Parent Directory
| 
Revision Log
Revision 1988 - (view) (download)
| 1 : | edgomez | 1382 | /***************************************************************************** |
| 2 : | * | ||
| 3 : | * XVID MPEG-4 VIDEO CODEC | ||
| 4 : | * - Motion Compensation related code - | ||
| 5 : | * | ||
| 6 : | * Copyright(C) 2002 Peter Ross <pross@xvid.org> | ||
| 7 : | * 2003 Christoph Lampert <gruel@web.de> | ||
| 8 : | * | ||
| 9 : | * This program is free software ; you can redistribute it and/or modify | ||
| 10 : | * it under the terms of the GNU General Public License as published by | ||
| 11 : | * the Free Software Foundation ; either version 2 of the License, or | ||
| 12 : | * (at your option) any later version. | ||
| 13 : | * | ||
| 14 : | * This program is distributed in the hope that it will be useful, | ||
| 15 : | * but WITHOUT ANY WARRANTY ; without even the implied warranty of | ||
| 16 : | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
| 17 : | * GNU General Public License for more details. | ||
| 18 : | * | ||
| 19 : | * You should have received a copy of the GNU General Public License | ||
| 20 : | * along with this program ; if not, write to the Free Software | ||
| 21 : | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA | ||
| 22 : | * | ||
| 23 : | Isibaar | 1988 | * $Id$ |
| 24 : | edgomez | 1382 | * |
| 25 : | ****************************************************************************/ | ||
| 26 : | suxen_drol | 118 | |
| 27 : | edgomez | 851 | #include <stdio.h> |
| 28 : | |||
| 29 : | Isibaar | 3 | #include "../encoder.h" |
| 30 : | #include "../utils/mbfunctions.h" | ||
| 31 : | #include "../image/interpolate8x8.h" | ||
| 32 : | edgomez | 1382 | #include "../image/qpel.h" |
| 33 : | Isibaar | 3 | #include "../utils/timer.h" |
| 34 : | suxen_drol | 118 | #include "motion.h" |
| 35 : | Isibaar | 3 | |
| 36 : | syskin | 1565 | /* |
| 37 : | edgomez | 1382 | * getref: calculate reference image pointer |
| 38 : | * the decision to use interpolation h/v/hv or the normal image is | ||
| 39 : | * based on dx & dy. | ||
| 40 : | */ | ||
| 41 : | edgomez | 851 | |
| 42 : | edgomez | 1382 | static __inline const uint8_t * |
| 43 : | get_ref(const uint8_t * const refn, | ||
| 44 : | const uint8_t * const refh, | ||
| 45 : | const uint8_t * const refv, | ||
| 46 : | const uint8_t * const refhv, | ||
| 47 : | const uint32_t x, | ||
| 48 : | const uint32_t y, | ||
| 49 : | const uint32_t block, | ||
| 50 : | const int32_t dx, | ||
| 51 : | const int32_t dy, | ||
| 52 : | const int32_t stride) | ||
| 53 : | { | ||
| 54 : | switch (((dx & 1) << 1) + (dy & 1)) { | ||
| 55 : | case 0: | ||
| 56 : | edgomez | 1556 | return refn + (int) (((int)x * (int)block + dx / 2) + ((int)y * (int)block + dy / 2) * (int)stride); |
| 57 : | edgomez | 1382 | case 1: |
| 58 : | edgomez | 1556 | return refv + (int) (((int)x * (int)block + dx / 2) + ((int)y * (int)block + (dy - 1) / 2) * (int)stride); |
| 59 : | edgomez | 1382 | case 2: |
| 60 : | edgomez | 1556 | return refh + (int) (((int)x * (int)block + (dx - 1) / 2) + ((int)y * (int)block + dy / 2) * (int)stride); |
| 61 : | edgomez | 1382 | default: |
| 62 : | edgomez | 1556 | return refhv + (int) (((int)x * (int)block + (dx - 1) / 2) + ((int)y * (int)block + (dy - 1) / 2) * (int)stride); |
| 63 : | edgomez | 1382 | } |
| 64 : | } | ||
| 65 : | |||
| 66 : | edgomez | 195 | static __inline void |
| 67 : | edgomez | 851 | compensate16x16_interpolate(int16_t * const dct_codes, |
| 68 : | uint8_t * const cur, | ||
| 69 : | const uint8_t * const ref, | ||
| 70 : | const uint8_t * const refh, | ||
| 71 : | const uint8_t * const refv, | ||
| 72 : | const uint8_t * const refhv, | ||
| 73 : | uint8_t * const tmp, | ||
| 74 : | uint32_t x, | ||
| 75 : | uint32_t y, | ||
| 76 : | const int32_t dx, | ||
| 77 : | const int32_t dy, | ||
| 78 : | const int32_t stride, | ||
| 79 : | const int quarterpel, | ||
| 80 : | const int32_t rounding) | ||
| 81 : | Isibaar | 3 | { |
| 82 : | edgomez | 851 | const uint8_t * ptr; |
| 83 : | Isibaar | 3 | |
| 84 : | |||
| 85 : | syskin | 1565 | if(quarterpel) { |
| 86 : | if ((dx&3) | (dy&3)) { | ||
| 87 : | interpolate16x16_quarterpel(tmp - y * stride - x, | ||
| 88 : | (uint8_t *) ref, tmp + 32, | ||
| 89 : | tmp + 64, tmp + 96, x, y, dx, dy, stride, rounding); | ||
| 90 : | ptr = tmp; | ||
| 91 : | } else ptr = ref + ((int)y + dy/4)*(int)stride + (int)x + dx/4; /* fullpixel position */ | ||
| 92 : | Isibaar | 3 | |
| 93 : | syskin | 1565 | } else ptr = get_ref(ref, refh, refv, refhv, x, y, 1, dx, dy, stride); |
| 94 : | edgomez | 851 | |
| 95 : | syskin | 1565 | transfer_8to16sub(dct_codes, cur + y * stride + x, |
| 96 : | ptr, stride); | ||
| 97 : | transfer_8to16sub(dct_codes+64, cur + y * stride + x + 8, | ||
| 98 : | ptr + 8, stride); | ||
| 99 : | transfer_8to16sub(dct_codes+128, cur + y * stride + x + 8*stride, | ||
| 100 : | ptr + 8*stride, stride); | ||
| 101 : | transfer_8to16sub(dct_codes+192, cur + y * stride + x + 8*stride+8, | ||
| 102 : | ptr + 8*stride + 8, stride); | ||
| 103 : | Isibaar | 3 | |
| 104 : | } | ||
| 105 : | |||
| 106 : | edgomez | 851 | static __inline void |
| 107 : | compensate8x8_interpolate( int16_t * const dct_codes, | ||
| 108 : | uint8_t * const cur, | ||
| 109 : | const uint8_t * const ref, | ||
| 110 : | const uint8_t * const refh, | ||
| 111 : | const uint8_t * const refv, | ||
| 112 : | const uint8_t * const refhv, | ||
| 113 : | uint8_t * const tmp, | ||
| 114 : | uint32_t x, | ||
| 115 : | uint32_t y, | ||
| 116 : | const int32_t dx, | ||
| 117 : | const int32_t dy, | ||
| 118 : | const int32_t stride, | ||
| 119 : | const int32_t quarterpel, | ||
| 120 : | const int32_t rounding) | ||
| 121 : | { | ||
| 122 : | const uint8_t * ptr; | ||
| 123 : | Isibaar | 3 | |
| 124 : | syskin | 1565 | if(quarterpel) { |
| 125 : | if ((dx&3) | (dy&3)) { | ||
| 126 : | interpolate8x8_quarterpel(tmp - y*stride - x, | ||
| 127 : | (uint8_t *) ref, tmp + 32, | ||
| 128 : | tmp + 64, tmp + 96, x, y, dx, dy, stride, rounding); | ||
| 129 : | ptr = tmp; | ||
| 130 : | } else ptr = ref + ((int)y + dy/4)*(int)stride + (int)x + dx/4; /* fullpixel position */ | ||
| 131 : | } else ptr = get_ref(ref, refh, refv, refhv, x, y, 1, dx, dy, stride); | ||
| 132 : | Isibaar | 3 | |
| 133 : | syskin | 1565 | transfer_8to16sub(dct_codes, cur + y * stride + x, ptr, stride); |
| 134 : | edgomez | 851 | } |
| 135 : | |||
| 136 : | |||
| 137 : | static void | ||
| 138 : | CompensateChroma( int dx, int dy, | ||
| 139 : | const int i, const int j, | ||
| 140 : | IMAGE * const Cur, | ||
| 141 : | const IMAGE * const Ref, | ||
| 142 : | uint8_t * const temp, | ||
| 143 : | int16_t * const coeff, | ||
| 144 : | const int32_t stride, | ||
| 145 : | syskin | 1565 | const int rounding) |
| 146 : | edgomez | 851 | { /* uv-block-based compensation */ |
| 147 : | |||
| 148 : | syskin | 1565 | transfer_8to16sub(coeff, Cur->u + 8 * j * stride + 8 * i, |
| 149 : | interpolate8x8_switch2(temp, Ref->u, 8 * i, 8 * j, | ||
| 150 : | dx, dy, stride, rounding), | ||
| 151 : | stride); | ||
| 152 : | transfer_8to16sub(coeff + 64, Cur->v + 8 * j * stride + 8 * i, | ||
| 153 : | interpolate8x8_switch2(temp, Ref->v, 8 * i, 8 * j, | ||
| 154 : | dx, dy, stride, rounding), | ||
| 155 : | stride); | ||
| 156 : | edgomez | 851 | } |
| 157 : | |||
| 158 : | edgomez | 195 | void |
| 159 : | MBMotionCompensation(MACROBLOCK * const mb, | ||
| 160 : | syskin | 935 | const uint32_t i, |
| 161 : | const uint32_t j, | ||
| 162 : | const IMAGE * const ref, | ||
| 163 : | const IMAGE * const refh, | ||
| 164 : | const IMAGE * const refv, | ||
| 165 : | const IMAGE * const refhv, | ||
| 166 : | const IMAGE * const refGMC, | ||
| 167 : | IMAGE * const cur, | ||
| 168 : | int16_t * dct_codes, | ||
| 169 : | const uint32_t width, | ||
| 170 : | const uint32_t height, | ||
| 171 : | const uint32_t edged_width, | ||
| 172 : | const int32_t quarterpel, | ||
| 173 : | Isibaar | 1913 | const int32_t rounding, |
| 174 : | Isibaar | 1928 | uint8_t * const tmp) |
| 175 : | Isibaar | 3 | { |
| 176 : | edgomez | 851 | int32_t dx; |
| 177 : | int32_t dy; | ||
| 178 : | Isibaar | 3 | |
| 179 : | syskin | 1565 | if (mb->mode == MODE_NOT_CODED) { /* quick copy for early SKIP */ |
| 180 : | edgomez | 851 | /* early SKIP is only activated in P-VOPs, not in S-VOPs, so mcsel can never be 1 */ |
| 181 : | Isibaar | 3 | |
| 182 : | edgomez | 851 | transfer16x16_copy(cur->y + 16 * (i + j * edged_width), |
| 183 : | syskin | 935 | ref->y + 16 * (i + j * edged_width), |
| 184 : | edged_width); | ||
| 185 : | |||
| 186 : | edgomez | 851 | transfer8x8_copy(cur->u + 8 * (i + j * edged_width/2), |
| 187 : | ref->u + 8 * (i + j * edged_width/2), | ||
| 188 : | edged_width / 2); | ||
| 189 : | transfer8x8_copy(cur->v + 8 * (i + j * edged_width/2), | ||
| 190 : | ref->v + 8 * (i + j * edged_width/2), | ||
| 191 : | edged_width / 2); | ||
| 192 : | return; | ||
| 193 : | } | ||
| 194 : | Isibaar | 3 | |
| 195 : | syskin | 935 | if ((mb->mode == MODE_NOT_CODED || mb->mode == MODE_INTER |
| 196 : | edgomez | 851 | || mb->mode == MODE_INTER_Q)) { |
| 197 : | chl | 437 | |
| 198 : | edgomez | 851 | if (mb->mcsel) { |
| 199 : | syskin | 935 | |
| 200 : | edgomez | 851 | /* call normal routine once, easier than "if (mcsel)"ing all the time */ |
| 201 : | syskin | 935 | |
| 202 : | edgomez | 851 | transfer_8to16sub(&dct_codes[0*64], cur->y + 16*j*edged_width + 16*i, |
| 203 : | syskin | 935 | refGMC->y + 16*j*edged_width + 16*i, edged_width); |
| 204 : | edgomez | 851 | transfer_8to16sub(&dct_codes[1*64], cur->y + 16*j*edged_width + 16*i+8, |
| 205 : | syskin | 935 | refGMC->y + 16*j*edged_width + 16*i+8, edged_width); |
| 206 : | edgomez | 851 | transfer_8to16sub(&dct_codes[2*64], cur->y + (16*j+8)*edged_width + 16*i, |
| 207 : | syskin | 935 | refGMC->y + (16*j+8)*edged_width + 16*i, edged_width); |
| 208 : | edgomez | 851 | transfer_8to16sub(&dct_codes[3*64], cur->y + (16*j+8)*edged_width + 16*i+8, |
| 209 : | syskin | 935 | refGMC->y + (16*j+8)*edged_width + 16*i+8, edged_width); |
| 210 : | chl | 437 | |
| 211 : | edgomez | 851 | transfer_8to16sub(&dct_codes[4 * 64], cur->u + 8 *j*edged_width/2 + 8*i, |
| 212 : | refGMC->u + 8 *j*edged_width/2 + 8*i, edged_width/2); | ||
| 213 : | |||
| 214 : | transfer_8to16sub(&dct_codes[5 * 64], cur->v + 8*j* edged_width/2 + 8*i, | ||
| 215 : | refGMC->v + 8*j* edged_width/2 + 8*i, edged_width/2); | ||
| 216 : | |||
| 217 : | return; | ||
| 218 : | } | ||
| 219 : | |||
| 220 : | /* ordinary compensation */ | ||
| 221 : | syskin | 935 | |
| 222 : | edgomez | 851 | dx = (quarterpel ? mb->qmvs[0].x : mb->mvs[0].x); |
| 223 : | dy = (quarterpel ? mb->qmvs[0].y : mb->mvs[0].y); | ||
| 224 : | |||
| 225 : | compensate16x16_interpolate(&dct_codes[0 * 64], cur->y, ref->y, refh->y, | ||
| 226 : | refv->y, refhv->y, tmp, 16 * i, 16 * j, dx, dy, | ||
| 227 : | syskin | 1565 | edged_width, quarterpel, rounding); |
| 228 : | edgomez | 851 | |
| 229 : | syskin | 935 | if (quarterpel) { dx /= 2; dy /= 2; } |
| 230 : | |||
| 231 : | edgomez | 851 | dx = (dx >> 1) + roundtab_79[dx & 0x3]; |
| 232 : | dy = (dy >> 1) + roundtab_79[dy & 0x3]; | ||
| 233 : | |||
| 234 : | edgomez | 1382 | } else { /* mode == MODE_INTER4V */ |
| 235 : | edgomez | 851 | int k, sumx = 0, sumy = 0; |
| 236 : | const VECTOR * const mvs = (quarterpel ? mb->qmvs : mb->mvs); | ||
| 237 : | |||
| 238 : | for (k = 0; k < 4; k++) { | ||
| 239 : | dx = mvs[k].x; | ||
| 240 : | dy = mvs[k].y; | ||
| 241 : | syskin | 935 | sumx += quarterpel ? dx/2 : dx; |
| 242 : | sumy += quarterpel ? dy/2 : dy; | ||
| 243 : | edgomez | 851 | |
| 244 : | compensate8x8_interpolate(&dct_codes[k * 64], cur->y, ref->y, refh->y, | ||
| 245 : | refv->y, refhv->y, tmp, 16 * i + 8*(k&1), 16 * j + 8*(k>>1), dx, | ||
| 246 : | syskin | 1565 | dy, edged_width, quarterpel, rounding); |
| 247 : | edgomez | 851 | } |
| 248 : | dx = (sumx >> 3) + roundtab_76[sumx & 0xf]; | ||
| 249 : | dy = (sumy >> 3) + roundtab_76[sumy & 0xf]; | ||
| 250 : | } | ||
| 251 : | |||
| 252 : | CompensateChroma(dx, dy, i, j, cur, ref, tmp, | ||
| 253 : | syskin | 1565 | &dct_codes[4 * 64], edged_width / 2, rounding); |
| 254 : | edgomez | 851 | } |
| 255 : | |||
| 256 : | |||
| 257 : | void | ||
| 258 : | MBMotionCompensationBVOP(MBParam * pParam, | ||
| 259 : | MACROBLOCK * const mb, | ||
| 260 : | const uint32_t i, | ||
| 261 : | const uint32_t j, | ||
| 262 : | IMAGE * const cur, | ||
| 263 : | const IMAGE * const f_ref, | ||
| 264 : | const IMAGE * const f_refh, | ||
| 265 : | const IMAGE * const f_refv, | ||
| 266 : | const IMAGE * const f_refhv, | ||
| 267 : | const IMAGE * const b_ref, | ||
| 268 : | const IMAGE * const b_refh, | ||
| 269 : | const IMAGE * const b_refv, | ||
| 270 : | const IMAGE * const b_refhv, | ||
| 271 : | Isibaar | 1913 | int16_t * dct_codes, |
| 272 : | Isibaar | 1928 | uint8_t * const tmp) |
| 273 : | edgomez | 851 | { |
| 274 : | const uint32_t edged_width = pParam->edged_width; | ||
| 275 : | int32_t dx, dy, b_dx, b_dy, sumx, sumy, b_sumx, b_sumy; | ||
| 276 : | int k; | ||
| 277 : | edgomez | 1382 | const int quarterpel = pParam->vol_flags & XVID_VOL_QUARTERPEL; |
| 278 : | edgomez | 851 | const uint8_t * ptr1, * ptr2; |
| 279 : | const VECTOR * const fmvs = (quarterpel ? mb->qmvs : mb->mvs); | ||
| 280 : | const VECTOR * const bmvs = (quarterpel ? mb->b_qmvs : mb->b_mvs); | ||
| 281 : | |||
| 282 : | switch (mb->mode) { | ||
| 283 : | case MODE_FORWARD: | ||
| 284 : | dx = fmvs->x; dy = fmvs->y; | ||
| 285 : | |||
| 286 : | compensate16x16_interpolate(&dct_codes[0 * 64], cur->y, f_ref->y, f_refh->y, | ||
| 287 : | f_refv->y, f_refhv->y, tmp, 16 * i, 16 * j, dx, | ||
| 288 : | syskin | 1565 | dy, edged_width, quarterpel, 0); |
| 289 : | edgomez | 851 | |
| 290 : | if (quarterpel) { dx /= 2; dy /= 2; } | ||
| 291 : | |||
| 292 : | CompensateChroma( (dx >> 1) + roundtab_79[dx & 0x3], | ||
| 293 : | (dy >> 1) + roundtab_79[dy & 0x3], | ||
| 294 : | i, j, cur, f_ref, tmp, | ||
| 295 : | syskin | 1565 | &dct_codes[4 * 64], edged_width / 2, 0); |
| 296 : | edgomez | 851 | |
| 297 : | return; | ||
| 298 : | |||
| 299 : | case MODE_BACKWARD: | ||
| 300 : | b_dx = bmvs->x; b_dy = bmvs->y; | ||
| 301 : | |||
| 302 : | syskin | 935 | compensate16x16_interpolate(&dct_codes[0 * 64], cur->y, b_ref->y, b_refh->y, |
| 303 : | edgomez | 851 | b_refv->y, b_refhv->y, tmp, 16 * i, 16 * j, b_dx, |
| 304 : | syskin | 1565 | b_dy, edged_width, quarterpel, 0); |
| 305 : | edgomez | 851 | |
| 306 : | if (quarterpel) { b_dx /= 2; b_dy /= 2; } | ||
| 307 : | |||
| 308 : | CompensateChroma( (b_dx >> 1) + roundtab_79[b_dx & 0x3], | ||
| 309 : | (b_dy >> 1) + roundtab_79[b_dy & 0x3], | ||
| 310 : | i, j, cur, b_ref, tmp, | ||
| 311 : | syskin | 1565 | &dct_codes[4 * 64], edged_width / 2, 0); |
| 312 : | edgomez | 851 | |
| 313 : | return; | ||
| 314 : | |||
| 315 : | syskin | 1565 | case MODE_INTERPOLATE: |
| 316 : | edgomez | 851 | case MODE_DIRECT_NO4V: |
| 317 : | dx = fmvs->x; dy = fmvs->y; | ||
| 318 : | b_dx = bmvs->x; b_dy = bmvs->y; | ||
| 319 : | |||
| 320 : | if (quarterpel) { | ||
| 321 : | syskin | 935 | |
| 322 : | edgomez | 851 | if ((dx&3) | (dy&3)) { |
| 323 : | interpolate16x16_quarterpel(tmp - i * 16 - j * 16 * edged_width, | ||
| 324 : | (uint8_t *) f_ref->y, tmp + 32, | ||
| 325 : | tmp + 64, tmp + 96, 16*i, 16*j, dx, dy, edged_width, 0); | ||
| 326 : | ptr1 = tmp; | ||
| 327 : | edgomez | 1556 | } else ptr1 = f_ref->y + (16*(int)j + dy/4)*(int)edged_width + 16*(int)i + dx/4; /* fullpixel position */ |
| 328 : | edgomez | 851 | |
| 329 : | if ((b_dx&3) | (b_dy&3)) { | ||
| 330 : | interpolate16x16_quarterpel(tmp - i * 16 - j * 16 * edged_width + 16, | ||
| 331 : | (uint8_t *) b_ref->y, tmp + 32, | ||
| 332 : | tmp + 64, tmp + 96, 16*i, 16*j, b_dx, b_dy, edged_width, 0); | ||
| 333 : | ptr2 = tmp + 16; | ||
| 334 : | edgomez | 1556 | } else ptr2 = b_ref->y + (16*(int)j + b_dy/4)*(int)edged_width + 16*(int)i + b_dx/4; /* fullpixel position */ |
| 335 : | edgomez | 851 | |
| 336 : | b_dx /= 2; | ||
| 337 : | b_dy /= 2; | ||
| 338 : | dx /= 2; | ||
| 339 : | dy /= 2; | ||
| 340 : | |||
| 341 : | } else { | ||
| 342 : | ptr1 = get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, | ||
| 343 : | i, j, 16, dx, dy, edged_width); | ||
| 344 : | |||
| 345 : | ptr2 = get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, | ||
| 346 : | i, j, 16, b_dx, b_dy, edged_width); | ||
| 347 : | } | ||
| 348 : | for (k = 0; k < 4; k++) | ||
| 349 : | transfer_8to16sub2(&dct_codes[k * 64], | ||
| 350 : | cur->y + (i * 16+(k&1)*8) + (j * 16+((k>>1)*8)) * edged_width, | ||
| 351 : | ptr1 + (k&1)*8 + (k>>1)*8*edged_width, | ||
| 352 : | ptr2 + (k&1)*8 + (k>>1)*8*edged_width, edged_width); | ||
| 353 : | |||
| 354 : | |||
| 355 : | dx = (dx >> 1) + roundtab_79[dx & 0x3]; | ||
| 356 : | dy = (dy >> 1) + roundtab_79[dy & 0x3]; | ||
| 357 : | |||
| 358 : | b_dx = (b_dx >> 1) + roundtab_79[b_dx & 0x3]; | ||
| 359 : | b_dy = (b_dy >> 1) + roundtab_79[b_dy & 0x3]; | ||
| 360 : | |||
| 361 : | break; | ||
| 362 : | syskin | 935 | |
| 363 : | edgomez | 1382 | default: /* MODE_DIRECT (or MODE_DIRECT_NONE_MV in case of bframes decoding) */ |
| 364 : | edgomez | 851 | sumx = sumy = b_sumx = b_sumy = 0; |
| 365 : | |||
| 366 : | for (k = 0; k < 4; k++) { | ||
| 367 : | syskin | 935 | |
| 368 : | edgomez | 851 | dx = fmvs[k].x; dy = fmvs[k].y; |
| 369 : | b_dx = bmvs[k].x; b_dy = bmvs[k].y; | ||
| 370 : | |||
| 371 : | if (quarterpel) { | ||
| 372 : | sumx += dx/2; sumy += dy/2; | ||
| 373 : | b_sumx += b_dx/2; b_sumy += b_dy/2; | ||
| 374 : | |||
| 375 : | if ((dx&3) | (dy&3)) { | ||
| 376 : | interpolate8x8_quarterpel(tmp - (i * 16+(k&1)*8) - (j * 16+((k>>1)*8)) * edged_width, | ||
| 377 : | syskin | 935 | (uint8_t *) f_ref->y, |
| 378 : | tmp + 32, tmp + 64, tmp + 96, | ||
| 379 : | edgomez | 851 | 16*i + (k&1)*8, 16*j + (k>>1)*8, dx, dy, edged_width, 0); |
| 380 : | ptr1 = tmp; | ||
| 381 : | edgomez | 1556 | } else ptr1 = f_ref->y + (16*(int)j + (k>>1)*8 + dy/4)*(int)edged_width + 16*(int)i + (k&1)*8 + dx/4; |
| 382 : | edgomez | 851 | |
| 383 : | if ((b_dx&3) | (b_dy&3)) { | ||
| 384 : | interpolate8x8_quarterpel(tmp - (i * 16+(k&1)*8) - (j * 16+((k>>1)*8)) * edged_width + 16, | ||
| 385 : | (uint8_t *) b_ref->y, | ||
| 386 : | syskin | 935 | tmp + 16, tmp + 32, tmp + 48, |
| 387 : | edgomez | 851 | 16*i + (k&1)*8, 16*j + (k>>1)*8, b_dx, b_dy, edged_width, 0); |
| 388 : | ptr2 = tmp + 16; | ||
| 389 : | edgomez | 1556 | } else ptr2 = b_ref->y + (16*(int)j + (k>>1)*8 + b_dy/4)*(int)edged_width + 16*(int)i + (k&1)*8 + b_dx/4; |
| 390 : | edgomez | 851 | } else { |
| 391 : | sumx += dx; sumy += dy; | ||
| 392 : | b_sumx += b_dx; b_sumy += b_dy; | ||
| 393 : | |||
| 394 : | syskin | 935 | ptr1 = get_ref(f_ref->y, f_refh->y, f_refv->y, f_refhv->y, |
| 395 : | edgomez | 851 | 2*i + (k&1), 2*j + (k>>1), 8, dx, dy, edged_width); |
| 396 : | syskin | 935 | ptr2 = get_ref(b_ref->y, b_refh->y, b_refv->y, b_refhv->y, |
| 397 : | edgomez | 851 | 2*i + (k&1), 2*j + (k>>1), 8, b_dx, b_dy, edged_width); |
| 398 : | } | ||
| 399 : | transfer_8to16sub2(&dct_codes[k * 64], | ||
| 400 : | cur->y + (i * 16+(k&1)*8) + (j * 16+((k>>1)*8)) * edged_width, | ||
| 401 : | ptr1, ptr2, edged_width); | ||
| 402 : | syskin | 935 | |
| 403 : | edgomez | 851 | } |
| 404 : | |||
| 405 : | dx = (sumx >> 3) + roundtab_76[sumx & 0xf]; | ||
| 406 : | dy = (sumy >> 3) + roundtab_76[sumy & 0xf]; | ||
| 407 : | b_dx = (b_sumx >> 3) + roundtab_76[b_sumx & 0xf]; | ||
| 408 : | b_dy = (b_sumy >> 3) + roundtab_76[b_sumy & 0xf]; | ||
| 409 : | |||
| 410 : | break; | ||
| 411 : | } | ||
| 412 : | |||
| 413 : | syskin | 1565 | /* block-based chroma interpolation for direct and interpolate modes */ |
| 414 : | edgomez | 851 | transfer_8to16sub2(&dct_codes[4 * 64], |
| 415 : | cur->u + (j * 8) * edged_width / 2 + (i * 8), | ||
| 416 : | interpolate8x8_switch2(tmp, b_ref->u, 8 * i, 8 * j, | ||
| 417 : | b_dx, b_dy, edged_width / 2, 0), | ||
| 418 : | interpolate8x8_switch2(tmp + 8, f_ref->u, 8 * i, 8 * j, | ||
| 419 : | dx, dy, edged_width / 2, 0), | ||
| 420 : | edged_width / 2); | ||
| 421 : | |||
| 422 : | transfer_8to16sub2(&dct_codes[5 * 64], | ||
| 423 : | cur->v + (j * 8) * edged_width / 2 + (i * 8), | ||
| 424 : | interpolate8x8_switch2(tmp, b_ref->v, 8 * i, 8 * j, | ||
| 425 : | b_dx, b_dy, edged_width / 2, 0), | ||
| 426 : | interpolate8x8_switch2(tmp + 8, f_ref->v, 8 * i, 8 * j, | ||
| 427 : | dx, dy, edged_width / 2, 0), | ||
| 428 : | edged_width / 2); | ||
| 429 : | } |
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |