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_list_table[X264_MBTYPE_MAX][2][2] =
96 {{0,0},{0,0}}, {{0,0},{0,0}}, {{0,0},{0,0}}, {{0,0},{0,0}}, /* INTRA */
97 {{1,1},{0,0}}, /* P_L0 */
98 {{0,0},{0,0}}, /* P_8x8 */
99 {{1,1},{0,0}}, /* P_SKIP */
100 {{0,0},{0,0}}, /* B_DIRECT */
101 {{1,1},{0,0}}, {{1,0},{0,1}}, {{1,1},{0,1}}, /* B_L0_* */
102 {{0,1},{1,0}}, {{0,0},{1,1}}, {{0,1},{1,1}}, /* B_L1_* */
103 {{1,1},{1,0}}, {{1,0},{1,1}}, {{1,1},{1,1}}, /* B_BI_* */
104 {{0,0},{0,0}}, /* B_8x8 */
105 {{0,0},{0,0}} /* B_SKIP */
108 #define IS_SUB4x4(type) ( (type ==D_L0_4x4)||(type ==D_L1_4x4)||(type ==D_BI_4x4))
109 #define IS_SUB4x8(type) ( (type ==D_L0_4x8)||(type ==D_L1_4x8)||(type ==D_BI_4x8))
110 #define IS_SUB8x4(type) ( (type ==D_L0_8x4)||(type ==D_L1_8x4)||(type ==D_BI_8x4))
111 #define IS_SUB8x8(type) ( (type ==D_L0_8x8)||(type ==D_L1_8x8)||(type ==D_BI_8x8)||(type ==D_DIRECT_8x8))
114 /* sub partition type for P_8x8 and B_8x8 */
120 /* sub partition type for B_8x8 only */
137 X264_PARTTYPE_MAX = 17,
140 static const uint8_t x264_mb_partition_listX_table[2][17] =
142 1, 1, 1, 1, /* D_L0_* */
143 0, 0, 0, 0, /* D_L1_* */
144 1, 1, 1, 1, /* D_BI_* */
145 0, /* D_DIRECT_8x8 */
146 0, 0, 0, 0 /* 8x8 .. 16x16 */
149 0, 0, 0, 0, /* D_L0_* */
150 1, 1, 1, 1, /* D_L1_* */
151 1, 1, 1, 1, /* D_BI_* */
152 0, /* D_DIRECT_8x8 */
153 0, 0, 0, 0 /* 8x8 .. 16x16 */
155 static const uint8_t x264_mb_partition_count_table[17] =
168 static const uint8_t x264_mb_partition_pixel_table[17] =
170 6, 4, 5, 3, 6, 4, 5, 3, 6, 4, 5, 3, 3, 3, 1, 2, 0
173 /* zigzags are transposed with respect to the tables in the standard */
174 static const uint8_t x264_zigzag_scan4[2][16] =
176 0, 4, 1, 2, 5, 8, 12, 9, 6, 3, 7, 10, 13, 14, 11, 15
179 0, 1, 4, 2, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
181 static const uint8_t x264_zigzag_scan8[2][64] =
183 0, 8, 1, 2, 9, 16, 24, 17, 10, 3, 4, 11, 18, 25, 32, 40,
184 33, 26, 19, 12, 5, 6, 13, 20, 27, 34, 41, 48, 56, 49, 42, 35,
185 28, 21, 14, 7, 15, 22, 29, 36, 43, 50, 57, 58, 51, 44, 37, 30,
186 23, 31, 38, 45, 52, 59, 60, 53, 46, 39, 47, 54, 61, 62, 55, 63
189 0, 1, 2, 8, 9, 3, 4, 10, 16, 11, 5, 6, 7, 12, 17, 24,
190 18, 13, 14, 15, 19, 25, 32, 26, 20, 21, 22, 23, 27, 33, 40, 34,
191 28, 29, 30, 31, 35, 41, 48, 42, 36, 37, 38, 39, 43, 49, 50, 44,
192 45, 46, 47, 51, 56, 57, 52, 53, 54, 55, 58, 59, 60, 61, 62, 63
195 static const uint8_t block_idx_x[16] =
197 0, 1, 0, 1, 2, 3, 2, 3, 0, 1, 0, 1, 2, 3, 2, 3
199 static const uint8_t block_idx_y[16] =
201 0, 0, 1, 1, 0, 0, 1, 1, 2, 2, 3, 3, 2, 2, 3, 3
203 static const uint8_t block_idx_xy[4][4] =
210 static const uint8_t block_idx_xy_1d[16] =
212 0, 1, 4, 5, 2, 3, 6, 7, 8, 9, 12, 13, 10, 11, 14, 15
214 static const uint8_t block_idx_yx_1d[16] =
216 0, 4, 1, 5, 8, 12, 9, 13, 2, 6, 3, 7, 10, 14, 11, 15
218 static const uint8_t block_idx_xy_fenc[16] =
220 0*4 + 0*4*FENC_STRIDE, 1*4 + 0*4*FENC_STRIDE,
221 0*4 + 1*4*FENC_STRIDE, 1*4 + 1*4*FENC_STRIDE,
222 2*4 + 0*4*FENC_STRIDE, 3*4 + 0*4*FENC_STRIDE,
223 2*4 + 1*4*FENC_STRIDE, 3*4 + 1*4*FENC_STRIDE,
224 0*4 + 2*4*FENC_STRIDE, 1*4 + 2*4*FENC_STRIDE,
225 0*4 + 3*4*FENC_STRIDE, 1*4 + 3*4*FENC_STRIDE,
226 2*4 + 2*4*FENC_STRIDE, 3*4 + 2*4*FENC_STRIDE,
227 2*4 + 3*4*FENC_STRIDE, 3*4 + 3*4*FENC_STRIDE
229 static const uint16_t block_idx_xy_fdec[16] =
231 0*4 + 0*4*FDEC_STRIDE, 1*4 + 0*4*FDEC_STRIDE,
232 0*4 + 1*4*FDEC_STRIDE, 1*4 + 1*4*FDEC_STRIDE,
233 2*4 + 0*4*FDEC_STRIDE, 3*4 + 0*4*FDEC_STRIDE,
234 2*4 + 1*4*FDEC_STRIDE, 3*4 + 1*4*FDEC_STRIDE,
235 0*4 + 2*4*FDEC_STRIDE, 1*4 + 2*4*FDEC_STRIDE,
236 0*4 + 3*4*FDEC_STRIDE, 1*4 + 3*4*FDEC_STRIDE,
237 2*4 + 2*4*FDEC_STRIDE, 3*4 + 2*4*FDEC_STRIDE,
238 2*4 + 3*4*FDEC_STRIDE, 3*4 + 3*4*FDEC_STRIDE
241 static const uint8_t i_chroma_qp_table[52+12*2] =
243 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
244 0, 1, 2, 3, 4, 5, 6, 7, 8, 9,
245 10, 11, 12, 13, 14, 15, 16, 17, 18, 19,
246 20, 21, 22, 23, 24, 25, 26, 27, 28, 29,
247 29, 30, 31, 32, 32, 33, 34, 34, 35, 35,
248 36, 36, 37, 37, 37, 38, 38, 38, 39, 39,
250 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39, 39,
253 enum cabac_ctx_block_cat_e
264 int x264_macroblock_cache_init( x264_t *h );
265 void x264_macroblock_slice_init( x264_t *h );
266 void x264_macroblock_cache_load( x264_t *h, int i_mb_x, int i_mb_y );
267 void x264_macroblock_cache_save( x264_t *h );
268 void x264_macroblock_cache_end( x264_t *h );
270 void x264_macroblock_bipred_init( x264_t *h );
272 void x264_prefetch_fenc( x264_t *h, x264_frame_t *fenc, int i_mb_x, int i_mb_y );
274 /* x264_mb_predict_mv_16x16:
275 * set mvp with predicted mv for D_16x16 block
276 * h->mb. need only valid values from other blocks */
277 void x264_mb_predict_mv_16x16( x264_t *h, int i_list, int i_ref, int16_t mvp[2] );
278 /* x264_mb_predict_mv_pskip:
279 * set mvp with predicted mv for P_SKIP
280 * h->mb. need only valid values from other blocks */
281 void x264_mb_predict_mv_pskip( x264_t *h, int16_t mv[2] );
282 /* x264_mb_predict_mv:
283 * set mvp with predicted mv for all blocks except SKIP and DIRECT
284 * h->mb. need valid ref/partition/sub of current block to be valid
285 * and valid mv/ref from other blocks. */
286 void x264_mb_predict_mv( x264_t *h, int i_list, int idx, int i_width, int16_t mvp[2] );
287 /* x264_mb_predict_mv_direct16x16:
288 * set h->mb.cache.mv and h->mb.cache.ref for B_SKIP or B_DIRECT
289 * h->mb. need only valid values from other blocks.
290 * return 1 on success, 0 on failure.
291 * if b_changed != NULL, set it to whether refs or mvs differ from
292 * before this functioncall. */
293 int x264_mb_predict_mv_direct16x16( x264_t *h, int *b_changed );
294 /* x264_mb_load_mv_direct8x8:
295 * set h->mb.cache.mv and h->mb.cache.ref for B_DIRECT
296 * must be called only after x264_mb_predict_mv_direct16x16 */
297 void x264_mb_load_mv_direct8x8( x264_t *h, int idx );
298 /* x264_mb_predict_mv_ref16x16:
299 * set mvc with D_16x16 prediction.
300 * uses all neighbors, even those that didn't end up using this ref.
301 * h->mb. need only valid values from other blocks */
302 void x264_mb_predict_mv_ref16x16( x264_t *h, int i_list, int i_ref, int16_t mvc[8][2], int *i_mvc );
304 void x264_mb_mc( x264_t *h );
305 void x264_mb_mc_8x8( x264_t *h, int i8 );
307 static ALWAYS_INLINE uint32_t pack16to32( int a, int b )
309 #ifdef WORDS_BIGENDIAN
315 static ALWAYS_INLINE uint32_t pack8to16( int a, int b )
317 #ifdef WORDS_BIGENDIAN
323 static ALWAYS_INLINE uint32_t pack8to32( int a, int b, int c, int d )
325 #ifdef WORDS_BIGENDIAN
326 return d + (c<<8) + (b<<16) + (a<<24);
328 return a + (b<<8) + (c<<16) + (d<<24);
331 static ALWAYS_INLINE uint32_t pack16to32_mask( int a, int b )
333 #ifdef WORDS_BIGENDIAN
334 return (b&0xFFFF) + (a<<16);
336 return (a&0xFFFF) + (b<<16);
339 static ALWAYS_INLINE void x264_macroblock_cache_rect1( void *dst, int width, int height, uint8_t val )
343 uint32_t val2 = val * 0x01010101;
344 ((uint32_t*)dst)[0] = val2;
345 if( height >= 2 ) ((uint32_t*)dst)[2] = val2;
346 if( height == 4 ) ((uint32_t*)dst)[4] = val2;
347 if( height == 4 ) ((uint32_t*)dst)[6] = val2;
351 uint32_t val2 = val * 0x0101;
352 ((uint16_t*)dst)[ 0] = val2;
353 if( height >= 2 ) ((uint16_t*)dst)[ 4] = val2;
354 if( height == 4 ) ((uint16_t*)dst)[ 8] = val2;
355 if( height == 4 ) ((uint16_t*)dst)[12] = val2;
358 static ALWAYS_INLINE void x264_macroblock_cache_rect4( void *dst, int width, int height, uint32_t val )
361 if( width == 1 || WORD_SIZE < 8 )
363 for( dy = 0; dy < height; dy++ )
365 ((uint32_t*)dst)[8*dy+0] = val;
366 if( width >= 2 ) ((uint32_t*)dst)[8*dy+1] = val;
367 if( width == 4 ) ((uint32_t*)dst)[8*dy+2] = val;
368 if( width == 4 ) ((uint32_t*)dst)[8*dy+3] = val;
373 uint64_t val64 = val + ((uint64_t)val<<32);
374 for( dy = 0; dy < height; dy++ )
376 ((uint64_t*)dst)[4*dy+0] = val64;
377 if( width == 4 ) ((uint64_t*)dst)[4*dy+1] = val64;
381 #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)
382 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 )
384 x264_macroblock_cache_rect4( &h->mb.cache.mv[i_list][X264_SCAN8_0+x+8*y], width, height, mv );
386 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 )
388 x264_macroblock_cache_rect4( &h->mb.cache.mvd[i_list][X264_SCAN8_0+x+8*y], width, height, mv );
390 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 )
392 x264_macroblock_cache_rect1( &h->mb.cache.ref[i_list][X264_SCAN8_0+x+8*y], width, height, ref );
394 static ALWAYS_INLINE void x264_macroblock_cache_skip( x264_t *h, int x, int y, int width, int height, int b_skip )
396 x264_macroblock_cache_rect1( &h->mb.cache.skip[X264_SCAN8_0+x+8*y], width, height, b_skip );
398 static ALWAYS_INLINE void x264_macroblock_cache_intra8x8_pred( x264_t *h, int x, int y, int i_mode )
400 int8_t *cache = &h->mb.cache.intra4x4_pred_mode[X264_SCAN8_0+x+8*y];
401 cache[0] = cache[1] = cache[8] = cache[9] = i_mode;
403 #define array_non_zero(a) array_non_zero_int(a, sizeof(a))
404 #define array_non_zero_int array_non_zero_int
405 static ALWAYS_INLINE int array_non_zero_int( void *v, int i_count )
407 union {uint16_t s[4]; uint64_t l;} *x = v;
410 else if(i_count == 16)
411 return !!(x[0].l|x[1].l);
412 else if(i_count == 32)
413 return !!(x[0].l|x[1].l|x[2].l|x[3].l);
417 i_count /= sizeof(uint64_t);
418 for( i = 0; i < i_count; i++ )
419 if( x[i].l ) return 1;
423 static inline int x264_mb_predict_intra4x4_mode( x264_t *h, int idx )
425 const int ma = h->mb.cache.intra4x4_pred_mode[x264_scan8[idx] - 1];
426 const int mb = h->mb.cache.intra4x4_pred_mode[x264_scan8[idx] - 8];
427 const int m = X264_MIN( x264_mb_pred_mode4x4_fix(ma),
428 x264_mb_pred_mode4x4_fix(mb) );
431 return I_PRED_4x4_DC;
435 static inline int x264_mb_predict_non_zero_code( x264_t *h, int idx )
437 const int za = h->mb.cache.non_zero_count[x264_scan8[idx] - 1];
438 const int zb = h->mb.cache.non_zero_count[x264_scan8[idx] - 8];
444 i_ret = ( i_ret + 1 ) >> 1;
448 /* x264_mb_transform_8x8_allowed:
449 * check whether any partition is smaller than 8x8 (or at least
450 * might be, according to just partition type.)
451 * doesn't check for cbp */
452 static inline int x264_mb_transform_8x8_allowed( x264_t *h )
454 // intra and skip are disallowed
455 // large partitions are allowed
456 // direct and 8x8 are conditional
457 static const uint8_t partition_tab[X264_MBTYPE_MAX] = {
458 0,0,0,0,1,2,0,1,1,1,1,1,1,1,1,1,1,1,0,
461 if( !h->pps->b_transform_8x8_mode )
463 if( h->mb.i_type != P_8x8 )
464 return partition_tab[h->mb.i_type];
465 return *(uint32_t*)h->mb.i_sub_partition == D_L0_8x8*0x01010101;