Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * include/asm-generic/xor.h |
| 3 | * |
| 4 | * Generic optimized RAID-5 checksumming functions. |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or modify |
| 7 | * it under the terms of the GNU General Public License as published by |
| 8 | * the Free Software Foundation; either version 2, or (at your option) |
| 9 | * any later version. |
| 10 | * |
| 11 | * You should have received a copy of the GNU General Public License |
| 12 | * (for example /usr/src/linux/COPYING); if not, write to the Free |
| 13 | * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
| 14 | */ |
| 15 | |
Linus Torvalds | 268bb0c | 2011-05-20 12:50:29 -0700 | [diff] [blame] | 16 | #include <linux/prefetch.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 17 | |
| 18 | static void |
| 19 | xor_8regs_2(unsigned long bytes, unsigned long *p1, unsigned long *p2) |
| 20 | { |
| 21 | long lines = bytes / (sizeof (long)) / 8; |
| 22 | |
| 23 | do { |
| 24 | p1[0] ^= p2[0]; |
| 25 | p1[1] ^= p2[1]; |
| 26 | p1[2] ^= p2[2]; |
| 27 | p1[3] ^= p2[3]; |
| 28 | p1[4] ^= p2[4]; |
| 29 | p1[5] ^= p2[5]; |
| 30 | p1[6] ^= p2[6]; |
| 31 | p1[7] ^= p2[7]; |
| 32 | p1 += 8; |
| 33 | p2 += 8; |
| 34 | } while (--lines > 0); |
| 35 | } |
| 36 | |
| 37 | static void |
| 38 | xor_8regs_3(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 39 | unsigned long *p3) |
| 40 | { |
| 41 | long lines = bytes / (sizeof (long)) / 8; |
| 42 | |
| 43 | do { |
| 44 | p1[0] ^= p2[0] ^ p3[0]; |
| 45 | p1[1] ^= p2[1] ^ p3[1]; |
| 46 | p1[2] ^= p2[2] ^ p3[2]; |
| 47 | p1[3] ^= p2[3] ^ p3[3]; |
| 48 | p1[4] ^= p2[4] ^ p3[4]; |
| 49 | p1[5] ^= p2[5] ^ p3[5]; |
| 50 | p1[6] ^= p2[6] ^ p3[6]; |
| 51 | p1[7] ^= p2[7] ^ p3[7]; |
| 52 | p1 += 8; |
| 53 | p2 += 8; |
| 54 | p3 += 8; |
| 55 | } while (--lines > 0); |
| 56 | } |
| 57 | |
| 58 | static void |
| 59 | xor_8regs_4(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 60 | unsigned long *p3, unsigned long *p4) |
| 61 | { |
| 62 | long lines = bytes / (sizeof (long)) / 8; |
| 63 | |
| 64 | do { |
| 65 | p1[0] ^= p2[0] ^ p3[0] ^ p4[0]; |
| 66 | p1[1] ^= p2[1] ^ p3[1] ^ p4[1]; |
| 67 | p1[2] ^= p2[2] ^ p3[2] ^ p4[2]; |
| 68 | p1[3] ^= p2[3] ^ p3[3] ^ p4[3]; |
| 69 | p1[4] ^= p2[4] ^ p3[4] ^ p4[4]; |
| 70 | p1[5] ^= p2[5] ^ p3[5] ^ p4[5]; |
| 71 | p1[6] ^= p2[6] ^ p3[6] ^ p4[6]; |
| 72 | p1[7] ^= p2[7] ^ p3[7] ^ p4[7]; |
| 73 | p1 += 8; |
| 74 | p2 += 8; |
| 75 | p3 += 8; |
| 76 | p4 += 8; |
| 77 | } while (--lines > 0); |
| 78 | } |
| 79 | |
| 80 | static void |
| 81 | xor_8regs_5(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 82 | unsigned long *p3, unsigned long *p4, unsigned long *p5) |
| 83 | { |
| 84 | long lines = bytes / (sizeof (long)) / 8; |
| 85 | |
| 86 | do { |
| 87 | p1[0] ^= p2[0] ^ p3[0] ^ p4[0] ^ p5[0]; |
| 88 | p1[1] ^= p2[1] ^ p3[1] ^ p4[1] ^ p5[1]; |
| 89 | p1[2] ^= p2[2] ^ p3[2] ^ p4[2] ^ p5[2]; |
| 90 | p1[3] ^= p2[3] ^ p3[3] ^ p4[3] ^ p5[3]; |
| 91 | p1[4] ^= p2[4] ^ p3[4] ^ p4[4] ^ p5[4]; |
| 92 | p1[5] ^= p2[5] ^ p3[5] ^ p4[5] ^ p5[5]; |
| 93 | p1[6] ^= p2[6] ^ p3[6] ^ p4[6] ^ p5[6]; |
| 94 | p1[7] ^= p2[7] ^ p3[7] ^ p4[7] ^ p5[7]; |
| 95 | p1 += 8; |
| 96 | p2 += 8; |
| 97 | p3 += 8; |
| 98 | p4 += 8; |
| 99 | p5 += 8; |
| 100 | } while (--lines > 0); |
| 101 | } |
| 102 | |
| 103 | static void |
| 104 | xor_32regs_2(unsigned long bytes, unsigned long *p1, unsigned long *p2) |
| 105 | { |
| 106 | long lines = bytes / (sizeof (long)) / 8; |
| 107 | |
| 108 | do { |
| 109 | register long d0, d1, d2, d3, d4, d5, d6, d7; |
| 110 | d0 = p1[0]; /* Pull the stuff into registers */ |
| 111 | d1 = p1[1]; /* ... in bursts, if possible. */ |
| 112 | d2 = p1[2]; |
| 113 | d3 = p1[3]; |
| 114 | d4 = p1[4]; |
| 115 | d5 = p1[5]; |
| 116 | d6 = p1[6]; |
| 117 | d7 = p1[7]; |
| 118 | d0 ^= p2[0]; |
| 119 | d1 ^= p2[1]; |
| 120 | d2 ^= p2[2]; |
| 121 | d3 ^= p2[3]; |
| 122 | d4 ^= p2[4]; |
| 123 | d5 ^= p2[5]; |
| 124 | d6 ^= p2[6]; |
| 125 | d7 ^= p2[7]; |
| 126 | p1[0] = d0; /* Store the result (in bursts) */ |
| 127 | p1[1] = d1; |
| 128 | p1[2] = d2; |
| 129 | p1[3] = d3; |
| 130 | p1[4] = d4; |
| 131 | p1[5] = d5; |
| 132 | p1[6] = d6; |
| 133 | p1[7] = d7; |
| 134 | p1 += 8; |
| 135 | p2 += 8; |
| 136 | } while (--lines > 0); |
| 137 | } |
| 138 | |
| 139 | static void |
| 140 | xor_32regs_3(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 141 | unsigned long *p3) |
| 142 | { |
| 143 | long lines = bytes / (sizeof (long)) / 8; |
| 144 | |
| 145 | do { |
| 146 | register long d0, d1, d2, d3, d4, d5, d6, d7; |
| 147 | d0 = p1[0]; /* Pull the stuff into registers */ |
| 148 | d1 = p1[1]; /* ... in bursts, if possible. */ |
| 149 | d2 = p1[2]; |
| 150 | d3 = p1[3]; |
| 151 | d4 = p1[4]; |
| 152 | d5 = p1[5]; |
| 153 | d6 = p1[6]; |
| 154 | d7 = p1[7]; |
| 155 | d0 ^= p2[0]; |
| 156 | d1 ^= p2[1]; |
| 157 | d2 ^= p2[2]; |
| 158 | d3 ^= p2[3]; |
| 159 | d4 ^= p2[4]; |
| 160 | d5 ^= p2[5]; |
| 161 | d6 ^= p2[6]; |
| 162 | d7 ^= p2[7]; |
| 163 | d0 ^= p3[0]; |
| 164 | d1 ^= p3[1]; |
| 165 | d2 ^= p3[2]; |
| 166 | d3 ^= p3[3]; |
| 167 | d4 ^= p3[4]; |
| 168 | d5 ^= p3[5]; |
| 169 | d6 ^= p3[6]; |
| 170 | d7 ^= p3[7]; |
| 171 | p1[0] = d0; /* Store the result (in bursts) */ |
| 172 | p1[1] = d1; |
| 173 | p1[2] = d2; |
| 174 | p1[3] = d3; |
| 175 | p1[4] = d4; |
| 176 | p1[5] = d5; |
| 177 | p1[6] = d6; |
| 178 | p1[7] = d7; |
| 179 | p1 += 8; |
| 180 | p2 += 8; |
| 181 | p3 += 8; |
| 182 | } while (--lines > 0); |
| 183 | } |
| 184 | |
| 185 | static void |
| 186 | xor_32regs_4(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 187 | unsigned long *p3, unsigned long *p4) |
| 188 | { |
| 189 | long lines = bytes / (sizeof (long)) / 8; |
| 190 | |
| 191 | do { |
| 192 | register long d0, d1, d2, d3, d4, d5, d6, d7; |
| 193 | d0 = p1[0]; /* Pull the stuff into registers */ |
| 194 | d1 = p1[1]; /* ... in bursts, if possible. */ |
| 195 | d2 = p1[2]; |
| 196 | d3 = p1[3]; |
| 197 | d4 = p1[4]; |
| 198 | d5 = p1[5]; |
| 199 | d6 = p1[6]; |
| 200 | d7 = p1[7]; |
| 201 | d0 ^= p2[0]; |
| 202 | d1 ^= p2[1]; |
| 203 | d2 ^= p2[2]; |
| 204 | d3 ^= p2[3]; |
| 205 | d4 ^= p2[4]; |
| 206 | d5 ^= p2[5]; |
| 207 | d6 ^= p2[6]; |
| 208 | d7 ^= p2[7]; |
| 209 | d0 ^= p3[0]; |
| 210 | d1 ^= p3[1]; |
| 211 | d2 ^= p3[2]; |
| 212 | d3 ^= p3[3]; |
| 213 | d4 ^= p3[4]; |
| 214 | d5 ^= p3[5]; |
| 215 | d6 ^= p3[6]; |
| 216 | d7 ^= p3[7]; |
| 217 | d0 ^= p4[0]; |
| 218 | d1 ^= p4[1]; |
| 219 | d2 ^= p4[2]; |
| 220 | d3 ^= p4[3]; |
| 221 | d4 ^= p4[4]; |
| 222 | d5 ^= p4[5]; |
| 223 | d6 ^= p4[6]; |
| 224 | d7 ^= p4[7]; |
| 225 | p1[0] = d0; /* Store the result (in bursts) */ |
| 226 | p1[1] = d1; |
| 227 | p1[2] = d2; |
| 228 | p1[3] = d3; |
| 229 | p1[4] = d4; |
| 230 | p1[5] = d5; |
| 231 | p1[6] = d6; |
| 232 | p1[7] = d7; |
| 233 | p1 += 8; |
| 234 | p2 += 8; |
| 235 | p3 += 8; |
| 236 | p4 += 8; |
| 237 | } while (--lines > 0); |
| 238 | } |
| 239 | |
| 240 | static void |
| 241 | xor_32regs_5(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 242 | unsigned long *p3, unsigned long *p4, unsigned long *p5) |
| 243 | { |
| 244 | long lines = bytes / (sizeof (long)) / 8; |
| 245 | |
| 246 | do { |
| 247 | register long d0, d1, d2, d3, d4, d5, d6, d7; |
| 248 | d0 = p1[0]; /* Pull the stuff into registers */ |
| 249 | d1 = p1[1]; /* ... in bursts, if possible. */ |
| 250 | d2 = p1[2]; |
| 251 | d3 = p1[3]; |
| 252 | d4 = p1[4]; |
| 253 | d5 = p1[5]; |
| 254 | d6 = p1[6]; |
| 255 | d7 = p1[7]; |
| 256 | d0 ^= p2[0]; |
| 257 | d1 ^= p2[1]; |
| 258 | d2 ^= p2[2]; |
| 259 | d3 ^= p2[3]; |
| 260 | d4 ^= p2[4]; |
| 261 | d5 ^= p2[5]; |
| 262 | d6 ^= p2[6]; |
| 263 | d7 ^= p2[7]; |
| 264 | d0 ^= p3[0]; |
| 265 | d1 ^= p3[1]; |
| 266 | d2 ^= p3[2]; |
| 267 | d3 ^= p3[3]; |
| 268 | d4 ^= p3[4]; |
| 269 | d5 ^= p3[5]; |
| 270 | d6 ^= p3[6]; |
| 271 | d7 ^= p3[7]; |
| 272 | d0 ^= p4[0]; |
| 273 | d1 ^= p4[1]; |
| 274 | d2 ^= p4[2]; |
| 275 | d3 ^= p4[3]; |
| 276 | d4 ^= p4[4]; |
| 277 | d5 ^= p4[5]; |
| 278 | d6 ^= p4[6]; |
| 279 | d7 ^= p4[7]; |
| 280 | d0 ^= p5[0]; |
| 281 | d1 ^= p5[1]; |
| 282 | d2 ^= p5[2]; |
| 283 | d3 ^= p5[3]; |
| 284 | d4 ^= p5[4]; |
| 285 | d5 ^= p5[5]; |
| 286 | d6 ^= p5[6]; |
| 287 | d7 ^= p5[7]; |
| 288 | p1[0] = d0; /* Store the result (in bursts) */ |
| 289 | p1[1] = d1; |
| 290 | p1[2] = d2; |
| 291 | p1[3] = d3; |
| 292 | p1[4] = d4; |
| 293 | p1[5] = d5; |
| 294 | p1[6] = d6; |
| 295 | p1[7] = d7; |
| 296 | p1 += 8; |
| 297 | p2 += 8; |
| 298 | p3 += 8; |
| 299 | p4 += 8; |
| 300 | p5 += 8; |
| 301 | } while (--lines > 0); |
| 302 | } |
| 303 | |
| 304 | static void |
| 305 | xor_8regs_p_2(unsigned long bytes, unsigned long *p1, unsigned long *p2) |
| 306 | { |
| 307 | long lines = bytes / (sizeof (long)) / 8 - 1; |
| 308 | prefetchw(p1); |
| 309 | prefetch(p2); |
| 310 | |
| 311 | do { |
| 312 | prefetchw(p1+8); |
| 313 | prefetch(p2+8); |
| 314 | once_more: |
| 315 | p1[0] ^= p2[0]; |
| 316 | p1[1] ^= p2[1]; |
| 317 | p1[2] ^= p2[2]; |
| 318 | p1[3] ^= p2[3]; |
| 319 | p1[4] ^= p2[4]; |
| 320 | p1[5] ^= p2[5]; |
| 321 | p1[6] ^= p2[6]; |
| 322 | p1[7] ^= p2[7]; |
| 323 | p1 += 8; |
| 324 | p2 += 8; |
| 325 | } while (--lines > 0); |
| 326 | if (lines == 0) |
| 327 | goto once_more; |
| 328 | } |
| 329 | |
| 330 | static void |
| 331 | xor_8regs_p_3(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 332 | unsigned long *p3) |
| 333 | { |
| 334 | long lines = bytes / (sizeof (long)) / 8 - 1; |
| 335 | prefetchw(p1); |
| 336 | prefetch(p2); |
| 337 | prefetch(p3); |
| 338 | |
| 339 | do { |
| 340 | prefetchw(p1+8); |
| 341 | prefetch(p2+8); |
| 342 | prefetch(p3+8); |
| 343 | once_more: |
| 344 | p1[0] ^= p2[0] ^ p3[0]; |
| 345 | p1[1] ^= p2[1] ^ p3[1]; |
| 346 | p1[2] ^= p2[2] ^ p3[2]; |
| 347 | p1[3] ^= p2[3] ^ p3[3]; |
| 348 | p1[4] ^= p2[4] ^ p3[4]; |
| 349 | p1[5] ^= p2[5] ^ p3[5]; |
| 350 | p1[6] ^= p2[6] ^ p3[6]; |
| 351 | p1[7] ^= p2[7] ^ p3[7]; |
| 352 | p1 += 8; |
| 353 | p2 += 8; |
| 354 | p3 += 8; |
| 355 | } while (--lines > 0); |
| 356 | if (lines == 0) |
| 357 | goto once_more; |
| 358 | } |
| 359 | |
| 360 | static void |
| 361 | xor_8regs_p_4(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 362 | unsigned long *p3, unsigned long *p4) |
| 363 | { |
| 364 | long lines = bytes / (sizeof (long)) / 8 - 1; |
| 365 | |
| 366 | prefetchw(p1); |
| 367 | prefetch(p2); |
| 368 | prefetch(p3); |
| 369 | prefetch(p4); |
| 370 | |
| 371 | do { |
| 372 | prefetchw(p1+8); |
| 373 | prefetch(p2+8); |
| 374 | prefetch(p3+8); |
| 375 | prefetch(p4+8); |
| 376 | once_more: |
| 377 | p1[0] ^= p2[0] ^ p3[0] ^ p4[0]; |
| 378 | p1[1] ^= p2[1] ^ p3[1] ^ p4[1]; |
| 379 | p1[2] ^= p2[2] ^ p3[2] ^ p4[2]; |
| 380 | p1[3] ^= p2[3] ^ p3[3] ^ p4[3]; |
| 381 | p1[4] ^= p2[4] ^ p3[4] ^ p4[4]; |
| 382 | p1[5] ^= p2[5] ^ p3[5] ^ p4[5]; |
| 383 | p1[6] ^= p2[6] ^ p3[6] ^ p4[6]; |
| 384 | p1[7] ^= p2[7] ^ p3[7] ^ p4[7]; |
| 385 | p1 += 8; |
| 386 | p2 += 8; |
| 387 | p3 += 8; |
| 388 | p4 += 8; |
| 389 | } while (--lines > 0); |
| 390 | if (lines == 0) |
| 391 | goto once_more; |
| 392 | } |
| 393 | |
| 394 | static void |
| 395 | xor_8regs_p_5(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 396 | unsigned long *p3, unsigned long *p4, unsigned long *p5) |
| 397 | { |
| 398 | long lines = bytes / (sizeof (long)) / 8 - 1; |
| 399 | |
| 400 | prefetchw(p1); |
| 401 | prefetch(p2); |
| 402 | prefetch(p3); |
| 403 | prefetch(p4); |
| 404 | prefetch(p5); |
| 405 | |
| 406 | do { |
| 407 | prefetchw(p1+8); |
| 408 | prefetch(p2+8); |
| 409 | prefetch(p3+8); |
| 410 | prefetch(p4+8); |
| 411 | prefetch(p5+8); |
| 412 | once_more: |
| 413 | p1[0] ^= p2[0] ^ p3[0] ^ p4[0] ^ p5[0]; |
| 414 | p1[1] ^= p2[1] ^ p3[1] ^ p4[1] ^ p5[1]; |
| 415 | p1[2] ^= p2[2] ^ p3[2] ^ p4[2] ^ p5[2]; |
| 416 | p1[3] ^= p2[3] ^ p3[3] ^ p4[3] ^ p5[3]; |
| 417 | p1[4] ^= p2[4] ^ p3[4] ^ p4[4] ^ p5[4]; |
| 418 | p1[5] ^= p2[5] ^ p3[5] ^ p4[5] ^ p5[5]; |
| 419 | p1[6] ^= p2[6] ^ p3[6] ^ p4[6] ^ p5[6]; |
| 420 | p1[7] ^= p2[7] ^ p3[7] ^ p4[7] ^ p5[7]; |
| 421 | p1 += 8; |
| 422 | p2 += 8; |
| 423 | p3 += 8; |
| 424 | p4 += 8; |
| 425 | p5 += 8; |
| 426 | } while (--lines > 0); |
| 427 | if (lines == 0) |
| 428 | goto once_more; |
| 429 | } |
| 430 | |
| 431 | static void |
| 432 | xor_32regs_p_2(unsigned long bytes, unsigned long *p1, unsigned long *p2) |
| 433 | { |
| 434 | long lines = bytes / (sizeof (long)) / 8 - 1; |
| 435 | |
| 436 | prefetchw(p1); |
| 437 | prefetch(p2); |
| 438 | |
| 439 | do { |
| 440 | register long d0, d1, d2, d3, d4, d5, d6, d7; |
| 441 | |
| 442 | prefetchw(p1+8); |
| 443 | prefetch(p2+8); |
| 444 | once_more: |
| 445 | d0 = p1[0]; /* Pull the stuff into registers */ |
| 446 | d1 = p1[1]; /* ... in bursts, if possible. */ |
| 447 | d2 = p1[2]; |
| 448 | d3 = p1[3]; |
| 449 | d4 = p1[4]; |
| 450 | d5 = p1[5]; |
| 451 | d6 = p1[6]; |
| 452 | d7 = p1[7]; |
| 453 | d0 ^= p2[0]; |
| 454 | d1 ^= p2[1]; |
| 455 | d2 ^= p2[2]; |
| 456 | d3 ^= p2[3]; |
| 457 | d4 ^= p2[4]; |
| 458 | d5 ^= p2[5]; |
| 459 | d6 ^= p2[6]; |
| 460 | d7 ^= p2[7]; |
| 461 | p1[0] = d0; /* Store the result (in bursts) */ |
| 462 | p1[1] = d1; |
| 463 | p1[2] = d2; |
| 464 | p1[3] = d3; |
| 465 | p1[4] = d4; |
| 466 | p1[5] = d5; |
| 467 | p1[6] = d6; |
| 468 | p1[7] = d7; |
| 469 | p1 += 8; |
| 470 | p2 += 8; |
| 471 | } while (--lines > 0); |
| 472 | if (lines == 0) |
| 473 | goto once_more; |
| 474 | } |
| 475 | |
| 476 | static void |
| 477 | xor_32regs_p_3(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 478 | unsigned long *p3) |
| 479 | { |
| 480 | long lines = bytes / (sizeof (long)) / 8 - 1; |
| 481 | |
| 482 | prefetchw(p1); |
| 483 | prefetch(p2); |
| 484 | prefetch(p3); |
| 485 | |
| 486 | do { |
| 487 | register long d0, d1, d2, d3, d4, d5, d6, d7; |
| 488 | |
| 489 | prefetchw(p1+8); |
| 490 | prefetch(p2+8); |
| 491 | prefetch(p3+8); |
| 492 | once_more: |
| 493 | d0 = p1[0]; /* Pull the stuff into registers */ |
| 494 | d1 = p1[1]; /* ... in bursts, if possible. */ |
| 495 | d2 = p1[2]; |
| 496 | d3 = p1[3]; |
| 497 | d4 = p1[4]; |
| 498 | d5 = p1[5]; |
| 499 | d6 = p1[6]; |
| 500 | d7 = p1[7]; |
| 501 | d0 ^= p2[0]; |
| 502 | d1 ^= p2[1]; |
| 503 | d2 ^= p2[2]; |
| 504 | d3 ^= p2[3]; |
| 505 | d4 ^= p2[4]; |
| 506 | d5 ^= p2[5]; |
| 507 | d6 ^= p2[6]; |
| 508 | d7 ^= p2[7]; |
| 509 | d0 ^= p3[0]; |
| 510 | d1 ^= p3[1]; |
| 511 | d2 ^= p3[2]; |
| 512 | d3 ^= p3[3]; |
| 513 | d4 ^= p3[4]; |
| 514 | d5 ^= p3[5]; |
| 515 | d6 ^= p3[6]; |
| 516 | d7 ^= p3[7]; |
| 517 | p1[0] = d0; /* Store the result (in bursts) */ |
| 518 | p1[1] = d1; |
| 519 | p1[2] = d2; |
| 520 | p1[3] = d3; |
| 521 | p1[4] = d4; |
| 522 | p1[5] = d5; |
| 523 | p1[6] = d6; |
| 524 | p1[7] = d7; |
| 525 | p1 += 8; |
| 526 | p2 += 8; |
| 527 | p3 += 8; |
| 528 | } while (--lines > 0); |
| 529 | if (lines == 0) |
| 530 | goto once_more; |
| 531 | } |
| 532 | |
| 533 | static void |
| 534 | xor_32regs_p_4(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 535 | unsigned long *p3, unsigned long *p4) |
| 536 | { |
| 537 | long lines = bytes / (sizeof (long)) / 8 - 1; |
| 538 | |
| 539 | prefetchw(p1); |
| 540 | prefetch(p2); |
| 541 | prefetch(p3); |
| 542 | prefetch(p4); |
| 543 | |
| 544 | do { |
| 545 | register long d0, d1, d2, d3, d4, d5, d6, d7; |
| 546 | |
| 547 | prefetchw(p1+8); |
| 548 | prefetch(p2+8); |
| 549 | prefetch(p3+8); |
| 550 | prefetch(p4+8); |
| 551 | once_more: |
| 552 | d0 = p1[0]; /* Pull the stuff into registers */ |
| 553 | d1 = p1[1]; /* ... in bursts, if possible. */ |
| 554 | d2 = p1[2]; |
| 555 | d3 = p1[3]; |
| 556 | d4 = p1[4]; |
| 557 | d5 = p1[5]; |
| 558 | d6 = p1[6]; |
| 559 | d7 = p1[7]; |
| 560 | d0 ^= p2[0]; |
| 561 | d1 ^= p2[1]; |
| 562 | d2 ^= p2[2]; |
| 563 | d3 ^= p2[3]; |
| 564 | d4 ^= p2[4]; |
| 565 | d5 ^= p2[5]; |
| 566 | d6 ^= p2[6]; |
| 567 | d7 ^= p2[7]; |
| 568 | d0 ^= p3[0]; |
| 569 | d1 ^= p3[1]; |
| 570 | d2 ^= p3[2]; |
| 571 | d3 ^= p3[3]; |
| 572 | d4 ^= p3[4]; |
| 573 | d5 ^= p3[5]; |
| 574 | d6 ^= p3[6]; |
| 575 | d7 ^= p3[7]; |
| 576 | d0 ^= p4[0]; |
| 577 | d1 ^= p4[1]; |
| 578 | d2 ^= p4[2]; |
| 579 | d3 ^= p4[3]; |
| 580 | d4 ^= p4[4]; |
| 581 | d5 ^= p4[5]; |
| 582 | d6 ^= p4[6]; |
| 583 | d7 ^= p4[7]; |
| 584 | p1[0] = d0; /* Store the result (in bursts) */ |
| 585 | p1[1] = d1; |
| 586 | p1[2] = d2; |
| 587 | p1[3] = d3; |
| 588 | p1[4] = d4; |
| 589 | p1[5] = d5; |
| 590 | p1[6] = d6; |
| 591 | p1[7] = d7; |
| 592 | p1 += 8; |
| 593 | p2 += 8; |
| 594 | p3 += 8; |
| 595 | p4 += 8; |
| 596 | } while (--lines > 0); |
| 597 | if (lines == 0) |
| 598 | goto once_more; |
| 599 | } |
| 600 | |
| 601 | static void |
| 602 | xor_32regs_p_5(unsigned long bytes, unsigned long *p1, unsigned long *p2, |
| 603 | unsigned long *p3, unsigned long *p4, unsigned long *p5) |
| 604 | { |
| 605 | long lines = bytes / (sizeof (long)) / 8 - 1; |
| 606 | |
| 607 | prefetchw(p1); |
| 608 | prefetch(p2); |
| 609 | prefetch(p3); |
| 610 | prefetch(p4); |
| 611 | prefetch(p5); |
| 612 | |
| 613 | do { |
| 614 | register long d0, d1, d2, d3, d4, d5, d6, d7; |
| 615 | |
| 616 | prefetchw(p1+8); |
| 617 | prefetch(p2+8); |
| 618 | prefetch(p3+8); |
| 619 | prefetch(p4+8); |
| 620 | prefetch(p5+8); |
| 621 | once_more: |
| 622 | d0 = p1[0]; /* Pull the stuff into registers */ |
| 623 | d1 = p1[1]; /* ... in bursts, if possible. */ |
| 624 | d2 = p1[2]; |
| 625 | d3 = p1[3]; |
| 626 | d4 = p1[4]; |
| 627 | d5 = p1[5]; |
| 628 | d6 = p1[6]; |
| 629 | d7 = p1[7]; |
| 630 | d0 ^= p2[0]; |
| 631 | d1 ^= p2[1]; |
| 632 | d2 ^= p2[2]; |
| 633 | d3 ^= p2[3]; |
| 634 | d4 ^= p2[4]; |
| 635 | d5 ^= p2[5]; |
| 636 | d6 ^= p2[6]; |
| 637 | d7 ^= p2[7]; |
| 638 | d0 ^= p3[0]; |
| 639 | d1 ^= p3[1]; |
| 640 | d2 ^= p3[2]; |
| 641 | d3 ^= p3[3]; |
| 642 | d4 ^= p3[4]; |
| 643 | d5 ^= p3[5]; |
| 644 | d6 ^= p3[6]; |
| 645 | d7 ^= p3[7]; |
| 646 | d0 ^= p4[0]; |
| 647 | d1 ^= p4[1]; |
| 648 | d2 ^= p4[2]; |
| 649 | d3 ^= p4[3]; |
| 650 | d4 ^= p4[4]; |
| 651 | d5 ^= p4[5]; |
| 652 | d6 ^= p4[6]; |
| 653 | d7 ^= p4[7]; |
| 654 | d0 ^= p5[0]; |
| 655 | d1 ^= p5[1]; |
| 656 | d2 ^= p5[2]; |
| 657 | d3 ^= p5[3]; |
| 658 | d4 ^= p5[4]; |
| 659 | d5 ^= p5[5]; |
| 660 | d6 ^= p5[6]; |
| 661 | d7 ^= p5[7]; |
| 662 | p1[0] = d0; /* Store the result (in bursts) */ |
| 663 | p1[1] = d1; |
| 664 | p1[2] = d2; |
| 665 | p1[3] = d3; |
| 666 | p1[4] = d4; |
| 667 | p1[5] = d5; |
| 668 | p1[6] = d6; |
| 669 | p1[7] = d7; |
| 670 | p1 += 8; |
| 671 | p2 += 8; |
| 672 | p3 += 8; |
| 673 | p4 += 8; |
| 674 | p5 += 8; |
| 675 | } while (--lines > 0); |
| 676 | if (lines == 0) |
| 677 | goto once_more; |
| 678 | } |
| 679 | |
| 680 | static struct xor_block_template xor_block_8regs = { |
| 681 | .name = "8regs", |
| 682 | .do_2 = xor_8regs_2, |
| 683 | .do_3 = xor_8regs_3, |
| 684 | .do_4 = xor_8regs_4, |
| 685 | .do_5 = xor_8regs_5, |
| 686 | }; |
| 687 | |
| 688 | static struct xor_block_template xor_block_32regs = { |
| 689 | .name = "32regs", |
| 690 | .do_2 = xor_32regs_2, |
| 691 | .do_3 = xor_32regs_3, |
| 692 | .do_4 = xor_32regs_4, |
| 693 | .do_5 = xor_32regs_5, |
| 694 | }; |
| 695 | |
Arnd Bergmann | 720fb19 | 2012-08-04 21:53:26 +0000 | [diff] [blame] | 696 | static struct xor_block_template xor_block_8regs_p __maybe_unused = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 697 | .name = "8regs_prefetch", |
| 698 | .do_2 = xor_8regs_p_2, |
| 699 | .do_3 = xor_8regs_p_3, |
| 700 | .do_4 = xor_8regs_p_4, |
| 701 | .do_5 = xor_8regs_p_5, |
| 702 | }; |
| 703 | |
Arnd Bergmann | 720fb19 | 2012-08-04 21:53:26 +0000 | [diff] [blame] | 704 | static struct xor_block_template xor_block_32regs_p __maybe_unused = { |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 705 | .name = "32regs_prefetch", |
| 706 | .do_2 = xor_32regs_p_2, |
| 707 | .do_3 = xor_32regs_p_3, |
| 708 | .do_4 = xor_32regs_p_4, |
| 709 | .do_5 = xor_32regs_p_5, |
| 710 | }; |
| 711 | |
| 712 | #define XOR_TRY_TEMPLATES \ |
| 713 | do { \ |
| 714 | xor_speed(&xor_block_8regs); \ |
| 715 | xor_speed(&xor_block_8regs_p); \ |
| 716 | xor_speed(&xor_block_32regs); \ |
| 717 | xor_speed(&xor_block_32regs_p); \ |
| 718 | } while (0) |