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 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685
#![cfg_attr(feature = "nightly", feature(step_trait))]
#![cfg_attr(feature = "nightly", allow(internal_features))]
#![cfg_attr(feature = "nightly", doc(rust_logo))]
#![cfg_attr(feature = "nightly", feature(rustdoc_internals))]
use std::fmt;
use std::num::{NonZeroUsize, ParseIntError};
use std::ops::{Add, AddAssign, Mul, RangeInclusive, Sub};
use std::str::FromStr;
use bitflags::bitflags;
use rustc_index::{Idx, IndexSlice, IndexVec};
#[cfg(feature = "nightly")]
use rustc_data_structures::stable_hasher::StableOrd;
#[cfg(feature = "nightly")]
use rustc_macros::HashStable_Generic;
#[cfg(feature = "nightly")]
use rustc_macros::{Decodable_Generic, Encodable_Generic};
#[cfg(feature = "nightly")]
use std::iter::Step;
mod layout;
pub use layout::LayoutCalculator;
/// Requirements for a `StableHashingContext` to be used in this crate.
/// This is a hack to allow using the `HashStable_Generic` derive macro
/// instead of implementing everything in `rustc_middle`.
pub trait HashStableContext {}
#[derive(Clone, Copy, PartialEq, Eq, Default)]
#[cfg_attr(feature = "nightly", derive(Encodable_Generic, Decodable_Generic, HashStable_Generic))]
pub struct ReprFlags(u8);
bitflags! {
impl ReprFlags: u8 {
const IS_C = 1 << 0;
const IS_SIMD = 1 << 1;
const IS_TRANSPARENT = 1 << 2;
// Internal only for now. If true, don't reorder fields.
const IS_LINEAR = 1 << 3;
// If true, the type's layout can be randomized using
// the seed stored in `ReprOptions.field_shuffle_seed`
const RANDOMIZE_LAYOUT = 1 << 4;
// Any of these flags being set prevent field reordering optimisation.
const IS_UNOPTIMISABLE = ReprFlags::IS_C.bits()
| ReprFlags::IS_SIMD.bits()
| ReprFlags::IS_LINEAR.bits();
}
}
// This is the same as `rustc_data_structures::external_bitflags_debug` but without the
// `rustc_data_structures` to make it build on stable.
impl std::fmt::Debug for ReprFlags {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
bitflags::parser::to_writer(self, f)
}
}
#[derive(Copy, Clone, Debug, Eq, PartialEq)]
#[cfg_attr(feature = "nightly", derive(Encodable_Generic, Decodable_Generic, HashStable_Generic))]
pub enum IntegerType {
/// Pointer-sized integer type, i.e. `isize` and `usize`. The field shows signedness, e.g.
/// `Pointer(true)` means `isize`.
Pointer(bool),
/// Fixed-sized integer type, e.g. `i8`, `u32`, `i128`. The bool field shows signedness, e.g.
/// `Fixed(I8, false)` means `u8`.
Fixed(Integer, bool),
}
impl IntegerType {
pub fn is_signed(&self) -> bool {
match self {
IntegerType::Pointer(b) => *b,
IntegerType::Fixed(_, b) => *b,
}
}
}
/// Represents the repr options provided by the user.
#[derive(Copy, Clone, Debug, Eq, PartialEq, Default)]
#[cfg_attr(feature = "nightly", derive(Encodable_Generic, Decodable_Generic, HashStable_Generic))]
pub struct ReprOptions {
pub int: Option<IntegerType>,
pub align: Option<Align>,
pub pack: Option<Align>,
pub flags: ReprFlags,
/// The seed to be used for randomizing a type's layout
///
/// Note: This could technically be a `u128` which would
/// be the "most accurate" hash as it'd encompass the item and crate
/// hash without loss, but it does pay the price of being larger.
/// Everything's a tradeoff, a 64-bit seed should be sufficient for our
/// purposes (primarily `-Z randomize-layout`)
pub field_shuffle_seed: u64,
}
impl ReprOptions {
#[inline]
pub fn simd(&self) -> bool {
self.flags.contains(ReprFlags::IS_SIMD)
}
#[inline]
pub fn c(&self) -> bool {
self.flags.contains(ReprFlags::IS_C)
}
#[inline]
pub fn packed(&self) -> bool {
self.pack.is_some()
}
#[inline]
pub fn transparent(&self) -> bool {
self.flags.contains(ReprFlags::IS_TRANSPARENT)
}
#[inline]
pub fn linear(&self) -> bool {
self.flags.contains(ReprFlags::IS_LINEAR)
}
/// Returns the discriminant type, given these `repr` options.
/// This must only be called on enums!
pub fn discr_type(&self) -> IntegerType {
self.int.unwrap_or(IntegerType::Pointer(true))
}
/// Returns `true` if this `#[repr()]` should inhabit "smart enum
/// layout" optimizations, such as representing `Foo<&T>` as a
/// single pointer.
pub fn inhibit_enum_layout_opt(&self) -> bool {
self.c() || self.int.is_some()
}
/// Returns `true` if this `#[repr()]` should inhibit struct field reordering
/// optimizations, such as with `repr(C)`, `repr(packed(1))`, or `repr(<int>)`.
pub fn inhibit_struct_field_reordering_opt(&self) -> bool {
if let Some(pack) = self.pack {
if pack.bytes() == 1 {
return true;
}
}
self.flags.intersects(ReprFlags::IS_UNOPTIMISABLE) || self.int.is_some()
}
/// Returns `true` if this type is valid for reordering and `-Z randomize-layout`
/// was enabled for its declaration crate.
pub fn can_randomize_type_layout(&self) -> bool {
!self.inhibit_struct_field_reordering_opt()
&& self.flags.contains(ReprFlags::RANDOMIZE_LAYOUT)
}
/// Returns `true` if this `#[repr()]` should inhibit union ABI optimisations.
pub fn inhibit_union_abi_opt(&self) -> bool {
self.c()
}
}
/// Parsed [Data layout](https://llvm.org/docs/LangRef.html#data-layout)
/// for a target, which contains everything needed to compute layouts.
#[derive(Debug, PartialEq, Eq)]
pub struct TargetDataLayout {
pub endian: Endian,
pub i1_align: AbiAndPrefAlign,
pub i8_align: AbiAndPrefAlign,
pub i16_align: AbiAndPrefAlign,
pub i32_align: AbiAndPrefAlign,
pub i64_align: AbiAndPrefAlign,
pub i128_align: AbiAndPrefAlign,
pub f16_align: AbiAndPrefAlign,
pub f32_align: AbiAndPrefAlign,
pub f64_align: AbiAndPrefAlign,
pub f128_align: AbiAndPrefAlign,
pub pointer_size: Size,
pub pointer_align: AbiAndPrefAlign,
pub aggregate_align: AbiAndPrefAlign,
/// Alignments for vector types.
pub vector_align: Vec<(Size, AbiAndPrefAlign)>,
pub instruction_address_space: AddressSpace,
/// Minimum size of #[repr(C)] enums (default c_int::BITS, usually 32)
/// Note: This isn't in LLVM's data layout string, it is `short_enum`
/// so the only valid spec for LLVM is c_int::BITS or 8
pub c_enum_min_size: Integer,
}
impl Default for TargetDataLayout {
/// Creates an instance of `TargetDataLayout`.
fn default() -> TargetDataLayout {
let align = |bits| Align::from_bits(bits).unwrap();
TargetDataLayout {
endian: Endian::Big,
i1_align: AbiAndPrefAlign::new(align(8)),
i8_align: AbiAndPrefAlign::new(align(8)),
i16_align: AbiAndPrefAlign::new(align(16)),
i32_align: AbiAndPrefAlign::new(align(32)),
i64_align: AbiAndPrefAlign { abi: align(32), pref: align(64) },
i128_align: AbiAndPrefAlign { abi: align(32), pref: align(64) },
f16_align: AbiAndPrefAlign::new(align(16)),
f32_align: AbiAndPrefAlign::new(align(32)),
f64_align: AbiAndPrefAlign::new(align(64)),
f128_align: AbiAndPrefAlign::new(align(128)),
pointer_size: Size::from_bits(64),
pointer_align: AbiAndPrefAlign::new(align(64)),
aggregate_align: AbiAndPrefAlign { abi: align(0), pref: align(64) },
vector_align: vec![
(Size::from_bits(64), AbiAndPrefAlign::new(align(64))),
(Size::from_bits(128), AbiAndPrefAlign::new(align(128))),
],
instruction_address_space: AddressSpace::DATA,
c_enum_min_size: Integer::I32,
}
}
}
pub enum TargetDataLayoutErrors<'a> {
InvalidAddressSpace { addr_space: &'a str, cause: &'a str, err: ParseIntError },
InvalidBits { kind: &'a str, bit: &'a str, cause: &'a str, err: ParseIntError },
MissingAlignment { cause: &'a str },
InvalidAlignment { cause: &'a str, err: AlignFromBytesError },
InconsistentTargetArchitecture { dl: &'a str, target: &'a str },
InconsistentTargetPointerWidth { pointer_size: u64, target: u32 },
InvalidBitsSize { err: String },
}
impl TargetDataLayout {
/// Parse data layout from an
/// [llvm data layout string](https://llvm.org/docs/LangRef.html#data-layout)
///
/// This function doesn't fill `c_enum_min_size` and it will always be `I32` since it can not be
/// determined from llvm string.
pub fn parse_from_llvm_datalayout_string<'a>(
input: &'a str,
) -> Result<TargetDataLayout, TargetDataLayoutErrors<'a>> {
// Parse an address space index from a string.
let parse_address_space = |s: &'a str, cause: &'a str| {
s.parse::<u32>().map(AddressSpace).map_err(|err| {
TargetDataLayoutErrors::InvalidAddressSpace { addr_space: s, cause, err }
})
};
// Parse a bit count from a string.
let parse_bits = |s: &'a str, kind: &'a str, cause: &'a str| {
s.parse::<u64>().map_err(|err| TargetDataLayoutErrors::InvalidBits {
kind,
bit: s,
cause,
err,
})
};
// Parse a size string.
let parse_size =
|s: &'a str, cause: &'a str| parse_bits(s, "size", cause).map(Size::from_bits);
// Parse an alignment string.
let parse_align = |s: &[&'a str], cause: &'a str| {
if s.is_empty() {
return Err(TargetDataLayoutErrors::MissingAlignment { cause });
}
let align_from_bits = |bits| {
Align::from_bits(bits)
.map_err(|err| TargetDataLayoutErrors::InvalidAlignment { cause, err })
};
let abi = parse_bits(s[0], "alignment", cause)?;
let pref = s.get(1).map_or(Ok(abi), |pref| parse_bits(pref, "alignment", cause))?;
Ok(AbiAndPrefAlign { abi: align_from_bits(abi)?, pref: align_from_bits(pref)? })
};
let mut dl = TargetDataLayout::default();
let mut i128_align_src = 64;
for spec in input.split('-') {
let spec_parts = spec.split(':').collect::<Vec<_>>();
match &*spec_parts {
["e"] => dl.endian = Endian::Little,
["E"] => dl.endian = Endian::Big,
[p] if p.starts_with('P') => {
dl.instruction_address_space = parse_address_space(&p[1..], "P")?
}
["a", ref a @ ..] => dl.aggregate_align = parse_align(a, "a")?,
["f16", ref a @ ..] => dl.f16_align = parse_align(a, "f16")?,
["f32", ref a @ ..] => dl.f32_align = parse_align(a, "f32")?,
["f64", ref a @ ..] => dl.f64_align = parse_align(a, "f64")?,
["f128", ref a @ ..] => dl.f128_align = parse_align(a, "f128")?,
// FIXME(erikdesjardins): we should be parsing nonzero address spaces
// this will require replacing TargetDataLayout::{pointer_size,pointer_align}
// with e.g. `fn pointer_size_in(AddressSpace)`
[p @ "p", s, ref a @ ..] | [p @ "p0", s, ref a @ ..] => {
dl.pointer_size = parse_size(s, p)?;
dl.pointer_align = parse_align(a, p)?;
}
[s, ref a @ ..] if s.starts_with('i') => {
let Ok(bits) = s[1..].parse::<u64>() else {
parse_size(&s[1..], "i")?; // For the user error.
continue;
};
let a = parse_align(a, s)?;
match bits {
1 => dl.i1_align = a,
8 => dl.i8_align = a,
16 => dl.i16_align = a,
32 => dl.i32_align = a,
64 => dl.i64_align = a,
_ => {}
}
if bits >= i128_align_src && bits <= 128 {
// Default alignment for i128 is decided by taking the alignment of
// largest-sized i{64..=128}.
i128_align_src = bits;
dl.i128_align = a;
}
}
[s, ref a @ ..] if s.starts_with('v') => {
let v_size = parse_size(&s[1..], "v")?;
let a = parse_align(a, s)?;
if let Some(v) = dl.vector_align.iter_mut().find(|v| v.0 == v_size) {
v.1 = a;
continue;
}
// No existing entry, add a new one.
dl.vector_align.push((v_size, a));
}
_ => {} // Ignore everything else.
}
}
Ok(dl)
}
/// Returns exclusive upper bound on object size.
///
/// The theoretical maximum object size is defined as the maximum positive `isize` value.
/// This ensures that the `offset` semantics remain well-defined by allowing it to correctly
/// index every address within an object along with one byte past the end, along with allowing
/// `isize` to store the difference between any two pointers into an object.
///
/// The upper bound on 64-bit currently needs to be lower because LLVM uses a 64-bit integer
/// to represent object size in bits. It would need to be 1 << 61 to account for this, but is
/// currently conservatively bounded to 1 << 47 as that is enough to cover the current usable
/// address space on 64-bit ARMv8 and x86_64.
#[inline]
pub fn obj_size_bound(&self) -> u64 {
match self.pointer_size.bits() {
16 => 1 << 15,
32 => 1 << 31,
64 => 1 << 47,
bits => panic!("obj_size_bound: unknown pointer bit size {bits}"),
}
}
#[inline]
pub fn ptr_sized_integer(&self) -> Integer {
use Integer::*;
match self.pointer_size.bits() {
16 => I16,
32 => I32,
64 => I64,
bits => panic!("ptr_sized_integer: unknown pointer bit size {bits}"),
}
}
#[inline]
pub fn vector_align(&self, vec_size: Size) -> AbiAndPrefAlign {
for &(size, align) in &self.vector_align {
if size == vec_size {
return align;
}
}
// Default to natural alignment, which is what LLVM does.
// That is, use the size, rounded up to a power of 2.
AbiAndPrefAlign::new(Align::from_bytes(vec_size.bytes().next_power_of_two()).unwrap())
}
}
pub trait HasDataLayout {
fn data_layout(&self) -> &TargetDataLayout;
}
impl HasDataLayout for TargetDataLayout {
#[inline]
fn data_layout(&self) -> &TargetDataLayout {
self
}
}
/// Endianness of the target, which must match cfg(target-endian).
#[derive(Copy, Clone, PartialEq, Eq)]
pub enum Endian {
Little,
Big,
}
impl Endian {
pub fn as_str(&self) -> &'static str {
match self {
Self::Little => "little",
Self::Big => "big",
}
}
}
impl fmt::Debug for Endian {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
f.write_str(self.as_str())
}
}
impl FromStr for Endian {
type Err = String;
fn from_str(s: &str) -> Result<Self, Self::Err> {
match s {
"little" => Ok(Self::Little),
"big" => Ok(Self::Big),
_ => Err(format!(r#"unknown endian: "{s}""#)),
}
}
}
/// Size of a type in bytes.
#[derive(Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
#[cfg_attr(feature = "nightly", derive(Encodable_Generic, Decodable_Generic, HashStable_Generic))]
pub struct Size {
raw: u64,
}
// Safety: Ord is implement as just comparing numerical values and numerical values
// are not changed by (de-)serialization.
#[cfg(feature = "nightly")]
unsafe impl StableOrd for Size {
const CAN_USE_UNSTABLE_SORT: bool = true;
}
// This is debug-printed a lot in larger structs, don't waste too much space there
impl fmt::Debug for Size {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "Size({} bytes)", self.bytes())
}
}
impl Size {
pub const ZERO: Size = Size { raw: 0 };
/// Rounds `bits` up to the next-higher byte boundary, if `bits` is
/// not a multiple of 8.
pub fn from_bits(bits: impl TryInto<u64>) -> Size {
let bits = bits.try_into().ok().unwrap();
// Avoid potential overflow from `bits + 7`.
Size { raw: bits / 8 + ((bits % 8) + 7) / 8 }
}
#[inline]
pub fn from_bytes(bytes: impl TryInto<u64>) -> Size {
let bytes: u64 = bytes.try_into().ok().unwrap();
Size { raw: bytes }
}
#[inline]
pub fn bytes(self) -> u64 {
self.raw
}
#[inline]
pub fn bytes_usize(self) -> usize {
self.bytes().try_into().unwrap()
}
#[inline]
pub fn bits(self) -> u64 {
#[cold]
fn overflow(bytes: u64) -> ! {
panic!("Size::bits: {bytes} bytes in bits doesn't fit in u64")
}
self.bytes().checked_mul(8).unwrap_or_else(|| overflow(self.bytes()))
}
#[inline]
pub fn bits_usize(self) -> usize {
self.bits().try_into().unwrap()
}
#[inline]
pub fn align_to(self, align: Align) -> Size {
let mask = align.bytes() - 1;
Size::from_bytes((self.bytes() + mask) & !mask)
}
#[inline]
pub fn is_aligned(self, align: Align) -> bool {
let mask = align.bytes() - 1;
self.bytes() & mask == 0
}
#[inline]
pub fn checked_add<C: HasDataLayout>(self, offset: Size, cx: &C) -> Option<Size> {
let dl = cx.data_layout();
let bytes = self.bytes().checked_add(offset.bytes())?;
if bytes < dl.obj_size_bound() { Some(Size::from_bytes(bytes)) } else { None }
}
#[inline]
pub fn checked_mul<C: HasDataLayout>(self, count: u64, cx: &C) -> Option<Size> {
let dl = cx.data_layout();
let bytes = self.bytes().checked_mul(count)?;
if bytes < dl.obj_size_bound() { Some(Size::from_bytes(bytes)) } else { None }
}
/// Truncates `value` to `self` bits and then sign-extends it to 128 bits
/// (i.e., if it is negative, fill with 1's on the left).
#[inline]
pub fn sign_extend(self, value: u128) -> u128 {
let size = self.bits();
if size == 0 {
// Truncated until nothing is left.
return 0;
}
// Sign-extend it.
let shift = 128 - size;
// Shift the unsigned value to the left, then shift back to the right as signed
// (essentially fills with sign bit on the left).
(((value << shift) as i128) >> shift) as u128
}
/// Truncates `value` to `self` bits.
#[inline]
pub fn truncate(self, value: u128) -> u128 {
let size = self.bits();
if size == 0 {
// Truncated until nothing is left.
return 0;
}
let shift = 128 - size;
// Truncate (shift left to drop out leftover values, shift right to fill with zeroes).
(value << shift) >> shift
}
#[inline]
pub fn signed_int_min(&self) -> i128 {
self.sign_extend(1_u128 << (self.bits() - 1)) as i128
}
#[inline]
pub fn signed_int_max(&self) -> i128 {
i128::MAX >> (128 - self.bits())
}
#[inline]
pub fn unsigned_int_max(&self) -> u128 {
u128::MAX >> (128 - self.bits())
}
}
// Panicking addition, subtraction and multiplication for convenience.
// Avoid during layout computation, return `LayoutError` instead.
impl Add for Size {
type Output = Size;
#[inline]
fn add(self, other: Size) -> Size {
Size::from_bytes(self.bytes().checked_add(other.bytes()).unwrap_or_else(|| {
panic!("Size::add: {} + {} doesn't fit in u64", self.bytes(), other.bytes())
}))
}
}
impl Sub for Size {
type Output = Size;
#[inline]
fn sub(self, other: Size) -> Size {
Size::from_bytes(self.bytes().checked_sub(other.bytes()).unwrap_or_else(|| {
panic!("Size::sub: {} - {} would result in negative size", self.bytes(), other.bytes())
}))
}
}
impl Mul<Size> for u64 {
type Output = Size;
#[inline]
fn mul(self, size: Size) -> Size {
size * self
}
}
impl Mul<u64> for Size {
type Output = Size;
#[inline]
fn mul(self, count: u64) -> Size {
match self.bytes().checked_mul(count) {
Some(bytes) => Size::from_bytes(bytes),
None => panic!("Size::mul: {} * {} doesn't fit in u64", self.bytes(), count),
}
}
}
impl AddAssign for Size {
#[inline]
fn add_assign(&mut self, other: Size) {
*self = *self + other;
}
}
#[cfg(feature = "nightly")]
impl Step for Size {
#[inline]
fn steps_between(start: &Self, end: &Self) -> Option<usize> {
u64::steps_between(&start.bytes(), &end.bytes())
}
#[inline]
fn forward_checked(start: Self, count: usize) -> Option<Self> {
u64::forward_checked(start.bytes(), count).map(Self::from_bytes)
}
#[inline]
fn forward(start: Self, count: usize) -> Self {
Self::from_bytes(u64::forward(start.bytes(), count))
}
#[inline]
unsafe fn forward_unchecked(start: Self, count: usize) -> Self {
Self::from_bytes(u64::forward_unchecked(start.bytes(), count))
}
#[inline]
fn backward_checked(start: Self, count: usize) -> Option<Self> {
u64::backward_checked(start.bytes(), count).map(Self::from_bytes)
}
#[inline]
fn backward(start: Self, count: usize) -> Self {
Self::from_bytes(u64::backward(start.bytes(), count))
}
#[inline]
unsafe fn backward_unchecked(start: Self, count: usize) -> Self {
Self::from_bytes(u64::backward_unchecked(start.bytes(), count))
}
}
/// Alignment of a type in bytes (always a power of two).
#[derive(Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
#[cfg_attr(feature = "nightly", derive(Encodable_Generic, Decodable_Generic, HashStable_Generic))]
pub struct Align {
pow2: u8,
}
// This is debug-printed a lot in larger structs, don't waste too much space there
impl fmt::Debug for Align {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(f, "Align({} bytes)", self.bytes())
}
}
#[derive(Clone, Copy)]
pub enum AlignFromBytesError {
NotPowerOfTwo(u64),
TooLarge(u64),
}
impl AlignFromBytesError {
pub fn diag_ident(self) -> &'static str {
match self {
Self::NotPowerOfTwo(_) => "not_power_of_two",
Self::TooLarge(_) => "too_large",
}
}
pub fn align(self) -> u64 {
let (Self::NotPowerOfTwo(align) | Self::TooLarge(align)) = self;
align
}
}
impl fmt::Debug for AlignFromBytesError {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
fmt::Display::fmt(self, f)
}
}
impl fmt::Display for AlignFromBytesError {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
match self {
AlignFromBytesError::NotPowerOfTwo(align) => write!(f, "`{align}` is not a power of 2"),
AlignFromBytesError::TooLarge(align) => write!(f, "`{align}` is too large"),
}
}
}
impl Align {
pub const ONE: Align = Align { pow2: 0 };
// LLVM has a maximal supported alignment of 2^29, we inherit that.
pub const MAX: Align = Align { pow2: 29 };
#[inline]
pub fn from_bits(bits: u64) -> Result<Align, AlignFromBytesError> {
Align::from_bytes(Size::from_bits(bits).bytes())
}
#[inline]
pub fn from_bytes(align: u64) -> Result<Align, AlignFromBytesError> {
// Treat an alignment of 0 bytes like 1-byte alignment.
if align == 0 {
return Ok(Align::ONE);
}
#[cold]
fn not_power_of_2(align: u64) -> AlignFromBytesError {
AlignFromBytesError::NotPowerOfTwo(align)
}
#[cold]
fn too_large(align: u64) -> AlignFromBytesError {
AlignFromBytesError::TooLarge(align)
}
let tz = align.trailing_zeros();
if align != (1 << tz) {
return Err(not_power_of_2(align));
}
let pow2 = tz as u8;
if pow2 > Self::MAX.pow2 {
return Err(too_large(align));
}
Ok(Align { pow2 })
}
#[inline]
pub fn bytes(self) -> u64 {
1 << self.pow2
}
#[inline]
pub fn bits(self) -> u64 {
self.bytes() * 8
}
/// Computes the best alignment possible for the given offset
/// (the largest power of two that the offset is a multiple of).
///
/// N.B., for an offset of `0`, this happens to return `2^64`.
#[inline]
pub fn max_for_offset(offset: Size) -> Align {
Align { pow2: offset.bytes().trailing_zeros() as u8 }
}
/// Lower the alignment, if necessary, such that the given offset
/// is aligned to it (the offset is a multiple of the alignment).
#[inline]
pub fn restrict_for_offset(self, offset: Size) -> Align {
self.min(Align::max_for_offset(offset))
}
}
/// A pair of alignments, ABI-mandated and preferred.
#[derive(Copy, Clone, PartialEq, Eq, Hash, Debug)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub struct AbiAndPrefAlign {
pub abi: Align,
pub pref: Align,
}
impl AbiAndPrefAlign {
#[inline]
pub fn new(align: Align) -> AbiAndPrefAlign {
AbiAndPrefAlign { abi: align, pref: align }
}
#[inline]
pub fn min(self, other: AbiAndPrefAlign) -> AbiAndPrefAlign {
AbiAndPrefAlign { abi: self.abi.min(other.abi), pref: self.pref.min(other.pref) }
}
#[inline]
pub fn max(self, other: AbiAndPrefAlign) -> AbiAndPrefAlign {
AbiAndPrefAlign { abi: self.abi.max(other.abi), pref: self.pref.max(other.pref) }
}
}
/// Integers, also used for enum discriminants.
#[derive(Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash, Debug)]
#[cfg_attr(feature = "nightly", derive(Encodable_Generic, Decodable_Generic, HashStable_Generic))]
pub enum Integer {
I8,
I16,
I32,
I64,
I128,
}
impl Integer {
#[inline]
pub fn size(self) -> Size {
use Integer::*;
match self {
I8 => Size::from_bytes(1),
I16 => Size::from_bytes(2),
I32 => Size::from_bytes(4),
I64 => Size::from_bytes(8),
I128 => Size::from_bytes(16),
}
}
/// Gets the Integer type from an IntegerType.
pub fn from_attr<C: HasDataLayout>(cx: &C, ity: IntegerType) -> Integer {
let dl = cx.data_layout();
match ity {
IntegerType::Pointer(_) => dl.ptr_sized_integer(),
IntegerType::Fixed(x, _) => x,
}
}
pub fn align<C: HasDataLayout>(self, cx: &C) -> AbiAndPrefAlign {
use Integer::*;
let dl = cx.data_layout();
match self {
I8 => dl.i8_align,
I16 => dl.i16_align,
I32 => dl.i32_align,
I64 => dl.i64_align,
I128 => dl.i128_align,
}
}
/// Returns the largest signed value that can be represented by this Integer.
#[inline]
pub fn signed_max(self) -> i128 {
use Integer::*;
match self {
I8 => i8::MAX as i128,
I16 => i16::MAX as i128,
I32 => i32::MAX as i128,
I64 => i64::MAX as i128,
I128 => i128::MAX,
}
}
/// Finds the smallest Integer type which can represent the signed value.
#[inline]
pub fn fit_signed(x: i128) -> Integer {
use Integer::*;
match x {
-0x0000_0000_0000_0080..=0x0000_0000_0000_007f => I8,
-0x0000_0000_0000_8000..=0x0000_0000_0000_7fff => I16,
-0x0000_0000_8000_0000..=0x0000_0000_7fff_ffff => I32,
-0x8000_0000_0000_0000..=0x7fff_ffff_ffff_ffff => I64,
_ => I128,
}
}
/// Finds the smallest Integer type which can represent the unsigned value.
#[inline]
pub fn fit_unsigned(x: u128) -> Integer {
use Integer::*;
match x {
0..=0x0000_0000_0000_00ff => I8,
0..=0x0000_0000_0000_ffff => I16,
0..=0x0000_0000_ffff_ffff => I32,
0..=0xffff_ffff_ffff_ffff => I64,
_ => I128,
}
}
/// Finds the smallest integer with the given alignment.
pub fn for_align<C: HasDataLayout>(cx: &C, wanted: Align) -> Option<Integer> {
use Integer::*;
let dl = cx.data_layout();
[I8, I16, I32, I64, I128].into_iter().find(|&candidate| {
wanted == candidate.align(dl).abi && wanted.bytes() == candidate.size().bytes()
})
}
/// Find the largest integer with the given alignment or less.
pub fn approximate_align<C: HasDataLayout>(cx: &C, wanted: Align) -> Integer {
use Integer::*;
let dl = cx.data_layout();
// FIXME(eddyb) maybe include I128 in the future, when it works everywhere.
for candidate in [I64, I32, I16] {
if wanted >= candidate.align(dl).abi && wanted.bytes() >= candidate.size().bytes() {
return candidate;
}
}
I8
}
// FIXME(eddyb) consolidate this and other methods that find the appropriate
// `Integer` given some requirements.
#[inline]
pub fn from_size(size: Size) -> Result<Self, String> {
match size.bits() {
8 => Ok(Integer::I8),
16 => Ok(Integer::I16),
32 => Ok(Integer::I32),
64 => Ok(Integer::I64),
128 => Ok(Integer::I128),
_ => Err(format!("rust does not support integers with {} bits", size.bits())),
}
}
}
/// Fundamental unit of memory access and layout.
#[derive(Copy, Clone, PartialEq, Eq, Hash, Debug)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub enum Primitive {
/// The `bool` is the signedness of the `Integer` type.
///
/// One would think we would not care about such details this low down,
/// but some ABIs are described in terms of C types and ISAs where the
/// integer arithmetic is done on {sign,zero}-extended registers, e.g.
/// a negative integer passed by zero-extension will appear positive in
/// the callee, and most operations on it will produce the wrong values.
Int(Integer, bool),
F16,
F32,
F64,
F128,
Pointer(AddressSpace),
}
impl Primitive {
pub fn size<C: HasDataLayout>(self, cx: &C) -> Size {
use Primitive::*;
let dl = cx.data_layout();
match self {
Int(i, _) => i.size(),
F16 => Size::from_bits(16),
F32 => Size::from_bits(32),
F64 => Size::from_bits(64),
F128 => Size::from_bits(128),
// FIXME(erikdesjardins): ignoring address space is technically wrong, pointers in
// different address spaces can have different sizes
// (but TargetDataLayout doesn't currently parse that part of the DL string)
Pointer(_) => dl.pointer_size,
}
}
pub fn align<C: HasDataLayout>(self, cx: &C) -> AbiAndPrefAlign {
use Primitive::*;
let dl = cx.data_layout();
match self {
Int(i, _) => i.align(dl),
F16 => dl.f16_align,
F32 => dl.f32_align,
F64 => dl.f64_align,
F128 => dl.f128_align,
// FIXME(erikdesjardins): ignoring address space is technically wrong, pointers in
// different address spaces can have different alignments
// (but TargetDataLayout doesn't currently parse that part of the DL string)
Pointer(_) => dl.pointer_align,
}
}
}
/// Inclusive wrap-around range of valid values, that is, if
/// start > end, it represents `start..=MAX`, followed by `0..=end`.
///
/// That is, for an i8 primitive, a range of `254..=2` means following
/// sequence:
///
/// 254 (-2), 255 (-1), 0, 1, 2
///
/// This is intended specifically to mirror LLVM’s `!range` metadata semantics.
#[derive(Clone, Copy, PartialEq, Eq, Hash)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub struct WrappingRange {
pub start: u128,
pub end: u128,
}
impl WrappingRange {
pub fn full(size: Size) -> Self {
Self { start: 0, end: size.unsigned_int_max() }
}
/// Returns `true` if `v` is contained in the range.
#[inline(always)]
pub fn contains(&self, v: u128) -> bool {
if self.start <= self.end {
self.start <= v && v <= self.end
} else {
self.start <= v || v <= self.end
}
}
/// Returns `self` with replaced `start`
#[inline(always)]
fn with_start(mut self, start: u128) -> Self {
self.start = start;
self
}
/// Returns `self` with replaced `end`
#[inline(always)]
fn with_end(mut self, end: u128) -> Self {
self.end = end;
self
}
/// Returns `true` if `size` completely fills the range.
#[inline]
fn is_full_for(&self, size: Size) -> bool {
let max_value = size.unsigned_int_max();
debug_assert!(self.start <= max_value && self.end <= max_value);
self.start == (self.end.wrapping_add(1) & max_value)
}
}
impl fmt::Debug for WrappingRange {
fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
if self.start > self.end {
write!(fmt, "(..={}) | ({}..)", self.end, self.start)?;
} else {
write!(fmt, "{}..={}", self.start, self.end)?;
}
Ok(())
}
}
/// Information about one scalar component of a Rust type.
#[derive(Clone, Copy, PartialEq, Eq, Hash, Debug)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub enum Scalar {
Initialized {
value: Primitive,
// FIXME(eddyb) always use the shortest range, e.g., by finding
// the largest space between two consecutive valid values and
// taking everything else as the (shortest) valid range.
valid_range: WrappingRange,
},
Union {
/// Even for unions, we need to use the correct registers for the kind of
/// values inside the union, so we keep the `Primitive` type around. We
/// also use it to compute the size of the scalar.
/// However, unions never have niches and even allow undef,
/// so there is no `valid_range`.
value: Primitive,
},
}
impl Scalar {
#[inline]
pub fn is_bool(&self) -> bool {
use Integer::*;
matches!(
self,
Scalar::Initialized {
value: Primitive::Int(I8, false),
valid_range: WrappingRange { start: 0, end: 1 }
}
)
}
/// Get the primitive representation of this type, ignoring the valid range and whether the
/// value is allowed to be undefined (due to being a union).
pub fn primitive(&self) -> Primitive {
match *self {
Scalar::Initialized { value, .. } | Scalar::Union { value } => value,
}
}
pub fn align(self, cx: &impl HasDataLayout) -> AbiAndPrefAlign {
self.primitive().align(cx)
}
pub fn size(self, cx: &impl HasDataLayout) -> Size {
self.primitive().size(cx)
}
#[inline]
pub fn to_union(&self) -> Self {
Self::Union { value: self.primitive() }
}
#[inline]
pub fn valid_range(&self, cx: &impl HasDataLayout) -> WrappingRange {
match *self {
Scalar::Initialized { valid_range, .. } => valid_range,
Scalar::Union { value } => WrappingRange::full(value.size(cx)),
}
}
#[inline]
/// Allows the caller to mutate the valid range. This operation will panic if attempted on a
/// union.
pub fn valid_range_mut(&mut self) -> &mut WrappingRange {
match self {
Scalar::Initialized { valid_range, .. } => valid_range,
Scalar::Union { .. } => panic!("cannot change the valid range of a union"),
}
}
/// Returns `true` if all possible numbers are valid, i.e `valid_range` covers the whole
/// layout.
#[inline]
pub fn is_always_valid<C: HasDataLayout>(&self, cx: &C) -> bool {
match *self {
Scalar::Initialized { valid_range, .. } => valid_range.is_full_for(self.size(cx)),
Scalar::Union { .. } => true,
}
}
/// Returns `true` if this type can be left uninit.
#[inline]
pub fn is_uninit_valid(&self) -> bool {
match *self {
Scalar::Initialized { .. } => false,
Scalar::Union { .. } => true,
}
}
}
// NOTE: This struct is generic over the FieldIdx for rust-analyzer usage.
/// Describes how the fields of a type are located in memory.
#[derive(PartialEq, Eq, Hash, Clone, Debug)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub enum FieldsShape<FieldIdx: Idx> {
/// Scalar primitives and `!`, which never have fields.
Primitive,
/// All fields start at no offset. The `usize` is the field count.
Union(NonZeroUsize),
/// Array/vector-like placement, with all fields of identical types.
Array { stride: Size, count: u64 },
/// Struct-like placement, with precomputed offsets.
///
/// Fields are guaranteed to not overlap, but note that gaps
/// before, between and after all the fields are NOT always
/// padding, and as such their contents may not be discarded.
/// For example, enum variants leave a gap at the start,
/// where the discriminant field in the enum layout goes.
Arbitrary {
/// Offsets for the first byte of each field,
/// ordered to match the source definition order.
/// This vector does not go in increasing order.
// FIXME(eddyb) use small vector optimization for the common case.
offsets: IndexVec<FieldIdx, Size>,
/// Maps source order field indices to memory order indices,
/// depending on how the fields were reordered (if at all).
/// This is a permutation, with both the source order and the
/// memory order using the same (0..n) index ranges.
///
/// Note that during computation of `memory_index`, sometimes
/// it is easier to operate on the inverse mapping (that is,
/// from memory order to source order), and that is usually
/// named `inverse_memory_index`.
///
// FIXME(eddyb) build a better abstraction for permutations, if possible.
// FIXME(camlorn) also consider small vector optimization here.
memory_index: IndexVec<FieldIdx, u32>,
},
}
impl<FieldIdx: Idx> FieldsShape<FieldIdx> {
#[inline]
pub fn count(&self) -> usize {
match *self {
FieldsShape::Primitive => 0,
FieldsShape::Union(count) => count.get(),
FieldsShape::Array { count, .. } => count.try_into().unwrap(),
FieldsShape::Arbitrary { ref offsets, .. } => offsets.len(),
}
}
#[inline]
pub fn offset(&self, i: usize) -> Size {
match *self {
FieldsShape::Primitive => {
unreachable!("FieldsShape::offset: `Primitive`s have no fields")
}
FieldsShape::Union(count) => {
assert!(i < count.get(), "tried to access field {i} of union with {count} fields");
Size::ZERO
}
FieldsShape::Array { stride, count } => {
let i = u64::try_from(i).unwrap();
assert!(i < count, "tried to access field {i} of array with {count} fields");
stride * i
}
FieldsShape::Arbitrary { ref offsets, .. } => offsets[FieldIdx::new(i)],
}
}
#[inline]
pub fn memory_index(&self, i: usize) -> usize {
match *self {
FieldsShape::Primitive => {
unreachable!("FieldsShape::memory_index: `Primitive`s have no fields")
}
FieldsShape::Union(_) | FieldsShape::Array { .. } => i,
FieldsShape::Arbitrary { ref memory_index, .. } => {
memory_index[FieldIdx::new(i)].try_into().unwrap()
}
}
}
/// Gets source indices of the fields by increasing offsets.
#[inline]
pub fn index_by_increasing_offset(&self) -> impl Iterator<Item = usize> + '_ {
let mut inverse_small = [0u8; 64];
let mut inverse_big = IndexVec::new();
let use_small = self.count() <= inverse_small.len();
// We have to write this logic twice in order to keep the array small.
if let FieldsShape::Arbitrary { ref memory_index, .. } = *self {
if use_small {
for (field_idx, &mem_idx) in memory_index.iter_enumerated() {
inverse_small[mem_idx as usize] = field_idx.index() as u8;
}
} else {
inverse_big = memory_index.invert_bijective_mapping();
}
}
(0..self.count()).map(move |i| match *self {
FieldsShape::Primitive | FieldsShape::Union(_) | FieldsShape::Array { .. } => i,
FieldsShape::Arbitrary { .. } => {
if use_small {
inverse_small[i] as usize
} else {
inverse_big[i as u32].index()
}
}
})
}
}
/// An identifier that specifies the address space that some operation
/// should operate on. Special address spaces have an effect on code generation,
/// depending on the target and the address spaces it implements.
#[derive(Copy, Clone, Debug, PartialEq, Eq, PartialOrd, Ord, Hash)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub struct AddressSpace(pub u32);
impl AddressSpace {
/// The default address space, corresponding to data space.
pub const DATA: Self = AddressSpace(0);
}
/// Describes how values of the type are passed by target ABIs,
/// in terms of categories of C types there are ABI rules for.
#[derive(Clone, Copy, PartialEq, Eq, Hash, Debug)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub enum Abi {
Uninhabited,
Scalar(Scalar),
ScalarPair(Scalar, Scalar),
Vector {
element: Scalar,
count: u64,
},
Aggregate {
/// If true, the size is exact, otherwise it's only a lower bound.
sized: bool,
},
}
impl Abi {
/// Returns `true` if the layout corresponds to an unsized type.
#[inline]
pub fn is_unsized(&self) -> bool {
match *self {
Abi::Uninhabited | Abi::Scalar(_) | Abi::ScalarPair(..) | Abi::Vector { .. } => false,
Abi::Aggregate { sized } => !sized,
}
}
#[inline]
pub fn is_sized(&self) -> bool {
!self.is_unsized()
}
/// Returns `true` if this is a single signed integer scalar
#[inline]
pub fn is_signed(&self) -> bool {
match self {
Abi::Scalar(scal) => match scal.primitive() {
Primitive::Int(_, signed) => signed,
_ => false,
},
_ => panic!("`is_signed` on non-scalar ABI {self:?}"),
}
}
/// Returns `true` if this is an uninhabited type
#[inline]
pub fn is_uninhabited(&self) -> bool {
matches!(*self, Abi::Uninhabited)
}
/// Returns `true` if this is a scalar type
#[inline]
pub fn is_scalar(&self) -> bool {
matches!(*self, Abi::Scalar(_))
}
/// Returns `true` if this is a bool
#[inline]
pub fn is_bool(&self) -> bool {
matches!(*self, Abi::Scalar(s) if s.is_bool())
}
/// Returns the fixed alignment of this ABI, if any is mandated.
pub fn inherent_align<C: HasDataLayout>(&self, cx: &C) -> Option<AbiAndPrefAlign> {
Some(match *self {
Abi::Scalar(s) => s.align(cx),
Abi::ScalarPair(s1, s2) => s1.align(cx).max(s2.align(cx)),
Abi::Vector { element, count } => {
cx.data_layout().vector_align(element.size(cx) * count)
}
Abi::Uninhabited | Abi::Aggregate { .. } => return None,
})
}
/// Returns the fixed size of this ABI, if any is mandated.
pub fn inherent_size<C: HasDataLayout>(&self, cx: &C) -> Option<Size> {
Some(match *self {
Abi::Scalar(s) => {
// No padding in scalars.
s.size(cx)
}
Abi::ScalarPair(s1, s2) => {
// May have some padding between the pair.
let field2_offset = s1.size(cx).align_to(s2.align(cx).abi);
(field2_offset + s2.size(cx)).align_to(self.inherent_align(cx)?.abi)
}
Abi::Vector { element, count } => {
// No padding in vectors, except possibly for trailing padding
// to make the size a multiple of align (e.g. for vectors of size 3).
(element.size(cx) * count).align_to(self.inherent_align(cx)?.abi)
}
Abi::Uninhabited | Abi::Aggregate { .. } => return None,
})
}
/// Discard validity range information and allow undef.
pub fn to_union(&self) -> Self {
match *self {
Abi::Scalar(s) => Abi::Scalar(s.to_union()),
Abi::ScalarPair(s1, s2) => Abi::ScalarPair(s1.to_union(), s2.to_union()),
Abi::Vector { element, count } => Abi::Vector { element: element.to_union(), count },
Abi::Uninhabited | Abi::Aggregate { .. } => Abi::Aggregate { sized: true },
}
}
pub fn eq_up_to_validity(&self, other: &Self) -> bool {
match (self, other) {
// Scalar, Vector, ScalarPair have `Scalar` in them where we ignore validity ranges.
// We do *not* ignore the sign since it matters for some ABIs (e.g. s390x).
(Abi::Scalar(l), Abi::Scalar(r)) => l.primitive() == r.primitive(),
(
Abi::Vector { element: element_l, count: count_l },
Abi::Vector { element: element_r, count: count_r },
) => element_l.primitive() == element_r.primitive() && count_l == count_r,
(Abi::ScalarPair(l1, l2), Abi::ScalarPair(r1, r2)) => {
l1.primitive() == r1.primitive() && l2.primitive() == r2.primitive()
}
// Everything else must be strictly identical.
_ => self == other,
}
}
}
// NOTE: This struct is generic over the FieldIdx and VariantIdx for rust-analyzer usage.
#[derive(PartialEq, Eq, Hash, Clone, Debug)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub enum Variants<FieldIdx: Idx, VariantIdx: Idx> {
/// Single enum variants, structs/tuples, unions, and all non-ADTs.
Single { index: VariantIdx },
/// Enum-likes with more than one inhabited variant: each variant comes with
/// a *discriminant* (usually the same as the variant index but the user can
/// assign explicit discriminant values). That discriminant is encoded
/// as a *tag* on the machine. The layout of each variant is
/// a struct, and they all have space reserved for the tag.
/// For enums, the tag is the sole field of the layout.
Multiple {
tag: Scalar,
tag_encoding: TagEncoding<VariantIdx>,
tag_field: usize,
variants: IndexVec<VariantIdx, LayoutS<FieldIdx, VariantIdx>>,
},
}
// NOTE: This struct is generic over the VariantIdx for rust-analyzer usage.
#[derive(PartialEq, Eq, Hash, Clone, Debug)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub enum TagEncoding<VariantIdx: Idx> {
/// The tag directly stores the discriminant, but possibly with a smaller layout
/// (so converting the tag to the discriminant can require sign extension).
Direct,
/// Niche (values invalid for a type) encoding the discriminant:
/// Discriminant and variant index coincide.
/// The variant `untagged_variant` contains a niche at an arbitrary
/// offset (field `tag_field` of the enum), which for a variant with
/// discriminant `d` is set to
/// `(d - niche_variants.start).wrapping_add(niche_start)`.
///
/// For example, `Option<(usize, &T)>` is represented such that
/// `None` has a null pointer for the second tuple field, and
/// `Some` is the identity function (with a non-null reference).
Niche {
untagged_variant: VariantIdx,
niche_variants: RangeInclusive<VariantIdx>,
niche_start: u128,
},
}
#[derive(Clone, Copy, PartialEq, Eq, Hash, Debug)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub struct Niche {
pub offset: Size,
pub value: Primitive,
pub valid_range: WrappingRange,
}
impl Niche {
pub fn from_scalar<C: HasDataLayout>(cx: &C, offset: Size, scalar: Scalar) -> Option<Self> {
let Scalar::Initialized { value, valid_range } = scalar else { return None };
let niche = Niche { offset, value, valid_range };
if niche.available(cx) > 0 { Some(niche) } else { None }
}
pub fn available<C: HasDataLayout>(&self, cx: &C) -> u128 {
let Self { value, valid_range: v, .. } = *self;
let size = value.size(cx);
assert!(size.bits() <= 128);
let max_value = size.unsigned_int_max();
// Find out how many values are outside the valid range.
let niche = v.end.wrapping_add(1)..v.start;
niche.end.wrapping_sub(niche.start) & max_value
}
pub fn reserve<C: HasDataLayout>(&self, cx: &C, count: u128) -> Option<(u128, Scalar)> {
assert!(count > 0);
let Self { value, valid_range: v, .. } = *self;
let size = value.size(cx);
assert!(size.bits() <= 128);
let max_value = size.unsigned_int_max();
let niche = v.end.wrapping_add(1)..v.start;
let available = niche.end.wrapping_sub(niche.start) & max_value;
if count > available {
return None;
}
// Extend the range of valid values being reserved by moving either `v.start` or `v.end`
// bound. Given an eventual `Option<T>`, we try to maximize the chance for `None` to occupy
// the niche of zero. This is accomplished by preferring enums with 2 variants(`count==1`)
// and always taking the shortest path to niche zero. Having `None` in niche zero can
// enable some special optimizations.
//
// Bound selection criteria:
// 1. Select closest to zero given wrapping semantics.
// 2. Avoid moving past zero if possible.
//
// In practice this means that enums with `count > 1` are unlikely to claim niche zero,
// since they have to fit perfectly. If niche zero is already reserved, the selection of
// bounds are of little interest.
let move_start = |v: WrappingRange| {
let start = v.start.wrapping_sub(count) & max_value;
Some((start, Scalar::Initialized { value, valid_range: v.with_start(start) }))
};
let move_end = |v: WrappingRange| {
let start = v.end.wrapping_add(1) & max_value;
let end = v.end.wrapping_add(count) & max_value;
Some((start, Scalar::Initialized { value, valid_range: v.with_end(end) }))
};
let distance_end_zero = max_value - v.end;
if v.start > v.end {
// zero is unavailable because wrapping occurs
move_end(v)
} else if v.start <= distance_end_zero {
if count <= v.start {
move_start(v)
} else {
// moved past zero, use other bound
move_end(v)
}
} else {
let end = v.end.wrapping_add(count) & max_value;
let overshot_zero = (1..=v.end).contains(&end);
if overshot_zero {
// moved past zero, use other bound
move_start(v)
} else {
move_end(v)
}
}
}
}
// NOTE: This struct is generic over the FieldIdx and VariantIdx for rust-analyzer usage.
#[derive(PartialEq, Eq, Hash, Clone)]
#[cfg_attr(feature = "nightly", derive(HashStable_Generic))]
pub struct LayoutS<FieldIdx: Idx, VariantIdx: Idx> {
/// Says where the fields are located within the layout.
pub fields: FieldsShape<FieldIdx>,
/// Encodes information about multi-variant layouts.
/// Even with `Multiple` variants, a layout still has its own fields! Those are then
/// shared between all variants. One of them will be the discriminant,
/// but e.g. coroutines can have more.
///
/// To access all fields of this layout, both `fields` and the fields of the active variant
/// must be taken into account.
pub variants: Variants<FieldIdx, VariantIdx>,
/// The `abi` defines how this data is passed between functions, and it defines
/// value restrictions via `valid_range`.
///
/// Note that this is entirely orthogonal to the recursive structure defined by
/// `variants` and `fields`; for example, `ManuallyDrop<Result<isize, isize>>` has
/// `Abi::ScalarPair`! So, even with non-`Aggregate` `abi`, `fields` and `variants`
/// have to be taken into account to find all fields of this layout.
pub abi: Abi,
/// The leaf scalar with the largest number of invalid values
/// (i.e. outside of its `valid_range`), if it exists.
pub largest_niche: Option<Niche>,
pub align: AbiAndPrefAlign,
pub size: Size,
/// The largest alignment explicitly requested with `repr(align)` on this type or any field.
/// Only used on i686-windows, where the argument passing ABI is different when alignment is
/// requested, even if the requested alignment is equal to the natural alignment.
pub max_repr_align: Option<Align>,
/// The alignment the type would have, ignoring any `repr(align)` but including `repr(packed)`.
/// Only used on aarch64-linux, where the argument passing ABI ignores the requested alignment
/// in some cases.
pub unadjusted_abi_align: Align,
}
impl<FieldIdx: Idx, VariantIdx: Idx> LayoutS<FieldIdx, VariantIdx> {
pub fn scalar<C: HasDataLayout>(cx: &C, scalar: Scalar) -> Self {
let largest_niche = Niche::from_scalar(cx, Size::ZERO, scalar);
let size = scalar.size(cx);
let align = scalar.align(cx);
LayoutS {
variants: Variants::Single { index: VariantIdx::new(0) },
fields: FieldsShape::Primitive,
abi: Abi::Scalar(scalar),
largest_niche,
size,
align,
max_repr_align: None,
unadjusted_abi_align: align.abi,
}
}
}
impl<FieldIdx: Idx, VariantIdx: Idx> fmt::Debug for LayoutS<FieldIdx, VariantIdx>
where
FieldsShape<FieldIdx>: fmt::Debug,
Variants<FieldIdx, VariantIdx>: fmt::Debug,
{
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
// This is how `Layout` used to print before it become
// `Interned<LayoutS>`. We print it like this to avoid having to update
// expected output in a lot of tests.
let LayoutS {
size,
align,
abi,
fields,
largest_niche,
variants,
max_repr_align,
unadjusted_abi_align,
} = self;
f.debug_struct("Layout")
.field("size", size)
.field("align", align)
.field("abi", abi)
.field("fields", fields)
.field("largest_niche", largest_niche)
.field("variants", variants)
.field("max_repr_align", max_repr_align)
.field("unadjusted_abi_align", unadjusted_abi_align)
.finish()
}
}
#[derive(Copy, Clone, PartialEq, Eq, Debug)]
pub enum PointerKind {
/// Shared reference. `frozen` indicates the absence of any `UnsafeCell`.
SharedRef { frozen: bool },
/// Mutable reference. `unpin` indicates the absence of any pinned data.
MutableRef { unpin: bool },
/// Box. `unpin` indicates the absence of any pinned data. `global` indicates whether this box
/// uses the global allocator or a custom one.
Box { unpin: bool, global: bool },
}
/// Note that this information is advisory only, and backends are free to ignore it.
/// It can only be used to encode potential optimizations, but no critical information.
#[derive(Copy, Clone, Debug)]
pub struct PointeeInfo {
pub size: Size,
pub align: Align,
/// If this is `None`, then this is a raw pointer, so size and alignment are not guaranteed to
/// be reliable.
pub safe: Option<PointerKind>,
}
impl<FieldIdx: Idx, VariantIdx: Idx> LayoutS<FieldIdx, VariantIdx> {
/// Returns `true` if the layout corresponds to an unsized type.
#[inline]
pub fn is_unsized(&self) -> bool {
self.abi.is_unsized()
}
#[inline]
pub fn is_sized(&self) -> bool {
self.abi.is_sized()
}
/// Returns `true` if the type is sized and a 1-ZST (meaning it has size 0 and alignment 1).
pub fn is_1zst(&self) -> bool {
self.is_sized() && self.size.bytes() == 0 && self.align.abi.bytes() == 1
}
/// Returns `true` if the type is a ZST and not unsized.
///
/// Note that this does *not* imply that the type is irrelevant for layout! It can still have
/// non-trivial alignment constraints. You probably want to use `is_1zst` instead.
pub fn is_zst(&self) -> bool {
match self.abi {
Abi::Scalar(_) | Abi::ScalarPair(..) | Abi::Vector { .. } => false,
Abi::Uninhabited => self.size.bytes() == 0,
Abi::Aggregate { sized } => sized && self.size.bytes() == 0,
}
}
/// Checks if these two `Layout` are equal enough to be considered "the same for all function
/// call ABIs". Note however that real ABIs depend on more details that are not reflected in the
/// `Layout`; the `PassMode` need to be compared as well.
pub fn eq_abi(&self, other: &Self) -> bool {
// The one thing that we are not capturing here is that for unsized types, the metadata must
// also have the same ABI, and moreover that the same metadata leads to the same size. The
// 2nd point is quite hard to check though.
self.size == other.size
&& self.is_sized() == other.is_sized()
&& self.abi.eq_up_to_validity(&other.abi)
&& self.abi.is_bool() == other.abi.is_bool()
&& self.align.abi == other.align.abi
&& self.max_repr_align == other.max_repr_align
&& self.unadjusted_abi_align == other.unadjusted_abi_align
}
}
#[derive(Copy, Clone, Debug)]
pub enum StructKind {
/// A tuple, closure, or univariant which cannot be coerced to unsized.
AlwaysSized,
/// A univariant, the last field of which may be coerced to unsized.
MaybeUnsized,
/// A univariant, but with a prefix of an arbitrary size & alignment (e.g., enum tag).
Prefixed(Size, Align),
}