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 ****************************************************************************/
96 # define MB_INTERLACED h->mb.b_interlaced
97 # define SLICE_MBAFF h->sh.b_mbaff
98 # define PARAM_INTERLACED h->param.b_interlaced
100 # define MB_INTERLACED 0
101 # define SLICE_MBAFF 0
102 # define PARAM_INTERLACED 0
105 /* Unions for type-punning.
106 * Mn: load or store n bits, aligned, native-endian
107 * CPn: copy n bits, aligned, native-endian
108 * we don't use memcpy for CPn because memcpy's args aren't assumed to be aligned */
109 typedef union { uint16_t i; uint8_t c[2]; } MAY_ALIAS x264_union16_t;
110 typedef union { uint32_t i; uint16_t b[2]; uint8_t c[4]; } MAY_ALIAS x264_union32_t;
111 typedef union { uint64_t i; uint32_t a[2]; uint16_t b[4]; uint8_t c[8]; } MAY_ALIAS x264_union64_t;
112 typedef struct { uint64_t i[2]; } x264_uint128_t;
113 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;
114 #define M16(src) (((x264_union16_t*)(src))->i)
115 #define M32(src) (((x264_union32_t*)(src))->i)
116 #define M64(src) (((x264_union64_t*)(src))->i)
117 #define M128(src) (((x264_union128_t*)(src))->i)
118 #define M128_ZERO ((x264_uint128_t){{0,0}})
119 #define CP16(dst,src) M16(dst) = M16(src)
120 #define CP32(dst,src) M32(dst) = M32(src)
121 #define CP64(dst,src) M64(dst) = M64(src)
122 #define CP128(dst,src) M128(dst) = M128(src)
125 typedef uint16_t pixel;
126 typedef uint64_t pixel4;
127 typedef int32_t dctcoef;
128 typedef uint32_t udctcoef;
130 # define PIXEL_SPLAT_X4(x) ((x)*0x0001000100010001ULL)
131 # define MPIXEL_X4(src) M64(src)
133 typedef uint8_t pixel;
134 typedef uint32_t pixel4;
135 typedef int16_t dctcoef;
136 typedef uint16_t udctcoef;
138 # define PIXEL_SPLAT_X4(x) ((x)*0x01010101U)
139 # define MPIXEL_X4(src) M32(src)
142 #define BIT_DEPTH X264_BIT_DEPTH
144 #define CPPIXEL_X4(dst,src) MPIXEL_X4(dst) = MPIXEL_X4(src)
146 #define X264_SCAN8_SIZE (6*8)
147 #define X264_SCAN8_LUMA_SIZE (5*8)
148 #define X264_SCAN8_0 (4+1*8)
150 static const unsigned x264_scan8[16+2*4+3] =
153 4+1*8, 5+1*8, 4+2*8, 5+2*8,
154 6+1*8, 7+1*8, 6+2*8, 7+2*8,
155 4+3*8, 5+3*8, 4+4*8, 5+4*8,
156 6+3*8, 7+3*8, 6+4*8, 7+4*8,
183 #include "bitstream.h"
193 #include "threadpool.h"
195 /****************************************************************************
197 ****************************************************************************/
198 /* x264_malloc : will do or emulate a memalign
199 * you have to use x264_free for buffers allocated with x264_malloc */
200 void *x264_malloc( int );
201 void x264_free( void * );
203 /* x264_slurp_file: malloc space for the whole file and read it */
204 char *x264_slurp_file( const char *filename );
206 /* mdate: return the current date in microsecond */
207 int64_t x264_mdate( void );
209 /* x264_param2string: return a (malloced) string containing most of
210 * the encoding options */
211 char *x264_param2string( x264_param_t *p, int b_res );
214 void x264_log( x264_t *h, int i_level, const char *psz_fmt, ... );
216 void x264_reduce_fraction( uint32_t *n, uint32_t *d );
217 void x264_reduce_fraction64( uint64_t *n, uint64_t *d );
218 void x264_cavlc_init( void );
219 void x264_cabac_init( void );
221 static ALWAYS_INLINE pixel x264_clip_pixel( int x )
223 return ( (x & ~PIXEL_MAX) ? (-x)>>31 & PIXEL_MAX : x );
226 static ALWAYS_INLINE int x264_clip3( int v, int i_min, int i_max )
228 return ( (v < i_min) ? i_min : (v > i_max) ? i_max : v );
231 static ALWAYS_INLINE double x264_clip3f( double v, double f_min, double f_max )
233 return ( (v < f_min) ? f_min : (v > f_max) ? f_max : v );
236 static ALWAYS_INLINE int x264_median( int a, int b, int c )
238 int t = (a-b)&((a-b)>>31);
241 b -= (b-c)&((b-c)>>31);
242 b += (a-b)&((a-b)>>31);
246 static ALWAYS_INLINE void x264_median_mv( int16_t *dst, int16_t *a, int16_t *b, int16_t *c )
248 dst[0] = x264_median( a[0], b[0], c[0] );
249 dst[1] = x264_median( a[1], b[1], c[1] );
252 static ALWAYS_INLINE int x264_predictor_difference( int16_t (*mvc)[2], intptr_t i_mvc )
255 for( int i = 0; i < i_mvc-1; i++ )
257 sum += abs( mvc[i][0] - mvc[i+1][0] )
258 + abs( mvc[i][1] - mvc[i+1][1] );
263 static ALWAYS_INLINE uint16_t x264_cabac_mvd_sum( uint8_t *mvdleft, uint8_t *mvdtop )
265 int amvd0 = abs(mvdleft[0]) + abs(mvdtop[0]);
266 int amvd1 = abs(mvdleft[1]) + abs(mvdtop[1]);
267 amvd0 = (amvd0 > 2) + (amvd0 > 32);
268 amvd1 = (amvd1 > 2) + (amvd1 > 32);
269 return amvd0 + (amvd1<<8);
272 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 )
274 for( int i = 0; i < i_mvc; i++ )
276 int mx = (mvc[i][0] + 2) >> 2;
277 int my = (mvc[i][1] + 2) >> 2;
278 dst[i][0] = x264_clip3( mx, mv_x_min, mv_x_max );
279 dst[i][1] = x264_clip3( my, mv_y_min, mv_y_max );
283 extern const uint8_t x264_exp2_lut[64];
284 extern const float x264_log2_lut[128];
285 extern const float x264_log2_lz_lut[32];
287 /* Not a general-purpose function; multiplies input by -1/6 to convert
289 static ALWAYS_INLINE int x264_exp2fix8( float x )
291 int i = x*(-64.f/6.f) + 512.5f;
292 if( i < 0 ) return 0;
293 if( i > 1023 ) return 0xffff;
294 return (x264_exp2_lut[i&63]+256) << (i>>6) >> 8;
297 static ALWAYS_INLINE float x264_log2( uint32_t x )
299 int lz = x264_clz( x );
300 return x264_log2_lut[(x<<lz>>24)&0x7f] + x264_log2_lz_lut[lz];
303 /****************************************************************************
305 ****************************************************************************/
313 static const char slice_type_to_char[] = { 'P', 'B', 'I' };
315 enum sei_payload_type_e
317 SEI_BUFFERING_PERIOD = 0,
319 SEI_PAN_SCAN_RECT = 2,
321 SEI_USER_DATA_REGISTERED = 4,
322 SEI_USER_DATA_UNREGISTERED = 5,
323 SEI_RECOVERY_POINT = 6,
324 SEI_DEC_REF_PIC_MARKING = 7,
325 SEI_FRAME_PACKING = 45,
345 int i_idr_pic_id; /* -1 if nal_type != 5 */
348 int i_delta_poc_bottom;
351 int i_redundant_pic_cnt;
353 int b_direct_spatial_mv_pred;
355 int b_num_ref_idx_override;
356 int i_num_ref_idx_l0_active;
357 int i_num_ref_idx_l1_active;
359 int b_ref_pic_list_reordering[2];
364 } ref_pic_list_order[2][X264_REF_MAX];
366 /* P-frame weighting */
367 x264_weight_t weight[X264_REF_MAX*2][3];
369 int i_mmco_remove_from_end;
370 int i_mmco_command_count;
371 struct /* struct for future expansion */
373 int i_difference_of_pic_nums;
375 } mmco[X264_REF_MAX];
377 int i_cabac_init_idc;
384 /* deblocking filter */
385 int i_disable_deblocking_filter_idc;
386 int i_alpha_c0_offset;
389 } x264_slice_header_t;
391 typedef struct x264_lookahead_t
393 volatile uint8_t b_exit_thread;
394 uint8_t b_thread_active;
395 uint8_t b_analyse_keyframe;
397 int i_slicetype_length;
398 x264_frame_t *last_nonb;
399 x264_pthread_t thread_handle;
400 x264_sync_frame_list_t ifbuf;
401 x264_sync_frame_list_t next;
402 x264_sync_frame_list_t ofbuf;
405 typedef struct x264_ratecontrol_t x264_ratecontrol_t;
407 typedef struct x264_left_table_t
411 uint8_t nnz_chroma[4];
418 /* encoder parameters */
421 x264_t *thread[X264_THREAD_MAX+1];
423 int i_thread_phase; /* which thread to use for the next frame */
424 int i_threadslice_start; /* first row in this thread slice */
425 int i_threadslice_end; /* row after the end of this thread slice */
426 x264_threadpool_t *threadpool;
428 /* bitstream output */
432 int i_nals_allocated;
434 int i_bitstream; /* size of p_bitstream */
435 uint8_t *p_bitstream; /* will hold data for all nal */
445 /**** thread synchronization starts here ****/
447 /* frame number/poc */
451 int i_thread_frames; /* Number of different frames being encoded by threads;
452 * 1 when sliced-threads is on. */
456 int64_t i_disp_fields; /* Number of displayed fields (both coded and implied via pic_struct) */
457 int i_disp_fields_last_frame;
458 int64_t i_prev_duration; /* Duration of previous frame */
459 int64_t i_coded_fields; /* Number of coded fields (both coded and implied via pic_struct) */
460 int64_t i_cpb_delay; /* Equal to number of fields preceding this field
461 * since last buffering_period SEI */
462 int64_t i_coded_fields_lookahead; /* Use separate counters for lookahead */
463 int64_t i_cpb_delay_lookahead;
465 int64_t i_cpb_delay_pir_offset;
467 int b_queued_intra_refresh;
468 int64_t i_last_idr_pts;
470 /* We use only one SPS and one PPS */
471 x264_sps_t sps_array[1];
472 x264_pps_t pps_array[1];
475 /* quantization matrix for decoding, [cqm][qp%6][coef] */
476 int (*dequant4_mf[4])[16]; /* [4][6][16] */
477 int (*dequant8_mf[2])[64]; /* [2][6][64] */
478 /* quantization matrix for trellis, [cqm][qp][coef] */
479 int (*unquant4_mf[4])[16]; /* [4][52][16] */
480 int (*unquant8_mf[2])[64]; /* [2][52][64] */
481 /* quantization matrix for deadzone */
482 udctcoef (*quant4_mf[4])[16]; /* [4][52][16] */
483 udctcoef (*quant8_mf[2])[64]; /* [2][52][64] */
484 udctcoef (*quant4_bias[4])[16]; /* [4][52][16] */
485 udctcoef (*quant8_bias[2])[64]; /* [2][52][64] */
486 udctcoef (*nr_offset_emergency)[3][64];
488 /* mv/ref cost arrays. */
489 uint16_t *cost_mv[QP_MAX+1];
490 uint16_t *cost_mv_fpel[QP_MAX+1][4];
492 const uint8_t *chroma_qp_table; /* includes both the nonlinear luma->chroma mapping and chroma_qp_offset */
495 x264_slice_header_t sh;
497 /* Slice header backup, for SEI_DEC_REF_PIC_MARKING */
499 x264_slice_header_t sh_backup;
506 /* Frames to be encoded (whose types have been decided) */
507 x264_frame_t **current;
508 /* Unused frames: 0 = fenc, 1 = fdec */
509 x264_frame_t **unused[2];
511 /* Unused blank frames (for duplicates) */
512 x264_frame_t **blank_unused;
514 /* frames used for reference + sentinels */
515 x264_frame_t *reference[X264_REF_MAX+2];
517 int i_last_keyframe; /* Frame number of the last keyframe */
518 int i_last_idr; /* Frame number of the last IDR (not RP)*/
519 int i_poc_last_open_gop; /* Poc of the I frame of the last open-gop. The value
520 * is only assigned during the period between that
521 * I frame and the next P or I frame, else -1 */
523 int i_input; /* Number of input frames already accepted */
525 int i_max_dpb; /* Number of frames allocated in the decoded picture buffer */
528 int i_delay; /* Number of frames buffered for B reordering */
530 int64_t i_bframe_delay_time;
532 int64_t i_prev_reordered_pts[2];
533 int64_t i_largest_pts;
534 int64_t i_second_largest_pts;
535 int b_have_lowres; /* Whether 1/2 resolution luma planes are being used */
536 int b_have_sub8x8_esa;
539 /* current frame being encoded */
542 /* frame being reconstructed */
545 /* references lists */
547 x264_frame_t *fref[2][X264_REF_MAX+3];
548 x264_frame_t *fref_nearest[2];
549 int b_ref_reorder[2];
552 int initial_cpb_removal_delay;
553 int initial_cpb_removal_delay_offset;
554 int64_t i_reordered_pts_delay;
556 /* Current MB DCT coeffs */
559 ALIGNED_16( dctcoef luma16x16_dc[16] );
560 ALIGNED_16( dctcoef chroma_dc[2][4] );
561 // FIXME share memory?
562 ALIGNED_16( dctcoef luma8x8[4][64] );
563 ALIGNED_16( dctcoef luma4x4[16+8][16] );
566 /* MB table and cache for current frame/mb */
571 int i_mb_count; /* number of mbs in a frame */
587 /* Search parameters */
592 int b_noise_reduction;
594 int i_psy_rd; /* Psy RD strength--fixed point value*/
595 int i_psy_trellis; /* Psy trellis strength--fixed point value*/
598 int b_adaptive_mbaff; /* MBAFF+subme 0 requires non-adaptive MBAFF i.e. all field mbs */
600 /* Allowed qpel MV range to stay within the picture + emulated edge pixels */
603 int mv_miny_row[3]; /* 0 == top progressive, 1 == bot progressive, 2 == interlaced */
605 /* Subpel MV range for motion search.
606 * same mv_min/max but includes levels' i_mv_range. */
609 int mv_miny_spel_row[3];
610 int mv_maxy_spel_row[3];
611 /* Fullpel MV range for motion search */
614 int mv_miny_fpel_row[3];
615 int mv_maxy_fpel_row[3];
617 /* neighboring MBs */
618 unsigned int i_neighbour;
619 unsigned int i_neighbour8[4]; /* neighbours of each 8x8 or 4x4 block that are available */
620 unsigned int i_neighbour4[16]; /* at the time the block is coded */
621 unsigned int i_neighbour_intra; /* for constrained intra pred */
622 unsigned int i_neighbour_frame; /* ignoring slice boundaries */
624 int i_mb_type_left[2];
625 int i_mb_type_topleft;
626 int i_mb_type_topright;
631 int i_mb_topright_xy;
635 const x264_left_table_t *left_index_table;
636 int i_mb_top_mbpair_xy;
637 int topleft_partition;
639 int field_decoding_flag;
641 /**** thread synchronization ends here ****/
642 /* subsequent variables are either thread-local or constant,
643 * and won't be copied from one thread to another */
646 int8_t *type; /* mb type */
647 uint8_t *partition; /* mb partition */
648 int8_t *qp; /* mb qp */
649 int16_t *cbp; /* mb cbp: 0x0?: luma, 0x?0: chroma, 0x100: luma dc, 0x0200 and 0x0400: chroma dc (all set for PCM)*/
650 int8_t (*intra4x4_pred_mode)[8]; /* intra4x4 pred mode. for non I4x4 set to I_PRED_4x4_DC(2) */
651 /* actually has only 7 entries; set to 8 for write-combining optimizations */
652 uint8_t (*non_zero_count)[16+4+4]; /* nzc. for I_PCM set to 16 */
653 int8_t *chroma_pred_mode; /* chroma_pred_mode. cabac only. for non intra I_PRED_CHROMA_DC(0) */
654 int16_t (*mv[2])[2]; /* mb mv. set to 0 for intra mb */
655 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 */
656 int8_t *ref[2]; /* mb ref. set to -1 if non used (intra or Lx only) */
657 int16_t (*mvr[2][X264_REF_MAX*2])[2];/* 16x16 mv for each possible ref */
658 int8_t *skipbp; /* block pattern for SKIP or DIRECT (sub)mbs. B-frames + cabac only */
659 int8_t *mb_transform_size; /* transform_size_8x8_flag of each mb */
660 uint16_t *slice_table; /* sh->first_mb of the slice that the indexed mb is part of
661 * NOTE: this will fail on resolutions above 2^16 MBs... */
664 /* buffer for weighted versions of the reference frames */
665 pixel *p_weight_buf[X264_REF_MAX];
670 ALIGNED_4( uint8_t i_sub_partition[4] );
676 int i_intra16x16_pred_mode;
677 int i_chroma_pred_mode;
679 /* skip flags for i4x4 and i8x8
680 * 0 = encode as normal.
681 * 1 (non-RD only) = the DCT is still in h->dct, restore fdec and skip reconstruction.
682 * 2 (RD only) = the DCT has since been overwritten by RD; restore that too. */
684 /* skip flag for motion compensation */
685 /* if we've already done MC, we don't need to do it again */
687 /* set to true if we are re-encoding a macroblock. */
689 int ip_offset; /* Used by PIR to offset the quantizer of intra-refresh blocks. */
691 int b_overflow; /* If CAVLC had a level code overflow during bitstream writing. */
695 /* space for p_fenc and p_fdec */
696 #define FENC_STRIDE 16
697 #define FDEC_STRIDE 32
698 ALIGNED_16( pixel fenc_buf[24*FENC_STRIDE] );
699 ALIGNED_16( pixel fdec_buf[27*FDEC_STRIDE] );
701 /* i4x4 and i8x8 backup data, for skipping the encode stage when possible */
702 ALIGNED_16( pixel i4x4_fdec_buf[16*16] );
703 ALIGNED_16( pixel i8x8_fdec_buf[16*16] );
704 ALIGNED_16( dctcoef i8x8_dct_buf[3][64] );
705 ALIGNED_16( dctcoef i4x4_dct_buf[15][16] );
706 uint32_t i4x4_nnz_buf[4];
707 uint32_t i8x8_nnz_buf[4];
711 /* Psy trellis DCT data */
712 ALIGNED_16( dctcoef fenc_dct8[4][64] );
713 ALIGNED_16( dctcoef fenc_dct4[16][16] );
715 /* Psy RD SATD/SA8D scores cache */
716 ALIGNED_16( uint64_t fenc_hadamard_cache[9] );
717 ALIGNED_16( uint32_t fenc_satd_cache[32] );
719 /* pointer over mb of the frame to be compressed */
720 pixel *p_fenc[3]; /* y,u,v */
721 /* pointer to the actual source frame, not a block copy */
722 pixel *p_fenc_plane[2]; /* y,uv */
724 /* pointer over mb of the frame to be reconstructed */
727 /* pointer over mb of the references */
729 pixel *p_fref[2][X264_REF_MAX*2][4+1]; /* last: yN, yH, yV, yHV, uv */
730 pixel *p_fref_w[X264_REF_MAX*2]; /* weighted fullpel luma */
731 uint16_t *p_integral[2][X264_REF_MAX];
740 /* real intra4x4_pred_mode if I_4X4 or I_8X8, I_PRED_4x4_DC if mb available, -1 if not */
741 ALIGNED_8( int8_t intra4x4_pred_mode[X264_SCAN8_LUMA_SIZE] );
743 /* i_non_zero_count if available else 0x80 */
744 ALIGNED_16( uint8_t non_zero_count[X264_SCAN8_SIZE] );
746 /* -1 if unused, -2 if unavailable */
747 ALIGNED_4( int8_t ref[2][X264_SCAN8_LUMA_SIZE] );
749 /* 0 if not available */
750 ALIGNED_16( int16_t mv[2][X264_SCAN8_LUMA_SIZE][2] );
751 ALIGNED_8( uint8_t mvd[2][X264_SCAN8_LUMA_SIZE][2] );
753 /* 1 if SKIP or DIRECT. set only for B-frames + CABAC */
754 ALIGNED_4( int8_t skip[X264_SCAN8_LUMA_SIZE] );
756 ALIGNED_4( int16_t direct_mv[2][4][2] );
757 ALIGNED_4( int8_t direct_ref[2][4] );
758 int direct_partition;
759 ALIGNED_4( int16_t pskip_mv[2] );
761 /* number of neighbors (top and left) that used 8x8 dct */
762 int i_neighbour_transform_size;
763 int i_neighbour_skip;
769 /* extra data required for mbaff in mv prediction */
770 int16_t topright_mv[2][3][2];
771 int8_t topright_ref[2][3];
775 int i_qp; /* current qp */
777 int i_last_qp; /* last qp */
778 int i_last_dqp; /* last delta qp */
779 int b_variable_qp; /* whether qp is allowed to vary per macroblock */
781 int b_direct_auto_read; /* take stats for --direct auto from the 2pass log */
782 int b_direct_auto_write; /* analyse direct modes, to use and/or save */
785 int i_trellis_lambda2[2][2]; /* [luma,chroma][inter,intra] */
787 int i_chroma_lambda2_offset;
789 /* B_direct and weighted prediction */
790 int16_t dist_scale_factor_buf[2][2][X264_REF_MAX*2][4];
791 int16_t (*dist_scale_factor)[4];
792 int8_t bipred_weight_buf[2][2][X264_REF_MAX*2][4];
793 int8_t (*bipred_weight)[4];
794 /* maps fref1[0]'s ref indices into the current list0 */
795 #define map_col_to_list0(col) h->mb.map_col_to_list0[(col)+2]
796 int8_t map_col_to_list0[X264_REF_MAX+2];
797 int ref_blind_dupe; /* The index of the blind reference frame duplicate. */
798 int8_t deblock_ref_table[X264_REF_MAX*2+2];
799 #define deblock_ref_table(x) h->mb.deblock_ref_table[(x)+2]
802 /* rate control encoding only */
803 x264_ratecontrol_t *rc;
808 /* Current frame stats */
811 /* MV bits (MV+Ref+Block Type) */
813 /* Texture bits (DCT coefs) */
822 int i_mb_count_8x8dct[2];
823 int i_mb_count_ref[2][X264_REF_MAX*2];
824 int i_mb_partition[17];
826 int i_mb_pred_mode[4][13];
828 /* Adaptive direct mv pred */
829 int i_direct_score[2];
835 /* Cumulated stats */
838 int i_frame_count[3];
839 int64_t i_frame_size[3];
840 double f_frame_qp[3];
841 int i_consecutive_bframes[X264_BFRAME_MAX+1];
843 double f_ssd_global[3];
844 double f_psnr_average[3];
845 double f_psnr_mean_y[3];
846 double f_psnr_mean_u[3];
847 double f_psnr_mean_v[3];
848 double f_ssim_mean_y[3];
849 double f_frame_duration[3];
851 int64_t i_mb_count[3][19];
852 int64_t i_mb_partition[2][17];
853 int64_t i_mb_count_8x8dct[2];
854 int64_t i_mb_count_ref[2][2][X264_REF_MAX*2];
856 int64_t i_mb_pred_mode[4][13];
857 int64_t i_mb_field[3];
859 int i_direct_score[2];
860 int i_direct_frames[2];
861 /* num p-frames weighted */
866 /* 0 = luma 4x4, 1 = luma 8x8, 2 = chroma 4x4 */
867 udctcoef (*nr_offset)[64];
868 uint32_t (*nr_residual_sum)[64];
871 ALIGNED_16( udctcoef nr_offset_denoise[3][64] );
872 ALIGNED_16( uint32_t nr_residual_sum_buf[2][3][64] );
873 uint32_t nr_count_buf[2][3];
875 /* Buffers that are allocated per-thread even in sliced threads. */
876 void *scratch_buffer; /* for any temporary storage that doesn't want repeated malloc */
877 pixel *intra_border_backup[5][2]; /* bottom pixels of the previous mb row, used for intra prediction after the framebuffer has been deblocked */
878 /* Deblock strength values are stored for each 4x4 partition. In MBAFF
879 * there are four extra values that need to be stored, located in [4][i]. */
880 uint8_t (*deblock_strength[2])[2][8][4];
882 /* CPU functions dependents */
883 x264_predict_t predict_16x16[4+3];
884 x264_predict_t predict_8x8c[4+3];
885 x264_predict8x8_t predict_8x8[9+3];
886 x264_predict_t predict_4x4[9+3];
887 x264_predict_8x8_filter_t predict_8x8_filter;
889 x264_pixel_function_t pixf;
890 x264_mc_functions_t mc;
891 x264_dct_function_t dctf;
892 x264_zigzag_function_t zigzagf;
893 x264_zigzag_function_t zigzagf_interlaced;
894 x264_zigzag_function_t zigzagf_progressive;
895 x264_quant_function_t quantf;
896 x264_deblock_function_t loopf;
897 x264_bitstream_function_t bsf;
900 struct visualize_t *visualize;
902 x264_lookahead_t *lookahead;
905 // included at the end because it needs x264_t
906 #include "macroblock.h"
908 #if ARCH_X86 || ARCH_X86_64
909 #include "x86/util.h"
912 #include "rectangle.h"