1 /*****************************************************************************
2 * deblock.c: deblocking
3 *****************************************************************************
4 * Copyright (C) 2003-2011 x264 project
6 * Authors: Laurent Aimar <fenrir@via.ecp.fr>
7 * Loren Merritt <lorenm@u.washington.edu>
8 * Fiona Glaser <fiona@x264.com>
9 * Henrik Gramner <hengar-6@student.ltu.se>
11 * This program is free software; you can redistribute it and/or modify
12 * it under the terms of the GNU General Public License as published by
13 * the Free Software Foundation; either version 2 of the License, or
14 * (at your option) any later version.
16 * This program is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19 * GNU General Public License for more details.
21 * You should have received a copy of the GNU General Public License
22 * along with this program; if not, write to the Free Software
23 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
25 * This program is also available under a commercial proprietary license.
26 * For more information, contact us at licensing@x264.com.
27 *****************************************************************************/
31 /* Deblocking filter */
32 static const uint8_t i_alpha_table[52+12*3] =
34 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
35 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
36 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
37 0, 0, 0, 0, 0, 0, 4, 4, 5, 6,
38 7, 8, 9, 10, 12, 13, 15, 17, 20, 22,
39 25, 28, 32, 36, 40, 45, 50, 56, 63, 71,
40 80, 90,101,113,127,144,162,182,203,226,
42 255,255,255,255,255,255,255,255,255,255,255,255,
44 static const uint8_t i_beta_table[52+12*3] =
46 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
47 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
48 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
49 0, 0, 0, 0, 0, 0, 2, 2, 2, 3,
50 3, 3, 3, 4, 4, 4, 6, 6, 7, 7,
51 8, 8, 9, 9, 10, 10, 11, 11, 12, 12,
52 13, 13, 14, 14, 15, 15, 16, 16, 17, 17,
54 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18, 18,
56 static const int8_t i_tc0_table[52+12*3][4] =
58 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
59 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
60 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
61 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
62 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
63 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 },
64 {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 0 }, {-1, 0, 0, 1 },
65 {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 0, 1 }, {-1, 0, 1, 1 }, {-1, 0, 1, 1 }, {-1, 1, 1, 1 },
66 {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 1 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 }, {-1, 1, 1, 2 },
67 {-1, 1, 1, 2 }, {-1, 1, 2, 3 }, {-1, 1, 2, 3 }, {-1, 2, 2, 3 }, {-1, 2, 2, 4 }, {-1, 2, 3, 4 },
68 {-1, 2, 3, 4 }, {-1, 3, 3, 5 }, {-1, 3, 4, 6 }, {-1, 3, 4, 6 }, {-1, 4, 5, 7 }, {-1, 4, 5, 8 },
69 {-1, 4, 6, 9 }, {-1, 5, 7,10 }, {-1, 6, 8,11 }, {-1, 6, 8,13 }, {-1, 7,10,14 }, {-1, 8,11,16 },
70 {-1, 9,12,18 }, {-1,10,13,20 }, {-1,11,15,23 }, {-1,13,17,25 },
71 {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
72 {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 }, {-1,13,17,25 },
74 #define alpha_table(x) i_alpha_table[(x)+24]
75 #define beta_table(x) i_beta_table[(x)+24]
76 #define tc0_table(x) i_tc0_table[(x)+24]
79 static ALWAYS_INLINE void deblock_edge_luma_c( pixel *pix, int xstride, int alpha, int beta, int8_t tc0 )
81 int p2 = pix[-3*xstride];
82 int p1 = pix[-2*xstride];
83 int p0 = pix[-1*xstride];
84 int q0 = pix[ 0*xstride];
85 int q1 = pix[ 1*xstride];
86 int q2 = pix[ 2*xstride];
88 if( abs( p0 - q0 ) < alpha && abs( p1 - p0 ) < beta && abs( q1 - q0 ) < beta )
92 if( abs( p2 - p0 ) < beta )
95 pix[-2*xstride] = p1 + x264_clip3( (( p2 + ((p0 + q0 + 1) >> 1)) >> 1) - p1, -tc0, tc0 );
98 if( abs( q2 - q0 ) < beta )
101 pix[ 1*xstride] = q1 + x264_clip3( (( q2 + ((p0 + q0 + 1) >> 1)) >> 1) - q1, -tc0, tc0 );
105 delta = x264_clip3( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
106 pix[-1*xstride] = x264_clip_pixel( p0 + delta ); /* p0' */
107 pix[ 0*xstride] = x264_clip_pixel( q0 - delta ); /* q0' */
110 static inline void deblock_luma_c( pixel *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0 )
112 for( int i = 0; i < 4; i++ )
119 for( int d = 0; d < 4; d++, pix += ystride )
120 deblock_edge_luma_c( pix, xstride, alpha, beta, tc0[i] );
123 static void deblock_h_luma_mbaff_c( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 )
125 for( int d = 0; d < 8; d++, pix += stride )
126 deblock_edge_luma_c( pix, 1, alpha, beta, tc0[d>>1] );
128 static void deblock_v_luma_c( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 )
130 deblock_luma_c( pix, stride, 1, alpha, beta, tc0 );
132 static void deblock_h_luma_c( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 )
134 deblock_luma_c( pix, 1, stride, alpha, beta, tc0 );
137 static ALWAYS_INLINE void deblock_edge_chroma_c( pixel *pix, int xstride, int alpha, int beta, int8_t tc )
139 int p1 = pix[-2*xstride];
140 int p0 = pix[-1*xstride];
141 int q0 = pix[ 0*xstride];
142 int q1 = pix[ 1*xstride];
144 if( abs( p0 - q0 ) < alpha && abs( p1 - p0 ) < beta && abs( q1 - q0 ) < beta )
146 int delta = x264_clip3( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
147 pix[-1*xstride] = x264_clip_pixel( p0 + delta ); /* p0' */
148 pix[ 0*xstride] = x264_clip_pixel( q0 - delta ); /* q0' */
151 static ALWAYS_INLINE void deblock_chroma_c( pixel *pix, int height, int xstride, int ystride, int alpha, int beta, int8_t *tc0 )
153 for( int i = 0; i < 4; i++ )
158 pix += height*ystride;
161 for( int d = 0; d < height; d++, pix += ystride-2 )
162 for( int e = 0; e < 2; e++, pix++ )
163 deblock_edge_chroma_c( pix, xstride, alpha, beta, tc0[i] );
166 static void deblock_h_chroma_mbaff_c( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 )
168 for( int i = 0; i < 4; i++, pix += stride )
169 deblock_edge_chroma_c( pix, 2, alpha, beta, tc0[i] );
171 static void deblock_h_chroma_422_mbaff_c( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 )
173 for( int i = 0; i < 8; i++, pix += stride )
174 deblock_edge_chroma_c( pix, 2, alpha, beta, tc0[i>>1] );
176 static void deblock_v_chroma_c( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 )
178 deblock_chroma_c( pix, 2, stride, 2, alpha, beta, tc0 );
180 static void deblock_h_chroma_c( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 )
182 deblock_chroma_c( pix, 2, 2, stride, alpha, beta, tc0 );
184 static void deblock_h_chroma_422_c( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 )
186 deblock_chroma_c( pix, 4, 2, stride, alpha, beta, tc0 );
189 static ALWAYS_INLINE void deblock_edge_luma_intra_c( pixel *pix, int xstride, int alpha, int beta )
191 int p2 = pix[-3*xstride];
192 int p1 = pix[-2*xstride];
193 int p0 = pix[-1*xstride];
194 int q0 = pix[ 0*xstride];
195 int q1 = pix[ 1*xstride];
196 int q2 = pix[ 2*xstride];
198 if( abs( p0 - q0 ) < alpha && abs( p1 - p0 ) < beta && abs( q1 - q0 ) < beta )
200 if( abs( p0 - q0 ) < ((alpha >> 2) + 2) )
202 if( abs( p2 - p0 ) < beta ) /* p0', p1', p2' */
204 const int p3 = pix[-4*xstride];
205 pix[-1*xstride] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
206 pix[-2*xstride] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
207 pix[-3*xstride] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
210 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
211 if( abs( q2 - q0 ) < beta ) /* q0', q1', q2' */
213 const int q3 = pix[3*xstride];
214 pix[0*xstride] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
215 pix[1*xstride] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
216 pix[2*xstride] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
219 pix[0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
223 pix[-1*xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
224 pix[ 0*xstride] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
228 static inline void deblock_luma_intra_c( pixel *pix, int xstride, int ystride, int alpha, int beta )
230 for( int d = 0; d < 16; d++, pix += ystride )
231 deblock_edge_luma_intra_c( pix, xstride, alpha, beta );
233 static void deblock_h_luma_intra_mbaff_c( pixel *pix, int ystride, int alpha, int beta )
235 for( int d = 0; d < 8; d++, pix += ystride )
236 deblock_edge_luma_intra_c( pix, 1, alpha, beta );
238 static void deblock_v_luma_intra_c( pixel *pix, int stride, int alpha, int beta )
240 deblock_luma_intra_c( pix, stride, 1, alpha, beta );
242 static void deblock_h_luma_intra_c( pixel *pix, int stride, int alpha, int beta )
244 deblock_luma_intra_c( pix, 1, stride, alpha, beta );
247 static ALWAYS_INLINE void deblock_edge_chroma_intra_c( pixel *pix, int xstride, int alpha, int beta )
249 int p1 = pix[-2*xstride];
250 int p0 = pix[-1*xstride];
251 int q0 = pix[ 0*xstride];
252 int q1 = pix[ 1*xstride];
254 if( abs( p0 - q0 ) < alpha && abs( p1 - p0 ) < beta && abs( q1 - q0 ) < beta )
256 pix[-1*xstride] = (2*p1 + p0 + q1 + 2) >> 2; /* p0' */
257 pix[ 0*xstride] = (2*q1 + q0 + p1 + 2) >> 2; /* q0' */
260 static ALWAYS_INLINE void deblock_chroma_intra_c( pixel *pix, int width, int height, int xstride, int ystride, int alpha, int beta )
262 for( int d = 0; d < height; d++, pix += ystride-2 )
263 for( int e = 0; e < width; e++, pix++ )
264 deblock_edge_chroma_intra_c( pix, xstride, alpha, beta );
266 static void deblock_h_chroma_intra_mbaff_c( pixel *pix, int stride, int alpha, int beta )
268 for( int i = 0; i < 4; i++, pix += stride )
269 deblock_edge_chroma_intra_c( pix, 2, alpha, beta );
271 static void deblock_h_chroma_422_intra_mbaff_c( pixel *pix, int stride, int alpha, int beta )
273 for( int i = 0; i < 8; i++, pix += stride )
274 deblock_edge_chroma_intra_c( pix, 2, alpha, beta );
276 static void deblock_v_chroma_intra_c( pixel *pix, int stride, int alpha, int beta )
278 deblock_chroma_intra_c( pix, 1, 16, stride, 2, alpha, beta );
280 static void deblock_h_chroma_intra_c( pixel *pix, int stride, int alpha, int beta )
282 deblock_chroma_intra_c( pix, 2, 8, 2, stride, alpha, beta );
284 static void deblock_h_chroma_422_intra_c( pixel *pix, int stride, int alpha, int beta )
286 deblock_chroma_intra_c( pix, 2, 16, 2, stride, alpha, beta );
289 static void deblock_strength_c( uint8_t nnz[X264_SCAN8_SIZE], int8_t ref[2][X264_SCAN8_LUMA_SIZE],
290 int16_t mv[2][X264_SCAN8_LUMA_SIZE][2], uint8_t bs[2][8][4], int mvy_limit,
293 for( int dir = 0; dir < 2; dir++ )
295 int s1 = dir ? 1 : 8;
296 int s2 = dir ? 8 : 1;
297 for( int edge = 0; edge < 4; edge++ )
298 for( int i = 0, loc = X264_SCAN8_0+edge*s2; i < 4; i++, loc += s1 )
301 if( nnz[loc] || nnz[locn] )
302 bs[dir][edge][i] = 2;
303 else if( ref[0][loc] != ref[0][locn] ||
304 abs( mv[0][loc][0] - mv[0][locn][0] ) >= 4 ||
305 abs( mv[0][loc][1] - mv[0][locn][1] ) >= mvy_limit ||
306 (bframe && (ref[1][loc] != ref[1][locn] ||
307 abs( mv[1][loc][0] - mv[1][locn][0] ) >= 4 ||
308 abs( mv[1][loc][1] - mv[1][locn][1] ) >= mvy_limit )))
310 bs[dir][edge][i] = 1;
313 bs[dir][edge][i] = 0;
318 static ALWAYS_INLINE void deblock_edge( x264_t *h, pixel *pix, int i_stride, uint8_t bS[4], int i_qp, int a, int b, int b_chroma, x264_deblock_inter_t pf_inter )
320 int index_a = i_qp + a;
321 int index_b = i_qp + b;
322 int alpha = alpha_table(index_a) << (BIT_DEPTH-8);
323 int beta = beta_table(index_b) << (BIT_DEPTH-8);
326 if( !M32(bS) || !alpha || !beta )
329 tc[0] = (tc0_table(index_a)[bS[0]] << (BIT_DEPTH-8)) + b_chroma;
330 tc[1] = (tc0_table(index_a)[bS[1]] << (BIT_DEPTH-8)) + b_chroma;
331 tc[2] = (tc0_table(index_a)[bS[2]] << (BIT_DEPTH-8)) + b_chroma;
332 tc[3] = (tc0_table(index_a)[bS[3]] << (BIT_DEPTH-8)) + b_chroma;
334 pf_inter( pix, i_stride, alpha, beta, tc );
337 static ALWAYS_INLINE void deblock_edge_intra( x264_t *h, pixel *pix, int i_stride, uint8_t bS[4], int i_qp, int a, int b, int b_chroma, x264_deblock_intra_t pf_intra )
339 int index_a = i_qp + a;
340 int index_b = i_qp + b;
341 int alpha = alpha_table(index_a) << (BIT_DEPTH-8);
342 int beta = beta_table(index_b) << (BIT_DEPTH-8);
344 if( !alpha || !beta )
347 pf_intra( pix, i_stride, alpha, beta );
350 static ALWAYS_INLINE void x264_macroblock_cache_load_neighbours_deblock( x264_t *h, int mb_x, int mb_y )
352 int deblock_on_slice_edges = h->sh.i_disable_deblocking_filter_idc != 2;
354 h->mb.i_neighbour = 0;
355 h->mb.i_mb_xy = mb_y * h->mb.i_mb_stride + mb_x;
356 h->mb.b_interlaced = PARAM_INTERLACED && h->mb.field[h->mb.i_mb_xy];
357 h->mb.i_mb_top_y = mb_y - (1 << MB_INTERLACED);
358 h->mb.i_mb_top_xy = mb_x + h->mb.i_mb_stride*h->mb.i_mb_top_y;
359 h->mb.i_mb_left_xy[1] =
360 h->mb.i_mb_left_xy[0] = h->mb.i_mb_xy - 1;
365 if( mb_x && h->mb.field[h->mb.i_mb_xy - 1] != MB_INTERLACED )
366 h->mb.i_mb_left_xy[0] -= h->mb.i_mb_stride;
370 if( h->mb.i_mb_top_xy >= 0 && MB_INTERLACED && !h->mb.field[h->mb.i_mb_top_xy] )
372 h->mb.i_mb_top_xy += h->mb.i_mb_stride;
375 if( mb_x && h->mb.field[h->mb.i_mb_xy - 1] != MB_INTERLACED )
376 h->mb.i_mb_left_xy[1] += h->mb.i_mb_stride;
380 if( mb_x > 0 && (deblock_on_slice_edges ||
381 h->mb.slice_table[h->mb.i_mb_left_xy[0]] == h->mb.slice_table[h->mb.i_mb_xy]) )
382 h->mb.i_neighbour |= MB_LEFT;
383 if( mb_y > MB_INTERLACED && (deblock_on_slice_edges
384 || h->mb.slice_table[h->mb.i_mb_top_xy] == h->mb.slice_table[h->mb.i_mb_xy]) )
385 h->mb.i_neighbour |= MB_TOP;
388 void x264_frame_deblock_row( x264_t *h, int mb_y )
390 int b_interlaced = SLICE_MBAFF;
391 int a = h->sh.i_alpha_c0_offset - QP_BD_OFFSET;
392 int b = h->sh.i_beta_offset - QP_BD_OFFSET;
393 int qp_thresh = 15 - X264_MIN( a, b ) - X264_MAX( 0, h->pps->i_chroma_qp_index_offset );
394 int stridey = h->fdec->i_stride[0];
395 int strideuv = h->fdec->i_stride[1];
396 int chroma444 = CHROMA444;
397 int chroma_height = 16 >> CHROMA_V_SHIFT;
398 intptr_t uvdiff = chroma444 ? h->fdec->plane[2] - h->fdec->plane[1] : 1;
400 for( int mb_x = 0; mb_x < h->mb.i_mb_width; mb_x += (~b_interlaced | mb_y)&1, mb_y ^= b_interlaced )
402 x264_prefetch_fenc( h, h->fdec, mb_x, mb_y );
403 x264_macroblock_cache_load_neighbours_deblock( h, mb_x, mb_y );
405 int mb_xy = h->mb.i_mb_xy;
406 int transform_8x8 = h->mb.mb_transform_size[h->mb.i_mb_xy];
407 int intra_cur = IS_INTRA( h->mb.type[mb_xy] );
408 uint8_t (*bs)[8][4] = h->deblock_strength[mb_y&1][mb_x];
410 pixel *pixy = h->fdec->plane[0] + 16*mb_y*stridey + 16*mb_x;
411 pixel *pixuv = h->fdec->plane[1] + chroma_height*mb_y*strideuv + 16*mb_x;
413 if( mb_y & MB_INTERLACED )
416 pixuv -= (chroma_height-1)*strideuv;
419 int stride2y = stridey << MB_INTERLACED;
420 int stride2uv = strideuv << MB_INTERLACED;
421 int qp = h->mb.qp[mb_xy];
422 int qpc = h->chroma_qp_table[qp];
423 int first_edge_only = (h->mb.partition[mb_xy] == D_16x16 && !h->mb.cbp[mb_xy] && !intra_cur) || qp <= qp_thresh;
425 #define FILTER( intra, dir, edge, qp, chroma_qp )\
428 if( !(edge & 1) || !transform_8x8 )\
430 deblock_edge##intra( h, pixy + 4*edge*(dir?stride2y:1),\
431 stride2y, bs[dir][edge], qp, a, b, 0,\
432 h->loopf.deblock_luma##intra[dir] );\
433 if( CHROMA_FORMAT == CHROMA_444 )\
435 deblock_edge##intra( h, pixuv + 4*edge*(dir?stride2uv:1),\
436 stride2uv, bs[dir][edge], chroma_qp, a, b, 0,\
437 h->loopf.deblock_luma##intra[dir] );\
438 deblock_edge##intra( h, pixuv + uvdiff + 4*edge*(dir?stride2uv:1),\
439 stride2uv, bs[dir][edge], chroma_qp, a, b, 0,\
440 h->loopf.deblock_luma##intra[dir] );\
442 else if( CHROMA_FORMAT == CHROMA_420 && !(edge & 1) )\
444 deblock_edge##intra( h, pixuv + edge*(dir?2*stride2uv:4),\
445 stride2uv, bs[dir][edge], chroma_qp, a, b, 1,\
446 h->loopf.deblock_chroma##intra[dir] );\
449 if( CHROMA_FORMAT == CHROMA_422 && (dir || !(edge & 1)) )\
451 deblock_edge##intra( h, pixuv + edge*(dir?4*stride2uv:4),\
452 stride2uv, bs[dir][edge], chroma_qp, a, b, 1,\
453 h->loopf.deblock_chroma##intra[dir] );\
457 if( h->mb.i_neighbour & MB_LEFT )
459 if( b_interlaced && h->mb.field[h->mb.i_mb_left_xy[0]] != MB_INTERLACED )
464 x264_deblock_inter_t luma_deblock = h->loopf.deblock_luma_mbaff;
465 x264_deblock_inter_t chroma_deblock = h->loopf.deblock_chroma_mbaff;
466 x264_deblock_intra_t luma_intra_deblock = h->loopf.deblock_luma_intra_mbaff;
467 x264_deblock_intra_t chroma_intra_deblock = h->loopf.deblock_chroma_intra_mbaff;
468 int c = chroma444 ? 0 : 1;
470 left_qp[0] = h->mb.qp[h->mb.i_mb_left_xy[0]];
471 luma_qp[0] = (qp + left_qp[0] + 1) >> 1;
472 chroma_qp[0] = (qpc + h->chroma_qp_table[left_qp[0]] + 1) >> 1;
473 if( intra_cur || IS_INTRA( h->mb.type[h->mb.i_mb_left_xy[0]] ) )
475 deblock_edge_intra( h, pixy, 2*stridey, bs[0][0], luma_qp[0], a, b, 0, luma_intra_deblock );
476 deblock_edge_intra( h, pixuv, 2*strideuv, bs[0][0], chroma_qp[0], a, b, c, chroma_intra_deblock );
477 deblock_edge_intra( h, pixuv + uvdiff, 2*strideuv, bs[0][0], chroma_qp[0], a, b, c, chroma_intra_deblock );
481 deblock_edge( h, pixy, 2*stridey, bs[0][0], luma_qp[0], a, b, 0, luma_deblock );
482 deblock_edge( h, pixuv, 2*strideuv, bs[0][0], chroma_qp[0], a, b, c, chroma_deblock );
483 deblock_edge( h, pixuv + uvdiff, 2*strideuv, bs[0][0], chroma_qp[0], a, b, c, chroma_deblock );
486 int offy = MB_INTERLACED ? 4 : 0;
487 int offuv = MB_INTERLACED ? 4-CHROMA_V_SHIFT : 0;
488 left_qp[1] = h->mb.qp[h->mb.i_mb_left_xy[1]];
489 luma_qp[1] = (qp + left_qp[1] + 1) >> 1;
490 chroma_qp[1] = (qpc + h->chroma_qp_table[left_qp[1]] + 1) >> 1;
491 if( intra_cur || IS_INTRA( h->mb.type[h->mb.i_mb_left_xy[1]] ) )
493 deblock_edge_intra( h, pixy + (stridey<<offy), 2*stridey, bs[0][4], luma_qp[1], a, b, 0, luma_intra_deblock );
494 deblock_edge_intra( h, pixuv + (strideuv<<offuv), 2*strideuv, bs[0][4], chroma_qp[1], a, b, c, chroma_intra_deblock );
495 deblock_edge_intra( h, pixuv + uvdiff + (strideuv<<offuv), 2*strideuv, bs[0][4], chroma_qp[1], a, b, c, chroma_intra_deblock );
499 deblock_edge( h, pixy + (stridey<<offy), 2*stridey, bs[0][4], luma_qp[1], a, b, 0, luma_deblock );
500 deblock_edge( h, pixuv + (strideuv<<offuv), 2*strideuv, bs[0][4], chroma_qp[1], a, b, c, chroma_deblock );
501 deblock_edge( h, pixuv + uvdiff + (strideuv<<offuv), 2*strideuv, bs[0][4], chroma_qp[1], a, b, c, chroma_deblock );
506 int qpl = h->mb.qp[h->mb.i_mb_xy-1];
507 int qp_left = (qp + qpl + 1) >> 1;
508 int qpc_left = (qpc + h->chroma_qp_table[qpl] + 1) >> 1;
509 int intra_left = IS_INTRA( h->mb.type[h->mb.i_mb_xy-1] );
511 if( intra_cur || intra_left )
512 FILTER( _intra, 0, 0, qp_left, qpc_left );
514 FILTER( , 0, 0, qp_left, qpc_left );
517 if( !first_edge_only )
519 FILTER( , 0, 1, qp, qpc );
520 FILTER( , 0, 2, qp, qpc );
521 FILTER( , 0, 3, qp, qpc );
524 if( h->mb.i_neighbour & MB_TOP )
526 if( b_interlaced && !(mb_y&1) && !MB_INTERLACED && h->mb.field[h->mb.i_mb_top_xy] )
528 int mbn_xy = mb_xy - 2 * h->mb.i_mb_stride;
530 for( int j = 0; j < 2; j++, mbn_xy += h->mb.i_mb_stride )
532 int qpt = h->mb.qp[mbn_xy];
533 int qp_top = (qp + qpt + 1) >> 1;
534 int qpc_top = (qpc + h->chroma_qp_table[qpt] + 1) >> 1;
535 int intra_top = IS_INTRA( h->mb.type[mbn_xy] );
536 if( intra_cur || intra_top )
537 M32( bs[1][4*j] ) = 0x03030303;
539 // deblock the first horizontal edge of the even rows, then the first horizontal edge of the odd rows
540 deblock_edge( h, pixy + j*stridey, 2* stridey, bs[1][4*j], qp_top, a, b, 0, h->loopf.deblock_luma[1] );
543 deblock_edge( h, pixuv + j*strideuv, 2*strideuv, bs[1][4*j], qpc_top, a, b, 0, h->loopf.deblock_luma[1] );
544 deblock_edge( h, pixuv + uvdiff + j*strideuv, 2*strideuv, bs[1][4*j], qpc_top, a, b, 0, h->loopf.deblock_luma[1] );
547 deblock_edge( h, pixuv + j*strideuv, 2*strideuv, bs[1][4*j], qpc_top, a, b, 1, h->loopf.deblock_chroma[1] );
552 int qpt = h->mb.qp[h->mb.i_mb_top_xy];
553 int qp_top = (qp + qpt + 1) >> 1;
554 int qpc_top = (qpc + h->chroma_qp_table[qpt] + 1) >> 1;
555 int intra_top = IS_INTRA( h->mb.type[h->mb.i_mb_top_xy] );
557 if( (!b_interlaced || (!MB_INTERLACED && !h->mb.field[h->mb.i_mb_top_xy]))
558 && (intra_cur || intra_top) )
560 FILTER( _intra, 1, 0, qp_top, qpc_top );
564 if( intra_cur || intra_top )
565 M32( bs[1][0] ) = 0x03030303;
566 FILTER( , 1, 0, qp_top, qpc_top );
571 if( !first_edge_only )
573 FILTER( , 1, 1, qp, qpc );
574 FILTER( , 1, 2, qp, qpc );
575 FILTER( , 1, 3, qp, qpc );
582 /* For deblock-aware RD.
584 * deblock macroblock edges
585 * support analysis partitions smaller than 16x16
586 * deblock chroma for 4:2:0/4:2:2
587 * handle duplicate refs correctly
588 * handle cavlc+8x8dct correctly
590 void x264_macroblock_deblock( x264_t *h )
592 int a = h->sh.i_alpha_c0_offset - QP_BD_OFFSET;
593 int b = h->sh.i_beta_offset - QP_BD_OFFSET;
594 int qp_thresh = 15 - X264_MIN( a, b ) - X264_MAX( 0, h->pps->i_chroma_qp_index_offset );
595 int intra_cur = IS_INTRA( h->mb.i_type );
597 int qpc = h->mb.i_chroma_qp;
598 if( (h->mb.i_partition == D_16x16 && !h->mb.i_cbp_luma && !intra_cur) || qp <= qp_thresh )
601 uint8_t (*bs)[8][4] = h->deblock_strength[h->mb.i_mb_y&1][h->mb.i_mb_x];
604 memset( &bs[0][1], 3, 3*4*sizeof(uint8_t) );
605 memset( &bs[1][1], 3, 3*4*sizeof(uint8_t) );
608 h->loopf.deblock_strength( h->mb.cache.non_zero_count, h->mb.cache.ref, h->mb.cache.mv,
609 bs, 4 >> MB_INTERLACED, h->sh.i_type == SLICE_TYPE_B );
611 int transform_8x8 = h->mb.b_transform_8x8;
613 #define FILTER( dir, edge )\
616 deblock_edge( h, h->mb.pic.p_fdec[0] + 4*edge*(dir?FDEC_STRIDE:1),\
617 FDEC_STRIDE, bs[dir][edge], qp, a, b, 0,\
618 h->loopf.deblock_luma[dir] );\
621 deblock_edge( h, h->mb.pic.p_fdec[1] + 4*edge*(dir?FDEC_STRIDE:1),\
622 FDEC_STRIDE, bs[dir][edge], qpc, a, b, 0,\
623 h->loopf.deblock_luma[dir] );\
624 deblock_edge( h, h->mb.pic.p_fdec[2] + 4*edge*(dir?FDEC_STRIDE:1),\
625 FDEC_STRIDE, bs[dir][edge], qpc, a, b, 0,\
626 h->loopf.deblock_luma[dir] );\
630 if( !transform_8x8 ) FILTER( 0, 1 );
632 if( !transform_8x8 ) FILTER( 0, 3 );
634 if( !transform_8x8 ) FILTER( 1, 1 );
636 if( !transform_8x8 ) FILTER( 1, 3 );
642 void x264_deblock_v_luma_sse2( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
643 void x264_deblock_v_luma_avx ( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
644 void x264_deblock_h_luma_sse2( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
645 void x264_deblock_h_luma_avx ( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
646 void x264_deblock_v_chroma_sse2( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
647 void x264_deblock_v_chroma_avx ( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
648 void x264_deblock_h_chroma_sse2( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
649 void x264_deblock_h_chroma_avx ( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
650 void x264_deblock_v_luma_intra_sse2( pixel *pix, int stride, int alpha, int beta );
651 void x264_deblock_v_luma_intra_avx ( pixel *pix, int stride, int alpha, int beta );
652 void x264_deblock_h_luma_intra_sse2( pixel *pix, int stride, int alpha, int beta );
653 void x264_deblock_h_luma_intra_avx ( pixel *pix, int stride, int alpha, int beta );
654 void x264_deblock_v_chroma_intra_sse2( pixel *pix, int stride, int alpha, int beta );
655 void x264_deblock_v_chroma_intra_avx ( pixel *pix, int stride, int alpha, int beta );
656 void x264_deblock_h_chroma_intra_sse2( pixel *pix, int stride, int alpha, int beta );
657 void x264_deblock_h_chroma_intra_avx ( pixel *pix, int stride, int alpha, int beta );
658 void x264_deblock_strength_mmx2 ( uint8_t nnz[X264_SCAN8_SIZE], int8_t ref[2][X264_SCAN8_LUMA_SIZE],
659 int16_t mv[2][X264_SCAN8_LUMA_SIZE][2], uint8_t bs[2][8][4],
660 int mvy_limit, int bframe );
661 void x264_deblock_strength_sse2 ( uint8_t nnz[X264_SCAN8_SIZE], int8_t ref[2][X264_SCAN8_LUMA_SIZE],
662 int16_t mv[2][X264_SCAN8_LUMA_SIZE][2], uint8_t bs[2][8][4],
663 int mvy_limit, int bframe );
664 void x264_deblock_strength_ssse3( uint8_t nnz[X264_SCAN8_SIZE], int8_t ref[2][X264_SCAN8_LUMA_SIZE],
665 int16_t mv[2][X264_SCAN8_LUMA_SIZE][2], uint8_t bs[2][8][4],
666 int mvy_limit, int bframe );
667 void x264_deblock_strength_avx ( uint8_t nnz[X264_SCAN8_SIZE], int8_t ref[2][X264_SCAN8_LUMA_SIZE],
668 int16_t mv[2][X264_SCAN8_LUMA_SIZE][2], uint8_t bs[2][8][4],
669 int mvy_limit, int bframe );
671 void x264_deblock_h_luma_mmx2( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
672 void x264_deblock_v8_luma_mmx2( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
673 void x264_deblock_v_chroma_mmx2( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
674 void x264_deblock_h_chroma_mmx2( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
675 void x264_deblock_h_luma_intra_mmx2( pixel *pix, int stride, int alpha, int beta );
676 void x264_deblock_v8_luma_intra_mmx2( uint8_t *pix, int stride, int alpha, int beta );
677 void x264_deblock_v_chroma_intra_mmx2( pixel *pix, int stride, int alpha, int beta );
678 void x264_deblock_h_chroma_intra_mmx2( pixel *pix, int stride, int alpha, int beta );
681 void x264_deblock_v_luma_mmx2( pixel *pix, int stride, int alpha, int beta, int8_t *tc0 );
682 void x264_deblock_v_luma_intra_mmx2( pixel *pix, int stride, int alpha, int beta );
684 // FIXME this wrapper has a significant cpu cost
685 static void x264_deblock_v_luma_mmx2( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 )
687 x264_deblock_v8_luma_mmx2( pix, stride, alpha, beta, tc0 );
688 x264_deblock_v8_luma_mmx2( pix+8, stride, alpha, beta, tc0+2 );
690 static void x264_deblock_v_luma_intra_mmx2( uint8_t *pix, int stride, int alpha, int beta )
692 x264_deblock_v8_luma_intra_mmx2( pix, stride, alpha, beta );
693 x264_deblock_v8_luma_intra_mmx2( pix+8, stride, alpha, beta );
695 #endif // HIGH_BIT_DEPTH
700 void x264_deblock_v_luma_altivec( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
701 void x264_deblock_h_luma_altivec( uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0 );
705 void x264_deblock_v_luma_neon( uint8_t *, int, int, int, int8_t * );
706 void x264_deblock_h_luma_neon( uint8_t *, int, int, int, int8_t * );
707 void x264_deblock_v_chroma_neon( uint8_t *, int, int, int, int8_t * );
708 void x264_deblock_h_chroma_neon( uint8_t *, int, int, int, int8_t * );
711 void x264_deblock_init( int cpu, x264_deblock_function_t *pf, int b_mbaff )
713 pf->deblock_luma[1] = deblock_v_luma_c;
714 pf->deblock_luma[0] = deblock_h_luma_c;
715 pf->deblock_chroma[1] = deblock_v_chroma_c;
716 pf->deblock_h_chroma_420 = deblock_h_chroma_c;
717 pf->deblock_h_chroma_422 = deblock_h_chroma_422_c;
718 pf->deblock_luma_intra[1] = deblock_v_luma_intra_c;
719 pf->deblock_luma_intra[0] = deblock_h_luma_intra_c;
720 pf->deblock_chroma_intra[1] = deblock_v_chroma_intra_c;
721 pf->deblock_h_chroma_420_intra = deblock_h_chroma_intra_c;
722 pf->deblock_h_chroma_422_intra = deblock_h_chroma_422_intra_c;
723 pf->deblock_luma_mbaff = deblock_h_luma_mbaff_c;
724 pf->deblock_chroma_420_mbaff = deblock_h_chroma_mbaff_c;
725 pf->deblock_chroma_422_mbaff = deblock_h_chroma_422_mbaff_c;
726 pf->deblock_luma_intra_mbaff = deblock_h_luma_intra_mbaff_c;
727 pf->deblock_chroma_420_intra_mbaff = deblock_h_chroma_intra_mbaff_c;
728 pf->deblock_chroma_422_intra_mbaff = deblock_h_chroma_422_intra_mbaff_c;
729 pf->deblock_strength = deblock_strength_c;
732 if( cpu&X264_CPU_MMX2 )
735 pf->deblock_luma[1] = x264_deblock_v_luma_mmx2;
736 pf->deblock_luma[0] = x264_deblock_h_luma_mmx2;
737 pf->deblock_chroma[1] = x264_deblock_v_chroma_mmx2;
738 pf->deblock_h_chroma_420 = x264_deblock_h_chroma_mmx2;
739 pf->deblock_luma_intra[1] = x264_deblock_v_luma_intra_mmx2;
740 pf->deblock_luma_intra[0] = x264_deblock_h_luma_intra_mmx2;
741 pf->deblock_chroma_intra[1] = x264_deblock_v_chroma_intra_mmx2;
742 pf->deblock_h_chroma_420_intra = x264_deblock_h_chroma_intra_mmx2;
744 pf->deblock_strength = x264_deblock_strength_mmx2;
745 if( cpu&X264_CPU_SSE2 )
747 pf->deblock_strength = x264_deblock_strength_sse2;
748 if( !(cpu&X264_CPU_STACK_MOD4) )
750 pf->deblock_luma[1] = x264_deblock_v_luma_sse2;
751 pf->deblock_luma[0] = x264_deblock_h_luma_sse2;
752 pf->deblock_chroma[1] = x264_deblock_v_chroma_sse2;
753 pf->deblock_h_chroma_420 = x264_deblock_h_chroma_sse2;
754 pf->deblock_luma_intra[1] = x264_deblock_v_luma_intra_sse2;
755 pf->deblock_luma_intra[0] = x264_deblock_h_luma_intra_sse2;
756 pf->deblock_chroma_intra[1] = x264_deblock_v_chroma_intra_sse2;
757 pf->deblock_h_chroma_420_intra = x264_deblock_h_chroma_intra_sse2;
760 if( cpu&X264_CPU_SSSE3 )
761 pf->deblock_strength = x264_deblock_strength_ssse3;
762 if( cpu&X264_CPU_AVX )
764 pf->deblock_strength = x264_deblock_strength_avx;
765 if( !(cpu&X264_CPU_STACK_MOD4) )
767 pf->deblock_luma[1] = x264_deblock_v_luma_avx;
768 pf->deblock_luma[0] = x264_deblock_h_luma_avx;
769 pf->deblock_chroma[1] = x264_deblock_v_chroma_avx;
770 pf->deblock_h_chroma_420 = x264_deblock_h_chroma_avx;
771 pf->deblock_luma_intra[1] = x264_deblock_v_luma_intra_avx;
772 pf->deblock_luma_intra[0] = x264_deblock_h_luma_intra_avx;
773 pf->deblock_chroma_intra[1] = x264_deblock_v_chroma_intra_avx;
774 pf->deblock_h_chroma_420_intra = x264_deblock_h_chroma_intra_avx;
782 if( cpu&X264_CPU_ALTIVEC )
784 pf->deblock_luma[1] = x264_deblock_v_luma_altivec;
785 pf->deblock_luma[0] = x264_deblock_h_luma_altivec;
787 #endif // HAVE_ALTIVEC
790 if( cpu&X264_CPU_NEON )
792 pf->deblock_luma[1] = x264_deblock_v_luma_neon;
793 pf->deblock_luma[0] = x264_deblock_h_luma_neon;
794 pf->deblock_chroma[1] = x264_deblock_v_chroma_neon;
795 pf->deblock_h_chroma_420 = x264_deblock_h_chroma_neon;
798 #endif // !HIGH_BIT_DEPTH