2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3 * Copyright (c) 2003-2011 Michael Niedermayer <michaelni@gmx.at>
5 * This file is part of Libav.
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * H.264 / AVC / MPEG4 part10 prediction functions.
25 * @author Michael Niedermayer <michaelni@gmx.at>
30 #include "bit_depth_template.c"
32 static void FUNCC(pred4x4_vertical)(uint8_t *_src, const uint8_t *topright, int _stride){
33 pixel *src = (pixel*)_src;
34 int stride = _stride/sizeof(pixel);
35 const pixel4 a= AV_RN4PA(src-stride);
37 AV_WN4PA(src+0*stride, a);
38 AV_WN4PA(src+1*stride, a);
39 AV_WN4PA(src+2*stride, a);
40 AV_WN4PA(src+3*stride, a);
43 static void FUNCC(pred4x4_horizontal)(uint8_t *_src, const uint8_t *topright, int _stride){
44 pixel *src = (pixel*)_src;
45 int stride = _stride/sizeof(pixel);
46 AV_WN4PA(src+0*stride, PIXEL_SPLAT_X4(src[-1+0*stride]));
47 AV_WN4PA(src+1*stride, PIXEL_SPLAT_X4(src[-1+1*stride]));
48 AV_WN4PA(src+2*stride, PIXEL_SPLAT_X4(src[-1+2*stride]));
49 AV_WN4PA(src+3*stride, PIXEL_SPLAT_X4(src[-1+3*stride]));
52 static void FUNCC(pred4x4_dc)(uint8_t *_src, const uint8_t *topright, int _stride){
53 pixel *src = (pixel*)_src;
54 int stride = _stride/sizeof(pixel);
55 const int dc= ( src[-stride] + src[1-stride] + src[2-stride] + src[3-stride]
56 + src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 4) >>3;
57 const pixel4 a = PIXEL_SPLAT_X4(dc);
59 AV_WN4PA(src+0*stride, a);
60 AV_WN4PA(src+1*stride, a);
61 AV_WN4PA(src+2*stride, a);
62 AV_WN4PA(src+3*stride, a);
65 static void FUNCC(pred4x4_left_dc)(uint8_t *_src, const uint8_t *topright, int _stride){
66 pixel *src = (pixel*)_src;
67 int stride = _stride/sizeof(pixel);
68 const int dc= ( src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 2) >>2;
69 const pixel4 a = PIXEL_SPLAT_X4(dc);
71 AV_WN4PA(src+0*stride, a);
72 AV_WN4PA(src+1*stride, a);
73 AV_WN4PA(src+2*stride, a);
74 AV_WN4PA(src+3*stride, a);
77 static void FUNCC(pred4x4_top_dc)(uint8_t *_src, const uint8_t *topright, int _stride){
78 pixel *src = (pixel*)_src;
79 int stride = _stride/sizeof(pixel);
80 const int dc= ( src[-stride] + src[1-stride] + src[2-stride] + src[3-stride] + 2) >>2;
81 const pixel4 a = PIXEL_SPLAT_X4(dc);
83 AV_WN4PA(src+0*stride, a);
84 AV_WN4PA(src+1*stride, a);
85 AV_WN4PA(src+2*stride, a);
86 AV_WN4PA(src+3*stride, a);
89 static void FUNCC(pred4x4_128_dc)(uint8_t *_src, const uint8_t *topright, int _stride){
90 pixel *src = (pixel*)_src;
91 int stride = _stride/sizeof(pixel);
92 const pixel4 a = PIXEL_SPLAT_X4(1<<(BIT_DEPTH-1));
94 AV_WN4PA(src+0*stride, a);
95 AV_WN4PA(src+1*stride, a);
96 AV_WN4PA(src+2*stride, a);
97 AV_WN4PA(src+3*stride, a);
100 static void FUNCC(pred4x4_127_dc)(uint8_t *_src, const uint8_t *topright, int _stride){
101 pixel *src = (pixel*)_src;
102 int stride = _stride/sizeof(pixel);
103 const pixel4 a = PIXEL_SPLAT_X4((1<<(BIT_DEPTH-1))-1);
105 AV_WN4PA(src+0*stride, a);
106 AV_WN4PA(src+1*stride, a);
107 AV_WN4PA(src+2*stride, a);
108 AV_WN4PA(src+3*stride, a);
111 static void FUNCC(pred4x4_129_dc)(uint8_t *_src, const uint8_t *topright, int _stride){
112 pixel *src = (pixel*)_src;
113 int stride = _stride/sizeof(pixel);
114 const pixel4 a = PIXEL_SPLAT_X4((1<<(BIT_DEPTH-1))+1);
116 AV_WN4PA(src+0*stride, a);
117 AV_WN4PA(src+1*stride, a);
118 AV_WN4PA(src+2*stride, a);
119 AV_WN4PA(src+3*stride, a);
123 #define LOAD_TOP_RIGHT_EDGE\
124 const unsigned av_unused t4 = topright[0];\
125 const unsigned av_unused t5 = topright[1];\
126 const unsigned av_unused t6 = topright[2];\
127 const unsigned av_unused t7 = topright[3];\
129 #define LOAD_DOWN_LEFT_EDGE\
130 const unsigned av_unused l4 = src[-1+4*stride];\
131 const unsigned av_unused l5 = src[-1+5*stride];\
132 const unsigned av_unused l6 = src[-1+6*stride];\
133 const unsigned av_unused l7 = src[-1+7*stride];\
135 #define LOAD_LEFT_EDGE\
136 const unsigned av_unused l0 = src[-1+0*stride];\
137 const unsigned av_unused l1 = src[-1+1*stride];\
138 const unsigned av_unused l2 = src[-1+2*stride];\
139 const unsigned av_unused l3 = src[-1+3*stride];\
141 #define LOAD_TOP_EDGE\
142 const unsigned av_unused t0 = src[ 0-1*stride];\
143 const unsigned av_unused t1 = src[ 1-1*stride];\
144 const unsigned av_unused t2 = src[ 2-1*stride];\
145 const unsigned av_unused t3 = src[ 3-1*stride];\
147 static void FUNCC(pred4x4_down_right)(uint8_t *_src, const uint8_t *topright, int _stride){
148 pixel *src = (pixel*)_src;
149 int stride = _stride/sizeof(pixel);
150 const int lt= src[-1-1*stride];
154 src[0+3*stride]=(l3 + 2*l2 + l1 + 2)>>2;
156 src[1+3*stride]=(l2 + 2*l1 + l0 + 2)>>2;
159 src[2+3*stride]=(l1 + 2*l0 + lt + 2)>>2;
163 src[3+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
166 src[3+2*stride]=(lt + 2*t0 + t1 + 2)>>2;
168 src[3+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
169 src[3+0*stride]=(t1 + 2*t2 + t3 + 2)>>2;
172 static void FUNCC(pred4x4_down_left)(uint8_t *_src, const uint8_t *_topright, int _stride){
173 pixel *src = (pixel*)_src;
174 const pixel *topright = (const pixel*)_topright;
175 int stride = _stride/sizeof(pixel);
180 src[0+0*stride]=(t0 + t2 + 2*t1 + 2)>>2;
182 src[0+1*stride]=(t1 + t3 + 2*t2 + 2)>>2;
185 src[0+2*stride]=(t2 + t4 + 2*t3 + 2)>>2;
189 src[0+3*stride]=(t3 + t5 + 2*t4 + 2)>>2;
192 src[1+3*stride]=(t4 + t6 + 2*t5 + 2)>>2;
194 src[2+3*stride]=(t5 + t7 + 2*t6 + 2)>>2;
195 src[3+3*stride]=(t6 + 3*t7 + 2)>>2;
198 static void FUNCC(pred4x4_vertical_right)(uint8_t *_src, const uint8_t *topright, int _stride){
199 pixel *src = (pixel*)_src;
200 int stride = _stride/sizeof(pixel);
201 const int lt= src[-1-1*stride];
206 src[1+2*stride]=(lt + t0 + 1)>>1;
208 src[2+2*stride]=(t0 + t1 + 1)>>1;
210 src[3+2*stride]=(t1 + t2 + 1)>>1;
211 src[3+0*stride]=(t2 + t3 + 1)>>1;
213 src[1+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
215 src[2+3*stride]=(lt + 2*t0 + t1 + 2)>>2;
217 src[3+3*stride]=(t0 + 2*t1 + t2 + 2)>>2;
218 src[3+1*stride]=(t1 + 2*t2 + t3 + 2)>>2;
219 src[0+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
220 src[0+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
223 static void FUNCC(pred4x4_vertical_left)(uint8_t *_src, const uint8_t *_topright, int _stride){
224 pixel *src = (pixel*)_src;
225 const pixel *topright = (const pixel*)_topright;
226 int stride = _stride/sizeof(pixel);
230 src[0+0*stride]=(t0 + t1 + 1)>>1;
232 src[0+2*stride]=(t1 + t2 + 1)>>1;
234 src[1+2*stride]=(t2 + t3 + 1)>>1;
236 src[2+2*stride]=(t3 + t4+ 1)>>1;
237 src[3+2*stride]=(t4 + t5+ 1)>>1;
238 src[0+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
240 src[0+3*stride]=(t1 + 2*t2 + t3 + 2)>>2;
242 src[1+3*stride]=(t2 + 2*t3 + t4 + 2)>>2;
244 src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
245 src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
248 static void FUNCC(pred4x4_horizontal_up)(uint8_t *_src, const uint8_t *topright, int _stride){
249 pixel *src = (pixel*)_src;
250 int stride = _stride/sizeof(pixel);
253 src[0+0*stride]=(l0 + l1 + 1)>>1;
254 src[1+0*stride]=(l0 + 2*l1 + l2 + 2)>>2;
256 src[0+1*stride]=(l1 + l2 + 1)>>1;
258 src[1+1*stride]=(l1 + 2*l2 + l3 + 2)>>2;
260 src[0+2*stride]=(l2 + l3 + 1)>>1;
262 src[1+2*stride]=(l2 + 2*l3 + l3 + 2)>>2;
271 static void FUNCC(pred4x4_horizontal_down)(uint8_t *_src, const uint8_t *topright, int _stride){
272 pixel *src = (pixel*)_src;
273 int stride = _stride/sizeof(pixel);
274 const int lt= src[-1-1*stride];
279 src[2+1*stride]=(lt + l0 + 1)>>1;
281 src[3+1*stride]=(l0 + 2*lt + t0 + 2)>>2;
282 src[2+0*stride]=(lt + 2*t0 + t1 + 2)>>2;
283 src[3+0*stride]=(t0 + 2*t1 + t2 + 2)>>2;
285 src[2+2*stride]=(l0 + l1 + 1)>>1;
287 src[3+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
289 src[2+3*stride]=(l1 + l2+ 1)>>1;
291 src[3+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
292 src[0+3*stride]=(l2 + l3 + 1)>>1;
293 src[1+3*stride]=(l1 + 2*l2 + l3 + 2)>>2;
296 static void FUNCC(pred16x16_vertical)(uint8_t *_src, int _stride){
298 pixel *src = (pixel*)_src;
299 int stride = _stride/sizeof(pixel);
300 const pixel4 a = AV_RN4PA(((pixel4*)(src-stride))+0);
301 const pixel4 b = AV_RN4PA(((pixel4*)(src-stride))+1);
302 const pixel4 c = AV_RN4PA(((pixel4*)(src-stride))+2);
303 const pixel4 d = AV_RN4PA(((pixel4*)(src-stride))+3);
306 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
307 AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
308 AV_WN4PA(((pixel4*)(src+i*stride))+2, c);
309 AV_WN4PA(((pixel4*)(src+i*stride))+3, d);
313 static void FUNCC(pred16x16_horizontal)(uint8_t *_src, int stride){
315 pixel *src = (pixel*)_src;
316 stride /= sizeof(pixel);
319 const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
321 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
322 AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
323 AV_WN4PA(((pixel4*)(src+i*stride))+2, a);
324 AV_WN4PA(((pixel4*)(src+i*stride))+3, a);
328 #define PREDICT_16x16_DC(v)\
329 for(i=0; i<16; i++){\
330 AV_WN4PA(src+ 0, v);\
331 AV_WN4PA(src+ 4, v);\
332 AV_WN4PA(src+ 8, v);\
333 AV_WN4PA(src+12, v);\
337 static void FUNCC(pred16x16_dc)(uint8_t *_src, int stride){
339 pixel *src = (pixel*)_src;
341 stride /= sizeof(pixel);
344 dc+= src[-1+i*stride];
351 dcsplat = PIXEL_SPLAT_X4((dc+16)>>5);
352 PREDICT_16x16_DC(dcsplat);
355 static void FUNCC(pred16x16_left_dc)(uint8_t *_src, int stride){
357 pixel *src = (pixel*)_src;
359 stride /= sizeof(pixel);
362 dc+= src[-1+i*stride];
365 dcsplat = PIXEL_SPLAT_X4((dc+8)>>4);
366 PREDICT_16x16_DC(dcsplat);
369 static void FUNCC(pred16x16_top_dc)(uint8_t *_src, int stride){
371 pixel *src = (pixel*)_src;
373 stride /= sizeof(pixel);
379 dcsplat = PIXEL_SPLAT_X4((dc+8)>>4);
380 PREDICT_16x16_DC(dcsplat);
383 #define PRED16x16_X(n, v) \
384 static void FUNCC(pred16x16_##n##_dc)(uint8_t *_src, int stride){\
386 pixel *src = (pixel*)_src;\
387 stride /= sizeof(pixel);\
388 PREDICT_16x16_DC(PIXEL_SPLAT_X4(v));\
391 PRED16x16_X(127, (1<<(BIT_DEPTH-1))-1);
392 PRED16x16_X(128, (1<<(BIT_DEPTH-1))+0);
393 PRED16x16_X(129, (1<<(BIT_DEPTH-1))+1);
395 static inline void FUNCC(pred16x16_plane_compat)(uint8_t *_src, int _stride, const int svq3, const int rv40){
399 pixel *src = (pixel*)_src;
400 int stride = _stride/sizeof(pixel);
401 const pixel * const src0 = src +7-stride;
402 const pixel * src1 = src +8*stride-1;
403 const pixel * src2 = src1-2*stride; // == src+6*stride-1;
404 int H = src0[1] - src0[-1];
405 int V = src1[0] - src2[ 0];
406 for(k=2; k<=8; ++k) {
407 src1 += stride; src2 -= stride;
408 H += k*(src0[k] - src0[-k]);
409 V += k*(src1[0] - src2[ 0]);
412 H = ( 5*(H/4) ) / 16;
413 V = ( 5*(V/4) ) / 16;
415 /* required for 100% accuracy */
418 H = ( H + (H>>2) ) >> 4;
419 V = ( V + (V>>2) ) >> 4;
425 a = 16*(src1[0] + src2[16] + 1) - 7*(V+H);
426 for(j=16; j>0; --j) {
429 for(i=-16; i<0; i+=4) {
430 src[16+i] = CLIP((b ) >> 5);
431 src[17+i] = CLIP((b+ H) >> 5);
432 src[18+i] = CLIP((b+2*H) >> 5);
433 src[19+i] = CLIP((b+3*H) >> 5);
440 static void FUNCC(pred16x16_plane)(uint8_t *src, int stride){
441 FUNCC(pred16x16_plane_compat)(src, stride, 0, 0);
444 static void FUNCC(pred8x8_vertical)(uint8_t *_src, int _stride){
446 pixel *src = (pixel*)_src;
447 int stride = _stride/sizeof(pixel);
448 const pixel4 a= AV_RN4PA(((pixel4*)(src-stride))+0);
449 const pixel4 b= AV_RN4PA(((pixel4*)(src-stride))+1);
452 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
453 AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
457 static void FUNCC(pred8x8_horizontal)(uint8_t *_src, int stride){
459 pixel *src = (pixel*)_src;
460 stride /= sizeof(pixel);
463 const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
464 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
465 AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
469 #define PRED8x8_X(n, v)\
470 static void FUNCC(pred8x8_##n##_dc)(uint8_t *_src, int stride){\
472 const pixel4 a = PIXEL_SPLAT_X4(v);\
473 pixel *src = (pixel*)_src;\
474 stride /= sizeof(pixel);\
476 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);\
477 AV_WN4PA(((pixel4*)(src+i*stride))+1, a);\
481 PRED8x8_X(127, (1<<(BIT_DEPTH-1))-1);
482 PRED8x8_X(128, (1<<(BIT_DEPTH-1))+0);
483 PRED8x8_X(129, (1<<(BIT_DEPTH-1))+1);
485 static void FUNCC(pred8x8_left_dc)(uint8_t *_src, int stride){
488 pixel4 dc0splat, dc2splat;
489 pixel *src = (pixel*)_src;
490 stride /= sizeof(pixel);
494 dc0+= src[-1+i*stride];
495 dc2+= src[-1+(i+4)*stride];
497 dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
498 dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
501 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
502 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc0splat);
505 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
506 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc2splat);
510 static void FUNCC(pred8x8_top_dc)(uint8_t *_src, int stride){
513 pixel4 dc0splat, dc1splat;
514 pixel *src = (pixel*)_src;
515 stride /= sizeof(pixel);
520 dc1+= src[4+i-stride];
522 dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
523 dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
526 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
527 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
530 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
531 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
535 static void FUNCC(pred8x8_dc)(uint8_t *_src, int stride){
538 pixel4 dc0splat, dc1splat, dc2splat, dc3splat;
539 pixel *src = (pixel*)_src;
540 stride /= sizeof(pixel);
544 dc0+= src[-1+i*stride] + src[i-stride];
545 dc1+= src[4+i-stride];
546 dc2+= src[-1+(i+4)*stride];
548 dc0splat = PIXEL_SPLAT_X4((dc0 + 4)>>3);
549 dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
550 dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
551 dc3splat = PIXEL_SPLAT_X4((dc1 + dc2 + 4)>>3);
554 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
555 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
558 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
559 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc3splat);
563 //the following 4 function should not be optimized!
564 static void FUNC(pred8x8_mad_cow_dc_l0t)(uint8_t *src, int stride){
565 FUNCC(pred8x8_top_dc)(src, stride);
566 FUNCC(pred4x4_dc)(src, NULL, stride);
569 static void FUNC(pred8x8_mad_cow_dc_0lt)(uint8_t *src, int stride){
570 FUNCC(pred8x8_dc)(src, stride);
571 FUNCC(pred4x4_top_dc)(src, NULL, stride);
574 static void FUNC(pred8x8_mad_cow_dc_l00)(uint8_t *src, int stride){
575 FUNCC(pred8x8_left_dc)(src, stride);
576 FUNCC(pred4x4_128_dc)(src + 4*stride , NULL, stride);
577 FUNCC(pred4x4_128_dc)(src + 4*stride + 4*sizeof(pixel), NULL, stride);
580 static void FUNC(pred8x8_mad_cow_dc_0l0)(uint8_t *src, int stride){
581 FUNCC(pred8x8_left_dc)(src, stride);
582 FUNCC(pred4x4_128_dc)(src , NULL, stride);
583 FUNCC(pred4x4_128_dc)(src + 4*sizeof(pixel), NULL, stride);
586 static void FUNCC(pred8x8_plane)(uint8_t *_src, int _stride){
590 pixel *src = (pixel*)_src;
591 int stride = _stride/sizeof(pixel);
592 const pixel * const src0 = src +3-stride;
593 const pixel * src1 = src +4*stride-1;
594 const pixel * src2 = src1-2*stride; // == src+2*stride-1;
595 int H = src0[1] - src0[-1];
596 int V = src1[0] - src2[ 0];
597 for(k=2; k<=4; ++k) {
598 src1 += stride; src2 -= stride;
599 H += k*(src0[k] - src0[-k]);
600 V += k*(src1[0] - src2[ 0]);
602 H = ( 17*H+16 ) >> 5;
603 V = ( 17*V+16 ) >> 5;
605 a = 16*(src1[0] + src2[8]+1) - 3*(V+H);
609 src[0] = CLIP((b ) >> 5);
610 src[1] = CLIP((b+ H) >> 5);
611 src[2] = CLIP((b+2*H) >> 5);
612 src[3] = CLIP((b+3*H) >> 5);
613 src[4] = CLIP((b+4*H) >> 5);
614 src[5] = CLIP((b+5*H) >> 5);
615 src[6] = CLIP((b+6*H) >> 5);
616 src[7] = CLIP((b+7*H) >> 5);
621 #define SRC(x,y) src[(x)+(y)*stride]
623 const int l##y = (SRC(-1,y-1) + 2*SRC(-1,y) + SRC(-1,y+1) + 2) >> 2;
624 #define PREDICT_8x8_LOAD_LEFT \
625 const int l0 = ((has_topleft ? SRC(-1,-1) : SRC(-1,0)) \
626 + 2*SRC(-1,0) + SRC(-1,1) + 2) >> 2; \
627 PL(1) PL(2) PL(3) PL(4) PL(5) PL(6) \
628 const int l7 av_unused = (SRC(-1,6) + 3*SRC(-1,7) + 2) >> 2
631 const int t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
632 #define PREDICT_8x8_LOAD_TOP \
633 const int t0 = ((has_topleft ? SRC(-1,-1) : SRC(0,-1)) \
634 + 2*SRC(0,-1) + SRC(1,-1) + 2) >> 2; \
635 PT(1) PT(2) PT(3) PT(4) PT(5) PT(6) \
636 const int t7 av_unused = ((has_topright ? SRC(8,-1) : SRC(7,-1)) \
637 + 2*SRC(7,-1) + SRC(6,-1) + 2) >> 2
640 t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
641 #define PREDICT_8x8_LOAD_TOPRIGHT \
642 int t8, t9, t10, t11, t12, t13, t14, t15; \
644 PTR(8) PTR(9) PTR(10) PTR(11) PTR(12) PTR(13) PTR(14) \
645 t15 = (SRC(14,-1) + 3*SRC(15,-1) + 2) >> 2; \
646 } else t8=t9=t10=t11=t12=t13=t14=t15= SRC(7,-1);
648 #define PREDICT_8x8_LOAD_TOPLEFT \
649 const int lt = (SRC(-1,0) + 2*SRC(-1,-1) + SRC(0,-1) + 2) >> 2
651 #define PREDICT_8x8_DC(v) \
653 for( y = 0; y < 8; y++ ) { \
654 AV_WN4PA(((pixel4*)src)+0, v); \
655 AV_WN4PA(((pixel4*)src)+1, v); \
659 static void FUNCC(pred8x8l_128_dc)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
661 pixel *src = (pixel*)_src;
662 int stride = _stride/sizeof(pixel);
664 PREDICT_8x8_DC(PIXEL_SPLAT_X4(1<<(BIT_DEPTH-1)));
666 static void FUNCC(pred8x8l_left_dc)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
668 pixel *src = (pixel*)_src;
669 int stride = _stride/sizeof(pixel);
671 PREDICT_8x8_LOAD_LEFT;
672 const pixel4 dc = PIXEL_SPLAT_X4((l0+l1+l2+l3+l4+l5+l6+l7+4) >> 3);
675 static void FUNCC(pred8x8l_top_dc)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
677 pixel *src = (pixel*)_src;
678 int stride = _stride/sizeof(pixel);
680 PREDICT_8x8_LOAD_TOP;
681 const pixel4 dc = PIXEL_SPLAT_X4((t0+t1+t2+t3+t4+t5+t6+t7+4) >> 3);
684 static void FUNCC(pred8x8l_dc)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
686 pixel *src = (pixel*)_src;
687 int stride = _stride/sizeof(pixel);
689 PREDICT_8x8_LOAD_LEFT;
690 PREDICT_8x8_LOAD_TOP;
691 const pixel4 dc = PIXEL_SPLAT_X4((l0+l1+l2+l3+l4+l5+l6+l7
692 +t0+t1+t2+t3+t4+t5+t6+t7+8) >> 4);
695 static void FUNCC(pred8x8l_horizontal)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
697 pixel *src = (pixel*)_src;
698 int stride = _stride/sizeof(pixel);
701 PREDICT_8x8_LOAD_LEFT;
702 #define ROW(y) a = PIXEL_SPLAT_X4(l##y); \
703 AV_WN4PA(src+y*stride, a); \
704 AV_WN4PA(src+y*stride+4, a);
705 ROW(0); ROW(1); ROW(2); ROW(3); ROW(4); ROW(5); ROW(6); ROW(7);
708 static void FUNCC(pred8x8l_vertical)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
711 pixel *src = (pixel*)_src;
712 int stride = _stride/sizeof(pixel);
715 PREDICT_8x8_LOAD_TOP;
724 a = AV_RN4PA(((pixel4*)src)+0);
725 b = AV_RN4PA(((pixel4*)src)+1);
726 for( y = 1; y < 8; y++ ) {
727 AV_WN4PA(((pixel4*)(src+y*stride))+0, a);
728 AV_WN4PA(((pixel4*)(src+y*stride))+1, b);
731 static void FUNCC(pred8x8l_down_left)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
733 pixel *src = (pixel*)_src;
734 int stride = _stride/sizeof(pixel);
735 PREDICT_8x8_LOAD_TOP;
736 PREDICT_8x8_LOAD_TOPRIGHT;
737 SRC(0,0)= (t0 + 2*t1 + t2 + 2) >> 2;
738 SRC(0,1)=SRC(1,0)= (t1 + 2*t2 + t3 + 2) >> 2;
739 SRC(0,2)=SRC(1,1)=SRC(2,0)= (t2 + 2*t3 + t4 + 2) >> 2;
740 SRC(0,3)=SRC(1,2)=SRC(2,1)=SRC(3,0)= (t3 + 2*t4 + t5 + 2) >> 2;
741 SRC(0,4)=SRC(1,3)=SRC(2,2)=SRC(3,1)=SRC(4,0)= (t4 + 2*t5 + t6 + 2) >> 2;
742 SRC(0,5)=SRC(1,4)=SRC(2,3)=SRC(3,2)=SRC(4,1)=SRC(5,0)= (t5 + 2*t6 + t7 + 2) >> 2;
743 SRC(0,6)=SRC(1,5)=SRC(2,4)=SRC(3,3)=SRC(4,2)=SRC(5,1)=SRC(6,0)= (t6 + 2*t7 + t8 + 2) >> 2;
744 SRC(0,7)=SRC(1,6)=SRC(2,5)=SRC(3,4)=SRC(4,3)=SRC(5,2)=SRC(6,1)=SRC(7,0)= (t7 + 2*t8 + t9 + 2) >> 2;
745 SRC(1,7)=SRC(2,6)=SRC(3,5)=SRC(4,4)=SRC(5,3)=SRC(6,2)=SRC(7,1)= (t8 + 2*t9 + t10 + 2) >> 2;
746 SRC(2,7)=SRC(3,6)=SRC(4,5)=SRC(5,4)=SRC(6,3)=SRC(7,2)= (t9 + 2*t10 + t11 + 2) >> 2;
747 SRC(3,7)=SRC(4,6)=SRC(5,5)=SRC(6,4)=SRC(7,3)= (t10 + 2*t11 + t12 + 2) >> 2;
748 SRC(4,7)=SRC(5,6)=SRC(6,5)=SRC(7,4)= (t11 + 2*t12 + t13 + 2) >> 2;
749 SRC(5,7)=SRC(6,6)=SRC(7,5)= (t12 + 2*t13 + t14 + 2) >> 2;
750 SRC(6,7)=SRC(7,6)= (t13 + 2*t14 + t15 + 2) >> 2;
751 SRC(7,7)= (t14 + 3*t15 + 2) >> 2;
753 static void FUNCC(pred8x8l_down_right)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
755 pixel *src = (pixel*)_src;
756 int stride = _stride/sizeof(pixel);
757 PREDICT_8x8_LOAD_TOP;
758 PREDICT_8x8_LOAD_LEFT;
759 PREDICT_8x8_LOAD_TOPLEFT;
760 SRC(0,7)= (l7 + 2*l6 + l5 + 2) >> 2;
761 SRC(0,6)=SRC(1,7)= (l6 + 2*l5 + l4 + 2) >> 2;
762 SRC(0,5)=SRC(1,6)=SRC(2,7)= (l5 + 2*l4 + l3 + 2) >> 2;
763 SRC(0,4)=SRC(1,5)=SRC(2,6)=SRC(3,7)= (l4 + 2*l3 + l2 + 2) >> 2;
764 SRC(0,3)=SRC(1,4)=SRC(2,5)=SRC(3,6)=SRC(4,7)= (l3 + 2*l2 + l1 + 2) >> 2;
765 SRC(0,2)=SRC(1,3)=SRC(2,4)=SRC(3,5)=SRC(4,6)=SRC(5,7)= (l2 + 2*l1 + l0 + 2) >> 2;
766 SRC(0,1)=SRC(1,2)=SRC(2,3)=SRC(3,4)=SRC(4,5)=SRC(5,6)=SRC(6,7)= (l1 + 2*l0 + lt + 2) >> 2;
767 SRC(0,0)=SRC(1,1)=SRC(2,2)=SRC(3,3)=SRC(4,4)=SRC(5,5)=SRC(6,6)=SRC(7,7)= (l0 + 2*lt + t0 + 2) >> 2;
768 SRC(1,0)=SRC(2,1)=SRC(3,2)=SRC(4,3)=SRC(5,4)=SRC(6,5)=SRC(7,6)= (lt + 2*t0 + t1 + 2) >> 2;
769 SRC(2,0)=SRC(3,1)=SRC(4,2)=SRC(5,3)=SRC(6,4)=SRC(7,5)= (t0 + 2*t1 + t2 + 2) >> 2;
770 SRC(3,0)=SRC(4,1)=SRC(5,2)=SRC(6,3)=SRC(7,4)= (t1 + 2*t2 + t3 + 2) >> 2;
771 SRC(4,0)=SRC(5,1)=SRC(6,2)=SRC(7,3)= (t2 + 2*t3 + t4 + 2) >> 2;
772 SRC(5,0)=SRC(6,1)=SRC(7,2)= (t3 + 2*t4 + t5 + 2) >> 2;
773 SRC(6,0)=SRC(7,1)= (t4 + 2*t5 + t6 + 2) >> 2;
774 SRC(7,0)= (t5 + 2*t6 + t7 + 2) >> 2;
776 static void FUNCC(pred8x8l_vertical_right)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
778 pixel *src = (pixel*)_src;
779 int stride = _stride/sizeof(pixel);
780 PREDICT_8x8_LOAD_TOP;
781 PREDICT_8x8_LOAD_LEFT;
782 PREDICT_8x8_LOAD_TOPLEFT;
783 SRC(0,6)= (l5 + 2*l4 + l3 + 2) >> 2;
784 SRC(0,7)= (l6 + 2*l5 + l4 + 2) >> 2;
785 SRC(0,4)=SRC(1,6)= (l3 + 2*l2 + l1 + 2) >> 2;
786 SRC(0,5)=SRC(1,7)= (l4 + 2*l3 + l2 + 2) >> 2;
787 SRC(0,2)=SRC(1,4)=SRC(2,6)= (l1 + 2*l0 + lt + 2) >> 2;
788 SRC(0,3)=SRC(1,5)=SRC(2,7)= (l2 + 2*l1 + l0 + 2) >> 2;
789 SRC(0,1)=SRC(1,3)=SRC(2,5)=SRC(3,7)= (l0 + 2*lt + t0 + 2) >> 2;
790 SRC(0,0)=SRC(1,2)=SRC(2,4)=SRC(3,6)= (lt + t0 + 1) >> 1;
791 SRC(1,1)=SRC(2,3)=SRC(3,5)=SRC(4,7)= (lt + 2*t0 + t1 + 2) >> 2;
792 SRC(1,0)=SRC(2,2)=SRC(3,4)=SRC(4,6)= (t0 + t1 + 1) >> 1;
793 SRC(2,1)=SRC(3,3)=SRC(4,5)=SRC(5,7)= (t0 + 2*t1 + t2 + 2) >> 2;
794 SRC(2,0)=SRC(3,2)=SRC(4,4)=SRC(5,6)= (t1 + t2 + 1) >> 1;
795 SRC(3,1)=SRC(4,3)=SRC(5,5)=SRC(6,7)= (t1 + 2*t2 + t3 + 2) >> 2;
796 SRC(3,0)=SRC(4,2)=SRC(5,4)=SRC(6,6)= (t2 + t3 + 1) >> 1;
797 SRC(4,1)=SRC(5,3)=SRC(6,5)=SRC(7,7)= (t2 + 2*t3 + t4 + 2) >> 2;
798 SRC(4,0)=SRC(5,2)=SRC(6,4)=SRC(7,6)= (t3 + t4 + 1) >> 1;
799 SRC(5,1)=SRC(6,3)=SRC(7,5)= (t3 + 2*t4 + t5 + 2) >> 2;
800 SRC(5,0)=SRC(6,2)=SRC(7,4)= (t4 + t5 + 1) >> 1;
801 SRC(6,1)=SRC(7,3)= (t4 + 2*t5 + t6 + 2) >> 2;
802 SRC(6,0)=SRC(7,2)= (t5 + t6 + 1) >> 1;
803 SRC(7,1)= (t5 + 2*t6 + t7 + 2) >> 2;
804 SRC(7,0)= (t6 + t7 + 1) >> 1;
806 static void FUNCC(pred8x8l_horizontal_down)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
808 pixel *src = (pixel*)_src;
809 int stride = _stride/sizeof(pixel);
810 PREDICT_8x8_LOAD_TOP;
811 PREDICT_8x8_LOAD_LEFT;
812 PREDICT_8x8_LOAD_TOPLEFT;
813 SRC(0,7)= (l6 + l7 + 1) >> 1;
814 SRC(1,7)= (l5 + 2*l6 + l7 + 2) >> 2;
815 SRC(0,6)=SRC(2,7)= (l5 + l6 + 1) >> 1;
816 SRC(1,6)=SRC(3,7)= (l4 + 2*l5 + l6 + 2) >> 2;
817 SRC(0,5)=SRC(2,6)=SRC(4,7)= (l4 + l5 + 1) >> 1;
818 SRC(1,5)=SRC(3,6)=SRC(5,7)= (l3 + 2*l4 + l5 + 2) >> 2;
819 SRC(0,4)=SRC(2,5)=SRC(4,6)=SRC(6,7)= (l3 + l4 + 1) >> 1;
820 SRC(1,4)=SRC(3,5)=SRC(5,6)=SRC(7,7)= (l2 + 2*l3 + l4 + 2) >> 2;
821 SRC(0,3)=SRC(2,4)=SRC(4,5)=SRC(6,6)= (l2 + l3 + 1) >> 1;
822 SRC(1,3)=SRC(3,4)=SRC(5,5)=SRC(7,6)= (l1 + 2*l2 + l3 + 2) >> 2;
823 SRC(0,2)=SRC(2,3)=SRC(4,4)=SRC(6,5)= (l1 + l2 + 1) >> 1;
824 SRC(1,2)=SRC(3,3)=SRC(5,4)=SRC(7,5)= (l0 + 2*l1 + l2 + 2) >> 2;
825 SRC(0,1)=SRC(2,2)=SRC(4,3)=SRC(6,4)= (l0 + l1 + 1) >> 1;
826 SRC(1,1)=SRC(3,2)=SRC(5,3)=SRC(7,4)= (lt + 2*l0 + l1 + 2) >> 2;
827 SRC(0,0)=SRC(2,1)=SRC(4,2)=SRC(6,3)= (lt + l0 + 1) >> 1;
828 SRC(1,0)=SRC(3,1)=SRC(5,2)=SRC(7,3)= (l0 + 2*lt + t0 + 2) >> 2;
829 SRC(2,0)=SRC(4,1)=SRC(6,2)= (t1 + 2*t0 + lt + 2) >> 2;
830 SRC(3,0)=SRC(5,1)=SRC(7,2)= (t2 + 2*t1 + t0 + 2) >> 2;
831 SRC(4,0)=SRC(6,1)= (t3 + 2*t2 + t1 + 2) >> 2;
832 SRC(5,0)=SRC(7,1)= (t4 + 2*t3 + t2 + 2) >> 2;
833 SRC(6,0)= (t5 + 2*t4 + t3 + 2) >> 2;
834 SRC(7,0)= (t6 + 2*t5 + t4 + 2) >> 2;
836 static void FUNCC(pred8x8l_vertical_left)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
838 pixel *src = (pixel*)_src;
839 int stride = _stride/sizeof(pixel);
840 PREDICT_8x8_LOAD_TOP;
841 PREDICT_8x8_LOAD_TOPRIGHT;
842 SRC(0,0)= (t0 + t1 + 1) >> 1;
843 SRC(0,1)= (t0 + 2*t1 + t2 + 2) >> 2;
844 SRC(0,2)=SRC(1,0)= (t1 + t2 + 1) >> 1;
845 SRC(0,3)=SRC(1,1)= (t1 + 2*t2 + t3 + 2) >> 2;
846 SRC(0,4)=SRC(1,2)=SRC(2,0)= (t2 + t3 + 1) >> 1;
847 SRC(0,5)=SRC(1,3)=SRC(2,1)= (t2 + 2*t3 + t4 + 2) >> 2;
848 SRC(0,6)=SRC(1,4)=SRC(2,2)=SRC(3,0)= (t3 + t4 + 1) >> 1;
849 SRC(0,7)=SRC(1,5)=SRC(2,3)=SRC(3,1)= (t3 + 2*t4 + t5 + 2) >> 2;
850 SRC(1,6)=SRC(2,4)=SRC(3,2)=SRC(4,0)= (t4 + t5 + 1) >> 1;
851 SRC(1,7)=SRC(2,5)=SRC(3,3)=SRC(4,1)= (t4 + 2*t5 + t6 + 2) >> 2;
852 SRC(2,6)=SRC(3,4)=SRC(4,2)=SRC(5,0)= (t5 + t6 + 1) >> 1;
853 SRC(2,7)=SRC(3,5)=SRC(4,3)=SRC(5,1)= (t5 + 2*t6 + t7 + 2) >> 2;
854 SRC(3,6)=SRC(4,4)=SRC(5,2)=SRC(6,0)= (t6 + t7 + 1) >> 1;
855 SRC(3,7)=SRC(4,5)=SRC(5,3)=SRC(6,1)= (t6 + 2*t7 + t8 + 2) >> 2;
856 SRC(4,6)=SRC(5,4)=SRC(6,2)=SRC(7,0)= (t7 + t8 + 1) >> 1;
857 SRC(4,7)=SRC(5,5)=SRC(6,3)=SRC(7,1)= (t7 + 2*t8 + t9 + 2) >> 2;
858 SRC(5,6)=SRC(6,4)=SRC(7,2)= (t8 + t9 + 1) >> 1;
859 SRC(5,7)=SRC(6,5)=SRC(7,3)= (t8 + 2*t9 + t10 + 2) >> 2;
860 SRC(6,6)=SRC(7,4)= (t9 + t10 + 1) >> 1;
861 SRC(6,7)=SRC(7,5)= (t9 + 2*t10 + t11 + 2) >> 2;
862 SRC(7,6)= (t10 + t11 + 1) >> 1;
863 SRC(7,7)= (t10 + 2*t11 + t12 + 2) >> 2;
865 static void FUNCC(pred8x8l_horizontal_up)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
867 pixel *src = (pixel*)_src;
868 int stride = _stride/sizeof(pixel);
869 PREDICT_8x8_LOAD_LEFT;
870 SRC(0,0)= (l0 + l1 + 1) >> 1;
871 SRC(1,0)= (l0 + 2*l1 + l2 + 2) >> 2;
872 SRC(0,1)=SRC(2,0)= (l1 + l2 + 1) >> 1;
873 SRC(1,1)=SRC(3,0)= (l1 + 2*l2 + l3 + 2) >> 2;
874 SRC(0,2)=SRC(2,1)=SRC(4,0)= (l2 + l3 + 1) >> 1;
875 SRC(1,2)=SRC(3,1)=SRC(5,0)= (l2 + 2*l3 + l4 + 2) >> 2;
876 SRC(0,3)=SRC(2,2)=SRC(4,1)=SRC(6,0)= (l3 + l4 + 1) >> 1;
877 SRC(1,3)=SRC(3,2)=SRC(5,1)=SRC(7,0)= (l3 + 2*l4 + l5 + 2) >> 2;
878 SRC(0,4)=SRC(2,3)=SRC(4,2)=SRC(6,1)= (l4 + l5 + 1) >> 1;
879 SRC(1,4)=SRC(3,3)=SRC(5,2)=SRC(7,1)= (l4 + 2*l5 + l6 + 2) >> 2;
880 SRC(0,5)=SRC(2,4)=SRC(4,3)=SRC(6,2)= (l5 + l6 + 1) >> 1;
881 SRC(1,5)=SRC(3,4)=SRC(5,3)=SRC(7,2)= (l5 + 2*l6 + l7 + 2) >> 2;
882 SRC(0,6)=SRC(2,5)=SRC(4,4)=SRC(6,3)= (l6 + l7 + 1) >> 1;
883 SRC(1,6)=SRC(3,5)=SRC(5,4)=SRC(7,3)= (l6 + 3*l7 + 2) >> 2;
884 SRC(0,7)=SRC(1,7)=SRC(2,6)=SRC(2,7)=SRC(3,6)=
885 SRC(3,7)=SRC(4,5)=SRC(4,6)=SRC(4,7)=SRC(5,5)=
886 SRC(5,6)=SRC(5,7)=SRC(6,4)=SRC(6,5)=SRC(6,6)=
887 SRC(6,7)=SRC(7,4)=SRC(7,5)=SRC(7,6)=SRC(7,7)= l7;
889 #undef PREDICT_8x8_LOAD_LEFT
890 #undef PREDICT_8x8_LOAD_TOP
891 #undef PREDICT_8x8_LOAD_TOPLEFT
892 #undef PREDICT_8x8_LOAD_TOPRIGHT
893 #undef PREDICT_8x8_DC
899 static void FUNCC(pred4x4_vertical_add)(uint8_t *_pix, const DCTELEM *_block, int stride){
901 pixel *pix = (pixel*)_pix;
902 const dctcoef *block = (const dctcoef*)_block;
903 stride /= sizeof(pixel);
907 pix[1*stride]= v += block[0];
908 pix[2*stride]= v += block[4];
909 pix[3*stride]= v += block[8];
910 pix[4*stride]= v + block[12];
916 static void FUNCC(pred4x4_horizontal_add)(uint8_t *_pix, const DCTELEM *_block, int stride){
918 pixel *pix = (pixel*)_pix;
919 const dctcoef *block = (const dctcoef*)_block;
920 stride /= sizeof(pixel);
923 pix[0]= v += block[0];
924 pix[1]= v += block[1];
925 pix[2]= v += block[2];
926 pix[3]= v + block[3];
932 static void FUNCC(pred8x8l_vertical_add)(uint8_t *_pix, const DCTELEM *_block, int stride){
934 pixel *pix = (pixel*)_pix;
935 const dctcoef *block = (const dctcoef*)_block;
936 stride /= sizeof(pixel);
940 pix[1*stride]= v += block[0];
941 pix[2*stride]= v += block[8];
942 pix[3*stride]= v += block[16];
943 pix[4*stride]= v += block[24];
944 pix[5*stride]= v += block[32];
945 pix[6*stride]= v += block[40];
946 pix[7*stride]= v += block[48];
947 pix[8*stride]= v + block[56];
953 static void FUNCC(pred8x8l_horizontal_add)(uint8_t *_pix, const DCTELEM *_block, int stride){
955 pixel *pix = (pixel*)_pix;
956 const dctcoef *block = (const dctcoef*)_block;
957 stride /= sizeof(pixel);
960 pix[0]= v += block[0];
961 pix[1]= v += block[1];
962 pix[2]= v += block[2];
963 pix[3]= v += block[3];
964 pix[4]= v += block[4];
965 pix[5]= v += block[5];
966 pix[6]= v += block[6];
967 pix[7]= v + block[7];
973 static void FUNCC(pred16x16_vertical_add)(uint8_t *pix, const int *block_offset, const DCTELEM *block, int stride){
976 FUNCC(pred4x4_vertical_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
979 static void FUNCC(pred16x16_horizontal_add)(uint8_t *pix, const int *block_offset, const DCTELEM *block, int stride){
982 FUNCC(pred4x4_horizontal_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
985 static void FUNCC(pred8x8_vertical_add)(uint8_t *pix, const int *block_offset, const DCTELEM *block, int stride){
988 FUNCC(pred4x4_vertical_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
991 static void FUNCC(pred8x8_horizontal_add)(uint8_t *pix, const int *block_offset, const DCTELEM *block, int stride){
994 FUNCC(pred4x4_horizontal_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);