1 //===- ConstantRange.cpp - ConstantRange implementation -------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // Represent a range of possible values that may occur when the program is run 10 // for an integral value. This keeps track of a lower and upper bound for the 11 // constant, which MAY wrap around the end of the numeric range. To do this, it 12 // keeps track of a [lower, upper) bound, which specifies an interval just like 13 // STL iterators. When used with boolean values, the following are important 14 // ranges (other integral ranges use min/max values for special range values): 15 // 16 // [F, F) = {} = Empty set 17 // [T, F) = {T} 18 // [F, T) = {F} 19 // [T, T) = {F, T} = Full set 20 // 21 //===----------------------------------------------------------------------===// 22 23 #include "llvm/ADT/APInt.h" 24 #include "llvm/Config/llvm-config.h" 25 #include "llvm/IR/ConstantRange.h" 26 #include "llvm/IR/Constants.h" 27 #include "llvm/IR/InstrTypes.h" 28 #include "llvm/IR/Instruction.h" 29 #include "llvm/IR/Metadata.h" 30 #include "llvm/IR/Operator.h" 31 #include "llvm/Support/Compiler.h" 32 #include "llvm/Support/Debug.h" 33 #include "llvm/Support/ErrorHandling.h" 34 #include "llvm/Support/KnownBits.h" 35 #include "llvm/Support/raw_ostream.h" 36 #include <algorithm> 37 #include <cassert> 38 #include <cstdint> 39 40 using namespace llvm; 41 42 ConstantRange::ConstantRange(uint32_t BitWidth, bool Full) 43 : Lower(Full ? APInt::getMaxValue(BitWidth) : APInt::getMinValue(BitWidth)), 44 Upper(Lower) {} 45 46 ConstantRange::ConstantRange(APInt V) 47 : Lower(std::move(V)), Upper(Lower + 1) {} 48 49 ConstantRange::ConstantRange(APInt L, APInt U) 50 : Lower(std::move(L)), Upper(std::move(U)) { 51 assert(Lower.getBitWidth() == Upper.getBitWidth() && 52 "ConstantRange with unequal bit widths"); 53 assert((Lower != Upper || (Lower.isMaxValue() || Lower.isMinValue())) && 54 "Lower == Upper, but they aren't min or max value!"); 55 } 56 57 ConstantRange ConstantRange::fromKnownBits(const KnownBits &Known, 58 bool IsSigned) { 59 assert(!Known.hasConflict() && "Expected valid KnownBits"); 60 61 if (Known.isUnknown()) 62 return getFull(Known.getBitWidth()); 63 64 // For unsigned ranges, or signed ranges with known sign bit, create a simple 65 // range between the smallest and largest possible value. 66 if (!IsSigned || Known.isNegative() || Known.isNonNegative()) 67 return ConstantRange(Known.One, ~Known.Zero + 1); 68 69 // If we don't know the sign bit, pick the lower bound as a negative number 70 // and the upper bound as a non-negative one. 71 APInt Lower = Known.One, Upper = ~Known.Zero; 72 Lower.setSignBit(); 73 Upper.clearSignBit(); 74 return ConstantRange(Lower, Upper + 1); 75 } 76 77 ConstantRange ConstantRange::makeAllowedICmpRegion(CmpInst::Predicate Pred, 78 const ConstantRange &CR) { 79 if (CR.isEmptySet()) 80 return CR; 81 82 uint32_t W = CR.getBitWidth(); 83 switch (Pred) { 84 default: 85 llvm_unreachable("Invalid ICmp predicate to makeAllowedICmpRegion()"); 86 case CmpInst::ICMP_EQ: 87 return CR; 88 case CmpInst::ICMP_NE: 89 if (CR.isSingleElement()) 90 return ConstantRange(CR.getUpper(), CR.getLower()); 91 return getFull(W); 92 case CmpInst::ICMP_ULT: { 93 APInt UMax(CR.getUnsignedMax()); 94 if (UMax.isMinValue()) 95 return getEmpty(W); 96 return ConstantRange(APInt::getMinValue(W), std::move(UMax)); 97 } 98 case CmpInst::ICMP_SLT: { 99 APInt SMax(CR.getSignedMax()); 100 if (SMax.isMinSignedValue()) 101 return getEmpty(W); 102 return ConstantRange(APInt::getSignedMinValue(W), std::move(SMax)); 103 } 104 case CmpInst::ICMP_ULE: 105 return getNonEmpty(APInt::getMinValue(W), CR.getUnsignedMax() + 1); 106 case CmpInst::ICMP_SLE: 107 return getNonEmpty(APInt::getSignedMinValue(W), CR.getSignedMax() + 1); 108 case CmpInst::ICMP_UGT: { 109 APInt UMin(CR.getUnsignedMin()); 110 if (UMin.isMaxValue()) 111 return getEmpty(W); 112 return ConstantRange(std::move(UMin) + 1, APInt::getNullValue(W)); 113 } 114 case CmpInst::ICMP_SGT: { 115 APInt SMin(CR.getSignedMin()); 116 if (SMin.isMaxSignedValue()) 117 return getEmpty(W); 118 return ConstantRange(std::move(SMin) + 1, APInt::getSignedMinValue(W)); 119 } 120 case CmpInst::ICMP_UGE: 121 return getNonEmpty(CR.getUnsignedMin(), APInt::getNullValue(W)); 122 case CmpInst::ICMP_SGE: 123 return getNonEmpty(CR.getSignedMin(), APInt::getSignedMinValue(W)); 124 } 125 } 126 127 ConstantRange ConstantRange::makeSatisfyingICmpRegion(CmpInst::Predicate Pred, 128 const ConstantRange &CR) { 129 // Follows from De-Morgan's laws: 130 // 131 // ~(~A union ~B) == A intersect B. 132 // 133 return makeAllowedICmpRegion(CmpInst::getInversePredicate(Pred), CR) 134 .inverse(); 135 } 136 137 ConstantRange ConstantRange::makeExactICmpRegion(CmpInst::Predicate Pred, 138 const APInt &C) { 139 // Computes the exact range that is equal to both the constant ranges returned 140 // by makeAllowedICmpRegion and makeSatisfyingICmpRegion. This is always true 141 // when RHS is a singleton such as an APInt and so the assert is valid. 142 // However for non-singleton RHS, for example ult [2,5) makeAllowedICmpRegion 143 // returns [0,4) but makeSatisfyICmpRegion returns [0,2). 144 // 145 assert(makeAllowedICmpRegion(Pred, C) == makeSatisfyingICmpRegion(Pred, C)); 146 return makeAllowedICmpRegion(Pred, C); 147 } 148 149 bool ConstantRange::getEquivalentICmp(CmpInst::Predicate &Pred, 150 APInt &RHS) const { 151 bool Success = false; 152 153 if (isFullSet() || isEmptySet()) { 154 Pred = isEmptySet() ? CmpInst::ICMP_ULT : CmpInst::ICMP_UGE; 155 RHS = APInt(getBitWidth(), 0); 156 Success = true; 157 } else if (auto *OnlyElt = getSingleElement()) { 158 Pred = CmpInst::ICMP_EQ; 159 RHS = *OnlyElt; 160 Success = true; 161 } else if (auto *OnlyMissingElt = getSingleMissingElement()) { 162 Pred = CmpInst::ICMP_NE; 163 RHS = *OnlyMissingElt; 164 Success = true; 165 } else if (getLower().isMinSignedValue() || getLower().isMinValue()) { 166 Pred = 167 getLower().isMinSignedValue() ? CmpInst::ICMP_SLT : CmpInst::ICMP_ULT; 168 RHS = getUpper(); 169 Success = true; 170 } else if (getUpper().isMinSignedValue() || getUpper().isMinValue()) { 171 Pred = 172 getUpper().isMinSignedValue() ? CmpInst::ICMP_SGE : CmpInst::ICMP_UGE; 173 RHS = getLower(); 174 Success = true; 175 } 176 177 assert((!Success || ConstantRange::makeExactICmpRegion(Pred, RHS) == *this) && 178 "Bad result!"); 179 180 return Success; 181 } 182 183 ConstantRange 184 ConstantRange::makeGuaranteedNoWrapRegion(Instruction::BinaryOps BinOp, 185 const ConstantRange &Other, 186 unsigned NoWrapKind) { 187 using OBO = OverflowingBinaryOperator; 188 189 // Computes the intersection of CR0 and CR1. It is different from 190 // intersectWith in that the ConstantRange returned will only contain elements 191 // in both CR0 and CR1 (i.e. SubsetIntersect(X, Y) is a *subset*, proper or 192 // not, of both X and Y). 193 auto SubsetIntersect = 194 [](const ConstantRange &CR0, const ConstantRange &CR1) { 195 return CR0.inverse().unionWith(CR1.inverse()).inverse(); 196 }; 197 198 assert(Instruction::isBinaryOp(BinOp) && "Binary operators only!"); 199 200 assert((NoWrapKind == OBO::NoSignedWrap || 201 NoWrapKind == OBO::NoUnsignedWrap) && 202 "NoWrapKind invalid!"); 203 204 unsigned BitWidth = Other.getBitWidth(); 205 ConstantRange Result(BitWidth, /* full */ true); 206 207 switch (BinOp) { 208 default: 209 // Conservative answer: empty set 210 return getEmpty(BitWidth); 211 212 case Instruction::Add: 213 if (auto *C = Other.getSingleElement()) 214 if (C->isNullValue()) 215 // Full set: nothing signed / unsigned wraps when added to 0. 216 return getFull(BitWidth); 217 218 if (NoWrapKind == OBO::NoUnsignedWrap) 219 return ConstantRange(APInt::getNullValue(BitWidth), 220 -Other.getUnsignedMax()); 221 222 if (NoWrapKind == OBO::NoSignedWrap) { 223 const APInt &SignedMin = Other.getSignedMin(); 224 const APInt &SignedMax = Other.getSignedMax(); 225 if (SignedMax.isStrictlyPositive()) 226 Result = SubsetIntersect( 227 Result, 228 ConstantRange(APInt::getSignedMinValue(BitWidth), 229 APInt::getSignedMinValue(BitWidth) - SignedMax)); 230 if (SignedMin.isNegative()) 231 Result = SubsetIntersect( 232 Result, 233 ConstantRange(APInt::getSignedMinValue(BitWidth) - SignedMin, 234 APInt::getSignedMinValue(BitWidth))); 235 } 236 return Result; 237 238 case Instruction::Sub: 239 if (auto *C = Other.getSingleElement()) 240 if (C->isNullValue()) 241 // Full set: nothing signed / unsigned wraps when subtracting 0. 242 return getFull(BitWidth); 243 244 if (NoWrapKind == OBO::NoUnsignedWrap) 245 return ConstantRange(Other.getUnsignedMax(), 246 APInt::getMinValue(BitWidth)); 247 248 if (NoWrapKind == OBO::NoSignedWrap) { 249 const APInt &SignedMin = Other.getSignedMin(); 250 const APInt &SignedMax = Other.getSignedMax(); 251 if (SignedMax.isStrictlyPositive()) 252 Result = SubsetIntersect( 253 Result, 254 ConstantRange(APInt::getSignedMinValue(BitWidth) + SignedMax, 255 APInt::getSignedMinValue(BitWidth))); 256 if (SignedMin.isNegative()) 257 Result = SubsetIntersect( 258 Result, 259 ConstantRange(APInt::getSignedMinValue(BitWidth), 260 APInt::getSignedMinValue(BitWidth) + SignedMin)); 261 } 262 return Result; 263 264 case Instruction::Mul: { 265 // Equivalent to calling makeGuaranteedNoWrapRegion() on [V, V+1). 266 const bool Unsigned = NoWrapKind == OBO::NoUnsignedWrap; 267 const auto makeSingleValueRegion = [Unsigned, 268 BitWidth](APInt V) -> ConstantRange { 269 // Handle special case for 0, -1 and 1. See the last for reason why we 270 // specialize -1 and 1. 271 if (V == 0 || V.isOneValue()) 272 return getFull(BitWidth); 273 274 APInt MinValue, MaxValue; 275 if (Unsigned) { 276 MinValue = APInt::getMinValue(BitWidth); 277 MaxValue = APInt::getMaxValue(BitWidth); 278 } else { 279 MinValue = APInt::getSignedMinValue(BitWidth); 280 MaxValue = APInt::getSignedMaxValue(BitWidth); 281 } 282 // e.g. Returning [-127, 127], represented as [-127, -128). 283 if (!Unsigned && V.isAllOnesValue()) 284 return ConstantRange(-MaxValue, MinValue); 285 286 APInt Lower, Upper; 287 if (!Unsigned && V.isNegative()) { 288 Lower = APIntOps::RoundingSDiv(MaxValue, V, APInt::Rounding::UP); 289 Upper = APIntOps::RoundingSDiv(MinValue, V, APInt::Rounding::DOWN); 290 } else if (Unsigned) { 291 Lower = APIntOps::RoundingUDiv(MinValue, V, APInt::Rounding::UP); 292 Upper = APIntOps::RoundingUDiv(MaxValue, V, APInt::Rounding::DOWN); 293 } else { 294 Lower = APIntOps::RoundingSDiv(MinValue, V, APInt::Rounding::UP); 295 Upper = APIntOps::RoundingSDiv(MaxValue, V, APInt::Rounding::DOWN); 296 } 297 // ConstantRange ctor take a half inclusive interval [Lower, Upper + 1). 298 // Upper + 1 is guanranteed not to overflow, because |divisor| > 1. 0, -1, 299 // and 1 are already handled as special cases. 300 return ConstantRange(Lower, Upper + 1); 301 }; 302 303 if (Unsigned) 304 return makeSingleValueRegion(Other.getUnsignedMax()); 305 306 return SubsetIntersect(makeSingleValueRegion(Other.getSignedMin()), 307 makeSingleValueRegion(Other.getSignedMax())); 308 } 309 } 310 } 311 312 bool ConstantRange::isFullSet() const { 313 return Lower == Upper && Lower.isMaxValue(); 314 } 315 316 bool ConstantRange::isEmptySet() const { 317 return Lower == Upper && Lower.isMinValue(); 318 } 319 320 bool ConstantRange::isWrappedSet() const { 321 return Lower.ugt(Upper) && !Upper.isNullValue(); 322 } 323 324 bool ConstantRange::isUpperWrapped() const { 325 return Lower.ugt(Upper); 326 } 327 328 bool ConstantRange::isSignWrappedSet() const { 329 return Lower.sgt(Upper) && !Upper.isMinSignedValue(); 330 } 331 332 bool ConstantRange::isUpperSignWrapped() const { 333 return Lower.sgt(Upper); 334 } 335 336 bool 337 ConstantRange::isSizeStrictlySmallerThan(const ConstantRange &Other) const { 338 assert(getBitWidth() == Other.getBitWidth()); 339 if (isFullSet()) 340 return false; 341 if (Other.isFullSet()) 342 return true; 343 return (Upper - Lower).ult(Other.Upper - Other.Lower); 344 } 345 346 bool 347 ConstantRange::isSizeLargerThan(uint64_t MaxSize) const { 348 assert(MaxSize && "MaxSize can't be 0."); 349 // If this a full set, we need special handling to avoid needing an extra bit 350 // to represent the size. 351 if (isFullSet()) 352 return APInt::getMaxValue(getBitWidth()).ugt(MaxSize - 1); 353 354 return (Upper - Lower).ugt(MaxSize); 355 } 356 357 bool ConstantRange::isAllNegative() const { 358 // Empty set is all negative, full set is not. 359 if (isEmptySet()) 360 return true; 361 if (isFullSet()) 362 return false; 363 364 return !isUpperSignWrapped() && !Upper.isStrictlyPositive(); 365 } 366 367 bool ConstantRange::isAllNonNegative() const { 368 // Empty and full set are automatically treated correctly. 369 return !isSignWrappedSet() && Lower.isNonNegative(); 370 } 371 372 APInt ConstantRange::getUnsignedMax() const { 373 if (isFullSet() || isUpperWrapped()) 374 return APInt::getMaxValue(getBitWidth()); 375 return getUpper() - 1; 376 } 377 378 APInt ConstantRange::getUnsignedMin() const { 379 if (isFullSet() || isWrappedSet()) 380 return APInt::getMinValue(getBitWidth()); 381 return getLower(); 382 } 383 384 APInt ConstantRange::getSignedMax() const { 385 if (isFullSet() || isUpperSignWrapped()) 386 return APInt::getSignedMaxValue(getBitWidth()); 387 return getUpper() - 1; 388 } 389 390 APInt ConstantRange::getSignedMin() const { 391 if (isFullSet() || isSignWrappedSet()) 392 return APInt::getSignedMinValue(getBitWidth()); 393 return getLower(); 394 } 395 396 bool ConstantRange::contains(const APInt &V) const { 397 if (Lower == Upper) 398 return isFullSet(); 399 400 if (!isUpperWrapped()) 401 return Lower.ule(V) && V.ult(Upper); 402 return Lower.ule(V) || V.ult(Upper); 403 } 404 405 bool ConstantRange::contains(const ConstantRange &Other) const { 406 if (isFullSet() || Other.isEmptySet()) return true; 407 if (isEmptySet() || Other.isFullSet()) return false; 408 409 if (!isUpperWrapped()) { 410 if (Other.isUpperWrapped()) 411 return false; 412 413 return Lower.ule(Other.getLower()) && Other.getUpper().ule(Upper); 414 } 415 416 if (!Other.isUpperWrapped()) 417 return Other.getUpper().ule(Upper) || 418 Lower.ule(Other.getLower()); 419 420 return Other.getUpper().ule(Upper) && Lower.ule(Other.getLower()); 421 } 422 423 ConstantRange ConstantRange::subtract(const APInt &Val) const { 424 assert(Val.getBitWidth() == getBitWidth() && "Wrong bit width"); 425 // If the set is empty or full, don't modify the endpoints. 426 if (Lower == Upper) 427 return *this; 428 return ConstantRange(Lower - Val, Upper - Val); 429 } 430 431 ConstantRange ConstantRange::difference(const ConstantRange &CR) const { 432 return intersectWith(CR.inverse()); 433 } 434 435 static ConstantRange getPreferredRange( 436 const ConstantRange &CR1, const ConstantRange &CR2, 437 ConstantRange::PreferredRangeType Type) { 438 if (Type == ConstantRange::Unsigned) { 439 if (!CR1.isWrappedSet() && CR2.isWrappedSet()) 440 return CR1; 441 if (CR1.isWrappedSet() && !CR2.isWrappedSet()) 442 return CR2; 443 } else if (Type == ConstantRange::Signed) { 444 if (!CR1.isSignWrappedSet() && CR2.isSignWrappedSet()) 445 return CR1; 446 if (CR1.isSignWrappedSet() && !CR2.isSignWrappedSet()) 447 return CR2; 448 } 449 450 if (CR1.isSizeStrictlySmallerThan(CR2)) 451 return CR1; 452 return CR2; 453 } 454 455 ConstantRange ConstantRange::intersectWith(const ConstantRange &CR, 456 PreferredRangeType Type) const { 457 assert(getBitWidth() == CR.getBitWidth() && 458 "ConstantRange types don't agree!"); 459 460 // Handle common cases. 461 if ( isEmptySet() || CR.isFullSet()) return *this; 462 if (CR.isEmptySet() || isFullSet()) return CR; 463 464 if (!isUpperWrapped() && CR.isUpperWrapped()) 465 return CR.intersectWith(*this, Type); 466 467 if (!isUpperWrapped() && !CR.isUpperWrapped()) { 468 if (Lower.ult(CR.Lower)) { 469 // L---U : this 470 // L---U : CR 471 if (Upper.ule(CR.Lower)) 472 return getEmpty(); 473 474 // L---U : this 475 // L---U : CR 476 if (Upper.ult(CR.Upper)) 477 return ConstantRange(CR.Lower, Upper); 478 479 // L-------U : this 480 // L---U : CR 481 return CR; 482 } 483 // L---U : this 484 // L-------U : CR 485 if (Upper.ult(CR.Upper)) 486 return *this; 487 488 // L-----U : this 489 // L-----U : CR 490 if (Lower.ult(CR.Upper)) 491 return ConstantRange(Lower, CR.Upper); 492 493 // L---U : this 494 // L---U : CR 495 return getEmpty(); 496 } 497 498 if (isUpperWrapped() && !CR.isUpperWrapped()) { 499 if (CR.Lower.ult(Upper)) { 500 // ------U L--- : this 501 // L--U : CR 502 if (CR.Upper.ult(Upper)) 503 return CR; 504 505 // ------U L--- : this 506 // L------U : CR 507 if (CR.Upper.ule(Lower)) 508 return ConstantRange(CR.Lower, Upper); 509 510 // ------U L--- : this 511 // L----------U : CR 512 return getPreferredRange(*this, CR, Type); 513 } 514 if (CR.Lower.ult(Lower)) { 515 // --U L---- : this 516 // L--U : CR 517 if (CR.Upper.ule(Lower)) 518 return getEmpty(); 519 520 // --U L---- : this 521 // L------U : CR 522 return ConstantRange(Lower, CR.Upper); 523 } 524 525 // --U L------ : this 526 // L--U : CR 527 return CR; 528 } 529 530 if (CR.Upper.ult(Upper)) { 531 // ------U L-- : this 532 // --U L------ : CR 533 if (CR.Lower.ult(Upper)) 534 return getPreferredRange(*this, CR, Type); 535 536 // ----U L-- : this 537 // --U L---- : CR 538 if (CR.Lower.ult(Lower)) 539 return ConstantRange(Lower, CR.Upper); 540 541 // ----U L---- : this 542 // --U L-- : CR 543 return CR; 544 } 545 if (CR.Upper.ule(Lower)) { 546 // --U L-- : this 547 // ----U L---- : CR 548 if (CR.Lower.ult(Lower)) 549 return *this; 550 551 // --U L---- : this 552 // ----U L-- : CR 553 return ConstantRange(CR.Lower, Upper); 554 } 555 556 // --U L------ : this 557 // ------U L-- : CR 558 return getPreferredRange(*this, CR, Type); 559 } 560 561 ConstantRange ConstantRange::unionWith(const ConstantRange &CR, 562 PreferredRangeType Type) const { 563 assert(getBitWidth() == CR.getBitWidth() && 564 "ConstantRange types don't agree!"); 565 566 if ( isFullSet() || CR.isEmptySet()) return *this; 567 if (CR.isFullSet() || isEmptySet()) return CR; 568 569 if (!isUpperWrapped() && CR.isUpperWrapped()) 570 return CR.unionWith(*this, Type); 571 572 if (!isUpperWrapped() && !CR.isUpperWrapped()) { 573 // L---U and L---U : this 574 // L---U L---U : CR 575 // result in one of 576 // L---------U 577 // -----U L----- 578 if (CR.Upper.ult(Lower) || Upper.ult(CR.Lower)) 579 return getPreferredRange( 580 ConstantRange(Lower, CR.Upper), ConstantRange(CR.Lower, Upper), Type); 581 582 APInt L = CR.Lower.ult(Lower) ? CR.Lower : Lower; 583 APInt U = (CR.Upper - 1).ugt(Upper - 1) ? CR.Upper : Upper; 584 585 if (L.isNullValue() && U.isNullValue()) 586 return getFull(); 587 588 return ConstantRange(std::move(L), std::move(U)); 589 } 590 591 if (!CR.isUpperWrapped()) { 592 // ------U L----- and ------U L----- : this 593 // L--U L--U : CR 594 if (CR.Upper.ule(Upper) || CR.Lower.uge(Lower)) 595 return *this; 596 597 // ------U L----- : this 598 // L---------U : CR 599 if (CR.Lower.ule(Upper) && Lower.ule(CR.Upper)) 600 return getFull(); 601 602 // ----U L---- : this 603 // L---U : CR 604 // results in one of 605 // ----------U L---- 606 // ----U L---------- 607 if (Upper.ult(CR.Lower) && CR.Upper.ult(Lower)) 608 return getPreferredRange( 609 ConstantRange(Lower, CR.Upper), ConstantRange(CR.Lower, Upper), Type); 610 611 // ----U L----- : this 612 // L----U : CR 613 if (Upper.ult(CR.Lower) && Lower.ule(CR.Upper)) 614 return ConstantRange(CR.Lower, Upper); 615 616 // ------U L---- : this 617 // L-----U : CR 618 assert(CR.Lower.ule(Upper) && CR.Upper.ult(Lower) && 619 "ConstantRange::unionWith missed a case with one range wrapped"); 620 return ConstantRange(Lower, CR.Upper); 621 } 622 623 // ------U L---- and ------U L---- : this 624 // -U L----------- and ------------U L : CR 625 if (CR.Lower.ule(Upper) || Lower.ule(CR.Upper)) 626 return getFull(); 627 628 APInt L = CR.Lower.ult(Lower) ? CR.Lower : Lower; 629 APInt U = CR.Upper.ugt(Upper) ? CR.Upper : Upper; 630 631 return ConstantRange(std::move(L), std::move(U)); 632 } 633 634 ConstantRange ConstantRange::castOp(Instruction::CastOps CastOp, 635 uint32_t ResultBitWidth) const { 636 switch (CastOp) { 637 default: 638 llvm_unreachable("unsupported cast type"); 639 case Instruction::Trunc: 640 return truncate(ResultBitWidth); 641 case Instruction::SExt: 642 return signExtend(ResultBitWidth); 643 case Instruction::ZExt: 644 return zeroExtend(ResultBitWidth); 645 case Instruction::BitCast: 646 return *this; 647 case Instruction::FPToUI: 648 case Instruction::FPToSI: 649 if (getBitWidth() == ResultBitWidth) 650 return *this; 651 else 652 return getFull(); 653 case Instruction::UIToFP: { 654 // TODO: use input range if available 655 auto BW = getBitWidth(); 656 APInt Min = APInt::getMinValue(BW).zextOrSelf(ResultBitWidth); 657 APInt Max = APInt::getMaxValue(BW).zextOrSelf(ResultBitWidth); 658 return ConstantRange(std::move(Min), std::move(Max)); 659 } 660 case Instruction::SIToFP: { 661 // TODO: use input range if available 662 auto BW = getBitWidth(); 663 APInt SMin = APInt::getSignedMinValue(BW).sextOrSelf(ResultBitWidth); 664 APInt SMax = APInt::getSignedMaxValue(BW).sextOrSelf(ResultBitWidth); 665 return ConstantRange(std::move(SMin), std::move(SMax)); 666 } 667 case Instruction::FPTrunc: 668 case Instruction::FPExt: 669 case Instruction::IntToPtr: 670 case Instruction::PtrToInt: 671 case Instruction::AddrSpaceCast: 672 // Conservatively return getFull set. 673 return getFull(); 674 }; 675 } 676 677 ConstantRange ConstantRange::zeroExtend(uint32_t DstTySize) const { 678 if (isEmptySet()) return getEmpty(DstTySize); 679 680 unsigned SrcTySize = getBitWidth(); 681 assert(SrcTySize < DstTySize && "Not a value extension"); 682 if (isFullSet() || isUpperWrapped()) { 683 // Change into [0, 1 << src bit width) 684 APInt LowerExt(DstTySize, 0); 685 if (!Upper) // special case: [X, 0) -- not really wrapping around 686 LowerExt = Lower.zext(DstTySize); 687 return ConstantRange(std::move(LowerExt), 688 APInt::getOneBitSet(DstTySize, SrcTySize)); 689 } 690 691 return ConstantRange(Lower.zext(DstTySize), Upper.zext(DstTySize)); 692 } 693 694 ConstantRange ConstantRange::signExtend(uint32_t DstTySize) const { 695 if (isEmptySet()) return getEmpty(DstTySize); 696 697 unsigned SrcTySize = getBitWidth(); 698 assert(SrcTySize < DstTySize && "Not a value extension"); 699 700 // special case: [X, INT_MIN) -- not really wrapping around 701 if (Upper.isMinSignedValue()) 702 return ConstantRange(Lower.sext(DstTySize), Upper.zext(DstTySize)); 703 704 if (isFullSet() || isSignWrappedSet()) { 705 return ConstantRange(APInt::getHighBitsSet(DstTySize,DstTySize-SrcTySize+1), 706 APInt::getLowBitsSet(DstTySize, SrcTySize-1) + 1); 707 } 708 709 return ConstantRange(Lower.sext(DstTySize), Upper.sext(DstTySize)); 710 } 711 712 ConstantRange ConstantRange::truncate(uint32_t DstTySize) const { 713 assert(getBitWidth() > DstTySize && "Not a value truncation"); 714 if (isEmptySet()) 715 return getEmpty(DstTySize); 716 if (isFullSet()) 717 return getFull(DstTySize); 718 719 APInt LowerDiv(Lower), UpperDiv(Upper); 720 ConstantRange Union(DstTySize, /*isFullSet=*/false); 721 722 // Analyze wrapped sets in their two parts: [0, Upper) \/ [Lower, MaxValue] 723 // We use the non-wrapped set code to analyze the [Lower, MaxValue) part, and 724 // then we do the union with [MaxValue, Upper) 725 if (isUpperWrapped()) { 726 // If Upper is greater than or equal to MaxValue(DstTy), it covers the whole 727 // truncated range. 728 if (Upper.getActiveBits() > DstTySize || 729 Upper.countTrailingOnes() == DstTySize) 730 return getFull(DstTySize); 731 732 Union = ConstantRange(APInt::getMaxValue(DstTySize),Upper.trunc(DstTySize)); 733 UpperDiv.setAllBits(); 734 735 // Union covers the MaxValue case, so return if the remaining range is just 736 // MaxValue(DstTy). 737 if (LowerDiv == UpperDiv) 738 return Union; 739 } 740 741 // Chop off the most significant bits that are past the destination bitwidth. 742 if (LowerDiv.getActiveBits() > DstTySize) { 743 // Mask to just the signficant bits and subtract from LowerDiv/UpperDiv. 744 APInt Adjust = LowerDiv & APInt::getBitsSetFrom(getBitWidth(), DstTySize); 745 LowerDiv -= Adjust; 746 UpperDiv -= Adjust; 747 } 748 749 unsigned UpperDivWidth = UpperDiv.getActiveBits(); 750 if (UpperDivWidth <= DstTySize) 751 return ConstantRange(LowerDiv.trunc(DstTySize), 752 UpperDiv.trunc(DstTySize)).unionWith(Union); 753 754 // The truncated value wraps around. Check if we can do better than fullset. 755 if (UpperDivWidth == DstTySize + 1) { 756 // Clear the MSB so that UpperDiv wraps around. 757 UpperDiv.clearBit(DstTySize); 758 if (UpperDiv.ult(LowerDiv)) 759 return ConstantRange(LowerDiv.trunc(DstTySize), 760 UpperDiv.trunc(DstTySize)).unionWith(Union); 761 } 762 763 return getFull(DstTySize); 764 } 765 766 ConstantRange ConstantRange::zextOrTrunc(uint32_t DstTySize) const { 767 unsigned SrcTySize = getBitWidth(); 768 if (SrcTySize > DstTySize) 769 return truncate(DstTySize); 770 if (SrcTySize < DstTySize) 771 return zeroExtend(DstTySize); 772 return *this; 773 } 774 775 ConstantRange ConstantRange::sextOrTrunc(uint32_t DstTySize) const { 776 unsigned SrcTySize = getBitWidth(); 777 if (SrcTySize > DstTySize) 778 return truncate(DstTySize); 779 if (SrcTySize < DstTySize) 780 return signExtend(DstTySize); 781 return *this; 782 } 783 784 ConstantRange ConstantRange::binaryOp(Instruction::BinaryOps BinOp, 785 const ConstantRange &Other) const { 786 assert(Instruction::isBinaryOp(BinOp) && "Binary operators only!"); 787 788 switch (BinOp) { 789 case Instruction::Add: 790 return add(Other); 791 case Instruction::Sub: 792 return sub(Other); 793 case Instruction::Mul: 794 return multiply(Other); 795 case Instruction::UDiv: 796 return udiv(Other); 797 case Instruction::URem: 798 return urem(Other); 799 case Instruction::Shl: 800 return shl(Other); 801 case Instruction::LShr: 802 return lshr(Other); 803 case Instruction::AShr: 804 return ashr(Other); 805 case Instruction::And: 806 return binaryAnd(Other); 807 case Instruction::Or: 808 return binaryOr(Other); 809 // Note: floating point operations applied to abstract ranges are just 810 // ideal integer operations with a lossy representation 811 case Instruction::FAdd: 812 return add(Other); 813 case Instruction::FSub: 814 return sub(Other); 815 case Instruction::FMul: 816 return multiply(Other); 817 default: 818 // Conservatively return getFull set. 819 return getFull(); 820 } 821 } 822 823 ConstantRange 824 ConstantRange::add(const ConstantRange &Other) const { 825 if (isEmptySet() || Other.isEmptySet()) 826 return getEmpty(); 827 if (isFullSet() || Other.isFullSet()) 828 return getFull(); 829 830 APInt NewLower = getLower() + Other.getLower(); 831 APInt NewUpper = getUpper() + Other.getUpper() - 1; 832 if (NewLower == NewUpper) 833 return getFull(); 834 835 ConstantRange X = ConstantRange(std::move(NewLower), std::move(NewUpper)); 836 if (X.isSizeStrictlySmallerThan(*this) || 837 X.isSizeStrictlySmallerThan(Other)) 838 // We've wrapped, therefore, full set. 839 return getFull(); 840 return X; 841 } 842 843 ConstantRange ConstantRange::addWithNoSignedWrap(const APInt &Other) const { 844 // Calculate the subset of this range such that "X + Other" is 845 // guaranteed not to wrap (overflow) for all X in this subset. 846 // makeGuaranteedNoWrapRegion will produce an exact NSW range. 847 auto NSWRange = ConstantRange::makeGuaranteedNoWrapRegion(BinaryOperator::Add, 848 ConstantRange(Other), 849 OverflowingBinaryOperator::NoSignedWrap); 850 auto NSWConstrainedRange = intersectWith(NSWRange); 851 852 return NSWConstrainedRange.add(ConstantRange(Other)); 853 } 854 855 ConstantRange 856 ConstantRange::sub(const ConstantRange &Other) const { 857 if (isEmptySet() || Other.isEmptySet()) 858 return getEmpty(); 859 if (isFullSet() || Other.isFullSet()) 860 return getFull(); 861 862 APInt NewLower = getLower() - Other.getUpper() + 1; 863 APInt NewUpper = getUpper() - Other.getLower(); 864 if (NewLower == NewUpper) 865 return getFull(); 866 867 ConstantRange X = ConstantRange(std::move(NewLower), std::move(NewUpper)); 868 if (X.isSizeStrictlySmallerThan(*this) || 869 X.isSizeStrictlySmallerThan(Other)) 870 // We've wrapped, therefore, full set. 871 return getFull(); 872 return X; 873 } 874 875 ConstantRange 876 ConstantRange::multiply(const ConstantRange &Other) const { 877 // TODO: If either operand is a single element and the multiply is known to 878 // be non-wrapping, round the result min and max value to the appropriate 879 // multiple of that element. If wrapping is possible, at least adjust the 880 // range according to the greatest power-of-two factor of the single element. 881 882 if (isEmptySet() || Other.isEmptySet()) 883 return getEmpty(); 884 885 // Multiplication is signedness-independent. However different ranges can be 886 // obtained depending on how the input ranges are treated. These different 887 // ranges are all conservatively correct, but one might be better than the 888 // other. We calculate two ranges; one treating the inputs as unsigned 889 // and the other signed, then return the smallest of these ranges. 890 891 // Unsigned range first. 892 APInt this_min = getUnsignedMin().zext(getBitWidth() * 2); 893 APInt this_max = getUnsignedMax().zext(getBitWidth() * 2); 894 APInt Other_min = Other.getUnsignedMin().zext(getBitWidth() * 2); 895 APInt Other_max = Other.getUnsignedMax().zext(getBitWidth() * 2); 896 897 ConstantRange Result_zext = ConstantRange(this_min * Other_min, 898 this_max * Other_max + 1); 899 ConstantRange UR = Result_zext.truncate(getBitWidth()); 900 901 // If the unsigned range doesn't wrap, and isn't negative then it's a range 902 // from one positive number to another which is as good as we can generate. 903 // In this case, skip the extra work of generating signed ranges which aren't 904 // going to be better than this range. 905 if (!UR.isUpperWrapped() && 906 (UR.getUpper().isNonNegative() || UR.getUpper().isMinSignedValue())) 907 return UR; 908 909 // Now the signed range. Because we could be dealing with negative numbers 910 // here, the lower bound is the smallest of the cartesian product of the 911 // lower and upper ranges; for example: 912 // [-1,4) * [-2,3) = min(-1*-2, -1*2, 3*-2, 3*2) = -6. 913 // Similarly for the upper bound, swapping min for max. 914 915 this_min = getSignedMin().sext(getBitWidth() * 2); 916 this_max = getSignedMax().sext(getBitWidth() * 2); 917 Other_min = Other.getSignedMin().sext(getBitWidth() * 2); 918 Other_max = Other.getSignedMax().sext(getBitWidth() * 2); 919 920 auto L = {this_min * Other_min, this_min * Other_max, 921 this_max * Other_min, this_max * Other_max}; 922 auto Compare = [](const APInt &A, const APInt &B) { return A.slt(B); }; 923 ConstantRange Result_sext(std::min(L, Compare), std::max(L, Compare) + 1); 924 ConstantRange SR = Result_sext.truncate(getBitWidth()); 925 926 return UR.isSizeStrictlySmallerThan(SR) ? UR : SR; 927 } 928 929 ConstantRange 930 ConstantRange::smax(const ConstantRange &Other) const { 931 // X smax Y is: range(smax(X_smin, Y_smin), 932 // smax(X_smax, Y_smax)) 933 if (isEmptySet() || Other.isEmptySet()) 934 return getEmpty(); 935 APInt NewL = APIntOps::smax(getSignedMin(), Other.getSignedMin()); 936 APInt NewU = APIntOps::smax(getSignedMax(), Other.getSignedMax()) + 1; 937 return getNonEmpty(std::move(NewL), std::move(NewU)); 938 } 939 940 ConstantRange 941 ConstantRange::umax(const ConstantRange &Other) const { 942 // X umax Y is: range(umax(X_umin, Y_umin), 943 // umax(X_umax, Y_umax)) 944 if (isEmptySet() || Other.isEmptySet()) 945 return getEmpty(); 946 APInt NewL = APIntOps::umax(getUnsignedMin(), Other.getUnsignedMin()); 947 APInt NewU = APIntOps::umax(getUnsignedMax(), Other.getUnsignedMax()) + 1; 948 return getNonEmpty(std::move(NewL), std::move(NewU)); 949 } 950 951 ConstantRange 952 ConstantRange::smin(const ConstantRange &Other) const { 953 // X smin Y is: range(smin(X_smin, Y_smin), 954 // smin(X_smax, Y_smax)) 955 if (isEmptySet() || Other.isEmptySet()) 956 return getEmpty(); 957 APInt NewL = APIntOps::smin(getSignedMin(), Other.getSignedMin()); 958 APInt NewU = APIntOps::smin(getSignedMax(), Other.getSignedMax()) + 1; 959 return getNonEmpty(std::move(NewL), std::move(NewU)); 960 } 961 962 ConstantRange 963 ConstantRange::umin(const ConstantRange &Other) const { 964 // X umin Y is: range(umin(X_umin, Y_umin), 965 // umin(X_umax, Y_umax)) 966 if (isEmptySet() || Other.isEmptySet()) 967 return getEmpty(); 968 APInt NewL = APIntOps::umin(getUnsignedMin(), Other.getUnsignedMin()); 969 APInt NewU = APIntOps::umin(getUnsignedMax(), Other.getUnsignedMax()) + 1; 970 return getNonEmpty(std::move(NewL), std::move(NewU)); 971 } 972 973 ConstantRange 974 ConstantRange::udiv(const ConstantRange &RHS) const { 975 if (isEmptySet() || RHS.isEmptySet() || RHS.getUnsignedMax().isNullValue()) 976 return getEmpty(); 977 978 APInt Lower = getUnsignedMin().udiv(RHS.getUnsignedMax()); 979 980 APInt RHS_umin = RHS.getUnsignedMin(); 981 if (RHS_umin.isNullValue()) { 982 // We want the lowest value in RHS excluding zero. Usually that would be 1 983 // except for a range in the form of [X, 1) in which case it would be X. 984 if (RHS.getUpper() == 1) 985 RHS_umin = RHS.getLower(); 986 else 987 RHS_umin = 1; 988 } 989 990 APInt Upper = getUnsignedMax().udiv(RHS_umin) + 1; 991 return getNonEmpty(std::move(Lower), std::move(Upper)); 992 } 993 994 ConstantRange ConstantRange::urem(const ConstantRange &RHS) const { 995 if (isEmptySet() || RHS.isEmptySet() || RHS.getUnsignedMax().isNullValue()) 996 return getEmpty(); 997 998 // L % R for L < R is L. 999 if (getUnsignedMax().ult(RHS.getUnsignedMin())) 1000 return *this; 1001 1002 // L % R is <= L and < R. 1003 APInt Upper = APIntOps::umin(getUnsignedMax(), RHS.getUnsignedMax() - 1) + 1; 1004 return getNonEmpty(APInt::getNullValue(getBitWidth()), std::move(Upper)); 1005 } 1006 1007 ConstantRange 1008 ConstantRange::binaryAnd(const ConstantRange &Other) const { 1009 if (isEmptySet() || Other.isEmptySet()) 1010 return getEmpty(); 1011 1012 // TODO: replace this with something less conservative 1013 1014 APInt umin = APIntOps::umin(Other.getUnsignedMax(), getUnsignedMax()); 1015 return getNonEmpty(APInt::getNullValue(getBitWidth()), std::move(umin) + 1); 1016 } 1017 1018 ConstantRange 1019 ConstantRange::binaryOr(const ConstantRange &Other) const { 1020 if (isEmptySet() || Other.isEmptySet()) 1021 return getEmpty(); 1022 1023 // TODO: replace this with something less conservative 1024 1025 APInt umax = APIntOps::umax(getUnsignedMin(), Other.getUnsignedMin()); 1026 return getNonEmpty(std::move(umax), APInt::getNullValue(getBitWidth())); 1027 } 1028 1029 ConstantRange 1030 ConstantRange::shl(const ConstantRange &Other) const { 1031 if (isEmptySet() || Other.isEmptySet()) 1032 return getEmpty(); 1033 1034 APInt max = getUnsignedMax(); 1035 APInt Other_umax = Other.getUnsignedMax(); 1036 1037 // If we are shifting by maximum amount of 1038 // zero return return the original range. 1039 if (Other_umax.isNullValue()) 1040 return *this; 1041 // there's overflow! 1042 if (Other_umax.ugt(max.countLeadingZeros())) 1043 return getFull(); 1044 1045 // FIXME: implement the other tricky cases 1046 1047 APInt min = getUnsignedMin(); 1048 min <<= Other.getUnsignedMin(); 1049 max <<= Other_umax; 1050 1051 return ConstantRange(std::move(min), std::move(max) + 1); 1052 } 1053 1054 ConstantRange 1055 ConstantRange::lshr(const ConstantRange &Other) const { 1056 if (isEmptySet() || Other.isEmptySet()) 1057 return getEmpty(); 1058 1059 APInt max = getUnsignedMax().lshr(Other.getUnsignedMin()) + 1; 1060 APInt min = getUnsignedMin().lshr(Other.getUnsignedMax()); 1061 return getNonEmpty(std::move(min), std::move(max)); 1062 } 1063 1064 ConstantRange 1065 ConstantRange::ashr(const ConstantRange &Other) const { 1066 if (isEmptySet() || Other.isEmptySet()) 1067 return getEmpty(); 1068 1069 // May straddle zero, so handle both positive and negative cases. 1070 // 'PosMax' is the upper bound of the result of the ashr 1071 // operation, when Upper of the LHS of ashr is a non-negative. 1072 // number. Since ashr of a non-negative number will result in a 1073 // smaller number, the Upper value of LHS is shifted right with 1074 // the minimum value of 'Other' instead of the maximum value. 1075 APInt PosMax = getSignedMax().ashr(Other.getUnsignedMin()) + 1; 1076 1077 // 'PosMin' is the lower bound of the result of the ashr 1078 // operation, when Lower of the LHS is a non-negative number. 1079 // Since ashr of a non-negative number will result in a smaller 1080 // number, the Lower value of LHS is shifted right with the 1081 // maximum value of 'Other'. 1082 APInt PosMin = getSignedMin().ashr(Other.getUnsignedMax()); 1083 1084 // 'NegMax' is the upper bound of the result of the ashr 1085 // operation, when Upper of the LHS of ashr is a negative number. 1086 // Since 'ashr' of a negative number will result in a bigger 1087 // number, the Upper value of LHS is shifted right with the 1088 // maximum value of 'Other'. 1089 APInt NegMax = getSignedMax().ashr(Other.getUnsignedMax()) + 1; 1090 1091 // 'NegMin' is the lower bound of the result of the ashr 1092 // operation, when Lower of the LHS of ashr is a negative number. 1093 // Since 'ashr' of a negative number will result in a bigger 1094 // number, the Lower value of LHS is shifted right with the 1095 // minimum value of 'Other'. 1096 APInt NegMin = getSignedMin().ashr(Other.getUnsignedMin()); 1097 1098 APInt max, min; 1099 if (getSignedMin().isNonNegative()) { 1100 // Upper and Lower of LHS are non-negative. 1101 min = PosMin; 1102 max = PosMax; 1103 } else if (getSignedMax().isNegative()) { 1104 // Upper and Lower of LHS are negative. 1105 min = NegMin; 1106 max = NegMax; 1107 } else { 1108 // Upper is non-negative and Lower is negative. 1109 min = NegMin; 1110 max = PosMax; 1111 } 1112 return getNonEmpty(std::move(min), std::move(max)); 1113 } 1114 1115 ConstantRange ConstantRange::uadd_sat(const ConstantRange &Other) const { 1116 if (isEmptySet() || Other.isEmptySet()) 1117 return getEmpty(); 1118 1119 APInt NewL = getUnsignedMin().uadd_sat(Other.getUnsignedMin()); 1120 APInt NewU = getUnsignedMax().uadd_sat(Other.getUnsignedMax()) + 1; 1121 return getNonEmpty(std::move(NewL), std::move(NewU)); 1122 } 1123 1124 ConstantRange ConstantRange::sadd_sat(const ConstantRange &Other) const { 1125 if (isEmptySet() || Other.isEmptySet()) 1126 return getEmpty(); 1127 1128 APInt NewL = getSignedMin().sadd_sat(Other.getSignedMin()); 1129 APInt NewU = getSignedMax().sadd_sat(Other.getSignedMax()) + 1; 1130 return getNonEmpty(std::move(NewL), std::move(NewU)); 1131 } 1132 1133 ConstantRange ConstantRange::usub_sat(const ConstantRange &Other) const { 1134 if (isEmptySet() || Other.isEmptySet()) 1135 return getEmpty(); 1136 1137 APInt NewL = getUnsignedMin().usub_sat(Other.getUnsignedMax()); 1138 APInt NewU = getUnsignedMax().usub_sat(Other.getUnsignedMin()) + 1; 1139 return getNonEmpty(std::move(NewL), std::move(NewU)); 1140 } 1141 1142 ConstantRange ConstantRange::ssub_sat(const ConstantRange &Other) const { 1143 if (isEmptySet() || Other.isEmptySet()) 1144 return getEmpty(); 1145 1146 APInt NewL = getSignedMin().ssub_sat(Other.getSignedMax()); 1147 APInt NewU = getSignedMax().ssub_sat(Other.getSignedMin()) + 1; 1148 return getNonEmpty(std::move(NewL), std::move(NewU)); 1149 } 1150 1151 ConstantRange ConstantRange::inverse() const { 1152 if (isFullSet()) 1153 return getEmpty(); 1154 if (isEmptySet()) 1155 return getFull(); 1156 return ConstantRange(Upper, Lower); 1157 } 1158 1159 ConstantRange::OverflowResult ConstantRange::unsignedAddMayOverflow( 1160 const ConstantRange &Other) const { 1161 if (isEmptySet() || Other.isEmptySet()) 1162 return OverflowResult::MayOverflow; 1163 1164 APInt Min = getUnsignedMin(), Max = getUnsignedMax(); 1165 APInt OtherMin = Other.getUnsignedMin(), OtherMax = Other.getUnsignedMax(); 1166 1167 // a u+ b overflows iff a u> ~b. 1168 if (Min.ugt(~OtherMin)) 1169 return OverflowResult::AlwaysOverflows; 1170 if (Max.ugt(~OtherMax)) 1171 return OverflowResult::MayOverflow; 1172 return OverflowResult::NeverOverflows; 1173 } 1174 1175 ConstantRange::OverflowResult ConstantRange::signedAddMayOverflow( 1176 const ConstantRange &Other) const { 1177 if (isEmptySet() || Other.isEmptySet()) 1178 return OverflowResult::MayOverflow; 1179 1180 APInt Min = getSignedMin(), Max = getSignedMax(); 1181 APInt OtherMin = Other.getSignedMin(), OtherMax = Other.getSignedMax(); 1182 1183 APInt SignedMin = APInt::getSignedMinValue(getBitWidth()); 1184 APInt SignedMax = APInt::getSignedMaxValue(getBitWidth()); 1185 1186 // a s+ b overflows high iff a s>=0 && b s>= 0 && a s> smax - b. 1187 // a s+ b overflows low iff a s< 0 && b s< 0 && a s< smin - b. 1188 if (Min.isNonNegative() && OtherMin.isNonNegative() && 1189 Min.sgt(SignedMax - OtherMin)) 1190 return OverflowResult::AlwaysOverflows; 1191 if (Max.isNegative() && OtherMax.isNegative() && 1192 Max.slt(SignedMin - OtherMax)) 1193 return OverflowResult::AlwaysOverflows; 1194 1195 if (Max.isNonNegative() && OtherMax.isNonNegative() && 1196 Max.sgt(SignedMax - OtherMax)) 1197 return OverflowResult::MayOverflow; 1198 if (Min.isNegative() && OtherMin.isNegative() && 1199 Min.slt(SignedMin - OtherMin)) 1200 return OverflowResult::MayOverflow; 1201 1202 return OverflowResult::NeverOverflows; 1203 } 1204 1205 ConstantRange::OverflowResult ConstantRange::unsignedSubMayOverflow( 1206 const ConstantRange &Other) const { 1207 if (isEmptySet() || Other.isEmptySet()) 1208 return OverflowResult::MayOverflow; 1209 1210 APInt Min = getUnsignedMin(), Max = getUnsignedMax(); 1211 APInt OtherMin = Other.getUnsignedMin(), OtherMax = Other.getUnsignedMax(); 1212 1213 // a u- b overflows iff a u< b. 1214 if (Max.ult(OtherMin)) 1215 return OverflowResult::AlwaysOverflows; 1216 if (Min.ult(OtherMax)) 1217 return OverflowResult::MayOverflow; 1218 return OverflowResult::NeverOverflows; 1219 } 1220 1221 ConstantRange::OverflowResult ConstantRange::signedSubMayOverflow( 1222 const ConstantRange &Other) const { 1223 if (isEmptySet() || Other.isEmptySet()) 1224 return OverflowResult::MayOverflow; 1225 1226 APInt Min = getSignedMin(), Max = getSignedMax(); 1227 APInt OtherMin = Other.getSignedMin(), OtherMax = Other.getSignedMax(); 1228 1229 APInt SignedMin = APInt::getSignedMinValue(getBitWidth()); 1230 APInt SignedMax = APInt::getSignedMaxValue(getBitWidth()); 1231 1232 // a s- b overflows high iff a s>=0 && b s< 0 && a s> smax + b. 1233 // a s- b overflows low iff a s< 0 && b s>= 0 && a s< smin + b. 1234 if (Min.isNonNegative() && OtherMax.isNegative() && 1235 Min.sgt(SignedMax + OtherMax)) 1236 return OverflowResult::AlwaysOverflows; 1237 if (Max.isNegative() && OtherMin.isNonNegative() && 1238 Max.slt(SignedMin + OtherMin)) 1239 return OverflowResult::AlwaysOverflows; 1240 1241 if (Max.isNonNegative() && OtherMin.isNegative() && 1242 Max.sgt(SignedMax + OtherMin)) 1243 return OverflowResult::MayOverflow; 1244 if (Min.isNegative() && OtherMax.isNonNegative() && 1245 Min.slt(SignedMin + OtherMax)) 1246 return OverflowResult::MayOverflow; 1247 1248 return OverflowResult::NeverOverflows; 1249 } 1250 1251 ConstantRange::OverflowResult ConstantRange::unsignedMulMayOverflow( 1252 const ConstantRange &Other) const { 1253 if (isEmptySet() || Other.isEmptySet()) 1254 return OverflowResult::MayOverflow; 1255 1256 APInt Min = getUnsignedMin(), Max = getUnsignedMax(); 1257 APInt OtherMin = Other.getUnsignedMin(), OtherMax = Other.getUnsignedMax(); 1258 bool Overflow; 1259 1260 (void) Min.umul_ov(OtherMin, Overflow); 1261 if (Overflow) 1262 return OverflowResult::AlwaysOverflows; 1263 1264 (void) Max.umul_ov(OtherMax, Overflow); 1265 if (Overflow) 1266 return OverflowResult::MayOverflow; 1267 1268 return OverflowResult::NeverOverflows; 1269 } 1270 1271 void ConstantRange::print(raw_ostream &OS) const { 1272 if (isFullSet()) 1273 OS << "full-set"; 1274 else if (isEmptySet()) 1275 OS << "empty-set"; 1276 else 1277 OS << "[" << Lower << "," << Upper << ")"; 1278 } 1279 1280 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP) 1281 LLVM_DUMP_METHOD void ConstantRange::dump() const { 1282 print(dbgs()); 1283 } 1284 #endif 1285 1286 ConstantRange llvm::getConstantRangeFromMetadata(const MDNode &Ranges) { 1287 const unsigned NumRanges = Ranges.getNumOperands() / 2; 1288 assert(NumRanges >= 1 && "Must have at least one range!"); 1289 assert(Ranges.getNumOperands() % 2 == 0 && "Must be a sequence of pairs"); 1290 1291 auto *FirstLow = mdconst::extract<ConstantInt>(Ranges.getOperand(0)); 1292 auto *FirstHigh = mdconst::extract<ConstantInt>(Ranges.getOperand(1)); 1293 1294 ConstantRange CR(FirstLow->getValue(), FirstHigh->getValue()); 1295 1296 for (unsigned i = 1; i < NumRanges; ++i) { 1297 auto *Low = mdconst::extract<ConstantInt>(Ranges.getOperand(2 * i + 0)); 1298 auto *High = mdconst::extract<ConstantInt>(Ranges.getOperand(2 * i + 1)); 1299 1300 // Note: unionWith will potentially create a range that contains values not 1301 // contained in any of the original N ranges. 1302 CR = CR.unionWith(ConstantRange(Low->getValue(), High->getValue())); 1303 } 1304 1305 return CR; 1306 } 1307