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 FFmpeg.
7 * FFmpeg 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 * FFmpeg 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 FFmpeg; 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,
35 pixel *src = (pixel*)_src;
36 int stride = _stride>>(sizeof(pixel)-1);
37 const pixel4 a= AV_RN4PA(src-stride);
39 AV_WN4PA(src+0*stride, a);
40 AV_WN4PA(src+1*stride, a);
41 AV_WN4PA(src+2*stride, a);
42 AV_WN4PA(src+3*stride, a);
45 static void FUNCC(pred4x4_horizontal)(uint8_t *_src, const uint8_t *topright,
48 pixel *src = (pixel*)_src;
49 int stride = _stride>>(sizeof(pixel)-1);
50 AV_WN4PA(src+0*stride, PIXEL_SPLAT_X4(src[-1+0*stride]));
51 AV_WN4PA(src+1*stride, PIXEL_SPLAT_X4(src[-1+1*stride]));
52 AV_WN4PA(src+2*stride, PIXEL_SPLAT_X4(src[-1+2*stride]));
53 AV_WN4PA(src+3*stride, PIXEL_SPLAT_X4(src[-1+3*stride]));
56 static void FUNCC(pred4x4_dc)(uint8_t *_src, const uint8_t *topright,
59 pixel *src = (pixel*)_src;
60 int stride = _stride>>(sizeof(pixel)-1);
61 const int dc= ( src[-stride] + src[1-stride] + src[2-stride] + src[3-stride]
62 + src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 4) >>3;
63 const pixel4 a = PIXEL_SPLAT_X4(dc);
65 AV_WN4PA(src+0*stride, a);
66 AV_WN4PA(src+1*stride, a);
67 AV_WN4PA(src+2*stride, a);
68 AV_WN4PA(src+3*stride, a);
71 static void FUNCC(pred4x4_left_dc)(uint8_t *_src, const uint8_t *topright,
74 pixel *src = (pixel*)_src;
75 int stride = _stride>>(sizeof(pixel)-1);
76 const int dc= ( src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 2) >>2;
77 const pixel4 a = PIXEL_SPLAT_X4(dc);
79 AV_WN4PA(src+0*stride, a);
80 AV_WN4PA(src+1*stride, a);
81 AV_WN4PA(src+2*stride, a);
82 AV_WN4PA(src+3*stride, a);
85 static void FUNCC(pred4x4_top_dc)(uint8_t *_src, const uint8_t *topright,
88 pixel *src = (pixel*)_src;
89 int stride = _stride>>(sizeof(pixel)-1);
90 const int dc= ( src[-stride] + src[1-stride] + src[2-stride] + src[3-stride] + 2) >>2;
91 const pixel4 a = PIXEL_SPLAT_X4(dc);
93 AV_WN4PA(src+0*stride, a);
94 AV_WN4PA(src+1*stride, a);
95 AV_WN4PA(src+2*stride, a);
96 AV_WN4PA(src+3*stride, a);
99 static void FUNCC(pred4x4_128_dc)(uint8_t *_src, const uint8_t *topright,
102 pixel *src = (pixel*)_src;
103 int stride = _stride>>(sizeof(pixel)-1);
104 const pixel4 a = PIXEL_SPLAT_X4(1<<(BIT_DEPTH-1));
106 AV_WN4PA(src+0*stride, a);
107 AV_WN4PA(src+1*stride, a);
108 AV_WN4PA(src+2*stride, a);
109 AV_WN4PA(src+3*stride, a);
112 static void FUNCC(pred4x4_127_dc)(uint8_t *_src, const uint8_t *topright,
115 pixel *src = (pixel*)_src;
116 int stride = _stride>>(sizeof(pixel)-1);
117 const pixel4 a = PIXEL_SPLAT_X4((1<<(BIT_DEPTH-1))-1);
119 AV_WN4PA(src+0*stride, a);
120 AV_WN4PA(src+1*stride, a);
121 AV_WN4PA(src+2*stride, a);
122 AV_WN4PA(src+3*stride, a);
125 static void FUNCC(pred4x4_129_dc)(uint8_t *_src, const uint8_t *topright,
128 pixel *src = (pixel*)_src;
129 int stride = _stride>>(sizeof(pixel)-1);
130 const pixel4 a = PIXEL_SPLAT_X4((1<<(BIT_DEPTH-1))+1);
132 AV_WN4PA(src+0*stride, a);
133 AV_WN4PA(src+1*stride, a);
134 AV_WN4PA(src+2*stride, a);
135 AV_WN4PA(src+3*stride, a);
139 #define LOAD_TOP_RIGHT_EDGE\
140 const unsigned av_unused t4 = topright[0];\
141 const unsigned av_unused t5 = topright[1];\
142 const unsigned av_unused t6 = topright[2];\
143 const unsigned av_unused t7 = topright[3];\
145 #define LOAD_DOWN_LEFT_EDGE\
146 const unsigned av_unused l4 = src[-1+4*stride];\
147 const unsigned av_unused l5 = src[-1+5*stride];\
148 const unsigned av_unused l6 = src[-1+6*stride];\
149 const unsigned av_unused l7 = src[-1+7*stride];\
151 #define LOAD_LEFT_EDGE\
152 const unsigned av_unused l0 = src[-1+0*stride];\
153 const unsigned av_unused l1 = src[-1+1*stride];\
154 const unsigned av_unused l2 = src[-1+2*stride];\
155 const unsigned av_unused l3 = src[-1+3*stride];\
157 #define LOAD_TOP_EDGE\
158 const unsigned av_unused t0 = src[ 0-1*stride];\
159 const unsigned av_unused t1 = src[ 1-1*stride];\
160 const unsigned av_unused t2 = src[ 2-1*stride];\
161 const unsigned av_unused t3 = src[ 3-1*stride];\
163 static void FUNCC(pred4x4_down_right)(uint8_t *_src, const uint8_t *topright,
166 pixel *src = (pixel*)_src;
167 int stride = _stride>>(sizeof(pixel)-1);
168 const int lt= src[-1-1*stride];
172 src[0+3*stride]=(l3 + 2*l2 + l1 + 2)>>2;
174 src[1+3*stride]=(l2 + 2*l1 + l0 + 2)>>2;
177 src[2+3*stride]=(l1 + 2*l0 + lt + 2)>>2;
181 src[3+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
184 src[3+2*stride]=(lt + 2*t0 + t1 + 2)>>2;
186 src[3+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
187 src[3+0*stride]=(t1 + 2*t2 + t3 + 2)>>2;
190 static void FUNCC(pred4x4_down_left)(uint8_t *_src, const uint8_t *_topright,
193 pixel *src = (pixel*)_src;
194 const pixel *topright = (const pixel*)_topright;
195 int stride = _stride>>(sizeof(pixel)-1);
200 src[0+0*stride]=(t0 + t2 + 2*t1 + 2)>>2;
202 src[0+1*stride]=(t1 + t3 + 2*t2 + 2)>>2;
205 src[0+2*stride]=(t2 + t4 + 2*t3 + 2)>>2;
209 src[0+3*stride]=(t3 + t5 + 2*t4 + 2)>>2;
212 src[1+3*stride]=(t4 + t6 + 2*t5 + 2)>>2;
214 src[2+3*stride]=(t5 + t7 + 2*t6 + 2)>>2;
215 src[3+3*stride]=(t6 + 3*t7 + 2)>>2;
218 static void FUNCC(pred4x4_vertical_right)(uint8_t *_src,
219 const uint8_t *topright,
222 pixel *src = (pixel*)_src;
223 int stride = _stride>>(sizeof(pixel)-1);
224 const int lt= src[-1-1*stride];
229 src[1+2*stride]=(lt + t0 + 1)>>1;
231 src[2+2*stride]=(t0 + t1 + 1)>>1;
233 src[3+2*stride]=(t1 + t2 + 1)>>1;
234 src[3+0*stride]=(t2 + t3 + 1)>>1;
236 src[1+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
238 src[2+3*stride]=(lt + 2*t0 + t1 + 2)>>2;
240 src[3+3*stride]=(t0 + 2*t1 + t2 + 2)>>2;
241 src[3+1*stride]=(t1 + 2*t2 + t3 + 2)>>2;
242 src[0+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
243 src[0+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
246 static void FUNCC(pred4x4_vertical_left)(uint8_t *_src,
247 const uint8_t *_topright,
250 pixel *src = (pixel*)_src;
251 const pixel *topright = (const pixel*)_topright;
252 int stride = _stride>>(sizeof(pixel)-1);
256 src[0+0*stride]=(t0 + t1 + 1)>>1;
258 src[0+2*stride]=(t1 + t2 + 1)>>1;
260 src[1+2*stride]=(t2 + t3 + 1)>>1;
262 src[2+2*stride]=(t3 + t4+ 1)>>1;
263 src[3+2*stride]=(t4 + t5+ 1)>>1;
264 src[0+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
266 src[0+3*stride]=(t1 + 2*t2 + t3 + 2)>>2;
268 src[1+3*stride]=(t2 + 2*t3 + t4 + 2)>>2;
270 src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
271 src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
274 static void FUNCC(pred4x4_horizontal_up)(uint8_t *_src, const uint8_t *topright,
277 pixel *src = (pixel*)_src;
278 int stride = _stride>>(sizeof(pixel)-1);
281 src[0+0*stride]=(l0 + l1 + 1)>>1;
282 src[1+0*stride]=(l0 + 2*l1 + l2 + 2)>>2;
284 src[0+1*stride]=(l1 + l2 + 1)>>1;
286 src[1+1*stride]=(l1 + 2*l2 + l3 + 2)>>2;
288 src[0+2*stride]=(l2 + l3 + 1)>>1;
290 src[1+2*stride]=(l2 + 2*l3 + l3 + 2)>>2;
299 static void FUNCC(pred4x4_horizontal_down)(uint8_t *_src,
300 const uint8_t *topright,
303 pixel *src = (pixel*)_src;
304 int stride = _stride>>(sizeof(pixel)-1);
305 const int lt= src[-1-1*stride];
310 src[2+1*stride]=(lt + l0 + 1)>>1;
312 src[3+1*stride]=(l0 + 2*lt + t0 + 2)>>2;
313 src[2+0*stride]=(lt + 2*t0 + t1 + 2)>>2;
314 src[3+0*stride]=(t0 + 2*t1 + t2 + 2)>>2;
316 src[2+2*stride]=(l0 + l1 + 1)>>1;
318 src[3+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
320 src[2+3*stride]=(l1 + l2+ 1)>>1;
322 src[3+3*stride]=(l0 + 2*l1 + l2 + 2)>>2;
323 src[0+3*stride]=(l2 + l3 + 1)>>1;
324 src[1+3*stride]=(l1 + 2*l2 + l3 + 2)>>2;
327 static void FUNCC(pred16x16_vertical)(uint8_t *_src, ptrdiff_t _stride)
330 pixel *src = (pixel*)_src;
331 int stride = _stride>>(sizeof(pixel)-1);
332 const pixel4 a = AV_RN4PA(((pixel4*)(src-stride))+0);
333 const pixel4 b = AV_RN4PA(((pixel4*)(src-stride))+1);
334 const pixel4 c = AV_RN4PA(((pixel4*)(src-stride))+2);
335 const pixel4 d = AV_RN4PA(((pixel4*)(src-stride))+3);
338 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
339 AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
340 AV_WN4PA(((pixel4*)(src+i*stride))+2, c);
341 AV_WN4PA(((pixel4*)(src+i*stride))+3, d);
345 static void FUNCC(pred16x16_horizontal)(uint8_t *_src, ptrdiff_t stride)
348 pixel *src = (pixel*)_src;
349 stride >>= sizeof(pixel)-1;
352 const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
354 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
355 AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
356 AV_WN4PA(((pixel4*)(src+i*stride))+2, a);
357 AV_WN4PA(((pixel4*)(src+i*stride))+3, a);
361 #define PREDICT_16x16_DC(v)\
362 for(i=0; i<16; i++){\
363 AV_WN4PA(src+ 0, v);\
364 AV_WN4PA(src+ 4, v);\
365 AV_WN4PA(src+ 8, v);\
366 AV_WN4PA(src+12, v);\
370 static void FUNCC(pred16x16_dc)(uint8_t *_src, ptrdiff_t stride)
373 pixel *src = (pixel*)_src;
375 stride >>= sizeof(pixel)-1;
378 dc+= src[-1+i*stride];
385 dcsplat = PIXEL_SPLAT_X4((dc+16)>>5);
386 PREDICT_16x16_DC(dcsplat);
389 static void FUNCC(pred16x16_left_dc)(uint8_t *_src, ptrdiff_t stride)
392 pixel *src = (pixel*)_src;
394 stride >>= sizeof(pixel)-1;
397 dc+= src[-1+i*stride];
400 dcsplat = PIXEL_SPLAT_X4((dc+8)>>4);
401 PREDICT_16x16_DC(dcsplat);
404 static void FUNCC(pred16x16_top_dc)(uint8_t *_src, ptrdiff_t stride)
407 pixel *src = (pixel*)_src;
409 stride >>= sizeof(pixel)-1;
415 dcsplat = PIXEL_SPLAT_X4((dc+8)>>4);
416 PREDICT_16x16_DC(dcsplat);
419 #define PRED16x16_X(n, v) \
420 static void FUNCC(pred16x16_##n##_dc)(uint8_t *_src, ptrdiff_t stride)\
423 pixel *src = (pixel*)_src;\
424 stride >>= sizeof(pixel)-1;\
425 PREDICT_16x16_DC(PIXEL_SPLAT_X4(v));\
428 PRED16x16_X(127, (1<<(BIT_DEPTH-1))-1)
429 PRED16x16_X(128, (1<<(BIT_DEPTH-1))+0)
430 PRED16x16_X(129, (1<<(BIT_DEPTH-1))+1)
432 static inline void FUNCC(pred16x16_plane_compat)(uint8_t *_src,
440 pixel *src = (pixel*)_src;
441 int stride = _stride>>(sizeof(pixel)-1);
442 const pixel * const src0 = src +7-stride;
443 const pixel * src1 = src +8*stride-1;
444 const pixel * src2 = src1-2*stride; // == src+6*stride-1;
445 int H = src0[1] - src0[-1];
446 int V = src1[0] - src2[ 0];
447 for(k=2; k<=8; ++k) {
448 src1 += stride; src2 -= stride;
449 H += k*(src0[k] - src0[-k]);
450 V += k*(src1[0] - src2[ 0]);
453 H = ( 5*(H/4) ) / 16;
454 V = ( 5*(V/4) ) / 16;
456 /* required for 100% accuracy */
459 H = ( H + (H>>2) ) >> 4;
460 V = ( V + (V>>2) ) >> 4;
466 a = 16*(src1[0] + src2[16] + 1) - 7*(V+H);
467 for(j=16; j>0; --j) {
470 for(i=-16; i<0; i+=4) {
471 src[16+i] = CLIP((b ) >> 5);
472 src[17+i] = CLIP((b+ H) >> 5);
473 src[18+i] = CLIP((b+2*H) >> 5);
474 src[19+i] = CLIP((b+3*H) >> 5);
481 static void FUNCC(pred16x16_plane)(uint8_t *src, ptrdiff_t stride)
483 FUNCC(pred16x16_plane_compat)(src, stride, 0, 0);
486 static void FUNCC(pred8x8_vertical)(uint8_t *_src, ptrdiff_t _stride)
489 pixel *src = (pixel*)_src;
490 int stride = _stride>>(sizeof(pixel)-1);
491 const pixel4 a= AV_RN4PA(((pixel4*)(src-stride))+0);
492 const pixel4 b= AV_RN4PA(((pixel4*)(src-stride))+1);
495 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
496 AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
500 static void FUNCC(pred8x16_vertical)(uint8_t *_src, ptrdiff_t _stride)
503 pixel *src = (pixel*)_src;
504 int stride = _stride>>(sizeof(pixel)-1);
505 const pixel4 a= AV_RN4PA(((pixel4*)(src-stride))+0);
506 const pixel4 b= AV_RN4PA(((pixel4*)(src-stride))+1);
509 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
510 AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
514 static void FUNCC(pred8x8_horizontal)(uint8_t *_src, ptrdiff_t stride)
517 pixel *src = (pixel*)_src;
518 stride >>= sizeof(pixel)-1;
521 const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
522 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
523 AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
527 static void FUNCC(pred8x16_horizontal)(uint8_t *_src, ptrdiff_t stride)
530 pixel *src = (pixel*)_src;
531 stride >>= sizeof(pixel)-1;
533 const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
534 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
535 AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
539 #define PRED8x8_X(n, v)\
540 static void FUNCC(pred8x8_##n##_dc)(uint8_t *_src, ptrdiff_t stride)\
543 const pixel4 a = PIXEL_SPLAT_X4(v);\
544 pixel *src = (pixel*)_src;\
545 stride >>= sizeof(pixel)-1;\
547 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);\
548 AV_WN4PA(((pixel4*)(src+i*stride))+1, a);\
552 PRED8x8_X(127, (1<<(BIT_DEPTH-1))-1)
553 PRED8x8_X(128, (1<<(BIT_DEPTH-1))+0)
554 PRED8x8_X(129, (1<<(BIT_DEPTH-1))+1)
556 static void FUNCC(pred8x16_128_dc)(uint8_t *_src, ptrdiff_t stride)
558 FUNCC(pred8x8_128_dc)(_src, stride);
559 FUNCC(pred8x8_128_dc)(_src+8*stride, stride);
562 static void FUNCC(pred8x8_left_dc)(uint8_t *_src, ptrdiff_t stride)
566 pixel4 dc0splat, dc2splat;
567 pixel *src = (pixel*)_src;
568 stride >>= sizeof(pixel)-1;
572 dc0+= src[-1+i*stride];
573 dc2+= src[-1+(i+4)*stride];
575 dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
576 dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
579 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
580 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc0splat);
583 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
584 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc2splat);
588 static void FUNCC(pred8x16_left_dc)(uint8_t *_src, ptrdiff_t stride)
590 FUNCC(pred8x8_left_dc)(_src, stride);
591 FUNCC(pred8x8_left_dc)(_src+8*stride, stride);
594 static void FUNCC(pred8x8_top_dc)(uint8_t *_src, ptrdiff_t stride)
598 pixel4 dc0splat, dc1splat;
599 pixel *src = (pixel*)_src;
600 stride >>= sizeof(pixel)-1;
605 dc1+= src[4+i-stride];
607 dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
608 dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
611 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
612 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
615 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
616 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
620 static void FUNCC(pred8x16_top_dc)(uint8_t *_src, ptrdiff_t stride)
624 pixel4 dc0splat, dc1splat;
625 pixel *src = (pixel*)_src;
626 stride >>= sizeof(pixel)-1;
631 dc1+= src[4+i-stride];
633 dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
634 dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
637 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
638 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
642 static void FUNCC(pred8x8_dc)(uint8_t *_src, ptrdiff_t stride)
646 pixel4 dc0splat, dc1splat, dc2splat, dc3splat;
647 pixel *src = (pixel*)_src;
648 stride >>= sizeof(pixel)-1;
652 dc0+= src[-1+i*stride] + src[i-stride];
653 dc1+= src[4+i-stride];
654 dc2+= src[-1+(i+4)*stride];
656 dc0splat = PIXEL_SPLAT_X4((dc0 + 4)>>3);
657 dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
658 dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
659 dc3splat = PIXEL_SPLAT_X4((dc1 + dc2 + 4)>>3);
662 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
663 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
666 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
667 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc3splat);
671 static void FUNCC(pred8x16_dc)(uint8_t *_src, ptrdiff_t stride)
674 int dc0, dc1, dc2, dc3, dc4;
675 pixel4 dc0splat, dc1splat, dc2splat, dc3splat, dc4splat, dc5splat, dc6splat, dc7splat;
676 pixel *src = (pixel*)_src;
677 stride >>= sizeof(pixel)-1;
679 dc0=dc1=dc2=dc3=dc4=0;
681 dc0+= src[-1+i*stride] + src[i-stride];
682 dc1+= src[4+i-stride];
683 dc2+= src[-1+(i+4)*stride];
684 dc3+= src[-1+(i+8)*stride];
685 dc4+= src[-1+(i+12)*stride];
687 dc0splat = PIXEL_SPLAT_X4((dc0 + 4)>>3);
688 dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
689 dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
690 dc3splat = PIXEL_SPLAT_X4((dc1 + dc2 + 4)>>3);
691 dc4splat = PIXEL_SPLAT_X4((dc3 + 2)>>2);
692 dc5splat = PIXEL_SPLAT_X4((dc1 + dc3 + 4)>>3);
693 dc6splat = PIXEL_SPLAT_X4((dc4 + 2)>>2);
694 dc7splat = PIXEL_SPLAT_X4((dc1 + dc4 + 4)>>3);
697 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
698 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
701 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
702 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc3splat);
705 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc4splat);
706 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc5splat);
708 for(i=12; i<16; i++){
709 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc6splat);
710 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc7splat);
714 //the following 4 function should not be optimized!
715 static void FUNC(pred8x8_mad_cow_dc_l0t)(uint8_t *src, ptrdiff_t stride)
717 FUNCC(pred8x8_top_dc)(src, stride);
718 FUNCC(pred4x4_dc)(src, NULL, stride);
721 static void FUNC(pred8x16_mad_cow_dc_l0t)(uint8_t *src, ptrdiff_t stride)
723 FUNCC(pred8x16_top_dc)(src, stride);
724 FUNCC(pred4x4_dc)(src, NULL, stride);
727 static void FUNC(pred8x8_mad_cow_dc_0lt)(uint8_t *src, ptrdiff_t stride)
729 FUNCC(pred8x8_dc)(src, stride);
730 FUNCC(pred4x4_top_dc)(src, NULL, stride);
733 static void FUNC(pred8x16_mad_cow_dc_0lt)(uint8_t *src, ptrdiff_t stride)
735 FUNCC(pred8x16_dc)(src, stride);
736 FUNCC(pred4x4_top_dc)(src, NULL, stride);
739 static void FUNC(pred8x8_mad_cow_dc_l00)(uint8_t *src, ptrdiff_t stride)
741 FUNCC(pred8x8_left_dc)(src, stride);
742 FUNCC(pred4x4_128_dc)(src + 4*stride , NULL, stride);
743 FUNCC(pred4x4_128_dc)(src + 4*stride + 4*sizeof(pixel), NULL, stride);
746 static void FUNC(pred8x16_mad_cow_dc_l00)(uint8_t *src, ptrdiff_t stride)
748 FUNCC(pred8x16_left_dc)(src, stride);
749 FUNCC(pred4x4_128_dc)(src + 4*stride , NULL, stride);
750 FUNCC(pred4x4_128_dc)(src + 4*stride + 4*sizeof(pixel), NULL, stride);
753 static void FUNC(pred8x8_mad_cow_dc_0l0)(uint8_t *src, ptrdiff_t stride)
755 FUNCC(pred8x8_left_dc)(src, stride);
756 FUNCC(pred4x4_128_dc)(src , NULL, stride);
757 FUNCC(pred4x4_128_dc)(src + 4*sizeof(pixel), NULL, stride);
760 static void FUNC(pred8x16_mad_cow_dc_0l0)(uint8_t *src, ptrdiff_t stride)
762 FUNCC(pred8x16_left_dc)(src, stride);
763 FUNCC(pred4x4_128_dc)(src , NULL, stride);
764 FUNCC(pred4x4_128_dc)(src + 4*sizeof(pixel), NULL, stride);
767 static void FUNCC(pred8x8_plane)(uint8_t *_src, ptrdiff_t _stride)
772 pixel *src = (pixel*)_src;
773 int stride = _stride>>(sizeof(pixel)-1);
774 const pixel * const src0 = src +3-stride;
775 const pixel * src1 = src +4*stride-1;
776 const pixel * src2 = src1-2*stride; // == src+2*stride-1;
777 int H = src0[1] - src0[-1];
778 int V = src1[0] - src2[ 0];
779 for(k=2; k<=4; ++k) {
780 src1 += stride; src2 -= stride;
781 H += k*(src0[k] - src0[-k]);
782 V += k*(src1[0] - src2[ 0]);
784 H = ( 17*H+16 ) >> 5;
785 V = ( 17*V+16 ) >> 5;
787 a = 16*(src1[0] + src2[8]+1) - 3*(V+H);
791 src[0] = CLIP((b ) >> 5);
792 src[1] = CLIP((b+ H) >> 5);
793 src[2] = CLIP((b+2*H) >> 5);
794 src[3] = CLIP((b+3*H) >> 5);
795 src[4] = CLIP((b+4*H) >> 5);
796 src[5] = CLIP((b+5*H) >> 5);
797 src[6] = CLIP((b+6*H) >> 5);
798 src[7] = CLIP((b+7*H) >> 5);
803 static void FUNCC(pred8x16_plane)(uint8_t *_src, ptrdiff_t _stride)
808 pixel *src = (pixel*)_src;
809 int stride = _stride>>(sizeof(pixel)-1);
810 const pixel * const src0 = src +3-stride;
811 const pixel * src1 = src +8*stride-1;
812 const pixel * src2 = src1-2*stride; // == src+6*stride-1;
813 int H = src0[1] - src0[-1];
814 int V = src1[0] - src2[ 0];
816 for (k = 2; k <= 4; ++k) {
817 src1 += stride; src2 -= stride;
818 H += k*(src0[k] - src0[-k]);
819 V += k*(src1[0] - src2[ 0]);
821 for (; k <= 8; ++k) {
822 src1 += stride; src2 -= stride;
823 V += k*(src1[0] - src2[0]);
829 a = 16*(src1[0] + src2[8] + 1) - 7*V - 3*H;
830 for(j=16; j>0; --j) {
833 src[0] = CLIP((b ) >> 5);
834 src[1] = CLIP((b+ H) >> 5);
835 src[2] = CLIP((b+2*H) >> 5);
836 src[3] = CLIP((b+3*H) >> 5);
837 src[4] = CLIP((b+4*H) >> 5);
838 src[5] = CLIP((b+5*H) >> 5);
839 src[6] = CLIP((b+6*H) >> 5);
840 src[7] = CLIP((b+7*H) >> 5);
845 #define SRC(x,y) src[(x)+(y)*stride]
847 const int l##y = (SRC(-1,y-1) + 2*SRC(-1,y) + SRC(-1,y+1) + 2) >> 2;
848 #define PREDICT_8x8_LOAD_LEFT \
849 const int l0 = ((has_topleft ? SRC(-1,-1) : SRC(-1,0)) \
850 + 2*SRC(-1,0) + SRC(-1,1) + 2) >> 2; \
851 PL(1) PL(2) PL(3) PL(4) PL(5) PL(6) \
852 const int l7 av_unused = (SRC(-1,6) + 3*SRC(-1,7) + 2) >> 2
855 const int t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
856 #define PREDICT_8x8_LOAD_TOP \
857 const int t0 = ((has_topleft ? SRC(-1,-1) : SRC(0,-1)) \
858 + 2*SRC(0,-1) + SRC(1,-1) + 2) >> 2; \
859 PT(1) PT(2) PT(3) PT(4) PT(5) PT(6) \
860 const int t7 av_unused = ((has_topright ? SRC(8,-1) : SRC(7,-1)) \
861 + 2*SRC(7,-1) + SRC(6,-1) + 2) >> 2
864 t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
865 #define PREDICT_8x8_LOAD_TOPRIGHT \
866 int t8, t9, t10, t11, t12, t13, t14, t15; \
868 PTR(8) PTR(9) PTR(10) PTR(11) PTR(12) PTR(13) PTR(14) \
869 t15 = (SRC(14,-1) + 3*SRC(15,-1) + 2) >> 2; \
870 } else t8=t9=t10=t11=t12=t13=t14=t15= SRC(7,-1);
872 #define PREDICT_8x8_LOAD_TOPLEFT \
873 const int lt = (SRC(-1,0) + 2*SRC(-1,-1) + SRC(0,-1) + 2) >> 2
875 #define PREDICT_8x8_DC(v) \
877 for( y = 0; y < 8; y++ ) { \
878 AV_WN4PA(((pixel4*)src)+0, v); \
879 AV_WN4PA(((pixel4*)src)+1, v); \
883 static void FUNCC(pred8x8l_128_dc)(uint8_t *_src, int has_topleft,
884 int has_topright, ptrdiff_t _stride)
886 pixel *src = (pixel*)_src;
887 int stride = _stride>>(sizeof(pixel)-1);
889 PREDICT_8x8_DC(PIXEL_SPLAT_X4(1<<(BIT_DEPTH-1)));
891 static void FUNCC(pred8x8l_left_dc)(uint8_t *_src, int has_topleft,
892 int has_topright, ptrdiff_t _stride)
894 pixel *src = (pixel*)_src;
895 int stride = _stride>>(sizeof(pixel)-1);
897 PREDICT_8x8_LOAD_LEFT;
898 const pixel4 dc = PIXEL_SPLAT_X4((l0+l1+l2+l3+l4+l5+l6+l7+4) >> 3);
901 static void FUNCC(pred8x8l_top_dc)(uint8_t *_src, int has_topleft,
902 int has_topright, ptrdiff_t _stride)
904 pixel *src = (pixel*)_src;
905 int stride = _stride>>(sizeof(pixel)-1);
907 PREDICT_8x8_LOAD_TOP;
908 const pixel4 dc = PIXEL_SPLAT_X4((t0+t1+t2+t3+t4+t5+t6+t7+4) >> 3);
911 static void FUNCC(pred8x8l_dc)(uint8_t *_src, int has_topleft,
912 int has_topright, ptrdiff_t _stride)
914 pixel *src = (pixel*)_src;
915 int stride = _stride>>(sizeof(pixel)-1);
917 PREDICT_8x8_LOAD_LEFT;
918 PREDICT_8x8_LOAD_TOP;
919 const pixel4 dc = PIXEL_SPLAT_X4((l0+l1+l2+l3+l4+l5+l6+l7
920 +t0+t1+t2+t3+t4+t5+t6+t7+8) >> 4);
923 static void FUNCC(pred8x8l_horizontal)(uint8_t *_src, int has_topleft,
924 int has_topright, ptrdiff_t _stride)
926 pixel *src = (pixel*)_src;
927 int stride = _stride>>(sizeof(pixel)-1);
930 PREDICT_8x8_LOAD_LEFT;
931 #define ROW(y) a = PIXEL_SPLAT_X4(l##y); \
932 AV_WN4PA(src+y*stride, a); \
933 AV_WN4PA(src+y*stride+4, a);
934 ROW(0); ROW(1); ROW(2); ROW(3); ROW(4); ROW(5); ROW(6); ROW(7);
937 static void FUNCC(pred8x8l_vertical)(uint8_t *_src, int has_topleft,
938 int has_topright, ptrdiff_t _stride)
941 pixel *src = (pixel*)_src;
942 int stride = _stride>>(sizeof(pixel)-1);
945 PREDICT_8x8_LOAD_TOP;
954 a = AV_RN4PA(((pixel4*)src)+0);
955 b = AV_RN4PA(((pixel4*)src)+1);
956 for( y = 1; y < 8; y++ ) {
957 AV_WN4PA(((pixel4*)(src+y*stride))+0, a);
958 AV_WN4PA(((pixel4*)(src+y*stride))+1, b);
961 static void FUNCC(pred8x8l_down_left)(uint8_t *_src, int has_topleft,
962 int has_topright, ptrdiff_t _stride)
964 pixel *src = (pixel*)_src;
965 int stride = _stride>>(sizeof(pixel)-1);
966 PREDICT_8x8_LOAD_TOP;
967 PREDICT_8x8_LOAD_TOPRIGHT;
968 SRC(0,0)= (t0 + 2*t1 + t2 + 2) >> 2;
969 SRC(0,1)=SRC(1,0)= (t1 + 2*t2 + t3 + 2) >> 2;
970 SRC(0,2)=SRC(1,1)=SRC(2,0)= (t2 + 2*t3 + t4 + 2) >> 2;
971 SRC(0,3)=SRC(1,2)=SRC(2,1)=SRC(3,0)= (t3 + 2*t4 + t5 + 2) >> 2;
972 SRC(0,4)=SRC(1,3)=SRC(2,2)=SRC(3,1)=SRC(4,0)= (t4 + 2*t5 + t6 + 2) >> 2;
973 SRC(0,5)=SRC(1,4)=SRC(2,3)=SRC(3,2)=SRC(4,1)=SRC(5,0)= (t5 + 2*t6 + t7 + 2) >> 2;
974 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;
975 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;
976 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;
977 SRC(2,7)=SRC(3,6)=SRC(4,5)=SRC(5,4)=SRC(6,3)=SRC(7,2)= (t9 + 2*t10 + t11 + 2) >> 2;
978 SRC(3,7)=SRC(4,6)=SRC(5,5)=SRC(6,4)=SRC(7,3)= (t10 + 2*t11 + t12 + 2) >> 2;
979 SRC(4,7)=SRC(5,6)=SRC(6,5)=SRC(7,4)= (t11 + 2*t12 + t13 + 2) >> 2;
980 SRC(5,7)=SRC(6,6)=SRC(7,5)= (t12 + 2*t13 + t14 + 2) >> 2;
981 SRC(6,7)=SRC(7,6)= (t13 + 2*t14 + t15 + 2) >> 2;
982 SRC(7,7)= (t14 + 3*t15 + 2) >> 2;
984 static void FUNCC(pred8x8l_down_right)(uint8_t *_src, int has_topleft,
985 int has_topright, ptrdiff_t _stride)
987 pixel *src = (pixel*)_src;
988 int stride = _stride>>(sizeof(pixel)-1);
989 PREDICT_8x8_LOAD_TOP;
990 PREDICT_8x8_LOAD_LEFT;
991 PREDICT_8x8_LOAD_TOPLEFT;
992 SRC(0,7)= (l7 + 2*l6 + l5 + 2) >> 2;
993 SRC(0,6)=SRC(1,7)= (l6 + 2*l5 + l4 + 2) >> 2;
994 SRC(0,5)=SRC(1,6)=SRC(2,7)= (l5 + 2*l4 + l3 + 2) >> 2;
995 SRC(0,4)=SRC(1,5)=SRC(2,6)=SRC(3,7)= (l4 + 2*l3 + l2 + 2) >> 2;
996 SRC(0,3)=SRC(1,4)=SRC(2,5)=SRC(3,6)=SRC(4,7)= (l3 + 2*l2 + l1 + 2) >> 2;
997 SRC(0,2)=SRC(1,3)=SRC(2,4)=SRC(3,5)=SRC(4,6)=SRC(5,7)= (l2 + 2*l1 + l0 + 2) >> 2;
998 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;
999 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;
1000 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;
1001 SRC(2,0)=SRC(3,1)=SRC(4,2)=SRC(5,3)=SRC(6,4)=SRC(7,5)= (t0 + 2*t1 + t2 + 2) >> 2;
1002 SRC(3,0)=SRC(4,1)=SRC(5,2)=SRC(6,3)=SRC(7,4)= (t1 + 2*t2 + t3 + 2) >> 2;
1003 SRC(4,0)=SRC(5,1)=SRC(6,2)=SRC(7,3)= (t2 + 2*t3 + t4 + 2) >> 2;
1004 SRC(5,0)=SRC(6,1)=SRC(7,2)= (t3 + 2*t4 + t5 + 2) >> 2;
1005 SRC(6,0)=SRC(7,1)= (t4 + 2*t5 + t6 + 2) >> 2;
1006 SRC(7,0)= (t5 + 2*t6 + t7 + 2) >> 2;
1008 static void FUNCC(pred8x8l_vertical_right)(uint8_t *_src, int has_topleft,
1009 int has_topright, ptrdiff_t _stride)
1011 pixel *src = (pixel*)_src;
1012 int stride = _stride>>(sizeof(pixel)-1);
1013 PREDICT_8x8_LOAD_TOP;
1014 PREDICT_8x8_LOAD_LEFT;
1015 PREDICT_8x8_LOAD_TOPLEFT;
1016 SRC(0,6)= (l5 + 2*l4 + l3 + 2) >> 2;
1017 SRC(0,7)= (l6 + 2*l5 + l4 + 2) >> 2;
1018 SRC(0,4)=SRC(1,6)= (l3 + 2*l2 + l1 + 2) >> 2;
1019 SRC(0,5)=SRC(1,7)= (l4 + 2*l3 + l2 + 2) >> 2;
1020 SRC(0,2)=SRC(1,4)=SRC(2,6)= (l1 + 2*l0 + lt + 2) >> 2;
1021 SRC(0,3)=SRC(1,5)=SRC(2,7)= (l2 + 2*l1 + l0 + 2) >> 2;
1022 SRC(0,1)=SRC(1,3)=SRC(2,5)=SRC(3,7)= (l0 + 2*lt + t0 + 2) >> 2;
1023 SRC(0,0)=SRC(1,2)=SRC(2,4)=SRC(3,6)= (lt + t0 + 1) >> 1;
1024 SRC(1,1)=SRC(2,3)=SRC(3,5)=SRC(4,7)= (lt + 2*t0 + t1 + 2) >> 2;
1025 SRC(1,0)=SRC(2,2)=SRC(3,4)=SRC(4,6)= (t0 + t1 + 1) >> 1;
1026 SRC(2,1)=SRC(3,3)=SRC(4,5)=SRC(5,7)= (t0 + 2*t1 + t2 + 2) >> 2;
1027 SRC(2,0)=SRC(3,2)=SRC(4,4)=SRC(5,6)= (t1 + t2 + 1) >> 1;
1028 SRC(3,1)=SRC(4,3)=SRC(5,5)=SRC(6,7)= (t1 + 2*t2 + t3 + 2) >> 2;
1029 SRC(3,0)=SRC(4,2)=SRC(5,4)=SRC(6,6)= (t2 + t3 + 1) >> 1;
1030 SRC(4,1)=SRC(5,3)=SRC(6,5)=SRC(7,7)= (t2 + 2*t3 + t4 + 2) >> 2;
1031 SRC(4,0)=SRC(5,2)=SRC(6,4)=SRC(7,6)= (t3 + t4 + 1) >> 1;
1032 SRC(5,1)=SRC(6,3)=SRC(7,5)= (t3 + 2*t4 + t5 + 2) >> 2;
1033 SRC(5,0)=SRC(6,2)=SRC(7,4)= (t4 + t5 + 1) >> 1;
1034 SRC(6,1)=SRC(7,3)= (t4 + 2*t5 + t6 + 2) >> 2;
1035 SRC(6,0)=SRC(7,2)= (t5 + t6 + 1) >> 1;
1036 SRC(7,1)= (t5 + 2*t6 + t7 + 2) >> 2;
1037 SRC(7,0)= (t6 + t7 + 1) >> 1;
1039 static void FUNCC(pred8x8l_horizontal_down)(uint8_t *_src, int has_topleft,
1040 int has_topright, ptrdiff_t _stride)
1042 pixel *src = (pixel*)_src;
1043 int stride = _stride>>(sizeof(pixel)-1);
1044 PREDICT_8x8_LOAD_TOP;
1045 PREDICT_8x8_LOAD_LEFT;
1046 PREDICT_8x8_LOAD_TOPLEFT;
1047 SRC(0,7)= (l6 + l7 + 1) >> 1;
1048 SRC(1,7)= (l5 + 2*l6 + l7 + 2) >> 2;
1049 SRC(0,6)=SRC(2,7)= (l5 + l6 + 1) >> 1;
1050 SRC(1,6)=SRC(3,7)= (l4 + 2*l5 + l6 + 2) >> 2;
1051 SRC(0,5)=SRC(2,6)=SRC(4,7)= (l4 + l5 + 1) >> 1;
1052 SRC(1,5)=SRC(3,6)=SRC(5,7)= (l3 + 2*l4 + l5 + 2) >> 2;
1053 SRC(0,4)=SRC(2,5)=SRC(4,6)=SRC(6,7)= (l3 + l4 + 1) >> 1;
1054 SRC(1,4)=SRC(3,5)=SRC(5,6)=SRC(7,7)= (l2 + 2*l3 + l4 + 2) >> 2;
1055 SRC(0,3)=SRC(2,4)=SRC(4,5)=SRC(6,6)= (l2 + l3 + 1) >> 1;
1056 SRC(1,3)=SRC(3,4)=SRC(5,5)=SRC(7,6)= (l1 + 2*l2 + l3 + 2) >> 2;
1057 SRC(0,2)=SRC(2,3)=SRC(4,4)=SRC(6,5)= (l1 + l2 + 1) >> 1;
1058 SRC(1,2)=SRC(3,3)=SRC(5,4)=SRC(7,5)= (l0 + 2*l1 + l2 + 2) >> 2;
1059 SRC(0,1)=SRC(2,2)=SRC(4,3)=SRC(6,4)= (l0 + l1 + 1) >> 1;
1060 SRC(1,1)=SRC(3,2)=SRC(5,3)=SRC(7,4)= (lt + 2*l0 + l1 + 2) >> 2;
1061 SRC(0,0)=SRC(2,1)=SRC(4,2)=SRC(6,3)= (lt + l0 + 1) >> 1;
1062 SRC(1,0)=SRC(3,1)=SRC(5,2)=SRC(7,3)= (l0 + 2*lt + t0 + 2) >> 2;
1063 SRC(2,0)=SRC(4,1)=SRC(6,2)= (t1 + 2*t0 + lt + 2) >> 2;
1064 SRC(3,0)=SRC(5,1)=SRC(7,2)= (t2 + 2*t1 + t0 + 2) >> 2;
1065 SRC(4,0)=SRC(6,1)= (t3 + 2*t2 + t1 + 2) >> 2;
1066 SRC(5,0)=SRC(7,1)= (t4 + 2*t3 + t2 + 2) >> 2;
1067 SRC(6,0)= (t5 + 2*t4 + t3 + 2) >> 2;
1068 SRC(7,0)= (t6 + 2*t5 + t4 + 2) >> 2;
1070 static void FUNCC(pred8x8l_vertical_left)(uint8_t *_src, int has_topleft,
1071 int has_topright, ptrdiff_t _stride)
1073 pixel *src = (pixel*)_src;
1074 int stride = _stride>>(sizeof(pixel)-1);
1075 PREDICT_8x8_LOAD_TOP;
1076 PREDICT_8x8_LOAD_TOPRIGHT;
1077 SRC(0,0)= (t0 + t1 + 1) >> 1;
1078 SRC(0,1)= (t0 + 2*t1 + t2 + 2) >> 2;
1079 SRC(0,2)=SRC(1,0)= (t1 + t2 + 1) >> 1;
1080 SRC(0,3)=SRC(1,1)= (t1 + 2*t2 + t3 + 2) >> 2;
1081 SRC(0,4)=SRC(1,2)=SRC(2,0)= (t2 + t3 + 1) >> 1;
1082 SRC(0,5)=SRC(1,3)=SRC(2,1)= (t2 + 2*t3 + t4 + 2) >> 2;
1083 SRC(0,6)=SRC(1,4)=SRC(2,2)=SRC(3,0)= (t3 + t4 + 1) >> 1;
1084 SRC(0,7)=SRC(1,5)=SRC(2,3)=SRC(3,1)= (t3 + 2*t4 + t5 + 2) >> 2;
1085 SRC(1,6)=SRC(2,4)=SRC(3,2)=SRC(4,0)= (t4 + t5 + 1) >> 1;
1086 SRC(1,7)=SRC(2,5)=SRC(3,3)=SRC(4,1)= (t4 + 2*t5 + t6 + 2) >> 2;
1087 SRC(2,6)=SRC(3,4)=SRC(4,2)=SRC(5,0)= (t5 + t6 + 1) >> 1;
1088 SRC(2,7)=SRC(3,5)=SRC(4,3)=SRC(5,1)= (t5 + 2*t6 + t7 + 2) >> 2;
1089 SRC(3,6)=SRC(4,4)=SRC(5,2)=SRC(6,0)= (t6 + t7 + 1) >> 1;
1090 SRC(3,7)=SRC(4,5)=SRC(5,3)=SRC(6,1)= (t6 + 2*t7 + t8 + 2) >> 2;
1091 SRC(4,6)=SRC(5,4)=SRC(6,2)=SRC(7,0)= (t7 + t8 + 1) >> 1;
1092 SRC(4,7)=SRC(5,5)=SRC(6,3)=SRC(7,1)= (t7 + 2*t8 + t9 + 2) >> 2;
1093 SRC(5,6)=SRC(6,4)=SRC(7,2)= (t8 + t9 + 1) >> 1;
1094 SRC(5,7)=SRC(6,5)=SRC(7,3)= (t8 + 2*t9 + t10 + 2) >> 2;
1095 SRC(6,6)=SRC(7,4)= (t9 + t10 + 1) >> 1;
1096 SRC(6,7)=SRC(7,5)= (t9 + 2*t10 + t11 + 2) >> 2;
1097 SRC(7,6)= (t10 + t11 + 1) >> 1;
1098 SRC(7,7)= (t10 + 2*t11 + t12 + 2) >> 2;
1100 static void FUNCC(pred8x8l_horizontal_up)(uint8_t *_src, int has_topleft,
1101 int has_topright, ptrdiff_t _stride)
1103 pixel *src = (pixel*)_src;
1104 int stride = _stride>>(sizeof(pixel)-1);
1105 PREDICT_8x8_LOAD_LEFT;
1106 SRC(0,0)= (l0 + l1 + 1) >> 1;
1107 SRC(1,0)= (l0 + 2*l1 + l2 + 2) >> 2;
1108 SRC(0,1)=SRC(2,0)= (l1 + l2 + 1) >> 1;
1109 SRC(1,1)=SRC(3,0)= (l1 + 2*l2 + l3 + 2) >> 2;
1110 SRC(0,2)=SRC(2,1)=SRC(4,0)= (l2 + l3 + 1) >> 1;
1111 SRC(1,2)=SRC(3,1)=SRC(5,0)= (l2 + 2*l3 + l4 + 2) >> 2;
1112 SRC(0,3)=SRC(2,2)=SRC(4,1)=SRC(6,0)= (l3 + l4 + 1) >> 1;
1113 SRC(1,3)=SRC(3,2)=SRC(5,1)=SRC(7,0)= (l3 + 2*l4 + l5 + 2) >> 2;
1114 SRC(0,4)=SRC(2,3)=SRC(4,2)=SRC(6,1)= (l4 + l5 + 1) >> 1;
1115 SRC(1,4)=SRC(3,3)=SRC(5,2)=SRC(7,1)= (l4 + 2*l5 + l6 + 2) >> 2;
1116 SRC(0,5)=SRC(2,4)=SRC(4,3)=SRC(6,2)= (l5 + l6 + 1) >> 1;
1117 SRC(1,5)=SRC(3,4)=SRC(5,3)=SRC(7,2)= (l5 + 2*l6 + l7 + 2) >> 2;
1118 SRC(0,6)=SRC(2,5)=SRC(4,4)=SRC(6,3)= (l6 + l7 + 1) >> 1;
1119 SRC(1,6)=SRC(3,5)=SRC(5,4)=SRC(7,3)= (l6 + 3*l7 + 2) >> 2;
1120 SRC(0,7)=SRC(1,7)=SRC(2,6)=SRC(2,7)=SRC(3,6)=
1121 SRC(3,7)=SRC(4,5)=SRC(4,6)=SRC(4,7)=SRC(5,5)=
1122 SRC(5,6)=SRC(5,7)=SRC(6,4)=SRC(6,5)=SRC(6,6)=
1123 SRC(6,7)=SRC(7,4)=SRC(7,5)=SRC(7,6)=SRC(7,7)= l7;
1125 #undef PREDICT_8x8_LOAD_LEFT
1126 #undef PREDICT_8x8_LOAD_TOP
1127 #undef PREDICT_8x8_LOAD_TOPLEFT
1128 #undef PREDICT_8x8_LOAD_TOPRIGHT
1129 #undef PREDICT_8x8_DC
1135 static void FUNCC(pred4x4_vertical_add)(uint8_t *_pix, int16_t *_block,
1139 pixel *pix = (pixel*)_pix;
1140 const dctcoef *block = (const dctcoef*)_block;
1141 stride >>= sizeof(pixel)-1;
1145 pix[1*stride]= v += block[0];
1146 pix[2*stride]= v += block[4];
1147 pix[3*stride]= v += block[8];
1148 pix[4*stride]= v + block[12];
1153 memset(_block, 0, sizeof(dctcoef) * 16);
1156 static void FUNCC(pred4x4_horizontal_add)(uint8_t *_pix, int16_t *_block,
1160 pixel *pix = (pixel*)_pix;
1161 const dctcoef *block = (const dctcoef*)_block;
1162 stride >>= sizeof(pixel)-1;
1165 pix[0]= v += block[0];
1166 pix[1]= v += block[1];
1167 pix[2]= v += block[2];
1168 pix[3]= v + block[3];
1173 memset(_block, 0, sizeof(dctcoef) * 16);
1176 static void FUNCC(pred8x8l_vertical_add)(uint8_t *_pix, int16_t *_block,
1180 pixel *pix = (pixel*)_pix;
1181 const dctcoef *block = (const dctcoef*)_block;
1182 stride >>= sizeof(pixel)-1;
1186 pix[1*stride]= v += block[0];
1187 pix[2*stride]= v += block[8];
1188 pix[3*stride]= v += block[16];
1189 pix[4*stride]= v += block[24];
1190 pix[5*stride]= v += block[32];
1191 pix[6*stride]= v += block[40];
1192 pix[7*stride]= v += block[48];
1193 pix[8*stride]= v + block[56];
1198 memset(_block, 0, sizeof(dctcoef) * 64);
1201 static void FUNCC(pred8x8l_horizontal_add)(uint8_t *_pix, int16_t *_block,
1205 pixel *pix = (pixel*)_pix;
1206 const dctcoef *block = (const dctcoef*)_block;
1207 stride >>= sizeof(pixel)-1;
1210 pix[0]= v += block[0];
1211 pix[1]= v += block[1];
1212 pix[2]= v += block[2];
1213 pix[3]= v += block[3];
1214 pix[4]= v += block[4];
1215 pix[5]= v += block[5];
1216 pix[6]= v += block[6];
1217 pix[7]= v + block[7];
1222 memset(_block, 0, sizeof(dctcoef) * 64);
1225 static void FUNCC(pred16x16_vertical_add)(uint8_t *pix, const int *block_offset,
1231 FUNCC(pred4x4_vertical_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1234 static void FUNCC(pred16x16_horizontal_add)(uint8_t *pix,
1235 const int *block_offset,
1241 FUNCC(pred4x4_horizontal_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1244 static void FUNCC(pred8x8_vertical_add)(uint8_t *pix, const int *block_offset,
1245 int16_t *block, ptrdiff_t stride)
1249 FUNCC(pred4x4_vertical_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1252 static void FUNCC(pred8x16_vertical_add)(uint8_t *pix, const int *block_offset,
1253 int16_t *block, ptrdiff_t stride)
1257 FUNCC(pred4x4_vertical_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1259 FUNCC(pred4x4_vertical_add)(pix + block_offset[i+4], block + i*16*sizeof(pixel), stride);
1262 static void FUNCC(pred8x8_horizontal_add)(uint8_t *pix, const int *block_offset,
1268 FUNCC(pred4x4_horizontal_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1271 static void FUNCC(pred8x16_horizontal_add)(uint8_t *pix,
1272 const int *block_offset,
1273 int16_t *block, ptrdiff_t stride)
1277 FUNCC(pred4x4_horizontal_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1279 FUNCC(pred4x4_horizontal_add)(pix + block_offset[i+4], block + i*16*sizeof(pixel), stride);