/usr/include/fflas-ffpack/fflas/fflas_simd/simd256_int64.inl is in fflas-ffpack-common 2.2.2-5.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 | /* -*- mode: C++; tab-width: 4; indent-tabs-mode: t; c-basic-offset: 4 -*- */
// vim:sts=4:sw=4:ts=4:noet:sr:cino=>s,f0,{0,g0,(0,\:0,t0,+0,=s
/*
* Copyright (C) 2014 the FFLAS-FFPACK group
*
* Written by Bastien Vialla<bastien.vialla@lirmm.fr>
* Brice Boyer (briceboyer) <boyer.brice@gmail.com>
* Romain Lebreton <romain.lebreton@lirmm.fr>
*
*
* ========LICENCE========
* This file is part of the library FFLAS-FFPACK.
*
* FFLAS-FFPACK is free software: you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
* ========LICENCE========
*.
*/
#ifndef __FFLASFFPACK_fflas_ffpack_utils_simd256_int64_INL
#define __FFLASFFPACK_fflas_ffpack_utils_simd256_int64_INL
#ifndef __FFLASFFPACK_HAVE_AVX2_INSTRUCTIONS
#error "You need AVX2 instructions to perform 256bits operations on int64_t"
#endif
/*
* Simd256 specialized for int64_t
*/
template <> struct Simd256_impl<true, true, true, 8> : public Simd256i_base {
/*
* alias to 256 bit simd register
*/
using vect_t = __m256i;
/*
* alias to 256 bit simd register
*/
using half_t = __m128i;
/*
* define the scalar type corresponding to the specialization
*/
using scalar_t = int64_t;
/*
* Simd128 for scalar_t, to deal half_t
*/
using simdHalf = Simd128<scalar_t>;
/*
* number of scalar_t in a simd register
*/
static const constexpr size_t vect_size = 4;
/*
* alignement required by scalar_t pointer to be loaded in a vect_t
*/
static const constexpr size_t alignment = 32;
/*
* Check if the pointer p is a multiple of alignemnt
*/
template <class T> static constexpr bool valid(T *p) { return (int64_t)p % alignment == 0; }
/*
* Check if the number n is a multiple of vect_size
*/
template <class T> static constexpr bool compliant(T n) { return n % vect_size == 0; }
/*
* Converter from vect_t to a tab.
* exple:
* Converter conv;
* conv.v = a;
* scalart_t x = conv.t[1]
*/
union Converter {
vect_t v;
scalar_t t[vect_size];
};
/*
* Broadcast 64-bit integer a to all elements of dst. This intrinsic may generate the vpbroadcastw.
* Return [x,x,x,x] int64_t
*/
static INLINE CONST vect_t set1(const scalar_t x) { return _mm256_set1_epi64x(x); }
/*
* Set packed 64-bit integers in dst with the supplied values.
* Return [x0,x1,x2,x3] int64_t
*/
static INLINE CONST vect_t set(const scalar_t x0, const scalar_t x1, const scalar_t x2, const scalar_t x3) {
return _mm256_set_epi64x(x3, x2, x1, x0);
}
/*
* Gather 64-bit integer elements with indexes idx[0], ..., idx[3] from the address p in vect_t.
* Return [p[idx[0]], p[idx[1]], p[idx[2]], p[idx[3]]] int64_t
*/
template <class T> static INLINE PURE vect_t gather(const scalar_t *const p, const T *const idx) {
return set(p[idx[0]], p[idx[1]], p[idx[2]], p[idx[3]]);
}
/*
* Load 256-bits of integer data from memory into dst.
* p must be aligned on a 32-byte boundary or a general-protection exception will be generated.
* Return [p[0],p[1],p[2],p[3]] int32_t
*/
static INLINE PURE vect_t load(const scalar_t *const p) {
return _mm256_load_si256(reinterpret_cast<const vect_t *>(p));
}
/*
* Load 256-bits of integer data from memory into dst.
* p does not need to be aligned on any particular boundary.
* Return [p[0],p[1],p[2],p[3]] int64_t
*/
static INLINE PURE vect_t loadu(const scalar_t *const p) {
return _mm256_loadu_si256(reinterpret_cast<const vect_t *>(p));
}
/*
* Store 256-bits of integer data from a into memory.
* p must be aligned on a 32-byte boundary or a general-protection exception will be generated.
*/
static INLINE void store(scalar_t *p, vect_t v) {
_mm256_store_si256(reinterpret_cast<vect_t *>(p), v);
}
/*
* Store 256-bits of integer data from a into memory.
* p does not need to be aligned on any particular boundary.
*/
static INLINE void storeu(scalar_t *p, vect_t v) {
_mm256_storeu_si256(reinterpret_cast<vect_t *>(p), v);
}
/*
* Store 256-bits of integer data from a into memory using a non-temporal memory hint.
* p must be aligned on a 32-byte boundary or a general-protection exception may be generated.
*/
static INLINE void stream(scalar_t *p, const vect_t v) {
_mm256_stream_si256(reinterpret_cast<vect_t *>(p), v);
}
/*
* Shift packed 64-bit integers in a left by s while shifting in zeros, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* Return : [a0 << s, a1 << s, a2 << s, a3 << s] int64_t
*/
static INLINE CONST vect_t sll(const vect_t a, const int s) { return _mm256_slli_epi64(a, s); }
/*
* Shift packed 64-bit integers in a right by s while shifting in zeros, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* Return : [a0 >> s, a1 >> s, a2 >> s, a3 >> s] int64_t
*/
static INLINE CONST vect_t srl(const vect_t a, const int s) { return _mm256_srli_epi64(a, s); }
/*
* Shift packed 64-bit integers in a right by s while shifting in sign bits, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* Return : [a0 >> s, a1 >> s, a2 >> s, a3 >> s] int64_t
*/
static INLINE CONST vect_t sra(const vect_t a, const int s) {
#ifdef __FFLASFFPACK_HAVE_AVX512F_INSTRUCTIONS
return _mm256_srai_epi64(a, s);
#else
const int b = 63 - s;
vect_t m = sll(set1(1), b);
vect_t x = srl(a, s);
vect_t result = sub(vxor(x, m), m); // result = x^m - m
return result;
#endif
}
/*
* Shuffle 64-bit integers in a using the control in imm8, and store the results in dst.
* Args : [a0, ..., a3] int32_t
* Return : [a[s[0..1]], ..., a[s[6..7]]] int32_t
*/
template<uint8_t s>
static INLINE CONST vect_t shuffle(const vect_t a) {
return _mm256_permute4x64_epi64(a, s);
}
/*
* Unpack and interleave 64-bit integers from the low half of a and b within 128-bit lanes, and store the results in dst.
* Args : [a0, a1, a2, a3] int64_t
[b0, b1, b2, b3] int64_t
* Return : [a0, b0, a2, b2] int64_t
*/
static INLINE CONST vect_t unpacklo_twice(const vect_t a, const vect_t b) { return _mm256_unpacklo_epi64(a, b); }
/*
* Unpack and interleave 64-bit integers from the high half of a and b within 128-bit lanes, and store the results in dst.
* Args : [a0, a1, a2, a3] int64_t
[b0, b1, b2, b3] int64_t
* Return : [a1, b1, a3, b3] int64_t
*/
static INLINE CONST vect_t unpackhi_twice(const vect_t a, const vect_t b) { return _mm256_unpackhi_epi64(a, b); }
/*
* Unpack and interleave 64-bit integers from the low half of a and b, and store the results in dst.
* Args : [a0, a1, a2, a3] int64_t
[b0, b1, b2, b3] int64_t
* Return : [a0, b0, a1, b1] int64_t
*/
static INLINE CONST vect_t unpacklo(const vect_t a, const vect_t b) {
vect_t a1 = shuffle<0xD8>(a); // 0xD8 = 3120 base_4 so a -> [a0,a2,a1,a3]
vect_t b1 = shuffle<0xD8>(b); // 0xD8 = 3120 base_4
return unpacklo_twice(a1, b1);
}
/*
* Unpack and interleave 64-bit integers from the high half of a and b, and store the results in dst.
* Args : [a0, a1, a2, a3] int64_t
[b0, b1, b2, b3] int64_t
* Return : [a2, b2, a3, b3] int64_t
*/
static INLINE CONST vect_t unpackhi(const vect_t a, const vect_t b) {
vect_t a1 = shuffle<0xD8>(a); // 0xD8 = 3120 base_4
vect_t b1 = shuffle<0xD8>(b); // 0xD8 = 3120 base_4
return unpackhi_twice(a1, b1);
}
/*
* Unpack and interleave 64-bit integers from the low then high half of a and b, and store the results in dst.
* Args : [a0, a1, a2, a3] int64_t
[b0, b1, b2, b3] int64_t
* Return : [a0, b0, a1, b1] int64_t
* [a2, b2, a3, b3] int64_t
*/
static INLINE CONST void unpacklohi(vect_t& l, vect_t& h, const vect_t a, const vect_t b) {
vect_t a1 = shuffle<0xD8>(a); // 0xD8 = 3120 base_4 so a -> [a0,a2,a1,a3]
vect_t b1 = shuffle<0xD8>(b); // 0xD8 = 3120 base_4
l = unpacklo_twice(a1, b1);
h = unpackhi_twice(a1, b1);
}
/*
* Blend packed 64-bit integers from a and b using control mask imm8, and store the results in dst.
* Args : [a0, a1, a2, a3] int64_t
[b0, b1, b2, b3] int64_t
* Return : [s[0]?a0:b0, , s[3]?a3:b3] int64_t
*/
template<uint8_t s>
static INLINE CONST vect_t blend(const vect_t a, const vect_t b) {
// _mm_blend_epi16 is faster than _mm_blend_epi32 and require SSE4.1 instead of AVX2
// We have to transform s = [d3 d2 d1 d0]_base2 to s1 = [d3 d3 d2 d2 d1 d1 d0 d0]_base2
constexpr uint8_t s1 = (s & 0x1) * 3 + (((s & 0x2) << 1)*3) + (((s & 0x4) << 2)*3) + (((s & 0x8) << 3)*3);
return _mm256_blend_epi32(a, b, s1);
}
/*
* Add packed 64-bits integer in a and b, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* Return : [a0+b0, a1+b1, a2+b2, a3+b3] int64_t
*/
static INLINE CONST vect_t add(const vect_t a, const vect_t b) { return _mm256_add_epi64(a, b); }
static INLINE vect_t addin(vect_t &a, const vect_t b) { return a = add(a, b); }
/*
* Subtract packed 64-bits integers in b from packed 64-bits integers in a, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* Return : [a0-b0, a1-b1, a2-b2, a3-b3] int64_t
*/
static INLINE CONST vect_t sub(const vect_t a, const vect_t b) { return _mm256_sub_epi64(a, b); }
static INLINE vect_t subin(vect_t &a, const vect_t b) { return a = sub(a, b); }
/*
* Multiply the packed 64-bits integers in a and b, producing intermediate 128-bit integers, and store the low 64
bits of the intermediate integers in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* Return : [a0*b0 smod 2^32, ..., a3*b3 smod 2^32] int64_t
* where (a smod p) is the signed representant of a modulo p, that is -p/2 <= (a smod p) < p/2
*/
static INLINE CONST vect_t mullo(vect_t a, vect_t b) {
#ifdef __FFLASFFPACK_HAVE_AVX512F_INSTRUCTIONS
return _mm256_mullo_epi64(a, b);
#else
//#pragma warning "The simd mullo function is emulate, it may impact the performances."
Converter ca, cb;
ca.v = a;
cb.v = b;
return set(ca.t[0] * cb.t[0], ca.t[1] * cb.t[1], ca.t[2] * cb.t[2], ca.t[3] * cb.t[3]);
#endif
}
static INLINE CONST vect_t mul(const vect_t a, const vect_t b) { return mullo(a, b); }
/*
* Multiply the packed 64-bits integers in a and b, producing intermediate 128-bit integers, and store the high 64
bits of the intermediate integers in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* Return : [Floor(a0*b0/2^64), ..., Floor(a3*b3/2^64)] int64_t
*/
#ifdef __FFLASFFPACK_HAVE_INT128
static INLINE CONST vect_t mulhi(vect_t a, vect_t b) {
//#pragma warning "The simd mulhi function is emulate, it may impact the performances."
// ugly solution, but it works.
// tested with gcc, clang, icc
Converter ca, cb;
ca.v = a;
cb.v = b;
return set((scalar_t)((int128_t(ca.t[0]) * cb.t[0]) >> 64), (scalar_t)((int128_t(ca.t[1]) * cb.t[1]) >> 64),
(scalar_t)((int128_t(ca.t[2]) * cb.t[2]) >> 64), (scalar_t)((int128_t(ca.t[3]) * cb.t[3]) >> 64));
}
#endif
/*
* Multiply the low 32-bits integers from each packed 64-bit element in a and b, and store the signed 64-bit results
in dst.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* Return : [(a0 smod 2^32)*(b0 smod 2^32), ..., (a3 smod 2^32)*(b3 smod 2^32)] int64_t
* where (a smod p) is the signed representant of a modulo p, that is -p/2 <= (a smod p) < p/2
*/
static INLINE CONST vect_t mulx(const vect_t a, const vect_t b) { return _mm256_mul_epi32(a, b); }
/*
* Multiply packed 64-bit integers in a and b, producing intermediate 128-bit integers, and add the low 64-bits of
the intermediate with c, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* [c0, c1, c2, c3] int64_t
* Return : [(a0*b0+c0) smod 2^64, ..., (a3*b3+c3) smod 2^64] int64_t
*/
static INLINE CONST vect_t fmadd(const vect_t c, const vect_t a, const vect_t b) { return add(c, mul(a, b)); }
static INLINE vect_t fmaddin(vect_t &c, const vect_t a, const vect_t b) { return c = fmadd(c, a, b); }
/*
* Multiply the low 32-bit integers from each packed 64-bit element in a and b,
* keep the signed 64-bit results and add the low 64-bits of c.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* [c0, c1, c2, c3] int64_t
* Return : [((a0 smod 2^32)*(b0 smod 2^32)+c0) smod 2^64, ...,
* ((a3 smod 2^32)*(b3 smod 2^32)+c3) smod 2^64] int64_t
*/
static INLINE CONST vect_t fmaddx(const vect_t c, const vect_t a, const vect_t b) { return add(c, mulx(a, b)); }
static INLINE vect_t fmaddxin(vect_t &c, const vect_t a, const vect_t b) { return c = fmaddx(c, a, b); }
/*
* Multiply the packed 64-bit integers in a and b, producing intermediate 128-bit integers,
* and substract the low 64 bits of the intermediate from elements of c.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* [c0, c1, c2, c3] int64_t
* Return : [(-a0*b0+c0) smod 2^64, ..., (-a3*b3+c3) smod 2^64] int64_t
*/
static INLINE CONST vect_t fnmadd(const vect_t c, const vect_t a, const vect_t b) { return sub(c, mul(a, b)); }
static INLINE vect_t fnmaddin(vect_t &c, const vect_t a, const vect_t b) { return c = fnmadd(c, a, b); }
/*
* Multiply the low 32-bit integers from each packed 64-bit element in a and b,
* keep the signed 64-bit results and substract them from elements of c.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* [c0, c1, c2, c3] int64_t
* Return : [(-(a0 smod 2^32)*(b0 smod 2^32)+c0) smod 2^64, ...,
* (-(a3 smod 2^32)*(b3 smod 2^32)+c3) smod 2^64] int64_t
*/
static INLINE CONST vect_t fnmaddx(const vect_t c, const vect_t a, const vect_t b) { return sub(c, mulx(a, b)); }
static INLINE vect_t fnmaddxin(vect_t &c, const vect_t a, const vect_t b) { return c = fnmaddx(c, a, b); }
/*
* Multiply the packed 64-bit integers in a and b, producing intermediate 128-bit integers,
* and substract elements of c to the low 64-bits of the intermediate.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* [c0, c1, c2, c3] int64_t
* Return : [(a0*b0-c0) smod 2^64, ..., (a3*b3-c3) smod 2^64] int64_t
*/
static INLINE CONST vect_t fmsub(const vect_t c, const vect_t a, const vect_t b) { return sub(mul(a, b), c); }
static INLINE vect_t fmsubin(vect_t &c, const vect_t a, const vect_t b) { return c = fmsub(c, a, b); }
/*
* Multiply the low 32-bit integers from each packed 64-bit element in a and b,
* keep the signed 64-bit results and substract elements of c from them.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* [c0, c1, c2, c3] int64_t
* Return : [((a0 smod 2^32)*(b0 smod 2^32)-c0) smod 2^64, ...,
* ((a3 smod 2^32)*(b3 smod 2^32)-c3) smod 2^64] int64_t
*/
static INLINE CONST vect_t fmsubx(const vect_t c, const vect_t a, const vect_t b) { return sub(mulx(a, b), c); }
static INLINE vect_t fmsubxin(vect_t &c, const vect_t a, const vect_t b) { return c = fmsubx(c, a, b); }
/*
* Compare packed 64-bits in a and b for equality, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* Return : [(a0==b0) ? 0xFFFFFFFFFFFFFFFF : 0, (a1==b1) ? 0xFFFFFFFFFFFFFFFF : 0,
(a2==b2) ? 0xFFFFFFFFFFFFFFFF : 0, (a3==b3) ? 0xFFFFFFFFFFFFFFFF : 0] int64_t
*/
static INLINE CONST vect_t eq(const vect_t a, const vect_t b) { return _mm256_cmpeq_epi64(a, b); }
/*
* Compare packed 64-bits in a and b for greater-than, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* Return : [(a0>b0) ? 0xFFFFFFFFFFFFFFFF : 0, (a1>b1) ? 0xFFFFFFFFFFFFFFFF : 0,
(a2>b2) ? 0xFFFFFFFFFFFFFFFF : 0, (a3>b3) ? 0xFFFFFFFFFFFFFFFF : 0] int64_t
*/
static INLINE CONST vect_t greater(const vect_t a, const vect_t b) {
return _mm256_cmpgt_epi64(a, b);
}
/*
* Compare packed 64-bits in a and b for lesser-than, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* Return : [(a0<b0) ? 0xFFFFFFFFFFFFFFFF : 0, (a1<b1) ? 0xFFFFFFFFFFFFFFFF : 0,
(a2<b2) ? 0xFFFFFFFFFFFFFFFF : 0, (a3<b3) ? 0xFFFFFFFFFFFFFFFF : 0] int64_t
*/
static INLINE CONST vect_t lesser(const vect_t a, const vect_t b) { return _mm256_cmpgt_epi64(b, a); }
/*
* Compare packed 64-bits in a and b for greater or equal than, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* Return : [(a0>=b0) ? 0xFFFFFFFFFFFFFFFF : 0, (a1>=b1) ? 0xFFFFFFFFFFFFFFFF : 0,
(a2>=b2) ? 0xFFFFFFFFFFFFFFFF : 0, (a3>=b3) ? 0xFFFFFFFFFFFFFFFF : 0,
(a4>=b4) ? 0xFFFFFFFFFFFFFFFF : 0, (a5>=b5) ? 0xFFFFFFFFFFFFFFFF : 0,
(a6>=b6) ? 0xFFFFFFFFFFFFFFFF : 0, (a7>=b7) ? 0xFFFFFFFFFFFFFFFF : 0] int64_t
*/
static INLINE CONST vect_t greater_eq(const vect_t a, const vect_t b) { return vor(greater(a, b), eq(a, b)); }
/*
* Compare packed 64-bits in a and b for lesser or equal than, and store the results in vect_t.
* Args : [a0, a1, a2, a3] int64_t
* [b0, b1, b2, b3] int64_t
* Return : [(a0<=b0) ? 0xFFFFFFFFFFFFFFFF : 0, (a1<=b1) ? 0xFFFFFFFFFFFFFFFF : 0,
(a2<=b2) ? 0xFFFFFFFFFFFFFFFF : 0, (a3<=b3) ? 0xFFFFFFFFFFFFFFFF : 0] int64_t
*/
static INLINE CONST vect_t lesser_eq(const vect_t a, const vect_t b) { return vor(lesser(a, b), eq(a, b)); }
/*
* Horizontally add 64-bits elements of a.
* Args : [a0, a1, a2, a3] int64_t
* Return : a0+a1+a2+a3 int64_t
*/
static INLINE CONST scalar_t hadd_to_scal(const vect_t a) {
Converter ca;
ca.v = a;
return scalar_t(ca.t[0] + ca.t[1] + ca.t[2] + ca.t[3]);
}
static INLINE CONST vect_t round(const vect_t a) { return a; }
static INLINE CONST vect_t signbits(const vect_t x) {
vect_t signBits = sub(zero(), srl(x, 4*sizeof(scalar_t)-1));
return signBits;
}
// mask the high 32 bits of a 64 bits, that is 00000000FFFFFFFF
static INLINE CONST vect_t mask_high() { return srl(_mm256_set1_epi8(-1), 32); }
static INLINE CONST vect_t mulhi_fast(vect_t x, vect_t y);
template <bool overflow, bool poweroftwo>
static INLINE vect_t mod(vect_t &C, const vect_t &P, const int8_t &shifter, const vect_t &magic, const vect_t &NEGP,
const vect_t &MIN, const vect_t &MAX, vect_t &Q, vect_t &T);
}; // Simd256_impl<true, true, true, 8>
/*
* Simd256 specialized for uint64_t
*/
template <> struct Simd256_impl<true, true, false, 8> : public Simd256_impl<true, true, true, 8> {
/*
* define the scalar type corresponding to the specialization
*/
using scalar_t = uint64_t;
/*
* Simd128 for scalar_t, to deal half_t
*/
using simdHalf = Simd128<scalar_t>;
/*
* Converter from vect_t to a tab.
* exple:
* Converter conv;
* conv.v = a;
* scalart_t x = conv.t[1]
*/
union Converter {
vect_t v;
scalar_t t[vect_size];
};
/*
* Broadcast 64-bit unsigned integer a to all elements of dst. This intrinsic may generate the vpbroadcastw.
* Return [x,x,x,x] uint64_t
*/
static INLINE CONST vect_t set1(const scalar_t x) { return _mm256_set1_epi64x(x); }
/*
* Set packed 64-bit unsigned integers in dst with the supplied values.
* Return [x0,x1,x2,x3] uint64_t
*/
static INLINE CONST vect_t set(const scalar_t x0, const scalar_t x1, const scalar_t x2, const scalar_t x3) {
return _mm256_set_epi64x(x3, x2, x1, x0);
}
/*
* Gather 64-bit unsigned integer elements with indexes idx[0], ..., idx[3] from the address p in vect_t.
* Return [p[idx[0]], p[idx[1]], p[idx[2]], p[idx[3]]] uint64_t
*/
template <class T> static INLINE PURE vect_t gather(const scalar_t *const p, const T *const idx) {
return set(p[idx[0]], p[idx[1]], p[idx[2]], p[idx[3]]);
}
/*
* Load 256-bits of unsigned integer data from memory into dst.
* p must be aligned on a 32-byte boundary or a general-protection exception will be generated.
* Return [p[0],p[1],p[2],p[3]] uint64_t
*/
static INLINE PURE vect_t load(const scalar_t *const p) {
return _mm256_load_si256(reinterpret_cast<const vect_t *>(p));
}
/*
* Load 256-bits of unsigned integer data from memory into dst.
* p does not need to be aligned on any particular boundary.
* Return [p[0],p[1],p[2],p[3]] uint64_t
*/
static INLINE PURE vect_t loadu(const scalar_t *const p) {
return _mm256_loadu_si256(reinterpret_cast<const vect_t *>(p));
}
/*
* Store 256-bits of unsigned integer data from a into memory.
* p must be aligned on a 32-byte boundary or a general-protection exception will be generated.
*/
static INLINE void store(scalar_t *p, vect_t v) {
_mm256_store_si256(reinterpret_cast<vect_t *>(p), v);
}
/*
* Store 256-bits of unsigned integer data from a into memory.
* p does not need to be aligned on any particular boundary.
*/
static INLINE void storeu(scalar_t *p, vect_t v) {
_mm256_storeu_si256(reinterpret_cast<vect_t *>(p), v);
}
/*
* Store 256-bits of unsigned integer data from a into memory using a non-temporal memory hint.
* p must be aligned on a 32-byte boundary or a general-protection exception may be generated.
*/
static INLINE void stream(scalar_t *p, const vect_t v) {
_mm256_stream_si256(reinterpret_cast<vect_t *>(p), v);
}
/*
* Shift packed 64-bit unsigned integers in a right by s while shifting in sign bits, and store the results in vect_t.
* Args : [a0, ..., a3] uint64_t
* Return : [Floor(a0/2^s), ..., Floor(a3/2^s)] uint64_t
*/
static INLINE CONST vect_t sra(const vect_t a, const int s) { return _mm256_srli_epi64(a, s); }
static INLINE CONST vect_t greater(vect_t a, vect_t b) {
vect_t x;
x = set1(-(static_cast<scalar_t>(1) << (sizeof(scalar_t) * 8 - 1)));
a = sub(x, a);
b = sub(x, b);
return _mm256_cmpgt_epi64(b,a);
}
static INLINE CONST vect_t lesser(vect_t a, vect_t b) {
vect_t x;
x = set1(-(static_cast<scalar_t>(1) << (sizeof(scalar_t) * 8 - 1)));
a = sub(x, a);
b = sub(x, b);
return _mm256_cmpgt_epi64(a, b);
}
static INLINE CONST vect_t greater_eq(const vect_t a, const vect_t b) { return vor(greater(a, b), eq(a, b)); }
static INLINE CONST vect_t lesser_eq(const vect_t a, const vect_t b) { return vor(lesser(a, b), eq(a, b)); }
/*
* Multiply the packed 64-bits integers in a and b, producing intermediate 128-bit integers, and store the low 64
bits of the intermediate integers in vect_t.
* Args : [a0, a1, a2, a3] uint64_t
[b0, b1, b2, b3] uint64_t
* Return : [a0*b0 mod 2^64, a1*b1 mod 2^64, a2*b2 mod 2^64, a3*b3 mod 2^64] uint64_t
*/
static INLINE CONST vect_t mullo(vect_t a, vect_t b) {
//#pragma warning "The simd mullo function is emulate, it may impact the performances."
Converter ca, cb;
ca.v = a;
cb.v = b;
return set(ca.t[0] * cb.t[0], ca.t[1] * cb.t[1], ca.t[2] * cb.t[2], ca.t[3] * cb.t[3]);
}
/*
* Multiply the packed 64-bits integers in a and b, producing intermediate 128-bit integers, and store the high 64
bits of the intermediate integers in vect_t.
* Args : [a0, a1, a2, a3] uint64_t
[b0, b1, b2, b3] uint64_t
* Return :
*/
#ifdef __FFLASFFPACK_HAVE_INT128
static INLINE CONST vect_t mulhi(vect_t a, vect_t b) {
//#pragma warning "The simd mulhi function is emulate, it may impact the performances."
// ugly solution, but it works.
// tested with gcc, clang, icc
Converter c0, c1;
c0.v = a;
c1.v = b;
return set((scalar_t)(((uint128_t)(c0.t[0]) * c1.t[0]) >> 64), (scalar_t)(((uint128_t)(c0.t[1]) * c1.t[1]) >> 64),
(scalar_t)(((uint128_t)(c0.t[2]) * c1.t[2]) >> 64), (scalar_t)(((uint128_t)(c0.t[3]) * c1.t[3]) >> 64));
}
#endif
/*
* Multiply the low 32-bits integers from each packed 64-bit element in a and b, and store the unsigned 64-bit
results in dst.
* Args : [a0, a1, a2, a3] uint64_t
[b0, b1, b2, b3] uint64_t
* Return : [a0*b0, a1*b1, a2*b2, a3*b3] uint64_t
*/
static INLINE CONST vect_t mulx(const vect_t a, const vect_t b) { return _mm256_mul_epu32(a, b); }
static INLINE CONST vect_t fmaddx(const vect_t c, const vect_t a, const vect_t b) { return add(c, mulx(a, b)); }
static INLINE vect_t fmaddxin(vect_t &c, const vect_t a, const vect_t b) { return c = fmaddx(c, a, b); }
static INLINE CONST vect_t fnmaddx(const vect_t c, const vect_t a, const vect_t b) { return sub(c, mulx(a, b)); }
static INLINE vect_t fnmaddxin(vect_t &c, const vect_t a, const vect_t b) { return c = fnmaddx(c, a, b); }
static INLINE CONST vect_t fmsubx(const vect_t c, const vect_t a, const vect_t b) { return sub(mulx(a, b), c); }
static INLINE vect_t fmsubxin(vect_t &c, const vect_t a, const vect_t b) { return c = fmsubx(c, a, b); }
/*
* Horizontally add 64-bits elements of a.
* Args : [a0, a1, a2, a3]
* Return : a0+a1+a2+a3
*/
static INLINE CONST scalar_t hadd_to_scal(const vect_t a) {
Converter ca;
ca.v = a;
return ca.t[0] + ca.t[1] + ca.t[2] + ca.t[3];
}
}; // Simd256_impl<true, true, false, 8>
#define vect_t Simd256_impl<true, true, true, 8>::vect_t
// warning : may be off by 1 multiple, but we save a mul...
INLINE CONST vect_t Simd256_impl<true, true, true, 8>::mulhi_fast(vect_t x, vect_t y) {
// unsigned mulhi starts:
// x1 = xy_high = mulhiu_fast(x,y)
const vect_t mask = mask_high();
vect_t x0 = vand(x, mask), x1 = srl(x, 32);
vect_t y0 = vand(y, mask), y1 = srl(y, 32);
x0 = Simd256_impl<true, true, false, 8>::mulx(x0, y1); // x0y1
y0 = Simd256_impl<true, true, false, 8>::mulx(x1, y0); // x1y0
y1 = Simd256_impl<true, true, false, 8>::mulx(x1, y1); // x1y1
x1 = vand(y0, mask);
y0 = srl(y0, 32); // x1y0_lo = x1 // y1yo_hi = y0
x1 = srl(add(x1, x0), 32);
y0 = add(y1, y0);
x1 = add(x1, y0);
// unsigned mulhi ends
// fixing signs
x0 = vand(signbits(x), y);
x1 = sub(x1, x0);
x0 = vand(signbits(y), x);
x1 = sub(x1, x0);
// end fixing
return x1;
}
template <bool overflow, bool poweroftwo>
INLINE vect_t Simd256_impl<true, true, true, 8>::mod(vect_t &C, const vect_t &P, const int8_t &shifter, const vect_t &magic, const vect_t &NEGP,
const vect_t &MIN, const vect_t &MAX, vect_t &Q, vect_t &T) {
#ifdef __INTEL_COMPILER
// Works fine with ICC 15.0.1 - A.B.
C = _mm256_rem_epi64(C, P);
#else
if (poweroftwo) {
Q = srl(C, 63);
vect_t un = set1(1);
T = sub(sll(un, shifter), un);
Q = add(C, vand(Q, T));
Q = sll(srl(Q, shifter), shifter);
C = sub(C, Q);
Q = vand(greater(zero(), Q), P);
C = add(C, Q);
} else {
Q = mulhi_fast(C, magic);
if (overflow) {
Q = add(Q, C);
}
Q = sra(Q, shifter);
vect_t q1 = Simd256_impl<true, true, false, 8>::mulx(Q, P);
vect_t q2 = sll(Simd256_impl<true, true, false, 8>::mulx(srl(Q, 32), P), 32);
C = sub(C, add(q1, q2));
T = greater_eq(C, P);
C = sub(C, vand(T, P));
}
#endif
NORML_MOD(C, P, NEGP, MIN, MAX, Q, T);
return C;
}
#undef vect_t
#endif // __FFLASFFPACK_fflas_ffpack_utils_simd256_int64_INL
|