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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269
// field.rs: part of the galois_2p8 Crate.
// Copyright 2018 Daniel Sweet. See the COPYRIGHT file at the top-level
// directory of this distribution.
//! Implements arithmetic operations over all `GF(2^8)` extensions.
//!
//! Galois (finite) fields are defined in one variable modulo some
//! prime number, or over algebraic extensions, where the members
//! are polynomials with coefficients in the one-variable field modulo
//! some irreducable polynomial.
//!
//! An irreducable polynomial is analogous to a prime number: it cannot
//! be factored as the product of two or more polynomials. Performing
//! polynomial arithmetic modulo an irreducable polynomial of degree `n`
//! ensures that all `2^n` values from `0` to `2^n - 1` are represented
//! within the extended field.
//!
//! Algebraic extensions to Galois fields can be expressed as operations
//! modulo several potential irreducable polynomials, except for the special
//! case of `GF(2^2)`, which can only be represented in terms of one
//! irreducable polynomial. This crate implements field arithmetic modulo
//! all possible irreducable polynomials capable of generating `GF(2^8)`.
use std::ptr;
#[cfg(all(feature="simd", target_arch="x86_64"))]
use std::arch::x86_64;
/// Represents an irreducable polynomial of `GF(2^8)`.
///
/// Each polynomial is named according to the nonzero positions of
/// the coefficients. Each digit after the `Poly` prefix corresponds to
/// the exponent of the variable for which a nonzero coefficient is
/// present. Recall that in `GF(2^x)`, the only possible coefficients are
/// either `0` or `1`.
///
/// For example, [`Poly84310`] represents `x^8 + x^4 + x^3 + x + 1`.
///
/// [`Poly84310`]: #variant.Poly84310
#[derive(Clone, Copy, Debug, Eq, Hash, Ord, PartialEq, PartialOrd)]
pub enum IrreducablePolynomial {
Poly84310 = 0x1b,
Poly84320 = 0x1d, // Primitive polynomial
Poly85310 = 0x2b, // Primitive polynomial
Poly85320 = 0x2d, // Primitive polynomial
Poly85430 = 0x39,
Poly8543210 = 0x3f,
Poly86320 = 0x4d, // Primitive polynomial
Poly8643210 = 0x5f, // Primitive polynomial
Poly86510 = 0x63, // Primitive polynomial
Poly86520 = 0x65, // Primitive polynomial
Poly86530 = 0x69, // Primitive polynomial
Poly86540 = 0x71, // Primitive polynomial
Poly8654210 = 0x77,
Poly8654310 = 0x7b,
Poly87210 = 0x87, // Primitive polynomial
Poly87310 = 0x8b,
Poly87320 = 0x8d, // Primitive polynomial
Poly8743210 = 0x9f,
Poly87510 = 0xa3,
Poly87530 = 0xa9, // Primitive polynomial
Poly87540 = 0xb1,
Poly8754320 = 0xbd,
Poly87610 = 0xc3, // Primitive polynomial
Poly8763210 = 0xcf, // Primitive polynomial
Poly8764210 = 0xd7,
Poly8764320 = 0xdd,
Poly8765210 = 0xe7, // Primitive polynomial
Poly8765410 = 0xf3,
Poly8765420 = 0xf5, // Primitive polynomial
Poly8765430 = 0xf9,
}
/// Contains all possible irreducable polynomials for `GF(2^8)`.
///
/// This array is exposed primarily for testing purposes, but may be used
/// to enumerate all possible values in the order of their declaration.
pub const POLYNOMIALS: [IrreducablePolynomial; 30] = [
IrreducablePolynomial::Poly84310,
IrreducablePolynomial::Poly84320,
IrreducablePolynomial::Poly85310,
IrreducablePolynomial::Poly85320,
IrreducablePolynomial::Poly85430,
IrreducablePolynomial::Poly8543210,
IrreducablePolynomial::Poly86320,
IrreducablePolynomial::Poly8643210,
IrreducablePolynomial::Poly86510,
IrreducablePolynomial::Poly86520,
IrreducablePolynomial::Poly86530,
IrreducablePolynomial::Poly86540,
IrreducablePolynomial::Poly8654210,
IrreducablePolynomial::Poly8654310,
IrreducablePolynomial::Poly87210,
IrreducablePolynomial::Poly87310,
IrreducablePolynomial::Poly87320,
IrreducablePolynomial::Poly8743210,
IrreducablePolynomial::Poly87510,
IrreducablePolynomial::Poly87530,
IrreducablePolynomial::Poly87540,
IrreducablePolynomial::Poly8754320,
IrreducablePolynomial::Poly87610,
IrreducablePolynomial::Poly8763210,
IrreducablePolynomial::Poly8764210,
IrreducablePolynomial::Poly8764320,
IrreducablePolynomial::Poly8765210,
IrreducablePolynomial::Poly8765410,
IrreducablePolynomial::Poly8765420,
IrreducablePolynomial::Poly8765430,
];
/// Contains the primitive polynomials of `GF(2^8)`.
///
/// In `GF(2^x)` wherein arithmetic operations are performed modulo a
/// polynomial, the polynomial is said to be primitive if for some alpha,
/// each nonzero member value of the field can be uniquely represented by
/// `alpha ^ p` for `p < 2^x`, where alpha is a root of the polynomial. That
/// is, for some root of the polynomial, every member of the field can
/// be represented as the exponentiation of said root.
///
/// In `GF(2^x)`, the only nontrivial prime root of any given
/// [`IrreducablePolynomial`] is two. We say "is primitive" as a shorthand
/// for meaning that the [`IrreducablePolynomial`] is primitive assuming
/// a root of two.
///
/// The use of primitive polynomials confers an immediate performance
/// benefit for single values: we can represent multiplication and
/// division as addition and subtraction within logarithms and exponents.
/// Additionally, some usages of Galois fields, e.g. Reed-Solomon syndrome
/// coding, require primitive polynomials to function properly.
///
/// This table is used by the [`is_primitive`] method. Specifically,
///
/// ```rust
/// # use galois_2p8::field::{IrreducablePolynomial, PRIMITIVES};
/// # let poly = IrreducablePolynomial::Poly84320;
/// # assert_eq!(
/// PRIMITIVES.binary_search(&poly).is_ok() == poly.is_primitive()
/// # , true);
/// ```
///
/// [`IrreducablePolynomial`]: enum.IrreducablePolynomial.html
/// [`is_primitive`]: enum.IrreducablePolynomial.html#method.is_primitive
pub const PRIMITIVES: [IrreducablePolynomial; 16] = [
IrreducablePolynomial::Poly84320,
IrreducablePolynomial::Poly85310,
IrreducablePolynomial::Poly85320,
IrreducablePolynomial::Poly86320,
IrreducablePolynomial::Poly8643210,
IrreducablePolynomial::Poly86510,
IrreducablePolynomial::Poly86520,
IrreducablePolynomial::Poly86530,
IrreducablePolynomial::Poly86540,
IrreducablePolynomial::Poly87210,
IrreducablePolynomial::Poly87320,
IrreducablePolynomial::Poly87530,
IrreducablePolynomial::Poly87610,
IrreducablePolynomial::Poly8763210,
IrreducablePolynomial::Poly8765210,
IrreducablePolynomial::Poly8765420,
];
impl IrreducablePolynomial {
/// Converts the [`IrreducablePolynomial`] to its binary representation.
///
/// In `GF(2^x)`, coefficients in the extension polynomial may only take
/// values of `0` or `1`. As a result, there is a natural mapping of values
/// in `GF(2^x)` to bits. `0b1011` maps to `x^3 + x^1 + 1`, for example.
/// The greatest degree that can be encoded by an eight-bit byte is
/// 7, as a consequence of the least significant bit being treated as
/// `x^0 == 1`. In order to represent a degree 8 polynomial, there must be
/// at least 9 bits available, so it can be encoded as a 16-bit value.
///
/// [`IrreducablePolynomial`]: enum.IrreducablePolynomial.html
pub fn to_u16(&self) -> u16 {
0x100 + ((*self) as u16)
}
/// Determines whether the [`IrreducablePolynomial`] is a primitive polynomial.
///
/// In `GF(2^x)` wherein arithmetic operations are performed modulo a
/// polynomial, the polynomial is said to be primitive if for some alpha,
/// each nonzero member value of the field can be uniquely represented by
/// `alpha ^ p` for `p < 2^x`, where alpha is a root of the polynomial. That
/// is, for some root of the polynomial, every member of the field can
/// be represented as the exponentiation of said root.
///
/// In `GF(2^x)`, the only nontrivial prime root of any given
/// [`IrreducablePolynomial`] is two. We say "is primitive" as a shorthand
/// for meaning that the [`IrreducablePolynomial`] is primitive assuming
/// a root of two.
///
/// The use of primitive polynomials confers an immediate performance
/// benefit for single values: we can represent multiplication and
/// division as addition and subtraction within logarithms and exponents.
/// Additionally, some usages of Galois fields, e.g. Reed-Solomon syndrome
/// coding, require primitive polynomials to function properly.
///
/// This method consults the [`PRIMITIVES`] table to determine if the
/// [`IrreducablePolynomial`] is actually primitive. That is to say,
///
/// ```rust
/// # use galois_2p8::field::{IrreducablePolynomial, PRIMITIVES};
/// # let poly = IrreducablePolynomial::Poly84320;
/// # assert_eq!(
/// poly.is_primitive() == PRIMITIVES.binary_search(&poly).is_ok()
/// # , true);
/// ```
///
/// [`IrreducablePolynomial`]: enum.IrreducablePolynomial.html
/// [`PRIMITIVES`]: constant.PRIMITIVES.html
pub fn is_primitive(&self) -> bool {
match PRIMITIVES.binary_search(self) {
Ok(_) => true,
Err(_) => false
}
}
}
/// Establishes `GF(2^8)` arithmetic for scalar and vector operands.
///
/// In all instances of `GF(2^8)`, over every possible [`IrreducablePolynomial`],
/// addition and subtraction is defined as XOR, as in `GF(2)`. Addition and
/// subtraction are accordingly provided as default implementations of
/// this trait.
///
/// Multiplication and division are more complicated, and the optimal strategy
/// for implementing them in a scalar context depends on whether the
/// [`IrreducablePolynomial`] over which the field is implemented is a primitive
/// polynomial.
///
/// Recall that if a `p: IrreducablePolynomial` is primitive, then all members of
/// the field in which operations are performed modulo `p` can be represented
/// as `2^n` for `n in [0..255]`, with the exception of `0`.
///
/// In these cases, we can represent multiplication and division as
/// addition and subtraction within logarithmic representations of the operands.
/// This requires fewer instructions to implement at the scalar level.
/// Note that this cannot be done for an [`IrreducablePolynomial`] that is not
/// also primitive. As a consequence, we provide two concrete implementations
/// of this trait: [`GeneralField`] and [`PrimitivePolynomialField`], where the
/// slightly faster logarithm arithmetic is only used in the latter.
///
/// This trait also exposes operations over vectors containing `GF(2^8)`
/// members.
///
/// Common operations over `GF(2^8)` operands can exploit long-word vector
/// operations as implemented by the target hardware. A trivial example
/// is the addition and subtraction of vectors: this is a simple bitwise
/// XOR across a very long word. This already functions as expected
/// in Rust 1.25 as a consequence of LLVM optimizations. A less trivial
/// example involves multiplication and division: vector processors require
/// a specialized long-word lookup function to implement these operations.
///
/// The `x86_64` architecture mandates SSE4.2 and earlier, as is found in the
/// earlier `x86` architecture; in SSE3, an intrinsic `_mm_shuffle_epi8` was
/// added that allows the entries of a vector register `a` to function as
/// indices of the vector register `b` in the lower four bits, effectively
/// implementing an accelerated 16-entry table lookup. These SSE3 intrinsics
/// are used for multiword operations if the `"simd"` feature is enabled.
///
/// As of Rust 1.27.2, code generation for AVX on the default ABI results in
/// the generation of incorrect code. Because of this, `galois_2p8` only uses
/// AVX 2 intrinsics for optimized multiplication and division if the `rustc`
/// target feature `avx2` is enabled, e.g. exporting
/// `RUSTFLAGS="-C target-feature=avx2` before running `rustc` or `cargo`.
///
/// [`IrreducablePolynomial`]: enum.IrreducablePolynomial.html
/// [`GeneralField`]: struct.GeneralField.html
/// [`PrimitivePolynomialField`]: struct.PrimitivePolynomialField.html
pub trait Field {
/// Returns the polynomial modulo which all operations are performed.
fn polynomial(&self) -> IrreducablePolynomial;
/// Returns the result of `src * scale` in this field.
fn mult(&self, src: u8, scale: u8) -> u8;
/// Returns the result of `src / scale` in this field.
///
/// Implementations of this method are expected to panic if the `scale`
/// argument is zero. The contents of the resulting error message are
/// not defined.
fn div(&self, src: u8, scale: u8) -> u8;
/// Returns the result of `2^x` in this field.
fn two_pow(&self, x: u8) -> u8;
/// Returns the result of `scale * 2^x` in this field.
fn mult_two_pow(&self, scale: u8, x: u8) -> u8;
/// Adds `scale * src[0..len]` into `dst[0..len]` in place.
unsafe fn add_ptr_scaled_len(
&self,
dst: *mut u8,
src: *const u8,
scale: u8,
len: usize
);
/// Multiplies `dst[0..len]` by `scale` in place.
unsafe fn mult_ptr_len(
&self,
dst: *mut u8,
scale: u8,
len: usize
);
/// Divides `dst[0..len]` by `scale` in place.
///
/// Implementations of this method are expected to panic if the `scale`
/// argument is zero. The contents of the resulting error message are
/// not defined.
unsafe fn div_ptr_len(
&self,
dst: *mut u8,
scale: u8,
len: usize
);
// Basic arithmetic
/// Adds `left` and `right`, returning their sum.
fn add(&self, left: u8, right: u8) -> u8 {
left ^ right
}
/// Subtracts `right` from `left`, returning the difference.
fn sub(&self, left: u8, right: u8) -> u8 {
left ^ right
}
// Multiword operations. These will eventually take advantage of SIMD
// intrinsics, when those become stable.
/// Adds `src[0..len]` into `dst[0..len]`.
unsafe fn add_ptr_len(
&self,
dst: *mut u8,
src: *const u8,
len: usize
) {
for i in 0..len {
let dst_ptr = dst.offset(i as isize);
let src_ptr = src.offset(i as isize);
*dst_ptr ^= *src_ptr;
}
}
/// Adds `src` into `dst` in place, over the smallest common length.
///
/// The length used in operation is set to the minimum of `src.len()` and
/// `dst.len()`.
fn add_multiword(
&self,
dst: &mut [u8],
src: &[u8]
) {
let dlen = dst.len();
let slen = src.len();
self.add_multiword_len(dst, src, slen.min(dlen));
}
/// Adds `src[0..len]` into `dst[0..len]`.
///
/// This method will panic if `src.len()` or `dst.len()` is less than
/// the supplied `len` parameter.
fn add_multiword_len(
&self,
dst: &mut [u8],
src: &[u8],
len: usize
) {
assert!(len <= dst.len());
assert!(len <= src.len());
unsafe {
self.add_ptr_len(dst.as_mut_ptr(), src.as_ptr(), len);
}
}
/// Adds `src * scale` into `dst` in place, over the smallest common length.
///
/// The length used in the operation is set to the minimum of `src.len()` and
/// `dst.len()`.
fn add_scaled_multiword(
&self,
dst: &mut [u8],
src: &[u8],
scale: u8
) {
let dlen = dst.len();
let slen = src.len();
self.add_scaled_multiword_len(dst, src, scale, slen.min(dlen));
}
/// Adds `src[0..len] * scale` into `dst[0..len]`.
///
/// This method will panic if `src.len()` or `dst.len()` is less than
/// the supplied `len` parameter.
fn add_scaled_multiword_len(
&self,
dst: &mut [u8],
src: &[u8],
scale: u8,
len: usize
) {
assert!(len <= dst.len());
assert!(len <= src.len());
unsafe {
self.add_ptr_scaled_len(dst.as_mut_ptr(), src.as_ptr(), scale, len);
}
}
/// Subtracts `src[0..len]` from `dst[0..len]` in place.
unsafe fn sub_ptr_len(
&self,
dst: *mut u8,
src: *const u8,
len: usize
) {
self.add_ptr_len(dst, src, len);
}
/// Subracts `scale * src[0..len]` from `dst[0..len]` in place.
unsafe fn sub_ptr_scaled_len(
&self,
dst: *mut u8,
src: *const u8,
scale: u8,
len: usize
) {
self.add_ptr_scaled_len(dst, src, scale, len);
}
/// Subtracts `src` from `dst` in place, over the smallest common length.
///
/// The length used in the operation is set to the minimum of `src.len()` and
/// `dst.len()`.
fn sub_multiword(
&self,
dst: &mut [u8],
src: &[u8]
) {
self.add_multiword(dst, src);
}
/// Subtracts `src[0..len]` from `dst[0..len]` in place.
///
/// This method will panic if `src.len()` or `dst.len()` is less than
/// the supplied `len` parameter.
fn sub_multiword_len(
&self,
dst: &mut [u8],
src: &[u8],
len: usize
) {
self.add_multiword_len(dst, src, len);
}
/// Subtracts `scale * src` from `dst` in place, over the smallest common length.
///
/// The length used in the operation is set to the minimum of `src.len()` and
/// `dst.len()`.
fn sub_scaled_multiword(
&self,
dst: &mut [u8],
src: &[u8],
scale: u8
) {
self.add_scaled_multiword(dst, src, scale);
}
/// Subtracts `scale * src[0..len]` from `dst[0..len]` in place.
///
/// This method will panic if `src.len()` or `dst.len()` is less than
/// the supplied `len` parameter.
fn sub_scaled_multiword_len(
&self,
dst: &mut [u8],
src: &[u8],
scale: u8,
len: usize
) {
self.add_scaled_multiword_len(dst, src, scale, len);
}
/// Multiplies `dst` by `scale` in place.
fn mult_multiword(
&self,
dst: &mut [u8],
scale: u8
) {
unsafe {
self.mult_ptr_len(dst.as_mut_ptr(), scale, dst.len());
}
}
/// Divides `dst` by `scale` in place.
///
/// This method will panic if `scale` is zero. The contents of the
/// resulting error message are not defined.
fn div_multiword(
&self,
dst: &mut [u8],
scale: u8
) {
unsafe {
self.div_ptr_len(dst.as_mut_ptr(), scale, dst.len());
}
}
}
// SIMD support. Currently limited to x86_64 and SSE or AVX 2.
// The simd_scale_vec and simd_scale_vec_into functions are expected
// to be used on all SIMD platforms.
// simd_scale_vec_writeback is x86_64 specific, so we don't use it in
// any client code, only in simd_scale_vec and simd_scale_vec_into.
#[cfg(all(feature="simd", target_arch="x86_64"))]
unsafe fn simd_scale_vec_writeback(
scale_table: *const u8,
mut dst: *mut u8,
mut src: *const u8,
mut len: usize,
scale: u8,
_write_func_32: impl Fn(*mut u8, x86_64::__m256i),
write_func_16: impl Fn(*mut u8, x86_64::__m128i)
) -> usize {
let scale_offset = scale_table.offset(scale as isize * 32);
let scale_reg_lower = x86_64::_mm_loadu_si128(
scale_offset as *const x86_64::__m128i
);
let scale_reg_upper = x86_64::_mm_loadu_si128(
scale_offset.offset(16) as *const x86_64::__m128i
);
let mask = x86_64::_mm_set1_epi8(0x0f);
// In Rust 1.27, the x86_64 SIMD support was marked "stable", but due to
// code generation bugs in LLVM, if the AVX ABI was not enabled, rustc
// would generate incorrect assembly. As of this writing (2018-11-27),
// there's a race to see who is actually fixing the bug: rustc or LLVM.
//
// Even assuming the code generation works properly, calling into the
// AVX intrinsic functions without targeting the AVX ABI is likely going
// to be slower due to function calling instead of just executing the
// assembly. So, for AVX 2 support at all, you'll have to enable the AVX 2
// ABI.
#[cfg(target_feature="avx2")]
{
if is_x86_feature_detected!("avx2") && len >= 32 {
let srl_256 = x86_64::_mm256_broadcastsi128_si256(scale_reg_lower);
let sru_256 = x86_64::_mm256_broadcastsi128_si256(scale_reg_upper);
let mask_256 = x86_64::_mm256_broadcastsi128_si256(mask);
while len >= 32 {
let window = x86_64::_mm256_loadu_si256(
src as *const x86_64::__m256i
);
let low_portion = x86_64::_mm256_and_si256(
window,
mask_256
);
let low_value = x86_64::_mm256_shuffle_epi8(
srl_256,
low_portion
);
let high_portion = x86_64::_mm256_and_si256(
x86_64::_mm256_srli_epi16(window, 4),
mask_256
);
let high_value = x86_64::_mm256_shuffle_epi8(
sru_256,
high_portion
);
_write_func_32(dst, x86_64::_mm256_xor_si256(
high_value, low_value
));
src = src.offset(32);
dst = dst.offset(32);
len -= 32;
}
}
}
while len >= 16 {
let window = x86_64::_mm_loadu_si128(
src as *const x86_64::__m128i
);
let low_portion = x86_64::_mm_and_si128(
window,
mask
);
let low_value = x86_64::_mm_shuffle_epi8(
scale_reg_lower,
low_portion
);
let high_portion = x86_64::_mm_and_si128(
x86_64::_mm_srli_epi16(window, 4),
mask
);
let high_value = x86_64::_mm_shuffle_epi8(
scale_reg_upper,
high_portion
);
write_func_16(dst, x86_64::_mm_xor_si128(high_value, low_value));
src = src.offset(16);
dst = dst.offset(16);
len -= 16;
}
return len;
}
#[cfg(all(feature="simd", target_arch="x86_64"))]
unsafe fn simd_scale_vec(
scale_table: *const u8,
dst: *mut u8,
src: *const u8,
len: usize,
scale: u8
) -> usize {
use std::arch::x86_64;
simd_scale_vec_writeback(
scale_table,
dst,
src,
len,
scale,
|dst_ptr, writeback| {
x86_64::_mm256_storeu_si256(
dst_ptr as *mut x86_64::__m256i,
writeback
)
},
|dst_ptr, writeback| {
x86_64::_mm_storeu_si128(
dst_ptr as *mut x86_64::__m128i,
writeback
)
}
)
}
#[cfg(all(feature="simd", target_arch="x86_64"))]
unsafe fn simd_scale_vec_into(
scale_table: *const u8,
dst: *mut u8,
src: *const u8,
len: usize,
scale: u8
) -> usize {
use std::arch::x86_64;
simd_scale_vec_writeback(
scale_table,
dst,
src,
len,
scale,
|dst, writeback| {
let dst_ptr = dst as *mut x86_64::__m256i;
let dst_value = x86_64::_mm256_loadu_si256(dst_ptr);
let added = x86_64::_mm256_xor_si256(writeback, dst_value);
x86_64::_mm256_storeu_si256(dst_ptr, added);
},
|dst, writeback| {
let dst_ptr = dst as *mut x86_64::__m128i;
let dst_value = x86_64::_mm_loadu_si128(dst_ptr);
let added = x86_64::_mm_xor_si128(writeback, dst_value);
x86_64::_mm_storeu_si128(dst_ptr, added);
}
)
}
// SIMD operations function best over a traditional multiply/divide table.
// In the case of GeneralField, this is easy enough to support, because
// those implementations use traditional multiply/divide tables anyway.
// In the case of PrimitivePolynomialField, we'll need extra multiply/
// divide tables in addition to the exponent/logarithm tables. However,
// the multiply/divide tables can be constructed the same way for both
// implementations.
fn construct_mult_div_tables(
poly: IrreducablePolynomial
) -> (Vec<u8>, Vec<u8>) {
let table_dim = 1 << 13;
let mut mult_table = Vec::with_capacity(table_dim);
let mut div_table = Vec::with_capacity(table_dim);
unsafe {
// We're fine with pushing to mult table, but div_table
// needs random access.
div_table.set_len(table_dim);
}
// Handle the mult table.
for alpha in 0..=255 {
for low in 0..16 {
let product = gf2_mult_mod(alpha, low, poly);
mult_table.push(product);
}
for lowshift in 0..16 {
let high = lowshift << 4;
let product = gf2_mult_mod(alpha, high, poly);
mult_table.push(product);
}
}
// These first 32 entries are n/0, which is undefined, but worth
// clearing out anyway in the event of some OOB access bug.
for i in 0..32 {
div_table[i] = 0;
}
// Handle the rest of the div table.
for a in 1..=255 {
let mt_offset = a * 32;
// Handle 0/n cases separately in div_table for the high-quad access:
// we're never going to see 0 fill out the high-quads from in a
// multiplication so we have to explicitly clear it out here.
div_table[mt_offset + 16] = 0;
for b_low_offset in 0..16 {
let prod_b_low = mult_table[mt_offset + b_low_offset];
for b_high_offset in 16..32 {
let prod_b_high = mult_table[mt_offset + b_high_offset];
let prod = prod_b_low ^ prod_b_high;
// In here, we have a * b = prod. So we can store
// both prod / a = b and prod / b = a, which is
// what we do with the div_b_offset and div_a_offset
// into the div table.
let b_high = ((b_high_offset - 16) << 4) as u8;
let b = b_high + (b_low_offset as u8);
if prod < 16 {
let div_b_offset = (b as usize) * 32 + prod as usize;
div_table[div_b_offset] = a as u8;
let div_a_offset = mt_offset + prod as usize;
div_table[div_a_offset] = b;
} else if prod & 0x0f == 0 {
let prod_offset = (prod >> 4) as usize + 16;
let div_b_offset = (b as usize) * 32 + prod_offset;
div_table[div_b_offset] = a as u8;
let div_a_offset = mt_offset + prod_offset;
div_table[div_a_offset] = b;
}
}
}
}
(mult_table, div_table)
}
/// Implements field arithmetic compatible with all [`IrreducablePolynomial`]s.
///
/// Recall that there are two strategies for optimizing field arithmetic in
/// `GF(2^8)`: accessing direct multiplication and division tables, and
/// manipulating logarithms and exponentials. The latter method requires fewer
/// operations, but is only possible if the given [`IrreducablePolynomial`] is
/// primitive.
///
/// This struct uses direct multiplication and division tables for its
/// operations, and is compatible with all [`IrreducablePolynomial`]s,
/// including primitive ones. Operations are expected to be less performant
/// than those implemented by [`PrimitivePolynomialField`], with the exception
/// of vector operations, which may be accelerated in both using multiplication
/// and division tables. See the [`Field`] documentation for more details
/// regarding vector operations.
///
/// [`IrreducablePolynomial`]: enum.IrreducablePolynomial.html
/// [`PrimitivePolynomialField`]: struct.PrimitivePolynomialField.html
/// [`Field`]: trait.Field.html
pub struct GeneralField {
modulo: IrreducablePolynomial,
mult_table: Vec<u8>,
div_table: Vec<u8>,
exp_table: Vec<u8>,
pmult_table: *const u8,
pdiv_table: *const u8,
pexp_table: *const u8
}
// This is ok because instance data is not modified after construction.
unsafe impl Sync for GeneralField {}
impl GeneralField {
/// Constructs a new `GeneralField` with all tables initialized.
pub fn new(poly: IrreducablePolynomial) -> Self {
let mut exp_table = Vec::with_capacity(1 << 8);
let mut two_x = 1;
let (mult_table, div_table) = construct_mult_div_tables(poly);
for alpha in 0..=255 {
exp_table.push(two_x);
two_x = gf2_mult_mod(alpha, 2, poly);
}
let mut ret = Self {
modulo: poly,
mult_table: mult_table,
div_table: div_table,
exp_table: exp_table,
pmult_table: ptr::null(),
pdiv_table: ptr::null(),
pexp_table: ptr::null()
};
ret.pmult_table = ret.mult_table.as_ptr();
ret.pdiv_table = ret.div_table.as_ptr();
ret.pexp_table = ret.exp_table.as_ptr();
ret
}
}
impl Field for GeneralField {
fn polynomial(&self) -> IrreducablePolynomial {
self.modulo
}
// For primitive polynomials, you can take the logarithms
// and perform addition and subtraction with them, but this
// is not true of non-primitive polynomials.
fn mult(&self, src: u8, scale: u8) -> u8 {
if src == 0 || scale == 0 {
return 0;
}
unsafe {
let src_low = src & 0x0f;
let src_high = (src & 0xf0) >> 4;
let basis = self.pmult_table.offset(scale as isize * 32);
let prod_low = *basis.offset(src_low as isize);
let prod_high = *basis.offset(16 + src_high as isize);
self.add(prod_high, prod_low)
}
}
fn div(&self, src: u8, scale: u8) -> u8 {
if scale == 0 {
panic!("Can't divide {}/0", src);
}
if src == 0 {
return 0;
}
unsafe {
let src_low = src & 0x0f;
let src_high = (src & 0xf0) >> 4;
let basis = self.pdiv_table.offset(scale as isize * 32);
let quot_low = *basis.offset(src_low as isize);
let quot_high = *basis.offset(16 + src_high as isize);
self.add(quot_low, quot_high)
}
}
// Special support for powers of 2
// It is important to note that powers of two are most useful when the
// polynomial is primitive: in these cases, powers of two from 0 through 255
// generate the field, and 2^256 is congruent to 1, so the powers are
// cyclical.
// Again, that's only for primitive polynomials.
fn two_pow(&self, x: u8) -> u8 {
unsafe {
*self.pexp_table.offset(x as isize)
}
}
fn mult_two_pow(&self, scale: u8, x: u8) -> u8 {
if scale == 0 {
return 0;
}
if x == 0 {
return scale;
}
self.mult(scale, self.two_pow(x))
}
unsafe fn add_ptr_scaled_len(
&self,
mut dst: *mut u8,
mut src: *const u8,
scale: u8,
mut len: usize
) {
if scale == 0 {
return;
}
if scale == 1 {
self.add_ptr_len(dst, src, len);
return;
}
#[cfg(feature="simd")]
{
let left = simd_scale_vec_into(
self.pmult_table,
dst,
src,
len,
scale
);
dst = dst.offset((len - left) as isize);
src = src.offset((len - left) as isize);
len = left;
}
while len > 0 {
*dst ^= self.mult(scale, *src);
dst = dst.offset(1);
src = src.offset(1);
len -= 1;
}
}
unsafe fn mult_ptr_len(
&self,
mut dst: *mut u8,
scale: u8,
mut len: usize
) {
if scale == 0 {
for i in 0..len {
*dst.offset(i as isize) = 0;
}
} else if scale != 1 {
#[cfg(feature="simd")]
{
let left = simd_scale_vec(
self.pmult_table,
dst,
dst,
len,
scale
);
dst = dst.offset((len - left) as isize);
len = left;
}
while len > 0 {
*dst = self.mult(*dst, scale);
dst = dst.offset(1);
len -= 1;
}
}
}
unsafe fn div_ptr_len(
&self,
mut dst: *mut u8,
scale: u8,
mut len: usize
) {
if scale == 0 {
panic!("Can't divide vector by 0");
} else if scale != 1 {
#[cfg(feature="simd")]
{
let left = simd_scale_vec(
self.pdiv_table,
dst,
dst,
len,
scale
);
dst = dst.offset((len - left) as isize);
len = left;
}
while len > 0 {
*dst = self.div(*dst, scale);
dst = dst.offset(1);
len -= 1;
}
}
}
}
/// Implements field arithmetic compatible with primitive [`IrreducablePolynomial`]s.
///
/// Recall that there are two strategies for optimizing field arithmetic in
/// `GF(2^8)`: accessing direct multiplication and division tables, and
/// manipulating logarithms and exponentials. The latter method requires fewer
/// operations, but is only possible if the given [`IrreducablePolynomial`] is
/// primitive.
///
/// This struct uses exponentiation and logarithm tables, and is only
/// compatible with primitive [`IrreducablePolynomial`]s. For an implementation
/// compatible with all [`IrreducablePolynomial`]s, see [`GeneralField`].
///
/// Note that this implementation may also use multiplication and division
/// tables for its vectorized operations. See the [`Field`] documentation
/// for more details.
///
/// [`IrreducablePolynomial`]: enum.IrreducablePolynomial.html
/// [`GeneralField`]: struct.GeneralField.html
/// [`Field`]: trait.Field.html
pub struct PrimitivePolynomialField {
modulo: IrreducablePolynomial,
exp_table: Vec<u8>,
log_table: Vec<u8>,
pexp_table: *const u8,
plog_table: *const u8,
// These fields are only used with SIMD support.
#[allow(dead_code)]
mult_table: Vec<u8>,
#[allow(dead_code)]
div_table: Vec<u8>,
#[allow(dead_code)]
pmult_table: *const u8,
#[allow(dead_code)]
pdiv_table: *const u8
}
// This is ok because instance data is not modified after construction.
unsafe impl Sync for PrimitivePolynomialField {}
impl PrimitivePolynomialField {
/// Constructs a new `PrimitivePolynomialField` with all tables initialized.
///
/// If the given `poly` argument is not primitive, this function returns
/// `None`; otherwise it returns `Some(f: PrimitivePolynomialField)`.
/// In situations where the use of `Option<PrimitivePolynomialField>` is
/// less ideal than incurring a panic, consider [`new_might_panic`].
///
/// [`new_might_panic`]: #method.new_might_panic
pub fn new(poly: IrreducablePolynomial) -> Option<Self> {
if !poly.is_primitive() {
return None;
}
let mut exp_table = Vec::with_capacity(510);
let mut log_table = Vec::with_capacity(255);
unsafe {
exp_table.set_len(510);
log_table.set_len(256);
}
let mut member = 1;
for x in 0..255 {
exp_table[x] = member;
exp_table[x + 255] = member;
log_table[member as usize] = x as u8;
member = gf2_mult_mod(member, 2, poly);
}
// This isn't used
log_table[0] = 0;
// These _have_ to remain mutable in general, because we might be
// altering them if SIMD is enabled.
// Note that Vec::with_capacity(0) does not allocate, so in the
// non-SIMD case we don't even trigger a heap allocation.
#[allow(unused_assignments, unused_mut)]
let mut mult_table = Vec::with_capacity(0);
#[allow(unused_assignments, unused_mut)]
let mut div_table = Vec::with_capacity(0);
#[cfg(feature = "simd")]
{
let tup = construct_mult_div_tables(poly);
mult_table = tup.0;
div_table = tup.1;
}
let mut ret = Self{
modulo: poly,
exp_table: exp_table,
log_table: log_table,
pexp_table: ptr::null(),
plog_table: ptr::null(),
mult_table: mult_table,
div_table: div_table,
pmult_table: ptr::null(),
pdiv_table: ptr::null()
};
ret.pexp_table = ret.exp_table.as_ptr();
ret.plog_table = ret.log_table.as_ptr();
#[cfg(feature = "simd")]
{
ret.pmult_table = ret.mult_table.as_ptr();
ret.pdiv_table = ret.div_table.as_ptr();
}
Some(ret)
}
/// Constructs a new `PrimitivePolynomialField` with all tables initialized.
///
/// If the given `poly` argument is not primitive, this function panics.
/// The contents of the resulting error message are not defined.
/// In situations where incurring a panic is less ideal than the use of
/// `Option<PrimitivePolynomialField>`, consider [`new`].
///
/// [`new`]: #method.new
pub fn new_might_panic(poly: IrreducablePolynomial) -> Self {
match Self::new(poly) {
Some(f) => f,
None => panic!("Polynomial {:?} is not primitive")
}
}
}
impl Field for PrimitivePolynomialField {
fn polynomial(&self) -> IrreducablePolynomial {
self.modulo
}
fn mult(&self, src: u8, scale: u8) -> u8 {
if src == 0 || scale == 0 {
return 0;
}
unsafe {
let log_src = *self.plog_table.offset(src as isize) as isize;
let log_scale = *self.plog_table.offset(scale as isize) as isize;
return *self.pexp_table.offset(log_src + log_scale);
}
}
fn div(&self, src: u8, scale: u8) -> u8 {
if scale == 0 {
panic!("Can't divide {}/0", src);
}
if src == 0 {
return 0;
}
unsafe {
let log_src = *self.plog_table.offset(src as isize) as isize;
let log_scale = *self.plog_table.offset(scale as isize) as isize;
let exp_offset = 255 + log_src - log_scale;
return *self.pexp_table.offset(exp_offset);
}
}
fn two_pow(&self, x: u8) -> u8 {
unsafe {
*self.pexp_table.offset(x as isize)
}
}
fn mult_two_pow(&self, scale: u8, x: u8) -> u8 {
if scale == 0 {
return 0;
}
unsafe {
let scale_log = *self.plog_table.offset(scale as isize) as isize;
return *self.pexp_table.offset(scale_log + x as isize);
}
}
unsafe fn add_ptr_scaled_len(
&self,
mut dst: *mut u8,
mut src: *const u8,
scale: u8,
mut len: usize
) {
if scale == 0 {
return;
}
if scale == 1 {
self.add_ptr_len(dst, src, len);
return;
}
#[cfg(feature = "simd")]
{
let left = simd_scale_vec_into(
self.pmult_table,
dst,
src,
len,
scale
);
dst = dst.offset((len - left) as isize);
src = src.offset((len - left) as isize);
len = left;
}
while len > 0 {
*dst ^= self.mult(scale, *src);
dst = dst.offset(1);
src = src.offset(1);
len -= 1;
}
}
unsafe fn mult_ptr_len(
&self,
mut dst: *mut u8,
scale: u8,
mut len: usize
) {
if scale == 0 {
for i in 0..len {
let dst_ptr = dst.offset(i as isize);
*dst_ptr = 0;
}
} else if scale != 1 {
#[cfg(feature = "simd")]
{
let left = simd_scale_vec(
self.pmult_table,
dst,
dst,
len,
scale
);
dst = dst.offset((len - left) as isize);
len = left;
}
while len > 0 {
*dst = self.mult(*dst, scale);
dst = dst.offset(1);
len -= 1;
}
}
}
unsafe fn div_ptr_len(
&self,
mut dst: *mut u8,
scale: u8,
mut len: usize
) {
if scale == 0 {
panic!("Cannot divide vector by 0");
} else if scale != 1 {
#[cfg(feature = "simd")]
{
let left = simd_scale_vec(
self.pdiv_table,
dst,
dst,
len,
scale
);
dst = dst.offset((len - left) as isize);
len = left;
}
while len > 0 {
*dst = self.div(*dst, scale);
dst = dst.offset(1);
len -= 1;
}
}
}
}
// Slow variants of GF(2^8) arithmetic, used for generating tables used
// by the fast variants.
// Exposed as pub(crate) for testing purposes.
pub(crate) fn gf2_mult_mod(
left: u8,
right: u8,
modulo: IrreducablePolynomial
) -> u8 {
gf2_mod(gf2_mult(left, right), modulo)
}
fn gf2_mult(left: u8, right: u8) -> u16 {
let mut ret: u16 = 0;
let mut left_working = left as u16;
for _ in 0..8 {
ret <<= 1;
if left_working & 0x80 != 0 {
ret ^= right as u16;
}
left_working <<= 1;
}
ret
}
fn gf2_mod(mut quotient: u16, poly: IrreducablePolynomial) -> u8 {
let sub = poly.to_u16();
let mut degree = log2_floor(quotient);
while degree > 7 {
let shift_by = degree - 8;
let sub_shifted = sub << shift_by;
quotient ^= sub_shifted;
degree = log2_floor(quotient);
}
quotient as u8
}
// Note that we treat 0 as 1 for the purposes of this function
fn log2_floor(mut target: u16) -> u8 {
let mut ret = 0;
for _ in 0..16 {
target >>= 1;
if target != 0 {
ret += 1;
} else {
break;
}
}
ret
}