1 //===- ConstantFold.cpp - LLVM constant folder ----------------------------===// 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 // This file implements folding of constants for LLVM. This implements the 10 // (internal) ConstantFold.h interface, which is used by the 11 // ConstantExpr::get* methods to automatically fold constants when possible. 12 // 13 // The current constant folding implementation is implemented in two pieces: the 14 // pieces that don't need DataLayout, and the pieces that do. This is to avoid 15 // a dependence in IR on Target. 16 // 17 //===----------------------------------------------------------------------===// 18 19 #include "llvm/IR/ConstantFold.h" 20 #include "llvm/ADT/APSInt.h" 21 #include "llvm/ADT/SmallVector.h" 22 #include "llvm/IR/Constants.h" 23 #include "llvm/IR/DerivedTypes.h" 24 #include "llvm/IR/Function.h" 25 #include "llvm/IR/GetElementPtrTypeIterator.h" 26 #include "llvm/IR/GlobalAlias.h" 27 #include "llvm/IR/GlobalVariable.h" 28 #include "llvm/IR/Instructions.h" 29 #include "llvm/IR/Module.h" 30 #include "llvm/IR/Operator.h" 31 #include "llvm/IR/PatternMatch.h" 32 #include "llvm/Support/ErrorHandling.h" 33 using namespace llvm; 34 using namespace llvm::PatternMatch; 35 36 //===----------------------------------------------------------------------===// 37 // ConstantFold*Instruction Implementations 38 //===----------------------------------------------------------------------===// 39 40 /// This function determines which opcode to use to fold two constant cast 41 /// expressions together. It uses CastInst::isEliminableCastPair to determine 42 /// the opcode. Consequently its just a wrapper around that function. 43 /// Determine if it is valid to fold a cast of a cast 44 static unsigned 45 foldConstantCastPair( 46 unsigned opc, ///< opcode of the second cast constant expression 47 ConstantExpr *Op, ///< the first cast constant expression 48 Type *DstTy ///< destination type of the first cast 49 ) { 50 assert(Op && Op->isCast() && "Can't fold cast of cast without a cast!"); 51 assert(DstTy && DstTy->isFirstClassType() && "Invalid cast destination type"); 52 assert(CastInst::isCast(opc) && "Invalid cast opcode"); 53 54 // The types and opcodes for the two Cast constant expressions 55 Type *SrcTy = Op->getOperand(0)->getType(); 56 Type *MidTy = Op->getType(); 57 Instruction::CastOps firstOp = Instruction::CastOps(Op->getOpcode()); 58 Instruction::CastOps secondOp = Instruction::CastOps(opc); 59 60 // Assume that pointers are never more than 64 bits wide, and only use this 61 // for the middle type. Otherwise we could end up folding away illegal 62 // bitcasts between address spaces with different sizes. 63 IntegerType *FakeIntPtrTy = Type::getInt64Ty(DstTy->getContext()); 64 65 // Let CastInst::isEliminableCastPair do the heavy lifting. 66 return CastInst::isEliminableCastPair(firstOp, secondOp, SrcTy, MidTy, DstTy, 67 nullptr, FakeIntPtrTy, nullptr); 68 } 69 70 static Constant *FoldBitCast(Constant *V, Type *DestTy) { 71 Type *SrcTy = V->getType(); 72 if (SrcTy == DestTy) 73 return V; // no-op cast 74 75 // Handle casts from one vector constant to another. We know that the src 76 // and dest type have the same size (otherwise its an illegal cast). 77 if (VectorType *DestPTy = dyn_cast<VectorType>(DestTy)) { 78 if (V->isAllOnesValue()) 79 return Constant::getAllOnesValue(DestTy); 80 81 // Canonicalize scalar-to-vector bitcasts into vector-to-vector bitcasts 82 // This allows for other simplifications (although some of them 83 // can only be handled by Analysis/ConstantFolding.cpp). 84 if (isa<ConstantInt>(V) || isa<ConstantFP>(V)) 85 return ConstantExpr::getBitCast(ConstantVector::get(V), DestPTy); 86 return nullptr; 87 } 88 89 // Handle integral constant input. 90 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 91 // See note below regarding the PPC_FP128 restriction. 92 if (DestTy->isFloatingPointTy() && !DestTy->isPPC_FP128Ty()) 93 return ConstantFP::get(DestTy->getContext(), 94 APFloat(DestTy->getFltSemantics(), 95 CI->getValue())); 96 97 // Otherwise, can't fold this (vector?) 98 return nullptr; 99 } 100 101 // Handle ConstantFP input: FP -> Integral. 102 if (ConstantFP *FP = dyn_cast<ConstantFP>(V)) { 103 // PPC_FP128 is really the sum of two consecutive doubles, where the first 104 // double is always stored first in memory, regardless of the target 105 // endianness. The memory layout of i128, however, depends on the target 106 // endianness, and so we can't fold this without target endianness 107 // information. This should instead be handled by 108 // Analysis/ConstantFolding.cpp 109 if (FP->getType()->isPPC_FP128Ty()) 110 return nullptr; 111 112 // Make sure dest type is compatible with the folded integer constant. 113 if (!DestTy->isIntegerTy()) 114 return nullptr; 115 116 return ConstantInt::get(FP->getContext(), 117 FP->getValueAPF().bitcastToAPInt()); 118 } 119 120 return nullptr; 121 } 122 123 124 /// V is an integer constant which only has a subset of its bytes used. 125 /// The bytes used are indicated by ByteStart (which is the first byte used, 126 /// counting from the least significant byte) and ByteSize, which is the number 127 /// of bytes used. 128 /// 129 /// This function analyzes the specified constant to see if the specified byte 130 /// range can be returned as a simplified constant. If so, the constant is 131 /// returned, otherwise null is returned. 132 static Constant *ExtractConstantBytes(Constant *C, unsigned ByteStart, 133 unsigned ByteSize) { 134 assert(C->getType()->isIntegerTy() && 135 (cast<IntegerType>(C->getType())->getBitWidth() & 7) == 0 && 136 "Non-byte sized integer input"); 137 [[maybe_unused]] unsigned CSize = cast<IntegerType>(C->getType())->getBitWidth()/8; 138 assert(ByteSize && "Must be accessing some piece"); 139 assert(ByteStart+ByteSize <= CSize && "Extracting invalid piece from input"); 140 assert(ByteSize != CSize && "Should not extract everything"); 141 142 // Constant Integers are simple. 143 if (ConstantInt *CI = dyn_cast<ConstantInt>(C)) { 144 APInt V = CI->getValue(); 145 if (ByteStart) 146 V.lshrInPlace(ByteStart*8); 147 V = V.trunc(ByteSize*8); 148 return ConstantInt::get(CI->getContext(), V); 149 } 150 151 // In the input is a constant expr, we might be able to recursively simplify. 152 // If not, we definitely can't do anything. 153 ConstantExpr *CE = dyn_cast<ConstantExpr>(C); 154 if (!CE) return nullptr; 155 156 switch (CE->getOpcode()) { 157 default: return nullptr; 158 case Instruction::Shl: { 159 ConstantInt *Amt = dyn_cast<ConstantInt>(CE->getOperand(1)); 160 if (!Amt) 161 return nullptr; 162 APInt ShAmt = Amt->getValue(); 163 // Cannot analyze non-byte shifts. 164 if ((ShAmt & 7) != 0) 165 return nullptr; 166 ShAmt.lshrInPlace(3); 167 168 // If the extract is known to be all zeros, return zero. 169 if (ShAmt.uge(ByteStart + ByteSize)) 170 return Constant::getNullValue( 171 IntegerType::get(CE->getContext(), ByteSize * 8)); 172 // If the extract is known to be fully in the input, extract it. 173 if (ShAmt.ule(ByteStart)) 174 return ExtractConstantBytes(CE->getOperand(0), 175 ByteStart - ShAmt.getZExtValue(), ByteSize); 176 177 // TODO: Handle the 'partially zero' case. 178 return nullptr; 179 } 180 } 181 } 182 183 static Constant *foldMaybeUndesirableCast(unsigned opc, Constant *V, 184 Type *DestTy) { 185 return ConstantExpr::isDesirableCastOp(opc) 186 ? ConstantExpr::getCast(opc, V, DestTy) 187 : ConstantFoldCastInstruction(opc, V, DestTy); 188 } 189 190 Constant *llvm::ConstantFoldCastInstruction(unsigned opc, Constant *V, 191 Type *DestTy) { 192 if (isa<PoisonValue>(V)) 193 return PoisonValue::get(DestTy); 194 195 if (isa<UndefValue>(V)) { 196 // zext(undef) = 0, because the top bits will be zero. 197 // sext(undef) = 0, because the top bits will all be the same. 198 // [us]itofp(undef) = 0, because the result value is bounded. 199 if (opc == Instruction::ZExt || opc == Instruction::SExt || 200 opc == Instruction::UIToFP || opc == Instruction::SIToFP) 201 return Constant::getNullValue(DestTy); 202 return UndefValue::get(DestTy); 203 } 204 205 if (V->isNullValue() && !DestTy->isX86_MMXTy() && !DestTy->isX86_AMXTy() && 206 opc != Instruction::AddrSpaceCast) 207 return Constant::getNullValue(DestTy); 208 209 // If the cast operand is a constant expression, there's a few things we can 210 // do to try to simplify it. 211 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) { 212 if (CE->isCast()) { 213 // Try hard to fold cast of cast because they are often eliminable. 214 if (unsigned newOpc = foldConstantCastPair(opc, CE, DestTy)) 215 return foldMaybeUndesirableCast(newOpc, CE->getOperand(0), DestTy); 216 } 217 } 218 219 // If the cast operand is a constant vector, perform the cast by 220 // operating on each element. In the cast of bitcasts, the element 221 // count may be mismatched; don't attempt to handle that here. 222 if ((isa<ConstantVector>(V) || isa<ConstantDataVector>(V)) && 223 DestTy->isVectorTy() && 224 cast<FixedVectorType>(DestTy)->getNumElements() == 225 cast<FixedVectorType>(V->getType())->getNumElements()) { 226 VectorType *DestVecTy = cast<VectorType>(DestTy); 227 Type *DstEltTy = DestVecTy->getElementType(); 228 // Fast path for splatted constants. 229 if (Constant *Splat = V->getSplatValue()) { 230 Constant *Res = foldMaybeUndesirableCast(opc, Splat, DstEltTy); 231 if (!Res) 232 return nullptr; 233 return ConstantVector::getSplat( 234 cast<VectorType>(DestTy)->getElementCount(), Res); 235 } 236 SmallVector<Constant *, 16> res; 237 Type *Ty = IntegerType::get(V->getContext(), 32); 238 for (unsigned i = 0, 239 e = cast<FixedVectorType>(V->getType())->getNumElements(); 240 i != e; ++i) { 241 Constant *C = ConstantExpr::getExtractElement(V, ConstantInt::get(Ty, i)); 242 Constant *Casted = foldMaybeUndesirableCast(opc, C, DstEltTy); 243 if (!Casted) 244 return nullptr; 245 res.push_back(Casted); 246 } 247 return ConstantVector::get(res); 248 } 249 250 // We actually have to do a cast now. Perform the cast according to the 251 // opcode specified. 252 switch (opc) { 253 default: 254 llvm_unreachable("Failed to cast constant expression"); 255 case Instruction::FPTrunc: 256 case Instruction::FPExt: 257 if (ConstantFP *FPC = dyn_cast<ConstantFP>(V)) { 258 bool ignored; 259 APFloat Val = FPC->getValueAPF(); 260 Val.convert(DestTy->getFltSemantics(), APFloat::rmNearestTiesToEven, 261 &ignored); 262 return ConstantFP::get(V->getContext(), Val); 263 } 264 return nullptr; // Can't fold. 265 case Instruction::FPToUI: 266 case Instruction::FPToSI: 267 if (ConstantFP *FPC = dyn_cast<ConstantFP>(V)) { 268 const APFloat &V = FPC->getValueAPF(); 269 bool ignored; 270 uint32_t DestBitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 271 APSInt IntVal(DestBitWidth, opc == Instruction::FPToUI); 272 if (APFloat::opInvalidOp == 273 V.convertToInteger(IntVal, APFloat::rmTowardZero, &ignored)) { 274 // Undefined behavior invoked - the destination type can't represent 275 // the input constant. 276 return PoisonValue::get(DestTy); 277 } 278 return ConstantInt::get(FPC->getContext(), IntVal); 279 } 280 return nullptr; // Can't fold. 281 case Instruction::UIToFP: 282 case Instruction::SIToFP: 283 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 284 const APInt &api = CI->getValue(); 285 APFloat apf(DestTy->getFltSemantics(), 286 APInt::getZero(DestTy->getPrimitiveSizeInBits())); 287 apf.convertFromAPInt(api, opc==Instruction::SIToFP, 288 APFloat::rmNearestTiesToEven); 289 return ConstantFP::get(V->getContext(), apf); 290 } 291 return nullptr; 292 case Instruction::ZExt: 293 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 294 uint32_t BitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 295 return ConstantInt::get(V->getContext(), 296 CI->getValue().zext(BitWidth)); 297 } 298 return nullptr; 299 case Instruction::SExt: 300 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 301 uint32_t BitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 302 return ConstantInt::get(V->getContext(), 303 CI->getValue().sext(BitWidth)); 304 } 305 return nullptr; 306 case Instruction::Trunc: { 307 if (V->getType()->isVectorTy()) 308 return nullptr; 309 310 uint32_t DestBitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 311 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 312 return ConstantInt::get(V->getContext(), 313 CI->getValue().trunc(DestBitWidth)); 314 } 315 316 // The input must be a constantexpr. See if we can simplify this based on 317 // the bytes we are demanding. Only do this if the source and dest are an 318 // even multiple of a byte. 319 if ((DestBitWidth & 7) == 0 && 320 (cast<IntegerType>(V->getType())->getBitWidth() & 7) == 0) 321 if (Constant *Res = ExtractConstantBytes(V, 0, DestBitWidth / 8)) 322 return Res; 323 324 return nullptr; 325 } 326 case Instruction::BitCast: 327 return FoldBitCast(V, DestTy); 328 case Instruction::AddrSpaceCast: 329 case Instruction::IntToPtr: 330 case Instruction::PtrToInt: 331 return nullptr; 332 } 333 } 334 335 Constant *llvm::ConstantFoldSelectInstruction(Constant *Cond, 336 Constant *V1, Constant *V2) { 337 // Check for i1 and vector true/false conditions. 338 if (Cond->isNullValue()) return V2; 339 if (Cond->isAllOnesValue()) return V1; 340 341 // If the condition is a vector constant, fold the result elementwise. 342 if (ConstantVector *CondV = dyn_cast<ConstantVector>(Cond)) { 343 auto *V1VTy = CondV->getType(); 344 SmallVector<Constant*, 16> Result; 345 Type *Ty = IntegerType::get(CondV->getContext(), 32); 346 for (unsigned i = 0, e = V1VTy->getNumElements(); i != e; ++i) { 347 Constant *V; 348 Constant *V1Element = ConstantExpr::getExtractElement(V1, 349 ConstantInt::get(Ty, i)); 350 Constant *V2Element = ConstantExpr::getExtractElement(V2, 351 ConstantInt::get(Ty, i)); 352 auto *Cond = cast<Constant>(CondV->getOperand(i)); 353 if (isa<PoisonValue>(Cond)) { 354 V = PoisonValue::get(V1Element->getType()); 355 } else if (V1Element == V2Element) { 356 V = V1Element; 357 } else if (isa<UndefValue>(Cond)) { 358 V = isa<UndefValue>(V1Element) ? V1Element : V2Element; 359 } else { 360 if (!isa<ConstantInt>(Cond)) break; 361 V = Cond->isNullValue() ? V2Element : V1Element; 362 } 363 Result.push_back(V); 364 } 365 366 // If we were able to build the vector, return it. 367 if (Result.size() == V1VTy->getNumElements()) 368 return ConstantVector::get(Result); 369 } 370 371 if (isa<PoisonValue>(Cond)) 372 return PoisonValue::get(V1->getType()); 373 374 if (isa<UndefValue>(Cond)) { 375 if (isa<UndefValue>(V1)) return V1; 376 return V2; 377 } 378 379 if (V1 == V2) return V1; 380 381 if (isa<PoisonValue>(V1)) 382 return V2; 383 if (isa<PoisonValue>(V2)) 384 return V1; 385 386 // If the true or false value is undef, we can fold to the other value as 387 // long as the other value isn't poison. 388 auto NotPoison = [](Constant *C) { 389 if (isa<PoisonValue>(C)) 390 return false; 391 392 // TODO: We can analyze ConstExpr by opcode to determine if there is any 393 // possibility of poison. 394 if (isa<ConstantExpr>(C)) 395 return false; 396 397 if (isa<ConstantInt>(C) || isa<GlobalVariable>(C) || isa<ConstantFP>(C) || 398 isa<ConstantPointerNull>(C) || isa<Function>(C)) 399 return true; 400 401 if (C->getType()->isVectorTy()) 402 return !C->containsPoisonElement() && !C->containsConstantExpression(); 403 404 // TODO: Recursively analyze aggregates or other constants. 405 return false; 406 }; 407 if (isa<UndefValue>(V1) && NotPoison(V2)) return V2; 408 if (isa<UndefValue>(V2) && NotPoison(V1)) return V1; 409 410 return nullptr; 411 } 412 413 Constant *llvm::ConstantFoldExtractElementInstruction(Constant *Val, 414 Constant *Idx) { 415 auto *ValVTy = cast<VectorType>(Val->getType()); 416 417 // extractelt poison, C -> poison 418 // extractelt C, undef -> poison 419 if (isa<PoisonValue>(Val) || isa<UndefValue>(Idx)) 420 return PoisonValue::get(ValVTy->getElementType()); 421 422 // extractelt undef, C -> undef 423 if (isa<UndefValue>(Val)) 424 return UndefValue::get(ValVTy->getElementType()); 425 426 auto *CIdx = dyn_cast<ConstantInt>(Idx); 427 if (!CIdx) 428 return nullptr; 429 430 if (auto *ValFVTy = dyn_cast<FixedVectorType>(Val->getType())) { 431 // ee({w,x,y,z}, wrong_value) -> poison 432 if (CIdx->uge(ValFVTy->getNumElements())) 433 return PoisonValue::get(ValFVTy->getElementType()); 434 } 435 436 // ee (gep (ptr, idx0, ...), idx) -> gep (ee (ptr, idx), ee (idx0, idx), ...) 437 if (auto *CE = dyn_cast<ConstantExpr>(Val)) { 438 if (auto *GEP = dyn_cast<GEPOperator>(CE)) { 439 SmallVector<Constant *, 8> Ops; 440 Ops.reserve(CE->getNumOperands()); 441 for (unsigned i = 0, e = CE->getNumOperands(); i != e; ++i) { 442 Constant *Op = CE->getOperand(i); 443 if (Op->getType()->isVectorTy()) { 444 Constant *ScalarOp = ConstantExpr::getExtractElement(Op, Idx); 445 if (!ScalarOp) 446 return nullptr; 447 Ops.push_back(ScalarOp); 448 } else 449 Ops.push_back(Op); 450 } 451 return CE->getWithOperands(Ops, ValVTy->getElementType(), false, 452 GEP->getSourceElementType()); 453 } else if (CE->getOpcode() == Instruction::InsertElement) { 454 if (const auto *IEIdx = dyn_cast<ConstantInt>(CE->getOperand(2))) { 455 if (APSInt::isSameValue(APSInt(IEIdx->getValue()), 456 APSInt(CIdx->getValue()))) { 457 return CE->getOperand(1); 458 } else { 459 return ConstantExpr::getExtractElement(CE->getOperand(0), CIdx); 460 } 461 } 462 } 463 } 464 465 if (Constant *C = Val->getAggregateElement(CIdx)) 466 return C; 467 468 // Lane < Splat minimum vector width => extractelt Splat(x), Lane -> x 469 if (CIdx->getValue().ult(ValVTy->getElementCount().getKnownMinValue())) { 470 if (Constant *SplatVal = Val->getSplatValue()) 471 return SplatVal; 472 } 473 474 return nullptr; 475 } 476 477 Constant *llvm::ConstantFoldInsertElementInstruction(Constant *Val, 478 Constant *Elt, 479 Constant *Idx) { 480 if (isa<UndefValue>(Idx)) 481 return PoisonValue::get(Val->getType()); 482 483 // Inserting null into all zeros is still all zeros. 484 // TODO: This is true for undef and poison splats too. 485 if (isa<ConstantAggregateZero>(Val) && Elt->isNullValue()) 486 return Val; 487 488 ConstantInt *CIdx = dyn_cast<ConstantInt>(Idx); 489 if (!CIdx) return nullptr; 490 491 // Do not iterate on scalable vector. The num of elements is unknown at 492 // compile-time. 493 if (isa<ScalableVectorType>(Val->getType())) 494 return nullptr; 495 496 auto *ValTy = cast<FixedVectorType>(Val->getType()); 497 498 unsigned NumElts = ValTy->getNumElements(); 499 if (CIdx->uge(NumElts)) 500 return PoisonValue::get(Val->getType()); 501 502 SmallVector<Constant*, 16> Result; 503 Result.reserve(NumElts); 504 auto *Ty = Type::getInt32Ty(Val->getContext()); 505 uint64_t IdxVal = CIdx->getZExtValue(); 506 for (unsigned i = 0; i != NumElts; ++i) { 507 if (i == IdxVal) { 508 Result.push_back(Elt); 509 continue; 510 } 511 512 Constant *C = ConstantExpr::getExtractElement(Val, ConstantInt::get(Ty, i)); 513 Result.push_back(C); 514 } 515 516 return ConstantVector::get(Result); 517 } 518 519 Constant *llvm::ConstantFoldShuffleVectorInstruction(Constant *V1, Constant *V2, 520 ArrayRef<int> Mask) { 521 auto *V1VTy = cast<VectorType>(V1->getType()); 522 unsigned MaskNumElts = Mask.size(); 523 auto MaskEltCount = 524 ElementCount::get(MaskNumElts, isa<ScalableVectorType>(V1VTy)); 525 Type *EltTy = V1VTy->getElementType(); 526 527 // Poison shuffle mask -> poison value. 528 if (all_of(Mask, [](int Elt) { return Elt == PoisonMaskElem; })) { 529 return PoisonValue::get(VectorType::get(EltTy, MaskEltCount)); 530 } 531 532 // If the mask is all zeros this is a splat, no need to go through all 533 // elements. 534 if (all_of(Mask, [](int Elt) { return Elt == 0; })) { 535 Type *Ty = IntegerType::get(V1->getContext(), 32); 536 Constant *Elt = 537 ConstantExpr::getExtractElement(V1, ConstantInt::get(Ty, 0)); 538 539 if (Elt->isNullValue()) { 540 auto *VTy = VectorType::get(EltTy, MaskEltCount); 541 return ConstantAggregateZero::get(VTy); 542 } else if (!MaskEltCount.isScalable()) 543 return ConstantVector::getSplat(MaskEltCount, Elt); 544 } 545 546 // Do not iterate on scalable vector. The num of elements is unknown at 547 // compile-time. 548 if (isa<ScalableVectorType>(V1VTy)) 549 return nullptr; 550 551 unsigned SrcNumElts = V1VTy->getElementCount().getKnownMinValue(); 552 553 // Loop over the shuffle mask, evaluating each element. 554 SmallVector<Constant*, 32> Result; 555 for (unsigned i = 0; i != MaskNumElts; ++i) { 556 int Elt = Mask[i]; 557 if (Elt == -1) { 558 Result.push_back(UndefValue::get(EltTy)); 559 continue; 560 } 561 Constant *InElt; 562 if (unsigned(Elt) >= SrcNumElts*2) 563 InElt = UndefValue::get(EltTy); 564 else if (unsigned(Elt) >= SrcNumElts) { 565 Type *Ty = IntegerType::get(V2->getContext(), 32); 566 InElt = 567 ConstantExpr::getExtractElement(V2, 568 ConstantInt::get(Ty, Elt - SrcNumElts)); 569 } else { 570 Type *Ty = IntegerType::get(V1->getContext(), 32); 571 InElt = ConstantExpr::getExtractElement(V1, ConstantInt::get(Ty, Elt)); 572 } 573 Result.push_back(InElt); 574 } 575 576 return ConstantVector::get(Result); 577 } 578 579 Constant *llvm::ConstantFoldExtractValueInstruction(Constant *Agg, 580 ArrayRef<unsigned> Idxs) { 581 // Base case: no indices, so return the entire value. 582 if (Idxs.empty()) 583 return Agg; 584 585 if (Constant *C = Agg->getAggregateElement(Idxs[0])) 586 return ConstantFoldExtractValueInstruction(C, Idxs.slice(1)); 587 588 return nullptr; 589 } 590 591 Constant *llvm::ConstantFoldInsertValueInstruction(Constant *Agg, 592 Constant *Val, 593 ArrayRef<unsigned> Idxs) { 594 // Base case: no indices, so replace the entire value. 595 if (Idxs.empty()) 596 return Val; 597 598 unsigned NumElts; 599 if (StructType *ST = dyn_cast<StructType>(Agg->getType())) 600 NumElts = ST->getNumElements(); 601 else 602 NumElts = cast<ArrayType>(Agg->getType())->getNumElements(); 603 604 SmallVector<Constant*, 32> Result; 605 for (unsigned i = 0; i != NumElts; ++i) { 606 Constant *C = Agg->getAggregateElement(i); 607 if (!C) return nullptr; 608 609 if (Idxs[0] == i) 610 C = ConstantFoldInsertValueInstruction(C, Val, Idxs.slice(1)); 611 612 Result.push_back(C); 613 } 614 615 if (StructType *ST = dyn_cast<StructType>(Agg->getType())) 616 return ConstantStruct::get(ST, Result); 617 return ConstantArray::get(cast<ArrayType>(Agg->getType()), Result); 618 } 619 620 Constant *llvm::ConstantFoldUnaryInstruction(unsigned Opcode, Constant *C) { 621 assert(Instruction::isUnaryOp(Opcode) && "Non-unary instruction detected"); 622 623 // Handle scalar UndefValue and scalable vector UndefValue. Fixed-length 624 // vectors are always evaluated per element. 625 bool IsScalableVector = isa<ScalableVectorType>(C->getType()); 626 bool HasScalarUndefOrScalableVectorUndef = 627 (!C->getType()->isVectorTy() || IsScalableVector) && isa<UndefValue>(C); 628 629 if (HasScalarUndefOrScalableVectorUndef) { 630 switch (static_cast<Instruction::UnaryOps>(Opcode)) { 631 case Instruction::FNeg: 632 return C; // -undef -> undef 633 case Instruction::UnaryOpsEnd: 634 llvm_unreachable("Invalid UnaryOp"); 635 } 636 } 637 638 // Constant should not be UndefValue, unless these are vector constants. 639 assert(!HasScalarUndefOrScalableVectorUndef && "Unexpected UndefValue"); 640 // We only have FP UnaryOps right now. 641 assert(!isa<ConstantInt>(C) && "Unexpected Integer UnaryOp"); 642 643 if (ConstantFP *CFP = dyn_cast<ConstantFP>(C)) { 644 const APFloat &CV = CFP->getValueAPF(); 645 switch (Opcode) { 646 default: 647 break; 648 case Instruction::FNeg: 649 return ConstantFP::get(C->getContext(), neg(CV)); 650 } 651 } else if (auto *VTy = dyn_cast<FixedVectorType>(C->getType())) { 652 653 Type *Ty = IntegerType::get(VTy->getContext(), 32); 654 // Fast path for splatted constants. 655 if (Constant *Splat = C->getSplatValue()) 656 if (Constant *Elt = ConstantFoldUnaryInstruction(Opcode, Splat)) 657 return ConstantVector::getSplat(VTy->getElementCount(), Elt); 658 659 // Fold each element and create a vector constant from those constants. 660 SmallVector<Constant *, 16> Result; 661 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 662 Constant *ExtractIdx = ConstantInt::get(Ty, i); 663 Constant *Elt = ConstantExpr::getExtractElement(C, ExtractIdx); 664 Constant *Res = ConstantFoldUnaryInstruction(Opcode, Elt); 665 if (!Res) 666 return nullptr; 667 Result.push_back(Res); 668 } 669 670 return ConstantVector::get(Result); 671 } 672 673 // We don't know how to fold this. 674 return nullptr; 675 } 676 677 Constant *llvm::ConstantFoldBinaryInstruction(unsigned Opcode, Constant *C1, 678 Constant *C2) { 679 assert(Instruction::isBinaryOp(Opcode) && "Non-binary instruction detected"); 680 681 // Simplify BinOps with their identity values first. They are no-ops and we 682 // can always return the other value, including undef or poison values. 683 if (Constant *Identity = ConstantExpr::getBinOpIdentity( 684 Opcode, C1->getType(), /*AllowRHSIdentity*/ false)) { 685 if (C1 == Identity) 686 return C2; 687 if (C2 == Identity) 688 return C1; 689 } else if (Constant *Identity = ConstantExpr::getBinOpIdentity( 690 Opcode, C1->getType(), /*AllowRHSIdentity*/ true)) { 691 if (C2 == Identity) 692 return C1; 693 } 694 695 // Binary operations propagate poison. 696 if (isa<PoisonValue>(C1) || isa<PoisonValue>(C2)) 697 return PoisonValue::get(C1->getType()); 698 699 // Handle scalar UndefValue and scalable vector UndefValue. Fixed-length 700 // vectors are always evaluated per element. 701 bool IsScalableVector = isa<ScalableVectorType>(C1->getType()); 702 bool HasScalarUndefOrScalableVectorUndef = 703 (!C1->getType()->isVectorTy() || IsScalableVector) && 704 (isa<UndefValue>(C1) || isa<UndefValue>(C2)); 705 if (HasScalarUndefOrScalableVectorUndef) { 706 switch (static_cast<Instruction::BinaryOps>(Opcode)) { 707 case Instruction::Xor: 708 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) 709 // Handle undef ^ undef -> 0 special case. This is a common 710 // idiom (misuse). 711 return Constant::getNullValue(C1->getType()); 712 [[fallthrough]]; 713 case Instruction::Add: 714 case Instruction::Sub: 715 return UndefValue::get(C1->getType()); 716 case Instruction::And: 717 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) // undef & undef -> undef 718 return C1; 719 return Constant::getNullValue(C1->getType()); // undef & X -> 0 720 case Instruction::Mul: { 721 // undef * undef -> undef 722 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) 723 return C1; 724 const APInt *CV; 725 // X * undef -> undef if X is odd 726 if (match(C1, m_APInt(CV)) || match(C2, m_APInt(CV))) 727 if ((*CV)[0]) 728 return UndefValue::get(C1->getType()); 729 730 // X * undef -> 0 otherwise 731 return Constant::getNullValue(C1->getType()); 732 } 733 case Instruction::SDiv: 734 case Instruction::UDiv: 735 // X / undef -> poison 736 // X / 0 -> poison 737 if (match(C2, m_CombineOr(m_Undef(), m_Zero()))) 738 return PoisonValue::get(C2->getType()); 739 // undef / X -> 0 otherwise 740 return Constant::getNullValue(C1->getType()); 741 case Instruction::URem: 742 case Instruction::SRem: 743 // X % undef -> poison 744 // X % 0 -> poison 745 if (match(C2, m_CombineOr(m_Undef(), m_Zero()))) 746 return PoisonValue::get(C2->getType()); 747 // undef % X -> 0 otherwise 748 return Constant::getNullValue(C1->getType()); 749 case Instruction::Or: // X | undef -> -1 750 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) // undef | undef -> undef 751 return C1; 752 return Constant::getAllOnesValue(C1->getType()); // undef | X -> ~0 753 case Instruction::LShr: 754 // X >>l undef -> poison 755 if (isa<UndefValue>(C2)) 756 return PoisonValue::get(C2->getType()); 757 // undef >>l X -> 0 758 return Constant::getNullValue(C1->getType()); 759 case Instruction::AShr: 760 // X >>a undef -> poison 761 if (isa<UndefValue>(C2)) 762 return PoisonValue::get(C2->getType()); 763 // TODO: undef >>a X -> poison if the shift is exact 764 // undef >>a X -> 0 765 return Constant::getNullValue(C1->getType()); 766 case Instruction::Shl: 767 // X << undef -> undef 768 if (isa<UndefValue>(C2)) 769 return PoisonValue::get(C2->getType()); 770 // undef << X -> 0 771 return Constant::getNullValue(C1->getType()); 772 case Instruction::FSub: 773 // -0.0 - undef --> undef (consistent with "fneg undef") 774 if (match(C1, m_NegZeroFP()) && isa<UndefValue>(C2)) 775 return C2; 776 [[fallthrough]]; 777 case Instruction::FAdd: 778 case Instruction::FMul: 779 case Instruction::FDiv: 780 case Instruction::FRem: 781 // [any flop] undef, undef -> undef 782 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) 783 return C1; 784 // [any flop] C, undef -> NaN 785 // [any flop] undef, C -> NaN 786 // We could potentially specialize NaN/Inf constants vs. 'normal' 787 // constants (possibly differently depending on opcode and operand). This 788 // would allow returning undef sometimes. But it is always safe to fold to 789 // NaN because we can choose the undef operand as NaN, and any FP opcode 790 // with a NaN operand will propagate NaN. 791 return ConstantFP::getNaN(C1->getType()); 792 case Instruction::BinaryOpsEnd: 793 llvm_unreachable("Invalid BinaryOp"); 794 } 795 } 796 797 // Neither constant should be UndefValue, unless these are vector constants. 798 assert((!HasScalarUndefOrScalableVectorUndef) && "Unexpected UndefValue"); 799 800 // Handle simplifications when the RHS is a constant int. 801 if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) { 802 switch (Opcode) { 803 case Instruction::Mul: 804 if (CI2->isZero()) 805 return C2; // X * 0 == 0 806 break; 807 case Instruction::UDiv: 808 case Instruction::SDiv: 809 if (CI2->isZero()) 810 return PoisonValue::get(CI2->getType()); // X / 0 == poison 811 break; 812 case Instruction::URem: 813 case Instruction::SRem: 814 if (CI2->isOne()) 815 return Constant::getNullValue(CI2->getType()); // X % 1 == 0 816 if (CI2->isZero()) 817 return PoisonValue::get(CI2->getType()); // X % 0 == poison 818 break; 819 case Instruction::And: 820 if (CI2->isZero()) 821 return C2; // X & 0 == 0 822 823 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 824 // If and'ing the address of a global with a constant, fold it. 825 if (CE1->getOpcode() == Instruction::PtrToInt && 826 isa<GlobalValue>(CE1->getOperand(0))) { 827 GlobalValue *GV = cast<GlobalValue>(CE1->getOperand(0)); 828 829 Align GVAlign; // defaults to 1 830 831 if (Module *TheModule = GV->getParent()) { 832 const DataLayout &DL = TheModule->getDataLayout(); 833 GVAlign = GV->getPointerAlignment(DL); 834 835 // If the function alignment is not specified then assume that it 836 // is 4. 837 // This is dangerous; on x86, the alignment of the pointer 838 // corresponds to the alignment of the function, but might be less 839 // than 4 if it isn't explicitly specified. 840 // However, a fix for this behaviour was reverted because it 841 // increased code size (see https://reviews.llvm.org/D55115) 842 // FIXME: This code should be deleted once existing targets have 843 // appropriate defaults 844 if (isa<Function>(GV) && !DL.getFunctionPtrAlign()) 845 GVAlign = Align(4); 846 } else if (isa<GlobalVariable>(GV)) { 847 GVAlign = cast<GlobalVariable>(GV)->getAlign().valueOrOne(); 848 } 849 850 if (GVAlign > 1) { 851 unsigned DstWidth = CI2->getBitWidth(); 852 unsigned SrcWidth = std::min(DstWidth, Log2(GVAlign)); 853 APInt BitsNotSet(APInt::getLowBitsSet(DstWidth, SrcWidth)); 854 855 // If checking bits we know are clear, return zero. 856 if ((CI2->getValue() & BitsNotSet) == CI2->getValue()) 857 return Constant::getNullValue(CI2->getType()); 858 } 859 } 860 } 861 break; 862 case Instruction::Or: 863 if (CI2->isMinusOne()) 864 return C2; // X | -1 == -1 865 break; 866 case Instruction::Xor: 867 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 868 switch (CE1->getOpcode()) { 869 default: 870 break; 871 case Instruction::ICmp: 872 case Instruction::FCmp: 873 // cmp pred ^ true -> cmp !pred 874 assert(CI2->isOne()); 875 CmpInst::Predicate pred = (CmpInst::Predicate)CE1->getPredicate(); 876 pred = CmpInst::getInversePredicate(pred); 877 return ConstantExpr::getCompare(pred, CE1->getOperand(0), 878 CE1->getOperand(1)); 879 } 880 } 881 break; 882 } 883 } else if (isa<ConstantInt>(C1)) { 884 // If C1 is a ConstantInt and C2 is not, swap the operands. 885 if (Instruction::isCommutative(Opcode)) 886 return ConstantExpr::isDesirableBinOp(Opcode) 887 ? ConstantExpr::get(Opcode, C2, C1) 888 : ConstantFoldBinaryInstruction(Opcode, C2, C1); 889 } 890 891 if (ConstantInt *CI1 = dyn_cast<ConstantInt>(C1)) { 892 if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) { 893 const APInt &C1V = CI1->getValue(); 894 const APInt &C2V = CI2->getValue(); 895 switch (Opcode) { 896 default: 897 break; 898 case Instruction::Add: 899 return ConstantInt::get(CI1->getContext(), C1V + C2V); 900 case Instruction::Sub: 901 return ConstantInt::get(CI1->getContext(), C1V - C2V); 902 case Instruction::Mul: 903 return ConstantInt::get(CI1->getContext(), C1V * C2V); 904 case Instruction::UDiv: 905 assert(!CI2->isZero() && "Div by zero handled above"); 906 return ConstantInt::get(CI1->getContext(), C1V.udiv(C2V)); 907 case Instruction::SDiv: 908 assert(!CI2->isZero() && "Div by zero handled above"); 909 if (C2V.isAllOnes() && C1V.isMinSignedValue()) 910 return PoisonValue::get(CI1->getType()); // MIN_INT / -1 -> poison 911 return ConstantInt::get(CI1->getContext(), C1V.sdiv(C2V)); 912 case Instruction::URem: 913 assert(!CI2->isZero() && "Div by zero handled above"); 914 return ConstantInt::get(CI1->getContext(), C1V.urem(C2V)); 915 case Instruction::SRem: 916 assert(!CI2->isZero() && "Div by zero handled above"); 917 if (C2V.isAllOnes() && C1V.isMinSignedValue()) 918 return PoisonValue::get(CI1->getType()); // MIN_INT % -1 -> poison 919 return ConstantInt::get(CI1->getContext(), C1V.srem(C2V)); 920 case Instruction::And: 921 return ConstantInt::get(CI1->getContext(), C1V & C2V); 922 case Instruction::Or: 923 return ConstantInt::get(CI1->getContext(), C1V | C2V); 924 case Instruction::Xor: 925 return ConstantInt::get(CI1->getContext(), C1V ^ C2V); 926 case Instruction::Shl: 927 if (C2V.ult(C1V.getBitWidth())) 928 return ConstantInt::get(CI1->getContext(), C1V.shl(C2V)); 929 return PoisonValue::get(C1->getType()); // too big shift is poison 930 case Instruction::LShr: 931 if (C2V.ult(C1V.getBitWidth())) 932 return ConstantInt::get(CI1->getContext(), C1V.lshr(C2V)); 933 return PoisonValue::get(C1->getType()); // too big shift is poison 934 case Instruction::AShr: 935 if (C2V.ult(C1V.getBitWidth())) 936 return ConstantInt::get(CI1->getContext(), C1V.ashr(C2V)); 937 return PoisonValue::get(C1->getType()); // too big shift is poison 938 } 939 } 940 941 switch (Opcode) { 942 case Instruction::SDiv: 943 case Instruction::UDiv: 944 case Instruction::URem: 945 case Instruction::SRem: 946 case Instruction::LShr: 947 case Instruction::AShr: 948 case Instruction::Shl: 949 if (CI1->isZero()) return C1; 950 break; 951 default: 952 break; 953 } 954 } else if (ConstantFP *CFP1 = dyn_cast<ConstantFP>(C1)) { 955 if (ConstantFP *CFP2 = dyn_cast<ConstantFP>(C2)) { 956 const APFloat &C1V = CFP1->getValueAPF(); 957 const APFloat &C2V = CFP2->getValueAPF(); 958 APFloat C3V = C1V; // copy for modification 959 switch (Opcode) { 960 default: 961 break; 962 case Instruction::FAdd: 963 (void)C3V.add(C2V, APFloat::rmNearestTiesToEven); 964 return ConstantFP::get(C1->getContext(), C3V); 965 case Instruction::FSub: 966 (void)C3V.subtract(C2V, APFloat::rmNearestTiesToEven); 967 return ConstantFP::get(C1->getContext(), C3V); 968 case Instruction::FMul: 969 (void)C3V.multiply(C2V, APFloat::rmNearestTiesToEven); 970 return ConstantFP::get(C1->getContext(), C3V); 971 case Instruction::FDiv: 972 (void)C3V.divide(C2V, APFloat::rmNearestTiesToEven); 973 return ConstantFP::get(C1->getContext(), C3V); 974 case Instruction::FRem: 975 (void)C3V.mod(C2V); 976 return ConstantFP::get(C1->getContext(), C3V); 977 } 978 } 979 } else if (auto *VTy = dyn_cast<VectorType>(C1->getType())) { 980 // Fast path for splatted constants. 981 if (Constant *C2Splat = C2->getSplatValue()) { 982 if (Instruction::isIntDivRem(Opcode) && C2Splat->isNullValue()) 983 return PoisonValue::get(VTy); 984 if (Constant *C1Splat = C1->getSplatValue()) { 985 Constant *Res = 986 ConstantExpr::isDesirableBinOp(Opcode) 987 ? ConstantExpr::get(Opcode, C1Splat, C2Splat) 988 : ConstantFoldBinaryInstruction(Opcode, C1Splat, C2Splat); 989 if (!Res) 990 return nullptr; 991 return ConstantVector::getSplat(VTy->getElementCount(), Res); 992 } 993 } 994 995 if (auto *FVTy = dyn_cast<FixedVectorType>(VTy)) { 996 // Fold each element and create a vector constant from those constants. 997 SmallVector<Constant*, 16> Result; 998 Type *Ty = IntegerType::get(FVTy->getContext(), 32); 999 for (unsigned i = 0, e = FVTy->getNumElements(); i != e; ++i) { 1000 Constant *ExtractIdx = ConstantInt::get(Ty, i); 1001 Constant *LHS = ConstantExpr::getExtractElement(C1, ExtractIdx); 1002 Constant *RHS = ConstantExpr::getExtractElement(C2, ExtractIdx); 1003 1004 // If any element of a divisor vector is zero, the whole op is poison. 1005 if (Instruction::isIntDivRem(Opcode) && RHS->isNullValue()) 1006 return PoisonValue::get(VTy); 1007 1008 Constant *Res = ConstantExpr::isDesirableBinOp(Opcode) 1009 ? ConstantExpr::get(Opcode, LHS, RHS) 1010 : ConstantFoldBinaryInstruction(Opcode, LHS, RHS); 1011 if (!Res) 1012 return nullptr; 1013 Result.push_back(Res); 1014 } 1015 1016 return ConstantVector::get(Result); 1017 } 1018 } 1019 1020 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1021 // There are many possible foldings we could do here. We should probably 1022 // at least fold add of a pointer with an integer into the appropriate 1023 // getelementptr. This will improve alias analysis a bit. 1024 1025 // Given ((a + b) + c), if (b + c) folds to something interesting, return 1026 // (a + (b + c)). 1027 if (Instruction::isAssociative(Opcode) && CE1->getOpcode() == Opcode) { 1028 Constant *T = ConstantExpr::get(Opcode, CE1->getOperand(1), C2); 1029 if (!isa<ConstantExpr>(T) || cast<ConstantExpr>(T)->getOpcode() != Opcode) 1030 return ConstantExpr::get(Opcode, CE1->getOperand(0), T); 1031 } 1032 } else if (isa<ConstantExpr>(C2)) { 1033 // If C2 is a constant expr and C1 isn't, flop them around and fold the 1034 // other way if possible. 1035 if (Instruction::isCommutative(Opcode)) 1036 return ConstantFoldBinaryInstruction(Opcode, C2, C1); 1037 } 1038 1039 // i1 can be simplified in many cases. 1040 if (C1->getType()->isIntegerTy(1)) { 1041 switch (Opcode) { 1042 case Instruction::Add: 1043 case Instruction::Sub: 1044 return ConstantExpr::getXor(C1, C2); 1045 case Instruction::Shl: 1046 case Instruction::LShr: 1047 case Instruction::AShr: 1048 // We can assume that C2 == 0. If it were one the result would be 1049 // undefined because the shift value is as large as the bitwidth. 1050 return C1; 1051 case Instruction::SDiv: 1052 case Instruction::UDiv: 1053 // We can assume that C2 == 1. If it were zero the result would be 1054 // undefined through division by zero. 1055 return C1; 1056 case Instruction::URem: 1057 case Instruction::SRem: 1058 // We can assume that C2 == 1. If it were zero the result would be 1059 // undefined through division by zero. 1060 return ConstantInt::getFalse(C1->getContext()); 1061 default: 1062 break; 1063 } 1064 } 1065 1066 // We don't know how to fold this. 1067 return nullptr; 1068 } 1069 1070 static ICmpInst::Predicate areGlobalsPotentiallyEqual(const GlobalValue *GV1, 1071 const GlobalValue *GV2) { 1072 auto isGlobalUnsafeForEquality = [](const GlobalValue *GV) { 1073 if (GV->isInterposable() || GV->hasGlobalUnnamedAddr()) 1074 return true; 1075 if (const auto *GVar = dyn_cast<GlobalVariable>(GV)) { 1076 Type *Ty = GVar->getValueType(); 1077 // A global with opaque type might end up being zero sized. 1078 if (!Ty->isSized()) 1079 return true; 1080 // A global with an empty type might lie at the address of any other 1081 // global. 1082 if (Ty->isEmptyTy()) 1083 return true; 1084 } 1085 return false; 1086 }; 1087 // Don't try to decide equality of aliases. 1088 if (!isa<GlobalAlias>(GV1) && !isa<GlobalAlias>(GV2)) 1089 if (!isGlobalUnsafeForEquality(GV1) && !isGlobalUnsafeForEquality(GV2)) 1090 return ICmpInst::ICMP_NE; 1091 return ICmpInst::BAD_ICMP_PREDICATE; 1092 } 1093 1094 /// This function determines if there is anything we can decide about the two 1095 /// constants provided. This doesn't need to handle simple things like integer 1096 /// comparisons, but should instead handle ConstantExprs and GlobalValues. 1097 /// If we can determine that the two constants have a particular relation to 1098 /// each other, we should return the corresponding ICmp predicate, otherwise 1099 /// return ICmpInst::BAD_ICMP_PREDICATE. 1100 static ICmpInst::Predicate evaluateICmpRelation(Constant *V1, Constant *V2) { 1101 assert(V1->getType() == V2->getType() && 1102 "Cannot compare different types of values!"); 1103 if (V1 == V2) return ICmpInst::ICMP_EQ; 1104 1105 // The following folds only apply to pointers. 1106 if (!V1->getType()->isPointerTy()) 1107 return ICmpInst::BAD_ICMP_PREDICATE; 1108 1109 // To simplify this code we canonicalize the relation so that the first 1110 // operand is always the most "complex" of the two. We consider simple 1111 // constants (like ConstantPointerNull) to be the simplest, followed by 1112 // BlockAddress, GlobalValues, and ConstantExpr's (the most complex). 1113 auto GetComplexity = [](Constant *V) { 1114 if (isa<ConstantExpr>(V)) 1115 return 3; 1116 if (isa<GlobalValue>(V)) 1117 return 2; 1118 if (isa<BlockAddress>(V)) 1119 return 1; 1120 return 0; 1121 }; 1122 if (GetComplexity(V1) < GetComplexity(V2)) { 1123 ICmpInst::Predicate SwappedRelation = evaluateICmpRelation(V2, V1); 1124 if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE) 1125 return ICmpInst::getSwappedPredicate(SwappedRelation); 1126 return ICmpInst::BAD_ICMP_PREDICATE; 1127 } 1128 1129 if (const BlockAddress *BA = dyn_cast<BlockAddress>(V1)) { 1130 // Now we know that the RHS is a BlockAddress or simple constant. 1131 if (const BlockAddress *BA2 = dyn_cast<BlockAddress>(V2)) { 1132 // Block address in another function can't equal this one, but block 1133 // addresses in the current function might be the same if blocks are 1134 // empty. 1135 if (BA2->getFunction() != BA->getFunction()) 1136 return ICmpInst::ICMP_NE; 1137 } else if (isa<ConstantPointerNull>(V2)) { 1138 return ICmpInst::ICMP_NE; 1139 } 1140 } else if (const GlobalValue *GV = dyn_cast<GlobalValue>(V1)) { 1141 // Now we know that the RHS is a GlobalValue, BlockAddress or simple 1142 // constant. 1143 if (const GlobalValue *GV2 = dyn_cast<GlobalValue>(V2)) { 1144 return areGlobalsPotentiallyEqual(GV, GV2); 1145 } else if (isa<BlockAddress>(V2)) { 1146 return ICmpInst::ICMP_NE; // Globals never equal labels. 1147 } else if (isa<ConstantPointerNull>(V2)) { 1148 // GlobalVals can never be null unless they have external weak linkage. 1149 // We don't try to evaluate aliases here. 1150 // NOTE: We should not be doing this constant folding if null pointer 1151 // is considered valid for the function. But currently there is no way to 1152 // query it from the Constant type. 1153 if (!GV->hasExternalWeakLinkage() && !isa<GlobalAlias>(GV) && 1154 !NullPointerIsDefined(nullptr /* F */, 1155 GV->getType()->getAddressSpace())) 1156 return ICmpInst::ICMP_UGT; 1157 } 1158 } else if (auto *CE1 = dyn_cast<ConstantExpr>(V1)) { 1159 // Ok, the LHS is known to be a constantexpr. The RHS can be any of a 1160 // constantexpr, a global, block address, or a simple constant. 1161 Constant *CE1Op0 = CE1->getOperand(0); 1162 1163 switch (CE1->getOpcode()) { 1164 case Instruction::GetElementPtr: { 1165 GEPOperator *CE1GEP = cast<GEPOperator>(CE1); 1166 // Ok, since this is a getelementptr, we know that the constant has a 1167 // pointer type. Check the various cases. 1168 if (isa<ConstantPointerNull>(V2)) { 1169 // If we are comparing a GEP to a null pointer, check to see if the base 1170 // of the GEP equals the null pointer. 1171 if (const GlobalValue *GV = dyn_cast<GlobalValue>(CE1Op0)) { 1172 // If its not weak linkage, the GVal must have a non-zero address 1173 // so the result is greater-than 1174 if (!GV->hasExternalWeakLinkage() && CE1GEP->isInBounds()) 1175 return ICmpInst::ICMP_UGT; 1176 } 1177 } else if (const GlobalValue *GV2 = dyn_cast<GlobalValue>(V2)) { 1178 if (const GlobalValue *GV = dyn_cast<GlobalValue>(CE1Op0)) { 1179 if (GV != GV2) { 1180 if (CE1GEP->hasAllZeroIndices()) 1181 return areGlobalsPotentiallyEqual(GV, GV2); 1182 return ICmpInst::BAD_ICMP_PREDICATE; 1183 } 1184 } 1185 } else if (const auto *CE2GEP = dyn_cast<GEPOperator>(V2)) { 1186 // By far the most common case to handle is when the base pointers are 1187 // obviously to the same global. 1188 const Constant *CE2Op0 = cast<Constant>(CE2GEP->getPointerOperand()); 1189 if (isa<GlobalValue>(CE1Op0) && isa<GlobalValue>(CE2Op0)) { 1190 // Don't know relative ordering, but check for inequality. 1191 if (CE1Op0 != CE2Op0) { 1192 if (CE1GEP->hasAllZeroIndices() && CE2GEP->hasAllZeroIndices()) 1193 return areGlobalsPotentiallyEqual(cast<GlobalValue>(CE1Op0), 1194 cast<GlobalValue>(CE2Op0)); 1195 return ICmpInst::BAD_ICMP_PREDICATE; 1196 } 1197 } 1198 } 1199 break; 1200 } 1201 default: 1202 break; 1203 } 1204 } 1205 1206 return ICmpInst::BAD_ICMP_PREDICATE; 1207 } 1208 1209 Constant *llvm::ConstantFoldCompareInstruction(CmpInst::Predicate Predicate, 1210 Constant *C1, Constant *C2) { 1211 Type *ResultTy; 1212 if (VectorType *VT = dyn_cast<VectorType>(C1->getType())) 1213 ResultTy = VectorType::get(Type::getInt1Ty(C1->getContext()), 1214 VT->getElementCount()); 1215 else 1216 ResultTy = Type::getInt1Ty(C1->getContext()); 1217 1218 // Fold FCMP_FALSE/FCMP_TRUE unconditionally. 1219 if (Predicate == FCmpInst::FCMP_FALSE) 1220 return Constant::getNullValue(ResultTy); 1221 1222 if (Predicate == FCmpInst::FCMP_TRUE) 1223 return Constant::getAllOnesValue(ResultTy); 1224 1225 // Handle some degenerate cases first 1226 if (isa<PoisonValue>(C1) || isa<PoisonValue>(C2)) 1227 return PoisonValue::get(ResultTy); 1228 1229 if (isa<UndefValue>(C1) || isa<UndefValue>(C2)) { 1230 bool isIntegerPredicate = ICmpInst::isIntPredicate(Predicate); 1231 // For EQ and NE, we can always pick a value for the undef to make the 1232 // predicate pass or fail, so we can return undef. 1233 // Also, if both operands are undef, we can return undef for int comparison. 1234 if (ICmpInst::isEquality(Predicate) || (isIntegerPredicate && C1 == C2)) 1235 return UndefValue::get(ResultTy); 1236 1237 // Otherwise, for integer compare, pick the same value as the non-undef 1238 // operand, and fold it to true or false. 1239 if (isIntegerPredicate) 1240 return ConstantInt::get(ResultTy, CmpInst::isTrueWhenEqual(Predicate)); 1241 1242 // Choosing NaN for the undef will always make unordered comparison succeed 1243 // and ordered comparison fails. 1244 return ConstantInt::get(ResultTy, CmpInst::isUnordered(Predicate)); 1245 } 1246 1247 if (C2->isNullValue()) { 1248 // The caller is expected to commute the operands if the constant expression 1249 // is C2. 1250 // C1 >= 0 --> true 1251 if (Predicate == ICmpInst::ICMP_UGE) 1252 return Constant::getAllOnesValue(ResultTy); 1253 // C1 < 0 --> false 1254 if (Predicate == ICmpInst::ICMP_ULT) 1255 return Constant::getNullValue(ResultTy); 1256 } 1257 1258 // If the comparison is a comparison between two i1's, simplify it. 1259 if (C1->getType()->isIntegerTy(1)) { 1260 switch (Predicate) { 1261 case ICmpInst::ICMP_EQ: 1262 if (isa<ConstantInt>(C2)) 1263 return ConstantExpr::getXor(C1, ConstantExpr::getNot(C2)); 1264 return ConstantExpr::getXor(ConstantExpr::getNot(C1), C2); 1265 case ICmpInst::ICMP_NE: 1266 return ConstantExpr::getXor(C1, C2); 1267 default: 1268 break; 1269 } 1270 } 1271 1272 if (isa<ConstantInt>(C1) && isa<ConstantInt>(C2)) { 1273 const APInt &V1 = cast<ConstantInt>(C1)->getValue(); 1274 const APInt &V2 = cast<ConstantInt>(C2)->getValue(); 1275 return ConstantInt::get(ResultTy, ICmpInst::compare(V1, V2, Predicate)); 1276 } else if (isa<ConstantFP>(C1) && isa<ConstantFP>(C2)) { 1277 const APFloat &C1V = cast<ConstantFP>(C1)->getValueAPF(); 1278 const APFloat &C2V = cast<ConstantFP>(C2)->getValueAPF(); 1279 return ConstantInt::get(ResultTy, FCmpInst::compare(C1V, C2V, Predicate)); 1280 } else if (auto *C1VTy = dyn_cast<VectorType>(C1->getType())) { 1281 1282 // Fast path for splatted constants. 1283 if (Constant *C1Splat = C1->getSplatValue()) 1284 if (Constant *C2Splat = C2->getSplatValue()) 1285 return ConstantVector::getSplat( 1286 C1VTy->getElementCount(), 1287 ConstantExpr::getCompare(Predicate, C1Splat, C2Splat)); 1288 1289 // Do not iterate on scalable vector. The number of elements is unknown at 1290 // compile-time. 1291 if (isa<ScalableVectorType>(C1VTy)) 1292 return nullptr; 1293 1294 // If we can constant fold the comparison of each element, constant fold 1295 // the whole vector comparison. 1296 SmallVector<Constant*, 4> ResElts; 1297 Type *Ty = IntegerType::get(C1->getContext(), 32); 1298 // Compare the elements, producing an i1 result or constant expr. 1299 for (unsigned I = 0, E = C1VTy->getElementCount().getKnownMinValue(); 1300 I != E; ++I) { 1301 Constant *C1E = 1302 ConstantExpr::getExtractElement(C1, ConstantInt::get(Ty, I)); 1303 Constant *C2E = 1304 ConstantExpr::getExtractElement(C2, ConstantInt::get(Ty, I)); 1305 1306 ResElts.push_back(ConstantExpr::getCompare(Predicate, C1E, C2E)); 1307 } 1308 1309 return ConstantVector::get(ResElts); 1310 } 1311 1312 if (C1->getType()->isFPOrFPVectorTy()) { 1313 if (C1 == C2) { 1314 // We know that C1 == C2 || isUnordered(C1, C2). 1315 if (Predicate == FCmpInst::FCMP_ONE) 1316 return ConstantInt::getFalse(ResultTy); 1317 else if (Predicate == FCmpInst::FCMP_UEQ) 1318 return ConstantInt::getTrue(ResultTy); 1319 } 1320 } else { 1321 // Evaluate the relation between the two constants, per the predicate. 1322 int Result = -1; // -1 = unknown, 0 = known false, 1 = known true. 1323 switch (evaluateICmpRelation(C1, C2)) { 1324 default: llvm_unreachable("Unknown relational!"); 1325 case ICmpInst::BAD_ICMP_PREDICATE: 1326 break; // Couldn't determine anything about these constants. 1327 case ICmpInst::ICMP_EQ: // We know the constants are equal! 1328 // If we know the constants are equal, we can decide the result of this 1329 // computation precisely. 1330 Result = ICmpInst::isTrueWhenEqual(Predicate); 1331 break; 1332 case ICmpInst::ICMP_ULT: 1333 switch (Predicate) { 1334 case ICmpInst::ICMP_ULT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_ULE: 1335 Result = 1; break; 1336 case ICmpInst::ICMP_UGT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_UGE: 1337 Result = 0; break; 1338 default: 1339 break; 1340 } 1341 break; 1342 case ICmpInst::ICMP_SLT: 1343 switch (Predicate) { 1344 case ICmpInst::ICMP_SLT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_SLE: 1345 Result = 1; break; 1346 case ICmpInst::ICMP_SGT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_SGE: 1347 Result = 0; break; 1348 default: 1349 break; 1350 } 1351 break; 1352 case ICmpInst::ICMP_UGT: 1353 switch (Predicate) { 1354 case ICmpInst::ICMP_UGT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_UGE: 1355 Result = 1; break; 1356 case ICmpInst::ICMP_ULT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_ULE: 1357 Result = 0; break; 1358 default: 1359 break; 1360 } 1361 break; 1362 case ICmpInst::ICMP_SGT: 1363 switch (Predicate) { 1364 case ICmpInst::ICMP_SGT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_SGE: 1365 Result = 1; break; 1366 case ICmpInst::ICMP_SLT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_SLE: 1367 Result = 0; break; 1368 default: 1369 break; 1370 } 1371 break; 1372 case ICmpInst::ICMP_ULE: 1373 if (Predicate == ICmpInst::ICMP_UGT) 1374 Result = 0; 1375 if (Predicate == ICmpInst::ICMP_ULT || Predicate == ICmpInst::ICMP_ULE) 1376 Result = 1; 1377 break; 1378 case ICmpInst::ICMP_SLE: 1379 if (Predicate == ICmpInst::ICMP_SGT) 1380 Result = 0; 1381 if (Predicate == ICmpInst::ICMP_SLT || Predicate == ICmpInst::ICMP_SLE) 1382 Result = 1; 1383 break; 1384 case ICmpInst::ICMP_UGE: 1385 if (Predicate == ICmpInst::ICMP_ULT) 1386 Result = 0; 1387 if (Predicate == ICmpInst::ICMP_UGT || Predicate == ICmpInst::ICMP_UGE) 1388 Result = 1; 1389 break; 1390 case ICmpInst::ICMP_SGE: 1391 if (Predicate == ICmpInst::ICMP_SLT) 1392 Result = 0; 1393 if (Predicate == ICmpInst::ICMP_SGT || Predicate == ICmpInst::ICMP_SGE) 1394 Result = 1; 1395 break; 1396 case ICmpInst::ICMP_NE: 1397 if (Predicate == ICmpInst::ICMP_EQ) 1398 Result = 0; 1399 if (Predicate == ICmpInst::ICMP_NE) 1400 Result = 1; 1401 break; 1402 } 1403 1404 // If we evaluated the result, return it now. 1405 if (Result != -1) 1406 return ConstantInt::get(ResultTy, Result); 1407 1408 if ((!isa<ConstantExpr>(C1) && isa<ConstantExpr>(C2)) || 1409 (C1->isNullValue() && !C2->isNullValue())) { 1410 // If C2 is a constant expr and C1 isn't, flip them around and fold the 1411 // other way if possible. 1412 // Also, if C1 is null and C2 isn't, flip them around. 1413 Predicate = ICmpInst::getSwappedPredicate(Predicate); 1414 return ConstantExpr::getICmp(Predicate, C2, C1); 1415 } 1416 } 1417 return nullptr; 1418 } 1419 1420 /// Test whether the given sequence of *normalized* indices is "inbounds". 1421 template<typename IndexTy> 1422 static bool isInBoundsIndices(ArrayRef<IndexTy> Idxs) { 1423 // No indices means nothing that could be out of bounds. 1424 if (Idxs.empty()) return true; 1425 1426 // If the first index is zero, it's in bounds. 1427 if (cast<Constant>(Idxs[0])->isNullValue()) return true; 1428 1429 // If the first index is one and all the rest are zero, it's in bounds, 1430 // by the one-past-the-end rule. 1431 if (auto *CI = dyn_cast<ConstantInt>(Idxs[0])) { 1432 if (!CI->isOne()) 1433 return false; 1434 } else { 1435 auto *CV = cast<ConstantDataVector>(Idxs[0]); 1436 CI = dyn_cast_or_null<ConstantInt>(CV->getSplatValue()); 1437 if (!CI || !CI->isOne()) 1438 return false; 1439 } 1440 1441 for (unsigned i = 1, e = Idxs.size(); i != e; ++i) 1442 if (!cast<Constant>(Idxs[i])->isNullValue()) 1443 return false; 1444 return true; 1445 } 1446 1447 /// Test whether a given ConstantInt is in-range for a SequentialType. 1448 static bool isIndexInRangeOfArrayType(uint64_t NumElements, 1449 const ConstantInt *CI) { 1450 // We cannot bounds check the index if it doesn't fit in an int64_t. 1451 if (CI->getValue().getSignificantBits() > 64) 1452 return false; 1453 1454 // A negative index or an index past the end of our sequential type is 1455 // considered out-of-range. 1456 int64_t IndexVal = CI->getSExtValue(); 1457 if (IndexVal < 0 || (IndexVal != 0 && (uint64_t)IndexVal >= NumElements)) 1458 return false; 1459 1460 // Otherwise, it is in-range. 1461 return true; 1462 } 1463 1464 // Combine Indices - If the source pointer to this getelementptr instruction 1465 // is a getelementptr instruction, combine the indices of the two 1466 // getelementptr instructions into a single instruction. 1467 static Constant *foldGEPOfGEP(GEPOperator *GEP, Type *PointeeTy, bool InBounds, 1468 ArrayRef<Value *> Idxs) { 1469 if (PointeeTy != GEP->getResultElementType()) 1470 return nullptr; 1471 1472 // Leave inrange handling to DL-aware constant folding. 1473 if (GEP->getInRange()) 1474 return nullptr; 1475 1476 Constant *Idx0 = cast<Constant>(Idxs[0]); 1477 if (Idx0->isNullValue()) { 1478 // Handle the simple case of a zero index. 1479 SmallVector<Value*, 16> NewIndices; 1480 NewIndices.reserve(Idxs.size() + GEP->getNumIndices()); 1481 NewIndices.append(GEP->idx_begin(), GEP->idx_end()); 1482 NewIndices.append(Idxs.begin() + 1, Idxs.end()); 1483 return ConstantExpr::getGetElementPtr( 1484 GEP->getSourceElementType(), cast<Constant>(GEP->getPointerOperand()), 1485 NewIndices, InBounds && GEP->isInBounds()); 1486 } 1487 1488 gep_type_iterator LastI = gep_type_end(GEP); 1489 for (gep_type_iterator I = gep_type_begin(GEP), E = gep_type_end(GEP); 1490 I != E; ++I) 1491 LastI = I; 1492 1493 // We can't combine GEPs if the last index is a struct type. 1494 if (!LastI.isSequential()) 1495 return nullptr; 1496 // We could perform the transform with non-constant index, but prefer leaving 1497 // it as GEP of GEP rather than GEP of add for now. 1498 ConstantInt *CI = dyn_cast<ConstantInt>(Idx0); 1499 if (!CI) 1500 return nullptr; 1501 1502 // TODO: This code may be extended to handle vectors as well. 1503 auto *LastIdx = cast<Constant>(GEP->getOperand(GEP->getNumOperands()-1)); 1504 Type *LastIdxTy = LastIdx->getType(); 1505 if (LastIdxTy->isVectorTy()) 1506 return nullptr; 1507 1508 SmallVector<Value*, 16> NewIndices; 1509 NewIndices.reserve(Idxs.size() + GEP->getNumIndices()); 1510 NewIndices.append(GEP->idx_begin(), GEP->idx_end() - 1); 1511 1512 // Add the last index of the source with the first index of the new GEP. 1513 // Make sure to handle the case when they are actually different types. 1514 if (LastIdxTy != Idx0->getType()) { 1515 unsigned CommonExtendedWidth = 1516 std::max(LastIdxTy->getIntegerBitWidth(), 1517 Idx0->getType()->getIntegerBitWidth()); 1518 CommonExtendedWidth = std::max(CommonExtendedWidth, 64U); 1519 1520 Type *CommonTy = 1521 Type::getIntNTy(LastIdxTy->getContext(), CommonExtendedWidth); 1522 if (Idx0->getType() != CommonTy) 1523 Idx0 = ConstantFoldCastInstruction(Instruction::SExt, Idx0, CommonTy); 1524 if (LastIdx->getType() != CommonTy) 1525 LastIdx = 1526 ConstantFoldCastInstruction(Instruction::SExt, LastIdx, CommonTy); 1527 if (!Idx0 || !LastIdx) 1528 return nullptr; 1529 } 1530 1531 NewIndices.push_back(ConstantExpr::get(Instruction::Add, Idx0, LastIdx)); 1532 NewIndices.append(Idxs.begin() + 1, Idxs.end()); 1533 1534 return ConstantExpr::getGetElementPtr( 1535 GEP->getSourceElementType(), cast<Constant>(GEP->getPointerOperand()), 1536 NewIndices, InBounds && GEP->isInBounds()); 1537 } 1538 1539 Constant *llvm::ConstantFoldGetElementPtr(Type *PointeeTy, Constant *C, 1540 bool InBounds, 1541 std::optional<ConstantRange> InRange, 1542 ArrayRef<Value *> Idxs) { 1543 if (Idxs.empty()) return C; 1544 1545 Type *GEPTy = GetElementPtrInst::getGEPReturnType( 1546 C, ArrayRef((Value *const *)Idxs.data(), Idxs.size())); 1547 1548 if (isa<PoisonValue>(C)) 1549 return PoisonValue::get(GEPTy); 1550 1551 if (isa<UndefValue>(C)) 1552 // If inbounds, we can choose an out-of-bounds pointer as a base pointer. 1553 return InBounds ? PoisonValue::get(GEPTy) : UndefValue::get(GEPTy); 1554 1555 auto IsNoOp = [&]() { 1556 // Avoid losing inrange information. 1557 if (InRange) 1558 return false; 1559 1560 return all_of(Idxs, [](Value *Idx) { 1561 Constant *IdxC = cast<Constant>(Idx); 1562 return IdxC->isNullValue() || isa<UndefValue>(IdxC); 1563 }); 1564 }; 1565 if (IsNoOp()) 1566 return GEPTy->isVectorTy() && !C->getType()->isVectorTy() 1567 ? ConstantVector::getSplat( 1568 cast<VectorType>(GEPTy)->getElementCount(), C) 1569 : C; 1570 1571 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) 1572 if (auto *GEP = dyn_cast<GEPOperator>(CE)) 1573 if (Constant *C = foldGEPOfGEP(GEP, PointeeTy, InBounds, Idxs)) 1574 return C; 1575 1576 // Check to see if any array indices are not within the corresponding 1577 // notional array or vector bounds. If so, try to determine if they can be 1578 // factored out into preceding dimensions. 1579 SmallVector<Constant *, 8> NewIdxs; 1580 Type *Ty = PointeeTy; 1581 Type *Prev = C->getType(); 1582 auto GEPIter = gep_type_begin(PointeeTy, Idxs); 1583 bool Unknown = 1584 !isa<ConstantInt>(Idxs[0]) && !isa<ConstantDataVector>(Idxs[0]); 1585 for (unsigned i = 1, e = Idxs.size(); i != e; 1586 Prev = Ty, Ty = (++GEPIter).getIndexedType(), ++i) { 1587 if (!isa<ConstantInt>(Idxs[i]) && !isa<ConstantDataVector>(Idxs[i])) { 1588 // We don't know if it's in range or not. 1589 Unknown = true; 1590 continue; 1591 } 1592 if (!isa<ConstantInt>(Idxs[i - 1]) && !isa<ConstantDataVector>(Idxs[i - 1])) 1593 // Skip if the type of the previous index is not supported. 1594 continue; 1595 if (isa<StructType>(Ty)) { 1596 // The verify makes sure that GEPs into a struct are in range. 1597 continue; 1598 } 1599 if (isa<VectorType>(Ty)) { 1600 // There can be awkward padding in after a non-power of two vector. 1601 Unknown = true; 1602 continue; 1603 } 1604 auto *STy = cast<ArrayType>(Ty); 1605 if (ConstantInt *CI = dyn_cast<ConstantInt>(Idxs[i])) { 1606 if (isIndexInRangeOfArrayType(STy->getNumElements(), CI)) 1607 // It's in range, skip to the next index. 1608 continue; 1609 if (CI->isNegative()) { 1610 // It's out of range and negative, don't try to factor it. 1611 Unknown = true; 1612 continue; 1613 } 1614 } else { 1615 auto *CV = cast<ConstantDataVector>(Idxs[i]); 1616 bool IsInRange = true; 1617 for (unsigned I = 0, E = CV->getNumElements(); I != E; ++I) { 1618 auto *CI = cast<ConstantInt>(CV->getElementAsConstant(I)); 1619 IsInRange &= isIndexInRangeOfArrayType(STy->getNumElements(), CI); 1620 if (CI->isNegative()) { 1621 Unknown = true; 1622 break; 1623 } 1624 } 1625 if (IsInRange || Unknown) 1626 // It's in range, skip to the next index. 1627 // It's out of range and negative, don't try to factor it. 1628 continue; 1629 } 1630 if (isa<StructType>(Prev)) { 1631 // It's out of range, but the prior dimension is a struct 1632 // so we can't do anything about it. 1633 Unknown = true; 1634 continue; 1635 } 1636 1637 // Determine the number of elements in our sequential type. 1638 uint64_t NumElements = STy->getArrayNumElements(); 1639 if (!NumElements) { 1640 Unknown = true; 1641 continue; 1642 } 1643 1644 // It's out of range, but we can factor it into the prior 1645 // dimension. 1646 NewIdxs.resize(Idxs.size()); 1647 1648 // Expand the current index or the previous index to a vector from a scalar 1649 // if necessary. 1650 Constant *CurrIdx = cast<Constant>(Idxs[i]); 1651 auto *PrevIdx = 1652 NewIdxs[i - 1] ? NewIdxs[i - 1] : cast<Constant>(Idxs[i - 1]); 1653 bool IsCurrIdxVector = CurrIdx->getType()->isVectorTy(); 1654 bool IsPrevIdxVector = PrevIdx->getType()->isVectorTy(); 1655 bool UseVector = IsCurrIdxVector || IsPrevIdxVector; 1656 1657 if (!IsCurrIdxVector && IsPrevIdxVector) 1658 CurrIdx = ConstantDataVector::getSplat( 1659 cast<FixedVectorType>(PrevIdx->getType())->getNumElements(), CurrIdx); 1660 1661 if (!IsPrevIdxVector && IsCurrIdxVector) 1662 PrevIdx = ConstantDataVector::getSplat( 1663 cast<FixedVectorType>(CurrIdx->getType())->getNumElements(), PrevIdx); 1664 1665 Constant *Factor = 1666 ConstantInt::get(CurrIdx->getType()->getScalarType(), NumElements); 1667 if (UseVector) 1668 Factor = ConstantDataVector::getSplat( 1669 IsPrevIdxVector 1670 ? cast<FixedVectorType>(PrevIdx->getType())->getNumElements() 1671 : cast<FixedVectorType>(CurrIdx->getType())->getNumElements(), 1672 Factor); 1673 1674 NewIdxs[i] = 1675 ConstantFoldBinaryInstruction(Instruction::SRem, CurrIdx, Factor); 1676 1677 Constant *Div = 1678 ConstantFoldBinaryInstruction(Instruction::SDiv, CurrIdx, Factor); 1679 1680 // We're working on either ConstantInt or vectors of ConstantInt, 1681 // so these should always fold. 1682 assert(NewIdxs[i] != nullptr && Div != nullptr && "Should have folded"); 1683 1684 unsigned CommonExtendedWidth = 1685 std::max(PrevIdx->getType()->getScalarSizeInBits(), 1686 Div->getType()->getScalarSizeInBits()); 1687 CommonExtendedWidth = std::max(CommonExtendedWidth, 64U); 1688 1689 // Before adding, extend both operands to i64 to avoid 1690 // overflow trouble. 1691 Type *ExtendedTy = Type::getIntNTy(Div->getContext(), CommonExtendedWidth); 1692 if (UseVector) 1693 ExtendedTy = FixedVectorType::get( 1694 ExtendedTy, 1695 IsPrevIdxVector 1696 ? cast<FixedVectorType>(PrevIdx->getType())->getNumElements() 1697 : cast<FixedVectorType>(CurrIdx->getType())->getNumElements()); 1698 1699 if (!PrevIdx->getType()->isIntOrIntVectorTy(CommonExtendedWidth)) 1700 PrevIdx = 1701 ConstantFoldCastInstruction(Instruction::SExt, PrevIdx, ExtendedTy); 1702 1703 if (!Div->getType()->isIntOrIntVectorTy(CommonExtendedWidth)) 1704 Div = ConstantFoldCastInstruction(Instruction::SExt, Div, ExtendedTy); 1705 1706 assert(PrevIdx && Div && "Should have folded"); 1707 NewIdxs[i - 1] = ConstantExpr::getAdd(PrevIdx, Div); 1708 } 1709 1710 // If we did any factoring, start over with the adjusted indices. 1711 if (!NewIdxs.empty()) { 1712 for (unsigned i = 0, e = Idxs.size(); i != e; ++i) 1713 if (!NewIdxs[i]) NewIdxs[i] = cast<Constant>(Idxs[i]); 1714 return ConstantExpr::getGetElementPtr(PointeeTy, C, NewIdxs, InBounds, 1715 InRange); 1716 } 1717 1718 // If all indices are known integers and normalized, we can do a simple 1719 // check for the "inbounds" property. 1720 if (!Unknown && !InBounds) 1721 if (auto *GV = dyn_cast<GlobalVariable>(C)) 1722 if (!GV->hasExternalWeakLinkage() && GV->getValueType() == PointeeTy && 1723 isInBoundsIndices(Idxs)) 1724 return ConstantExpr::getGetElementPtr(PointeeTy, C, Idxs, 1725 /*InBounds=*/true, InRange); 1726 1727 return nullptr; 1728 } 1729