rsh1aors_n.asm revision 1.1.1.1.8.1
1 dnl IA-64 mpn_rsh1add_n/mpn_rsh1sub_n -- rp[] = (up[] +- vp[]) >> 1. 2 3 dnl Contributed to the GNU project by Torbjorn Granlund. 4 5 dnl Copyright 2003, 2004, 2005 Free Software Foundation, Inc. 6 7 dnl This file is part of the GNU MP Library. 8 9 dnl The GNU MP Library is free software; you can redistribute it and/or modify 10 dnl it under the terms of the GNU Lesser General Public License as published 11 dnl by the Free Software Foundation; either version 3 of the License, or (at 12 dnl your option) any later version. 13 14 dnl The GNU MP Library is distributed in the hope that it will be useful, but 15 dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 16 dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public 17 dnl License for more details. 18 19 dnl You should have received a copy of the GNU Lesser General Public License 20 dnl along with the GNU MP Library. If not, see http://www.gnu.org/licenses/. 21 22 include(`../config.m4') 23 24 C cycles/limb 25 C Itanium: 2.5 26 C Itanium 2: 1.5 27 28 C TODO 29 C * Rewrite function entry code using aorslsh1_n.asm style. 30 C * Micro-optimize feed-in and wind-down code. 31 32 C INPUT PARAMETERS 33 define(`rp',`r32') 34 define(`up',`r33') 35 define(`vp',`r34') 36 define(`n',`r35') 37 38 ifdef(`OPERATION_rsh1add_n',` 39 define(ADDSUB, add) 40 define(PRED, ltu) 41 define(INCR, 1) 42 define(LIM, -1) 43 define(func, mpn_rsh1add_n) 44 ') 45 ifdef(`OPERATION_rsh1sub_n',` 46 define(ADDSUB, sub) 47 define(PRED, gtu) 48 define(INCR, -1) 49 define(LIM, 0) 50 define(func, mpn_rsh1sub_n) 51 ') 52 53 C Some useful aliases for registers we use 54 define(`u0',`r14') define(`u1',`r15') define(`u2',`r16') define(`u3',`r17') 55 define(`v0',`r18') define(`v1',`r19') define(`v2',`r20') define(`v3',`r21') 56 define(`w0',`r22') define(`w1',`r23') define(`w2',`r24') define(`w3',`r25') 57 define(`x0',`r26') define(`x1',`r9') define(`x2',`r30') define(`x3',`r31') 58 59 MULFUNC_PROLOGUE(mpn_rsh1add_n mpn_rsh1sub_n) 60 61 ASM_START() 62 PROLOGUE(func) 63 .prologue 64 .save ar.lc, r2 65 .body 66 ifdef(`HAVE_ABI_32',` 67 addp4 rp = 0, rp C M I 68 addp4 up = 0, up C M I 69 addp4 vp = 0, vp C M I 70 zxt4 n = n C I 71 ;; 72 ') 73 {.mmi; ld8 r11 = [vp], 8 C M01 74 ld8 r10 = [up], 8 C M01 75 mov.i r2 = ar.lc C I0 76 }{.mmi; and r14 = 3, n C M I 77 cmp.lt p15, p0 = 4, n C M I 78 add n = -4, n C M I 79 ;; 80 }{.mmi; cmp.eq p6, p0 = 1, r14 C M I 81 cmp.eq p7, p0 = 2, r14 C M I 82 cmp.eq p8, p0 = 3, r14 C M I 83 }{.bbb 84 (p6) br.dptk .Lb01 C B 85 (p7) br.dptk .Lb10 C B 86 (p8) br.dptk .Lb11 C B 87 } 88 89 .Lb00: ld8 v0 = [vp], 8 C M01 90 ld8 u0 = [up], 8 C M01 91 shr.u n = n, 2 C I0 92 ;; 93 ld8 v1 = [vp], 8 C M01 94 ld8 u1 = [up], 8 C M01 95 ADDSUB w3 = r10, r11 C M I 96 ;; 97 ld8 v2 = [vp], 8 C M01 98 ld8 u2 = [up], 8 C M01 99 (p15) br.dpnt .grt4 C B 100 ;; 101 102 cmp.PRED p7, p0 = w3, r10 C M I 103 and r8 = 1, w3 C M I 104 ADDSUB w0 = u0, v0 C M I 105 ;; 106 cmp.PRED p8, p0 = w0, u0 C M I 107 ADDSUB w1 = u1, v1 C M I 108 ;; 109 cmp.PRED p9, p0 = w1, u1 C M I 110 (p7) cmp.eq.or p8, p0 = LIM, w0 C M I 111 (p7) add w0 = INCR, w0 C M I 112 ;; 113 shrp x3 = w0, w3, 1 C I0 114 ADDSUB w2 = u2, v2 C M I 115 (p8) cmp.eq.or p9, p0 = LIM, w1 C M I 116 (p8) add w1 = INCR, w1 C M I 117 br .Lcj4 C B 118 119 .grt4: ld8 v3 = [vp], 8 C M01 120 cmp.PRED p7, p0 = w3, r10 C M I 121 ld8 u3 = [up], 8 C M01 122 and r8 = 1, w3 C M I 123 ;; 124 ADDSUB w0 = u0, v0 C M I 125 ld8 v0 = [vp], 8 C M01 126 add n = -1, n 127 ;; 128 cmp.PRED p8, p0 = w0, u0 C M I 129 ld8 u0 = [up], 8 C M01 130 ADDSUB w1 = u1, v1 C M I 131 ;; 132 ld8 v1 = [vp], 8 C M01 133 mov.i ar.lc = n C I0 134 cmp.PRED p9, p0 = w1, u1 C M I 135 ld8 u1 = [up], 8 C M01 136 (p7) cmp.eq.or p8, p0 = LIM, w0 C M I 137 (p7) add w0 = INCR, w0 C M I 138 ;; 139 ADDSUB w2 = u2, v2 C M I 140 ld8 v2 = [vp], 8 C M01 141 shrp x3 = w0, w3, 1 C I0 142 (p8) cmp.eq.or p9, p0 = LIM, w1 C M I 143 (p8) add w1 = INCR, w1 C M I 144 br .LL00 C B 145 146 147 .Lb01: ADDSUB w2 = r10, r11 C M I 148 shr.u n = n, 2 C I0 149 (p15) br.dpnt .grt1 C B 150 ;; 151 152 cmp.PRED p6, p7 = w2, r10 C M I 153 shr.u x2 = w2, 1 C I0 154 and r8 = 1, w2 C M I 155 ;; 156 (p6) dep x2 = -1, x2, 63, 1 C I0 157 br .Lcj1 C B 158 159 .grt1: ld8 v3 = [vp], 8 C M01 160 ld8 u3 = [up], 8 C M01 161 ;; 162 ld8 v0 = [vp], 8 C M01 163 ld8 u0 = [up], 8 C M01 164 mov.i ar.lc = n C FIXME swap with next I0 165 ;; 166 ld8 v1 = [vp], 8 C M01 167 ld8 u1 = [up], 8 C M01 168 ;; 169 ld8 v2 = [vp], 8 C M01 170 ld8 u2 = [up], 8 C M01 171 cmp.PRED p6, p0 = w2, r10 C M I 172 and r8 = 1, w2 C M I 173 ADDSUB w3 = u3, v3 C M I 174 br.cloop.dptk .grt5 C B 175 ;; 176 177 cmp.PRED p7, p0 = w3, u3 C M I 178 ;; 179 ADDSUB w0 = u0, v0 C M I 180 (p6) cmp.eq.or p7, p0 = LIM, w3 C M I 181 (p6) add w3 = INCR, w3 C M I 182 ;; 183 cmp.PRED p8, p0 = w0, u0 C M I 184 shrp x2 = w3, w2, 1 C I0 185 ADDSUB w1 = u1, v1 C M I 186 ;; 187 cmp.PRED p9, p0 = w1, u1 C M I 188 (p7) cmp.eq.or p8, p0 = LIM, w0 C M I 189 (p7) add w0 = INCR, w0 C M I 190 br .Lcj5 C B 191 192 .grt5: ld8 v3 = [vp], 8 C M01 193 cmp.PRED p7, p0 = w3, u3 C M I 194 ld8 u3 = [up], 8 C M01 195 ;; 196 ADDSUB w0 = u0, v0 C M I 197 ld8 v0 = [vp], 8 C M01 198 (p6) cmp.eq.or p7, p0 = LIM, w3 C M I 199 (p6) add w3 = INCR, w3 C M I 200 ;; 201 cmp.PRED p8, p0 = w0, u0 C M I 202 shrp x2 = w3, w2, 1 C I0 203 ld8 u0 = [up], 8 C M01 204 ADDSUB w1 = u1, v1 C M I 205 ;; 206 ld8 v1 = [vp], 8 C M01 207 cmp.PRED p9, p0 = w1, u1 C M I 208 ld8 u1 = [up], 8 C M01 209 (p7) cmp.eq.or p8, p0 = LIM, w0 C M I 210 (p7) add w0 = INCR, w0 C M I 211 br .LL01 C B 212 213 214 .Lb10: ld8 v2 = [vp], 8 C M01 215 ld8 u2 = [up], 8 C M01 216 shr.u n = n, 2 C I0 217 ADDSUB w1 = r10, r11 C M I 218 (p15) br.dpnt .grt2 C B 219 ;; 220 221 cmp.PRED p9, p0 = w1, r10 C M I 222 and r8 = 1, w1 C M I 223 ADDSUB w2 = u2, v2 C M I 224 ;; 225 cmp.PRED p6, p0 = w2, u2 C M I 226 ;; 227 (p9) cmp.eq.or p6, p0 = LIM, w2 C M I 228 (p9) add w2 = INCR, w2 C M I 229 ;; 230 shrp x1 = w2, w1, 1 C I0 231 shr.u x2 = w2, 1 C I0 232 br .Lcj2 C B 233 234 .grt2: ld8 v3 = [vp], 8 C M01 235 ld8 u3 = [up], 8 C M01 236 ;; 237 ld8 v0 = [vp], 8 C M01 238 ld8 u0 = [up], 8 C M01 239 mov.i ar.lc = n C I0 240 ;; 241 ld8 v1 = [vp], 8 C M01 242 cmp.PRED p9, p0 = w1, r10 C M I 243 ld8 u1 = [up], 8 C M01 244 and r8 = 1, w1 C M I 245 ;; 246 ADDSUB w2 = u2, v2 C M I 247 ld8 v2 = [vp], 8 C M01 248 ;; 249 cmp.PRED p6, p0 = w2, u2 C M I 250 ld8 u2 = [up], 8 C M01 251 ADDSUB w3 = u3, v3 C M I 252 br.cloop.dptk .grt6 C B 253 ;; 254 255 cmp.PRED p7, p0 = w3, u3 C M I 256 (p9) cmp.eq.or p6, p0 = LIM, w2 C M I 257 (p9) add w2 = INCR, w2 C M I 258 ;; 259 shrp x1 = w2, w1, 1 C I0 260 ADDSUB w0 = u0, v0 C M I 261 (p6) cmp.eq.or p7, p0 = LIM, w3 C M I 262 (p6) add w3 = INCR, w3 C M I 263 br .Lcj6 C B 264 265 .grt6: ld8 v3 = [vp], 8 C M01 266 cmp.PRED p7, p0 = w3, u3 C M I 267 ld8 u3 = [up], 8 C M01 268 (p9) cmp.eq.or p6, p0 = LIM, w2 C M I 269 (p9) add w2 = INCR, w2 C M I 270 ;; 271 shrp x1 = w2, w1, 1 C I0 272 ADDSUB w0 = u0, v0 C M I 273 ld8 v0 = [vp], 8 C M01 274 (p6) cmp.eq.or p7, p0 = LIM, w3 C M I 275 (p6) add w3 = INCR, w3 C M I 276 br .LL10 C B 277 278 279 .Lb11: ld8 v1 = [vp], 8 C M01 280 ld8 u1 = [up], 8 C M01 281 shr.u n = n, 2 C I0 282 ;; 283 ld8 v2 = [vp], 8 C M01 284 ld8 u2 = [up], 8 C M01 285 ADDSUB w0 = r10, r11 C M I 286 (p15) br.dpnt .grt3 C B 287 ;; 288 289 cmp.PRED p8, p0 = w0, r10 C M I 290 ADDSUB w1 = u1, v1 C M I 291 and r8 = 1, w0 C M I 292 ;; 293 cmp.PRED p9, p0 = w1, u1 C M I 294 ;; 295 ADDSUB w2 = u2, v2 C M I 296 (p8) cmp.eq.or p9, p0 = LIM, w1 C M I 297 (p8) add w1 = INCR, w1 C M I 298 ;; 299 cmp.PRED p6, p0 = w2, u2 C M I 300 shrp x0 = w1, w0, 1 C I0 301 ;; 302 (p9) cmp.eq.or p6, p0 = LIM, w2 C M I 303 (p9) add w2 = INCR, w2 C M I 304 br .Lcj3 C B 305 306 .grt3: ld8 v3 = [vp], 8 C M01 307 ld8 u3 = [up], 8 C M01 308 ;; 309 ld8 v0 = [vp], 8 C M01 310 mov.i ar.lc = n C I0 311 cmp.PRED p8, p0 = w0, r10 C M I 312 ld8 u0 = [up], 8 C M01 313 ADDSUB w1 = u1, v1 C M I 314 and r8 = 1, w0 C M I 315 ;; 316 ld8 v1 = [vp], 8 C M01 317 cmp.PRED p9, p0 = w1, u1 C M I 318 ld8 u1 = [up], 8 C M01 319 ;; 320 ADDSUB w2 = u2, v2 C M I 321 ld8 v2 = [vp], 8 C M01 322 (p8) cmp.eq.or p9, p0 = LIM, w1 C M I 323 (p8) add w1 = INCR, w1 C M I 324 ;; 325 cmp.PRED p6, p0 = w2, u2 C M I 326 shrp x0 = w1, w0, 1 C I0 327 ld8 u2 = [up], 8 C M01 328 ADDSUB w3 = u3, v3 C M I 329 br.cloop.dptk .grt7 C B 330 ;; 331 332 cmp.PRED p7, p0 = w3, u3 C M I 333 (p9) cmp.eq.or p6, p0 = LIM, w2 C M I 334 (p9) add w2 = INCR, w2 C M I 335 br .Lcj7 C B 336 337 .grt7: ld8 v3 = [vp], 8 C M01 338 cmp.PRED p7, p0 = w3, u3 C M I 339 ld8 u3 = [up], 8 C M01 340 (p9) cmp.eq.or p6, p0 = LIM, w2 C M I 341 (p9) add w2 = INCR, w2 C M I 342 br .LL11 C B 343 344 345 C *** MAIN LOOP START *** 346 ALIGN(32) 347 .Loop: st8 [rp] = x3, 8 C M23 348 ld8 v3 = [vp], 8 C M01 349 cmp.PRED p7, p0 = w3, u3 C M I 350 ld8 u3 = [up], 8 C M01 351 (p9) cmp.eq.or p6, p0 = LIM, w2 C M I 352 (p9) add w2 = INCR, w2 C M I 353 ;; 354 .LL11: st8 [rp] = x0, 8 C M23 355 shrp x1 = w2, w1, 1 C I0 356 ADDSUB w0 = u0, v0 C M I 357 ld8 v0 = [vp], 8 C M01 358 (p6) cmp.eq.or p7, p0 = LIM, w3 C M I 359 (p6) add w3 = INCR, w3 C M I 360 ;; 361 .LL10: cmp.PRED p8, p0 = w0, u0 C M I 362 shrp x2 = w3, w2, 1 C I0 363 nop.b 0 364 ld8 u0 = [up], 8 C M01 365 ADDSUB w1 = u1, v1 C M I 366 nop.b 0 367 ;; 368 st8 [rp] = x1, 8 C M23 369 ld8 v1 = [vp], 8 C M01 370 cmp.PRED p9, p0 = w1, u1 C M I 371 ld8 u1 = [up], 8 C M01 372 (p7) cmp.eq.or p8, p0 = LIM, w0 C M I 373 (p7) add w0 = INCR, w0 C M I 374 ;; 375 .LL01: st8 [rp] = x2, 8 C M23 376 shrp x3 = w0, w3, 1 C I0 377 ADDSUB w2 = u2, v2 C M I 378 ld8 v2 = [vp], 8 C M01 379 (p8) cmp.eq.or p9, p0 = LIM, w1 C M I 380 (p8) add w1 = INCR, w1 C M I 381 ;; 382 .LL00: cmp.PRED p6, p0 = w2, u2 C M I 383 shrp x0 = w1, w0, 1 C I0 384 nop.b 0 385 ld8 u2 = [up], 8 C M01 386 ADDSUB w3 = u3, v3 C M I 387 br.cloop.dptk .Loop C B 388 ;; 389 C *** MAIN LOOP END *** 390 391 .Lskip: st8 [rp] = x3, 8 C M23 392 cmp.PRED p7, p0 = w3, u3 C M I 393 (p9) cmp.eq.or p6, p0 = LIM, w2 C M I 394 (p9) add w2 = INCR, w2 C M I 395 ;; 396 .Lcj7: st8 [rp] = x0, 8 C M23 397 shrp x1 = w2, w1, 1 C I0 398 ADDSUB w0 = u0, v0 C M I 399 (p6) cmp.eq.or p7, p0 = LIM, w3 C M I 400 (p6) add w3 = INCR, w3 C M I 401 ;; 402 .Lcj6: cmp.PRED p8, p0 = w0, u0 C M I 403 shrp x2 = w3, w2, 1 C I0 404 ADDSUB w1 = u1, v1 C M I 405 ;; 406 st8 [rp] = x1, 8 C M23 407 cmp.PRED p9, p0 = w1, u1 C M I 408 (p7) cmp.eq.or p8, p0 = LIM, w0 C M I 409 (p7) add w0 = INCR, w0 C M I 410 ;; 411 .Lcj5: st8 [rp] = x2, 8 C M23 412 shrp x3 = w0, w3, 1 C I0 413 ADDSUB w2 = u2, v2 C M I 414 (p8) cmp.eq.or p9, p0 = LIM, w1 C M I 415 (p8) add w1 = INCR, w1 C M I 416 ;; 417 .Lcj4: cmp.PRED p6, p0 = w2, u2 C M I 418 shrp x0 = w1, w0, 1 C I0 419 ;; 420 st8 [rp] = x3, 8 C M23 421 (p9) cmp.eq.or p6, p0 = LIM, w2 C M I 422 (p9) add w2 = INCR, w2 C M I 423 ;; 424 .Lcj3: st8 [rp] = x0, 8 C M23 425 shrp x1 = w2, w1, 1 C I0 426 shr.u x2 = w2, 1 C I0 427 ;; 428 .Lcj2: st8 [rp] = x1, 8 C M23 429 (p6) dep x2 = -1, x2, 63, 1 C I0 430 ;; 431 .Lcj1: st8 [rp] = x2 C M23 432 mov.i ar.lc = r2 C I0 433 br.ret.sptk.many b0 C B 434 EPILOGUE() 435
Indexes created Sun Mar 01 05:31:48 UTC 2026