10 |
#define ABS(X) (((X)>0)?(X):-(X)) |
#define ABS(X) (((X)>0)?(X):-(X)) |
11 |
#define SIGN(X) (((X)>0)?1:-1) |
#define SIGN(X) (((X)>0)?1:-1) |
12 |
|
|
13 |
|
|
14 |
static __inline void |
static __inline void |
15 |
compensate8x8_halfpel(int16_t * const dct_codes, |
compensate8x8_halfpel(int16_t * const dct_codes, |
16 |
uint8_t * const cur, |
uint8_t * const cur, |
24 |
const int dy, |
const int dy, |
25 |
const uint32_t stride) |
const uint32_t stride) |
26 |
{ |
{ |
27 |
int32_t ddx, ddy; |
const uint8_t * reference; |
28 |
|
|
29 |
switch (((dx & 1) << 1) + (dy & 1)) // ((dx%2)?2:0)+((dy%2)?1:0) |
switch (((dx & 1) << 1) + (dy & 1)) // ((dx%2)?2:0)+((dy%2)?1:0) |
30 |
{ |
{ |
31 |
case 0: |
case 0: reference = ref + ((y + dy / 2) * stride + x + dx / 2); break; |
32 |
ddx = dx / 2; |
case 1: reference = refv + ((y + (dy-1) / 2) * stride + x + dx / 2); break; |
33 |
ddy = dy / 2; |
case 2: reference = refh + ((y + dy / 2) * stride + x + (dx-1) / 2); break; |
|
transfer_8to16sub(dct_codes, cur + y * stride + x, |
|
|
ref + (y + ddy) * stride + x + ddx, stride); |
|
|
break; |
|
|
|
|
|
case 1: |
|
|
ddx = dx / 2; |
|
|
ddy = (dy - 1) / 2; |
|
|
transfer_8to16sub(dct_codes, cur + y * stride + x, |
|
|
refv + (y + ddy) * stride + x + ddx, stride); |
|
|
break; |
|
|
|
|
|
case 2: |
|
|
ddx = (dx - 1) / 2; |
|
|
ddy = dy / 2; |
|
|
transfer_8to16sub(dct_codes, cur + y * stride + x, |
|
|
refh + (y + ddy) * stride + x + ddx, stride); |
|
|
break; |
|
|
|
|
34 |
default: // case 3: |
default: // case 3: |
35 |
ddx = (dx - 1) / 2; |
reference = refhv + ((y + (dy-1) / 2) * stride + x + (dx-1) / 2); break; |
|
ddy = (dy - 1) / 2; |
|
|
transfer_8to16sub(dct_codes, cur + y * stride + x, |
|
|
refhv + (y + ddy) * stride + x + ddx, stride); |
|
|
break; |
|
36 |
} |
} |
37 |
|
transfer_8to16sub(dct_codes, cur + y * stride + x, |
38 |
|
reference, stride); |
39 |
} |
} |
40 |
|
|
|
|
|
|
|
|
41 |
void |
void |
42 |
MBMotionCompensation(MACROBLOCK * const mb, |
MBMotionCompensation(MACROBLOCK * const mb, |
43 |
const uint32_t i, |
const uint32_t i, |
56 |
static const uint32_t roundtab[16] = |
static const uint32_t roundtab[16] = |
57 |
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
{ 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2 }; |
58 |
|
|
59 |
|
if (mb->mode == MODE_NOT_CODED) { |
60 |
|
transfer8x8_copy(cur->y + 16 * (i + j * edged_width), |
61 |
|
ref->y + 16 * (i + j * edged_width), |
62 |
|
edged_width); |
63 |
|
transfer8x8_copy(cur->y + 16 * (i + j * edged_width) + 8, |
64 |
|
ref->y + 16 * (i + j * edged_width) + 8, |
65 |
|
edged_width); |
66 |
|
transfer8x8_copy(cur->y + 16 * (i + j * edged_width) + 8 * edged_width, |
67 |
|
ref->y + 16 * (i + j * edged_width) + 8 * edged_width, |
68 |
|
edged_width); |
69 |
|
transfer8x8_copy(cur->y + 16 * (i + j * edged_width) + 8 * (edged_width+1), |
70 |
|
ref->y + 16 * (i + j * edged_width) + 8 * (edged_width+1), |
71 |
|
edged_width); |
72 |
|
|
73 |
|
transfer8x8_copy(cur->u + 8 * (i + j * edged_width/2), |
74 |
|
ref->u + 8 * (i + j * edged_width/2), |
75 |
|
edged_width / 2); |
76 |
|
transfer8x8_copy(cur->v + 8 * (i + j * edged_width/2), |
77 |
|
ref->v + 8 * (i + j * edged_width/2), |
78 |
|
edged_width / 2); |
79 |
|
return; |
80 |
|
} |
81 |
|
|
82 |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
if (mb->mode == MODE_INTER || mb->mode == MODE_INTER_Q) { |
83 |
int32_t dx = mb->mvs[0].x; |
int32_t dx = mb->mvs[0].x; |
99 |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
100 |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
101 |
|
|
|
/* uv-image-based compensation */ |
|
|
#ifdef BFRAMES |
|
|
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, ref->u, refh->u, |
|
|
refv->u, refhv->u, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2); |
|
|
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, ref->v, refh->v, |
|
|
refv->v, refhv->v, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2); |
|
|
#else |
|
102 |
/* uv-block-based compensation */ |
/* uv-block-based compensation */ |
|
interpolate8x8_switch(refv->u, ref->u, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2, rounding); |
|
103 |
transfer_8to16sub(&dct_codes[4 * 64], |
transfer_8to16sub(&dct_codes[4 * 64], |
104 |
cur->u + 8 * j * edged_width / 2 + 8 * i, |
cur->u + 8 * j * edged_width / 2 + 8 * i, |
105 |
refv->u + 8 * j * edged_width / 2 + 8 * i, |
interpolate8x8_switch2(refv->u, ref->u, 8 * i, 8 * j, |
106 |
|
dx, dy, edged_width / 2, rounding), |
107 |
edged_width / 2); |
edged_width / 2); |
108 |
|
|
|
interpolate8x8_switch(refv->v, ref->v, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2, rounding); |
|
109 |
transfer_8to16sub(&dct_codes[5 * 64], |
transfer_8to16sub(&dct_codes[5 * 64], |
110 |
cur->v + 8 * j * edged_width / 2 + 8 * i, |
cur->v + 8 * j * edged_width / 2 + 8 * i, |
111 |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
interpolate8x8_switch2(refv->u, ref->v, 8 * i, 8 * j, |
112 |
|
dx, dy, edged_width / 2, rounding), |
113 |
edged_width / 2); |
edged_width / 2); |
114 |
#endif |
|
115 |
} else // mode == MODE_INTER4V |
} else { // mode == MODE_INTER4V |
|
{ |
|
116 |
int32_t sum, dx, dy; |
int32_t sum, dx, dy; |
117 |
|
|
118 |
compensate8x8_halfpel(&dct_codes[0 * 64], cur->y, ref->y, refh->y, |
compensate8x8_halfpel(&dct_codes[0 * 64], cur->y, ref->y, refh->y, |
136 |
dy = (sum ? SIGN(sum) * |
dy = (sum ? SIGN(sum) * |
137 |
(roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
(roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2) : 0); |
138 |
|
|
|
/* uv-image-based compensation */ |
|
|
#ifdef BFRAMES |
|
|
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, ref->u, refh->u, |
|
|
refv->u, refhv->u, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2); |
|
|
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, ref->v, refh->v, |
|
|
refv->v, refhv->v, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2); |
|
|
#else |
|
139 |
/* uv-block-based compensation */ |
/* uv-block-based compensation */ |
|
interpolate8x8_switch(refv->u, ref->u, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2, rounding); |
|
140 |
transfer_8to16sub(&dct_codes[4 * 64], |
transfer_8to16sub(&dct_codes[4 * 64], |
141 |
cur->u + 8 * j * edged_width / 2 + 8 * i, |
cur->u + 8 * j * edged_width / 2 + 8 * i, |
142 |
refv->u + 8 * j * edged_width / 2 + 8 * i, |
interpolate8x8_switch2(refv->u, ref->u, 8 * i, 8 * j, |
143 |
|
dx, dy, edged_width / 2, rounding), |
144 |
edged_width / 2); |
edged_width / 2); |
145 |
|
|
|
interpolate8x8_switch(refv->v, ref->v, 8 * i, 8 * j, dx, dy, |
|
|
edged_width / 2, rounding); |
|
146 |
transfer_8to16sub(&dct_codes[5 * 64], |
transfer_8to16sub(&dct_codes[5 * 64], |
147 |
cur->v + 8 * j * edged_width / 2 + 8 * i, |
cur->v + 8 * j * edged_width / 2 + 8 * i, |
148 |
refv->v + 8 * j * edged_width / 2 + 8 * i, |
interpolate8x8_switch2(refv->u, ref->v, 8 * i, 8 * j, |
149 |
|
dx, dy, edged_width / 2, rounding), |
150 |
|
|
151 |
edged_width / 2); |
edged_width / 2); |
|
#endif |
|
152 |
} |
} |
153 |
} |
} |
154 |
|
|
175 |
const int32_t edged_width = pParam->edged_width; |
const int32_t edged_width = pParam->edged_width; |
176 |
int32_t dx, dy; |
int32_t dx, dy; |
177 |
int32_t b_dx, b_dy; |
int32_t b_dx, b_dy; |
178 |
|
int k,sum; |
179 |
int x = i; |
int x = i; |
180 |
int y = j; |
int y = j; |
181 |
|
|
182 |
|
|
|
|
|
183 |
switch (mb->mode) { |
switch (mb->mode) { |
184 |
case MODE_FORWARD: |
case MODE_FORWARD: |
185 |
dx = mb->mvs[0].x; |
dx = mb->mvs[0].x; |
186 |
dy = mb->mvs[0].y; |
dy = mb->mvs[0].y; |
187 |
|
|
188 |
transfer_8to16sub_c(&dct_codes[0 * 64], |
transfer_8to16sub(&dct_codes[0 * 64], |
189 |
cur->y + (j * 16) * edged_width + (i * 16), |
cur->y + (j * 16) * edged_width + (i * 16), |
190 |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
191 |
i * 16, j * 16, 1, dx, dy, edged_width), |
i * 16, j * 16, 1, dx, dy, edged_width), |
197 |
i * 16 + 8, j * 16, 1, dx, dy, edged_width), |
i * 16 + 8, j * 16, 1, dx, dy, edged_width), |
198 |
edged_width); |
edged_width); |
199 |
|
|
200 |
transfer_8to16sub_c(&dct_codes[2 * 64], |
transfer_8to16sub(&dct_codes[2 * 64], |
201 |
cur->y + (j * 16 + 8) * edged_width + (i * 16), |
cur->y + (j * 16 + 8) * edged_width + (i * 16), |
202 |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
203 |
i * 16, j * 16 + 8, 1, dx, dy, |
i * 16, j * 16 + 8, 1, dx, dy, edged_width), |
204 |
edged_width), edged_width); |
edged_width); |
205 |
|
|
206 |
transfer_8to16sub(&dct_codes[3 * 64], |
transfer_8to16sub(&dct_codes[3 * 64], |
207 |
cur->y + (j * 16 + 8) * edged_width + (i * 16 + 8), |
cur->y + (j * 16 + 8) * edged_width + (i * 16 + 8), |
208 |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
209 |
i * 16 + 8, j * 16 + 8, 1, dx, dy, |
i * 16 + 8, j * 16 + 8, 1, dx, dy, edged_width), |
210 |
edged_width), edged_width); |
edged_width); |
211 |
|
|
212 |
|
|
213 |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
214 |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
215 |
|
|
216 |
/* uv-image-based compensation */ |
/* uv-block-based compensation */ |
217 |
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, f_ref->u, f_refh->u, |
transfer_8to16sub(&dct_codes[4 * 64], |
218 |
f_refv->u, f_refhv->u, 8 * i, 8 * j, dx, dy, |
cur->u + 8 * j * edged_width / 2 + 8 * i, |
219 |
|
interpolate8x8_switch2(f_refv->u, f_ref->u, 8 * i, 8 * j, |
220 |
|
dx, dy, edged_width / 2, 0), |
221 |
|
|
222 |
edged_width / 2); |
edged_width / 2); |
223 |
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, f_ref->v, f_refh->v, |
|
224 |
f_refv->v, f_refhv->v, 8 * i, 8 * j, dx, dy, |
transfer_8to16sub(&dct_codes[5 * 64], |
225 |
|
cur->v + 8 * j * edged_width / 2 + 8 * i, |
226 |
|
interpolate8x8_switch2(f_refv->u, f_ref->v, 8 * i, 8 * j, |
227 |
|
dx, dy, edged_width / 2, 0), |
228 |
|
|
229 |
edged_width / 2); |
edged_width / 2); |
230 |
|
|
231 |
break; |
break; |
234 |
b_dx = mb->b_mvs[0].x; |
b_dx = mb->b_mvs[0].x; |
235 |
b_dy = mb->b_mvs[0].y; |
b_dy = mb->b_mvs[0].y; |
236 |
|
|
237 |
transfer_8to16sub_c(&dct_codes[0 * 64], |
transfer_8to16sub(&dct_codes[0 * 64], |
238 |
cur->y + (j * 16) * edged_width + (i * 16), |
cur->y + (j * 16) * edged_width + (i * 16), |
239 |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
240 |
i * 16, j * 16, 1, b_dx, b_dy, |
i * 16, j * 16, 1, b_dx, b_dy, |
246 |
i * 16 + 8, j * 16, 1, b_dx, b_dy, |
i * 16 + 8, j * 16, 1, b_dx, b_dy, |
247 |
edged_width), edged_width); |
edged_width), edged_width); |
248 |
|
|
249 |
transfer_8to16sub_c(&dct_codes[2 * 64], |
transfer_8to16sub(&dct_codes[2 * 64], |
250 |
cur->y + (j * 16 + 8) * edged_width + (i * 16), |
cur->y + (j * 16 + 8) * edged_width + (i * 16), |
251 |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
252 |
i * 16, j * 16 + 8, 1, b_dx, b_dy, |
i * 16, j * 16 + 8, 1, b_dx, b_dy, |
261 |
b_dx = (b_dx & 3) ? (b_dx >> 1) | 1 : b_dx / 2; |
b_dx = (b_dx & 3) ? (b_dx >> 1) | 1 : b_dx / 2; |
262 |
b_dy = (b_dy & 3) ? (b_dy >> 1) | 1 : b_dy / 2; |
b_dy = (b_dy & 3) ? (b_dy >> 1) | 1 : b_dy / 2; |
263 |
|
|
264 |
/* uv-image-based compensation */ |
/* uv-block-based compensation */ |
265 |
compensate8x8_halfpel(&dct_codes[4 * 64], cur->u, b_ref->u, b_refh->u, |
transfer_8to16sub(&dct_codes[4 * 64], |
266 |
b_refv->u, b_refhv->u, 8 * i, 8 * j, b_dx, b_dy, |
cur->u + 8 * j * edged_width / 2 + 8 * i, |
267 |
|
interpolate8x8_switch2(f_refv->u, b_ref->u, 8 * i, 8 * j, |
268 |
|
b_dx, b_dy, edged_width / 2, 0), |
269 |
|
|
270 |
edged_width / 2); |
edged_width / 2); |
271 |
compensate8x8_halfpel(&dct_codes[5 * 64], cur->v, b_ref->v, b_refh->v, |
|
272 |
b_refv->v, b_refhv->v, 8 * i, 8 * j, b_dx, b_dy, |
transfer_8to16sub(&dct_codes[5 * 64], |
273 |
|
cur->v + 8 * j * edged_width / 2 + 8 * i, |
274 |
|
interpolate8x8_switch2(f_refv->u, b_ref->v, 8 * i, 8 * j, |
275 |
|
b_dx, b_dy, edged_width / 2, 0), |
276 |
|
|
277 |
edged_width / 2); |
edged_width / 2); |
278 |
|
|
279 |
break; |
break; |
280 |
|
|
281 |
|
|
282 |
case MODE_INTERPOLATE: |
case MODE_INTERPOLATE: /* _could_ use DIRECT, but would be overkill (no 4MV there) */ |
283 |
|
case MODE_DIRECT_NO4V: |
284 |
|
|
285 |
dx = mb->mvs[0].x; |
dx = mb->mvs[0].x; |
286 |
dy = mb->mvs[0].y; |
dy = mb->mvs[0].y; |
287 |
|
|
288 |
b_dx = mb->b_mvs[0].x; |
b_dx = mb->b_mvs[0].x; |
289 |
b_dy = mb->b_mvs[0].y; |
b_dy = mb->b_mvs[0].y; |
290 |
|
|
291 |
transfer_8to16sub2_c(&dct_codes[0 * 64], |
for (k = 0; k < 4; k++) { |
292 |
cur->y + (i * 16) + (j * 16) * edged_width, |
transfer_8to16sub2(&dct_codes[k * 64], |
293 |
get_ref(f_ref->y, f_refh->y, f_refv->y, |
cur->y + (i * 16+(k&1)*8) + (j * 16+((k>>1)*8)) * edged_width, |
|
f_refhv->y, 16 * i, 16 * j, 1, dx, dy, |
|
|
edged_width), get_ref(b_ref->y, b_refh->y, |
|
|
b_refv->y, |
|
|
b_refhv->y, 16 * i, |
|
|
16 * j, 1, b_dx, |
|
|
b_dy, edged_width), |
|
|
edged_width); |
|
|
|
|
|
transfer_8to16sub2_c(&dct_codes[1 * 64], |
|
|
cur->y + (i * 16 + 8) + (j * 16) * edged_width, |
|
294 |
get_ref(f_ref->y, f_refh->y, f_refv->y, |
get_ref(f_ref->y, f_refh->y, f_refv->y, |
295 |
f_refhv->y, 16 * i + 8, 16 * j, 1, dx, dy, |
f_refhv->y, 2*i + (k&1), 2*j + (k>>1), 8, dx, dy, |
|
edged_width), get_ref(b_ref->y, b_refh->y, |
|
|
b_refv->y, |
|
|
b_refhv->y, |
|
|
16 * i + 8, 16 * j, |
|
|
1, b_dx, b_dy, |
|
296 |
edged_width), |
edged_width), |
297 |
edged_width); |
get_ref(b_ref->y, b_refh->y, b_refv->y, |
298 |
|
b_refhv->y, 2*i + (k&1), 2 * j+(k>>1), 8, b_dx, b_dy, |
|
transfer_8to16sub2_c(&dct_codes[2 * 64], |
|
|
cur->y + (i * 16) + (j * 16 + 8) * edged_width, |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, |
|
|
f_refhv->y, 16 * i, 16 * j + 8, 1, dx, dy, |
|
|
edged_width), get_ref(b_ref->y, b_refh->y, |
|
|
b_refv->y, |
|
|
b_refhv->y, 16 * i, |
|
|
16 * j + 8, 1, b_dx, |
|
|
b_dy, edged_width), |
|
|
edged_width); |
|
|
|
|
|
transfer_8to16sub2_c(&dct_codes[3 * 64], |
|
|
cur->y + (i * 16 + 8) + (j * 16 + |
|
|
8) * edged_width, |
|
|
get_ref(f_ref->y, f_refh->y, f_refv->y, |
|
|
f_refhv->y, 16 * i + 8, 16 * j + 8, 1, dx, |
|
|
dy, edged_width), get_ref(b_ref->y, |
|
|
b_refh->y, |
|
|
b_refv->y, |
|
|
b_refhv->y, |
|
|
16 * i + 8, |
|
|
16 * j + 8, 1, |
|
|
b_dx, b_dy, |
|
299 |
edged_width), |
edged_width), |
300 |
edged_width); |
edged_width); |
301 |
|
} |
302 |
|
|
303 |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
dx = (dx & 3) ? (dx >> 1) | 1 : dx / 2; |
304 |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
dy = (dy & 3) ? (dy >> 1) | 1 : dy / 2; |
306 |
b_dx = (b_dx & 3) ? (b_dx >> 1) | 1 : b_dx / 2; |
b_dx = (b_dx & 3) ? (b_dx >> 1) | 1 : b_dx / 2; |
307 |
b_dy = (b_dy & 3) ? (b_dy >> 1) | 1 : b_dy / 2; |
b_dy = (b_dy & 3) ? (b_dy >> 1) | 1 : b_dy / 2; |
308 |
|
|
309 |
transfer_8to16sub2_c(&dct_codes[4 * 64], |
transfer_8to16sub2(&dct_codes[4 * 64], |
310 |
cur->u + (y * 8) * edged_width / 2 + (x * 8), |
cur->u + (y * 8) * edged_width / 2 + (x * 8), |
311 |
get_ref(f_ref->u, f_refh->u, f_refv->u, |
interpolate8x8_switch2(f_refv->u, b_ref->u, 8 * i, 8 * j, |
312 |
f_refhv->u, 8 * i, 8 * j, 1, dx, dy, |
b_dx, b_dy, edged_width / 2, 0), |
313 |
edged_width / 2), get_ref(b_ref->u, |
interpolate8x8_switch2(f_refv->u + 8, f_ref->u, 8 * i, 8 * j, |
314 |
b_refh->u, |
dx, dy, edged_width / 2, 0), |
|
b_refv->u, |
|
|
b_refhv->u, |
|
|
8 * i, 8 * j, 1, |
|
|
b_dx, b_dy, |
|
|
edged_width / |
|
|
2), |
|
315 |
edged_width / 2); |
edged_width / 2); |
316 |
|
|
317 |
transfer_8to16sub2_c(&dct_codes[5 * 64], |
transfer_8to16sub2(&dct_codes[5 * 64], |
318 |
cur->v + (y * 8) * edged_width / 2 + (x * 8), |
cur->v + (y * 8) * edged_width / 2 + (x * 8), |
319 |
get_ref(f_ref->v, f_refh->v, f_refv->v, |
interpolate8x8_switch2(f_refv->u, b_ref->v, 8 * i, 8 * j, |
320 |
f_refhv->v, 8 * i, 8 * j, 1, dx, dy, |
b_dx, b_dy, edged_width / 2, 0), |
321 |
edged_width / 2), get_ref(b_ref->v, |
interpolate8x8_switch2(f_refv->u + 8, f_ref->v, 8 * i, 8 * j, |
322 |
b_refh->v, |
dx, dy, edged_width / 2, 0), |
|
b_refv->v, |
|
|
b_refhv->v, |
|
|
8 * i, 8 * j, 1, |
|
|
b_dx, b_dy, |
|
|
edged_width / |
|
|
2), |
|
323 |
edged_width / 2); |
edged_width / 2); |
324 |
|
|
325 |
break; |
break; |
326 |
|
|
327 |
case MODE_DIRECT: |
case MODE_DIRECT: |
328 |
// todo |
|
329 |
break; |
for (k=0;k<4;k++) |
330 |
|
{ |
331 |
|
dx = mb->mvs[k].x; |
332 |
|
dy = mb->mvs[k].y; |
333 |
|
|
334 |
|
b_dx = mb->b_mvs[k].x; |
335 |
|
b_dy = mb->b_mvs[k].y; |
336 |
|
|
337 |
|
// fprintf(stderr,"Direct Vector %d -- %d:%d %d:%d\n",k,dx,dy,b_dx,b_dy); |
338 |
|
|
339 |
|
transfer_8to16sub2(&dct_codes[k * 64], |
340 |
|
cur->y + (i*16 + (k&1)*8) + (j*16 + (k>>1)*8 ) * edged_width, |
341 |
|
get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
342 |
|
2*i + (k&1), 2*j + (k>>1), 8, dx, dy, |
343 |
|
edged_width), |
344 |
|
get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
345 |
|
2*i + (k&1), 2*j + (k>>1), 8, b_dx, b_dy, |
346 |
|
edged_width), |
347 |
|
edged_width); |
348 |
} |
} |
349 |
|
|
350 |
|
sum = mb->mvs[0].x + mb->mvs[1].x + mb->mvs[2].x + mb->mvs[3].x; |
351 |
|
dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
352 |
|
|
353 |
|
sum = mb->mvs[0].y + mb->mvs[1].y + mb->mvs[2].y + mb->mvs[3].y; |
354 |
|
dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
355 |
|
|
356 |
|
|
357 |
|
sum = mb->b_mvs[0].x + mb->b_mvs[1].x + mb->b_mvs[2].x + mb->b_mvs[3].x; |
358 |
|
b_dx = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
359 |
|
|
360 |
|
sum = mb->b_mvs[0].y + mb->b_mvs[1].y + mb->b_mvs[2].y + mb->b_mvs[3].y; |
361 |
|
b_dy = (sum == 0 ? 0 : SIGN(sum) * (roundtab[ABS(sum) % 16] + (ABS(sum) / 16) * 2)); |
362 |
|
|
363 |
|
/* // for QPel don't forget to always do |
364 |
|
|
365 |
|
if (quarterpel) |
366 |
|
sum /= 2; |
367 |
|
*/ |
368 |
|
transfer_8to16sub2(&dct_codes[4 * 64], |
369 |
|
cur->u + (y * 8) * edged_width / 2 + (x * 8), |
370 |
|
interpolate8x8_switch2(f_refv->u, b_ref->u, 8 * i, 8 * j, |
371 |
|
b_dx, b_dy, edged_width / 2, 0), |
372 |
|
interpolate8x8_switch2(f_refv->u + 8, f_ref->u, 8 * i, 8 * j, dx, dy, |
373 |
|
edged_width / 2, 0), |
374 |
|
edged_width / 2); |
375 |
|
|
376 |
|
transfer_8to16sub2(&dct_codes[5 * 64], |
377 |
|
cur->v + (y * 8) * edged_width / 2 + (x * 8), |
378 |
|
interpolate8x8_switch2(f_refv->u, b_ref->v, 8 * i, 8 * j, |
379 |
|
b_dx, b_dy, edged_width / 2, 0), |
380 |
|
interpolate8x8_switch2(f_refv->u + 8, f_ref->v, 8 * i, 8 * j, |
381 |
|
dx, dy, edged_width / 2, 0), |
382 |
|
edged_width / 2); |
383 |
|
|
384 |
|
|
385 |
|
break; |
386 |
|
} |
387 |
} |
} |