1 |
|
#ifndef _INTERPOLATE8X8_H_ |
2 |
|
#define _INTERPOLATE8X8_H_ |
3 |
|
|
4 |
#include "../utils/mem_transfer.h" |
#include "../utils/mem_transfer.h" |
5 |
|
|
6 |
typedef void (INTERPOLATE8X8)(uint8_t * const dst, |
typedef void (INTERPOLATE8X8)(uint8_t * const dst, |
9 |
const uint32_t rounding); |
const uint32_t rounding); |
10 |
typedef INTERPOLATE8X8 * INTERPOLATE8X8_PTR; |
typedef INTERPOLATE8X8 * INTERPOLATE8X8_PTR; |
11 |
|
|
12 |
|
typedef void (INTERPOLATE8X8_AVG2) (uint8_t *dst, |
13 |
|
const uint8_t *src1, |
14 |
|
const uint8_t *src2, |
15 |
|
const uint32_t stride, |
16 |
|
const uint32_t rounding, |
17 |
|
const uint32_t height); |
18 |
|
typedef INTERPOLATE8X8_AVG2 *INTERPOLATE8X8_AVG2_PTR; |
19 |
|
|
20 |
|
typedef void (INTERPOLATE8X8_AVG4) (uint8_t *dst, |
21 |
|
const uint8_t *src1, |
22 |
|
const uint8_t *src2, |
23 |
|
const uint8_t *src3, |
24 |
|
const uint8_t *src4, |
25 |
|
const uint32_t stride, |
26 |
|
const uint32_t rounding); |
27 |
|
typedef INTERPOLATE8X8_AVG4 *INTERPOLATE8X8_AVG4_PTR; |
28 |
|
|
29 |
|
typedef void (INTERPOLATE_LOWPASS) (uint8_t *dst, |
30 |
|
uint8_t *src, |
31 |
|
int32_t stride, |
32 |
|
int32_t rounding); |
33 |
|
|
34 |
|
typedef INTERPOLATE_LOWPASS *INTERPOLATE_LOWPASS_PTR; |
35 |
|
|
36 |
|
typedef void (INTERPOLATE_LOWPASS_HV) (uint8_t *dst1, |
37 |
|
uint8_t *dst2, |
38 |
|
uint8_t *src, |
39 |
|
int32_t stride, |
40 |
|
int32_t rounding); |
41 |
|
|
42 |
|
typedef INTERPOLATE_LOWPASS_HV *INTERPOLATE_LOWPASS_HV_PTR; |
43 |
|
|
44 |
|
typedef void (INTERPOLATE8X8_6TAP_LOWPASS) (uint8_t *dst, |
45 |
|
uint8_t *src, |
46 |
|
int32_t stride, |
47 |
|
int32_t rounding); |
48 |
|
|
49 |
|
typedef INTERPOLATE8X8_6TAP_LOWPASS *INTERPOLATE8X8_6TAP_LOWPASS_PTR; |
50 |
|
|
51 |
extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_h; |
extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_h; |
52 |
extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_v; |
extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_v; |
53 |
extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_hv; |
extern INTERPOLATE8X8_PTR interpolate8x8_halfpel_hv; |
54 |
|
|
55 |
|
extern INTERPOLATE8X8_AVG2_PTR interpolate8x8_avg2; |
56 |
|
extern INTERPOLATE8X8_AVG4_PTR interpolate8x8_avg4; |
57 |
|
|
58 |
|
extern INTERPOLATE_LOWPASS_PTR interpolate8x8_lowpass_h; |
59 |
|
extern INTERPOLATE_LOWPASS_PTR interpolate8x8_lowpass_v; |
60 |
|
|
61 |
|
extern INTERPOLATE_LOWPASS_PTR interpolate16x16_lowpass_h; |
62 |
|
extern INTERPOLATE_LOWPASS_PTR interpolate16x16_lowpass_v; |
63 |
|
|
64 |
|
extern INTERPOLATE_LOWPASS_HV_PTR interpolate8x8_lowpass_hv; |
65 |
|
extern INTERPOLATE_LOWPASS_HV_PTR interpolate16x16_lowpass_hv; |
66 |
|
|
67 |
|
extern INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_h; |
68 |
|
extern INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_v; |
69 |
|
|
70 |
INTERPOLATE8X8 interpolate8x8_halfpel_h_c; |
INTERPOLATE8X8 interpolate8x8_halfpel_h_c; |
71 |
INTERPOLATE8X8 interpolate8x8_halfpel_v_c; |
INTERPOLATE8X8 interpolate8x8_halfpel_v_c; |
72 |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_c; |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_c; |
83 |
INTERPOLATE8X8 interpolate8x8_halfpel_v_3dn; |
INTERPOLATE8X8 interpolate8x8_halfpel_v_3dn; |
84 |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dn; |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dn; |
85 |
|
|
86 |
static __inline void interpolate8x8_switch(uint8_t * const cur, |
INTERPOLATE8X8 interpolate8x8_halfpel_h_ia64; |
87 |
|
INTERPOLATE8X8 interpolate8x8_halfpel_v_ia64; |
88 |
|
INTERPOLATE8X8 interpolate8x8_halfpel_hv_ia64; |
89 |
|
|
90 |
|
INTERPOLATE8X8_AVG2 interpolate8x8_avg2_c; |
91 |
|
INTERPOLATE8X8_AVG4 interpolate8x8_avg4_c; |
92 |
|
|
93 |
|
INTERPOLATE8X8_AVG2 interpolate8x8_avg2_mmx; |
94 |
|
INTERPOLATE8X8_AVG4 interpolate8x8_avg4_mmx; |
95 |
|
|
96 |
|
INTERPOLATE_LOWPASS interpolate8x8_lowpass_h_c; |
97 |
|
INTERPOLATE_LOWPASS interpolate8x8_lowpass_v_c; |
98 |
|
|
99 |
|
INTERPOLATE_LOWPASS interpolate16x16_lowpass_h_c; |
100 |
|
INTERPOLATE_LOWPASS interpolate16x16_lowpass_v_c; |
101 |
|
|
102 |
|
INTERPOLATE_LOWPASS_HV interpolate8x8_lowpass_hv_c; |
103 |
|
INTERPOLATE_LOWPASS_HV interpolate16x16_lowpass_hv_c; |
104 |
|
|
105 |
|
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_h_c; |
106 |
|
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_v_c; |
107 |
|
|
108 |
|
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_h_mmx; |
109 |
|
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_v_mmx; |
110 |
|
|
111 |
|
static __inline void |
112 |
|
interpolate8x8_switch(uint8_t * const cur, |
113 |
const uint8_t * const refn, |
const uint8_t * const refn, |
114 |
const uint32_t x, const uint32_t y, |
const uint32_t x, |
115 |
const int32_t dx, const int dy, |
const uint32_t y, |
116 |
|
const int32_t dx, |
117 |
|
const int dy, |
118 |
const uint32_t stride, |
const uint32_t stride, |
119 |
const uint32_t rounding) |
const uint32_t rounding) |
120 |
{ |
{ |
125 |
case 0 : |
case 0 : |
126 |
ddx = dx/2; |
ddx = dx/2; |
127 |
ddy = dy/2; |
ddy = dy/2; |
128 |
transfer8x8_copy(cur + y*stride + x, refn + (y+ddy)*stride + x + ddx, stride); |
transfer8x8_copy(cur + y * stride + x, |
129 |
|
refn + (int)((y + ddy) * stride + x + ddx), stride); |
130 |
break; |
break; |
131 |
|
|
132 |
case 1 : |
case 1 : |
133 |
ddx = dx/2; |
ddx = dx/2; |
134 |
ddy = (dy-1)/2; |
ddy = (dy-1)/2; |
135 |
interpolate8x8_halfpel_v(cur + y*stride + x, |
interpolate8x8_halfpel_v(cur + y*stride + x, |
136 |
refn + (y+ddy)*stride + x + ddx, stride, rounding); |
refn + (int)((y + ddy) * stride + x + ddx), stride, |
137 |
|
rounding); |
138 |
break; |
break; |
139 |
|
|
140 |
case 2 : |
case 2 : |
141 |
ddx = (dx-1)/2; |
ddx = (dx-1)/2; |
142 |
ddy = dy/2; |
ddy = dy/2; |
143 |
interpolate8x8_halfpel_h(cur + y*stride + x, |
interpolate8x8_halfpel_h(cur + y*stride + x, |
144 |
refn + (y+ddy)*stride + x + ddx, stride, rounding); |
refn + (int)((y + ddy) * stride + x + ddx), stride, |
145 |
|
rounding); |
146 |
break; |
break; |
147 |
|
|
148 |
default : |
default : |
149 |
ddx = (dx-1)/2; |
ddx = (dx-1)/2; |
150 |
ddy = (dy-1)/2; |
ddy = (dy-1)/2; |
151 |
interpolate8x8_halfpel_hv(cur + y*stride + x, |
interpolate8x8_halfpel_hv(cur + y*stride + x, |
152 |
refn + (y+ddy)*stride + x + ddx, stride, rounding); |
refn + (int)((y + ddy) * stride + x + ddx), stride, |
153 |
|
rounding); |
154 |
break; |
break; |
155 |
} |
} |
156 |
} |
} |
157 |
|
|
158 |
|
|
159 |
void interpolate8x8_c(uint8_t * const dst, |
static __inline void |
160 |
const uint8_t * const src, |
interpolate16x16_switch(uint8_t * const cur, |
161 |
|
const uint8_t * const refn, |
162 |
|
const uint32_t x, |
163 |
|
const uint32_t y, |
164 |
|
const int32_t dx, |
165 |
|
const int dy, |
166 |
|
const uint32_t stride, |
167 |
|
const uint32_t rounding) |
168 |
|
{ |
169 |
|
interpolate8x8_switch(cur, refn, x, y, dx, dy, stride, rounding); |
170 |
|
interpolate8x8_switch(cur, refn, x+8, y, dx, dy, stride, rounding); |
171 |
|
interpolate8x8_switch(cur, refn, x, y+8, dx, dy, stride, rounding); |
172 |
|
interpolate8x8_switch(cur, refn, x+8, y+8, dx, dy, stride, rounding); |
173 |
|
} |
174 |
|
|
175 |
|
|
176 |
|
static __inline void |
177 |
|
interpolate32x32_switch(uint8_t * const cur, |
178 |
|
const uint8_t * const refn, |
179 |
|
const uint32_t x, |
180 |
|
const uint32_t y, |
181 |
|
const int32_t dx, |
182 |
|
const int dy, |
183 |
|
const uint32_t stride, |
184 |
|
const uint32_t rounding) |
185 |
|
{ |
186 |
|
interpolate16x16_switch(cur, refn, x, y, dx, dy, stride, rounding); |
187 |
|
interpolate16x16_switch(cur, refn, x+16, y, dx, dy, stride, rounding); |
188 |
|
interpolate16x16_switch(cur, refn, x, y+16, dx, dy, stride, rounding); |
189 |
|
interpolate16x16_switch(cur, refn, x+16, y+16, dx, dy, stride, rounding); |
190 |
|
} |
191 |
|
|
192 |
|
|
193 |
|
static __inline uint8_t * |
194 |
|
interpolate8x8_switch2(uint8_t * const buffer, |
195 |
|
const uint8_t * const refn, |
196 |
|
const uint32_t x, |
197 |
|
const uint32_t y, |
198 |
|
const int32_t dx, |
199 |
|
const int dy, |
200 |
|
const uint32_t stride, |
201 |
|
const uint32_t rounding) |
202 |
|
{ |
203 |
|
int32_t ddx, ddy; |
204 |
|
|
205 |
|
switch (((dx & 1) << 1) + (dy & 1)) // ((dx%2)?2:0)+((dy%2)?1:0) |
206 |
|
{ |
207 |
|
case 0: |
208 |
|
return (uint8_t *)refn + (int)((y + dy/2) * stride + x + dx/2); |
209 |
|
|
210 |
|
case 1: |
211 |
|
ddx = dx / 2; |
212 |
|
ddy = (dy - 1) / 2; |
213 |
|
interpolate8x8_halfpel_v(buffer, |
214 |
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
215 |
|
rounding); |
216 |
|
break; |
217 |
|
|
218 |
|
case 2: |
219 |
|
ddx = (dx - 1) / 2; |
220 |
|
ddy = dy / 2; |
221 |
|
interpolate8x8_halfpel_h(buffer, |
222 |
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
223 |
|
rounding); |
224 |
|
break; |
225 |
|
|
226 |
|
default: |
227 |
|
ddx = (dx - 1) / 2; |
228 |
|
ddy = (dy - 1) / 2; |
229 |
|
interpolate8x8_halfpel_hv(buffer, |
230 |
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
231 |
|
rounding); |
232 |
|
break; |
233 |
|
} |
234 |
|
return buffer; |
235 |
|
} |
236 |
|
|
237 |
|
static __inline void interpolate8x8_quarterpel(uint8_t * const cur, |
238 |
|
uint8_t * const refn, |
239 |
|
uint8_t * const refh, |
240 |
|
uint8_t * const refv, |
241 |
|
uint8_t * const refhv, |
242 |
|
const uint32_t x, const uint32_t y, |
243 |
|
const int32_t dx, const int dy, |
244 |
|
const uint32_t stride, |
245 |
|
const uint32_t rounding) |
246 |
|
{ |
247 |
|
const int32_t xRef = x*4 + dx; |
248 |
|
const int32_t yRef = y*4 + dy; |
249 |
|
|
250 |
|
uint8_t *src, *dst; |
251 |
|
uint8_t *halfpel_h, *halfpel_v, *halfpel_hv; |
252 |
|
int32_t x_int, y_int, x_frac, y_frac; |
253 |
|
|
254 |
|
x_int = xRef/4; |
255 |
|
if (xRef < 0 && xRef % 4) |
256 |
|
x_int--; |
257 |
|
|
258 |
|
x_frac = xRef - (4*x_int); |
259 |
|
|
260 |
|
y_int = yRef/4; |
261 |
|
if (yRef < 0 && yRef % 4) |
262 |
|
y_int--; |
263 |
|
|
264 |
|
y_frac = yRef - (4*y_int); |
265 |
|
|
266 |
|
src = refn + y_int * stride + x_int; |
267 |
|
halfpel_h = refh; |
268 |
|
halfpel_v = refv; |
269 |
|
halfpel_hv = refhv; |
270 |
|
|
271 |
|
dst = cur + y * stride + x; |
272 |
|
|
273 |
|
switch((y_frac << 2) | (x_frac)) { |
274 |
|
|
275 |
|
case 0: |
276 |
|
transfer8x8_copy(dst, src, stride); |
277 |
|
break; |
278 |
|
|
279 |
|
case 1: |
280 |
|
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
281 |
|
interpolate8x8_avg2(dst, src, halfpel_h, stride, rounding, 8); |
282 |
|
break; |
283 |
|
|
284 |
|
case 2: |
285 |
|
interpolate8x8_lowpass_h(dst, src, stride, rounding); |
286 |
|
break; |
287 |
|
|
288 |
|
case 3: |
289 |
|
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
290 |
|
interpolate8x8_avg2(dst, src + 1, halfpel_h, stride, rounding, 8); |
291 |
|
break; |
292 |
|
|
293 |
|
case 4: |
294 |
|
interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); |
295 |
|
interpolate8x8_avg2(dst, src, halfpel_v, stride, rounding, 8); |
296 |
|
break; |
297 |
|
|
298 |
|
case 5: |
299 |
|
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
300 |
|
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
301 |
|
interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
302 |
|
interpolate8x8_avg2(dst, halfpel_v, halfpel_hv, stride, rounding, 8); |
303 |
|
break; |
304 |
|
|
305 |
|
case 6: |
306 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
307 |
|
interpolate8x8_avg2(dst, halfpel_h, halfpel_hv, stride, rounding, 8); |
308 |
|
break; |
309 |
|
|
310 |
|
case 7: |
311 |
|
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
312 |
|
interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9); |
313 |
|
interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
314 |
|
interpolate8x8_avg2(dst, halfpel_v, halfpel_hv, stride, rounding, 8); |
315 |
|
break; |
316 |
|
|
317 |
|
case 8: |
318 |
|
interpolate8x8_lowpass_v(dst, src, stride, rounding); |
319 |
|
break; |
320 |
|
|
321 |
|
case 9: |
322 |
|
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
323 |
|
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
324 |
|
interpolate8x8_lowpass_v(dst, halfpel_v, stride, rounding); |
325 |
|
break; |
326 |
|
|
327 |
|
case 10: |
328 |
|
interpolate8x8_lowpass_hv(dst, halfpel_h, src, stride, rounding); |
329 |
|
break; |
330 |
|
|
331 |
|
case 11: |
332 |
|
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
333 |
|
interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9); |
334 |
|
interpolate8x8_lowpass_v(dst, halfpel_v, stride, rounding); |
335 |
|
break; |
336 |
|
|
337 |
|
case 12: |
338 |
|
interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); |
339 |
|
interpolate8x8_avg2(dst, src+stride, halfpel_v, stride, rounding, 8); |
340 |
|
break; |
341 |
|
|
342 |
|
case 13: |
343 |
|
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
344 |
|
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
345 |
|
interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
346 |
|
interpolate8x8_avg2(dst, halfpel_v+stride, halfpel_hv, stride, rounding, 8); |
347 |
|
break; |
348 |
|
|
349 |
|
case 14: |
350 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
351 |
|
interpolate8x8_avg2(dst, halfpel_h+stride, halfpel_hv, stride, rounding, 8); |
352 |
|
break; |
353 |
|
|
354 |
|
case 15: |
355 |
|
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
356 |
|
interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9); |
357 |
|
interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
358 |
|
interpolate8x8_avg2(dst, halfpel_hv, halfpel_v + stride, stride, rounding, 8); |
359 |
|
break; |
360 |
|
} |
361 |
|
} |
362 |
|
|
363 |
|
static __inline void interpolate16x16_quarterpel(uint8_t * const cur, |
364 |
|
uint8_t * const refn, |
365 |
|
uint8_t * const refh, |
366 |
|
uint8_t * const refv, |
367 |
|
uint8_t * const refhv, |
368 |
const uint32_t x, const uint32_t y, |
const uint32_t x, const uint32_t y, |
369 |
const uint32_t stride); |
const int32_t dx, const int dy, |
370 |
|
const uint32_t stride, |
371 |
|
const uint32_t rounding) |
372 |
|
{ |
373 |
|
const int32_t xRef = x*4 + dx; |
374 |
|
const int32_t yRef = y*4 + dy; |
375 |
|
|
376 |
|
uint8_t *src, *dst; |
377 |
|
uint8_t *halfpel_h, *halfpel_v, *halfpel_hv; |
378 |
|
int32_t x_int, y_int, x_frac, y_frac; |
379 |
|
|
380 |
|
x_int = xRef/4; |
381 |
|
if (xRef < 0 && xRef % 4) |
382 |
|
x_int--; |
383 |
|
|
384 |
|
x_frac = xRef - (4*x_int); |
385 |
|
|
386 |
|
y_int = yRef/4; |
387 |
|
if (yRef < 0 && yRef % 4) |
388 |
|
y_int--; |
389 |
|
|
390 |
|
y_frac = yRef - (4*y_int); |
391 |
|
|
392 |
|
src = refn + y_int * stride + x_int; |
393 |
|
halfpel_h = refh; |
394 |
|
halfpel_v = refv; |
395 |
|
halfpel_hv = refhv; |
396 |
|
|
397 |
|
dst = cur + y * stride + x; |
398 |
|
|
399 |
|
switch((y_frac << 2) | (x_frac)) { |
400 |
|
|
401 |
|
case 0: |
402 |
|
transfer8x8_copy(dst, src, stride); |
403 |
|
transfer8x8_copy(dst+8, src+8, stride); |
404 |
|
transfer8x8_copy(dst+8*stride, src+8*stride, stride); |
405 |
|
transfer8x8_copy(dst+8*stride+8, src+8*stride+8, stride); |
406 |
|
break; |
407 |
|
|
408 |
|
case 1: |
409 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
410 |
|
interpolate8x8_avg2(dst, src, halfpel_h, stride, rounding, 8); |
411 |
|
interpolate8x8_avg2(dst+8, src+8, halfpel_h+8, stride, rounding, 8); |
412 |
|
interpolate8x8_avg2(dst+8*stride, src+8*stride, halfpel_h+8*stride, stride, rounding, 8); |
413 |
|
interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 8); |
414 |
|
break; |
415 |
|
|
416 |
|
case 2: |
417 |
|
interpolate16x16_lowpass_h(dst, src, stride, rounding); |
418 |
|
break; |
419 |
|
|
420 |
|
case 3: |
421 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
422 |
|
interpolate8x8_avg2(dst, src + 1, halfpel_h, stride, rounding, 8); |
423 |
|
interpolate8x8_avg2(dst+8, src + 8 + 1, halfpel_h+8, stride, rounding, 8); |
424 |
|
interpolate8x8_avg2(dst+8*stride, src + 8*stride + 1, halfpel_h+8*stride, stride, rounding, 8); |
425 |
|
interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8 + 1, halfpel_h+8*stride+8, stride, rounding, 8); |
426 |
|
break; |
427 |
|
|
428 |
|
case 4: |
429 |
|
interpolate16x16_lowpass_v(halfpel_v, src, stride, rounding); |
430 |
|
interpolate8x8_avg2(dst, src, halfpel_v, stride, rounding, 8); |
431 |
|
interpolate8x8_avg2(dst+8, src+8, halfpel_v+8, stride, rounding, 8); |
432 |
|
interpolate8x8_avg2(dst+8*stride, src+8*stride, halfpel_v+8*stride, stride, rounding, 8); |
433 |
|
interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); |
434 |
|
break; |
435 |
|
|
436 |
|
case 5: |
437 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
438 |
|
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
439 |
|
interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9); |
440 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
441 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
442 |
|
|
443 |
|
interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
444 |
|
interpolate8x8_avg2(dst, halfpel_hv, halfpel_v, stride, rounding, 8); |
445 |
|
interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+8, stride, rounding, 8); |
446 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+8*stride, stride, rounding, 8); |
447 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); |
448 |
|
break; |
449 |
|
|
450 |
|
case 6: |
451 |
|
interpolate16x16_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
452 |
|
interpolate8x8_avg2(dst, halfpel_h, halfpel_hv, stride, rounding, 8); |
453 |
|
interpolate8x8_avg2(dst+8, halfpel_h+8, halfpel_hv+8, stride, rounding, 8); |
454 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_h+8*stride, halfpel_hv+8*stride, stride, rounding, 8); |
455 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_h+8*stride+8, halfpel_hv+8*stride+8, stride, rounding, 8); |
456 |
|
break; |
457 |
|
|
458 |
|
case 7: |
459 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
460 |
|
interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9); |
461 |
|
interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9); |
462 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
463 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
464 |
|
|
465 |
|
interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
466 |
|
interpolate8x8_avg2(dst, halfpel_hv, halfpel_v, stride, rounding, 8); |
467 |
|
interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+8, stride, rounding, 8); |
468 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+8*stride, stride, rounding, 8); |
469 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); |
470 |
|
break; |
471 |
|
|
472 |
|
case 8: |
473 |
|
interpolate16x16_lowpass_v(dst, src, stride, rounding); |
474 |
|
break; |
475 |
|
|
476 |
|
case 9: |
477 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
478 |
|
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
479 |
|
interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9); |
480 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
481 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
482 |
|
interpolate16x16_lowpass_v(dst, halfpel_v, stride, rounding); |
483 |
|
break; |
484 |
|
|
485 |
|
case 10: |
486 |
|
interpolate16x16_lowpass_hv(dst, halfpel_h, src, stride, rounding); |
487 |
|
break; |
488 |
|
|
489 |
|
case 11: |
490 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
491 |
|
interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9); |
492 |
|
interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9); |
493 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
494 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
495 |
|
interpolate16x16_lowpass_v(dst, halfpel_v, stride, rounding); |
496 |
|
break; |
497 |
|
|
498 |
|
case 12: |
499 |
|
interpolate16x16_lowpass_v(halfpel_v, src, stride, rounding); |
500 |
|
interpolate8x8_avg2(dst, src+stride, halfpel_v, stride, rounding, 8); |
501 |
|
interpolate8x8_avg2(dst+8, src+stride+8, halfpel_v+8, stride, rounding, 8); |
502 |
|
interpolate8x8_avg2(dst+8*stride, src+stride+8*stride, halfpel_v+8*stride, stride, rounding, 8); |
503 |
|
interpolate8x8_avg2(dst+8*stride+8, src+stride+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); |
504 |
|
break; |
505 |
|
|
506 |
|
case 13: |
507 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
508 |
|
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
509 |
|
interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9); |
510 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
511 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
512 |
|
|
513 |
|
interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
514 |
|
interpolate8x8_avg2(dst, halfpel_hv, halfpel_v+stride, stride, rounding, 8); |
515 |
|
interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+stride+8, stride, rounding, 8); |
516 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+stride+8*stride, stride, rounding, 8); |
517 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+stride+8*stride+8, stride, rounding, 8); |
518 |
|
break; |
519 |
|
|
520 |
|
case 14: |
521 |
|
interpolate16x16_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
522 |
|
interpolate8x8_avg2(dst, halfpel_h+stride, halfpel_hv, stride, rounding, 8); |
523 |
|
interpolate8x8_avg2(dst+8, halfpel_h+stride+8, halfpel_hv+8, stride, rounding, 8); |
524 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_h+stride+8*stride, halfpel_hv+8*stride, stride, rounding, 8); |
525 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_h+stride+8*stride+8, halfpel_hv+8*stride+8, stride, rounding, 8); |
526 |
|
break; |
527 |
|
|
528 |
|
case 15: |
529 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
530 |
|
interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9); |
531 |
|
interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9); |
532 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
533 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
534 |
|
|
535 |
|
interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
536 |
|
interpolate8x8_avg2(dst, halfpel_hv, halfpel_v+stride, stride, rounding, 8); |
537 |
|
interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+stride+8, stride, rounding, 8); |
538 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+stride+8*stride, stride, rounding, 8); |
539 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+stride+8*stride+8, stride, rounding, 8); |
540 |
|
break; |
541 |
|
} |
542 |
|
} |
543 |
|
#endif |