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 TargetData, and the pieces that do. This is to avoid 16 // a dependence in VMCore on Target. 17 // 18 //===----------------------------------------------------------------------===// 19 20 #include "ConstantFold.h" 21 #include "llvm/Constants.h" 22 #include "llvm/Instructions.h" 23 #include "llvm/DerivedTypes.h" 24 #include "llvm/Function.h" 25 #include "llvm/GlobalAlias.h" 26 #include "llvm/GlobalVariable.h" 27 #include "llvm/Operator.h" 28 #include "llvm/ADT/SmallVector.h" 29 #include "llvm/Support/Compiler.h" 30 #include "llvm/Support/ErrorHandling.h" 31 #include "llvm/Support/GetElementPtrTypeIterator.h" 32 #include "llvm/Support/ManagedStatic.h" 33 #include "llvm/Support/MathExtras.h" 34 #include <limits> 35 using namespace llvm; 36 37 //===----------------------------------------------------------------------===// 38 // ConstantFold*Instruction Implementations 39 //===----------------------------------------------------------------------===// 40 41 /// BitCastConstantVector - Convert the specified ConstantVector node to the 42 /// specified vector type. At this point, we know that the elements of the 43 /// input vector constant are all simple integer or FP values. 44 static Constant *BitCastConstantVector(ConstantVector *CV, 45 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->getNumOperands()) 55 return 0; 56 57 // Check to verify that all elements of the input are simple. 58 for (unsigned i = 0; i != NumElts; ++i) { 59 if (!isa<ConstantInt>(CV->getOperand(i)) && 60 !isa<ConstantFP>(CV->getOperand(i))) 61 return 0; 62 } 63 64 // Bitcast each element now. 65 std::vector<Constant*> Result; 66 Type *DstEltTy = DstTy->getElementType(); 67 for (unsigned i = 0; i != NumElts; ++i) 68 Result.push_back(ConstantExpr::getBitCast(CV->getOperand(i), 69 DstEltTy)); 70 return ConstantVector::get(Result); 71 } 72 73 /// This function determines which opcode to use to fold two constant cast 74 /// expressions together. It uses CastInst::isEliminableCastPair to determine 75 /// the opcode. Consequently its just a wrapper around that function. 76 /// @brief Determine if it is valid to fold a cast of a cast 77 static unsigned 78 foldConstantCastPair( 79 unsigned opc, ///< opcode of the second cast constant expression 80 ConstantExpr *Op, ///< the first cast constant expression 81 Type *DstTy ///< desintation type of the first cast 82 ) { 83 assert(Op && Op->isCast() && "Can't fold cast of cast without a cast!"); 84 assert(DstTy && DstTy->isFirstClassType() && "Invalid cast destination type"); 85 assert(CastInst::isCast(opc) && "Invalid cast opcode"); 86 87 // The the types and opcodes for the two Cast constant expressions 88 Type *SrcTy = Op->getOperand(0)->getType(); 89 Type *MidTy = Op->getType(); 90 Instruction::CastOps firstOp = Instruction::CastOps(Op->getOpcode()); 91 Instruction::CastOps secondOp = Instruction::CastOps(opc); 92 93 // Let CastInst::isEliminableCastPair do the heavy lifting. 94 return CastInst::isEliminableCastPair(firstOp, secondOp, SrcTy, MidTy, DstTy, 95 Type::getInt64Ty(DstTy->getContext())); 96 } 97 98 static Constant *FoldBitCast(Constant *V, Type *DestTy) { 99 Type *SrcTy = V->getType(); 100 if (SrcTy == DestTy) 101 return V; // no-op cast 102 103 // Check to see if we are casting a pointer to an aggregate to a pointer to 104 // the first element. If so, return the appropriate GEP instruction. 105 if (PointerType *PTy = dyn_cast<PointerType>(V->getType())) 106 if (PointerType *DPTy = dyn_cast<PointerType>(DestTy)) 107 if (PTy->getAddressSpace() == DPTy->getAddressSpace()) { 108 SmallVector<Value*, 8> IdxList; 109 Value *Zero = 110 Constant::getNullValue(Type::getInt32Ty(DPTy->getContext())); 111 IdxList.push_back(Zero); 112 Type *ElTy = PTy->getElementType(); 113 while (ElTy != DPTy->getElementType()) { 114 if (StructType *STy = dyn_cast<StructType>(ElTy)) { 115 if (STy->getNumElements() == 0) break; 116 ElTy = STy->getElementType(0); 117 IdxList.push_back(Zero); 118 } else if (SequentialType *STy = 119 dyn_cast<SequentialType>(ElTy)) { 120 if (ElTy->isPointerTy()) break; // Can't index into pointers! 121 ElTy = STy->getElementType(); 122 IdxList.push_back(Zero); 123 } else { 124 break; 125 } 126 } 127 128 if (ElTy == DPTy->getElementType()) 129 // This GEP is inbounds because all indices are zero. 130 return ConstantExpr::getInBoundsGetElementPtr(V, IdxList); 131 } 132 133 // Handle casts from one vector constant to another. We know that the src 134 // and dest type have the same size (otherwise its an illegal cast). 135 if (VectorType *DestPTy = dyn_cast<VectorType>(DestTy)) { 136 if (VectorType *SrcTy = dyn_cast<VectorType>(V->getType())) { 137 assert(DestPTy->getBitWidth() == SrcTy->getBitWidth() && 138 "Not cast between same sized vectors!"); 139 SrcTy = NULL; 140 // First, check for null. Undef is already handled. 141 if (isa<ConstantAggregateZero>(V)) 142 return Constant::getNullValue(DestTy); 143 144 if (ConstantVector *CV = dyn_cast<ConstantVector>(V)) 145 return BitCastConstantVector(CV, DestPTy); 146 } 147 148 // Canonicalize scalar-to-vector bitcasts into vector-to-vector bitcasts 149 // This allows for other simplifications (although some of them 150 // can only be handled by Analysis/ConstantFolding.cpp). 151 if (isa<ConstantInt>(V) || isa<ConstantFP>(V)) 152 return ConstantExpr::getBitCast(ConstantVector::get(V), DestPTy); 153 } 154 155 // Finally, implement bitcast folding now. The code below doesn't handle 156 // bitcast right. 157 if (isa<ConstantPointerNull>(V)) // ptr->ptr cast. 158 return ConstantPointerNull::get(cast<PointerType>(DestTy)); 159 160 // Handle integral constant input. 161 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 162 if (DestTy->isIntegerTy()) 163 // Integral -> Integral. This is a no-op because the bit widths must 164 // be the same. Consequently, we just fold to V. 165 return V; 166 167 if (DestTy->isFloatingPointTy()) 168 return ConstantFP::get(DestTy->getContext(), 169 APFloat(CI->getValue(), 170 !DestTy->isPPC_FP128Ty())); 171 172 // Otherwise, can't fold this (vector?) 173 return 0; 174 } 175 176 // Handle ConstantFP input: FP -> Integral. 177 if (ConstantFP *FP = dyn_cast<ConstantFP>(V)) 178 return ConstantInt::get(FP->getContext(), 179 FP->getValueAPF().bitcastToAPInt()); 180 181 return 0; 182 } 183 184 185 /// ExtractConstantBytes - V is an integer constant which only has a subset of 186 /// its bytes used. The bytes used are indicated by ByteStart (which is the 187 /// first byte used, counting from the least significant byte) and ByteSize, 188 /// which is the number of bytes used. 189 /// 190 /// This function analyzes the specified constant to see if the specified byte 191 /// range can be returned as a simplified constant. If so, the constant is 192 /// returned, otherwise null is returned. 193 /// 194 static Constant *ExtractConstantBytes(Constant *C, unsigned ByteStart, 195 unsigned ByteSize) { 196 assert(C->getType()->isIntegerTy() && 197 (cast<IntegerType>(C->getType())->getBitWidth() & 7) == 0 && 198 "Non-byte sized integer input"); 199 unsigned CSize = cast<IntegerType>(C->getType())->getBitWidth()/8; 200 assert(ByteSize && "Must be accessing some piece"); 201 assert(ByteStart+ByteSize <= CSize && "Extracting invalid piece from input"); 202 assert(ByteSize != CSize && "Should not extract everything"); 203 204 // Constant Integers are simple. 205 if (ConstantInt *CI = dyn_cast<ConstantInt>(C)) { 206 APInt V = CI->getValue(); 207 if (ByteStart) 208 V = V.lshr(ByteStart*8); 209 V = V.trunc(ByteSize*8); 210 return ConstantInt::get(CI->getContext(), V); 211 } 212 213 // In the input is a constant expr, we might be able to recursively simplify. 214 // If not, we definitely can't do anything. 215 ConstantExpr *CE = dyn_cast<ConstantExpr>(C); 216 if (CE == 0) return 0; 217 218 switch (CE->getOpcode()) { 219 default: return 0; 220 case Instruction::Or: { 221 Constant *RHS = ExtractConstantBytes(CE->getOperand(1), ByteStart,ByteSize); 222 if (RHS == 0) 223 return 0; 224 225 // X | -1 -> -1. 226 if (ConstantInt *RHSC = dyn_cast<ConstantInt>(RHS)) 227 if (RHSC->isAllOnesValue()) 228 return RHSC; 229 230 Constant *LHS = ExtractConstantBytes(CE->getOperand(0), ByteStart,ByteSize); 231 if (LHS == 0) 232 return 0; 233 return ConstantExpr::getOr(LHS, RHS); 234 } 235 case Instruction::And: { 236 Constant *RHS = ExtractConstantBytes(CE->getOperand(1), ByteStart,ByteSize); 237 if (RHS == 0) 238 return 0; 239 240 // X & 0 -> 0. 241 if (RHS->isNullValue()) 242 return RHS; 243 244 Constant *LHS = ExtractConstantBytes(CE->getOperand(0), ByteStart,ByteSize); 245 if (LHS == 0) 246 return 0; 247 return ConstantExpr::getAnd(LHS, RHS); 248 } 249 case Instruction::LShr: { 250 ConstantInt *Amt = dyn_cast<ConstantInt>(CE->getOperand(1)); 251 if (Amt == 0) 252 return 0; 253 unsigned ShAmt = Amt->getZExtValue(); 254 // Cannot analyze non-byte shifts. 255 if ((ShAmt & 7) != 0) 256 return 0; 257 ShAmt >>= 3; 258 259 // If the extract is known to be all zeros, return zero. 260 if (ByteStart >= CSize-ShAmt) 261 return Constant::getNullValue(IntegerType::get(CE->getContext(), 262 ByteSize*8)); 263 // If the extract is known to be fully in the input, extract it. 264 if (ByteStart+ByteSize+ShAmt <= CSize) 265 return ExtractConstantBytes(CE->getOperand(0), ByteStart+ShAmt, ByteSize); 266 267 // TODO: Handle the 'partially zero' case. 268 return 0; 269 } 270 271 case Instruction::Shl: { 272 ConstantInt *Amt = dyn_cast<ConstantInt>(CE->getOperand(1)); 273 if (Amt == 0) 274 return 0; 275 unsigned ShAmt = Amt->getZExtValue(); 276 // Cannot analyze non-byte shifts. 277 if ((ShAmt & 7) != 0) 278 return 0; 279 ShAmt >>= 3; 280 281 // If the extract is known to be all zeros, return zero. 282 if (ByteStart+ByteSize <= ShAmt) 283 return Constant::getNullValue(IntegerType::get(CE->getContext(), 284 ByteSize*8)); 285 // If the extract is known to be fully in the input, extract it. 286 if (ByteStart >= ShAmt) 287 return ExtractConstantBytes(CE->getOperand(0), ByteStart-ShAmt, ByteSize); 288 289 // TODO: Handle the 'partially zero' case. 290 return 0; 291 } 292 293 case Instruction::ZExt: { 294 unsigned SrcBitSize = 295 cast<IntegerType>(CE->getOperand(0)->getType())->getBitWidth(); 296 297 // If extracting something that is completely zero, return 0. 298 if (ByteStart*8 >= SrcBitSize) 299 return Constant::getNullValue(IntegerType::get(CE->getContext(), 300 ByteSize*8)); 301 302 // If exactly extracting the input, return it. 303 if (ByteStart == 0 && ByteSize*8 == SrcBitSize) 304 return CE->getOperand(0); 305 306 // If extracting something completely in the input, if if the input is a 307 // multiple of 8 bits, recurse. 308 if ((SrcBitSize&7) == 0 && (ByteStart+ByteSize)*8 <= SrcBitSize) 309 return ExtractConstantBytes(CE->getOperand(0), ByteStart, ByteSize); 310 311 // Otherwise, if extracting a subset of the input, which is not multiple of 312 // 8 bits, do a shift and trunc to get the bits. 313 if ((ByteStart+ByteSize)*8 < SrcBitSize) { 314 assert((SrcBitSize&7) && "Shouldn't get byte sized case here"); 315 Constant *Res = CE->getOperand(0); 316 if (ByteStart) 317 Res = ConstantExpr::getLShr(Res, 318 ConstantInt::get(Res->getType(), ByteStart*8)); 319 return ConstantExpr::getTrunc(Res, IntegerType::get(C->getContext(), 320 ByteSize*8)); 321 } 322 323 // TODO: Handle the 'partially zero' case. 324 return 0; 325 } 326 } 327 } 328 329 /// getFoldedSizeOf - Return a ConstantExpr with type DestTy for sizeof 330 /// on Ty, with any known factors factored out. If Folded is false, 331 /// return null if no factoring was possible, to avoid endlessly 332 /// bouncing an unfoldable expression back into the top-level folder. 333 /// 334 static Constant *getFoldedSizeOf(Type *Ty, Type *DestTy, 335 bool Folded) { 336 if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 337 Constant *N = ConstantInt::get(DestTy, ATy->getNumElements()); 338 Constant *E = getFoldedSizeOf(ATy->getElementType(), DestTy, true); 339 return ConstantExpr::getNUWMul(E, N); 340 } 341 342 if (StructType *STy = dyn_cast<StructType>(Ty)) 343 if (!STy->isPacked()) { 344 unsigned NumElems = STy->getNumElements(); 345 // An empty struct has size zero. 346 if (NumElems == 0) 347 return ConstantExpr::getNullValue(DestTy); 348 // Check for a struct with all members having the same size. 349 Constant *MemberSize = 350 getFoldedSizeOf(STy->getElementType(0), DestTy, true); 351 bool AllSame = true; 352 for (unsigned i = 1; i != NumElems; ++i) 353 if (MemberSize != 354 getFoldedSizeOf(STy->getElementType(i), DestTy, true)) { 355 AllSame = false; 356 break; 357 } 358 if (AllSame) { 359 Constant *N = ConstantInt::get(DestTy, NumElems); 360 return ConstantExpr::getNUWMul(MemberSize, N); 361 } 362 } 363 364 // Pointer size doesn't depend on the pointee type, so canonicalize them 365 // to an arbitrary pointee. 366 if (PointerType *PTy = dyn_cast<PointerType>(Ty)) 367 if (!PTy->getElementType()->isIntegerTy(1)) 368 return 369 getFoldedSizeOf(PointerType::get(IntegerType::get(PTy->getContext(), 1), 370 PTy->getAddressSpace()), 371 DestTy, true); 372 373 // If there's no interesting folding happening, bail so that we don't create 374 // a constant that looks like it needs folding but really doesn't. 375 if (!Folded) 376 return 0; 377 378 // Base case: Get a regular sizeof expression. 379 Constant *C = ConstantExpr::getSizeOf(Ty); 380 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 381 DestTy, false), 382 C, DestTy); 383 return C; 384 } 385 386 /// getFoldedAlignOf - Return a ConstantExpr with type DestTy for alignof 387 /// on Ty, with any known factors factored out. If Folded is false, 388 /// return null if no factoring was possible, to avoid endlessly 389 /// bouncing an unfoldable expression back into the top-level folder. 390 /// 391 static Constant *getFoldedAlignOf(Type *Ty, Type *DestTy, 392 bool Folded) { 393 // The alignment of an array is equal to the alignment of the 394 // array element. Note that this is not always true for vectors. 395 if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 396 Constant *C = ConstantExpr::getAlignOf(ATy->getElementType()); 397 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 398 DestTy, 399 false), 400 C, DestTy); 401 return C; 402 } 403 404 if (StructType *STy = dyn_cast<StructType>(Ty)) { 405 // Packed structs always have an alignment of 1. 406 if (STy->isPacked()) 407 return ConstantInt::get(DestTy, 1); 408 409 // Otherwise, struct alignment is the maximum alignment of any member. 410 // Without target data, we can't compare much, but we can check to see 411 // if all the members have the same alignment. 412 unsigned NumElems = STy->getNumElements(); 413 // An empty struct has minimal alignment. 414 if (NumElems == 0) 415 return ConstantInt::get(DestTy, 1); 416 // Check for a struct with all members having the same alignment. 417 Constant *MemberAlign = 418 getFoldedAlignOf(STy->getElementType(0), DestTy, true); 419 bool AllSame = true; 420 for (unsigned i = 1; i != NumElems; ++i) 421 if (MemberAlign != getFoldedAlignOf(STy->getElementType(i), DestTy, true)) { 422 AllSame = false; 423 break; 424 } 425 if (AllSame) 426 return MemberAlign; 427 } 428 429 // Pointer alignment doesn't depend on the pointee type, so canonicalize them 430 // to an arbitrary pointee. 431 if (PointerType *PTy = dyn_cast<PointerType>(Ty)) 432 if (!PTy->getElementType()->isIntegerTy(1)) 433 return 434 getFoldedAlignOf(PointerType::get(IntegerType::get(PTy->getContext(), 435 1), 436 PTy->getAddressSpace()), 437 DestTy, true); 438 439 // If there's no interesting folding happening, bail so that we don't create 440 // a constant that looks like it needs folding but really doesn't. 441 if (!Folded) 442 return 0; 443 444 // Base case: Get a regular alignof expression. 445 Constant *C = ConstantExpr::getAlignOf(Ty); 446 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 447 DestTy, false), 448 C, DestTy); 449 return C; 450 } 451 452 /// getFoldedOffsetOf - Return a ConstantExpr with type DestTy for offsetof 453 /// on Ty and FieldNo, with any known factors factored out. If Folded is false, 454 /// return null if no factoring was possible, to avoid endlessly 455 /// bouncing an unfoldable expression back into the top-level folder. 456 /// 457 static Constant *getFoldedOffsetOf(Type *Ty, Constant *FieldNo, 458 Type *DestTy, 459 bool Folded) { 460 if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 461 Constant *N = ConstantExpr::getCast(CastInst::getCastOpcode(FieldNo, false, 462 DestTy, false), 463 FieldNo, DestTy); 464 Constant *E = getFoldedSizeOf(ATy->getElementType(), DestTy, true); 465 return ConstantExpr::getNUWMul(E, N); 466 } 467 468 if (StructType *STy = dyn_cast<StructType>(Ty)) 469 if (!STy->isPacked()) { 470 unsigned NumElems = STy->getNumElements(); 471 // An empty struct has no members. 472 if (NumElems == 0) 473 return 0; 474 // Check for a struct with all members having the same size. 475 Constant *MemberSize = 476 getFoldedSizeOf(STy->getElementType(0), DestTy, true); 477 bool AllSame = true; 478 for (unsigned i = 1; i != NumElems; ++i) 479 if (MemberSize != 480 getFoldedSizeOf(STy->getElementType(i), DestTy, true)) { 481 AllSame = false; 482 break; 483 } 484 if (AllSame) { 485 Constant *N = ConstantExpr::getCast(CastInst::getCastOpcode(FieldNo, 486 false, 487 DestTy, 488 false), 489 FieldNo, DestTy); 490 return ConstantExpr::getNUWMul(MemberSize, N); 491 } 492 } 493 494 // If there's no interesting folding happening, bail so that we don't create 495 // a constant that looks like it needs folding but really doesn't. 496 if (!Folded) 497 return 0; 498 499 // Base case: Get a regular offsetof expression. 500 Constant *C = ConstantExpr::getOffsetOf(Ty, FieldNo); 501 C = ConstantExpr::getCast(CastInst::getCastOpcode(C, false, 502 DestTy, false), 503 C, DestTy); 504 return C; 505 } 506 507 Constant *llvm::ConstantFoldCastInstruction(unsigned opc, Constant *V, 508 Type *DestTy) { 509 if (isa<UndefValue>(V)) { 510 // zext(undef) = 0, because the top bits will be zero. 511 // sext(undef) = 0, because the top bits will all be the same. 512 // [us]itofp(undef) = 0, because the result value is bounded. 513 if (opc == Instruction::ZExt || opc == Instruction::SExt || 514 opc == Instruction::UIToFP || opc == Instruction::SIToFP) 515 return Constant::getNullValue(DestTy); 516 return UndefValue::get(DestTy); 517 } 518 519 // No compile-time operations on this type yet. 520 if (V->getType()->isPPC_FP128Ty() || DestTy->isPPC_FP128Ty()) 521 return 0; 522 523 if (V->isNullValue() && !DestTy->isX86_MMXTy()) 524 return Constant::getNullValue(DestTy); 525 526 // If the cast operand is a constant expression, there's a few things we can 527 // do to try to simplify it. 528 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) { 529 if (CE->isCast()) { 530 // Try hard to fold cast of cast because they are often eliminable. 531 if (unsigned newOpc = foldConstantCastPair(opc, CE, DestTy)) 532 return ConstantExpr::getCast(newOpc, CE->getOperand(0), DestTy); 533 } else if (CE->getOpcode() == Instruction::GetElementPtr) { 534 // If all of the indexes in the GEP are null values, there is no pointer 535 // adjustment going on. We might as well cast the source pointer. 536 bool isAllNull = true; 537 for (unsigned i = 1, e = CE->getNumOperands(); i != e; ++i) 538 if (!CE->getOperand(i)->isNullValue()) { 539 isAllNull = false; 540 break; 541 } 542 if (isAllNull) 543 // This is casting one pointer type to another, always BitCast 544 return ConstantExpr::getPointerCast(CE->getOperand(0), DestTy); 545 } 546 } 547 548 // If the cast operand is a constant vector, perform the cast by 549 // operating on each element. In the cast of bitcasts, the element 550 // count may be mismatched; don't attempt to handle that here. 551 if (ConstantVector *CV = dyn_cast<ConstantVector>(V)) 552 if (DestTy->isVectorTy() && 553 cast<VectorType>(DestTy)->getNumElements() == 554 CV->getType()->getNumElements()) { 555 std::vector<Constant*> res; 556 VectorType *DestVecTy = cast<VectorType>(DestTy); 557 Type *DstEltTy = DestVecTy->getElementType(); 558 for (unsigned i = 0, e = CV->getType()->getNumElements(); i != e; ++i) 559 res.push_back(ConstantExpr::getCast(opc, 560 CV->getOperand(i), DstEltTy)); 561 return ConstantVector::get(res); 562 } 563 564 // We actually have to do a cast now. Perform the cast according to the 565 // opcode specified. 566 switch (opc) { 567 default: 568 llvm_unreachable("Failed to cast constant expression"); 569 case Instruction::FPTrunc: 570 case Instruction::FPExt: 571 if (ConstantFP *FPC = dyn_cast<ConstantFP>(V)) { 572 bool ignored; 573 APFloat Val = FPC->getValueAPF(); 574 Val.convert(DestTy->isFloatTy() ? APFloat::IEEEsingle : 575 DestTy->isDoubleTy() ? APFloat::IEEEdouble : 576 DestTy->isX86_FP80Ty() ? APFloat::x87DoubleExtended : 577 DestTy->isFP128Ty() ? APFloat::IEEEquad : 578 APFloat::Bogus, 579 APFloat::rmNearestTiesToEven, &ignored); 580 return ConstantFP::get(V->getContext(), Val); 581 } 582 return 0; // Can't fold. 583 case Instruction::FPToUI: 584 case Instruction::FPToSI: 585 if (ConstantFP *FPC = dyn_cast<ConstantFP>(V)) { 586 const APFloat &V = FPC->getValueAPF(); 587 bool ignored; 588 uint64_t x[2]; 589 uint32_t DestBitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 590 (void) V.convertToInteger(x, DestBitWidth, opc==Instruction::FPToSI, 591 APFloat::rmTowardZero, &ignored); 592 APInt Val(DestBitWidth, x); 593 return ConstantInt::get(FPC->getContext(), Val); 594 } 595 return 0; // Can't fold. 596 case Instruction::IntToPtr: //always treated as unsigned 597 if (V->isNullValue()) // Is it an integral null value? 598 return ConstantPointerNull::get(cast<PointerType>(DestTy)); 599 return 0; // Other pointer types cannot be casted 600 case Instruction::PtrToInt: // always treated as unsigned 601 // Is it a null pointer value? 602 if (V->isNullValue()) 603 return ConstantInt::get(DestTy, 0); 604 // If this is a sizeof-like expression, pull out multiplications by 605 // known factors to expose them to subsequent folding. If it's an 606 // alignof-like expression, factor out known factors. 607 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) 608 if (CE->getOpcode() == Instruction::GetElementPtr && 609 CE->getOperand(0)->isNullValue()) { 610 Type *Ty = 611 cast<PointerType>(CE->getOperand(0)->getType())->getElementType(); 612 if (CE->getNumOperands() == 2) { 613 // Handle a sizeof-like expression. 614 Constant *Idx = CE->getOperand(1); 615 bool isOne = isa<ConstantInt>(Idx) && cast<ConstantInt>(Idx)->isOne(); 616 if (Constant *C = getFoldedSizeOf(Ty, DestTy, !isOne)) { 617 Idx = ConstantExpr::getCast(CastInst::getCastOpcode(Idx, true, 618 DestTy, false), 619 Idx, DestTy); 620 return ConstantExpr::getMul(C, Idx); 621 } 622 } else if (CE->getNumOperands() == 3 && 623 CE->getOperand(1)->isNullValue()) { 624 // Handle an alignof-like expression. 625 if (StructType *STy = dyn_cast<StructType>(Ty)) 626 if (!STy->isPacked()) { 627 ConstantInt *CI = cast<ConstantInt>(CE->getOperand(2)); 628 if (CI->isOne() && 629 STy->getNumElements() == 2 && 630 STy->getElementType(0)->isIntegerTy(1)) { 631 return getFoldedAlignOf(STy->getElementType(1), DestTy, false); 632 } 633 } 634 // Handle an offsetof-like expression. 635 if (Ty->isStructTy() || Ty->isArrayTy()) { 636 if (Constant *C = getFoldedOffsetOf(Ty, CE->getOperand(2), 637 DestTy, false)) 638 return C; 639 } 640 } 641 } 642 // Other pointer types cannot be casted 643 return 0; 644 case Instruction::UIToFP: 645 case Instruction::SIToFP: 646 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 647 APInt api = CI->getValue(); 648 APFloat apf(APInt::getNullValue(DestTy->getPrimitiveSizeInBits()), true); 649 (void)apf.convertFromAPInt(api, 650 opc==Instruction::SIToFP, 651 APFloat::rmNearestTiesToEven); 652 return ConstantFP::get(V->getContext(), apf); 653 } 654 return 0; 655 case Instruction::ZExt: 656 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 657 uint32_t BitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 658 return ConstantInt::get(V->getContext(), 659 CI->getValue().zext(BitWidth)); 660 } 661 return 0; 662 case Instruction::SExt: 663 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 664 uint32_t BitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 665 return ConstantInt::get(V->getContext(), 666 CI->getValue().sext(BitWidth)); 667 } 668 return 0; 669 case Instruction::Trunc: { 670 uint32_t DestBitWidth = cast<IntegerType>(DestTy)->getBitWidth(); 671 if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) { 672 return ConstantInt::get(V->getContext(), 673 CI->getValue().trunc(DestBitWidth)); 674 } 675 676 // The input must be a constantexpr. See if we can simplify this based on 677 // the bytes we are demanding. Only do this if the source and dest are an 678 // even multiple of a byte. 679 if ((DestBitWidth & 7) == 0 && 680 (cast<IntegerType>(V->getType())->getBitWidth() & 7) == 0) 681 if (Constant *Res = ExtractConstantBytes(V, 0, DestBitWidth / 8)) 682 return Res; 683 684 return 0; 685 } 686 case Instruction::BitCast: 687 return FoldBitCast(V, DestTy); 688 } 689 } 690 691 Constant *llvm::ConstantFoldSelectInstruction(Constant *Cond, 692 Constant *V1, Constant *V2) { 693 if (ConstantInt *CB = dyn_cast<ConstantInt>(Cond)) 694 return CB->getZExtValue() ? V1 : V2; 695 696 // Check for zero aggregate and ConstantVector of zeros 697 if (Cond->isNullValue()) return V2; 698 699 if (ConstantVector* CondV = dyn_cast<ConstantVector>(Cond)) { 700 701 if (CondV->isAllOnesValue()) return V1; 702 703 VectorType *VTy = cast<VectorType>(V1->getType()); 704 ConstantVector *CP1 = dyn_cast<ConstantVector>(V1); 705 ConstantVector *CP2 = dyn_cast<ConstantVector>(V2); 706 707 if ((CP1 || isa<ConstantAggregateZero>(V1)) && 708 (CP2 || isa<ConstantAggregateZero>(V2))) { 709 710 // Find the element type of the returned vector 711 Type *EltTy = VTy->getElementType(); 712 unsigned NumElem = VTy->getNumElements(); 713 std::vector<Constant*> Res(NumElem); 714 715 bool Valid = true; 716 for (unsigned i = 0; i < NumElem; ++i) { 717 ConstantInt* c = dyn_cast<ConstantInt>(CondV->getOperand(i)); 718 if (!c) { 719 Valid = false; 720 break; 721 } 722 Constant *C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 723 Constant *C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 724 Res[i] = c->getZExtValue() ? C1 : C2; 725 } 726 // If we were able to build the vector, return it 727 if (Valid) return ConstantVector::get(Res); 728 } 729 } 730 731 732 if (isa<UndefValue>(Cond)) { 733 if (isa<UndefValue>(V1)) return V1; 734 return V2; 735 } 736 if (isa<UndefValue>(V1)) return V2; 737 if (isa<UndefValue>(V2)) return V1; 738 if (V1 == V2) return V1; 739 740 if (ConstantExpr *TrueVal = dyn_cast<ConstantExpr>(V1)) { 741 if (TrueVal->getOpcode() == Instruction::Select) 742 if (TrueVal->getOperand(0) == Cond) 743 return ConstantExpr::getSelect(Cond, TrueVal->getOperand(1), V2); 744 } 745 if (ConstantExpr *FalseVal = dyn_cast<ConstantExpr>(V2)) { 746 if (FalseVal->getOpcode() == Instruction::Select) 747 if (FalseVal->getOperand(0) == Cond) 748 return ConstantExpr::getSelect(Cond, V1, FalseVal->getOperand(2)); 749 } 750 751 return 0; 752 } 753 754 Constant *llvm::ConstantFoldExtractElementInstruction(Constant *Val, 755 Constant *Idx) { 756 if (isa<UndefValue>(Val)) // ee(undef, x) -> undef 757 return UndefValue::get(cast<VectorType>(Val->getType())->getElementType()); 758 if (Val->isNullValue()) // ee(zero, x) -> zero 759 return Constant::getNullValue( 760 cast<VectorType>(Val->getType())->getElementType()); 761 762 if (ConstantVector *CVal = dyn_cast<ConstantVector>(Val)) { 763 if (ConstantInt *CIdx = dyn_cast<ConstantInt>(Idx)) { 764 uint64_t Index = CIdx->getZExtValue(); 765 if (Index >= CVal->getNumOperands()) 766 // ee({w,x,y,z}, wrong_value) -> undef 767 return UndefValue::get(cast<VectorType>(Val->getType())->getElementType()); 768 return CVal->getOperand(CIdx->getZExtValue()); 769 } else if (isa<UndefValue>(Idx)) { 770 // ee({w,x,y,z}, undef) -> undef 771 return UndefValue::get(cast<VectorType>(Val->getType())->getElementType()); 772 } 773 } 774 return 0; 775 } 776 777 Constant *llvm::ConstantFoldInsertElementInstruction(Constant *Val, 778 Constant *Elt, 779 Constant *Idx) { 780 ConstantInt *CIdx = dyn_cast<ConstantInt>(Idx); 781 if (!CIdx) return 0; 782 APInt idxVal = CIdx->getValue(); 783 if (isa<UndefValue>(Val)) { 784 // Insertion of scalar constant into vector undef 785 // Optimize away insertion of undef 786 if (isa<UndefValue>(Elt)) 787 return Val; 788 // Otherwise break the aggregate undef into multiple undefs and do 789 // the insertion 790 unsigned numOps = 791 cast<VectorType>(Val->getType())->getNumElements(); 792 std::vector<Constant*> Ops; 793 Ops.reserve(numOps); 794 for (unsigned i = 0; i < numOps; ++i) { 795 Constant *Op = 796 (idxVal == i) ? Elt : UndefValue::get(Elt->getType()); 797 Ops.push_back(Op); 798 } 799 return ConstantVector::get(Ops); 800 } 801 if (isa<ConstantAggregateZero>(Val)) { 802 // Insertion of scalar constant into vector aggregate zero 803 // Optimize away insertion of zero 804 if (Elt->isNullValue()) 805 return Val; 806 // Otherwise break the aggregate zero into multiple zeros and do 807 // the insertion 808 unsigned numOps = 809 cast<VectorType>(Val->getType())->getNumElements(); 810 std::vector<Constant*> Ops; 811 Ops.reserve(numOps); 812 for (unsigned i = 0; i < numOps; ++i) { 813 Constant *Op = 814 (idxVal == i) ? Elt : Constant::getNullValue(Elt->getType()); 815 Ops.push_back(Op); 816 } 817 return ConstantVector::get(Ops); 818 } 819 if (ConstantVector *CVal = dyn_cast<ConstantVector>(Val)) { 820 // Insertion of scalar constant into vector constant 821 std::vector<Constant*> Ops; 822 Ops.reserve(CVal->getNumOperands()); 823 for (unsigned i = 0; i < CVal->getNumOperands(); ++i) { 824 Constant *Op = 825 (idxVal == i) ? Elt : cast<Constant>(CVal->getOperand(i)); 826 Ops.push_back(Op); 827 } 828 return ConstantVector::get(Ops); 829 } 830 831 return 0; 832 } 833 834 /// GetVectorElement - If C is a ConstantVector, ConstantAggregateZero or Undef 835 /// return the specified element value. Otherwise return null. 836 static Constant *GetVectorElement(Constant *C, unsigned EltNo) { 837 if (ConstantVector *CV = dyn_cast<ConstantVector>(C)) 838 return CV->getOperand(EltNo); 839 840 Type *EltTy = cast<VectorType>(C->getType())->getElementType(); 841 if (isa<ConstantAggregateZero>(C)) 842 return Constant::getNullValue(EltTy); 843 if (isa<UndefValue>(C)) 844 return UndefValue::get(EltTy); 845 return 0; 846 } 847 848 Constant *llvm::ConstantFoldShuffleVectorInstruction(Constant *V1, 849 Constant *V2, 850 Constant *Mask) { 851 // Undefined shuffle mask -> undefined value. 852 if (isa<UndefValue>(Mask)) return UndefValue::get(V1->getType()); 853 854 unsigned MaskNumElts = cast<VectorType>(Mask->getType())->getNumElements(); 855 unsigned SrcNumElts = cast<VectorType>(V1->getType())->getNumElements(); 856 Type *EltTy = cast<VectorType>(V1->getType())->getElementType(); 857 858 // Loop over the shuffle mask, evaluating each element. 859 SmallVector<Constant*, 32> Result; 860 for (unsigned i = 0; i != MaskNumElts; ++i) { 861 Constant *InElt = GetVectorElement(Mask, i); 862 if (InElt == 0) return 0; 863 864 if (isa<UndefValue>(InElt)) 865 InElt = UndefValue::get(EltTy); 866 else if (ConstantInt *CI = dyn_cast<ConstantInt>(InElt)) { 867 unsigned Elt = CI->getZExtValue(); 868 if (Elt >= SrcNumElts*2) 869 InElt = UndefValue::get(EltTy); 870 else if (Elt >= SrcNumElts) 871 InElt = GetVectorElement(V2, Elt - SrcNumElts); 872 else 873 InElt = GetVectorElement(V1, Elt); 874 if (InElt == 0) return 0; 875 } else { 876 // Unknown value. 877 return 0; 878 } 879 Result.push_back(InElt); 880 } 881 882 return ConstantVector::get(Result); 883 } 884 885 Constant *llvm::ConstantFoldExtractValueInstruction(Constant *Agg, 886 ArrayRef<unsigned> Idxs) { 887 // Base case: no indices, so return the entire value. 888 if (Idxs.empty()) 889 return Agg; 890 891 if (isa<UndefValue>(Agg)) // ev(undef, x) -> undef 892 return UndefValue::get(ExtractValueInst::getIndexedType(Agg->getType(), 893 Idxs)); 894 895 if (isa<ConstantAggregateZero>(Agg)) // ev(0, x) -> 0 896 return 897 Constant::getNullValue(ExtractValueInst::getIndexedType(Agg->getType(), 898 Idxs)); 899 900 // Otherwise recurse. 901 if (ConstantStruct *CS = dyn_cast<ConstantStruct>(Agg)) 902 return ConstantFoldExtractValueInstruction(CS->getOperand(Idxs[0]), 903 Idxs.slice(1)); 904 905 if (ConstantArray *CA = dyn_cast<ConstantArray>(Agg)) 906 return ConstantFoldExtractValueInstruction(CA->getOperand(Idxs[0]), 907 Idxs.slice(1)); 908 ConstantVector *CV = cast<ConstantVector>(Agg); 909 return ConstantFoldExtractValueInstruction(CV->getOperand(Idxs[0]), 910 Idxs.slice(1)); 911 } 912 913 Constant *llvm::ConstantFoldInsertValueInstruction(Constant *Agg, 914 Constant *Val, 915 ArrayRef<unsigned> Idxs) { 916 // Base case: no indices, so replace the entire value. 917 if (Idxs.empty()) 918 return Val; 919 920 if (isa<UndefValue>(Agg)) { 921 // Insertion of constant into aggregate undef 922 // Optimize away insertion of undef. 923 if (isa<UndefValue>(Val)) 924 return Agg; 925 926 // Otherwise break the aggregate undef into multiple undefs and do 927 // the insertion. 928 CompositeType *AggTy = cast<CompositeType>(Agg->getType()); 929 unsigned numOps; 930 if (ArrayType *AR = dyn_cast<ArrayType>(AggTy)) 931 numOps = AR->getNumElements(); 932 else 933 numOps = cast<StructType>(AggTy)->getNumElements(); 934 935 std::vector<Constant*> Ops(numOps); 936 for (unsigned i = 0; i < numOps; ++i) { 937 Type *MemberTy = AggTy->getTypeAtIndex(i); 938 Constant *Op = 939 (Idxs[0] == i) ? 940 ConstantFoldInsertValueInstruction(UndefValue::get(MemberTy), 941 Val, Idxs.slice(1)) : 942 UndefValue::get(MemberTy); 943 Ops[i] = Op; 944 } 945 946 if (StructType* ST = dyn_cast<StructType>(AggTy)) 947 return ConstantStruct::get(ST, Ops); 948 return ConstantArray::get(cast<ArrayType>(AggTy), Ops); 949 } 950 951 if (isa<ConstantAggregateZero>(Agg)) { 952 // Insertion of constant into aggregate zero 953 // Optimize away insertion of zero. 954 if (Val->isNullValue()) 955 return Agg; 956 957 // Otherwise break the aggregate zero into multiple zeros and do 958 // the insertion. 959 CompositeType *AggTy = cast<CompositeType>(Agg->getType()); 960 unsigned numOps; 961 if (ArrayType *AR = dyn_cast<ArrayType>(AggTy)) 962 numOps = AR->getNumElements(); 963 else 964 numOps = cast<StructType>(AggTy)->getNumElements(); 965 966 std::vector<Constant*> Ops(numOps); 967 for (unsigned i = 0; i < numOps; ++i) { 968 Type *MemberTy = AggTy->getTypeAtIndex(i); 969 Constant *Op = 970 (Idxs[0] == i) ? 971 ConstantFoldInsertValueInstruction(Constant::getNullValue(MemberTy), 972 Val, Idxs.slice(1)) : 973 Constant::getNullValue(MemberTy); 974 Ops[i] = Op; 975 } 976 977 if (StructType *ST = dyn_cast<StructType>(AggTy)) 978 return ConstantStruct::get(ST, Ops); 979 return ConstantArray::get(cast<ArrayType>(AggTy), Ops); 980 } 981 982 if (isa<ConstantStruct>(Agg) || isa<ConstantArray>(Agg)) { 983 // Insertion of constant into aggregate constant. 984 std::vector<Constant*> Ops(Agg->getNumOperands()); 985 for (unsigned i = 0; i < Agg->getNumOperands(); ++i) { 986 Constant *Op = cast<Constant>(Agg->getOperand(i)); 987 if (Idxs[0] == i) 988 Op = ConstantFoldInsertValueInstruction(Op, Val, Idxs.slice(1)); 989 Ops[i] = Op; 990 } 991 992 if (StructType* ST = dyn_cast<StructType>(Agg->getType())) 993 return ConstantStruct::get(ST, Ops); 994 return ConstantArray::get(cast<ArrayType>(Agg->getType()), Ops); 995 } 996 997 return 0; 998 } 999 1000 1001 Constant *llvm::ConstantFoldBinaryInstruction(unsigned Opcode, 1002 Constant *C1, Constant *C2) { 1003 // No compile-time operations on this type yet. 1004 if (C1->getType()->isPPC_FP128Ty()) 1005 return 0; 1006 1007 // Handle UndefValue up front. 1008 if (isa<UndefValue>(C1) || isa<UndefValue>(C2)) { 1009 switch (Opcode) { 1010 case Instruction::Xor: 1011 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) 1012 // Handle undef ^ undef -> 0 special case. This is a common 1013 // idiom (misuse). 1014 return Constant::getNullValue(C1->getType()); 1015 // Fallthrough 1016 case Instruction::Add: 1017 case Instruction::Sub: 1018 return UndefValue::get(C1->getType()); 1019 case Instruction::And: 1020 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) // undef & undef -> undef 1021 return C1; 1022 return Constant::getNullValue(C1->getType()); // undef & X -> 0 1023 case Instruction::Mul: { 1024 ConstantInt *CI; 1025 // X * undef -> undef if X is odd or undef 1026 if (((CI = dyn_cast<ConstantInt>(C1)) && CI->getValue()[0]) || 1027 ((CI = dyn_cast<ConstantInt>(C2)) && CI->getValue()[0]) || 1028 (isa<UndefValue>(C1) && isa<UndefValue>(C2))) 1029 return UndefValue::get(C1->getType()); 1030 1031 // X * undef -> 0 otherwise 1032 return Constant::getNullValue(C1->getType()); 1033 } 1034 case Instruction::UDiv: 1035 case Instruction::SDiv: 1036 // undef / 1 -> undef 1037 if (Opcode == Instruction::UDiv || Opcode == Instruction::SDiv) 1038 if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) 1039 if (CI2->isOne()) 1040 return C1; 1041 // FALL THROUGH 1042 case Instruction::URem: 1043 case Instruction::SRem: 1044 if (!isa<UndefValue>(C2)) // undef / X -> 0 1045 return Constant::getNullValue(C1->getType()); 1046 return C2; // X / undef -> undef 1047 case Instruction::Or: // X | undef -> -1 1048 if (isa<UndefValue>(C1) && isa<UndefValue>(C2)) // undef | undef -> undef 1049 return C1; 1050 return Constant::getAllOnesValue(C1->getType()); // undef | X -> ~0 1051 case Instruction::LShr: 1052 if (isa<UndefValue>(C2) && isa<UndefValue>(C1)) 1053 return C1; // undef lshr undef -> undef 1054 return Constant::getNullValue(C1->getType()); // X lshr undef -> 0 1055 // undef lshr X -> 0 1056 case Instruction::AShr: 1057 if (!isa<UndefValue>(C2)) // undef ashr X --> all ones 1058 return Constant::getAllOnesValue(C1->getType()); 1059 else if (isa<UndefValue>(C1)) 1060 return C1; // undef ashr undef -> undef 1061 else 1062 return C1; // X ashr undef --> X 1063 case Instruction::Shl: 1064 if (isa<UndefValue>(C2) && isa<UndefValue>(C1)) 1065 return C1; // undef shl undef -> undef 1066 // undef << X -> 0 or X << undef -> 0 1067 return Constant::getNullValue(C1->getType()); 1068 } 1069 } 1070 1071 // Handle simplifications when the RHS is a constant int. 1072 if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) { 1073 switch (Opcode) { 1074 case Instruction::Add: 1075 if (CI2->equalsInt(0)) return C1; // X + 0 == X 1076 break; 1077 case Instruction::Sub: 1078 if (CI2->equalsInt(0)) return C1; // X - 0 == X 1079 break; 1080 case Instruction::Mul: 1081 if (CI2->equalsInt(0)) return C2; // X * 0 == 0 1082 if (CI2->equalsInt(1)) 1083 return C1; // X * 1 == X 1084 break; 1085 case Instruction::UDiv: 1086 case Instruction::SDiv: 1087 if (CI2->equalsInt(1)) 1088 return C1; // X / 1 == X 1089 if (CI2->equalsInt(0)) 1090 return UndefValue::get(CI2->getType()); // X / 0 == undef 1091 break; 1092 case Instruction::URem: 1093 case Instruction::SRem: 1094 if (CI2->equalsInt(1)) 1095 return Constant::getNullValue(CI2->getType()); // X % 1 == 0 1096 if (CI2->equalsInt(0)) 1097 return UndefValue::get(CI2->getType()); // X % 0 == undef 1098 break; 1099 case Instruction::And: 1100 if (CI2->isZero()) return C2; // X & 0 == 0 1101 if (CI2->isAllOnesValue()) 1102 return C1; // X & -1 == X 1103 1104 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1105 // (zext i32 to i64) & 4294967295 -> (zext i32 to i64) 1106 if (CE1->getOpcode() == Instruction::ZExt) { 1107 unsigned DstWidth = CI2->getType()->getBitWidth(); 1108 unsigned SrcWidth = 1109 CE1->getOperand(0)->getType()->getPrimitiveSizeInBits(); 1110 APInt PossiblySetBits(APInt::getLowBitsSet(DstWidth, SrcWidth)); 1111 if ((PossiblySetBits & CI2->getValue()) == PossiblySetBits) 1112 return C1; 1113 } 1114 1115 // If and'ing the address of a global with a constant, fold it. 1116 if (CE1->getOpcode() == Instruction::PtrToInt && 1117 isa<GlobalValue>(CE1->getOperand(0))) { 1118 GlobalValue *GV = cast<GlobalValue>(CE1->getOperand(0)); 1119 1120 // Functions are at least 4-byte aligned. 1121 unsigned GVAlign = GV->getAlignment(); 1122 if (isa<Function>(GV)) 1123 GVAlign = std::max(GVAlign, 4U); 1124 1125 if (GVAlign > 1) { 1126 unsigned DstWidth = CI2->getType()->getBitWidth(); 1127 unsigned SrcWidth = std::min(DstWidth, Log2_32(GVAlign)); 1128 APInt BitsNotSet(APInt::getLowBitsSet(DstWidth, SrcWidth)); 1129 1130 // If checking bits we know are clear, return zero. 1131 if ((CI2->getValue() & BitsNotSet) == CI2->getValue()) 1132 return Constant::getNullValue(CI2->getType()); 1133 } 1134 } 1135 } 1136 break; 1137 case Instruction::Or: 1138 if (CI2->equalsInt(0)) return C1; // X | 0 == X 1139 if (CI2->isAllOnesValue()) 1140 return C2; // X | -1 == -1 1141 break; 1142 case Instruction::Xor: 1143 if (CI2->equalsInt(0)) return C1; // X ^ 0 == X 1144 1145 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1146 switch (CE1->getOpcode()) { 1147 default: break; 1148 case Instruction::ICmp: 1149 case Instruction::FCmp: 1150 // cmp pred ^ true -> cmp !pred 1151 assert(CI2->equalsInt(1)); 1152 CmpInst::Predicate pred = (CmpInst::Predicate)CE1->getPredicate(); 1153 pred = CmpInst::getInversePredicate(pred); 1154 return ConstantExpr::getCompare(pred, CE1->getOperand(0), 1155 CE1->getOperand(1)); 1156 } 1157 } 1158 break; 1159 case Instruction::AShr: 1160 // ashr (zext C to Ty), C2 -> lshr (zext C, CSA), C2 1161 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) 1162 if (CE1->getOpcode() == Instruction::ZExt) // Top bits known zero. 1163 return ConstantExpr::getLShr(C1, C2); 1164 break; 1165 } 1166 } else if (isa<ConstantInt>(C1)) { 1167 // If C1 is a ConstantInt and C2 is not, swap the operands. 1168 if (Instruction::isCommutative(Opcode)) 1169 return ConstantExpr::get(Opcode, C2, C1); 1170 } 1171 1172 // At this point we know neither constant is an UndefValue. 1173 if (ConstantInt *CI1 = dyn_cast<ConstantInt>(C1)) { 1174 if (ConstantInt *CI2 = dyn_cast<ConstantInt>(C2)) { 1175 using namespace APIntOps; 1176 const APInt &C1V = CI1->getValue(); 1177 const APInt &C2V = CI2->getValue(); 1178 switch (Opcode) { 1179 default: 1180 break; 1181 case Instruction::Add: 1182 return ConstantInt::get(CI1->getContext(), C1V + C2V); 1183 case Instruction::Sub: 1184 return ConstantInt::get(CI1->getContext(), C1V - C2V); 1185 case Instruction::Mul: 1186 return ConstantInt::get(CI1->getContext(), C1V * C2V); 1187 case Instruction::UDiv: 1188 assert(!CI2->isNullValue() && "Div by zero handled above"); 1189 return ConstantInt::get(CI1->getContext(), C1V.udiv(C2V)); 1190 case Instruction::SDiv: 1191 assert(!CI2->isNullValue() && "Div by zero handled above"); 1192 if (C2V.isAllOnesValue() && C1V.isMinSignedValue()) 1193 return UndefValue::get(CI1->getType()); // MIN_INT / -1 -> undef 1194 return ConstantInt::get(CI1->getContext(), C1V.sdiv(C2V)); 1195 case Instruction::URem: 1196 assert(!CI2->isNullValue() && "Div by zero handled above"); 1197 return ConstantInt::get(CI1->getContext(), C1V.urem(C2V)); 1198 case Instruction::SRem: 1199 assert(!CI2->isNullValue() && "Div by zero handled above"); 1200 if (C2V.isAllOnesValue() && C1V.isMinSignedValue()) 1201 return UndefValue::get(CI1->getType()); // MIN_INT % -1 -> undef 1202 return ConstantInt::get(CI1->getContext(), C1V.srem(C2V)); 1203 case Instruction::And: 1204 return ConstantInt::get(CI1->getContext(), C1V & C2V); 1205 case Instruction::Or: 1206 return ConstantInt::get(CI1->getContext(), C1V | C2V); 1207 case Instruction::Xor: 1208 return ConstantInt::get(CI1->getContext(), C1V ^ C2V); 1209 case Instruction::Shl: { 1210 uint32_t shiftAmt = C2V.getZExtValue(); 1211 if (shiftAmt < C1V.getBitWidth()) 1212 return ConstantInt::get(CI1->getContext(), C1V.shl(shiftAmt)); 1213 else 1214 return UndefValue::get(C1->getType()); // too big shift is undef 1215 } 1216 case Instruction::LShr: { 1217 uint32_t shiftAmt = C2V.getZExtValue(); 1218 if (shiftAmt < C1V.getBitWidth()) 1219 return ConstantInt::get(CI1->getContext(), C1V.lshr(shiftAmt)); 1220 else 1221 return UndefValue::get(C1->getType()); // too big shift is undef 1222 } 1223 case Instruction::AShr: { 1224 uint32_t shiftAmt = C2V.getZExtValue(); 1225 if (shiftAmt < C1V.getBitWidth()) 1226 return ConstantInt::get(CI1->getContext(), C1V.ashr(shiftAmt)); 1227 else 1228 return UndefValue::get(C1->getType()); // too big shift is undef 1229 } 1230 } 1231 } 1232 1233 switch (Opcode) { 1234 case Instruction::SDiv: 1235 case Instruction::UDiv: 1236 case Instruction::URem: 1237 case Instruction::SRem: 1238 case Instruction::LShr: 1239 case Instruction::AShr: 1240 case Instruction::Shl: 1241 if (CI1->equalsInt(0)) return C1; 1242 break; 1243 default: 1244 break; 1245 } 1246 } else if (ConstantFP *CFP1 = dyn_cast<ConstantFP>(C1)) { 1247 if (ConstantFP *CFP2 = dyn_cast<ConstantFP>(C2)) { 1248 APFloat C1V = CFP1->getValueAPF(); 1249 APFloat C2V = CFP2->getValueAPF(); 1250 APFloat C3V = C1V; // copy for modification 1251 switch (Opcode) { 1252 default: 1253 break; 1254 case Instruction::FAdd: 1255 (void)C3V.add(C2V, APFloat::rmNearestTiesToEven); 1256 return ConstantFP::get(C1->getContext(), C3V); 1257 case Instruction::FSub: 1258 (void)C3V.subtract(C2V, APFloat::rmNearestTiesToEven); 1259 return ConstantFP::get(C1->getContext(), C3V); 1260 case Instruction::FMul: 1261 (void)C3V.multiply(C2V, APFloat::rmNearestTiesToEven); 1262 return ConstantFP::get(C1->getContext(), C3V); 1263 case Instruction::FDiv: 1264 (void)C3V.divide(C2V, APFloat::rmNearestTiesToEven); 1265 return ConstantFP::get(C1->getContext(), C3V); 1266 case Instruction::FRem: 1267 (void)C3V.mod(C2V, APFloat::rmNearestTiesToEven); 1268 return ConstantFP::get(C1->getContext(), C3V); 1269 } 1270 } 1271 } else if (VectorType *VTy = dyn_cast<VectorType>(C1->getType())) { 1272 ConstantVector *CP1 = dyn_cast<ConstantVector>(C1); 1273 ConstantVector *CP2 = dyn_cast<ConstantVector>(C2); 1274 if ((CP1 != NULL || isa<ConstantAggregateZero>(C1)) && 1275 (CP2 != NULL || isa<ConstantAggregateZero>(C2))) { 1276 std::vector<Constant*> Res; 1277 Type* EltTy = VTy->getElementType(); 1278 Constant *C1 = 0; 1279 Constant *C2 = 0; 1280 switch (Opcode) { 1281 default: 1282 break; 1283 case Instruction::Add: 1284 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1285 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1286 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1287 Res.push_back(ConstantExpr::getAdd(C1, C2)); 1288 } 1289 return ConstantVector::get(Res); 1290 case Instruction::FAdd: 1291 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1292 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1293 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1294 Res.push_back(ConstantExpr::getFAdd(C1, C2)); 1295 } 1296 return ConstantVector::get(Res); 1297 case Instruction::Sub: 1298 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1299 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1300 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1301 Res.push_back(ConstantExpr::getSub(C1, C2)); 1302 } 1303 return ConstantVector::get(Res); 1304 case Instruction::FSub: 1305 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1306 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1307 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1308 Res.push_back(ConstantExpr::getFSub(C1, C2)); 1309 } 1310 return ConstantVector::get(Res); 1311 case Instruction::Mul: 1312 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1313 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1314 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1315 Res.push_back(ConstantExpr::getMul(C1, C2)); 1316 } 1317 return ConstantVector::get(Res); 1318 case Instruction::FMul: 1319 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1320 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1321 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1322 Res.push_back(ConstantExpr::getFMul(C1, C2)); 1323 } 1324 return ConstantVector::get(Res); 1325 case Instruction::UDiv: 1326 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1327 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1328 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1329 Res.push_back(ConstantExpr::getUDiv(C1, C2)); 1330 } 1331 return ConstantVector::get(Res); 1332 case Instruction::SDiv: 1333 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1334 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1335 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1336 Res.push_back(ConstantExpr::getSDiv(C1, C2)); 1337 } 1338 return ConstantVector::get(Res); 1339 case Instruction::FDiv: 1340 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1341 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1342 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1343 Res.push_back(ConstantExpr::getFDiv(C1, C2)); 1344 } 1345 return ConstantVector::get(Res); 1346 case Instruction::URem: 1347 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1348 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1349 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1350 Res.push_back(ConstantExpr::getURem(C1, C2)); 1351 } 1352 return ConstantVector::get(Res); 1353 case Instruction::SRem: 1354 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1355 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1356 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1357 Res.push_back(ConstantExpr::getSRem(C1, C2)); 1358 } 1359 return ConstantVector::get(Res); 1360 case Instruction::FRem: 1361 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1362 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1363 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1364 Res.push_back(ConstantExpr::getFRem(C1, C2)); 1365 } 1366 return ConstantVector::get(Res); 1367 case Instruction::And: 1368 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1369 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1370 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1371 Res.push_back(ConstantExpr::getAnd(C1, C2)); 1372 } 1373 return ConstantVector::get(Res); 1374 case Instruction::Or: 1375 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1376 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1377 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1378 Res.push_back(ConstantExpr::getOr(C1, C2)); 1379 } 1380 return ConstantVector::get(Res); 1381 case Instruction::Xor: 1382 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1383 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1384 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1385 Res.push_back(ConstantExpr::getXor(C1, C2)); 1386 } 1387 return ConstantVector::get(Res); 1388 case Instruction::LShr: 1389 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1390 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1391 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1392 Res.push_back(ConstantExpr::getLShr(C1, C2)); 1393 } 1394 return ConstantVector::get(Res); 1395 case Instruction::AShr: 1396 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1397 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1398 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1399 Res.push_back(ConstantExpr::getAShr(C1, C2)); 1400 } 1401 return ConstantVector::get(Res); 1402 case Instruction::Shl: 1403 for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i) { 1404 C1 = CP1 ? CP1->getOperand(i) : Constant::getNullValue(EltTy); 1405 C2 = CP2 ? CP2->getOperand(i) : Constant::getNullValue(EltTy); 1406 Res.push_back(ConstantExpr::getShl(C1, C2)); 1407 } 1408 return ConstantVector::get(Res); 1409 } 1410 } 1411 } 1412 1413 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 1414 // There are many possible foldings we could do here. We should probably 1415 // at least fold add of a pointer with an integer into the appropriate 1416 // getelementptr. This will improve alias analysis a bit. 1417 1418 // Given ((a + b) + c), if (b + c) folds to something interesting, return 1419 // (a + (b + c)). 1420 if (Instruction::isAssociative(Opcode) && CE1->getOpcode() == Opcode) { 1421 Constant *T = ConstantExpr::get(Opcode, CE1->getOperand(1), C2); 1422 if (!isa<ConstantExpr>(T) || cast<ConstantExpr>(T)->getOpcode() != Opcode) 1423 return ConstantExpr::get(Opcode, CE1->getOperand(0), T); 1424 } 1425 } else if (isa<ConstantExpr>(C2)) { 1426 // If C2 is a constant expr and C1 isn't, flop them around and fold the 1427 // other way if possible. 1428 if (Instruction::isCommutative(Opcode)) 1429 return ConstantFoldBinaryInstruction(Opcode, C2, C1); 1430 } 1431 1432 // i1 can be simplified in many cases. 1433 if (C1->getType()->isIntegerTy(1)) { 1434 switch (Opcode) { 1435 case Instruction::Add: 1436 case Instruction::Sub: 1437 return ConstantExpr::getXor(C1, C2); 1438 case Instruction::Mul: 1439 return ConstantExpr::getAnd(C1, C2); 1440 case Instruction::Shl: 1441 case Instruction::LShr: 1442 case Instruction::AShr: 1443 // We can assume that C2 == 0. If it were one the result would be 1444 // undefined because the shift value is as large as the bitwidth. 1445 return C1; 1446 case Instruction::SDiv: 1447 case Instruction::UDiv: 1448 // We can assume that C2 == 1. If it were zero the result would be 1449 // undefined through division by zero. 1450 return C1; 1451 case Instruction::URem: 1452 case Instruction::SRem: 1453 // We can assume that C2 == 1. If it were zero the result would be 1454 // undefined through division by zero. 1455 return ConstantInt::getFalse(C1->getContext()); 1456 default: 1457 break; 1458 } 1459 } 1460 1461 // We don't know how to fold this. 1462 return 0; 1463 } 1464 1465 /// isZeroSizedType - This type is zero sized if its an array or structure of 1466 /// zero sized types. The only leaf zero sized type is an empty structure. 1467 static bool isMaybeZeroSizedType(Type *Ty) { 1468 if (StructType *STy = dyn_cast<StructType>(Ty)) { 1469 if (STy->isOpaque()) return true; // Can't say. 1470 1471 // If all of elements have zero size, this does too. 1472 for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) 1473 if (!isMaybeZeroSizedType(STy->getElementType(i))) return false; 1474 return true; 1475 1476 } else if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) { 1477 return isMaybeZeroSizedType(ATy->getElementType()); 1478 } 1479 return false; 1480 } 1481 1482 /// IdxCompare - Compare the two constants as though they were getelementptr 1483 /// indices. This allows coersion of the types to be the same thing. 1484 /// 1485 /// If the two constants are the "same" (after coersion), return 0. If the 1486 /// first is less than the second, return -1, if the second is less than the 1487 /// first, return 1. If the constants are not integral, return -2. 1488 /// 1489 static int IdxCompare(Constant *C1, Constant *C2, Type *ElTy) { 1490 if (C1 == C2) return 0; 1491 1492 // Ok, we found a different index. If they are not ConstantInt, we can't do 1493 // anything with them. 1494 if (!isa<ConstantInt>(C1) || !isa<ConstantInt>(C2)) 1495 return -2; // don't know! 1496 1497 // Ok, we have two differing integer indices. Sign extend them to be the same 1498 // type. Long is always big enough, so we use it. 1499 if (!C1->getType()->isIntegerTy(64)) 1500 C1 = ConstantExpr::getSExt(C1, Type::getInt64Ty(C1->getContext())); 1501 1502 if (!C2->getType()->isIntegerTy(64)) 1503 C2 = ConstantExpr::getSExt(C2, Type::getInt64Ty(C1->getContext())); 1504 1505 if (C1 == C2) return 0; // They are equal 1506 1507 // If the type being indexed over is really just a zero sized type, there is 1508 // no pointer difference being made here. 1509 if (isMaybeZeroSizedType(ElTy)) 1510 return -2; // dunno. 1511 1512 // If they are really different, now that they are the same type, then we 1513 // found a difference! 1514 if (cast<ConstantInt>(C1)->getSExtValue() < 1515 cast<ConstantInt>(C2)->getSExtValue()) 1516 return -1; 1517 else 1518 return 1; 1519 } 1520 1521 /// evaluateFCmpRelation - This function determines if there is anything we can 1522 /// decide about the two constants provided. This doesn't need to handle simple 1523 /// things like ConstantFP comparisons, but should instead handle ConstantExprs. 1524 /// If we can determine that the two constants have a particular relation to 1525 /// each other, we should return the corresponding FCmpInst predicate, 1526 /// otherwise return FCmpInst::BAD_FCMP_PREDICATE. This is used below in 1527 /// ConstantFoldCompareInstruction. 1528 /// 1529 /// To simplify this code we canonicalize the relation so that the first 1530 /// operand is always the most "complex" of the two. We consider ConstantFP 1531 /// to be the simplest, and ConstantExprs to be the most complex. 1532 static FCmpInst::Predicate evaluateFCmpRelation(Constant *V1, Constant *V2) { 1533 assert(V1->getType() == V2->getType() && 1534 "Cannot compare values of different types!"); 1535 1536 // No compile-time operations on this type yet. 1537 if (V1->getType()->isPPC_FP128Ty()) 1538 return FCmpInst::BAD_FCMP_PREDICATE; 1539 1540 // Handle degenerate case quickly 1541 if (V1 == V2) return FCmpInst::FCMP_OEQ; 1542 1543 if (!isa<ConstantExpr>(V1)) { 1544 if (!isa<ConstantExpr>(V2)) { 1545 // We distilled thisUse the standard constant folder for a few cases 1546 ConstantInt *R = 0; 1547 R = dyn_cast<ConstantInt>( 1548 ConstantExpr::getFCmp(FCmpInst::FCMP_OEQ, V1, V2)); 1549 if (R && !R->isZero()) 1550 return FCmpInst::FCMP_OEQ; 1551 R = dyn_cast<ConstantInt>( 1552 ConstantExpr::getFCmp(FCmpInst::FCMP_OLT, V1, V2)); 1553 if (R && !R->isZero()) 1554 return FCmpInst::FCMP_OLT; 1555 R = dyn_cast<ConstantInt>( 1556 ConstantExpr::getFCmp(FCmpInst::FCMP_OGT, V1, V2)); 1557 if (R && !R->isZero()) 1558 return FCmpInst::FCMP_OGT; 1559 1560 // Nothing more we can do 1561 return FCmpInst::BAD_FCMP_PREDICATE; 1562 } 1563 1564 // If the first operand is simple and second is ConstantExpr, swap operands. 1565 FCmpInst::Predicate SwappedRelation = evaluateFCmpRelation(V2, V1); 1566 if (SwappedRelation != FCmpInst::BAD_FCMP_PREDICATE) 1567 return FCmpInst::getSwappedPredicate(SwappedRelation); 1568 } else { 1569 // Ok, the LHS is known to be a constantexpr. The RHS can be any of a 1570 // constantexpr or a simple constant. 1571 ConstantExpr *CE1 = cast<ConstantExpr>(V1); 1572 switch (CE1->getOpcode()) { 1573 case Instruction::FPTrunc: 1574 case Instruction::FPExt: 1575 case Instruction::UIToFP: 1576 case Instruction::SIToFP: 1577 // We might be able to do something with these but we don't right now. 1578 break; 1579 default: 1580 break; 1581 } 1582 } 1583 // There are MANY other foldings that we could perform here. They will 1584 // probably be added on demand, as they seem needed. 1585 return FCmpInst::BAD_FCMP_PREDICATE; 1586 } 1587 1588 /// evaluateICmpRelation - This function determines if there is anything we can 1589 /// decide about the two constants provided. This doesn't need to handle simple 1590 /// things like integer comparisons, but should instead handle ConstantExprs 1591 /// and GlobalValues. If we can determine that the two constants have a 1592 /// particular relation to each other, we should return the corresponding ICmp 1593 /// predicate, otherwise return ICmpInst::BAD_ICMP_PREDICATE. 1594 /// 1595 /// To simplify this code we canonicalize the relation so that the first 1596 /// operand is always the most "complex" of the two. We consider simple 1597 /// constants (like ConstantInt) to be the simplest, followed by 1598 /// GlobalValues, followed by ConstantExpr's (the most complex). 1599 /// 1600 static ICmpInst::Predicate evaluateICmpRelation(Constant *V1, Constant *V2, 1601 bool isSigned) { 1602 assert(V1->getType() == V2->getType() && 1603 "Cannot compare different types of values!"); 1604 if (V1 == V2) return ICmpInst::ICMP_EQ; 1605 1606 if (!isa<ConstantExpr>(V1) && !isa<GlobalValue>(V1) && 1607 !isa<BlockAddress>(V1)) { 1608 if (!isa<GlobalValue>(V2) && !isa<ConstantExpr>(V2) && 1609 !isa<BlockAddress>(V2)) { 1610 // We distilled this down to a simple case, use the standard constant 1611 // folder. 1612 ConstantInt *R = 0; 1613 ICmpInst::Predicate pred = ICmpInst::ICMP_EQ; 1614 R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2)); 1615 if (R && !R->isZero()) 1616 return pred; 1617 pred = isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; 1618 R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2)); 1619 if (R && !R->isZero()) 1620 return pred; 1621 pred = isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1622 R = dyn_cast<ConstantInt>(ConstantExpr::getICmp(pred, V1, V2)); 1623 if (R && !R->isZero()) 1624 return pred; 1625 1626 // If we couldn't figure it out, bail. 1627 return ICmpInst::BAD_ICMP_PREDICATE; 1628 } 1629 1630 // If the first operand is simple, swap operands. 1631 ICmpInst::Predicate SwappedRelation = 1632 evaluateICmpRelation(V2, V1, isSigned); 1633 if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE) 1634 return ICmpInst::getSwappedPredicate(SwappedRelation); 1635 1636 } else if (const GlobalValue *GV = dyn_cast<GlobalValue>(V1)) { 1637 if (isa<ConstantExpr>(V2)) { // Swap as necessary. 1638 ICmpInst::Predicate SwappedRelation = 1639 evaluateICmpRelation(V2, V1, isSigned); 1640 if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE) 1641 return ICmpInst::getSwappedPredicate(SwappedRelation); 1642 return ICmpInst::BAD_ICMP_PREDICATE; 1643 } 1644 1645 // Now we know that the RHS is a GlobalValue, BlockAddress or simple 1646 // constant (which, since the types must match, means that it's a 1647 // ConstantPointerNull). 1648 if (const GlobalValue *GV2 = dyn_cast<GlobalValue>(V2)) { 1649 // Don't try to decide equality of aliases. 1650 if (!isa<GlobalAlias>(GV) && !isa<GlobalAlias>(GV2)) 1651 if (!GV->hasExternalWeakLinkage() || !GV2->hasExternalWeakLinkage()) 1652 return ICmpInst::ICMP_NE; 1653 } else if (isa<BlockAddress>(V2)) { 1654 return ICmpInst::ICMP_NE; // Globals never equal labels. 1655 } else { 1656 assert(isa<ConstantPointerNull>(V2) && "Canonicalization guarantee!"); 1657 // GlobalVals can never be null unless they have external weak linkage. 1658 // We don't try to evaluate aliases here. 1659 if (!GV->hasExternalWeakLinkage() && !isa<GlobalAlias>(GV)) 1660 return ICmpInst::ICMP_NE; 1661 } 1662 } else if (const BlockAddress *BA = dyn_cast<BlockAddress>(V1)) { 1663 if (isa<ConstantExpr>(V2)) { // Swap as necessary. 1664 ICmpInst::Predicate SwappedRelation = 1665 evaluateICmpRelation(V2, V1, isSigned); 1666 if (SwappedRelation != ICmpInst::BAD_ICMP_PREDICATE) 1667 return ICmpInst::getSwappedPredicate(SwappedRelation); 1668 return ICmpInst::BAD_ICMP_PREDICATE; 1669 } 1670 1671 // Now we know that the RHS is a GlobalValue, BlockAddress or simple 1672 // constant (which, since the types must match, means that it is a 1673 // ConstantPointerNull). 1674 if (const BlockAddress *BA2 = dyn_cast<BlockAddress>(V2)) { 1675 // Block address in another function can't equal this one, but block 1676 // addresses in the current function might be the same if blocks are 1677 // empty. 1678 if (BA2->getFunction() != BA->getFunction()) 1679 return ICmpInst::ICMP_NE; 1680 } else { 1681 // Block addresses aren't null, don't equal the address of globals. 1682 assert((isa<ConstantPointerNull>(V2) || isa<GlobalValue>(V2)) && 1683 "Canonicalization guarantee!"); 1684 return ICmpInst::ICMP_NE; 1685 } 1686 } else { 1687 // Ok, the LHS is known to be a constantexpr. The RHS can be any of a 1688 // constantexpr, a global, block address, or a simple constant. 1689 ConstantExpr *CE1 = cast<ConstantExpr>(V1); 1690 Constant *CE1Op0 = CE1->getOperand(0); 1691 1692 switch (CE1->getOpcode()) { 1693 case Instruction::Trunc: 1694 case Instruction::FPTrunc: 1695 case Instruction::FPExt: 1696 case Instruction::FPToUI: 1697 case Instruction::FPToSI: 1698 break; // We can't evaluate floating point casts or truncations. 1699 1700 case Instruction::UIToFP: 1701 case Instruction::SIToFP: 1702 case Instruction::BitCast: 1703 case Instruction::ZExt: 1704 case Instruction::SExt: 1705 // If the cast is not actually changing bits, and the second operand is a 1706 // null pointer, do the comparison with the pre-casted value. 1707 if (V2->isNullValue() && 1708 (CE1->getType()->isPointerTy() || CE1->getType()->isIntegerTy())) { 1709 if (CE1->getOpcode() == Instruction::ZExt) isSigned = false; 1710 if (CE1->getOpcode() == Instruction::SExt) isSigned = true; 1711 return evaluateICmpRelation(CE1Op0, 1712 Constant::getNullValue(CE1Op0->getType()), 1713 isSigned); 1714 } 1715 break; 1716 1717 case Instruction::GetElementPtr: 1718 // Ok, since this is a getelementptr, we know that the constant has a 1719 // pointer type. Check the various cases. 1720 if (isa<ConstantPointerNull>(V2)) { 1721 // If we are comparing a GEP to a null pointer, check to see if the base 1722 // of the GEP equals the null pointer. 1723 if (const GlobalValue *GV = dyn_cast<GlobalValue>(CE1Op0)) { 1724 if (GV->hasExternalWeakLinkage()) 1725 // Weak linkage GVals could be zero or not. We're comparing that 1726 // to null pointer so its greater-or-equal 1727 return isSigned ? ICmpInst::ICMP_SGE : ICmpInst::ICMP_UGE; 1728 else 1729 // If its not weak linkage, the GVal must have a non-zero address 1730 // so the result is greater-than 1731 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1732 } else if (isa<ConstantPointerNull>(CE1Op0)) { 1733 // If we are indexing from a null pointer, check to see if we have any 1734 // non-zero indices. 1735 for (unsigned i = 1, e = CE1->getNumOperands(); i != e; ++i) 1736 if (!CE1->getOperand(i)->isNullValue()) 1737 // Offsetting from null, must not be equal. 1738 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1739 // Only zero indexes from null, must still be zero. 1740 return ICmpInst::ICMP_EQ; 1741 } 1742 // Otherwise, we can't really say if the first operand is null or not. 1743 } else if (const GlobalValue *GV2 = dyn_cast<GlobalValue>(V2)) { 1744 if (isa<ConstantPointerNull>(CE1Op0)) { 1745 if (GV2->hasExternalWeakLinkage()) 1746 // Weak linkage GVals could be zero or not. We're comparing it to 1747 // a null pointer, so its less-or-equal 1748 return isSigned ? ICmpInst::ICMP_SLE : ICmpInst::ICMP_ULE; 1749 else 1750 // If its not weak linkage, the GVal must have a non-zero address 1751 // so the result is less-than 1752 return isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; 1753 } else if (const GlobalValue *GV = dyn_cast<GlobalValue>(CE1Op0)) { 1754 if (GV == GV2) { 1755 // If this is a getelementptr of the same global, then it must be 1756 // different. Because the types must match, the getelementptr could 1757 // only have at most one index, and because we fold getelementptr's 1758 // with a single zero index, it must be nonzero. 1759 assert(CE1->getNumOperands() == 2 && 1760 !CE1->getOperand(1)->isNullValue() && 1761 "Surprising getelementptr!"); 1762 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1763 } else { 1764 // If they are different globals, we don't know what the value is, 1765 // but they can't be equal. 1766 return ICmpInst::ICMP_NE; 1767 } 1768 } 1769 } else { 1770 ConstantExpr *CE2 = cast<ConstantExpr>(V2); 1771 Constant *CE2Op0 = CE2->getOperand(0); 1772 1773 // There are MANY other foldings that we could perform here. They will 1774 // probably be added on demand, as they seem needed. 1775 switch (CE2->getOpcode()) { 1776 default: break; 1777 case Instruction::GetElementPtr: 1778 // By far the most common case to handle is when the base pointers are 1779 // obviously to the same or different globals. 1780 if (isa<GlobalValue>(CE1Op0) && isa<GlobalValue>(CE2Op0)) { 1781 if (CE1Op0 != CE2Op0) // Don't know relative ordering, but not equal 1782 return ICmpInst::ICMP_NE; 1783 // Ok, we know that both getelementptr instructions are based on the 1784 // same global. From this, we can precisely determine the relative 1785 // ordering of the resultant pointers. 1786 unsigned i = 1; 1787 1788 // The logic below assumes that the result of the comparison 1789 // can be determined by finding the first index that differs. 1790 // This doesn't work if there is over-indexing in any 1791 // subsequent indices, so check for that case first. 1792 if (!CE1->isGEPWithNoNotionalOverIndexing() || 1793 !CE2->isGEPWithNoNotionalOverIndexing()) 1794 return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal. 1795 1796 // Compare all of the operands the GEP's have in common. 1797 gep_type_iterator GTI = gep_type_begin(CE1); 1798 for (;i != CE1->getNumOperands() && i != CE2->getNumOperands(); 1799 ++i, ++GTI) 1800 switch (IdxCompare(CE1->getOperand(i), 1801 CE2->getOperand(i), GTI.getIndexedType())) { 1802 case -1: return isSigned ? ICmpInst::ICMP_SLT:ICmpInst::ICMP_ULT; 1803 case 1: return isSigned ? ICmpInst::ICMP_SGT:ICmpInst::ICMP_UGT; 1804 case -2: return ICmpInst::BAD_ICMP_PREDICATE; 1805 } 1806 1807 // Ok, we ran out of things they have in common. If any leftovers 1808 // are non-zero then we have a difference, otherwise we are equal. 1809 for (; i < CE1->getNumOperands(); ++i) 1810 if (!CE1->getOperand(i)->isNullValue()) { 1811 if (isa<ConstantInt>(CE1->getOperand(i))) 1812 return isSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT; 1813 else 1814 return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal. 1815 } 1816 1817 for (; i < CE2->getNumOperands(); ++i) 1818 if (!CE2->getOperand(i)->isNullValue()) { 1819 if (isa<ConstantInt>(CE2->getOperand(i))) 1820 return isSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT; 1821 else 1822 return ICmpInst::BAD_ICMP_PREDICATE; // Might be equal. 1823 } 1824 return ICmpInst::ICMP_EQ; 1825 } 1826 } 1827 } 1828 default: 1829 break; 1830 } 1831 } 1832 1833 return ICmpInst::BAD_ICMP_PREDICATE; 1834 } 1835 1836 Constant *llvm::ConstantFoldCompareInstruction(unsigned short pred, 1837 Constant *C1, Constant *C2) { 1838 Type *ResultTy; 1839 if (VectorType *VT = dyn_cast<VectorType>(C1->getType())) 1840 ResultTy = VectorType::get(Type::getInt1Ty(C1->getContext()), 1841 VT->getNumElements()); 1842 else 1843 ResultTy = Type::getInt1Ty(C1->getContext()); 1844 1845 // Fold FCMP_FALSE/FCMP_TRUE unconditionally. 1846 if (pred == FCmpInst::FCMP_FALSE) 1847 return Constant::getNullValue(ResultTy); 1848 1849 if (pred == FCmpInst::FCMP_TRUE) 1850 return Constant::getAllOnesValue(ResultTy); 1851 1852 // Handle some degenerate cases first 1853 if (isa<UndefValue>(C1) || isa<UndefValue>(C2)) { 1854 // For EQ and NE, we can always pick a value for the undef to make the 1855 // predicate pass or fail, so we can return undef. 1856 // Also, if both operands are undef, we can return undef. 1857 if (ICmpInst::isEquality(ICmpInst::Predicate(pred)) || 1858 (isa<UndefValue>(C1) && isa<UndefValue>(C2))) 1859 return UndefValue::get(ResultTy); 1860 // Otherwise, pick the same value as the non-undef operand, and fold 1861 // it to true or false. 1862 return ConstantInt::get(ResultTy, CmpInst::isTrueWhenEqual(pred)); 1863 } 1864 1865 // No compile-time operations on this type yet. 1866 if (C1->getType()->isPPC_FP128Ty()) 1867 return 0; 1868 1869 // icmp eq/ne(null,GV) -> false/true 1870 if (C1->isNullValue()) { 1871 if (const GlobalValue *GV = dyn_cast<GlobalValue>(C2)) 1872 // Don't try to evaluate aliases. External weak GV can be null. 1873 if (!isa<GlobalAlias>(GV) && !GV->hasExternalWeakLinkage()) { 1874 if (pred == ICmpInst::ICMP_EQ) 1875 return ConstantInt::getFalse(C1->getContext()); 1876 else if (pred == ICmpInst::ICMP_NE) 1877 return ConstantInt::getTrue(C1->getContext()); 1878 } 1879 // icmp eq/ne(GV,null) -> false/true 1880 } else if (C2->isNullValue()) { 1881 if (const GlobalValue *GV = dyn_cast<GlobalValue>(C1)) 1882 // Don't try to evaluate aliases. External weak GV can be null. 1883 if (!isa<GlobalAlias>(GV) && !GV->hasExternalWeakLinkage()) { 1884 if (pred == ICmpInst::ICMP_EQ) 1885 return ConstantInt::getFalse(C1->getContext()); 1886 else if (pred == ICmpInst::ICMP_NE) 1887 return ConstantInt::getTrue(C1->getContext()); 1888 } 1889 } 1890 1891 // If the comparison is a comparison between two i1's, simplify it. 1892 if (C1->getType()->isIntegerTy(1)) { 1893 switch(pred) { 1894 case ICmpInst::ICMP_EQ: 1895 if (isa<ConstantInt>(C2)) 1896 return ConstantExpr::getXor(C1, ConstantExpr::getNot(C2)); 1897 return ConstantExpr::getXor(ConstantExpr::getNot(C1), C2); 1898 case ICmpInst::ICMP_NE: 1899 return ConstantExpr::getXor(C1, C2); 1900 default: 1901 break; 1902 } 1903 } 1904 1905 if (isa<ConstantInt>(C1) && isa<ConstantInt>(C2)) { 1906 APInt V1 = cast<ConstantInt>(C1)->getValue(); 1907 APInt V2 = cast<ConstantInt>(C2)->getValue(); 1908 switch (pred) { 1909 default: llvm_unreachable("Invalid ICmp Predicate"); return 0; 1910 case ICmpInst::ICMP_EQ: return ConstantInt::get(ResultTy, V1 == V2); 1911 case ICmpInst::ICMP_NE: return ConstantInt::get(ResultTy, V1 != V2); 1912 case ICmpInst::ICMP_SLT: return ConstantInt::get(ResultTy, V1.slt(V2)); 1913 case ICmpInst::ICMP_SGT: return ConstantInt::get(ResultTy, V1.sgt(V2)); 1914 case ICmpInst::ICMP_SLE: return ConstantInt::get(ResultTy, V1.sle(V2)); 1915 case ICmpInst::ICMP_SGE: return ConstantInt::get(ResultTy, V1.sge(V2)); 1916 case ICmpInst::ICMP_ULT: return ConstantInt::get(ResultTy, V1.ult(V2)); 1917 case ICmpInst::ICMP_UGT: return ConstantInt::get(ResultTy, V1.ugt(V2)); 1918 case ICmpInst::ICMP_ULE: return ConstantInt::get(ResultTy, V1.ule(V2)); 1919 case ICmpInst::ICMP_UGE: return ConstantInt::get(ResultTy, V1.uge(V2)); 1920 } 1921 } else if (isa<ConstantFP>(C1) && isa<ConstantFP>(C2)) { 1922 APFloat C1V = cast<ConstantFP>(C1)->getValueAPF(); 1923 APFloat C2V = cast<ConstantFP>(C2)->getValueAPF(); 1924 APFloat::cmpResult R = C1V.compare(C2V); 1925 switch (pred) { 1926 default: llvm_unreachable("Invalid FCmp Predicate"); return 0; 1927 case FCmpInst::FCMP_FALSE: return Constant::getNullValue(ResultTy); 1928 case FCmpInst::FCMP_TRUE: return Constant::getAllOnesValue(ResultTy); 1929 case FCmpInst::FCMP_UNO: 1930 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered); 1931 case FCmpInst::FCMP_ORD: 1932 return ConstantInt::get(ResultTy, R!=APFloat::cmpUnordered); 1933 case FCmpInst::FCMP_UEQ: 1934 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered || 1935 R==APFloat::cmpEqual); 1936 case FCmpInst::FCMP_OEQ: 1937 return ConstantInt::get(ResultTy, R==APFloat::cmpEqual); 1938 case FCmpInst::FCMP_UNE: 1939 return ConstantInt::get(ResultTy, R!=APFloat::cmpEqual); 1940 case FCmpInst::FCMP_ONE: 1941 return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan || 1942 R==APFloat::cmpGreaterThan); 1943 case FCmpInst::FCMP_ULT: 1944 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered || 1945 R==APFloat::cmpLessThan); 1946 case FCmpInst::FCMP_OLT: 1947 return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan); 1948 case FCmpInst::FCMP_UGT: 1949 return ConstantInt::get(ResultTy, R==APFloat::cmpUnordered || 1950 R==APFloat::cmpGreaterThan); 1951 case FCmpInst::FCMP_OGT: 1952 return ConstantInt::get(ResultTy, R==APFloat::cmpGreaterThan); 1953 case FCmpInst::FCMP_ULE: 1954 return ConstantInt::get(ResultTy, R!=APFloat::cmpGreaterThan); 1955 case FCmpInst::FCMP_OLE: 1956 return ConstantInt::get(ResultTy, R==APFloat::cmpLessThan || 1957 R==APFloat::cmpEqual); 1958 case FCmpInst::FCMP_UGE: 1959 return ConstantInt::get(ResultTy, R!=APFloat::cmpLessThan); 1960 case FCmpInst::FCMP_OGE: 1961 return ConstantInt::get(ResultTy, R==APFloat::cmpGreaterThan || 1962 R==APFloat::cmpEqual); 1963 } 1964 } else if (C1->getType()->isVectorTy()) { 1965 SmallVector<Constant*, 16> C1Elts, C2Elts; 1966 C1->getVectorElements(C1Elts); 1967 C2->getVectorElements(C2Elts); 1968 if (C1Elts.empty() || C2Elts.empty()) 1969 return 0; 1970 1971 // If we can constant fold the comparison of each element, constant fold 1972 // the whole vector comparison. 1973 SmallVector<Constant*, 4> ResElts; 1974 // Compare the elements, producing an i1 result or constant expr. 1975 for (unsigned i = 0, e = C1Elts.size(); i != e; ++i) 1976 ResElts.push_back(ConstantExpr::getCompare(pred, C1Elts[i], C2Elts[i])); 1977 1978 return ConstantVector::get(ResElts); 1979 } 1980 1981 if (C1->getType()->isFloatingPointTy()) { 1982 int Result = -1; // -1 = unknown, 0 = known false, 1 = known true. 1983 switch (evaluateFCmpRelation(C1, C2)) { 1984 default: llvm_unreachable("Unknown relation!"); 1985 case FCmpInst::FCMP_UNO: 1986 case FCmpInst::FCMP_ORD: 1987 case FCmpInst::FCMP_UEQ: 1988 case FCmpInst::FCMP_UNE: 1989 case FCmpInst::FCMP_ULT: 1990 case FCmpInst::FCMP_UGT: 1991 case FCmpInst::FCMP_ULE: 1992 case FCmpInst::FCMP_UGE: 1993 case FCmpInst::FCMP_TRUE: 1994 case FCmpInst::FCMP_FALSE: 1995 case FCmpInst::BAD_FCMP_PREDICATE: 1996 break; // Couldn't determine anything about these constants. 1997 case FCmpInst::FCMP_OEQ: // We know that C1 == C2 1998 Result = (pred == FCmpInst::FCMP_UEQ || pred == FCmpInst::FCMP_OEQ || 1999 pred == FCmpInst::FCMP_ULE || pred == FCmpInst::FCMP_OLE || 2000 pred == FCmpInst::FCMP_UGE || pred == FCmpInst::FCMP_OGE); 2001 break; 2002 case FCmpInst::FCMP_OLT: // We know that C1 < C2 2003 Result = (pred == FCmpInst::FCMP_UNE || pred == FCmpInst::FCMP_ONE || 2004 pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT || 2005 pred == FCmpInst::FCMP_ULE || pred == FCmpInst::FCMP_OLE); 2006 break; 2007 case FCmpInst::FCMP_OGT: // We know that C1 > C2 2008 Result = (pred == FCmpInst::FCMP_UNE || pred == FCmpInst::FCMP_ONE || 2009 pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT || 2010 pred == FCmpInst::FCMP_UGE || pred == FCmpInst::FCMP_OGE); 2011 break; 2012 case FCmpInst::FCMP_OLE: // We know that C1 <= C2 2013 // We can only partially decide this relation. 2014 if (pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT) 2015 Result = 0; 2016 else if (pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT) 2017 Result = 1; 2018 break; 2019 case FCmpInst::FCMP_OGE: // We known that C1 >= C2 2020 // We can only partially decide this relation. 2021 if (pred == FCmpInst::FCMP_ULT || pred == FCmpInst::FCMP_OLT) 2022 Result = 0; 2023 else if (pred == FCmpInst::FCMP_UGT || pred == FCmpInst::FCMP_OGT) 2024 Result = 1; 2025 break; 2026 case FCmpInst::FCMP_ONE: // We know that C1 != C2 2027 // We can only partially decide this relation. 2028 if (pred == FCmpInst::FCMP_OEQ || pred == FCmpInst::FCMP_UEQ) 2029 Result = 0; 2030 else if (pred == FCmpInst::FCMP_ONE || pred == FCmpInst::FCMP_UNE) 2031 Result = 1; 2032 break; 2033 } 2034 2035 // If we evaluated the result, return it now. 2036 if (Result != -1) 2037 return ConstantInt::get(ResultTy, Result); 2038 2039 } else { 2040 // Evaluate the relation between the two constants, per the predicate. 2041 int Result = -1; // -1 = unknown, 0 = known false, 1 = known true. 2042 switch (evaluateICmpRelation(C1, C2, CmpInst::isSigned(pred))) { 2043 default: llvm_unreachable("Unknown relational!"); 2044 case ICmpInst::BAD_ICMP_PREDICATE: 2045 break; // Couldn't determine anything about these constants. 2046 case ICmpInst::ICMP_EQ: // We know the constants are equal! 2047 // If we know the constants are equal, we can decide the result of this 2048 // computation precisely. 2049 Result = ICmpInst::isTrueWhenEqual((ICmpInst::Predicate)pred); 2050 break; 2051 case ICmpInst::ICMP_ULT: 2052 switch (pred) { 2053 case ICmpInst::ICMP_ULT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_ULE: 2054 Result = 1; break; 2055 case ICmpInst::ICMP_UGT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_UGE: 2056 Result = 0; break; 2057 } 2058 break; 2059 case ICmpInst::ICMP_SLT: 2060 switch (pred) { 2061 case ICmpInst::ICMP_SLT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_SLE: 2062 Result = 1; break; 2063 case ICmpInst::ICMP_SGT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_SGE: 2064 Result = 0; break; 2065 } 2066 break; 2067 case ICmpInst::ICMP_UGT: 2068 switch (pred) { 2069 case ICmpInst::ICMP_UGT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_UGE: 2070 Result = 1; break; 2071 case ICmpInst::ICMP_ULT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_ULE: 2072 Result = 0; break; 2073 } 2074 break; 2075 case ICmpInst::ICMP_SGT: 2076 switch (pred) { 2077 case ICmpInst::ICMP_SGT: case ICmpInst::ICMP_NE: case ICmpInst::ICMP_SGE: 2078 Result = 1; break; 2079 case ICmpInst::ICMP_SLT: case ICmpInst::ICMP_EQ: case ICmpInst::ICMP_SLE: 2080 Result = 0; break; 2081 } 2082 break; 2083 case ICmpInst::ICMP_ULE: 2084 if (pred == ICmpInst::ICMP_UGT) Result = 0; 2085 if (pred == ICmpInst::ICMP_ULT || pred == ICmpInst::ICMP_ULE) Result = 1; 2086 break; 2087 case ICmpInst::ICMP_SLE: 2088 if (pred == ICmpInst::ICMP_SGT) Result = 0; 2089 if (pred == ICmpInst::ICMP_SLT || pred == ICmpInst::ICMP_SLE) Result = 1; 2090 break; 2091 case ICmpInst::ICMP_UGE: 2092 if (pred == ICmpInst::ICMP_ULT) Result = 0; 2093 if (pred == ICmpInst::ICMP_UGT || pred == ICmpInst::ICMP_UGE) Result = 1; 2094 break; 2095 case ICmpInst::ICMP_SGE: 2096 if (pred == ICmpInst::ICMP_SLT) Result = 0; 2097 if (pred == ICmpInst::ICMP_SGT || pred == ICmpInst::ICMP_SGE) Result = 1; 2098 break; 2099 case ICmpInst::ICMP_NE: 2100 if (pred == ICmpInst::ICMP_EQ) Result = 0; 2101 if (pred == ICmpInst::ICMP_NE) Result = 1; 2102 break; 2103 } 2104 2105 // If we evaluated the result, return it now. 2106 if (Result != -1) 2107 return ConstantInt::get(ResultTy, Result); 2108 2109 // If the right hand side is a bitcast, try using its inverse to simplify 2110 // it by moving it to the left hand side. We can't do this if it would turn 2111 // a vector compare into a scalar compare or visa versa. 2112 if (ConstantExpr *CE2 = dyn_cast<ConstantExpr>(C2)) { 2113 Constant *CE2Op0 = CE2->getOperand(0); 2114 if (CE2->getOpcode() == Instruction::BitCast && 2115 CE2->getType()->isVectorTy() == CE2Op0->getType()->isVectorTy()) { 2116 Constant *Inverse = ConstantExpr::getBitCast(C1, CE2Op0->getType()); 2117 return ConstantExpr::getICmp(pred, Inverse, CE2Op0); 2118 } 2119 } 2120 2121 // If the left hand side is an extension, try eliminating it. 2122 if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { 2123 if ((CE1->getOpcode() == Instruction::SExt && ICmpInst::isSigned(pred)) || 2124 (CE1->getOpcode() == Instruction::ZExt && !ICmpInst::isSigned(pred))){ 2125 Constant *CE1Op0 = CE1->getOperand(0); 2126 Constant *CE1Inverse = ConstantExpr::getTrunc(CE1, CE1Op0->getType()); 2127 if (CE1Inverse == CE1Op0) { 2128 // Check whether we can safely truncate the right hand side. 2129 Constant *C2Inverse = ConstantExpr::getTrunc(C2, CE1Op0->getType()); 2130 if (ConstantExpr::getZExt(C2Inverse, C2->getType()) == C2) { 2131 return ConstantExpr::getICmp(pred, CE1Inverse, C2Inverse); 2132 } 2133 } 2134 } 2135 } 2136 2137 if ((!isa<ConstantExpr>(C1) && isa<ConstantExpr>(C2)) || 2138 (C1->isNullValue() && !C2->isNullValue())) { 2139 // If C2 is a constant expr and C1 isn't, flip them around and fold the 2140 // other way if possible. 2141 // Also, if C1 is null and C2 isn't, flip them around. 2142 pred = ICmpInst::getSwappedPredicate((ICmpInst::Predicate)pred); 2143 return ConstantExpr::getICmp(pred, C2, C1); 2144 } 2145 } 2146 return 0; 2147 } 2148 2149 /// isInBoundsIndices - Test whether the given sequence of *normalized* indices 2150 /// is "inbounds". 2151 template<typename IndexTy> 2152 static bool isInBoundsIndices(ArrayRef<IndexTy> Idxs) { 2153 // No indices means nothing that could be out of bounds. 2154 if (Idxs.empty()) return true; 2155 2156 // If the first index is zero, it's in bounds. 2157 if (cast<Constant>(Idxs[0])->isNullValue()) return true; 2158 2159 // If the first index is one and all the rest are zero, it's in bounds, 2160 // by the one-past-the-end rule. 2161 if (!cast<ConstantInt>(Idxs[0])->isOne()) 2162 return false; 2163 for (unsigned i = 1, e = Idxs.size(); i != e; ++i) 2164 if (!cast<Constant>(Idxs[i])->isNullValue()) 2165 return false; 2166 return true; 2167 } 2168 2169 template<typename IndexTy> 2170 static Constant *ConstantFoldGetElementPtrImpl(Constant *C, 2171 bool inBounds, 2172 ArrayRef<IndexTy> Idxs) { 2173 if (Idxs.empty()) return C; 2174 Constant *Idx0 = cast<Constant>(Idxs[0]); 2175 if ((Idxs.size() == 1 && Idx0->isNullValue())) 2176 return C; 2177 2178 if (isa<UndefValue>(C)) { 2179 PointerType *Ptr = cast<PointerType>(C->getType()); 2180 Type *Ty = GetElementPtrInst::getIndexedType(Ptr, Idxs); 2181 assert(Ty != 0 && "Invalid indices for GEP!"); 2182 return UndefValue::get(PointerType::get(Ty, Ptr->getAddressSpace())); 2183 } 2184 2185 if (C->isNullValue()) { 2186 bool isNull = true; 2187 for (unsigned i = 0, e = Idxs.size(); i != e; ++i) 2188 if (!cast<Constant>(Idxs[i])->isNullValue()) { 2189 isNull = false; 2190 break; 2191 } 2192 if (isNull) { 2193 PointerType *Ptr = cast<PointerType>(C->getType()); 2194 Type *Ty = GetElementPtrInst::getIndexedType(Ptr, Idxs); 2195 assert(Ty != 0 && "Invalid indices for GEP!"); 2196 return ConstantPointerNull::get(PointerType::get(Ty, 2197 Ptr->getAddressSpace())); 2198 } 2199 } 2200 2201 if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) { 2202 // Combine Indices - If the source pointer to this getelementptr instruction 2203 // is a getelementptr instruction, combine the indices of the two 2204 // getelementptr instructions into a single instruction. 2205 // 2206 if (CE->getOpcode() == Instruction::GetElementPtr) { 2207 Type *LastTy = 0; 2208 for (gep_type_iterator I = gep_type_begin(CE), E = gep_type_end(CE); 2209 I != E; ++I) 2210 LastTy = *I; 2211 2212 if ((LastTy && LastTy->isArrayTy()) || Idx0->isNullValue()) { 2213 SmallVector<Value*, 16> NewIndices; 2214 NewIndices.reserve(Idxs.size() + CE->getNumOperands()); 2215 for (unsigned i = 1, e = CE->getNumOperands()-1; i != e; ++i) 2216 NewIndices.push_back(CE->getOperand(i)); 2217 2218 // Add the last index of the source with the first index of the new GEP. 2219 // Make sure to handle the case when they are actually different types. 2220 Constant *Combined = CE->getOperand(CE->getNumOperands()-1); 2221 // Otherwise it must be an array. 2222 if (!Idx0->isNullValue()) { 2223 Type *IdxTy = Combined->getType(); 2224 if (IdxTy != Idx0->getType()) { 2225 Type *Int64Ty = Type::getInt64Ty(IdxTy->getContext()); 2226 Constant *C1 = ConstantExpr::getSExtOrBitCast(Idx0, Int64Ty); 2227 Constant *C2 = ConstantExpr::getSExtOrBitCast(Combined, Int64Ty); 2228 Combined = ConstantExpr::get(Instruction::Add, C1, C2); 2229 } else { 2230 Combined = 2231 ConstantExpr::get(Instruction::Add, Idx0, Combined); 2232 } 2233 } 2234 2235 NewIndices.push_back(Combined); 2236 NewIndices.append(Idxs.begin() + 1, Idxs.end()); 2237 return 2238 ConstantExpr::getGetElementPtr(CE->getOperand(0), NewIndices, 2239 inBounds && 2240 cast<GEPOperator>(CE)->isInBounds()); 2241 } 2242 } 2243 2244 // Implement folding of: 2245 // i32* getelementptr ([2 x i32]* bitcast ([3 x i32]* %X to [2 x i32]*), 2246 // i64 0, i64 0) 2247 // To: i32* getelementptr ([3 x i32]* %X, i64 0, i64 0) 2248 // 2249 if (CE->isCast() && Idxs.size() > 1 && Idx0->isNullValue()) { 2250 if (PointerType *SPT = 2251 dyn_cast<PointerType>(CE->getOperand(0)->getType())) 2252 if (ArrayType *SAT = dyn_cast<ArrayType>(SPT->getElementType())) 2253 if (ArrayType *CAT = 2254 dyn_cast<ArrayType>(cast<PointerType>(C->getType())->getElementType())) 2255 if (CAT->getElementType() == SAT->getElementType()) 2256 return 2257 ConstantExpr::getGetElementPtr((Constant*)CE->getOperand(0), 2258 Idxs, inBounds); 2259 } 2260 } 2261 2262 // Check to see if any array indices are not within the corresponding 2263 // notional array bounds. If so, try to determine if they can be factored 2264 // out into preceding dimensions. 2265 bool Unknown = false; 2266 SmallVector<Constant *, 8> NewIdxs; 2267 Type *Ty = C->getType(); 2268 Type *Prev = 0; 2269 for (unsigned i = 0, e = Idxs.size(); i != e; 2270 Prev = Ty, Ty = cast<CompositeType>(Ty)->getTypeAtIndex(Idxs[i]), ++i) { 2271 if (ConstantInt *CI = dyn_cast<ConstantInt>(Idxs[i])) { 2272 if (ArrayType *ATy = dyn_cast<ArrayType>(Ty)) 2273 if (ATy->getNumElements() <= INT64_MAX && 2274 ATy->getNumElements() != 0 && 2275 CI->getSExtValue() >= (int64_t)ATy->getNumElements()) { 2276 if (isa<SequentialType>(Prev)) { 2277 // It's out of range, but we can factor it into the prior 2278 // dimension. 2279 NewIdxs.resize(Idxs.size()); 2280 ConstantInt *Factor = ConstantInt::get(CI->getType(), 2281 ATy->getNumElements()); 2282 NewIdxs[i] = ConstantExpr::getSRem(CI, Factor); 2283 2284 Constant *PrevIdx = cast<Constant>(Idxs[i-1]); 2285 Constant *Div = ConstantExpr::getSDiv(CI, Factor); 2286 2287 // Before adding, extend both operands to i64 to avoid 2288 // overflow trouble. 2289 if (!PrevIdx->getType()->isIntegerTy(64)) 2290 PrevIdx = ConstantExpr::getSExt(PrevIdx, 2291 Type::getInt64Ty(Div->getContext())); 2292 if (!Div->getType()->isIntegerTy(64)) 2293 Div = ConstantExpr::getSExt(Div, 2294 Type::getInt64Ty(Div->getContext())); 2295 2296 NewIdxs[i-1] = ConstantExpr::getAdd(PrevIdx, Div); 2297 } else { 2298 // It's out of range, but the prior dimension is a struct 2299 // so we can't do anything about it. 2300 Unknown = true; 2301 } 2302 } 2303 } else { 2304 // We don't know if it's in range or not. 2305 Unknown = true; 2306 } 2307 } 2308 2309 // If we did any factoring, start over with the adjusted indices. 2310 if (!NewIdxs.empty()) { 2311 for (unsigned i = 0, e = Idxs.size(); i != e; ++i) 2312 if (!NewIdxs[i]) NewIdxs[i] = cast<Constant>(Idxs[i]); 2313 return ConstantExpr::getGetElementPtr(C, NewIdxs, inBounds); 2314 } 2315 2316 // If all indices are known integers and normalized, we can do a simple 2317 // check for the "inbounds" property. 2318 if (!Unknown && !inBounds && 2319 isa<GlobalVariable>(C) && isInBoundsIndices(Idxs)) 2320 return ConstantExpr::getInBoundsGetElementPtr(C, Idxs); 2321 2322 return 0; 2323 } 2324 2325 Constant *llvm::ConstantFoldGetElementPtr(Constant *C, 2326 bool inBounds, 2327 ArrayRef<Constant *> Idxs) { 2328 return ConstantFoldGetElementPtrImpl(C, inBounds, Idxs); 2329 } 2330 2331 Constant *llvm::ConstantFoldGetElementPtr(Constant *C, 2332 bool inBounds, 2333 ArrayRef<Value *> Idxs) { 2334 return ConstantFoldGetElementPtrImpl(C, inBounds, Idxs); 2335 } 2336