min;
|
|
||||||
VP56mv mv_max;
|
|
||||||
|
|
||||||
int mbskip_enabled;
|
|
||||||
int sign_bias[4]; ///< one state [0, 1] per ref frame type
|
|
||||||
int ref_count[3];
|
|
||||||
|
|
||||||
/**
|
|
||||||
* Base parameters for segmentation, i.e. per-macroblock parameters.
|
|
||||||
* These must be kept unchanged even if segmentation is not used for
|
|
||||||
* a frame, since the values persist between interframes.
|
|
||||||
*/
|
|
||||||
struct {
|
|
||||||
int enabled;
|
|
||||||
int absolute_vals;
|
|
||||||
int update_map;
|
|
||||||
int8_t base_quant[4];
|
|
||||||
int8_t filter_level[4]; ///< base loop filter level
|
|
||||||
} segmentation;
|
|
||||||
|
|
||||||
/**
|
|
||||||
* Macroblocks can have one of 4 different quants in a frame when
|
|
||||||
* segmentation is enabled.
|
|
||||||
* If segmentation is disabled, only the first segment's values are used.
|
|
||||||
*/
|
|
||||||
struct {
|
|
||||||
// [0] - DC qmul [1] - AC qmul
|
|
||||||
int16_t luma_qmul[2];
|
|
||||||
int16_t luma_dc_qmul[2]; ///< luma dc-only block quant
|
|
||||||
int16_t chroma_qmul[2];
|
|
||||||
} qmat[4];
|
|
||||||
|
|
||||||
struct {
|
|
||||||
int simple;
|
|
||||||
int level;
|
|
||||||
int sharpness;
|
|
||||||
} filter;
|
|
||||||
|
|
||||||
struct {
|
|
||||||
int enabled; ///< whether each mb can have a different strength based on mode/ref
|
|
||||||
|
|
||||||
/**
|
|
||||||
* filter strength adjustment for the following macroblock modes:
|
|
||||||
* [0-3] - i16x16 (always zero)
|
|
||||||
* [4] - i4x4
|
|
||||||
* [5] - zero mv
|
|
||||||
* [6] - inter modes except for zero or split mv
|
|
||||||
* [7] - split mv
|
|
||||||
* i16x16 modes never have any adjustment
|
|
||||||
*/
|
|
||||||
int8_t mode[VP8_MVMODE_SPLIT+1];
|
|
||||||
|
|
||||||
/**
|
|
||||||
* filter strength adjustment for macroblocks that reference:
|
|
||||||
* [0] - intra / VP56_FRAME_CURRENT
|
|
||||||
* [1] - VP56_FRAME_PREVIOUS
|
|
||||||
* [2] - VP56_FRAME_GOLDEN
|
|
||||||
* [3] - altref / VP56_FRAME_GOLDEN2
|
|
||||||
*/
|
|
||||||
int8_t ref[4];
|
|
||||||
} lf_delta;
|
|
||||||
|
|
||||||
/**
|
/**
|
||||||
* These are all of the updatable probabilities for binary decisions.
|
* These are all of the updatable probabilities for binary decisions.
|
||||||
* They are only implictly reset on keyframes, making it quite likely
|
* They are only implictly reset on keyframes, making it quite likely
|
||||||
|
@ -236,6 +212,30 @@ typedef struct {
|
||||||
uint8_t token[4][16][3][NUM_DCT_TOKENS-1];
|
uint8_t token[4][16][3][NUM_DCT_TOKENS-1];
|
||||||
uint8_t mvc[2][19];
|
uint8_t mvc[2][19];
|
||||||
} prob[2];
|
} prob[2];
|
||||||
|
|
||||||
|
VP8Macroblock *macroblocks_base;
|
||||||
|
int invisible;
|
||||||
|
int update_last; ///< update VP56_FRAME_PREVIOUS with the current one
|
||||||
|
int update_golden; ///< VP56_FRAME_NONE if not updated, or which frame to copy if so
|
||||||
|
int update_altref;
|
||||||
|
|
||||||
|
/**
|
||||||
|
* If this flag is not set, all the probability updates
|
||||||
|
* are discarded after this frame is decoded.
|
||||||
|
*/
|
||||||
|
int update_probabilities;
|
||||||
|
|
||||||
|
/**
|
||||||
|
* All coefficients are contained in separate arith coding contexts.
|
||||||
|
* There can be 1, 2, 4, or 8 of these after the header context.
|
||||||
|
*/
|
||||||
|
int num_coeff_partitions;
|
||||||
|
VP56RangeCoder coeff_partition[8];
|
||||||
|
DSPContext dsp;
|
||||||
|
VP8DSPContext vp8dsp;
|
||||||
|
H264PredContext hpc;
|
||||||
|
vp8_mc_func put_pixels_tab[3][3][3];
|
||||||
|
AVFrame frames[4];
|
||||||
} VP8Context;
|
} VP8Context;
|
||||||
|
|
||||||
#endif
|
#endif
|
||||||
|
|
Loading…
Reference in New Issue