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; |
break; |
89 |
} |
} |
90 |
} |
} |
91 |
|
|
92 |
|
|
93 |
void interpolate8x8_c(uint8_t * const dst, |
static __inline void interpolate8x8_quarterpel(uint8_t * const cur, |
94 |
const uint8_t * const src, |
uint8_t * const refn, |
95 |
const uint32_t x, const uint32_t y, |
const uint32_t x, const uint32_t y, |
96 |
const uint32_t stride); |
const int32_t dx, const int dy, |
97 |
|
const uint32_t stride, |
98 |
|
const uint32_t rounding) |
99 |
|
{ |
100 |
|
const int32_t xRef = x*4 + dx; |
101 |
|
const int32_t yRef = y*4 + dy; |
102 |
|
|
103 |
|
uint8_t *src, *dst; |
104 |
|
int32_t x_int, y_int, x_frac, y_frac; |
105 |
|
|
106 |
|
uint8_t halfpel_h[72]; |
107 |
|
uint8_t halfpel_v[64]; |
108 |
|
uint8_t halfpel_hv[64]; |
109 |
|
|
110 |
|
x_int = xRef/4; |
111 |
|
if (xRef < 0 && xRef % 4) |
112 |
|
x_int--; |
113 |
|
|
114 |
|
x_frac = xRef - (4*x_int); |
115 |
|
|
116 |
|
y_int = yRef/4; |
117 |
|
if (yRef < 0 && yRef % 4) |
118 |
|
y_int--; |
119 |
|
|
120 |
|
y_frac = yRef - (4*y_int); |
121 |
|
|
122 |
|
src = refn + y_int * stride + x_int; |
123 |
|
dst = cur + y * stride + x; |
124 |
|
|
125 |
|
switch((y_frac << 2) | (x_frac)) { |
126 |
|
|
127 |
|
case 0: |
128 |
|
transfer8x8_copy(dst, src, stride); |
129 |
|
break; |
130 |
|
|
131 |
|
case 1: |
132 |
|
interpolate8x8_lowpass_h(halfpel_h, src, 8, stride, rounding); |
133 |
|
interpolate8x8_bilinear2(dst, src, halfpel_h, stride, stride, rounding); |
134 |
|
break; |
135 |
|
|
136 |
|
case 2: |
137 |
|
interpolate8x8_lowpass_h(dst, src, stride, stride, rounding); |
138 |
|
break; |
139 |
|
|
140 |
|
case 3: |
141 |
|
interpolate8x8_lowpass_h(halfpel_h, src, 8, stride, rounding); |
142 |
|
interpolate8x8_bilinear2(dst, src+1, halfpel_h, stride, stride, rounding); |
143 |
|
break; |
144 |
|
|
145 |
|
case 4: |
146 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
147 |
|
interpolate8x8_bilinear2(dst, src, halfpel_v, stride, stride, rounding); |
148 |
|
break; |
149 |
|
|
150 |
|
case 5: |
151 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
152 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
153 |
|
interpolate8x8_bilinear4(dst, src, halfpel_h, halfpel_v, halfpel_hv, stride, rounding); |
154 |
|
break; |
155 |
|
|
156 |
|
case 6: |
157 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
158 |
|
interpolate8x8_bilinear2(dst, halfpel_h, halfpel_hv, stride, 8, 1-rounding); |
159 |
|
break; |
160 |
|
|
161 |
|
case 7: |
162 |
|
interpolate8x8_lowpass_v(halfpel_v, src+1, 8, stride, 16-rounding); |
163 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
164 |
|
interpolate8x8_bilinear4(dst, src+1, halfpel_h, halfpel_v, halfpel_hv, stride, rounding); |
165 |
|
break; |
166 |
|
|
167 |
|
case 8: |
168 |
|
interpolate8x8_lowpass_v(dst, src, stride, stride, rounding); |
169 |
|
break; |
170 |
|
|
171 |
|
case 9: |
172 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, 16-rounding); |
173 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
174 |
|
interpolate8x8_bilinear2(dst, halfpel_v, halfpel_hv, stride, 8, rounding); |
175 |
|
break; |
176 |
|
|
177 |
|
case 10: |
178 |
|
interpolate8x8_lowpass_hv(dst, halfpel_h, src, stride, 8, stride, rounding); |
179 |
|
break; |
180 |
|
|
181 |
|
case 11: |
182 |
|
interpolate8x8_lowpass_v(halfpel_v, src+1, 8, stride, 16-rounding); |
183 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
184 |
|
interpolate8x8_bilinear2(dst, halfpel_v, halfpel_hv, stride, 8, rounding); |
185 |
|
break; |
186 |
|
|
187 |
|
case 12: |
188 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
189 |
|
interpolate8x8_bilinear2(dst, src+stride, halfpel_v, stride, stride, rounding); |
190 |
|
break; |
191 |
|
|
192 |
|
case 13: |
193 |
|
interpolate8x8_lowpass_v(halfpel_v, src, 8, stride, rounding); |
194 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
195 |
|
interpolate8x8_bilinear4(dst, src+stride, halfpel_h+8, halfpel_v, halfpel_hv, stride, rounding); |
196 |
|
break; |
197 |
|
|
198 |
|
case 14: |
199 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
200 |
|
interpolate8x8_bilinear2(dst, halfpel_h+8, halfpel_hv, stride, 8, rounding); |
201 |
|
break; |
202 |
|
|
203 |
|
case 15: |
204 |
|
interpolate8x8_lowpass_v(halfpel_v, src+1, 8, stride, rounding); |
205 |
|
interpolate8x8_lowpass_hv(halfpel_hv, halfpel_h, src, 8, 8, stride, rounding); |
206 |
|
interpolate8x8_bilinear4(dst, src+stride+1, halfpel_h+8, halfpel_v, halfpel_hv, stride, rounding); |
207 |
|
break; |
208 |
|
} |
209 |
|
} |