1 /*****************************************************************************
2 * macroblock.h: h264 encoder library
3 *****************************************************************************
4 * Copyright (C) 2005-2008 x264 project
6 * Authors: Loren Merritt <lorenm@u.washington.edu>
7 * Laurent Aimar <fenrir@via.ecp.fr>
8 * Fiona Glaser <fiona@x264.com>
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License as published by
12 * the Free Software Foundation; either version 2 of the License, or
13 * (at your option) any later version.
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
18 * GNU General Public License for more details.
20 * You should have received a copy of the GNU General Public License
21 * along with this program; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
23 *****************************************************************************/
25 #ifndef X264_MACROBLOCK_H
26 #define X264_MACROBLOCK_H
28 enum macroblock_position_e
40 static const uint8_t x264_pred_i4x4_neighbors[12] =
42 MB_TOP, // I_PRED_4x4_V
43 MB_LEFT, // I_PRED_4x4_H
44 MB_LEFT | MB_TOP, // I_PRED_4x4_DC
45 MB_TOP | MB_TOPRIGHT, // I_PRED_4x4_DDL
46 MB_LEFT | MB_TOPLEFT | MB_TOP, // I_PRED_4x4_DDR
47 MB_LEFT | MB_TOPLEFT | MB_TOP, // I_PRED_4x4_VR
48 MB_LEFT | MB_TOPLEFT | MB_TOP, // I_PRED_4x4_HD
49 MB_TOP | MB_TOPRIGHT, // I_PRED_4x4_VL
50 MB_LEFT, // I_PRED_4x4_HU
51 MB_LEFT, // I_PRED_4x4_DC_LEFT
52 MB_TOP, // I_PRED_4x4_DC_TOP
53 0 // I_PRED_4x4_DC_128
57 /* XXX mb_type isn't the one written in the bitstream -> only internal usage */
58 #define IS_INTRA(type) ( (type) == I_4x4 || (type) == I_8x8 || (type) == I_16x16 || (type) == I_PCM )
59 #define IS_SKIP(type) ( (type) == P_SKIP || (type) == B_SKIP )
60 #define IS_DIRECT(type) ( (type) == B_DIRECT )
87 static const uint8_t x264_mb_type_fix[X264_MBTYPE_MAX] =
89 I_4x4, I_4x4, I_16x16, I_PCM,
91 B_DIRECT, B_L0_L0, B_L0_L1, B_L0_BI, B_L1_L0, B_L1_L1,
92 B_L1_BI, B_BI_L0, B_BI_L1, B_BI_BI, B_8x8, B_SKIP
94 static const uint8_t x264_mb_type_list0_table[X264_MBTYPE_MAX][2] =
96 {0,0}, {0,0}, {0,0}, {0,0}, /* INTRA */
100 {0,0}, /* B_DIRECT */
101 {1,1}, {1,0}, {1,1}, /* B_L0_* */
102 {0,1}, {0,0}, {0,1}, /* B_L1_* */
103 {1,1}, {1,0}, {1,1}, /* B_BI_* */
107 static const uint8_t x264_mb_type_list1_table[X264_MBTYPE_MAX][2] =
109 {0,0}, {0,0}, {0,0}, {0,0}, /* INTRA */
113 {0,0}, /* B_DIRECT */
114 {0,0}, {0,1}, {0,1}, /* B_L0_* */
115 {1,0}, {1,1}, {1,1}, /* B_L1_* */
116 {1,0}, {1,1}, {1,1}, /* B_BI_* */
121 #define IS_SUB4x4(type) ( (type ==D_L0_4x4)||(type ==D_L1_4x4)||(type ==D_BI_4x4))
122 #define IS_SUB4x8(type) ( (type ==D_L0_4x8)||(type ==D_L1_4x8)||(type ==D_BI_4x8))
123 #define IS_SUB8x4(type) ( (type ==D_L0_8x4)||(type ==D_L1_8x4)||(type ==D_BI_8x4))
124 #define IS_SUB8x8(type) ( (type ==D_L0_8x8)||(type ==D_L1_8x8)||(type ==D_BI_8x8)||(type ==D_DIRECT_8x8))
127 /* sub partition type for P_8x8 and B_8x8 */
133 /* sub partition type for B_8x8 only */
150 X264_PARTTYPE_MAX = 17,
153 static const uint8_t x264_mb_partition_listX_table[2][17] =
155 1, 1, 1, 1, /* D_L0_* */
156 0, 0, 0, 0, /* D_L1_* */
157 1, 1, 1, 1, /* D_BI_* */
158 0, /* D_DIRECT_8x8 */
159 0, 0, 0, 0 /* 8x8 .. 16x16 */
162 0, 0, 0, 0, /* D_L0_* */
163 1, 1, 1, 1, /* D_L1_* */
164 1, 1, 1, 1, /* D_BI_* */
165 0, /* D_DIRECT_8x8 */
166 0, 0, 0, 0 /* 8x8 .. 16x16 */
168 static const uint8_t x264_mb_partition_count_table[17] =
181 static const uint8_t x264_mb_partition_pixel_table[17] =
183 6, 4, 5, 3, 6, 4, 5, 3, 6, 4, 5, 3, 3, 3, 1, 2, 0
186 /* zigzags are transposed with respect to the tables in the standard */
187 static const uint8_t x264_zigzag_scan4[2][16] =
189 0, 4, 1, 2, 5, 8, 12, 9, 6, 3, 7, 10, 13, 14, 11, 15
192 0, 1, 4, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
194 static const uint8_t x264_zigzag_scan8[2][64] =
196 0, 8, 1, 2, 9, 16, 24, 17, 10, 3, 4, 11, 18, 25, 32, 40,
197 33, 26, 19, 12, 5, 6, 13, 20, 27, 34, 41, 48, 56, 49, 42, 35,
198 28, 21, 14, 7, 15, 22, 29, 36, 43, 50, 57, 58, 51, 44, 37, 30,
199 23, 31, 38, 45, 52, 59, 60, 53, 46, 39, 47, 54, 61, 62, 55, 63
202 0, 1, 2, 8, 9, 3, 4, 10, 16, 11, 5, 6, 7, 12, 17, 24,
203 18, 13, 14, 15, 19, 25, 32, 26, 20, 21, 22, 23, 27, 33, 40, 34,
204 28, 29, 30, 31, 35, 41, 48, 42, 36, 37, 38, 39, 43, 49, 50, 44,
205 45, 46, 47, 51, 56, 57, 52, 53, 54, 55, 58, 59, 60, 61, 62, 63
208 static const uint8_t block_idx_x[16] =
210 0, 1, 0, 1, 2, 3, 2, 3, 0, 1, 0, 1, 2, 3, 2, 3
212 static const uint8_t block_idx_y[16] =
214 0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3, 2, 2, 3, 3
216 static const uint8_t block_idx_xy[4][4] =
223 static const uint8_t block_idx_xy_1d[16] =
225 0, 1, 4, 5, 2, 3, 6, 7, 8, 9, 12, 13, 10, 11, 14, 15
227 static const uint8_t block_idx_yx_1d[16] =
229 0, 4, 1, 5, 8, 12, 9, 13, 2, 6, 3, 7, 10, 14, 11, 15
231 static const uint8_t block_idx_xy_fenc[16] =
233 0*4 + 0*4*FENC_STRIDE, 1*4 + 0*4*FENC_STRIDE,
234 0*4 + 1*4*FENC_STRIDE, 1*4 + 1*4*FENC_STRIDE,
235 2*4 + 0*4*FENC_STRIDE, 3*4 + 0*4*FENC_STRIDE,
236 2*4 + 1*4*FENC_STRIDE, 3*4 + 1*4*FENC_STRIDE,
237 0*4 + 2*4*FENC_STRIDE, 1*4 + 2*4*FENC_STRIDE,
238 0*4 + 3*4*FENC_STRIDE, 1*4 + 3*4*FENC_STRIDE,
239 2*4 + 2*4*FENC_STRIDE, 3*4 + 2*4*FENC_STRIDE,
240 2*4 + 3*4*FENC_STRIDE, 3*4 + 3*4*FENC_STRIDE
242 static const uint16_t block_idx_xy_fdec[16] =
244 0*4 + 0*4*FDEC_STRIDE, 1*4 + 0*4*FDEC_STRIDE,
245 0*4 + 1*4*FDEC_STRIDE, 1*4 + 1*4*FDEC_STRIDE,
246 2*4 + 0*4*FDEC_STRIDE, 3*4 + 0*4*FDEC_STRIDE,
247 2*4 + 1*4*FDEC_STRIDE, 3*4 + 1*4*FDEC_STRIDE,
248 0*4 + 2*4*FDEC_STRIDE, 1*4 + 2*4*FDEC_STRIDE,
249 0*4 + 3*4*FDEC_STRIDE, 1*4 + 3*4*FDEC_STRIDE,
250 2*4 + 2*4*FDEC_STRIDE, 3*4 + 2*4*FDEC_STRIDE,
251 2*4 + 3*4*FDEC_STRIDE, 3*4 + 3*4*FDEC_STRIDE
254 static const uint8_t i_chroma_qp_table[52+12*2] =
256 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
257 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,
258 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
259 20, 21, 22, 23, 24, 25, 26, 27, 28, 29,
260 29, 30, 31, 32, 32, 33, 34, 34, 35, 35,
261 36, 36, 37, 37, 37, 38, 38, 38, 39, 39,
263 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39,
266 enum cabac_ctx_block_cat_e
277 int x264_macroblock_cache_init( x264_t *h );
278 void x264_macroblock_slice_init( x264_t *h );
279 void x264_macroblock_cache_load( x264_t *h, int i_mb_x, int i_mb_y );
280 void x264_macroblock_cache_save( x264_t *h );
281 void x264_macroblock_cache_end( x264_t *h );
283 void x264_macroblock_bipred_init( x264_t *h );
285 void x264_prefetch_fenc( x264_t *h, x264_frame_t *fenc, int i_mb_x, int i_mb_y );
287 /* x264_mb_predict_mv_16x16:
288 * set mvp with predicted mv for D_16x16 block
289 * h->mb. need only valid values from other blocks */
290 void x264_mb_predict_mv_16x16( x264_t *h, int i_list, int i_ref, int16_t mvp[2] );
291 /* x264_mb_predict_mv_pskip:
292 * set mvp with predicted mv for P_SKIP
293 * h->mb. need only valid values from other blocks */
294 void x264_mb_predict_mv_pskip( x264_t *h, int16_t mv[2] );
295 /* x264_mb_predict_mv:
296 * set mvp with predicted mv for all blocks except SKIP and DIRECT
297 * h->mb. need valid ref/partition/sub of current block to be valid
298 * and valid mv/ref from other blocks. */
299 void x264_mb_predict_mv( x264_t *h, int i_list, int idx, int i_width, int16_t mvp[2] );
300 /* x264_mb_predict_mv_direct16x16:
301 * set h->mb.cache.mv and h->mb.cache.ref for B_SKIP or B_DIRECT
302 * h->mb. need only valid values from other blocks.
303 * return 1 on success, 0 on failure.
304 * if b_changed != NULL, set it to whether refs or mvs differ from
305 * before this functioncall. */
306 int x264_mb_predict_mv_direct16x16( x264_t *h, int *b_changed );
307 /* x264_mb_load_mv_direct8x8:
308 * set h->mb.cache.mv and h->mb.cache.ref for B_DIRECT
309 * must be called only after x264_mb_predict_mv_direct16x16 */
310 void x264_mb_load_mv_direct8x8( x264_t *h, int idx );
311 /* x264_mb_predict_mv_ref16x16:
312 * set mvc with D_16x16 prediction.
313 * uses all neighbors, even those that didn't end up using this ref.
314 * h->mb. need only valid values from other blocks */
315 void x264_mb_predict_mv_ref16x16( x264_t *h, int i_list, int i_ref, int16_t mvc[8][2], int *i_mvc );
317 void x264_mb_mc( x264_t *h );
318 void x264_mb_mc_8x8( x264_t *h, int i8 );
320 static ALWAYS_INLINE uint32_t pack16to32( int a, int b )
322 #ifdef WORDS_BIGENDIAN
328 static ALWAYS_INLINE uint32_t pack8to16( int a, int b )
330 #ifdef WORDS_BIGENDIAN
336 static ALWAYS_INLINE uint32_t pack8to32( int a, int b, int c, int d )
338 #ifdef WORDS_BIGENDIAN
339 return d + (c<<8) + (b<<16) + (a<<24);
341 return a + (b<<8) + (c<<16) + (d<<24);
344 static ALWAYS_INLINE uint32_t pack16to32_mask( int a, int b )
346 #ifdef WORDS_BIGENDIAN
347 return (b&0xFFFF) + (a<<16);
349 return (a&0xFFFF) + (b<<16);
352 static ALWAYS_INLINE void x264_macroblock_cache_rect1( void *dst, int width, int height, uint8_t val )
357 uint32_t val2 = val * 0x01010101;
358 for( dy = 0; dy < height; dy++ )
359 ((uint32_t*)dst)[2*dy] = val2;
363 uint32_t val2 = val * 0x0101;
364 for( dy = 0; dy < height; dy++ )
365 ((uint16_t*)dst)[4*dy] = val2;
368 static ALWAYS_INLINE void x264_macroblock_cache_rect4( void *dst, int width, int height, uint32_t val )
371 if( width == 1 || WORD_SIZE < 8 )
373 for( dy = 0; dy < height; dy++ )
374 for( dx = 0; dx < width; dx++ )
375 ((uint32_t*)dst)[dx+8*dy] = val;
379 uint64_t val64 = val + ((uint64_t)val<<32);
380 for( dy = 0; dy < height; dy++ )
381 for( dx = 0; dx < width/2; dx++ )
382 ((uint64_t*)dst)[dx+4*dy] = val64;
385 #define x264_macroblock_cache_mv_ptr(a,x,y,w,h,l,mv) x264_macroblock_cache_mv(a,x,y,w,h,l,*(uint32_t*)mv)
386 static ALWAYS_INLINE void x264_macroblock_cache_mv( x264_t *h, int x, int y, int width, int height, int i_list, uint32_t mv )
388 x264_macroblock_cache_rect4( &h->mb.cache.mv[i_list][X264_SCAN8_0+x+8*y], width, height, mv );
390 static ALWAYS_INLINE void x264_macroblock_cache_mvd( x264_t *h, int x, int y, int width, int height, int i_list, uint32_t mv )
392 x264_macroblock_cache_rect4( &h->mb.cache.mvd[i_list][X264_SCAN8_0+x+8*y], width, height, mv );
394 static ALWAYS_INLINE void x264_macroblock_cache_ref( x264_t *h, int x, int y, int width, int height, int i_list, uint8_t ref )
396 x264_macroblock_cache_rect1( &h->mb.cache.ref[i_list][X264_SCAN8_0+x+8*y], width, height, ref );
398 static ALWAYS_INLINE void x264_macroblock_cache_skip( x264_t *h, int x, int y, int width, int height, int b_skip )
400 x264_macroblock_cache_rect1( &h->mb.cache.skip[X264_SCAN8_0+x+8*y], width, height, b_skip );
402 static ALWAYS_INLINE void x264_macroblock_cache_intra8x8_pred( x264_t *h, int x, int y, int i_mode )
404 int8_t *cache = &h->mb.cache.intra4x4_pred_mode[X264_SCAN8_0+x+8*y];
405 cache[0] = cache[1] = cache[8] = cache[9] = i_mode;
407 #define array_non_zero(a) array_non_zero_int(a, sizeof(a))
408 #define array_non_zero_int array_non_zero_int_c
409 static ALWAYS_INLINE int array_non_zero_int_c( void *v, int i_count )
414 else if(i_count == 16)
415 return !!(x[0]|x[1]);
416 else if(i_count == 32)
417 return !!(x[0]|x[1]|x[2]|x[3]);
421 i_count /= sizeof(uint64_t);
422 for( i = 0; i < i_count; i++ )
427 static inline int x264_mb_predict_intra4x4_mode( x264_t *h, int idx )
429 const int ma = h->mb.cache.intra4x4_pred_mode[x264_scan8[idx] - 1];
430 const int mb = h->mb.cache.intra4x4_pred_mode[x264_scan8[idx] - 8];
431 const int m = X264_MIN( x264_mb_pred_mode4x4_fix(ma),
432 x264_mb_pred_mode4x4_fix(mb) );
435 return I_PRED_4x4_DC;
439 static inline int x264_mb_predict_non_zero_code( x264_t *h, int idx )
441 const int za = h->mb.cache.non_zero_count[x264_scan8[idx] - 1];
442 const int zb = h->mb.cache.non_zero_count[x264_scan8[idx] - 8];
448 i_ret = ( i_ret + 1 ) >> 1;
452 /* x264_mb_transform_8x8_allowed:
453 * check whether any partition is smaller than 8x8 (or at least
454 * might be, according to just partition type.)
455 * doesn't check for cbp */
456 static inline int x264_mb_transform_8x8_allowed( x264_t *h )
458 // intra and skip are disallowed
459 // large partitions are allowed
460 // direct and 8x8 are conditional
461 static const uint8_t partition_tab[X264_MBTYPE_MAX] = {
462 0,0,0,0,1,2,0,2,1,1,1,1,1,1,1,1,1,2,0,
466 if( !h->pps->b_transform_8x8_mode )
468 p = partition_tab[h->mb.i_type];
471 else if( h->mb.i_type == B_DIRECT )
472 return h->sps->b_direct8x8_inference;
473 else if( h->mb.i_type == P_8x8 )
475 if( !(h->param.analyse.inter & X264_ANALYSE_PSUB8x8) )
478 if( h->mb.i_sub_partition[i] != D_L0_8x8 )
484 // x264 currently doesn't use sub-8x8 B partitions, so don't check for them
485 if( h->sps->b_direct8x8_inference )
488 if( h->mb.i_sub_partition[i] == D_DIRECT_8x8 )