1 /*****************************************************************************
2 * common.h: misc common functions
3 *****************************************************************************
4 * Copyright (C) 2003-2011 x264 project
6 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
7 * Loren Merritt <lorenm@u.washington.edu>
9 * This program is free software; you can redistribute it and/or modify
10 * it under the terms of the GNU General Public License as published by
11 * the Free Software Foundation; either version 2 of the License, or
12 * (at your option) any later version.
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 * GNU General Public License for more details.
19 * You should have received a copy of the GNU General Public License
20 * along with this program; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
23 * This program is also available under a commercial proprietary license.
24 * For more information, contact us at licensing@x264.com.
25 *****************************************************************************/
30 /****************************************************************************
32 ****************************************************************************/
33 #define X264_MIN(a,b) ( (a)<(b) ? (a) : (b) )
34 #define X264_MAX(a,b) ( (a)>(b) ? (a) : (b) )
35 #define X264_MIN3(a,b,c) X264_MIN((a),X264_MIN((b),(c)))
36 #define X264_MAX3(a,b,c) X264_MAX((a),X264_MAX((b),(c)))
37 #define X264_MIN4(a,b,c,d) X264_MIN((a),X264_MIN3((b),(c),(d)))
38 #define X264_MAX4(a,b,c,d) X264_MAX((a),X264_MAX3((b),(c),(d)))
39 #define XCHG(type,a,b) do{ type t = a; a = b; b = t; } while(0)
40 #define IS_DISPOSABLE(type) ( type == X264_TYPE_B )
41 #define FIX8(f) ((int)(f*(1<<8)+.5))
42 #define ALIGN(x,a) (((x)+((a)-1))&~((a)-1))
44 #define CHECKED_MALLOC( var, size )\
46 var = x264_malloc( size );\
50 #define CHECKED_MALLOCZERO( var, size )\
52 CHECKED_MALLOC( var, size );\
53 memset( var, 0, size );\
56 #define X264_BFRAME_MAX 16
57 #define X264_REF_MAX 16
58 #define X264_THREAD_MAX 128
59 #define X264_PCM_COST (384*BIT_DEPTH+16)
60 #define X264_LOOKAHEAD_MAX 250
61 #define QP_BD_OFFSET (6*(BIT_DEPTH-8))
62 #define QP_MAX_SPEC (51+QP_BD_OFFSET)
63 #define QP_MAX (QP_MAX_SPEC+18)
64 #define QP_MAX_MAX (51+2*6+18)
65 #define PIXEL_MAX ((1 << BIT_DEPTH)-1)
66 // arbitrary, but low because SATD scores are 1/4 normal
67 #define X264_LOOKAHEAD_QP (12+QP_BD_OFFSET)
68 #define SPEC_QP(x) X264_MIN((x), QP_MAX_SPEC)
70 // number of pixels (per thread) in progress at any given time.
71 // 16 for the macroblock in progress + 3 for deblocking + 3 for motion compensation filter + 2 for extra safety
72 #define X264_THREAD_HEIGHT 24
74 /* WEIGHTP_FAKE is set when mb_tree & psy are enabled, but normal weightp is disabled
75 * (such as in baseline). It checks for fades in lookahead and adjusts qp accordingly
76 * to increase quality. Defined as (-1) so that if(i_weighted_pred > 0) is true only when
77 * real weights are being used. */
79 #define X264_WEIGHTP_FAKE (-1)
81 #define NALU_OVERHEAD 5 // startcode + NAL type costs 5 bytes per frame
82 #define FILLER_OVERHEAD (NALU_OVERHEAD+1)
84 /****************************************************************************
86 ****************************************************************************/
95 /* Unions for type-punning.
96 * Mn: load or store n bits, aligned, native-endian
97 * CPn: copy n bits, aligned, native-endian
98 * we don't use memcpy for CPn because memcpy's args aren't assumed to be aligned */
99 typedef union { uint16_t i; uint8_t c[2]; } MAY_ALIAS x264_union16_t;
100 typedef union { uint32_t i; uint16_t b[2]; uint8_t c[4]; } MAY_ALIAS x264_union32_t;
101 typedef union { uint64_t i; uint32_t a[2]; uint16_t b[4]; uint8_t c[8]; } MAY_ALIAS x264_union64_t;
102 typedef struct { uint64_t i[2]; } x264_uint128_t;
103 typedef union { x264_uint128_t i; uint64_t a[2]; uint32_t b[4]; uint16_t c[8]; uint8_t d[16]; } MAY_ALIAS x264_union128_t;
104 #define M16(src) (((x264_union16_t*)(src))->i)
105 #define M32(src) (((x264_union32_t*)(src))->i)
106 #define M64(src) (((x264_union64_t*)(src))->i)
107 #define M128(src) (((x264_union128_t*)(src))->i)
108 #define M128_ZERO ((x264_uint128_t){{0,0}})
109 #define CP16(dst,src) M16(dst) = M16(src)
110 #define CP32(dst,src) M32(dst) = M32(src)
111 #define CP64(dst,src) M64(dst) = M64(src)
112 #define CP128(dst,src) M128(dst) = M128(src)
115 typedef uint16_t pixel;
116 typedef uint64_t pixel4;
117 typedef int32_t dctcoef;
118 typedef uint32_t udctcoef;
120 # define PIXEL_SPLAT_X4(x) ((x)*0x0001000100010001ULL)
121 # define MPIXEL_X4(src) M64(src)
123 typedef uint8_t pixel;
124 typedef uint32_t pixel4;
125 typedef int16_t dctcoef;
126 typedef uint16_t udctcoef;
128 # define PIXEL_SPLAT_X4(x) ((x)*0x01010101U)
129 # define MPIXEL_X4(src) M32(src)
132 #define BIT_DEPTH X264_BIT_DEPTH
134 #define CPPIXEL_X4(dst,src) MPIXEL_X4(dst) = MPIXEL_X4(src)
136 #define X264_SCAN8_SIZE (6*8)
137 #define X264_SCAN8_LUMA_SIZE (5*8)
138 #define X264_SCAN8_0 (4+1*8)
140 static const unsigned x264_scan8[16+2*4+3] =
143 4+1*8, 5+1*8, 4+2*8, 5+2*8,
144 6+1*8, 7+1*8, 6+2*8, 7+2*8,
145 4+3*8, 5+3*8, 4+4*8, 5+4*8,
146 6+3*8, 7+3*8, 6+4*8, 7+4*8,
173 #include "bitstream.h"
183 #include "threadpool.h"
185 /****************************************************************************
187 ****************************************************************************/
188 /* x264_malloc : will do or emulate a memalign
189 * you have to use x264_free for buffers allocated with x264_malloc */
190 void *x264_malloc( int );
191 void x264_free( void * );
193 /* x264_slurp_file: malloc space for the whole file and read it */
194 char *x264_slurp_file( const char *filename );
196 /* mdate: return the current date in microsecond */
197 int64_t x264_mdate( void );
199 /* x264_param2string: return a (malloced) string containing most of
200 * the encoding options */
201 char *x264_param2string( x264_param_t *p, int b_res );
204 void x264_log( x264_t *h, int i_level, const char *psz_fmt, ... );
206 void x264_reduce_fraction( uint32_t *n, uint32_t *d );
207 void x264_reduce_fraction64( uint64_t *n, uint64_t *d );
208 void x264_cavlc_init( void );
209 void x264_cabac_init( void );
211 static ALWAYS_INLINE pixel x264_clip_pixel( int x )
213 return ( (x & ~PIXEL_MAX) ? (-x)>>31 & PIXEL_MAX : x );
216 static ALWAYS_INLINE int x264_clip3( int v, int i_min, int i_max )
218 return ( (v < i_min) ? i_min : (v > i_max) ? i_max : v );
221 static ALWAYS_INLINE double x264_clip3f( double v, double f_min, double f_max )
223 return ( (v < f_min) ? f_min : (v > f_max) ? f_max : v );
226 static ALWAYS_INLINE int x264_median( int a, int b, int c )
228 int t = (a-b)&((a-b)>>31);
231 b -= (b-c)&((b-c)>>31);
232 b += (a-b)&((a-b)>>31);
236 static ALWAYS_INLINE void x264_median_mv( int16_t *dst, int16_t *a, int16_t *b, int16_t *c )
238 dst[0] = x264_median( a[0], b[0], c[0] );
239 dst[1] = x264_median( a[1], b[1], c[1] );
242 static ALWAYS_INLINE int x264_predictor_difference( int16_t (*mvc)[2], intptr_t i_mvc )
245 for( int i = 0; i < i_mvc-1; i++ )
247 sum += abs( mvc[i][0] - mvc[i+1][0] )
248 + abs( mvc[i][1] - mvc[i+1][1] );
253 static ALWAYS_INLINE uint16_t x264_cabac_mvd_sum( uint8_t *mvdleft, uint8_t *mvdtop )
255 int amvd0 = abs(mvdleft[0]) + abs(mvdtop[0]);
256 int amvd1 = abs(mvdleft[1]) + abs(mvdtop[1]);
257 amvd0 = (amvd0 > 2) + (amvd0 > 32);
258 amvd1 = (amvd1 > 2) + (amvd1 > 32);
259 return amvd0 + (amvd1<<8);
262 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 )
264 for( int i = 0; i < i_mvc; i++ )
266 int mx = (mvc[i][0] + 2) >> 2;
267 int my = (mvc[i][1] + 2) >> 2;
268 dst[i][0] = x264_clip3( mx, mv_x_min, mv_x_max );
269 dst[i][1] = x264_clip3( my, mv_y_min, mv_y_max );
273 extern const uint8_t x264_exp2_lut[64];
274 extern const float x264_log2_lut[128];
275 extern const float x264_log2_lz_lut[32];
277 /* Not a general-purpose function; multiplies input by -1/6 to convert
279 static ALWAYS_INLINE int x264_exp2fix8( float x )
281 int i = x*(-64.f/6.f) + 512.5f;
282 if( i < 0 ) return 0;
283 if( i > 1023 ) return 0xffff;
284 return (x264_exp2_lut[i&63]+256) << (i>>6) >> 8;
287 static ALWAYS_INLINE float x264_log2( uint32_t x )
289 int lz = x264_clz( x );
290 return x264_log2_lut[(x<<lz>>24)&0x7f] + x264_log2_lz_lut[lz];
293 /****************************************************************************
295 ****************************************************************************/
303 static const char slice_type_to_char[] = { 'P', 'B', 'I' };
305 enum sei_payload_type_e
307 SEI_BUFFERING_PERIOD = 0,
309 SEI_PAN_SCAN_RECT = 2,
311 SEI_USER_DATA_REGISTERED = 4,
312 SEI_USER_DATA_UNREGISTERED = 5,
313 SEI_RECOVERY_POINT = 6,
314 SEI_DEC_REF_PIC_MARKING = 7,
315 SEI_FRAME_PACKING = 45,
335 int i_idr_pic_id; /* -1 if nal_type != 5 */
338 int i_delta_poc_bottom;
341 int i_redundant_pic_cnt;
343 int b_direct_spatial_mv_pred;
345 int b_num_ref_idx_override;
346 int i_num_ref_idx_l0_active;
347 int i_num_ref_idx_l1_active;
349 int b_ref_pic_list_reordering[2];
354 } ref_pic_list_order[2][X264_REF_MAX];
356 /* P-frame weighting */
357 x264_weight_t weight[X264_REF_MAX*2][3];
359 int i_mmco_remove_from_end;
360 int i_mmco_command_count;
361 struct /* struct for future expansion */
363 int i_difference_of_pic_nums;
365 } mmco[X264_REF_MAX];
367 int i_cabac_init_idc;
374 /* deblocking filter */
375 int i_disable_deblocking_filter_idc;
376 int i_alpha_c0_offset;
379 } x264_slice_header_t;
381 typedef struct x264_lookahead_t
383 volatile uint8_t b_exit_thread;
384 uint8_t b_thread_active;
385 uint8_t b_analyse_keyframe;
387 int i_slicetype_length;
388 x264_frame_t *last_nonb;
389 x264_pthread_t thread_handle;
390 x264_sync_frame_list_t ifbuf;
391 x264_sync_frame_list_t next;
392 x264_sync_frame_list_t ofbuf;
395 typedef struct x264_ratecontrol_t x264_ratecontrol_t;
397 typedef struct x264_left_table_t
401 uint8_t nnz_chroma[4];
408 /* encoder parameters */
411 x264_t *thread[X264_THREAD_MAX+1];
413 int i_thread_phase; /* which thread to use for the next frame */
414 int i_threadslice_start; /* first row in this thread slice */
415 int i_threadslice_end; /* row after the end of this thread slice */
416 x264_threadpool_t *threadpool;
418 /* bitstream output */
422 int i_nals_allocated;
424 int i_bitstream; /* size of p_bitstream */
425 uint8_t *p_bitstream; /* will hold data for all nal */
435 /**** thread synchronization starts here ****/
437 /* frame number/poc */
441 int i_thread_frames; /* Number of different frames being encoded by threads;
442 * 1 when sliced-threads is on. */
446 int64_t i_disp_fields; /* Number of displayed fields (both coded and implied via pic_struct) */
447 int i_disp_fields_last_frame;
448 int64_t i_prev_duration; /* Duration of previous frame */
449 int64_t i_coded_fields; /* Number of coded fields (both coded and implied via pic_struct) */
450 int64_t i_cpb_delay; /* Equal to number of fields preceding this field
451 * since last buffering_period SEI */
452 int64_t i_coded_fields_lookahead; /* Use separate counters for lookahead */
453 int64_t i_cpb_delay_lookahead;
455 int64_t i_cpb_delay_pir_offset;
457 int b_queued_intra_refresh;
458 int64_t i_last_idr_pts;
460 /* We use only one SPS and one PPS */
461 x264_sps_t sps_array[1];
462 x264_pps_t pps_array[1];
465 /* quantization matrix for decoding, [cqm][qp%6][coef] */
466 int (*dequant4_mf[4])[16]; /* [4][6][16] */
467 int (*dequant8_mf[2])[64]; /* [2][6][64] */
468 /* quantization matrix for trellis, [cqm][qp][coef] */
469 int (*unquant4_mf[4])[16]; /* [4][52][16] */
470 int (*unquant8_mf[2])[64]; /* [2][52][64] */
471 /* quantization matrix for deadzone */
472 udctcoef (*quant4_mf[4])[16]; /* [4][52][16] */
473 udctcoef (*quant8_mf[2])[64]; /* [2][52][64] */
474 udctcoef (*quant4_bias[4])[16]; /* [4][52][16] */
475 udctcoef (*quant8_bias[2])[64]; /* [2][52][64] */
476 udctcoef (*nr_offset_emergency)[3][64];
478 /* mv/ref cost arrays. */
479 uint16_t *cost_mv[QP_MAX+1];
480 uint16_t *cost_mv_fpel[QP_MAX+1][4];
482 const uint8_t *chroma_qp_table; /* includes both the nonlinear luma->chroma mapping and chroma_qp_offset */
485 x264_slice_header_t sh;
487 /* Slice header backup, for SEI_DEC_REF_PIC_MARKING */
489 x264_slice_header_t sh_backup;
496 /* Frames to be encoded (whose types have been decided) */
497 x264_frame_t **current;
498 /* Unused frames: 0 = fenc, 1 = fdec */
499 x264_frame_t **unused[2];
501 /* Unused blank frames (for duplicates) */
502 x264_frame_t **blank_unused;
504 /* frames used for reference + sentinels */
505 x264_frame_t *reference[X264_REF_MAX+2];
507 int i_last_keyframe; /* Frame number of the last keyframe */
508 int i_last_idr; /* Frame number of the last IDR (not RP)*/
509 int i_poc_last_open_gop; /* Poc of the I frame of the last open-gop. The value
510 * is only assigned during the period between that
511 * I frame and the next P or I frame, else -1 */
513 int i_input; /* Number of input frames already accepted */
515 int i_max_dpb; /* Number of frames allocated in the decoded picture buffer */
518 int i_delay; /* Number of frames buffered for B reordering */
520 int64_t i_bframe_delay_time;
522 int64_t i_prev_reordered_pts[2];
523 int64_t i_largest_pts;
524 int64_t i_second_largest_pts;
525 int b_have_lowres; /* Whether 1/2 resolution luma planes are being used */
526 int b_have_sub8x8_esa;
529 /* current frame being encoded */
532 /* frame being reconstructed */
535 /* references lists */
537 x264_frame_t *fref[2][X264_REF_MAX+3];
538 x264_frame_t *fref_nearest[2];
539 int b_ref_reorder[2];
542 int initial_cpb_removal_delay;
543 int initial_cpb_removal_delay_offset;
544 int64_t i_reordered_pts_delay;
546 /* Current MB DCT coeffs */
549 ALIGNED_16( dctcoef luma16x16_dc[16] );
550 ALIGNED_16( dctcoef chroma_dc[2][4] );
551 // FIXME share memory?
552 ALIGNED_16( dctcoef luma8x8[4][64] );
553 ALIGNED_16( dctcoef luma4x4[16+8][16] );
556 /* MB table and cache for current frame/mb */
561 int i_mb_count; /* number of mbs in a frame */
577 /* Search parameters */
582 int b_noise_reduction;
584 int i_psy_rd; /* Psy RD strength--fixed point value*/
585 int i_psy_trellis; /* Psy trellis strength--fixed point value*/
588 int b_adaptive_mbaff; /* MBAFF+subme 0 requires non-adaptive MBAFF i.e. all field mbs */
590 /* Allowed qpel MV range to stay within the picture + emulated edge pixels */
593 /* Subpel MV range for motion search.
594 * same mv_min/max but includes levels' i_mv_range. */
597 /* Fullpel MV range for motion search */
601 /* neighboring MBs */
602 unsigned int i_neighbour;
603 unsigned int i_neighbour8[4]; /* neighbours of each 8x8 or 4x4 block that are available */
604 unsigned int i_neighbour4[16]; /* at the time the block is coded */
605 unsigned int i_neighbour_intra; /* for constrained intra pred */
606 unsigned int i_neighbour_frame; /* ignoring slice boundaries */
608 int i_mb_type_left[2];
609 int i_mb_type_topleft;
610 int i_mb_type_topright;
615 int i_mb_topright_xy;
619 x264_left_table_t *left_index_table;
620 int topleft_partition;
622 /**** thread synchronization ends here ****/
623 /* subsequent variables are either thread-local or constant,
624 * and won't be copied from one thread to another */
627 int8_t *type; /* mb type */
628 uint8_t *partition; /* mb partition */
629 int8_t *qp; /* mb qp */
630 int16_t *cbp; /* mb cbp: 0x0?: luma, 0x?0: chroma, 0x100: luma dc, 0x0200 and 0x0400: chroma dc (all set for PCM)*/
631 int8_t (*intra4x4_pred_mode)[8]; /* intra4x4 pred mode. for non I4x4 set to I_PRED_4x4_DC(2) */
632 /* actually has only 7 entries; set to 8 for write-combining optimizations */
633 uint8_t (*non_zero_count)[16+4+4]; /* nzc. for I_PCM set to 16 */
634 int8_t *chroma_pred_mode; /* chroma_pred_mode. cabac only. for non intra I_PRED_CHROMA_DC(0) */
635 int16_t (*mv[2])[2]; /* mb mv. set to 0 for intra mb */
636 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 */
637 int8_t *ref[2]; /* mb ref. set to -1 if non used (intra or Lx only) */
638 int16_t (*mvr[2][X264_REF_MAX*2])[2];/* 16x16 mv for each possible ref */
639 int8_t *skipbp; /* block pattern for SKIP or DIRECT (sub)mbs. B-frames + cabac only */
640 int8_t *mb_transform_size; /* transform_size_8x8_flag of each mb */
641 uint16_t *slice_table; /* sh->first_mb of the slice that the indexed mb is part of
642 * NOTE: this will fail on resolutions above 2^16 MBs... */
645 /* buffer for weighted versions of the reference frames */
646 pixel *p_weight_buf[X264_REF_MAX];
651 ALIGNED_4( uint8_t i_sub_partition[4] );
657 int i_intra16x16_pred_mode;
658 int i_chroma_pred_mode;
660 /* skip flags for i4x4 and i8x8
661 * 0 = encode as normal.
662 * 1 (non-RD only) = the DCT is still in h->dct, restore fdec and skip reconstruction.
663 * 2 (RD only) = the DCT has since been overwritten by RD; restore that too. */
665 /* skip flag for motion compensation */
666 /* if we've already done MC, we don't need to do it again */
668 /* set to true if we are re-encoding a macroblock. */
670 int ip_offset; /* Used by PIR to offset the quantizer of intra-refresh blocks. */
672 int b_overflow; /* If CAVLC had a level code overflow during bitstream writing. */
676 /* space for p_fenc and p_fdec */
677 #define FENC_STRIDE 16
678 #define FDEC_STRIDE 32
679 ALIGNED_16( pixel fenc_buf[24*FENC_STRIDE] );
680 ALIGNED_16( pixel fdec_buf[27*FDEC_STRIDE] );
682 /* i4x4 and i8x8 backup data, for skipping the encode stage when possible */
683 ALIGNED_16( pixel i4x4_fdec_buf[16*16] );
684 ALIGNED_16( pixel i8x8_fdec_buf[16*16] );
685 ALIGNED_16( dctcoef i8x8_dct_buf[3][64] );
686 ALIGNED_16( dctcoef i4x4_dct_buf[15][16] );
687 uint32_t i4x4_nnz_buf[4];
688 uint32_t i8x8_nnz_buf[4];
692 /* Psy trellis DCT data */
693 ALIGNED_16( dctcoef fenc_dct8[4][64] );
694 ALIGNED_16( dctcoef fenc_dct4[16][16] );
696 /* Psy RD SATD/SA8D scores cache */
697 ALIGNED_16( uint64_t fenc_hadamard_cache[9] );
698 ALIGNED_16( uint32_t fenc_satd_cache[32] );
700 /* pointer over mb of the frame to be compressed */
701 pixel *p_fenc[3]; /* y,u,v */
702 /* pointer to the actual source frame, not a block copy */
703 pixel *p_fenc_plane[2]; /* y,uv */
705 /* pointer over mb of the frame to be reconstructed */
708 /* pointer over mb of the references */
710 pixel *p_fref[2][X264_REF_MAX*2][4+1]; /* last: yN, yH, yV, yHV, uv */
711 pixel *p_fref_w[X264_REF_MAX*2]; /* weighted fullpel luma */
712 uint16_t *p_integral[2][X264_REF_MAX];
721 /* real intra4x4_pred_mode if I_4X4 or I_8X8, I_PRED_4x4_DC if mb available, -1 if not */
722 ALIGNED_8( int8_t intra4x4_pred_mode[X264_SCAN8_LUMA_SIZE] );
724 /* i_non_zero_count if available else 0x80 */
725 ALIGNED_16( uint8_t non_zero_count[X264_SCAN8_SIZE] );
727 /* -1 if unused, -2 if unavailable */
728 ALIGNED_4( int8_t ref[2][X264_SCAN8_LUMA_SIZE] );
730 /* 0 if not available */
731 ALIGNED_16( int16_t mv[2][X264_SCAN8_LUMA_SIZE][2] );
732 ALIGNED_8( uint8_t mvd[2][X264_SCAN8_LUMA_SIZE][2] );
734 /* 1 if SKIP or DIRECT. set only for B-frames + CABAC */
735 ALIGNED_4( int8_t skip[X264_SCAN8_LUMA_SIZE] );
737 ALIGNED_4( int16_t direct_mv[2][4][2] );
738 ALIGNED_4( int8_t direct_ref[2][4] );
739 int direct_partition;
740 ALIGNED_4( int16_t pskip_mv[2] );
742 /* number of neighbors (top and left) that used 8x8 dct */
743 int i_neighbour_transform_size;
744 int i_neighbour_interlaced;
752 int i_qp; /* current qp */
754 int i_last_qp; /* last qp */
755 int i_last_dqp; /* last delta qp */
756 int b_variable_qp; /* whether qp is allowed to vary per macroblock */
758 int b_direct_auto_read; /* take stats for --direct auto from the 2pass log */
759 int b_direct_auto_write; /* analyse direct modes, to use and/or save */
762 int i_trellis_lambda2[2][2]; /* [luma,chroma][inter,intra] */
764 int i_chroma_lambda2_offset;
766 /* B_direct and weighted prediction */
767 int16_t dist_scale_factor_buf[2][X264_REF_MAX*2][4];
768 int16_t (*dist_scale_factor)[4];
769 int8_t bipred_weight_buf[2][X264_REF_MAX*2][4];
770 int8_t (*bipred_weight)[4];
771 /* maps fref1[0]'s ref indices into the current list0 */
772 #define map_col_to_list0(col) h->mb.map_col_to_list0[(col)+2]
773 int8_t map_col_to_list0[X264_REF_MAX+2];
774 int ref_blind_dupe; /* The index of the blind reference frame duplicate. */
775 int8_t deblock_ref_table[X264_REF_MAX*2+2];
776 #define deblock_ref_table(x) h->mb.deblock_ref_table[(x)+2]
779 /* rate control encoding only */
780 x264_ratecontrol_t *rc;
785 /* Current frame stats */
788 /* MV bits (MV+Ref+Block Type) */
790 /* Texture bits (DCT coefs) */
799 int i_mb_count_8x8dct[2];
800 int i_mb_count_ref[2][X264_REF_MAX*2];
801 int i_mb_partition[17];
803 int i_mb_pred_mode[4][13];
804 /* Adaptive direct mv pred */
805 int i_direct_score[2];
811 /* Cumulated stats */
814 int i_frame_count[3];
815 int64_t i_frame_size[3];
816 double f_frame_qp[3];
817 int i_consecutive_bframes[X264_BFRAME_MAX+1];
819 double f_ssd_global[3];
820 double f_psnr_average[3];
821 double f_psnr_mean_y[3];
822 double f_psnr_mean_u[3];
823 double f_psnr_mean_v[3];
824 double f_ssim_mean_y[3];
825 double f_frame_duration[3];
827 int64_t i_mb_count[3][19];
828 int64_t i_mb_partition[2][17];
829 int64_t i_mb_count_8x8dct[2];
830 int64_t i_mb_count_ref[2][2][X264_REF_MAX*2];
832 int64_t i_mb_pred_mode[4][13];
834 int i_direct_score[2];
835 int i_direct_frames[2];
836 /* num p-frames weighted */
841 /* 0 = luma 4x4, 1 = luma 8x8, 2 = chroma 4x4 */
842 udctcoef (*nr_offset)[64];
843 uint32_t (*nr_residual_sum)[64];
846 ALIGNED_16( udctcoef nr_offset_denoise[3][64] );
847 ALIGNED_16( uint32_t nr_residual_sum_buf[2][3][64] );
848 uint32_t nr_count_buf[2][3];
850 /* Buffers that are allocated per-thread even in sliced threads. */
851 void *scratch_buffer; /* for any temporary storage that doesn't want repeated malloc */
852 pixel *intra_border_backup[5][2]; /* bottom pixels of the previous mb row, used for intra prediction after the framebuffer has been deblocked */
853 uint8_t (*deblock_strength[2])[2][4][4];
855 /* CPU functions dependents */
856 x264_predict_t predict_16x16[4+3];
857 x264_predict_t predict_8x8c[4+3];
858 x264_predict8x8_t predict_8x8[9+3];
859 x264_predict_t predict_4x4[9+3];
860 x264_predict_8x8_filter_t predict_8x8_filter;
862 x264_pixel_function_t pixf;
863 x264_mc_functions_t mc;
864 x264_dct_function_t dctf;
865 x264_zigzag_function_t zigzagf;
866 x264_quant_function_t quantf;
867 x264_deblock_function_t loopf;
868 x264_bitstream_function_t bsf;
871 struct visualize_t *visualize;
873 x264_lookahead_t *lookahead;
876 // included at the end because it needs x264_t
877 #include "macroblock.h"
879 #if ARCH_X86 || ARCH_X86_64
880 #include "x86/util.h"
883 #include "rectangle.h"