h264: 4:2:2 intra decoding support
[libav.git] / libavcodec / h264pred_template.c
1 /*
2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3 * Copyright (c) 2003-2011 Michael Niedermayer <michaelni@gmx.at>
4 *
5 * This file is part of Libav.
6 *
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.
11 *
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.
16 *
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
20 */
21
22 /**
23 * @file
24 * H.264 / AVC / MPEG4 part10 prediction functions.
25 * @author Michael Niedermayer <michaelni@gmx.at>
26 */
27
28 #include "mathops.h"
29
30 #include "bit_depth_template.c"
31
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);
36
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);
41 }
42
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]));
50 }
51
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);
58
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);
63 }
64
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);
70
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);
75 }
76
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);
82
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);
87 }
88
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));
93
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);
98 }
99
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);
104
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);
109 }
110
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);
115
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);
120 }
121
122
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];\
128
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];\
134
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];\
140
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];\
146
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];
151 LOAD_TOP_EDGE
152 LOAD_LEFT_EDGE
153
154 src[0+3*stride]=(l3 + 2*l2 + l1 + 2)>>2;
155 src[0+2*stride]=
156 src[1+3*stride]=(l2 + 2*l1 + l0 + 2)>>2;
157 src[0+1*stride]=
158 src[1+2*stride]=
159 src[2+3*stride]=(l1 + 2*l0 + lt + 2)>>2;
160 src[0+0*stride]=
161 src[1+1*stride]=
162 src[2+2*stride]=
163 src[3+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
164 src[1+0*stride]=
165 src[2+1*stride]=
166 src[3+2*stride]=(lt + 2*t0 + t1 + 2)>>2;
167 src[2+0*stride]=
168 src[3+1*stride]=(t0 + 2*t1 + t2 + 2)>>2;
169 src[3+0*stride]=(t1 + 2*t2 + t3 + 2)>>2;
170 }
171
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);
176 LOAD_TOP_EDGE
177 LOAD_TOP_RIGHT_EDGE
178 // LOAD_LEFT_EDGE
179
180 src[0+0*stride]=(t0 + t2 + 2*t1 + 2)>>2;
181 src[1+0*stride]=
182 src[0+1*stride]=(t1 + t3 + 2*t2 + 2)>>2;
183 src[2+0*stride]=
184 src[1+1*stride]=
185 src[0+2*stride]=(t2 + t4 + 2*t3 + 2)>>2;
186 src[3+0*stride]=
187 src[2+1*stride]=
188 src[1+2*stride]=
189 src[0+3*stride]=(t3 + t5 + 2*t4 + 2)>>2;
190 src[3+1*stride]=
191 src[2+2*stride]=
192 src[1+3*stride]=(t4 + t6 + 2*t5 + 2)>>2;
193 src[3+2*stride]=
194 src[2+3*stride]=(t5 + t7 + 2*t6 + 2)>>2;
195 src[3+3*stride]=(t6 + 3*t7 + 2)>>2;
196 }
197
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];
202 LOAD_TOP_EDGE
203 LOAD_LEFT_EDGE
204
205 src[0+0*stride]=
206 src[1+2*stride]=(lt + t0 + 1)>>1;
207 src[1+0*stride]=
208 src[2+2*stride]=(t0 + t1 + 1)>>1;
209 src[2+0*stride]=
210 src[3+2*stride]=(t1 + t2 + 1)>>1;
211 src[3+0*stride]=(t2 + t3 + 1)>>1;
212 src[0+1*stride]=
213 src[1+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
214 src[1+1*stride]=
215 src[2+3*stride]=(lt + 2*t0 + t1 + 2)>>2;
216 src[2+1*stride]=
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;
221 }
222
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);
227 LOAD_TOP_EDGE
228 LOAD_TOP_RIGHT_EDGE
229
230 src[0+0*stride]=(t0 + t1 + 1)>>1;
231 src[1+0*stride]=
232 src[0+2*stride]=(t1 + t2 + 1)>>1;
233 src[2+0*stride]=
234 src[1+2*stride]=(t2 + t3 + 1)>>1;
235 src[3+0*stride]=
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;
239 src[1+1*stride]=
240 src[0+3*stride]=(t1 + 2*t2 + t3 + 2)>>2;
241 src[2+1*stride]=
242 src[1+3*stride]=(t2 + 2*t3 + t4 + 2)>>2;
243 src[3+1*stride]=
244 src[2+3*stride]=(t3 + 2*t4 + t5 + 2)>>2;
245 src[3+3*stride]=(t4 + 2*t5 + t6 + 2)>>2;
246 }
247
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);
251 LOAD_LEFT_EDGE
252
253 src[0+0*stride]=(l0 + l1 + 1)>>1;
254 src[1+0*stride]=(l0 + 2*l1 + l2 + 2)>>2;
255 src[2+0*stride]=
256 src[0+1*stride]=(l1 + l2 + 1)>>1;
257 src[3+0*stride]=
258 src[1+1*stride]=(l1 + 2*l2 + l3 + 2)>>2;
259 src[2+1*stride]=
260 src[0+2*stride]=(l2 + l3 + 1)>>1;
261 src[3+1*stride]=
262 src[1+2*stride]=(l2 + 2*l3 + l3 + 2)>>2;
263 src[3+2*stride]=
264 src[1+3*stride]=
265 src[0+3*stride]=
266 src[2+2*stride]=
267 src[2+3*stride]=
268 src[3+3*stride]=l3;
269 }
270
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];
275 LOAD_TOP_EDGE
276 LOAD_LEFT_EDGE
277
278 src[0+0*stride]=
279 src[2+1*stride]=(lt + l0 + 1)>>1;
280 src[1+0*stride]=
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;
284 src[0+1*stride]=
285 src[2+2*stride]=(l0 + l1 + 1)>>1;
286 src[1+1*stride]=
287 src[3+2*stride]=(lt + 2*l0 + l1 + 2)>>2;
288 src[0+2*stride]=
289 src[2+3*stride]=(l1 + l2+ 1)>>1;
290 src[1+2*stride]=
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;
294 }
295
296 static void FUNCC(pred16x16_vertical)(uint8_t *_src, int _stride){
297 int i;
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);
304
305 for(i=0; i<16; i++){
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);
310 }
311 }
312
313 static void FUNCC(pred16x16_horizontal)(uint8_t *_src, int stride){
314 int i;
315 pixel *src = (pixel*)_src;
316 stride /= sizeof(pixel);
317
318 for(i=0; i<16; i++){
319 const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
320
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);
325 }
326 }
327
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);\
334 src += stride;\
335 }
336
337 static void FUNCC(pred16x16_dc)(uint8_t *_src, int stride){
338 int i, dc=0;
339 pixel *src = (pixel*)_src;
340 pixel4 dcsplat;
341 stride /= sizeof(pixel);
342
343 for(i=0;i<16; i++){
344 dc+= src[-1+i*stride];
345 }
346
347 for(i=0;i<16; i++){
348 dc+= src[i-stride];
349 }
350
351 dcsplat = PIXEL_SPLAT_X4((dc+16)>>5);
352 PREDICT_16x16_DC(dcsplat);
353 }
354
355 static void FUNCC(pred16x16_left_dc)(uint8_t *_src, int stride){
356 int i, dc=0;
357 pixel *src = (pixel*)_src;
358 pixel4 dcsplat;
359 stride /= sizeof(pixel);
360
361 for(i=0;i<16; i++){
362 dc+= src[-1+i*stride];
363 }
364
365 dcsplat = PIXEL_SPLAT_X4((dc+8)>>4);
366 PREDICT_16x16_DC(dcsplat);
367 }
368
369 static void FUNCC(pred16x16_top_dc)(uint8_t *_src, int stride){
370 int i, dc=0;
371 pixel *src = (pixel*)_src;
372 pixel4 dcsplat;
373 stride /= sizeof(pixel);
374
375 for(i=0;i<16; i++){
376 dc+= src[i-stride];
377 }
378
379 dcsplat = PIXEL_SPLAT_X4((dc+8)>>4);
380 PREDICT_16x16_DC(dcsplat);
381 }
382
383 #define PRED16x16_X(n, v) \
384 static void FUNCC(pred16x16_##n##_dc)(uint8_t *_src, int stride){\
385 int i;\
386 pixel *src = (pixel*)_src;\
387 stride /= sizeof(pixel);\
388 PREDICT_16x16_DC(PIXEL_SPLAT_X4(v));\
389 }
390
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);
394
395 static inline void FUNCC(pred16x16_plane_compat)(uint8_t *_src, int _stride, const int svq3, const int rv40){
396 int i, j, k;
397 int a;
398 INIT_CLIP
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]);
410 }
411 if(svq3){
412 H = ( 5*(H/4) ) / 16;
413 V = ( 5*(V/4) ) / 16;
414
415 /* required for 100% accuracy */
416 i = H; H = V; V = i;
417 }else if(rv40){
418 H = ( H + (H>>2) ) >> 4;
419 V = ( V + (V>>2) ) >> 4;
420 }else{
421 H = ( 5*H+32 ) >> 6;
422 V = ( 5*V+32 ) >> 6;
423 }
424
425 a = 16*(src1[0] + src2[16] + 1) - 7*(V+H);
426 for(j=16; j>0; --j) {
427 int b = a;
428 a += V;
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);
434 b += 4*H;
435 }
436 src += stride;
437 }
438 }
439
440 static void FUNCC(pred16x16_plane)(uint8_t *src, int stride){
441 FUNCC(pred16x16_plane_compat)(src, stride, 0, 0);
442 }
443
444 static void FUNCC(pred8x8_vertical)(uint8_t *_src, int _stride){
445 int i;
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);
450
451 for(i=0; i<8; i++){
452 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
453 AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
454 }
455 }
456
457 static void FUNCC(pred8x16_vertical)(uint8_t *_src, int _stride){
458 int i;
459 pixel *src = (pixel*)_src;
460 int stride = _stride>>(sizeof(pixel)-1);
461 const pixel4 a= AV_RN4PA(((pixel4*)(src-stride))+0);
462 const pixel4 b= AV_RN4PA(((pixel4*)(src-stride))+1);
463
464 for(i=0; i<16; i++){
465 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
466 AV_WN4PA(((pixel4*)(src+i*stride))+1, b);
467 }
468 }
469
470 static void FUNCC(pred8x8_horizontal)(uint8_t *_src, int stride){
471 int i;
472 pixel *src = (pixel*)_src;
473 stride /= sizeof(pixel);
474
475 for(i=0; i<8; i++){
476 const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
477 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
478 AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
479 }
480 }
481
482 static void FUNCC(pred8x16_horizontal)(uint8_t *_src, int stride){
483 int i;
484 pixel *src = (pixel*)_src;
485 stride >>= sizeof(pixel)-1;
486 for(i=0; i<16; i++){
487 const pixel4 a = PIXEL_SPLAT_X4(src[-1+i*stride]);
488 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);
489 AV_WN4PA(((pixel4*)(src+i*stride))+1, a);
490 }
491 }
492
493 #define PRED8x8_X(n, v)\
494 static void FUNCC(pred8x8_##n##_dc)(uint8_t *_src, int stride){\
495 int i;\
496 const pixel4 a = PIXEL_SPLAT_X4(v);\
497 pixel *src = (pixel*)_src;\
498 stride /= sizeof(pixel);\
499 for(i=0; i<8; i++){\
500 AV_WN4PA(((pixel4*)(src+i*stride))+0, a);\
501 AV_WN4PA(((pixel4*)(src+i*stride))+1, a);\
502 }\
503 }
504
505 PRED8x8_X(127, (1<<(BIT_DEPTH-1))-1);
506 PRED8x8_X(128, (1<<(BIT_DEPTH-1))+0);
507 PRED8x8_X(129, (1<<(BIT_DEPTH-1))+1);
508
509 static void FUNCC(pred8x16_128_dc)(uint8_t *_src, int stride){
510 FUNCC(pred8x8_128_dc)(_src, stride);
511 FUNCC(pred8x8_128_dc)(_src+8*stride, stride);
512 }
513
514 static void FUNCC(pred8x8_left_dc)(uint8_t *_src, int stride){
515 int i;
516 int dc0, dc2;
517 pixel4 dc0splat, dc2splat;
518 pixel *src = (pixel*)_src;
519 stride /= sizeof(pixel);
520
521 dc0=dc2=0;
522 for(i=0;i<4; i++){
523 dc0+= src[-1+i*stride];
524 dc2+= src[-1+(i+4)*stride];
525 }
526 dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
527 dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
528
529 for(i=0; i<4; i++){
530 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
531 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc0splat);
532 }
533 for(i=4; i<8; i++){
534 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
535 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc2splat);
536 }
537 }
538
539 static void FUNCC(pred8x16_left_dc)(uint8_t *_src, int stride){
540 FUNCC(pred8x8_left_dc)(_src, stride);
541 FUNCC(pred8x8_left_dc)(_src+8*stride, stride);
542 }
543
544 static void FUNCC(pred8x8_top_dc)(uint8_t *_src, int stride){
545 int i;
546 int dc0, dc1;
547 pixel4 dc0splat, dc1splat;
548 pixel *src = (pixel*)_src;
549 stride /= sizeof(pixel);
550
551 dc0=dc1=0;
552 for(i=0;i<4; i++){
553 dc0+= src[i-stride];
554 dc1+= src[4+i-stride];
555 }
556 dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
557 dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
558
559 for(i=0; i<4; i++){
560 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
561 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
562 }
563 for(i=4; i<8; i++){
564 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
565 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
566 }
567 }
568
569 static void FUNCC(pred8x16_top_dc)(uint8_t *_src, int stride){
570 int i;
571 int dc0, dc1;
572 pixel4 dc0splat, dc1splat;
573 pixel *src = (pixel*)_src;
574 stride >>= sizeof(pixel)-1;
575
576 dc0=dc1=0;
577 for(i=0;i<4; i++){
578 dc0+= src[i-stride];
579 dc1+= src[4+i-stride];
580 }
581 dc0splat = PIXEL_SPLAT_X4((dc0 + 2)>>2);
582 dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
583
584 for(i=0; i<16; i++){
585 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
586 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
587 }
588 }
589
590 static void FUNCC(pred8x8_dc)(uint8_t *_src, int stride){
591 int i;
592 int dc0, dc1, dc2;
593 pixel4 dc0splat, dc1splat, dc2splat, dc3splat;
594 pixel *src = (pixel*)_src;
595 stride /= sizeof(pixel);
596
597 dc0=dc1=dc2=0;
598 for(i=0;i<4; i++){
599 dc0+= src[-1+i*stride] + src[i-stride];
600 dc1+= src[4+i-stride];
601 dc2+= src[-1+(i+4)*stride];
602 }
603 dc0splat = PIXEL_SPLAT_X4((dc0 + 4)>>3);
604 dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
605 dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
606 dc3splat = PIXEL_SPLAT_X4((dc1 + dc2 + 4)>>3);
607
608 for(i=0; i<4; i++){
609 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
610 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
611 }
612 for(i=4; i<8; i++){
613 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
614 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc3splat);
615 }
616 }
617
618 static void FUNCC(pred8x16_dc)(uint8_t *_src, int stride){
619 int i;
620 int dc0, dc1, dc2, dc3, dc4;
621 pixel4 dc0splat, dc1splat, dc2splat, dc3splat, dc4splat, dc5splat, dc6splat, dc7splat;
622 pixel *src = (pixel*)_src;
623 stride >>= sizeof(pixel)-1;
624
625 dc0=dc1=dc2=dc3=dc4=0;
626 for(i=0;i<4; i++){
627 dc0+= src[-1+i*stride] + src[i-stride];
628 dc1+= src[4+i-stride];
629 dc2+= src[-1+(i+4)*stride];
630 dc3+= src[-1+(i+8)*stride];
631 dc4+= src[-1+(i+12)*stride];
632 }
633 dc0splat = PIXEL_SPLAT_X4((dc0 + 4)>>3);
634 dc1splat = PIXEL_SPLAT_X4((dc1 + 2)>>2);
635 dc2splat = PIXEL_SPLAT_X4((dc2 + 2)>>2);
636 dc3splat = PIXEL_SPLAT_X4((dc1 + dc2 + 4)>>3);
637 dc4splat = PIXEL_SPLAT_X4((dc3 + 2)>>2);
638 dc5splat = PIXEL_SPLAT_X4((dc1 + dc3 + 4)>>3);
639 dc6splat = PIXEL_SPLAT_X4((dc4 + 2)>>2);
640 dc7splat = PIXEL_SPLAT_X4((dc1 + dc4 + 4)>>3);
641
642 for(i=0; i<4; i++){
643 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc0splat);
644 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc1splat);
645 }
646 for(i=4; i<8; i++){
647 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc2splat);
648 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc3splat);
649 }
650 for(i=8; i<12; i++){
651 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc4splat);
652 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc5splat);
653 }
654 for(i=12; i<16; i++){
655 AV_WN4PA(((pixel4*)(src+i*stride))+0, dc6splat);
656 AV_WN4PA(((pixel4*)(src+i*stride))+1, dc7splat);
657 }
658 }
659
660 //the following 4 function should not be optimized!
661 static void FUNC(pred8x8_mad_cow_dc_l0t)(uint8_t *src, int stride){
662 FUNCC(pred8x8_top_dc)(src, stride);
663 FUNCC(pred4x4_dc)(src, NULL, stride);
664 }
665
666 static void FUNC(pred8x8_mad_cow_dc_0lt)(uint8_t *src, int stride){
667 FUNCC(pred8x8_dc)(src, stride);
668 FUNCC(pred4x4_top_dc)(src, NULL, stride);
669 }
670
671 static void FUNC(pred8x8_mad_cow_dc_l00)(uint8_t *src, int stride){
672 FUNCC(pred8x8_left_dc)(src, stride);
673 FUNCC(pred4x4_128_dc)(src + 4*stride , NULL, stride);
674 FUNCC(pred4x4_128_dc)(src + 4*stride + 4*sizeof(pixel), NULL, stride);
675 }
676
677 static void FUNC(pred8x8_mad_cow_dc_0l0)(uint8_t *src, int stride){
678 FUNCC(pred8x8_left_dc)(src, stride);
679 FUNCC(pred4x4_128_dc)(src , NULL, stride);
680 FUNCC(pred4x4_128_dc)(src + 4*sizeof(pixel), NULL, stride);
681 }
682
683 static void FUNCC(pred8x8_plane)(uint8_t *_src, int _stride){
684 int j, k;
685 int a;
686 INIT_CLIP
687 pixel *src = (pixel*)_src;
688 int stride = _stride/sizeof(pixel);
689 const pixel * const src0 = src +3-stride;
690 const pixel * src1 = src +4*stride-1;
691 const pixel * src2 = src1-2*stride; // == src+2*stride-1;
692 int H = src0[1] - src0[-1];
693 int V = src1[0] - src2[ 0];
694 for(k=2; k<=4; ++k) {
695 src1 += stride; src2 -= stride;
696 H += k*(src0[k] - src0[-k]);
697 V += k*(src1[0] - src2[ 0]);
698 }
699 H = ( 17*H+16 ) >> 5;
700 V = ( 17*V+16 ) >> 5;
701
702 a = 16*(src1[0] + src2[8]+1) - 3*(V+H);
703 for(j=8; j>0; --j) {
704 int b = a;
705 a += V;
706 src[0] = CLIP((b ) >> 5);
707 src[1] = CLIP((b+ H) >> 5);
708 src[2] = CLIP((b+2*H) >> 5);
709 src[3] = CLIP((b+3*H) >> 5);
710 src[4] = CLIP((b+4*H) >> 5);
711 src[5] = CLIP((b+5*H) >> 5);
712 src[6] = CLIP((b+6*H) >> 5);
713 src[7] = CLIP((b+7*H) >> 5);
714 src += stride;
715 }
716 }
717
718 static void FUNCC(pred8x16_plane)(uint8_t *_src, int _stride){
719 int j, k;
720 int a;
721 INIT_CLIP
722 pixel *src = (pixel*)_src;
723 int stride = _stride>>(sizeof(pixel)-1);
724 const pixel * const src0 = src +3-stride;
725 const pixel * src1 = src +8*stride-1;
726 const pixel * src2 = src1-2*stride; // == src+6*stride-1;
727 int H = src0[1] - src0[-1];
728 int V = src1[0] - src2[ 0];
729
730 for (k = 2; k <= 4; ++k) {
731 src1 += stride; src2 -= stride;
732 H += k*(src0[k] - src0[-k]);
733 V += k*(src1[0] - src2[ 0]);
734 }
735 for (; k <= 8; ++k) {
736 src1 += stride; src2 -= stride;
737 V += k*(src1[0] - src2[0]);
738 }
739
740 H = (17*H+16) >> 5;
741 V = (5*V+32) >> 6;
742
743 a = 16*(src1[0] + src2[8] + 1) - 7*V - 3*H;
744 for(j=16; j>0; --j) {
745 int b = a;
746 a += V;
747 src[0] = CLIP((b ) >> 5);
748 src[1] = CLIP((b+ H) >> 5);
749 src[2] = CLIP((b+2*H) >> 5);
750 src[3] = CLIP((b+3*H) >> 5);
751 src[4] = CLIP((b+4*H) >> 5);
752 src[5] = CLIP((b+5*H) >> 5);
753 src[6] = CLIP((b+6*H) >> 5);
754 src[7] = CLIP((b+7*H) >> 5);
755 src += stride;
756 }
757 }
758
759 #define SRC(x,y) src[(x)+(y)*stride]
760 #define PL(y) \
761 const int l##y = (SRC(-1,y-1) + 2*SRC(-1,y) + SRC(-1,y+1) + 2) >> 2;
762 #define PREDICT_8x8_LOAD_LEFT \
763 const int l0 = ((has_topleft ? SRC(-1,-1) : SRC(-1,0)) \
764 + 2*SRC(-1,0) + SRC(-1,1) + 2) >> 2; \
765 PL(1) PL(2) PL(3) PL(4) PL(5) PL(6) \
766 const int l7 av_unused = (SRC(-1,6) + 3*SRC(-1,7) + 2) >> 2
767
768 #define PT(x) \
769 const int t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
770 #define PREDICT_8x8_LOAD_TOP \
771 const int t0 = ((has_topleft ? SRC(-1,-1) : SRC(0,-1)) \
772 + 2*SRC(0,-1) + SRC(1,-1) + 2) >> 2; \
773 PT(1) PT(2) PT(3) PT(4) PT(5) PT(6) \
774 const int t7 av_unused = ((has_topright ? SRC(8,-1) : SRC(7,-1)) \
775 + 2*SRC(7,-1) + SRC(6,-1) + 2) >> 2
776
777 #define PTR(x) \
778 t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
779 #define PREDICT_8x8_LOAD_TOPRIGHT \
780 int t8, t9, t10, t11, t12, t13, t14, t15; \
781 if(has_topright) { \
782 PTR(8) PTR(9) PTR(10) PTR(11) PTR(12) PTR(13) PTR(14) \
783 t15 = (SRC(14,-1) + 3*SRC(15,-1) + 2) >> 2; \
784 } else t8=t9=t10=t11=t12=t13=t14=t15= SRC(7,-1);
785
786 #define PREDICT_8x8_LOAD_TOPLEFT \
787 const int lt = (SRC(-1,0) + 2*SRC(-1,-1) + SRC(0,-1) + 2) >> 2
788
789 #define PREDICT_8x8_DC(v) \
790 int y; \
791 for( y = 0; y < 8; y++ ) { \
792 AV_WN4PA(((pixel4*)src)+0, v); \
793 AV_WN4PA(((pixel4*)src)+1, v); \
794 src += stride; \
795 }
796
797 static void FUNCC(pred8x8l_128_dc)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
798 {
799 pixel *src = (pixel*)_src;
800 int stride = _stride/sizeof(pixel);
801
802 PREDICT_8x8_DC(PIXEL_SPLAT_X4(1<<(BIT_DEPTH-1)));
803 }
804 static void FUNCC(pred8x8l_left_dc)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
805 {
806 pixel *src = (pixel*)_src;
807 int stride = _stride/sizeof(pixel);
808
809 PREDICT_8x8_LOAD_LEFT;
810 const pixel4 dc = PIXEL_SPLAT_X4((l0+l1+l2+l3+l4+l5+l6+l7+4) >> 3);
811 PREDICT_8x8_DC(dc);
812 }
813 static void FUNCC(pred8x8l_top_dc)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
814 {
815 pixel *src = (pixel*)_src;
816 int stride = _stride/sizeof(pixel);
817
818 PREDICT_8x8_LOAD_TOP;
819 const pixel4 dc = PIXEL_SPLAT_X4((t0+t1+t2+t3+t4+t5+t6+t7+4) >> 3);
820 PREDICT_8x8_DC(dc);
821 }
822 static void FUNCC(pred8x8l_dc)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
823 {
824 pixel *src = (pixel*)_src;
825 int stride = _stride/sizeof(pixel);
826
827 PREDICT_8x8_LOAD_LEFT;
828 PREDICT_8x8_LOAD_TOP;
829 const pixel4 dc = PIXEL_SPLAT_X4((l0+l1+l2+l3+l4+l5+l6+l7
830 +t0+t1+t2+t3+t4+t5+t6+t7+8) >> 4);
831 PREDICT_8x8_DC(dc);
832 }
833 static void FUNCC(pred8x8l_horizontal)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
834 {
835 pixel *src = (pixel*)_src;
836 int stride = _stride/sizeof(pixel);
837 pixel4 a;
838
839 PREDICT_8x8_LOAD_LEFT;
840 #define ROW(y) a = PIXEL_SPLAT_X4(l##y); \
841 AV_WN4PA(src+y*stride, a); \
842 AV_WN4PA(src+y*stride+4, a);
843 ROW(0); ROW(1); ROW(2); ROW(3); ROW(4); ROW(5); ROW(6); ROW(7);
844 #undef ROW
845 }
846 static void FUNCC(pred8x8l_vertical)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
847 {
848 int y;
849 pixel *src = (pixel*)_src;
850 int stride = _stride/sizeof(pixel);
851 pixel4 a, b;
852
853 PREDICT_8x8_LOAD_TOP;
854 src[0] = t0;
855 src[1] = t1;
856 src[2] = t2;
857 src[3] = t3;
858 src[4] = t4;
859 src[5] = t5;
860 src[6] = t6;
861 src[7] = t7;
862 a = AV_RN4PA(((pixel4*)src)+0);
863 b = AV_RN4PA(((pixel4*)src)+1);
864 for( y = 1; y < 8; y++ ) {
865 AV_WN4PA(((pixel4*)(src+y*stride))+0, a);
866 AV_WN4PA(((pixel4*)(src+y*stride))+1, b);
867 }
868 }
869 static void FUNCC(pred8x8l_down_left)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
870 {
871 pixel *src = (pixel*)_src;
872 int stride = _stride/sizeof(pixel);
873 PREDICT_8x8_LOAD_TOP;
874 PREDICT_8x8_LOAD_TOPRIGHT;
875 SRC(0,0)= (t0 + 2*t1 + t2 + 2) >> 2;
876 SRC(0,1)=SRC(1,0)= (t1 + 2*t2 + t3 + 2) >> 2;
877 SRC(0,2)=SRC(1,1)=SRC(2,0)= (t2 + 2*t3 + t4 + 2) >> 2;
878 SRC(0,3)=SRC(1,2)=SRC(2,1)=SRC(3,0)= (t3 + 2*t4 + t5 + 2) >> 2;
879 SRC(0,4)=SRC(1,3)=SRC(2,2)=SRC(3,1)=SRC(4,0)= (t4 + 2*t5 + t6 + 2) >> 2;
880 SRC(0,5)=SRC(1,4)=SRC(2,3)=SRC(3,2)=SRC(4,1)=SRC(5,0)= (t5 + 2*t6 + t7 + 2) >> 2;
881 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;
882 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;
883 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;
884 SRC(2,7)=SRC(3,6)=SRC(4,5)=SRC(5,4)=SRC(6,3)=SRC(7,2)= (t9 + 2*t10 + t11 + 2) >> 2;
885 SRC(3,7)=SRC(4,6)=SRC(5,5)=SRC(6,4)=SRC(7,3)= (t10 + 2*t11 + t12 + 2) >> 2;
886 SRC(4,7)=SRC(5,6)=SRC(6,5)=SRC(7,4)= (t11 + 2*t12 + t13 + 2) >> 2;
887 SRC(5,7)=SRC(6,6)=SRC(7,5)= (t12 + 2*t13 + t14 + 2) >> 2;
888 SRC(6,7)=SRC(7,6)= (t13 + 2*t14 + t15 + 2) >> 2;
889 SRC(7,7)= (t14 + 3*t15 + 2) >> 2;
890 }
891 static void FUNCC(pred8x8l_down_right)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
892 {
893 pixel *src = (pixel*)_src;
894 int stride = _stride/sizeof(pixel);
895 PREDICT_8x8_LOAD_TOP;
896 PREDICT_8x8_LOAD_LEFT;
897 PREDICT_8x8_LOAD_TOPLEFT;
898 SRC(0,7)= (l7 + 2*l6 + l5 + 2) >> 2;
899 SRC(0,6)=SRC(1,7)= (l6 + 2*l5 + l4 + 2) >> 2;
900 SRC(0,5)=SRC(1,6)=SRC(2,7)= (l5 + 2*l4 + l3 + 2) >> 2;
901 SRC(0,4)=SRC(1,5)=SRC(2,6)=SRC(3,7)= (l4 + 2*l3 + l2 + 2) >> 2;
902 SRC(0,3)=SRC(1,4)=SRC(2,5)=SRC(3,6)=SRC(4,7)= (l3 + 2*l2 + l1 + 2) >> 2;
903 SRC(0,2)=SRC(1,3)=SRC(2,4)=SRC(3,5)=SRC(4,6)=SRC(5,7)= (l2 + 2*l1 + l0 + 2) >> 2;
904 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;
905 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;
906 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;
907 SRC(2,0)=SRC(3,1)=SRC(4,2)=SRC(5,3)=SRC(6,4)=SRC(7,5)= (t0 + 2*t1 + t2 + 2) >> 2;
908 SRC(3,0)=SRC(4,1)=SRC(5,2)=SRC(6,3)=SRC(7,4)= (t1 + 2*t2 + t3 + 2) >> 2;
909 SRC(4,0)=SRC(5,1)=SRC(6,2)=SRC(7,3)= (t2 + 2*t3 + t4 + 2) >> 2;
910 SRC(5,0)=SRC(6,1)=SRC(7,2)= (t3 + 2*t4 + t5 + 2) >> 2;
911 SRC(6,0)=SRC(7,1)= (t4 + 2*t5 + t6 + 2) >> 2;
912 SRC(7,0)= (t5 + 2*t6 + t7 + 2) >> 2;
913 }
914 static void FUNCC(pred8x8l_vertical_right)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
915 {
916 pixel *src = (pixel*)_src;
917 int stride = _stride/sizeof(pixel);
918 PREDICT_8x8_LOAD_TOP;
919 PREDICT_8x8_LOAD_LEFT;
920 PREDICT_8x8_LOAD_TOPLEFT;
921 SRC(0,6)= (l5 + 2*l4 + l3 + 2) >> 2;
922 SRC(0,7)= (l6 + 2*l5 + l4 + 2) >> 2;
923 SRC(0,4)=SRC(1,6)= (l3 + 2*l2 + l1 + 2) >> 2;
924 SRC(0,5)=SRC(1,7)= (l4 + 2*l3 + l2 + 2) >> 2;
925 SRC(0,2)=SRC(1,4)=SRC(2,6)= (l1 + 2*l0 + lt + 2) >> 2;
926 SRC(0,3)=SRC(1,5)=SRC(2,7)= (l2 + 2*l1 + l0 + 2) >> 2;
927 SRC(0,1)=SRC(1,3)=SRC(2,5)=SRC(3,7)= (l0 + 2*lt + t0 + 2) >> 2;
928 SRC(0,0)=SRC(1,2)=SRC(2,4)=SRC(3,6)= (lt + t0 + 1) >> 1;
929 SRC(1,1)=SRC(2,3)=SRC(3,5)=SRC(4,7)= (lt + 2*t0 + t1 + 2) >> 2;
930 SRC(1,0)=SRC(2,2)=SRC(3,4)=SRC(4,6)= (t0 + t1 + 1) >> 1;
931 SRC(2,1)=SRC(3,3)=SRC(4,5)=SRC(5,7)= (t0 + 2*t1 + t2 + 2) >> 2;
932 SRC(2,0)=SRC(3,2)=SRC(4,4)=SRC(5,6)= (t1 + t2 + 1) >> 1;
933 SRC(3,1)=SRC(4,3)=SRC(5,5)=SRC(6,7)= (t1 + 2*t2 + t3 + 2) >> 2;
934 SRC(3,0)=SRC(4,2)=SRC(5,4)=SRC(6,6)= (t2 + t3 + 1) >> 1;
935 SRC(4,1)=SRC(5,3)=SRC(6,5)=SRC(7,7)= (t2 + 2*t3 + t4 + 2) >> 2;
936 SRC(4,0)=SRC(5,2)=SRC(6,4)=SRC(7,6)= (t3 + t4 + 1) >> 1;
937 SRC(5,1)=SRC(6,3)=SRC(7,5)= (t3 + 2*t4 + t5 + 2) >> 2;
938 SRC(5,0)=SRC(6,2)=SRC(7,4)= (t4 + t5 + 1) >> 1;
939 SRC(6,1)=SRC(7,3)= (t4 + 2*t5 + t6 + 2) >> 2;
940 SRC(6,0)=SRC(7,2)= (t5 + t6 + 1) >> 1;
941 SRC(7,1)= (t5 + 2*t6 + t7 + 2) >> 2;
942 SRC(7,0)= (t6 + t7 + 1) >> 1;
943 }
944 static void FUNCC(pred8x8l_horizontal_down)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
945 {
946 pixel *src = (pixel*)_src;
947 int stride = _stride/sizeof(pixel);
948 PREDICT_8x8_LOAD_TOP;
949 PREDICT_8x8_LOAD_LEFT;
950 PREDICT_8x8_LOAD_TOPLEFT;
951 SRC(0,7)= (l6 + l7 + 1) >> 1;
952 SRC(1,7)= (l5 + 2*l6 + l7 + 2) >> 2;
953 SRC(0,6)=SRC(2,7)= (l5 + l6 + 1) >> 1;
954 SRC(1,6)=SRC(3,7)= (l4 + 2*l5 + l6 + 2) >> 2;
955 SRC(0,5)=SRC(2,6)=SRC(4,7)= (l4 + l5 + 1) >> 1;
956 SRC(1,5)=SRC(3,6)=SRC(5,7)= (l3 + 2*l4 + l5 + 2) >> 2;
957 SRC(0,4)=SRC(2,5)=SRC(4,6)=SRC(6,7)= (l3 + l4 + 1) >> 1;
958 SRC(1,4)=SRC(3,5)=SRC(5,6)=SRC(7,7)= (l2 + 2*l3 + l4 + 2) >> 2;
959 SRC(0,3)=SRC(2,4)=SRC(4,5)=SRC(6,6)= (l2 + l3 + 1) >> 1;
960 SRC(1,3)=SRC(3,4)=SRC(5,5)=SRC(7,6)= (l1 + 2*l2 + l3 + 2) >> 2;
961 SRC(0,2)=SRC(2,3)=SRC(4,4)=SRC(6,5)= (l1 + l2 + 1) >> 1;
962 SRC(1,2)=SRC(3,3)=SRC(5,4)=SRC(7,5)= (l0 + 2*l1 + l2 + 2) >> 2;
963 SRC(0,1)=SRC(2,2)=SRC(4,3)=SRC(6,4)= (l0 + l1 + 1) >> 1;
964 SRC(1,1)=SRC(3,2)=SRC(5,3)=SRC(7,4)= (lt + 2*l0 + l1 + 2) >> 2;
965 SRC(0,0)=SRC(2,1)=SRC(4,2)=SRC(6,3)= (lt + l0 + 1) >> 1;
966 SRC(1,0)=SRC(3,1)=SRC(5,2)=SRC(7,3)= (l0 + 2*lt + t0 + 2) >> 2;
967 SRC(2,0)=SRC(4,1)=SRC(6,2)= (t1 + 2*t0 + lt + 2) >> 2;
968 SRC(3,0)=SRC(5,1)=SRC(7,2)= (t2 + 2*t1 + t0 + 2) >> 2;
969 SRC(4,0)=SRC(6,1)= (t3 + 2*t2 + t1 + 2) >> 2;
970 SRC(5,0)=SRC(7,1)= (t4 + 2*t3 + t2 + 2) >> 2;
971 SRC(6,0)= (t5 + 2*t4 + t3 + 2) >> 2;
972 SRC(7,0)= (t6 + 2*t5 + t4 + 2) >> 2;
973 }
974 static void FUNCC(pred8x8l_vertical_left)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
975 {
976 pixel *src = (pixel*)_src;
977 int stride = _stride/sizeof(pixel);
978 PREDICT_8x8_LOAD_TOP;
979 PREDICT_8x8_LOAD_TOPRIGHT;
980 SRC(0,0)= (t0 + t1 + 1) >> 1;
981 SRC(0,1)= (t0 + 2*t1 + t2 + 2) >> 2;
982 SRC(0,2)=SRC(1,0)= (t1 + t2 + 1) >> 1;
983 SRC(0,3)=SRC(1,1)= (t1 + 2*t2 + t3 + 2) >> 2;
984 SRC(0,4)=SRC(1,2)=SRC(2,0)= (t2 + t3 + 1) >> 1;
985 SRC(0,5)=SRC(1,3)=SRC(2,1)= (t2 + 2*t3 + t4 + 2) >> 2;
986 SRC(0,6)=SRC(1,4)=SRC(2,2)=SRC(3,0)= (t3 + t4 + 1) >> 1;
987 SRC(0,7)=SRC(1,5)=SRC(2,3)=SRC(3,1)= (t3 + 2*t4 + t5 + 2) >> 2;
988 SRC(1,6)=SRC(2,4)=SRC(3,2)=SRC(4,0)= (t4 + t5 + 1) >> 1;
989 SRC(1,7)=SRC(2,5)=SRC(3,3)=SRC(4,1)= (t4 + 2*t5 + t6 + 2) >> 2;
990 SRC(2,6)=SRC(3,4)=SRC(4,2)=SRC(5,0)= (t5 + t6 + 1) >> 1;
991 SRC(2,7)=SRC(3,5)=SRC(4,3)=SRC(5,1)= (t5 + 2*t6 + t7 + 2) >> 2;
992 SRC(3,6)=SRC(4,4)=SRC(5,2)=SRC(6,0)= (t6 + t7 + 1) >> 1;
993 SRC(3,7)=SRC(4,5)=SRC(5,3)=SRC(6,1)= (t6 + 2*t7 + t8 + 2) >> 2;
994 SRC(4,6)=SRC(5,4)=SRC(6,2)=SRC(7,0)= (t7 + t8 + 1) >> 1;
995 SRC(4,7)=SRC(5,5)=SRC(6,3)=SRC(7,1)= (t7 + 2*t8 + t9 + 2) >> 2;
996 SRC(5,6)=SRC(6,4)=SRC(7,2)= (t8 + t9 + 1) >> 1;
997 SRC(5,7)=SRC(6,5)=SRC(7,3)= (t8 + 2*t9 + t10 + 2) >> 2;
998 SRC(6,6)=SRC(7,4)= (t9 + t10 + 1) >> 1;
999 SRC(6,7)=SRC(7,5)= (t9 + 2*t10 + t11 + 2) >> 2;
1000 SRC(7,6)= (t10 + t11 + 1) >> 1;
1001 SRC(7,7)= (t10 + 2*t11 + t12 + 2) >> 2;
1002 }
1003 static void FUNCC(pred8x8l_horizontal_up)(uint8_t *_src, int has_topleft, int has_topright, int _stride)
1004 {
1005 pixel *src = (pixel*)_src;
1006 int stride = _stride/sizeof(pixel);
1007 PREDICT_8x8_LOAD_LEFT;
1008 SRC(0,0)= (l0 + l1 + 1) >> 1;
1009 SRC(1,0)= (l0 + 2*l1 + l2 + 2) >> 2;
1010 SRC(0,1)=SRC(2,0)= (l1 + l2 + 1) >> 1;
1011 SRC(1,1)=SRC(3,0)= (l1 + 2*l2 + l3 + 2) >> 2;
1012 SRC(0,2)=SRC(2,1)=SRC(4,0)= (l2 + l3 + 1) >> 1;
1013 SRC(1,2)=SRC(3,1)=SRC(5,0)= (l2 + 2*l3 + l4 + 2) >> 2;
1014 SRC(0,3)=SRC(2,2)=SRC(4,1)=SRC(6,0)= (l3 + l4 + 1) >> 1;
1015 SRC(1,3)=SRC(3,2)=SRC(5,1)=SRC(7,0)= (l3 + 2*l4 + l5 + 2) >> 2;
1016 SRC(0,4)=SRC(2,3)=SRC(4,2)=SRC(6,1)= (l4 + l5 + 1) >> 1;
1017 SRC(1,4)=SRC(3,3)=SRC(5,2)=SRC(7,1)= (l4 + 2*l5 + l6 + 2) >> 2;
1018 SRC(0,5)=SRC(2,4)=SRC(4,3)=SRC(6,2)= (l5 + l6 + 1) >> 1;
1019 SRC(1,5)=SRC(3,4)=SRC(5,3)=SRC(7,2)= (l5 + 2*l6 + l7 + 2) >> 2;
1020 SRC(0,6)=SRC(2,5)=SRC(4,4)=SRC(6,3)= (l6 + l7 + 1) >> 1;
1021 SRC(1,6)=SRC(3,5)=SRC(5,4)=SRC(7,3)= (l6 + 3*l7 + 2) >> 2;
1022 SRC(0,7)=SRC(1,7)=SRC(2,6)=SRC(2,7)=SRC(3,6)=
1023 SRC(3,7)=SRC(4,5)=SRC(4,6)=SRC(4,7)=SRC(5,5)=
1024 SRC(5,6)=SRC(5,7)=SRC(6,4)=SRC(6,5)=SRC(6,6)=
1025 SRC(6,7)=SRC(7,4)=SRC(7,5)=SRC(7,6)=SRC(7,7)= l7;
1026 }
1027 #undef PREDICT_8x8_LOAD_LEFT
1028 #undef PREDICT_8x8_LOAD_TOP
1029 #undef PREDICT_8x8_LOAD_TOPLEFT
1030 #undef PREDICT_8x8_LOAD_TOPRIGHT
1031 #undef PREDICT_8x8_DC
1032 #undef PTR
1033 #undef PT
1034 #undef PL
1035 #undef SRC
1036
1037 static void FUNCC(pred4x4_vertical_add)(uint8_t *_pix, const DCTELEM *_block, int stride){
1038 int i;
1039 pixel *pix = (pixel*)_pix;
1040 const dctcoef *block = (const dctcoef*)_block;
1041 stride /= sizeof(pixel);
1042 pix -= stride;
1043 for(i=0; i<4; i++){
1044 pixel v = pix[0];
1045 pix[1*stride]= v += block[0];
1046 pix[2*stride]= v += block[4];
1047 pix[3*stride]= v += block[8];
1048 pix[4*stride]= v + block[12];
1049 pix++;
1050 block++;
1051 }
1052 }
1053
1054 static void FUNCC(pred4x4_horizontal_add)(uint8_t *_pix, const DCTELEM *_block, int stride){
1055 int i;
1056 pixel *pix = (pixel*)_pix;
1057 const dctcoef *block = (const dctcoef*)_block;
1058 stride /= sizeof(pixel);
1059 for(i=0; i<4; i++){
1060 pixel v = pix[-1];
1061 pix[0]= v += block[0];
1062 pix[1]= v += block[1];
1063 pix[2]= v += block[2];
1064 pix[3]= v + block[3];
1065 pix+= stride;
1066 block+= 4;
1067 }
1068 }
1069
1070 static void FUNCC(pred8x8l_vertical_add)(uint8_t *_pix, const DCTELEM *_block, int stride){
1071 int i;
1072 pixel *pix = (pixel*)_pix;
1073 const dctcoef *block = (const dctcoef*)_block;
1074 stride /= sizeof(pixel);
1075 pix -= stride;
1076 for(i=0; i<8; i++){
1077 pixel v = pix[0];
1078 pix[1*stride]= v += block[0];
1079 pix[2*stride]= v += block[8];
1080 pix[3*stride]= v += block[16];
1081 pix[4*stride]= v += block[24];
1082 pix[5*stride]= v += block[32];
1083 pix[6*stride]= v += block[40];
1084 pix[7*stride]= v += block[48];
1085 pix[8*stride]= v + block[56];
1086 pix++;
1087 block++;
1088 }
1089 }
1090
1091 static void FUNCC(pred8x8l_horizontal_add)(uint8_t *_pix, const DCTELEM *_block, int stride){
1092 int i;
1093 pixel *pix = (pixel*)_pix;
1094 const dctcoef *block = (const dctcoef*)_block;
1095 stride /= sizeof(pixel);
1096 for(i=0; i<8; i++){
1097 pixel v = pix[-1];
1098 pix[0]= v += block[0];
1099 pix[1]= v += block[1];
1100 pix[2]= v += block[2];
1101 pix[3]= v += block[3];
1102 pix[4]= v += block[4];
1103 pix[5]= v += block[5];
1104 pix[6]= v += block[6];
1105 pix[7]= v + block[7];
1106 pix+= stride;
1107 block+= 8;
1108 }
1109 }
1110
1111 static void FUNCC(pred16x16_vertical_add)(uint8_t *pix, const int *block_offset, const DCTELEM *block, int stride){
1112 int i;
1113 for(i=0; i<16; i++)
1114 FUNCC(pred4x4_vertical_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1115 }
1116
1117 static void FUNCC(pred16x16_horizontal_add)(uint8_t *pix, const int *block_offset, const DCTELEM *block, int stride){
1118 int i;
1119 for(i=0; i<16; i++)
1120 FUNCC(pred4x4_horizontal_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1121 }
1122
1123 static void FUNCC(pred8x8_vertical_add)(uint8_t *pix, const int *block_offset, const DCTELEM *block, int stride){
1124 int i;
1125 for(i=0; i<4; i++)
1126 FUNCC(pred4x4_vertical_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1127 }
1128
1129 static void FUNCC(pred8x8_horizontal_add)(uint8_t *pix, const int *block_offset, const DCTELEM *block, int stride){
1130 int i;
1131 for(i=0; i<4; i++)
1132 FUNCC(pred4x4_horizontal_add)(pix + block_offset[i], block + i*16*sizeof(pixel), stride);
1133 }