26 |
INTERPOLATE8X8 interpolate8x8_halfpel_v_3dn; |
INTERPOLATE8X8 interpolate8x8_halfpel_v_3dn; |
27 |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dn; |
INTERPOLATE8X8 interpolate8x8_halfpel_hv_3dn; |
28 |
|
|
29 |
static __inline void interpolate8x8_switch(uint8_t * const cur, |
INTERPOLATE8X8 interpolate8x8_halfpel_h_ia64; |
30 |
|
INTERPOLATE8X8 interpolate8x8_halfpel_v_ia64; |
31 |
|
INTERPOLATE8X8 interpolate8x8_halfpel_hv_ia64; |
32 |
|
|
33 |
|
void interpolate8x8_lowpass_h(uint8_t *dst, uint8_t *src, int32_t dst_stride, int32_t src_stride, int32_t rounding); |
34 |
|
void interpolate8x8_lowpass_v(uint8_t *dst, uint8_t *src, int32_t dst_stride, int32_t src_stride, int32_t rounding); |
35 |
|
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); |
36 |
|
void interpolate8x8_bilinear2(uint8_t *dst, uint8_t *src1, uint8_t *src2, int32_t dst_stride, int32_t src_stride, int32_t rounding); |
37 |
|
void interpolate8x8_bilinear4(uint8_t *dst, uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4, int32_t stride, int32_t rounding); |
38 |
|
|
39 |
|
void interpolate8x8_c(uint8_t * const dst, |
40 |
|
const uint8_t * const src, |
41 |
|
const uint32_t x, |
42 |
|
const uint32_t y, |
43 |
|
const uint32_t stride); |
44 |
|
|
45 |
|
static __inline void |
46 |
|
interpolate8x8_switch(uint8_t * const cur, |
47 |
const uint8_t * const refn, |
const uint8_t * const refn, |
48 |
const uint32_t x, const uint32_t y, |
const uint32_t x, |
49 |
const int32_t dx, const int dy, |
const uint32_t y, |
50 |
|
const int32_t dx, |
51 |
|
const int dy, |
52 |
const uint32_t stride, |
const uint32_t stride, |
53 |
const uint32_t rounding) |
const uint32_t rounding) |
54 |
{ |
{ |
59 |
case 0 : |
case 0 : |
60 |
ddx = dx/2; |
ddx = dx/2; |
61 |
ddy = dy/2; |
ddy = dy/2; |
62 |
transfer8x8_copy(cur + y*stride + x, refn + (y+ddy)*stride + x + ddx, stride); |
transfer8x8_copy(cur + y * stride + x, |
63 |
|
refn + (int)((y + ddy) * stride + x + ddx), stride); |
64 |
break; |
break; |
65 |
|
|
66 |
case 1 : |
case 1 : |
67 |
ddx = dx/2; |
ddx = dx/2; |
68 |
ddy = (dy-1)/2; |
ddy = (dy-1)/2; |
69 |
interpolate8x8_halfpel_v(cur + y*stride + x, |
interpolate8x8_halfpel_v(cur + y*stride + x, |
70 |
refn + (y+ddy)*stride + x + ddx, stride, rounding); |
refn + (int)((y + ddy) * stride + x + ddx), stride, |
71 |
|
rounding); |
72 |
break; |
break; |
73 |
|
|
74 |
case 2 : |
case 2 : |
75 |
ddx = (dx-1)/2; |
ddx = (dx-1)/2; |
76 |
ddy = dy/2; |
ddy = dy/2; |
77 |
interpolate8x8_halfpel_h(cur + y*stride + x, |
interpolate8x8_halfpel_h(cur + y*stride + x, |
78 |
refn + (y+ddy)*stride + x + ddx, stride, rounding); |
refn + (int)((y + ddy) * stride + x + ddx), stride, |
79 |
|
rounding); |
80 |
break; |
break; |
81 |
|
|
82 |
default : |
default : |
83 |
ddx = (dx-1)/2; |
ddx = (dx-1)/2; |
84 |
ddy = (dy-1)/2; |
ddy = (dy-1)/2; |
85 |
interpolate8x8_halfpel_hv(cur + y*stride + x, |
interpolate8x8_halfpel_hv(cur + y*stride + x, |
86 |
refn + (y+ddy)*stride + x + ddx, stride, rounding); |
refn + (int)((y + ddy) * stride + x + ddx), stride, |
87 |
|
rounding); |
88 |
|
break; |
89 |
|
} |
90 |
|
} |
91 |
|
|
92 |
|
static __inline uint8_t * |
93 |
|
interpolate8x8_switch2(uint8_t * const buffer, |
94 |
|
const uint8_t * const refn, |
95 |
|
const uint32_t x, |
96 |
|
const uint32_t y, |
97 |
|
const int32_t dx, |
98 |
|
const int dy, |
99 |
|
const uint32_t stride, |
100 |
|
const uint32_t rounding) |
101 |
|
{ |
102 |
|
int32_t ddx, ddy; |
103 |
|
|
104 |
|
switch (((dx & 1) << 1) + (dy & 1)) // ((dx%2)?2:0)+((dy%2)?1:0) |
105 |
|
{ |
106 |
|
case 0: |
107 |
|
return (uint8_t *)refn + (int)((y + dy/2) * stride + x + dx/2); |
108 |
|
|
109 |
|
case 1: |
110 |
|
ddx = dx / 2; |
111 |
|
ddy = (dy - 1) / 2; |
112 |
|
interpolate8x8_halfpel_v(buffer, |
113 |
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
114 |
|
rounding); |
115 |
|
break; |
116 |
|
|
117 |
|
case 2: |
118 |
|
ddx = (dx - 1) / 2; |
119 |
|
ddy = dy / 2; |
120 |
|
interpolate8x8_halfpel_h(buffer, |
121 |
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
122 |
|
rounding); |
123 |
|
break; |
124 |
|
|
125 |
|
default: |
126 |
|
ddx = (dx - 1) / 2; |
127 |
|
ddy = (dy - 1) / 2; |
128 |
|
interpolate8x8_halfpel_hv(buffer, |
129 |
|
refn + (int)((y + ddy) * stride + x + ddx), stride, |
130 |
|
rounding); |
131 |
|
break; |
132 |
|
} |
133 |
|
return buffer; |
134 |
|
} |
135 |
|
|
136 |
|
|
137 |
|
|
138 |
|
static void |
139 |
|
interpolate8x8_quarterpel(uint8_t * const cur, |
140 |
|
uint8_t * const refn, |
141 |
|
const uint32_t x, const uint32_t y, |
142 |
|
const int32_t dx, const int dy, |
143 |
|
const uint32_t stride, |
144 |
|
const uint32_t rounding) |
145 |
|
{ |
146 |
|
const int32_t xRef = x*4 + dx; |
147 |
|
const int32_t yRef = y*4 + dy; |
148 |
|
|
149 |
|
uint8_t *src, *dst; |
150 |
|
int32_t x_int, y_int, x_frac, y_frac; |
151 |
|
|
152 |
|
uint8_t halfpel_h[72]; |
153 |
|
uint8_t halfpel_v[64]; |
154 |
|
uint8_t halfpel_hv[64]; |
155 |
|
|
156 |
|
x_int = xRef/4; |
157 |
|
if (xRef < 0 && xRef % 4) |
158 |
|
x_int--; |
159 |
|
|
160 |
|
x_frac = xRef - (4*x_int); |
161 |
|
|
162 |
|
y_int = yRef/4; |
163 |
|
if (yRef < 0 && yRef % 4) |
164 |
|
y_int--; |
165 |
|
|
166 |
|
y_frac = yRef - (4*y_int); |
167 |
|
|
168 |
|
src = refn + y_int * stride + x_int; |
169 |
|
dst = cur + y * stride + x; |
170 |
|
|
171 |
|
switch((y_frac << 2) | (x_frac)) { |
172 |
|
|
173 |
|
case 0: |
174 |
|
transfer8x8_copy(dst, src, stride); |
175 |
|
break; |
176 |
|
|
177 |
|
case 1: |
178 |
|
interpolate8x8_lowpass_h(halfpel_h, src, 8, stride, rounding); |
179 |
|
interpolate8x8_bilinear2(dst, src, halfpel_h, stride, stride, rounding); |
180 |
|
break; |
181 |
|
|
182 |
|
case 2: |
183 |
|
interpolate8x8_lowpass_h(dst, src, stride, stride, rounding); |
184 |
|
break; |
185 |
|
|
186 |
|
case 3: |
187 |
|
interpolate8x8_lowpass_h(halfpel_h, src, 8, stride, rounding); |
188 |
|
interpolate8x8_bilinear2(dst, src+1, halfpel_h, stride, stride, rounding); |
189 |
|
break; |
190 |
|
|
191 |
|
case 4: |
192 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
193 |
|
interpolate8x8_bilinear2(dst, src, halfpel_v, stride, stride, rounding); |
194 |
|
break; |
195 |
|
|
196 |
|
case 5: |
197 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
198 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
199 |
|
interpolate8x8_bilinear4(dst, src, halfpel_h, halfpel_v, halfpel_hv, stride, rounding); |
200 |
|
break; |
201 |
|
|
202 |
|
case 6: |
203 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
204 |
|
interpolate8x8_bilinear2(dst, halfpel_h, halfpel_hv, stride, 8, 1-rounding); |
205 |
|
break; |
206 |
|
|
207 |
|
case 7: |
208 |
|
interpolate8x8_lowpass_v(halfpel_v, src+1, 8, stride, 16-rounding); |
209 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
210 |
|
interpolate8x8_bilinear4(dst, src+1, halfpel_h, halfpel_v, halfpel_hv, stride, rounding); |
211 |
|
break; |
212 |
|
|
213 |
|
case 8: |
214 |
|
interpolate8x8_lowpass_v(dst, src, stride, stride, rounding); |
215 |
|
break; |
216 |
|
|
217 |
|
case 9: |
218 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, 16-rounding); |
219 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
220 |
|
interpolate8x8_bilinear2(dst, halfpel_v, halfpel_hv, stride, 8, rounding); |
221 |
|
break; |
222 |
|
|
223 |
|
case 10: |
224 |
|
interpolate8x8_lowpass_hv(dst, halfpel_h, src, stride, 8, stride, rounding); |
225 |
|
break; |
226 |
|
|
227 |
|
case 11: |
228 |
|
interpolate8x8_lowpass_v(halfpel_v, src+1, 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 12: |
234 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
235 |
|
interpolate8x8_bilinear2(dst, src+stride, halfpel_v, stride, stride, rounding); |
236 |
|
break; |
237 |
|
|
238 |
|
case 13: |
239 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
240 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
241 |
|
interpolate8x8_bilinear4(dst, src+stride, halfpel_h+8, halfpel_v, halfpel_hv, stride, rounding); |
242 |
|
break; |
243 |
|
|
244 |
|
case 14: |
245 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
246 |
|
interpolate8x8_bilinear2(dst, halfpel_h+8, halfpel_hv, stride, 8, rounding); |
247 |
|
break; |
248 |
|
|
249 |
|
case 15: |
250 |
|
interpolate8x8_lowpass_v(halfpel_v, src+1, 8, stride, rounding); |
251 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
252 |
|
interpolate8x8_bilinear4(dst, src+stride+1, halfpel_h+8, halfpel_v, halfpel_hv, stride, rounding); |
253 |
break; |
break; |
254 |
} |
} |
255 |
} |
} |