X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=common%2Fcommon.h;h=64947ed1dc4cdb03061e5ae3f313faea0c1d38e3;hb=94b9141609d17ebbeb3184a8a5fc0660725a4cf2;hp=e75f743a13dae0eb73eaa3fce0378415731b3289;hpb=4e105e079314b2fe04742d5605ffb0d961c16813;p=x264 diff --git a/common/common.h b/common/common.h index e75f743a..64947ed1 100644 --- a/common/common.h +++ b/common/common.h @@ -1,7 +1,7 @@ /***************************************************************************** - * common.h: h264 encoder + * common.h: misc common functions ***************************************************************************** - * Copyright (C) 2003-2008 x264 project + * Copyright (C) 2003-2011 x264 project * * Authors: Laurent Aimar * Loren Merritt @@ -19,6 +19,9 @@ * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA. + * + * This program is also available under a commercial proprietary license. + * For more information, contact us at licensing@x264.com. *****************************************************************************/ #ifndef X264_COMMON_H @@ -36,6 +39,7 @@ #define XCHG(type,a,b) do{ type t = a; a = b; b = t; } while(0) #define IS_DISPOSABLE(type) ( type == X264_TYPE_B ) #define FIX8(f) ((int)(f*(1<<8)+.5)) +#define ALIGN(x,a) (((x)+((a)-1))&~((a)-1)) #define CHECKED_MALLOC( var, size )\ do {\ @@ -50,11 +54,18 @@ do {\ } while( 0 ) #define X264_BFRAME_MAX 16 +#define X264_REF_MAX 16 #define X264_THREAD_MAX 128 -#define X264_PCM_COST (386*8) +#define X264_PCM_COST (384*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 QP_MAX (QP_MAX_SPEC+18) +#define QP_MAX_MAX (51+2*6+18) +#define PIXEL_MAX ((1 << BIT_DEPTH)-1) // arbitrary, but low because SATD scores are 1/4 normal -#define X264_LOOKAHEAD_QP 12 +#define X264_LOOKAHEAD_QP (12+QP_BD_OFFSET) +#define SPEC_QP(x) X264_MIN((x), QP_MAX_SPEC) // number of pixels (per thread) in progress at any given time. // 16 for the macroblock in progress + 3 for deblocking + 3 for motion compensation filter + 2 for extra safety @@ -100,8 +111,66 @@ typedef union { x264_uint128_t i; uint64_t a[2]; uint32_t b[4]; uint16_t c[8]; u #define CP64(dst,src) M64(dst) = M64(src) #define CP128(dst,src) M128(dst) = M128(src) +#if HIGH_BIT_DEPTH + typedef uint16_t pixel; + typedef uint64_t pixel4; + typedef int32_t dctcoef; + typedef uint32_t udctcoef; + +# define PIXEL_SPLAT_X4(x) ((x)*0x0001000100010001ULL) +# define MPIXEL_X4(src) M64(src) +#else + typedef uint8_t pixel; + typedef uint32_t pixel4; + typedef int16_t dctcoef; + typedef uint16_t udctcoef; + +# define PIXEL_SPLAT_X4(x) ((x)*0x01010101U) +# define MPIXEL_X4(src) M32(src) +#endif + +#define BIT_DEPTH X264_BIT_DEPTH + +#define CPPIXEL_X4(dst,src) MPIXEL_X4(dst) = MPIXEL_X4(src) + +#define X264_SCAN8_SIZE (6*8) +#define X264_SCAN8_LUMA_SIZE (5*8) +#define X264_SCAN8_0 (4+1*8) + +static const unsigned x264_scan8[16+2*4+3] = +{ + /* Luma */ + 4+1*8, 5+1*8, 4+2*8, 5+2*8, + 6+1*8, 7+1*8, 6+2*8, 7+2*8, + 4+3*8, 5+3*8, 4+4*8, 5+4*8, + 6+3*8, 7+3*8, 6+4*8, 7+4*8, + + /* Cb */ + 1+1*8, 2+1*8, + 1+2*8, 2+2*8, + + /* Cr */ + 1+4*8, 2+4*8, + 1+5*8, 2+5*8, + + /* Luma DC */ + 4+5*8, + + /* Chroma DC */ + 6+5*8, 7+5*8 +}; +/* + 0 1 2 3 4 5 6 7 + 0 + 1 B B L L L L + 2 B B L L L L + 3 L L L L + 4 R R L L L L + 5 R R Dy DuDv +*/ + #include "x264.h" -#include "bs.h" +#include "bitstream.h" #include "set.h" #include "predict.h" #include "pixel.h" @@ -110,6 +179,8 @@ typedef union { x264_uint128_t i; uint64_t a[2]; uint32_t b[4]; uint16_t c[8]; u #include "dct.h" #include "cabac.h" #include "quant.h" +#include "cpu.h" +#include "threadpool.h" /**************************************************************************** * General functions @@ -129,17 +200,17 @@ int64_t x264_mdate( void ); * the encoding options */ char *x264_param2string( x264_param_t *p, int b_res ); -int x264_nal_encode( uint8_t *dst, x264_nal_t *nal, int b_annexb, int b_long_startcode ); - /* log */ void x264_log( x264_t *h, int i_level, const char *psz_fmt, ... ); void x264_reduce_fraction( uint32_t *n, uint32_t *d ); -void x264_init_vlc_tables(); +void x264_reduce_fraction64( uint64_t *n, uint64_t *d ); +void x264_cavlc_init( void ); +void x264_cabac_init( void ); -static ALWAYS_INLINE uint8_t x264_clip_uint8( int x ) +static ALWAYS_INLINE pixel x264_clip_pixel( int x ) { - return x&(~255) ? (-x)>>31 : x; + return ( (x & ~PIXEL_MAX) ? (-x)>>31 & PIXEL_MAX : x ); } static ALWAYS_INLINE int x264_clip3( int v, int i_min, int i_max ) @@ -188,14 +259,14 @@ static ALWAYS_INLINE uint16_t x264_cabac_mvd_sum( uint8_t *mvdleft, uint8_t *mvd return amvd0 + (amvd1<<8); } -static void ALWAYS_INLINE x264_predictor_roundclip( int16_t (*mvc)[2], int i_mvc, int mv_x_min, int mv_x_max, int mv_y_min, int mv_y_max ) +static void ALWAYS_INLINE x264_predictor_roundclip( int16_t (*dst)[2], int16_t (*mvc)[2], int i_mvc, int mv_x_min, int mv_x_max, int mv_y_min, int mv_y_max ) { for( int i = 0; i < i_mvc; i++ ) { int mx = (mvc[i][0] + 2) >> 2; int my = (mvc[i][1] + 2) >> 2; - mvc[i][0] = x264_clip3( mx, mv_x_min, mv_x_max ); - mvc[i][0] = x264_clip3( my, mv_y_min, mv_y_max ); + dst[i][0] = x264_clip3( mx, mv_x_min, mv_x_max ); + dst[i][1] = x264_clip3( my, mv_y_min, mv_y_max ); } } @@ -227,11 +298,9 @@ enum slice_type_e SLICE_TYPE_P = 0, SLICE_TYPE_B = 1, SLICE_TYPE_I = 2, - SLICE_TYPE_SP = 3, - SLICE_TYPE_SI = 4 }; -static const char slice_type_to_char[] = { 'P', 'B', 'I', 'S', 'S' }; +static const char slice_type_to_char[] = { 'P', 'B', 'I' }; enum sei_payload_type_e { @@ -242,6 +311,8 @@ enum sei_payload_type_e SEI_USER_DATA_REGISTERED = 4, SEI_USER_DATA_UNREGISTERED = 5, SEI_RECOVERY_POINT = 6, + SEI_DEC_REF_PIC_MARKING = 7, + SEI_FRAME_PACKING = 45, }; typedef struct @@ -275,16 +346,15 @@ typedef struct int i_num_ref_idx_l0_active; int i_num_ref_idx_l1_active; - int b_ref_pic_list_reordering_l0; - int b_ref_pic_list_reordering_l1; + int b_ref_pic_list_reordering[2]; struct { int idc; int arg; - } ref_pic_list_order[2][16]; + } ref_pic_list_order[2][X264_REF_MAX]; /* P-frame weighting */ - x264_weight_t weight[32][3]; + x264_weight_t weight[X264_REF_MAX*2][3]; int i_mmco_remove_from_end; int i_mmco_command_count; @@ -292,7 +362,7 @@ typedef struct { int i_difference_of_pic_nums; int i_poc; - } mmco[16]; + } mmco[X264_REF_MAX]; int i_cabac_init_idc; @@ -316,50 +386,22 @@ typedef struct x264_lookahead_t int i_last_keyframe; int i_slicetype_length; x264_frame_t *last_nonb; - x264_synch_frame_list_t ifbuf; - x264_synch_frame_list_t next; - x264_synch_frame_list_t ofbuf; + x264_pthread_t thread_handle; + x264_sync_frame_list_t ifbuf; + x264_sync_frame_list_t next; + x264_sync_frame_list_t ofbuf; } x264_lookahead_t; -/* From ffmpeg - */ -#define X264_SCAN8_SIZE (6*8) -#define X264_SCAN8_LUMA_SIZE (5*8) -#define X264_SCAN8_0 (4+1*8) +typedef struct x264_ratecontrol_t x264_ratecontrol_t; -static const int x264_scan8[16+2*4+3] = +typedef struct x264_left_table_t { - /* Luma */ - 4+1*8, 5+1*8, 4+2*8, 5+2*8, - 6+1*8, 7+1*8, 6+2*8, 7+2*8, - 4+3*8, 5+3*8, 4+4*8, 5+4*8, - 6+3*8, 7+3*8, 6+4*8, 7+4*8, - - /* Cb */ - 1+1*8, 2+1*8, - 1+2*8, 2+2*8, - - /* Cr */ - 1+4*8, 2+4*8, - 1+5*8, 2+5*8, - - /* Luma DC */ - 4+5*8, - - /* Chroma DC */ - 6+5*8, 7+5*8 -}; -/* - 0 1 2 3 4 5 6 7 - 0 - 1 B B L L L L - 2 B B L L L L - 3 L L L L - 4 R R L L L L - 5 R R Dy DuDv -*/ - -typedef struct x264_ratecontrol_t x264_ratecontrol_t; + uint8_t intra[4]; + uint8_t nnz[4]; + uint8_t nnz_chroma[4]; + uint8_t mv[4]; + uint8_t ref[4]; +} x264_left_table_t; struct x264_t { @@ -367,11 +409,11 @@ struct x264_t x264_param_t param; x264_t *thread[X264_THREAD_MAX+1]; - x264_pthread_t thread_handle; int b_thread_active; int i_thread_phase; /* which thread to use for the next frame */ int i_threadslice_start; /* first row in this thread slice */ int i_threadslice_end; /* row after the end of this thread slice */ + x264_threadpool_t *threadpool; /* bitstream output */ struct @@ -387,6 +429,9 @@ struct x264_t uint8_t *nal_buffer; int nal_buffer_size; + x264_sps_t *sps; + x264_pps_t *pps; + /**** thread synchronization starts here ****/ /* frame number/poc */ @@ -398,25 +443,25 @@ struct x264_t int i_nal_type; int i_nal_ref_idc; - int i_disp_fields; /* Number of displayed fields (both coded and implied via pic_struct) */ + int64_t i_disp_fields; /* Number of displayed fields (both coded and implied via pic_struct) */ int i_disp_fields_last_frame; - int i_prev_duration; /* Duration of previous frame */ - int i_coded_fields; /* Number of coded fields (both coded and implied via pic_struct) */ - int i_cpb_delay; /* Equal to number of fields preceding this field + int64_t i_prev_duration; /* Duration of previous frame */ + int64_t i_coded_fields; /* Number of coded fields (both coded and implied via pic_struct) */ + int64_t i_cpb_delay; /* Equal to number of fields preceding this field * since last buffering_period SEI */ - int i_coded_fields_lookahead; /* Use separate counters for lookahead */ - int i_cpb_delay_lookahead; + int64_t i_coded_fields_lookahead; /* Use separate counters for lookahead */ + int64_t i_cpb_delay_lookahead; + + int64_t i_cpb_delay_pir_offset; + + int b_queued_intra_refresh; + int64_t i_last_idr_pts; /* We use only one SPS and one PPS */ x264_sps_t sps_array[1]; - x264_sps_t *sps; x264_pps_t pps_array[1]; - x264_pps_t *pps; int i_idr_pic_id; - /* Timebase multiplier for DTS compression */ - int i_dts_compress_multiplier; - /* quantization matrix for decoding, [cqm][qp%6][coef] */ int (*dequant4_mf[4])[16]; /* [4][6][16] */ int (*dequant8_mf[2])[64]; /* [2][6][64] */ @@ -424,22 +469,25 @@ struct x264_t int (*unquant4_mf[4])[16]; /* [4][52][16] */ int (*unquant8_mf[2])[64]; /* [2][52][64] */ /* quantization matrix for deadzone */ - uint16_t (*quant4_mf[4])[16]; /* [4][52][16] */ - uint16_t (*quant8_mf[2])[64]; /* [2][52][64] */ - uint16_t (*quant4_bias[4])[16]; /* [4][52][16] */ - uint16_t (*quant8_bias[2])[64]; /* [2][52][64] */ + udctcoef (*quant4_mf[4])[16]; /* [4][52][16] */ + udctcoef (*quant8_mf[2])[64]; /* [2][52][64] */ + udctcoef (*quant4_bias[4])[16]; /* [4][52][16] */ + udctcoef (*quant8_bias[2])[64]; /* [2][52][64] */ + udctcoef (*nr_offset_emergency)[3][64]; - /* mv/ref cost arrays. Indexed by lambda instead of - * qp because, due to rounding, some quantizers share - * lambdas. This saves memory. */ - uint16_t *cost_mv[92]; - uint16_t *cost_mv_fpel[92][4]; + /* mv/ref cost arrays. */ + uint16_t *cost_mv[QP_MAX+1]; + uint16_t *cost_mv_fpel[QP_MAX+1][4]; const uint8_t *chroma_qp_table; /* includes both the nonlinear luma->chroma mapping and chroma_qp_offset */ /* Slice header */ x264_slice_header_t sh; + /* Slice header backup, for SEI_DEC_REF_PIC_MARKING */ + int b_sh_backup; + x264_slice_header_t sh_backup; + /* cabac context */ x264_cabac_t cabac; @@ -454,9 +502,13 @@ struct x264_t x264_frame_t **blank_unused; /* frames used for reference + sentinels */ - x264_frame_t *reference[16+2]; + x264_frame_t *reference[X264_REF_MAX+2]; - int i_last_keyframe; /* Frame number of the last keyframe */ + int i_last_keyframe; /* Frame number of the last keyframe */ + int i_last_idr; /* Frame number of the last IDR (not RP)*/ + int i_poc_last_open_gop; /* Poc of the I frame of the last open-gop. The value + * is only assigned during the period between that + * I frame and the next P or I frame, else -1 */ int i_input; /* Number of input frames already accepted */ @@ -466,7 +518,7 @@ struct x264_t int i_delay; /* Number of frames buffered for B reordering */ int i_bframe_delay; int64_t i_bframe_delay_time; - int64_t i_init_delta; + int64_t i_first_pts; int64_t i_prev_reordered_pts[2]; int64_t i_largest_pts; int64_t i_second_largest_pts; @@ -481,36 +533,39 @@ struct x264_t x264_frame_t *fdec; /* references lists */ - int i_ref0; - x264_frame_t *fref0[16+3]; /* ref list 0 */ - int i_ref1; - x264_frame_t *fref1[16+3]; /* ref list 1 */ + int i_ref[2]; + x264_frame_t *fref[2][X264_REF_MAX+3]; + x264_frame_t *fref_nearest[2]; int b_ref_reorder[2]; /* hrd */ int initial_cpb_removal_delay; int initial_cpb_removal_delay_offset; - int64_t first_pts; + int64_t i_reordered_pts_delay; /* Current MB DCT coeffs */ struct { - ALIGNED_16( int16_t luma16x16_dc[16] ); - ALIGNED_16( int16_t chroma_dc[2][4] ); + ALIGNED_16( dctcoef luma16x16_dc[16] ); + ALIGNED_16( dctcoef chroma_dc[2][4] ); // FIXME share memory? - ALIGNED_16( int16_t luma8x8[4][64] ); - ALIGNED_16( int16_t luma4x4[16+8][16] ); + ALIGNED_16( dctcoef luma8x8[4][64] ); + ALIGNED_16( dctcoef luma4x4[16+8][16] ); } dct; /* MB table and cache for current frame/mb */ struct { + int i_mb_width; + int i_mb_height; int i_mb_count; /* number of mbs in a frame */ /* Strides */ int i_mb_stride; int i_b8_stride; int i_b4_stride; + int left_b8[2]; + int left_b4[2]; /* Current index */ int i_mb_x; @@ -530,17 +585,24 @@ struct x264_t int i_psy_trellis; /* Psy trellis strength--fixed point value*/ int b_interlaced; + int b_adaptive_mbaff; /* MBAFF+subme 0 requires non-adaptive MBAFF i.e. all field mbs */ /* Allowed qpel MV range to stay within the picture + emulated edge pixels */ int mv_min[2]; int mv_max[2]; + int mv_miny_row[3]; /* 0 == top progressive, 1 == bot progressive, 2 == interlaced */ + int mv_maxy_row[3]; /* Subpel MV range for motion search. * same mv_min/max but includes levels' i_mv_range. */ int mv_min_spel[2]; int mv_max_spel[2]; + int mv_miny_spel_row[3]; + int mv_maxy_spel_row[3]; /* Fullpel MV range for motion search */ int mv_min_fpel[2]; int mv_max_fpel[2]; + int mv_miny_fpel_row[3]; + int mv_maxy_fpel_row[3]; /* neighboring MBs */ unsigned int i_neighbour; @@ -549,14 +611,22 @@ struct x264_t unsigned int i_neighbour_intra; /* for constrained intra pred */ unsigned int i_neighbour_frame; /* ignoring slice boundaries */ int i_mb_type_top; - int i_mb_type_left; + int i_mb_type_left[2]; int i_mb_type_topleft; int i_mb_type_topright; int i_mb_prev_xy; - int i_mb_left_xy; + int i_mb_left_xy[2]; int i_mb_top_xy; int i_mb_topleft_xy; int i_mb_topright_xy; + int i_mb_top_y; + int i_mb_topleft_y; + int i_mb_topright_y; + x264_left_table_t *left_index_table; + int i_mb_top_mbpair_xy; + int topleft_partition; + int b_allow_skip; + int field_decoding_flag; /**** thread synchronization ends here ****/ /* subsequent variables are either thread-local or constant, @@ -574,14 +644,15 @@ struct x264_t int16_t (*mv[2])[2]; /* mb mv. set to 0 for intra mb */ uint8_t (*mvd[2])[8][2]; /* absolute value of mb mv difference with predict, clipped to [0,33]. set to 0 if intra. cabac only */ int8_t *ref[2]; /* mb ref. set to -1 if non used (intra or Lx only) */ - int16_t (*mvr[2][32])[2]; /* 16x16 mv for each possible ref */ + int16_t (*mvr[2][X264_REF_MAX*2])[2];/* 16x16 mv for each possible ref */ int8_t *skipbp; /* block pattern for SKIP or DIRECT (sub)mbs. B-frames + cabac only */ int8_t *mb_transform_size; /* transform_size_8x8_flag of each mb */ uint16_t *slice_table; /* sh->first_mb of the slice that the indexed mb is part of * NOTE: this will fail on resolutions above 2^16 MBs... */ + uint8_t *field; /* buffer for weighted versions of the reference frames */ - uint8_t *p_weight_buf[16]; + pixel *p_weight_buf[X264_REF_MAX]; /* current value */ int i_type; @@ -606,46 +677,48 @@ struct x264_t /* set to true if we are re-encoding a macroblock. */ int b_reencode_mb; int ip_offset; /* Used by PIR to offset the quantizer of intra-refresh blocks. */ + int b_deblock_rdo; + int b_overflow; /* If CAVLC had a level code overflow during bitstream writing. */ struct { /* space for p_fenc and p_fdec */ #define FENC_STRIDE 16 #define FDEC_STRIDE 32 - ALIGNED_16( uint8_t fenc_buf[24*FENC_STRIDE] ); - ALIGNED_16( uint8_t fdec_buf[27*FDEC_STRIDE] ); + ALIGNED_16( pixel fenc_buf[24*FENC_STRIDE] ); + ALIGNED_16( pixel fdec_buf[27*FDEC_STRIDE] ); /* i4x4 and i8x8 backup data, for skipping the encode stage when possible */ - ALIGNED_16( uint8_t i4x4_fdec_buf[16*16] ); - ALIGNED_16( uint8_t i8x8_fdec_buf[16*16] ); - ALIGNED_16( int16_t i8x8_dct_buf[3][64] ); - ALIGNED_16( int16_t i4x4_dct_buf[15][16] ); + ALIGNED_16( pixel i4x4_fdec_buf[16*16] ); + ALIGNED_16( pixel i8x8_fdec_buf[16*16] ); + ALIGNED_16( dctcoef i8x8_dct_buf[3][64] ); + ALIGNED_16( dctcoef 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 */ - ALIGNED_16( int16_t fenc_dct8[4][64] ); - ALIGNED_16( int16_t fenc_dct4[16][16] ); + ALIGNED_16( dctcoef fenc_dct8[4][64] ); + ALIGNED_16( dctcoef fenc_dct4[16][16] ); /* Psy RD SATD/SA8D scores cache */ ALIGNED_16( uint64_t fenc_hadamard_cache[9] ); ALIGNED_16( uint32_t fenc_satd_cache[32] ); /* pointer over mb of the frame to be compressed */ - uint8_t *p_fenc[3]; + pixel *p_fenc[3]; /* y,u,v */ /* pointer to the actual source frame, not a block copy */ - uint8_t *p_fenc_plane[3]; + pixel *p_fenc_plane[2]; /* y,uv */ /* pointer over mb of the frame to be reconstructed */ - uint8_t *p_fdec[3]; + pixel *p_fdec[3]; /* pointer over mb of the references */ int i_fref[2]; - uint8_t *p_fref[2][32][4+2]; /* last: lN, lH, lV, lHV, cU, cV */ - uint8_t *p_fref_w[32]; /* weighted fullpel luma */ - uint16_t *p_integral[2][16]; + pixel *p_fref[2][X264_REF_MAX*2][4+1]; /* last: yN, yH, yV, yHV, uv */ + pixel *p_fref_w[X264_REF_MAX*2]; /* weighted fullpel luma */ + uint16_t *p_integral[2][X264_REF_MAX]; /* fref stride */ int i_stride[3]; @@ -658,7 +731,7 @@ struct x264_t ALIGNED_8( int8_t intra4x4_pred_mode[X264_SCAN8_LUMA_SIZE] ); /* i_non_zero_count if available else 0x80 */ - ALIGNED_4( uint8_t non_zero_count[X264_SCAN8_SIZE] ); + ALIGNED_16( uint8_t non_zero_count[X264_SCAN8_SIZE] ); /* -1 if unused, -2 if unavailable */ ALIGNED_4( int8_t ref[2][X264_SCAN8_LUMA_SIZE] ); @@ -677,11 +750,15 @@ struct x264_t /* number of neighbors (top and left) that used 8x8 dct */ int i_neighbour_transform_size; - int i_neighbour_interlaced; + int i_neighbour_skip; /* neighbor CBPs */ int i_cbp_top; int i_cbp_left; + + /* extra data required for mbaff in mv prediction */ + int16_t topright_mv[2][3][2]; + int8_t topright_ref[2][3]; } cache; /* */ @@ -700,14 +777,16 @@ struct x264_t int i_chroma_lambda2_offset; /* B_direct and weighted prediction */ - int16_t dist_scale_factor_buf[2][32][4]; + int16_t dist_scale_factor_buf[2][X264_REF_MAX*2][4]; int16_t (*dist_scale_factor)[4]; - int8_t bipred_weight_buf[2][32][4]; + int8_t bipred_weight_buf[2][X264_REF_MAX*2][4]; int8_t (*bipred_weight)[4]; /* maps fref1[0]'s ref indices into the current list0 */ #define map_col_to_list0(col) h->mb.map_col_to_list0[(col)+2] - int8_t map_col_to_list0[18]; + int8_t map_col_to_list0[X264_REF_MAX+2]; int ref_blind_dupe; /* The index of the blind reference frame duplicate. */ + int8_t deblock_ref_table[X264_REF_MAX*2+2]; +#define deblock_ref_table(x) h->mb.deblock_ref_table[(x)+2] } mb; /* rate control encoding only */ @@ -731,10 +810,10 @@ struct x264_t int i_mb_count_p; int i_mb_count_skip; int i_mb_count_8x8dct[2]; - int i_mb_count_ref[2][32]; + 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[3][13]; + int i_mb_pred_mode[4][13]; /* Adaptive direct mv pred */ int i_direct_score[2]; /* Metrics */ @@ -745,39 +824,48 @@ struct x264_t /* Cumulated stats */ /* per slice info */ - int i_frame_count[5]; - int64_t i_frame_size[5]; - double f_frame_qp[5]; + int i_frame_count[3]; + int64_t i_frame_size[3]; + double f_frame_qp[3]; int i_consecutive_bframes[X264_BFRAME_MAX+1]; /* */ - 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_psnr_mean_v[5]; - double f_ssim_mean_y[5]; + double f_ssd_global[3]; + double f_psnr_average[3]; + double f_psnr_mean_y[3]; + double f_psnr_mean_u[3]; + double f_psnr_mean_v[3]; + double f_ssim_mean_y[3]; + double f_frame_duration[3]; /* */ - int64_t i_mb_count[5][19]; + int64_t i_mb_count[3][19]; int64_t i_mb_partition[2][17]; int64_t i_mb_count_8x8dct[2]; - int64_t i_mb_count_ref[2][2][32]; + int64_t i_mb_count_ref[2][2][X264_REF_MAX*2]; int64_t i_mb_cbp[6]; - int64_t i_mb_pred_mode[3][13]; + int64_t i_mb_pred_mode[4][13]; /* */ int i_direct_score[2]; int i_direct_frames[2]; /* num p-frames weighted */ - int i_wpred[3]; + int i_wpred[2]; } stat; - ALIGNED_16( uint32_t nr_residual_sum[2][64] ); - ALIGNED_16( uint16_t nr_offset[2][64] ); - uint32_t nr_count[2]; + /* 0 = luma 4x4, 1 = luma 8x8, 2 = chroma 4x4 */ + udctcoef (*nr_offset)[64]; + uint32_t (*nr_residual_sum)[64]; + uint32_t *nr_count; + + ALIGNED_16( udctcoef nr_offset_denoise[3][64] ); + ALIGNED_16( uint32_t nr_residual_sum_buf[2][3][64] ); + uint32_t nr_count_buf[2][3]; /* Buffers that are allocated per-thread even in sliced threads. */ void *scratch_buffer; /* for any temporary storage that doesn't want repeated malloc */ - uint8_t *intra_border_backup[2][3]; /* bottom pixels of the previous mb row, used for intra prediction after the framebuffer has been deblocked */ + pixel *intra_border_backup[5][2]; /* 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]. */ + uint8_t (*deblock_strength[2])[2][8][4]; /* CPU functions dependents */ x264_predict_t predict_16x16[4+3]; @@ -792,8 +880,9 @@ struct x264_t x264_zigzag_function_t zigzagf; x264_quant_function_t quantf; x264_deblock_function_t loopf; + x264_bitstream_function_t bsf; -#ifdef HAVE_VISUALIZE +#if HAVE_VISUALIZE struct visualize_t *visualize; #endif x264_lookahead_t *lookahead; @@ -801,11 +890,12 @@ struct x264_t // included at the end because it needs x264_t #include "macroblock.h" -#include "rectangle.h" -#ifdef HAVE_MMX +#if ARCH_X86 || ARCH_X86_64 #include "x86/util.h" #endif +#include "rectangle.h" + #endif