1 |
|
/***************************************************************************** |
2 |
|
* |
3 |
|
* XVID MPEG-4 VIDEO CODEC |
4 |
|
* - Interpolation related header - |
5 |
|
* |
6 |
|
* Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
7 |
|
* |
8 |
|
* This program is free software ; you can redistribute it and/or modify |
9 |
|
* it under the terms of the GNU General Public License as published by |
10 |
|
* the Free Software Foundation ; either version 2 of the License, or |
11 |
|
* (at your option) any later version. |
12 |
|
* |
13 |
|
* This program is distributed in the hope that it will be useful, |
14 |
|
* but WITHOUT ANY WARRANTY ; without even the implied warranty of |
15 |
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
16 |
|
* GNU General Public License for more details. |
17 |
|
* |
18 |
|
* You should have received a copy of the GNU General Public License |
19 |
|
* along with this program ; if not, write to the Free Software |
20 |
|
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
21 |
|
* |
22 |
|
* $Id$ |
23 |
|
* |
24 |
|
****************************************************************************/ |
25 |
|
|
26 |
#ifndef _INTERPOLATE8X8_H_ |
#ifndef _INTERPOLATE8X8_H_ |
27 |
#define _INTERPOLATE8X8_H_ |
#define _INTERPOLATE8X8_H_ |
28 |
|
|
38 |
const uint8_t *src1, |
const uint8_t *src1, |
39 |
const uint8_t *src2, |
const uint8_t *src2, |
40 |
const uint32_t stride, |
const uint32_t stride, |
41 |
const uint32_t rounding); |
const uint32_t rounding, |
42 |
|
const uint32_t height); |
43 |
typedef INTERPOLATE8X8_AVG2 *INTERPOLATE8X8_AVG2_PTR; |
typedef INTERPOLATE8X8_AVG2 *INTERPOLATE8X8_AVG2_PTR; |
44 |
|
|
45 |
typedef void (INTERPOLATE8X8_AVG4) (uint8_t *dst, |
typedef void (INTERPOLATE8X8_AVG4) (uint8_t *dst, |
51 |
const uint32_t rounding); |
const uint32_t rounding); |
52 |
typedef INTERPOLATE8X8_AVG4 *INTERPOLATE8X8_AVG4_PTR; |
typedef INTERPOLATE8X8_AVG4 *INTERPOLATE8X8_AVG4_PTR; |
53 |
|
|
54 |
typedef void (INTERPOLATE8X8_LOWPASS) (uint8_t *dst, |
typedef void (INTERPOLATE_LOWPASS) (uint8_t *dst, |
55 |
uint8_t *src, |
uint8_t *src, |
56 |
int32_t stride, |
int32_t stride, |
57 |
int32_t rounding); |
int32_t rounding); |
58 |
|
|
59 |
typedef INTERPOLATE8X8_LOWPASS *INTERPOLATE8X8_LOWPASS_PTR; |
typedef INTERPOLATE_LOWPASS *INTERPOLATE_LOWPASS_PTR; |
60 |
|
|
61 |
typedef void (INTERPOLATE8X8_LOWPASS_HV) (uint8_t *dst1, |
typedef void (INTERPOLATE_LOWPASS_HV) (uint8_t *dst1, |
62 |
uint8_t *dst2, |
uint8_t *dst2, |
63 |
uint8_t *src, |
uint8_t *src, |
64 |
int32_t stride, |
int32_t stride, |
65 |
int32_t rounding); |
int32_t rounding); |
66 |
|
|
67 |
typedef INTERPOLATE8X8_LOWPASS_HV *INTERPOLATE8X8_LOWPASS_HV_PTR; |
typedef INTERPOLATE_LOWPASS_HV *INTERPOLATE_LOWPASS_HV_PTR; |
68 |
|
|
69 |
typedef void (INTERPOLATE8X8_6TAP_LOWPASS) (uint8_t *dst, |
typedef void (INTERPOLATE8X8_6TAP_LOWPASS) (uint8_t *dst, |
70 |
uint8_t *src, |
uint8_t *src, |
80 |
extern INTERPOLATE8X8_AVG2_PTR interpolate8x8_avg2; |
extern INTERPOLATE8X8_AVG2_PTR interpolate8x8_avg2; |
81 |
extern INTERPOLATE8X8_AVG4_PTR interpolate8x8_avg4; |
extern INTERPOLATE8X8_AVG4_PTR interpolate8x8_avg4; |
82 |
|
|
83 |
extern INTERPOLATE8X8_LOWPASS_PTR interpolate8x8_lowpass_h; |
extern INTERPOLATE_LOWPASS_PTR interpolate8x8_lowpass_h; |
84 |
extern INTERPOLATE8X8_LOWPASS_PTR interpolate8x8_lowpass_v; |
extern INTERPOLATE_LOWPASS_PTR interpolate8x8_lowpass_v; |
85 |
|
|
86 |
|
extern INTERPOLATE_LOWPASS_PTR interpolate16x16_lowpass_h; |
87 |
|
extern INTERPOLATE_LOWPASS_PTR interpolate16x16_lowpass_v; |
88 |
|
|
89 |
extern INTERPOLATE8X8_LOWPASS_HV_PTR interpolate8x8_lowpass_hv; |
extern INTERPOLATE_LOWPASS_HV_PTR interpolate8x8_lowpass_hv; |
90 |
|
extern INTERPOLATE_LOWPASS_HV_PTR interpolate16x16_lowpass_hv; |
91 |
|
|
92 |
extern INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_h; |
extern INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_h; |
93 |
extern INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_v; |
extern INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_v; |
96 |
INTERPOLATE8X8 interpolate8x8_halfpel_v_c; |
INTERPOLATE8X8 interpolate8x8_halfpel_v_c; |
97 |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_c; |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_c; |
98 |
|
|
99 |
|
#ifdef ARCH_IS_IA32 |
100 |
INTERPOLATE8X8 interpolate8x8_halfpel_h_mmx; |
INTERPOLATE8X8 interpolate8x8_halfpel_h_mmx; |
101 |
INTERPOLATE8X8 interpolate8x8_halfpel_v_mmx; |
INTERPOLATE8X8 interpolate8x8_halfpel_v_mmx; |
102 |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_mmx; |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_mmx; |
109 |
INTERPOLATE8X8 interpolate8x8_halfpel_v_3dn; |
INTERPOLATE8X8 interpolate8x8_halfpel_v_3dn; |
110 |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dn; |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dn; |
111 |
|
|
112 |
|
INTERPOLATE8X8 interpolate8x8_halfpel_h_3dne; |
113 |
|
INTERPOLATE8X8 interpolate8x8_halfpel_v_3dne; |
114 |
|
INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dne; |
115 |
|
#endif |
116 |
|
|
117 |
|
#ifdef ARCH_IS_IA64 |
118 |
INTERPOLATE8X8 interpolate8x8_halfpel_h_ia64; |
INTERPOLATE8X8 interpolate8x8_halfpel_h_ia64; |
119 |
INTERPOLATE8X8 interpolate8x8_halfpel_v_ia64; |
INTERPOLATE8X8 interpolate8x8_halfpel_v_ia64; |
120 |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_ia64; |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_ia64; |
121 |
|
#endif |
122 |
|
|
123 |
INTERPOLATE8X8_AVG2 interpolate8x8_avg2_c; |
INTERPOLATE8X8_AVG2 interpolate8x8_avg2_c; |
124 |
INTERPOLATE8X8_AVG4 interpolate8x8_avg4_c; |
INTERPOLATE8X8_AVG4 interpolate8x8_avg4_c; |
125 |
|
|
126 |
|
#ifdef ARCH_IS_IA32 |
127 |
INTERPOLATE8X8_AVG2 interpolate8x8_avg2_mmx; |
INTERPOLATE8X8_AVG2 interpolate8x8_avg2_mmx; |
128 |
INTERPOLATE8X8_AVG4 interpolate8x8_avg4_mmx; |
INTERPOLATE8X8_AVG4 interpolate8x8_avg4_mmx; |
129 |
|
#endif |
130 |
|
|
131 |
|
INTERPOLATE_LOWPASS interpolate8x8_lowpass_h_c; |
132 |
|
INTERPOLATE_LOWPASS interpolate8x8_lowpass_v_c; |
133 |
|
|
134 |
INTERPOLATE8X8_LOWPASS interpolate8x8_lowpass_h_c; |
INTERPOLATE_LOWPASS interpolate16x16_lowpass_h_c; |
135 |
INTERPOLATE8X8_LOWPASS interpolate8x8_lowpass_v_c; |
INTERPOLATE_LOWPASS interpolate16x16_lowpass_v_c; |
136 |
INTERPOLATE8X8_LOWPASS_HV interpolate8x8_lowpass_hv_c; |
|
137 |
|
INTERPOLATE_LOWPASS_HV interpolate8x8_lowpass_hv_c; |
138 |
|
INTERPOLATE_LOWPASS_HV interpolate16x16_lowpass_hv_c; |
139 |
|
|
140 |
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_h_c; |
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_h_c; |
141 |
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_v_c; |
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_v_c; |
142 |
|
|
143 |
|
#ifdef ARCH_IS_IA32 |
144 |
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_h_mmx; |
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_h_mmx; |
145 |
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_v_mmx; |
INTERPOLATE8X8_6TAP_LOWPASS interpolate8x8_6tap_lowpass_v_mmx; |
146 |
|
#endif |
147 |
|
|
148 |
static __inline void |
static __inline void |
149 |
interpolate8x8_switch(uint8_t * const cur, |
interpolate8x8_switch(uint8_t * const cur, |
155 |
const uint32_t stride, |
const uint32_t stride, |
156 |
const uint32_t rounding) |
const uint32_t rounding) |
157 |
{ |
{ |
|
int32_t ddx, ddy; |
|
158 |
|
|
159 |
switch (((dx & 1) << 1) + (dy & 1)) // ((dx%2)?2:0)+((dy%2)?1:0) |
const uint8_t * const src = refn + (int)((y + (dy>>1)) * stride + x + (dx>>1)); |
160 |
{ |
uint8_t * const dst = cur + (int)(y * stride + x); |
161 |
|
|
162 |
|
switch (((dx & 1) << 1) + (dy & 1)) { /* ((dx%2)?2:0)+((dy%2)?1:0) */ |
163 |
case 0: |
case 0: |
164 |
ddx = dx / 2; |
transfer8x8_copy(dst, src, stride); |
|
ddy = dy / 2; |
|
|
transfer8x8_copy(cur + y * stride + x, |
|
|
refn + (int)((y + ddy) * stride + x + ddx), stride); |
|
165 |
break; |
break; |
|
|
|
166 |
case 1: |
case 1: |
167 |
ddx = dx / 2; |
interpolate8x8_halfpel_v(dst, src, stride, rounding); |
|
ddy = (dy - 1) / 2; |
|
|
interpolate8x8_halfpel_v(cur + y * stride + x, |
|
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
|
|
rounding); |
|
168 |
break; |
break; |
|
|
|
169 |
case 2: |
case 2: |
170 |
ddx = (dx - 1) / 2; |
interpolate8x8_halfpel_h(dst, src, stride, rounding); |
|
ddy = dy / 2; |
|
|
interpolate8x8_halfpel_h(cur + y * stride + x, |
|
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
|
|
rounding); |
|
171 |
break; |
break; |
|
|
|
172 |
default: |
default: |
173 |
ddx = (dx - 1) / 2; |
interpolate8x8_halfpel_hv(dst, src, stride, rounding); |
|
ddy = (dy - 1) / 2; |
|
|
interpolate8x8_halfpel_hv(cur + y * stride + x, |
|
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
|
|
rounding); |
|
174 |
break; |
break; |
175 |
} |
} |
176 |
} |
} |
177 |
|
|
178 |
static __inline uint8_t * |
|
179 |
interpolate8x8_switch2(uint8_t * const buffer, |
static __inline void |
180 |
|
interpolate16x16_switch(uint8_t * const cur, |
181 |
const uint8_t * const refn, |
const uint8_t * const refn, |
182 |
const uint32_t x, |
const uint32_t x, |
183 |
const uint32_t y, |
const uint32_t y, |
186 |
const uint32_t stride, |
const uint32_t stride, |
187 |
const uint32_t rounding) |
const uint32_t rounding) |
188 |
{ |
{ |
189 |
int32_t ddx, ddy; |
interpolate8x8_switch(cur, refn, x, y, dx, dy, stride, rounding); |
190 |
|
interpolate8x8_switch(cur, refn, x+8, y, dx, dy, stride, rounding); |
191 |
|
interpolate8x8_switch(cur, refn, x, y+8, dx, dy, stride, rounding); |
192 |
|
interpolate8x8_switch(cur, refn, x+8, y+8, dx, dy, stride, rounding); |
193 |
|
} |
194 |
|
|
195 |
|
|
196 |
switch (((dx & 1) << 1) + (dy & 1)) // ((dx%2)?2:0)+((dy%2)?1:0) |
static __inline void |
197 |
|
interpolate32x32_switch(uint8_t * const cur, |
198 |
|
const uint8_t * const refn, |
199 |
|
const uint32_t x, |
200 |
|
const uint32_t y, |
201 |
|
const int32_t dx, |
202 |
|
const int dy, |
203 |
|
const uint32_t stride, |
204 |
|
const uint32_t rounding) |
205 |
{ |
{ |
206 |
case 0: |
interpolate16x16_switch(cur, refn, x, y, dx, dy, stride, rounding); |
207 |
return (uint8_t *)refn + (int)((y + dy/2) * stride + x + dx/2); |
interpolate16x16_switch(cur, refn, x+16, y, dx, dy, stride, rounding); |
208 |
|
interpolate16x16_switch(cur, refn, x, y+16, dx, dy, stride, rounding); |
209 |
|
interpolate16x16_switch(cur, refn, x+16, y+16, dx, dy, stride, rounding); |
210 |
|
} |
211 |
|
|
212 |
|
|
213 |
|
static __inline uint8_t * |
214 |
|
interpolate8x8_switch2(uint8_t * const buffer, |
215 |
|
const uint8_t * const refn, |
216 |
|
const int x, |
217 |
|
const int y, |
218 |
|
const int dx, |
219 |
|
const int dy, |
220 |
|
const uint32_t stride, |
221 |
|
const uint32_t rounding) |
222 |
|
{ |
223 |
|
|
224 |
|
const uint8_t * const src = refn + (int)((y + (dy>>1)) * stride + x + (dx>>1)); |
225 |
|
|
226 |
|
switch (((dx & 1) << 1) + (dy & 1)) { /* ((dx%2)?2:0)+((dy%2)?1:0) */ |
227 |
|
case 0: |
228 |
|
return (uint8_t *)src; |
229 |
case 1: |
case 1: |
230 |
ddx = dx / 2; |
interpolate8x8_halfpel_v(buffer, src, stride, rounding); |
|
ddy = (dy - 1) / 2; |
|
|
interpolate8x8_halfpel_v(buffer, |
|
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
|
|
rounding); |
|
231 |
break; |
break; |
|
|
|
232 |
case 2: |
case 2: |
233 |
ddx = (dx - 1) / 2; |
interpolate8x8_halfpel_h(buffer, src, stride, rounding); |
|
ddy = dy / 2; |
|
|
interpolate8x8_halfpel_h(buffer, |
|
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
|
|
rounding); |
|
234 |
break; |
break; |
|
|
|
235 |
default: |
default: |
236 |
ddx = (dx - 1) / 2; |
interpolate8x8_halfpel_hv(buffer, src, stride, rounding); |
|
ddy = (dy - 1) / 2; |
|
|
interpolate8x8_halfpel_hv(buffer, |
|
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
|
|
rounding); |
|
237 |
break; |
break; |
238 |
} |
} |
239 |
return buffer; |
return buffer; |
240 |
} |
} |
241 |
|
|
|
|
|
242 |
static __inline void interpolate8x8_quarterpel(uint8_t * const cur, |
static __inline void interpolate8x8_quarterpel(uint8_t * const cur, |
243 |
uint8_t * const refn, |
uint8_t * const refn, |
244 |
uint8_t * const refh, |
uint8_t * const refh, |
283 |
|
|
284 |
case 1: |
case 1: |
285 |
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
286 |
interpolate8x8_avg2(dst, src, halfpel_h, stride, rounding); |
interpolate8x8_avg2(dst, src, halfpel_h, stride, rounding, 8); |
287 |
break; |
break; |
288 |
|
|
289 |
case 2: |
case 2: |
292 |
|
|
293 |
case 3: |
case 3: |
294 |
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
295 |
interpolate8x8_avg2(dst, src + 1, halfpel_h, stride, rounding); |
interpolate8x8_avg2(dst, src + 1, halfpel_h, stride, rounding, 8); |
296 |
break; |
break; |
297 |
|
|
298 |
case 4: |
case 4: |
299 |
interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); |
interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); |
300 |
interpolate8x8_avg2(dst, src, halfpel_v, stride, rounding); |
interpolate8x8_avg2(dst, src, halfpel_v, stride, rounding, 8); |
301 |
break; |
break; |
302 |
|
|
303 |
case 5: |
case 5: |
304 |
interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); |
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
305 |
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
306 |
interpolate8x8_avg4(dst, src, halfpel_h, halfpel_v, halfpel_hv, stride, rounding); |
interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
307 |
|
interpolate8x8_avg2(dst, halfpel_v, halfpel_hv, stride, rounding, 8); |
308 |
break; |
break; |
309 |
|
|
310 |
case 6: |
case 6: |
311 |
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
312 |
interpolate8x8_avg2(dst, halfpel_h, halfpel_hv, stride, rounding); |
interpolate8x8_avg2(dst, halfpel_h, halfpel_hv, stride, rounding, 8); |
313 |
break; |
break; |
314 |
|
|
315 |
case 7: |
case 7: |
316 |
interpolate8x8_lowpass_v(halfpel_v, src+1, stride, rounding); |
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
317 |
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9); |
318 |
interpolate8x8_avg4(dst, src+1, halfpel_h, halfpel_v, halfpel_hv, stride, rounding); |
interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
319 |
|
interpolate8x8_avg2(dst, halfpel_v, halfpel_hv, stride, rounding, 8); |
320 |
break; |
break; |
321 |
|
|
322 |
case 8: |
case 8: |
324 |
break; |
break; |
325 |
|
|
326 |
case 9: |
case 9: |
327 |
interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); |
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
328 |
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
329 |
interpolate8x8_avg2(dst, halfpel_v, halfpel_hv, stride, rounding); |
interpolate8x8_lowpass_v(dst, halfpel_v, stride, rounding); |
330 |
break; |
break; |
331 |
|
|
332 |
case 10: |
case 10: |
334 |
break; |
break; |
335 |
|
|
336 |
case 11: |
case 11: |
337 |
interpolate8x8_lowpass_v(halfpel_v, src+1, stride, rounding); |
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
338 |
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9); |
339 |
interpolate8x8_avg2(dst, halfpel_v, halfpel_hv, stride, rounding); |
interpolate8x8_lowpass_v(dst, halfpel_v, stride, rounding); |
340 |
break; |
break; |
341 |
|
|
342 |
case 12: |
case 12: |
343 |
interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); |
interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); |
344 |
interpolate8x8_avg2(dst, src+stride, halfpel_v, stride, rounding); |
interpolate8x8_avg2(dst, src+stride, halfpel_v, stride, rounding, 8); |
345 |
break; |
break; |
346 |
|
|
347 |
case 13: |
case 13: |
348 |
interpolate8x8_lowpass_v(halfpel_v, src, stride, rounding); |
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
349 |
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
350 |
interpolate8x8_avg4(dst, src+stride, halfpel_h+stride, halfpel_v, halfpel_hv, stride, rounding); |
interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
351 |
|
interpolate8x8_avg2(dst, halfpel_v+stride, halfpel_hv, stride, rounding, 8); |
352 |
break; |
break; |
353 |
|
|
354 |
case 14: |
case 14: |
355 |
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
356 |
interpolate8x8_avg2(dst, halfpel_h+stride, halfpel_hv, stride, rounding); |
interpolate8x8_avg2(dst, halfpel_h+stride, halfpel_hv, stride, rounding, 8); |
357 |
break; |
break; |
358 |
|
|
359 |
case 15: |
case 15: |
360 |
interpolate8x8_lowpass_v(halfpel_v, src+1, stride, rounding); |
interpolate8x8_lowpass_h(halfpel_h, src, stride, rounding); |
361 |
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
interpolate8x8_avg2(halfpel_v, src + 1, halfpel_h, stride, rounding, 9); |
362 |
interpolate8x8_avg4(dst, src+stride+1, halfpel_h+stride, halfpel_v, halfpel_hv, stride, rounding); |
interpolate8x8_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
363 |
|
interpolate8x8_avg2(dst, halfpel_hv, halfpel_v + stride, stride, rounding, 8); |
364 |
|
break; |
365 |
|
} |
366 |
|
} |
367 |
|
|
368 |
|
static __inline void interpolate16x16_quarterpel(uint8_t * const cur, |
369 |
|
uint8_t * const refn, |
370 |
|
uint8_t * const refh, |
371 |
|
uint8_t * const refv, |
372 |
|
uint8_t * const refhv, |
373 |
|
const uint32_t x, const uint32_t y, |
374 |
|
const int32_t dx, const int dy, |
375 |
|
const uint32_t stride, |
376 |
|
const uint32_t rounding) |
377 |
|
{ |
378 |
|
const int32_t xRef = x*4 + dx; |
379 |
|
const int32_t yRef = y*4 + dy; |
380 |
|
|
381 |
|
uint8_t *src, *dst; |
382 |
|
uint8_t *halfpel_h, *halfpel_v, *halfpel_hv; |
383 |
|
int32_t x_int, y_int, x_frac, y_frac; |
384 |
|
|
385 |
|
x_int = xRef/4; |
386 |
|
if (xRef < 0 && xRef % 4) |
387 |
|
x_int--; |
388 |
|
|
389 |
|
x_frac = xRef - (4*x_int); |
390 |
|
|
391 |
|
y_int = yRef/4; |
392 |
|
if (yRef < 0 && yRef % 4) |
393 |
|
y_int--; |
394 |
|
|
395 |
|
y_frac = yRef - (4*y_int); |
396 |
|
|
397 |
|
src = refn + y_int * stride + x_int; |
398 |
|
halfpel_h = refh; |
399 |
|
halfpel_v = refv; |
400 |
|
halfpel_hv = refhv; |
401 |
|
|
402 |
|
dst = cur + y * stride + x; |
403 |
|
|
404 |
|
switch((y_frac << 2) | (x_frac)) { |
405 |
|
|
406 |
|
case 0: |
407 |
|
transfer16x16_copy(dst, src, stride); |
408 |
|
break; |
409 |
|
|
410 |
|
case 1: |
411 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
412 |
|
interpolate8x8_avg2(dst, src, halfpel_h, stride, rounding, 8); |
413 |
|
interpolate8x8_avg2(dst+8, src+8, halfpel_h+8, stride, rounding, 8); |
414 |
|
interpolate8x8_avg2(dst+8*stride, src+8*stride, halfpel_h+8*stride, stride, rounding, 8); |
415 |
|
interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 8); |
416 |
|
break; |
417 |
|
|
418 |
|
case 2: |
419 |
|
interpolate16x16_lowpass_h(dst, src, stride, rounding); |
420 |
|
break; |
421 |
|
|
422 |
|
case 3: |
423 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
424 |
|
interpolate8x8_avg2(dst, src + 1, halfpel_h, stride, rounding, 8); |
425 |
|
interpolate8x8_avg2(dst+8, src + 8 + 1, halfpel_h+8, stride, rounding, 8); |
426 |
|
interpolate8x8_avg2(dst+8*stride, src + 8*stride + 1, halfpel_h+8*stride, stride, rounding, 8); |
427 |
|
interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8 + 1, halfpel_h+8*stride+8, stride, rounding, 8); |
428 |
|
break; |
429 |
|
|
430 |
|
case 4: |
431 |
|
interpolate16x16_lowpass_v(halfpel_v, src, stride, rounding); |
432 |
|
interpolate8x8_avg2(dst, src, halfpel_v, stride, rounding, 8); |
433 |
|
interpolate8x8_avg2(dst+8, src+8, halfpel_v+8, stride, rounding, 8); |
434 |
|
interpolate8x8_avg2(dst+8*stride, src+8*stride, halfpel_v+8*stride, stride, rounding, 8); |
435 |
|
interpolate8x8_avg2(dst+8*stride+8, src+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); |
436 |
|
break; |
437 |
|
|
438 |
|
case 5: |
439 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
440 |
|
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
441 |
|
interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9); |
442 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
443 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
444 |
|
|
445 |
|
interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
446 |
|
interpolate8x8_avg2(dst, halfpel_hv, halfpel_v, stride, rounding, 8); |
447 |
|
interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+8, stride, rounding, 8); |
448 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+8*stride, stride, rounding, 8); |
449 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); |
450 |
|
break; |
451 |
|
|
452 |
|
case 6: |
453 |
|
interpolate16x16_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
454 |
|
interpolate8x8_avg2(dst, halfpel_h, halfpel_hv, stride, rounding, 8); |
455 |
|
interpolate8x8_avg2(dst+8, halfpel_h+8, halfpel_hv+8, stride, rounding, 8); |
456 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_h+8*stride, halfpel_hv+8*stride, stride, rounding, 8); |
457 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_h+8*stride+8, halfpel_hv+8*stride+8, stride, rounding, 8); |
458 |
|
break; |
459 |
|
|
460 |
|
case 7: |
461 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
462 |
|
interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9); |
463 |
|
interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9); |
464 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
465 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
466 |
|
|
467 |
|
interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
468 |
|
interpolate8x8_avg2(dst, halfpel_hv, halfpel_v, stride, rounding, 8); |
469 |
|
interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+8, stride, rounding, 8); |
470 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+8*stride, stride, rounding, 8); |
471 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); |
472 |
|
break; |
473 |
|
|
474 |
|
case 8: |
475 |
|
interpolate16x16_lowpass_v(dst, src, stride, rounding); |
476 |
|
break; |
477 |
|
|
478 |
|
case 9: |
479 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
480 |
|
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
481 |
|
interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9); |
482 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
483 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
484 |
|
interpolate16x16_lowpass_v(dst, halfpel_v, stride, rounding); |
485 |
|
break; |
486 |
|
|
487 |
|
case 10: |
488 |
|
interpolate16x16_lowpass_hv(dst, halfpel_h, src, stride, rounding); |
489 |
|
break; |
490 |
|
|
491 |
|
case 11: |
492 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
493 |
|
interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9); |
494 |
|
interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9); |
495 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
496 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
497 |
|
interpolate16x16_lowpass_v(dst, halfpel_v, stride, rounding); |
498 |
|
break; |
499 |
|
|
500 |
|
case 12: |
501 |
|
interpolate16x16_lowpass_v(halfpel_v, src, stride, rounding); |
502 |
|
interpolate8x8_avg2(dst, src+stride, halfpel_v, stride, rounding, 8); |
503 |
|
interpolate8x8_avg2(dst+8, src+stride+8, halfpel_v+8, stride, rounding, 8); |
504 |
|
interpolate8x8_avg2(dst+8*stride, src+stride+8*stride, halfpel_v+8*stride, stride, rounding, 8); |
505 |
|
interpolate8x8_avg2(dst+8*stride+8, src+stride+8*stride+8, halfpel_v+8*stride+8, stride, rounding, 8); |
506 |
|
break; |
507 |
|
|
508 |
|
case 13: |
509 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
510 |
|
interpolate8x8_avg2(halfpel_v, src, halfpel_h, stride, rounding, 9); |
511 |
|
interpolate8x8_avg2(halfpel_v+8, src + 8, halfpel_h+8, stride, rounding, 9); |
512 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
513 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
514 |
|
|
515 |
|
interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
516 |
|
interpolate8x8_avg2(dst, halfpel_hv, halfpel_v+stride, stride, rounding, 8); |
517 |
|
interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+stride+8, stride, rounding, 8); |
518 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+stride+8*stride, stride, rounding, 8); |
519 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+stride+8*stride+8, stride, rounding, 8); |
520 |
|
break; |
521 |
|
|
522 |
|
case 14: |
523 |
|
interpolate16x16_lowpass_hv(halfpel_hv, halfpel_h, src, stride, rounding); |
524 |
|
interpolate8x8_avg2(dst, halfpel_h+stride, halfpel_hv, stride, rounding, 8); |
525 |
|
interpolate8x8_avg2(dst+8, halfpel_h+stride+8, halfpel_hv+8, stride, rounding, 8); |
526 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_h+stride+8*stride, halfpel_hv+8*stride, stride, rounding, 8); |
527 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_h+stride+8*stride+8, halfpel_hv+8*stride+8, stride, rounding, 8); |
528 |
|
break; |
529 |
|
|
530 |
|
case 15: |
531 |
|
interpolate16x16_lowpass_h(halfpel_h, src, stride, rounding); |
532 |
|
interpolate8x8_avg2(halfpel_v, src+1, halfpel_h, stride, rounding, 9); |
533 |
|
interpolate8x8_avg2(halfpel_v+8, src+1 + 8, halfpel_h+8, stride, rounding, 9); |
534 |
|
interpolate8x8_avg2(halfpel_v+8*stride, src+1 + 8*stride, halfpel_h+8*stride, stride, rounding, 9); |
535 |
|
interpolate8x8_avg2(halfpel_v+8*stride+8, src+1+8*stride+8, halfpel_h+8*stride+8, stride, rounding, 9); |
536 |
|
|
537 |
|
interpolate16x16_lowpass_v(halfpel_hv, halfpel_v, stride, rounding); |
538 |
|
interpolate8x8_avg2(dst, halfpel_hv, halfpel_v+stride, stride, rounding, 8); |
539 |
|
interpolate8x8_avg2(dst+8, halfpel_hv+8, halfpel_v+stride+8, stride, rounding, 8); |
540 |
|
interpolate8x8_avg2(dst+8*stride, halfpel_hv+8*stride, halfpel_v+stride+8*stride, stride, rounding, 8); |
541 |
|
interpolate8x8_avg2(dst+8*stride+8, halfpel_hv+8*stride+8, halfpel_v+stride+8*stride+8, stride, rounding, 8); |
542 |
break; |
break; |
543 |
} |
} |
544 |
} |
} |