ConstantFolding.cpp revision 62d327e241f59735eed836e98c8f11ffaa6cf6f8
1//===-- ConstantFolding.cpp - Fold instructions into constants ------------===//
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 defines routines for folding instructions into constants.
11//
12// Also, to supplement the basic VMCore ConstantExpr simplifications,
13// this file defines some additional folding routines that can make use of
14// TargetData information. These functions cannot go in VMCore due to library
15// dependency issues.
16//
17//===----------------------------------------------------------------------===//
18
19#include "llvm/Analysis/ConstantFolding.h"
20#include "llvm/Constants.h"
21#include "llvm/DerivedTypes.h"
22#include "llvm/Function.h"
23#include "llvm/GlobalVariable.h"
24#include "llvm/Instructions.h"
25#include "llvm/Intrinsics.h"
26#include "llvm/LLVMContext.h"
27#include "llvm/Analysis/ValueTracking.h"
28#include "llvm/Target/TargetData.h"
29#include "llvm/ADT/SmallVector.h"
30#include "llvm/ADT/StringMap.h"
31#include "llvm/Support/ErrorHandling.h"
32#include "llvm/Support/GetElementPtrTypeIterator.h"
33#include "llvm/Support/MathExtras.h"
34#include <cerrno>
35#include <cmath>
36using namespace llvm;
37
38//===----------------------------------------------------------------------===//
39// Constant Folding internal helper functions
40//===----------------------------------------------------------------------===//
41
42/// IsConstantOffsetFromGlobal - If this constant is actually a constant offset
43/// from a global, return the global and the constant.  Because of
44/// constantexprs, this function is recursive.
45static bool IsConstantOffsetFromGlobal(Constant *C, GlobalValue *&GV,
46                                       int64_t &Offset, const TargetData &TD) {
47  // Trivial case, constant is the global.
48  if ((GV = dyn_cast<GlobalValue>(C))) {
49    Offset = 0;
50    return true;
51  }
52
53  // Otherwise, if this isn't a constant expr, bail out.
54  ConstantExpr *CE = dyn_cast<ConstantExpr>(C);
55  if (!CE) return false;
56
57  // Look through ptr->int and ptr->ptr casts.
58  if (CE->getOpcode() == Instruction::PtrToInt ||
59      CE->getOpcode() == Instruction::BitCast)
60    return IsConstantOffsetFromGlobal(CE->getOperand(0), GV, Offset, TD);
61
62  // i32* getelementptr ([5 x i32]* @a, i32 0, i32 5)
63  if (CE->getOpcode() == Instruction::GetElementPtr) {
64    // Cannot compute this if the element type of the pointer is missing size
65    // info.
66    if (!cast<PointerType>(CE->getOperand(0)->getType())
67                 ->getElementType()->isSized())
68      return false;
69
70    // If the base isn't a global+constant, we aren't either.
71    if (!IsConstantOffsetFromGlobal(CE->getOperand(0), GV, Offset, TD))
72      return false;
73
74    // Otherwise, add any offset that our operands provide.
75    gep_type_iterator GTI = gep_type_begin(CE);
76    for (User::const_op_iterator i = CE->op_begin() + 1, e = CE->op_end();
77         i != e; ++i, ++GTI) {
78      ConstantInt *CI = dyn_cast<ConstantInt>(*i);
79      if (!CI) return false;  // Index isn't a simple constant?
80      if (CI->getZExtValue() == 0) continue;  // Not adding anything.
81
82      if (const StructType *ST = dyn_cast<StructType>(*GTI)) {
83        // N = N + Offset
84        Offset += TD.getStructLayout(ST)->getElementOffset(CI->getZExtValue());
85      } else {
86        const SequentialType *SQT = cast<SequentialType>(*GTI);
87        Offset += TD.getTypeAllocSize(SQT->getElementType())*CI->getSExtValue();
88      }
89    }
90    return true;
91  }
92
93  return false;
94}
95
96/// ConstantFoldLoadFromConstPtr - Return the value that a load from C would
97/// produce if it is constant and determinable.  If this is not determinable,
98/// return null.
99Constant *llvm::ConstantFoldLoadFromConstPtr(Constant *C,
100                                             const TargetData *TD) {
101  // First, try the easy cases:
102  if (GlobalVariable *GV = dyn_cast<GlobalVariable>(C))
103    if (GV->isConstant() && GV->hasDefinitiveInitializer())
104      return GV->getInitializer();
105
106  if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
107    if (CE->getOpcode() == Instruction::GetElementPtr) {
108      if (GlobalVariable *GV = dyn_cast<GlobalVariable>(CE->getOperand(0)))
109        if (GV->isConstant() && GV->hasDefinitiveInitializer())
110          if (Constant *V =
111               ConstantFoldLoadThroughGEPConstantExpr(GV->getInitializer(), CE))
112            return V;
113    }
114
115    // Instead of loading constant c string, use corresponding integer value
116    // directly if string length is small enough.
117    std::string Str;
118    if (TD && GetConstantStringInfo(CE->getOperand(0), Str) && !Str.empty()) {
119      unsigned len = Str.length();
120      const Type *Ty = cast<PointerType>(CE->getType())->getElementType();
121      unsigned numBits = Ty->getPrimitiveSizeInBits();
122      // Replace LI with immediate integer store.
123      if ((numBits >> 3) == len + 1) {
124        APInt StrVal(numBits, 0);
125        APInt SingleChar(numBits, 0);
126        if (TD->isLittleEndian()) {
127          for (signed i = len-1; i >= 0; i--) {
128            SingleChar = (uint64_t) Str[i] & UCHAR_MAX;
129            StrVal = (StrVal << 8) | SingleChar;
130          }
131        } else {
132          for (unsigned i = 0; i < len; i++) {
133            SingleChar = (uint64_t) Str[i] & UCHAR_MAX;
134            StrVal = (StrVal << 8) | SingleChar;
135          }
136          // Append NULL at the end.
137          SingleChar = 0;
138          StrVal = (StrVal << 8) | SingleChar;
139        }
140        return ConstantInt::get(CE->getContext(), StrVal);
141      }
142    }
143  }
144
145  return 0;
146}
147
148static Constant *ConstantFoldLoadInst(const LoadInst *LI, const TargetData *TD){
149  if (LI->isVolatile()) return 0;
150
151  if (Constant *C = dyn_cast<Constant>(LI->getOperand(0)))
152    return ConstantFoldLoadFromConstPtr(C, TD);
153
154  return 0;
155}
156
157/// SymbolicallyEvaluateBinop - One of Op0/Op1 is a constant expression.
158/// Attempt to symbolically evaluate the result of a binary operator merging
159/// these together.  If target data info is available, it is provided as TD,
160/// otherwise TD is null.
161static Constant *SymbolicallyEvaluateBinop(unsigned Opc, Constant *Op0,
162                                           Constant *Op1, const TargetData *TD,
163                                           LLVMContext &Context){
164  // SROA
165
166  // Fold (and 0xffffffff00000000, (shl x, 32)) -> shl.
167  // Fold (lshr (or X, Y), 32) -> (lshr [X/Y], 32) if one doesn't contribute
168  // bits.
169
170
171  // If the constant expr is something like &A[123] - &A[4].f, fold this into a
172  // constant.  This happens frequently when iterating over a global array.
173  if (Opc == Instruction::Sub && TD) {
174    GlobalValue *GV1, *GV2;
175    int64_t Offs1, Offs2;
176
177    if (IsConstantOffsetFromGlobal(Op0, GV1, Offs1, *TD))
178      if (IsConstantOffsetFromGlobal(Op1, GV2, Offs2, *TD) &&
179          GV1 == GV2) {
180        // (&GV+C1) - (&GV+C2) -> C1-C2, pointer arithmetic cannot overflow.
181        return ConstantInt::get(Op0->getType(), Offs1-Offs2);
182      }
183  }
184
185  return 0;
186}
187
188/// SymbolicallyEvaluateGEP - If we can symbolically evaluate the specified GEP
189/// constant expression, do so.
190static Constant *SymbolicallyEvaluateGEP(Constant* const* Ops, unsigned NumOps,
191                                         const Type *ResultTy,
192                                         LLVMContext &Context,
193                                         const TargetData *TD) {
194  Constant *Ptr = Ops[0];
195  if (!TD || !cast<PointerType>(Ptr->getType())->getElementType()->isSized())
196    return 0;
197
198  unsigned BitWidth = TD->getTypeSizeInBits(TD->getIntPtrType(Context));
199  APInt BasePtr(BitWidth, 0);
200  bool BaseIsInt = true;
201  if (!Ptr->isNullValue()) {
202    // If this is a inttoptr from a constant int, we can fold this as the base,
203    // otherwise we can't.
204    if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ptr))
205      if (CE->getOpcode() == Instruction::IntToPtr)
206        if (ConstantInt *Base = dyn_cast<ConstantInt>(CE->getOperand(0))) {
207          BasePtr = Base->getValue();
208          BasePtr.zextOrTrunc(BitWidth);
209        }
210
211    if (BasePtr == 0)
212      BaseIsInt = false;
213  }
214
215  // If this is a constant expr gep that is effectively computing an
216  // "offsetof", fold it into 'cast int Size to T*' instead of 'gep 0, 0, 12'
217  for (unsigned i = 1; i != NumOps; ++i)
218    if (!isa<ConstantInt>(Ops[i]))
219      return 0;
220
221  APInt Offset = APInt(BitWidth,
222                       TD->getIndexedOffset(Ptr->getType(),
223                                            (Value**)Ops+1, NumOps-1));
224  // If the base value for this address is a literal integer value, fold the
225  // getelementptr to the resulting integer value casted to the pointer type.
226  if (BaseIsInt) {
227    Constant *C = ConstantInt::get(Context, Offset+BasePtr);
228    return ConstantExpr::getIntToPtr(C, ResultTy);
229  }
230
231  // Otherwise form a regular getelementptr. Recompute the indices so that
232  // we eliminate over-indexing of the notional static type array bounds.
233  // This makes it easy to determine if the getelementptr is "inbounds".
234  // Also, this helps GlobalOpt do SROA on GlobalVariables.
235  const Type *Ty = Ptr->getType();
236  SmallVector<Constant*, 32> NewIdxs;
237  do {
238    if (const SequentialType *ATy = dyn_cast<SequentialType>(Ty)) {
239      // The only pointer indexing we'll do is on the first index of the GEP.
240      if (isa<PointerType>(ATy) && !NewIdxs.empty())
241        break;
242      // Determine which element of the array the offset points into.
243      APInt ElemSize(BitWidth, TD->getTypeAllocSize(ATy->getElementType()));
244      if (ElemSize == 0)
245        return 0;
246      APInt NewIdx = Offset.udiv(ElemSize);
247      Offset -= NewIdx * ElemSize;
248      NewIdxs.push_back(ConstantInt::get(TD->getIntPtrType(Context), NewIdx));
249      Ty = ATy->getElementType();
250    } else if (const StructType *STy = dyn_cast<StructType>(Ty)) {
251      // Determine which field of the struct the offset points into. The
252      // getZExtValue is at least as safe as the StructLayout API because we
253      // know the offset is within the struct at this point.
254      const StructLayout &SL = *TD->getStructLayout(STy);
255      unsigned ElIdx = SL.getElementContainingOffset(Offset.getZExtValue());
256      NewIdxs.push_back(ConstantInt::get(Type::getInt32Ty(Context), ElIdx));
257      Offset -= APInt(BitWidth, SL.getElementOffset(ElIdx));
258      Ty = STy->getTypeAtIndex(ElIdx);
259    } else {
260      // We've reached some non-indexable type.
261      break;
262    }
263  } while (Ty != cast<PointerType>(ResultTy)->getElementType());
264
265  // If we haven't used up the entire offset by descending the static
266  // type, then the offset is pointing into the middle of an indivisible
267  // member, so we can't simplify it.
268  if (Offset != 0)
269    return 0;
270
271  // Create a GEP.
272  Constant *C =
273    ConstantExpr::getGetElementPtr(Ptr, &NewIdxs[0], NewIdxs.size());
274  assert(cast<PointerType>(C->getType())->getElementType() == Ty &&
275         "Computed GetElementPtr has unexpected type!");
276
277  // If we ended up indexing a member with a type that doesn't match
278  // the type of what the original indices indexed, add a cast.
279  if (Ty != cast<PointerType>(ResultTy)->getElementType())
280    C = ConstantExpr::getBitCast(C, ResultTy);
281
282  return C;
283}
284
285/// FoldBitCast - Constant fold bitcast, symbolically evaluating it with
286/// targetdata.  Return 0 if unfoldable.
287static Constant *FoldBitCast(Constant *C, const Type *DestTy,
288                             const TargetData &TD, LLVMContext &Context) {
289  // If this is a bitcast from constant vector -> vector, fold it.
290  if (ConstantVector *CV = dyn_cast<ConstantVector>(C)) {
291    if (const VectorType *DestVTy = dyn_cast<VectorType>(DestTy)) {
292      // If the element types match, VMCore can fold it.
293      unsigned NumDstElt = DestVTy->getNumElements();
294      unsigned NumSrcElt = CV->getNumOperands();
295      if (NumDstElt == NumSrcElt)
296        return 0;
297
298      const Type *SrcEltTy = CV->getType()->getElementType();
299      const Type *DstEltTy = DestVTy->getElementType();
300
301      // Otherwise, we're changing the number of elements in a vector, which
302      // requires endianness information to do the right thing.  For example,
303      //    bitcast (<2 x i64> <i64 0, i64 1> to <4 x i32>)
304      // folds to (little endian):
305      //    <4 x i32> <i32 0, i32 0, i32 1, i32 0>
306      // and to (big endian):
307      //    <4 x i32> <i32 0, i32 0, i32 0, i32 1>
308
309      // First thing is first.  We only want to think about integer here, so if
310      // we have something in FP form, recast it as integer.
311      if (DstEltTy->isFloatingPoint()) {
312        // Fold to an vector of integers with same size as our FP type.
313        unsigned FPWidth = DstEltTy->getPrimitiveSizeInBits();
314        const Type *DestIVTy = VectorType::get(
315                                 IntegerType::get(Context, FPWidth), NumDstElt);
316        // Recursively handle this integer conversion, if possible.
317        C = FoldBitCast(C, DestIVTy, TD, Context);
318        if (!C) return 0;
319
320        // Finally, VMCore can handle this now that #elts line up.
321        return ConstantExpr::getBitCast(C, DestTy);
322      }
323
324      // Okay, we know the destination is integer, if the input is FP, convert
325      // it to integer first.
326      if (SrcEltTy->isFloatingPoint()) {
327        unsigned FPWidth = SrcEltTy->getPrimitiveSizeInBits();
328        const Type *SrcIVTy = VectorType::get(
329                                 IntegerType::get(Context, FPWidth), NumSrcElt);
330        // Ask VMCore to do the conversion now that #elts line up.
331        C = ConstantExpr::getBitCast(C, SrcIVTy);
332        CV = dyn_cast<ConstantVector>(C);
333        if (!CV) return 0;  // If VMCore wasn't able to fold it, bail out.
334      }
335
336      // Now we know that the input and output vectors are both integer vectors
337      // of the same size, and that their #elements is not the same.  Do the
338      // conversion here, which depends on whether the input or output has
339      // more elements.
340      bool isLittleEndian = TD.isLittleEndian();
341
342      SmallVector<Constant*, 32> Result;
343      if (NumDstElt < NumSrcElt) {
344        // Handle: bitcast (<4 x i32> <i32 0, i32 1, i32 2, i32 3> to <2 x i64>)
345        Constant *Zero = Constant::getNullValue(DstEltTy);
346        unsigned Ratio = NumSrcElt/NumDstElt;
347        unsigned SrcBitSize = SrcEltTy->getPrimitiveSizeInBits();
348        unsigned SrcElt = 0;
349        for (unsigned i = 0; i != NumDstElt; ++i) {
350          // Build each element of the result.
351          Constant *Elt = Zero;
352          unsigned ShiftAmt = isLittleEndian ? 0 : SrcBitSize*(Ratio-1);
353          for (unsigned j = 0; j != Ratio; ++j) {
354            Constant *Src = dyn_cast<ConstantInt>(CV->getOperand(SrcElt++));
355            if (!Src) return 0;  // Reject constantexpr elements.
356
357            // Zero extend the element to the right size.
358            Src = ConstantExpr::getZExt(Src, Elt->getType());
359
360            // Shift it to the right place, depending on endianness.
361            Src = ConstantExpr::getShl(Src,
362                             ConstantInt::get(Src->getType(), ShiftAmt));
363            ShiftAmt += isLittleEndian ? SrcBitSize : -SrcBitSize;
364
365            // Mix it in.
366            Elt = ConstantExpr::getOr(Elt, Src);
367          }
368          Result.push_back(Elt);
369        }
370      } else {
371        // Handle: bitcast (<2 x i64> <i64 0, i64 1> to <4 x i32>)
372        unsigned Ratio = NumDstElt/NumSrcElt;
373        unsigned DstBitSize = DstEltTy->getPrimitiveSizeInBits();
374
375        // Loop over each source value, expanding into multiple results.
376        for (unsigned i = 0; i != NumSrcElt; ++i) {
377          Constant *Src = dyn_cast<ConstantInt>(CV->getOperand(i));
378          if (!Src) return 0;  // Reject constantexpr elements.
379
380          unsigned ShiftAmt = isLittleEndian ? 0 : DstBitSize*(Ratio-1);
381          for (unsigned j = 0; j != Ratio; ++j) {
382            // Shift the piece of the value into the right place, depending on
383            // endianness.
384            Constant *Elt = ConstantExpr::getLShr(Src,
385                            ConstantInt::get(Src->getType(), ShiftAmt));
386            ShiftAmt += isLittleEndian ? DstBitSize : -DstBitSize;
387
388            // Truncate and remember this piece.
389            Result.push_back(ConstantExpr::getTrunc(Elt, DstEltTy));
390          }
391        }
392      }
393
394      return ConstantVector::get(Result.data(), Result.size());
395    }
396  }
397
398  return 0;
399}
400
401
402//===----------------------------------------------------------------------===//
403// Constant Folding public APIs
404//===----------------------------------------------------------------------===//
405
406
407/// ConstantFoldInstruction - Attempt to constant fold the specified
408/// instruction.  If successful, the constant result is returned, if not, null
409/// is returned.  Note that this function can only fail when attempting to fold
410/// instructions like loads and stores, which have no constant expression form.
411///
412Constant *llvm::ConstantFoldInstruction(Instruction *I, LLVMContext &Context,
413                                        const TargetData *TD) {
414  if (PHINode *PN = dyn_cast<PHINode>(I)) {
415    if (PN->getNumIncomingValues() == 0)
416      return UndefValue::get(PN->getType());
417
418    Constant *Result = dyn_cast<Constant>(PN->getIncomingValue(0));
419    if (Result == 0) return 0;
420
421    // Handle PHI nodes specially here...
422    for (unsigned i = 1, e = PN->getNumIncomingValues(); i != e; ++i)
423      if (PN->getIncomingValue(i) != Result && PN->getIncomingValue(i) != PN)
424        return 0;   // Not all the same incoming constants...
425
426    // If we reach here, all incoming values are the same constant.
427    return Result;
428  }
429
430  // Scan the operand list, checking to see if they are all constants, if so,
431  // hand off to ConstantFoldInstOperands.
432  SmallVector<Constant*, 8> Ops;
433  for (User::op_iterator i = I->op_begin(), e = I->op_end(); i != e; ++i)
434    if (Constant *Op = dyn_cast<Constant>(*i))
435      Ops.push_back(Op);
436    else
437      return 0;  // All operands not constant!
438
439  if (const CmpInst *CI = dyn_cast<CmpInst>(I))
440    return ConstantFoldCompareInstOperands(CI->getPredicate(),
441                                           Ops.data(), Ops.size(),
442                                           Context, TD);
443
444  if (const LoadInst *LI = dyn_cast<LoadInst>(I))
445    return ConstantFoldLoadInst(LI, TD);
446
447  return ConstantFoldInstOperands(I->getOpcode(), I->getType(),
448                                  Ops.data(), Ops.size(), Context, TD);
449}
450
451/// ConstantFoldConstantExpression - Attempt to fold the constant expression
452/// using the specified TargetData.  If successful, the constant result is
453/// result is returned, if not, null is returned.
454Constant *llvm::ConstantFoldConstantExpression(ConstantExpr *CE,
455                                               LLVMContext &Context,
456                                               const TargetData *TD) {
457  SmallVector<Constant*, 8> Ops;
458  for (User::op_iterator i = CE->op_begin(), e = CE->op_end(); i != e; ++i)
459    Ops.push_back(cast<Constant>(*i));
460
461  if (CE->isCompare())
462    return ConstantFoldCompareInstOperands(CE->getPredicate(),
463                                           Ops.data(), Ops.size(),
464                                           Context, TD);
465  return ConstantFoldInstOperands(CE->getOpcode(), CE->getType(),
466                                  Ops.data(), Ops.size(), Context, TD);
467}
468
469/// ConstantFoldInstOperands - Attempt to constant fold an instruction with the
470/// specified opcode and operands.  If successful, the constant result is
471/// returned, if not, null is returned.  Note that this function can fail when
472/// attempting to fold instructions like loads and stores, which have no
473/// constant expression form.
474///
475Constant *llvm::ConstantFoldInstOperands(unsigned Opcode, const Type *DestTy,
476                                         Constant* const* Ops, unsigned NumOps,
477                                         LLVMContext &Context,
478                                         const TargetData *TD) {
479  // Handle easy binops first.
480  if (Instruction::isBinaryOp(Opcode)) {
481    if (isa<ConstantExpr>(Ops[0]) || isa<ConstantExpr>(Ops[1]))
482      if (Constant *C = SymbolicallyEvaluateBinop(Opcode, Ops[0], Ops[1], TD,
483                                                  Context))
484        return C;
485
486    return ConstantExpr::get(Opcode, Ops[0], Ops[1]);
487  }
488
489  switch (Opcode) {
490  default: return 0;
491  case Instruction::Call:
492    if (Function *F = dyn_cast<Function>(Ops[0]))
493      if (canConstantFoldCallTo(F))
494        return ConstantFoldCall(F, Ops+1, NumOps-1);
495    return 0;
496  case Instruction::ICmp:
497  case Instruction::FCmp:
498    llvm_unreachable("This function is invalid for compares: no predicate specified");
499  case Instruction::PtrToInt:
500    // If the input is a inttoptr, eliminate the pair.  This requires knowing
501    // the width of a pointer, so it can't be done in ConstantExpr::getCast.
502    if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ops[0])) {
503      if (TD && CE->getOpcode() == Instruction::IntToPtr) {
504        Constant *Input = CE->getOperand(0);
505        unsigned InWidth = Input->getType()->getScalarSizeInBits();
506        if (TD->getPointerSizeInBits() < InWidth) {
507          Constant *Mask =
508            ConstantInt::get(Context, APInt::getLowBitsSet(InWidth,
509                                                  TD->getPointerSizeInBits()));
510          Input = ConstantExpr::getAnd(Input, Mask);
511        }
512        // Do a zext or trunc to get to the dest size.
513        return ConstantExpr::getIntegerCast(Input, DestTy, false);
514      }
515    }
516    return ConstantExpr::getCast(Opcode, Ops[0], DestTy);
517  case Instruction::IntToPtr:
518    // If the input is a ptrtoint, turn the pair into a ptr to ptr bitcast if
519    // the int size is >= the ptr size.  This requires knowing the width of a
520    // pointer, so it can't be done in ConstantExpr::getCast.
521    if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ops[0])) {
522      if (TD &&
523          TD->getPointerSizeInBits() <=
524          CE->getType()->getScalarSizeInBits()) {
525        if (CE->getOpcode() == Instruction::PtrToInt) {
526          Constant *Input = CE->getOperand(0);
527          Constant *C = FoldBitCast(Input, DestTy, *TD, Context);
528          return C ? C : ConstantExpr::getBitCast(Input, DestTy);
529        }
530        // If there's a constant offset added to the integer value before
531        // it is casted back to a pointer, see if the expression can be
532        // converted into a GEP.
533        if (CE->getOpcode() == Instruction::Add)
534          if (ConstantInt *L = dyn_cast<ConstantInt>(CE->getOperand(0)))
535            if (ConstantExpr *R = dyn_cast<ConstantExpr>(CE->getOperand(1)))
536              if (R->getOpcode() == Instruction::PtrToInt)
537                if (GlobalVariable *GV =
538                      dyn_cast<GlobalVariable>(R->getOperand(0))) {
539                  const PointerType *GVTy = cast<PointerType>(GV->getType());
540                  if (const ArrayType *AT =
541                        dyn_cast<ArrayType>(GVTy->getElementType())) {
542                    const Type *ElTy = AT->getElementType();
543                    uint64_t AllocSize = TD->getTypeAllocSize(ElTy);
544                    APInt PSA(L->getValue().getBitWidth(), AllocSize);
545                    if (ElTy == cast<PointerType>(DestTy)->getElementType() &&
546                        L->getValue().urem(PSA) == 0) {
547                      APInt ElemIdx = L->getValue().udiv(PSA);
548                      if (ElemIdx.ult(APInt(ElemIdx.getBitWidth(),
549                                            AT->getNumElements()))) {
550                        Constant *Index[] = {
551                          Constant::getNullValue(CE->getType()),
552                          ConstantInt::get(Context, ElemIdx)
553                        };
554                        return
555                        ConstantExpr::getGetElementPtr(GV, &Index[0], 2);
556                      }
557                    }
558                  }
559                }
560      }
561    }
562    return ConstantExpr::getCast(Opcode, Ops[0], DestTy);
563  case Instruction::Trunc:
564  case Instruction::ZExt:
565  case Instruction::SExt:
566  case Instruction::FPTrunc:
567  case Instruction::FPExt:
568  case Instruction::UIToFP:
569  case Instruction::SIToFP:
570  case Instruction::FPToUI:
571  case Instruction::FPToSI:
572      return ConstantExpr::getCast(Opcode, Ops[0], DestTy);
573  case Instruction::BitCast:
574    if (TD)
575      if (Constant *C = FoldBitCast(Ops[0], DestTy, *TD, Context))
576        return C;
577    return ConstantExpr::getBitCast(Ops[0], DestTy);
578  case Instruction::Select:
579    return ConstantExpr::getSelect(Ops[0], Ops[1], Ops[2]);
580  case Instruction::ExtractElement:
581    return ConstantExpr::getExtractElement(Ops[0], Ops[1]);
582  case Instruction::InsertElement:
583    return ConstantExpr::getInsertElement(Ops[0], Ops[1], Ops[2]);
584  case Instruction::ShuffleVector:
585    return ConstantExpr::getShuffleVector(Ops[0], Ops[1], Ops[2]);
586  case Instruction::GetElementPtr:
587    if (Constant *C = SymbolicallyEvaluateGEP(Ops, NumOps, DestTy, Context, TD))
588      return C;
589
590    return ConstantExpr::getGetElementPtr(Ops[0], Ops+1, NumOps-1);
591  }
592}
593
594/// ConstantFoldCompareInstOperands - Attempt to constant fold a compare
595/// instruction (icmp/fcmp) with the specified operands.  If it fails, it
596/// returns a constant expression of the specified operands.
597///
598Constant *llvm::ConstantFoldCompareInstOperands(unsigned Predicate,
599                                                Constant*const * Ops,
600                                                unsigned NumOps,
601                                                LLVMContext &Context,
602                                                const TargetData *TD) {
603  // fold: icmp (inttoptr x), null         -> icmp x, 0
604  // fold: icmp (ptrtoint x), 0            -> icmp x, null
605  // fold: icmp (inttoptr x), (inttoptr y) -> icmp trunc/zext x, trunc/zext y
606  // fold: icmp (ptrtoint x), (ptrtoint y) -> icmp x, y
607  //
608  // ConstantExpr::getCompare cannot do this, because it doesn't have TD
609  // around to know if bit truncation is happening.
610  if (ConstantExpr *CE0 = dyn_cast<ConstantExpr>(Ops[0])) {
611    if (TD && Ops[1]->isNullValue()) {
612      const Type *IntPtrTy = TD->getIntPtrType(Context);
613      if (CE0->getOpcode() == Instruction::IntToPtr) {
614        // Convert the integer value to the right size to ensure we get the
615        // proper extension or truncation.
616        Constant *C = ConstantExpr::getIntegerCast(CE0->getOperand(0),
617                                                   IntPtrTy, false);
618        Constant *NewOps[] = { C, Constant::getNullValue(C->getType()) };
619        return ConstantFoldCompareInstOperands(Predicate, NewOps, 2,
620                                               Context, TD);
621      }
622
623      // Only do this transformation if the int is intptrty in size, otherwise
624      // there is a truncation or extension that we aren't modeling.
625      if (CE0->getOpcode() == Instruction::PtrToInt &&
626          CE0->getType() == IntPtrTy) {
627        Constant *C = CE0->getOperand(0);
628        Constant *NewOps[] = { C, Constant::getNullValue(C->getType()) };
629        // FIXME!
630        return ConstantFoldCompareInstOperands(Predicate, NewOps, 2,
631                                               Context, TD);
632      }
633    }
634
635    if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(Ops[1])) {
636      if (TD && CE0->getOpcode() == CE1->getOpcode()) {
637        const Type *IntPtrTy = TD->getIntPtrType(Context);
638
639        if (CE0->getOpcode() == Instruction::IntToPtr) {
640          // Convert the integer value to the right size to ensure we get the
641          // proper extension or truncation.
642          Constant *C0 = ConstantExpr::getIntegerCast(CE0->getOperand(0),
643                                                      IntPtrTy, false);
644          Constant *C1 = ConstantExpr::getIntegerCast(CE1->getOperand(0),
645                                                      IntPtrTy, false);
646          Constant *NewOps[] = { C0, C1 };
647          return ConstantFoldCompareInstOperands(Predicate, NewOps, 2,
648                                                 Context, TD);
649        }
650
651        // Only do this transformation if the int is intptrty in size, otherwise
652        // there is a truncation or extension that we aren't modeling.
653        if ((CE0->getOpcode() == Instruction::PtrToInt &&
654             CE0->getType() == IntPtrTy &&
655             CE0->getOperand(0)->getType() == CE1->getOperand(0)->getType())) {
656          Constant *NewOps[] = {
657            CE0->getOperand(0), CE1->getOperand(0)
658          };
659          return ConstantFoldCompareInstOperands(Predicate, NewOps, 2,
660                                                 Context, TD);
661        }
662      }
663    }
664  }
665  return ConstantExpr::getCompare(Predicate, Ops[0], Ops[1]);
666}
667
668
669/// ConstantFoldLoadThroughGEPConstantExpr - Given a constant and a
670/// getelementptr constantexpr, return the constant value being addressed by the
671/// constant expression, or null if something is funny and we can't decide.
672Constant *llvm::ConstantFoldLoadThroughGEPConstantExpr(Constant *C,
673                                                       ConstantExpr *CE) {
674  if (CE->getOperand(1) != Constant::getNullValue(CE->getOperand(1)->getType()))
675    return 0;  // Do not allow stepping over the value!
676
677  // Loop over all of the operands, tracking down which value we are
678  // addressing...
679  gep_type_iterator I = gep_type_begin(CE), E = gep_type_end(CE);
680  for (++I; I != E; ++I)
681    if (const StructType *STy = dyn_cast<StructType>(*I)) {
682      ConstantInt *CU = cast<ConstantInt>(I.getOperand());
683      assert(CU->getZExtValue() < STy->getNumElements() &&
684             "Struct index out of range!");
685      unsigned El = (unsigned)CU->getZExtValue();
686      if (ConstantStruct *CS = dyn_cast<ConstantStruct>(C)) {
687        C = CS->getOperand(El);
688      } else if (isa<ConstantAggregateZero>(C)) {
689        C = Constant::getNullValue(STy->getElementType(El));
690      } else if (isa<UndefValue>(C)) {
691        C = UndefValue::get(STy->getElementType(El));
692      } else {
693        return 0;
694      }
695    } else if (ConstantInt *CI = dyn_cast<ConstantInt>(I.getOperand())) {
696      if (const ArrayType *ATy = dyn_cast<ArrayType>(*I)) {
697        if (CI->getZExtValue() >= ATy->getNumElements())
698         return 0;
699        if (ConstantArray *CA = dyn_cast<ConstantArray>(C))
700          C = CA->getOperand(CI->getZExtValue());
701        else if (isa<ConstantAggregateZero>(C))
702          C = Constant::getNullValue(ATy->getElementType());
703        else if (isa<UndefValue>(C))
704          C = UndefValue::get(ATy->getElementType());
705        else
706          return 0;
707      } else if (const VectorType *VTy = dyn_cast<VectorType>(*I)) {
708        if (CI->getZExtValue() >= VTy->getNumElements())
709          return 0;
710        if (ConstantVector *CP = dyn_cast<ConstantVector>(C))
711          C = CP->getOperand(CI->getZExtValue());
712        else if (isa<ConstantAggregateZero>(C))
713          C = Constant::getNullValue(VTy->getElementType());
714        else if (isa<UndefValue>(C))
715          C = UndefValue::get(VTy->getElementType());
716        else
717          return 0;
718      } else {
719        return 0;
720      }
721    } else {
722      return 0;
723    }
724  return C;
725}
726
727
728//===----------------------------------------------------------------------===//
729//  Constant Folding for Calls
730//
731
732/// canConstantFoldCallTo - Return true if its even possible to fold a call to
733/// the specified function.
734bool
735llvm::canConstantFoldCallTo(const Function *F) {
736  switch (F->getIntrinsicID()) {
737  case Intrinsic::sqrt:
738  case Intrinsic::powi:
739  case Intrinsic::bswap:
740  case Intrinsic::ctpop:
741  case Intrinsic::ctlz:
742  case Intrinsic::cttz:
743  case Intrinsic::uadd_with_overflow:
744  case Intrinsic::usub_with_overflow:
745  case Intrinsic::sadd_with_overflow:
746  case Intrinsic::ssub_with_overflow:
747    return true;
748  default:
749    return false;
750  case 0: break;
751  }
752
753  if (!F->hasName()) return false;
754  StringRef Name = F->getName();
755
756  // In these cases, the check of the length is required.  We don't want to
757  // return true for a name like "cos\0blah" which strcmp would return equal to
758  // "cos", but has length 8.
759  switch (Name[0]) {
760  default: return false;
761  case 'a':
762    return Name == "acos" || Name == "asin" ||
763      Name == "atan" || Name == "atan2";
764  case 'c':
765    return Name == "cos" || Name == "ceil" || Name == "cosf" || Name == "cosh";
766  case 'e':
767    return Name == "exp";
768  case 'f':
769    return Name == "fabs" || Name == "fmod" || Name == "floor";
770  case 'l':
771    return Name == "log" || Name == "log10";
772  case 'p':
773    return Name == "pow";
774  case 's':
775    return Name == "sin" || Name == "sinh" || Name == "sqrt" ||
776      Name == "sinf" || Name == "sqrtf";
777  case 't':
778    return Name == "tan" || Name == "tanh";
779  }
780}
781
782static Constant *ConstantFoldFP(double (*NativeFP)(double), double V,
783                                const Type *Ty, LLVMContext &Context) {
784  errno = 0;
785  V = NativeFP(V);
786  if (errno != 0) {
787    errno = 0;
788    return 0;
789  }
790
791  if (Ty->isFloatTy())
792    return ConstantFP::get(Context, APFloat((float)V));
793  if (Ty->isDoubleTy())
794    return ConstantFP::get(Context, APFloat(V));
795  llvm_unreachable("Can only constant fold float/double");
796  return 0; // dummy return to suppress warning
797}
798
799static Constant *ConstantFoldBinaryFP(double (*NativeFP)(double, double),
800                                      double V, double W,
801                                      const Type *Ty,
802                                      LLVMContext &Context) {
803  errno = 0;
804  V = NativeFP(V, W);
805  if (errno != 0) {
806    errno = 0;
807    return 0;
808  }
809
810  if (Ty->isFloatTy())
811    return ConstantFP::get(Context, APFloat((float)V));
812  if (Ty->isDoubleTy())
813    return ConstantFP::get(Context, APFloat(V));
814  llvm_unreachable("Can only constant fold float/double");
815  return 0; // dummy return to suppress warning
816}
817
818/// ConstantFoldCall - Attempt to constant fold a call to the specified function
819/// with the specified arguments, returning null if unsuccessful.
820Constant *
821llvm::ConstantFoldCall(Function *F,
822                       Constant *const *Operands, unsigned NumOperands) {
823  if (!F->hasName()) return 0;
824  LLVMContext &Context = F->getContext();
825  StringRef Name = F->getName();
826
827  const Type *Ty = F->getReturnType();
828  if (NumOperands == 1) {
829    if (ConstantFP *Op = dyn_cast<ConstantFP>(Operands[0])) {
830      if (!Ty->isFloatTy() && !Ty->isDoubleTy())
831        return 0;
832      /// Currently APFloat versions of these functions do not exist, so we use
833      /// the host native double versions.  Float versions are not called
834      /// directly but for all these it is true (float)(f((double)arg)) ==
835      /// f(arg).  Long double not supported yet.
836      double V = Ty->isFloatTy() ? (double)Op->getValueAPF().convertToFloat() :
837                                     Op->getValueAPF().convertToDouble();
838      switch (Name[0]) {
839      case 'a':
840        if (Name == "acos")
841          return ConstantFoldFP(acos, V, Ty, Context);
842        else if (Name == "asin")
843          return ConstantFoldFP(asin, V, Ty, Context);
844        else if (Name == "atan")
845          return ConstantFoldFP(atan, V, Ty, Context);
846        break;
847      case 'c':
848        if (Name == "ceil")
849          return ConstantFoldFP(ceil, V, Ty, Context);
850        else if (Name == "cos")
851          return ConstantFoldFP(cos, V, Ty, Context);
852        else if (Name == "cosh")
853          return ConstantFoldFP(cosh, V, Ty, Context);
854        else if (Name == "cosf")
855          return ConstantFoldFP(cos, V, Ty, Context);
856        break;
857      case 'e':
858        if (Name == "exp")
859          return ConstantFoldFP(exp, V, Ty, Context);
860        break;
861      case 'f':
862        if (Name == "fabs")
863          return ConstantFoldFP(fabs, V, Ty, Context);
864        else if (Name == "floor")
865          return ConstantFoldFP(floor, V, Ty, Context);
866        break;
867      case 'l':
868        if (Name == "log" && V > 0)
869          return ConstantFoldFP(log, V, Ty, Context);
870        else if (Name == "log10" && V > 0)
871          return ConstantFoldFP(log10, V, Ty, Context);
872        else if (Name == "llvm.sqrt.f32" ||
873                 Name == "llvm.sqrt.f64") {
874          if (V >= -0.0)
875            return ConstantFoldFP(sqrt, V, Ty, Context);
876          else // Undefined
877            return Constant::getNullValue(Ty);
878        }
879        break;
880      case 's':
881        if (Name == "sin")
882          return ConstantFoldFP(sin, V, Ty, Context);
883        else if (Name == "sinh")
884          return ConstantFoldFP(sinh, V, Ty, Context);
885        else if (Name == "sqrt" && V >= 0)
886          return ConstantFoldFP(sqrt, V, Ty, Context);
887        else if (Name == "sqrtf" && V >= 0)
888          return ConstantFoldFP(sqrt, V, Ty, Context);
889        else if (Name == "sinf")
890          return ConstantFoldFP(sin, V, Ty, Context);
891        break;
892      case 't':
893        if (Name == "tan")
894          return ConstantFoldFP(tan, V, Ty, Context);
895        else if (Name == "tanh")
896          return ConstantFoldFP(tanh, V, Ty, Context);
897        break;
898      default:
899        break;
900      }
901      return 0;
902    }
903
904
905    if (ConstantInt *Op = dyn_cast<ConstantInt>(Operands[0])) {
906      if (Name.startswith("llvm.bswap"))
907        return ConstantInt::get(Context, Op->getValue().byteSwap());
908      else if (Name.startswith("llvm.ctpop"))
909        return ConstantInt::get(Ty, Op->getValue().countPopulation());
910      else if (Name.startswith("llvm.cttz"))
911        return ConstantInt::get(Ty, Op->getValue().countTrailingZeros());
912      else if (Name.startswith("llvm.ctlz"))
913        return ConstantInt::get(Ty, Op->getValue().countLeadingZeros());
914      return 0;
915    }
916
917    return 0;
918  }
919
920  if (NumOperands == 2) {
921    if (ConstantFP *Op1 = dyn_cast<ConstantFP>(Operands[0])) {
922      if (!Ty->isFloatTy() && !Ty->isDoubleTy())
923        return 0;
924      double Op1V = Ty->isFloatTy() ?
925                      (double)Op1->getValueAPF().convertToFloat() :
926                      Op1->getValueAPF().convertToDouble();
927      if (ConstantFP *Op2 = dyn_cast<ConstantFP>(Operands[1])) {
928        if (Op2->getType() != Op1->getType())
929          return 0;
930
931        double Op2V = Ty->isFloatTy() ?
932                      (double)Op2->getValueAPF().convertToFloat():
933                      Op2->getValueAPF().convertToDouble();
934
935        if (Name == "pow")
936          return ConstantFoldBinaryFP(pow, Op1V, Op2V, Ty, Context);
937        if (Name == "fmod")
938          return ConstantFoldBinaryFP(fmod, Op1V, Op2V, Ty, Context);
939        if (Name == "atan2")
940          return ConstantFoldBinaryFP(atan2, Op1V, Op2V, Ty, Context);
941      } else if (ConstantInt *Op2C = dyn_cast<ConstantInt>(Operands[1])) {
942        if (Name == "llvm.powi.f32")
943          return ConstantFP::get(Context, APFloat((float)std::pow((float)Op1V,
944                                                 (int)Op2C->getZExtValue())));
945        if (Name == "llvm.powi.f64")
946          return ConstantFP::get(Context, APFloat((double)std::pow((double)Op1V,
947                                                 (int)Op2C->getZExtValue())));
948      }
949      return 0;
950    }
951
952
953    if (ConstantInt *Op1 = dyn_cast<ConstantInt>(Operands[0])) {
954      if (ConstantInt *Op2 = dyn_cast<ConstantInt>(Operands[1])) {
955        switch (F->getIntrinsicID()) {
956        default: break;
957        case Intrinsic::uadd_with_overflow: {
958          Constant *Res = ConstantExpr::getAdd(Op1, Op2);           // result.
959          Constant *Ops[] = {
960            Res, ConstantExpr::getICmp(CmpInst::ICMP_ULT, Res, Op1) // overflow.
961          };
962          return ConstantStruct::get(F->getContext(), Ops, 2, false);
963        }
964        case Intrinsic::usub_with_overflow: {
965          Constant *Res = ConstantExpr::getSub(Op1, Op2);           // result.
966          Constant *Ops[] = {
967            Res, ConstantExpr::getICmp(CmpInst::ICMP_UGT, Res, Op1) // overflow.
968          };
969          return ConstantStruct::get(F->getContext(), Ops, 2, false);
970        }
971        case Intrinsic::sadd_with_overflow: {
972          Constant *Res = ConstantExpr::getAdd(Op1, Op2);           // result.
973          Constant *Overflow = ConstantExpr::getSelect(
974              ConstantExpr::getICmp(CmpInst::ICMP_SGT,
975                ConstantInt::get(Op1->getType(), 0), Op1),
976              ConstantExpr::getICmp(CmpInst::ICMP_SGT, Res, Op2),
977              ConstantExpr::getICmp(CmpInst::ICMP_SLT, Res, Op2)); // overflow.
978
979          Constant *Ops[] = { Res, Overflow };
980          return ConstantStruct::get(F->getContext(), Ops, 2, false);
981        }
982        case Intrinsic::ssub_with_overflow: {
983          Constant *Res = ConstantExpr::getSub(Op1, Op2);           // result.
984          Constant *Overflow = ConstantExpr::getSelect(
985              ConstantExpr::getICmp(CmpInst::ICMP_SGT,
986                ConstantInt::get(Op2->getType(), 0), Op2),
987              ConstantExpr::getICmp(CmpInst::ICMP_SLT, Res, Op1),
988              ConstantExpr::getICmp(CmpInst::ICMP_SGT, Res, Op1)); // overflow.
989
990          Constant *Ops[] = { Res, Overflow };
991          return ConstantStruct::get(F->getContext(), Ops, 2, false);
992        }
993        }
994      }
995
996      return 0;
997    }
998    return 0;
999  }
1000  return 0;
1001}
1002
1003