Parent Directory
| 
Revision Log
Revision 1632 - (view) (download)
| 1 : | edgomez | 212 | /***************************************************************************** |
| 2 : | edgomez | 195 | * |
| 3 : | edgomez | 212 | * XVID MPEG-4 VIDEO CODEC |
| 4 : | edgomez | 851 | * - 8bit<->16bit transfer - |
| 5 : | edgomez | 195 | * |
| 6 : | edgomez | 1382 | * Copyright(C) 2001-2003 Peter Ross <pross@xvid.org> |
| 7 : | * | ||
| 8 : | edgomez | 851 | * This program is free software ; you can redistribute it and/or modify |
| 9 : | * it under the terms of the GNU General Public License as published by | ||
| 10 : | * the Free Software Foundation ; either version 2 of the License, or | ||
| 11 : | edgomez | 212 | * (at your option) any later version. |
| 12 : | edgomez | 195 | * |
| 13 : | edgomez | 212 | * This program is distributed in the hope that it will be useful, |
| 14 : | edgomez | 851 | * but WITHOUT ANY WARRANTY ; without even the implied warranty of |
| 15 : | edgomez | 212 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 : | * GNU General Public License for more details. | ||
| 17 : | edgomez | 195 | * |
| 18 : | edgomez | 212 | * You should have received a copy of the GNU General Public License |
| 19 : | edgomez | 851 | * along with this program ; if not, write to the Free Software |
| 20 : | edgomez | 212 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA |
| 21 : | edgomez | 195 | * |
| 22 : | suxen_drol | 1632 | * $Id: mem_transfer.c,v 1.16 2005-09-13 12:12:15 suxen_drol Exp $ |
| 23 : | edgomez | 195 | * |
| 24 : | edgomez | 212 | ****************************************************************************/ |
| 25 : | edgomez | 195 | |
| 26 : | edgomez | 212 | #include "../global.h" |
| 27 : | #include "mem_transfer.h" | ||
| 28 : | edgomez | 195 | |
| 29 : | edgomez | 212 | /* Function pointers - Initialized in the xvid.c module */ |
| 30 : | |||
| 31 : | TRANSFER_8TO16COPY_PTR transfer_8to16copy; | ||
| 32 : | TRANSFER_16TO8COPY_PTR transfer_16to8copy; | ||
| 33 : | |||
| 34 : | edgomez | 195 | TRANSFER_8TO16SUB_PTR transfer_8to16sub; |
| 35 : | edgomez | 851 | TRANSFER_8TO16SUBRO_PTR transfer_8to16subro; |
| 36 : | edgomez | 195 | TRANSFER_8TO16SUB2_PTR transfer_8to16sub2; |
| 37 : | syskin | 1583 | TRANSFER_8TO16SUB2RO_PTR transfer_8to16sub2ro; |
| 38 : | edgomez | 195 | TRANSFER_16TO8ADD_PTR transfer_16to8add; |
| 39 : | |||
| 40 : | edgomez | 212 | TRANSFER8X8_COPY_PTR transfer8x8_copy; |
| 41 : | suxen_drol | 1632 | TRANSFER8X4_COPY_PTR transfer8x4_copy; |
| 42 : | edgomez | 195 | |
| 43 : | Skal | 1620 | #define USE_REFERENCE_C |
| 44 : | edgomez | 212 | |
| 45 : | edgomez | 195 | /***************************************************************************** |
| 46 : | * | ||
| 47 : | edgomez | 212 | * All these functions are used to transfer data from a 8 bit data array |
| 48 : | edgomez | 195 | * to a 16 bit data array. |
| 49 : | edgomez | 212 | * |
| 50 : | edgomez | 195 | * This is typically used during motion compensation, that's why some |
| 51 : | * functions also do the addition/substraction of another buffer during the | ||
| 52 : | * so called transfer. | ||
| 53 : | * | ||
| 54 : | edgomez | 212 | ****************************************************************************/ |
| 55 : | |||
| 56 : | /* | ||
| 57 : | * SRC - the source buffer | ||
| 58 : | * DST - the destination buffer | ||
| 59 : | * | ||
| 60 : | * Then the function does the 8->16 bit transfer and this serie of operations : | ||
| 61 : | * | ||
| 62 : | * SRC (8bit) = SRC | ||
| 63 : | * DST (16bit) = SRC | ||
| 64 : | edgomez | 195 | */ |
| 65 : | edgomez | 212 | void |
| 66 : | transfer_8to16copy_c(int16_t * const dst, | ||
| 67 : | const uint8_t * const src, | ||
| 68 : | uint32_t stride) | ||
| 69 : | { | ||
| 70 : | Skal | 1620 | int i, j; |
| 71 : | edgomez | 212 | for (j = 0; j < 8; j++) { |
| 72 : | for (i = 0; i < 8; i++) { | ||
| 73 : | edgomez | 195 | dst[j * 8 + i] = (int16_t) src[j * stride + i]; |
| 74 : | Isibaar | 1627 | } |
| 75 : | edgomez | 195 | } |
| 76 : | } | ||
| 77 : | |||
| 78 : | /* | ||
| 79 : | * SRC - the source buffer | ||
| 80 : | * DST - the destination buffer | ||
| 81 : | * | ||
| 82 : | * Then the function does the 8->16 bit transfer and this serie of operations : | ||
| 83 : | * | ||
| 84 : | * SRC (16bit) = SRC | ||
| 85 : | * DST (8bit) = max(min(SRC, 255), 0) | ||
| 86 : | */ | ||
| 87 : | edgomez | 212 | void |
| 88 : | transfer_16to8copy_c(uint8_t * const dst, | ||
| 89 : | const int16_t * const src, | ||
| 90 : | uint32_t stride) | ||
| 91 : | { | ||
| 92 : | Skal | 1620 | int i, j; |
| 93 : | edgomez | 195 | |
| 94 : | edgomez | 212 | for (j = 0; j < 8; j++) { |
| 95 : | for (i = 0; i < 8; i++) { | ||
| 96 : | Skal | 1620 | #ifdef USE_REFERENCE_C |
| 97 : | edgomez | 195 | int16_t pixel = src[j * 8 + i]; |
| 98 : | |||
| 99 : | if (pixel < 0) { | ||
| 100 : | pixel = 0; | ||
| 101 : | } else if (pixel > 255) { | ||
| 102 : | pixel = 255; | ||
| 103 : | } | ||
| 104 : | dst[j * stride + i] = (uint8_t) pixel; | ||
| 105 : | Skal | 1620 | #else |
| 106 : | const int16_t pixel = src[j * 8 + i]; | ||
| 107 : | const uint8_t value = (uint8_t)( (pixel&~255) ? (-pixel)>>(8*sizeof(pixel)-1) : pixel ); | ||
| 108 : | dst[j*stride + i] = value; | ||
| 109 : | #endif | ||
| 110 : | } | ||
| 111 : | edgomez | 195 | } |
| 112 : | } | ||
| 113 : | |||
| 114 : | |||
| 115 : | |||
| 116 : | |||
| 117 : | /* | ||
| 118 : | * C - the current buffer | ||
| 119 : | * R - the reference buffer | ||
| 120 : | * DCT - the dct coefficient buffer | ||
| 121 : | * | ||
| 122 : | * Then the function does the 8->16 bit transfer and this serie of operations : | ||
| 123 : | * | ||
| 124 : | * R (8bit) = R | ||
| 125 : | * C (8bit) = R | ||
| 126 : | * DCT (16bit) = C - R | ||
| 127 : | edgomez | 212 | */ |
| 128 : | void | ||
| 129 : | transfer_8to16sub_c(int16_t * const dct, | ||
| 130 : | uint8_t * const cur, | ||
| 131 : | const uint8_t * ref, | ||
| 132 : | const uint32_t stride) | ||
| 133 : | { | ||
| 134 : | Skal | 1620 | int i, j; |
| 135 : | edgomez | 212 | |
| 136 : | for (j = 0; j < 8; j++) { | ||
| 137 : | edgomez | 195 | for (i = 0; i < 8; i++) { |
| 138 : | Skal | 1620 | const uint8_t c = cur[j * stride + i]; |
| 139 : | const uint8_t r = ref[j * stride + i]; | ||
| 140 : | edgomez | 195 | |
| 141 : | cur[j * stride + i] = r; | ||
| 142 : | dct[j * 8 + i] = (int16_t) c - (int16_t) r; | ||
| 143 : | } | ||
| 144 : | } | ||
| 145 : | } | ||
| 146 : | |||
| 147 : | |||
| 148 : | edgomez | 851 | void |
| 149 : | transfer_8to16subro_c(int16_t * const dct, | ||
| 150 : | const uint8_t * const cur, | ||
| 151 : | const uint8_t * ref, | ||
| 152 : | const uint32_t stride) | ||
| 153 : | { | ||
| 154 : | Skal | 1620 | int i, j; |
| 155 : | edgomez | 851 | |
| 156 : | for (j = 0; j < 8; j++) { | ||
| 157 : | for (i = 0; i < 8; i++) { | ||
| 158 : | Skal | 1620 | const uint8_t c = cur[j * stride + i]; |
| 159 : | const uint8_t r = ref[j * stride + i]; | ||
| 160 : | edgomez | 851 | dct[j * 8 + i] = (int16_t) c - (int16_t) r; |
| 161 : | } | ||
| 162 : | } | ||
| 163 : | } | ||
| 164 : | |||
| 165 : | |||
| 166 : | |||
| 167 : | edgomez | 195 | /* |
| 168 : | * C - the current buffer | ||
| 169 : | * R1 - the 1st reference buffer | ||
| 170 : | * R2 - the 2nd reference buffer | ||
| 171 : | * DCT - the dct coefficient buffer | ||
| 172 : | * | ||
| 173 : | * Then the function does the 8->16 bit transfer and this serie of operations : | ||
| 174 : | * | ||
| 175 : | * R1 (8bit) = R1 | ||
| 176 : | * R2 (8bit) = R2 | ||
| 177 : | edgomez | 1382 | * R (temp) = min((R1 + R2)/2, 255) |
| 178 : | * DCT (16bit)= C - R | ||
| 179 : | * C (8bit) = R | ||
| 180 : | edgomez | 212 | */ |
| 181 : | void | ||
| 182 : | transfer_8to16sub2_c(int16_t * const dct, | ||
| 183 : | uint8_t * const cur, | ||
| 184 : | const uint8_t * ref1, | ||
| 185 : | const uint8_t * ref2, | ||
| 186 : | const uint32_t stride) | ||
| 187 : | { | ||
| 188 : | uint32_t i, j; | ||
| 189 : | |||
| 190 : | edgomez | 195 | for (j = 0; j < 8; j++) { |
| 191 : | for (i = 0; i < 8; i++) { | ||
| 192 : | Skal | 1620 | const uint8_t c = cur[j * stride + i]; |
| 193 : | const uint8_t r = (ref1[j * stride + i] + ref2[j * stride + i] + 1) >> 1; | ||
| 194 : | edgomez | 1382 | cur[j * stride + i] = r; |
| 195 : | edgomez | 195 | dct[j * 8 + i] = (int16_t) c - (int16_t) r; |
| 196 : | } | ||
| 197 : | } | ||
| 198 : | } | ||
| 199 : | |||
| 200 : | syskin | 1583 | void |
| 201 : | transfer_8to16sub2ro_c(int16_t * const dct, | ||
| 202 : | const uint8_t * const cur, | ||
| 203 : | const uint8_t * ref1, | ||
| 204 : | const uint8_t * ref2, | ||
| 205 : | const uint32_t stride) | ||
| 206 : | { | ||
| 207 : | uint32_t i, j; | ||
| 208 : | edgomez | 195 | |
| 209 : | syskin | 1583 | for (j = 0; j < 8; j++) { |
| 210 : | for (i = 0; i < 8; i++) { | ||
| 211 : | Skal | 1620 | const uint8_t c = cur[j * stride + i]; |
| 212 : | const uint8_t r = (ref1[j * stride + i] + ref2[j * stride + i] + 1) >> 1; | ||
| 213 : | syskin | 1583 | dct[j * 8 + i] = (int16_t) c - (int16_t) r; |
| 214 : | } | ||
| 215 : | } | ||
| 216 : | } | ||
| 217 : | |||
| 218 : | |||
| 219 : | edgomez | 195 | /* |
| 220 : | * SRC - the source buffer | ||
| 221 : | * DST - the destination buffer | ||
| 222 : | * | ||
| 223 : | * Then the function does the 16->8 bit transfer and this serie of operations : | ||
| 224 : | * | ||
| 225 : | * SRC (16bit) = SRC | ||
| 226 : | * DST (8bit) = max(min(DST+SRC, 255), 0) | ||
| 227 : | */ | ||
| 228 : | edgomez | 212 | void |
| 229 : | transfer_16to8add_c(uint8_t * const dst, | ||
| 230 : | const int16_t * const src, | ||
| 231 : | uint32_t stride) | ||
| 232 : | { | ||
| 233 : | Skal | 1620 | int i, j; |
| 234 : | edgomez | 212 | |
| 235 : | for (j = 0; j < 8; j++) { | ||
| 236 : | for (i = 0; i < 8; i++) { | ||
| 237 : | Skal | 1620 | #ifdef USE_REFERENCE_C |
| 238 : | edgomez | 195 | int16_t pixel = (int16_t) dst[j * stride + i] + src[j * 8 + i]; |
| 239 : | |||
| 240 : | if (pixel < 0) { | ||
| 241 : | pixel = 0; | ||
| 242 : | } else if (pixel > 255) { | ||
| 243 : | pixel = 255; | ||
| 244 : | } | ||
| 245 : | dst[j * stride + i] = (uint8_t) pixel; | ||
| 246 : | Skal | 1620 | #else |
| 247 : | const int16_t pixel = (int16_t) dst[j * stride + i] + src[j * 8 + i]; | ||
| 248 : | const uint8_t value = (uint8_t)( (pixel&~255) ? (-pixel)>>(8*sizeof(pixel)-1) : pixel ); | ||
| 249 : | dst[j*stride + i] = value; | ||
| 250 : | #endif | ||
| 251 : | |||
| 252 : | edgomez | 195 | } |
| 253 : | } | ||
| 254 : | } | ||
| 255 : | |||
| 256 : | /* | ||
| 257 : | * SRC - the source buffer | ||
| 258 : | * DST - the destination buffer | ||
| 259 : | * | ||
| 260 : | * Then the function does the 8->8 bit transfer and this serie of operations : | ||
| 261 : | * | ||
| 262 : | * SRC (8bit) = SRC | ||
| 263 : | * DST (8bit) = SRC | ||
| 264 : | */ | ||
| 265 : | edgomez | 212 | void |
| 266 : | transfer8x8_copy_c(uint8_t * const dst, | ||
| 267 : | const uint8_t * const src, | ||
| 268 : | const uint32_t stride) | ||
| 269 : | { | ||
| 270 : | Skal | 1620 | int j, i; |
| 271 : | edgomez | 195 | |
| 272 : | Isibaar | 1612 | for (j = 0; j < 8; ++j) { |
| 273 : | uint8_t *d = dst + j*stride; | ||
| 274 : | const uint8_t *s = src + j*stride; | ||
| 275 : | |||
| 276 : | for (i = 0; i < 8; ++i) | ||
| 277 : | { | ||
| 278 : | *d++ = *s++; | ||
| 279 : | } | ||
| 280 : | edgomez | 195 | } |
| 281 : | } | ||
| 282 : | suxen_drol | 1632 | |
| 283 : | /* | ||
| 284 : | * SRC - the source buffer | ||
| 285 : | * DST - the destination buffer | ||
| 286 : | * | ||
| 287 : | * Then the function does the 8->8 bit transfer and this serie of operations : | ||
| 288 : | * | ||
| 289 : | * SRC (8bit) = SRC | ||
| 290 : | * DST (8bit) = SRC | ||
| 291 : | */ | ||
| 292 : | void | ||
| 293 : | transfer8x4_copy_c(uint8_t * const dst, | ||
| 294 : | const uint8_t * const src, | ||
| 295 : | const uint32_t stride) | ||
| 296 : | { | ||
| 297 : | uint32_t j; | ||
| 298 : | |||
| 299 : | for (j = 0; j < 4; j++) { | ||
| 300 : | uint32_t *d= (uint32_t*)(dst + j*stride); | ||
| 301 : | const uint32_t *s = (const uint32_t*)(src + j*stride); | ||
| 302 : | *(d+0) = *(s+0); | ||
| 303 : | *(d+1) = *(s+1); | ||
| 304 : | } | ||
| 305 : | } |
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |