1 /*****************************************************************************
2 * deblock.c: h264 encoder library
3 *****************************************************************************
4 * Copyright (C) 2003-2008 x264 project
6 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
7 * Loren Merritt <lorenm@u.washington.edu>
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 *****************************************************************************/
27 /* cavlc + 8x8 transform stores nnz per 16 coeffs for the purpose of
28 * entropy coding, but per 64 coeffs for the purpose of deblocking */
29 static void munge_cavlc_nnz_row( x264_t *h, int mb_y, uint8_t (*buf)[16] )
31 uint32_t (*src)[6] = (uint32_t(*)[6])h->mb.non_zero_count + mb_y * h->sps->i_mb_width;
32 int8_t *transform = h->mb.mb_transform_size + mb_y * h->sps->i_mb_width;
33 for( int x = 0; x<h->sps->i_mb_width; x++ )
35 memcpy( buf+x, src+x, 16 );
38 int nnz = src[x][0] | src[x][1];
39 src[x][0] = src[x][1] = ((uint16_t)nnz ? 0x0101 : 0) + (nnz>>16 ? 0x01010000 : 0);
40 nnz = src[x][2] | src[x][3];
41 src[x][2] = src[x][3] = ((uint16_t)nnz ? 0x0101 : 0) + (nnz>>16 ? 0x01010000 : 0);
46 static void restore_cavlc_nnz_row( x264_t *h, int mb_y, uint8_t (*buf)[16] )
48 uint8_t (*dst)[24] = h->mb.non_zero_count + mb_y * h->sps->i_mb_width;
49 for( int x = 0; x < h->sps->i_mb_width; x++ )
50 memcpy( dst+x, buf+x, 16 );
53 static void munge_cavlc_nnz( x264_t *h, int mb_y, uint8_t (*buf)[16], void (*func)(x264_t*, int, uint8_t (*)[16]) )
57 func( h, mb_y-1, buf + h->sps->i_mb_width );
60 func( h, mb_y+1, buf + h->sps->i_mb_width * 2 );
62 func( h, mb_y-2, buf + h->sps->i_mb_width * 3 );
67 /* Deblocking filter */
68 static const uint8_t i_alpha_table[52+12*2] =
70 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
71 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
72 0, 0, 0, 0, 0, 0, 4, 4, 5, 6,
73 7, 8, 9, 10, 12, 13, 15, 17, 20, 22,
74 25, 28, 32, 36, 40, 45, 50, 56, 63, 71,
75 80, 90,101,113,127,144,162,182,203,226,
77 255,255,255,255,255,255,255,255,255,255,255,255,
79 static const uint8_t i_beta_table[52+12*2] =
81 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
82 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
83 0, 0, 0, 0, 0, 0, 2, 2, 2, 3,
84 3, 3, 3, 4, 4, 4, 6, 6, 7, 7,
85 8, 8, 9, 9, 10, 10, 11, 11, 12, 12,
86 13, 13, 14, 14, 15, 15, 16, 16, 17, 17,
88 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,
90 static const int8_t i_tc0_table[52+12*2][4] =
92 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
93 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
94 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
95 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
96 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 1 },
97 {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 1, 1 }, {-1, 0, 1, 1 }, {-1, 1, 1, 1 },
98 {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 },
99 {-1, 1, 1, 2 }, {-1, 1, 2, 3 }, {-1, 1, 2, 3 }, {-1, 2, 2, 3 }, {-1, 2, 2, 4 }, {-1, 2, 3, 4 },
100 {-1, 2, 3, 4 }, {-1, 3, 3, 5 }, {-1, 3, 4, 6 }, {-1, 3, 4, 6 }, {-1, 4, 5, 7 }, {-1, 4, 5, 8 },
101 {-1, 4, 6, 9 }, {-1, 5, 7,10 }, {-1, 6, 8,11 }, {-1, 6, 8,13 }, {-1, 7,10,14 }, {-1, 8,11,16 },
102 {-1, 9,12,18 }, {-1,10,13,20 }, {-1,11,15,23 }, {-1,13,17,25 },
103 {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
104 {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
106 #define alpha_table(x) i_alpha_table[(x)+12]
107 #define beta_table(x) i_beta_table[(x)+12]
108 #define tc0_table(x) i_tc0_table[(x)+12]
111 static inline void deblock_luma_c( uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0 )
113 for( int i = 0; i < 4; i++ )
120 for( int d = 0; d < 4; d++ )
122 int p2 = pix[-3*xstride];
123 int p1 = pix[-2*xstride];
124 int p0 = pix[-1*xstride];
125 int q0 = pix[ 0*xstride];
126 int q1 = pix[ 1*xstride];
127 int q2 = pix[ 2*xstride];
129 if( abs( p0 - q0 ) < alpha && abs( p1 - p0 ) < beta && abs( q1 - q0 ) < beta )
133 if( abs( p2 - p0 ) < beta )
136 pix[-2*xstride] = p1 + x264_clip3( (( p2 + ((p0 + q0 + 1) >> 1)) >> 1) - p1, -tc0[i], tc0[i] );
139 if( abs( q2 - q0 ) < beta )
142 pix[ 1*xstride] = q1 + x264_clip3( (( q2 + ((p0 + q0 + 1) >> 1)) >> 1) - q1, -tc0[i], tc0[i] );
146 delta = x264_clip3( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
147 pix[-1*xstride] = x264_clip_uint8( p0 + delta ); /* p0' */
148 pix[ 0*xstride] = x264_clip_uint8( q0 - delta ); /* q0' */
154 static void deblock_v_luma_c( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
156 deblock_luma_c( pix, stride, 1, alpha, beta, tc0 );
158 static void deblock_h_luma_c( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
160 deblock_luma_c( pix, 1, stride, alpha, beta, tc0 );
163 static inline void deblock_chroma_c( uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0 )
165 for( int i = 0; i < 4; i++ )
173 for( int d = 0; d < 2; d++ )
175 int p1 = pix[-2*xstride];
176 int p0 = pix[-1*xstride];
177 int q0 = pix[ 0*xstride];
178 int q1 = pix[ 1*xstride];
180 if( abs( p0 - q0 ) < alpha && abs( p1 - p0 ) < beta && abs( q1 - q0 ) < beta )
182 int delta = x264_clip3( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
183 pix[-1*xstride] = x264_clip_uint8( p0 + delta ); /* p0' */
184 pix[ 0*xstride] = x264_clip_uint8( q0 - delta ); /* q0' */
190 static void deblock_v_chroma_c( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
192 deblock_chroma_c( pix, stride, 1, alpha, beta, tc0 );
194 static void deblock_h_chroma_c( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
196 deblock_chroma_c( pix, 1, stride, alpha, beta, tc0 );
199 static inline void deblock_luma_intra_c( uint8_t *pix, int xstride, int ystride, int alpha, int beta )
201 for( int d = 0; d < 16; d++ )
203 int p2 = pix[-3*xstride];
204 int p1 = pix[-2*xstride];
205 int p0 = pix[-1*xstride];
206 int q0 = pix[ 0*xstride];
207 int q1 = pix[ 1*xstride];
208 int q2 = pix[ 2*xstride];
210 if( abs( p0 - q0 ) < alpha && abs( p1 - p0 ) < beta && abs( q1 - q0 ) < beta )
212 if(abs( p0 - q0 ) < ((alpha >> 2) + 2) )
214 if( abs( p2 - p0 ) < beta ) /* p0', p1', p2' */
216 const int p3 = pix[-4*xstride];
217 pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
218 pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
219 pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
222 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
223 if( abs( q2 - q0 ) < beta ) /* q0', q1', q2' */
225 const int q3 = pix[3*xstride];
226 pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
227 pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
228 pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
231 pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
235 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
236 pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
242 static void deblock_v_luma_intra_c( uint8_t *pix, int stride, int alpha, int beta )
244 deblock_luma_intra_c( pix, stride, 1, alpha, beta );
246 static void deblock_h_luma_intra_c( uint8_t *pix, int stride, int alpha, int beta )
248 deblock_luma_intra_c( pix, 1, stride, alpha, beta );
251 static inline void deblock_chroma_intra_c( uint8_t *pix, int xstride, int ystride, int alpha, int beta )
253 for( int d = 0; d < 8; d++ )
255 int p1 = pix[-2*xstride];
256 int p0 = pix[-1*xstride];
257 int q0 = pix[ 0*xstride];
258 int q1 = pix[ 1*xstride];
260 if( abs( p0 - q0 ) < alpha && abs( p1 - p0 ) < beta && abs( q1 - q0 ) < beta )
262 pix[-1*xstride] = (2*p1 + p0 + q1 + 2) >> 2; /* p0' */
263 pix[ 0*xstride] = (2*q1 + q0 + p1 + 2) >> 2; /* q0' */
268 static void deblock_v_chroma_intra_c( uint8_t *pix, int stride, int alpha, int beta )
270 deblock_chroma_intra_c( pix, stride, 1, alpha, beta );
272 static void deblock_h_chroma_intra_c( uint8_t *pix, int stride, int alpha, int beta )
274 deblock_chroma_intra_c( pix, 1, stride, alpha, beta );
277 static inline void deblock_edge( x264_t *h, uint8_t *pix1, uint8_t *pix2, int i_stride, uint8_t bS[4], int i_qp, int b_chroma, x264_deblock_inter_t pf_inter )
279 int index_a = i_qp + h->sh.i_alpha_c0_offset;
280 int alpha = alpha_table(index_a);
281 int beta = beta_table(i_qp + h->sh.i_beta_offset);
284 if( !alpha || !beta )
287 tc[0] = tc0_table(index_a)[bS[0]] + b_chroma;
288 tc[1] = tc0_table(index_a)[bS[1]] + b_chroma;
289 tc[2] = tc0_table(index_a)[bS[2]] + b_chroma;
290 tc[3] = tc0_table(index_a)[bS[3]] + b_chroma;
292 pf_inter( pix1, i_stride, alpha, beta, tc );
294 pf_inter( pix2, i_stride, alpha, beta, tc );
297 static inline void deblock_edge_intra( x264_t *h, uint8_t *pix1, uint8_t *pix2, int i_stride, uint8_t bS[4], int i_qp, int b_chroma, x264_deblock_intra_t pf_intra )
299 int alpha = alpha_table(i_qp + h->sh.i_alpha_c0_offset);
300 int beta = beta_table(i_qp + h->sh.i_beta_offset);
302 if( !alpha || !beta )
305 pf_intra( pix1, i_stride, alpha, beta );
307 pf_intra( pix2, i_stride, alpha, beta );
310 void x264_frame_deblock_row( x264_t *h, int mb_y )
312 int s8x8 = 2 * h->mb.i_mb_stride;
313 int s4x4 = 4 * h->mb.i_mb_stride;
314 int b_interlaced = h->sh.b_mbaff;
315 int mvy_limit = 4 >> b_interlaced;
316 int qp_thresh = 15 - X264_MIN(h->sh.i_alpha_c0_offset, h->sh.i_beta_offset) - X264_MAX(0, h->param.analyse.i_chroma_qp_offset);
317 int no_sub8x8 = !(h->param.analyse.inter & X264_ANALYSE_PSUB8x8);
319 int stridey = h->fdec->i_stride[0];
320 int stride2y = stridey << b_interlaced;
321 int strideuv = h->fdec->i_stride[1];
322 int stride2uv = strideuv << b_interlaced;
323 int deblock_on_slice_edges = h->sh.i_disable_deblocking_filter_idc != 2;
325 #define ref_table(x) ref_table[x+2]
326 uint8_t (*nnz_backup)[16] = h->scratch_buffer;
330 for( int i = 0; i < (h->i_ref0 << h->sh.b_mbaff); i++ )
332 if( !h->mb.b_interlaced )
333 ref_table(i) = h->fref0[i]->i_poc;
335 ref_table(i) = h->fref0[i>>1]->i_poc + (i&1);
338 if( !h->pps->b_cabac && h->pps->b_transform_8x8_mode )
339 munge_cavlc_nnz( h, mb_y, nnz_backup, munge_cavlc_nnz_row );
341 for( mb_x = 0; mb_x < h->sps->i_mb_width; mb_x += (~b_interlaced | mb_y)&1, mb_y ^= b_interlaced )
343 int mb_xy = mb_y * h->mb.i_mb_stride + mb_x;
344 int mb_8x8 = 2 * s8x8 * mb_y + 2 * mb_x;
345 int mb_4x4 = 4 * s4x4 * mb_y + 4 * mb_x;
346 int b_8x8_transform = h->mb.mb_transform_size[mb_xy];
347 int i_qp = h->mb.qp[mb_xy];
348 int i_edge_end = (h->mb.type[mb_xy] == P_SKIP) ? 1 : 4;
349 uint8_t *pixy = h->fdec->plane[0] + 16*mb_y*stridey + 16*mb_x;
350 uint8_t *pixu = h->fdec->plane[1] + 8*mb_y*strideuv + 8*mb_x;
351 uint8_t *pixv = h->fdec->plane[2] + 8*mb_y*strideuv + 8*mb_x;
352 if( b_interlaced && (mb_y&1) )
359 x264_prefetch_fenc( h, h->fdec, mb_x, mb_y );
361 if( i_qp <= qp_thresh )
364 #define FILTER_DIR(intra, i_dir)\
367 i_qpn= h->mb.qp[mbn_xy];\
371 deblock_edge##intra( h, pixy + 4*i_edge, NULL,\
372 stride2y, bS, (i_qp+i_qpn+1) >> 1, 0,\
373 h->loopf.deblock_h_luma##intra );\
377 int i_qpc = (h->chroma_qp_table[i_qp] + h->chroma_qp_table[i_qpn] + 1) >> 1;\
378 deblock_edge##intra( h, pixu + 2*i_edge, pixv + 2*i_edge,\
379 stride2uv, bS, i_qpc, 1,\
380 h->loopf.deblock_h_chroma##intra );\
385 /* horizontal edge */\
386 deblock_edge##intra( h, pixy + 4*i_edge*stride2y, NULL,\
387 stride2y, bS, (i_qp+i_qpn+1) >> 1, 0,\
388 h->loopf.deblock_v_luma##intra );\
392 int i_qpc = (h->chroma_qp_table[i_qp] + h->chroma_qp_table[i_qpn] + 1) >> 1;\
393 deblock_edge##intra( h, pixu + 2*i_edge*stride2uv, pixv + 2*i_edge*stride2uv,\
394 stride2uv, bS, i_qpc, 1,\
395 h->loopf.deblock_v_chroma##intra );\
400 #define DEBLOCK_STRENGTH(i_dir)\
402 /* *** Get bS for each 4px for the current edge *** */\
403 if( IS_INTRA( h->mb.type[mb_xy] ) || IS_INTRA( h->mb.type[mbn_xy]) )\
404 M32( bS ) = 0x03030303;\
407 M32( bS ) = 0x00000000;\
408 for( int i = 0; i < 4; i++ )\
410 int x = i_dir == 0 ? i_edge : i;\
411 int y = i_dir == 0 ? i : i_edge;\
412 int xn = i_dir == 0 ? (x - 1)&0x03 : x;\
413 int yn = i_dir == 0 ? y : (y - 1)&0x03;\
414 if( h->mb.non_zero_count[mb_xy][x+y*4] != 0 ||\
415 h->mb.non_zero_count[mbn_xy][xn+yn*4] != 0 )\
417 else if(!(i_edge&no_sub8x8))\
419 if((i&no_sub8x8) && bS[i-1] != 2)\
423 int i8p= mb_8x8+(x>>1)+(y>>1)*s8x8;\
424 int i8q= mbn_8x8+(xn>>1)+(yn>>1)*s8x8;\
425 int i4p= mb_4x4+x+y*s4x4;\
426 int i4q= mbn_4x4+xn+yn*s4x4;\
427 if((!(ref_table(h->mb.ref[0][i8p]) == ref_table(h->mb.ref[0][i8q])) ||\
428 abs( h->mb.mv[0][i4p][0] - h->mb.mv[0][i4q][0] ) >= 4 ||\
429 abs( h->mb.mv[0][i4p][1] - h->mb.mv[0][i4q][1] ) >= mvy_limit ) ||\
430 (h->sh.i_type == SLICE_TYPE_B &&\
431 (h->mb.ref[1][i8p] != h->mb.ref[1][i8q] ||\
432 abs( h->mb.mv[1][i4p][0] - h->mb.mv[1][i4q][0] ) >= 4 ||\
433 abs( h->mb.mv[1][i4p][1] - h->mb.mv[1][i4q][1] ) >= mvy_limit )))\
443 /* i_dir == 0 -> vertical edge
444 * i_dir == 1 -> horizontal edge */
445 #define DEBLOCK_DIR(i_dir)\
448 int i_qpn, mbn_xy, mbn_8x8, mbn_4x4;\
449 ALIGNED_4( uint8_t bS[4] ); /* filtering strength */\
450 /* We don't have to consider the MBAFF case of a slice breaking in the middle\
451 * of a row because x264 doesn't support that case. If we add support for that,\
452 * this will have to become significantly more complex. */\
453 if( i_dir == 0 && (mb_x == 0 || (!deblock_on_slice_edges &&\
454 h->mb.slice_table[mb_xy] != h->mb.slice_table[mb_xy-1])) )\
456 if( i_dir == 1 && (mb_y <= b_interlaced || (!deblock_on_slice_edges &&\
457 h->mb.slice_table[mb_xy] != h->mb.slice_table[mb_xy-(h->mb.i_mb_stride<<b_interlaced)])) )\
460 i_edge+= b_8x8_transform;\
463 mbn_xy = i_dir == 0 ? mb_xy - 1 : mb_xy - h->mb.i_mb_stride;\
464 mbn_8x8 = i_dir == 0 ? mb_8x8 - 2 : mb_8x8 - 2 * s8x8;\
465 mbn_4x4 = i_dir == 0 ? mb_4x4 - 4 : mb_4x4 - 4 * s4x4;\
466 if( b_interlaced && i_dir == 1 )\
468 mbn_xy -= h->mb.i_mb_stride;\
469 mbn_8x8 -= 2 * s8x8;\
470 mbn_4x4 -= 4 * s4x4;\
472 else if( IS_INTRA( h->mb.type[mb_xy] ) || IS_INTRA( h->mb.type[mbn_xy]) )\
474 FILTER_DIR( _intra, i_dir );\
477 DEBLOCK_STRENGTH(i_dir);\
479 FILTER_DIR( , i_dir);\
481 i_edge += b_8x8_transform+1;\
486 for( ; i_edge < i_edge_end; i_edge+=b_8x8_transform+1 )\
488 DEBLOCK_STRENGTH(i_dir);\
490 FILTER_DIR( , i_dir);\
498 if( !h->pps->b_cabac && h->pps->b_transform_8x8_mode )
499 munge_cavlc_nnz( h, mb_y, nnz_backup, restore_cavlc_nnz_row );
502 void x264_frame_deblock( x264_t *h )
504 for( int mb_y = 0; mb_y < h->sps->i_mb_height; mb_y += 1 + h->sh.b_mbaff )
505 x264_frame_deblock_row( h, mb_y );
509 void x264_deblock_v_chroma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
510 void x264_deblock_h_chroma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
511 void x264_deblock_v_chroma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );
512 void x264_deblock_h_chroma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );
514 void x264_deblock_v_luma_sse2( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
515 void x264_deblock_h_luma_sse2( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
516 void x264_deblock_v_luma_intra_sse2( uint8_t *pix, int stride, int alpha, int beta );
517 void x264_deblock_h_luma_intra_sse2( uint8_t *pix, int stride, int alpha, int beta );
519 void x264_deblock_h_luma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
520 void x264_deblock_v8_luma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
521 void x264_deblock_h_luma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );
522 void x264_deblock_v8_luma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta );
524 static void x264_deblock_v_luma_mmxext( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
526 x264_deblock_v8_luma_mmxext( pix, stride, alpha, beta, tc0 );
527 x264_deblock_v8_luma_mmxext( pix+8, stride, alpha, beta, tc0+2 );
529 static void x264_deblock_v_luma_intra_mmxext( uint8_t *pix, int stride, int alpha, int beta )
531 x264_deblock_v8_luma_intra_mmxext( pix, stride, alpha, beta );
532 x264_deblock_v8_luma_intra_mmxext( pix+8, stride, alpha, beta );
538 void x264_deblock_v_luma_altivec( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
539 void x264_deblock_h_luma_altivec( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
543 void x264_deblock_v_luma_neon( uint8_t *, int, int, int, int8_t * );
544 void x264_deblock_h_luma_neon( uint8_t *, int, int, int, int8_t * );
545 void x264_deblock_v_chroma_neon( uint8_t *, int, int, int, int8_t * );
546 void x264_deblock_h_chroma_neon( uint8_t *, int, int, int, int8_t * );
549 void x264_deblock_init( int cpu, x264_deblock_function_t *pf )
551 pf->deblock_v_luma = deblock_v_luma_c;
552 pf->deblock_h_luma = deblock_h_luma_c;
553 pf->deblock_v_chroma = deblock_v_chroma_c;
554 pf->deblock_h_chroma = deblock_h_chroma_c;
555 pf->deblock_v_luma_intra = deblock_v_luma_intra_c;
556 pf->deblock_h_luma_intra = deblock_h_luma_intra_c;
557 pf->deblock_v_chroma_intra = deblock_v_chroma_intra_c;
558 pf->deblock_h_chroma_intra = deblock_h_chroma_intra_c;
561 if( cpu&X264_CPU_MMXEXT )
563 pf->deblock_v_chroma = x264_deblock_v_chroma_mmxext;
564 pf->deblock_h_chroma = x264_deblock_h_chroma_mmxext;
565 pf->deblock_v_chroma_intra = x264_deblock_v_chroma_intra_mmxext;
566 pf->deblock_h_chroma_intra = x264_deblock_h_chroma_intra_mmxext;
568 pf->deblock_v_luma = x264_deblock_v_luma_mmxext;
569 pf->deblock_h_luma = x264_deblock_h_luma_mmxext;
570 pf->deblock_v_luma_intra = x264_deblock_v_luma_intra_mmxext;
571 pf->deblock_h_luma_intra = x264_deblock_h_luma_intra_mmxext;
573 if( (cpu&X264_CPU_SSE2) && !(cpu&X264_CPU_STACK_MOD4) )
575 pf->deblock_v_luma = x264_deblock_v_luma_sse2;
576 pf->deblock_h_luma = x264_deblock_h_luma_sse2;
577 pf->deblock_v_luma_intra = x264_deblock_v_luma_intra_sse2;
578 pf->deblock_h_luma_intra = x264_deblock_h_luma_intra_sse2;
584 if( cpu&X264_CPU_ALTIVEC )
586 pf->deblock_v_luma = x264_deblock_v_luma_altivec;
587 pf->deblock_h_luma = x264_deblock_h_luma_altivec;
589 #endif // HAVE_ALTIVEC
592 if( cpu&X264_CPU_NEON )
594 pf->deblock_v_luma = x264_deblock_v_luma_neon;
595 pf->deblock_h_luma = x264_deblock_h_luma_neon;
596 pf->deblock_v_chroma = x264_deblock_v_chroma_neon;
597 pf->deblock_h_chroma = x264_deblock_h_chroma_neon;