1 |
|
/***************************************************************************** |
2 |
|
* |
3 |
|
* XVID MPEG-4 VIDEO CODEC |
4 |
|
* - 8x8 block-based halfpel interpolation - headers |
5 |
|
* |
6 |
|
* Copyright(C) 2002 Peter Ross <pross@xvid.org> |
7 |
|
* |
8 |
|
* This file is part of XviD, a free MPEG-4 video encoder/decoder |
9 |
|
* |
10 |
|
* XviD is free software; you can redistribute it and/or modify it |
11 |
|
* under the terms of the GNU General Public License as published by |
12 |
|
* the Free Software Foundation; either version 2 of the License, or |
13 |
|
* (at your option) any later version. |
14 |
|
* |
15 |
|
* This program is distributed in the hope that it will be useful, |
16 |
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
17 |
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
18 |
|
* GNU General Public License for more details. |
19 |
|
* |
20 |
|
* You should have received a copy of the GNU General Public License |
21 |
|
* along with this program; if not, write to the Free Software |
22 |
|
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
23 |
|
* |
24 |
|
* Under section 8 of the GNU General Public License, the copyright |
25 |
|
* holders of XVID explicitly forbid distribution in the following |
26 |
|
* countries: |
27 |
|
* |
28 |
|
* - Japan |
29 |
|
* - United States of America |
30 |
|
* |
31 |
|
* Linking XviD statically or dynamically with other modules is making a |
32 |
|
* combined work based on XviD. Thus, the terms and conditions of the |
33 |
|
* GNU General Public License cover the whole combination. |
34 |
|
* |
35 |
|
* As a special exception, the copyright holders of XviD give you |
36 |
|
* permission to link XviD with independent modules that communicate with |
37 |
|
* XviD solely through the VFW1.1 and DShow interfaces, regardless of the |
38 |
|
* license terms of these independent modules, and to copy and distribute |
39 |
|
* the resulting combined work under terms of your choice, provided that |
40 |
|
* every copy of the combined work is accompanied by a complete copy of |
41 |
|
* the source code of XviD (the version of XviD used to produce the |
42 |
|
* combined work), being distributed under the terms of the GNU General |
43 |
|
* Public License plus this exception. An independent module is a module |
44 |
|
* which is not derived from or based on XviD. |
45 |
|
* |
46 |
|
* Note that people who make modified versions of XviD are not obligated |
47 |
|
* to grant this special exception for their modified versions; it is |
48 |
|
* their choice whether to do so. The GNU General Public License gives |
49 |
|
* permission to release a modified version without this exception; this |
50 |
|
* exception also makes it possible to release a modified version which |
51 |
|
* carries forward this exception. |
52 |
|
* |
53 |
|
* $Id$ |
54 |
|
* |
55 |
|
****************************************************************************/ |
56 |
|
|
57 |
#include "../utils/mem_transfer.h" |
#include "../utils/mem_transfer.h" |
58 |
|
|
59 |
typedef void (INTERPOLATE8X8)(uint8_t * const dst, |
typedef void (INTERPOLATE8X8)(uint8_t * const dst, |
82 |
INTERPOLATE8X8 interpolate8x8_halfpel_v_3dn; |
INTERPOLATE8X8 interpolate8x8_halfpel_v_3dn; |
83 |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dn; |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dn; |
84 |
|
|
85 |
static __inline void interpolate8x8_switch(uint8_t * const cur, |
INTERPOLATE8X8 interpolate8x8_halfpel_h_ia64; |
86 |
|
INTERPOLATE8X8 interpolate8x8_halfpel_v_ia64; |
87 |
|
INTERPOLATE8X8 interpolate8x8_halfpel_hv_ia64; |
88 |
|
|
89 |
|
void interpolate8x8_lowpass_h(uint8_t *dst, uint8_t *src, int32_t dst_stride, int32_t src_stride, int32_t rounding); |
90 |
|
void interpolate8x8_lowpass_v(uint8_t *dst, uint8_t *src, int32_t dst_stride, int32_t src_stride, int32_t rounding); |
91 |
|
void interpolate8x8_lowpass_hv(uint8_t *dst1, uint8_t *dst2, uint8_t *src, int32_t dst1_stride, int32_t dst2_stride, int32_t src_stride, int32_t rounding); |
92 |
|
void interpolate8x8_bilinear2(uint8_t *dst, uint8_t *src1, uint8_t *src2, int32_t dst_stride, int32_t src_stride, int32_t rounding); |
93 |
|
void interpolate8x8_bilinear4(uint8_t *dst, uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4, int32_t stride, int32_t rounding); |
94 |
|
|
95 |
|
void interpolate8x8_c(uint8_t * const dst, |
96 |
|
const uint8_t * const src, |
97 |
|
const uint32_t x, |
98 |
|
const uint32_t y, |
99 |
|
const uint32_t stride); |
100 |
|
|
101 |
|
static __inline void |
102 |
|
interpolate8x8_switch(uint8_t * const cur, |
103 |
const uint8_t * const refn, |
const uint8_t * const refn, |
104 |
const uint32_t x, const uint32_t y, |
const uint32_t x, |
105 |
const int32_t dx, const int dy, |
const uint32_t y, |
106 |
|
const int32_t dx, |
107 |
|
const int dy, |
108 |
const uint32_t stride, |
const uint32_t stride, |
109 |
const uint32_t rounding) |
const uint32_t rounding) |
110 |
{ |
{ |
115 |
case 0 : |
case 0 : |
116 |
ddx = dx/2; |
ddx = dx/2; |
117 |
ddy = dy/2; |
ddy = dy/2; |
118 |
transfer8x8_copy(cur + y*stride + x, refn + (y+ddy)*stride + x + ddx, stride); |
transfer8x8_copy(cur + y * stride + x, |
119 |
|
refn + (int)((y + ddy) * stride + x + ddx), stride); |
120 |
break; |
break; |
121 |
|
|
122 |
case 1 : |
case 1 : |
123 |
ddx = dx/2; |
ddx = dx/2; |
124 |
ddy = (dy-1)/2; |
ddy = (dy-1)/2; |
125 |
interpolate8x8_halfpel_v(cur + y*stride + x, |
interpolate8x8_halfpel_v(cur + y*stride + x, |
126 |
refn + (y+ddy)*stride + x + ddx, stride, rounding); |
refn + (int)((y + ddy) * stride + x + ddx), stride, |
127 |
|
rounding); |
128 |
break; |
break; |
129 |
|
|
130 |
case 2 : |
case 2 : |
131 |
ddx = (dx-1)/2; |
ddx = (dx-1)/2; |
132 |
ddy = dy/2; |
ddy = dy/2; |
133 |
interpolate8x8_halfpel_h(cur + y*stride + x, |
interpolate8x8_halfpel_h(cur + y*stride + x, |
134 |
refn + (y+ddy)*stride + x + ddx, stride, rounding); |
refn + (int)((y + ddy) * stride + x + ddx), stride, |
135 |
|
rounding); |
136 |
break; |
break; |
137 |
|
|
138 |
default : |
default : |
139 |
ddx = (dx-1)/2; |
ddx = (dx-1)/2; |
140 |
ddy = (dy-1)/2; |
ddy = (dy-1)/2; |
141 |
interpolate8x8_halfpel_hv(cur + y*stride + x, |
interpolate8x8_halfpel_hv(cur + y*stride + x, |
142 |
refn + (y+ddy)*stride + x + ddx, stride, rounding); |
refn + (int)((y + ddy) * stride + x + ddx), stride, |
143 |
|
rounding); |
144 |
break; |
break; |
145 |
} |
} |
146 |
} |
} |
147 |
|
|
148 |
|
|
149 |
void interpolate8x8_c(uint8_t * const dst, |
static __inline void interpolate8x8_quarterpel(uint8_t * const cur, |
150 |
const uint8_t * const src, |
uint8_t * const refn, |
151 |
const uint32_t x, const uint32_t y, |
const uint32_t x, const uint32_t y, |
152 |
const uint32_t stride); |
const int32_t dx, const int dy, |
153 |
|
const uint32_t stride, |
154 |
|
const uint32_t rounding) |
155 |
|
{ |
156 |
|
const int32_t xRef = x*4 + dx; |
157 |
|
const int32_t yRef = y*4 + dy; |
158 |
|
|
159 |
|
uint8_t *src, *dst; |
160 |
|
int32_t x_int, y_int, x_frac, y_frac; |
161 |
|
|
162 |
|
uint8_t halfpel_h[72]; |
163 |
|
uint8_t halfpel_v[64]; |
164 |
|
uint8_t halfpel_hv[64]; |
165 |
|
|
166 |
|
x_int = xRef/4; |
167 |
|
if (xRef < 0 && xRef % 4) |
168 |
|
x_int--; |
169 |
|
|
170 |
|
x_frac = xRef - (4*x_int); |
171 |
|
|
172 |
|
y_int = yRef/4; |
173 |
|
if (yRef < 0 && yRef % 4) |
174 |
|
y_int--; |
175 |
|
|
176 |
|
y_frac = yRef - (4*y_int); |
177 |
|
|
178 |
|
src = refn + y_int * stride + x_int; |
179 |
|
dst = cur + y * stride + x; |
180 |
|
|
181 |
|
switch((y_frac << 2) | (x_frac)) { |
182 |
|
|
183 |
|
case 0: |
184 |
|
transfer8x8_copy(dst, src, stride); |
185 |
|
break; |
186 |
|
|
187 |
|
case 1: |
188 |
|
interpolate8x8_lowpass_h(halfpel_h, src, 8, stride, rounding); |
189 |
|
interpolate8x8_bilinear2(dst, src, halfpel_h, stride, stride, rounding); |
190 |
|
break; |
191 |
|
|
192 |
|
case 2: |
193 |
|
interpolate8x8_lowpass_h(dst, src, stride, stride, rounding); |
194 |
|
break; |
195 |
|
|
196 |
|
case 3: |
197 |
|
interpolate8x8_lowpass_h(halfpel_h, src, 8, stride, rounding); |
198 |
|
interpolate8x8_bilinear2(dst, src+1, halfpel_h, stride, stride, rounding); |
199 |
|
break; |
200 |
|
|
201 |
|
case 4: |
202 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
203 |
|
interpolate8x8_bilinear2(dst, src, halfpel_v, stride, stride, rounding); |
204 |
|
break; |
205 |
|
|
206 |
|
case 5: |
207 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
208 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
209 |
|
interpolate8x8_bilinear4(dst, src, halfpel_h, halfpel_v, halfpel_hv, stride, rounding); |
210 |
|
break; |
211 |
|
|
212 |
|
case 6: |
213 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
214 |
|
interpolate8x8_bilinear2(dst, halfpel_h, halfpel_hv, stride, 8, 1-rounding); |
215 |
|
break; |
216 |
|
|
217 |
|
case 7: |
218 |
|
interpolate8x8_lowpass_v(halfpel_v, src+1, 8, stride, 16-rounding); |
219 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
220 |
|
interpolate8x8_bilinear4(dst, src+1, halfpel_h, halfpel_v, halfpel_hv, stride, rounding); |
221 |
|
break; |
222 |
|
|
223 |
|
case 8: |
224 |
|
interpolate8x8_lowpass_v(dst, src, stride, stride, rounding); |
225 |
|
break; |
226 |
|
|
227 |
|
case 9: |
228 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, 16-rounding); |
229 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
230 |
|
interpolate8x8_bilinear2(dst, halfpel_v, halfpel_hv, stride, 8, rounding); |
231 |
|
break; |
232 |
|
|
233 |
|
case 10: |
234 |
|
interpolate8x8_lowpass_hv(dst, halfpel_h, src, stride, 8, stride, rounding); |
235 |
|
break; |
236 |
|
|
237 |
|
case 11: |
238 |
|
interpolate8x8_lowpass_v(halfpel_v, src+1, 8, stride, 16-rounding); |
239 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
240 |
|
interpolate8x8_bilinear2(dst, halfpel_v, halfpel_hv, stride, 8, rounding); |
241 |
|
break; |
242 |
|
|
243 |
|
case 12: |
244 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
245 |
|
interpolate8x8_bilinear2(dst, src+stride, halfpel_v, stride, stride, rounding); |
246 |
|
break; |
247 |
|
|
248 |
|
case 13: |
249 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
250 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
251 |
|
interpolate8x8_bilinear4(dst, src+stride, halfpel_h+8, halfpel_v, halfpel_hv, stride, rounding); |
252 |
|
break; |
253 |
|
|
254 |
|
case 14: |
255 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
256 |
|
interpolate8x8_bilinear2(dst, halfpel_h+8, halfpel_hv, stride, 8, rounding); |
257 |
|
break; |
258 |
|
|
259 |
|
case 15: |
260 |
|
interpolate8x8_lowpass_v(halfpel_v, src+1, 8, stride, rounding); |
261 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
262 |
|
interpolate8x8_bilinear4(dst, src+stride+1, halfpel_h+8, halfpel_v, halfpel_hv, stride, rounding); |
263 |
|
break; |
264 |
|
} |
265 |
|
} |