Parent Directory
| 
Revision Log
Revision 1.2 - (view) (download)
| 1 : | Isibaar | 1.1 | ;/************************************************************************** |
| 2 : | ; * | ||
| 3 : | ; * XVID MPEG-4 VIDEO CODEC | ||
| 4 : | ; * mmx 8bit<->16bit transfers | ||
| 5 : | ; * | ||
| 6 : | ; * This program is an implementation of a part of one or more MPEG-4 | ||
| 7 : | ; * Video tools as specified in ISO/IEC 14496-2 standard. Those intending | ||
| 8 : | ; * to use this software module in hardware or software products are | ||
| 9 : | ; * advised that its use may infringe existing patents or copyrights, and | ||
| 10 : | ; * any such use would be at such party's own risk. The original | ||
| 11 : | ; * developer of this software module and his/her company, and subsequent | ||
| 12 : | ; * editors and their companies, will have no liability for use of this | ||
| 13 : | ; * software or modifications or derivatives thereof. | ||
| 14 : | ; * | ||
| 15 : | ; * This program is free software; you can redistribute it and/or modify | ||
| 16 : | ; * it under the terms of the GNU General Public License as published by | ||
| 17 : | ; * the Free Software Foundation; either version 2 of the License, or | ||
| 18 : | ; * (at your option) any later version. | ||
| 19 : | ; * | ||
| 20 : | ; * This program is distributed in the hope that it will be useful, | ||
| 21 : | ; * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 22 : | ; * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | ||
| 23 : | ; * GNU General Public License for more details. | ||
| 24 : | ; * | ||
| 25 : | ; * You should have received a copy of the GNU General Public License | ||
| 26 : | ; * along with this program; if not, write to the Free Software | ||
| 27 : | ; * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. | ||
| 28 : | ; * | ||
| 29 : | ; *************************************************************************/ | ||
| 30 : | |||
| 31 : | ;/************************************************************************** | ||
| 32 : | ; * | ||
| 33 : | ; * History: | ||
| 34 : | ; * | ||
| 35 : | ; * 07.01.2002 merge functions from compensate_mmx; rename functions | ||
| 36 : | ; * 07.11.2001 initial version; (c)2001 peter ross <pross@cs.rmit.edu.au> | ||
| 37 : | ; * | ||
| 38 : | ; *************************************************************************/ | ||
| 39 : | |||
| 40 : | |||
| 41 : | bits 32 | ||
| 42 : | |||
| 43 : | %macro cglobal 1 | ||
| 44 : | %ifdef PREFIX | ||
| 45 : | global _%1 | ||
| 46 : | %define %1 _%1 | ||
| 47 : | %else | ||
| 48 : | global %1 | ||
| 49 : | %endif | ||
| 50 : | %endmacro | ||
| 51 : | |||
| 52 : | |||
| 53 : | section .text | ||
| 54 : | |||
| 55 : | |||
| 56 : | ;=========================================================================== | ||
| 57 : | ; | ||
| 58 : | ; void transfer_8to16copy_mmx(int16_t * const dst, | ||
| 59 : | ; const uint8_t * const src, | ||
| 60 : | ; uint32_t stride); | ||
| 61 : | ; | ||
| 62 : | ;=========================================================================== | ||
| 63 : | |||
| 64 : | align 16 | ||
| 65 : | cglobal transfer_8to16copy_mmx | ||
| 66 : | transfer_8to16copy_mmx | ||
| 67 : | |||
| 68 : | push esi | ||
| 69 : | push edi | ||
| 70 : | |||
| 71 : | mov edi, [esp + 8 + 4] ; dst | ||
| 72 : | mov esi, [esp + 8 + 8] ; src | ||
| 73 : | mov ecx, [esp + 8 + 12] ; stride | ||
| 74 : | |||
| 75 : | pxor mm7, mm7 ; mm7 = zero | ||
| 76 : | |||
| 77 : | mov eax, 8 | ||
| 78 : | |||
| 79 : | .loop | ||
| 80 : | movq mm0, [esi] | ||
| 81 : | movq mm1, mm0 | ||
| 82 : | punpcklbw mm0, mm7 ; mm01 = unpack([src]) | ||
| 83 : | punpckhbw mm1, mm7 | ||
| 84 : | |||
| 85 : | movq [edi], mm0 ; [dst] = mm01 | ||
| 86 : | movq [edi + 8], mm1 | ||
| 87 : | |||
| 88 : | add edi, 16 | ||
| 89 : | add esi, ecx | ||
| 90 : | dec eax | ||
| 91 : | jnz .loop | ||
| 92 : | |||
| 93 : | pop edi | ||
| 94 : | pop esi | ||
| 95 : | |||
| 96 : | ret | ||
| 97 : | |||
| 98 : | |||
| 99 : | |||
| 100 : | ;=========================================================================== | ||
| 101 : | ; | ||
| 102 : | ; void transfer_16to8copy_mmx(uint8_t * const dst, | ||
| 103 : | ; const int16_t * const src, | ||
| 104 : | ; uint32_t stride); | ||
| 105 : | ; | ||
| 106 : | ;=========================================================================== | ||
| 107 : | |||
| 108 : | align 16 | ||
| 109 : | cglobal transfer_16to8copy_mmx | ||
| 110 : | transfer_16to8copy_mmx | ||
| 111 : | |||
| 112 : | push esi | ||
| 113 : | push edi | ||
| 114 : | |||
| 115 : | mov edi, [esp + 8 + 4] ; dst | ||
| 116 : | mov esi, [esp + 8 + 8] ; src | ||
| 117 : | mov ecx, [esp + 8 + 12] ; stride | ||
| 118 : | |||
| 119 : | mov eax, 8 | ||
| 120 : | |||
| 121 : | .loop | ||
| 122 : | movq mm0, [esi] | ||
| 123 : | packuswb mm0, [esi + 8] ; mm0 = pack([src]) | ||
| 124 : | |||
| 125 : | movq [edi], mm0 ; [dst] = mm0 | ||
| 126 : | |||
| 127 : | add esi, 16 | ||
| 128 : | add edi, ecx | ||
| 129 : | dec eax | ||
| 130 : | jnz .loop | ||
| 131 : | |||
| 132 : | pop edi | ||
| 133 : | pop esi | ||
| 134 : | |||
| 135 : | ret | ||
| 136 : | |||
| 137 : | |||
| 138 : | ;=========================================================================== | ||
| 139 : | ; | ||
| 140 : | ; void transfer_8to16sub_mmx(int16_t * const dct, | ||
| 141 : | ; uint8_t * const cur, | ||
| 142 : | ; const uint8_t * const ref, | ||
| 143 : | ; const uint32_t stride); | ||
| 144 : | ; | ||
| 145 : | ;=========================================================================== | ||
| 146 : | ;/************************************************************************** | ||
| 147 : | ; * | ||
| 148 : | ; * History: | ||
| 149 : | ; * | ||
| 150 : | ; * 27.12.2001 renamed from 'compensate' to 'transfer_8to16sub' | ||
| 151 : | ; * 02.12.2001 loop unrolled, code runs 10% faster now (Isibaar) | ||
| 152 : | ; * 30.11.2001 16 pixels are processed per iteration (Isibaar) | ||
| 153 : | ; * 30.11.2001 .text missing | ||
| 154 : | ; * 06.11.2001 inital version; (c)2001 peter ross <pross@cs.rmit.edu.au> | ||
| 155 : | ; * | ||
| 156 : | ; *************************************************************************/ | ||
| 157 : | |||
| 158 : | align 16 | ||
| 159 : | cglobal transfer_8to16sub_mmx | ||
| 160 : | transfer_8to16sub_mmx | ||
| 161 : | push esi | ||
| 162 : | push edi | ||
| 163 : | push ebx | ||
| 164 : | |||
| 165 : | mov edi, [esp + 12 + 4] ; dct [out] | ||
| 166 : | mov edx, [esp + 12 + 8] ; cur [in/out] | ||
| 167 : | mov esi, [esp + 12 + 12] ; ref [in] | ||
| 168 : | mov ecx, [esp + 12 + 16] ; stride [in] | ||
| 169 : | |||
| 170 : | mov eax, edx ; cur -> eax | ||
| 171 : | mov ebx, esi ; ref -> ebx | ||
| 172 : | add eax, ecx ; cur + stride | ||
| 173 : | add ebx, ecx ; ref + stride | ||
| 174 : | |||
| 175 : | shl ecx, 1 | ||
| 176 : | |||
| 177 : | pxor mm7, mm7 ; mm7 = zero | ||
| 178 : | |||
| 179 : | movq mm0, [edx] ; mm01 = [cur] | ||
| 180 : | movq mm1, mm0 | ||
| 181 : | |||
| 182 : | punpcklbw mm0, mm7 | ||
| 183 : | punpckhbw mm1, mm7 | ||
| 184 : | |||
| 185 : | movq mm4, [eax] | ||
| 186 : | movq mm5, mm4 | ||
| 187 : | |||
| 188 : | punpcklbw mm4, mm7 | ||
| 189 : | punpckhbw mm5, mm7 | ||
| 190 : | |||
| 191 : | movq mm2, [esi] ; mm23 = [ref] | ||
| 192 : | movq mm3, mm2 | ||
| 193 : | |||
| 194 : | movq mm6, [ebx] | ||
| 195 : | |||
| 196 : | movq [edx], mm2 ; [cur] = [ref] | ||
| 197 : | movq [eax], mm6 | ||
| 198 : | |||
| 199 : | punpcklbw mm2, mm7 | ||
| 200 : | punpckhbw mm3, mm7 | ||
| 201 : | |||
| 202 : | psubsw mm0, mm2 ; mm01 -= mm23 | ||
| 203 : | |||
| 204 : | movq mm2, mm6 | ||
| 205 : | |||
| 206 : | punpcklbw mm2, mm7 | ||
| 207 : | punpckhbw mm6, mm7 | ||
| 208 : | |||
| 209 : | psubsw mm1, mm3 | ||
| 210 : | |||
| 211 : | psubsw mm4, mm2 | ||
| 212 : | psubsw mm5, mm6 | ||
| 213 : | |||
| 214 : | movq [edi], mm0 ; dct[] = mm01 | ||
| 215 : | movq [edi + 8], mm1 | ||
| 216 : | movq [edi + 16], mm4 | ||
| 217 : | movq [edi + 24], mm5 | ||
| 218 : | |||
| 219 : | add edx, ecx | ||
| 220 : | add esi, ecx | ||
| 221 : | add eax, ecx | ||
| 222 : | add ebx, ecx | ||
| 223 : | |||
| 224 : | movq mm0, [edx] ; mm01 = [cur] | ||
| 225 : | movq mm1, mm0 | ||
| 226 : | |||
| 227 : | punpcklbw mm0, mm7 | ||
| 228 : | punpckhbw mm1, mm7 | ||
| 229 : | |||
| 230 : | movq mm4, [eax] | ||
| 231 : | movq mm5, mm4 | ||
| 232 : | |||
| 233 : | punpcklbw mm4, mm7 | ||
| 234 : | punpckhbw mm5, mm7 | ||
| 235 : | |||
| 236 : | movq mm2, [esi] ; mm23 = [ref] | ||
| 237 : | movq mm3, mm2 | ||
| 238 : | |||
| 239 : | movq mm6, [ebx] | ||
| 240 : | |||
| 241 : | movq [edx], mm2 ; [cur] = [ref] | ||
| 242 : | movq [eax], mm6 | ||
| 243 : | |||
| 244 : | punpcklbw mm2, mm7 | ||
| 245 : | punpckhbw mm3, mm7 | ||
| 246 : | |||
| 247 : | psubsw mm0, mm2 ; mm01 -= mm23 | ||
| 248 : | |||
| 249 : | movq mm2, mm6 | ||
| 250 : | |||
| 251 : | punpcklbw mm2, mm7 | ||
| 252 : | punpckhbw mm6, mm7 | ||
| 253 : | |||
| 254 : | psubsw mm1, mm3 | ||
| 255 : | |||
| 256 : | psubsw mm4, mm2 | ||
| 257 : | psubsw mm5, mm6 | ||
| 258 : | |||
| 259 : | movq [edi + 32], mm0 ; dct[] = mm01 | ||
| 260 : | movq [edi + 40], mm1 | ||
| 261 : | movq [edi + 48], mm4 | ||
| 262 : | movq [edi + 56], mm5 | ||
| 263 : | |||
| 264 : | add edx, ecx | ||
| 265 : | add esi, ecx | ||
| 266 : | add eax, ecx | ||
| 267 : | add ebx, ecx | ||
| 268 : | |||
| 269 : | movq mm0, [edx] ; mm01 = [cur] | ||
| 270 : | movq mm1, mm0 | ||
| 271 : | |||
| 272 : | punpcklbw mm0, mm7 | ||
| 273 : | punpckhbw mm1, mm7 | ||
| 274 : | |||
| 275 : | movq mm4, [eax] | ||
| 276 : | movq mm5, mm4 | ||
| 277 : | |||
| 278 : | punpcklbw mm4, mm7 | ||
| 279 : | punpckhbw mm5, mm7 | ||
| 280 : | |||
| 281 : | movq mm2, [esi] ; mm23 = [ref] | ||
| 282 : | movq mm3, mm2 | ||
| 283 : | |||
| 284 : | movq mm6, [ebx] | ||
| 285 : | |||
| 286 : | movq [edx], mm2 ; [cur] = [ref] | ||
| 287 : | movq [eax], mm6 | ||
| 288 : | |||
| 289 : | punpcklbw mm2, mm7 | ||
| 290 : | punpckhbw mm3, mm7 | ||
| 291 : | |||
| 292 : | psubsw mm0, mm2 ; mm01 -= mm23 | ||
| 293 : | |||
| 294 : | movq mm2, mm6 | ||
| 295 : | |||
| 296 : | punpcklbw mm2, mm7 | ||
| 297 : | punpckhbw mm6, mm7 | ||
| 298 : | |||
| 299 : | psubsw mm1, mm3 | ||
| 300 : | |||
| 301 : | psubsw mm4, mm2 | ||
| 302 : | psubsw mm5, mm6 | ||
| 303 : | |||
| 304 : | movq [edi + 64], mm0 ; dct[] = mm01 | ||
| 305 : | movq [edi + 72], mm1 | ||
| 306 : | movq [edi + 80], mm4 | ||
| 307 : | movq [edi + 88], mm5 | ||
| 308 : | |||
| 309 : | add edx, ecx | ||
| 310 : | add esi, ecx | ||
| 311 : | add eax, ecx | ||
| 312 : | add ebx, ecx | ||
| 313 : | |||
| 314 : | movq mm0, [edx] ; mm01 = [cur] | ||
| 315 : | movq mm1, mm0 | ||
| 316 : | |||
| 317 : | punpcklbw mm0, mm7 | ||
| 318 : | punpckhbw mm1, mm7 | ||
| 319 : | |||
| 320 : | movq mm4, [eax] | ||
| 321 : | movq mm5, mm4 | ||
| 322 : | |||
| 323 : | punpcklbw mm4, mm7 | ||
| 324 : | punpckhbw mm5, mm7 | ||
| 325 : | |||
| 326 : | movq mm2, [esi] ; mm23 = [ref] | ||
| 327 : | movq mm3, mm2 | ||
| 328 : | |||
| 329 : | movq mm6, [ebx] | ||
| 330 : | |||
| 331 : | movq [edx], mm2 ; [cur] = [ref] | ||
| 332 : | movq [eax], mm6 | ||
| 333 : | |||
| 334 : | punpcklbw mm2, mm7 | ||
| 335 : | punpckhbw mm3, mm7 | ||
| 336 : | |||
| 337 : | psubsw mm0, mm2 ; mm01 -= mm23 | ||
| 338 : | |||
| 339 : | movq mm2, mm6 | ||
| 340 : | |||
| 341 : | punpcklbw mm2, mm7 | ||
| 342 : | punpckhbw mm6, mm7 | ||
| 343 : | |||
| 344 : | psubsw mm1, mm3 | ||
| 345 : | |||
| 346 : | psubsw mm4, mm2 | ||
| 347 : | psubsw mm5, mm6 | ||
| 348 : | |||
| 349 : | movq [edi + 96], mm0 ; dct[] = mm01 | ||
| 350 : | movq [edi + 104], mm1 | ||
| 351 : | movq [edi + 112], mm4 | ||
| 352 : | movq [edi + 120], mm5 | ||
| 353 : | |||
| 354 : | pop ebx | ||
| 355 : | pop edi | ||
| 356 : | pop esi | ||
| 357 : | |||
| 358 : | ret | ||
| 359 : | |||
| 360 : | |||
| 361 : | edgomez | 1.2 | ;=========================================================================== |
| 362 : | ; | ||
| 363 : | ; void transfer_8to16sub2_xmm(int16_t * const dct, | ||
| 364 : | ; uint8_t * const cur, | ||
| 365 : | ; const uint8_t * ref1, | ||
| 366 : | ; const uint8_t * ref2, | ||
| 367 : | ; const uint32_t stride); | ||
| 368 : | ; | ||
| 369 : | ;=========================================================================== | ||
| 370 : | |||
| 371 : | align 16 | ||
| 372 : | cglobal transfer_8to16sub2_xmm | ||
| 373 : | transfer_8to16sub2_xmm | ||
| 374 : | |||
| 375 : | push edi | ||
| 376 : | push esi | ||
| 377 : | push ebx | ||
| 378 : | |||
| 379 : | mov edi, [esp + 12 + 4] ; edi = &dct | ||
| 380 : | mov esi, [esp + 12 + 8] ; esi = &cur | ||
| 381 : | mov ebx, [esp + 12 + 12] ; ebx = &ref1 | ||
| 382 : | mov edx, [esp + 12 + 16] ; edx = &ref2 | ||
| 383 : | mov eax, [esp + 12 + 20] ; eax = stride | ||
| 384 : | |||
| 385 : | pxor mm7, mm7 ; mm7 = 0 | ||
| 386 : | shl eax, 1 ; eax = stride<<1 | ||
| 387 : | add eax, 16 | ||
| 388 : | |||
| 389 : | ; Row processing | ||
| 390 : | ; One row at a time | ||
| 391 : | movq mm0, [esi + 0] ; mm0 = cur row | ||
| 392 : | movq mm2, [ebx + 0] ; mm2 = ref1 row | ||
| 393 : | movq mm3, [edx + 0] ; mm3 = ref2 row | ||
| 394 : | movq mm1, mm0 ; mm1 = cur row | ||
| 395 : | |||
| 396 : | pavgb mm2, mm3 ; mm2 = (ref1 + ref2 + 1)/2 (== avg) | ||
| 397 : | punpcklbw mm0, mm7 ; mm0 = cur(3-0) <-> 16bit | ||
| 398 : | |||
| 399 : | movq mm3,mm2 ; mm3 = avg | ||
| 400 : | punpckhbw mm1, mm7 ; mm1 = cur(7-4) <-> 16bit | ||
| 401 : | |||
| 402 : | punpcklbw mm2, mm7 ; mm2 = avg(3-0) <-> 16bit | ||
| 403 : | punpckhbw mm3, mm7 ; mm3 = avg(7-4) <-> 16bit | ||
| 404 : | |||
| 405 : | psubw mm0, mm2 ; mm0 = cur(3-0) - avg(3-0) | ||
| 406 : | psubw mm1, mm3 ; mm1 = cur(7-4) - avg(7-4) | ||
| 407 : | |||
| 408 : | movq [edi + 0], mm0 ; dct(3-0) = mm0 | ||
| 409 : | movq [edi + 8], mm1 ; dct(7-4) = mm1 | ||
| 410 : | |||
| 411 : | ; Increment all pointers | ||
| 412 : | add edi, eax ; edi = &(next dct row) | ||
| 413 : | |||
| 414 : | ; Row processing | ||
| 415 : | ; One row at a time | ||
| 416 : | movq mm0, [esi + 8] ; mm0 = cur row | ||
| 417 : | movq mm2, [ebx + 8] ; mm2 = ref1 row | ||
| 418 : | movq mm3, [edx + 8] ; mm3 = ref2 row | ||
| 419 : | movq mm1, mm0 ; mm1 = cur row | ||
| 420 : | |||
| 421 : | pavgb mm2, mm3 ; mm2 = (ref1 + ref2 + 1)/2 (== avg) | ||
| 422 : | punpcklbw mm0, mm7 ; mm0 = cur(3-0) <-> 16bit | ||
| 423 : | |||
| 424 : | movq mm3,mm2 ; mm3 = avg | ||
| 425 : | punpckhbw mm1, mm7 ; mm1 = cur(7-4) <-> 16bit | ||
| 426 : | |||
| 427 : | punpcklbw mm2, mm7 ; mm2 = avg(3-0) <-> 16bit | ||
| 428 : | punpckhbw mm3, mm7 ; mm3 = avg(7-4) <-> 16bit | ||
| 429 : | |||
| 430 : | psubw mm0, mm2 ; mm0 = cur(3-0) - avg(3-0) | ||
| 431 : | psubw mm1, mm3 ; mm1 = cur(7-4) - avg(7-4) | ||
| 432 : | |||
| 433 : | movq [edi + 0], mm0 ; dct(3-0) = mm0 | ||
| 434 : | movq [edi + 8], mm1 ; dct(7-4) = mm1 | ||
| 435 : | |||
| 436 : | ; Increment all pointers | ||
| 437 : | add edi, eax ; edi = &(next dct row) | ||
| 438 : | |||
| 439 : | ; Row processing | ||
| 440 : | ; One row at a time | ||
| 441 : | movq mm0, [esi + 16] ; mm0 = cur row | ||
| 442 : | movq mm2, [ebx + 16] ; mm2 = ref1 row | ||
| 443 : | movq mm3, [edx + 16] ; mm3 = ref2 row | ||
| 444 : | movq mm1, mm0 ; mm1 = cur row | ||
| 445 : | |||
| 446 : | pavgb mm2, mm3 ; mm2 = (ref1 + ref2 + 1)/2 (== avg) | ||
| 447 : | punpcklbw mm0, mm7 ; mm0 = cur(3-0) <-> 16bit | ||
| 448 : | |||
| 449 : | movq mm3,mm2 ; mm3 = avg | ||
| 450 : | punpckhbw mm1, mm7 ; mm1 = cur(7-4) <-> 16bit | ||
| 451 : | |||
| 452 : | punpcklbw mm2, mm7 ; mm2 = avg(3-0) <-> 16bit | ||
| 453 : | punpckhbw mm3, mm7 ; mm3 = avg(7-4) <-> 16bit | ||
| 454 : | |||
| 455 : | psubw mm0, mm2 ; mm0 = cur(3-0) - avg(3-0) | ||
| 456 : | psubw mm1, mm3 ; mm1 = cur(7-4) - avg(7-4) | ||
| 457 : | |||
| 458 : | movq [edi + 0], mm0 ; dct(3-0) = mm0 | ||
| 459 : | movq [edi + 8], mm1 ; dct(7-4) = mm1 | ||
| 460 : | |||
| 461 : | ; Increment all pointers | ||
| 462 : | add edi, eax ; edi = &(next dct row) | ||
| 463 : | |||
| 464 : | ; Row processing | ||
| 465 : | ; One row at a time | ||
| 466 : | movq mm0, [esi + 24] ; mm0 = cur row | ||
| 467 : | movq mm2, [ebx + 24] ; mm2 = ref1 row | ||
| 468 : | movq mm3, [edx + 24] ; mm3 = ref2 row | ||
| 469 : | movq mm1, mm0 ; mm1 = cur row | ||
| 470 : | |||
| 471 : | pavgb mm2, mm3 ; mm2 = (ref1 + ref2 + 1)/2 (== avg) | ||
| 472 : | punpcklbw mm0, mm7 ; mm0 = cur(3-0) <-> 16bit | ||
| 473 : | |||
| 474 : | movq mm3,mm2 ; mm3 = avg | ||
| 475 : | punpckhbw mm1, mm7 ; mm1 = cur(7-4) <-> 16bit | ||
| 476 : | |||
| 477 : | punpcklbw mm2, mm7 ; mm2 = avg(3-0) <-> 16bit | ||
| 478 : | punpckhbw mm3, mm7 ; mm3 = avg(7-4) <-> 16bit | ||
| 479 : | |||
| 480 : | psubw mm0, mm2 ; mm0 = cur(3-0) - avg(3-0) | ||
| 481 : | psubw mm1, mm3 ; mm1 = cur(7-4) - avg(7-4) | ||
| 482 : | |||
| 483 : | movq [edi + 0], mm0 ; dct(3-0) = mm0 | ||
| 484 : | movq [edi + 8], mm1 ; dct(7-4) = mm1 | ||
| 485 : | |||
| 486 : | ; Increment all pointers | ||
| 487 : | add edi, eax ; edi = &(next dct row) | ||
| 488 : | |||
| 489 : | ; Row processing | ||
| 490 : | ; One row at a time | ||
| 491 : | movq mm0, [esi + 32] ; mm0 = cur row | ||
| 492 : | movq mm2, [ebx + 32] ; mm2 = ref1 row | ||
| 493 : | movq mm3, [edx + 32] ; mm3 = ref2 row | ||
| 494 : | movq mm1, mm0 ; mm1 = cur row | ||
| 495 : | |||
| 496 : | pavgb mm2, mm3 ; mm2 = (ref1 + ref2 + 1)/2 (== avg) | ||
| 497 : | punpcklbw mm0, mm7 ; mm0 = cur(3-0) <-> 16bit | ||
| 498 : | |||
| 499 : | movq mm3,mm2 ; mm3 = avg | ||
| 500 : | punpckhbw mm1, mm7 ; mm1 = cur(7-4) <-> 16bit | ||
| 501 : | |||
| 502 : | punpcklbw mm2, mm7 ; mm2 = avg(3-0) <-> 16bit | ||
| 503 : | punpckhbw mm3, mm7 ; mm3 = avg(7-4) <-> 16bit | ||
| 504 : | |||
| 505 : | psubw mm0, mm2 ; mm0 = cur(3-0) - avg(3-0) | ||
| 506 : | psubw mm1, mm3 ; mm1 = cur(7-4) - avg(7-4) | ||
| 507 : | |||
| 508 : | movq [edi + 0], mm0 ; dct(3-0) = mm0 | ||
| 509 : | movq [edi + 8], mm1 ; dct(7-4) = mm1 | ||
| 510 : | |||
| 511 : | ; Increment all pointers | ||
| 512 : | add edi, eax ; edi = &(next dct row) | ||
| 513 : | |||
| 514 : | ; Row processing | ||
| 515 : | ; One row at a time | ||
| 516 : | movq mm0, [esi + 40] ; mm0 = cur row | ||
| 517 : | movq mm2, [ebx + 40] ; mm2 = ref1 row | ||
| 518 : | movq mm3, [edx + 40] ; mm3 = ref2 row | ||
| 519 : | movq mm1, mm0 ; mm1 = cur row | ||
| 520 : | |||
| 521 : | pavgb mm2, mm3 ; mm2 = (ref1 + ref2 + 1)/2 (== avg) | ||
| 522 : | punpcklbw mm0, mm7 ; mm0 = cur(3-0) <-> 16bit | ||
| 523 : | |||
| 524 : | movq mm3,mm2 ; mm3 = avg | ||
| 525 : | punpckhbw mm1, mm7 ; mm1 = cur(7-4) <-> 16bit | ||
| 526 : | |||
| 527 : | punpcklbw mm2, mm7 ; mm2 = avg(3-0) <-> 16bit | ||
| 528 : | punpckhbw mm3, mm7 ; mm3 = avg(7-4) <-> 16bit | ||
| 529 : | |||
| 530 : | psubw mm0, mm2 ; mm0 = cur(3-0) - avg(3-0) | ||
| 531 : | psubw mm1, mm3 ; mm1 = cur(7-4) - avg(7-4) | ||
| 532 : | |||
| 533 : | movq [edi + 0], mm0 ; dct(3-0) = mm0 | ||
| 534 : | movq [edi + 8], mm1 ; dct(7-4) = mm1 | ||
| 535 : | |||
| 536 : | ; Increment all pointers | ||
| 537 : | add edi, eax ; edi = &(next dct row) | ||
| 538 : | |||
| 539 : | ; Row processing | ||
| 540 : | ; One row at a time | ||
| 541 : | movq mm0, [esi + 48] ; mm0 = cur row | ||
| 542 : | movq mm2, [ebx + 48] ; mm2 = ref1 row | ||
| 543 : | movq mm3, [edx + 48] ; mm3 = ref2 row | ||
| 544 : | movq mm1, mm0 ; mm1 = cur row | ||
| 545 : | |||
| 546 : | pavgb mm2, mm3 ; mm2 = (ref1 + ref2 + 1)/2 (== avg) | ||
| 547 : | punpcklbw mm0, mm7 ; mm0 = cur(3-0) <-> 16bit | ||
| 548 : | |||
| 549 : | movq mm3,mm2 ; mm3 = avg | ||
| 550 : | punpckhbw mm1, mm7 ; mm1 = cur(7-4) <-> 16bit | ||
| 551 : | |||
| 552 : | punpcklbw mm2, mm7 ; mm2 = avg(3-0) <-> 16bit | ||
| 553 : | punpckhbw mm3, mm7 ; mm3 = avg(7-4) <-> 16bit | ||
| 554 : | |||
| 555 : | psubw mm0, mm2 ; mm0 = cur(3-0) - avg(3-0) | ||
| 556 : | psubw mm1, mm3 ; mm1 = cur(7-4) - avg(7-4) | ||
| 557 : | |||
| 558 : | movq [edi + 0], mm0 ; dct(3-0) = mm0 | ||
| 559 : | movq [edi + 8], mm1 ; dct(7-4) = mm1 | ||
| 560 : | |||
| 561 : | ; Increment all pointers | ||
| 562 : | add edi, eax ; edi = &(next dct row) | ||
| 563 : | |||
| 564 : | ; Row processing | ||
| 565 : | ; One row at a time | ||
| 566 : | movq mm0, [esi + 56] ; mm0 = cur row | ||
| 567 : | movq mm2, [ebx + 56] ; mm2 = ref1 row | ||
| 568 : | movq mm3, [edx + 56] ; mm3 = ref2 row | ||
| 569 : | movq mm1, mm0 ; mm1 = cur row | ||
| 570 : | |||
| 571 : | pavgb mm2, mm3 ; mm2 = (ref1 + ref2 + 1)/2 (== avg) | ||
| 572 : | punpcklbw mm0, mm7 ; mm0 = cur(3-0) <-> 16bit | ||
| 573 : | |||
| 574 : | movq mm3,mm2 ; mm3 = avg | ||
| 575 : | punpckhbw mm1, mm7 ; mm1 = cur(7-4) <-> 16bit | ||
| 576 : | |||
| 577 : | punpcklbw mm2, mm7 ; mm2 = avg(3-0) <-> 16bit | ||
| 578 : | punpckhbw mm3, mm7 ; mm3 = avg(7-4) <-> 16bit | ||
| 579 : | |||
| 580 : | psubw mm0, mm2 ; mm0 = cur(3-0) - avg(3-0) | ||
| 581 : | psubw mm1, mm3 ; mm1 = cur(7-4) - avg(7-4) | ||
| 582 : | |||
| 583 : | movq [edi + 0], mm0 ; dct(3-0) = mm0 | ||
| 584 : | movq [edi + 8], mm1 ; dct(7-4) = mm1 | ||
| 585 : | |||
| 586 : | ; Exit | ||
| 587 : | |||
| 588 : | pop ebx | ||
| 589 : | pop esi | ||
| 590 : | pop edi | ||
| 591 : | |||
| 592 : | ret | ||
| 593 : | Isibaar | 1.1 | |
| 594 : | ;=========================================================================== | ||
| 595 : | ; | ||
| 596 : | ; void transfer_16to8add_mmx(uint8_t * const dst, | ||
| 597 : | ; const int16_t * const src, | ||
| 598 : | ; uint32_t stride); | ||
| 599 : | ; | ||
| 600 : | ;=========================================================================== | ||
| 601 : | |||
| 602 : | align 16 | ||
| 603 : | cglobal transfer_16to8add_mmx | ||
| 604 : | transfer_16to8add_mmx | ||
| 605 : | |||
| 606 : | push esi | ||
| 607 : | push edi | ||
| 608 : | |||
| 609 : | mov edi, [esp + 8 + 4] ; dst | ||
| 610 : | mov esi, [esp + 8 + 8] ; src | ||
| 611 : | mov ecx, [esp + 8 + 12] ; stride | ||
| 612 : | |||
| 613 : | pxor mm7, mm7 | ||
| 614 : | |||
| 615 : | mov eax, 8 | ||
| 616 : | |||
| 617 : | .loop | ||
| 618 : | movq mm0, [edi] | ||
| 619 : | movq mm1, mm0 | ||
| 620 : | punpcklbw mm0, mm7 ; mm23 = unpack([dst]) | ||
| 621 : | punpckhbw mm1, mm7 | ||
| 622 : | |||
| 623 : | movq mm2, [esi] ; mm01 = [src] | ||
| 624 : | movq mm3, [esi + 8] | ||
| 625 : | |||
| 626 : | paddsw mm0, mm2 ; mm01 += mm23 | ||
| 627 : | paddsw mm1, mm3 | ||
| 628 : | |||
| 629 : | packuswb mm0, mm1 ; [dst] = pack(mm01) | ||
| 630 : | movq [edi], mm0 | ||
| 631 : | |||
| 632 : | add esi, 16 | ||
| 633 : | add edi, ecx | ||
| 634 : | dec eax | ||
| 635 : | jnz .loop | ||
| 636 : | |||
| 637 : | pop edi | ||
| 638 : | pop esi | ||
| 639 : | |||
| 640 : | ret | ||
| 641 : | |||
| 642 : | |||
| 643 : | ;=========================================================================== | ||
| 644 : | ; | ||
| 645 : | ; void transfer8x8_copy_mmx(uint8_t * const dst, | ||
| 646 : | ; const uint8_t * const src, | ||
| 647 : | ; const uint32_t stride); | ||
| 648 : | ; | ||
| 649 : | ; | ||
| 650 : | ;=========================================================================== | ||
| 651 : | |||
| 652 : | align 16 | ||
| 653 : | cglobal transfer8x8_copy_mmx | ||
| 654 : | transfer8x8_copy_mmx | ||
| 655 : | push esi | ||
| 656 : | push edi | ||
| 657 : | |||
| 658 : | mov edi, [esp + 8 + 4] ; dst [out] | ||
| 659 : | mov esi, [esp + 8 + 8] ; src [in] | ||
| 660 : | mov eax, [esp + 8 + 12] ; stride [in] | ||
| 661 : | |||
| 662 : | movq mm0, [esi] | ||
| 663 : | movq mm1, [esi+eax] | ||
| 664 : | movq [edi], mm0 | ||
| 665 : | movq [edi+eax], mm1 | ||
| 666 : | |||
| 667 : | add esi, eax | ||
| 668 : | add edi, eax | ||
| 669 : | add esi, eax | ||
| 670 : | add edi, eax | ||
| 671 : | |||
| 672 : | movq mm0, [esi] | ||
| 673 : | movq mm1, [esi+eax] | ||
| 674 : | movq [edi], mm0 | ||
| 675 : | movq [edi+eax], mm1 | ||
| 676 : | |||
| 677 : | add esi, eax | ||
| 678 : | add edi, eax | ||
| 679 : | add esi, eax | ||
| 680 : | add edi, eax | ||
| 681 : | |||
| 682 : | movq mm0, [esi] | ||
| 683 : | movq mm1, [esi+eax] | ||
| 684 : | movq [edi], mm0 | ||
| 685 : | movq [edi+eax], mm1 | ||
| 686 : | |||
| 687 : | add esi, eax | ||
| 688 : | add edi, eax | ||
| 689 : | add esi, eax | ||
| 690 : | add edi, eax | ||
| 691 : | |||
| 692 : | movq mm0, [esi] | ||
| 693 : | movq mm1, [esi+eax] | ||
| 694 : | movq [edi], mm0 | ||
| 695 : | movq [edi+eax], mm1 | ||
| 696 : | |||
| 697 : | add esi, eax | ||
| 698 : | add edi, eax | ||
| 699 : | add esi, eax | ||
| 700 : | add edi, eax | ||
| 701 : | |||
| 702 : | pop edi | ||
| 703 : | pop esi | ||
| 704 : | |||
| 705 : | ret |
| No admin address has been configured | ViewVC Help |
| Powered by ViewVC 1.0.4 |