| 1 |
dnl AMD64 mpn_sqr_basecase. |
| 2 |
|
| 3 |
dnl Contributed to the GNU project by Torbjorn Granlund. |
| 4 |
|
| 5 |
dnl Copyright 2008, 2009, 2011, 2012 Free Software Foundation, Inc. |
| 6 |
|
| 7 |
dnl This file is part of the GNU MP Library. |
| 8 |
dnl |
| 9 |
dnl The GNU MP Library is free software; you can redistribute it and/or modify |
| 10 |
dnl it under the terms of either: |
| 11 |
dnl |
| 12 |
dnl * the GNU Lesser General Public License as published by the Free |
| 13 |
dnl Software Foundation; either version 3 of the License, or (at your |
| 14 |
dnl option) any later version. |
| 15 |
dnl |
| 16 |
dnl or |
| 17 |
dnl |
| 18 |
dnl * the GNU General Public License as published by the Free Software |
| 19 |
dnl Foundation; either version 2 of the License, or (at your option) any |
| 20 |
dnl later version. |
| 21 |
dnl |
| 22 |
dnl or both in parallel, as here. |
| 23 |
dnl |
| 24 |
dnl The GNU MP Library is distributed in the hope that it will be useful, but |
| 25 |
dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY |
| 26 |
dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License |
| 27 |
dnl for more details. |
| 28 |
dnl |
| 29 |
dnl You should have received copies of the GNU General Public License and the |
| 30 |
dnl GNU Lesser General Public License along with the GNU MP Library. If not, |
| 31 |
dnl see https://www.gnu.org/licenses/. |
| 32 |
|
| 33 |
include(`../config.m4') |
| 34 |
|
| 35 |
C The inner loops of this code are the result of running a code generation and |
| 36 |
C optimization tool suite written by David Harvey and Torbjorn Granlund. |
| 37 |
|
| 38 |
C NOTES |
| 39 |
C * There is a major stupidity in that we call mpn_mul_1 initially, for a |
| 40 |
C large trip count. Instead, we should follow the generic/sqr_basecase.c |
| 41 |
C code which uses addmul_2s from the start, conditionally leaving a 1x1 |
| 42 |
C multiply to the end. (In assembly code, one would stop invoking |
| 43 |
C addmul_2s loops when perhaps 3x2s respectively a 2x2s remains.) |
| 44 |
C * Another stupidity is in the sqr_diag_addlsh1 code. It does not need to |
| 45 |
C save/restore carry, instead it can propagate into the high product word. |
| 46 |
C * Align more labels, should shave off a few cycles. |
| 47 |
C * We can safely use 32-bit size operations, since operands with (2^32) |
| 48 |
C limbs will lead to non-termination in practice. |
| 49 |
C * The jump table could probably be optimized, at least for non-pic. |
| 50 |
C * The special code for n <= 4 was quickly written. It is probably too |
| 51 |
C large and unnecessarily slow. |
| 52 |
C * Consider combining small cases code so that the n=k-1 code jumps into the |
| 53 |
C middle of the n=k code. |
| 54 |
C * Avoid saving registers for small cases code. |
| 55 |
C * Needed variables: |
| 56 |
C n r11 input size |
| 57 |
C i r8 work left, initially n |
| 58 |
C j r9 inner loop count |
| 59 |
C r15 unused |
| 60 |
C v0 r13 |
| 61 |
C v1 r14 |
| 62 |
C rp rdi |
| 63 |
C up rsi |
| 64 |
C w0 rbx |
| 65 |
C w1 rcx |
| 66 |
C w2 rbp |
| 67 |
C w3 r10 |
| 68 |
C tp r12 |
| 69 |
C lo rax |
| 70 |
C hi rdx |
| 71 |
C rsp |
| 72 |
|
| 73 |
C INPUT PARAMETERS |
| 74 |
define(`rp', `%rdi') |
| 75 |
define(`up', `%rsi') |
| 76 |
define(`n_param', `%rdx') |
| 77 |
|
| 78 |
define(`n', `%r11') |
| 79 |
define(`tp', `%r12') |
| 80 |
define(`i', `%r8') |
| 81 |
define(`j', `%r9') |
| 82 |
define(`v0', `%r13') |
| 83 |
define(`v1', `%r14') |
| 84 |
define(`w0', `%rbx') |
| 85 |
define(`w1', `%rcx') |
| 86 |
define(`w2', `%rbp') |
| 87 |
define(`w3', `%r10') |
| 88 |
|
| 89 |
ABI_SUPPORT(DOS64) |
| 90 |
ABI_SUPPORT(STD64) |
| 91 |
|
| 92 |
ASM_START() |
| 93 |
TEXT |
| 94 |
ALIGN(16) |
| 95 |
PROLOGUE(mpn_sqr_basecase) |
| 96 |
FUNC_ENTRY(3) |
| 97 |
mov R32(n_param), R32(%rcx) |
| 98 |
mov R32(n_param), R32(n) C free original n register (rdx) |
| 99 |
|
| 100 |
add $-40, %rsp |
| 101 |
|
| 102 |
and $3, R32(%rcx) |
| 103 |
cmp $4, R32(n_param) |
| 104 |
lea 4(%rcx), %r8 |
| 105 |
|
| 106 |
mov %rbx, 32(%rsp) |
| 107 |
mov %rbp, 24(%rsp) |
| 108 |
mov %r12, 16(%rsp) |
| 109 |
mov %r13, 8(%rsp) |
| 110 |
mov %r14, (%rsp) |
| 111 |
|
| 112 |
cmovg %r8, %rcx |
| 113 |
|
| 114 |
lea L(tab)(%rip), %rax |
| 115 |
ifdef(`PIC', |
| 116 |
` movslq (%rax,%rcx,4), %r10 |
| 117 |
add %r10, %rax |
| 118 |
jmp *%rax |
| 119 |
',` |
| 120 |
jmp *(%rax,%rcx,8) |
| 121 |
') |
| 122 |
JUMPTABSECT |
| 123 |
ALIGN(8) |
| 124 |
L(tab): JMPENT( L(4), L(tab)) |
| 125 |
JMPENT( L(1), L(tab)) |
| 126 |
JMPENT( L(2), L(tab)) |
| 127 |
JMPENT( L(3), L(tab)) |
| 128 |
JMPENT( L(0m4), L(tab)) |
| 129 |
JMPENT( L(1m4), L(tab)) |
| 130 |
JMPENT( L(2m4), L(tab)) |
| 131 |
JMPENT( L(3m4), L(tab)) |
| 132 |
TEXT |
| 133 |
|
| 134 |
L(1): mov (up), %rax |
| 135 |
mul %rax |
| 136 |
add $40, %rsp |
| 137 |
mov %rax, (rp) |
| 138 |
mov %rdx, 8(rp) |
| 139 |
FUNC_EXIT() |
| 140 |
ret |
| 141 |
|
| 142 |
L(2): mov (up), %rax |
| 143 |
mov %rax, %r8 |
| 144 |
mul %rax |
| 145 |
mov 8(up), %r11 |
| 146 |
mov %rax, (rp) |
| 147 |
mov %r11, %rax |
| 148 |
mov %rdx, %r9 |
| 149 |
mul %rax |
| 150 |
add $40, %rsp |
| 151 |
mov %rax, %r10 |
| 152 |
mov %r11, %rax |
| 153 |
mov %rdx, %r11 |
| 154 |
mul %r8 |
| 155 |
xor %r8, %r8 |
| 156 |
add %rax, %r9 |
| 157 |
adc %rdx, %r10 |
| 158 |
adc %r8, %r11 |
| 159 |
add %rax, %r9 |
| 160 |
mov %r9, 8(rp) |
| 161 |
adc %rdx, %r10 |
| 162 |
mov %r10, 16(rp) |
| 163 |
adc %r8, %r11 |
| 164 |
mov %r11, 24(rp) |
| 165 |
FUNC_EXIT() |
| 166 |
ret |
| 167 |
|
| 168 |
L(3): mov (up), %rax |
| 169 |
mov %rax, %r10 |
| 170 |
mul %rax |
| 171 |
mov 8(up), %r11 |
| 172 |
mov %rax, (rp) |
| 173 |
mov %r11, %rax |
| 174 |
mov %rdx, 8(rp) |
| 175 |
mul %rax |
| 176 |
mov 16(up), %rcx |
| 177 |
mov %rax, 16(rp) |
| 178 |
mov %rcx, %rax |
| 179 |
mov %rdx, 24(rp) |
| 180 |
mul %rax |
| 181 |
mov %rax, 32(rp) |
| 182 |
mov %rdx, 40(rp) |
| 183 |
|
| 184 |
mov %r11, %rax |
| 185 |
mul %r10 |
| 186 |
mov %rax, %r8 |
| 187 |
mov %rcx, %rax |
| 188 |
mov %rdx, %r9 |
| 189 |
mul %r10 |
| 190 |
xor %r10, %r10 |
| 191 |
add %rax, %r9 |
| 192 |
mov %r11, %rax |
| 193 |
mov %r10, %r11 |
| 194 |
adc %rdx, %r10 |
| 195 |
|
| 196 |
mul %rcx |
| 197 |
add $40, %rsp |
| 198 |
add %rax, %r10 |
| 199 |
adc %r11, %rdx |
| 200 |
add %r8, %r8 |
| 201 |
adc %r9, %r9 |
| 202 |
adc %r10, %r10 |
| 203 |
adc %rdx, %rdx |
| 204 |
adc %r11, %r11 |
| 205 |
add %r8, 8(rp) |
| 206 |
adc %r9, 16(rp) |
| 207 |
adc %r10, 24(rp) |
| 208 |
adc %rdx, 32(rp) |
| 209 |
adc %r11, 40(rp) |
| 210 |
FUNC_EXIT() |
| 211 |
ret |
| 212 |
|
| 213 |
L(4): mov (up), %rax |
| 214 |
mov %rax, %r11 |
| 215 |
mul %rax |
| 216 |
mov 8(up), %rbx |
| 217 |
mov %rax, (rp) |
| 218 |
mov %rbx, %rax |
| 219 |
mov %rdx, 8(rp) |
| 220 |
mul %rax |
| 221 |
mov %rax, 16(rp) |
| 222 |
mov %rdx, 24(rp) |
| 223 |
mov 16(up), %rax |
| 224 |
mul %rax |
| 225 |
mov %rax, 32(rp) |
| 226 |
mov %rdx, 40(rp) |
| 227 |
mov 24(up), %rax |
| 228 |
mul %rax |
| 229 |
mov %rax, 48(rp) |
| 230 |
mov %rbx, %rax |
| 231 |
mov %rdx, 56(rp) |
| 232 |
|
| 233 |
mul %r11 |
| 234 |
add $32, %rsp |
| 235 |
mov %rax, %r8 |
| 236 |
mov %rdx, %r9 |
| 237 |
mov 16(up), %rax |
| 238 |
mul %r11 |
| 239 |
xor %r10, %r10 |
| 240 |
add %rax, %r9 |
| 241 |
adc %rdx, %r10 |
| 242 |
mov 24(up), %rax |
| 243 |
mul %r11 |
| 244 |
xor %r11, %r11 |
| 245 |
add %rax, %r10 |
| 246 |
adc %rdx, %r11 |
| 247 |
mov 16(up), %rax |
| 248 |
mul %rbx |
| 249 |
xor %rcx, %rcx |
| 250 |
add %rax, %r10 |
| 251 |
adc %rdx, %r11 |
| 252 |
adc $0, %rcx |
| 253 |
mov 24(up), %rax |
| 254 |
mul %rbx |
| 255 |
pop %rbx |
| 256 |
add %rax, %r11 |
| 257 |
adc %rdx, %rcx |
| 258 |
mov 16(up), %rdx |
| 259 |
mov 24(up), %rax |
| 260 |
mul %rdx |
| 261 |
add %rax, %rcx |
| 262 |
adc $0, %rdx |
| 263 |
|
| 264 |
add %r8, %r8 |
| 265 |
adc %r9, %r9 |
| 266 |
adc %r10, %r10 |
| 267 |
adc %r11, %r11 |
| 268 |
adc %rcx, %rcx |
| 269 |
mov $0, R32(%rax) |
| 270 |
adc %rdx, %rdx |
| 271 |
|
| 272 |
adc %rax, %rax |
| 273 |
add %r8, 8(rp) |
| 274 |
adc %r9, 16(rp) |
| 275 |
adc %r10, 24(rp) |
| 276 |
adc %r11, 32(rp) |
| 277 |
adc %rcx, 40(rp) |
| 278 |
adc %rdx, 48(rp) |
| 279 |
adc %rax, 56(rp) |
| 280 |
FUNC_EXIT() |
| 281 |
ret |
| 282 |
|
| 283 |
|
| 284 |
L(0m4): |
| 285 |
lea -16(rp,n,8), tp C point tp in middle of result operand |
| 286 |
mov (up), v0 |
| 287 |
mov 8(up), %rax |
| 288 |
lea (up,n,8), up C point up at end of input operand |
| 289 |
|
| 290 |
lea -4(n), i |
| 291 |
C Function mpn_mul_1_m3(tp, up - i, i, up[-i - 1]) |
| 292 |
xor R32(j), R32(j) |
| 293 |
sub n, j |
| 294 |
|
| 295 |
mul v0 |
| 296 |
xor R32(w2), R32(w2) |
| 297 |
mov %rax, w0 |
| 298 |
mov 16(up,j,8), %rax |
| 299 |
mov %rdx, w3 |
| 300 |
jmp L(L3) |
| 301 |
|
| 302 |
ALIGN(16) |
| 303 |
L(mul_1_m3_top): |
| 304 |
add %rax, w2 |
| 305 |
mov w3, (tp,j,8) |
| 306 |
mov (up,j,8), %rax |
| 307 |
adc %rdx, w1 |
| 308 |
xor R32(w0), R32(w0) |
| 309 |
mul v0 |
| 310 |
xor R32(w3), R32(w3) |
| 311 |
mov w2, 8(tp,j,8) |
| 312 |
add %rax, w1 |
| 313 |
adc %rdx, w0 |
| 314 |
mov 8(up,j,8), %rax |
| 315 |
mov w1, 16(tp,j,8) |
| 316 |
xor R32(w2), R32(w2) |
| 317 |
mul v0 |
| 318 |
add %rax, w0 |
| 319 |
mov 16(up,j,8), %rax |
| 320 |
adc %rdx, w3 |
| 321 |
L(L3): xor R32(w1), R32(w1) |
| 322 |
mul v0 |
| 323 |
add %rax, w3 |
| 324 |
mov 24(up,j,8), %rax |
| 325 |
adc %rdx, w2 |
| 326 |
mov w0, 24(tp,j,8) |
| 327 |
mul v0 |
| 328 |
add $4, j |
| 329 |
js L(mul_1_m3_top) |
| 330 |
|
| 331 |
add %rax, w2 |
| 332 |
mov w3, (tp) |
| 333 |
adc %rdx, w1 |
| 334 |
mov w2, 8(tp) |
| 335 |
mov w1, 16(tp) |
| 336 |
|
| 337 |
lea eval(2*8)(tp), tp C tp += 2 |
| 338 |
lea -8(up), up |
| 339 |
jmp L(dowhile) |
| 340 |
|
| 341 |
|
| 342 |
L(1m4): |
| 343 |
lea 8(rp,n,8), tp C point tp in middle of result operand |
| 344 |
mov (up), v0 C u0 |
| 345 |
mov 8(up), %rax C u1 |
| 346 |
lea 8(up,n,8), up C point up at end of input operand |
| 347 |
|
| 348 |
lea -3(n), i |
| 349 |
C Function mpn_mul_2s_m0(tp, up - i, i, up - i - 1) |
| 350 |
lea -3(n), j |
| 351 |
neg j |
| 352 |
|
| 353 |
mov %rax, v1 C u1 |
| 354 |
mul v0 C u0 * u1 |
| 355 |
mov %rdx, w1 |
| 356 |
xor R32(w2), R32(w2) |
| 357 |
mov %rax, 8(rp) |
| 358 |
jmp L(m0) |
| 359 |
|
| 360 |
ALIGN(16) |
| 361 |
L(mul_2_m0_top): |
| 362 |
mul v1 |
| 363 |
add %rax, w0 |
| 364 |
adc %rdx, w1 |
| 365 |
mov -24(up,j,8), %rax |
| 366 |
mov $0, R32(w2) |
| 367 |
mul v0 |
| 368 |
add %rax, w0 |
| 369 |
mov -24(up,j,8), %rax |
| 370 |
adc %rdx, w1 |
| 371 |
adc $0, R32(w2) |
| 372 |
mul v1 C v1 * u0 |
| 373 |
add %rax, w1 |
| 374 |
mov w0, -24(tp,j,8) |
| 375 |
adc %rdx, w2 |
| 376 |
L(m0): mov -16(up,j,8), %rax C u2, u6 ... |
| 377 |
mul v0 C u0 * u2 |
| 378 |
mov $0, R32(w3) |
| 379 |
add %rax, w1 |
| 380 |
adc %rdx, w2 |
| 381 |
mov -16(up,j,8), %rax |
| 382 |
adc $0, R32(w3) |
| 383 |
mov $0, R32(w0) |
| 384 |
mov w1, -16(tp,j,8) |
| 385 |
mul v1 |
| 386 |
add %rax, w2 |
| 387 |
mov -8(up,j,8), %rax |
| 388 |
adc %rdx, w3 |
| 389 |
mov $0, R32(w1) |
| 390 |
mul v0 |
| 391 |
add %rax, w2 |
| 392 |
mov -8(up,j,8), %rax |
| 393 |
adc %rdx, w3 |
| 394 |
adc $0, R32(w0) |
| 395 |
mul v1 |
| 396 |
add %rax, w3 |
| 397 |
mov w2, -8(tp,j,8) |
| 398 |
adc %rdx, w0 |
| 399 |
L(m2x): mov (up,j,8), %rax |
| 400 |
mul v0 |
| 401 |
add %rax, w3 |
| 402 |
adc %rdx, w0 |
| 403 |
adc $0, R32(w1) |
| 404 |
add $4, j |
| 405 |
mov -32(up,j,8), %rax |
| 406 |
mov w3, -32(tp,j,8) |
| 407 |
js L(mul_2_m0_top) |
| 408 |
|
| 409 |
mul v1 |
| 410 |
add %rax, w0 |
| 411 |
adc %rdx, w1 |
| 412 |
mov w0, -8(tp) |
| 413 |
mov w1, (tp) |
| 414 |
|
| 415 |
lea -16(up), up |
| 416 |
lea eval(3*8-24)(tp), tp C tp += 3 |
| 417 |
jmp L(dowhile_end) |
| 418 |
|
| 419 |
|
| 420 |
L(2m4): |
| 421 |
lea -16(rp,n,8), tp C point tp in middle of result operand |
| 422 |
mov (up), v0 |
| 423 |
mov 8(up), %rax |
| 424 |
lea (up,n,8), up C point up at end of input operand |
| 425 |
|
| 426 |
lea -4(n), i |
| 427 |
C Function mpn_mul_1_m1(tp, up - (i - 1), i - 1, up[-i]) |
| 428 |
lea -2(n), j |
| 429 |
neg j |
| 430 |
|
| 431 |
mul v0 |
| 432 |
mov %rax, w2 |
| 433 |
mov (up,j,8), %rax |
| 434 |
mov %rdx, w1 |
| 435 |
jmp L(L1) |
| 436 |
|
| 437 |
ALIGN(16) |
| 438 |
L(mul_1_m1_top): |
| 439 |
add %rax, w2 |
| 440 |
mov w3, (tp,j,8) |
| 441 |
mov (up,j,8), %rax |
| 442 |
adc %rdx, w1 |
| 443 |
L(L1): xor R32(w0), R32(w0) |
| 444 |
mul v0 |
| 445 |
xor R32(w3), R32(w3) |
| 446 |
mov w2, 8(tp,j,8) |
| 447 |
add %rax, w1 |
| 448 |
adc %rdx, w0 |
| 449 |
mov 8(up,j,8), %rax |
| 450 |
mov w1, 16(tp,j,8) |
| 451 |
xor R32(w2), R32(w2) |
| 452 |
mul v0 |
| 453 |
add %rax, w0 |
| 454 |
mov 16(up,j,8), %rax |
| 455 |
adc %rdx, w3 |
| 456 |
xor R32(w1), R32(w1) |
| 457 |
mul v0 |
| 458 |
add %rax, w3 |
| 459 |
mov 24(up,j,8), %rax |
| 460 |
adc %rdx, w2 |
| 461 |
mov w0, 24(tp,j,8) |
| 462 |
mul v0 |
| 463 |
add $4, j |
| 464 |
js L(mul_1_m1_top) |
| 465 |
|
| 466 |
add %rax, w2 |
| 467 |
mov w3, (tp) |
| 468 |
adc %rdx, w1 |
| 469 |
mov w2, 8(tp) |
| 470 |
mov w1, 16(tp) |
| 471 |
|
| 472 |
lea eval(2*8)(tp), tp C tp += 2 |
| 473 |
lea -8(up), up |
| 474 |
jmp L(dowhile_mid) |
| 475 |
|
| 476 |
|
| 477 |
L(3m4): |
| 478 |
lea 8(rp,n,8), tp C point tp in middle of result operand |
| 479 |
mov (up), v0 C u0 |
| 480 |
mov 8(up), %rax C u1 |
| 481 |
lea 8(up,n,8), up C point up at end of input operand |
| 482 |
|
| 483 |
lea -5(n), i |
| 484 |
C Function mpn_mul_2s_m2(tp, up - i + 1, i - 1, up - i) |
| 485 |
lea -1(n), j |
| 486 |
neg j |
| 487 |
|
| 488 |
mov %rax, v1 C u1 |
| 489 |
mul v0 C u0 * u1 |
| 490 |
mov %rdx, w3 |
| 491 |
xor R32(w0), R32(w0) |
| 492 |
xor R32(w1), R32(w1) |
| 493 |
mov %rax, 8(rp) |
| 494 |
jmp L(m2) |
| 495 |
|
| 496 |
ALIGN(16) |
| 497 |
L(mul_2_m2_top): |
| 498 |
mul v1 |
| 499 |
add %rax, w0 |
| 500 |
adc %rdx, w1 |
| 501 |
mov -24(up,j,8), %rax |
| 502 |
mov $0, R32(w2) |
| 503 |
mul v0 |
| 504 |
add %rax, w0 |
| 505 |
mov -24(up,j,8), %rax |
| 506 |
adc %rdx, w1 |
| 507 |
adc $0, R32(w2) |
| 508 |
mul v1 C v1 * u0 |
| 509 |
add %rax, w1 |
| 510 |
mov w0, -24(tp,j,8) |
| 511 |
adc %rdx, w2 |
| 512 |
mov -16(up,j,8), %rax |
| 513 |
mul v0 |
| 514 |
mov $0, R32(w3) |
| 515 |
add %rax, w1 |
| 516 |
adc %rdx, w2 |
| 517 |
mov -16(up,j,8), %rax |
| 518 |
adc $0, R32(w3) |
| 519 |
mov $0, R32(w0) |
| 520 |
mov w1, -16(tp,j,8) |
| 521 |
mul v1 |
| 522 |
add %rax, w2 |
| 523 |
mov -8(up,j,8), %rax |
| 524 |
adc %rdx, w3 |
| 525 |
mov $0, R32(w1) |
| 526 |
mul v0 |
| 527 |
add %rax, w2 |
| 528 |
mov -8(up,j,8), %rax |
| 529 |
adc %rdx, w3 |
| 530 |
adc $0, R32(w0) |
| 531 |
mul v1 |
| 532 |
add %rax, w3 |
| 533 |
mov w2, -8(tp,j,8) |
| 534 |
adc %rdx, w0 |
| 535 |
L(m2): mov (up,j,8), %rax |
| 536 |
mul v0 |
| 537 |
add %rax, w3 |
| 538 |
adc %rdx, w0 |
| 539 |
adc $0, R32(w1) |
| 540 |
add $4, j |
| 541 |
mov -32(up,j,8), %rax |
| 542 |
mov w3, -32(tp,j,8) |
| 543 |
js L(mul_2_m2_top) |
| 544 |
|
| 545 |
mul v1 |
| 546 |
add %rax, w0 |
| 547 |
adc %rdx, w1 |
| 548 |
mov w0, -8(tp) |
| 549 |
mov w1, (tp) |
| 550 |
|
| 551 |
lea -16(up), up |
| 552 |
jmp L(dowhile_mid) |
| 553 |
|
| 554 |
L(dowhile): |
| 555 |
C Function mpn_addmul_2s_m2(tp, up - (i - 1), i - 1, up - i) |
| 556 |
lea 4(i), j |
| 557 |
neg j |
| 558 |
|
| 559 |
mov 16(up,j,8), v0 |
| 560 |
mov 24(up,j,8), v1 |
| 561 |
mov 24(up,j,8), %rax |
| 562 |
mul v0 |
| 563 |
xor R32(w3), R32(w3) |
| 564 |
add %rax, 24(tp,j,8) |
| 565 |
adc %rdx, w3 |
| 566 |
xor R32(w0), R32(w0) |
| 567 |
xor R32(w1), R32(w1) |
| 568 |
jmp L(am2) |
| 569 |
|
| 570 |
ALIGN(16) |
| 571 |
L(addmul_2_m2_top): |
| 572 |
add w3, (tp,j,8) |
| 573 |
adc %rax, w0 |
| 574 |
mov 8(up,j,8), %rax |
| 575 |
adc %rdx, w1 |
| 576 |
mov $0, R32(w2) |
| 577 |
mul v0 |
| 578 |
add %rax, w0 |
| 579 |
mov 8(up,j,8), %rax |
| 580 |
adc %rdx, w1 |
| 581 |
adc $0, R32(w2) |
| 582 |
mul v1 C v1 * u0 |
| 583 |
add w0, 8(tp,j,8) |
| 584 |
adc %rax, w1 |
| 585 |
adc %rdx, w2 |
| 586 |
mov 16(up,j,8), %rax |
| 587 |
mov $0, R32(w3) |
| 588 |
mul v0 C v0 * u1 |
| 589 |
add %rax, w1 |
| 590 |
mov 16(up,j,8), %rax |
| 591 |
adc %rdx, w2 |
| 592 |
adc $0, R32(w3) |
| 593 |
mul v1 C v1 * u1 |
| 594 |
add w1, 16(tp,j,8) |
| 595 |
adc %rax, w2 |
| 596 |
mov 24(up,j,8), %rax |
| 597 |
adc %rdx, w3 |
| 598 |
mul v0 |
| 599 |
mov $0, R32(w0) |
| 600 |
add %rax, w2 |
| 601 |
adc %rdx, w3 |
| 602 |
mov $0, R32(w1) |
| 603 |
mov 24(up,j,8), %rax |
| 604 |
adc $0, R32(w0) |
| 605 |
mul v1 |
| 606 |
add w2, 24(tp,j,8) |
| 607 |
adc %rax, w3 |
| 608 |
adc %rdx, w0 |
| 609 |
L(am2): mov 32(up,j,8), %rax |
| 610 |
mul v0 |
| 611 |
add %rax, w3 |
| 612 |
mov 32(up,j,8), %rax |
| 613 |
adc %rdx, w0 |
| 614 |
adc $0, R32(w1) |
| 615 |
mul v1 |
| 616 |
add $4, j |
| 617 |
js L(addmul_2_m2_top) |
| 618 |
|
| 619 |
add w3, (tp) |
| 620 |
adc %rax, w0 |
| 621 |
adc %rdx, w1 |
| 622 |
mov w0, 8(tp) |
| 623 |
mov w1, 16(tp) |
| 624 |
|
| 625 |
lea eval(2*8)(tp), tp C tp += 2 |
| 626 |
|
| 627 |
add $-2, R32(i) C i -= 2 |
| 628 |
|
| 629 |
L(dowhile_mid): |
| 630 |
C Function mpn_addmul_2s_m0(tp, up - (i - 1), i - 1, up - i) |
| 631 |
lea 2(i), j |
| 632 |
neg j |
| 633 |
|
| 634 |
mov (up,j,8), v0 |
| 635 |
mov 8(up,j,8), v1 |
| 636 |
mov 8(up,j,8), %rax |
| 637 |
mul v0 |
| 638 |
xor R32(w1), R32(w1) |
| 639 |
add %rax, 8(tp,j,8) |
| 640 |
adc %rdx, w1 |
| 641 |
xor R32(w2), R32(w2) |
| 642 |
jmp L(20) |
| 643 |
|
| 644 |
ALIGN(16) |
| 645 |
L(addmul_2_m0_top): |
| 646 |
add w3, (tp,j,8) |
| 647 |
adc %rax, w0 |
| 648 |
mov 8(up,j,8), %rax |
| 649 |
adc %rdx, w1 |
| 650 |
mov $0, R32(w2) |
| 651 |
mul v0 |
| 652 |
add %rax, w0 |
| 653 |
mov 8(up,j,8), %rax |
| 654 |
adc %rdx, w1 |
| 655 |
adc $0, R32(w2) |
| 656 |
mul v1 C v1 * u0 |
| 657 |
add w0, 8(tp,j,8) |
| 658 |
adc %rax, w1 |
| 659 |
adc %rdx, w2 |
| 660 |
L(20): mov 16(up,j,8), %rax |
| 661 |
mov $0, R32(w3) |
| 662 |
mul v0 C v0 * u1 |
| 663 |
add %rax, w1 |
| 664 |
mov 16(up,j,8), %rax |
| 665 |
adc %rdx, w2 |
| 666 |
adc $0, R32(w3) |
| 667 |
mul v1 C v1 * u1 |
| 668 |
add w1, 16(tp,j,8) |
| 669 |
adc %rax, w2 |
| 670 |
mov 24(up,j,8), %rax |
| 671 |
adc %rdx, w3 |
| 672 |
mul v0 |
| 673 |
mov $0, R32(w0) |
| 674 |
add %rax, w2 |
| 675 |
adc %rdx, w3 |
| 676 |
mov $0, R32(w1) |
| 677 |
mov 24(up,j,8), %rax |
| 678 |
adc $0, R32(w0) |
| 679 |
mul v1 |
| 680 |
add w2, 24(tp,j,8) |
| 681 |
adc %rax, w3 |
| 682 |
adc %rdx, w0 |
| 683 |
mov 32(up,j,8), %rax |
| 684 |
mul v0 |
| 685 |
add %rax, w3 |
| 686 |
mov 32(up,j,8), %rax |
| 687 |
adc %rdx, w0 |
| 688 |
adc $0, R32(w1) |
| 689 |
mul v1 |
| 690 |
add $4, j |
| 691 |
js L(addmul_2_m0_top) |
| 692 |
|
| 693 |
add w3, (tp) |
| 694 |
adc %rax, w0 |
| 695 |
adc %rdx, w1 |
| 696 |
mov w0, 8(tp) |
| 697 |
mov w1, 16(tp) |
| 698 |
|
| 699 |
lea eval(2*8)(tp), tp C tp += 2 |
| 700 |
L(dowhile_end): |
| 701 |
|
| 702 |
add $-2, R32(i) C i -= 2 |
| 703 |
jne L(dowhile) |
| 704 |
|
| 705 |
C Function mpn_addmul_2s_2 |
| 706 |
mov -16(up), v0 |
| 707 |
mov -8(up), v1 |
| 708 |
mov -8(up), %rax |
| 709 |
mul v0 |
| 710 |
xor R32(w3), R32(w3) |
| 711 |
add %rax, -8(tp) |
| 712 |
adc %rdx, w3 |
| 713 |
xor R32(w0), R32(w0) |
| 714 |
xor R32(w1), R32(w1) |
| 715 |
mov (up), %rax |
| 716 |
mul v0 |
| 717 |
add %rax, w3 |
| 718 |
mov (up), %rax |
| 719 |
adc %rdx, w0 |
| 720 |
mul v1 |
| 721 |
add w3, (tp) |
| 722 |
adc %rax, w0 |
| 723 |
adc %rdx, w1 |
| 724 |
mov w0, 8(tp) |
| 725 |
mov w1, 16(tp) |
| 726 |
|
| 727 |
C Function mpn_sqr_diag_addlsh1 |
| 728 |
lea -4(n,n), j |
| 729 |
|
| 730 |
mov 8(rp), %r11 |
| 731 |
lea -8(up), up |
| 732 |
lea (rp,j,8), rp |
| 733 |
neg j |
| 734 |
mov (up,j,4), %rax |
| 735 |
mul %rax |
| 736 |
test $2, R8(j) |
| 737 |
jnz L(odd) |
| 738 |
|
| 739 |
L(evn): add %r11, %r11 |
| 740 |
sbb R32(%rbx), R32(%rbx) C save CF |
| 741 |
add %rdx, %r11 |
| 742 |
mov %rax, (rp,j,8) |
| 743 |
jmp L(d0) |
| 744 |
|
| 745 |
L(odd): add %r11, %r11 |
| 746 |
sbb R32(%rbp), R32(%rbp) C save CF |
| 747 |
add %rdx, %r11 |
| 748 |
mov %rax, (rp,j,8) |
| 749 |
lea -2(j), j |
| 750 |
jmp L(d1) |
| 751 |
|
| 752 |
ALIGN(16) |
| 753 |
L(top): mov (up,j,4), %rax |
| 754 |
mul %rax |
| 755 |
add R32(%rbp), R32(%rbp) C restore carry |
| 756 |
adc %rax, %r10 |
| 757 |
adc %rdx, %r11 |
| 758 |
mov %r10, (rp,j,8) |
| 759 |
L(d0): mov %r11, 8(rp,j,8) |
| 760 |
mov 16(rp,j,8), %r10 |
| 761 |
adc %r10, %r10 |
| 762 |
mov 24(rp,j,8), %r11 |
| 763 |
adc %r11, %r11 |
| 764 |
nop |
| 765 |
sbb R32(%rbp), R32(%rbp) C save CF |
| 766 |
mov 8(up,j,4), %rax |
| 767 |
mul %rax |
| 768 |
add R32(%rbx), R32(%rbx) C restore carry |
| 769 |
adc %rax, %r10 |
| 770 |
adc %rdx, %r11 |
| 771 |
mov %r10, 16(rp,j,8) |
| 772 |
L(d1): mov %r11, 24(rp,j,8) |
| 773 |
mov 32(rp,j,8), %r10 |
| 774 |
adc %r10, %r10 |
| 775 |
mov 40(rp,j,8), %r11 |
| 776 |
adc %r11, %r11 |
| 777 |
sbb R32(%rbx), R32(%rbx) C save CF |
| 778 |
add $4, j |
| 779 |
js L(top) |
| 780 |
|
| 781 |
mov (up), %rax |
| 782 |
mul %rax |
| 783 |
add R32(%rbp), R32(%rbp) C restore carry |
| 784 |
adc %rax, %r10 |
| 785 |
adc %rdx, %r11 |
| 786 |
mov %r10, (rp) |
| 787 |
mov %r11, 8(rp) |
| 788 |
mov 16(rp), %r10 |
| 789 |
adc %r10, %r10 |
| 790 |
sbb R32(%rbp), R32(%rbp) C save CF |
| 791 |
neg R32(%rbp) |
| 792 |
mov 8(up), %rax |
| 793 |
mul %rax |
| 794 |
add R32(%rbx), R32(%rbx) C restore carry |
| 795 |
adc %rax, %r10 |
| 796 |
adc %rbp, %rdx |
| 797 |
mov %r10, 16(rp) |
| 798 |
mov %rdx, 24(rp) |
| 799 |
|
| 800 |
pop %r14 |
| 801 |
pop %r13 |
| 802 |
pop %r12 |
| 803 |
pop %rbp |
| 804 |
pop %rbx |
| 805 |
FUNC_EXIT() |
| 806 |
ret |
| 807 |
EPILOGUE() |