1 //===- ConstantFold.cpp - LLVM constant folder ----------------------------===// 2 // 3 // The LLVM Compiler Infrastructure 4 // 5 // This file is distributed under the University of Illinois Open Source 6 // License. See LICENSE.TXT for details. 7 // 8 //===----------------------------------------------------------------------===// 9 // 10 // This file implements folding of constants for LLVM. This implements the 11 // (internal) ConstantFold.h interface, which is used by the 12 // ConstantExpr::get* methods to automatically fold constants when possible. 13 // 14 // The current constant folding implementation is implemented in two pieces: the 15 // pieces that don't need DataLayout, and the pieces that do. This is to avoid 16 // a dependence in IR on Target. 17 // 18 //===----------------------------------------------------------------------===// 19 20 #include "ConstantFold.h" 21 #include "llvm/ADT/APSInt.h" 22 #include "llvm/ADT/SmallVector.h" 23 #include "llvm/IR/Constants.h" 24 #include "llvm/IR/DerivedTypes.h" 25 #include "llvm/IR/Function.h" 26 #include "llvm/IR/GetElementPtrTypeIterator.h" 27 #include "llvm/IR/GlobalAlias.h" 28 #include "llvm/IR/GlobalVariable.h" 29 #include "llvm/IR/Instructions.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 /// @brief 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 unsigned ShAmt = Amt->getZExtValue(); 272 // Cannot analyze non-byte shifts. 273 if ((ShAmt & 7) != 0) 274 return nullptr; 275 ShAmt >>= 3; 276 277 // If the extract is known to be all zeros, return zero. 278 if (ByteStart >= CSize-ShAmt) 279 return Constant::getNullValue(IntegerType::get(CE->getContext(), 280 ByteSize*8)); 281 // If the extract is known to be fully in the input, extract it. 282 if (ByteStart+ByteSize+ShAmt <= CSize) 283 return ExtractConstantBytes(CE->getOperand(0), ByteStart+ShAmt, ByteSize); 284 285 // TODO: Handle the 'partially zero' case. 286 return nullptr; 287 } 288 289 case Instruction::Shl: { 290 ConstantInt *Amt = dyn_cast<ConstantInt>(CE->getOperand(1)); 291 if (!Amt) 292 return nullptr; 293 unsigned ShAmt = Amt->getZExtValue(); 294 // Cannot analyze non-byte shifts. 295 if ((ShAmt & 7) != 0) 296 return nullptr; 297 ShAmt >>= 3; 298 299 // If the extract is known to be all zeros, return zero. 300 if (ByteStart+ByteSize <= ShAmt) 301 return Constant::getNullValue(IntegerType::get(CE->getContext(), 302 ByteSize*8)); 303 // If the extract is known to be fully in the input, extract it. 304 if (ByteStart >= ShAmt) 305 return ExtractConstantBytes(CE->getOperand(0), ByteStart-ShAmt, ByteSize); 306 307 // TODO: Handle the 'partially zero' case. 308 return nullptr; 309 } 310 311 case Instruction::ZExt: { 312 unsigned SrcBitSize = 313 cast<IntegerType>(CE->getOperand(0)->getType())->getBitWidth(); 314 315 // If extracting something that is completely zero, return 0. 316 if (ByteStart*8 >= SrcBitSize) 317 return Constant::getNullValue(IntegerType::get(CE->getContext(), 318 ByteSize*8)); 319 320 // If exactly extracting the input, return it. 321 if (ByteStart == 0 && ByteSize*8 == SrcBitSize) 322 return CE->getOperand(0); 323 324 // If extracting something completely in the input, if if the input is a 325 // multiple of 8 bits, recurse. 326 if ((SrcBitSize&7) == 0 && (ByteStart+ByteSize)*8 <= SrcBitSize) 327 return ExtractConstantBytes(CE->getOperand(0), ByteStart, ByteSize); 328 329 // Otherwise, if extracting a subset of the input, which is not multiple of 330 // 8 bits, do a shift and trunc to get the bits. 331 if ((ByteStart+ByteSize)*8 < SrcBitSize) { 332 assert((SrcBitSize&7) && "Shouldn't get byte sized case here"); 333 Constant *Res = CE->getOperand(0); 334 if (ByteStart) 335 Res = ConstantExpr::getLShr(Res, 336 ConstantInt::get(Res->getType(), ByteStart*8)); 337 return ConstantExpr::getTrunc(Res, IntegerType::get(C->getContext(), 338 ByteSize*8)); 339 } 340 341 // TODO: Handle the 'partially zero' case. 342 return nullptr; 343 } 344 } 345 } 346 347 /// Return a ConstantExpr with type DestTy for sizeof on Ty, with any known 348 /// factors factored out. If Folded is false, return null if no factoring was 349 /// possible, to avoid endlessly bouncing an unfoldable expression back into the 350 /// top-level folder. 351 static Constant *getFoldedSizeOf(Type *Ty, Type *DestTy, bool Folded) { 352 if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 353 Constant *N = ConstantInt::get(DestTy, ATy->getNumElements()); 354 Constant *E = getFoldedSizeOf(ATy->getElementType(), DestTy, true); 355 return ConstantExpr::getNUWMul(E, N); 356 } 357 358 if (StructType *STy = dyn_cast<StructType>(Ty)) 359 if (!STy->isPacked()) { 360 unsigned NumElems = STy->getNumElements(); 361 // An empty struct has size zero. 362 if (NumElems == 0) 363 return ConstantExpr::getNullValue(DestTy); 364 // Check for a struct with all members having the same size. 365 Constant *MemberSize = 366 getFoldedSizeOf(STy->getElementType(0), DestTy, true); 367 bool AllSame = true; 368 for (unsigned i = 1; i != NumElems; ++i) 369 if (MemberSize != 370 getFoldedSizeOf(STy->getElementType(i), DestTy, true)) { 371 AllSame = false; 372 break; 373 } 374 if (AllSame) { 375 Constant *N = ConstantInt::get(DestTy, NumElems); 376 return ConstantExpr::getNUWMul(MemberSize, N); 377 } 378 } 379 380 // Pointer size doesn't depend on the pointee type, so canonicalize them 381 // to an arbitrary pointee. 382 if (PointerType *PTy = dyn_cast<PointerType>(Ty)) 383 if (!PTy->getElementType()->isIntegerTy(1)) 384 return 385 getFoldedSizeOf(PointerType::get(IntegerType::get(PTy->getContext(), 1), 386 PTy->getAddressSpace()), 387 DestTy, true); 388 389 // If there's no interesting folding happening, bail so that we don't create 390 // a constant that looks like it needs folding but really doesn't. 391 if (!Folded) 392 return nullptr; 393 394 // Base case: Get a regular sizeof expression. 395 Constant *C = ConstantExpr::getSizeOf(Ty); 396 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 397 DestTy, false), 398 C, DestTy); 399 return C; 400 } 401 402 /// Return a ConstantExpr with type DestTy for alignof on Ty, with any known 403 /// factors factored out. If Folded is false, return null if no factoring was 404 /// possible, to avoid endlessly bouncing an unfoldable expression back into the 405 /// top-level folder. 406 static Constant *getFoldedAlignOf(Type *Ty, Type *DestTy, bool Folded) { 407 // The alignment of an array is equal to the alignment of the 408 // array element. Note that this is not always true for vectors. 409 if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 410 Constant *C = ConstantExpr::getAlignOf(ATy->getElementType()); 411 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 412 DestTy, 413 false), 414 C, DestTy); 415 return C; 416 } 417 418 if (StructType *STy = dyn_cast<StructType>(Ty)) { 419 // Packed structs always have an alignment of 1. 420 if (STy->isPacked()) 421 return ConstantInt::get(DestTy, 1); 422 423 // Otherwise, struct alignment is the maximum alignment of any member. 424 // Without target data, we can't compare much, but we can check to see 425 // if all the members have the same alignment. 426 unsigned NumElems = STy->getNumElements(); 427 // An empty struct has minimal alignment. 428 if (NumElems == 0) 429 return ConstantInt::get(DestTy, 1); 430 // Check for a struct with all members having the same alignment. 431 Constant *MemberAlign = 432 getFoldedAlignOf(STy->getElementType(0), DestTy, true); 433 bool AllSame = true; 434 for (unsigned i = 1; i != NumElems; ++i) 435 if (MemberAlign != getFoldedAlignOf(STy->getElementType(i), DestTy, true)) { 436 AllSame = false; 437 break; 438 } 439 if (AllSame) 440 return MemberAlign; 441 } 442 443 // Pointer alignment doesn't depend on the pointee type, so canonicalize them 444 // to an arbitrary pointee. 445 if (PointerType *PTy = dyn_cast<PointerType>(Ty)) 446 if (!PTy->getElementType()->isIntegerTy(1)) 447 return 448 getFoldedAlignOf(PointerType::get(IntegerType::get(PTy->getContext(), 449 1), 450 PTy->getAddressSpace()), 451 DestTy, true); 452 453 // If there's no interesting folding happening, bail so that we don't create 454 // a constant that looks like it needs folding but really doesn't. 455 if (!Folded) 456 return nullptr; 457 458 // Base case: Get a regular alignof expression. 459 Constant *C = ConstantExpr::getAlignOf(Ty); 460 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 461 DestTy, false), 462 C, DestTy); 463 return C; 464 } 465 466 /// Return a ConstantExpr with type DestTy for offsetof on Ty and FieldNo, with 467 /// any known factors factored out. If Folded is false, return null if no 468 /// factoring was possible, to avoid endlessly bouncing an unfoldable expression 469 /// back into the top-level folder. 470 static Constant *getFoldedOffsetOf(Type *Ty, Constant *FieldNo, Type *DestTy, 471 bool Folded) { 472 if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 473 Constant *N = ConstantExpr::getCast(CastInst::getCastOpcode(FieldNo, false, 474 DestTy, false), 475 FieldNo, DestTy); 476 Constant *E = getFoldedSizeOf(ATy->getElementType(), DestTy, true); 477 return ConstantExpr::getNUWMul(E, N); 478 } 479 480 if (StructType *STy = dyn_cast<StructType>(Ty)) 481 if (!STy->isPacked()) { 482 unsigned NumElems = STy->getNumElements(); 483 // An empty struct has no members. 484 if (NumElems == 0) 485 return nullptr; 486 // Check for a struct with all members having the same size. 487 Constant *MemberSize = 488 getFoldedSizeOf(STy->getElementType(0), DestTy, true); 489 bool AllSame = true; 490 for (unsigned i = 1; i != NumElems; ++i) 491 if (MemberSize != 492 getFoldedSizeOf(STy->getElementType(i), DestTy, true)) { 493 AllSame = false; 494 break; 495 } 496 if (AllSame) { 497 Constant *N = ConstantExpr::getCast(CastInst::getCastOpcode(FieldNo, 498 false, 499 DestTy, 500 false), 501 FieldNo, DestTy); 502 return ConstantExpr::getNUWMul(MemberSize, N); 503 } 504 } 505 506 // If there's no interesting folding happening, bail so that we don't create 507 // a constant that looks like it needs folding but really doesn't. 508 if (!Folded) 509 return nullptr; 510 511 // Base case: Get a regular offsetof expression. 512 Constant *C = ConstantExpr::getOffsetOf(Ty, FieldNo); 513 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 514 DestTy, false), 515 C, DestTy); 516 return C; 517 } 518 519 Constant *llvm::ConstantFoldCastInstruction(unsigned opc, Constant *V, 520 Type *DestTy) { 521 if (isa<UndefValue>(V)) { 522 // zext(undef) = 0, because the top bits will be zero. 523 // sext(undef) = 0, because the top bits will all be the same. 524 // [us]itofp(undef) = 0, because the result value is bounded. 525 if (opc == Instruction::ZExt || opc == Instruction::SExt || 526 opc == Instruction::UIToFP || opc == Instruction::SIToFP) 527 return Constant::getNullValue(DestTy); 528 return UndefValue::get(DestTy); 529 } 530 531 if (V->isNullValue() && !DestTy->isX86_MMXTy() && 532 opc != Instruction::AddrSpaceCast) 533 return Constant::getNullValue(DestTy); 534 535 // If the cast operand is a constant expression, there's a few things we can 536 // do to try to simplify it. 537 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) { 538 if (CE->isCast()) { 539 // Try hard to fold cast of cast because they are often eliminable. 540 if (unsigned newOpc = foldConstantCastPair(opc, CE, DestTy)) 541 return ConstantExpr::getCast(newOpc, CE->getOperand(0), DestTy); 542 } else if (CE->getOpcode() == Instruction::GetElementPtr && 543 // Do not fold addrspacecast (gep 0, .., 0). It might make the 544 // addrspacecast uncanonicalized. 545 opc != Instruction::AddrSpaceCast && 546 // Do not fold bitcast (gep) with inrange index, as this loses 547 // information. 548 !cast<GEPOperator>(CE)->getInRangeIndex().hasValue()) { 549 // If all of the indexes in the GEP are null values, there is no pointer 550 // adjustment going on. We might as well cast the source pointer. 551 bool isAllNull = true; 552 for (unsigned i = 1, e = CE->getNumOperands(); i != e; ++i) 553 if (!CE->getOperand(i)->isNullValue()) { 554 isAllNull = false; 555 break; 556 } 557 if (isAllNull) 558 // This is casting one pointer type to another, always BitCast 559 return ConstantExpr::getPointerCast(CE->getOperand(0), DestTy); 560 } 561 } 562 563 // If the cast operand is a constant vector, perform the cast by 564 // operating on each element. In the cast of bitcasts, the element 565 // count may be mismatched; don't attempt to handle that here. 566 if ((isa<ConstantVector>(V) || isa<ConstantDataVector>(V)) && 567 DestTy->isVectorTy() && 568 DestTy->getVectorNumElements() == V->getType()->getVectorNumElements()) { 569 SmallVector<Constant*, 16> res; 570 VectorType *DestVecTy = cast<VectorType>(DestTy); 571 Type *DstEltTy = DestVecTy->getElementType(); 572 Type *Ty = IntegerType::get(V->getContext(), 32); 573 for (unsigned i = 0, e = V->getType()->getVectorNumElements(); i != e; ++i) { 574 Constant *C = 575 ConstantExpr::getExtractElement(V, ConstantInt::get(Ty, i)); 576 res.push_back(ConstantExpr::getCast(opc, C, DstEltTy)); 577 } 578 return ConstantVector::get(res); 579 } 580 581 // We actually have to do a cast now. Perform the cast according to the 582 // opcode specified. 583 switch (opc) { 584 default: 585 llvm_unreachable("Failed to cast constant expression"); 586 case Instruction::FPTrunc: 587 case Instruction::FPExt: 588 if (ConstantFP *FPC = dyn_cast<ConstantFP>(V)) { 589 bool ignored; 590 APFloat Val = FPC->getValueAPF(); 591 Val.convert(DestTy->isHalfTy() ? APFloat::IEEEhalf() : 592 DestTy->isFloatTy() ? APFloat::IEEEsingle() : 593 DestTy->isDoubleTy() ? APFloat::IEEEdouble() : 594 DestTy->isX86_FP80Ty() ? APFloat::x87DoubleExtended() : 595 DestTy->isFP128Ty() ? APFloat::IEEEquad() : 596 DestTy->isPPC_FP128Ty() ? APFloat::PPCDoubleDouble() : 597 APFloat::Bogus(), 598 APFloat::rmNearestTiesToEven, &ignored); 599 return ConstantFP::get(V->getContext(), Val); 600 } 601 return nullptr; // Can't fold. 602 case Instruction::FPToUI: 603 case Instruction::FPToSI: 604 if (ConstantFP *FPC = dyn_cast<ConstantFP>(V)) { 605 const APFloat &V = FPC->getValueAPF(); 606 bool ignored; 607 uint32_t DestBitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 608 APSInt IntVal(DestBitWidth, opc == Instruction::FPToUI); 609 if (APFloat::opInvalidOp == 610 V.convertToInteger(IntVal, APFloat::rmTowardZero, &ignored)) { 611 // Undefined behavior invoked - the destination type can't represent 612 // the input constant. 613 return UndefValue::get(DestTy); 614 } 615 return ConstantInt::get(FPC->getContext(), IntVal); 616 } 617 return nullptr; // Can't fold. 618 case Instruction::IntToPtr: //always treated as unsigned 619 if (V->isNullValue()) // Is it an integral null value? 620 return ConstantPointerNull::get(cast<PointerType>(DestTy)); 621 return nullptr; // Other pointer types cannot be casted 622 case Instruction::PtrToInt: // always treated as unsigned 623 // Is it a null pointer value? 624 if (V->isNullValue()) 625 return ConstantInt::get(DestTy, 0); 626 // If this is a sizeof-like expression, pull out multiplications by 627 // known factors to expose them to subsequent folding. If it's an 628 // alignof-like expression, factor out known factors. 629 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) 630 if (CE->getOpcode() == Instruction::GetElementPtr && 631 CE->getOperand(0)->isNullValue()) { 632 // FIXME: Looks like getFoldedSizeOf(), getFoldedOffsetOf() and 633 // getFoldedAlignOf() don't handle the case when DestTy is a vector of 634 // pointers yet. We end up in asserts in CastInst::getCastOpcode (see 635 // test/Analysis/ConstantFolding/cast-vector.ll). I've only seen this 636 // happen in one "real" C-code test case, so it does not seem to be an 637 // important optimization to handle vectors here. For now, simply bail 638 // out. 639 if (DestTy->isVectorTy()) 640 return nullptr; 641 GEPOperator *GEPO = cast<GEPOperator>(CE); 642 Type *Ty = GEPO->getSourceElementType(); 643 if (CE->getNumOperands() == 2) { 644 // Handle a sizeof-like expression. 645 Constant *Idx = CE->getOperand(1); 646 bool isOne = isa<ConstantInt>(Idx) && cast<ConstantInt>(Idx)->isOne(); 647 if (Constant *C = getFoldedSizeOf(Ty, DestTy, !isOne)) { 648 Idx = ConstantExpr::getCast(CastInst::getCastOpcode(Idx, true, 649 DestTy, false), 650 Idx, DestTy); 651 return ConstantExpr::getMul(C, Idx); 652 } 653 } else if (CE->getNumOperands() == 3 && 654 CE->getOperand(1)->isNullValue()) { 655 // Handle an alignof-like expression. 656 if (StructType *STy = dyn_cast<StructType>(Ty)) 657 if (!STy->isPacked()) { 658 ConstantInt *CI = cast<ConstantInt>(CE->getOperand(2)); 659 if (CI->isOne() && 660 STy->getNumElements() == 2 && 661 STy->getElementType(0)->isIntegerTy(1)) { 662 return getFoldedAlignOf(STy->getElementType(1), DestTy, false); 663 } 664 } 665 // Handle an offsetof-like expression. 666 if (Ty->isStructTy() || Ty->isArrayTy()) { 667 if (Constant *C = getFoldedOffsetOf(Ty, CE->getOperand(2), 668 DestTy, false)) 669 return C; 670 } 671 } 672 } 673 // Other pointer types cannot be casted 674 return nullptr; 675 case Instruction::UIToFP: 676 case Instruction::SIToFP: 677 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 678 const APInt &api = CI->getValue(); 679 APFloat apf(DestTy->getFltSemantics(), 680 APInt::getNullValue(DestTy->getPrimitiveSizeInBits())); 681 if (APFloat::opOverflow & 682 apf.convertFromAPInt(api, opc==Instruction::SIToFP, 683 APFloat::rmNearestTiesToEven)) { 684 // Undefined behavior invoked - the destination type can't represent 685 // the input constant. 686 return UndefValue::get(DestTy); 687 } 688 return ConstantFP::get(V->getContext(), apf); 689 } 690 return nullptr; 691 case Instruction::ZExt: 692 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 693 uint32_t BitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 694 return ConstantInt::get(V->getContext(), 695 CI->getValue().zext(BitWidth)); 696 } 697 return nullptr; 698 case Instruction::SExt: 699 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 700 uint32_t BitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 701 return ConstantInt::get(V->getContext(), 702 CI->getValue().sext(BitWidth)); 703 } 704 return nullptr; 705 case Instruction::Trunc: { 706 if (V->getType()->isVectorTy()) 707 return nullptr; 708 709 uint32_t DestBitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 710 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 711 return ConstantInt::get(V->getContext(), 712 CI->getValue().trunc(DestBitWidth)); 713 } 714 715 // The input must be a constantexpr. See if we can simplify this based on 716 // the bytes we are demanding. Only do this if the source and dest are an 717 // even multiple of a byte. 718 if ((DestBitWidth & 7) == 0 && 719 (cast<IntegerType>(V->getType())->getBitWidth() & 7) == 0) 720 if (Constant *Res = ExtractConstantBytes(V, 0, DestBitWidth / 8)) 721 return Res; 722 723 return nullptr; 724 } 725 case Instruction::BitCast: 726 return FoldBitCast(V, DestTy); 727 case Instruction::AddrSpaceCast: 728 return nullptr; 729 } 730 } 731 732 Constant *llvm::ConstantFoldSelectInstruction(Constant *Cond, 733 Constant *V1, Constant *V2) { 734 // Check for i1 and vector true/false conditions. 735 if (Cond->isNullValue()) return V2; 736 if (Cond->isAllOnesValue()) return V1; 737 738 // If the condition is a vector constant, fold the result elementwise. 739 if (ConstantVector *CondV = dyn_cast<ConstantVector>(Cond)) { 740 SmallVector<Constant*, 16> Result; 741 Type *Ty = IntegerType::get(CondV->getContext(), 32); 742 for (unsigned i = 0, e = V1->getType()->getVectorNumElements(); i != e;++i){ 743 Constant *V; 744 Constant *V1Element = ConstantExpr::getExtractElement(V1, 745 ConstantInt::get(Ty, i)); 746 Constant *V2Element = ConstantExpr::getExtractElement(V2, 747 ConstantInt::get(Ty, i)); 748 Constant *Cond = dyn_cast<Constant>(CondV->getOperand(i)); 749 if (V1Element == V2Element) { 750 V = V1Element; 751 } else if (isa<UndefValue>(Cond)) { 752 V = isa<UndefValue>(V1Element) ? V1Element : V2Element; 753 } else { 754 if (!isa<ConstantInt>(Cond)) break; 755 V = Cond->isNullValue() ? V2Element : V1Element; 756 } 757 Result.push_back(V); 758 } 759 760 // If we were able to build the vector, return it. 761 if (Result.size() == V1->getType()->getVectorNumElements()) 762 return ConstantVector::get(Result); 763 } 764 765 if (isa<UndefValue>(Cond)) { 766 if (isa<UndefValue>(V1)) return V1; 767 return V2; 768 } 769 if (isa<UndefValue>(V1)) return V2; 770 if (isa<UndefValue>(V2)) return V1; 771 if (V1 == V2) return V1; 772 773 if (ConstantExpr *TrueVal = dyn_cast<ConstantExpr>(V1)) { 774 if (TrueVal->getOpcode() == Instruction::Select) 775 if (TrueVal->getOperand(0) == Cond) 776 return ConstantExpr::getSelect(Cond, TrueVal->getOperand(1), V2); 777 } 778 if (ConstantExpr *FalseVal = dyn_cast<ConstantExpr>(V2)) { 779 if (FalseVal->getOpcode() == Instruction::Select) 780 if (FalseVal->getOperand(0) == Cond) 781 return ConstantExpr::getSelect(Cond, V1, FalseVal->getOperand(2)); 782 } 783 784 return nullptr; 785 } 786 787 Constant *llvm::ConstantFoldExtractElementInstruction(Constant *Val, 788 Constant *Idx) { 789 if (isa<UndefValue>(Val)) // ee(undef, x) -> undef 790 return UndefValue::get(Val->getType()->getVectorElementType()); 791 if (Val->isNullValue()) // ee(zero, x) -> zero 792 return Constant::getNullValue(Val->getType()->getVectorElementType()); 793 // ee({w,x,y,z}, undef) -> undef 794 if (isa<UndefValue>(Idx)) 795 return UndefValue::get(Val->getType()->getVectorElementType()); 796 797 if (ConstantInt *CIdx = dyn_cast<ConstantInt>(Idx)) { 798 // ee({w,x,y,z}, wrong_value) -> undef 799 if (CIdx->uge(Val->getType()->getVectorNumElements())) 800 return UndefValue::get(Val->getType()->getVectorElementType()); 801 return Val->getAggregateElement(CIdx->getZExtValue()); 802 } 803 return nullptr; 804 } 805 806 Constant *llvm::ConstantFoldInsertElementInstruction(Constant *Val, 807 Constant *Elt, 808 Constant *Idx) { 809 if (isa<UndefValue>(Idx)) 810 return UndefValue::get(Val->getType()); 811 812 ConstantInt *CIdx = dyn_cast<ConstantInt>(Idx); 813 if (!CIdx) return nullptr; 814 815 unsigned NumElts = Val->getType()->getVectorNumElements(); 816 if (CIdx->uge(NumElts)) 817 return UndefValue::get(Val->getType()); 818 819 SmallVector<Constant*, 16> Result; 820 Result.reserve(NumElts); 821 auto *Ty = Type::getInt32Ty(Val->getContext()); 822 uint64_t IdxVal = CIdx->getZExtValue(); 823 for (unsigned i = 0; i != NumElts; ++i) { 824 if (i == IdxVal) { 825 Result.push_back(Elt); 826 continue; 827 } 828 829 Constant *C = ConstantExpr::getExtractElement(Val, ConstantInt::get(Ty, i)); 830 Result.push_back(C); 831 } 832 833 return ConstantVector::get(Result); 834 } 835 836 Constant *llvm::ConstantFoldShuffleVectorInstruction(Constant *V1, 837 Constant *V2, 838 Constant *Mask) { 839 unsigned MaskNumElts = Mask->getType()->getVectorNumElements(); 840 Type *EltTy = V1->getType()->getVectorElementType(); 841 842 // Undefined shuffle mask -> undefined value. 843 if (isa<UndefValue>(Mask)) 844 return UndefValue::get(VectorType::get(EltTy, MaskNumElts)); 845 846 // Don't break the bitcode reader hack. 847 if (isa<ConstantExpr>(Mask)) return nullptr; 848 849 unsigned SrcNumElts = V1->getType()->getVectorNumElements(); 850 851 // Loop over the shuffle mask, evaluating each element. 852 SmallVector<Constant*, 32> Result; 853 for (unsigned i = 0; i != MaskNumElts; ++i) { 854 int Elt = ShuffleVectorInst::getMaskValue(Mask, i); 855 if (Elt == -1) { 856 Result.push_back(UndefValue::get(EltTy)); 857 continue; 858 } 859 Constant *InElt; 860 if (unsigned(Elt) >= SrcNumElts*2) 861 InElt = UndefValue::get(EltTy); 862 else if (unsigned(Elt) >= SrcNumElts) { 863 Type *Ty = IntegerType::get(V2->getContext(), 32); 864 InElt = 865 ConstantExpr::getExtractElement(V2, 866 ConstantInt::get(Ty, Elt - SrcNumElts)); 867 } else { 868 Type *Ty = IntegerType::get(V1->getContext(), 32); 869 InElt = ConstantExpr::getExtractElement(V1, ConstantInt::get(Ty, Elt)); 870 } 871 Result.push_back(InElt); 872 } 873 874 return ConstantVector::get(Result); 875 } 876 877 Constant *llvm::ConstantFoldExtractValueInstruction(Constant *Agg, 878 ArrayRef<unsigned> Idxs) { 879 // Base case: no indices, so return the entire value. 880 if (Idxs.empty()) 881 return Agg; 882 883 if (Constant *C = Agg->getAggregateElement(Idxs[0])) 884 return ConstantFoldExtractValueInstruction(C, Idxs.slice(1)); 885 886 return nullptr; 887 } 888 889 Constant *llvm::ConstantFoldInsertValueInstruction(Constant *Agg, 890 Constant *Val, 891 ArrayRef<unsigned> Idxs) { 892 // Base case: no indices, so replace the entire value. 893 if (Idxs.empty()) 894 return Val; 895 896 unsigned NumElts; 897 if (StructType *ST = dyn_cast<StructType>(Agg->getType())) 898 NumElts = ST->getNumElements(); 899 else 900 NumElts = cast<SequentialType>(Agg->getType())->getNumElements(); 901 902 SmallVector<Constant*, 32> Result; 903 for (unsigned i = 0; i != NumElts; ++i) { 904 Constant *C = Agg->getAggregateElement(i); 905 if (!C) return nullptr; 906 907 if (Idxs[0] == i) 908 C = ConstantFoldInsertValueInstruction(C, Val, Idxs.slice(1)); 909 910 Result.push_back(C); 911 } 912 913 if (StructType *ST = dyn_cast<StructType>(Agg->getType())) 914 return ConstantStruct::get(ST, Result); 915 if (ArrayType *AT = dyn_cast<ArrayType>(Agg->getType())) 916 return ConstantArray::get(AT, Result); 917 return ConstantVector::get(Result); 918 } 919 920 921 Constant *llvm::ConstantFoldBinaryInstruction(unsigned Opcode, 922 Constant *C1, Constant *C2) { 923 assert(Instruction::isBinaryOp(Opcode) && "Non-binary instruction detected"); 924 925 // Handle UndefValue up front. 926 if (isa<UndefValue>(C1) || isa<UndefValue>(C2)) { 927 switch (static_cast<Instruction::BinaryOps>(Opcode)) { 928 case Instruction::Xor: 929 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) 930 // Handle undef ^ undef -> 0 special case. This is a common 931 // idiom (misuse). 932 return Constant::getNullValue(C1->getType()); 933 LLVM_FALLTHROUGH; 934 case Instruction::Add: 935 case Instruction::Sub: 936 return UndefValue::get(C1->getType()); 937 case Instruction::And: 938 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) // undef & undef -> undef 939 return C1; 940 return Constant::getNullValue(C1->getType()); // undef & X -> 0 941 case Instruction::Mul: { 942 // undef * undef -> undef 943 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) 944 return C1; 945 const APInt *CV; 946 // X * undef -> undef if X is odd 947 if (match(C1, m_APInt(CV)) || match(C2, m_APInt(CV))) 948 if ((*CV)[0]) 949 return UndefValue::get(C1->getType()); 950 951 // X * undef -> 0 otherwise 952 return Constant::getNullValue(C1->getType()); 953 } 954 case Instruction::SDiv: 955 case Instruction::UDiv: 956 // X / undef -> undef 957 if (isa<UndefValue>(C2)) 958 return C2; 959 // undef / 0 -> undef 960 // undef / 1 -> undef 961 if (match(C2, m_Zero()) || match(C2, m_One())) 962 return C1; 963 // undef / X -> 0 otherwise 964 return Constant::getNullValue(C1->getType()); 965 case Instruction::URem: 966 case Instruction::SRem: 967 // X % undef -> undef 968 if (match(C2, m_Undef())) 969 return C2; 970 // undef % 0 -> undef 971 if (match(C2, m_Zero())) 972 return C1; 973 // undef % X -> 0 otherwise 974 return Constant::getNullValue(C1->getType()); 975 case Instruction::Or: // X | undef -> -1 976 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) // undef | undef -> undef 977 return C1; 978 return Constant::getAllOnesValue(C1->getType()); // undef | X -> ~0 979 case Instruction::LShr: 980 // X >>l undef -> undef 981 if (isa<UndefValue>(C2)) 982 return C2; 983 // undef >>l 0 -> undef 984 if (match(C2, m_Zero())) 985 return C1; 986 // undef >>l X -> 0 987 return Constant::getNullValue(C1->getType()); 988 case Instruction::AShr: 989 // X >>a undef -> undef 990 if (isa<UndefValue>(C2)) 991 return C2; 992 // undef >>a 0 -> undef 993 if (match(C2, m_Zero())) 994 return C1; 995 // TODO: undef >>a X -> undef if the shift is exact 996 // undef >>a X -> 0 997 return Constant::getNullValue(C1->getType()); 998 case Instruction::Shl: 999 // X << undef -> undef 1000 if (isa<UndefValue>(C2)) 1001 return C2; 1002 // undef << 0 -> undef 1003 if (match(C2, m_Zero())) 1004 return C1; 1005 // undef << X -> 0 1006 return Constant::getNullValue(C1->getType()); 1007 case Instruction::FAdd: 1008 case Instruction::FSub: 1009 case Instruction::FMul: 1010 case Instruction::FDiv: 1011 case Instruction::FRem: 1012 // TODO: UNDEF handling for binary float instructions. 1013 return nullptr; 1014 case Instruction::BinaryOpsEnd: 1015 llvm_unreachable("Invalid BinaryOp"); 1016 } 1017 } 1018 1019 // At this point neither constant should be an UndefValue. 1020 assert(!isa<UndefValue>(C1) && !isa<UndefValue>(C2) && 1021 "Unexpected UndefValue"); 1022 1023 // Handle simplifications when the RHS is a constant int. 1024 if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) { 1025 switch (Opcode) { 1026 case Instruction::Add: 1027 if (CI2->isZero()) return C1; // X + 0 == X 1028 break; 1029 case Instruction::Sub: 1030 if (CI2->isZero()) return C1; // X - 0 == X 1031 break; 1032 case Instruction::Mul: 1033 if (CI2->isZero()) return C2; // X * 0 == 0 1034 if (CI2->isOne()) 1035 return C1; // X * 1 == X 1036 break; 1037 case Instruction::UDiv: 1038 case Instruction::SDiv: 1039 if (CI2->isOne()) 1040 return C1; // X / 1 == X 1041 if (CI2->isZero()) 1042 return UndefValue::get(CI2->getType()); // X / 0 == undef 1043 break; 1044 case Instruction::URem: 1045 case Instruction::SRem: 1046 if (CI2->isOne()) 1047 return Constant::getNullValue(CI2->getType()); // X % 1 == 0 1048 if (CI2->isZero()) 1049 return UndefValue::get(CI2->getType()); // X % 0 == undef 1050 break; 1051 case Instruction::And: 1052 if (CI2->isZero()) return C2; // X & 0 == 0 1053 if (CI2->isMinusOne()) 1054 return C1; // X & -1 == X 1055 1056 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1057 // (zext i32 to i64) & 4294967295 -> (zext i32 to i64) 1058 if (CE1->getOpcode() == Instruction::ZExt) { 1059 unsigned DstWidth = CI2->getType()->getBitWidth(); 1060 unsigned SrcWidth = 1061 CE1->getOperand(0)->getType()->getPrimitiveSizeInBits(); 1062 APInt PossiblySetBits(APInt::getLowBitsSet(DstWidth, SrcWidth)); 1063 if ((PossiblySetBits & CI2->getValue()) == PossiblySetBits) 1064 return C1; 1065 } 1066 1067 // If and'ing the address of a global with a constant, fold it. 1068 if (CE1->getOpcode() == Instruction::PtrToInt && 1069 isa<GlobalValue>(CE1->getOperand(0))) { 1070 GlobalValue *GV = cast<GlobalValue>(CE1->getOperand(0)); 1071 1072 // Functions are at least 4-byte aligned. 1073 unsigned GVAlign = GV->getAlignment(); 1074 if (isa<Function>(GV)) 1075 GVAlign = std::max(GVAlign, 4U); 1076 1077 if (GVAlign > 1) { 1078 unsigned DstWidth = CI2->getType()->getBitWidth(); 1079 unsigned SrcWidth = std::min(DstWidth, Log2_32(GVAlign)); 1080 APInt BitsNotSet(APInt::getLowBitsSet(DstWidth, SrcWidth)); 1081 1082 // If checking bits we know are clear, return zero. 1083 if ((CI2->getValue() & BitsNotSet) == CI2->getValue()) 1084 return Constant::getNullValue(CI2->getType()); 1085 } 1086 } 1087 } 1088 break; 1089 case Instruction::Or: 1090 if (CI2->isZero()) return C1; // X | 0 == X 1091 if (CI2->isMinusOne()) 1092 return C2; // X | -1 == -1 1093 break; 1094 case Instruction::Xor: 1095 if (CI2->isZero()) return C1; // X ^ 0 == X 1096 1097 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1098 switch (CE1->getOpcode()) { 1099 default: break; 1100 case Instruction::ICmp: 1101 case Instruction::FCmp: 1102 // cmp pred ^ true -> cmp !pred 1103 assert(CI2->isOne()); 1104 CmpInst::Predicate pred = (CmpInst::Predicate)CE1->getPredicate(); 1105 pred = CmpInst::getInversePredicate(pred); 1106 return ConstantExpr::getCompare(pred, CE1->getOperand(0), 1107 CE1->getOperand(1)); 1108 } 1109 } 1110 break; 1111 case Instruction::AShr: 1112 // ashr (zext C to Ty), C2 -> lshr (zext C, CSA), C2 1113 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) 1114 if (CE1->getOpcode() == Instruction::ZExt) // Top bits known zero. 1115 return ConstantExpr::getLShr(C1, C2); 1116 break; 1117 } 1118 } else if (isa<ConstantInt>(C1)) { 1119 // If C1 is a ConstantInt and C2 is not, swap the operands. 1120 if (Instruction::isCommutative(Opcode)) 1121 return ConstantExpr::get(Opcode, C2, C1); 1122 } 1123 1124 if (ConstantInt *CI1 = dyn_cast<ConstantInt>(C1)) { 1125 if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) { 1126 const APInt &C1V = CI1->getValue(); 1127 const APInt &C2V = CI2->getValue(); 1128 switch (Opcode) { 1129 default: 1130 break; 1131 case Instruction::Add: 1132 return ConstantInt::get(CI1->getContext(), C1V + C2V); 1133 case Instruction::Sub: 1134 return ConstantInt::get(CI1->getContext(), C1V - C2V); 1135 case Instruction::Mul: 1136 return ConstantInt::get(CI1->getContext(), C1V * C2V); 1137 case Instruction::UDiv: 1138 assert(!CI2->isZero() && "Div by zero handled above"); 1139 return ConstantInt::get(CI1->getContext(), C1V.udiv(C2V)); 1140 case Instruction::SDiv: 1141 assert(!CI2->isZero() && "Div by zero handled above"); 1142 if (C2V.isAllOnesValue() && C1V.isMinSignedValue()) 1143 return UndefValue::get(CI1->getType()); // MIN_INT / -1 -> undef 1144 return ConstantInt::get(CI1->getContext(), C1V.sdiv(C2V)); 1145 case Instruction::URem: 1146 assert(!CI2->isZero() && "Div by zero handled above"); 1147 return ConstantInt::get(CI1->getContext(), C1V.urem(C2V)); 1148 case Instruction::SRem: 1149 assert(!CI2->isZero() && "Div by zero handled above"); 1150 if (C2V.isAllOnesValue() && C1V.isMinSignedValue()) 1151 return UndefValue::get(CI1->getType()); // MIN_INT % -1 -> undef 1152 return ConstantInt::get(CI1->getContext(), C1V.srem(C2V)); 1153 case Instruction::And: 1154 return ConstantInt::get(CI1->getContext(), C1V & C2V); 1155 case Instruction::Or: 1156 return ConstantInt::get(CI1->getContext(), C1V | C2V); 1157 case Instruction::Xor: 1158 return ConstantInt::get(CI1->getContext(), C1V ^ C2V); 1159 case Instruction::Shl: 1160 if (C2V.ult(C1V.getBitWidth())) 1161 return ConstantInt::get(CI1->getContext(), C1V.shl(C2V)); 1162 return UndefValue::get(C1->getType()); // too big shift is undef 1163 case Instruction::LShr: 1164 if (C2V.ult(C1V.getBitWidth())) 1165 return ConstantInt::get(CI1->getContext(), C1V.lshr(C2V)); 1166 return UndefValue::get(C1->getType()); // too big shift is undef 1167 case Instruction::AShr: 1168 if (C2V.ult(C1V.getBitWidth())) 1169 return ConstantInt::get(CI1->getContext(), C1V.ashr(C2V)); 1170 return UndefValue::get(C1->getType()); // too big shift is undef 1171 } 1172 } 1173 1174 switch (Opcode) { 1175 case Instruction::SDiv: 1176 case Instruction::UDiv: 1177 case Instruction::URem: 1178 case Instruction::SRem: 1179 case Instruction::LShr: 1180 case Instruction::AShr: 1181 case Instruction::Shl: 1182 if (CI1->isZero()) return C1; 1183 break; 1184 default: 1185 break; 1186 } 1187 } else if (ConstantFP *CFP1 = dyn_cast<ConstantFP>(C1)) { 1188 if (ConstantFP *CFP2 = dyn_cast<ConstantFP>(C2)) { 1189 const APFloat &C1V = CFP1->getValueAPF(); 1190 const APFloat &C2V = CFP2->getValueAPF(); 1191 APFloat C3V = C1V; // copy for modification 1192 switch (Opcode) { 1193 default: 1194 break; 1195 case Instruction::FAdd: 1196 (void)C3V.add(C2V, APFloat::rmNearestTiesToEven); 1197 return ConstantFP::get(C1->getContext(), C3V); 1198 case Instruction::FSub: 1199 (void)C3V.subtract(C2V, APFloat::rmNearestTiesToEven); 1200 return ConstantFP::get(C1->getContext(), C3V); 1201 case Instruction::FMul: 1202 (void)C3V.multiply(C2V, APFloat::rmNearestTiesToEven); 1203 return ConstantFP::get(C1->getContext(), C3V); 1204 case Instruction::FDiv: 1205 (void)C3V.divide(C2V, APFloat::rmNearestTiesToEven); 1206 return ConstantFP::get(C1->getContext(), C3V); 1207 case Instruction::FRem: 1208 (void)C3V.mod(C2V); 1209 return ConstantFP::get(C1->getContext(), C3V); 1210 } 1211 } 1212 } else if (VectorType *VTy = dyn_cast<VectorType>(C1->getType())) { 1213 // Perform elementwise folding. 1214 SmallVector<Constant*, 16> Result; 1215 Type *Ty = IntegerType::get(VTy->getContext(), 32); 1216 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1217 Constant *ExtractIdx = ConstantInt::get(Ty, i); 1218 Constant *LHS = ConstantExpr::getExtractElement(C1, ExtractIdx); 1219 Constant *RHS = ConstantExpr::getExtractElement(C2, ExtractIdx); 1220 1221 // If any element of a divisor vector is zero, the whole op is undef. 1222 if ((Opcode == Instruction::SDiv || Opcode == Instruction::UDiv || 1223 Opcode == Instruction::SRem || Opcode == Instruction::URem) && 1224 RHS->isNullValue()) 1225 return UndefValue::get(VTy); 1226 1227 Result.push_back(ConstantExpr::get(Opcode, LHS, RHS)); 1228 } 1229 1230 return ConstantVector::get(Result); 1231 } 1232 1233 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1234 // There are many possible foldings we could do here. We should probably 1235 // at least fold add of a pointer with an integer into the appropriate 1236 // getelementptr. This will improve alias analysis a bit. 1237 1238 // Given ((a + b) + c), if (b + c) folds to something interesting, return 1239 // (a + (b + c)). 1240 if (Instruction::isAssociative(Opcode) && CE1->getOpcode() == Opcode) { 1241 Constant *T = ConstantExpr::get(Opcode, CE1->getOperand(1), C2); 1242 if (!isa<ConstantExpr>(T) || cast<ConstantExpr>(T)->getOpcode() != Opcode) 1243 return ConstantExpr::get(Opcode, CE1->getOperand(0), T); 1244 } 1245 } else if (isa<ConstantExpr>(C2)) { 1246 // If C2 is a constant expr and C1 isn't, flop them around and fold the 1247 // other way if possible. 1248 if (Instruction::isCommutative(Opcode)) 1249 return ConstantFoldBinaryInstruction(Opcode, C2, C1); 1250 } 1251 1252 // i1 can be simplified in many cases. 1253 if (C1->getType()->isIntegerTy(1)) { 1254 switch (Opcode) { 1255 case Instruction::Add: 1256 case Instruction::Sub: 1257 return ConstantExpr::getXor(C1, C2); 1258 case Instruction::Mul: 1259 return ConstantExpr::getAnd(C1, C2); 1260 case Instruction::Shl: 1261 case Instruction::LShr: 1262 case Instruction::AShr: 1263 // We can assume that C2 == 0. If it were one the result would be 1264 // undefined because the shift value is as large as the bitwidth. 1265 return C1; 1266 case Instruction::SDiv: 1267 case Instruction::UDiv: 1268 // We can assume that C2 == 1. If it were zero the result would be 1269 // undefined through division by zero. 1270 return C1; 1271 case Instruction::URem: 1272 case Instruction::SRem: 1273 // We can assume that C2 == 1. If it were zero the result would be 1274 // undefined through division by zero. 1275 return ConstantInt::getFalse(C1->getContext()); 1276 default: 1277 break; 1278 } 1279 } 1280 1281 // We don't know how to fold this. 1282 return nullptr; 1283 } 1284 1285 /// This type is zero-sized if it's an array or structure of zero-sized types. 1286 /// The only leaf zero-sized type is an empty structure. 1287 static bool isMaybeZeroSizedType(Type *Ty) { 1288 if (StructType *STy = dyn_cast<StructType>(Ty)) { 1289 if (STy->isOpaque()) return true; // Can't say. 1290 1291 // If all of elements have zero size, this does too. 1292 for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) 1293 if (!isMaybeZeroSizedType(STy->getElementType(i))) return false; 1294 return true; 1295 1296 } else if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 1297 return isMaybeZeroSizedType(ATy->getElementType()); 1298 } 1299 return false; 1300 } 1301 1302 /// Compare the two constants as though they were getelementptr indices. 1303 /// This allows coercion of the types to be the same thing. 1304 /// 1305 /// If the two constants are the "same" (after coercion), return 0. If the 1306 /// first is less than the second, return -1, if the second is less than the 1307 /// first, return 1. If the constants are not integral, return -2. 1308 /// 1309 static int IdxCompare(Constant *C1, Constant *C2, Type *ElTy) { 1310 if (C1 == C2) return 0; 1311 1312 // Ok, we found a different index. If they are not ConstantInt, we can't do 1313 // anything with them. 1314 if (!isa<ConstantInt>(C1) || !isa<ConstantInt>(C2)) 1315 return -2; // don't know! 1316 1317 // We cannot compare the indices if they don't fit in an int64_t. 1318 if (cast<ConstantInt>(C1)->getValue().getActiveBits() > 64 || 1319 cast<ConstantInt>(C2)->getValue().getActiveBits() > 64) 1320 return -2; // don't know! 1321 1322 // Ok, we have two differing integer indices. Sign extend them to be the same 1323 // type. 1324 int64_t C1Val = cast<ConstantInt>(C1)->getSExtValue(); 1325 int64_t C2Val = cast<ConstantInt>(C2)->getSExtValue(); 1326 1327 if (C1Val == C2Val) return 0; // They are equal 1328 1329 // If the type being indexed over is really just a zero sized type, there is 1330 // no pointer difference being made here. 1331 if (isMaybeZeroSizedType(ElTy)) 1332 return -2; // dunno. 1333 1334 // If they are really different, now that they are the same type, then we 1335 // found a difference! 1336 if (C1Val < C2Val) 1337 return -1; 1338 else 1339 return 1; 1340 } 1341 1342 /// This function determines if there is anything we can decide about the two 1343 /// constants provided. This doesn't need to handle simple things like 1344 /// ConstantFP comparisons, but should instead handle ConstantExprs. 1345 /// If we can determine that the two constants have a particular relation to 1346 /// each other, we should return the corresponding FCmpInst predicate, 1347 /// otherwise return FCmpInst::BAD_FCMP_PREDICATE. This is used below in 1348 /// ConstantFoldCompareInstruction. 1349 /// 1350 /// To simplify this code we canonicalize the relation so that the first 1351 /// operand is always the most "complex" of the two. We consider ConstantFP 1352 /// to be the simplest, and ConstantExprs to be the most complex. 1353 static FCmpInst::Predicate evaluateFCmpRelation(Constant *V1, Constant *V2) { 1354 assert(V1->getType() == V2->getType() && 1355 "Cannot compare values of different types!"); 1356 1357 // Handle degenerate case quickly 1358 if (V1 == V2) return FCmpInst::FCMP_OEQ; 1359 1360 if (!isa<ConstantExpr>(V1)) { 1361 if (!isa<ConstantExpr>(V2)) { 1362 // Simple case, use the standard constant folder. 1363 ConstantInt *R = nullptr; 1364 R = dyn_cast<ConstantInt>( 1365 ConstantExpr::getFCmp(FCmpInst::FCMP_OEQ, V1, V2)); 1366 if (R && !R->isZero()) 1367 return FCmpInst::FCMP_OEQ; 1368 R = dyn_cast<ConstantInt>( 1369 ConstantExpr::getFCmp(FCmpInst::FCMP_OLT, V1, V2)); 1370 if (R && !R->isZero()) 1371 return FCmpInst::FCMP_OLT; 1372 R = dyn_cast<ConstantInt>( 1373 ConstantExpr::getFCmp(FCmpInst::FCMP_OGT, V1, V2)); 1374 if (R && !R->isZero()) 1375 return FCmpInst::FCMP_OGT; 1376 1377 // Nothing more we can do 1378 return FCmpInst::BAD_FCMP_PREDICATE; 1379 } 1380 1381 // If the first operand is simple and second is ConstantExpr, swap operands. 1382 FCmpInst::Predicate SwappedRelation = evaluateFCmpRelation(V2, V1); 1383 if (SwappedRelation != FCmpInst::BAD_FCMP_PREDICATE) 1384 return FCmpInst::getSwappedPredicate(SwappedRelation); 1385 } else { 1386 // Ok, the LHS is known to be a constantexpr. The RHS can be any of a 1387 // constantexpr or a simple constant. 1388 ConstantExpr *CE1 = cast<ConstantExpr>(V1); 1389 switch (CE1->getOpcode()) { 1390 case Instruction::FPTrunc: 1391 case Instruction::FPExt: 1392 case Instruction::UIToFP: 1393 case Instruction::SIToFP: 1394 // We might be able to do something with these but we don't right now. 1395 break; 1396 default: 1397 break; 1398 } 1399 } 1400 // There are MANY other foldings that we could perform here. They will 1401 // probably be added on demand, as they seem needed. 1402 return FCmpInst::BAD_FCMP_PREDICATE; 1403 } 1404 1405 static ICmpInst::Predicate areGlobalsPotentiallyEqual(const GlobalValue *GV1, 1406 const GlobalValue *GV2) { 1407 auto isGlobalUnsafeForEquality = [](const GlobalValue *GV) { 1408 if (GV->hasExternalWeakLinkage() || GV->hasWeakAnyLinkage()) 1409 return true; 1410 if (const auto *GVar = dyn_cast<GlobalVariable>(GV)) { 1411 Type *Ty = GVar->getValueType(); 1412 // A global with opaque type might end up being zero sized. 1413 if (!Ty->isSized()) 1414 return true; 1415 // A global with an empty type might lie at the address of any other 1416 // global. 1417 if (Ty->isEmptyTy()) 1418 return true; 1419 } 1420 return false; 1421 }; 1422 // Don't try to decide equality of aliases. 1423 if (!isa<GlobalAlias>(GV1) && !isa<GlobalAlias>(GV2)) 1424 if (!isGlobalUnsafeForEquality(GV1) && !isGlobalUnsafeForEquality(GV2)) 1425 return ICmpInst::ICMP_NE; 1426 return ICmpInst::BAD_ICMP_PREDICATE; 1427 } 1428 1429 /// This function determines if there is anything we can decide about the two 1430 /// constants provided. This doesn't need to handle simple things like integer 1431 /// comparisons, but should instead handle ConstantExprs and GlobalValues. 1432 /// If we can determine that the two constants have a particular relation to 1433 /// each other, we should return the corresponding ICmp predicate, otherwise 1434 /// return ICmpInst::BAD_ICMP_PREDICATE. 1435 /// 1436 /// To simplify this code we canonicalize the relation so that the first 1437 /// operand is always the most "complex" of the two. We consider simple 1438 /// constants (like ConstantInt) to be the simplest, followed by 1439 /// GlobalValues, followed by ConstantExpr's (the most complex). 1440 /// 1441 static ICmpInst::Predicate evaluateICmpRelation(Constant *V1, Constant *V2, 1442 bool isSigned) { 1443 assert(V1->getType() == V2->getType() && 1444 "Cannot compare different types of values!"); 1445 if (V1 == V2) return ICmpInst::ICMP_EQ; 1446 1447 if (!isa<ConstantExpr>(V1) && !isa<GlobalValue>(V1) && 1448 !isa<BlockAddress>(V1)) { 1449 if (!isa<GlobalValue>(V2) && !isa<ConstantExpr>(V2) && 1450 !isa<BlockAddress>(V2)) { 1451 // We distilled this down to a simple case, use the standard constant 1452 // folder. 1453 ConstantInt *R = nullptr; 1454 ICmpInst::Predicate pred = ICmpInst::ICMP_EQ; 1455 R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2)); 1456 if (R && !R->isZero()) 1457 return pred; 1458 pred = isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; 1459 R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2)); 1460 if (R && !R->isZero()) 1461 return pred; 1462 pred = isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1463 R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2)); 1464 if (R && !R->isZero()) 1465 return pred; 1466 1467 // If we couldn't figure it out, bail. 1468 return ICmpInst::BAD_ICMP_PREDICATE; 1469 } 1470 1471 // If the first operand is simple, swap operands. 1472 ICmpInst::Predicate SwappedRelation = 1473 evaluateICmpRelation(V2, V1, isSigned); 1474 if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE) 1475 return ICmpInst::getSwappedPredicate(SwappedRelation); 1476 1477 } else if (const GlobalValue *GV = dyn_cast<GlobalValue>(V1)) { 1478 if (isa<ConstantExpr>(V2)) { // Swap as necessary. 1479 ICmpInst::Predicate SwappedRelation = 1480 evaluateICmpRelation(V2, V1, isSigned); 1481 if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE) 1482 return ICmpInst::getSwappedPredicate(SwappedRelation); 1483 return ICmpInst::BAD_ICMP_PREDICATE; 1484 } 1485 1486 // Now we know that the RHS is a GlobalValue, BlockAddress or simple 1487 // constant (which, since the types must match, means that it's a 1488 // ConstantPointerNull). 1489 if (const GlobalValue *GV2 = dyn_cast<GlobalValue>(V2)) { 1490 return areGlobalsPotentiallyEqual(GV, GV2); 1491 } else if (isa<BlockAddress>(V2)) { 1492 return ICmpInst::ICMP_NE; // Globals never equal labels. 1493 } else { 1494 assert(isa<ConstantPointerNull>(V2) && "Canonicalization guarantee!"); 1495 // GlobalVals can never be null unless they have external weak linkage. 1496 // We don't try to evaluate aliases here. 1497 if (!GV->hasExternalWeakLinkage() && !isa<GlobalAlias>(GV)) 1498 return ICmpInst::ICMP_NE; 1499 } 1500 } else if (const BlockAddress *BA = dyn_cast<BlockAddress>(V1)) { 1501 if (isa<ConstantExpr>(V2)) { // Swap as necessary. 1502 ICmpInst::Predicate SwappedRelation = 1503 evaluateICmpRelation(V2, V1, isSigned); 1504 if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE) 1505 return ICmpInst::getSwappedPredicate(SwappedRelation); 1506 return ICmpInst::BAD_ICMP_PREDICATE; 1507 } 1508 1509 // Now we know that the RHS is a GlobalValue, BlockAddress or simple 1510 // constant (which, since the types must match, means that it is a 1511 // ConstantPointerNull). 1512 if (const BlockAddress *BA2 = dyn_cast<BlockAddress>(V2)) { 1513 // Block address in another function can't equal this one, but block 1514 // addresses in the current function might be the same if blocks are 1515 // empty. 1516 if (BA2->getFunction() != BA->getFunction()) 1517 return ICmpInst::ICMP_NE; 1518 } else { 1519 // Block addresses aren't null, don't equal the address of globals. 1520 assert((isa<ConstantPointerNull>(V2) || isa<GlobalValue>(V2)) && 1521 "Canonicalization guarantee!"); 1522 return ICmpInst::ICMP_NE; 1523 } 1524 } else { 1525 // Ok, the LHS is known to be a constantexpr. The RHS can be any of a 1526 // constantexpr, a global, block address, or a simple constant. 1527 ConstantExpr *CE1 = cast<ConstantExpr>(V1); 1528 Constant *CE1Op0 = CE1->getOperand(0); 1529 1530 switch (CE1->getOpcode()) { 1531 case Instruction::Trunc: 1532 case Instruction::FPTrunc: 1533 case Instruction::FPExt: 1534 case Instruction::FPToUI: 1535 case Instruction::FPToSI: 1536 break; // We can't evaluate floating point casts or truncations. 1537 1538 case Instruction::UIToFP: 1539 case Instruction::SIToFP: 1540 case Instruction::BitCast: 1541 case Instruction::ZExt: 1542 case Instruction::SExt: 1543 // We can't evaluate floating point casts or truncations. 1544 if (CE1Op0->getType()->isFloatingPointTy()) 1545 break; 1546 1547 // If the cast is not actually changing bits, and the second operand is a 1548 // null pointer, do the comparison with the pre-casted value. 1549 if (V2->isNullValue() && 1550 (CE1->getType()->isPointerTy() || CE1->getType()->isIntegerTy())) { 1551 if (CE1->getOpcode() == Instruction::ZExt) isSigned = false; 1552 if (CE1->getOpcode() == Instruction::SExt) isSigned = true; 1553 return evaluateICmpRelation(CE1Op0, 1554 Constant::getNullValue(CE1Op0->getType()), 1555 isSigned); 1556 } 1557 break; 1558 1559 case Instruction::GetElementPtr: { 1560 GEPOperator *CE1GEP = cast<GEPOperator>(CE1); 1561 // Ok, since this is a getelementptr, we know that the constant has a 1562 // pointer type. Check the various cases. 1563 if (isa<ConstantPointerNull>(V2)) { 1564 // If we are comparing a GEP to a null pointer, check to see if the base 1565 // of the GEP equals the null pointer. 1566 if (const GlobalValue *GV = dyn_cast<GlobalValue>(CE1Op0)) { 1567 if (GV->hasExternalWeakLinkage()) 1568 // Weak linkage GVals could be zero or not. We're comparing that 1569 // to null pointer so its greater-or-equal 1570 return isSigned ? ICmpInst::ICMP_SGE : ICmpInst::ICMP_UGE; 1571 else 1572 // If its not weak linkage, the GVal must have a non-zero address 1573 // so the result is greater-than 1574 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1575 } else if (isa<ConstantPointerNull>(CE1Op0)) { 1576 // If we are indexing from a null pointer, check to see if we have any 1577 // non-zero indices. 1578 for (unsigned i = 1, e = CE1->getNumOperands(); i != e; ++i) 1579 if (!CE1->getOperand(i)->isNullValue()) 1580 // Offsetting from null, must not be equal. 1581 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1582 // Only zero indexes from null, must still be zero. 1583 return ICmpInst::ICMP_EQ; 1584 } 1585 // Otherwise, we can't really say if the first operand is null or not. 1586 } else if (const GlobalValue *GV2 = dyn_cast<GlobalValue>(V2)) { 1587 if (isa<ConstantPointerNull>(CE1Op0)) { 1588 if (GV2->hasExternalWeakLinkage()) 1589 // Weak linkage GVals could be zero or not. We're comparing it to 1590 // a null pointer, so its less-or-equal 1591 return isSigned ? ICmpInst::ICMP_SLE : ICmpInst::ICMP_ULE; 1592 else 1593 // If its not weak linkage, the GVal must have a non-zero address 1594 // so the result is less-than 1595 return isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; 1596 } else if (const GlobalValue *GV = dyn_cast<GlobalValue>(CE1Op0)) { 1597 if (GV == GV2) { 1598 // If this is a getelementptr of the same global, then it must be 1599 // different. Because the types must match, the getelementptr could 1600 // only have at most one index, and because we fold getelementptr's 1601 // with a single zero index, it must be nonzero. 1602 assert(CE1->getNumOperands() == 2 && 1603 !CE1->getOperand(1)->isNullValue() && 1604 "Surprising getelementptr!"); 1605 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1606 } else { 1607 if (CE1GEP->hasAllZeroIndices()) 1608 return areGlobalsPotentiallyEqual(GV, GV2); 1609 return ICmpInst::BAD_ICMP_PREDICATE; 1610 } 1611 } 1612 } else { 1613 ConstantExpr *CE2 = cast<ConstantExpr>(V2); 1614 Constant *CE2Op0 = CE2->getOperand(0); 1615 1616 // There are MANY other foldings that we could perform here. They will 1617 // probably be added on demand, as they seem needed. 1618 switch (CE2->getOpcode()) { 1619 default: break; 1620 case Instruction::GetElementPtr: 1621 // By far the most common case to handle is when the base pointers are 1622 // obviously to the same global. 1623 if (isa<GlobalValue>(CE1Op0) && isa<GlobalValue>(CE2Op0)) { 1624 // Don't know relative ordering, but check for inequality. 1625 if (CE1Op0 != CE2Op0) { 1626 GEPOperator *CE2GEP = cast<GEPOperator>(CE2); 1627 if (CE1GEP->hasAllZeroIndices() && CE2GEP->hasAllZeroIndices()) 1628 return areGlobalsPotentiallyEqual(cast<GlobalValue>(CE1Op0), 1629 cast<GlobalValue>(CE2Op0)); 1630 return ICmpInst::BAD_ICMP_PREDICATE; 1631 } 1632 // Ok, we know that both getelementptr instructions are based on the 1633 // same global. From this, we can precisely determine the relative 1634 // ordering of the resultant pointers. 1635 unsigned i = 1; 1636 1637 // The logic below assumes that the result of the comparison 1638 // can be determined by finding the first index that differs. 1639 // This doesn't work if there is over-indexing in any 1640 // subsequent indices, so check for that case first. 1641 if (!CE1->isGEPWithNoNotionalOverIndexing() || 1642 !CE2->isGEPWithNoNotionalOverIndexing()) 1643 return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal. 1644 1645 // Compare all of the operands the GEP's have in common. 1646 gep_type_iterator GTI = gep_type_begin(CE1); 1647 for (;i != CE1->getNumOperands() && i != CE2->getNumOperands(); 1648 ++i, ++GTI) 1649 switch (IdxCompare(CE1->getOperand(i), 1650 CE2->getOperand(i), GTI.getIndexedType())) { 1651 case -1: return isSigned ? ICmpInst::ICMP_SLT:ICmpInst::ICMP_ULT; 1652 case 1: return isSigned ? ICmpInst::ICMP_SGT:ICmpInst::ICMP_UGT; 1653 case -2: return ICmpInst::BAD_ICMP_PREDICATE; 1654 } 1655 1656 // Ok, we ran out of things they have in common. If any leftovers 1657 // are non-zero then we have a difference, otherwise we are equal. 1658 for (; i < CE1->getNumOperands(); ++i) 1659 if (!CE1->getOperand(i)->isNullValue()) { 1660 if (isa<ConstantInt>(CE1->getOperand(i))) 1661 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1662 else 1663 return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal. 1664 } 1665 1666 for (; i < CE2->getNumOperands(); ++i) 1667 if (!CE2->getOperand(i)->isNullValue()) { 1668 if (isa<ConstantInt>(CE2->getOperand(i))) 1669 return isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; 1670 else 1671 return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal. 1672 } 1673 return ICmpInst::ICMP_EQ; 1674 } 1675 } 1676 } 1677 break; 1678 } 1679 default: 1680 break; 1681 } 1682 } 1683 1684 return ICmpInst::BAD_ICMP_PREDICATE; 1685 } 1686 1687 Constant *llvm::ConstantFoldCompareInstruction(unsigned short pred, 1688 Constant *C1, Constant *C2) { 1689 Type *ResultTy; 1690 if (VectorType *VT = dyn_cast<VectorType>(C1->getType())) 1691 ResultTy = VectorType::get(Type::getInt1Ty(C1->getContext()), 1692 VT->getNumElements()); 1693 else 1694 ResultTy = Type::getInt1Ty(C1->getContext()); 1695 1696 // Fold FCMP_FALSE/FCMP_TRUE unconditionally. 1697 if (pred == FCmpInst::FCMP_FALSE) 1698 return Constant::getNullValue(ResultTy); 1699 1700 if (pred == FCmpInst::FCMP_TRUE) 1701 return Constant::getAllOnesValue(ResultTy); 1702 1703 // Handle some degenerate cases first 1704 if (isa<UndefValue>(C1) || isa<UndefValue>(C2)) { 1705 CmpInst::Predicate Predicate = CmpInst::Predicate(pred); 1706 bool isIntegerPredicate = ICmpInst::isIntPredicate(Predicate); 1707 // For EQ and NE, we can always pick a value for the undef to make the 1708 // predicate pass or fail, so we can return undef. 1709 // Also, if both operands are undef, we can return undef for int comparison. 1710 if (ICmpInst::isEquality(Predicate) || (isIntegerPredicate && C1 == C2)) 1711 return UndefValue::get(ResultTy); 1712 1713 // Otherwise, for integer compare, pick the same value as the non-undef 1714 // operand, and fold it to true or false. 1715 if (isIntegerPredicate) 1716 return ConstantInt::get(ResultTy, CmpInst::isTrueWhenEqual(Predicate)); 1717 1718 // Choosing NaN for the undef will always make unordered comparison succeed 1719 // and ordered comparison fails. 1720 return ConstantInt::get(ResultTy, CmpInst::isUnordered(Predicate)); 1721 } 1722 1723 // icmp eq/ne(null,GV) -> false/true 1724 if (C1->isNullValue()) { 1725 if (const GlobalValue *GV = dyn_cast<GlobalValue>(C2)) 1726 // Don't try to evaluate aliases. External weak GV can be null. 1727 if (!isa<GlobalAlias>(GV) && !GV->hasExternalWeakLinkage()) { 1728 if (pred == ICmpInst::ICMP_EQ) 1729 return ConstantInt::getFalse(C1->getContext()); 1730 else if (pred == ICmpInst::ICMP_NE) 1731 return ConstantInt::getTrue(C1->getContext()); 1732 } 1733 // icmp eq/ne(GV,null) -> false/true 1734 } else if (C2->isNullValue()) { 1735 if (const GlobalValue *GV = dyn_cast<GlobalValue>(C1)) 1736 // Don't try to evaluate aliases. External weak GV can be null. 1737 if (!isa<GlobalAlias>(GV) && !GV->hasExternalWeakLinkage()) { 1738 if (pred == ICmpInst::ICMP_EQ) 1739 return ConstantInt::getFalse(C1->getContext()); 1740 else if (pred == ICmpInst::ICMP_NE) 1741 return ConstantInt::getTrue(C1->getContext()); 1742 } 1743 } 1744 1745 // If the comparison is a comparison between two i1's, simplify it. 1746 if (C1->getType()->isIntegerTy(1)) { 1747 switch(pred) { 1748 case ICmpInst::ICMP_EQ: 1749 if (isa<ConstantInt>(C2)) 1750 return ConstantExpr::getXor(C1, ConstantExpr::getNot(C2)); 1751 return ConstantExpr::getXor(ConstantExpr::getNot(C1), C2); 1752 case ICmpInst::ICMP_NE: 1753 return ConstantExpr::getXor(C1, C2); 1754 default: 1755 break; 1756 } 1757 } 1758 1759 if (isa<ConstantInt>(C1) && isa<ConstantInt>(C2)) { 1760 const APInt &V1 = cast<ConstantInt>(C1)->getValue(); 1761 const APInt &V2 = cast<ConstantInt>(C2)->getValue(); 1762 switch (pred) { 1763 default: llvm_unreachable("Invalid ICmp Predicate"); 1764 case ICmpInst::ICMP_EQ: return ConstantInt::get(ResultTy, V1 == V2); 1765 case ICmpInst::ICMP_NE: return ConstantInt::get(ResultTy, V1 != V2); 1766 case ICmpInst::ICMP_SLT: return ConstantInt::get(ResultTy, V1.slt(V2)); 1767 case ICmpInst::ICMP_SGT: return ConstantInt::get(ResultTy, V1.sgt(V2)); 1768 case ICmpInst::ICMP_SLE: return ConstantInt::get(ResultTy, V1.sle(V2)); 1769 case ICmpInst::ICMP_SGE: return ConstantInt::get(ResultTy, V1.sge(V2)); 1770 case ICmpInst::ICMP_ULT: return ConstantInt::get(ResultTy, V1.ult(V2)); 1771 case ICmpInst::ICMP_UGT: return ConstantInt::get(ResultTy, V1.ugt(V2)); 1772 case ICmpInst::ICMP_ULE: return ConstantInt::get(ResultTy, V1.ule(V2)); 1773 case ICmpInst::ICMP_UGE: return ConstantInt::get(ResultTy, V1.uge(V2)); 1774 } 1775 } else if (isa<ConstantFP>(C1) && isa<ConstantFP>(C2)) { 1776 const APFloat &C1V = cast<ConstantFP>(C1)->getValueAPF(); 1777 const APFloat &C2V = cast<ConstantFP>(C2)->getValueAPF(); 1778 APFloat::cmpResult R = C1V.compare(C2V); 1779 switch (pred) { 1780 default: llvm_unreachable("Invalid FCmp Predicate"); 1781 case FCmpInst::FCMP_FALSE: return Constant::getNullValue(ResultTy); 1782 case FCmpInst::FCMP_TRUE: return Constant::getAllOnesValue(ResultTy); 1783 case FCmpInst::FCMP_UNO: 1784 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered); 1785 case FCmpInst::FCMP_ORD: 1786 return ConstantInt::get(ResultTy, R!=APFloat::cmpUnordered); 1787 case FCmpInst::FCMP_UEQ: 1788 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered || 1789 R==APFloat::cmpEqual); 1790 case FCmpInst::FCMP_OEQ: 1791 return ConstantInt::get(ResultTy, R==APFloat::cmpEqual); 1792 case FCmpInst::FCMP_UNE: 1793 return ConstantInt::get(ResultTy, R!=APFloat::cmpEqual); 1794 case FCmpInst::FCMP_ONE: 1795 return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan || 1796 R==APFloat::cmpGreaterThan); 1797 case FCmpInst::FCMP_ULT: 1798 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered || 1799 R==APFloat::cmpLessThan); 1800 case FCmpInst::FCMP_OLT: 1801 return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan); 1802 case FCmpInst::FCMP_UGT: 1803 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered || 1804 R==APFloat::cmpGreaterThan); 1805 case FCmpInst::FCMP_OGT: 1806 return ConstantInt::get(ResultTy, R==APFloat::cmpGreaterThan); 1807 case FCmpInst::FCMP_ULE: 1808 return ConstantInt::get(ResultTy, R!=APFloat::cmpGreaterThan); 1809 case FCmpInst::FCMP_OLE: 1810 return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan || 1811 R==APFloat::cmpEqual); 1812 case FCmpInst::FCMP_UGE: 1813 return ConstantInt::get(ResultTy, R!=APFloat::cmpLessThan); 1814 case FCmpInst::FCMP_OGE: 1815 return ConstantInt::get(ResultTy, R==APFloat::cmpGreaterThan || 1816 R==APFloat::cmpEqual); 1817 } 1818 } else if (C1->getType()->isVectorTy()) { 1819 // If we can constant fold the comparison of each element, constant fold 1820 // the whole vector comparison. 1821 SmallVector<Constant*, 4> ResElts; 1822 Type *Ty = IntegerType::get(C1->getContext(), 32); 1823 // Compare the elements, producing an i1 result or constant expr. 1824 for (unsigned i = 0, e = C1->getType()->getVectorNumElements(); i != e;++i){ 1825 Constant *C1E = 1826 ConstantExpr::getExtractElement(C1, ConstantInt::get(Ty, i)); 1827 Constant *C2E = 1828 ConstantExpr::getExtractElement(C2, ConstantInt::get(Ty, i)); 1829 1830 ResElts.push_back(ConstantExpr::getCompare(pred, C1E, C2E)); 1831 } 1832 1833 return ConstantVector::get(ResElts); 1834 } 1835 1836 if (C1->getType()->isFloatingPointTy() && 1837 // Only call evaluateFCmpRelation if we have a constant expr to avoid 1838 // infinite recursive loop 1839 (isa<ConstantExpr>(C1) || isa<ConstantExpr>(C2))) { 1840 int Result = -1; // -1 = unknown, 0 = known false, 1 = known true. 1841 switch (evaluateFCmpRelation(C1, C2)) { 1842 default: llvm_unreachable("Unknown relation!"); 1843 case FCmpInst::FCMP_UNO: 1844 case FCmpInst::FCMP_ORD: 1845 case FCmpInst::FCMP_UEQ: 1846 case FCmpInst::FCMP_UNE: 1847 case FCmpInst::FCMP_ULT: 1848 case FCmpInst::FCMP_UGT: 1849 case FCmpInst::FCMP_ULE: 1850 case FCmpInst::FCMP_UGE: 1851 case FCmpInst::FCMP_TRUE: 1852 case FCmpInst::FCMP_FALSE: 1853 case FCmpInst::BAD_FCMP_PREDICATE: 1854 break; // Couldn't determine anything about these constants. 1855 case FCmpInst::FCMP_OEQ: // We know that C1 == C2 1856 Result = (pred == FCmpInst::FCMP_UEQ || pred == FCmpInst::FCMP_OEQ || 1857 pred == FCmpInst::FCMP_ULE || pred == FCmpInst::FCMP_OLE || 1858 pred == FCmpInst::FCMP_UGE || pred == FCmpInst::FCMP_OGE); 1859 break; 1860 case FCmpInst::FCMP_OLT: // We know that C1 < C2 1861 Result = (pred == FCmpInst::FCMP_UNE || pred == FCmpInst::FCMP_ONE || 1862 pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT || 1863 pred == FCmpInst::FCMP_ULE || pred == FCmpInst::FCMP_OLE); 1864 break; 1865 case FCmpInst::FCMP_OGT: // We know that C1 > C2 1866 Result = (pred == FCmpInst::FCMP_UNE || pred == FCmpInst::FCMP_ONE || 1867 pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT || 1868 pred == FCmpInst::FCMP_UGE || pred == FCmpInst::FCMP_OGE); 1869 break; 1870 case FCmpInst::FCMP_OLE: // We know that C1 <= C2 1871 // We can only partially decide this relation. 1872 if (pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT) 1873 Result = 0; 1874 else if (pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT) 1875 Result = 1; 1876 break; 1877 case FCmpInst::FCMP_OGE: // We known that C1 >= C2 1878 // We can only partially decide this relation. 1879 if (pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT) 1880 Result = 0; 1881 else if (pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT) 1882 Result = 1; 1883 break; 1884 case FCmpInst::FCMP_ONE: // We know that C1 != C2 1885 // We can only partially decide this relation. 1886 if (pred == FCmpInst::FCMP_OEQ || pred == FCmpInst::FCMP_UEQ) 1887 Result = 0; 1888 else if (pred == FCmpInst::FCMP_ONE || pred == FCmpInst::FCMP_UNE) 1889 Result = 1; 1890 break; 1891 } 1892 1893 // If we evaluated the result, return it now. 1894 if (Result != -1) 1895 return ConstantInt::get(ResultTy, Result); 1896 1897 } else { 1898 // Evaluate the relation between the two constants, per the predicate. 1899 int Result = -1; // -1 = unknown, 0 = known false, 1 = known true. 1900 switch (evaluateICmpRelation(C1, C2, 1901 CmpInst::isSigned((CmpInst::Predicate)pred))) { 1902 default: llvm_unreachable("Unknown relational!"); 1903 case ICmpInst::BAD_ICMP_PREDICATE: 1904 break; // Couldn't determine anything about these constants. 1905 case ICmpInst::ICMP_EQ: // We know the constants are equal! 1906 // If we know the constants are equal, we can decide the result of this 1907 // computation precisely. 1908 Result = ICmpInst::isTrueWhenEqual((ICmpInst::Predicate)pred); 1909 break; 1910 case ICmpInst::ICMP_ULT: 1911 switch (pred) { 1912 case ICmpInst::ICMP_ULT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_ULE: 1913 Result = 1; break; 1914 case ICmpInst::ICMP_UGT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_UGE: 1915 Result = 0; break; 1916 } 1917 break; 1918 case ICmpInst::ICMP_SLT: 1919 switch (pred) { 1920 case ICmpInst::ICMP_SLT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_SLE: 1921 Result = 1; break; 1922 case ICmpInst::ICMP_SGT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_SGE: 1923 Result = 0; break; 1924 } 1925 break; 1926 case ICmpInst::ICMP_UGT: 1927 switch (pred) { 1928 case ICmpInst::ICMP_UGT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_UGE: 1929 Result = 1; break; 1930 case ICmpInst::ICMP_ULT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_ULE: 1931 Result = 0; break; 1932 } 1933 break; 1934 case ICmpInst::ICMP_SGT: 1935 switch (pred) { 1936 case ICmpInst::ICMP_SGT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_SGE: 1937 Result = 1; break; 1938 case ICmpInst::ICMP_SLT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_SLE: 1939 Result = 0; break; 1940 } 1941 break; 1942 case ICmpInst::ICMP_ULE: 1943 if (pred == ICmpInst::ICMP_UGT) Result = 0; 1944 if (pred == ICmpInst::ICMP_ULT || pred == ICmpInst::ICMP_ULE) Result = 1; 1945 break; 1946 case ICmpInst::ICMP_SLE: 1947 if (pred == ICmpInst::ICMP_SGT) Result = 0; 1948 if (pred == ICmpInst::ICMP_SLT || pred == ICmpInst::ICMP_SLE) Result = 1; 1949 break; 1950 case ICmpInst::ICMP_UGE: 1951 if (pred == ICmpInst::ICMP_ULT) Result = 0; 1952 if (pred == ICmpInst::ICMP_UGT || pred == ICmpInst::ICMP_UGE) Result = 1; 1953 break; 1954 case ICmpInst::ICMP_SGE: 1955 if (pred == ICmpInst::ICMP_SLT) Result = 0; 1956 if (pred == ICmpInst::ICMP_SGT || pred == ICmpInst::ICMP_SGE) Result = 1; 1957 break; 1958 case ICmpInst::ICMP_NE: 1959 if (pred == ICmpInst::ICMP_EQ) Result = 0; 1960 if (pred == ICmpInst::ICMP_NE) Result = 1; 1961 break; 1962 } 1963 1964 // If we evaluated the result, return it now. 1965 if (Result != -1) 1966 return ConstantInt::get(ResultTy, Result); 1967 1968 // If the right hand side is a bitcast, try using its inverse to simplify 1969 // it by moving it to the left hand side. We can't do this if it would turn 1970 // a vector compare into a scalar compare or visa versa. 1971 if (ConstantExpr *CE2 = dyn_cast<ConstantExpr>(C2)) { 1972 Constant *CE2Op0 = CE2->getOperand(0); 1973 if (CE2->getOpcode() == Instruction::BitCast && 1974 CE2->getType()->isVectorTy() == CE2Op0->getType()->isVectorTy()) { 1975 Constant *Inverse = ConstantExpr::getBitCast(C1, CE2Op0->getType()); 1976 return ConstantExpr::getICmp(pred, Inverse, CE2Op0); 1977 } 1978 } 1979 1980 // If the left hand side is an extension, try eliminating it. 1981 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1982 if ((CE1->getOpcode() == Instruction::SExt && 1983 ICmpInst::isSigned((ICmpInst::Predicate)pred)) || 1984 (CE1->getOpcode() == Instruction::ZExt && 1985 !ICmpInst::isSigned((ICmpInst::Predicate)pred))){ 1986 Constant *CE1Op0 = CE1->getOperand(0); 1987 Constant *CE1Inverse = ConstantExpr::getTrunc(CE1, CE1Op0->getType()); 1988 if (CE1Inverse == CE1Op0) { 1989 // Check whether we can safely truncate the right hand side. 1990 Constant *C2Inverse = ConstantExpr::getTrunc(C2, CE1Op0->getType()); 1991 if (ConstantExpr::getCast(CE1->getOpcode(), C2Inverse, 1992 C2->getType()) == C2) 1993 return ConstantExpr::getICmp(pred, CE1Inverse, C2Inverse); 1994 } 1995 } 1996 } 1997 1998 if ((!isa<ConstantExpr>(C1) && isa<ConstantExpr>(C2)) || 1999 (C1->isNullValue() && !C2->isNullValue())) { 2000 // If C2 is a constant expr and C1 isn't, flip them around and fold the 2001 // other way if possible. 2002 // Also, if C1 is null and C2 isn't, flip them around. 2003 pred = ICmpInst::getSwappedPredicate((ICmpInst::Predicate)pred); 2004 return ConstantExpr::getICmp(pred, C2, C1); 2005 } 2006 } 2007 return nullptr; 2008 } 2009 2010 /// Test whether the given sequence of *normalized* indices is "inbounds". 2011 template<typename IndexTy> 2012 static bool isInBoundsIndices(ArrayRef<IndexTy> Idxs) { 2013 // No indices means nothing that could be out of bounds. 2014 if (Idxs.empty()) return true; 2015 2016 // If the first index is zero, it's in bounds. 2017 if (cast<Constant>(Idxs[0])->isNullValue()) return true; 2018 2019 // If the first index is one and all the rest are zero, it's in bounds, 2020 // by the one-past-the-end rule. 2021 if (!cast<ConstantInt>(Idxs[0])->isOne()) 2022 return false; 2023 for (unsigned i = 1, e = Idxs.size(); i != e; ++i) 2024 if (!cast<Constant>(Idxs[i])->isNullValue()) 2025 return false; 2026 return true; 2027 } 2028 2029 /// Test whether a given ConstantInt is in-range for a SequentialType. 2030 static bool isIndexInRangeOfArrayType(uint64_t NumElements, 2031 const ConstantInt *CI) { 2032 // We cannot bounds check the index if it doesn't fit in an int64_t. 2033 if (CI->getValue().getActiveBits() > 64) 2034 return false; 2035 2036 // A negative index or an index past the end of our sequential type is 2037 // considered out-of-range. 2038 int64_t IndexVal = CI->getSExtValue(); 2039 if (IndexVal < 0 || (NumElements > 0 && (uint64_t)IndexVal >= NumElements)) 2040 return false; 2041 2042 // Otherwise, it is in-range. 2043 return true; 2044 } 2045 2046 Constant *llvm::ConstantFoldGetElementPtr(Type *PointeeTy, Constant *C, 2047 bool InBounds, 2048 Optional<unsigned> InRangeIndex, 2049 ArrayRef<Value *> Idxs) { 2050 if (Idxs.empty()) return C; 2051 2052 if (isa<UndefValue>(C)) { 2053 Type *GEPTy = GetElementPtrInst::getGEPReturnType( 2054 C, makeArrayRef((Value * const *)Idxs.data(), Idxs.size())); 2055 return UndefValue::get(GEPTy); 2056 } 2057 2058 Constant *Idx0 = cast<Constant>(Idxs[0]); 2059 if (Idxs.size() == 1 && (Idx0->isNullValue() || isa<UndefValue>(Idx0))) 2060 return C; 2061 2062 if (C->isNullValue()) { 2063 bool isNull = true; 2064 for (unsigned i = 0, e = Idxs.size(); i != e; ++i) 2065 if (!isa<UndefValue>(Idxs[i]) && 2066 !cast<Constant>(Idxs[i])->isNullValue()) { 2067 isNull = false; 2068 break; 2069 } 2070 if (isNull) { 2071 PointerType *PtrTy = cast<PointerType>(C->getType()->getScalarType()); 2072 Type *Ty = GetElementPtrInst::getIndexedType(PointeeTy, Idxs); 2073 2074 assert(Ty && "Invalid indices for GEP!"); 2075 Type *OrigGEPTy = PointerType::get(Ty, PtrTy->getAddressSpace()); 2076 Type *GEPTy = PointerType::get(Ty, PtrTy->getAddressSpace()); 2077 if (VectorType *VT = dyn_cast<VectorType>(C->getType())) 2078 GEPTy = VectorType::get(OrigGEPTy, VT->getNumElements()); 2079 2080 // The GEP returns a vector of pointers when one of more of 2081 // its arguments is a vector. 2082 for (unsigned i = 0, e = Idxs.size(); i != e; ++i) { 2083 if (auto *VT = dyn_cast<VectorType>(Idxs[i]->getType())) { 2084 GEPTy = VectorType::get(OrigGEPTy, VT->getNumElements()); 2085 break; 2086 } 2087 } 2088 2089 return Constant::getNullValue(GEPTy); 2090 } 2091 } 2092 2093 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) { 2094 // Combine Indices - If the source pointer to this getelementptr instruction 2095 // is a getelementptr instruction, combine the indices of the two 2096 // getelementptr instructions into a single instruction. 2097 // 2098 if (CE->getOpcode() == Instruction::GetElementPtr) { 2099 gep_type_iterator LastI = gep_type_end(CE); 2100 for (gep_type_iterator I = gep_type_begin(CE), E = gep_type_end(CE); 2101 I != E; ++I) 2102 LastI = I; 2103 2104 // We cannot combine indices if doing so would take us outside of an 2105 // array or vector. Doing otherwise could trick us if we evaluated such a 2106 // GEP as part of a load. 2107 // 2108 // e.g. Consider if the original GEP was: 2109 // i8* getelementptr ({ [2 x i8], i32, i8, [3 x i8] }* @main.c, 2110 // i32 0, i32 0, i64 0) 2111 // 2112 // If we then tried to offset it by '8' to get to the third element, 2113 // an i8, we should *not* get: 2114 // i8* getelementptr ({ [2 x i8], i32, i8, [3 x i8] }* @main.c, 2115 // i32 0, i32 0, i64 8) 2116 // 2117 // This GEP tries to index array element '8 which runs out-of-bounds. 2118 // Subsequent evaluation would get confused and produce erroneous results. 2119 // 2120 // The following prohibits such a GEP from being formed by checking to see 2121 // if the index is in-range with respect to an array. 2122 // TODO: This code may be extended to handle vectors as well. 2123 bool PerformFold = false; 2124 if (Idx0->isNullValue()) 2125 PerformFold = true; 2126 else if (LastI.isSequential()) 2127 if (ConstantInt *CI = dyn_cast<ConstantInt>(Idx0)) 2128 PerformFold = (!LastI.isBoundedSequential() || 2129 isIndexInRangeOfArrayType( 2130 LastI.getSequentialNumElements(), CI)) && 2131 !CE->getOperand(CE->getNumOperands() - 1) 2132 ->getType() 2133 ->isVectorTy(); 2134 2135 if (PerformFold) { 2136 SmallVector<Value*, 16> NewIndices; 2137 NewIndices.reserve(Idxs.size() + CE->getNumOperands()); 2138 NewIndices.append(CE->op_begin() + 1, CE->op_end() - 1); 2139 2140 // Add the last index of the source with the first index of the new GEP. 2141 // Make sure to handle the case when they are actually different types. 2142 Constant *Combined = CE->getOperand(CE->getNumOperands()-1); 2143 // Otherwise it must be an array. 2144 if (!Idx0->isNullValue()) { 2145 Type *IdxTy = Combined->getType(); 2146 if (IdxTy != Idx0->getType()) { 2147 unsigned CommonExtendedWidth = 2148 std::max(IdxTy->getIntegerBitWidth(), 2149 Idx0->getType()->getIntegerBitWidth()); 2150 CommonExtendedWidth = std::max(CommonExtendedWidth, 64U); 2151 2152 Type *CommonTy = 2153 Type::getIntNTy(IdxTy->getContext(), CommonExtendedWidth); 2154 Constant *C1 = ConstantExpr::getSExtOrBitCast(Idx0, CommonTy); 2155 Constant *C2 = ConstantExpr::getSExtOrBitCast(Combined, CommonTy); 2156 Combined = ConstantExpr::get(Instruction::Add, C1, C2); 2157 } else { 2158 Combined = 2159 ConstantExpr::get(Instruction::Add, Idx0, Combined); 2160 } 2161 } 2162 2163 NewIndices.push_back(Combined); 2164 NewIndices.append(Idxs.begin() + 1, Idxs.end()); 2165 2166 // The combined GEP normally inherits its index inrange attribute from 2167 // the inner GEP, but if the inner GEP's last index was adjusted by the 2168 // outer GEP, any inbounds attribute on that index is invalidated. 2169 Optional<unsigned> IRIndex = cast<GEPOperator>(CE)->getInRangeIndex(); 2170 if (IRIndex && *IRIndex == CE->getNumOperands() - 2 && !Idx0->isNullValue()) 2171 IRIndex = None; 2172 2173 return ConstantExpr::getGetElementPtr( 2174 cast<GEPOperator>(CE)->getSourceElementType(), CE->getOperand(0), 2175 NewIndices, InBounds && cast<GEPOperator>(CE)->isInBounds(), 2176 IRIndex); 2177 } 2178 } 2179 2180 // Attempt to fold casts to the same type away. For example, folding: 2181 // 2182 // i32* getelementptr ([2 x i32]* bitcast ([3 x i32]* %X to [2 x i32]*), 2183 // i64 0, i64 0) 2184 // into: 2185 // 2186 // i32* getelementptr ([3 x i32]* %X, i64 0, i64 0) 2187 // 2188 // Don't fold if the cast is changing address spaces. 2189 if (CE->isCast() && Idxs.size() > 1 && Idx0->isNullValue()) { 2190 PointerType *SrcPtrTy = 2191 dyn_cast<PointerType>(CE->getOperand(0)->getType()); 2192 PointerType *DstPtrTy = dyn_cast<PointerType>(CE->getType()); 2193 if (SrcPtrTy && DstPtrTy) { 2194 ArrayType *SrcArrayTy = 2195 dyn_cast<ArrayType>(SrcPtrTy->getElementType()); 2196 ArrayType *DstArrayTy = 2197 dyn_cast<ArrayType>(DstPtrTy->getElementType()); 2198 if (SrcArrayTy && DstArrayTy 2199 && SrcArrayTy->getElementType() == DstArrayTy->getElementType() 2200 && SrcPtrTy->getAddressSpace() == DstPtrTy->getAddressSpace()) 2201 return ConstantExpr::getGetElementPtr(SrcArrayTy, 2202 (Constant *)CE->getOperand(0), 2203 Idxs, InBounds, InRangeIndex); 2204 } 2205 } 2206 } 2207 2208 // Check to see if any array indices are not within the corresponding 2209 // notional array or vector bounds. If so, try to determine if they can be 2210 // factored out into preceding dimensions. 2211 SmallVector<Constant *, 8> NewIdxs; 2212 Type *Ty = PointeeTy; 2213 Type *Prev = C->getType(); 2214 bool Unknown = 2215 !isa<ConstantInt>(Idxs[0]) && !isa<ConstantDataVector>(Idxs[0]); 2216 for (unsigned i = 1, e = Idxs.size(); i != e; 2217 Prev = Ty, Ty = cast<CompositeType>(Ty)->getTypeAtIndex(Idxs[i]), ++i) { 2218 if (!isa<ConstantInt>(Idxs[i]) && !isa<ConstantDataVector>(Idxs[i])) { 2219 // We don't know if it's in range or not. 2220 Unknown = true; 2221 continue; 2222 } 2223 if (!isa<ConstantInt>(Idxs[i - 1]) && !isa<ConstantDataVector>(Idxs[i - 1])) 2224 // Skip if the type of the previous index is not supported. 2225 continue; 2226 if (InRangeIndex && i == *InRangeIndex + 1) { 2227 // If an index is marked inrange, we cannot apply this canonicalization to 2228 // the following index, as that will cause the inrange index to point to 2229 // the wrong element. 2230 continue; 2231 } 2232 if (isa<StructType>(Ty)) { 2233 // The verify makes sure that GEPs into a struct are in range. 2234 continue; 2235 } 2236 auto *STy = cast<SequentialType>(Ty); 2237 if (isa<VectorType>(STy)) { 2238 // There can be awkward padding in after a non-power of two vector. 2239 Unknown = true; 2240 continue; 2241 } 2242 if (ConstantInt *CI = dyn_cast<ConstantInt>(Idxs[i])) { 2243 if (isIndexInRangeOfArrayType(STy->getNumElements(), CI)) 2244 // It's in range, skip to the next index. 2245 continue; 2246 if (CI->getSExtValue() < 0) { 2247 // It's out of range and negative, don't try to factor it. 2248 Unknown = true; 2249 continue; 2250 } 2251 } else { 2252 auto *CV = cast<ConstantDataVector>(Idxs[i]); 2253 bool InRange = true; 2254 for (unsigned I = 0, E = CV->getNumElements(); I != E; ++I) { 2255 auto *CI = cast<ConstantInt>(CV->getElementAsConstant(I)); 2256 InRange &= isIndexInRangeOfArrayType(STy->getNumElements(), CI); 2257 if (CI->getSExtValue() < 0) { 2258 Unknown = true; 2259 break; 2260 } 2261 } 2262 if (InRange || Unknown) 2263 // It's in range, skip to the next index. 2264 // It's out of range and negative, don't try to factor it. 2265 continue; 2266 } 2267 if (isa<StructType>(Prev)) { 2268 // It's out of range, but the prior dimension is a struct 2269 // so we can't do anything about it. 2270 Unknown = true; 2271 continue; 2272 } 2273 // It's out of range, but we can factor it into the prior 2274 // dimension. 2275 NewIdxs.resize(Idxs.size()); 2276 // Determine the number of elements in our sequential type. 2277 uint64_t NumElements = STy->getArrayNumElements(); 2278 2279 // Expand the current index or the previous index to a vector from a scalar 2280 // if necessary. 2281 Constant *CurrIdx = cast<Constant>(Idxs[i]); 2282 auto *PrevIdx = 2283 NewIdxs[i - 1] ? NewIdxs[i - 1] : cast<Constant>(Idxs[i - 1]); 2284 bool IsCurrIdxVector = CurrIdx->getType()->isVectorTy(); 2285 bool IsPrevIdxVector = PrevIdx->getType()->isVectorTy(); 2286 bool UseVector = IsCurrIdxVector || IsPrevIdxVector; 2287 2288 if (!IsCurrIdxVector && IsPrevIdxVector) 2289 CurrIdx = ConstantDataVector::getSplat( 2290 PrevIdx->getType()->getVectorNumElements(), CurrIdx); 2291 2292 if (!IsPrevIdxVector && IsCurrIdxVector) 2293 PrevIdx = ConstantDataVector::getSplat( 2294 CurrIdx->getType()->getVectorNumElements(), PrevIdx); 2295 2296 Constant *Factor = 2297 ConstantInt::get(CurrIdx->getType()->getScalarType(), NumElements); 2298 if (UseVector) 2299 Factor = ConstantDataVector::getSplat( 2300 IsPrevIdxVector ? PrevIdx->getType()->getVectorNumElements() 2301 : CurrIdx->getType()->getVectorNumElements(), 2302 Factor); 2303 2304 NewIdxs[i] = ConstantExpr::getSRem(CurrIdx, Factor); 2305 2306 Constant *Div = ConstantExpr::getSDiv(CurrIdx, Factor); 2307 2308 unsigned CommonExtendedWidth = 2309 std::max(PrevIdx->getType()->getScalarSizeInBits(), 2310 Div->getType()->getScalarSizeInBits()); 2311 CommonExtendedWidth = std::max(CommonExtendedWidth, 64U); 2312 2313 // Before adding, extend both operands to i64 to avoid 2314 // overflow trouble. 2315 Type *ExtendedTy = Type::getIntNTy(Div->getContext(), CommonExtendedWidth); 2316 if (UseVector) 2317 ExtendedTy = VectorType::get( 2318 ExtendedTy, IsPrevIdxVector 2319 ? PrevIdx->getType()->getVectorNumElements() 2320 : CurrIdx->getType()->getVectorNumElements()); 2321 2322 if (!PrevIdx->getType()->isIntOrIntVectorTy(CommonExtendedWidth)) 2323 PrevIdx = ConstantExpr::getSExt(PrevIdx, ExtendedTy); 2324 2325 if (!Div->getType()->isIntOrIntVectorTy(CommonExtendedWidth)) 2326 Div = ConstantExpr::getSExt(Div, ExtendedTy); 2327 2328 NewIdxs[i - 1] = ConstantExpr::getAdd(PrevIdx, Div); 2329 } 2330 2331 // If we did any factoring, start over with the adjusted indices. 2332 if (!NewIdxs.empty()) { 2333 for (unsigned i = 0, e = Idxs.size(); i != e; ++i) 2334 if (!NewIdxs[i]) NewIdxs[i] = cast<Constant>(Idxs[i]); 2335 return ConstantExpr::getGetElementPtr(PointeeTy, C, NewIdxs, InBounds, 2336 InRangeIndex); 2337 } 2338 2339 // If all indices are known integers and normalized, we can do a simple 2340 // check for the "inbounds" property. 2341 if (!Unknown && !InBounds) 2342 if (auto *GV = dyn_cast<GlobalVariable>(C)) 2343 if (!GV->hasExternalWeakLinkage() && isInBoundsIndices(Idxs)) 2344 return ConstantExpr::getGetElementPtr(PointeeTy, C, Idxs, 2345 /*InBounds=*/true, InRangeIndex); 2346 2347 return nullptr; 2348 } 2349