/*****************************************************************************
* common.h: misc common functions
*****************************************************************************
- * Copyright (C) 2003-2011 x264 project
+ * Copyright (C) 2003-2012 x264 project
*
* Authors: Laurent Aimar <fenrir@via.ecp.fr>
* Loren Merritt <lorenm@u.washington.edu>
#define X264_BFRAME_MAX 16
#define X264_REF_MAX 16
#define X264_THREAD_MAX 128
-#define X264_PCM_COST ((384<<CHROMA444)*BIT_DEPTH+16)
+#define X264_LOOKAHEAD_THREAD_MAX 16
+#define X264_PCM_COST (FRAME_SIZE(256*BIT_DEPTH)+16)
#define X264_LOOKAHEAD_MAX 250
#define QP_BD_OFFSET (6*(BIT_DEPTH-8))
#define QP_MAX_SPEC (51+QP_BD_OFFSET)
# define PARAM_INTERLACED 0
#endif
-#define CHROMA444 (h->sps->i_chroma_format_idc == 3)
+#ifdef CHROMA_FORMAT
+# define CHROMA_H_SHIFT (CHROMA_FORMAT == CHROMA_420 || CHROMA_FORMAT == CHROMA_422)
+# define CHROMA_V_SHIFT (CHROMA_FORMAT == CHROMA_420)
+#else
+# define CHROMA_FORMAT h->sps->i_chroma_format_idc
+# define CHROMA_H_SHIFT h->mb.chroma_h_shift
+# define CHROMA_V_SHIFT h->mb.chroma_v_shift
+#endif
+
+#define CHROMA_SIZE(s) ((s)>>(CHROMA_H_SHIFT+CHROMA_V_SHIFT))
+#define FRAME_SIZE(s) ((s)+2*CHROMA_SIZE(s))
+#define CHROMA444 (CHROMA_FORMAT == CHROMA_444)
/* Unions for type-punning.
* Mn: load or store n bits, aligned, native-endian
void x264_reduce_fraction( uint32_t *n, uint32_t *d );
void x264_reduce_fraction64( uint64_t *n, uint64_t *d );
-void x264_cavlc_init( void );
+void x264_cavlc_init( x264_t *h );
void x264_cabac_init( x264_t *h );
static ALWAYS_INLINE pixel x264_clip_pixel( int x )
} ref_pic_list_order[2][X264_REF_MAX];
/* P-frame weighting */
+ int b_weighted_pred;
x264_weight_t weight[X264_REF_MAX*2][3];
int i_mmco_remove_from_end;
uint8_t ref[4];
} x264_left_table_t;
+/* Current frame stats */
+typedef struct
+{
+ /* 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[19];
+ int i_mb_count_i;
+ int i_mb_count_p;
+ int i_mb_count_skip;
+ int i_mb_count_8x8dct[2];
+ int i_mb_count_ref[2][X264_REF_MAX*2];
+ int i_mb_partition[17];
+ int i_mb_cbp[6];
+ int i_mb_pred_mode[4][13];
+ int i_mb_field[3];
+ /* Adaptive direct mv pred */
+ int i_direct_score[2];
+ /* Metrics */
+ int64_t i_ssd[3];
+ double f_ssim;
+ int i_ssim_cnt;
+} x264_frame_stat_t;
+
struct x264_t
{
/* encoder parameters */
x264_param_t param;
x264_t *thread[X264_THREAD_MAX+1];
+ x264_t *lookahead_thread[X264_LOOKAHEAD_THREAD_MAX];
int b_thread_active;
int i_thread_phase; /* which thread to use for the next frame */
+ int i_thread_idx; /* which thread this is */
int i_threadslice_start; /* first row in this thread slice */
int i_threadslice_end; /* row after the end of this thread slice */
+ int i_threadslice_pass; /* which pass of encoding we are on */
x264_threadpool_t *threadpool;
+ x264_threadpool_t *lookaheadpool;
+ x264_pthread_mutex_t mutex;
+ x264_pthread_cond_t cv;
/* bitstream output */
struct
int64_t i_cpb_delay_lookahead;
int64_t i_cpb_delay_pir_offset;
+ int64_t i_cpb_delay_pir_offset_next;
int b_queued_intra_refresh;
int64_t i_last_idr_pts;
udctcoef (*quant8_mf[4])[64]; /* [4][52][64] */
udctcoef (*quant4_bias[4])[16]; /* [4][52][16] */
udctcoef (*quant8_bias[4])[64]; /* [4][52][64] */
+ udctcoef (*quant4_bias0[4])[16]; /* [4][52][16] */
+ udctcoef (*quant8_bias0[4])[64]; /* [4][52][64] */
udctcoef (*nr_offset_emergency)[4][64];
/* mv/ref cost arrays. */
struct
{
ALIGNED_16( dctcoef luma16x16_dc[3][16] );
- ALIGNED_16( dctcoef chroma_dc[2][4] );
+ ALIGNED_16( dctcoef chroma_dc[2][8] );
// FIXME share memory?
ALIGNED_16( dctcoef luma8x8[12][64] );
ALIGNED_16( dctcoef luma4x4[16*3][16] );
int i_mb_height;
int i_mb_count; /* number of mbs in a frame */
+ /* Chroma subsampling */
+ int chroma_h_shift;
+ int chroma_v_shift;
+
/* Strides */
int i_mb_stride;
int i_b8_stride;
/* extra data required for mbaff in mv prediction */
int16_t topright_mv[2][3][2];
int8_t topright_ref[2][3];
+
+ /* current mb deblock strength */
+ uint8_t (*deblock_strength)[8][4];
} cache;
/* */
struct
{
/* Current frame stats */
- struct
- {
- /* 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[19];
- int i_mb_count_i;
- int i_mb_count_p;
- int i_mb_count_skip;
- int i_mb_count_8x8dct[2];
- int i_mb_count_ref[2][X264_REF_MAX*2];
- int i_mb_partition[17];
- int i_mb_cbp[6];
- int i_mb_pred_mode[4][13];
- int i_mb_field[3];
- /* Adaptive direct mv pred */
- int i_direct_score[2];
- /* Metrics */
- int64_t i_ssd[3];
- double f_ssim;
- int i_ssim_cnt;
- } frame;
+ x264_frame_stat_t frame;
/* Cumulated stats */
ALIGNED_16( uint32_t nr_residual_sum_buf[2][4][64] );
uint32_t nr_count_buf[2][4];
+ uint8_t luma2chroma_pixel[7]; /* Subsampled pixel size */
+
/* Buffers that are allocated per-thread even in sliced threads. */
void *scratch_buffer; /* for any temporary storage that doesn't want repeated malloc */
+ void *scratch_buffer2; /* if the first one's already in use */
pixel *intra_border_backup[5][3]; /* bottom pixels of the previous mb row, used for intra prediction after the framebuffer has been deblocked */
/* Deblock strength values are stored for each 4x4 partition. In MBAFF
* there are four extra values that need to be stored, located in [4][i]. */
/* 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_t predict_chroma[4+3];
+ x264_predict_t predict_8x8c[4+3];
+ x264_predict_t predict_8x16c[4+3];
x264_predict_8x8_filter_t predict_8x8_filter;
x264_pixel_function_t pixf;