void x264_log( x264_t *h, int i_level, const char *psz_fmt, ... );
void x264_reduce_fraction( int *n, int *d );
+void x264_init_vlc_tables();
static inline uint8_t x264_clip_uint8( int x )
{
#define X264_SCAN8_SIZE (6*8)
#define X264_SCAN8_0 (4+1*8)
-static const int x264_scan8[16+2*4] =
+static const int x264_scan8[16+2*4+3] =
{
/* Luma */
4+1*8, 5+1*8, 4+2*8, 5+2*8,
/* Cr */
1+4*8, 2+4*8,
1+5*8, 2+5*8,
+
+ /* Luma DC */
+ 4+5*8,
+
+ /* Chroma DC */
+ 5+5*8, 6+5*8
};
/*
0 1 2 3 4 5 6 7
2 B B L L L L
3 L L L L
4 R R L L L L
- 5 R R
+ 5 R R DyDuDv
*/
typedef struct x264_ratecontrol_t x264_ratecontrol_t;
int i_frame_size;
} out;
+ /**** thread synchronization starts here ****/
+
/* frame number/poc */
int i_frame;
uint16_t (*quant4_bias[4])[16]; /* [4][52][16] */
uint16_t (*quant8_bias[2])[64]; /* [2][52][64] */
+ const uint8_t *chroma_qp_table; /* includes both the nonlinear luma->chroma mapping and chroma_qp_offset */
+
DECLARE_ALIGNED_16( uint32_t nr_residual_sum[2][64] );
DECLARE_ALIGNED_16( uint16_t nr_offset[2][64] );
uint32_t nr_count[2];
struct
{
/* Frames to be encoded (whose types have been decided) */
- x264_frame_t *current[X264_BFRAME_MAX+3];
+ x264_frame_t *current[X264_BFRAME_MAX*4+3];
/* Temporary buffer (frames types not yet decided) */
- x264_frame_t *next[X264_BFRAME_MAX+3];
+ x264_frame_t *next[X264_BFRAME_MAX*4+3];
/* Unused frames */
- x264_frame_t *unused[X264_BFRAME_MAX + X264_THREAD_MAX*2 + 16+4];
+ x264_frame_t *unused[X264_BFRAME_MAX*4 + X264_THREAD_MAX*2 + 16+4];
/* For adaptive B decision */
x264_frame_t *last_nonb;
int i_max_ref1;
int i_delay; /* Number of frames buffered for B reordering */
int b_have_lowres; /* Whether 1/2 resolution luma planes are being used */
+ int b_have_sub8x8_esa;
} frames;
/* current frame being encoded */
int i_mb_xy;
int i_b8_xy;
int i_b4_xy;
-
+
/* Search parameters */
int i_me_method;
int i_subpel_refine;
int b_chroma_me;
int b_trellis;
int b_noise_reduction;
+ int i_psy_rd; /* Psy RD strength--fixed point value*/
+ int i_psy_trellis; /* Psy trellis strength--fixed point value*/
int b_interlaced;
unsigned int i_neighbour;
unsigned int i_neighbour8[4]; /* neighbours of each 8x8 or 4x4 block that are available */
unsigned int i_neighbour4[16]; /* at the time the block is coded */
- int i_mb_type_top;
- int i_mb_type_left;
- int i_mb_type_topleft;
- int i_mb_type_topright;
+ int i_mb_type_top;
+ int i_mb_type_left;
+ int i_mb_type_topleft;
+ int i_mb_type_topright;
int i_mb_prev_xy;
int i_mb_top_xy;
+ /**** thread synchronization ends here ****/
+ /* subsequent variables are either thread-local or constant,
+ * and won't be copied from one thread to another */
+
/* mb table */
int8_t *type; /* mb type */
int8_t *qp; /* mb qp */
/* current value */
int i_type;
int i_partition;
- int i_sub_partition[4];
+ DECLARE_ALIGNED_4( uint8_t i_sub_partition[4] );
int b_transform_8x8;
int i_cbp_luma;
DECLARE_ALIGNED_16( uint8_t fenc_buf[24*FENC_STRIDE] );
DECLARE_ALIGNED_16( uint8_t fdec_buf[27*FDEC_STRIDE] );
- /* i4x4 and i8x8 backup data, for skipping the encode stage when possible */
+ /* i4x4 and i8x8 backup data, for skipping the encode stage when possible */
DECLARE_ALIGNED_16( uint8_t i4x4_fdec_buf[16*16] );
DECLARE_ALIGNED_16( uint8_t i8x8_fdec_buf[16*16] );
DECLARE_ALIGNED_16( int16_t i8x8_dct_buf[3][64] );
DECLARE_ALIGNED_16( int16_t i4x4_dct_buf[15][16] );
+ uint32_t i4x4_nnz_buf[4];
+ uint32_t i8x8_nnz_buf[4];
+ int i4x4_cbp;
+ int i8x8_cbp;
+
+ /* Psy trellis DCT data */
+ DECLARE_ALIGNED_16( int16_t fenc_dct8[4][64] );
+ DECLARE_ALIGNED_16( int16_t fenc_dct4[16][16] );
+
+ /* Psy RD SATD scores */
+ int fenc_satd[4][4];
+ int fenc_satd_sum;
+ int fenc_sa8d[2][2];
+ int fenc_sa8d_sum;
/* pointer over mb of the frame to be compressed */
uint8_t *p_fenc[3];
+ /* pointer to the actual source frame, not a block copy */
+ uint8_t *p_fenc_plane[3];
/* pointer over mb of the frame to be reconstructed */
uint8_t *p_fdec[3];
/* number of neighbors (top and left) that used 8x8 dct */
int i_neighbour_transform_size;
int i_neighbour_interlaced;
+
+ /* neighbor CBPs */
+ int i_cbp_top;
+ int i_cbp_left;
} cache;
/* */
/* Current frame stats */
struct
{
- /* Headers bits (MV+Ref+MB Block Type */
- int i_hdr_bits;
- /* Texture bits (Intra/Predicted) */
- int i_itex_bits;
- int i_ptex_bits;
+ /* MV bits (MV+Ref+Block Type) */
+ int i_mv_bits;
+ /* Texture bits (DCT coefs) */
+ int i_tex_bits;
/* ? */
int i_misc_bits;
/* MB type counts */
int i_mb_count_p;
int i_mb_count_skip;
int i_mb_count_8x8dct[2];
- int i_mb_count_size[7];
- int i_mb_count_ref[32];
- /* Estimated (SATD) cost as Intra/Predicted frame */
- /* XXX: both omit the cost of MBs coded as P_SKIP */
- int i_intra_cost;
- int i_inter_cost;
- int i_mbs_analysed;
+ int i_mb_count_ref[2][32];
+ int i_mb_partition[17];
+ int i_mb_cbp[6];
/* Adaptive direct mv pred */
int i_direct_score[2];
/* Metrics */
int i_slice_count[5];
int64_t i_slice_size[5];
double f_slice_qp[5];
+ int i_consecutive_bframes[X264_BFRAME_MAX+1];
/* */
- int64_t i_sqe_global[5];
+ int64_t i_ssd_global[5];
double f_psnr_average[5];
double f_psnr_mean_y[5];
double f_psnr_mean_u[5];
double f_ssim_mean_y[5];
/* */
int64_t i_mb_count[5][19];
+ int64_t i_mb_partition[2][17];
int64_t i_mb_count_8x8dct[2];
- int64_t i_mb_count_size[2][7];
- int64_t i_mb_count_ref[2][32];
+ int64_t i_mb_count_ref[2][2][32];
+ int64_t i_mb_cbp[6];
/* */
int i_direct_score[2];
int i_direct_frames[2];
} stat;
+ void *scratch_buffer; /* for any temporary storage that doesn't want repeated malloc */
+
/* CPU functions dependents */
x264_predict_t predict_16x16[4+3];
x264_predict_t predict_8x8c[4+3];
x264_predict8x8_t predict_8x8[9+3];
x264_predict_t predict_4x4[9+3];
+ x264_predict_8x8_filter_t predict_8x8_filter;
x264_pixel_function_t pixf;
x264_mc_functions_t mc;