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 "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 #include "llvm/Support/ManagedStatic.h" 34 #include "llvm/Support/MathExtras.h" 35 using namespace llvm; 36 using namespace llvm::PatternMatch; 37 38 //===----------------------------------------------------------------------===// 39 // ConstantFold*Instruction Implementations 40 //===----------------------------------------------------------------------===// 41 42 /// Convert the specified vector Constant node to the specified vector type. 43 /// At this point, we know that the elements of the input vector constant are 44 /// all simple integer or FP values. 45 static Constant *BitCastConstantVector(Constant *CV, VectorType *DstTy) { 46 47 if (CV->isAllOnesValue()) return Constant::getAllOnesValue(DstTy); 48 if (CV->isNullValue()) return Constant::getNullValue(DstTy); 49 50 // If this cast changes element count then we can't handle it here: 51 // doing so requires endianness information. This should be handled by 52 // Analysis/ConstantFolding.cpp 53 unsigned NumElts = DstTy->getNumElements(); 54 if (NumElts != CV->getType()->getVectorNumElements()) 55 return nullptr; 56 57 Type *DstEltTy = DstTy->getElementType(); 58 59 SmallVector<Constant*, 16> Result; 60 Type *Ty = IntegerType::get(CV->getContext(), 32); 61 for (unsigned i = 0; i != NumElts; ++i) { 62 Constant *C = 63 ConstantExpr::getExtractElement(CV, ConstantInt::get(Ty, i)); 64 C = ConstantExpr::getBitCast(C, DstEltTy); 65 Result.push_back(C); 66 } 67 68 return ConstantVector::get(Result); 69 } 70 71 /// This function determines which opcode to use to fold two constant cast 72 /// expressions together. It uses CastInst::isEliminableCastPair to determine 73 /// the opcode. Consequently its just a wrapper around that function. 74 /// Determine if it is valid to fold a cast of a cast 75 static unsigned 76 foldConstantCastPair( 77 unsigned opc, ///< opcode of the second cast constant expression 78 ConstantExpr *Op, ///< the first cast constant expression 79 Type *DstTy ///< destination type of the first cast 80 ) { 81 assert(Op && Op->isCast() && "Can't fold cast of cast without a cast!"); 82 assert(DstTy && DstTy->isFirstClassType() && "Invalid cast destination type"); 83 assert(CastInst::isCast(opc) && "Invalid cast opcode"); 84 85 // The types and opcodes for the two Cast constant expressions 86 Type *SrcTy = Op->getOperand(0)->getType(); 87 Type *MidTy = Op->getType(); 88 Instruction::CastOps firstOp = Instruction::CastOps(Op->getOpcode()); 89 Instruction::CastOps secondOp = Instruction::CastOps(opc); 90 91 // Assume that pointers are never more than 64 bits wide, and only use this 92 // for the middle type. Otherwise we could end up folding away illegal 93 // bitcasts between address spaces with different sizes. 94 IntegerType *FakeIntPtrTy = Type::getInt64Ty(DstTy->getContext()); 95 96 // Let CastInst::isEliminableCastPair do the heavy lifting. 97 return CastInst::isEliminableCastPair(firstOp, secondOp, SrcTy, MidTy, DstTy, 98 nullptr, FakeIntPtrTy, nullptr); 99 } 100 101 static Constant *FoldBitCast(Constant *V, Type *DestTy) { 102 Type *SrcTy = V->getType(); 103 if (SrcTy == DestTy) 104 return V; // no-op cast 105 106 // Check to see if we are casting a pointer to an aggregate to a pointer to 107 // the first element. If so, return the appropriate GEP instruction. 108 if (PointerType *PTy = dyn_cast<PointerType>(V->getType())) 109 if (PointerType *DPTy = dyn_cast<PointerType>(DestTy)) 110 if (PTy->getAddressSpace() == DPTy->getAddressSpace() 111 && PTy->getElementType()->isSized()) { 112 SmallVector<Value*, 8> IdxList; 113 Value *Zero = 114 Constant::getNullValue(Type::getInt32Ty(DPTy->getContext())); 115 IdxList.push_back(Zero); 116 Type *ElTy = PTy->getElementType(); 117 while (ElTy != DPTy->getElementType()) { 118 if (StructType *STy = dyn_cast<StructType>(ElTy)) { 119 if (STy->getNumElements() == 0) break; 120 ElTy = STy->getElementType(0); 121 IdxList.push_back(Zero); 122 } else if (SequentialType *STy = 123 dyn_cast<SequentialType>(ElTy)) { 124 ElTy = STy->getElementType(); 125 IdxList.push_back(Zero); 126 } else { 127 break; 128 } 129 } 130 131 if (ElTy == DPTy->getElementType()) 132 // This GEP is inbounds because all indices are zero. 133 return ConstantExpr::getInBoundsGetElementPtr(PTy->getElementType(), 134 V, IdxList); 135 } 136 137 // Handle casts from one vector constant to another. We know that the src 138 // and dest type have the same size (otherwise its an illegal cast). 139 if (VectorType *DestPTy = dyn_cast<VectorType>(DestTy)) { 140 if (VectorType *SrcTy = dyn_cast<VectorType>(V->getType())) { 141 assert(DestPTy->getBitWidth() == SrcTy->getBitWidth() && 142 "Not cast between same sized vectors!"); 143 SrcTy = nullptr; 144 // First, check for null. Undef is already handled. 145 if (isa<ConstantAggregateZero>(V)) 146 return Constant::getNullValue(DestTy); 147 148 // Handle ConstantVector and ConstantAggregateVector. 149 return BitCastConstantVector(V, DestPTy); 150 } 151 152 // Canonicalize scalar-to-vector bitcasts into vector-to-vector bitcasts 153 // This allows for other simplifications (although some of them 154 // can only be handled by Analysis/ConstantFolding.cpp). 155 if (isa<ConstantInt>(V) || isa<ConstantFP>(V)) 156 return ConstantExpr::getBitCast(ConstantVector::get(V), DestPTy); 157 } 158 159 // Finally, implement bitcast folding now. The code below doesn't handle 160 // bitcast right. 161 if (isa<ConstantPointerNull>(V)) // ptr->ptr cast. 162 return ConstantPointerNull::get(cast<PointerType>(DestTy)); 163 164 // Handle integral constant input. 165 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 166 if (DestTy->isIntegerTy()) 167 // Integral -> Integral. This is a no-op because the bit widths must 168 // be the same. Consequently, we just fold to V. 169 return V; 170 171 // See note below regarding the PPC_FP128 restriction. 172 if (DestTy->isFloatingPointTy() && !DestTy->isPPC_FP128Ty()) 173 return ConstantFP::get(DestTy->getContext(), 174 APFloat(DestTy->getFltSemantics(), 175 CI->getValue())); 176 177 // Otherwise, can't fold this (vector?) 178 return nullptr; 179 } 180 181 // Handle ConstantFP input: FP -> Integral. 182 if (ConstantFP *FP = dyn_cast<ConstantFP>(V)) { 183 // PPC_FP128 is really the sum of two consecutive doubles, where the first 184 // double is always stored first in memory, regardless of the target 185 // endianness. The memory layout of i128, however, depends on the target 186 // endianness, and so we can't fold this without target endianness 187 // information. This should instead be handled by 188 // Analysis/ConstantFolding.cpp 189 if (FP->getType()->isPPC_FP128Ty()) 190 return nullptr; 191 192 // Make sure dest type is compatible with the folded integer constant. 193 if (!DestTy->isIntegerTy()) 194 return nullptr; 195 196 return ConstantInt::get(FP->getContext(), 197 FP->getValueAPF().bitcastToAPInt()); 198 } 199 200 return nullptr; 201 } 202 203 204 /// V is an integer constant which only has a subset of its bytes used. 205 /// The bytes used are indicated by ByteStart (which is the first byte used, 206 /// counting from the least significant byte) and ByteSize, which is the number 207 /// of bytes used. 208 /// 209 /// This function analyzes the specified constant to see if the specified byte 210 /// range can be returned as a simplified constant. If so, the constant is 211 /// returned, otherwise null is returned. 212 static Constant *ExtractConstantBytes(Constant *C, unsigned ByteStart, 213 unsigned ByteSize) { 214 assert(C->getType()->isIntegerTy() && 215 (cast<IntegerType>(C->getType())->getBitWidth() & 7) == 0 && 216 "Non-byte sized integer input"); 217 unsigned CSize = cast<IntegerType>(C->getType())->getBitWidth()/8; 218 assert(ByteSize && "Must be accessing some piece"); 219 assert(ByteStart+ByteSize <= CSize && "Extracting invalid piece from input"); 220 assert(ByteSize != CSize && "Should not extract everything"); 221 222 // Constant Integers are simple. 223 if (ConstantInt *CI = dyn_cast<ConstantInt>(C)) { 224 APInt V = CI->getValue(); 225 if (ByteStart) 226 V.lshrInPlace(ByteStart*8); 227 V = V.trunc(ByteSize*8); 228 return ConstantInt::get(CI->getContext(), V); 229 } 230 231 // In the input is a constant expr, we might be able to recursively simplify. 232 // If not, we definitely can't do anything. 233 ConstantExpr *CE = dyn_cast<ConstantExpr>(C); 234 if (!CE) return nullptr; 235 236 switch (CE->getOpcode()) { 237 default: return nullptr; 238 case Instruction::Or: { 239 Constant *RHS = ExtractConstantBytes(CE->getOperand(1), ByteStart,ByteSize); 240 if (!RHS) 241 return nullptr; 242 243 // X | -1 -> -1. 244 if (ConstantInt *RHSC = dyn_cast<ConstantInt>(RHS)) 245 if (RHSC->isMinusOne()) 246 return RHSC; 247 248 Constant *LHS = ExtractConstantBytes(CE->getOperand(0), ByteStart,ByteSize); 249 if (!LHS) 250 return nullptr; 251 return ConstantExpr::getOr(LHS, RHS); 252 } 253 case Instruction::And: { 254 Constant *RHS = ExtractConstantBytes(CE->getOperand(1), ByteStart,ByteSize); 255 if (!RHS) 256 return nullptr; 257 258 // X & 0 -> 0. 259 if (RHS->isNullValue()) 260 return RHS; 261 262 Constant *LHS = ExtractConstantBytes(CE->getOperand(0), ByteStart,ByteSize); 263 if (!LHS) 264 return nullptr; 265 return ConstantExpr::getAnd(LHS, RHS); 266 } 267 case Instruction::LShr: { 268 ConstantInt *Amt = dyn_cast<ConstantInt>(CE->getOperand(1)); 269 if (!Amt) 270 return nullptr; 271 APInt ShAmt = Amt->getValue(); 272 // Cannot analyze non-byte shifts. 273 if ((ShAmt & 7) != 0) 274 return nullptr; 275 ShAmt.lshrInPlace(3); 276 277 // If the extract is known to be all zeros, return zero. 278 if (ShAmt.uge(CSize - ByteStart)) 279 return Constant::getNullValue( 280 IntegerType::get(CE->getContext(), ByteSize * 8)); 281 // If the extract is known to be fully in the input, extract it. 282 if (ShAmt.ule(CSize - (ByteStart + ByteSize))) 283 return ExtractConstantBytes(CE->getOperand(0), 284 ByteStart + ShAmt.getZExtValue(), ByteSize); 285 286 // TODO: Handle the 'partially zero' case. 287 return nullptr; 288 } 289 290 case Instruction::Shl: { 291 ConstantInt *Amt = dyn_cast<ConstantInt>(CE->getOperand(1)); 292 if (!Amt) 293 return nullptr; 294 APInt ShAmt = Amt->getValue(); 295 // Cannot analyze non-byte shifts. 296 if ((ShAmt & 7) != 0) 297 return nullptr; 298 ShAmt.lshrInPlace(3); 299 300 // If the extract is known to be all zeros, return zero. 301 if (ShAmt.uge(ByteStart + ByteSize)) 302 return Constant::getNullValue( 303 IntegerType::get(CE->getContext(), ByteSize * 8)); 304 // If the extract is known to be fully in the input, extract it. 305 if (ShAmt.ule(ByteStart)) 306 return ExtractConstantBytes(CE->getOperand(0), 307 ByteStart - ShAmt.getZExtValue(), ByteSize); 308 309 // TODO: Handle the 'partially zero' case. 310 return nullptr; 311 } 312 313 case Instruction::ZExt: { 314 unsigned SrcBitSize = 315 cast<IntegerType>(CE->getOperand(0)->getType())->getBitWidth(); 316 317 // If extracting something that is completely zero, return 0. 318 if (ByteStart*8 >= SrcBitSize) 319 return Constant::getNullValue(IntegerType::get(CE->getContext(), 320 ByteSize*8)); 321 322 // If exactly extracting the input, return it. 323 if (ByteStart == 0 && ByteSize*8 == SrcBitSize) 324 return CE->getOperand(0); 325 326 // If extracting something completely in the input, if the input is a 327 // multiple of 8 bits, recurse. 328 if ((SrcBitSize&7) == 0 && (ByteStart+ByteSize)*8 <= SrcBitSize) 329 return ExtractConstantBytes(CE->getOperand(0), ByteStart, ByteSize); 330 331 // Otherwise, if extracting a subset of the input, which is not multiple of 332 // 8 bits, do a shift and trunc to get the bits. 333 if ((ByteStart+ByteSize)*8 < SrcBitSize) { 334 assert((SrcBitSize&7) && "Shouldn't get byte sized case here"); 335 Constant *Res = CE->getOperand(0); 336 if (ByteStart) 337 Res = ConstantExpr::getLShr(Res, 338 ConstantInt::get(Res->getType(), ByteStart*8)); 339 return ConstantExpr::getTrunc(Res, IntegerType::get(C->getContext(), 340 ByteSize*8)); 341 } 342 343 // TODO: Handle the 'partially zero' case. 344 return nullptr; 345 } 346 } 347 } 348 349 /// Return a ConstantExpr with type DestTy for sizeof on Ty, with any known 350 /// factors factored out. If Folded is false, return null if no factoring was 351 /// possible, to avoid endlessly bouncing an unfoldable expression back into the 352 /// top-level folder. 353 static Constant *getFoldedSizeOf(Type *Ty, Type *DestTy, bool Folded) { 354 if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 355 Constant *N = ConstantInt::get(DestTy, ATy->getNumElements()); 356 Constant *E = getFoldedSizeOf(ATy->getElementType(), DestTy, true); 357 return ConstantExpr::getNUWMul(E, N); 358 } 359 360 if (StructType *STy = dyn_cast<StructType>(Ty)) 361 if (!STy->isPacked()) { 362 unsigned NumElems = STy->getNumElements(); 363 // An empty struct has size zero. 364 if (NumElems == 0) 365 return ConstantExpr::getNullValue(DestTy); 366 // Check for a struct with all members having the same size. 367 Constant *MemberSize = 368 getFoldedSizeOf(STy->getElementType(0), DestTy, true); 369 bool AllSame = true; 370 for (unsigned i = 1; i != NumElems; ++i) 371 if (MemberSize != 372 getFoldedSizeOf(STy->getElementType(i), DestTy, true)) { 373 AllSame = false; 374 break; 375 } 376 if (AllSame) { 377 Constant *N = ConstantInt::get(DestTy, NumElems); 378 return ConstantExpr::getNUWMul(MemberSize, N); 379 } 380 } 381 382 // Pointer size doesn't depend on the pointee type, so canonicalize them 383 // to an arbitrary pointee. 384 if (PointerType *PTy = dyn_cast<PointerType>(Ty)) 385 if (!PTy->getElementType()->isIntegerTy(1)) 386 return 387 getFoldedSizeOf(PointerType::get(IntegerType::get(PTy->getContext(), 1), 388 PTy->getAddressSpace()), 389 DestTy, true); 390 391 // If there's no interesting folding happening, bail so that we don't create 392 // a constant that looks like it needs folding but really doesn't. 393 if (!Folded) 394 return nullptr; 395 396 // Base case: Get a regular sizeof expression. 397 Constant *C = ConstantExpr::getSizeOf(Ty); 398 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 399 DestTy, false), 400 C, DestTy); 401 return C; 402 } 403 404 /// Return a ConstantExpr with type DestTy for alignof on Ty, with any known 405 /// factors factored out. If Folded is false, return null if no factoring was 406 /// possible, to avoid endlessly bouncing an unfoldable expression back into the 407 /// top-level folder. 408 static Constant *getFoldedAlignOf(Type *Ty, Type *DestTy, bool Folded) { 409 // The alignment of an array is equal to the alignment of the 410 // array element. Note that this is not always true for vectors. 411 if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 412 Constant *C = ConstantExpr::getAlignOf(ATy->getElementType()); 413 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 414 DestTy, 415 false), 416 C, DestTy); 417 return C; 418 } 419 420 if (StructType *STy = dyn_cast<StructType>(Ty)) { 421 // Packed structs always have an alignment of 1. 422 if (STy->isPacked()) 423 return ConstantInt::get(DestTy, 1); 424 425 // Otherwise, struct alignment is the maximum alignment of any member. 426 // Without target data, we can't compare much, but we can check to see 427 // if all the members have the same alignment. 428 unsigned NumElems = STy->getNumElements(); 429 // An empty struct has minimal alignment. 430 if (NumElems == 0) 431 return ConstantInt::get(DestTy, 1); 432 // Check for a struct with all members having the same alignment. 433 Constant *MemberAlign = 434 getFoldedAlignOf(STy->getElementType(0), DestTy, true); 435 bool AllSame = true; 436 for (unsigned i = 1; i != NumElems; ++i) 437 if (MemberAlign != getFoldedAlignOf(STy->getElementType(i), DestTy, true)) { 438 AllSame = false; 439 break; 440 } 441 if (AllSame) 442 return MemberAlign; 443 } 444 445 // Pointer alignment doesn't depend on the pointee type, so canonicalize them 446 // to an arbitrary pointee. 447 if (PointerType *PTy = dyn_cast<PointerType>(Ty)) 448 if (!PTy->getElementType()->isIntegerTy(1)) 449 return 450 getFoldedAlignOf(PointerType::get(IntegerType::get(PTy->getContext(), 451 1), 452 PTy->getAddressSpace()), 453 DestTy, true); 454 455 // If there's no interesting folding happening, bail so that we don't create 456 // a constant that looks like it needs folding but really doesn't. 457 if (!Folded) 458 return nullptr; 459 460 // Base case: Get a regular alignof expression. 461 Constant *C = ConstantExpr::getAlignOf(Ty); 462 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 463 DestTy, false), 464 C, DestTy); 465 return C; 466 } 467 468 /// Return a ConstantExpr with type DestTy for offsetof on Ty and FieldNo, with 469 /// any known factors factored out. If Folded is false, return null if no 470 /// factoring was possible, to avoid endlessly bouncing an unfoldable expression 471 /// back into the top-level folder. 472 static Constant *getFoldedOffsetOf(Type *Ty, Constant *FieldNo, Type *DestTy, 473 bool Folded) { 474 if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 475 Constant *N = ConstantExpr::getCast(CastInst::getCastOpcode(FieldNo, false, 476 DestTy, false), 477 FieldNo, DestTy); 478 Constant *E = getFoldedSizeOf(ATy->getElementType(), DestTy, true); 479 return ConstantExpr::getNUWMul(E, N); 480 } 481 482 if (StructType *STy = dyn_cast<StructType>(Ty)) 483 if (!STy->isPacked()) { 484 unsigned NumElems = STy->getNumElements(); 485 // An empty struct has no members. 486 if (NumElems == 0) 487 return nullptr; 488 // Check for a struct with all members having the same size. 489 Constant *MemberSize = 490 getFoldedSizeOf(STy->getElementType(0), DestTy, true); 491 bool AllSame = true; 492 for (unsigned i = 1; i != NumElems; ++i) 493 if (MemberSize != 494 getFoldedSizeOf(STy->getElementType(i), DestTy, true)) { 495 AllSame = false; 496 break; 497 } 498 if (AllSame) { 499 Constant *N = ConstantExpr::getCast(CastInst::getCastOpcode(FieldNo, 500 false, 501 DestTy, 502 false), 503 FieldNo, DestTy); 504 return ConstantExpr::getNUWMul(MemberSize, N); 505 } 506 } 507 508 // If there's no interesting folding happening, bail so that we don't create 509 // a constant that looks like it needs folding but really doesn't. 510 if (!Folded) 511 return nullptr; 512 513 // Base case: Get a regular offsetof expression. 514 Constant *C = ConstantExpr::getOffsetOf(Ty, FieldNo); 515 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 516 DestTy, false), 517 C, DestTy); 518 return C; 519 } 520 521 Constant *llvm::ConstantFoldCastInstruction(unsigned opc, Constant *V, 522 Type *DestTy) { 523 if (isa<UndefValue>(V)) { 524 // zext(undef) = 0, because the top bits will be zero. 525 // sext(undef) = 0, because the top bits will all be the same. 526 // [us]itofp(undef) = 0, because the result value is bounded. 527 if (opc == Instruction::ZExt || opc == Instruction::SExt || 528 opc == Instruction::UIToFP || opc == Instruction::SIToFP) 529 return Constant::getNullValue(DestTy); 530 return UndefValue::get(DestTy); 531 } 532 533 if (V->isNullValue() && !DestTy->isX86_MMXTy() && 534 opc != Instruction::AddrSpaceCast) 535 return Constant::getNullValue(DestTy); 536 537 // If the cast operand is a constant expression, there's a few things we can 538 // do to try to simplify it. 539 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) { 540 if (CE->isCast()) { 541 // Try hard to fold cast of cast because they are often eliminable. 542 if (unsigned newOpc = foldConstantCastPair(opc, CE, DestTy)) 543 return ConstantExpr::getCast(newOpc, CE->getOperand(0), DestTy); 544 } else if (CE->getOpcode() == Instruction::GetElementPtr && 545 // Do not fold addrspacecast (gep 0, .., 0). It might make the 546 // addrspacecast uncanonicalized. 547 opc != Instruction::AddrSpaceCast && 548 // Do not fold bitcast (gep) with inrange index, as this loses 549 // information. 550 !cast<GEPOperator>(CE)->getInRangeIndex().hasValue() && 551 // Do not fold if the gep type is a vector, as bitcasting 552 // operand 0 of a vector gep will result in a bitcast between 553 // different sizes. 554 !CE->getType()->isVectorTy()) { 555 // If all of the indexes in the GEP are null values, there is no pointer 556 // adjustment going on. We might as well cast the source pointer. 557 bool isAllNull = true; 558 for (unsigned i = 1, e = CE->getNumOperands(); i != e; ++i) 559 if (!CE->getOperand(i)->isNullValue()) { 560 isAllNull = false; 561 break; 562 } 563 if (isAllNull) 564 // This is casting one pointer type to another, always BitCast 565 return ConstantExpr::getPointerCast(CE->getOperand(0), DestTy); 566 } 567 } 568 569 // If the cast operand is a constant vector, perform the cast by 570 // operating on each element. In the cast of bitcasts, the element 571 // count may be mismatched; don't attempt to handle that here. 572 if ((isa<ConstantVector>(V) || isa<ConstantDataVector>(V)) && 573 DestTy->isVectorTy() && 574 DestTy->getVectorNumElements() == V->getType()->getVectorNumElements()) { 575 SmallVector<Constant*, 16> res; 576 VectorType *DestVecTy = cast<VectorType>(DestTy); 577 Type *DstEltTy = DestVecTy->getElementType(); 578 Type *Ty = IntegerType::get(V->getContext(), 32); 579 for (unsigned i = 0, e = V->getType()->getVectorNumElements(); i != e; ++i) { 580 Constant *C = 581 ConstantExpr::getExtractElement(V, ConstantInt::get(Ty, i)); 582 res.push_back(ConstantExpr::getCast(opc, C, DstEltTy)); 583 } 584 return ConstantVector::get(res); 585 } 586 587 // We actually have to do a cast now. Perform the cast according to the 588 // opcode specified. 589 switch (opc) { 590 default: 591 llvm_unreachable("Failed to cast constant expression"); 592 case Instruction::FPTrunc: 593 case Instruction::FPExt: 594 if (ConstantFP *FPC = dyn_cast<ConstantFP>(V)) { 595 bool ignored; 596 APFloat Val = FPC->getValueAPF(); 597 Val.convert(DestTy->isHalfTy() ? APFloat::IEEEhalf() : 598 DestTy->isFloatTy() ? APFloat::IEEEsingle() : 599 DestTy->isDoubleTy() ? APFloat::IEEEdouble() : 600 DestTy->isX86_FP80Ty() ? APFloat::x87DoubleExtended() : 601 DestTy->isFP128Ty() ? APFloat::IEEEquad() : 602 DestTy->isPPC_FP128Ty() ? APFloat::PPCDoubleDouble() : 603 APFloat::Bogus(), 604 APFloat::rmNearestTiesToEven, &ignored); 605 return ConstantFP::get(V->getContext(), Val); 606 } 607 return nullptr; // Can't fold. 608 case Instruction::FPToUI: 609 case Instruction::FPToSI: 610 if (ConstantFP *FPC = dyn_cast<ConstantFP>(V)) { 611 const APFloat &V = FPC->getValueAPF(); 612 bool ignored; 613 uint32_t DestBitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 614 APSInt IntVal(DestBitWidth, opc == Instruction::FPToUI); 615 if (APFloat::opInvalidOp == 616 V.convertToInteger(IntVal, APFloat::rmTowardZero, &ignored)) { 617 // Undefined behavior invoked - the destination type can't represent 618 // the input constant. 619 return UndefValue::get(DestTy); 620 } 621 return ConstantInt::get(FPC->getContext(), IntVal); 622 } 623 return nullptr; // Can't fold. 624 case Instruction::IntToPtr: //always treated as unsigned 625 if (V->isNullValue()) // Is it an integral null value? 626 return ConstantPointerNull::get(cast<PointerType>(DestTy)); 627 return nullptr; // Other pointer types cannot be casted 628 case Instruction::PtrToInt: // always treated as unsigned 629 // Is it a null pointer value? 630 if (V->isNullValue()) 631 return ConstantInt::get(DestTy, 0); 632 // If this is a sizeof-like expression, pull out multiplications by 633 // known factors to expose them to subsequent folding. If it's an 634 // alignof-like expression, factor out known factors. 635 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) 636 if (CE->getOpcode() == Instruction::GetElementPtr && 637 CE->getOperand(0)->isNullValue()) { 638 // FIXME: Looks like getFoldedSizeOf(), getFoldedOffsetOf() and 639 // getFoldedAlignOf() don't handle the case when DestTy is a vector of 640 // pointers yet. We end up in asserts in CastInst::getCastOpcode (see 641 // test/Analysis/ConstantFolding/cast-vector.ll). I've only seen this 642 // happen in one "real" C-code test case, so it does not seem to be an 643 // important optimization to handle vectors here. For now, simply bail 644 // out. 645 if (DestTy->isVectorTy()) 646 return nullptr; 647 GEPOperator *GEPO = cast<GEPOperator>(CE); 648 Type *Ty = GEPO->getSourceElementType(); 649 if (CE->getNumOperands() == 2) { 650 // Handle a sizeof-like expression. 651 Constant *Idx = CE->getOperand(1); 652 bool isOne = isa<ConstantInt>(Idx) && cast<ConstantInt>(Idx)->isOne(); 653 if (Constant *C = getFoldedSizeOf(Ty, DestTy, !isOne)) { 654 Idx = ConstantExpr::getCast(CastInst::getCastOpcode(Idx, true, 655 DestTy, false), 656 Idx, DestTy); 657 return ConstantExpr::getMul(C, Idx); 658 } 659 } else if (CE->getNumOperands() == 3 && 660 CE->getOperand(1)->isNullValue()) { 661 // Handle an alignof-like expression. 662 if (StructType *STy = dyn_cast<StructType>(Ty)) 663 if (!STy->isPacked()) { 664 ConstantInt *CI = cast<ConstantInt>(CE->getOperand(2)); 665 if (CI->isOne() && 666 STy->getNumElements() == 2 && 667 STy->getElementType(0)->isIntegerTy(1)) { 668 return getFoldedAlignOf(STy->getElementType(1), DestTy, false); 669 } 670 } 671 // Handle an offsetof-like expression. 672 if (Ty->isStructTy() || Ty->isArrayTy()) { 673 if (Constant *C = getFoldedOffsetOf(Ty, CE->getOperand(2), 674 DestTy, false)) 675 return C; 676 } 677 } 678 } 679 // Other pointer types cannot be casted 680 return nullptr; 681 case Instruction::UIToFP: 682 case Instruction::SIToFP: 683 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 684 const APInt &api = CI->getValue(); 685 APFloat apf(DestTy->getFltSemantics(), 686 APInt::getNullValue(DestTy->getPrimitiveSizeInBits())); 687 apf.convertFromAPInt(api, opc==Instruction::SIToFP, 688 APFloat::rmNearestTiesToEven); 689 return ConstantFP::get(V->getContext(), apf); 690 } 691 return nullptr; 692 case Instruction::ZExt: 693 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 694 uint32_t BitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 695 return ConstantInt::get(V->getContext(), 696 CI->getValue().zext(BitWidth)); 697 } 698 return nullptr; 699 case Instruction::SExt: 700 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 701 uint32_t BitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 702 return ConstantInt::get(V->getContext(), 703 CI->getValue().sext(BitWidth)); 704 } 705 return nullptr; 706 case Instruction::Trunc: { 707 if (V->getType()->isVectorTy()) 708 return nullptr; 709 710 uint32_t DestBitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 711 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 712 return ConstantInt::get(V->getContext(), 713 CI->getValue().trunc(DestBitWidth)); 714 } 715 716 // The input must be a constantexpr. See if we can simplify this based on 717 // the bytes we are demanding. Only do this if the source and dest are an 718 // even multiple of a byte. 719 if ((DestBitWidth & 7) == 0 && 720 (cast<IntegerType>(V->getType())->getBitWidth() & 7) == 0) 721 if (Constant *Res = ExtractConstantBytes(V, 0, DestBitWidth / 8)) 722 return Res; 723 724 return nullptr; 725 } 726 case Instruction::BitCast: 727 return FoldBitCast(V, DestTy); 728 case Instruction::AddrSpaceCast: 729 return nullptr; 730 } 731 } 732 733 Constant *llvm::ConstantFoldSelectInstruction(Constant *Cond, 734 Constant *V1, Constant *V2) { 735 // Check for i1 and vector true/false conditions. 736 if (Cond->isNullValue()) return V2; 737 if (Cond->isAllOnesValue()) return V1; 738 739 // If the condition is a vector constant, fold the result elementwise. 740 if (ConstantVector *CondV = dyn_cast<ConstantVector>(Cond)) { 741 SmallVector<Constant*, 16> Result; 742 Type *Ty = IntegerType::get(CondV->getContext(), 32); 743 for (unsigned i = 0, e = V1->getType()->getVectorNumElements(); i != e;++i){ 744 Constant *V; 745 Constant *V1Element = ConstantExpr::getExtractElement(V1, 746 ConstantInt::get(Ty, i)); 747 Constant *V2Element = ConstantExpr::getExtractElement(V2, 748 ConstantInt::get(Ty, i)); 749 Constant *Cond = dyn_cast<Constant>(CondV->getOperand(i)); 750 if (V1Element == V2Element) { 751 V = V1Element; 752 } else if (isa<UndefValue>(Cond)) { 753 V = isa<UndefValue>(V1Element) ? V1Element : V2Element; 754 } else { 755 if (!isa<ConstantInt>(Cond)) break; 756 V = Cond->isNullValue() ? V2Element : V1Element; 757 } 758 Result.push_back(V); 759 } 760 761 // If we were able to build the vector, return it. 762 if (Result.size() == V1->getType()->getVectorNumElements()) 763 return ConstantVector::get(Result); 764 } 765 766 if (isa<UndefValue>(Cond)) { 767 if (isa<UndefValue>(V1)) return V1; 768 return V2; 769 } 770 if (isa<UndefValue>(V1)) return V2; 771 if (isa<UndefValue>(V2)) return V1; 772 if (V1 == V2) return V1; 773 774 if (ConstantExpr *TrueVal = dyn_cast<ConstantExpr>(V1)) { 775 if (TrueVal->getOpcode() == Instruction::Select) 776 if (TrueVal->getOperand(0) == Cond) 777 return ConstantExpr::getSelect(Cond, TrueVal->getOperand(1), V2); 778 } 779 if (ConstantExpr *FalseVal = dyn_cast<ConstantExpr>(V2)) { 780 if (FalseVal->getOpcode() == Instruction::Select) 781 if (FalseVal->getOperand(0) == Cond) 782 return ConstantExpr::getSelect(Cond, V1, FalseVal->getOperand(2)); 783 } 784 785 return nullptr; 786 } 787 788 Constant *llvm::ConstantFoldExtractElementInstruction(Constant *Val, 789 Constant *Idx) { 790 if (isa<UndefValue>(Val)) // ee(undef, x) -> undef 791 return UndefValue::get(Val->getType()->getVectorElementType()); 792 if (Val->isNullValue()) // ee(zero, x) -> zero 793 return Constant::getNullValue(Val->getType()->getVectorElementType()); 794 // ee({w,x,y,z}, undef) -> undef 795 if (isa<UndefValue>(Idx)) 796 return UndefValue::get(Val->getType()->getVectorElementType()); 797 798 if (ConstantInt *CIdx = dyn_cast<ConstantInt>(Idx)) { 799 // ee({w,x,y,z}, wrong_value) -> undef 800 if (CIdx->uge(Val->getType()->getVectorNumElements())) 801 return UndefValue::get(Val->getType()->getVectorElementType()); 802 return Val->getAggregateElement(CIdx->getZExtValue()); 803 } 804 return nullptr; 805 } 806 807 Constant *llvm::ConstantFoldInsertElementInstruction(Constant *Val, 808 Constant *Elt, 809 Constant *Idx) { 810 if (isa<UndefValue>(Idx)) 811 return UndefValue::get(Val->getType()); 812 813 ConstantInt *CIdx = dyn_cast<ConstantInt>(Idx); 814 if (!CIdx) return nullptr; 815 816 unsigned NumElts = Val->getType()->getVectorNumElements(); 817 if (CIdx->uge(NumElts)) 818 return UndefValue::get(Val->getType()); 819 820 SmallVector<Constant*, 16> Result; 821 Result.reserve(NumElts); 822 auto *Ty = Type::getInt32Ty(Val->getContext()); 823 uint64_t IdxVal = CIdx->getZExtValue(); 824 for (unsigned i = 0; i != NumElts; ++i) { 825 if (i == IdxVal) { 826 Result.push_back(Elt); 827 continue; 828 } 829 830 Constant *C = ConstantExpr::getExtractElement(Val, ConstantInt::get(Ty, i)); 831 Result.push_back(C); 832 } 833 834 return ConstantVector::get(Result); 835 } 836 837 Constant *llvm::ConstantFoldShuffleVectorInstruction(Constant *V1, 838 Constant *V2, 839 Constant *Mask) { 840 unsigned MaskNumElts = Mask->getType()->getVectorNumElements(); 841 Type *EltTy = V1->getType()->getVectorElementType(); 842 843 // Undefined shuffle mask -> undefined value. 844 if (isa<UndefValue>(Mask)) 845 return UndefValue::get(VectorType::get(EltTy, MaskNumElts)); 846 847 // Don't break the bitcode reader hack. 848 if (isa<ConstantExpr>(Mask)) return nullptr; 849 850 unsigned SrcNumElts = V1->getType()->getVectorNumElements(); 851 852 // Loop over the shuffle mask, evaluating each element. 853 SmallVector<Constant*, 32> Result; 854 for (unsigned i = 0; i != MaskNumElts; ++i) { 855 int Elt = ShuffleVectorInst::getMaskValue(Mask, i); 856 if (Elt == -1) { 857 Result.push_back(UndefValue::get(EltTy)); 858 continue; 859 } 860 Constant *InElt; 861 if (unsigned(Elt) >= SrcNumElts*2) 862 InElt = UndefValue::get(EltTy); 863 else if (unsigned(Elt) >= SrcNumElts) { 864 Type *Ty = IntegerType::get(V2->getContext(), 32); 865 InElt = 866 ConstantExpr::getExtractElement(V2, 867 ConstantInt::get(Ty, Elt - SrcNumElts)); 868 } else { 869 Type *Ty = IntegerType::get(V1->getContext(), 32); 870 InElt = ConstantExpr::getExtractElement(V1, ConstantInt::get(Ty, Elt)); 871 } 872 Result.push_back(InElt); 873 } 874 875 return ConstantVector::get(Result); 876 } 877 878 Constant *llvm::ConstantFoldExtractValueInstruction(Constant *Agg, 879 ArrayRef<unsigned> Idxs) { 880 // Base case: no indices, so return the entire value. 881 if (Idxs.empty()) 882 return Agg; 883 884 if (Constant *C = Agg->getAggregateElement(Idxs[0])) 885 return ConstantFoldExtractValueInstruction(C, Idxs.slice(1)); 886 887 return nullptr; 888 } 889 890 Constant *llvm::ConstantFoldInsertValueInstruction(Constant *Agg, 891 Constant *Val, 892 ArrayRef<unsigned> Idxs) { 893 // Base case: no indices, so replace the entire value. 894 if (Idxs.empty()) 895 return Val; 896 897 unsigned NumElts; 898 if (StructType *ST = dyn_cast<StructType>(Agg->getType())) 899 NumElts = ST->getNumElements(); 900 else 901 NumElts = cast<SequentialType>(Agg->getType())->getNumElements(); 902 903 SmallVector<Constant*, 32> Result; 904 for (unsigned i = 0; i != NumElts; ++i) { 905 Constant *C = Agg->getAggregateElement(i); 906 if (!C) return nullptr; 907 908 if (Idxs[0] == i) 909 C = ConstantFoldInsertValueInstruction(C, Val, Idxs.slice(1)); 910 911 Result.push_back(C); 912 } 913 914 if (StructType *ST = dyn_cast<StructType>(Agg->getType())) 915 return ConstantStruct::get(ST, Result); 916 if (ArrayType *AT = dyn_cast<ArrayType>(Agg->getType())) 917 return ConstantArray::get(AT, Result); 918 return ConstantVector::get(Result); 919 } 920 921 Constant *llvm::ConstantFoldUnaryInstruction(unsigned Opcode, Constant *C) { 922 assert(Instruction::isUnaryOp(Opcode) && "Non-unary instruction detected"); 923 924 // Handle scalar UndefValue. Vectors are always evaluated per element. 925 bool HasScalarUndef = !C->getType()->isVectorTy() && isa<UndefValue>(C); 926 927 if (HasScalarUndef) { 928 switch (static_cast<Instruction::UnaryOps>(Opcode)) { 929 case Instruction::FNeg: 930 return C; // -undef -> undef 931 case Instruction::UnaryOpsEnd: 932 llvm_unreachable("Invalid UnaryOp"); 933 } 934 } 935 936 // Constant should not be UndefValue, unless these are vector constants. 937 assert(!HasScalarUndef && "Unexpected UndefValue"); 938 // We only have FP UnaryOps right now. 939 assert(!isa<ConstantInt>(C) && "Unexpected Integer UnaryOp"); 940 941 if (ConstantFP *CFP = dyn_cast<ConstantFP>(C)) { 942 const APFloat &CV = CFP->getValueAPF(); 943 switch (Opcode) { 944 default: 945 break; 946 case Instruction::FNeg: 947 return ConstantFP::get(C->getContext(), neg(CV)); 948 } 949 } else if (VectorType *VTy = dyn_cast<VectorType>(C->getType())) { 950 // Fold each element and create a vector constant from those constants. 951 SmallVector<Constant*, 16> Result; 952 Type *Ty = IntegerType::get(VTy->getContext(), 32); 953 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 954 Constant *ExtractIdx = ConstantInt::get(Ty, i); 955 Constant *Elt = ConstantExpr::getExtractElement(C, ExtractIdx); 956 957 Result.push_back(ConstantExpr::get(Opcode, Elt)); 958 } 959 960 return ConstantVector::get(Result); 961 } 962 963 // We don't know how to fold this. 964 return nullptr; 965 } 966 967 Constant *llvm::ConstantFoldBinaryInstruction(unsigned Opcode, Constant *C1, 968 Constant *C2) { 969 assert(Instruction::isBinaryOp(Opcode) && "Non-binary instruction detected"); 970 971 // Handle scalar UndefValue. Vectors are always evaluated per element. 972 bool HasScalarUndef = !C1->getType()->isVectorTy() && 973 (isa<UndefValue>(C1) || isa<UndefValue>(C2)); 974 if (HasScalarUndef) { 975 switch (static_cast<Instruction::BinaryOps>(Opcode)) { 976 case Instruction::Xor: 977 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) 978 // Handle undef ^ undef -> 0 special case. This is a common 979 // idiom (misuse). 980 return Constant::getNullValue(C1->getType()); 981 LLVM_FALLTHROUGH; 982 case Instruction::Add: 983 case Instruction::Sub: 984 return UndefValue::get(C1->getType()); 985 case Instruction::And: 986 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) // undef & undef -> undef 987 return C1; 988 return Constant::getNullValue(C1->getType()); // undef & X -> 0 989 case Instruction::Mul: { 990 // undef * undef -> undef 991 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) 992 return C1; 993 const APInt *CV; 994 // X * undef -> undef if X is odd 995 if (match(C1, m_APInt(CV)) || match(C2, m_APInt(CV))) 996 if ((*CV)[0]) 997 return UndefValue::get(C1->getType()); 998 999 // X * undef -> 0 otherwise 1000 return Constant::getNullValue(C1->getType()); 1001 } 1002 case Instruction::SDiv: 1003 case Instruction::UDiv: 1004 // X / undef -> undef 1005 if (isa<UndefValue>(C2)) 1006 return C2; 1007 // undef / 0 -> undef 1008 // undef / 1 -> undef 1009 if (match(C2, m_Zero()) || match(C2, m_One())) 1010 return C1; 1011 // undef / X -> 0 otherwise 1012 return Constant::getNullValue(C1->getType()); 1013 case Instruction::URem: 1014 case Instruction::SRem: 1015 // X % undef -> undef 1016 if (match(C2, m_Undef())) 1017 return C2; 1018 // undef % 0 -> undef 1019 if (match(C2, m_Zero())) 1020 return C1; 1021 // undef % X -> 0 otherwise 1022 return Constant::getNullValue(C1->getType()); 1023 case Instruction::Or: // X | undef -> -1 1024 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) // undef | undef -> undef 1025 return C1; 1026 return Constant::getAllOnesValue(C1->getType()); // undef | X -> ~0 1027 case Instruction::LShr: 1028 // X >>l undef -> undef 1029 if (isa<UndefValue>(C2)) 1030 return C2; 1031 // undef >>l 0 -> undef 1032 if (match(C2, m_Zero())) 1033 return C1; 1034 // undef >>l X -> 0 1035 return Constant::getNullValue(C1->getType()); 1036 case Instruction::AShr: 1037 // X >>a undef -> undef 1038 if (isa<UndefValue>(C2)) 1039 return C2; 1040 // undef >>a 0 -> undef 1041 if (match(C2, m_Zero())) 1042 return C1; 1043 // TODO: undef >>a X -> undef if the shift is exact 1044 // undef >>a X -> 0 1045 return Constant::getNullValue(C1->getType()); 1046 case Instruction::Shl: 1047 // X << undef -> undef 1048 if (isa<UndefValue>(C2)) 1049 return C2; 1050 // undef << 0 -> undef 1051 if (match(C2, m_Zero())) 1052 return C1; 1053 // undef << X -> 0 1054 return Constant::getNullValue(C1->getType()); 1055 case Instruction::FAdd: 1056 case Instruction::FSub: 1057 case Instruction::FMul: 1058 case Instruction::FDiv: 1059 case Instruction::FRem: 1060 // [any flop] undef, undef -> undef 1061 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) 1062 return C1; 1063 // [any flop] C, undef -> NaN 1064 // [any flop] undef, C -> NaN 1065 // We could potentially specialize NaN/Inf constants vs. 'normal' 1066 // constants (possibly differently depending on opcode and operand). This 1067 // would allow returning undef sometimes. But it is always safe to fold to 1068 // NaN because we can choose the undef operand as NaN, and any FP opcode 1069 // with a NaN operand will propagate NaN. 1070 return ConstantFP::getNaN(C1->getType()); 1071 case Instruction::BinaryOpsEnd: 1072 llvm_unreachable("Invalid BinaryOp"); 1073 } 1074 } 1075 1076 // Neither constant should be UndefValue, unless these are vector constants. 1077 assert(!HasScalarUndef && "Unexpected UndefValue"); 1078 1079 // Handle simplifications when the RHS is a constant int. 1080 if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) { 1081 switch (Opcode) { 1082 case Instruction::Add: 1083 if (CI2->isZero()) return C1; // X + 0 == X 1084 break; 1085 case Instruction::Sub: 1086 if (CI2->isZero()) return C1; // X - 0 == X 1087 break; 1088 case Instruction::Mul: 1089 if (CI2->isZero()) return C2; // X * 0 == 0 1090 if (CI2->isOne()) 1091 return C1; // X * 1 == X 1092 break; 1093 case Instruction::UDiv: 1094 case Instruction::SDiv: 1095 if (CI2->isOne()) 1096 return C1; // X / 1 == X 1097 if (CI2->isZero()) 1098 return UndefValue::get(CI2->getType()); // X / 0 == undef 1099 break; 1100 case Instruction::URem: 1101 case Instruction::SRem: 1102 if (CI2->isOne()) 1103 return Constant::getNullValue(CI2->getType()); // X % 1 == 0 1104 if (CI2->isZero()) 1105 return UndefValue::get(CI2->getType()); // X % 0 == undef 1106 break; 1107 case Instruction::And: 1108 if (CI2->isZero()) return C2; // X & 0 == 0 1109 if (CI2->isMinusOne()) 1110 return C1; // X & -1 == X 1111 1112 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1113 // (zext i32 to i64) & 4294967295 -> (zext i32 to i64) 1114 if (CE1->getOpcode() == Instruction::ZExt) { 1115 unsigned DstWidth = CI2->getType()->getBitWidth(); 1116 unsigned SrcWidth = 1117 CE1->getOperand(0)->getType()->getPrimitiveSizeInBits(); 1118 APInt PossiblySetBits(APInt::getLowBitsSet(DstWidth, SrcWidth)); 1119 if ((PossiblySetBits & CI2->getValue()) == PossiblySetBits) 1120 return C1; 1121 } 1122 1123 // If and'ing the address of a global with a constant, fold it. 1124 if (CE1->getOpcode() == Instruction::PtrToInt && 1125 isa<GlobalValue>(CE1->getOperand(0))) { 1126 GlobalValue *GV = cast<GlobalValue>(CE1->getOperand(0)); 1127 1128 unsigned GVAlign; 1129 1130 if (Module *TheModule = GV->getParent()) { 1131 GVAlign = GV->getPointerAlignment(TheModule->getDataLayout()); 1132 1133 // If the function alignment is not specified then assume that it 1134 // is 4. 1135 // This is dangerous; on x86, the alignment of the pointer 1136 // corresponds to the alignment of the function, but might be less 1137 // than 4 if it isn't explicitly specified. 1138 // However, a fix for this behaviour was reverted because it 1139 // increased code size (see https://reviews.llvm.org/D55115) 1140 // FIXME: This code should be deleted once existing targets have 1141 // appropriate defaults 1142 if (GVAlign == 0U && isa<Function>(GV)) 1143 GVAlign = 4U; 1144 } else if (isa<Function>(GV)) { 1145 // Without a datalayout we have to assume the worst case: that the 1146 // function pointer isn't aligned at all. 1147 GVAlign = 0U; 1148 } else { 1149 GVAlign = GV->getAlignment(); 1150 } 1151 1152 if (GVAlign > 1) { 1153 unsigned DstWidth = CI2->getType()->getBitWidth(); 1154 unsigned SrcWidth = std::min(DstWidth, Log2_32(GVAlign)); 1155 APInt BitsNotSet(APInt::getLowBitsSet(DstWidth, SrcWidth)); 1156 1157 // If checking bits we know are clear, return zero. 1158 if ((CI2->getValue() & BitsNotSet) == CI2->getValue()) 1159 return Constant::getNullValue(CI2->getType()); 1160 } 1161 } 1162 } 1163 break; 1164 case Instruction::Or: 1165 if (CI2->isZero()) return C1; // X | 0 == X 1166 if (CI2->isMinusOne()) 1167 return C2; // X | -1 == -1 1168 break; 1169 case Instruction::Xor: 1170 if (CI2->isZero()) return C1; // X ^ 0 == X 1171 1172 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1173 switch (CE1->getOpcode()) { 1174 default: break; 1175 case Instruction::ICmp: 1176 case Instruction::FCmp: 1177 // cmp pred ^ true -> cmp !pred 1178 assert(CI2->isOne()); 1179 CmpInst::Predicate pred = (CmpInst::Predicate)CE1->getPredicate(); 1180 pred = CmpInst::getInversePredicate(pred); 1181 return ConstantExpr::getCompare(pred, CE1->getOperand(0), 1182 CE1->getOperand(1)); 1183 } 1184 } 1185 break; 1186 case Instruction::AShr: 1187 // ashr (zext C to Ty), C2 -> lshr (zext C, CSA), C2 1188 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) 1189 if (CE1->getOpcode() == Instruction::ZExt) // Top bits known zero. 1190 return ConstantExpr::getLShr(C1, C2); 1191 break; 1192 } 1193 } else if (isa<ConstantInt>(C1)) { 1194 // If C1 is a ConstantInt and C2 is not, swap the operands. 1195 if (Instruction::isCommutative(Opcode)) 1196 return ConstantExpr::get(Opcode, C2, C1); 1197 } 1198 1199 if (ConstantInt *CI1 = dyn_cast<ConstantInt>(C1)) { 1200 if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) { 1201 const APInt &C1V = CI1->getValue(); 1202 const APInt &C2V = CI2->getValue(); 1203 switch (Opcode) { 1204 default: 1205 break; 1206 case Instruction::Add: 1207 return ConstantInt::get(CI1->getContext(), C1V + C2V); 1208 case Instruction::Sub: 1209 return ConstantInt::get(CI1->getContext(), C1V - C2V); 1210 case Instruction::Mul: 1211 return ConstantInt::get(CI1->getContext(), C1V * C2V); 1212 case Instruction::UDiv: 1213 assert(!CI2->isZero() && "Div by zero handled above"); 1214 return ConstantInt::get(CI1->getContext(), C1V.udiv(C2V)); 1215 case Instruction::SDiv: 1216 assert(!CI2->isZero() && "Div by zero handled above"); 1217 if (C2V.isAllOnesValue() && C1V.isMinSignedValue()) 1218 return UndefValue::get(CI1->getType()); // MIN_INT / -1 -> undef 1219 return ConstantInt::get(CI1->getContext(), C1V.sdiv(C2V)); 1220 case Instruction::URem: 1221 assert(!CI2->isZero() && "Div by zero handled above"); 1222 return ConstantInt::get(CI1->getContext(), C1V.urem(C2V)); 1223 case Instruction::SRem: 1224 assert(!CI2->isZero() && "Div by zero handled above"); 1225 if (C2V.isAllOnesValue() && C1V.isMinSignedValue()) 1226 return UndefValue::get(CI1->getType()); // MIN_INT % -1 -> undef 1227 return ConstantInt::get(CI1->getContext(), C1V.srem(C2V)); 1228 case Instruction::And: 1229 return ConstantInt::get(CI1->getContext(), C1V & C2V); 1230 case Instruction::Or: 1231 return ConstantInt::get(CI1->getContext(), C1V | C2V); 1232 case Instruction::Xor: 1233 return ConstantInt::get(CI1->getContext(), C1V ^ C2V); 1234 case Instruction::Shl: 1235 if (C2V.ult(C1V.getBitWidth())) 1236 return ConstantInt::get(CI1->getContext(), C1V.shl(C2V)); 1237 return UndefValue::get(C1->getType()); // too big shift is undef 1238 case Instruction::LShr: 1239 if (C2V.ult(C1V.getBitWidth())) 1240 return ConstantInt::get(CI1->getContext(), C1V.lshr(C2V)); 1241 return UndefValue::get(C1->getType()); // too big shift is undef 1242 case Instruction::AShr: 1243 if (C2V.ult(C1V.getBitWidth())) 1244 return ConstantInt::get(CI1->getContext(), C1V.ashr(C2V)); 1245 return UndefValue::get(C1->getType()); // too big shift is undef 1246 } 1247 } 1248 1249 switch (Opcode) { 1250 case Instruction::SDiv: 1251 case Instruction::UDiv: 1252 case Instruction::URem: 1253 case Instruction::SRem: 1254 case Instruction::LShr: 1255 case Instruction::AShr: 1256 case Instruction::Shl: 1257 if (CI1->isZero()) return C1; 1258 break; 1259 default: 1260 break; 1261 } 1262 } else if (ConstantFP *CFP1 = dyn_cast<ConstantFP>(C1)) { 1263 if (ConstantFP *CFP2 = dyn_cast<ConstantFP>(C2)) { 1264 const APFloat &C1V = CFP1->getValueAPF(); 1265 const APFloat &C2V = CFP2->getValueAPF(); 1266 APFloat C3V = C1V; // copy for modification 1267 switch (Opcode) { 1268 default: 1269 break; 1270 case Instruction::FAdd: 1271 (void)C3V.add(C2V, APFloat::rmNearestTiesToEven); 1272 return ConstantFP::get(C1->getContext(), C3V); 1273 case Instruction::FSub: 1274 (void)C3V.subtract(C2V, APFloat::rmNearestTiesToEven); 1275 return ConstantFP::get(C1->getContext(), C3V); 1276 case Instruction::FMul: 1277 (void)C3V.multiply(C2V, APFloat::rmNearestTiesToEven); 1278 return ConstantFP::get(C1->getContext(), C3V); 1279 case Instruction::FDiv: 1280 (void)C3V.divide(C2V, APFloat::rmNearestTiesToEven); 1281 return ConstantFP::get(C1->getContext(), C3V); 1282 case Instruction::FRem: 1283 (void)C3V.mod(C2V); 1284 return ConstantFP::get(C1->getContext(), C3V); 1285 } 1286 } 1287 } else if (VectorType *VTy = dyn_cast<VectorType>(C1->getType())) { 1288 // Fold each element and create a vector constant from those constants. 1289 SmallVector<Constant*, 16> Result; 1290 Type *Ty = IntegerType::get(VTy->getContext(), 32); 1291 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1292 Constant *ExtractIdx = ConstantInt::get(Ty, i); 1293 Constant *LHS = ConstantExpr::getExtractElement(C1, ExtractIdx); 1294 Constant *RHS = ConstantExpr::getExtractElement(C2, ExtractIdx); 1295 1296 // If any element of a divisor vector is zero, the whole op is undef. 1297 if (Instruction::isIntDivRem(Opcode) && RHS->isNullValue()) 1298 return UndefValue::get(VTy); 1299 1300 Result.push_back(ConstantExpr::get(Opcode, LHS, RHS)); 1301 } 1302 1303 return ConstantVector::get(Result); 1304 } 1305 1306 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1307 // There are many possible foldings we could do here. We should probably 1308 // at least fold add of a pointer with an integer into the appropriate 1309 // getelementptr. This will improve alias analysis a bit. 1310 1311 // Given ((a + b) + c), if (b + c) folds to something interesting, return 1312 // (a + (b + c)). 1313 if (Instruction::isAssociative(Opcode) && CE1->getOpcode() == Opcode) { 1314 Constant *T = ConstantExpr::get(Opcode, CE1->getOperand(1), C2); 1315 if (!isa<ConstantExpr>(T) || cast<ConstantExpr>(T)->getOpcode() != Opcode) 1316 return ConstantExpr::get(Opcode, CE1->getOperand(0), T); 1317 } 1318 } else if (isa<ConstantExpr>(C2)) { 1319 // If C2 is a constant expr and C1 isn't, flop them around and fold the 1320 // other way if possible. 1321 if (Instruction::isCommutative(Opcode)) 1322 return ConstantFoldBinaryInstruction(Opcode, C2, C1); 1323 } 1324 1325 // i1 can be simplified in many cases. 1326 if (C1->getType()->isIntegerTy(1)) { 1327 switch (Opcode) { 1328 case Instruction::Add: 1329 case Instruction::Sub: 1330 return ConstantExpr::getXor(C1, C2); 1331 case Instruction::Mul: 1332 return ConstantExpr::getAnd(C1, C2); 1333 case Instruction::Shl: 1334 case Instruction::LShr: 1335 case Instruction::AShr: 1336 // We can assume that C2 == 0. If it were one the result would be 1337 // undefined because the shift value is as large as the bitwidth. 1338 return C1; 1339 case Instruction::SDiv: 1340 case Instruction::UDiv: 1341 // We can assume that C2 == 1. If it were zero the result would be 1342 // undefined through division by zero. 1343 return C1; 1344 case Instruction::URem: 1345 case Instruction::SRem: 1346 // We can assume that C2 == 1. If it were zero the result would be 1347 // undefined through division by zero. 1348 return ConstantInt::getFalse(C1->getContext()); 1349 default: 1350 break; 1351 } 1352 } 1353 1354 // We don't know how to fold this. 1355 return nullptr; 1356 } 1357 1358 /// This type is zero-sized if it's an array or structure of zero-sized types. 1359 /// The only leaf zero-sized type is an empty structure. 1360 static bool isMaybeZeroSizedType(Type *Ty) { 1361 if (StructType *STy = dyn_cast<StructType>(Ty)) { 1362 if (STy->isOpaque()) return true; // Can't say. 1363 1364 // If all of elements have zero size, this does too. 1365 for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) 1366 if (!isMaybeZeroSizedType(STy->getElementType(i))) return false; 1367 return true; 1368 1369 } else if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 1370 return isMaybeZeroSizedType(ATy->getElementType()); 1371 } 1372 return false; 1373 } 1374 1375 /// Compare the two constants as though they were getelementptr indices. 1376 /// This allows coercion of the types to be the same thing. 1377 /// 1378 /// If the two constants are the "same" (after coercion), return 0. If the 1379 /// first is less than the second, return -1, if the second is less than the 1380 /// first, return 1. If the constants are not integral, return -2. 1381 /// 1382 static int IdxCompare(Constant *C1, Constant *C2, Type *ElTy) { 1383 if (C1 == C2) return 0; 1384 1385 // Ok, we found a different index. If they are not ConstantInt, we can't do 1386 // anything with them. 1387 if (!isa<ConstantInt>(C1) || !isa<ConstantInt>(C2)) 1388 return -2; // don't know! 1389 1390 // We cannot compare the indices if they don't fit in an int64_t. 1391 if (cast<ConstantInt>(C1)->getValue().getActiveBits() > 64 || 1392 cast<ConstantInt>(C2)->getValue().getActiveBits() > 64) 1393 return -2; // don't know! 1394 1395 // Ok, we have two differing integer indices. Sign extend them to be the same 1396 // type. 1397 int64_t C1Val = cast<ConstantInt>(C1)->getSExtValue(); 1398 int64_t C2Val = cast<ConstantInt>(C2)->getSExtValue(); 1399 1400 if (C1Val == C2Val) return 0; // They are equal 1401 1402 // If the type being indexed over is really just a zero sized type, there is 1403 // no pointer difference being made here. 1404 if (isMaybeZeroSizedType(ElTy)) 1405 return -2; // dunno. 1406 1407 // If they are really different, now that they are the same type, then we 1408 // found a difference! 1409 if (C1Val < C2Val) 1410 return -1; 1411 else 1412 return 1; 1413 } 1414 1415 /// This function determines if there is anything we can decide about the two 1416 /// constants provided. This doesn't need to handle simple things like 1417 /// ConstantFP comparisons, but should instead handle ConstantExprs. 1418 /// If we can determine that the two constants have a particular relation to 1419 /// each other, we should return the corresponding FCmpInst predicate, 1420 /// otherwise return FCmpInst::BAD_FCMP_PREDICATE. This is used below in 1421 /// ConstantFoldCompareInstruction. 1422 /// 1423 /// To simplify this code we canonicalize the relation so that the first 1424 /// operand is always the most "complex" of the two. We consider ConstantFP 1425 /// to be the simplest, and ConstantExprs to be the most complex. 1426 static FCmpInst::Predicate evaluateFCmpRelation(Constant *V1, Constant *V2) { 1427 assert(V1->getType() == V2->getType() && 1428 "Cannot compare values of different types!"); 1429 1430 // We do not know if a constant expression will evaluate to a number or NaN. 1431 // Therefore, we can only say that the relation is unordered or equal. 1432 if (V1 == V2) return FCmpInst::FCMP_UEQ; 1433 1434 if (!isa<ConstantExpr>(V1)) { 1435 if (!isa<ConstantExpr>(V2)) { 1436 // Simple case, use the standard constant folder. 1437 ConstantInt *R = nullptr; 1438 R = dyn_cast<ConstantInt>( 1439 ConstantExpr::getFCmp(FCmpInst::FCMP_OEQ, V1, V2)); 1440 if (R && !R->isZero()) 1441 return FCmpInst::FCMP_OEQ; 1442 R = dyn_cast<ConstantInt>( 1443 ConstantExpr::getFCmp(FCmpInst::FCMP_OLT, V1, V2)); 1444 if (R && !R->isZero()) 1445 return FCmpInst::FCMP_OLT; 1446 R = dyn_cast<ConstantInt>( 1447 ConstantExpr::getFCmp(FCmpInst::FCMP_OGT, V1, V2)); 1448 if (R && !R->isZero()) 1449 return FCmpInst::FCMP_OGT; 1450 1451 // Nothing more we can do 1452 return FCmpInst::BAD_FCMP_PREDICATE; 1453 } 1454 1455 // If the first operand is simple and second is ConstantExpr, swap operands. 1456 FCmpInst::Predicate SwappedRelation = evaluateFCmpRelation(V2, V1); 1457 if (SwappedRelation != FCmpInst::BAD_FCMP_PREDICATE) 1458 return FCmpInst::getSwappedPredicate(SwappedRelation); 1459 } else { 1460 // Ok, the LHS is known to be a constantexpr. The RHS can be any of a 1461 // constantexpr or a simple constant. 1462 ConstantExpr *CE1 = cast<ConstantExpr>(V1); 1463 switch (CE1->getOpcode()) { 1464 case Instruction::FPTrunc: 1465 case Instruction::FPExt: 1466 case Instruction::UIToFP: 1467 case Instruction::SIToFP: 1468 // We might be able to do something with these but we don't right now. 1469 break; 1470 default: 1471 break; 1472 } 1473 } 1474 // There are MANY other foldings that we could perform here. They will 1475 // probably be added on demand, as they seem needed. 1476 return FCmpInst::BAD_FCMP_PREDICATE; 1477 } 1478 1479 static ICmpInst::Predicate areGlobalsPotentiallyEqual(const GlobalValue *GV1, 1480 const GlobalValue *GV2) { 1481 auto isGlobalUnsafeForEquality = [](const GlobalValue *GV) { 1482 if (GV->hasExternalWeakLinkage() || GV->hasWeakAnyLinkage()) 1483 return true; 1484 if (const auto *GVar = dyn_cast<GlobalVariable>(GV)) { 1485 Type *Ty = GVar->getValueType(); 1486 // A global with opaque type might end up being zero sized. 1487 if (!Ty->isSized()) 1488 return true; 1489 // A global with an empty type might lie at the address of any other 1490 // global. 1491 if (Ty->isEmptyTy()) 1492 return true; 1493 } 1494 return false; 1495 }; 1496 // Don't try to decide equality of aliases. 1497 if (!isa<GlobalAlias>(GV1) && !isa<GlobalAlias>(GV2)) 1498 if (!isGlobalUnsafeForEquality(GV1) && !isGlobalUnsafeForEquality(GV2)) 1499 return ICmpInst::ICMP_NE; 1500 return ICmpInst::BAD_ICMP_PREDICATE; 1501 } 1502 1503 /// This function determines if there is anything we can decide about the two 1504 /// constants provided. This doesn't need to handle simple things like integer 1505 /// comparisons, but should instead handle ConstantExprs and GlobalValues. 1506 /// If we can determine that the two constants have a particular relation to 1507 /// each other, we should return the corresponding ICmp predicate, otherwise 1508 /// return ICmpInst::BAD_ICMP_PREDICATE. 1509 /// 1510 /// To simplify this code we canonicalize the relation so that the first 1511 /// operand is always the most "complex" of the two. We consider simple 1512 /// constants (like ConstantInt) to be the simplest, followed by 1513 /// GlobalValues, followed by ConstantExpr's (the most complex). 1514 /// 1515 static ICmpInst::Predicate evaluateICmpRelation(Constant *V1, Constant *V2, 1516 bool isSigned) { 1517 assert(V1->getType() == V2->getType() && 1518 "Cannot compare different types of values!"); 1519 if (V1 == V2) return ICmpInst::ICMP_EQ; 1520 1521 if (!isa<ConstantExpr>(V1) && !isa<GlobalValue>(V1) && 1522 !isa<BlockAddress>(V1)) { 1523 if (!isa<GlobalValue>(V2) && !isa<ConstantExpr>(V2) && 1524 !isa<BlockAddress>(V2)) { 1525 // We distilled this down to a simple case, use the standard constant 1526 // folder. 1527 ConstantInt *R = nullptr; 1528 ICmpInst::Predicate pred = ICmpInst::ICMP_EQ; 1529 R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2)); 1530 if (R && !R->isZero()) 1531 return pred; 1532 pred = isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; 1533 R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2)); 1534 if (R && !R->isZero()) 1535 return pred; 1536 pred = isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1537 R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2)); 1538 if (R && !R->isZero()) 1539 return pred; 1540 1541 // If we couldn't figure it out, bail. 1542 return ICmpInst::BAD_ICMP_PREDICATE; 1543 } 1544 1545 // If the first operand is simple, swap operands. 1546 ICmpInst::Predicate SwappedRelation = 1547 evaluateICmpRelation(V2, V1, isSigned); 1548 if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE) 1549 return ICmpInst::getSwappedPredicate(SwappedRelation); 1550 1551 } else if (const GlobalValue *GV = dyn_cast<GlobalValue>(V1)) { 1552 if (isa<ConstantExpr>(V2)) { // Swap as necessary. 1553 ICmpInst::Predicate SwappedRelation = 1554 evaluateICmpRelation(V2, V1, isSigned); 1555 if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE) 1556 return ICmpInst::getSwappedPredicate(SwappedRelation); 1557 return ICmpInst::BAD_ICMP_PREDICATE; 1558 } 1559 1560 // Now we know that the RHS is a GlobalValue, BlockAddress or simple 1561 // constant (which, since the types must match, means that it's a 1562 // ConstantPointerNull). 1563 if (const GlobalValue *GV2 = dyn_cast<GlobalValue>(V2)) { 1564 return areGlobalsPotentiallyEqual(GV, GV2); 1565 } else if (isa<BlockAddress>(V2)) { 1566 return ICmpInst::ICMP_NE; // Globals never equal labels. 1567 } else { 1568 assert(isa<ConstantPointerNull>(V2) && "Canonicalization guarantee!"); 1569 // GlobalVals can never be null unless they have external weak linkage. 1570 // We don't try to evaluate aliases here. 1571 // NOTE: We should not be doing this constant folding if null pointer 1572 // is considered valid for the function. But currently there is no way to 1573 // query it from the Constant type. 1574 if (!GV->hasExternalWeakLinkage() && !isa<GlobalAlias>(GV) && 1575 !NullPointerIsDefined(nullptr /* F */, 1576 GV->getType()->getAddressSpace())) 1577 return ICmpInst::ICMP_NE; 1578 } 1579 } else if (const BlockAddress *BA = dyn_cast<BlockAddress>(V1)) { 1580 if (isa<ConstantExpr>(V2)) { // Swap as necessary. 1581 ICmpInst::Predicate SwappedRelation = 1582 evaluateICmpRelation(V2, V1, isSigned); 1583 if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE) 1584 return ICmpInst::getSwappedPredicate(SwappedRelation); 1585 return ICmpInst::BAD_ICMP_PREDICATE; 1586 } 1587 1588 // Now we know that the RHS is a GlobalValue, BlockAddress or simple 1589 // constant (which, since the types must match, means that it is a 1590 // ConstantPointerNull). 1591 if (const BlockAddress *BA2 = dyn_cast<BlockAddress>(V2)) { 1592 // Block address in another function can't equal this one, but block 1593 // addresses in the current function might be the same if blocks are 1594 // empty. 1595 if (BA2->getFunction() != BA->getFunction()) 1596 return ICmpInst::ICMP_NE; 1597 } else { 1598 // Block addresses aren't null, don't equal the address of globals. 1599 assert((isa<ConstantPointerNull>(V2) || isa<GlobalValue>(V2)) && 1600 "Canonicalization guarantee!"); 1601 return ICmpInst::ICMP_NE; 1602 } 1603 } else { 1604 // Ok, the LHS is known to be a constantexpr. The RHS can be any of a 1605 // constantexpr, a global, block address, or a simple constant. 1606 ConstantExpr *CE1 = cast<ConstantExpr>(V1); 1607 Constant *CE1Op0 = CE1->getOperand(0); 1608 1609 switch (CE1->getOpcode()) { 1610 case Instruction::Trunc: 1611 case Instruction::FPTrunc: 1612 case Instruction::FPExt: 1613 case Instruction::FPToUI: 1614 case Instruction::FPToSI: 1615 break; // We can't evaluate floating point casts or truncations. 1616 1617 case Instruction::UIToFP: 1618 case Instruction::SIToFP: 1619 case Instruction::BitCast: 1620 case Instruction::ZExt: 1621 case Instruction::SExt: 1622 // We can't evaluate floating point casts or truncations. 1623 if (CE1Op0->getType()->isFPOrFPVectorTy()) 1624 break; 1625 1626 // If the cast is not actually changing bits, and the second operand is a 1627 // null pointer, do the comparison with the pre-casted value. 1628 if (V2->isNullValue() && CE1->getType()->isIntOrPtrTy()) { 1629 if (CE1->getOpcode() == Instruction::ZExt) isSigned = false; 1630 if (CE1->getOpcode() == Instruction::SExt) isSigned = true; 1631 return evaluateICmpRelation(CE1Op0, 1632 Constant::getNullValue(CE1Op0->getType()), 1633 isSigned); 1634 } 1635 break; 1636 1637 case Instruction::GetElementPtr: { 1638 GEPOperator *CE1GEP = cast<GEPOperator>(CE1); 1639 // Ok, since this is a getelementptr, we know that the constant has a 1640 // pointer type. Check the various cases. 1641 if (isa<ConstantPointerNull>(V2)) { 1642 // If we are comparing a GEP to a null pointer, check to see if the base 1643 // of the GEP equals the null pointer. 1644 if (const GlobalValue *GV = dyn_cast<GlobalValue>(CE1Op0)) { 1645 if (GV->hasExternalWeakLinkage()) 1646 // Weak linkage GVals could be zero or not. We're comparing that 1647 // to null pointer so its greater-or-equal 1648 return isSigned ? ICmpInst::ICMP_SGE : ICmpInst::ICMP_UGE; 1649 else 1650 // If its not weak linkage, the GVal must have a non-zero address 1651 // so the result is greater-than 1652 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1653 } else if (isa<ConstantPointerNull>(CE1Op0)) { 1654 // If we are indexing from a null pointer, check to see if we have any 1655 // non-zero indices. 1656 for (unsigned i = 1, e = CE1->getNumOperands(); i != e; ++i) 1657 if (!CE1->getOperand(i)->isNullValue()) 1658 // Offsetting from null, must not be equal. 1659 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1660 // Only zero indexes from null, must still be zero. 1661 return ICmpInst::ICMP_EQ; 1662 } 1663 // Otherwise, we can't really say if the first operand is null or not. 1664 } else if (const GlobalValue *GV2 = dyn_cast<GlobalValue>(V2)) { 1665 if (isa<ConstantPointerNull>(CE1Op0)) { 1666 if (GV2->hasExternalWeakLinkage()) 1667 // Weak linkage GVals could be zero or not. We're comparing it to 1668 // a null pointer, so its less-or-equal 1669 return isSigned ? ICmpInst::ICMP_SLE : ICmpInst::ICMP_ULE; 1670 else 1671 // If its not weak linkage, the GVal must have a non-zero address 1672 // so the result is less-than 1673 return isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; 1674 } else if (const GlobalValue *GV = dyn_cast<GlobalValue>(CE1Op0)) { 1675 if (GV == GV2) { 1676 // If this is a getelementptr of the same global, then it must be 1677 // different. Because the types must match, the getelementptr could 1678 // only have at most one index, and because we fold getelementptr's 1679 // with a single zero index, it must be nonzero. 1680 assert(CE1->getNumOperands() == 2 && 1681 !CE1->getOperand(1)->isNullValue() && 1682 "Surprising getelementptr!"); 1683 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1684 } else { 1685 if (CE1GEP->hasAllZeroIndices()) 1686 return areGlobalsPotentiallyEqual(GV, GV2); 1687 return ICmpInst::BAD_ICMP_PREDICATE; 1688 } 1689 } 1690 } else { 1691 ConstantExpr *CE2 = cast<ConstantExpr>(V2); 1692 Constant *CE2Op0 = CE2->getOperand(0); 1693 1694 // There are MANY other foldings that we could perform here. They will 1695 // probably be added on demand, as they seem needed. 1696 switch (CE2->getOpcode()) { 1697 default: break; 1698 case Instruction::GetElementPtr: 1699 // By far the most common case to handle is when the base pointers are 1700 // obviously to the same global. 1701 if (isa<GlobalValue>(CE1Op0) && isa<GlobalValue>(CE2Op0)) { 1702 // Don't know relative ordering, but check for inequality. 1703 if (CE1Op0 != CE2Op0) { 1704 GEPOperator *CE2GEP = cast<GEPOperator>(CE2); 1705 if (CE1GEP->hasAllZeroIndices() && CE2GEP->hasAllZeroIndices()) 1706 return areGlobalsPotentiallyEqual(cast<GlobalValue>(CE1Op0), 1707 cast<GlobalValue>(CE2Op0)); 1708 return ICmpInst::BAD_ICMP_PREDICATE; 1709 } 1710 // Ok, we know that both getelementptr instructions are based on the 1711 // same global. From this, we can precisely determine the relative 1712 // ordering of the resultant pointers. 1713 unsigned i = 1; 1714 1715 // The logic below assumes that the result of the comparison 1716 // can be determined by finding the first index that differs. 1717 // This doesn't work if there is over-indexing in any 1718 // subsequent indices, so check for that case first. 1719 if (!CE1->isGEPWithNoNotionalOverIndexing() || 1720 !CE2->isGEPWithNoNotionalOverIndexing()) 1721 return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal. 1722 1723 // Compare all of the operands the GEP's have in common. 1724 gep_type_iterator GTI = gep_type_begin(CE1); 1725 for (;i != CE1->getNumOperands() && i != CE2->getNumOperands(); 1726 ++i, ++GTI) 1727 switch (IdxCompare(CE1->getOperand(i), 1728 CE2->getOperand(i), GTI.getIndexedType())) { 1729 case -1: return isSigned ? ICmpInst::ICMP_SLT:ICmpInst::ICMP_ULT; 1730 case 1: return isSigned ? ICmpInst::ICMP_SGT:ICmpInst::ICMP_UGT; 1731 case -2: return ICmpInst::BAD_ICMP_PREDICATE; 1732 } 1733 1734 // Ok, we ran out of things they have in common. If any leftovers 1735 // are non-zero then we have a difference, otherwise we are equal. 1736 for (; i < CE1->getNumOperands(); ++i) 1737 if (!CE1->getOperand(i)->isNullValue()) { 1738 if (isa<ConstantInt>(CE1->getOperand(i))) 1739 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1740 else 1741 return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal. 1742 } 1743 1744 for (; i < CE2->getNumOperands(); ++i) 1745 if (!CE2->getOperand(i)->isNullValue()) { 1746 if (isa<ConstantInt>(CE2->getOperand(i))) 1747 return isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; 1748 else 1749 return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal. 1750 } 1751 return ICmpInst::ICMP_EQ; 1752 } 1753 } 1754 } 1755 break; 1756 } 1757 default: 1758 break; 1759 } 1760 } 1761 1762 return ICmpInst::BAD_ICMP_PREDICATE; 1763 } 1764 1765 Constant *llvm::ConstantFoldCompareInstruction(unsigned short pred, 1766 Constant *C1, Constant *C2) { 1767 Type *ResultTy; 1768 if (VectorType *VT = dyn_cast<VectorType>(C1->getType())) 1769 ResultTy = VectorType::get(Type::getInt1Ty(C1->getContext()), 1770 VT->getNumElements()); 1771 else 1772 ResultTy = Type::getInt1Ty(C1->getContext()); 1773 1774 // Fold FCMP_FALSE/FCMP_TRUE unconditionally. 1775 if (pred == FCmpInst::FCMP_FALSE) 1776 return Constant::getNullValue(ResultTy); 1777 1778 if (pred == FCmpInst::FCMP_TRUE) 1779 return Constant::getAllOnesValue(ResultTy); 1780 1781 // Handle some degenerate cases first 1782 if (isa<UndefValue>(C1) || isa<UndefValue>(C2)) { 1783 CmpInst::Predicate Predicate = CmpInst::Predicate(pred); 1784 bool isIntegerPredicate = ICmpInst::isIntPredicate(Predicate); 1785 // For EQ and NE, we can always pick a value for the undef to make the 1786 // predicate pass or fail, so we can return undef. 1787 // Also, if both operands are undef, we can return undef for int comparison. 1788 if (ICmpInst::isEquality(Predicate) || (isIntegerPredicate && C1 == C2)) 1789 return UndefValue::get(ResultTy); 1790 1791 // Otherwise, for integer compare, pick the same value as the non-undef 1792 // operand, and fold it to true or false. 1793 if (isIntegerPredicate) 1794 return ConstantInt::get(ResultTy, CmpInst::isTrueWhenEqual(Predicate)); 1795 1796 // Choosing NaN for the undef will always make unordered comparison succeed 1797 // and ordered comparison fails. 1798 return ConstantInt::get(ResultTy, CmpInst::isUnordered(Predicate)); 1799 } 1800 1801 // icmp eq/ne(null,GV) -> false/true 1802 if (C1->isNullValue()) { 1803 if (const GlobalValue *GV = dyn_cast<GlobalValue>(C2)) 1804 // Don't try to evaluate aliases. External weak GV can be null. 1805 if (!isa<GlobalAlias>(GV) && !GV->hasExternalWeakLinkage() && 1806 !NullPointerIsDefined(nullptr /* F */, 1807 GV->getType()->getAddressSpace())) { 1808 if (pred == ICmpInst::ICMP_EQ) 1809 return ConstantInt::getFalse(C1->getContext()); 1810 else if (pred == ICmpInst::ICMP_NE) 1811 return ConstantInt::getTrue(C1->getContext()); 1812 } 1813 // icmp eq/ne(GV,null) -> false/true 1814 } else if (C2->isNullValue()) { 1815 if (const GlobalValue *GV = dyn_cast<GlobalValue>(C1)) 1816 // Don't try to evaluate aliases. External weak GV can be null. 1817 if (!isa<GlobalAlias>(GV) && !GV->hasExternalWeakLinkage() && 1818 !NullPointerIsDefined(nullptr /* F */, 1819 GV->getType()->getAddressSpace())) { 1820 if (pred == ICmpInst::ICMP_EQ) 1821 return ConstantInt::getFalse(C1->getContext()); 1822 else if (pred == ICmpInst::ICMP_NE) 1823 return ConstantInt::getTrue(C1->getContext()); 1824 } 1825 } 1826 1827 // If the comparison is a comparison between two i1's, simplify it. 1828 if (C1->getType()->isIntegerTy(1)) { 1829 switch(pred) { 1830 case ICmpInst::ICMP_EQ: 1831 if (isa<ConstantInt>(C2)) 1832 return ConstantExpr::getXor(C1, ConstantExpr::getNot(C2)); 1833 return ConstantExpr::getXor(ConstantExpr::getNot(C1), C2); 1834 case ICmpInst::ICMP_NE: 1835 return ConstantExpr::getXor(C1, C2); 1836 default: 1837 break; 1838 } 1839 } 1840 1841 if (isa<ConstantInt>(C1) && isa<ConstantInt>(C2)) { 1842 const APInt &V1 = cast<ConstantInt>(C1)->getValue(); 1843 const APInt &V2 = cast<ConstantInt>(C2)->getValue(); 1844 switch (pred) { 1845 default: llvm_unreachable("Invalid ICmp Predicate"); 1846 case ICmpInst::ICMP_EQ: return ConstantInt::get(ResultTy, V1 == V2); 1847 case ICmpInst::ICMP_NE: return ConstantInt::get(ResultTy, V1 != V2); 1848 case ICmpInst::ICMP_SLT: return ConstantInt::get(ResultTy, V1.slt(V2)); 1849 case ICmpInst::ICMP_SGT: return ConstantInt::get(ResultTy, V1.sgt(V2)); 1850 case ICmpInst::ICMP_SLE: return ConstantInt::get(ResultTy, V1.sle(V2)); 1851 case ICmpInst::ICMP_SGE: return ConstantInt::get(ResultTy, V1.sge(V2)); 1852 case ICmpInst::ICMP_ULT: return ConstantInt::get(ResultTy, V1.ult(V2)); 1853 case ICmpInst::ICMP_UGT: return ConstantInt::get(ResultTy, V1.ugt(V2)); 1854 case ICmpInst::ICMP_ULE: return ConstantInt::get(ResultTy, V1.ule(V2)); 1855 case ICmpInst::ICMP_UGE: return ConstantInt::get(ResultTy, V1.uge(V2)); 1856 } 1857 } else if (isa<ConstantFP>(C1) && isa<ConstantFP>(C2)) { 1858 const APFloat &C1V = cast<ConstantFP>(C1)->getValueAPF(); 1859 const APFloat &C2V = cast<ConstantFP>(C2)->getValueAPF(); 1860 APFloat::cmpResult R = C1V.compare(C2V); 1861 switch (pred) { 1862 default: llvm_unreachable("Invalid FCmp Predicate"); 1863 case FCmpInst::FCMP_FALSE: return Constant::getNullValue(ResultTy); 1864 case FCmpInst::FCMP_TRUE: return Constant::getAllOnesValue(ResultTy); 1865 case FCmpInst::FCMP_UNO: 1866 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered); 1867 case FCmpInst::FCMP_ORD: 1868 return ConstantInt::get(ResultTy, R!=APFloat::cmpUnordered); 1869 case FCmpInst::FCMP_UEQ: 1870 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered || 1871 R==APFloat::cmpEqual); 1872 case FCmpInst::FCMP_OEQ: 1873 return ConstantInt::get(ResultTy, R==APFloat::cmpEqual); 1874 case FCmpInst::FCMP_UNE: 1875 return ConstantInt::get(ResultTy, R!=APFloat::cmpEqual); 1876 case FCmpInst::FCMP_ONE: 1877 return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan || 1878 R==APFloat::cmpGreaterThan); 1879 case FCmpInst::FCMP_ULT: 1880 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered || 1881 R==APFloat::cmpLessThan); 1882 case FCmpInst::FCMP_OLT: 1883 return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan); 1884 case FCmpInst::FCMP_UGT: 1885 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered || 1886 R==APFloat::cmpGreaterThan); 1887 case FCmpInst::FCMP_OGT: 1888 return ConstantInt::get(ResultTy, R==APFloat::cmpGreaterThan); 1889 case FCmpInst::FCMP_ULE: 1890 return ConstantInt::get(ResultTy, R!=APFloat::cmpGreaterThan); 1891 case FCmpInst::FCMP_OLE: 1892 return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan || 1893 R==APFloat::cmpEqual); 1894 case FCmpInst::FCMP_UGE: 1895 return ConstantInt::get(ResultTy, R!=APFloat::cmpLessThan); 1896 case FCmpInst::FCMP_OGE: 1897 return ConstantInt::get(ResultTy, R==APFloat::cmpGreaterThan || 1898 R==APFloat::cmpEqual); 1899 } 1900 } else if (C1->getType()->isVectorTy()) { 1901 // If we can constant fold the comparison of each element, constant fold 1902 // the whole vector comparison. 1903 SmallVector<Constant*, 4> ResElts; 1904 Type *Ty = IntegerType::get(C1->getContext(), 32); 1905 // Compare the elements, producing an i1 result or constant expr. 1906 for (unsigned i = 0, e = C1->getType()->getVectorNumElements(); i != e;++i){ 1907 Constant *C1E = 1908 ConstantExpr::getExtractElement(C1, ConstantInt::get(Ty, i)); 1909 Constant *C2E = 1910 ConstantExpr::getExtractElement(C2, ConstantInt::get(Ty, i)); 1911 1912 ResElts.push_back(ConstantExpr::getCompare(pred, C1E, C2E)); 1913 } 1914 1915 return ConstantVector::get(ResElts); 1916 } 1917 1918 if (C1->getType()->isFloatingPointTy() && 1919 // Only call evaluateFCmpRelation if we have a constant expr to avoid 1920 // infinite recursive loop 1921 (isa<ConstantExpr>(C1) || isa<ConstantExpr>(C2))) { 1922 int Result = -1; // -1 = unknown, 0 = known false, 1 = known true. 1923 switch (evaluateFCmpRelation(C1, C2)) { 1924 default: llvm_unreachable("Unknown relation!"); 1925 case FCmpInst::FCMP_UNO: 1926 case FCmpInst::FCMP_ORD: 1927 case FCmpInst::FCMP_UNE: 1928 case FCmpInst::FCMP_ULT: 1929 case FCmpInst::FCMP_UGT: 1930 case FCmpInst::FCMP_ULE: 1931 case FCmpInst::FCMP_UGE: 1932 case FCmpInst::FCMP_TRUE: 1933 case FCmpInst::FCMP_FALSE: 1934 case FCmpInst::BAD_FCMP_PREDICATE: 1935 break; // Couldn't determine anything about these constants. 1936 case FCmpInst::FCMP_OEQ: // We know that C1 == C2 1937 Result = (pred == FCmpInst::FCMP_UEQ || pred == FCmpInst::FCMP_OEQ || 1938 pred == FCmpInst::FCMP_ULE || pred == FCmpInst::FCMP_OLE || 1939 pred == FCmpInst::FCMP_UGE || pred == FCmpInst::FCMP_OGE); 1940 break; 1941 case FCmpInst::FCMP_OLT: // We know that C1 < C2 1942 Result = (pred == FCmpInst::FCMP_UNE || pred == FCmpInst::FCMP_ONE || 1943 pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT || 1944 pred == FCmpInst::FCMP_ULE || pred == FCmpInst::FCMP_OLE); 1945 break; 1946 case FCmpInst::FCMP_OGT: // We know that C1 > C2 1947 Result = (pred == FCmpInst::FCMP_UNE || pred == FCmpInst::FCMP_ONE || 1948 pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT || 1949 pred == FCmpInst::FCMP_UGE || pred == FCmpInst::FCMP_OGE); 1950 break; 1951 case FCmpInst::FCMP_OLE: // We know that C1 <= C2 1952 // We can only partially decide this relation. 1953 if (pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT) 1954 Result = 0; 1955 else if (pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT) 1956 Result = 1; 1957 break; 1958 case FCmpInst::FCMP_OGE: // We known that C1 >= C2 1959 // We can only partially decide this relation. 1960 if (pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT) 1961 Result = 0; 1962 else if (pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT) 1963 Result = 1; 1964 break; 1965 case FCmpInst::FCMP_ONE: // We know that C1 != C2 1966 // We can only partially decide this relation. 1967 if (pred == FCmpInst::FCMP_OEQ || pred == FCmpInst::FCMP_UEQ) 1968 Result = 0; 1969 else if (pred == FCmpInst::FCMP_ONE || pred == FCmpInst::FCMP_UNE) 1970 Result = 1; 1971 break; 1972 case FCmpInst::FCMP_UEQ: // We know that C1 == C2 || isUnordered(C1, C2). 1973 // We can only partially decide this relation. 1974 if (pred == FCmpInst::FCMP_ONE) 1975 Result = 0; 1976 else if (pred == FCmpInst::FCMP_UEQ) 1977 Result = 1; 1978 break; 1979 } 1980 1981 // If we evaluated the result, return it now. 1982 if (Result != -1) 1983 return ConstantInt::get(ResultTy, Result); 1984 1985 } else { 1986 // Evaluate the relation between the two constants, per the predicate. 1987 int Result = -1; // -1 = unknown, 0 = known false, 1 = known true. 1988 switch (evaluateICmpRelation(C1, C2, 1989 CmpInst::isSigned((CmpInst::Predicate)pred))) { 1990 default: llvm_unreachable("Unknown relational!"); 1991 case ICmpInst::BAD_ICMP_PREDICATE: 1992 break; // Couldn't determine anything about these constants. 1993 case ICmpInst::ICMP_EQ: // We know the constants are equal! 1994 // If we know the constants are equal, we can decide the result of this 1995 // computation precisely. 1996 Result = ICmpInst::isTrueWhenEqual((ICmpInst::Predicate)pred); 1997 break; 1998 case ICmpInst::ICMP_ULT: 1999 switch (pred) { 2000 case ICmpInst::ICMP_ULT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_ULE: 2001 Result = 1; break; 2002 case ICmpInst::ICMP_UGT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_UGE: 2003 Result = 0; break; 2004 } 2005 break; 2006 case ICmpInst::ICMP_SLT: 2007 switch (pred) { 2008 case ICmpInst::ICMP_SLT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_SLE: 2009 Result = 1; break; 2010 case ICmpInst::ICMP_SGT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_SGE: 2011 Result = 0; break; 2012 } 2013 break; 2014 case ICmpInst::ICMP_UGT: 2015 switch (pred) { 2016 case ICmpInst::ICMP_UGT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_UGE: 2017 Result = 1; break; 2018 case ICmpInst::ICMP_ULT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_ULE: 2019 Result = 0; break; 2020 } 2021 break; 2022 case ICmpInst::ICMP_SGT: 2023 switch (pred) { 2024 case ICmpInst::ICMP_SGT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_SGE: 2025 Result = 1; break; 2026 case ICmpInst::ICMP_SLT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_SLE: 2027 Result = 0; break; 2028 } 2029 break; 2030 case ICmpInst::ICMP_ULE: 2031 if (pred == ICmpInst::ICMP_UGT) Result = 0; 2032 if (pred == ICmpInst::ICMP_ULT || pred == ICmpInst::ICMP_ULE) Result = 1; 2033 break; 2034 case ICmpInst::ICMP_SLE: 2035 if (pred == ICmpInst::ICMP_SGT) Result = 0; 2036 if (pred == ICmpInst::ICMP_SLT || pred == ICmpInst::ICMP_SLE) Result = 1; 2037 break; 2038 case ICmpInst::ICMP_UGE: 2039 if (pred == ICmpInst::ICMP_ULT) Result = 0; 2040 if (pred == ICmpInst::ICMP_UGT || pred == ICmpInst::ICMP_UGE) Result = 1; 2041 break; 2042 case ICmpInst::ICMP_SGE: 2043 if (pred == ICmpInst::ICMP_SLT) Result = 0; 2044 if (pred == ICmpInst::ICMP_SGT || pred == ICmpInst::ICMP_SGE) Result = 1; 2045 break; 2046 case ICmpInst::ICMP_NE: 2047 if (pred == ICmpInst::ICMP_EQ) Result = 0; 2048 if (pred == ICmpInst::ICMP_NE) Result = 1; 2049 break; 2050 } 2051 2052 // If we evaluated the result, return it now. 2053 if (Result != -1) 2054 return ConstantInt::get(ResultTy, Result); 2055 2056 // If the right hand side is a bitcast, try using its inverse to simplify 2057 // it by moving it to the left hand side. We can't do this if it would turn 2058 // a vector compare into a scalar compare or visa versa, or if it would turn 2059 // the operands into FP values. 2060 if (ConstantExpr *CE2 = dyn_cast<ConstantExpr>(C2)) { 2061 Constant *CE2Op0 = CE2->getOperand(0); 2062 if (CE2->getOpcode() == Instruction::BitCast && 2063 CE2->getType()->isVectorTy() == CE2Op0->getType()->isVectorTy() && 2064 !CE2Op0->getType()->isFPOrFPVectorTy()) { 2065 Constant *Inverse = ConstantExpr::getBitCast(C1, CE2Op0->getType()); 2066 return ConstantExpr::getICmp(pred, Inverse, CE2Op0); 2067 } 2068 } 2069 2070 // If the left hand side is an extension, try eliminating it. 2071 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 2072 if ((CE1->getOpcode() == Instruction::SExt && 2073 ICmpInst::isSigned((ICmpInst::Predicate)pred)) || 2074 (CE1->getOpcode() == Instruction::ZExt && 2075 !ICmpInst::isSigned((ICmpInst::Predicate)pred))){ 2076 Constant *CE1Op0 = CE1->getOperand(0); 2077 Constant *CE1Inverse = ConstantExpr::getTrunc(CE1, CE1Op0->getType()); 2078 if (CE1Inverse == CE1Op0) { 2079 // Check whether we can safely truncate the right hand side. 2080 Constant *C2Inverse = ConstantExpr::getTrunc(C2, CE1Op0->getType()); 2081 if (ConstantExpr::getCast(CE1->getOpcode(), C2Inverse, 2082 C2->getType()) == C2) 2083 return ConstantExpr::getICmp(pred, CE1Inverse, C2Inverse); 2084 } 2085 } 2086 } 2087 2088 if ((!isa<ConstantExpr>(C1) && isa<ConstantExpr>(C2)) || 2089 (C1->isNullValue() && !C2->isNullValue())) { 2090 // If C2 is a constant expr and C1 isn't, flip them around and fold the 2091 // other way if possible. 2092 // Also, if C1 is null and C2 isn't, flip them around. 2093 pred = ICmpInst::getSwappedPredicate((ICmpInst::Predicate)pred); 2094 return ConstantExpr::getICmp(pred, C2, C1); 2095 } 2096 } 2097 return nullptr; 2098 } 2099 2100 /// Test whether the given sequence of *normalized* indices is "inbounds". 2101 template<typename IndexTy> 2102 static bool isInBoundsIndices(ArrayRef<IndexTy> Idxs) { 2103 // No indices means nothing that could be out of bounds. 2104 if (Idxs.empty()) return true; 2105 2106 // If the first index is zero, it's in bounds. 2107 if (cast<Constant>(Idxs[0])->isNullValue()) return true; 2108 2109 // If the first index is one and all the rest are zero, it's in bounds, 2110 // by the one-past-the-end rule. 2111 if (auto *CI = dyn_cast<ConstantInt>(Idxs[0])) { 2112 if (!CI->isOne()) 2113 return false; 2114 } else { 2115 auto *CV = cast<ConstantDataVector>(Idxs[0]); 2116 CI = dyn_cast_or_null<ConstantInt>(CV->getSplatValue()); 2117 if (!CI || !CI->isOne()) 2118 return false; 2119 } 2120 2121 for (unsigned i = 1, e = Idxs.size(); i != e; ++i) 2122 if (!cast<Constant>(Idxs[i])->isNullValue()) 2123 return false; 2124 return true; 2125 } 2126 2127 /// Test whether a given ConstantInt is in-range for a SequentialType. 2128 static bool isIndexInRangeOfArrayType(uint64_t NumElements, 2129 const ConstantInt *CI) { 2130 // We cannot bounds check the index if it doesn't fit in an int64_t. 2131 if (CI->getValue().getMinSignedBits() > 64) 2132 return false; 2133 2134 // A negative index or an index past the end of our sequential type is 2135 // considered out-of-range. 2136 int64_t IndexVal = CI->getSExtValue(); 2137 if (IndexVal < 0 || (NumElements > 0 && (uint64_t)IndexVal >= NumElements)) 2138 return false; 2139 2140 // Otherwise, it is in-range. 2141 return true; 2142 } 2143 2144 Constant *llvm::ConstantFoldGetElementPtr(Type *PointeeTy, Constant *C, 2145 bool InBounds, 2146 Optional<unsigned> InRangeIndex, 2147 ArrayRef<Value *> Idxs) { 2148 if (Idxs.empty()) return C; 2149 2150 Type *GEPTy = GetElementPtrInst::getGEPReturnType( 2151 PointeeTy, C, makeArrayRef((Value *const *)Idxs.data(), Idxs.size())); 2152 2153 if (isa<UndefValue>(C)) 2154 return UndefValue::get(GEPTy); 2155 2156 Constant *Idx0 = cast<Constant>(Idxs[0]); 2157 if (Idxs.size() == 1 && (Idx0->isNullValue() || isa<UndefValue>(Idx0))) 2158 return GEPTy->isVectorTy() && !C->getType()->isVectorTy() 2159 ? ConstantVector::getSplat( 2160 cast<VectorType>(GEPTy)->getNumElements(), C) 2161 : C; 2162 2163 if (C->isNullValue()) { 2164 bool isNull = true; 2165 for (unsigned i = 0, e = Idxs.size(); i != e; ++i) 2166 if (!isa<UndefValue>(Idxs[i]) && 2167 !cast<Constant>(Idxs[i])->isNullValue()) { 2168 isNull = false; 2169 break; 2170 } 2171 if (isNull) { 2172 PointerType *PtrTy = cast<PointerType>(C->getType()->getScalarType()); 2173 Type *Ty = GetElementPtrInst::getIndexedType(PointeeTy, Idxs); 2174 2175 assert(Ty && "Invalid indices for GEP!"); 2176 Type *OrigGEPTy = PointerType::get(Ty, PtrTy->getAddressSpace()); 2177 Type *GEPTy = PointerType::get(Ty, PtrTy->getAddressSpace()); 2178 if (VectorType *VT = dyn_cast<VectorType>(C->getType())) 2179 GEPTy = VectorType::get(OrigGEPTy, VT->getNumElements()); 2180 2181 // The GEP returns a vector of pointers when one of more of 2182 // its arguments is a vector. 2183 for (unsigned i = 0, e = Idxs.size(); i != e; ++i) { 2184 if (auto *VT = dyn_cast<VectorType>(Idxs[i]->getType())) { 2185 GEPTy = VectorType::get(OrigGEPTy, VT->getNumElements()); 2186 break; 2187 } 2188 } 2189 2190 return Constant::getNullValue(GEPTy); 2191 } 2192 } 2193 2194 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) { 2195 // Combine Indices - If the source pointer to this getelementptr instruction 2196 // is a getelementptr instruction, combine the indices of the two 2197 // getelementptr instructions into a single instruction. 2198 // 2199 if (CE->getOpcode() == Instruction::GetElementPtr) { 2200 gep_type_iterator LastI = gep_type_end(CE); 2201 for (gep_type_iterator I = gep_type_begin(CE), E = gep_type_end(CE); 2202 I != E; ++I) 2203 LastI = I; 2204 2205 // We cannot combine indices if doing so would take us outside of an 2206 // array or vector. Doing otherwise could trick us if we evaluated such a 2207 // GEP as part of a load. 2208 // 2209 // e.g. Consider if the original GEP was: 2210 // i8* getelementptr ({ [2 x i8], i32, i8, [3 x i8] }* @main.c, 2211 // i32 0, i32 0, i64 0) 2212 // 2213 // If we then tried to offset it by '8' to get to the third element, 2214 // an i8, we should *not* get: 2215 // i8* getelementptr ({ [2 x i8], i32, i8, [3 x i8] }* @main.c, 2216 // i32 0, i32 0, i64 8) 2217 // 2218 // This GEP tries to index array element '8 which runs out-of-bounds. 2219 // Subsequent evaluation would get confused and produce erroneous results. 2220 // 2221 // The following prohibits such a GEP from being formed by checking to see 2222 // if the index is in-range with respect to an array. 2223 // TODO: This code may be extended to handle vectors as well. 2224 bool PerformFold = false; 2225 if (Idx0->isNullValue()) 2226 PerformFold = true; 2227 else if (LastI.isSequential()) 2228 if (ConstantInt *CI = dyn_cast<ConstantInt>(Idx0)) 2229 PerformFold = (!LastI.isBoundedSequential() || 2230 isIndexInRangeOfArrayType( 2231 LastI.getSequentialNumElements(), CI)) && 2232 !CE->getOperand(CE->getNumOperands() - 1) 2233 ->getType() 2234 ->isVectorTy(); 2235 2236 if (PerformFold) { 2237 SmallVector<Value*, 16> NewIndices; 2238 NewIndices.reserve(Idxs.size() + CE->getNumOperands()); 2239 NewIndices.append(CE->op_begin() + 1, CE->op_end() - 1); 2240 2241 // Add the last index of the source with the first index of the new GEP. 2242 // Make sure to handle the case when they are actually different types. 2243 Constant *Combined = CE->getOperand(CE->getNumOperands()-1); 2244 // Otherwise it must be an array. 2245 if (!Idx0->isNullValue()) { 2246 Type *IdxTy = Combined->getType(); 2247 if (IdxTy != Idx0->getType()) { 2248 unsigned CommonExtendedWidth = 2249 std::max(IdxTy->getIntegerBitWidth(), 2250 Idx0->getType()->getIntegerBitWidth()); 2251 CommonExtendedWidth = std::max(CommonExtendedWidth, 64U); 2252 2253 Type *CommonTy = 2254 Type::getIntNTy(IdxTy->getContext(), CommonExtendedWidth); 2255 Constant *C1 = ConstantExpr::getSExtOrBitCast(Idx0, CommonTy); 2256 Constant *C2 = ConstantExpr::getSExtOrBitCast(Combined, CommonTy); 2257 Combined = ConstantExpr::get(Instruction::Add, C1, C2); 2258 } else { 2259 Combined = 2260 ConstantExpr::get(Instruction::Add, Idx0, Combined); 2261 } 2262 } 2263 2264 NewIndices.push_back(Combined); 2265 NewIndices.append(Idxs.begin() + 1, Idxs.end()); 2266 2267 // The combined GEP normally inherits its index inrange attribute from 2268 // the inner GEP, but if the inner GEP's last index was adjusted by the 2269 // outer GEP, any inbounds attribute on that index is invalidated. 2270 Optional<unsigned> IRIndex = cast<GEPOperator>(CE)->getInRangeIndex(); 2271 if (IRIndex && *IRIndex == CE->getNumOperands() - 2 && !Idx0->isNullValue()) 2272 IRIndex = None; 2273 2274 return ConstantExpr::getGetElementPtr( 2275 cast<GEPOperator>(CE)->getSourceElementType(), CE->getOperand(0), 2276 NewIndices, InBounds && cast<GEPOperator>(CE)->isInBounds(), 2277 IRIndex); 2278 } 2279 } 2280 2281 // Attempt to fold casts to the same type away. For example, folding: 2282 // 2283 // i32* getelementptr ([2 x i32]* bitcast ([3 x i32]* %X to [2 x i32]*), 2284 // i64 0, i64 0) 2285 // into: 2286 // 2287 // i32* getelementptr ([3 x i32]* %X, i64 0, i64 0) 2288 // 2289 // Don't fold if the cast is changing address spaces. 2290 if (CE->isCast() && Idxs.size() > 1 && Idx0->isNullValue()) { 2291 PointerType *SrcPtrTy = 2292 dyn_cast<PointerType>(CE->getOperand(0)->getType()); 2293 PointerType *DstPtrTy = dyn_cast<PointerType>(CE->getType()); 2294 if (SrcPtrTy && DstPtrTy) { 2295 ArrayType *SrcArrayTy = 2296 dyn_cast<ArrayType>(SrcPtrTy->getElementType()); 2297 ArrayType *DstArrayTy = 2298 dyn_cast<ArrayType>(DstPtrTy->getElementType()); 2299 if (SrcArrayTy && DstArrayTy 2300 && SrcArrayTy->getElementType() == DstArrayTy->getElementType() 2301 && SrcPtrTy->getAddressSpace() == DstPtrTy->getAddressSpace()) 2302 return ConstantExpr::getGetElementPtr(SrcArrayTy, 2303 (Constant *)CE->getOperand(0), 2304 Idxs, InBounds, InRangeIndex); 2305 } 2306 } 2307 } 2308 2309 // Check to see if any array indices are not within the corresponding 2310 // notional array or vector bounds. If so, try to determine if they can be 2311 // factored out into preceding dimensions. 2312 SmallVector<Constant *, 8> NewIdxs; 2313 Type *Ty = PointeeTy; 2314 Type *Prev = C->getType(); 2315 bool Unknown = 2316 !isa<ConstantInt>(Idxs[0]) && !isa<ConstantDataVector>(Idxs[0]); 2317 for (unsigned i = 1, e = Idxs.size(); i != e; 2318 Prev = Ty, Ty = cast<CompositeType>(Ty)->getTypeAtIndex(Idxs[i]), ++i) { 2319 if (!isa<ConstantInt>(Idxs[i]) && !isa<ConstantDataVector>(Idxs[i])) { 2320 // We don't know if it's in range or not. 2321 Unknown = true; 2322 continue; 2323 } 2324 if (!isa<ConstantInt>(Idxs[i - 1]) && !isa<ConstantDataVector>(Idxs[i - 1])) 2325 // Skip if the type of the previous index is not supported. 2326 continue; 2327 if (InRangeIndex && i == *InRangeIndex + 1) { 2328 // If an index is marked inrange, we cannot apply this canonicalization to 2329 // the following index, as that will cause the inrange index to point to 2330 // the wrong element. 2331 continue; 2332 } 2333 if (isa<StructType>(Ty)) { 2334 // The verify makes sure that GEPs into a struct are in range. 2335 continue; 2336 } 2337 auto *STy = cast<SequentialType>(Ty); 2338 if (isa<VectorType>(STy)) { 2339 // There can be awkward padding in after a non-power of two vector. 2340 Unknown = true; 2341 continue; 2342 } 2343 if (ConstantInt *CI = dyn_cast<ConstantInt>(Idxs[i])) { 2344 if (isIndexInRangeOfArrayType(STy->getNumElements(), CI)) 2345 // It's in range, skip to the next index. 2346 continue; 2347 if (CI->getSExtValue() < 0) { 2348 // It's out of range and negative, don't try to factor it. 2349 Unknown = true; 2350 continue; 2351 } 2352 } else { 2353 auto *CV = cast<ConstantDataVector>(Idxs[i]); 2354 bool InRange = true; 2355 for (unsigned I = 0, E = CV->getNumElements(); I != E; ++I) { 2356 auto *CI = cast<ConstantInt>(CV->getElementAsConstant(I)); 2357 InRange &= isIndexInRangeOfArrayType(STy->getNumElements(), CI); 2358 if (CI->getSExtValue() < 0) { 2359 Unknown = true; 2360 break; 2361 } 2362 } 2363 if (InRange || Unknown) 2364 // It's in range, skip to the next index. 2365 // It's out of range and negative, don't try to factor it. 2366 continue; 2367 } 2368 if (isa<StructType>(Prev)) { 2369 // It's out of range, but the prior dimension is a struct 2370 // so we can't do anything about it. 2371 Unknown = true; 2372 continue; 2373 } 2374 // It's out of range, but we can factor it into the prior 2375 // dimension. 2376 NewIdxs.resize(Idxs.size()); 2377 // Determine the number of elements in our sequential type. 2378 uint64_t NumElements = STy->getArrayNumElements(); 2379 2380 // Expand the current index or the previous index to a vector from a scalar 2381 // if necessary. 2382 Constant *CurrIdx = cast<Constant>(Idxs[i]); 2383 auto *PrevIdx = 2384 NewIdxs[i - 1] ? NewIdxs[i - 1] : cast<Constant>(Idxs[i - 1]); 2385 bool IsCurrIdxVector = CurrIdx->getType()->isVectorTy(); 2386 bool IsPrevIdxVector = PrevIdx->getType()->isVectorTy(); 2387 bool UseVector = IsCurrIdxVector || IsPrevIdxVector; 2388 2389 if (!IsCurrIdxVector && IsPrevIdxVector) 2390 CurrIdx = ConstantDataVector::getSplat( 2391 PrevIdx->getType()->getVectorNumElements(), CurrIdx); 2392 2393 if (!IsPrevIdxVector && IsCurrIdxVector) 2394 PrevIdx = ConstantDataVector::getSplat( 2395 CurrIdx->getType()->getVectorNumElements(), PrevIdx); 2396 2397 Constant *Factor = 2398 ConstantInt::get(CurrIdx->getType()->getScalarType(), NumElements); 2399 if (UseVector) 2400 Factor = ConstantDataVector::getSplat( 2401 IsPrevIdxVector ? PrevIdx->getType()->getVectorNumElements() 2402 : CurrIdx->getType()->getVectorNumElements(), 2403 Factor); 2404 2405 NewIdxs[i] = ConstantExpr::getSRem(CurrIdx, Factor); 2406 2407 Constant *Div = ConstantExpr::getSDiv(CurrIdx, Factor); 2408 2409 unsigned CommonExtendedWidth = 2410 std::max(PrevIdx->getType()->getScalarSizeInBits(), 2411 Div->getType()->getScalarSizeInBits()); 2412 CommonExtendedWidth = std::max(CommonExtendedWidth, 64U); 2413 2414 // Before adding, extend both operands to i64 to avoid 2415 // overflow trouble. 2416 Type *ExtendedTy = Type::getIntNTy(Div->getContext(), CommonExtendedWidth); 2417 if (UseVector) 2418 ExtendedTy = VectorType::get( 2419 ExtendedTy, IsPrevIdxVector 2420 ? PrevIdx->getType()->getVectorNumElements() 2421 : CurrIdx->getType()->getVectorNumElements()); 2422 2423 if (!PrevIdx->getType()->isIntOrIntVectorTy(CommonExtendedWidth)) 2424 PrevIdx = ConstantExpr::getSExt(PrevIdx, ExtendedTy); 2425 2426 if (!Div->getType()->isIntOrIntVectorTy(CommonExtendedWidth)) 2427 Div = ConstantExpr::getSExt(Div, ExtendedTy); 2428 2429 NewIdxs[i - 1] = ConstantExpr::getAdd(PrevIdx, Div); 2430 } 2431 2432 // If we did any factoring, start over with the adjusted indices. 2433 if (!NewIdxs.empty()) { 2434 for (unsigned i = 0, e = Idxs.size(); i != e; ++i) 2435 if (!NewIdxs[i]) NewIdxs[i] = cast<Constant>(Idxs[i]); 2436 return ConstantExpr::getGetElementPtr(PointeeTy, C, NewIdxs, InBounds, 2437 InRangeIndex); 2438 } 2439 2440 // If all indices are known integers and normalized, we can do a simple 2441 // check for the "inbounds" property. 2442 if (!Unknown && !InBounds) 2443 if (auto *GV = dyn_cast<GlobalVariable>(C)) 2444 if (!GV->hasExternalWeakLinkage() && isInBoundsIndices(Idxs)) 2445 return ConstantExpr::getGetElementPtr(PointeeTy, C, Idxs, 2446 /*InBounds=*/true, InRangeIndex); 2447 2448 return nullptr; 2449 } 2450