1 |
/************************************************************************** |
/***************************************************************************** |
2 |
* |
* |
3 |
* XVID MPEG-4 VIDEO CODEC |
* XVID MPEG-4 VIDEO CODEC |
4 |
* 8x8 block-based halfpel interpolation |
* - 8x8 block-based halfpel interpolation - |
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 |
* 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 |
* it under the terms of the GNU General Public License as published by |
17 |
* |
* |
18 |
* You should have received a copy of the GNU General Public License |
* You should have received a copy of the GNU General Public License |
19 |
* along with this program; if not, write to the Free Software |
* along with this program; if not, write to the Free Software |
20 |
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
|
* |
|
|
*************************************************************************/ |
|
|
|
|
|
/************************************************************************** |
|
|
* |
|
|
* History: |
|
21 |
* |
* |
22 |
* 05.10.2002 new bilinear and qpel interpolation code - Isibaar |
* $Id$ |
|
* 27.12.2001 modified "compensate_halfpel" |
|
|
* 05.11.2001 initial version; (c)2001 peter ross <pross@cs.rmit.edu.au> |
|
23 |
* |
* |
24 |
*************************************************************************/ |
****************************************************************************/ |
|
|
|
25 |
|
|
26 |
#include "../portab.h" |
#include "../portab.h" |
27 |
#include "../global.h" |
#include "../global.h" |
28 |
#include "interpolate8x8.h" |
#include "interpolate8x8.h" |
29 |
|
|
30 |
// function pointers |
/* function pointers */ |
31 |
INTERPOLATE8X8_PTR interpolate8x8_halfpel_h; |
INTERPOLATE8X8_PTR interpolate8x8_halfpel_h; |
32 |
INTERPOLATE8X8_PTR interpolate8x8_halfpel_v; |
INTERPOLATE8X8_PTR interpolate8x8_halfpel_v; |
33 |
INTERPOLATE8X8_PTR interpolate8x8_halfpel_hv; |
INTERPOLATE8X8_PTR interpolate8x8_halfpel_hv; |
34 |
|
|
35 |
|
INTERPOLATE8X8_PTR interpolate8x8_halfpel_add; |
36 |
|
INTERPOLATE8X8_PTR interpolate8x8_halfpel_h_add; |
37 |
|
INTERPOLATE8X8_PTR interpolate8x8_halfpel_v_add; |
38 |
|
INTERPOLATE8X8_PTR interpolate8x8_halfpel_hv_add; |
39 |
|
|
40 |
|
|
41 |
INTERPOLATE8X8_AVG2_PTR interpolate8x8_avg2; |
INTERPOLATE8X8_AVG2_PTR interpolate8x8_avg2; |
42 |
INTERPOLATE8X8_AVG4_PTR interpolate8x8_avg4; |
INTERPOLATE8X8_AVG4_PTR interpolate8x8_avg4; |
43 |
|
|
53 |
INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_h; |
INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_h; |
54 |
INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_v; |
INTERPOLATE8X8_6TAP_LOWPASS_PTR interpolate8x8_6tap_lowpass_v; |
55 |
|
|
56 |
void interpolate8x8_avg2_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint32_t stride, const uint32_t rounding, const uint32_t height) |
void __inline |
57 |
|
interpolate8x8_avg2_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint32_t stride, const uint32_t rounding, const uint32_t height) |
58 |
{ |
{ |
59 |
uint32_t i; |
uint32_t i; |
60 |
const int32_t round = 1 - rounding; |
const int32_t round = 1 - rounding; |
61 |
|
|
62 |
for(i = 0; i < height; i++) |
for(i = 0; i < height; i++) { |
|
{ |
|
63 |
dst[0] = (src1[0] + src2[0] + round) >> 1; |
dst[0] = (src1[0] + src2[0] + round) >> 1; |
64 |
dst[1] = (src1[1] + src2[1] + round) >> 1; |
dst[1] = (src1[1] + src2[1] + round) >> 1; |
65 |
dst[2] = (src1[2] + src2[2] + round) >> 1; |
dst[2] = (src1[2] + src2[2] + round) >> 1; |
75 |
} |
} |
76 |
} |
} |
77 |
|
|
78 |
|
void |
79 |
|
interpolate8x8_halfpel_add_c(uint8_t *dst, const uint8_t *src, const uint32_t stride, const uint32_t rounding) |
80 |
|
{ |
81 |
|
interpolate8x8_avg2_c(dst, dst, src, stride, 0, 8); |
82 |
|
} |
83 |
|
|
84 |
void interpolate8x8_avg4_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4, const uint32_t stride, const uint32_t rounding) |
void interpolate8x8_avg4_c(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, const uint8_t *src3, const uint8_t *src4, const uint32_t stride, const uint32_t rounding) |
85 |
{ |
{ |
86 |
int32_t i; |
int32_t i; |
87 |
const int32_t round = 2 - rounding; |
const int32_t round = 2 - rounding; |
88 |
|
|
89 |
for(i = 0; i < 8; i++) |
for(i = 0; i < 8; i++) { |
|
{ |
|
90 |
dst[0] = (src1[0] + src2[0] + src3[0] + src4[0] + round) >> 2; |
dst[0] = (src1[0] + src2[0] + src3[0] + src4[0] + round) >> 2; |
91 |
dst[1] = (src1[1] + src2[1] + src3[1] + src4[1] + round) >> 2; |
dst[1] = (src1[1] + src2[1] + src3[1] + src4[1] + round) >> 2; |
92 |
dst[2] = (src1[2] + src2[2] + src3[2] + src4[2] + round) >> 2; |
dst[2] = (src1[2] + src2[2] + src3[2] + src4[2] + round) >> 2; |
104 |
} |
} |
105 |
} |
} |
106 |
|
|
107 |
// dst = interpolate(src) |
/* dst = interpolate(src) */ |
108 |
|
|
109 |
void |
void |
110 |
interpolate8x8_halfpel_h_c(uint8_t * const dst, |
interpolate8x8_halfpel_h_c(uint8_t * const dst, |
112 |
const uint32_t stride, |
const uint32_t stride, |
113 |
const uint32_t rounding) |
const uint32_t rounding) |
114 |
{ |
{ |
115 |
intptr_t j; |
uintptr_t j; |
116 |
|
|
117 |
if (rounding) |
if (rounding) { |
118 |
for (j = 7*stride; j >= 0; j-=stride) |
for (j = 0; j < 8*stride; j+=stride) { |
|
{ |
|
119 |
dst[j + 0] = (uint8_t)((src[j + 0] + src[j + 1] )>>1); |
dst[j + 0] = (uint8_t)((src[j + 0] + src[j + 1] )>>1); |
120 |
dst[j + 1] = (uint8_t)((src[j + 1] + src[j + 2] )>>1); |
dst[j + 1] = (uint8_t)((src[j + 1] + src[j + 2] )>>1); |
121 |
dst[j + 2] = (uint8_t)((src[j + 2] + src[j + 3] )>>1); |
dst[j + 2] = (uint8_t)((src[j + 2] + src[j + 3] )>>1); |
125 |
dst[j + 6] = (uint8_t)((src[j + 6] + src[j + 7] )>>1); |
dst[j + 6] = (uint8_t)((src[j + 6] + src[j + 7] )>>1); |
126 |
dst[j + 7] = (uint8_t)((src[j + 7] + src[j + 8] )>>1); |
dst[j + 7] = (uint8_t)((src[j + 7] + src[j + 8] )>>1); |
127 |
} |
} |
128 |
else |
} else { |
129 |
for (j = 0; j < 8*stride; j+=stride) /* forward or backwards? Who knows ... */ |
for (j = 0; j < 8*stride; j+=stride) { |
|
{ |
|
130 |
dst[j + 0] = (uint8_t)((src[j + 0] + src[j + 1] + 1)>>1); |
dst[j + 0] = (uint8_t)((src[j + 0] + src[j + 1] + 1)>>1); |
131 |
dst[j + 1] = (uint8_t)((src[j + 1] + src[j + 2] + 1)>>1); |
dst[j + 1] = (uint8_t)((src[j + 1] + src[j + 2] + 1)>>1); |
132 |
dst[j + 2] = (uint8_t)((src[j + 2] + src[j + 3] + 1)>>1); |
dst[j + 2] = (uint8_t)((src[j + 2] + src[j + 3] + 1)>>1); |
137 |
dst[j + 7] = (uint8_t)((src[j + 7] + src[j + 8] + 1)>>1); |
dst[j + 7] = (uint8_t)((src[j + 7] + src[j + 8] + 1)>>1); |
138 |
} |
} |
139 |
} |
} |
140 |
|
} |
141 |
|
|
142 |
|
/* dst = (dst + interpolate(src)/2 */ |
143 |
|
|
144 |
|
void |
145 |
|
interpolate8x8_halfpel_h_add_c(uint8_t * const dst, |
146 |
|
const uint8_t * const src, |
147 |
|
const uint32_t stride, |
148 |
|
const uint32_t rounding) |
149 |
|
{ |
150 |
|
uintptr_t j; |
151 |
|
|
152 |
|
if (rounding) { |
153 |
|
for (j = 0; j < 8*stride; j+=stride) { |
154 |
|
dst[j + 0] = (uint8_t)((((src[j + 0] + src[j + 1] )>>1) + dst[j+0] + 1)>>1); |
155 |
|
dst[j + 1] = (uint8_t)((((src[j + 1] + src[j + 2] )>>1) + dst[j+1] + 1)>>1); |
156 |
|
dst[j + 2] = (uint8_t)((((src[j + 2] + src[j + 3] )>>1) + dst[j+2] + 1)>>1); |
157 |
|
dst[j + 3] = (uint8_t)((((src[j + 3] + src[j + 4] )>>1) + dst[j+3] + 1)>>1); |
158 |
|
dst[j + 4] = (uint8_t)((((src[j + 4] + src[j + 5] )>>1) + dst[j+4] + 1)>>1); |
159 |
|
dst[j + 5] = (uint8_t)((((src[j + 5] + src[j + 6] )>>1) + dst[j+5] + 1)>>1); |
160 |
|
dst[j + 6] = (uint8_t)((((src[j + 6] + src[j + 7] )>>1) + dst[j+6] + 1)>>1); |
161 |
|
dst[j + 7] = (uint8_t)((((src[j + 7] + src[j + 8] )>>1) + dst[j+7] + 1)>>1); |
162 |
|
} |
163 |
|
} else { |
164 |
|
for (j = 0; j < 8*stride; j+=stride) { |
165 |
|
dst[j + 0] = (uint8_t)((((src[j + 0] + src[j + 1] + 1)>>1) + dst[j+0] + 1)>>1); |
166 |
|
dst[j + 1] = (uint8_t)((((src[j + 1] + src[j + 2] + 1)>>1) + dst[j+1] + 1)>>1); |
167 |
|
dst[j + 2] = (uint8_t)((((src[j + 2] + src[j + 3] + 1)>>1) + dst[j+2] + 1)>>1); |
168 |
|
dst[j + 3] = (uint8_t)((((src[j + 3] + src[j + 4] + 1)>>1) + dst[j+3] + 1)>>1); |
169 |
|
dst[j + 4] = (uint8_t)((((src[j + 4] + src[j + 5] + 1)>>1) + dst[j+4] + 1)>>1); |
170 |
|
dst[j + 5] = (uint8_t)((((src[j + 5] + src[j + 6] + 1)>>1) + dst[j+5] + 1)>>1); |
171 |
|
dst[j + 6] = (uint8_t)((((src[j + 6] + src[j + 7] + 1)>>1) + dst[j+6] + 1)>>1); |
172 |
|
dst[j + 7] = (uint8_t)((((src[j + 7] + src[j + 8] + 1)>>1) + dst[j+7] + 1)>>1); |
173 |
|
} |
174 |
|
} |
175 |
|
} |
176 |
|
|
177 |
|
/* dst = interpolate(src) */ |
178 |
|
|
179 |
void |
void |
180 |
interpolate8x8_halfpel_v_c(uint8_t * const dst, |
interpolate8x8_halfpel_v_c(uint8_t * const dst, |
182 |
const uint32_t stride, |
const uint32_t stride, |
183 |
const uint32_t rounding) |
const uint32_t rounding) |
184 |
{ |
{ |
185 |
intptr_t j; |
uintptr_t j; |
|
// const uint8_t * const src2 = src+stride; /* using a second pointer is _not_ faster here */ |
|
186 |
|
|
187 |
if (rounding) |
|
188 |
for (j = 0; j < 8*stride; j+=stride) /* forward is better. Some automatic prefetch perhaps. */ |
if (rounding) { |
189 |
{ |
for (j = 0; j < 8*stride; j+=stride) { |
190 |
dst[j + 0] = (uint8_t)((src[j + 0] + src[j + stride + 0] )>>1); |
dst[j + 0] = (uint8_t)((src[j + 0] + src[j + stride + 0] )>>1); |
191 |
dst[j + 1] = (uint8_t)((src[j + 1] + src[j + stride + 1] )>>1); |
dst[j + 1] = (uint8_t)((src[j + 1] + src[j + stride + 1] )>>1); |
192 |
dst[j + 2] = (uint8_t)((src[j + 2] + src[j + stride + 2] )>>1); |
dst[j + 2] = (uint8_t)((src[j + 2] + src[j + stride + 2] )>>1); |
196 |
dst[j + 6] = (uint8_t)((src[j + 6] + src[j + stride + 6] )>>1); |
dst[j + 6] = (uint8_t)((src[j + 6] + src[j + stride + 6] )>>1); |
197 |
dst[j + 7] = (uint8_t)((src[j + 7] + src[j + stride + 7] )>>1); |
dst[j + 7] = (uint8_t)((src[j + 7] + src[j + stride + 7] )>>1); |
198 |
} |
} |
199 |
else |
} else { |
200 |
for (j = 0; j < 8*stride; j+=stride) |
for (j = 0; j < 8*stride; j+=stride) { |
|
{ |
|
201 |
dst[j + 0] = (uint8_t)((src[j + 0] + src[j + stride + 0] + 1)>>1); |
dst[j + 0] = (uint8_t)((src[j + 0] + src[j + stride + 0] + 1)>>1); |
202 |
dst[j + 1] = (uint8_t)((src[j + 1] + src[j + stride + 1] + 1)>>1); |
dst[j + 1] = (uint8_t)((src[j + 1] + src[j + stride + 1] + 1)>>1); |
203 |
dst[j + 2] = (uint8_t)((src[j + 2] + src[j + stride + 2] + 1)>>1); |
dst[j + 2] = (uint8_t)((src[j + 2] + src[j + stride + 2] + 1)>>1); |
208 |
dst[j + 7] = (uint8_t)((src[j + 7] + src[j + stride + 7] + 1)>>1); |
dst[j + 7] = (uint8_t)((src[j + 7] + src[j + stride + 7] + 1)>>1); |
209 |
} |
} |
210 |
} |
} |
211 |
|
} |
212 |
|
|
213 |
|
/* dst = (dst + interpolate(src))/2 */ |
214 |
|
|
215 |
void |
void |
216 |
interpolate8x8_halfpel_hv_c(uint8_t * const dst, |
interpolate8x8_halfpel_v_add_c(uint8_t * const dst, |
217 |
const uint8_t * const src, |
const uint8_t * const src, |
218 |
const uint32_t stride, |
const uint32_t stride, |
219 |
const uint32_t rounding) |
const uint32_t rounding) |
220 |
{ |
{ |
221 |
intptr_t j; |
uintptr_t j; |
222 |
|
|
223 |
|
|
224 |
|
if (rounding) { |
225 |
|
for (j = 0; j < 8*stride; j+=stride) { |
226 |
|
dst[j + 0] = (uint8_t)((((src[j + 0] + src[j + stride + 0] )>>1) + dst[j+0] + 1)>>1); |
227 |
|
dst[j + 1] = (uint8_t)((((src[j + 1] + src[j + stride + 1] )>>1) + dst[j+1] + 1)>>1); |
228 |
|
dst[j + 2] = (uint8_t)((((src[j + 2] + src[j + stride + 2] )>>1) + dst[j+2] + 1)>>1); |
229 |
|
dst[j + 3] = (uint8_t)((((src[j + 3] + src[j + stride + 3] )>>1) + dst[j+3] + 1)>>1); |
230 |
|
dst[j + 4] = (uint8_t)((((src[j + 4] + src[j + stride + 4] )>>1) + dst[j+4] + 1)>>1); |
231 |
|
dst[j + 5] = (uint8_t)((((src[j + 5] + src[j + stride + 5] )>>1) + dst[j+5] + 1)>>1); |
232 |
|
dst[j + 6] = (uint8_t)((((src[j + 6] + src[j + stride + 6] )>>1) + dst[j+6] + 1)>>1); |
233 |
|
dst[j + 7] = (uint8_t)((((src[j + 7] + src[j + stride + 7] )>>1) + dst[j+7] + 1)>>1); |
234 |
|
} |
235 |
|
} else { |
236 |
|
for (j = 0; j < 8*stride; j+=stride) { |
237 |
|
dst[j + 0] = (uint8_t)((((src[j + 0] + src[j + stride + 0] + 1)>>1) + dst[j+0] + 1)>>1); |
238 |
|
dst[j + 1] = (uint8_t)((((src[j + 1] + src[j + stride + 1] + 1)>>1) + dst[j+1] + 1)>>1); |
239 |
|
dst[j + 2] = (uint8_t)((((src[j + 2] + src[j + stride + 2] + 1)>>1) + dst[j+2] + 1)>>1); |
240 |
|
dst[j + 3] = (uint8_t)((((src[j + 3] + src[j + stride + 3] + 1)>>1) + dst[j+3] + 1)>>1); |
241 |
|
dst[j + 4] = (uint8_t)((((src[j + 4] + src[j + stride + 4] + 1)>>1) + dst[j+4] + 1)>>1); |
242 |
|
dst[j + 5] = (uint8_t)((((src[j + 5] + src[j + stride + 5] + 1)>>1) + dst[j+5] + 1)>>1); |
243 |
|
dst[j + 6] = (uint8_t)((((src[j + 6] + src[j + stride + 6] + 1)>>1) + dst[j+6] + 1)>>1); |
244 |
|
dst[j + 7] = (uint8_t)((((src[j + 7] + src[j + stride + 7] + 1)>>1) + dst[j+7] + 1)>>1); |
245 |
|
} |
246 |
|
} |
247 |
|
} |
248 |
|
|
249 |
|
/* dst = interpolate(src) */ |
250 |
|
|
251 |
if (rounding) |
void |
252 |
for (j = 7*stride; j >= 0; j-=stride) |
interpolate8x8_halfpel_hv_c(uint8_t * const dst, |
253 |
|
const uint8_t * const src, |
254 |
|
const uint32_t stride, |
255 |
|
const uint32_t rounding) |
256 |
{ |
{ |
257 |
|
uintptr_t j; |
258 |
|
|
259 |
|
if (rounding) { |
260 |
|
for (j = 0; j < 8*stride; j+=stride) { |
261 |
dst[j + 0] = (uint8_t)((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +1)>>2); |
dst[j + 0] = (uint8_t)((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +1)>>2); |
262 |
dst[j + 1] = (uint8_t)((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +1)>>2); |
dst[j + 1] = (uint8_t)((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +1)>>2); |
263 |
dst[j + 2] = (uint8_t)((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +1)>>2); |
dst[j + 2] = (uint8_t)((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +1)>>2); |
267 |
dst[j + 6] = (uint8_t)((src[j+6] + src[j+7] + src[j+stride+6] + src[j+stride+7] +1)>>2); |
dst[j + 6] = (uint8_t)((src[j+6] + src[j+7] + src[j+stride+6] + src[j+stride+7] +1)>>2); |
268 |
dst[j + 7] = (uint8_t)((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +1)>>2); |
dst[j + 7] = (uint8_t)((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +1)>>2); |
269 |
} |
} |
270 |
else |
} else { |
271 |
for (j = 7*stride; j >= 0; j-=stride) |
for (j = 0; j < 8*stride; j+=stride) { |
|
{ |
|
272 |
dst[j + 0] = (uint8_t)((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +2)>>2); |
dst[j + 0] = (uint8_t)((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +2)>>2); |
273 |
dst[j + 1] = (uint8_t)((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +2)>>2); |
dst[j + 1] = (uint8_t)((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +2)>>2); |
274 |
dst[j + 2] = (uint8_t)((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +2)>>2); |
dst[j + 2] = (uint8_t)((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +2)>>2); |
279 |
dst[j + 7] = (uint8_t)((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +2)>>2); |
dst[j + 7] = (uint8_t)((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +2)>>2); |
280 |
} |
} |
281 |
} |
} |
282 |
|
} |
283 |
|
|
284 |
|
/* dst = (interpolate(src) + dst)/2 */ |
285 |
|
|
286 |
|
void |
287 |
|
interpolate8x8_halfpel_hv_add_c(uint8_t * const dst, |
288 |
|
const uint8_t * const src, |
289 |
|
const uint32_t stride, |
290 |
|
const uint32_t rounding) |
291 |
|
{ |
292 |
|
uintptr_t j; |
293 |
|
|
294 |
|
if (rounding) { |
295 |
|
for (j = 0; j < 8*stride; j+=stride) { |
296 |
|
dst[j + 0] = (uint8_t)((((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +1)>>2) + dst[j+0])>>1); |
297 |
|
dst[j + 1] = (uint8_t)((((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +1)>>2) + dst[j+1])>>1); |
298 |
|
dst[j + 2] = (uint8_t)((((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +1)>>2) + dst[j+2])>>1); |
299 |
|
dst[j + 3] = (uint8_t)((((src[j+3] + src[j+4] + src[j+stride+3] + src[j+stride+4] +1)>>2) + dst[j+3])>>1); |
300 |
|
dst[j + 4] = (uint8_t)((((src[j+4] + src[j+5] + src[j+stride+4] + src[j+stride+5] +1)>>2) + dst[j+4])>>1); |
301 |
|
dst[j + 5] = (uint8_t)((((src[j+5] + src[j+6] + src[j+stride+5] + src[j+stride+6] +1)>>2) + dst[j+5])>>1); |
302 |
|
dst[j + 6] = (uint8_t)((((src[j+6] + src[j+7] + src[j+stride+6] + src[j+stride+7] +1)>>2) + dst[j+6])>>1); |
303 |
|
dst[j + 7] = (uint8_t)((((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +1)>>2) + dst[j+7])>>1); |
304 |
|
} |
305 |
|
} else { |
306 |
|
for (j = 0; j < 8*stride; j+=stride) { |
307 |
|
dst[j + 0] = (uint8_t)((((src[j+0] + src[j+1] + src[j+stride+0] + src[j+stride+1] +2)>>2) + dst[j+0] + 1)>>1); |
308 |
|
dst[j + 1] = (uint8_t)((((src[j+1] + src[j+2] + src[j+stride+1] + src[j+stride+2] +2)>>2) + dst[j+1] + 1)>>1); |
309 |
|
dst[j + 2] = (uint8_t)((((src[j+2] + src[j+3] + src[j+stride+2] + src[j+stride+3] +2)>>2) + dst[j+2] + 1)>>1); |
310 |
|
dst[j + 3] = (uint8_t)((((src[j+3] + src[j+4] + src[j+stride+3] + src[j+stride+4] +2)>>2) + dst[j+3] + 1)>>1); |
311 |
|
dst[j + 4] = (uint8_t)((((src[j+4] + src[j+5] + src[j+stride+4] + src[j+stride+5] +2)>>2) + dst[j+4] + 1)>>1); |
312 |
|
dst[j + 5] = (uint8_t)((((src[j+5] + src[j+6] + src[j+stride+5] + src[j+stride+6] +2)>>2) + dst[j+5] + 1)>>1); |
313 |
|
dst[j + 6] = (uint8_t)((((src[j+6] + src[j+7] + src[j+stride+6] + src[j+stride+7] +2)>>2) + dst[j+6] + 1)>>1); |
314 |
|
dst[j + 7] = (uint8_t)((((src[j+7] + src[j+8] + src[j+stride+7] + src[j+stride+8] +2)>>2) + dst[j+7] + 1)>>1); |
315 |
|
} |
316 |
|
} |
317 |
|
} |
318 |
|
|
319 |
/************************************************************* |
/************************************************************* |
320 |
* QPEL STUFF STARTS HERE * |
* QPEL STUFF STARTS HERE * |