/// know the sizes of the potential memory references.
static uint64_t const UnknownSize = ~UINT64_C(0);
- /// getDataLayout - Return a pointer to the current DataLayout object, or
- /// null if no DataLayout object is available.
- ///
- const DataLayout *getDataLayout() const { return DL; }
-
/// getTargetLibraryInfo - Return a pointer to the current TargetLibraryInfo
/// object, or null if no TargetLibraryInfo object is available.
///
/// Note that this fails if not all of the operands are constant. Otherwise,
/// this function can only fail when attempting to fold instructions like loads
/// and stores, which have no constant expression form.
-Constant *ConstantFoldInstruction(Instruction *I,
- const DataLayout *TD = nullptr,
- const TargetLibraryInfo *TLI = nullptr);
+ Constant *ConstantFoldInstruction(Instruction *I, const DataLayout &DL,
+ const TargetLibraryInfo *TLI = nullptr);
/// ConstantFoldConstantExpression - Attempt to fold the constant expression
/// using the specified DataLayout. If successful, the constant result is
/// result is returned, if not, null is returned.
-Constant *ConstantFoldConstantExpression(const ConstantExpr *CE,
- const DataLayout *TD = nullptr,
- const TargetLibraryInfo *TLI =nullptr);
+ Constant *
+ ConstantFoldConstantExpression(const ConstantExpr *CE, const DataLayout &DL,
+ const TargetLibraryInfo *TLI = nullptr);
/// ConstantFoldInstOperands - Attempt to constant fold an instruction with the
/// specified operands. If successful, the constant result is returned, if not,
/// fold instructions like loads and stores, which have no constant expression
/// form.
///
-Constant *ConstantFoldInstOperands(unsigned Opcode, Type *DestTy,
- ArrayRef<Constant *> Ops,
- const DataLayout *TD = nullptr,
- const TargetLibraryInfo *TLI = nullptr);
+ Constant *ConstantFoldInstOperands(unsigned Opcode, Type *DestTy,
+ ArrayRef<Constant *> Ops,
+ const DataLayout &DL,
+ const TargetLibraryInfo *TLI = nullptr);
/// ConstantFoldCompareInstOperands - Attempt to constant fold a compare
/// instruction (icmp/fcmp) with the specified operands. If it fails, it
/// returns a constant expression of the specified operands.
///
-Constant *ConstantFoldCompareInstOperands(unsigned Predicate,
- Constant *LHS, Constant *RHS,
- const DataLayout *TD = nullptr,
- const TargetLibraryInfo *TLI=nullptr);
+ Constant *
+ ConstantFoldCompareInstOperands(unsigned Predicate, Constant *LHS,
+ Constant *RHS, const DataLayout &DL,
+ const TargetLibraryInfo *TLI = nullptr);
/// ConstantFoldInsertValueInstruction - Attempt to constant fold an insertvalue
/// instruction with the specified operands and indices. The constant result is
/// ConstantFoldLoadFromConstPtr - Return the value that a load from C would
/// produce if it is constant and determinable. If this is not determinable,
/// return null.
-Constant *ConstantFoldLoadFromConstPtr(Constant *C,
- const DataLayout *TD = nullptr);
+Constant *ConstantFoldLoadFromConstPtr(Constant *C, const DataLayout &DL);
/// ConstantFoldLoadThroughGEPConstantExpr - Given a constant and a
/// getelementptr constantexpr, return the constant value being addressed by the
LoopInfo *LI;
DominatorTree *DT;
ScalarEvolution *SE;
- const DataLayout *DL;
SmallPtrSet<Instruction*,16> Processed;
/// IVUses - A list of all tracked IV uses of induction variable expressions
/// SimplifyAddInst - Given operands for an Add, see if we can
/// fold the result. If not, this returns null.
Value *SimplifyAddInst(Value *LHS, Value *RHS, bool isNSW, bool isNUW,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifySubInst - Given operands for a Sub, see if we can
/// fold the result. If not, this returns null.
Value *SimplifySubInst(Value *LHS, Value *RHS, bool isNSW, bool isNUW,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// Given operands for an FAdd, see if we can fold the result. If not, this
/// returns null.
Value *SimplifyFAddInst(Value *LHS, Value *RHS, FastMathFlags FMF,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// Given operands for an FSub, see if we can fold the result. If not, this
/// returns null.
Value *SimplifyFSubInst(Value *LHS, Value *RHS, FastMathFlags FMF,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// Given operands for an FMul, see if we can fold the result. If not, this
/// returns null.
Value *SimplifyFMulInst(Value *LHS, Value *RHS, FastMathFlags FMF,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyMulInst - Given operands for a Mul, see if we can
/// fold the result. If not, this returns null.
- Value *SimplifyMulInst(Value *LHS, Value *RHS, const DataLayout *TD = nullptr,
+ Value *SimplifyMulInst(Value *LHS, Value *RHS, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifySDivInst - Given operands for an SDiv, see if we can
/// fold the result. If not, this returns null.
- Value *SimplifySDivInst(Value *LHS, Value *RHS,
- const DataLayout *TD = nullptr,
+ Value *SimplifySDivInst(Value *LHS, Value *RHS, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyUDivInst - Given operands for a UDiv, see if we can
/// fold the result. If not, this returns null.
- Value *SimplifyUDivInst(Value *LHS, Value *RHS,
- const DataLayout *TD = nullptr,
+ Value *SimplifyUDivInst(Value *LHS, Value *RHS, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyFDivInst - Given operands for an FDiv, see if we can
/// fold the result. If not, this returns null.
Value *SimplifyFDivInst(Value *LHS, Value *RHS, FastMathFlags FMF,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifySRemInst - Given operands for an SRem, see if we can
/// fold the result. If not, this returns null.
- Value *SimplifySRemInst(Value *LHS, Value *RHS,
- const DataLayout *TD = nullptr,
+ Value *SimplifySRemInst(Value *LHS, Value *RHS, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyURemInst - Given operands for a URem, see if we can
/// fold the result. If not, this returns null.
- Value *SimplifyURemInst(Value *LHS, Value *RHS,
- const DataLayout *TD = nullptr,
+ Value *SimplifyURemInst(Value *LHS, Value *RHS, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyFRemInst - Given operands for an FRem, see if we can
/// fold the result. If not, this returns null.
Value *SimplifyFRemInst(Value *LHS, Value *RHS, FastMathFlags FMF,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyShlInst - Given operands for a Shl, see if we can
/// fold the result. If not, this returns null.
Value *SimplifyShlInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyLShrInst - Given operands for a LShr, see if we can
/// fold the result. If not, this returns null.
Value *SimplifyLShrInst(Value *Op0, Value *Op1, bool isExact,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyAShrInst - Given operands for a AShr, see if we can
/// fold the result. If not, this returns null.
Value *SimplifyAShrInst(Value *Op0, Value *Op1, bool isExact,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyAndInst - Given operands for an And, see if we can
/// fold the result. If not, this returns null.
- Value *SimplifyAndInst(Value *LHS, Value *RHS, const DataLayout *TD = nullptr,
+ Value *SimplifyAndInst(Value *LHS, Value *RHS, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyOrInst - Given operands for an Or, see if we can
/// fold the result. If not, this returns null.
- Value *SimplifyOrInst(Value *LHS, Value *RHS, const DataLayout *TD = nullptr,
+ Value *SimplifyOrInst(Value *LHS, Value *RHS, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyXorInst - Given operands for a Xor, see if we can
/// fold the result. If not, this returns null.
- Value *SimplifyXorInst(Value *LHS, Value *RHS, const DataLayout *TD = nullptr,
+ Value *SimplifyXorInst(Value *LHS, Value *RHS, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyICmpInst - Given operands for an ICmpInst, see if we can
/// fold the result. If not, this returns null.
Value *SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyFCmpInst - Given operands for an FCmpInst, see if we can
/// fold the result. If not, this returns null.
Value *SimplifyFCmpInst(unsigned Predicate, Value *LHS, Value *RHS,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifySelectInst - Given operands for a SelectInst, see if we can fold
/// the result. If not, this returns null.
Value *SimplifySelectInst(Value *Cond, Value *TrueVal, Value *FalseVal,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyGEPInst - Given operands for an GetElementPtrInst, see if we can
/// fold the result. If not, this returns null.
- Value *SimplifyGEPInst(ArrayRef<Value *> Ops, const DataLayout *TD = nullptr,
+ Value *SimplifyGEPInst(ArrayRef<Value *> Ops, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyInsertValueInst - Given operands for an InsertValueInst, see if we
/// can fold the result. If not, this returns null.
Value *SimplifyInsertValueInst(Value *Agg, Value *Val,
- ArrayRef<unsigned> Idxs,
- const DataLayout *TD = nullptr,
+ ArrayRef<unsigned> Idxs, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyTruncInst - Given operands for an TruncInst, see if we can fold
/// the result. If not, this returns null.
- Value *SimplifyTruncInst(Value *Op, Type *Ty, const DataLayout *TD = nullptr,
+ Value *SimplifyTruncInst(Value *Op, Type *Ty, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyCmpInst - Given operands for a CmpInst, see if we can
/// fold the result. If not, this returns null.
Value *SimplifyCmpInst(unsigned Predicate, Value *LHS, Value *RHS,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyBinOp - Given operands for a BinaryOperator, see if we can
/// fold the result. If not, this returns null.
Value *SimplifyBinOp(unsigned Opcode, Value *LHS, Value *RHS,
- const DataLayout *TD = nullptr,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// In contrast to SimplifyBinOp, try to use FastMathFlag when folding the
/// result. In case we don't need FastMathFlags, simply fall to SimplifyBinOp.
Value *SimplifyFPBinOp(unsigned Opcode, Value *LHS, Value *RHS,
- const FastMathFlags &FMF,
- const DataLayout *TD = nullptr,
+ const FastMathFlags &FMF, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
///
/// If this call could not be simplified returns null.
Value *SimplifyCall(Value *V, User::op_iterator ArgBegin,
- User::op_iterator ArgEnd, const DataLayout *TD = nullptr,
+ User::op_iterator ArgEnd, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// result.
///
/// If this call could not be simplified returns null.
- Value *SimplifyCall(Value *V, ArrayRef<Value *> Args,
- const DataLayout *TD = nullptr,
+ Value *SimplifyCall(Value *V, ArrayRef<Value *> Args, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr,
/// SimplifyInstruction - See if we can compute a simplified version of this
/// instruction. If not, this returns null.
- Value *SimplifyInstruction(Instruction *I, const DataLayout *TD = nullptr,
+ Value *SimplifyInstruction(Instruction *I, const DataLayout &DL,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr);
///
/// The function returns true if any simplifications were performed.
bool replaceAndRecursivelySimplify(Instruction *I, Value *SimpleV,
- const DataLayout *TD = nullptr,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr);
/// of the users impacted. It returns true if any simplifications were
/// performed.
bool recursivelySimplifyInstruction(Instruction *I,
- const DataLayout *TD = nullptr,
const TargetLibraryInfo *TLI = nullptr,
const DominatorTree *DT = nullptr,
AssumptionCache *AC = nullptr);
/// This pass computes, caches, and vends lazy value constraint information.
class LazyValueInfo : public FunctionPass {
AssumptionCache *AC;
- const DataLayout *DL;
class TargetLibraryInfo *TLI;
DominatorTree *DT;
void *PImpl;
/// specified pointer, we do a quick local scan of the basic block containing
/// ScanFrom, to determine if the address is already accessed.
bool isSafeToLoadUnconditionally(Value *V, Instruction *ScanFrom,
- unsigned Align,
- const DataLayout *TD = nullptr);
+ unsigned Align);
/// FindAvailableLoadedValue - Scan the ScanBB block backwards (starting at
/// the instruction before ScanFrom) checking to see if we have the value at
SmallVector<unsigned, 2> AliasSetId;
};
- LoopAccessInfo(Loop *L, ScalarEvolution *SE, const DataLayout *DL,
+ LoopAccessInfo(Loop *L, ScalarEvolution *SE, const DataLayout &DL,
const TargetLibraryInfo *TLI, AliasAnalysis *AA,
DominatorTree *DT, const ValueToValueMap &Strides);
RuntimePointerCheck PtrRtCheck;
Loop *TheLoop;
ScalarEvolution *SE;
- const DataLayout *DL;
+ const DataLayout &DL;
const TargetLibraryInfo *TLI;
AliasAnalysis *AA;
DominatorTree *DT;
// The used analysis passes.
ScalarEvolution *SE;
- const DataLayout *DL;
const TargetLibraryInfo *TLI;
AliasAnalysis *AA;
DominatorTree *DT;
/// then return that multiple. For non-array mallocs, the multiple is
/// constant 1. Otherwise, return NULL for mallocs whose array size cannot be
/// determined.
-Value *getMallocArraySize(CallInst *CI, const DataLayout *DL,
+Value *getMallocArraySize(CallInst *CI, const DataLayout &DL,
const TargetLibraryInfo *TLI,
bool LookThroughSExt = false);
-
//===----------------------------------------------------------------------===//
// calloc Call Utility Functions.
//
/// underlying object pointed to by Ptr.
/// If RoundToAlign is true, then Size is rounded up to the aligment of allocas,
/// byval arguments, and global variables.
-bool getObjectSize(const Value *Ptr, uint64_t &Size, const DataLayout *DL,
+bool getObjectSize(const Value *Ptr, uint64_t &Size, const DataLayout &DL,
const TargetLibraryInfo *TLI, bool RoundToAlign = false);
-
-
typedef std::pair<APInt, APInt> SizeOffsetType;
/// \brief Evaluate the size and offset of an object pointed to by a Value*
class ObjectSizeOffsetVisitor
: public InstVisitor<ObjectSizeOffsetVisitor, SizeOffsetType> {
- const DataLayout *DL;
+ const DataLayout &DL;
const TargetLibraryInfo *TLI;
bool RoundToAlign;
unsigned IntTyBits;
}
public:
- ObjectSizeOffsetVisitor(const DataLayout *DL, const TargetLibraryInfo *TLI,
+ ObjectSizeOffsetVisitor(const DataLayout &DL, const TargetLibraryInfo *TLI,
LLVMContext &Context, bool RoundToAlign = false);
SizeOffsetType compute(Value *V);
typedef DenseMap<const Value*, WeakEvalType> CacheMapTy;
typedef SmallPtrSet<const Value*, 8> PtrSetTy;
- const DataLayout *DL;
+ const DataLayout &DL;
const TargetLibraryInfo *TLI;
LLVMContext &Context;
BuilderTy Builder;
SizeOffsetEvalType compute_(Value *V);
public:
- ObjectSizeOffsetEvaluator(const DataLayout *DL, const TargetLibraryInfo *TLI,
+ ObjectSizeOffsetEvaluator(const DataLayout &DL, const TargetLibraryInfo *TLI,
LLVMContext &Context, bool RoundToAlign = false);
SizeOffsetEvalType compute(Value *V);
class CallSite;
class AliasAnalysis;
class AssumptionCache;
- class DataLayout;
class MemoryDependenceAnalysis;
class PredIteratorCache;
class DominatorTree;
/// Current AA implementation, just a cache.
AliasAnalysis *AA;
- const DataLayout *DL;
DominatorTree *DT;
AssumptionCache *AC;
std::unique_ptr<PredIteratorCache> PredCache;
static unsigned getLoadLoadClobberFullWidthSize(const Value *MemLocBase,
int64_t MemLocOffs,
unsigned MemLocSize,
- const LoadInst *LI,
- const DataLayout &DL);
+ const LoadInst *LI);
private:
MemDepResult getCallSiteDependencyFrom(CallSite C, bool isReadOnlyCall,
class PHITransAddr {
/// Addr - The actual address we're analyzing.
Value *Addr;
-
- /// The DataLayout we are playing with if known, otherwise null.
- const DataLayout *DL;
+
+ /// The DataLayout we are playing with.
+ const DataLayout &DL;
/// TLI - The target library info if known, otherwise null.
const TargetLibraryInfo *TLI;
/// InstInputs - The inputs for our symbolic address.
SmallVector<Instruction*, 4> InstInputs;
public:
- PHITransAddr(Value *addr, const DataLayout *DL, AssumptionCache *AC)
+ PHITransAddr(Value *addr, const DataLayout &DL, AssumptionCache *AC)
: Addr(addr), DL(DL), TLI(nullptr), AC(AC) {
// If the address is an instruction, the whole thing is considered an input.
if (Instruction *I = dyn_cast<Instruction>(Addr))
///
LoopInfo *LI;
- /// The DataLayout information for the target we are targeting.
- ///
- const DataLayout *DL;
-
/// TLI - The target library information for the target we are targeting.
///
TargetLibraryInfo *TLI;
/// memory.
class SCEVExpander : public SCEVVisitor<SCEVExpander, Value*> {
ScalarEvolution &SE;
+ const DataLayout &DL;
// New instructions receive a name to identifies them with the current pass.
const char* IVName;
public:
/// SCEVExpander - Construct a SCEVExpander in "canonical" mode.
- explicit SCEVExpander(ScalarEvolution &se, const char *name)
- : SE(se), IVName(name), IVIncInsertLoop(nullptr), IVIncInsertPos(nullptr),
- CanonicalMode(true), LSRMode(false),
- Builder(se.getContext(), TargetFolder(se.DL)) {
+ explicit SCEVExpander(ScalarEvolution &se, const DataLayout &DL,
+ const char *name)
+ : SE(se), DL(DL), IVName(name), IVIncInsertLoop(nullptr),
+ IVIncInsertPos(nullptr), CanonicalMode(true), LSRMode(false),
+ Builder(se.getContext(), TargetFolder(DL)) {
#ifndef NDEBUG
DebugType = "";
#endif
/// TargetFolder - Create constants with target dependent folding.
class TargetFolder {
- const DataLayout *DL;
+ const DataLayout &DL;
/// Fold - Fold the constant using target specific information.
Constant *Fold(Constant *C) const {
}
public:
- explicit TargetFolder(const DataLayout *DL) : DL(DL) {}
+ explicit TargetFolder(const DataLayout &DL) : DL(DL) {}
//===--------------------------------------------------------------------===//
// Binary Operators
/// them in the KnownZero/KnownOne bit sets.
///
/// This function is defined on values with integer type, values with pointer
- /// type (but only if TD is non-null), and vectors of integers. In the case
+ /// type, and vectors of integers. In the case
/// where V is a vector, the known zero and known one values are the
/// same width as the vector element, and the bit is set only if it is true
/// for all of the elements in the vector.
void computeKnownBits(Value *V, APInt &KnownZero, APInt &KnownOne,
- const DataLayout *TD = nullptr, unsigned Depth = 0,
+ const DataLayout &DL, unsigned Depth = 0,
AssumptionCache *AC = nullptr,
const Instruction *CxtI = nullptr,
const DominatorTree *DT = nullptr);
/// ComputeSignBit - Determine whether the sign bit is known to be zero or
/// one. Convenience wrapper around computeKnownBits.
void ComputeSignBit(Value *V, bool &KnownZero, bool &KnownOne,
- const DataLayout *TD = nullptr, unsigned Depth = 0,
+ const DataLayout &DL, unsigned Depth = 0,
AssumptionCache *AC = nullptr,
const Instruction *CxtI = nullptr,
const DominatorTree *DT = nullptr);
/// element is known to be a power of two when defined. Supports values with
/// integer or pointer type and vectors of integers. If 'OrZero' is set then
/// returns true if the given value is either a power of two or zero.
- bool isKnownToBeAPowerOfTwo(Value *V, bool OrZero = false, unsigned Depth = 0,
+ bool isKnownToBeAPowerOfTwo(Value *V, const DataLayout &DL,
+ bool OrZero = false, unsigned Depth = 0,
AssumptionCache *AC = nullptr,
const Instruction *CxtI = nullptr,
const DominatorTree *DT = nullptr);
/// when defined. For vectors return true if every element is known to be
/// non-zero when defined. Supports values with integer or pointer type and
/// vectors of integers.
- bool isKnownNonZero(Value *V, const DataLayout *TD = nullptr,
- unsigned Depth = 0, AssumptionCache *AC = nullptr,
+ bool isKnownNonZero(Value *V, const DataLayout &DL, unsigned Depth = 0,
+ AssumptionCache *AC = nullptr,
const Instruction *CxtI = nullptr,
const DominatorTree *DT = nullptr);
/// zero for bits that V cannot have.
///
/// This function is defined on values with integer type, values with pointer
- /// type (but only if TD is non-null), and vectors of integers. In the case
+ /// type, and vectors of integers. In the case
/// where V is a vector, the mask, known zero, and known one values are the
/// same width as the vector element, and the bit is set only if it is true
/// for all of the elements in the vector.
- bool MaskedValueIsZero(Value *V, const APInt &Mask,
- const DataLayout *TD = nullptr, unsigned Depth = 0,
- AssumptionCache *AC = nullptr,
+ bool MaskedValueIsZero(Value *V, const APInt &Mask, const DataLayout &DL,
+ unsigned Depth = 0, AssumptionCache *AC = nullptr,
const Instruction *CxtI = nullptr,
const DominatorTree *DT = nullptr);
///
/// 'Op' must have a scalar integer type.
///
- unsigned ComputeNumSignBits(Value *Op, const DataLayout *TD = nullptr,
+ unsigned ComputeNumSignBits(Value *Op, const DataLayout &DL,
unsigned Depth = 0, AssumptionCache *AC = nullptr,
const Instruction *CxtI = nullptr,
const DominatorTree *DT = nullptr);
/// it can be expressed as a base pointer plus a constant offset. Return the
/// base and offset to the caller.
Value *GetPointerBaseWithConstantOffset(Value *Ptr, int64_t &Offset,
- const DataLayout *TD);
+ const DataLayout &DL);
static inline const Value *
GetPointerBaseWithConstantOffset(const Value *Ptr, int64_t &Offset,
- const DataLayout *TD) {
- return GetPointerBaseWithConstantOffset(const_cast<Value*>(Ptr), Offset,TD);
+ const DataLayout &DL) {
+ return GetPointerBaseWithConstantOffset(const_cast<Value *>(Ptr), Offset,
+ DL);
}
/// getConstantStringInfo - This function computes the length of a
/// being addressed. Note that the returned value has pointer type if the
/// specified value does. If the MaxLookup value is non-zero, it limits the
/// number of instructions to be stripped off.
- Value *GetUnderlyingObject(Value *V, const DataLayout *TD = nullptr,
+ Value *GetUnderlyingObject(Value *V, const DataLayout &DL,
unsigned MaxLookup = 6);
- static inline const Value *
- GetUnderlyingObject(const Value *V, const DataLayout *TD = nullptr,
- unsigned MaxLookup = 6) {
- return GetUnderlyingObject(const_cast<Value *>(V), TD, MaxLookup);
+ static inline const Value *GetUnderlyingObject(const Value *V,
+ const DataLayout &DL,
+ unsigned MaxLookup = 6) {
+ return GetUnderlyingObject(const_cast<Value *>(V), DL, MaxLookup);
}
/// GetUnderlyingObjects - This method is similar to GetUnderlyingObject
/// except that it can look through phi and select instructions and return
/// multiple objects.
- void GetUnderlyingObjects(Value *V,
- SmallVectorImpl<Value *> &Objects,
- const DataLayout *TD = nullptr,
- unsigned MaxLookup = 6);
+ void GetUnderlyingObjects(Value *V, SmallVectorImpl<Value *> &Objects,
+ const DataLayout &DL, unsigned MaxLookup = 6);
/// onlyUsedByLifetimeMarkers - Return true if the only users of this pointer
/// are lifetime markers.
/// the correct dominance relationships for the operands and users hold.
/// However, this method can return true for instructions that read memory;
/// for such instructions, moving them may change the resulting value.
- bool isSafeToSpeculativelyExecute(const Value *V,
- const DataLayout *TD = nullptr);
+ bool isSafeToSpeculativelyExecute(const Value *V);
/// isKnownNonNull - Return true if this pointer couldn't possibly be null by
/// its definition. This returns true for allocas, non-extern-weak globals
/// assume intrinsic, I, at the point in the control-flow identified by the
/// context instruction, CxtI.
bool isValidAssumeForContext(const Instruction *I, const Instruction *CxtI,
- const DataLayout *DL = nullptr,
const DominatorTree *DT = nullptr);
enum class OverflowResult { AlwaysOverflows, MayOverflow, NeverOverflows };
OverflowResult computeOverflowForUnsignedMul(Value *LHS, Value *RHS,
- const DataLayout *DL,
+ const DataLayout &DL,
AssumptionCache *AC,
const Instruction *CxtI,
const DominatorTree *DT);
OverflowResult computeOverflowForUnsignedAdd(Value *LHS, Value *RHS,
- const DataLayout *DL,
+ const DataLayout &DL,
AssumptionCache *AC,
const Instruction *CxtI,
const DominatorTree *DT);
}
/// \brief Fetch the type representing a pointer to an integer value.
- IntegerType* getIntPtrTy(const DataLayout *DL, unsigned AddrSpace = 0) {
- return DL->getIntPtrType(Context, AddrSpace);
+ IntegerType *getIntPtrTy(const DataLayout &DL, unsigned AddrSpace = 0) {
+ return DL.getIntPtrType(Context, AddrSpace);
}
//===--------------------------------------------------------------------===//
"trying to create an alignment assumption on a non-pointer?");
PointerType *PtrTy = cast<PointerType>(PtrValue->getType());
- Type *IntPtrTy = getIntPtrTy(&DL, PtrTy->getAddressSpace());
+ Type *IntPtrTy = getIntPtrTy(DL, PtrTy->getAddressSpace());
Value *PtrIntValue = CreatePtrToInt(PtrValue, IntPtrTy, "ptrint");
Value *Mask = ConstantInt::get(IntPtrTy,
/// This ensures that any pointer<->integer cast has enough bits in the
/// integer and any other cast is a bitcast.
static bool isBitOrNoopPointerCastable(
- Type *SrcTy, ///< The Type from which the value should be cast.
- Type *DestTy, ///< The Type to which the value should be cast.
- const DataLayout *Layout = 0 ///< Optional DataLayout.
- );
+ Type *SrcTy, ///< The Type from which the value should be cast.
+ Type *DestTy, ///< The Type to which the value should be cast.
+ const DataLayout &DL);
/// Returns the opcode necessary to cast Val into Ty using usual casting
/// rules.
) const;
/// @brief Determine if this cast is a no-op cast.
- bool isNoopCast(
- const DataLayout *DL ///< DataLayout to get the Int Ptr type from.
- ) const;
+ ///
+ /// \param DL is the DataLayout to get the Int Ptr type from.
+ bool isNoopCast(const DataLayout &DL) const;
/// Determine how a pair of casts can be eliminated, if they can be at all.
/// This is a helper function for both CastInst and ConstantExpr.
///
/// Test if this value is always a pointer to allocated and suitably aligned
/// memory for a simple load or store.
- bool isDereferenceablePointer(const DataLayout *DL = nullptr) const;
+ bool isDereferenceablePointer(const DataLayout &DL) const;
/// \brief Translate PHI node to its predecessor from the given basic block.
///
public:
const TargetMachine &getTargetMachine() const { return TM; }
- const DataLayout *getDataLayout() const { return DL; }
+ const DataLayout *getDataLayout() const { return TM.getDataLayout(); }
bool isBigEndian() const { return !IsLittleEndian; }
bool isLittleEndian() const { return IsLittleEndian; }
private:
const TargetMachine &TM;
- const DataLayout *DL;
/// True if this is a little endian target.
bool IsLittleEndian;
bool containsGlobalOffset(uint64_t Offset) const;
- bool containsValue(const DataLayout *DL,
+ bool containsValue(const DataLayout &DL,
const DenseMap<GlobalVariable *, uint64_t> &GlobalLayout,
Value *V, uint64_t COffset = 0) const;
};
/// EmitStrLen - Emit a call to the strlen function to the builder, for the
/// specified pointer. Ptr is required to be some pointer type, and the
/// return value has 'intptr_t' type.
- Value *EmitStrLen(Value *Ptr, IRBuilder<> &B, const DataLayout *TD,
+ Value *EmitStrLen(Value *Ptr, IRBuilder<> &B, const DataLayout &DL,
const TargetLibraryInfo *TLI);
/// EmitStrNLen - Emit a call to the strnlen function to the builder, for the
/// specified pointer. Ptr is required to be some pointer type, MaxLen must
/// be of size_t type, and the return value has 'intptr_t' type.
Value *EmitStrNLen(Value *Ptr, Value *MaxLen, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI);
+ const DataLayout &DL, const TargetLibraryInfo *TLI);
/// EmitStrChr - Emit a call to the strchr function to the builder, for the
/// specified pointer and character. Ptr is required to be some pointer type,
/// and the return value has 'i8*' type.
- Value *EmitStrChr(Value *Ptr, char C, IRBuilder<> &B, const DataLayout *TD,
+ Value *EmitStrChr(Value *Ptr, char C, IRBuilder<> &B,
const TargetLibraryInfo *TLI);
/// EmitStrNCmp - Emit a call to the strncmp function to the builder.
Value *EmitStrNCmp(Value *Ptr1, Value *Ptr2, Value *Len, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI);
+ const DataLayout &DL, const TargetLibraryInfo *TLI);
/// EmitStrCpy - Emit a call to the strcpy function to the builder, for the
/// specified pointer arguments.
Value *EmitStrCpy(Value *Dst, Value *Src, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI,
- StringRef Name = "strcpy");
+ const TargetLibraryInfo *TLI, StringRef Name = "strcpy");
/// EmitStrNCpy - Emit a call to the strncpy function to the builder, for the
/// specified pointer arguments and length.
Value *EmitStrNCpy(Value *Dst, Value *Src, Value *Len, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI,
- StringRef Name = "strncpy");
+ const TargetLibraryInfo *TLI, StringRef Name = "strncpy");
/// EmitMemCpyChk - Emit a call to the __memcpy_chk function to the builder.
/// This expects that the Len and ObjSize have type 'intptr_t' and Dst/Src
/// are pointers.
Value *EmitMemCpyChk(Value *Dst, Value *Src, Value *Len, Value *ObjSize,
- IRBuilder<> &B, const DataLayout *TD,
+ IRBuilder<> &B, const DataLayout &DL,
const TargetLibraryInfo *TLI);
/// EmitMemChr - Emit a call to the memchr function. This assumes that Ptr is
/// a pointer, Val is an i32 value, and Len is an 'intptr_t' value.
Value *EmitMemChr(Value *Ptr, Value *Val, Value *Len, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI);
+ const DataLayout &DL, const TargetLibraryInfo *TLI);
/// EmitMemCmp - Emit a call to the memcmp function.
Value *EmitMemCmp(Value *Ptr1, Value *Ptr2, Value *Len, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI);
+ const DataLayout &DL, const TargetLibraryInfo *TLI);
/// EmitUnaryFloatFnCall - Emit a call to the unary function named 'Name'
/// (e.g. 'floor'). This function is known to take a single of type matching
/// EmitPutChar - Emit a call to the putchar function. This assumes that Char
/// is an integer.
- Value *EmitPutChar(Value *Char, IRBuilder<> &B, const DataLayout *TD,
- const TargetLibraryInfo *TLI);
+ Value *EmitPutChar(Value *Char, IRBuilder<> &B, const TargetLibraryInfo *TLI);
/// EmitPutS - Emit a call to the puts function. This assumes that Str is
/// some pointer.
- Value *EmitPutS(Value *Str, IRBuilder<> &B, const DataLayout *TD,
- const TargetLibraryInfo *TLI);
+ Value *EmitPutS(Value *Str, IRBuilder<> &B, const TargetLibraryInfo *TLI);
/// EmitFPutC - Emit a call to the fputc function. This assumes that Char is
/// an i32, and File is a pointer to FILE.
Value *EmitFPutC(Value *Char, Value *File, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI);
+ const TargetLibraryInfo *TLI);
/// EmitFPutS - Emit a call to the puts function. Str is required to be a
/// pointer and File is a pointer to FILE.
- Value *EmitFPutS(Value *Str, Value *File, IRBuilder<> &B, const DataLayout *TD,
+ Value *EmitFPutS(Value *Str, Value *File, IRBuilder<> &B,
const TargetLibraryInfo *TLI);
/// EmitFWrite - Emit a call to the fwrite function. This assumes that Ptr is
/// a pointer, Size is an 'intptr_t', and File is a pointer to FILE.
Value *EmitFWrite(Value *Ptr, Value *Size, Value *File, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI);
+ const DataLayout &DL, const TargetLibraryInfo *TLI);
}
#endif
SmallVectorImpl<ReturnInst*> &Returns,
const char *NameSuffix = "",
ClonedCodeInfo *CodeInfo = nullptr,
- const DataLayout *DL = nullptr,
CloningDirector *Director = nullptr);
SmallVectorImpl<ReturnInst*> &Returns,
const char *NameSuffix = "",
ClonedCodeInfo *CodeInfo = nullptr,
- const DataLayout *DL = nullptr,
Instruction *TheCall = nullptr);
/// InlineFunctionInfo - This class captures the data input to the
///
/// This returns true if it changed the code, note that it can delete
/// instructions in other blocks as well in this block.
-bool SimplifyInstructionsInBlock(BasicBlock *BB, const DataLayout *TD = nullptr,
+bool SimplifyInstructionsInBlock(BasicBlock *BB,
const TargetLibraryInfo *TLI = nullptr);
//===----------------------------------------------------------------------===//
///
/// .. and delete the predecessor corresponding to the '1', this will attempt to
/// recursively fold the 'and' to 0.
-void RemovePredecessorAndSimplify(BasicBlock *BB, BasicBlock *Pred,
- DataLayout *TD = nullptr);
+void RemovePredecessorAndSimplify(BasicBlock *BB, BasicBlock *Pred);
/// MergeBasicBlockIntoOnlyPred - BB is a block with one predecessor and its
/// predecessor is known to have one successor (BB!). Eliminate the edge
/// the basic block that was pointed to.
///
bool SimplifyCFG(BasicBlock *BB, const TargetTransformInfo &TTI,
- unsigned BonusInstThreshold, const DataLayout *TD = nullptr,
- AssumptionCache *AC = nullptr);
+ unsigned BonusInstThreshold, AssumptionCache *AC = nullptr);
/// FlatternCFG - This function is used to flatten a CFG. For
/// example, it uses parallel-and and parallel-or mode to collapse
/// and if a predecessor branches to us and one of our successors, fold the
/// setcc into the predecessor and use logical operations to pick the right
/// destination.
-bool FoldBranchToCommonDest(BranchInst *BI, const DataLayout *DL = nullptr,
- unsigned BonusInstThreshold = 1);
+bool FoldBranchToCommonDest(BranchInst *BI, unsigned BonusInstThreshold = 1);
/// DemoteRegToStack - This function takes a virtual register computed by an
/// Instruction and replaces it with a slot in the stack frame, allocated via
/// and it is more than the alignment of the ultimate object, see if we can
/// increase the alignment of the ultimate object, making this check succeed.
unsigned getOrEnforceKnownAlignment(Value *V, unsigned PrefAlign,
- const DataLayout *TD = nullptr,
- AssumptionCache *AC = nullptr,
+ const DataLayout &DL,
const Instruction *CxtI = nullptr,
+ AssumptionCache *AC = nullptr,
const DominatorTree *DT = nullptr);
/// getKnownAlignment - Try to infer an alignment for the specified pointer.
-static inline unsigned getKnownAlignment(Value *V,
- const DataLayout *TD = nullptr,
- AssumptionCache *AC = nullptr,
+static inline unsigned getKnownAlignment(Value *V, const DataLayout &DL,
const Instruction *CxtI = nullptr,
+ AssumptionCache *AC = nullptr,
const DominatorTree *DT = nullptr) {
- return getOrEnforceKnownAlignment(V, 0, TD, AC, CxtI, DT);
+ return getOrEnforceKnownAlignment(V, 0, DL, CxtI, AC, DT);
}
/// EmitGEPOffset - Given a getelementptr instruction/constantexpr, emit the
/// in the base pointer). Return the result as a signed integer of intptr size.
/// When NoAssumptions is true, no assumptions about index computation not
/// overflowing is made.
-template<typename IRBuilderTy>
-Value *EmitGEPOffset(IRBuilderTy *Builder, const DataLayout &TD, User *GEP,
+template <typename IRBuilderTy>
+Value *EmitGEPOffset(IRBuilderTy *Builder, const DataLayout &DL, User *GEP,
bool NoAssumptions = false) {
GEPOperator *GEPOp = cast<GEPOperator>(GEP);
- Type *IntPtrTy = TD.getIntPtrType(GEP->getType());
+ Type *IntPtrTy = DL.getIntPtrType(GEP->getType());
Value *Result = Constant::getNullValue(IntPtrTy);
// If the GEP is inbounds, we know that none of the addressing operations will
for (User::op_iterator i = GEP->op_begin() + 1, e = GEP->op_end(); i != e;
++i, ++GTI) {
Value *Op = *i;
- uint64_t Size = TD.getTypeAllocSize(GTI.getIndexedType()) & PtrSizeMask;
+ uint64_t Size = DL.getTypeAllocSize(GTI.getIndexedType()) & PtrSizeMask;
if (Constant *OpC = dyn_cast<Constant>(Op)) {
if (OpC->isZeroValue())
continue;
OpC = OpC->getSplatValue();
uint64_t OpValue = cast<ConstantInt>(OpC)->getZExtValue();
- Size = TD.getStructLayout(STy)->getElementOffset(OpValue);
+ Size = DL.getStructLayout(STy)->getElementOffset(OpValue);
if (Size)
Result = Builder->CreateAdd(Result, ConstantInt::get(IntPtrTy, Size),
/// passed into it.
bool simplifyLoop(Loop *L, DominatorTree *DT, LoopInfo *LI, Pass *PP,
AliasAnalysis *AA = nullptr, ScalarEvolution *SE = nullptr,
- const DataLayout *DL = nullptr,
AssumptionCache *AC = nullptr);
/// \brief Put loop into LCSSA form.
/// dominated by the specified block, and that are in the current loop) in
/// reverse depth first order w.r.t the DominatorTree. This allows us to visit
/// uses before definitions, allowing us to sink a loop body in one pass without
-/// iteration. Takes DomTreeNode, AliasAnalysis, LoopInfo, DominatorTree,
-/// DataLayout, TargetLibraryInfo, Loop, AliasSet information for all
-/// instructions of the loop and loop safety information as arguments.
-/// It returns changed status.
+/// iteration. Takes DomTreeNode, AliasAnalysis, LoopInfo, DominatorTree,
+/// DataLayout, TargetLibraryInfo, Loop, AliasSet information for all
+/// instructions of the loop and loop safety information as arguments.
+/// It returns changed status.
bool sinkRegion(DomTreeNode *, AliasAnalysis *, LoopInfo *, DominatorTree *,
- const DataLayout *, TargetLibraryInfo *, Loop *,
- AliasSetTracker *, LICMSafetyInfo *);
+ TargetLibraryInfo *, Loop *, AliasSetTracker *,
+ LICMSafetyInfo *);
/// \brief Walk the specified region of the CFG (defined by all blocks
/// dominated by the specified block, and that are in the current loop) in depth
/// TargetLibraryInfo, Loop, AliasSet information for all instructions of the
/// loop and loop safety information as arguments. It returns changed status.
bool hoistRegion(DomTreeNode *, AliasAnalysis *, LoopInfo *, DominatorTree *,
- const DataLayout *, TargetLibraryInfo *, Loop *,
- AliasSetTracker *, LICMSafetyInfo *);
+ TargetLibraryInfo *, Loop *, AliasSetTracker *,
+ LICMSafetyInfo *);
/// \brief Try to promote memory values to scalars by sinking stores out of
/// the loop and moving loads to before the loop. We do this by looping over
/// is unknown) by passing true for OnlyLowerUnknownSize.
class FortifiedLibCallSimplifier {
private:
- const DataLayout *DL;
const TargetLibraryInfo *TLI;
bool OnlyLowerUnknownSize;
public:
- FortifiedLibCallSimplifier(const DataLayout *DL, const TargetLibraryInfo *TLI,
+ FortifiedLibCallSimplifier(const TargetLibraryInfo *TLI,
bool OnlyLowerUnknownSize = false);
/// \brief Take the given call instruction and return a more
class LibCallSimplifier {
private:
FortifiedLibCallSimplifier FortifiedSimplifier;
- const DataLayout *DL;
+ const DataLayout &DL;
const TargetLibraryInfo *TLI;
bool UnsafeFPShrink;
function_ref<void(Instruction *, Value *)> Replacer;
void replaceAllUsesWith(Instruction *I, Value *With);
public:
- LibCallSimplifier(const DataLayout *TD, const TargetLibraryInfo *TLI,
+ LibCallSimplifier(const DataLayout &DL, const TargetLibraryInfo *TLI,
function_ref<void(Instruction *, Value *)> Replacer =
&replaceAllUsesWithDefault);
AliasAnalysis::callCapturesBefore(const Instruction *I,
const AliasAnalysis::Location &MemLoc,
DominatorTree *DT) {
- if (!DT || !DL) return AliasAnalysis::ModRef;
+ if (!DT)
+ return AliasAnalysis::ModRef;
- const Value *Object = GetUnderlyingObject(MemLoc.Ptr, DL);
+ const Value *Object = GetUnderlyingObject(MemLoc.Ptr, *DL);
if (!isIdentifiedObject(Object) || isa<GlobalValue>(Object) ||
isa<Constant>(Object))
return AliasAnalysis::ModRef;
const TargetLibraryInfo &TLI,
bool RoundToAlign = false) {
uint64_t Size;
- if (getObjectSize(V, Size, &DL, &TLI, RoundToAlign))
+ if (getObjectSize(V, Size, DL, &TLI, RoundToAlign))
return Size;
return AliasAnalysis::UnknownSize;
}
case Instruction::Or:
// X|C == X+C if all the bits in C are unset in X. Otherwise we can't
// analyze it.
- if (!MaskedValueIsZero(BOp->getOperand(0), RHSC->getValue(), &DL, 0, AC,
+ if (!MaskedValueIsZero(BOp->getOperand(0), RHSC->getValue(), DL, 0, AC,
BOp, DT))
break;
// FALL THROUGH.
static const Value *
DecomposeGEPExpression(const Value *V, int64_t &BaseOffs,
SmallVectorImpl<VariableGEPIndex> &VarIndices,
- bool &MaxLookupReached, const DataLayout *DL,
+ bool &MaxLookupReached, const DataLayout &DL,
AssumptionCache *AC, DominatorTree *DT) {
// Limit recursion depth to limit compile time in crazy cases.
unsigned MaxLookup = MaxLookupSearchDepth;
if (!GEPOp->getOperand(0)->getType()->getPointerElementType()->isSized())
return V;
- // If we are lacking DataLayout information, we can't compute the offets of
- // elements computed by GEPs. However, we can handle bitcast equivalent
- // GEPs.
- if (!DL) {
- if (!GEPOp->hasAllZeroIndices())
- return V;
- V = GEPOp->getOperand(0);
- continue;
- }
-
unsigned AS = GEPOp->getPointerAddressSpace();
// Walk the indices of the GEP, accumulating them into BaseOff/VarIndices.
gep_type_iterator GTI = gep_type_begin(GEPOp);
unsigned FieldNo = cast<ConstantInt>(Index)->getZExtValue();
if (FieldNo == 0) continue;
- BaseOffs += DL->getStructLayout(STy)->getElementOffset(FieldNo);
+ BaseOffs += DL.getStructLayout(STy)->getElementOffset(FieldNo);
continue;
}
// For an array/pointer, add the element offset, explicitly scaled.
if (ConstantInt *CIdx = dyn_cast<ConstantInt>(Index)) {
if (CIdx->isZero()) continue;
- BaseOffs += DL->getTypeAllocSize(*GTI)*CIdx->getSExtValue();
+ BaseOffs += DL.getTypeAllocSize(*GTI) * CIdx->getSExtValue();
continue;
}
- uint64_t Scale = DL->getTypeAllocSize(*GTI);
+ uint64_t Scale = DL.getTypeAllocSize(*GTI);
ExtensionKind Extension = EK_NotExtended;
// If the integer type is smaller than the pointer size, it is implicitly
// sign extended to pointer size.
unsigned Width = Index->getType()->getIntegerBitWidth();
- if (DL->getPointerSizeInBits(AS) > Width)
+ if (DL.getPointerSizeInBits(AS) > Width)
Extension = EK_SignExt;
// Use GetLinearExpression to decompose the index into a C1*V+C2 form.
APInt IndexScale(Width, 0), IndexOffset(Width, 0);
- Index = GetLinearExpression(Index, IndexScale, IndexOffset, Extension,
- *DL, 0, AC, DT);
+ Index = GetLinearExpression(Index, IndexScale, IndexOffset, Extension, DL,
+ 0, AC, DT);
// The GEP index scale ("Scale") scales C1*V+C2, yielding (C1*V+C2)*Scale.
// This gives us an aggregate computation of (C1*Scale)*V + C2*Scale.
// Make sure that we have a scale that makes sense for this target's
// pointer size.
- if (unsigned ShiftBits = 64 - DL->getPointerSizeInBits(AS)) {
+ if (unsigned ShiftBits = 64 - DL.getPointerSizeInBits(AS)) {
Scale <<= ShiftBits;
Scale = (int64_t)Scale >> ShiftBits;
}
SmallVector<const Value *, 16> Worklist;
Worklist.push_back(Loc.Ptr);
do {
- const Value *V = GetUnderlyingObject(Worklist.pop_back_val(), DL);
+ const Value *V = GetUnderlyingObject(Worklist.pop_back_val(), *DL);
if (!Visited.insert(V).second) {
Visited.clear();
return AliasAnalysis::pointsToConstantMemory(Loc, OrLocal);
assert(notDifferentParent(CS.getInstruction(), Loc.Ptr) &&
"AliasAnalysis query involving multiple functions!");
- const Value *Object = GetUnderlyingObject(Loc.Ptr, DL);
+ const Value *Object = GetUnderlyingObject(Loc.Ptr, *DL);
// If this is a tail call and Loc.Ptr points to a stack location, we know that
// the tail call cannot access or modify the local stack.
SmallVector<VariableGEPIndex, 4> GEP2VariableIndices;
const Value *GEP2BasePtr =
DecomposeGEPExpression(GEP2, GEP2BaseOffset, GEP2VariableIndices,
- GEP2MaxLookupReached, DL, AC2, DT);
+ GEP2MaxLookupReached, *DL, AC2, DT);
const Value *GEP1BasePtr =
DecomposeGEPExpression(GEP1, GEP1BaseOffset, GEP1VariableIndices,
- GEP1MaxLookupReached, DL, AC1, DT);
+ GEP1MaxLookupReached, *DL, AC1, DT);
// DecomposeGEPExpression and GetUnderlyingObject should return the
// same result except when DecomposeGEPExpression has no DataLayout.
if (GEP1BasePtr != UnderlyingV1 || GEP2BasePtr != UnderlyingV2) {
// about the relation of the resulting pointer.
const Value *GEP1BasePtr =
DecomposeGEPExpression(GEP1, GEP1BaseOffset, GEP1VariableIndices,
- GEP1MaxLookupReached, DL, AC1, DT);
+ GEP1MaxLookupReached, *DL, AC1, DT);
int64_t GEP2BaseOffset;
bool GEP2MaxLookupReached;
SmallVector<VariableGEPIndex, 4> GEP2VariableIndices;
const Value *GEP2BasePtr =
DecomposeGEPExpression(GEP2, GEP2BaseOffset, GEP2VariableIndices,
- GEP2MaxLookupReached, DL, AC2, DT);
+ GEP2MaxLookupReached, *DL, AC2, DT);
// DecomposeGEPExpression and GetUnderlyingObject should return the
// same result except when DecomposeGEPExpression has no DataLayout.
const Value *GEP1BasePtr =
DecomposeGEPExpression(GEP1, GEP1BaseOffset, GEP1VariableIndices,
- GEP1MaxLookupReached, DL, AC1, DT);
+ GEP1MaxLookupReached, *DL, AC1, DT);
// DecomposeGEPExpression and GetUnderlyingObject should return the
// same result except when DecomposeGEPExpression has no DataLayout.
const Value *V = GEP1VariableIndices[i].V;
bool SignKnownZero, SignKnownOne;
- ComputeSignBit(const_cast<Value *>(V), SignKnownZero, SignKnownOne, DL,
+ ComputeSignBit(const_cast<Value *>(V), SignKnownZero, SignKnownOne, *DL,
0, AC1, nullptr, DT);
// Zero-extension widens the variable, and so forces the sign
return NoAlias; // Scalars cannot alias each other
// Figure out what objects these things are pointing to if we can.
- const Value *O1 = GetUnderlyingObject(V1, DL, MaxLookupSearchDepth);
- const Value *O2 = GetUnderlyingObject(V2, DL, MaxLookupSearchDepth);
+ const Value *O1 = GetUnderlyingObject(V1, *DL, MaxLookupSearchDepth);
+ const Value *O2 = GetUnderlyingObject(V2, *DL, MaxLookupSearchDepth);
// Null values in the default address space don't point to any object, so they
// don't alias any other pointer.
/// Constant fold bitcast, symbolically evaluating it with DataLayout.
/// This always returns a non-null constant, but it may be a
/// ConstantExpr if unfoldable.
-static Constant *FoldBitCast(Constant *C, Type *DestTy,
- const DataLayout &TD) {
+static Constant *FoldBitCast(Constant *C, Type *DestTy, const DataLayout &DL) {
// Catch the obvious splat cases.
if (C->isNullValue() && !DestTy->isX86_MMXTy())
return Constant::getNullValue(DestTy);
// Now that we know that the input value is a vector of integers, just shift
// and insert them into our result.
- unsigned BitShift = TD.getTypeAllocSizeInBits(SrcEltTy);
+ unsigned BitShift = DL.getTypeAllocSizeInBits(SrcEltTy);
APInt Result(IT->getBitWidth(), 0);
for (unsigned i = 0; i != NumSrcElts; ++i) {
Result <<= BitShift;
- if (TD.isLittleEndian())
+ if (DL.isLittleEndian())
Result |= CDV->getElementAsInteger(NumSrcElts-i-1);
else
Result |= CDV->getElementAsInteger(i);
// vector so the code below can handle it uniformly.
if (isa<ConstantFP>(C) || isa<ConstantInt>(C)) {
Constant *Ops = C; // don't take the address of C!
- return FoldBitCast(ConstantVector::get(Ops), DestTy, TD);
+ return FoldBitCast(ConstantVector::get(Ops), DestTy, DL);
}
// If this is a bitcast from constant vector -> vector, fold it.
Type *DestIVTy =
VectorType::get(IntegerType::get(C->getContext(), FPWidth), NumDstElt);
// Recursively handle this integer conversion, if possible.
- C = FoldBitCast(C, DestIVTy, TD);
+ C = FoldBitCast(C, DestIVTy, DL);
// Finally, IR can handle this now that #elts line up.
return ConstantExpr::getBitCast(C, DestTy);
// of the same size, and that their #elements is not the same. Do the
// conversion here, which depends on whether the input or output has
// more elements.
- bool isLittleEndian = TD.isLittleEndian();
+ bool isLittleEndian = DL.isLittleEndian();
SmallVector<Constant*, 32> Result;
if (NumDstElt < NumSrcElt) {
// Handle: bitcast (<2 x i64> <i64 0, i64 1> to <4 x i32>)
unsigned Ratio = NumDstElt/NumSrcElt;
- unsigned DstBitSize = TD.getTypeSizeInBits(DstEltTy);
+ unsigned DstBitSize = DL.getTypeSizeInBits(DstEltTy);
// Loop over each source value, expanding into multiple results.
for (unsigned i = 0; i != NumSrcElt; ++i) {
/// If this constant is a constant offset from a global, return the global and
/// the constant. Because of constantexprs, this function is recursive.
static bool IsConstantOffsetFromGlobal(Constant *C, GlobalValue *&GV,
- APInt &Offset, const DataLayout &TD) {
+ APInt &Offset, const DataLayout &DL) {
// Trivial case, constant is the global.
if ((GV = dyn_cast<GlobalValue>(C))) {
- unsigned BitWidth = TD.getPointerTypeSizeInBits(GV->getType());
+ unsigned BitWidth = DL.getPointerTypeSizeInBits(GV->getType());
Offset = APInt(BitWidth, 0);
return true;
}
if (CE->getOpcode() == Instruction::PtrToInt ||
CE->getOpcode() == Instruction::BitCast ||
CE->getOpcode() == Instruction::AddrSpaceCast)
- return IsConstantOffsetFromGlobal(CE->getOperand(0), GV, Offset, TD);
+ return IsConstantOffsetFromGlobal(CE->getOperand(0), GV, Offset, DL);
// i32* getelementptr ([5 x i32]* @a, i32 0, i32 5)
GEPOperator *GEP = dyn_cast<GEPOperator>(CE);
if (!GEP)
return false;
- unsigned BitWidth = TD.getPointerTypeSizeInBits(GEP->getType());
+ unsigned BitWidth = DL.getPointerTypeSizeInBits(GEP->getType());
APInt TmpOffset(BitWidth, 0);
// If the base isn't a global+constant, we aren't either.
- if (!IsConstantOffsetFromGlobal(CE->getOperand(0), GV, TmpOffset, TD))
+ if (!IsConstantOffsetFromGlobal(CE->getOperand(0), GV, TmpOffset, DL))
return false;
// Otherwise, add any offset that our operands provide.
- if (!GEP->accumulateConstantOffset(TD, TmpOffset))
+ if (!GEP->accumulateConstantOffset(DL, TmpOffset))
return false;
Offset = TmpOffset;
/// Recursive helper to read bits out of global. C is the constant being copied
/// out of. ByteOffset is an offset into C. CurPtr is the pointer to copy
/// results into and BytesLeft is the number of bytes left in
-/// the CurPtr buffer. TD is the target data.
+/// the CurPtr buffer. DL is the DataLayout.
static bool ReadDataFromGlobal(Constant *C, uint64_t ByteOffset,
unsigned char *CurPtr, unsigned BytesLeft,
- const DataLayout &TD) {
- assert(ByteOffset <= TD.getTypeAllocSize(C->getType()) &&
+ const DataLayout &DL) {
+ assert(ByteOffset <= DL.getTypeAllocSize(C->getType()) &&
"Out of range access");
// If this element is zero or undefined, we can just return since *CurPtr is
for (unsigned i = 0; i != BytesLeft && ByteOffset != IntBytes; ++i) {
int n = ByteOffset;
- if (!TD.isLittleEndian())
+ if (!DL.isLittleEndian())
n = IntBytes - n - 1;
CurPtr[i] = (unsigned char)(Val >> (n * 8));
++ByteOffset;
if (ConstantFP *CFP = dyn_cast<ConstantFP>(C)) {
if (CFP->getType()->isDoubleTy()) {
- C = FoldBitCast(C, Type::getInt64Ty(C->getContext()), TD);
- return ReadDataFromGlobal(C, ByteOffset, CurPtr, BytesLeft, TD);
+ C = FoldBitCast(C, Type::getInt64Ty(C->getContext()), DL);
+ return ReadDataFromGlobal(C, ByteOffset, CurPtr, BytesLeft, DL);
}
if (CFP->getType()->isFloatTy()){
- C = FoldBitCast(C, Type::getInt32Ty(C->getContext()), TD);
- return ReadDataFromGlobal(C, ByteOffset, CurPtr, BytesLeft, TD);
+ C = FoldBitCast(C, Type::getInt32Ty(C->getContext()), DL);
+ return ReadDataFromGlobal(C, ByteOffset, CurPtr, BytesLeft, DL);
}
if (CFP->getType()->isHalfTy()){
- C = FoldBitCast(C, Type::getInt16Ty(C->getContext()), TD);
- return ReadDataFromGlobal(C, ByteOffset, CurPtr, BytesLeft, TD);
+ C = FoldBitCast(C, Type::getInt16Ty(C->getContext()), DL);
+ return ReadDataFromGlobal(C, ByteOffset, CurPtr, BytesLeft, DL);
}
return false;
}
if (ConstantStruct *CS = dyn_cast<ConstantStruct>(C)) {
- const StructLayout *SL = TD.getStructLayout(CS->getType());
+ const StructLayout *SL = DL.getStructLayout(CS->getType());
unsigned Index = SL->getElementContainingOffset(ByteOffset);
uint64_t CurEltOffset = SL->getElementOffset(Index);
ByteOffset -= CurEltOffset;
while (1) {
// If the element access is to the element itself and not to tail padding,
// read the bytes from the element.
- uint64_t EltSize = TD.getTypeAllocSize(CS->getOperand(Index)->getType());
+ uint64_t EltSize = DL.getTypeAllocSize(CS->getOperand(Index)->getType());
if (ByteOffset < EltSize &&
!ReadDataFromGlobal(CS->getOperand(Index), ByteOffset, CurPtr,
- BytesLeft, TD))
+ BytesLeft, DL))
return false;
++Index;
if (isa<ConstantArray>(C) || isa<ConstantVector>(C) ||
isa<ConstantDataSequential>(C)) {
Type *EltTy = C->getType()->getSequentialElementType();
- uint64_t EltSize = TD.getTypeAllocSize(EltTy);
+ uint64_t EltSize = DL.getTypeAllocSize(EltTy);
uint64_t Index = ByteOffset / EltSize;
uint64_t Offset = ByteOffset - Index * EltSize;
uint64_t NumElts;
for (; Index != NumElts; ++Index) {
if (!ReadDataFromGlobal(C->getAggregateElement(Index), Offset, CurPtr,
- BytesLeft, TD))
+ BytesLeft, DL))
return false;
uint64_t BytesWritten = EltSize - Offset;
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
if (CE->getOpcode() == Instruction::IntToPtr &&
- CE->getOperand(0)->getType() == TD.getIntPtrType(CE->getType())) {
+ CE->getOperand(0)->getType() == DL.getIntPtrType(CE->getType())) {
return ReadDataFromGlobal(CE->getOperand(0), ByteOffset, CurPtr,
- BytesLeft, TD);
+ BytesLeft, DL);
}
}
}
static Constant *FoldReinterpretLoadFromConstPtr(Constant *C,
- const DataLayout &TD) {
+ const DataLayout &DL) {
PointerType *PTy = cast<PointerType>(C->getType());
Type *LoadTy = PTy->getElementType();
IntegerType *IntType = dyn_cast<IntegerType>(LoadTy);
MapTy = Type::getInt64PtrTy(C->getContext(), AS);
else if (LoadTy->isVectorTy()) {
MapTy = PointerType::getIntNPtrTy(C->getContext(),
- TD.getTypeAllocSizeInBits(LoadTy),
- AS);
+ DL.getTypeAllocSizeInBits(LoadTy), AS);
} else
return nullptr;
- C = FoldBitCast(C, MapTy, TD);
- if (Constant *Res = FoldReinterpretLoadFromConstPtr(C, TD))
- return FoldBitCast(Res, LoadTy, TD);
+ C = FoldBitCast(C, MapTy, DL);
+ if (Constant *Res = FoldReinterpretLoadFromConstPtr(C, DL))
+ return FoldBitCast(Res, LoadTy, DL);
return nullptr;
}
GlobalValue *GVal;
APInt Offset;
- if (!IsConstantOffsetFromGlobal(C, GVal, Offset, TD))
+ if (!IsConstantOffsetFromGlobal(C, GVal, Offset, DL))
return nullptr;
GlobalVariable *GV = dyn_cast<GlobalVariable>(GVal);
// If we're not accessing anything in this constant, the result is undefined.
if (Offset.getZExtValue() >=
- TD.getTypeAllocSize(GV->getInitializer()->getType()))
+ DL.getTypeAllocSize(GV->getInitializer()->getType()))
return UndefValue::get(IntType);
unsigned char RawBytes[32] = {0};
if (!ReadDataFromGlobal(GV->getInitializer(), Offset.getZExtValue(), RawBytes,
- BytesLoaded, TD))
+ BytesLoaded, DL))
return nullptr;
APInt ResultVal = APInt(IntType->getBitWidth(), 0);
- if (TD.isLittleEndian()) {
+ if (DL.isLittleEndian()) {
ResultVal = RawBytes[BytesLoaded - 1];
for (unsigned i = 1; i != BytesLoaded; ++i) {
ResultVal <<= 8;
}
static Constant *ConstantFoldLoadThroughBitcast(ConstantExpr *CE,
- const DataLayout *DL) {
- if (!DL)
- return nullptr;
+ const DataLayout &DL) {
auto *DestPtrTy = dyn_cast<PointerType>(CE->getType());
if (!DestPtrTy)
return nullptr;
// If the type sizes are the same and a cast is legal, just directly
// cast the constant.
- if (DL->getTypeSizeInBits(DestTy) == DL->getTypeSizeInBits(SrcTy)) {
+ if (DL.getTypeSizeInBits(DestTy) == DL.getTypeSizeInBits(SrcTy)) {
Instruction::CastOps Cast = Instruction::BitCast;
// If we are going from a pointer to int or vice versa, we spell the cast
// differently.
/// Return the value that a load from C would produce if it is constant and
/// determinable. If this is not determinable, return null.
Constant *llvm::ConstantFoldLoadFromConstPtr(Constant *C,
- const DataLayout *TD) {
+ const DataLayout &DL) {
// First, try the easy cases:
if (GlobalVariable *GV = dyn_cast<GlobalVariable>(C))
if (GV->isConstant() && GV->hasDefinitiveInitializer())
}
if (CE->getOpcode() == Instruction::BitCast)
- if (Constant *LoadedC = ConstantFoldLoadThroughBitcast(CE, TD))
+ if (Constant *LoadedC = ConstantFoldLoadThroughBitcast(CE, DL))
return LoadedC;
// Instead of loading constant c string, use corresponding integer value
// directly if string length is small enough.
StringRef Str;
- if (TD && getConstantStringInfo(CE, Str) && !Str.empty()) {
+ if (getConstantStringInfo(CE, Str) && !Str.empty()) {
unsigned StrLen = Str.size();
Type *Ty = cast<PointerType>(CE->getType())->getElementType();
unsigned NumBits = Ty->getPrimitiveSizeInBits();
(isa<IntegerType>(Ty) || Ty->isFloatingPointTy())) {
APInt StrVal(NumBits, 0);
APInt SingleChar(NumBits, 0);
- if (TD->isLittleEndian()) {
+ if (DL.isLittleEndian()) {
for (signed i = StrLen-1; i >= 0; i--) {
SingleChar = (uint64_t) Str[i] & UCHAR_MAX;
StrVal = (StrVal << 8) | SingleChar;
// If this load comes from anywhere in a constant global, and if the global
// is all undef or zero, we know what it loads.
if (GlobalVariable *GV =
- dyn_cast<GlobalVariable>(GetUnderlyingObject(CE, TD))) {
+ dyn_cast<GlobalVariable>(GetUnderlyingObject(CE, DL))) {
if (GV->isConstant() && GV->hasDefinitiveInitializer()) {
Type *ResTy = cast<PointerType>(C->getType())->getElementType();
if (GV->getInitializer()->isNullValue())
}
// Try hard to fold loads from bitcasted strange and non-type-safe things.
- if (TD)
- return FoldReinterpretLoadFromConstPtr(CE, *TD);
- return nullptr;
+ return FoldReinterpretLoadFromConstPtr(CE, DL);
}
-static Constant *ConstantFoldLoadInst(const LoadInst *LI, const DataLayout *TD){
+static Constant *ConstantFoldLoadInst(const LoadInst *LI,
+ const DataLayout &DL) {
if (LI->isVolatile()) return nullptr;
if (Constant *C = dyn_cast<Constant>(LI->getOperand(0)))
- return ConstantFoldLoadFromConstPtr(C, TD);
+ return ConstantFoldLoadFromConstPtr(C, DL);
return nullptr;
}
/// these together. If target data info is available, it is provided as DL,
/// otherwise DL is null.
static Constant *SymbolicallyEvaluateBinop(unsigned Opc, Constant *Op0,
- Constant *Op1, const DataLayout *DL){
+ Constant *Op1,
+ const DataLayout &DL) {
// SROA
// Fold (and 0xffffffff00000000, (shl x, 32)) -> shl.
// Fold (lshr (or X, Y), 32) -> (lshr [X/Y], 32) if one doesn't contribute
// bits.
-
- if (Opc == Instruction::And && DL) {
- unsigned BitWidth = DL->getTypeSizeInBits(Op0->getType()->getScalarType());
+ if (Opc == Instruction::And) {
+ unsigned BitWidth = DL.getTypeSizeInBits(Op0->getType()->getScalarType());
APInt KnownZero0(BitWidth, 0), KnownOne0(BitWidth, 0);
APInt KnownZero1(BitWidth, 0), KnownOne1(BitWidth, 0);
computeKnownBits(Op0, KnownZero0, KnownOne0, DL);
// If the constant expr is something like &A[123] - &A[4].f, fold this into a
// constant. This happens frequently when iterating over a global array.
- if (Opc == Instruction::Sub && DL) {
+ if (Opc == Instruction::Sub) {
GlobalValue *GV1, *GV2;
APInt Offs1, Offs2;
- if (IsConstantOffsetFromGlobal(Op0, GV1, Offs1, *DL))
- if (IsConstantOffsetFromGlobal(Op1, GV2, Offs2, *DL) &&
- GV1 == GV2) {
- unsigned OpSize = DL->getTypeSizeInBits(Op0->getType());
+ if (IsConstantOffsetFromGlobal(Op0, GV1, Offs1, DL))
+ if (IsConstantOffsetFromGlobal(Op1, GV2, Offs2, DL) && GV1 == GV2) {
+ unsigned OpSize = DL.getTypeSizeInBits(Op0->getType());
// (&GV+C1) - (&GV+C2) -> C1-C2, pointer arithmetic cannot overflow.
// PtrToInt may change the bitwidth so we have convert to the right size
/// If array indices are not pointer-sized integers, explicitly cast them so
/// that they aren't implicitly casted by the getelementptr.
-static Constant *CastGEPIndices(ArrayRef<Constant *> Ops,
- Type *ResultTy, const DataLayout *TD,
+static Constant *CastGEPIndices(ArrayRef<Constant *> Ops, Type *ResultTy,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI) {
- if (!TD)
- return nullptr;
-
- Type *IntPtrTy = TD->getIntPtrType(ResultTy);
+ Type *IntPtrTy = DL.getIntPtrType(ResultTy);
bool Any = false;
SmallVector<Constant*, 32> NewIdxs;
Constant *C = ConstantExpr::getGetElementPtr(Ops[0], NewIdxs);
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
- if (Constant *Folded = ConstantFoldConstantExpression(CE, TD, TLI))
+ if (Constant *Folded = ConstantFoldConstantExpression(CE, DL, TLI))
C = Folded;
}
/// If we can symbolically evaluate the GEP constant expression, do so.
static Constant *SymbolicallyEvaluateGEP(ArrayRef<Constant *> Ops,
- Type *ResultTy, const DataLayout *TD,
+ Type *ResultTy, const DataLayout &DL,
const TargetLibraryInfo *TLI) {
Constant *Ptr = Ops[0];
- if (!TD || !Ptr->getType()->getPointerElementType()->isSized() ||
+ if (!Ptr->getType()->getPointerElementType()->isSized() ||
!Ptr->getType()->isPointerTy())
return nullptr;
- Type *IntPtrTy = TD->getIntPtrType(Ptr->getType());
+ Type *IntPtrTy = DL.getIntPtrType(Ptr->getType());
Type *ResultElementTy = ResultTy->getPointerElementType();
// If this is a constant expr gep that is effectively computing an
Res = ConstantExpr::getSub(Res, CE->getOperand(1));
Res = ConstantExpr::getIntToPtr(Res, ResultTy);
if (ConstantExpr *ResCE = dyn_cast<ConstantExpr>(Res))
- Res = ConstantFoldConstantExpression(ResCE, TD, TLI);
+ Res = ConstantFoldConstantExpression(ResCE, DL, TLI);
return Res;
}
}
return nullptr;
}
- unsigned BitWidth = TD->getTypeSizeInBits(IntPtrTy);
+ unsigned BitWidth = DL.getTypeSizeInBits(IntPtrTy);
APInt Offset =
- APInt(BitWidth, TD->getIndexedOffset(Ptr->getType(),
- makeArrayRef((Value *const*)
- Ops.data() + 1,
- Ops.size() - 1)));
+ APInt(BitWidth,
+ DL.getIndexedOffset(
+ Ptr->getType(),
+ makeArrayRef((Value * const *)Ops.data() + 1, Ops.size() - 1)));
Ptr = StripPtrCastKeepAS(Ptr);
// If this is a GEP of a GEP, fold it all into a single GEP.
break;
Ptr = cast<Constant>(GEP->getOperand(0));
- Offset += APInt(BitWidth,
- TD->getIndexedOffset(Ptr->getType(), NestedOps));
+ Offset += APInt(BitWidth, DL.getIndexedOffset(Ptr->getType(), NestedOps));
Ptr = StripPtrCastKeepAS(Ptr);
}
}
// Determine which element of the array the offset points into.
- APInt ElemSize(BitWidth, TD->getTypeAllocSize(ATy->getElementType()));
+ APInt ElemSize(BitWidth, DL.getTypeAllocSize(ATy->getElementType()));
if (ElemSize == 0)
// The element size is 0. This may be [0 x Ty]*, so just use a zero
// index for this level and proceed to the next level to see if it can
// can't re-form this GEP in a regular form, so bail out. The pointer
// operand likely went through casts that are necessary to make the GEP
// sensible.
- const StructLayout &SL = *TD->getStructLayout(STy);
+ const StructLayout &SL = *DL.getStructLayout(STy);
if (Offset.uge(SL.getSizeInBytes()))
break;
// If we ended up indexing a member with a type that doesn't match
// the type of what the original indices indexed, add a cast.
if (Ty != ResultElementTy)
- C = FoldBitCast(C, ResultTy, *TD);
+ C = FoldBitCast(C, ResultTy, DL);
return C;
}
/// Note that this fails if not all of the operands are constant. Otherwise,
/// this function can only fail when attempting to fold instructions like loads
/// and stores, which have no constant expression form.
-Constant *llvm::ConstantFoldInstruction(Instruction *I,
- const DataLayout *TD,
+Constant *llvm::ConstantFoldInstruction(Instruction *I, const DataLayout &DL,
const TargetLibraryInfo *TLI) {
// Handle PHI nodes quickly here...
if (PHINode *PN = dyn_cast<PHINode>(I)) {
return nullptr;
// Fold the PHI's operands.
if (ConstantExpr *NewC = dyn_cast<ConstantExpr>(C))
- C = ConstantFoldConstantExpression(NewC, TD, TLI);
+ C = ConstantFoldConstantExpression(NewC, DL, TLI);
// If the incoming value is a different constant to
// the one we saw previously, then give up.
if (CommonValue && C != CommonValue)
// Fold the Instruction's operands.
if (ConstantExpr *NewCE = dyn_cast<ConstantExpr>(Op))
- Op = ConstantFoldConstantExpression(NewCE, TD, TLI);
+ Op = ConstantFoldConstantExpression(NewCE, DL, TLI);
Ops.push_back(Op);
}
if (const CmpInst *CI = dyn_cast<CmpInst>(I))
return ConstantFoldCompareInstOperands(CI->getPredicate(), Ops[0], Ops[1],
- TD, TLI);
+ DL, TLI);
if (const LoadInst *LI = dyn_cast<LoadInst>(I))
- return ConstantFoldLoadInst(LI, TD);
+ return ConstantFoldLoadInst(LI, DL);
if (InsertValueInst *IVI = dyn_cast<InsertValueInst>(I)) {
return ConstantExpr::getInsertValue(
EVI->getIndices());
}
- return ConstantFoldInstOperands(I->getOpcode(), I->getType(), Ops, TD, TLI);
+ return ConstantFoldInstOperands(I->getOpcode(), I->getType(), Ops, DL, TLI);
}
static Constant *
-ConstantFoldConstantExpressionImpl(const ConstantExpr *CE, const DataLayout *TD,
+ConstantFoldConstantExpressionImpl(const ConstantExpr *CE, const DataLayout &DL,
const TargetLibraryInfo *TLI,
SmallPtrSetImpl<ConstantExpr *> &FoldedOps) {
SmallVector<Constant *, 8> Ops;
// a ConstantExpr, we don't have to process it again.
if (ConstantExpr *NewCE = dyn_cast<ConstantExpr>(NewC)) {
if (FoldedOps.insert(NewCE).second)
- NewC = ConstantFoldConstantExpressionImpl(NewCE, TD, TLI, FoldedOps);
+ NewC = ConstantFoldConstantExpressionImpl(NewCE, DL, TLI, FoldedOps);
}
Ops.push_back(NewC);
}
if (CE->isCompare())
return ConstantFoldCompareInstOperands(CE->getPredicate(), Ops[0], Ops[1],
- TD, TLI);
- return ConstantFoldInstOperands(CE->getOpcode(), CE->getType(), Ops, TD, TLI);
+ DL, TLI);
+ return ConstantFoldInstOperands(CE->getOpcode(), CE->getType(), Ops, DL, TLI);
}
/// Attempt to fold the constant expression
/// using the specified DataLayout. If successful, the constant result is
/// result is returned, if not, null is returned.
Constant *llvm::ConstantFoldConstantExpression(const ConstantExpr *CE,
- const DataLayout *TD,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI) {
SmallPtrSet<ConstantExpr *, 4> FoldedOps;
- return ConstantFoldConstantExpressionImpl(CE, TD, TLI, FoldedOps);
+ return ConstantFoldConstantExpressionImpl(CE, DL, TLI, FoldedOps);
}
/// Attempt to constant fold an instruction with the
///
Constant *llvm::ConstantFoldInstOperands(unsigned Opcode, Type *DestTy,
ArrayRef<Constant *> Ops,
- const DataLayout *TD,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI) {
// Handle easy binops first.
if (Instruction::isBinaryOp(Opcode)) {
if (isa<ConstantExpr>(Ops[0]) || isa<ConstantExpr>(Ops[1])) {
- if (Constant *C = SymbolicallyEvaluateBinop(Opcode, Ops[0], Ops[1], TD))
+ if (Constant *C = SymbolicallyEvaluateBinop(Opcode, Ops[0], Ops[1], DL))
return C;
}
// If the input is a inttoptr, eliminate the pair. This requires knowing
// the width of a pointer, so it can't be done in ConstantExpr::getCast.
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ops[0])) {
- if (TD && CE->getOpcode() == Instruction::IntToPtr) {
+ if (CE->getOpcode() == Instruction::IntToPtr) {
Constant *Input = CE->getOperand(0);
unsigned InWidth = Input->getType()->getScalarSizeInBits();
- unsigned PtrWidth = TD->getPointerTypeSizeInBits(CE->getType());
+ unsigned PtrWidth = DL.getPointerTypeSizeInBits(CE->getType());
if (PtrWidth < InWidth) {
Constant *Mask =
ConstantInt::get(CE->getContext(),
// This requires knowing the width of a pointer, so it can't be done in
// ConstantExpr::getCast.
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ops[0])) {
- if (TD && CE->getOpcode() == Instruction::PtrToInt) {
+ if (CE->getOpcode() == Instruction::PtrToInt) {
Constant *SrcPtr = CE->getOperand(0);
- unsigned SrcPtrSize = TD->getPointerTypeSizeInBits(SrcPtr->getType());
+ unsigned SrcPtrSize = DL.getPointerTypeSizeInBits(SrcPtr->getType());
unsigned MidIntSize = CE->getType()->getScalarSizeInBits();
if (MidIntSize >= SrcPtrSize) {
unsigned SrcAS = SrcPtr->getType()->getPointerAddressSpace();
if (SrcAS == DestTy->getPointerAddressSpace())
- return FoldBitCast(CE->getOperand(0), DestTy, *TD);
+ return FoldBitCast(CE->getOperand(0), DestTy, DL);
}
}
}
case Instruction::AddrSpaceCast:
return ConstantExpr::getCast(Opcode, Ops[0], DestTy);
case Instruction::BitCast:
- if (TD)
- return FoldBitCast(Ops[0], DestTy, *TD);
- return ConstantExpr::getBitCast(Ops[0], DestTy);
+ return FoldBitCast(Ops[0], DestTy, DL);
case Instruction::Select:
return ConstantExpr::getSelect(Ops[0], Ops[1], Ops[2]);
case Instruction::ExtractElement:
case Instruction::ShuffleVector:
return ConstantExpr::getShuffleVector(Ops[0], Ops[1], Ops[2]);
case Instruction::GetElementPtr:
- if (Constant *C = CastGEPIndices(Ops, DestTy, TD, TLI))
+ if (Constant *C = CastGEPIndices(Ops, DestTy, DL, TLI))
return C;
- if (Constant *C = SymbolicallyEvaluateGEP(Ops, DestTy, TD, TLI))
+ if (Constant *C = SymbolicallyEvaluateGEP(Ops, DestTy, DL, TLI))
return C;
return ConstantExpr::getGetElementPtr(Ops[0], Ops.slice(1));
/// returns a constant expression of the specified operands.
Constant *llvm::ConstantFoldCompareInstOperands(unsigned Predicate,
Constant *Ops0, Constant *Ops1,
- const DataLayout *TD,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI) {
// fold: icmp (inttoptr x), null -> icmp x, 0
// fold: icmp (ptrtoint x), 0 -> icmp x, null
// fold: icmp (inttoptr x), (inttoptr y) -> icmp trunc/zext x, trunc/zext y
// fold: icmp (ptrtoint x), (ptrtoint y) -> icmp x, y
//
- // ConstantExpr::getCompare cannot do this, because it doesn't have TD
+ // FIXME: The following comment is out of data and the DataLayout is here now.
+ // ConstantExpr::getCompare cannot do this, because it doesn't have DL
// around to know if bit truncation is happening.
if (ConstantExpr *CE0 = dyn_cast<ConstantExpr>(Ops0)) {
- if (TD && Ops1->isNullValue()) {
+ if (Ops1->isNullValue()) {
if (CE0->getOpcode() == Instruction::IntToPtr) {
- Type *IntPtrTy = TD->getIntPtrType(CE0->getType());
+ Type *IntPtrTy = DL.getIntPtrType(CE0->getType());
// Convert the integer value to the right size to ensure we get the
// proper extension or truncation.
Constant *C = ConstantExpr::getIntegerCast(CE0->getOperand(0),
IntPtrTy, false);
Constant *Null = Constant::getNullValue(C->getType());
- return ConstantFoldCompareInstOperands(Predicate, C, Null, TD, TLI);
+ return ConstantFoldCompareInstOperands(Predicate, C, Null, DL, TLI);
}
// Only do this transformation if the int is intptrty in size, otherwise
// there is a truncation or extension that we aren't modeling.
if (CE0->getOpcode() == Instruction::PtrToInt) {
- Type *IntPtrTy = TD->getIntPtrType(CE0->getOperand(0)->getType());
+ Type *IntPtrTy = DL.getIntPtrType(CE0->getOperand(0)->getType());
if (CE0->getType() == IntPtrTy) {
Constant *C = CE0->getOperand(0);
Constant *Null = Constant::getNullValue(C->getType());
- return ConstantFoldCompareInstOperands(Predicate, C, Null, TD, TLI);
+ return ConstantFoldCompareInstOperands(Predicate, C, Null, DL, TLI);
}
}
}
if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(Ops1)) {
- if (TD && CE0->getOpcode() == CE1->getOpcode()) {
+ if (CE0->getOpcode() == CE1->getOpcode()) {
if (CE0->getOpcode() == Instruction::IntToPtr) {
- Type *IntPtrTy = TD->getIntPtrType(CE0->getType());
+ Type *IntPtrTy = DL.getIntPtrType(CE0->getType());
// Convert the integer value to the right size to ensure we get the
// proper extension or truncation.
IntPtrTy, false);
Constant *C1 = ConstantExpr::getIntegerCast(CE1->getOperand(0),
IntPtrTy, false);
- return ConstantFoldCompareInstOperands(Predicate, C0, C1, TD, TLI);
+ return ConstantFoldCompareInstOperands(Predicate, C0, C1, DL, TLI);
}
// Only do this transformation if the int is intptrty in size, otherwise
// there is a truncation or extension that we aren't modeling.
if (CE0->getOpcode() == Instruction::PtrToInt) {
- Type *IntPtrTy = TD->getIntPtrType(CE0->getOperand(0)->getType());
+ Type *IntPtrTy = DL.getIntPtrType(CE0->getOperand(0)->getType());
if (CE0->getType() == IntPtrTy &&
CE0->getOperand(0)->getType() == CE1->getOperand(0)->getType()) {
- return ConstantFoldCompareInstOperands(Predicate,
- CE0->getOperand(0),
- CE1->getOperand(0),
- TD,
- TLI);
+ return ConstantFoldCompareInstOperands(
+ Predicate, CE0->getOperand(0), CE1->getOperand(0), DL, TLI);
}
}
}
// icmp ne (or x, y), 0 -> (icmp ne x, 0) | (icmp ne y, 0)
if ((Predicate == ICmpInst::ICMP_EQ || Predicate == ICmpInst::ICMP_NE) &&
CE0->getOpcode() == Instruction::Or && Ops1->isNullValue()) {
- Constant *LHS =
- ConstantFoldCompareInstOperands(Predicate, CE0->getOperand(0), Ops1,
- TD, TLI);
- Constant *RHS =
- ConstantFoldCompareInstOperands(Predicate, CE0->getOperand(1), Ops1,
- TD, TLI);
+ Constant *LHS = ConstantFoldCompareInstOperands(
+ Predicate, CE0->getOperand(0), Ops1, DL, TLI);
+ Constant *RHS = ConstantFoldCompareInstOperands(
+ Predicate, CE0->getOperand(1), Ops1, DL, TLI);
unsigned OpC =
Predicate == ICmpInst::ICMP_EQ ? Instruction::And : Instruction::Or;
Constant *Ops[] = { LHS, RHS };
- return ConstantFoldInstOperands(OpC, LHS->getType(), Ops, TD, TLI);
+ return ConstantFoldInstOperands(OpC, LHS->getType(), Ops, DL, TLI);
}
}
#include "llvm/Analysis/ScalarEvolutionExpressions.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/IR/InstIterator.h"
+#include "llvm/IR/Module.h"
#include "llvm/IR/Operator.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
OS << "!\n";
}
-
-
-static
-AliasAnalysis::AliasResult underlyingObjectsAlias(AliasAnalysis *AA,
- const Value *A,
- const Value *B) {
- const Value *AObj = GetUnderlyingObject(A);
- const Value *BObj = GetUnderlyingObject(B);
+static AliasAnalysis::AliasResult underlyingObjectsAlias(AliasAnalysis *AA,
+ const DataLayout &DL,
+ const Value *A,
+ const Value *B) {
+ const Value *AObj = GetUnderlyingObject(A, DL);
+ const Value *BObj = GetUnderlyingObject(B, DL);
return AA->alias(AObj, AA->getTypeStoreSize(AObj->getType()),
BObj, AA->getTypeStoreSize(BObj->getType()));
}
Value *SrcPtr = getPointerOperand(Src);
Value *DstPtr = getPointerOperand(Dst);
- switch (underlyingObjectsAlias(AA, DstPtr, SrcPtr)) {
+ switch (underlyingObjectsAlias(AA, F->getParent()->getDataLayout(), DstPtr,
+ SrcPtr)) {
case AliasAnalysis::MayAlias:
case AliasAnalysis::PartialAlias:
// cannot analyse objects if we don't understand their aliasing.
assert(isLoadOrStore(Dst));
Value *SrcPtr = getPointerOperand(Src);
Value *DstPtr = getPointerOperand(Dst);
- assert(underlyingObjectsAlias(AA, DstPtr, SrcPtr) ==
- AliasAnalysis::MustAlias);
+ assert(underlyingObjectsAlias(AA, F->getParent()->getDataLayout(), DstPtr,
+ SrcPtr) == AliasAnalysis::MustAlias);
// establish loop nesting levels
establishNestingLevels(Src, Dst);
continue;
// Check the value being stored.
- Value *Ptr = GetUnderlyingObject(SI->getOperand(0));
+ Value *Ptr = GetUnderlyingObject(SI->getOperand(0),
+ GV->getParent()->getDataLayout());
if (!isAllocLikeFn(Ptr, TLI))
return false; // Too hard to analyze.
GlobalsModRef::alias(const Location &LocA,
const Location &LocB) {
// Get the base object these pointers point to.
- const Value *UV1 = GetUnderlyingObject(LocA.Ptr);
- const Value *UV2 = GetUnderlyingObject(LocB.Ptr);
+ const Value *UV1 = GetUnderlyingObject(LocA.Ptr, *DL);
+ const Value *UV2 = GetUnderlyingObject(LocB.Ptr, *DL);
// If either of the underlying values is a global, they may be non-addr-taken
// globals, which we can answer queries about.
// If we are asking for mod/ref info of a direct call with a pointer to a
// global we are tracking, return information if we have it.
+ const DataLayout &DL = CS.getCaller()->getParent()->getDataLayout();
if (const GlobalValue *GV =
- dyn_cast<GlobalValue>(GetUnderlyingObject(Loc.Ptr)))
+ dyn_cast<GlobalValue>(GetUnderlyingObject(Loc.Ptr, DL)))
if (GV->hasLocalLinkage())
if (const Function *F = CS.getCalledFunction())
if (NonAddressTakenGlobals.count(GV))
typedef InstVisitor<CallAnalyzer, bool> Base;
friend class InstVisitor<CallAnalyzer, bool>;
- // DataLayout if available, or null.
- const DataLayout *const DL;
-
/// The TargetTransformInfo available for this compilation.
const TargetTransformInfo &TTI;
bool visitUnreachableInst(UnreachableInst &I);
public:
- CallAnalyzer(const DataLayout *DL, const TargetTransformInfo &TTI,
- AssumptionCacheTracker *ACT, Function &Callee, int Threshold)
- : DL(DL), TTI(TTI), ACT(ACT), F(Callee), Threshold(Threshold), Cost(0),
+ CallAnalyzer(const TargetTransformInfo &TTI, AssumptionCacheTracker *ACT,
+ Function &Callee, int Threshold)
+ : TTI(TTI), ACT(ACT), F(Callee), Threshold(Threshold), Cost(0),
IsCallerRecursive(false), IsRecursiveCall(false),
ExposesReturnsTwice(false), HasDynamicAlloca(false),
ContainsNoDuplicateCall(false), HasReturn(false), HasIndirectBr(false),
/// Returns false if unable to compute the offset for any reason. Respects any
/// simplified values known during the analysis of this callsite.
bool CallAnalyzer::accumulateGEPOffset(GEPOperator &GEP, APInt &Offset) {
- if (!DL)
- return false;
-
- unsigned IntPtrWidth = DL->getPointerSizeInBits();
+ const DataLayout &DL = F.getParent()->getDataLayout();
+ unsigned IntPtrWidth = DL.getPointerSizeInBits();
assert(IntPtrWidth == Offset.getBitWidth());
for (gep_type_iterator GTI = gep_type_begin(GEP), GTE = gep_type_end(GEP);
// Handle a struct index, which adds its field offset to the pointer.
if (StructType *STy = dyn_cast<StructType>(*GTI)) {
unsigned ElementIdx = OpC->getZExtValue();
- const StructLayout *SL = DL->getStructLayout(STy);
+ const StructLayout *SL = DL.getStructLayout(STy);
Offset += APInt(IntPtrWidth, SL->getElementOffset(ElementIdx));
continue;
}
- APInt TypeSize(IntPtrWidth, DL->getTypeAllocSize(GTI.getIndexedType()));
+ APInt TypeSize(IntPtrWidth, DL.getTypeAllocSize(GTI.getIndexedType()));
Offset += OpC->getValue().sextOrTrunc(IntPtrWidth) * TypeSize;
}
return true;
// Accumulate the allocated size.
if (I.isStaticAlloca()) {
+ const DataLayout &DL = F.getParent()->getDataLayout();
Type *Ty = I.getAllocatedType();
- AllocatedSize += (DL ? DL->getTypeAllocSize(Ty) :
- Ty->getPrimitiveSizeInBits());
+ AllocatedSize += DL.getTypeAllocSize(Ty);
}
// We will happily inline static alloca instructions.
// Try to fold GEPs of constant-offset call site argument pointers. This
// requires target data and inbounds GEPs.
- if (DL && I.isInBounds()) {
+ if (I.isInBounds()) {
// Check if we have a base + offset for the pointer.
Value *Ptr = I.getPointerOperand();
std::pair<Value *, APInt> BaseAndOffset = ConstantOffsetPtrs.lookup(Ptr);
// Track base/offset pairs when converted to a plain integer provided the
// integer is large enough to represent the pointer.
unsigned IntegerSize = I.getType()->getScalarSizeInBits();
- const DataLayout &DL = I.getModule()->getDataLayout();
+ const DataLayout &DL = F.getParent()->getDataLayout();
if (IntegerSize >= DL.getPointerSizeInBits()) {
std::pair<Value *, APInt> BaseAndOffset
= ConstantOffsetPtrs.lookup(I.getOperand(0));
// modifications provided the integer is not too large.
Value *Op = I.getOperand(0);
unsigned IntegerSize = Op->getType()->getScalarSizeInBits();
- const DataLayout &DL = I.getModule()->getDataLayout();
+ const DataLayout &DL = F.getParent()->getDataLayout();
if (IntegerSize <= DL.getPointerSizeInBits()) {
std::pair<Value *, APInt> BaseAndOffset = ConstantOffsetPtrs.lookup(Op);
if (BaseAndOffset.first)
Constant *COp = dyn_cast<Constant>(Operand);
if (!COp)
COp = SimplifiedValues.lookup(Operand);
- if (COp)
+ if (COp) {
+ const DataLayout &DL = F.getParent()->getDataLayout();
if (Constant *C = ConstantFoldInstOperands(I.getOpcode(), I.getType(),
COp, DL)) {
SimplifiedValues[&I] = C;
return true;
}
+ }
// Disable any SROA on the argument to arbitrary unary operators.
disableSROA(Operand);
bool CallAnalyzer::visitBinaryOperator(BinaryOperator &I) {
Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
+ const DataLayout &DL = F.getParent()->getDataLayout();
if (!isa<Constant>(LHS))
if (Constant *SimpleLHS = SimplifiedValues.lookup(LHS))
LHS = SimpleLHS;
// during devirtualization and so we want to give it a hefty bonus for
// inlining, but cap that bonus in the event that inlining wouldn't pan
// out. Pretend to inline the function, with a custom threshold.
- CallAnalyzer CA(DL, TTI, ACT, *F, InlineConstants::IndirectCallThreshold);
+ CallAnalyzer CA(TTI, ACT, *F, InlineConstants::IndirectCallThreshold);
if (CA.analyzeCall(CS)) {
// We were able to inline the indirect call! Subtract the cost from the
// bonus we want to apply, but don't go below zero.
/// returns 0 if V is not a pointer, and returns the constant '0' if there are
/// no constant offsets applied.
ConstantInt *CallAnalyzer::stripAndComputeInBoundsConstantOffsets(Value *&V) {
- if (!DL || !V->getType()->isPointerTy())
+ if (!V->getType()->isPointerTy())
return nullptr;
- unsigned IntPtrWidth = DL->getPointerSizeInBits();
+ const DataLayout &DL = F.getParent()->getDataLayout();
+ unsigned IntPtrWidth = DL.getPointerSizeInBits();
APInt Offset = APInt::getNullValue(IntPtrWidth);
// Even though we don't look through PHI nodes, we could be called on an
assert(V->getType()->isPointerTy() && "Unexpected operand type!");
} while (Visited.insert(V).second);
- Type *IntPtrTy = DL->getIntPtrType(V->getContext());
+ Type *IntPtrTy = DL.getIntPtrType(V->getContext());
return cast<ConstantInt>(ConstantInt::get(IntPtrTy, Offset));
}
assert(NumVectorInstructions == 0);
FiftyPercentVectorBonus = Threshold;
TenPercentVectorBonus = Threshold / 2;
+ const DataLayout &DL = F.getParent()->getDataLayout();
// Give out bonuses per argument, as the instructions setting them up will
// be gone after inlining.
for (unsigned I = 0, E = CS.arg_size(); I != E; ++I) {
- if (DL && CS.isByValArgument(I)) {
+ if (CS.isByValArgument(I)) {
// We approximate the number of loads and stores needed by dividing the
// size of the byval type by the target's pointer size.
PointerType *PTy = cast<PointerType>(CS.getArgument(I)->getType());
- unsigned TypeSize = DL->getTypeSizeInBits(PTy->getElementType());
- unsigned PointerSize = DL->getPointerSizeInBits();
+ unsigned TypeSize = DL.getTypeSizeInBits(PTy->getElementType());
+ unsigned PointerSize = DL.getPointerSizeInBits();
// Ceiling division.
unsigned NumStores = (TypeSize + PointerSize - 1) / PointerSize;
DEBUG(llvm::dbgs() << " Analyzing call of " << Callee->getName()
<< "...\n");
- CallAnalyzer CA(&Callee->getParent()->getDataLayout(), TTIWP->getTTI(*Callee),
- ACT, *Callee, Threshold);
+ CallAnalyzer CA(TTIWP->getTTI(*Callee), ACT, *Callee, Threshold);
bool ShouldInline = CA.analyzeCall(CS);
DEBUG(CA.dump());
/// return true. Otherwise, return false.
bool IVUsers::AddUsersImpl(Instruction *I,
SmallPtrSetImpl<Loop*> &SimpleLoopNests) {
+ const DataLayout &DL = I->getModule()->getDataLayout();
+
// Add this IV user to the Processed set before returning false to ensure that
// all IV users are members of the set. See IVUsers::isIVUserOrOperand.
if (!Processed.insert(I).second)
// IVUsers is used by LSR which assumes that all SCEV expressions are safe to
// pass to SCEVExpander. Expressions are not safe to expand if they represent
// operations that are not safe to speculate, namely integer division.
- if (!isa<PHINode>(I) && !isSafeToSpeculativelyExecute(I, DL))
+ if (!isa<PHINode>(I) && !isSafeToSpeculativelyExecute(I))
return false;
// LSR is not APInt clean, do not touch integers bigger than 64-bits.
// Also avoid creating IVs of non-native types. For example, we don't want a
// 64-bit IV in 32-bit code just because the loop has one 64-bit cast.
uint64_t Width = SE->getTypeSizeInBits(I->getType());
- if (Width > 64 || (DL && !DL->isLegalInteger(Width)))
+ if (Width > 64 || !DL.isLegalInteger(Width))
return false;
// Get the symbolic expression for this instruction.
LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
SE = &getAnalysis<ScalarEvolution>();
- DL = &L->getHeader()->getModule()->getDataLayout();
// Find all uses of induction variables in this loop, and categorize
// them by stride. Start by finding all of the PHI nodes in the header for
namespace {
struct Query {
- const DataLayout *DL;
+ const DataLayout &DL;
const TargetLibraryInfo *TLI;
const DominatorTree *DT;
AssumptionCache *AC;
const Instruction *CxtI;
- Query(const DataLayout *DL, const TargetLibraryInfo *tli,
+ Query(const DataLayout &DL, const TargetLibraryInfo *tli,
const DominatorTree *dt, AssumptionCache *ac = nullptr,
const Instruction *cxti = nullptr)
: DL(DL), TLI(tli), DT(dt), AC(ac), CxtI(cxti) {}
}
Value *llvm::SimplifyAddInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return ::SimplifyAddInst(Op0, Op1, isNSW, isNUW, Query(DL, TLI, DT, AC, CxtI),
/// This is very similar to GetPointerBaseWithConstantOffset except it doesn't
/// follow non-inbounds geps. This allows it to remain usable for icmp ult/etc.
/// folding.
-static Constant *stripAndComputeConstantOffsets(const DataLayout *DL,
- Value *&V,
+static Constant *stripAndComputeConstantOffsets(const DataLayout &DL, Value *&V,
bool AllowNonInbounds = false) {
assert(V->getType()->getScalarType()->isPointerTy());
- // Without DataLayout, just be conservative for now. Theoretically, more could
- // be done in this case.
- if (!DL)
- return ConstantInt::get(IntegerType::get(V->getContext(), 64), 0);
-
- Type *IntPtrTy = DL->getIntPtrType(V->getType())->getScalarType();
+ Type *IntPtrTy = DL.getIntPtrType(V->getType())->getScalarType();
APInt Offset = APInt::getNullValue(IntPtrTy->getIntegerBitWidth());
// Even though we don't look through PHI nodes, we could be called on an
do {
if (GEPOperator *GEP = dyn_cast<GEPOperator>(V)) {
if ((!AllowNonInbounds && !GEP->isInBounds()) ||
- !GEP->accumulateConstantOffset(*DL, Offset))
+ !GEP->accumulateConstantOffset(DL, Offset))
break;
V = GEP->getPointerOperand();
} else if (Operator::getOpcode(V) == Instruction::BitCast) {
/// \brief Compute the constant difference between two pointer values.
/// If the difference is not a constant, returns zero.
-static Constant *computePointerDifference(const DataLayout *DL,
- Value *LHS, Value *RHS) {
+static Constant *computePointerDifference(const DataLayout &DL, Value *LHS,
+ Value *RHS) {
Constant *LHSOffset = stripAndComputeConstantOffsets(DL, LHS);
Constant *RHSOffset = stripAndComputeConstantOffsets(DL, RHS);
}
Value *llvm::SimplifySubInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return ::SimplifySubInst(Op0, Op1, isNSW, isNUW, Query(DL, TLI, DT, AC, CxtI),
}
Value *llvm::SimplifyFAddInst(Value *Op0, Value *Op1, FastMathFlags FMF,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
}
Value *llvm::SimplifyFSubInst(Value *Op0, Value *Op1, FastMathFlags FMF,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
}
Value *llvm::SimplifyFMulInst(Value *Op0, Value *Op1, FastMathFlags FMF,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
RecursionLimit);
}
-Value *llvm::SimplifyMulInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyMulInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return nullptr;
}
-Value *llvm::SimplifySDivInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifySDivInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return nullptr;
}
-Value *llvm::SimplifyUDivInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyUDivInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
}
Value *llvm::SimplifyFDivInst(Value *Op0, Value *Op1, FastMathFlags FMF,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return nullptr;
}
-Value *llvm::SimplifySRemInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifySRemInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return nullptr;
}
-Value *llvm::SimplifyURemInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyURemInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
}
Value *llvm::SimplifyFRemInst(Value *Op0, Value *Op1, FastMathFlags FMF,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
}
Value *llvm::SimplifyShlInst(Value *Op0, Value *Op1, bool isNSW, bool isNUW,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return ::SimplifyShlInst(Op0, Op1, isNSW, isNUW, Query(DL, TLI, DT, AC, CxtI),
}
Value *llvm::SimplifyLShrInst(Value *Op0, Value *Op1, bool isExact,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
}
Value *llvm::SimplifyAShrInst(Value *Op0, Value *Op1, bool isExact,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
// A & (-A) = A if A is a power of two or zero.
if (match(Op0, m_Neg(m_Specific(Op1))) ||
match(Op1, m_Neg(m_Specific(Op0)))) {
- if (isKnownToBeAPowerOfTwo(Op0, /*OrZero*/ true, 0, Q.AC, Q.CxtI, Q.DT))
+ if (isKnownToBeAPowerOfTwo(Op0, Q.DL, /*OrZero*/ true, 0, Q.AC, Q.CxtI,
+ Q.DT))
return Op0;
- if (isKnownToBeAPowerOfTwo(Op1, /*OrZero*/ true, 0, Q.AC, Q.CxtI, Q.DT))
+ if (isKnownToBeAPowerOfTwo(Op1, Q.DL, /*OrZero*/ true, 0, Q.AC, Q.CxtI,
+ Q.DT))
return Op1;
}
return nullptr;
}
-Value *llvm::SimplifyAndInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyAndInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return nullptr;
}
-Value *llvm::SimplifyOrInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyOrInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return nullptr;
}
-Value *llvm::SimplifyXorInst(Value *Op0, Value *Op1, const DataLayout *DL,
+Value *llvm::SimplifyXorInst(Value *Op0, Value *Op1, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
// If the C and C++ standards are ever made sufficiently restrictive in this
// area, it may be possible to update LLVM's semantics accordingly and reinstate
// this optimization.
-static Constant *computePointerICmp(const DataLayout *DL,
+static Constant *computePointerICmp(const DataLayout &DL,
const TargetLibraryInfo *TLI,
- CmpInst::Predicate Pred,
- Value *LHS, Value *RHS) {
+ CmpInst::Predicate Pred, Value *LHS,
+ Value *RHS) {
// First, skip past any trivial no-ops.
LHS = LHS->stripPointerCasts();
RHS = RHS->stripPointerCasts();
// Turn icmp (ptrtoint x), (ptrtoint/constant) into a compare of the input
// if the integer type is the same size as the pointer type.
- if (MaxRecurse &&
Q.DL && isa<PtrToIntInst>(LI) &&
- Q.DL->getTypeSizeInBits(SrcTy) == DstTy->getPrimitiveSizeInBits()) {
+ if (MaxRecurse && isa<PtrToIntInst>(LI) &&
+ Q.DL.getTypeSizeInBits(SrcTy) == DstTy->getPrimitiveSizeInBits()) {
if (Constant *RHSC = dyn_cast<Constant>(RHS)) {
// Transfer the cast to the constant.
if (Value *V = SimplifyICmpInst(Pred, SrcOp,
}
Value *llvm::SimplifyICmpInst(unsigned Predicate, Value *LHS, Value *RHS,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
Instruction *CxtI) {
}
Value *llvm::SimplifyFCmpInst(unsigned Predicate, Value *LHS, Value *RHS,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
}
Value *llvm::SimplifySelectInst(Value *Cond, Value *TrueVal, Value *FalseVal,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return Ops[0];
Type *Ty = PtrTy->getElementType();
- if (Q.DL && Ty->isSized()) {
+ if (Ty->isSized()) {
Value *P;
uint64_t C;
- uint64_t TyAllocSize = Q.DL->getTypeAllocSize(Ty);
+ uint64_t TyAllocSize = Q.DL.getTypeAllocSize(Ty);
// getelementptr P, N -> P if P points to a type of zero size.
if (TyAllocSize == 0)
return Ops[0];
// The following transforms are only safe if the ptrtoint cast
// doesn't truncate the pointers.
if (Ops[1]->getType()->getScalarSizeInBits() ==
- Q.DL->getPointerSizeInBits(AS)) {
+ Q.DL.getPointerSizeInBits(AS)) {
auto PtrToIntOrZero = [GEPTy](Value *P) -> Value * {
if (match(P, m_Zero()))
return Constant::getNullValue(GEPTy);
return ConstantExpr::getGetElementPtr(cast<Constant>(Ops[0]), Ops.slice(1));
}
-Value *llvm::SimplifyGEPInst(ArrayRef<Value *> Ops, const DataLayout *DL,
+Value *llvm::SimplifyGEPInst(ArrayRef<Value *> Ops, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
}
Value *llvm::SimplifyInsertValueInst(
- Value *Agg, Value *Val, ArrayRef<unsigned> Idxs, const DataLayout *DL,
+ Value *Agg, Value *Val, ArrayRef<unsigned> Idxs, const DataLayout &DL,
const TargetLibraryInfo *TLI, const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return ::SimplifyInsertValueInst(Agg, Val, Idxs, Query(DL, TLI, DT, AC, CxtI),
return nullptr;
}
-Value *llvm::SimplifyTruncInst(Value *Op, Type *Ty, const DataLayout *DL,
+Value *llvm::SimplifyTruncInst(Value *Op, Type *Ty, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
}
Value *llvm::SimplifyBinOp(unsigned Opcode, Value *LHS, Value *RHS,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return ::SimplifyBinOp(Opcode, LHS, RHS, Query(DL, TLI, DT, AC, CxtI),
}
Value *llvm::SimplifyFPBinOp(unsigned Opcode, Value *LHS, Value *RHS,
- const FastMathFlags &FMF, const DataLayout *DL,
+ const FastMathFlags &FMF, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
}
Value *llvm::SimplifyCmpInst(unsigned Predicate, Value *LHS, Value *RHS,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return ::SimplifyCmpInst(Predicate, LHS, RHS, Query(DL, TLI, DT, AC, CxtI),
}
Value *llvm::SimplifyCall(Value *V, User::op_iterator ArgBegin,
- User::op_iterator ArgEnd, const DataLayout *DL,
+ User::op_iterator ArgEnd, const DataLayout &DL,
const TargetLibraryInfo *TLI, const DominatorTree *DT,
AssumptionCache *AC, const Instruction *CxtI) {
return ::SimplifyCall(V, ArgBegin, ArgEnd, Query(DL, TLI, DT, AC, CxtI),
}
Value *llvm::SimplifyCall(Value *V, ArrayRef<Value *> Args,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC,
const Instruction *CxtI) {
return ::SimplifyCall(V, Args.begin(), Args.end(),
/// SimplifyInstruction - See if we can compute a simplified version of this
/// instruction. If not, this returns null.
-Value *llvm::SimplifyInstruction(Instruction *I, const DataLayout *DL,
+Value *llvm::SimplifyInstruction(Instruction *I, const DataLayout &DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT, AssumptionCache *AC) {
Value *Result;
/// This routine returns 'true' only when *it* simplifies something. The passed
/// in simplified value does not count toward this.
static bool replaceAndRecursivelySimplifyImpl(Instruction *I, Value *SimpleV,
- const DataLayout *DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT,
AssumptionCache *AC) {
bool Simplified = false;
SmallSetVector<Instruction *, 8> Worklist;
+ const DataLayout &DL = I->getModule()->getDataLayout();
// If we have an explicit value to collapse to, do that round of the
// simplification loop by hand initially.
return Simplified;
}
-bool llvm::recursivelySimplifyInstruction(Instruction *I, const DataLayout *DL,
+bool llvm::recursivelySimplifyInstruction(Instruction *I,
const TargetLibraryInfo *TLI,
const DominatorTree *DT,
AssumptionCache *AC) {
- return replaceAndRecursivelySimplifyImpl(I, nullptr, DL, TLI, DT, AC);
+ return replaceAndRecursivelySimplifyImpl(I, nullptr, TLI, DT, AC);
}
bool llvm::replaceAndRecursivelySimplify(Instruction *I, Value *SimpleV,
- const DataLayout *DL,
const TargetLibraryInfo *TLI,
const DominatorTree *DT,
AssumptionCache *AC) {
assert(I != SimpleV && "replaceAndRecursivelySimplify(X,X) is not valid!");
assert(SimpleV && "Must provide a simplified value.");
- return replaceAndRecursivelySimplifyImpl(I, SimpleV, DL, TLI, DT, AC);
+ return replaceAndRecursivelySimplifyImpl(I, SimpleV, TLI, DT, AC);
}
/// Merge the specified lattice value into this one, updating this
/// one and returning true if anything changed.
- bool mergeIn(const LVILatticeVal &RHS) {
+ bool mergeIn(const LVILatticeVal &RHS, const DataLayout &DL) {
if (RHS.isUndefined() || isOverdefined()) return false;
if (RHS.isOverdefined()) return markOverdefined();
// Unless we can prove that the two Constants are different, we must
// move to overdefined.
- // FIXME: use DataLayout/TargetLibraryInfo for smarter constant folding.
- if (ConstantInt *Res = dyn_cast<ConstantInt>(
- ConstantFoldCompareInstOperands(CmpInst::ICMP_NE,
- getConstant(),
- RHS.getNotConstant())))
+ if (ConstantInt *Res =
+ dyn_cast<ConstantInt>(ConstantFoldCompareInstOperands(
+ CmpInst::ICMP_NE, getConstant(), RHS.getNotConstant(), DL)))
if (Res->isOne())
return markNotConstant(RHS.getNotConstant());
// Unless we can prove that the two Constants are different, we must
// move to overdefined.
- // FIXME: use DataLayout/TargetLibraryInfo for smarter constant folding.
- if (ConstantInt *Res = dyn_cast<ConstantInt>(
- ConstantFoldCompareInstOperands(CmpInst::ICMP_NE,
- getNotConstant(),
- RHS.getConstant())))
+ if (ConstantInt *Res =
+ dyn_cast<ConstantInt>(ConstantFoldCompareInstOperands(
+ CmpInst::ICMP_NE, getNotConstant(), RHS.getConstant(), DL)))
if (Res->isOne())
return false;
return true;
}
- /// A pointer to the cache of @llvm.assume calls.
- AssumptionCache *AC;
- /// An optional DL pointer.
- const DataLayout *DL;
- /// An optional DT pointer.
- DominatorTree *DT;
-
+ AssumptionCache *AC; ///< A pointer to the cache of @llvm.assume calls.
+ const DataLayout &DL; ///< A mandatory DataLayout
+ DominatorTree *DT; ///< An optional DT pointer.
+
friend struct LVIValueHandle;
void insertResult(Value *Val, BasicBlock *BB, const LVILatticeVal &Result) {
OverDefinedCache.clear();
}
- LazyValueInfoCache(AssumptionCache *AC, const DataLayout *DL = nullptr,
+ LazyValueInfoCache(AssumptionCache *AC, const DataLayout &DL,
DominatorTree *DT = nullptr)
: AC(AC), DL(DL), DT(DT) {}
};
static bool InstructionDereferencesPointer(Instruction *I, Value *Ptr) {
if (LoadInst *L = dyn_cast<LoadInst>(I)) {
return L->getPointerAddressSpace() == 0 &&
- GetUnderlyingObject(L->getPointerOperand()) == Ptr;
+ GetUnderlyingObject(L->getPointerOperand(),
+ L->getModule()->getDataLayout()) == Ptr;
}
if (StoreInst *S = dyn_cast<StoreInst>(I)) {
return S->getPointerAddressSpace() == 0 &&
- GetUnderlyingObject(S->getPointerOperand()) == Ptr;
+ GetUnderlyingObject(S->getPointerOperand(),
+ S->getModule()->getDataLayout()) == Ptr;
}
if (MemIntrinsic *MI = dyn_cast<MemIntrinsic>(I)) {
if (MI->isVolatile()) return false;
if (!Len || Len->isZero()) return false;
if (MI->getDestAddressSpace() == 0)
- if (GetUnderlyingObject(MI->getRawDest()) == Ptr)
+ if (GetUnderlyingObject(MI->getRawDest(),
+ MI->getModule()->getDataLayout()) == Ptr)
return true;
if (MemTransferInst *MTI = dyn_cast<MemTransferInst>(MI))
if (MTI->getSourceAddressSpace() == 0)
- if (GetUnderlyingObject(MTI->getRawSource()) == Ptr)
+ if (GetUnderlyingObject(MTI->getRawSource(),
+ MTI->getModule()->getDataLayout()) == Ptr)
return true;
}
return false;
if (isKnownNonNull(Val)) {
NotNull = true;
} else {
- Value *UnderlyingVal = GetUnderlyingObject(Val);
+ const DataLayout &DL = BB->getModule()->getDataLayout();
+ Value *UnderlyingVal = GetUnderlyingObject(Val, DL);
// If 'GetUnderlyingObject' didn't converge, skip it. It won't converge
// inside InstructionDereferencesPointer either.
- if (UnderlyingVal == GetUnderlyingObject(UnderlyingVal, nullptr, 1)) {
+ if (UnderlyingVal == GetUnderlyingObject(UnderlyingVal, DL, 1)) {
for (Instruction &I : *BB) {
if (InstructionDereferencesPointer(&I, UnderlyingVal)) {
NotNull = true;
if (EdgesMissing)
continue;
- Result.mergeIn(EdgeResult);
+ Result.mergeIn(EdgeResult, DL);
// If we hit overdefined, exit early. The BlockVals entry is already set
// to overdefined.
if (EdgesMissing)
continue;
- Result.mergeIn(EdgeResult);
+ Result.mergeIn(EdgeResult, DL);
// If we hit overdefined, exit early. The BlockVals entry is already set
// to overdefined.
if (!AssumeVH)
continue;
auto *I = cast<CallInst>(AssumeVH);
- if (!isValidAssumeForContext(I, BBI, DL, DT))
+ if (!isValidAssumeForContext(I, BBI, DT))
continue;
Value *C = I->getArgOperand(0);
if (BBLV.isOverdefined())
BBLV = Result;
else
- BBLV.mergeIn(Result);
+ BBLV.mergeIn(Result, DL);
}
}
}
/// This lazily constructs the LazyValueInfoCache.
static LazyValueInfoCache &getCache(void *&PImpl, AssumptionCache *AC,
- const DataLayout *DL = nullptr,
+ const DataLayout *DL,
DominatorTree *DT = nullptr) {
- if (!PImpl)
- PImpl = new LazyValueInfoCache(AC, DL, DT);
+ if (!PImpl) {
+ assert(DL && "getCache() called with a null DataLayout");
+ PImpl = new LazyValueInfoCache(AC, *DL, DT);
+ }
return *static_cast<LazyValueInfoCache*>(PImpl);
}
bool LazyValueInfo::runOnFunction(Function &F) {
AC = &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
+ const DataLayout &DL = F.getParent()->getDataLayout();
DominatorTreeWrapperPass *DTWP =
getAnalysisIfAvailable<DominatorTreeWrapperPass>();
DT = DTWP ? &DTWP->getDomTree() : nullptr;
- DL = &F.getParent()->getDataLayout();
-
TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
if (PImpl)
- getCache(PImpl, AC, DL, DT).clear();
+ getCache(PImpl, AC, &DL, DT).clear();
// Fully lazy.
return false;
void LazyValueInfo::releaseMemory() {
// If the cache was allocated, free it.
if (PImpl) {
- delete &getCache(PImpl, AC);
+ delete &getCache(PImpl, AC, nullptr);
PImpl = nullptr;
}
}
Constant *LazyValueInfo::getConstant(Value *V, BasicBlock *BB,
Instruction *CxtI) {
+ const DataLayout &DL = BB->getModule()->getDataLayout();
LVILatticeVal Result =
- getCache(PImpl, AC, DL, DT).getValueInBlock(V, BB, CxtI);
+ getCache(PImpl, AC, &DL, DT).getValueInBlock(V, BB, CxtI);
if (Result.isConstant())
return Result.getConstant();
Constant *LazyValueInfo::getConstantOnEdge(Value *V, BasicBlock *FromBB,
BasicBlock *ToBB,
Instruction *CxtI) {
+ const DataLayout &DL = FromBB->getModule()->getDataLayout();
LVILatticeVal Result =
- getCache(PImpl, AC, DL, DT).getValueOnEdge(V, FromBB, ToBB, CxtI);
+ getCache(PImpl, AC, &DL, DT).getValueOnEdge(V, FromBB, ToBB, CxtI);
if (Result.isConstant())
return Result.getConstant();
return nullptr;
}
-static LazyValueInfo::Tristate
-getPredicateResult(unsigned Pred, Constant *C, LVILatticeVal &Result,
- const DataLayout *DL, TargetLibraryInfo *TLI) {
+static LazyValueInfo::Tristate getPredicateResult(unsigned Pred, Constant *C,
+ LVILatticeVal &Result,
+ const DataLayout &DL,
+ TargetLibraryInfo *TLI) {
// If we know the value is a constant, evaluate the conditional.
Constant *Res = nullptr;
LazyValueInfo::getPredicateOnEdge(unsigned Pred, Value *V, Constant *C,
BasicBlock *FromBB, BasicBlock *ToBB,
Instruction *CxtI) {
+ const DataLayout &DL = FromBB->getModule()->getDataLayout();
LVILatticeVal Result =
- getCache(PImpl, AC, DL, DT).getValueOnEdge(V, FromBB, ToBB, CxtI);
+ getCache(PImpl, AC, &DL, DT).getValueOnEdge(V, FromBB, ToBB, CxtI);
return getPredicateResult(Pred, C, Result, DL, TLI);
}
LazyValueInfo::Tristate
LazyValueInfo::getPredicateAt(unsigned Pred, Value *V, Constant *C,
Instruction *CxtI) {
- LVILatticeVal Result = getCache(PImpl, AC, DL, DT).getValueAt(V, CxtI);
+ const DataLayout &DL = CxtI->getModule()->getDataLayout();
+ LVILatticeVal Result = getCache(PImpl, AC, &DL, DT).getValueAt(V, CxtI);
return getPredicateResult(Pred, C, Result, DL, TLI);
}
void LazyValueInfo::threadEdge(BasicBlock *PredBB, BasicBlock *OldSucc,
BasicBlock *NewSucc) {
- if (PImpl)
- getCache(PImpl, AC, DL, DT).threadEdge(PredBB, OldSucc, NewSucc);
+ if (PImpl) {
+ const DataLayout &DL = PredBB->getModule()->getDataLayout();
+ getCache(PImpl, AC, &DL, DT).threadEdge(PredBB, OldSucc, NewSucc);
+ }
}
void LazyValueInfo::eraseBlock(BasicBlock *BB) {
- if (PImpl)
- getCache(PImpl, AC, DL, DT).eraseBlock(BB);
+ if (PImpl) {
+ const DataLayout &DL = BB->getModule()->getDataLayout();
+ getCache(PImpl, AC, &DL, DT).eraseBlock(BB);
+ }
}
void visitInsertElementInst(InsertElementInst &I);
void visitUnreachableInst(UnreachableInst &I);
- Value *findValue(Value *V, bool OffsetOk) const;
- Value *findValueImpl(Value *V, bool OffsetOk,
+ Value *findValue(Value *V, const DataLayout &DL, bool OffsetOk) const;
+ Value *findValueImpl(Value *V, const DataLayout &DL, bool OffsetOk,
SmallPtrSetImpl<Value *> &Visited) const;
public:
AliasAnalysis *AA;
AssumptionCache *AC;
DominatorTree *DT;
- const DataLayout *DL;
TargetLibraryInfo *TLI;
std::string Messages;
AA = &getAnalysis<AliasAnalysis>();
AC = &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
- DL = &F.getParent()->getDataLayout();
TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
visit(F);
dbgs() << MessagesStr.str();
void Lint::visitCallSite(CallSite CS) {
Instruction &I = *CS.getInstruction();
Value *Callee = CS.getCalledValue();
+ const DataLayout &DL = CS->getModule()->getDataLayout();
visitMemoryReference(I, Callee, AliasAnalysis::UnknownSize,
0, nullptr, MemRef::Callee);
- if (Function *F = dyn_cast<Function>(findValue(Callee, /*OffsetOk=*/false))) {
+ if (Function *F = dyn_cast<Function>(findValue(Callee, DL,
+ /*OffsetOk=*/false))) {
Assert(CS.getCallingConv() == F->getCallingConv(),
"Undefined behavior: Caller and callee calling convention differ",
&I);
Type *Ty =
cast<PointerType>(Formal->getType())->getElementType();
visitMemoryReference(I, Actual, AA->getTypeStoreSize(Ty),
- DL ? DL->getABITypeAlignment(Ty) : 0,
- Ty, MemRef::Read | MemRef::Write);
+ DL.getABITypeAlignment(Ty), Ty,
+ MemRef::Read | MemRef::Write);
}
}
}
if (CS.isCall() && cast<CallInst>(CS.getInstruction())->isTailCall())
for (CallSite::arg_iterator AI = CS.arg_begin(), AE = CS.arg_end();
AI != AE; ++AI) {
- Value *Obj = findValue(*AI, /*OffsetOk=*/true);
+ Value *Obj = findValue(*AI, DL, /*OffsetOk=*/true);
Assert(!isa<AllocaInst>(Obj),
- "Undefined behavior: Call with \"tail\" keyword references alloca",
+ "Undefined behavior: Call with \"tail\" keyword references "
+ "alloca",
&I);
}
// overlap is not distinguished from the case where nothing is known.
uint64_t Size = 0;
if (const ConstantInt *Len =
- dyn_cast<ConstantInt>(findValue(MCI->getLength(),
- /*OffsetOk=*/false)))
+ dyn_cast<ConstantInt>(findValue(MCI->getLength(), DL,
+ /*OffsetOk=*/false)))
if (Len->getValue().isIntN(32))
Size = Len->getValue().getZExtValue();
Assert(AA->alias(MCI->getSource(), Size, MCI->getDest(), Size) !=
"Unusual: Return statement in function with noreturn attribute", &I);
if (Value *V = I.getReturnValue()) {
- Value *Obj = findValue(V, /*OffsetOk=*/true);
+ Value *Obj =
+ findValue(V, F->getParent()->getDataLayout(), /*OffsetOk=*/true);
Assert(!isa<AllocaInst>(Obj), "Unusual: Returning alloca value", &I);
}
}
if (Size == 0)
return;
- Value *UnderlyingObject = findValue(Ptr, /*OffsetOk=*/true);
+ Value *UnderlyingObject =
+ findValue(Ptr, I.getModule()->getDataLayout(), /*OffsetOk=*/true);
Assert(!isa<ConstantPointerNull>(UnderlyingObject),
"Undefined behavior: Null pointer dereference", &I);
Assert(!isa<UndefValue>(UnderlyingObject),
// Check for buffer overflows and misalignment.
// Only handles memory references that read/write something simple like an
// alloca instruction or a global variable.
+ auto &DL = I.getModule()->getDataLayout();
int64_t Offset = 0;
if (Value *Base = GetPointerBaseWithConstantOffset(Ptr, Offset, DL)) {
// OK, so the access is to a constant offset from Ptr. Check that Ptr is
if (AllocaInst *AI = dyn_cast<AllocaInst>(Base)) {
Type *ATy = AI->getAllocatedType();
- if (DL && !AI->isArrayAllocation() && ATy->isSized())
- BaseSize = DL->getTypeAllocSize(ATy);
+ if (!AI->isArrayAllocation() && ATy->isSized())
+ BaseSize = DL.getTypeAllocSize(ATy);
BaseAlign = AI->getAlignment();
- if (DL && BaseAlign == 0 && ATy->isSized())
- BaseAlign = DL->getABITypeAlignment(ATy);
+ if (BaseAlign == 0 && ATy->isSized())
+ BaseAlign = DL.getABITypeAlignment(ATy);
} else if (GlobalVariable *GV = dyn_cast<GlobalVariable>(Base)) {
// If the global may be defined differently in another compilation unit
// then don't warn about funky memory accesses.
if (GV->hasDefinitiveInitializer()) {
Type *GTy = GV->getType()->getElementType();
- if (DL && GTy->isSized())
- BaseSize = DL->getTypeAllocSize(GTy);
+ if (GTy->isSized())
+ BaseSize = DL.getTypeAllocSize(GTy);
BaseAlign = GV->getAlignment();
- if (DL && BaseAlign == 0 && GTy->isSized())
- BaseAlign = DL->getABITypeAlignment(GTy);
+ if (BaseAlign == 0 && GTy->isSized())
+ BaseAlign = DL.getABITypeAlignment(GTy);
}
}
// Accesses that say that the memory is more aligned than it is are not
// defined.
- if (DL && Align == 0 && Ty && Ty->isSized())
- Align = DL->getABITypeAlignment(Ty);
+ if (Align == 0 && Ty && Ty->isSized())
+ Align = DL.getABITypeAlignment(Ty);
Assert(!BaseAlign || Align <= MinAlign(BaseAlign, Offset),
"Undefined behavior: Memory reference address is misaligned", &I);
}
}
void Lint::visitLShr(BinaryOperator &I) {
- if (ConstantInt *CI =
- dyn_cast<ConstantInt>(findValue(I.getOperand(1), /*OffsetOk=*/false)))
+ if (ConstantInt *CI = dyn_cast<ConstantInt>(
+ findValue(I.getOperand(1), I.getModule()->getDataLayout(),
+ /*OffsetOk=*/false)))
Assert(CI->getValue().ult(cast<IntegerType>(I.getType())->getBitWidth()),
"Undefined result: Shift count out of range", &I);
}
void Lint::visitAShr(BinaryOperator &I) {
- if (ConstantInt *CI =
- dyn_cast<ConstantInt>(findValue(I.getOperand(1), /*OffsetOk=*/false)))
+ if (ConstantInt *CI = dyn_cast<ConstantInt>(findValue(
+ I.getOperand(1), I.getModule()->getDataLayout(), /*OffsetOk=*/false)))
Assert(CI->getValue().ult(cast<IntegerType>(I.getType())->getBitWidth()),
"Undefined result: Shift count out of range", &I);
}
void Lint::visitShl(BinaryOperator &I) {
- if (ConstantInt *CI =
- dyn_cast<ConstantInt>(findValue(I.getOperand(1), /*OffsetOk=*/false)))
+ if (ConstantInt *CI = dyn_cast<ConstantInt>(findValue(
+ I.getOperand(1), I.getModule()->getDataLayout(), /*OffsetOk=*/false)))
Assert(CI->getValue().ult(cast<IntegerType>(I.getType())->getBitWidth()),
"Undefined result: Shift count out of range", &I);
}
II);
}
-static bool isZero(Value *V, const DataLayout *DL, DominatorTree *DT,
+static bool isZero(Value *V, const DataLayout &DL, DominatorTree *DT,
AssumptionCache *AC) {
// Assume undef could be zero.
if (isa<UndefValue>(V))
}
void Lint::visitSDiv(BinaryOperator &I) {
- Assert(!isZero(I.getOperand(1), DL, DT, AC),
+ Assert(!isZero(I.getOperand(1), I.getModule()->getDataLayout(), DT, AC),
"Undefined behavior: Division by zero", &I);
}
void Lint::visitUDiv(BinaryOperator &I) {
- Assert(!isZero(I.getOperand(1), DL, DT, AC),
+ Assert(!isZero(I.getOperand(1), I.getModule()->getDataLayout(), DT, AC),
"Undefined behavior: Division by zero", &I);
}
void Lint::visitSRem(BinaryOperator &I) {
- Assert(!isZero(I.getOperand(1), DL, DT, AC),
+ Assert(!isZero(I.getOperand(1), I.getModule()->getDataLayout(), DT, AC),
"Undefined behavior: Division by zero", &I);
}
void Lint::visitURem(BinaryOperator &I) {
- Assert(!isZero(I.getOperand(1), DL, DT, AC),
+ Assert(!isZero(I.getOperand(1), I.getModule()->getDataLayout(), DT, AC),
"Undefined behavior: Division by zero", &I);
}
}
void Lint::visitExtractElementInst(ExtractElementInst &I) {
- if (ConstantInt *CI =
- dyn_cast<ConstantInt>(findValue(I.getIndexOperand(),
- /*OffsetOk=*/false)))
+ if (ConstantInt *CI = dyn_cast<ConstantInt>(
+ findValue(I.getIndexOperand(), I.getModule()->getDataLayout(),
+ /*OffsetOk=*/false)))
Assert(CI->getValue().ult(I.getVectorOperandType()->getNumElements()),
"Undefined result: extractelement index out of range", &I);
}
void Lint::visitInsertElementInst(InsertElementInst &I) {
- if (ConstantInt *CI =
- dyn_cast<ConstantInt>(findValue(I.getOperand(2),
- /*OffsetOk=*/false)))
+ if (ConstantInt *CI = dyn_cast<ConstantInt>(
+ findValue(I.getOperand(2), I.getModule()->getDataLayout(),
+ /*OffsetOk=*/false)))
Assert(CI->getValue().ult(I.getType()->getNumElements()),
"Undefined result: insertelement index out of range", &I);
}
/// Most analysis passes don't require this logic, because instcombine
/// will simplify most of these kinds of things away. But it's a goal of
/// this Lint pass to be useful even on non-optimized IR.
-Value *Lint::findValue(Value *V, bool OffsetOk) const {
+Value *Lint::findValue(Value *V, const DataLayout &DL, bool OffsetOk) const {
SmallPtrSet<Value *, 4> Visited;
- return findValueImpl(V, OffsetOk, Visited);
+ return findValueImpl(V, DL, OffsetOk, Visited);
}
/// findValueImpl - Implementation helper for findValue.
-Value *Lint::findValueImpl(Value *V, bool OffsetOk,
+Value *Lint::findValueImpl(Value *V, const DataLayout &DL, bool OffsetOk,
SmallPtrSetImpl<Value *> &Visited) const {
// Detect self-referential values.
if (!Visited.insert(V).second)
break;
if (Value *U = FindAvailableLoadedValue(L->getPointerOperand(),
BB, BBI, 6, AA))
- return findValueImpl(U, OffsetOk, Visited);
+ return findValueImpl(U, DL, OffsetOk, Visited);
if (BBI != BB->begin()) break;
BB = BB->getUniquePredecessor();
if (!BB) break;
} else if (PHINode *PN = dyn_cast<PHINode>(V)) {
if (Value *W = PN->hasConstantValue())
if (W != V)
- return findValueImpl(W, OffsetOk, Visited);
+ return findValueImpl(W, DL, OffsetOk, Visited);
} else if (CastInst *CI = dyn_cast<CastInst>(V)) {
if (CI->isNoopCast(DL))
- return findValueImpl(CI->getOperand(0), OffsetOk, Visited);
+ return findValueImpl(CI->getOperand(0), DL, OffsetOk, Visited);
} else if (ExtractValueInst *Ex = dyn_cast<ExtractValueInst>(V)) {
if (Value *W = FindInsertedValue(Ex->getAggregateOperand(),
Ex->getIndices()))
if (W != V)
- return findValueImpl(W, OffsetOk, Visited);
+ return findValueImpl(W, DL, OffsetOk, Visited);
} else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) {
// Same as above, but for ConstantExpr instead of Instruction.
if (Instruction::isCast(CE->getOpcode())) {
if (CastInst::isNoopCast(Instruction::CastOps(CE->getOpcode()),
- CE->getOperand(0)->getType(),
- CE->getType(),
- DL ? DL->getIntPtrType(V->getType()) :
- Type::getInt64Ty(V->getContext())))
- return findValueImpl(CE->getOperand(0), OffsetOk, Visited);
+ CE->getOperand(0)->getType(), CE->getType(),
+ DL.getIntPtrType(V->getType())))
+ return findValueImpl(CE->getOperand(0), DL, OffsetOk, Visited);
} else if (CE->getOpcode() == Instruction::ExtractValue) {
ArrayRef<unsigned> Indices = CE->getIndices();
if (Value *W = FindInsertedValue(CE->getOperand(0), Indices))
if (W != V)
- return findValueImpl(W, OffsetOk, Visited);
+ return findValueImpl(W, DL, OffsetOk, Visited);
}
}
// As a last resort, try SimplifyInstruction or constant folding.
if (Instruction *Inst = dyn_cast<Instruction>(V)) {
if (Value *W = SimplifyInstruction(Inst, DL, TLI, DT, AC))
- return findValueImpl(W, OffsetOk, Visited);
+ return findValueImpl(W, DL, OffsetOk, Visited);
} else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V)) {
if (Value *W = ConstantFoldConstantExpression(CE, DL, TLI))
if (W != V)
- return findValueImpl(W, OffsetOk, Visited);
+ return findValueImpl(W, DL, OffsetOk, Visited);
}
return V;
/// This uses the pointee type to determine how many bytes need to be safe to
/// load from the pointer.
bool llvm::isSafeToLoadUnconditionally(Value *V, Instruction *ScanFrom,
- unsigned Align, const DataLayout *DL) {
+ unsigned Align) {
+ const DataLayout &DL = ScanFrom->getModule()->getDataLayout();
int64_t ByteOffset = 0;
Value *Base = V;
Base = GetPointerBaseWithConstantOffset(V, ByteOffset, DL);
}
PointerType *AddrTy = cast<PointerType>(V->getType());
- uint64_t LoadSize = DL ? DL->getTypeStoreSize(AddrTy->getElementType()) : 0;
+ uint64_t LoadSize = DL.getTypeStoreSize(AddrTy->getElementType());
// If we found a base allocated type from either an alloca or global variable,
// try to see if we are definitively within the allocated region. We need to
// know the size of the base type and the loaded type to do anything in this
- // case, so only try this when we have the DataLayout available.
- if (BaseType && BaseType->isSized() && DL) {
+ // case.
+ if (BaseType && BaseType->isSized()) {
if (BaseAlign == 0)
- BaseAlign = DL->getPrefTypeAlignment(BaseType);
+ BaseAlign = DL.getPrefTypeAlignment(BaseType);
if (Align <= BaseAlign) {
// Check if the load is within the bounds of the underlying object.
- if (ByteOffset + LoadSize <= DL->getTypeAllocSize(BaseType) &&
+ if (ByteOffset + LoadSize <= DL.getTypeAllocSize(BaseType) &&
(Align == 0 || (ByteOffset % Align) == 0))
return true;
}
else
continue;
- // Handle trivial cases even w/o DataLayout or other work.
+ // Handle trivial cases.
if (AccessedPtr == V)
return true;
- if (!DL)
- continue;
-
auto *AccessedTy = cast<PointerType>(AccessedPtr->getType());
if (AreEquivalentAddressValues(AccessedPtr->stripPointerCasts(), V) &&
- LoadSize <= DL->getTypeStoreSize(AccessedTy->getElementType()))
+ LoadSize <= DL.getTypeStoreSize(AccessedTy->getElementType()))
return true;
}
return false;
Type *AccessTy = cast<PointerType>(Ptr->getType())->getElementType();
- // Try to get the DataLayout for this module. This may be null, in which case
- // the optimizations will be limited.
const DataLayout &DL = ScanBB->getModule()->getDataLayout();
// Try to get the store size for the type.
if (LoadInst *LI = dyn_cast<LoadInst>(Inst))
if (AreEquivalentAddressValues(
LI->getPointerOperand()->stripPointerCasts(), StrippedPtr) &&
- CastInst::isBitOrNoopPointerCastable(LI->getType(), AccessTy, &DL)) {
+ CastInst::isBitOrNoopPointerCastable(LI->getType(), AccessTy, DL)) {
if (AATags)
LI->getAAMetadata(*AATags);
return LI;
// those cases are unlikely.)
if (AreEquivalentAddressValues(StorePtr, StrippedPtr) &&
CastInst::isBitOrNoopPointerCastable(SI->getValueOperand()->getType(),
- AccessTy, &DL)) {
+ AccessTy, DL)) {
if (AATags)
SI->getAAMetadata(*AATags);
return SI->getOperand(0);
/// \brief Set of potential dependent memory accesses.
typedef EquivalenceClasses<MemAccessInfo> DepCandidates;
- AccessAnalysis(const DataLayout *Dl, AliasAnalysis *AA, DepCandidates &DA) :
- DL(Dl), AST(*AA), DepCands(DA), IsRTCheckNeeded(false) {}
+ AccessAnalysis(const DataLayout &Dl, AliasAnalysis *AA, DepCandidates &DA)
+ : DL(Dl), AST(*AA), DepCands(DA), IsRTCheckNeeded(false) {}
/// \brief Register a load and whether it is only read from.
void addLoad(AliasAnalysis::Location &Loc, bool IsReadOnly) {
/// Set of all accesses.
PtrAccessSet Accesses;
+ const DataLayout &DL;
+
/// Set of accesses that need a further dependence check.
MemAccessInfoSet CheckDeps;
/// Set of pointers that are read only.
SmallPtrSet<Value*, 16> ReadOnlyPtr;
- const DataLayout *DL;
-
/// An alias set tracker to partition the access set by underlying object and
//intrinsic property (such as TBAA metadata).
AliasSetTracker AST;
/// \brief Check the stride of the pointer and ensure that it does not wrap in
/// the address space.
-static int isStridedPtr(ScalarEvolution *SE, const DataLayout *DL, Value *Ptr,
- const Loop *Lp, const ValueToValueMap &StridesMap);
+static int isStridedPtr(ScalarEvolution *SE, Value *Ptr, const Loop *Lp,
+ const ValueToValueMap &StridesMap);
bool AccessAnalysis::canCheckPtrAtRT(
LoopAccessInfo::RuntimePointerCheck &RtCheck, unsigned &NumComparisons,
++NumReadPtrChecks;
if (hasComputableBounds(SE, StridesMap, Ptr) &&
- // When we run after a failing dependency check we have to make sure we
- // don't have wrapping pointers.
+ // When we run after a failing dependency check we have to make sure
+ // we don't have wrapping pointers.
(!ShouldCheckStride ||
- isStridedPtr(SE, DL, Ptr, TheLoop, StridesMap) == 1)) {
+ isStridedPtr(SE, Ptr, TheLoop, StridesMap) == 1)) {
// The id of the dependence set.
unsigned DepId;
typedef PointerIntPair<Value *, 1, bool> MemAccessInfo;
typedef SmallPtrSet<MemAccessInfo, 8> MemAccessInfoSet;
- MemoryDepChecker(ScalarEvolution *Se, const DataLayout *Dl, const Loop *L)
- : SE(Se), DL(Dl), InnermostLoop(L), AccessIdx(0),
+ MemoryDepChecker(ScalarEvolution *Se, const Loop *L)
+ : SE(Se), InnermostLoop(L), AccessIdx(0),
ShouldRetryWithRuntimeCheck(false) {}
/// \brief Register the location (instructions are given increasing numbers)
private:
ScalarEvolution *SE;
- const DataLayout *DL;
const Loop *InnermostLoop;
/// \brief Maps access locations (ptr, read/write) to program order.
}
/// \brief Check whether the access through \p Ptr has a constant stride.
-static int isStridedPtr(ScalarEvolution *SE, const DataLayout *DL, Value *Ptr,
- const Loop *Lp, const ValueToValueMap &StridesMap) {
+static int isStridedPtr(ScalarEvolution *SE, Value *Ptr, const Loop *Lp,
+ const ValueToValueMap &StridesMap) {
const Type *Ty = Ptr->getType();
assert(Ty->isPointerTy() && "Unexpected non-ptr");
return 0;
}
- int64_t Size = DL->getTypeAllocSize(PtrTy->getElementType());
+ auto &DL = Lp->getHeader()->getModule()->getDataLayout();
+ int64_t Size = DL.getTypeAllocSize(PtrTy->getElementType());
const APInt &APStepVal = C->getValue()->getValue();
// Huge step value - give up.
const SCEV *AScev = replaceSymbolicStrideSCEV(SE, Strides, APtr);
const SCEV *BScev = replaceSymbolicStrideSCEV(SE, Strides, BPtr);
- int StrideAPtr = isStridedPtr(SE, DL, APtr, InnermostLoop, Strides);
- int StrideBPtr = isStridedPtr(SE, DL, BPtr, InnermostLoop, Strides);
+ int StrideAPtr = isStridedPtr(SE, APtr, InnermostLoop, Strides);
+ int StrideBPtr = isStridedPtr(SE, BPtr, InnermostLoop, Strides);
const SCEV *Src = AScev;
const SCEV *Sink = BScev;
Type *ATy = APtr->getType()->getPointerElementType();
Type *BTy = BPtr->getType()->getPointerElementType();
- unsigned TypeByteSize = DL->getTypeAllocSize(ATy);
+ auto &DL = InnermostLoop->getHeader()->getModule()->getDataLayout();
+ unsigned TypeByteSize = DL.getTypeAllocSize(ATy);
// Negative distances are not plausible dependencies.
const APInt &Val = C->getValue()->getValue();
PtrRtCheck.Need = false;
const bool IsAnnotatedParallel = TheLoop->isAnnotatedParallel();
- MemoryDepChecker DepChecker(SE, DL, TheLoop);
+ MemoryDepChecker DepChecker(SE, TheLoop);
// For each block.
for (Loop::block_iterator bb = TheLoop->block_begin(),
}
AccessAnalysis::DepCandidates DependentAccesses;
- AccessAnalysis Accesses(DL, AA, DependentAccesses);
+ AccessAnalysis Accesses(TheLoop->getHeader()->getModule()->getDataLayout(),
+ AA, DependentAccesses);
// Holds the analyzed pointers. We don't want to call GetUnderlyingObjects
// multiple times on the same object. If the ptr is accessed twice, once
// read a few words, modify, and write a few words, and some of the
// words may be written to the same address.
bool IsReadOnlyPtr = false;
- if (Seen.insert(Ptr).second ||
- !isStridedPtr(SE, DL, Ptr, TheLoop, Strides)) {
+ if (Seen.insert(Ptr).second || !isStridedPtr(SE, Ptr, TheLoop, Strides)) {
++NumReads;
IsReadOnlyPtr = true;
}
SmallVector<TrackingVH<Value> , 2> Ends;
LLVMContext &Ctx = Loc->getContext();
- SCEVExpander Exp(*SE, "induction");
+ SCEVExpander Exp(*SE, DL, "induction");
Instruction *FirstInst = nullptr;
for (unsigned i = 0; i < NumPointers; ++i) {
}
LoopAccessInfo::LoopAccessInfo(Loop *L, ScalarEvolution *SE,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI, AliasAnalysis *AA,
DominatorTree *DT,
const ValueToValueMap &Strides)
#endif
if (!LAI) {
+ const DataLayout &DL = L->getHeader()->getModule()->getDataLayout();
LAI = llvm::make_unique<LoopAccessInfo>(L, SE, DL, TLI, AA, DT, Strides);
#ifndef NDEBUG
LAI->NumSymbolicStrides = Strides.size();
bool LoopAccessAnalysis::runOnFunction(Function &F) {
SE = &getAnalysis<ScalarEvolution>();
- DL = &F.getParent()->getDataLayout();
auto *TLIP = getAnalysisIfAvailable<TargetLibraryInfoWrapperPass>();
TLI = TLIP ? &TLIP->getTLI() : nullptr;
AA = &getAnalysis<AliasAnalysis>();
for (auto &I: inst_range(F)) {
if (LoadInst *LI = dyn_cast<LoadInst>(&I)) {
Value *PO = LI->getPointerOperand();
- if (PO->isDereferenceablePointer(&DL))
+ if (PO->isDereferenceablePointer(DL))
Vec.push_back(PO);
}
}
return isMallocLikeFn(I, TLI) ? dyn_cast<CallInst>(I) : nullptr;
}
-static Value *computeArraySize(const CallInst *CI, const DataLayout *DL,
+static Value *computeArraySize(const CallInst *CI, const DataLayout &DL,
const TargetLibraryInfo *TLI,
bool LookThroughSExt = false) {
if (!CI)
// The size of the malloc's result type must be known to determine array size.
Type *T = getMallocAllocatedType(CI, TLI);
- if (!T || !T->isSized() || !DL)
+ if (!T || !T->isSized())
return nullptr;
- unsigned ElementSize = DL->getTypeAllocSize(T);
+ unsigned ElementSize = DL.getTypeAllocSize(T);
if (StructType *ST = dyn_cast<StructType>(T))
- ElementSize = DL->getStructLayout(ST)->getSizeInBytes();
+ ElementSize = DL.getStructLayout(ST)->getSizeInBytes();
// If malloc call's arg can be determined to be a multiple of ElementSize,
// return the multiple. Otherwise, return NULL.
/// then return that multiple. For non-array mallocs, the multiple is
/// constant 1. Otherwise, return NULL for mallocs whose array size cannot be
/// determined.
-Value *llvm::getMallocArraySize(CallInst *CI, const DataLayout *DL,
+Value *llvm::getMallocArraySize(CallInst *CI, const DataLayout &DL,
const TargetLibraryInfo *TLI,
bool LookThroughSExt) {
assert(isMallocLikeFn(CI, TLI) && "getMallocArraySize and not malloc call");
/// object size in Size if successful, and false otherwise.
/// If RoundToAlign is true, then Size is rounded up to the aligment of allocas,
/// byval arguments, and global variables.
-bool llvm::getObjectSize(const Value *Ptr, uint64_t &Size, const DataLayout *DL,
+bool llvm::getObjectSize(const Value *Ptr, uint64_t &Size, const DataLayout &DL,
const TargetLibraryInfo *TLI, bool RoundToAlign) {
- if (!DL)
- return false;
-
ObjectSizeOffsetVisitor Visitor(DL, TLI, Ptr->getContext(), RoundToAlign);
SizeOffsetType Data = Visitor.compute(const_cast<Value*>(Ptr));
if (!Visitor.bothKnown(Data))
return Size;
}
-ObjectSizeOffsetVisitor::ObjectSizeOffsetVisitor(const DataLayout *DL,
+ObjectSizeOffsetVisitor::ObjectSizeOffsetVisitor(const DataLayout &DL,
const TargetLibraryInfo *TLI,
LLVMContext &Context,
bool RoundToAlign)
-: DL(DL), TLI(TLI), RoundToAlign(RoundToAlign) {
+ : DL(DL), TLI(TLI), RoundToAlign(RoundToAlign) {
// Pointer size must be rechecked for each object visited since it could have
// a different address space.
}
SizeOffsetType ObjectSizeOffsetVisitor::compute(Value *V) {
- IntTyBits = DL->getPointerTypeSizeInBits(V->getType());
+ IntTyBits = DL.getPointerTypeSizeInBits(V->getType());
Zero = APInt::getNullValue(IntTyBits);
V = V->stripPointerCasts();
if (!I.getAllocatedType()->isSized())
return unknown();
- APInt Size(IntTyBits, DL->getTypeAllocSize(I.getAllocatedType()));
+ APInt Size(IntTyBits, DL.getTypeAllocSize(I.getAllocatedType()));
if (!I.isArrayAllocation())
return std::make_pair(align(Size, I.getAlignment()), Zero);
return unknown();
}
PointerType *PT = cast<PointerType>(A.getType());
- APInt Size(IntTyBits, DL->getTypeAllocSize(PT->getElementType()));
+ APInt Size(IntTyBits, DL.getTypeAllocSize(PT->getElementType()));
return std::make_pair(align(Size, A.getParamAlignment()), Zero);
}
SizeOffsetType ObjectSizeOffsetVisitor::visitGEPOperator(GEPOperator &GEP) {
SizeOffsetType PtrData = compute(GEP.getPointerOperand());
APInt Offset(IntTyBits, 0);
- if (!bothKnown(PtrData) || !GEP.accumulateConstantOffset(*DL, Offset))
+ if (!bothKnown(PtrData) || !GEP.accumulateConstantOffset(DL, Offset))
return unknown();
return std::make_pair(PtrData.first, PtrData.second + Offset);
if (!GV.hasDefinitiveInitializer())
return unknown();
- APInt Size(IntTyBits, DL->getTypeAllocSize(GV.getType()->getElementType()));
+ APInt Size(IntTyBits, DL.getTypeAllocSize(GV.getType()->getElementType()));
return std::make_pair(align(Size, GV.getAlignment()), Zero);
}
return unknown();
}
-ObjectSizeOffsetEvaluator::ObjectSizeOffsetEvaluator(const DataLayout *DL,
- const TargetLibraryInfo *TLI,
- LLVMContext &Context,
- bool RoundToAlign)
-: DL(DL), TLI(TLI), Context(Context), Builder(Context, TargetFolder(DL)),
- RoundToAlign(RoundToAlign) {
+ObjectSizeOffsetEvaluator::ObjectSizeOffsetEvaluator(
+ const DataLayout &DL, const TargetLibraryInfo *TLI, LLVMContext &Context,
+ bool RoundToAlign)
+ : DL(DL), TLI(TLI), Context(Context), Builder(Context, TargetFolder(DL)),
+ RoundToAlign(RoundToAlign) {
// IntTy and Zero must be set for each compute() since the address space may
// be different for later objects.
}
SizeOffsetEvalType ObjectSizeOffsetEvaluator::compute(Value *V) {
// XXX - Are vectors of pointers possible here?
- IntTy = cast<IntegerType>(DL->getIntPtrType(V->getType()));
+ IntTy = cast<IntegerType>(DL.getIntPtrType(V->getType()));
Zero = ConstantInt::get(IntTy, 0);
SizeOffsetEvalType Result = compute_(V);
assert(I.isArrayAllocation());
Value *ArraySize = I.getArraySize();
Value *Size = ConstantInt::get(ArraySize->getType(),
- DL->getTypeAllocSize(I.getAllocatedType()));
+ DL.getTypeAllocSize(I.getAllocatedType()));
Size = Builder.CreateMul(Size, ArraySize);
return std::make_pair(Size, Zero);
}
if (!bothKnown(PtrData))
return unknown();
- Value *Offset = EmitGEPOffset(&Builder, *DL, &GEP, /*NoAssumptions=*/true);
+ Value *Offset = EmitGEPOffset(&Builder, DL, &GEP, /*NoAssumptions=*/true);
Offset = Builder.CreateAdd(PtrData.second, Offset);
return std::make_pair(PtrData.first, Offset);
}
bool MemoryDependenceAnalysis::runOnFunction(Function &F) {
AA = &getAnalysis<AliasAnalysis>();
AC = &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
- DL = &F.getParent()->getDataLayout();
DominatorTreeWrapperPass *DTWP =
getAnalysisIfAvailable<DominatorTreeWrapperPass>();
DT = DTWP ? &DTWP->getDomTree() : nullptr;
///
/// MemLocBase, MemLocOffset are lazily computed here the first time the
/// base/offs of memloc is needed.
-static bool
-isLoadLoadClobberIfExtendedToFullWidth(const AliasAnalysis::Location &MemLoc,
- const Value *&MemLocBase,
- int64_t &MemLocOffs,
- const LoadInst *LI,
- const DataLayout *DL) {
- // If we have no target data, we can't do this.
- if (!DL) return false;
+static bool isLoadLoadClobberIfExtendedToFullWidth(
+ const AliasAnalysis::Location &MemLoc, const Value *&MemLocBase,
+ int64_t &MemLocOffs, const LoadInst *LI) {
+ const DataLayout &DL = LI->getModule()->getDataLayout();
// If we haven't already computed the base/offset of MemLoc, do so now.
if (!MemLocBase)
MemLocBase = GetPointerBaseWithConstantOffset(MemLoc.Ptr, MemLocOffs, DL);
- unsigned Size = MemoryDependenceAnalysis::
- getLoadLoadClobberFullWidthSize(MemLocBase, MemLocOffs, MemLoc.Size,
- LI, *DL);
+ unsigned Size = MemoryDependenceAnalysis::getLoadLoadClobberFullWidthSize(
+ MemLocBase, MemLocOffs, MemLoc.Size, LI);
return Size != 0;
}
/// 2) safe for the target, and 3) would provide the specified memory
/// location value, then this function returns the size in bytes of the
/// load width to use. If not, this returns zero.
-unsigned MemoryDependenceAnalysis::
-getLoadLoadClobberFullWidthSize(const Value *MemLocBase, int64_t MemLocOffs,
- unsigned MemLocSize, const LoadInst *LI,
- const DataLayout &DL) {
+unsigned MemoryDependenceAnalysis::getLoadLoadClobberFullWidthSize(
+ const Value *MemLocBase, int64_t MemLocOffs, unsigned MemLocSize,
+ const LoadInst *LI) {
// We can only extend simple integer loads.
if (!isa<IntegerType>(LI->getType()) || !LI->isSimple()) return 0;
if (LI->getParent()->getParent()->hasFnAttribute(Attribute::SanitizeThread))
return 0;
+ const DataLayout &DL = LI->getModule()->getDataLayout();
+
// Get the base of this load.
int64_t LIOffs = 0;
const Value *LIBase =
- GetPointerBaseWithConstantOffset(LI->getPointerOperand(), LIOffs, &DL);
+ GetPointerBaseWithConstantOffset(LI->getPointerOperand(), LIOffs, DL);
// If the two pointers are not based on the same pointer, we can't tell that
// they are related.
isInvariantLoad = true;
}
+ const DataLayout &DL = BB->getModule()->getDataLayout();
+
// Walk backwards through the basic block, looking for dependencies.
while (ScanIt != BB->begin()) {
Instruction *Inst = --ScanIt;
// location is 1 byte at P+1). If so, return it as a load/load
// clobber result, allowing the client to decide to widen the load if
// it wants to.
- if (IntegerType *ITy = dyn_cast<IntegerType>(LI->getType()))
- if (LI->getAlignment()*8 > ITy->getPrimitiveSizeInBits() &&
+ if (IntegerType *ITy = dyn_cast<IntegerType>(LI->getType())) {
+ if (LI->getAlignment() * 8 > ITy->getPrimitiveSizeInBits() &&
isLoadLoadClobberIfExtendedToFullWidth(MemLoc, MemLocBase,
- MemLocOffset, LI, DL))
+ MemLocOffset, LI))
return MemDepResult::getClobber(Inst);
-
+ }
continue;
}
const_cast<Value *>(Loc.Ptr)));
return;
}
-
-
+ const DataLayout &DL = FromBB->getModule()->getDataLayout();
PHITransAddr Address(const_cast<Value *>(Loc.Ptr), DL, AC);
// This is the set of blocks we've inspected, and the pointer we consider in
}
const SCEV *ScalarEvolution::getSizeOfExpr(Type *IntTy, Type *AllocTy) {
- // If we have DataLayout, we can bypass creating a target-independent
+ // We can bypass creating a target-independent
// constant expression and then folding it back into a ConstantInt.
// This is just a compile-time optimization.
- if (DL)
- return getConstant(IntTy, DL->getTypeAllocSize(AllocTy));
-
- Constant *C = ConstantExpr::getSizeOf(AllocTy);
- if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C))
- if (Constant *Folded = ConstantFoldConstantExpression(CE, DL, TLI))
- C = Folded;
- Type *Ty = getEffectiveSCEVType(PointerType::getUnqual(AllocTy));
- assert(Ty == IntTy && "Effective SCEV type doesn't match");
- return getTruncateOrZeroExtend(getSCEV(C), Ty);
+ return getConstant(IntTy,
+ F->getParent()->getDataLayout().getTypeAllocSize(AllocTy));
}
const SCEV *ScalarEvolution::getOffsetOfExpr(Type *IntTy,
StructType *STy,
unsigned FieldNo) {
- // If we have DataLayout, we can bypass creating a target-independent
+ // We can bypass creating a target-independent
// constant expression and then folding it back into a ConstantInt.
// This is just a compile-time optimization.
- if (DL) {
- return getConstant(IntTy,
- DL->getStructLayout(STy)->getElementOffset(FieldNo));
- }
-
- Constant *C = ConstantExpr::getOffsetOf(STy, FieldNo);
- if (ConstantExpr *CE = dyn_cast<ConstantExpr>(C))
- if (Constant *Folded = ConstantFoldConstantExpression(CE, DL, TLI))
- C = Folded;
-
- Type *Ty = getEffectiveSCEVType(PointerType::getUnqual(STy));
- return getTruncateOrZeroExtend(getSCEV(C), Ty);
+ return getConstant(
+ IntTy,
+ F->getParent()->getDataLayout().getStructLayout(STy)->getElementOffset(
+ FieldNo));
}
const SCEV *ScalarEvolution::getUnknown(Value *V) {
/// for which isSCEVable must return true.
uint64_t ScalarEvolution::getTypeSizeInBits(Type *Ty) const {
assert(isSCEVable(Ty) && "Type is not SCEVable!");
-
- // If we have a DataLayout, use it!
- if (DL)
- return DL->getTypeSizeInBits(Ty);
-
- // Integer types have fixed sizes.
- if (Ty->isIntegerTy())
- return Ty->getPrimitiveSizeInBits();
-
- // The only other support type is pointer. Without DataLayout, conservatively
- // assume pointers are 64-bit.
- assert(Ty->isPointerTy() && "isSCEVable permitted a non-SCEVable type!");
- return 64;
+ return F->getParent()->getDataLayout().getTypeSizeInBits(Ty);
}
/// getEffectiveSCEVType - Return a type with the same bitwidth as
// The only other support type is pointer.
assert(Ty->isPointerTy() && "Unexpected non-pointer non-integer type!");
-
- if (DL)
- return DL->getIntPtrType(Ty);
-
- // Without DataLayout, conservatively assume pointers are 64-bit.
- return Type::getInt64Ty(getContext());
+ return F->getParent()->getDataLayout().getIntPtrType(Ty);
}
const SCEV *ScalarEvolution::getCouldNotCompute() {
// PHI's incoming blocks are in a different loop, in which case doing so
// risks breaking LCSSA form. Instcombine would normally zap these, but
// it doesn't have DominatorTree information, so it may miss cases.
- if (Value *V = SimplifyInstruction(PN, DL, TLI, DT, AC))
+ if (Value *V =
+ SimplifyInstruction(PN, F->getParent()->getDataLayout(), TLI, DT, AC))
if (LI->replacementPreservesLCSSAForm(PN, V))
return getSCEV(V);
// For a SCEVUnknown, ask ValueTracking.
unsigned BitWidth = getTypeSizeInBits(U->getType());
APInt Zeros(BitWidth, 0), Ones(BitWidth, 0);
- computeKnownBits(U->getValue(), Zeros, Ones, DL, 0, AC, nullptr, DT);
+ computeKnownBits(U->getValue(), Zeros, Ones,
+ F->getParent()->getDataLayout(), 0, AC, nullptr, DT);
return Zeros.countTrailingOnes();
}
// Split here to avoid paying the compile-time cost of calling both
// computeKnownBits and ComputeNumSignBits. This restriction can be lifted
// if needed.
-
+ const DataLayout &DL = F->getParent()->getDataLayout();
if (SignHint == ScalarEvolution::HINT_RANGE_UNSIGNED) {
// For a SCEVUnknown, ask ValueTracking.
APInt Zeros(BitWidth, 0), Ones(BitWidth, 0);
} else {
assert(SignHint == ScalarEvolution::HINT_RANGE_SIGNED &&
"generalize as needed!");
- if (U->getValue()->getType()->isIntegerTy() || DL) {
- unsigned NS = ComputeNumSignBits(U->getValue(), DL, 0, AC, nullptr, DT);
- if (NS > 1)
- ConservativeResult = ConservativeResult.intersectWith(ConstantRange(
- APInt::getSignedMinValue(BitWidth).ashr(NS - 1),
- APInt::getSignedMaxValue(BitWidth).ashr(NS - 1) + 1));
- }
+ unsigned NS = ComputeNumSignBits(U->getValue(), DL, 0, AC, nullptr, DT);
+ if (NS > 1)
+ ConservativeResult = ConservativeResult.intersectWith(
+ ConstantRange(APInt::getSignedMinValue(BitWidth).ashr(NS - 1),
+ APInt::getSignedMaxValue(BitWidth).ashr(NS - 1) + 1));
}
return setRange(U, SignHint, ConservativeResult);
unsigned TZ = A.countTrailingZeros();
unsigned BitWidth = A.getBitWidth();
APInt KnownZero(BitWidth, 0), KnownOne(BitWidth, 0);
- computeKnownBits(U->getOperand(0), KnownZero, KnownOne, DL, 0, AC,
- nullptr, DT);
+ computeKnownBits(U->getOperand(0), KnownZero, KnownOne,
+ F->getParent()->getDataLayout(), 0, AC, nullptr, DT);
APInt EffectiveMask =
APInt::getLowBitsSet(BitWidth, BitWidth - LZ - TZ).shl(TZ);
/// reason, return null.
static Constant *EvaluateExpression(Value *V, const Loop *L,
DenseMap<Instruction *, Constant *> &Vals,
- const DataLayout *DL,
+ const DataLayout &DL,
const TargetLibraryInfo *TLI) {
// Convenient constant check, but redundant for recursive calls.
if (Constant *C = dyn_cast<Constant>(V)) return C;
unsigned NumIterations = BEs.getZExtValue(); // must be in range
unsigned IterationNum = 0;
+ const DataLayout &DL = F->getParent()->getDataLayout();
for (; ; ++IterationNum) {
if (IterationNum == NumIterations)
return RetVal = CurrentIterVals[PN]; // Got exit value!
// Compute the value of the PHIs for the next iteration.
// EvaluateExpression adds non-phi values to the CurrentIterVals map.
DenseMap<Instruction *, Constant *> NextIterVals;
- Constant *NextPHI = EvaluateExpression(BEValue, L, CurrentIterVals, DL,
- TLI);
+ Constant *NextPHI =
+ EvaluateExpression(BEValue, L, CurrentIterVals, DL, TLI);
if (!NextPHI)
return nullptr; // Couldn't evaluate!
NextIterVals[PN] = NextPHI;
// Okay, we find a PHI node that defines the trip count of this loop. Execute
// the loop symbolically to determine when the condition gets a value of
// "ExitWhen".
-
unsigned MaxIterations = MaxBruteForceIterations; // Limit analysis.
+ const DataLayout &DL = F->getParent()->getDataLayout();
for (unsigned IterationNum = 0; IterationNum != MaxIterations;++IterationNum){
- ConstantInt *CondVal =
- dyn_cast_or_null<ConstantInt>(EvaluateExpression(Cond, L, CurrentIterVals,
- DL, TLI));
+ ConstantInt *CondVal = dyn_cast_or_null<ConstantInt>(
+ EvaluateExpression(Cond, L, CurrentIterVals, DL, TLI));
// Couldn't symbolically evaluate.
if (!CondVal) return getCouldNotCompute();
// Check to see if getSCEVAtScope actually made an improvement.
if (MadeImprovement) {
Constant *C = nullptr;
+ const DataLayout &DL = F->getParent()->getDataLayout();
if (const CmpInst *CI = dyn_cast<CmpInst>(I))
- C = ConstantFoldCompareInstOperands(CI->getPredicate(),
- Operands[0], Operands[1], DL,
- TLI);
+ C = ConstantFoldCompareInstOperands(CI->getPredicate(), Operands[0],
+ Operands[1], DL, TLI);
else if (const LoadInst *LI = dyn_cast<LoadInst>(I)) {
if (!LI->isVolatile())
C = ConstantFoldLoadFromConstPtr(Operands[0], DL);
} else
- C = ConstantFoldInstOperands(I->getOpcode(), I->getType(),
- Operands, DL, TLI);
+ C = ConstantFoldInstOperands(I->getOpcode(), I->getType(), Operands,
+ DL, TLI);
if (!C) return V;
return getSCEV(C);
}
this->F = &F;
AC = &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
- DL = &F.getParent()->getDataLayout();
TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
return false;
/// TODO: When ScalarEvolution gets a SCEVSDivExpr, this can be made
/// unnecessary; in its place, just signed-divide Ops[i] by the scale and
/// check to see if the divide was folded.
-static bool FactorOutConstant(const SCEV *&S,
- const SCEV *&Remainder,
- const SCEV *Factor,
- ScalarEvolution &SE,
- const DataLayout *DL) {
+static bool FactorOutConstant(const SCEV *&S, const SCEV *&Remainder,
+ const SCEV *Factor, ScalarEvolution &SE,
+ const DataLayout &DL) {
// Everything is divisible by one.
if (Factor->isOne())
return true;
// In a Mul, check if there is a constant operand which is a multiple
// of the given factor.
if (const SCEVMulExpr *M = dyn_cast<SCEVMulExpr>(S)) {
- if (DL) {
- // With DataLayout, the size is known. Check if there is a constant
- // operand which is a multiple of the given factor. If so, we can
- // factor it.
- const SCEVConstant *FC = cast<SCEVConstant>(Factor);
- if (const SCEVConstant *C = dyn_cast<SCEVConstant>(M->getOperand(0)))
- if (!C->getValue()->getValue().srem(FC->getValue()->getValue())) {
- SmallVector<const SCEV *, 4> NewMulOps(M->op_begin(), M->op_end());
- NewMulOps[0] =
- SE.getConstant(C->getValue()->getValue().sdiv(
- FC->getValue()->getValue()));
- S = SE.getMulExpr(NewMulOps);
- return true;
- }
- } else {
- // Without DataLayout, check if Factor can be factored out of any of the
- // Mul's operands. If so, we can just remove it.
- for (unsigned i = 0, e = M->getNumOperands(); i != e; ++i) {
- const SCEV *SOp = M->getOperand(i);
- const SCEV *Remainder = SE.getConstant(SOp->getType(), 0);
- if (FactorOutConstant(SOp, Remainder, Factor, SE, DL) &&
- Remainder->isZero()) {
- SmallVector<const SCEV *, 4> NewMulOps(M->op_begin(), M->op_end());
- NewMulOps[i] = SOp;
- S = SE.getMulExpr(NewMulOps);
- return true;
- }
+ // Size is known, check if there is a constant operand which is a multiple
+ // of the given factor. If so, we can factor it.
+ const SCEVConstant *FC = cast<SCEVConstant>(Factor);
+ if (const SCEVConstant *C = dyn_cast<SCEVConstant>(M->getOperand(0)))
+ if (!C->getValue()->getValue().srem(FC->getValue()->getValue())) {
+ SmallVector<const SCEV *, 4> NewMulOps(M->op_begin(), M->op_end());
+ NewMulOps[0] = SE.getConstant(
+ C->getValue()->getValue().sdiv(FC->getValue()->getValue()));
+ S = SE.getMulExpr(NewMulOps);
+ return true;
}
- }
}
// In an AddRec, check if both start and step are divisible.
// without the other.
SplitAddRecs(Ops, Ty, SE);
- Type *IntPtrTy = SE.DL
- ? SE.DL->getIntPtrType(PTy)
- : Type::getInt64Ty(PTy->getContext());
+ Type *IntPtrTy = DL.getIntPtrType(PTy);
// Descend down the pointer's type and attempt to convert the other
// operands into GEP indices, at each level. The first index in a GEP
for (unsigned i = 0, e = Ops.size(); i != e; ++i) {
const SCEV *Op = Ops[i];
const SCEV *Remainder = SE.getConstant(Ty, 0);
- if (FactorOutConstant(Op, Remainder, ElSize, SE, SE.DL)) {
+ if (FactorOutConstant(Op, Remainder, ElSize, SE, DL)) {
// Op now has ElSize factored out.
ScaledOps.push_back(Op);
if (!Remainder->isZero())
bool FoundFieldNo = false;
// An empty struct has no fields.
if (STy->getNumElements() == 0) break;
- if (SE.DL) {
- // With DataLayout, field offsets are known. See if a constant offset
- // falls within any of the struct fields.
- if (Ops.empty()) break;
- if (const SCEVConstant *C = dyn_cast<SCEVConstant>(Ops[0]))
- if (SE.getTypeSizeInBits(C->getType()) <= 64) {
- const StructLayout &SL = *SE.DL->getStructLayout(STy);
- uint64_t FullOffset = C->getValue()->getZExtValue();
- if (FullOffset < SL.getSizeInBytes()) {
- unsigned ElIdx = SL.getElementContainingOffset(FullOffset);
- GepIndices.push_back(
- ConstantInt::get(Type::getInt32Ty(Ty->getContext()), ElIdx));
- ElTy = STy->getTypeAtIndex(ElIdx);
- Ops[0] =
+ // Field offsets are known. See if a constant offset falls within any of
+ // the struct fields.
+ if (Ops.empty())
+ break;
+ if (const SCEVConstant *C = dyn_cast<SCEVConstant>(Ops[0]))
+ if (SE.getTypeSizeInBits(C->getType()) <= 64) {
+ const StructLayout &SL = *DL.getStructLayout(STy);
+ uint64_t FullOffset = C->getValue()->getZExtValue();
+ if (FullOffset < SL.getSizeInBytes()) {
+ unsigned ElIdx = SL.getElementContainingOffset(FullOffset);
+ GepIndices.push_back(
+ ConstantInt::get(Type::getInt32Ty(Ty->getContext()), ElIdx));
+ ElTy = STy->getTypeAtIndex(ElIdx);
+ Ops[0] =
SE.getConstant(Ty, FullOffset - SL.getElementOffset(ElIdx));
- AnyNonZeroIndices = true;
- FoundFieldNo = true;
- }
- }
- } else {
- // Without DataLayout, just check for an offsetof expression of the
- // appropriate struct type.
- for (unsigned i = 0, e = Ops.size(); i != e; ++i)
- if (const SCEVUnknown *U = dyn_cast<SCEVUnknown>(Ops[i])) {
- Type *CTy;
- Constant *FieldNo;
- if (U->isOffsetOf(CTy, FieldNo) && CTy == STy) {
- GepIndices.push_back(FieldNo);
- ElTy =
- STy->getTypeAtIndex(cast<ConstantInt>(FieldNo)->getZExtValue());
- Ops[i] = SE.getConstant(Ty, 0);
- AnyNonZeroIndices = true;
- FoundFieldNo = true;
- break;
- }
+ AnyNonZeroIndices = true;
+ FoundFieldNo = true;
}
- }
+ }
// If no struct field offsets were found, tentatively assume that
// field zero was selected (since the zero offset would obviously
// be folded away).
// Fold constant phis. They may be congruent to other constant phis and
// would confuse the logic below that expects proper IVs.
- if (Value *V = SimplifyInstruction(Phi, SE.DL, SE.TLI, SE.DT, SE.AC)) {
+ if (Value *V = SimplifyInstruction(Phi, DL, SE.TLI, SE.DT, SE.AC)) {
Phi->replaceAllUsesWith(V);
DeadInsts.push_back(Phi);
++NumElim;
/// Returns the bitwidth of the given scalar or pointer type (if unknown returns
/// 0). For vector types, returns the element type's bitwidth.
-static unsigned getBitWidth(Type *Ty, const DataLayout *TD) {
+static unsigned getBitWidth(Type *Ty, const DataLayout &DL) {
if (unsigned BitWidth = Ty->getScalarSizeInBits())
return BitWidth;
- return TD ? TD->getPointerTypeSizeInBits(Ty) : 0;
+ return DL.getPointerTypeSizeInBits(Ty);
}
// Many of these functions have internal versions that take an assumption
}
static void computeKnownBits(Value *V, APInt &KnownZero, APInt &KnownOne,
- const DataLayout *TD, unsigned Depth,
- const Query &Q);
+ const DataLayout &DL, unsigned Depth,
+ const Query &Q);
void llvm::computeKnownBits(Value *V, APInt &KnownZero, APInt &KnownOne,
- const DataLayout *TD, unsigned Depth,
+ const DataLayout &DL, unsigned Depth,
AssumptionCache *AC, const Instruction *CxtI,
const DominatorTree *DT) {
- ::computeKnownBits(V, KnownZero, KnownOne, TD, Depth,
+ ::computeKnownBits(V, KnownZero, KnownOne, DL, Depth,
Query(AC, safeCxtI(V, CxtI), DT));
}
static void ComputeSignBit(Value *V, bool &KnownZero, bool &KnownOne,
- const DataLayout *TD, unsigned Depth,
- const Query &Q);
+ const DataLayout &DL, unsigned Depth,
+ const Query &Q);
void llvm::ComputeSignBit(Value *V, bool &KnownZero, bool &KnownOne,
- const DataLayout *TD, unsigned Depth,
+ const DataLayout &DL, unsigned Depth,
AssumptionCache *AC, const Instruction *CxtI,
const DominatorTree *DT) {
- ::ComputeSignBit(V, KnownZero, KnownOne, TD, Depth,
+ ::ComputeSignBit(V, KnownZero, KnownOne, DL, Depth,
Query(AC, safeCxtI(V, CxtI), DT));
}
static bool isKnownToBeAPowerOfTwo(Value *V, bool OrZero, unsigned Depth,
- const Query &Q);
+ const Query &Q, const DataLayout &DL);
-bool llvm::isKnownToBeAPowerOfTwo(Value *V, bool OrZero, unsigned Depth,
- AssumptionCache *AC, const Instruction *CxtI,
+bool llvm::isKnownToBeAPowerOfTwo(Value *V, const DataLayout &DL, bool OrZero,
+ unsigned Depth, AssumptionCache *AC,
+ const Instruction *CxtI,
const DominatorTree *DT) {
return ::isKnownToBeAPowerOfTwo(V, OrZero, Depth,
- Query(AC, safeCxtI(V, CxtI), DT));
+ Query(AC, safeCxtI(V, CxtI), DT), DL);
}
-static bool isKnownNonZero(Value *V, const DataLayout *TD, unsigned Depth,
+static bool isKnownNonZero(Value *V, const DataLayout &DL, unsigned Depth,
const Query &Q);
-bool llvm::isKnownNonZero(Value *V, const DataLayout *TD, unsigned Depth,
+bool llvm::isKnownNonZero(Value *V, const DataLayout &DL, unsigned Depth,
AssumptionCache *AC, const Instruction *CxtI,
const DominatorTree *DT) {
- return ::isKnownNonZero(V, TD, Depth, Query(AC, safeCxtI(V, CxtI), DT));
+ return ::isKnownNonZero(V, DL, Depth, Query(AC, safeCxtI(V, CxtI), DT));
}
-static bool MaskedValueIsZero(Value *V, const APInt &Mask,
- const DataLayout *TD, unsigned Depth,
- const Query &Q);
+static bool MaskedValueIsZero(Value *V, const APInt &Mask, const DataLayout &DL,
+ unsigned Depth, const Query &Q);
-bool llvm::MaskedValueIsZero(Value *V, const APInt &Mask, const DataLayout *TD,
+bool llvm::MaskedValueIsZero(Value *V, const APInt &Mask, const DataLayout &DL,
unsigned Depth, AssumptionCache *AC,
const Instruction *CxtI, const DominatorTree *DT) {
- return ::MaskedValueIsZero(V, Mask, TD, Depth,
+ return ::MaskedValueIsZero(V, Mask, DL, Depth,
Query(AC, safeCxtI(V, CxtI), DT));
}
-static unsigned ComputeNumSignBits(Value *V, const DataLayout *TD,
+static unsigned ComputeNumSignBits(Value *V, const DataLayout &DL,
unsigned Depth, const Query &Q);
-unsigned llvm::ComputeNumSignBits(Value *V, const DataLayout *TD,
+unsigned llvm::ComputeNumSignBits(Value *V, const DataLayout &DL,
unsigned Depth, AssumptionCache *AC,
const Instruction *CxtI,
const DominatorTree *DT) {
- return ::ComputeNumSignBits(V, TD, Depth, Query(AC, safeCxtI(V, CxtI), DT));
+ return ::ComputeNumSignBits(V, DL, Depth, Query(AC, safeCxtI(V, CxtI), DT));
}
static void computeKnownBitsAddSub(bool Add, Value *Op0, Value *Op1, bool NSW,
APInt &KnownZero, APInt &KnownOne,
APInt &KnownZero2, APInt &KnownOne2,
- const DataLayout *TD, unsigned Depth,
+ const DataLayout &DL, unsigned Depth,
const Query &Q) {
if (!Add) {
if (ConstantInt *CLHS = dyn_cast<ConstantInt>(Op0)) {
unsigned NLZ = (CLHS->getValue()+1).countLeadingZeros();
// NLZ can't be BitWidth with no sign bit
APInt MaskV = APInt::getHighBitsSet(BitWidth, NLZ+1);
- computeKnownBits(Op1, KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(Op1, KnownZero2, KnownOne2, DL, Depth + 1, Q);
// If all of the MaskV bits are known to be zero, then we know the
// output top bits are zero, because we now know that the output is
// If an initial sequence of bits in the result is not needed, the
// corresponding bits in the operands are not needed.
APInt LHSKnownZero(BitWidth, 0), LHSKnownOne(BitWidth, 0);
- computeKnownBits(Op0, LHSKnownZero, LHSKnownOne, TD, Depth+1, Q);
- computeKnownBits(Op1, KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(Op0, LHSKnownZero, LHSKnownOne, DL, Depth + 1, Q);
+ computeKnownBits(Op1, KnownZero2, KnownOne2, DL, Depth + 1, Q);
// Carry in a 1 for a subtract, rather than a 0.
APInt CarryIn(BitWidth, 0);
static void computeKnownBitsMul(Value *Op0, Value *Op1, bool NSW,
APInt &KnownZero, APInt &KnownOne,
APInt &KnownZero2, APInt &KnownOne2,
- const DataLayout *TD, unsigned Depth,
+ const DataLayout &DL, unsigned Depth,
const Query &Q) {
unsigned BitWidth = KnownZero.getBitWidth();
- computeKnownBits(Op1, KnownZero, KnownOne, TD, Depth+1, Q);
- computeKnownBits(Op0, KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(Op1, KnownZero, KnownOne, DL, Depth + 1, Q);
+ computeKnownBits(Op0, KnownZero2, KnownOne2, DL, Depth + 1, Q);
bool isKnownNegative = false;
bool isKnownNonNegative = false;
// negative or zero.
if (!isKnownNonNegative)
isKnownNegative = (isKnownNegativeOp1 && isKnownNonNegativeOp0 &&
- isKnownNonZero(Op0, TD, Depth, Q)) ||
+ isKnownNonZero(Op0, DL, Depth, Q)) ||
(isKnownNegativeOp0 && isKnownNonNegativeOp1 &&
- isKnownNonZero(Op1, TD, Depth, Q));
+ isKnownNonZero(Op1, DL, Depth, Q));
}
}
return false;
}
-static bool isValidAssumeForContext(Value *V, const Query &Q,
- const DataLayout *DL) {
+static bool isValidAssumeForContext(Value *V, const Query &Q) {
Instruction *Inv = cast<Instruction>(V);
// There are two restrictions on the use of an assume:
for (BasicBlock::const_iterator I =
std::next(BasicBlock::const_iterator(Q.CxtI)),
IE(Inv); I != IE; ++I)
- if (!isSafeToSpeculativelyExecute(I, DL) &&
- !isAssumeLikeIntrinsic(I))
+ if (!isSafeToSpeculativelyExecute(I) && !isAssumeLikeIntrinsic(I))
return false;
return !isEphemeralValueOf(Inv, Q.CxtI);
for (BasicBlock::const_iterator I =
std::next(BasicBlock::const_iterator(Q.CxtI)),
IE(Inv); I != IE; ++I)
- if (!isSafeToSpeculativelyExecute(I, DL) &&
- !isAssumeLikeIntrinsic(I))
+ if (!isSafeToSpeculativelyExecute(I) && !isAssumeLikeIntrinsic(I))
return false;
return !isEphemeralValueOf(Inv, Q.CxtI);
bool llvm::isValidAssumeForContext(const Instruction *I,
const Instruction *CxtI,
- const DataLayout *DL,
const DominatorTree *DT) {
- return ::isValidAssumeForContext(const_cast<Instruction*>(I),
- Query(nullptr, CxtI, DT), DL);
+ return ::isValidAssumeForContext(const_cast<Instruction *>(I),
+ Query(nullptr, CxtI, DT));
}
template<typename LHS, typename RHS>
}
static void computeKnownBitsFromAssume(Value *V, APInt &KnownZero,
- APInt &KnownOne,
- const DataLayout *DL,
+ APInt &KnownOne, const DataLayout &DL,
unsigned Depth, const Query &Q) {
// Use of assumptions is context-sensitive. If we don't have a context, we
// cannot use them!
Value *Arg = I->getArgOperand(0);
- if (Arg == V &&
- isValidAssumeForContext(I, Q, DL)) {
+ if (Arg == V && isValidAssumeForContext(I, Q)) {
assert(BitWidth == 1 && "assume operand is not i1?");
KnownZero.clearAllBits();
KnownOne.setAllBits();
ConstantInt *C;
// assume(v = a)
if (match(Arg, m_c_ICmp(Pred, m_V, m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
KnownZero |= RHSKnownZero;
KnownOne |= RHSKnownOne;
// assume(v & b = a)
- } else if (match(Arg, m_c_ICmp(Pred, m_c_And(m_V, m_Value(B)),
- m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ } else if (match(Arg,
+ m_c_ICmp(Pred, m_c_And(m_V, m_Value(B)), m_Value(A))) &&
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
APInt MaskKnownZero(BitWidth, 0), MaskKnownOne(BitWidth, 0);
// assume(~(v & b) = a)
} else if (match(Arg, m_c_ICmp(Pred, m_Not(m_c_And(m_V, m_Value(B))),
m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
APInt MaskKnownZero(BitWidth, 0), MaskKnownOne(BitWidth, 0);
KnownZero |= RHSKnownOne & MaskKnownOne;
KnownOne |= RHSKnownZero & MaskKnownOne;
// assume(v | b = a)
- } else if (match(Arg, m_c_ICmp(Pred, m_c_Or(m_V, m_Value(B)),
- m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ } else if (match(Arg,
+ m_c_ICmp(Pred, m_c_Or(m_V, m_Value(B)), m_Value(A))) &&
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
APInt BKnownZero(BitWidth, 0), BKnownOne(BitWidth, 0);
// assume(~(v | b) = a)
} else if (match(Arg, m_c_ICmp(Pred, m_Not(m_c_Or(m_V, m_Value(B))),
m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
APInt BKnownZero(BitWidth, 0), BKnownOne(BitWidth, 0);
KnownZero |= RHSKnownOne & BKnownZero;
KnownOne |= RHSKnownZero & BKnownZero;
// assume(v ^ b = a)
- } else if (match(Arg, m_c_ICmp(Pred, m_c_Xor(m_V, m_Value(B)),
- m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ } else if (match(Arg,
+ m_c_ICmp(Pred, m_c_Xor(m_V, m_Value(B)), m_Value(A))) &&
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
APInt BKnownZero(BitWidth, 0), BKnownOne(BitWidth, 0);
// assume(~(v ^ b) = a)
} else if (match(Arg, m_c_ICmp(Pred, m_Not(m_c_Xor(m_V, m_Value(B))),
m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
APInt BKnownZero(BitWidth, 0), BKnownOne(BitWidth, 0);
// assume(v << c = a)
} else if (match(Arg, m_c_ICmp(Pred, m_Shl(m_V, m_ConstantInt(C)),
m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
// For those bits in RHS that are known, we can propagate them to known
// assume(~(v << c) = a)
} else if (match(Arg, m_c_ICmp(Pred, m_Not(m_Shl(m_V, m_ConstantInt(C))),
m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
// For those bits in RHS that are known, we can propagate them inverted
// assume(v >> c = a)
} else if (match(Arg,
m_c_ICmp(Pred, m_CombineOr(m_LShr(m_V, m_ConstantInt(C)),
- m_AShr(m_V,
- m_ConstantInt(C))),
- m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ m_AShr(m_V, m_ConstantInt(C))),
+ m_Value(A))) &&
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
// For those bits in RHS that are known, we can propagate them to known
KnownOne |= RHSKnownOne << C->getZExtValue();
// assume(~(v >> c) = a)
} else if (match(Arg, m_c_ICmp(Pred, m_Not(m_CombineOr(
- m_LShr(m_V, m_ConstantInt(C)),
- m_AShr(m_V, m_ConstantInt(C)))),
+ m_LShr(m_V, m_ConstantInt(C)),
+ m_AShr(m_V, m_ConstantInt(C)))),
m_Value(A))) &&
- Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_EQ && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
// For those bits in RHS that are known, we can propagate them inverted
KnownOne |= RHSKnownZero << C->getZExtValue();
// assume(v >=_s c) where c is non-negative
} else if (match(Arg, m_ICmp(Pred, m_V, m_Value(A))) &&
- Pred == ICmpInst::ICMP_SGE &&
- isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_SGE && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
}
// assume(v >_s c) where c is at least -1.
} else if (match(Arg, m_ICmp(Pred, m_V, m_Value(A))) &&
- Pred == ICmpInst::ICMP_SGT &&
- isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_SGT && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
}
// assume(v <=_s c) where c is negative
} else if (match(Arg, m_ICmp(Pred, m_V, m_Value(A))) &&
- Pred == ICmpInst::ICMP_SLE &&
- isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_SLE && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
}
// assume(v <_s c) where c is non-positive
} else if (match(Arg, m_ICmp(Pred, m_V, m_Value(A))) &&
- Pred == ICmpInst::ICMP_SLT &&
- isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_SLT && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
}
// assume(v <=_u c)
} else if (match(Arg, m_ICmp(Pred, m_V, m_Value(A))) &&
- Pred == ICmpInst::ICMP_ULE &&
- isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_ULE && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
APInt::getHighBitsSet(BitWidth, RHSKnownZero.countLeadingOnes());
// assume(v <_u c)
} else if (match(Arg, m_ICmp(Pred, m_V, m_Value(A))) &&
- Pred == ICmpInst::ICMP_ULT &&
- isValidAssumeForContext(I, Q, DL)) {
+ Pred == ICmpInst::ICMP_ULT && isValidAssumeForContext(I, Q)) {
APInt RHSKnownZero(BitWidth, 0), RHSKnownOne(BitWidth, 0);
computeKnownBits(A, RHSKnownZero, RHSKnownOne, DL, Depth+1, Query(Q, I));
// Whatever high bits in c are zero are known to be zero (if c is a power
// of 2, then one more).
- if (isKnownToBeAPowerOfTwo(A, false, Depth+1, Query(Q, I)))
+ if (isKnownToBeAPowerOfTwo(A, false, Depth + 1, Query(Q, I), DL))
KnownZero |=
APInt::getHighBitsSet(BitWidth, RHSKnownZero.countLeadingOnes()+1);
else
/// this won't lose us code quality.
///
/// This function is defined on values with integer type, values with pointer
-/// type (but only if TD is non-null), and vectors of integers. In the case
+/// type, and vectors of integers. In the case
/// where V is a vector, known zero, and known one values are the
/// same width as the vector element, and the bit is set only if it is true
/// for all of the elements in the vector.
void computeKnownBits(Value *V, APInt &KnownZero, APInt &KnownOne,
- const DataLayout *TD, unsigned Depth,
- const Query &Q) {
+ const DataLayout &DL, unsigned Depth, const Query &Q) {
assert(V && "No Value?");
assert(Depth <= MaxDepth && "Limit Search Depth");
unsigned BitWidth = KnownZero.getBitWidth();
assert((V->getType()->isIntOrIntVectorTy() ||
V->getType()->getScalarType()->isPointerTy()) &&
"Not integer or pointer type!");
- assert((!TD ||
- TD->getTypeSizeInBits(V->getType()->getScalarType()) == BitWidth) &&
+ assert((DL.getTypeSizeInBits(V->getType()->getScalarType()) == BitWidth) &&
(!V->getType()->isIntOrIntVectorTy() ||
V->getType()->getScalarSizeInBits() == BitWidth) &&
KnownZero.getBitWidth() == BitWidth &&
// The address of an aligned GlobalValue has trailing zeros.
if (auto *GO = dyn_cast<GlobalObject>(V)) {
unsigned Align = GO->getAlignment();
- if (Align == 0 && TD) {
+ if (Align == 0) {
if (auto *GVar = dyn_cast<GlobalVariable>(GO)) {
Type *ObjectType = GVar->getType()->getElementType();
if (ObjectType->isSized()) {
// it the preferred alignment. Otherwise, we have to assume that it
// may only have the minimum ABI alignment.
if (!GVar->isDeclaration() && !GVar->isWeakForLinker())
- Align = TD->getPreferredAlignment(GVar);
+ Align = DL.getPreferredAlignment(GVar);
else
- Align = TD->getABITypeAlignment(ObjectType);
+ Align = DL.getABITypeAlignment(ObjectType);
}
}
}
if (Argument *A = dyn_cast<Argument>(V)) {
unsigned Align = A->getType()->isPointerTy() ? A->getParamAlignment() : 0;
- if (!Align && TD && A->hasStructRetAttr()) {
+ if (!Align && A->hasStructRetAttr()) {
// An sret parameter has at least the ABI alignment of the return type.
Type *EltTy = cast<PointerType>(A->getType())->getElementType();
if (EltTy->isSized())
- Align = TD->getABITypeAlignment(EltTy);
+ Align = DL.getABITypeAlignment(EltTy);
}
if (Align)
// Don't give up yet... there might be an assumption that provides more
// information...
- computeKnownBitsFromAssume(V, KnownZero, KnownOne, TD, Depth, Q);
+ computeKnownBitsFromAssume(V, KnownZero, KnownOne, DL, Depth, Q);
return;
}
// the bits of its aliasee.
if (GlobalAlias *GA = dyn_cast<GlobalAlias>(V)) {
if (!GA->mayBeOverridden())
- computeKnownBits(GA->getAliasee(), KnownZero, KnownOne, TD, Depth + 1, Q);
+ computeKnownBits(GA->getAliasee(), KnownZero, KnownOne, DL, Depth + 1, Q);
return;
}
// Check whether a nearby assume intrinsic can determine some known bits.
- computeKnownBitsFromAssume(V, KnownZero, KnownOne, TD, Depth, Q);
+ computeKnownBitsFromAssume(V, KnownZero, KnownOne, DL, Depth, Q);
Operator *I = dyn_cast<Operator>(V);
if (!I) return;
break;
case Instruction::And: {
// If either the LHS or the RHS are Zero, the result is zero.
- computeKnownBits(I->getOperand(1), KnownZero, KnownOne, TD, Depth+1, Q);
- computeKnownBits(I->getOperand(0), KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(1), KnownZero, KnownOne, DL, Depth + 1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero2, KnownOne2, DL, Depth + 1, Q);
// Output known-1 bits are only known if set in both the LHS & RHS.
KnownOne &= KnownOne2;
break;
}
case Instruction::Or: {
- computeKnownBits(I->getOperand(1), KnownZero, KnownOne, TD, Depth+1, Q);
- computeKnownBits(I->getOperand(0), KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(1), KnownZero, KnownOne, DL, Depth + 1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero2, KnownOne2, DL, Depth + 1, Q);
// Output known-0 bits are only known if clear in both the LHS & RHS.
KnownZero &= KnownZero2;
break;
}
case Instruction::Xor: {
- computeKnownBits(I->getOperand(1), KnownZero, KnownOne, TD, Depth+1, Q);
- computeKnownBits(I->getOperand(0), KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(1), KnownZero, KnownOne, DL, Depth + 1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero2, KnownOne2, DL, Depth + 1, Q);
// Output known-0 bits are known if clear or set in both the LHS & RHS.
APInt KnownZeroOut = (KnownZero & KnownZero2) | (KnownOne & KnownOne2);
}
case Instruction::Mul: {
bool NSW = cast<OverflowingBinaryOperator>(I)->hasNoSignedWrap();
- computeKnownBitsMul(I->getOperand(0), I->getOperand(1), NSW,
- KnownZero, KnownOne, KnownZero2, KnownOne2, TD,
- Depth, Q);
+ computeKnownBitsMul(I->getOperand(0), I->getOperand(1), NSW, KnownZero,
+ KnownOne, KnownZero2, KnownOne2, DL, Depth, Q);
break;
}
case Instruction::UDiv: {
// For the purposes of computing leading zeros we can conservatively
// treat a udiv as a logical right shift by the power of 2 known to
// be less than the denominator.
- computeKnownBits(I->getOperand(0), KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero2, KnownOne2, DL, Depth + 1, Q);
unsigned LeadZ = KnownZero2.countLeadingOnes();
KnownOne2.clearAllBits();
KnownZero2.clearAllBits();
- computeKnownBits(I->getOperand(1), KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(1), KnownZero2, KnownOne2, DL, Depth + 1, Q);
unsigned RHSUnknownLeadingOnes = KnownOne2.countLeadingZeros();
if (RHSUnknownLeadingOnes != BitWidth)
LeadZ = std::min(BitWidth,
break;
}
case Instruction::Select:
- computeKnownBits(I->getOperand(2), KnownZero, KnownOne, TD, Depth+1, Q);
- computeKnownBits(I->getOperand(1), KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(2), KnownZero, KnownOne, DL, Depth + 1, Q);
+ computeKnownBits(I->getOperand(1), KnownZero2, KnownOne2, DL, Depth + 1, Q);
// Only known if known in both the LHS and RHS.
KnownOne &= KnownOne2;
case Instruction::PtrToInt:
case Instruction::IntToPtr:
case Instruction::AddrSpaceCast: // Pointers could be different sizes.
- // We can't handle these if we don't know the pointer size.
- if (!TD) break;
// FALL THROUGH and handle them the same as zext/trunc.
case Instruction::ZExt:
case Instruction::Trunc: {
unsigned SrcBitWidth;
// Note that we handle pointer operands here because of inttoptr/ptrtoint
// which fall through here.
- if(TD) {
- SrcBitWidth = TD->getTypeSizeInBits(SrcTy->getScalarType());
- } else {
- SrcBitWidth = SrcTy->getScalarSizeInBits();
- if (!SrcBitWidth) break;
- }
+ SrcBitWidth = DL.getTypeSizeInBits(SrcTy->getScalarType());
assert(SrcBitWidth && "SrcBitWidth can't be zero");
KnownZero = KnownZero.zextOrTrunc(SrcBitWidth);
KnownOne = KnownOne.zextOrTrunc(SrcBitWidth);
- computeKnownBits(I->getOperand(0), KnownZero, KnownOne, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero, KnownOne, DL, Depth + 1, Q);
KnownZero = KnownZero.zextOrTrunc(BitWidth);
KnownOne = KnownOne.zextOrTrunc(BitWidth);
// Any top bits are known to be zero.
// TODO: For now, not handling conversions like:
// (bitcast i64 %x to <2 x i32>)
!I->getType()->isVectorTy()) {
- computeKnownBits(I->getOperand(0), KnownZero, KnownOne, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero, KnownOne, DL, Depth + 1, Q);
break;
}
break;
KnownZero = KnownZero.trunc(SrcBitWidth);
KnownOne = KnownOne.trunc(SrcBitWidth);
- computeKnownBits(I->getOperand(0), KnownZero, KnownOne, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero, KnownOne, DL, Depth + 1, Q);
KnownZero = KnownZero.zext(BitWidth);
KnownOne = KnownOne.zext(BitWidth);
// (shl X, C1) & C2 == 0 iff (X & C2 >>u C1) == 0
if (ConstantInt *SA = dyn_cast<ConstantInt>(I->getOperand(1))) {
uint64_t ShiftAmt = SA->getLimitedValue(BitWidth);
- computeKnownBits(I->getOperand(0), KnownZero, KnownOne, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero, KnownOne, DL, Depth + 1, Q);
KnownZero <<= ShiftAmt;
KnownOne <<= ShiftAmt;
KnownZero |= APInt::getLowBitsSet(BitWidth, ShiftAmt); // low bits known 0
uint64_t ShiftAmt = SA->getLimitedValue(BitWidth);
// Unsigned shift right.
- computeKnownBits(I->getOperand(0), KnownZero, KnownOne, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero, KnownOne, DL, Depth + 1, Q);
KnownZero = APIntOps::lshr(KnownZero, ShiftAmt);
KnownOne = APIntOps::lshr(KnownOne, ShiftAmt);
// high bits known zero.
uint64_t ShiftAmt = SA->getLimitedValue(BitWidth-1);
// Signed shift right.
- computeKnownBits(I->getOperand(0), KnownZero, KnownOne, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero, KnownOne, DL, Depth + 1, Q);
KnownZero = APIntOps::lshr(KnownZero, ShiftAmt);
KnownOne = APIntOps::lshr(KnownOne, ShiftAmt);
case Instruction::Sub: {
bool NSW = cast<OverflowingBinaryOperator>(I)->hasNoSignedWrap();
computeKnownBitsAddSub(false, I->getOperand(0), I->getOperand(1), NSW,
- KnownZero, KnownOne, KnownZero2, KnownOne2, TD,
- Depth, Q);
+ KnownZero, KnownOne, KnownZero2, KnownOne2, DL,
+ Depth, Q);
break;
}
case Instruction::Add: {
bool NSW = cast<OverflowingBinaryOperator>(I)->hasNoSignedWrap();
computeKnownBitsAddSub(true, I->getOperand(0), I->getOperand(1), NSW,
- KnownZero, KnownOne, KnownZero2, KnownOne2, TD,
- Depth, Q);
+ KnownZero, KnownOne, KnownZero2, KnownOne2, DL,
+ Depth, Q);
break;
}
case Instruction::SRem:
APInt RA = Rem->getValue().abs();
if (RA.isPowerOf2()) {
APInt LowBits = RA - 1;
- computeKnownBits(I->getOperand(0), KnownZero2, KnownOne2, TD,
- Depth+1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero2, KnownOne2, DL, Depth + 1,
+ Q);
// The low bits of the first operand are unchanged by the srem.
KnownZero = KnownZero2 & LowBits;
// remainder is zero.
if (KnownZero.isNonNegative()) {
APInt LHSKnownZero(BitWidth, 0), LHSKnownOne(BitWidth, 0);
- computeKnownBits(I->getOperand(0), LHSKnownZero, LHSKnownOne, TD,
- Depth+1, Q);
+ computeKnownBits(I->getOperand(0), LHSKnownZero, LHSKnownOne, DL,
+ Depth + 1, Q);
// If it's known zero, our sign bit is also zero.
if (LHSKnownZero.isNegative())
KnownZero.setBit(BitWidth - 1);
APInt RA = Rem->getValue();
if (RA.isPowerOf2()) {
APInt LowBits = (RA - 1);
- computeKnownBits(I->getOperand(0), KnownZero, KnownOne, TD,
- Depth+1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero, KnownOne, DL, Depth + 1,
+ Q);
KnownZero |= ~LowBits;
KnownOne &= LowBits;
break;
// Since the result is less than or equal to either operand, any leading
// zero bits in either operand must also exist in the result.
- computeKnownBits(I->getOperand(0), KnownZero, KnownOne, TD, Depth+1, Q);
- computeKnownBits(I->getOperand(1), KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(I->getOperand(0), KnownZero, KnownOne, DL, Depth + 1, Q);
+ computeKnownBits(I->getOperand(1), KnownZero2, KnownOne2, DL, Depth + 1, Q);
unsigned Leaders = std::max(KnownZero.countLeadingOnes(),
KnownZero2.countLeadingOnes());
case Instruction::Alloca: {
AllocaInst *AI = cast<AllocaInst>(V);
unsigned Align = AI->getAlignment();
- if (Align == 0 && TD)
- Align = TD->getABITypeAlignment(AI->getType()->getElementType());
+ if (Align == 0)
+ Align = DL.getABITypeAlignment(AI->getType()->getElementType());
if (Align > 0)
KnownZero = APInt::getLowBitsSet(BitWidth, countTrailingZeros(Align));
// Analyze all of the subscripts of this getelementptr instruction
// to determine if we can prove known low zero bits.
APInt LocalKnownZero(BitWidth, 0), LocalKnownOne(BitWidth, 0);
- computeKnownBits(I->getOperand(0), LocalKnownZero, LocalKnownOne, TD,
- Depth+1, Q);
+ computeKnownBits(I->getOperand(0), LocalKnownZero, LocalKnownOne, DL,
+ Depth + 1, Q);
unsigned TrailZ = LocalKnownZero.countTrailingOnes();
gep_type_iterator GTI = gep_type_begin(I);
Value *Index = I->getOperand(i);
if (StructType *STy = dyn_cast<StructType>(*GTI)) {
// Handle struct member offset arithmetic.
- if (!TD) {
- TrailZ = 0;
- break;
- }
// Handle case when index is vector zeroinitializer
Constant *CIndex = cast<Constant>(Index);
Index = CIndex->getSplatValue();
unsigned Idx = cast<ConstantInt>(Index)->getZExtValue();
- const StructLayout *SL = TD->getStructLayout(STy);
+ const StructLayout *SL = DL.getStructLayout(STy);
uint64_t Offset = SL->getElementOffset(Idx);
TrailZ = std::min<unsigned>(TrailZ,
countTrailingZeros(Offset));
break;
}
unsigned GEPOpiBits = Index->getType()->getScalarSizeInBits();
- uint64_t TypeSize = TD ? TD->getTypeAllocSize(IndexedTy) : 1;
+ uint64_t TypeSize = DL.getTypeAllocSize(IndexedTy);
LocalKnownZero = LocalKnownOne = APInt(GEPOpiBits, 0);
- computeKnownBits(Index, LocalKnownZero, LocalKnownOne, TD, Depth+1, Q);
+ computeKnownBits(Index, LocalKnownZero, LocalKnownOne, DL, Depth + 1,
+ Q);
TrailZ = std::min(TrailZ,
unsigned(countTrailingZeros(TypeSize) +
LocalKnownZero.countTrailingOnes()));
break;
// Ok, we have a PHI of the form L op= R. Check for low
// zero bits.
- computeKnownBits(R, KnownZero2, KnownOne2, TD, Depth+1, Q);
+ computeKnownBits(R, KnownZero2, KnownOne2, DL, Depth + 1, Q);
// We need to take the minimum number of known bits
APInt KnownZero3(KnownZero), KnownOne3(KnownOne);
- computeKnownBits(L, KnownZero3, KnownOne3, TD, Depth+1, Q);
+ computeKnownBits(L, KnownZero3, KnownOne3, DL, Depth + 1, Q);
KnownZero = APInt::getLowBitsSet(BitWidth,
std::min(KnownZero2.countTrailingOnes(),
KnownOne2 = APInt(BitWidth, 0);
// Recurse, but cap the recursion to one level, because we don't
// want to waste time spinning around in loops.
- computeKnownBits(P->getIncomingValue(i), KnownZero2, KnownOne2, TD,
- MaxDepth-1, Q);
+ computeKnownBits(P->getIncomingValue(i), KnownZero2, KnownOne2, DL,
+ MaxDepth - 1, Q);
KnownZero &= KnownZero2;
KnownOne &= KnownOne2;
// If all bits have been ruled out, there's no need to check
case Intrinsic::sadd_with_overflow:
computeKnownBitsAddSub(true, II->getArgOperand(0),
II->getArgOperand(1), false, KnownZero,
- KnownOne, KnownZero2, KnownOne2, TD, Depth, Q);
+ KnownOne, KnownZero2, KnownOne2, DL, Depth, Q);
break;
case Intrinsic::usub_with_overflow:
case Intrinsic::ssub_with_overflow:
computeKnownBitsAddSub(false, II->getArgOperand(0),
II->getArgOperand(1), false, KnownZero,
- KnownOne, KnownZero2, KnownOne2, TD, Depth, Q);
+ KnownOne, KnownZero2, KnownOne2, DL, Depth, Q);
break;
case Intrinsic::umul_with_overflow:
case Intrinsic::smul_with_overflow:
- computeKnownBitsMul(II->getArgOperand(0), II->getArgOperand(1),
- false, KnownZero, KnownOne,
- KnownZero2, KnownOne2, TD, Depth, Q);
+ computeKnownBitsMul(II->getArgOperand(0), II->getArgOperand(1), false,
+ KnownZero, KnownOne, KnownZero2, KnownOne2, DL,
+ Depth, Q);
break;
}
}
/// Determine whether the sign bit is known to be zero or one.
/// Convenience wrapper around computeKnownBits.
void ComputeSignBit(Value *V, bool &KnownZero, bool &KnownOne,
- const DataLayout *TD, unsigned Depth,
- const Query &Q) {
- unsigned BitWidth = getBitWidth(V->getType(), TD);
+ const DataLayout &DL, unsigned Depth, const Query &Q) {
+ unsigned BitWidth = getBitWidth(V->getType(), DL);
if (!BitWidth) {
KnownZero = false;
KnownOne = false;
}
APInt ZeroBits(BitWidth, 0);
APInt OneBits(BitWidth, 0);
- computeKnownBits(V, ZeroBits, OneBits, TD, Depth, Q);
+ computeKnownBits(V, ZeroBits, OneBits, DL, Depth, Q);
KnownOne = OneBits[BitWidth - 1];
KnownZero = ZeroBits[BitWidth - 1];
}
/// be a power of two when defined. Supports values with integer or pointer
/// types and vectors of integers.
bool isKnownToBeAPowerOfTwo(Value *V, bool OrZero, unsigned Depth,
- const Query &Q) {
+ const Query &Q, const DataLayout &DL) {
if (Constant *C = dyn_cast<Constant>(V)) {
if (C->isNullValue())
return OrZero;
// A shift of a power of two is a power of two or zero.
if (OrZero && (match(V, m_Shl(m_Value(X), m_Value())) ||
match(V, m_Shr(m_Value(X), m_Value()))))
- return isKnownToBeAPowerOfTwo(X, /*OrZero*/true, Depth, Q);
+ return isKnownToBeAPowerOfTwo(X, /*OrZero*/ true, Depth, Q, DL);
if (ZExtInst *ZI = dyn_cast<ZExtInst>(V))
- return isKnownToBeAPowerOfTwo(ZI->getOperand(0), OrZero, Depth, Q);
+ return isKnownToBeAPowerOfTwo(ZI->getOperand(0), OrZero, Depth, Q, DL);
if (SelectInst *SI = dyn_cast<SelectInst>(V))
- return
- isKnownToBeAPowerOfTwo(SI->getTrueValue(), OrZero, Depth, Q) &&
- isKnownToBeAPowerOfTwo(SI->getFalseValue(), OrZero, Depth, Q);
+ return isKnownToBeAPowerOfTwo(SI->getTrueValue(), OrZero, Depth, Q, DL) &&
+ isKnownToBeAPowerOfTwo(SI->getFalseValue(), OrZero, Depth, Q, DL);
if (OrZero && match(V, m_And(m_Value(X), m_Value(Y)))) {
// A power of two and'd with anything is a power of two or zero.
- if (isKnownToBeAPowerOfTwo(X, /*OrZero*/true, Depth, Q) ||
- isKnownToBeAPowerOfTwo(Y, /*OrZero*/true, Depth, Q))
+ if (isKnownToBeAPowerOfTwo(X, /*OrZero*/ true, Depth, Q, DL) ||
+ isKnownToBeAPowerOfTwo(Y, /*OrZero*/ true, Depth, Q, DL))
return true;
// X & (-X) is always a power of two or zero.
if (match(X, m_Neg(m_Specific(Y))) || match(Y, m_Neg(m_Specific(X))))
if (OrZero || VOBO->hasNoUnsignedWrap() || VOBO->hasNoSignedWrap()) {
if (match(X, m_And(m_Specific(Y), m_Value())) ||
match(X, m_And(m_Value(), m_Specific(Y))))
- if (isKnownToBeAPowerOfTwo(Y, OrZero, Depth, Q))
+ if (isKnownToBeAPowerOfTwo(Y, OrZero, Depth, Q, DL))
return true;
if (match(Y, m_And(m_Specific(X), m_Value())) ||
match(Y, m_And(m_Value(), m_Specific(X))))
- if (isKnownToBeAPowerOfTwo(X, OrZero, Depth, Q))
+ if (isKnownToBeAPowerOfTwo(X, OrZero, Depth, Q, DL))
return true;
unsigned BitWidth = V->getType()->getScalarSizeInBits();
APInt LHSZeroBits(BitWidth, 0), LHSOneBits(BitWidth, 0);
- computeKnownBits(X, LHSZeroBits, LHSOneBits, nullptr, Depth, Q);
+ computeKnownBits(X, LHSZeroBits, LHSOneBits, DL, Depth, Q);
APInt RHSZeroBits(BitWidth, 0), RHSOneBits(BitWidth, 0);
- computeKnownBits(Y, RHSZeroBits, RHSOneBits, nullptr, Depth, Q);
+ computeKnownBits(Y, RHSZeroBits, RHSOneBits, DL, Depth, Q);
// If i8 V is a power of two or zero:
// ZeroBits: 1 1 1 0 1 1 1 1
// ~ZeroBits: 0 0 0 1 0 0 0 0
if (match(V, m_Exact(m_LShr(m_Value(), m_Value()))) ||
match(V, m_Exact(m_UDiv(m_Value(), m_Value())))) {
return isKnownToBeAPowerOfTwo(cast<Operator>(V)->getOperand(0), OrZero,
- Depth, Q);
+ Depth, Q, DL);
}
return false;
/// to be non-null.
///
/// Currently this routine does not support vector GEPs.
-static bool isGEPKnownNonNull(GEPOperator *GEP, const DataLayout *DL,
+static bool isGEPKnownNonNull(GEPOperator *GEP, const DataLayout &DL,
unsigned Depth, const Query &Q) {
if (!GEP->isInBounds() || GEP->getPointerAddressSpace() != 0)
return false;
if (isKnownNonZero(GEP->getPointerOperand(), DL, Depth, Q))
return true;
- // Past this, if we don't have DataLayout, we can't do much.
- if (!DL)
- return false;
-
// Walk the GEP operands and see if any operand introduces a non-zero offset.
// If so, then the GEP cannot produce a null pointer, as doing so would
// inherently violate the inbounds contract within address space zero.
if (StructType *STy = dyn_cast<StructType>(*GTI)) {
ConstantInt *OpC = cast<ConstantInt>(GTI.getOperand());
unsigned ElementIdx = OpC->getZExtValue();
- const StructLayout *SL = DL->getStructLayout(STy);
+ const StructLayout *SL = DL.getStructLayout(STy);
uint64_t ElementOffset = SL->getElementOffset(ElementIdx);
if (ElementOffset > 0)
return true;
}
// If we have a zero-sized type, the index doesn't matter. Keep looping.
- if (DL->getTypeAllocSize(GTI.getIndexedType()) == 0)
+ if (DL.getTypeAllocSize(GTI.getIndexedType()) == 0)
continue;
// Fast path the constant operand case both for efficiency and so we don't
/// For vectors return true if every element is known to be non-zero when
/// defined. Supports values with integer or pointer type and vectors of
/// integers.
-bool isKnownNonZero(Value *V, const DataLayout *TD, unsigned Depth,
+bool isKnownNonZero(Value *V, const DataLayout &DL, unsigned Depth,
const Query &Q) {
if (Constant *C = dyn_cast<Constant>(V)) {
if (C->isNullValue())
if (isKnownNonNull(V))
return true;
if (GEPOperator *GEP = dyn_cast<GEPOperator>(V))
- if (isGEPKnownNonNull(GEP, TD, Depth, Q))
+ if (isGEPKnownNonNull(GEP, DL, Depth, Q))
return true;
}
- unsigned BitWidth = getBitWidth(V->getType()->getScalarType(), TD);
+ unsigned BitWidth = getBitWidth(V->getType()->getScalarType(), DL);
// X | Y != 0 if X != 0 or Y != 0.
Value *X = nullptr, *Y = nullptr;
if (match(V, m_Or(m_Value(X), m_Value(Y))))
- return isKnownNonZero(X, TD, Depth, Q) ||
- isKnownNonZero(Y, TD, Depth, Q);
+ return isKnownNonZero(X, DL, Depth, Q) || isKnownNonZero(Y, DL, Depth, Q);
// ext X != 0 if X != 0.
if (isa<SExtInst>(V) || isa<ZExtInst>(V))
- return isKnownNonZero(cast<Instruction>(V)->getOperand(0), TD, Depth, Q);
+ return isKnownNonZero(cast<Instruction>(V)->getOperand(0), DL, Depth, Q);
// shl X, Y != 0 if X is odd. Note that the value of the shift is undefined
// if the lowest bit is shifted off the end.
// shl nuw can't remove any non-zero bits.
OverflowingBinaryOperator *BO = cast<OverflowingBinaryOperator>(V);
if (BO->hasNoUnsignedWrap())
- return isKnownNonZero(X, TD, Depth, Q);
+ return isKnownNonZero(X, DL, Depth, Q);
APInt KnownZero(BitWidth, 0);
APInt KnownOne(BitWidth, 0);
- computeKnownBits(X, KnownZero, KnownOne, TD, Depth, Q);
+ computeKnownBits(X, KnownZero, KnownOne, DL, Depth, Q);
if (KnownOne[0])
return true;
}
// shr exact can only shift out zero bits.
PossiblyExactOperator *BO = cast<PossiblyExactOperator>(V);
if (BO->isExact())
- return isKnownNonZero(X, TD, Depth, Q);
+ return isKnownNonZero(X, DL, Depth, Q);
bool XKnownNonNegative, XKnownNegative;
- ComputeSignBit(X, XKnownNonNegative, XKnownNegative, TD, Depth, Q);
+ ComputeSignBit(X, XKnownNonNegative, XKnownNegative, DL, Depth, Q);
if (XKnownNegative)
return true;
}
// div exact can only produce a zero if the dividend is zero.
else if (match(V, m_Exact(m_IDiv(m_Value(X), m_Value())))) {
- return isKnownNonZero(X, TD, Depth, Q);
+ return isKnownNonZero(X, DL, Depth, Q);
}
// X + Y.
else if (match(V, m_Add(m_Value(X), m_Value(Y)))) {
bool XKnownNonNegative, XKnownNegative;
bool YKnownNonNegative, YKnownNegative;
- ComputeSignBit(X, XKnownNonNegative, XKnownNegative, TD, Depth, Q);
- ComputeSignBit(Y, YKnownNonNegative, YKnownNegative, TD, Depth, Q);
+ ComputeSignBit(X, XKnownNonNegative, XKnownNegative, DL, Depth, Q);
+ ComputeSignBit(Y, YKnownNonNegative, YKnownNegative, DL, Depth, Q);
// If X and Y are both non-negative (as signed values) then their sum is not
// zero unless both X and Y are zero.
if (XKnownNonNegative && YKnownNonNegative)
- if (isKnownNonZero(X, TD, Depth, Q) ||
- isKnownNonZero(Y, TD, Depth, Q))
+ if (isKnownNonZero(X, DL, Depth, Q) || isKnownNonZero(Y, DL, Depth, Q))
return true;
// If X and Y are both negative (as signed values) then their sum is not
APInt Mask = APInt::getSignedMaxValue(BitWidth);
// The sign bit of X is set. If some other bit is set then X is not equal
// to INT_MIN.
- computeKnownBits(X, KnownZero, KnownOne, TD, Depth, Q);
+ computeKnownBits(X, KnownZero, KnownOne, DL, Depth, Q);
if ((KnownOne & Mask) != 0)
return true;
// The sign bit of Y is set. If some other bit is set then Y is not equal
// to INT_MIN.
- computeKnownBits(Y, KnownZero, KnownOne, TD, Depth, Q);
+ computeKnownBits(Y, KnownZero, KnownOne, DL, Depth, Q);
if ((KnownOne & Mask) != 0)
return true;
}
// The sum of a non-negative number and a power of two is not zero.
if (XKnownNonNegative &&
- isKnownToBeAPowerOfTwo(Y, /*OrZero*/false, Depth, Q))
+ isKnownToBeAPowerOfTwo(Y, /*OrZero*/ false, Depth, Q, DL))
return true;
if (YKnownNonNegative &&
- isKnownToBeAPowerOfTwo(X, /*OrZero*/false, Depth, Q))
+ isKnownToBeAPowerOfTwo(X, /*OrZero*/ false, Depth, Q, DL))
return true;
}
// X * Y.
// If X and Y are non-zero then so is X * Y as long as the multiplication
// does not overflow.
if ((BO->hasNoSignedWrap() || BO->hasNoUnsignedWrap()) &&
- isKnownNonZero(X, TD, Depth, Q) &&
- isKnownNonZero(Y, TD, Depth, Q))
+ isKnownNonZero(X, DL, Depth, Q) && isKnownNonZero(Y, DL, Depth, Q))
return true;
}
// (C ? X : Y) != 0 if X != 0 and Y != 0.
else if (SelectInst *SI = dyn_cast<SelectInst>(V)) {
- if (isKnownNonZero(SI->getTrueValue(), TD, Depth, Q) &&
- isKnownNonZero(SI->getFalseValue(), TD, Depth, Q))
+ if (isKnownNonZero(SI->getTrueValue(), DL, Depth, Q) &&
+ isKnownNonZero(SI->getFalseValue(), DL, Depth, Q))
return true;
}
if (!BitWidth) return false;
APInt KnownZero(BitWidth, 0);
APInt KnownOne(BitWidth, 0);
- computeKnownBits(V, KnownZero, KnownOne, TD, Depth, Q);
+ computeKnownBits(V, KnownZero, KnownOne, DL, Depth, Q);
return KnownOne != 0;
}
/// cannot have.
///
/// This function is defined on values with integer type, values with pointer
-/// type (but only if TD is non-null), and vectors of integers. In the case
+/// type, and vectors of integers. In the case
/// where V is a vector, the mask, known zero, and known one values are the
/// same width as the vector element, and the bit is set only if it is true
/// for all of the elements in the vector.
-bool MaskedValueIsZero(Value *V, const APInt &Mask,
- const DataLayout *TD, unsigned Depth,
- const Query &Q) {
+bool MaskedValueIsZero(Value *V, const APInt &Mask, const DataLayout &DL,
+ unsigned Depth, const Query &Q) {
APInt KnownZero(Mask.getBitWidth(), 0), KnownOne(Mask.getBitWidth(), 0);
- computeKnownBits(V, KnownZero, KnownOne, TD, Depth, Q);
+ computeKnownBits(V, KnownZero, KnownOne, DL, Depth, Q);
return (KnownZero & Mask) == Mask;
}
///
/// 'Op' must have a scalar integer type.
///
-unsigned ComputeNumSignBits(Value *V, const DataLayout *TD,
- unsigned Depth, const Query &Q) {
- assert((TD || V->getType()->isIntOrIntVectorTy()) &&
- "ComputeNumSignBits requires a DataLayout object to operate "
- "on non-integer values!");
- Type *Ty = V->getType();
- unsigned TyBits = TD ? TD->getTypeSizeInBits(V->getType()->getScalarType()) :
- Ty->getScalarSizeInBits();
+unsigned ComputeNumSignBits(Value *V, const DataLayout &DL, unsigned Depth,
+ const Query &Q) {
+ unsigned TyBits = DL.getTypeSizeInBits(V->getType()->getScalarType());
unsigned Tmp, Tmp2;
unsigned FirstAnswer = 1;
default: break;
case Instruction::SExt:
Tmp = TyBits - U->getOperand(0)->getType()->getScalarSizeInBits();
- return ComputeNumSignBits(U->getOperand(0), TD, Depth+1, Q) + Tmp;
+ return ComputeNumSignBits(U->getOperand(0), DL, Depth + 1, Q) + Tmp;
case Instruction::SDiv: {
const APInt *Denominator;
break;
// Calculate the incoming numerator bits.
- unsigned NumBits = ComputeNumSignBits(U->getOperand(0), TD, Depth+1, Q);
+ unsigned NumBits = ComputeNumSignBits(U->getOperand(0), DL, Depth + 1, Q);
// Add floor(log(C)) bits to the numerator bits.
return std::min(TyBits, NumBits + Denominator->logBase2());
// Calculate the incoming numerator bits. SRem by a positive constant
// can't lower the number of sign bits.
- unsigned NumrBits = ComputeNumSignBits(U->getOperand(0), TD, Depth+1, Q);
+ unsigned NumrBits =
+ ComputeNumSignBits(U->getOperand(0), DL, Depth + 1, Q);
// Calculate the leading sign bit constraints by examining the
// denominator. The remainder is in the range 0..C-1, which is
}
case Instruction::AShr: {
- Tmp = ComputeNumSignBits(U->getOperand(0), TD, Depth+1, Q);
+ Tmp = ComputeNumSignBits(U->getOperand(0), DL, Depth + 1, Q);
// ashr X, C -> adds C sign bits. Vectors too.
const APInt *ShAmt;
if (match(U->getOperand(1), m_APInt(ShAmt))) {
const APInt *ShAmt;
if (match(U->getOperand(1), m_APInt(ShAmt))) {
// shl destroys sign bits.
- Tmp = ComputeNumSignBits(U->getOperand(0), TD, Depth+1, Q);
+ Tmp = ComputeNumSignBits(U->getOperand(0), DL, Depth + 1, Q);
Tmp2 = ShAmt->getZExtValue();
if (Tmp2 >= TyBits || // Bad shift.
Tmp2 >= Tmp) break; // Shifted all sign bits out.
case Instruction::Or:
case Instruction::Xor: // NOT is handled here.
// Logical binary ops preserve the number of sign bits at the worst.
- Tmp = ComputeNumSignBits(U->getOperand(0), TD, Depth+1, Q);
+ Tmp = ComputeNumSignBits(U->getOperand(0), DL, Depth + 1, Q);
if (Tmp != 1) {
- Tmp2 = ComputeNumSignBits(U->getOperand(1), TD, Depth+1, Q);
+ Tmp2 = ComputeNumSignBits(U->getOperand(1), DL, Depth + 1, Q);
FirstAnswer = std::min(Tmp, Tmp2);
// We computed what we know about the sign bits as our first
// answer. Now proceed to the generic code that uses
break;
case Instruction::Select:
- Tmp = ComputeNumSignBits(U->getOperand(1), TD, Depth+1, Q);
+ Tmp = ComputeNumSignBits(U->getOperand(1), DL, Depth + 1, Q);
if (Tmp == 1) return 1; // Early out.
- Tmp2 = ComputeNumSignBits(U->getOperand(2), TD, Depth+1, Q);
+ Tmp2 = ComputeNumSignBits(U->getOperand(2), DL, Depth + 1, Q);
return std::min(Tmp, Tmp2);
case Instruction::Add:
// Add can have at most one carry bit. Thus we know that the output
// is, at worst, one more bit than the inputs.
- Tmp = ComputeNumSignBits(U->getOperand(0), TD, Depth+1, Q);
+ Tmp = ComputeNumSignBits(U->getOperand(0), DL, Depth + 1, Q);
if (Tmp == 1) return 1; // Early out.
// Special case decrementing a value (ADD X, -1):
if (const auto *CRHS = dyn_cast<Constant>(U->getOperand(1)))
if (CRHS->isAllOnesValue()) {
APInt KnownZero(TyBits, 0), KnownOne(TyBits, 0);
- computeKnownBits(U->getOperand(0), KnownZero, KnownOne, TD, Depth+1, Q);
+ computeKnownBits(U->getOperand(0), KnownZero, KnownOne, DL, Depth + 1,
+ Q);
// If the input is known to be 0 or 1, the output is 0/-1, which is all
// sign bits set.
return Tmp;
}
- Tmp2 = ComputeNumSignBits(U->getOperand(1), TD, Depth+1, Q);
+ Tmp2 = ComputeNumSignBits(U->getOperand(1), DL, Depth + 1, Q);
if (Tmp2 == 1) return 1;
return std::min(Tmp, Tmp2)-1;
case Instruction::Sub:
- Tmp2 = ComputeNumSignBits(U->getOperand(1), TD, Depth+1, Q);
+ Tmp2 = ComputeNumSignBits(U->getOperand(1), DL, Depth + 1, Q);
if (Tmp2 == 1) return 1;
// Handle NEG.
if (const auto *CLHS = dyn_cast<Constant>(U->getOperand(0)))
if (CLHS->isNullValue()) {
APInt KnownZero(TyBits, 0), KnownOne(TyBits, 0);
- computeKnownBits(U->getOperand(1), KnownZero, KnownOne, TD, Depth+1, Q);
+ computeKnownBits(U->getOperand(1), KnownZero, KnownOne, DL, Depth + 1,
+ Q);
// If the input is known to be 0 or 1, the output is 0/-1, which is all
// sign bits set.
if ((KnownZero | APInt(TyBits, 1)).isAllOnesValue())
// Sub can have at most one carry bit. Thus we know that the output
// is, at worst, one more bit than the inputs.
- Tmp = ComputeNumSignBits(U->getOperand(0), TD, Depth+1, Q);
+ Tmp = ComputeNumSignBits(U->getOperand(0), DL, Depth + 1, Q);
if (Tmp == 1) return 1; // Early out.
return std::min(Tmp, Tmp2)-1;
// Take the minimum of all incoming values. This can't infinitely loop
// because of our depth threshold.
- Tmp = ComputeNumSignBits(PN->getIncomingValue(0), TD, Depth+1, Q);
+ Tmp = ComputeNumSignBits(PN->getIncomingValue(0), DL, Depth + 1, Q);
for (unsigned i = 1, e = NumIncomingValues; i != e; ++i) {
if (Tmp == 1) return Tmp;
- Tmp = std::min(Tmp,
- ComputeNumSignBits(PN->getIncomingValue(i), TD,
- Depth+1, Q));
+ Tmp = std::min(
+ Tmp, ComputeNumSignBits(PN->getIncomingValue(i), DL, Depth + 1, Q));
}
return Tmp;
}
// use this information.
APInt KnownZero(TyBits, 0), KnownOne(TyBits, 0);
APInt Mask;
- computeKnownBits(V, KnownZero, KnownOne, TD, Depth, Q);
+ computeKnownBits(V, KnownZero, KnownOne, DL, Depth, Q);
if (KnownZero.isNegative()) { // sign bit is 0
Mask = KnownZero;
/// Analyze the specified pointer to see if it can be expressed as a base
/// pointer plus a constant offset. Return the base and offset to the caller.
Value *llvm::GetPointerBaseWithConstantOffset(Value *Ptr, int64_t &Offset,
- const DataLayout *DL) {
- // Without DataLayout, conservatively assume 64-bit offsets, which is
- // the widest we support.
- unsigned BitWidth = DL ? DL->getPointerTypeSizeInBits(Ptr->getType()) : 64;
+ const DataLayout &DL) {
+ unsigned BitWidth = DL.getPointerTypeSizeInBits(Ptr->getType());
APInt ByteOffset(BitWidth, 0);
while (1) {
if (Ptr->getType()->isVectorTy())
break;
if (GEPOperator *GEP = dyn_cast<GEPOperator>(Ptr)) {
- if (DL) {
- APInt GEPOffset(BitWidth, 0);
- if (!GEP->accumulateConstantOffset(*DL, GEPOffset))
- break;
+ APInt GEPOffset(BitWidth, 0);
+ if (!GEP->accumulateConstantOffset(DL, GEPOffset))
+ break;
- ByteOffset += GEPOffset;
- }
+ ByteOffset += GEPOffset;
Ptr = GEP->getPointerOperand();
} else if (Operator::getOpcode(Ptr) == Instruction::BitCast ||
return Len == ~0ULL ? 1 : Len;
}
-Value *
-llvm::GetUnderlyingObject(Value *V, const DataLayout *TD, unsigned MaxLookup) {
+Value *llvm::GetUnderlyingObject(Value *V, const DataLayout &DL,
+ unsigned MaxLookup) {
if (!V->getType()->isPointerTy())
return V;
for (unsigned Count = 0; MaxLookup == 0 || Count < MaxLookup; ++Count) {
// See if InstructionSimplify knows any relevant tricks.
if (Instruction *I = dyn_cast<Instruction>(V))
// TODO: Acquire a DominatorTree and AssumptionCache and use them.
- if (Value *Simplified = SimplifyInstruction(I, TD, nullptr)) {
+ if (Value *Simplified = SimplifyInstruction(I, DL, nullptr)) {
V = Simplified;
continue;
}
return V;
}
-void
-llvm::GetUnderlyingObjects(Value *V,
- SmallVectorImpl<Value *> &Objects,
- const DataLayout *TD,
- unsigned MaxLookup) {
+void llvm::GetUnderlyingObjects(Value *V, SmallVectorImpl<Value *> &Objects,
+ const DataLayout &DL, unsigned MaxLookup) {
SmallPtrSet<Value *, 4> Visited;
SmallVector<Value *, 4> Worklist;
Worklist.push_back(V);
do {
Value *P = Worklist.pop_back_val();
- P = GetUnderlyingObject(P, TD, MaxLookup);
+ P = GetUnderlyingObject(P, DL, MaxLookup);
if (!Visited.insert(P).second)
continue;
return true;
}
-bool llvm::isSafeToSpeculativelyExecute(const Value *V,
- const DataLayout *TD) {
+bool llvm::isSafeToSpeculativelyExecute(const Value *V) {
const Operator *Inst = dyn_cast<Operator>(V);
if (!Inst)
return false;
// Speculative load may create a race that did not exist in the source.
LI->getParent()->getParent()->hasFnAttribute(Attribute::SanitizeThread))
return false;
- return LI->getPointerOperand()->isDereferenceablePointer(TD);
+ const DataLayout &DL = LI->getModule()->getDataLayout();
+ return LI->getPointerOperand()->isDereferenceablePointer(DL);
}
case Instruction::Call: {
if (const IntrinsicInst *II = dyn_cast<IntrinsicInst>(Inst)) {
}
OverflowResult llvm::computeOverflowForUnsignedMul(Value *LHS, Value *RHS,
- const DataLayout *DL,
+ const DataLayout &DL,
AssumptionCache *AC,
const Instruction *CxtI,
const DominatorTree *DT) {
}
OverflowResult llvm::computeOverflowForUnsignedAdd(Value *LHS, Value *RHS,
- const DataLayout *DL,
+ const DataLayout &DL,
AssumptionCache *AC,
const Instruction *CxtI,
const DominatorTree *DT) {
/// getGVAlignmentLog2 - Return the alignment to use for the specified global
/// value in log2 form. This rounds up to the preferred alignment if possible
/// and legal.
-static unsigned getGVAlignmentLog2(const GlobalValue *GV, const DataLayout &TD,
+static unsigned getGVAlignmentLog2(const GlobalValue *GV, const DataLayout &DL,
unsigned InBits = 0) {
unsigned NumBits = 0;
if (const GlobalVariable *GVar = dyn_cast<GlobalVariable>(GV))
- NumBits = TD.getPreferredAlignmentLog(GVar);
+ NumBits = DL.getPreferredAlignmentLog(GVar);
// If InBits is specified, round it to it.
if (InBits > NumBits)
// If the code isn't optimized, there may be outstanding folding
// opportunities. Attempt to fold the expression using DataLayout as a
// last resort before giving up.
- if (Constant *C = ConstantFoldConstantExpression(
- CE, TM.getDataLayout()))
+ if (Constant *C = ConstantFoldConstantExpression(CE, *TM.getDataLayout()))
if (C != CE)
return lowerConstant(C);
// If the constant expression's size is greater than 64-bits, then we have
// to emit the value in chunks. Try to constant fold the value and emit it
// that way.
- Constant *New = ConstantFoldConstantExpression(CE, DL);
+ Constant *New = ConstantFoldConstantExpression(CE, *DL);
if (New && New != CE)
return emitGlobalConstantImpl(New, AP);
}
WeakVH IterHandle(CurInstIterator);
replaceAndRecursivelySimplify(CI, RetVal,
- TLI ? TLI->getDataLayout() : nullptr,
TLInfo, ModifiedDT ? nullptr : DT);
// If the iterator instruction was recursively deleted, start over at the
// From here on out we're working with named functions.
if (!CI->getCalledFunction()) return false;
- // We'll need DataLayout from here on out.
- const DataLayout *TD = TLI ? TLI->getDataLayout() : nullptr;
- if (!TD) return false;
-
// Lower all default uses of _chk calls. This is very similar
// to what InstCombineCalls does, but here we are only lowering calls
// to fortified library functions (e.g. __memcpy_chk) that have the default
// "don't know" as the objectsize. Anything else should be left alone.
- FortifiedLibCallSimplifier Simplifier(TD, TLInfo, true);
+ FortifiedLibCallSimplifier Simplifier(TLInfo, true);
if (Value *V = Simplifier.optimizeCall(CI)) {
CI->replaceAllUsesWith(V);
CI->eraseFromParent();
// It is possible for very late stage optimizations (such as SimplifyCFG)
// to introduce PHI nodes too late to be cleaned up. If we detect such a
// trivial PHI, go ahead and zap it here.
- if (Value *V = SimplifyInstruction(P, TLI ? TLI->getDataLayout() : nullptr,
- TLInfo, DT)) {
+ const DataLayout &DL = I->getModule()->getDataLayout();
+ if (Value *V = SimplifyInstruction(P, DL, TLInfo, DT)) {
P->replaceAllUsesWith(V);
P->eraseFromParent();
++NumPHIsElim;
BasicBlock *BB = RI->getParent();
new UnreachableInst(Ctx, RI);
RI->eraseFromParent();
- SimplifyCFG(BB, TTI, 1, TLI->getDataLayout());
+ SimplifyCFG(BB, TTI, 1);
}
}
Resumes.resize(ResumesLeft);
// DataLayout.
if (isa<PointerType>(A->getType()))
A = ConstantFoldInstOperands(Instruction::PtrToInt, IntTy,
- const_cast<Constant*>(A), TD);
+ const_cast<Constant *>(A), *TD);
else if (A->getType() != IntTy)
A = ConstantFoldInstOperands(Instruction::BitCast, IntTy,
- const_cast<Constant*>(A), TD);
+ const_cast<Constant *>(A), *TD);
if (isa<PointerType>(B->getType()))
B = ConstantFoldInstOperands(Instruction::PtrToInt, IntTy,
- const_cast<Constant*>(B), TD);
+ const_cast<Constant *>(B), *TD);
else if (B->getType() != IntTy)
B = ConstantFoldInstOperands(Instruction::BitCast, IntTy,
- const_cast<Constant*>(B), TD);
+ const_cast<Constant *>(B), *TD);
return A == B;
}
/// getUnderlyingObjects - This is a wrapper around GetUnderlyingObjects
/// and adds support for basic ptrtoint+arithmetic+inttoptr sequences.
static void getUnderlyingObjects(const Value *V,
- SmallVectorImpl<Value *> &Objects) {
+ SmallVectorImpl<Value *> &Objects,
+ const DataLayout &DL) {
SmallPtrSet<const Value *, 16> Visited;
SmallVector<const Value *, 4> Working(1, V);
do {
V = Working.pop_back_val();
SmallVector<Value *, 4> Objs;
- GetUnderlyingObjects(const_cast<Value *>(V), Objs);
+ GetUnderlyingObjects(const_cast<Value *>(V), Objs, DL);
for (SmallVectorImpl<Value *>::iterator I = Objs.begin(), IE = Objs.end();
I != IE; ++I) {
/// object, return the Value for that object.
static void getUnderlyingObjectsForInstr(const MachineInstr *MI,
const MachineFrameInfo *MFI,
- UnderlyingObjectsVector &Objects) {
+ UnderlyingObjectsVector &Objects,
+ const DataLayout &DL) {
if (!MI->hasOneMemOperand() ||
(!(*MI->memoperands_begin())->getValue() &&
!(*MI->memoperands_begin())->getPseudoValue()) ||
return;
SmallVector<Value *, 4> Objs;
- getUnderlyingObjects(V, Objs);
+ getUnderlyingObjects(V, Objs, DL);
for (SmallVectorImpl<Value *>::iterator I = Objs.begin(), IE = Objs.end();
I != IE; ++I) {
// This MI might have either incomplete info, or known to be unsafe
// to deal with (i.e. volatile object).
static inline bool isUnsafeMemoryObject(MachineInstr *MI,
- const MachineFrameInfo *MFI) {
+ const MachineFrameInfo *MFI,
+ const DataLayout &DL) {
if (!MI || MI->memoperands_empty())
return true;
// We purposefully do no check for hasOneMemOperand() here
return true;
SmallVector<Value *, 4> Objs;
- getUnderlyingObjects(V, Objs);
+ getUnderlyingObjects(V, Objs, DL);
for (SmallVectorImpl<Value *>::iterator I = Objs.begin(),
IE = Objs.end(); I != IE; ++I) {
// Does this pointer refer to a distinct and identifiable object?
/// these two MIs be reordered during scheduling from memory dependency
/// point of view.
static bool MIsNeedChainEdge(AliasAnalysis *AA, const MachineFrameInfo *MFI,
- MachineInstr *MIa,
+ const DataLayout &DL, MachineInstr *MIa,
MachineInstr *MIb) {
const MachineFunction *MF = MIa->getParent()->getParent();
const TargetInstrInfo *TII = MF->getSubtarget().getInstrInfo();
if (!MIa->hasOneMemOperand() || !MIb->hasOneMemOperand())
return true;
- if (isUnsafeMemoryObject(MIa, MFI) || isUnsafeMemoryObject(MIb, MFI))
+ if (isUnsafeMemoryObject(MIa, MFI, DL) || isUnsafeMemoryObject(MIb, MFI, DL))
return true;
// If we are dealing with two "normal" loads, we do not need an edge
/// This recursive function iterates over chain deps of SUb looking for
/// "latest" node that needs a chain edge to SUa.
-static unsigned
-iterateChainSucc(AliasAnalysis *AA, const MachineFrameInfo *MFI,
- SUnit *SUa, SUnit *SUb, SUnit *ExitSU, unsigned *Depth,
- SmallPtrSetImpl<const SUnit*> &Visited) {
+static unsigned iterateChainSucc(AliasAnalysis *AA, const MachineFrameInfo *MFI,
+ const DataLayout &DL, SUnit *SUa, SUnit *SUb,
+ SUnit *ExitSU, unsigned *Depth,
+ SmallPtrSetImpl<const SUnit *> &Visited) {
if (!SUa || !SUb || SUb == ExitSU)
return *Depth;
// add that edge to the predecessors chain of SUb,
// and stop descending.
if (*Depth > 200 ||
- MIsNeedChainEdge(AA, MFI, SUa->getInstr(), SUb->getInstr())) {
+ MIsNeedChainEdge(AA, MFI, DL, SUa->getInstr(), SUb->getInstr())) {
SUb->addPred(SDep(SUa, SDep::MayAliasMem));
return *Depth;
}
for (SUnit::const_succ_iterator I = SUb->Succs.begin(), E = SUb->Succs.end();
I != E; ++I)
if (I->isNormalMemoryOrBarrier())
- iterateChainSucc (AA, MFI, SUa, I->getSUnit(), ExitSU, Depth, Visited);
+ iterateChainSucc(AA, MFI, DL, SUa, I->getSUnit(), ExitSU, Depth, Visited);
return *Depth;
}
/// checks whether SU can be aliasing any node dominated
/// by it.
static void adjustChainDeps(AliasAnalysis *AA, const MachineFrameInfo *MFI,
- SUnit *SU, SUnit *ExitSU, std::set<SUnit *> &CheckList,
+ const DataLayout &DL, SUnit *SU, SUnit *ExitSU,
+ std::set<SUnit *> &CheckList,
unsigned LatencyToLoad) {
if (!SU)
return;
I != IE; ++I) {
if (SU == *I)
continue;
- if (MIsNeedChainEdge(AA, MFI, SU->getInstr(), (*I)->getInstr())) {
+ if (MIsNeedChainEdge(AA, MFI, DL, SU->getInstr(), (*I)->getInstr())) {
SDep Dep(SU, SDep::MayAliasMem);
Dep.setLatency(((*I)->getInstr()->mayLoad()) ? LatencyToLoad : 0);
(*I)->addPred(Dep);
for (SUnit::const_succ_iterator J = (*I)->Succs.begin(),
JE = (*I)->Succs.end(); J != JE; ++J)
if (J->isNormalMemoryOrBarrier())
- iterateChainSucc (AA, MFI, SU, J->getSUnit(),
- ExitSU, &Depth, Visited);
+ iterateChainSucc(AA, MFI, DL, SU, J->getSUnit(), ExitSU, &Depth,
+ Visited);
}
}
/// Check whether two objects need a chain edge, if so, add it
/// otherwise remember the rejected SU.
-static inline
-void addChainDependency (AliasAnalysis *AA, const MachineFrameInfo *MFI,
- SUnit *SUa, SUnit *SUb,
- std::set<SUnit *> &RejectList,
- unsigned TrueMemOrderLatency = 0,
- bool isNormalMemory = false) {
+static inline void addChainDependency(AliasAnalysis *AA,
+ const MachineFrameInfo *MFI,
+ const DataLayout &DL, SUnit *SUa,
+ SUnit *SUb, std::set<SUnit *> &RejectList,
+ unsigned TrueMemOrderLatency = 0,
+ bool isNormalMemory = false) {
// If this is a false dependency,
// do not add the edge, but rememeber the rejected node.
- if (MIsNeedChainEdge(AA, MFI, SUa->getInstr(), SUb->getInstr())) {
+ if (MIsNeedChainEdge(AA, MFI, DL, SUa->getInstr(), SUb->getInstr())) {
SDep Dep(SUa, isNormalMemory ? SDep::MayAliasMem : SDep::Barrier);
Dep.setLatency(TrueMemOrderLatency);
SUb->addPred(Dep);
BarrierChain = SU;
// This is a barrier event that acts as a pivotal node in the DAG,
// so it is safe to clear list of exposed nodes.
- adjustChainDeps(AA, MFI, SU, &ExitSU, RejectMemNodes,
+ adjustChainDeps(AA, MFI, *TM.getDataLayout(), SU, &ExitSU, RejectMemNodes,
TrueMemOrderLatency);
RejectMemNodes.clear();
NonAliasMemDefs.clear();
unsigned ChainLatency = 0;
if (AliasChain->getInstr()->mayLoad())
ChainLatency = TrueMemOrderLatency;
- addChainDependency(AAForDep, MFI, SU, AliasChain, RejectMemNodes,
- ChainLatency);
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, AliasChain,
+ RejectMemNodes, ChainLatency);
}
AliasChain = SU;
for (unsigned k = 0, m = PendingLoads.size(); k != m; ++k)
- addChainDependency(AAForDep, MFI, SU, PendingLoads[k], RejectMemNodes,
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU,
+ PendingLoads[k], RejectMemNodes,
TrueMemOrderLatency);
for (MapVector<ValueType, std::vector<SUnit *> >::iterator I =
AliasMemDefs.begin(), E = AliasMemDefs.end(); I != E; ++I) {
for (unsigned i = 0, e = I->second.size(); i != e; ++i)
- addChainDependency(AAForDep, MFI, SU, I->second[i], RejectMemNodes);
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU,
+ I->second[i], RejectMemNodes);
}
for (MapVector<ValueType, std::vector<SUnit *> >::iterator I =
AliasMemUses.begin(), E = AliasMemUses.end(); I != E; ++I) {
for (unsigned i = 0, e = I->second.size(); i != e; ++i)
- addChainDependency(AAForDep, MFI, SU, I->second[i], RejectMemNodes,
- TrueMemOrderLatency);
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU,
+ I->second[i], RejectMemNodes, TrueMemOrderLatency);
}
- adjustChainDeps(AA, MFI, SU, &ExitSU, RejectMemNodes,
+ adjustChainDeps(AA, MFI, *TM.getDataLayout(), SU, &ExitSU, RejectMemNodes,
TrueMemOrderLatency);
PendingLoads.clear();
AliasMemDefs.clear();
BarrierChain->addPred(SDep(SU, SDep::Barrier));
UnderlyingObjectsVector Objs;
- getUnderlyingObjectsForInstr(MI, MFI, Objs);
+ getUnderlyingObjectsForInstr(MI, MFI, Objs, *TM.getDataLayout());
if (Objs.empty()) {
// Treat all other stores conservatively.
((ThisMayAlias) ? AliasMemDefs.end() : NonAliasMemDefs.end());
if (I != IE) {
for (unsigned i = 0, e = I->second.size(); i != e; ++i)
- addChainDependency(AAForDep, MFI, SU, I->second[i], RejectMemNodes,
- 0, true);
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU,
+ I->second[i], RejectMemNodes, 0, true);
// If we're not using AA, then we only need one store per object.
if (!AAForDep)
((ThisMayAlias) ? AliasMemUses.end() : NonAliasMemUses.end());
if (J != JE) {
for (unsigned i = 0, e = J->second.size(); i != e; ++i)
- addChainDependency(AAForDep, MFI, SU, J->second[i], RejectMemNodes,
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU,
+ J->second[i], RejectMemNodes,
TrueMemOrderLatency, true);
J->second.clear();
}
// Add dependencies from all the PendingLoads, i.e. loads
// with no underlying object.
for (unsigned k = 0, m = PendingLoads.size(); k != m; ++k)
- addChainDependency(AAForDep, MFI, SU, PendingLoads[k], RejectMemNodes,
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU,
+ PendingLoads[k], RejectMemNodes,
TrueMemOrderLatency);
// Add dependence on alias chain, if needed.
if (AliasChain)
- addChainDependency(AAForDep, MFI, SU, AliasChain, RejectMemNodes);
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, AliasChain,
+ RejectMemNodes);
}
- adjustChainDeps(AA, MFI, SU, &ExitSU, RejectMemNodes,
+ adjustChainDeps(AA, MFI, *TM.getDataLayout(), SU, &ExitSU, RejectMemNodes,
TrueMemOrderLatency);
} else if (MI->mayLoad()) {
bool MayAlias = true;
// Invariant load, no chain dependencies needed!
} else {
UnderlyingObjectsVector Objs;
- getUnderlyingObjectsForInstr(MI, MFI, Objs);
+ getUnderlyingObjectsForInstr(MI, MFI, Objs, *TM.getDataLayout());
if (Objs.empty()) {
// A load with no underlying object. Depend on all
for (MapVector<ValueType, std::vector<SUnit *> >::iterator I =
AliasMemDefs.begin(), E = AliasMemDefs.end(); I != E; ++I)
for (unsigned i = 0, e = I->second.size(); i != e; ++i)
- addChainDependency(AAForDep, MFI, SU, I->second[i],
- RejectMemNodes);
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU,
+ I->second[i], RejectMemNodes);
PendingLoads.push_back(SU);
MayAlias = true;
((ThisMayAlias) ? AliasMemDefs.end() : NonAliasMemDefs.end());
if (I != IE)
for (unsigned i = 0, e = I->second.size(); i != e; ++i)
- addChainDependency(AAForDep, MFI, SU, I->second[i],
- RejectMemNodes, 0, true);
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU,
+ I->second[i], RejectMemNodes, 0, true);
if (ThisMayAlias)
AliasMemUses[V].push_back(SU);
else
NonAliasMemUses[V].push_back(SU);
}
if (MayAlias)
- adjustChainDeps(AA, MFI, SU, &ExitSU, RejectMemNodes, /*Latency=*/0);
+ adjustChainDeps(AA, MFI, *TM.getDataLayout(), SU, &ExitSU,
+ RejectMemNodes, /*Latency=*/0);
// Add dependencies on alias and barrier chains, if needed.
if (MayAlias && AliasChain)
- addChainDependency(AAForDep, MFI, SU, AliasChain, RejectMemNodes);
+ addChainDependency(AAForDep, MFI, *TM.getDataLayout(), SU, AliasChain,
+ RejectMemNodes);
if (BarrierChain)
BarrierChain->addPred(SDep(SU, SDep::Barrier));
}
if (TLI->isGAPlusOffset(Ptr.getNode(), GV, GVOffset)) {
unsigned PtrWidth = TLI->getPointerTypeSizeInBits(GV->getType());
APInt KnownZero(PtrWidth, 0), KnownOne(PtrWidth, 0);
- llvm::computeKnownBits(const_cast<GlobalValue*>(GV), KnownZero, KnownOne,
- TLI->getDataLayout());
+ llvm::computeKnownBits(const_cast<GlobalValue *>(GV), KnownZero, KnownOne,
+ *TLI->getDataLayout());
unsigned AlignBits = KnownZero.countTrailingOnes();
unsigned Align = AlignBits ? 1 << std::min(31U, AlignBits) : 0;
if (Align)
return nullptr;
SmallVector<Value *, 4> Allocas;
- GetUnderlyingObjects(I.getArgOperand(1), Allocas, DL);
+ GetUnderlyingObjects(I.getArgOperand(1), Allocas, *DL);
for (SmallVectorImpl<Value*>::iterator Object = Allocas.begin(),
E = Allocas.end(); Object != E; ++Object) {
LoadInput = ConstantExpr::getBitCast(const_cast<Constant *>(LoadInput),
PointerType::getUnqual(LoadTy));
- if (const Constant *LoadCst =
- ConstantFoldLoadFromConstPtr(const_cast<Constant *>(LoadInput),
- Builder.DL))
+ if (const Constant *LoadCst = ConstantFoldLoadFromConstPtr(
+ const_cast<Constant *>(LoadInput), *Builder.DL))
return Builder.getValue(LoadCst);
}
}
/// NOTE: The TargetMachine owns TLOF.
-TargetLoweringBase::TargetLoweringBase(const TargetMachine &tm)
- : TM(tm), DL(TM.getDataLayout()) {
+TargetLoweringBase::TargetLoweringBase(const TargetMachine &tm) : TM(tm) {
initActions();
// Perform these initializations only once.
- IsLittleEndian = DL->isLittleEndian();
+ IsLittleEndian = getDataLayout()->isLittleEndian();
MaxStoresPerMemset = MaxStoresPerMemcpy = MaxStoresPerMemmove = 8;
MaxStoresPerMemsetOptSize = MaxStoresPerMemcpyOptSize
= MaxStoresPerMemmoveOptSize = 4;
}
unsigned TargetLoweringBase::getPointerSizeInBits(uint32_t AS) const {
- return DL->getPointerSizeInBits(AS);
+ return getDataLayout()->getPointerSizeInBits(AS);
}
unsigned TargetLoweringBase::getPointerTypeSizeInBits(Type *Ty) const {
}
MVT TargetLoweringBase::getScalarShiftAmountTy(EVT LHSTy) const {
- return MVT::getIntegerVT(8*DL->getPointerSize(0));
+ return MVT::getIntegerVT(8 * getDataLayout()->getPointerSize(0));
}
EVT TargetLoweringBase::getShiftAmountTy(EVT LHSTy) const {
/// function arguments in the caller parameter area. This is the actual
/// alignment, not its logarithm.
unsigned TargetLoweringBase::getByValTypeAlignment(Type *Ty) const {
- return DL->getABITypeAlignment(Ty);
+ return getDataLayout()->getABITypeAlignment(Ty);
}
//===----------------------------------------------------------------------===//
BasicBlock::iterator II = LPad;
- CloneAndPruneIntoFromInst(
- Handler, SrcFn, ++II, VMap,
- /*ModuleLevelChanges=*/false, Returns, "", &InlinedFunctionInfo,
- &SrcFn->getParent()->getDataLayout(), Director.get());
+ CloneAndPruneIntoFromInst(Handler, SrcFn, ++II, VMap,
+ /*ModuleLevelChanges=*/false, Returns, "",
+ &InlinedFunctionInfo, Director.get());
// Move all the instructions in the first cloned block into our entry block.
BasicBlock *FirstClonedBB = std::next(Function::iterator(Entry));
return isNoopCast(getOpcode(), getOperand(0)->getType(), getType(), IntPtrTy);
}
-bool CastInst::isNoopCast(const DataLayout *DL) const {
- if (!DL) {
- // Assume maximum pointer size.
- return isNoopCast(Type::getInt64Ty(getContext()));
- }
-
+bool CastInst::isNoopCast(const DataLayout &DL) const {
Type *PtrOpTy = nullptr;
if (getOpcode() == Instruction::PtrToInt)
PtrOpTy = getOperand(0)->getType();
else if (getOpcode() == Instruction::IntToPtr)
PtrOpTy = getType();
- Type *IntPtrTy = PtrOpTy
- ? DL->getIntPtrType(PtrOpTy)
- : DL->getIntPtrType(getContext(), 0);
+ Type *IntPtrTy =
+ PtrOpTy ? DL.getIntPtrType(PtrOpTy) : DL.getIntPtrType(getContext(), 0);
return isNoopCast(getOpcode(), getOperand(0)->getType(), getType(), IntPtrTy);
}
}
bool CastInst::isBitOrNoopPointerCastable(Type *SrcTy, Type *DestTy,
- const DataLayout *DL) {
+ const DataLayout &DL) {
if (auto *PtrTy = dyn_cast<PointerType>(SrcTy))
if (auto *IntTy = dyn_cast<IntegerType>(DestTy))
- return DL && IntTy->getBitWidth() == DL->getPointerTypeSizeInBits(PtrTy);
+ return IntTy->getBitWidth() == DL.getPointerTypeSizeInBits(PtrTy);
if (auto *PtrTy = dyn_cast<PointerType>(DestTy))
if (auto *IntTy = dyn_cast<IntegerType>(SrcTy))
- return DL && IntTy->getBitWidth() == DL->getPointerTypeSizeInBits(PtrTy);
+ return IntTy->getBitWidth() == DL.getPointerTypeSizeInBits(PtrTy);
return isBitCastable(SrcTy, DestTy);
}
/// Microsoft fastcall and stdcall functions require a suffix on their name
/// indicating the number of words of arguments they take.
static void addByteCountSuffix(raw_ostream &OS, const Function *F,
- const DataLayout &TD) {
+ const DataLayout &DL) {
// Calculate arguments size total.
unsigned ArgWords = 0;
for (Function::const_arg_iterator AI = F->arg_begin(), AE = F->arg_end();
if (AI->hasByValOrInAllocaAttr())
Ty = cast<PointerType>(Ty)->getElementType();
// Size should be aligned to pointer size.
- unsigned PtrSize = TD.getPointerSize();
- ArgWords += RoundUpToAlignment(TD.getTypeAllocSize(Ty), PtrSize);
+ unsigned PtrSize = DL.getPointerSize();
+ ArgWords += RoundUpToAlignment(DL.getTypeAllocSize(Ty), PtrSize);
}
OS << '@' << ArgWords;
///
/// Test if V is always a pointer to allocated and suitably aligned memory for
/// a simple load or store.
-static bool isDereferenceablePointer(const Value *V, const DataLayout *DL,
+static bool isDereferenceablePointer(const Value *V, const DataLayout &DL,
SmallPtrSetImpl<const Value *> &Visited) {
// Note that it is not safe to speculate into a malloc'd region because
// malloc may return null.
// to a type of smaller size (or the same size), and the alignment
// is at least as large as for the resulting pointer type, then
// we can look through the bitcast.
- if (DL)
- if (const BitCastOperator *BC = dyn_cast<BitCastOperator>(V)) {
- Type *STy = BC->getSrcTy()->getPointerElementType(),
- *DTy = BC->getDestTy()->getPointerElementType();
- if (STy->isSized() && DTy->isSized() &&
- (DL->getTypeStoreSize(STy) >=
- DL->getTypeStoreSize(DTy)) &&
- (DL->getABITypeAlignment(STy) >=
- DL->getABITypeAlignment(DTy)))
- return isDereferenceablePointer(BC->getOperand(0), DL, Visited);
- }
+ if (const BitCastOperator *BC = dyn_cast<BitCastOperator>(V)) {
+ Type *STy = BC->getSrcTy()->getPointerElementType(),
+ *DTy = BC->getDestTy()->getPointerElementType();
+ if (STy->isSized() && DTy->isSized() &&
+ (DL.getTypeStoreSize(STy) >= DL.getTypeStoreSize(DTy)) &&
+ (DL.getABITypeAlignment(STy) >= DL.getABITypeAlignment(DTy)))
+ return isDereferenceablePointer(BC->getOperand(0), DL, Visited);
+ }
// Global variables which can't collapse to null are ok.
if (const GlobalVariable *GV = dyn_cast<GlobalVariable>(V))
return true;
else if (uint64_t Bytes = A->getDereferenceableBytes()) {
Type *Ty = V->getType()->getPointerElementType();
- if (Ty->isSized() && DL && DL->getTypeStoreSize(Ty) <= Bytes)
+ if (Ty->isSized() && DL.getTypeStoreSize(Ty) <= Bytes)
return true;
}
if (ImmutableCallSite CS = V) {
if (uint64_t Bytes = CS.getDereferenceableBytes(0)) {
Type *Ty = V->getType()->getPointerElementType();
- if (Ty->isSized() && DL && DL->getTypeStoreSize(Ty) <= Bytes)
+ if (Ty->isSized() && DL.getTypeStoreSize(Ty) <= Bytes)
return true;
}
}
return false;
}
-bool Value::isDereferenceablePointer(const DataLayout *DL) const {
+bool Value::isDereferenceablePointer(const DataLayout &DL) const {
// When dereferenceability information is provided by a dereferenceable
// attribute, we know exactly how many bytes are dereferenceable. If we can
// determine the exact offset to the attributed variable, we can use that
// information here.
Type *Ty = getType()->getPointerElementType();
- if (Ty->isSized() && DL) {
- APInt Offset(DL->getTypeStoreSizeInBits(getType()), 0);
- const Value *BV = stripAndAccumulateInBoundsConstantOffsets(*DL, Offset);
+ if (Ty->isSized()) {
+ APInt Offset(DL.getTypeStoreSizeInBits(getType()), 0);
+ const Value *BV = stripAndAccumulateInBoundsConstantOffsets(DL, Offset);
APInt DerefBytes(Offset.getBitWidth(), 0);
if (const Argument *A = dyn_cast<Argument>(BV))
DerefBytes = CS.getDereferenceableBytes(0);
if (DerefBytes.getBoolValue() && Offset.isNonNegative()) {
- if (DerefBytes.uge(Offset + DL->getTypeStoreSize(Ty)))
+ if (DerefBytes.uge(Offset + DL.getTypeStoreSize(Ty)))
return true;
}
}
return false;
}
- // FIXME: Make datalayout mandatory and just use getDataLayout().
- DataLayout DL(Dest.getParent());
-
+ const DataLayout &DL = Dest.getParent()->getDataLayout();
uint64_t DestSize = DL.getTypeAllocSize(Dest.getType()->getElementType());
uint64_t SrcSize = DL.getTypeAllocSize(Src.getType()->getElementType());
LinkFromSrc = SrcSize > DestSize;
/// memory instruction can be moved to a delay slot.
class MemDefsUses : public InspectMemInstr {
public:
- MemDefsUses(const MachineFrameInfo *MFI);
+ MemDefsUses(const DataLayout &DL, const MachineFrameInfo *MFI);
private:
typedef PointerUnion<const Value *, const PseudoSourceValue *> ValueType;
const MachineFrameInfo *MFI;
SmallPtrSet<ValueType, 4> Uses, Defs;
+ const DataLayout &DL;
/// Flags indicating whether loads or stores with no underlying objects have
/// been seen.
return true;
}
-MemDefsUses::MemDefsUses(const MachineFrameInfo *MFI_)
- : InspectMemInstr(false), MFI(MFI_), SeenNoObjLoad(false),
- SeenNoObjStore(false) {}
+MemDefsUses::MemDefsUses(const DataLayout &DL, const MachineFrameInfo *MFI_)
+ : InspectMemInstr(false), MFI(MFI_), DL(DL), SeenNoObjLoad(false),
+ SeenNoObjStore(false) {}
bool MemDefsUses::hasHazard_(const MachineInstr &MI) {
bool HasHazard = false;
const Value *V = (*MI.memoperands_begin())->getValue();
SmallVector<Value *, 4> Objs;
- GetUnderlyingObjects(const_cast<Value *>(V), Objs);
+ GetUnderlyingObjects(const_cast<Value *>(V), Objs, DL);
for (SmallVectorImpl<Value *>::iterator I = Objs.begin(), E = Objs.end();
I != E; ++I) {
return false;
RegDefsUses RegDU(*MBB.getParent()->getSubtarget().getRegisterInfo());
- MemDefsUses MemDU(MBB.getParent()->getFrameInfo());
+ MemDefsUses MemDU(*TM.getDataLayout(), MBB.getParent()->getFrameInfo());
ReverseIter Filler;
RegDU.init(*Slot);
IM.reset(new LoadFromStackOrConst());
} else {
const MachineFrameInfo *MFI = MBB.getParent()->getFrameInfo();
- IM.reset(new MemDefsUses(MFI));
+ IM.reset(new MemDefsUses(*TM.getDataLayout(), MFI));
}
if (!searchRange(MBB, SuccBB->begin(), SuccBB->end(), RegDU, *IM, Slot,
break;
} else if (const ConstantExpr *Cexpr = dyn_cast<ConstantExpr>(CPV)) {
if (const ConstantInt *constInt = dyn_cast<ConstantInt>(
- ConstantFoldConstantExpression(Cexpr, TD))) {
+ ConstantFoldConstantExpression(Cexpr, *TD))) {
int int32 = (int)(constInt->getZExtValue());
ptr = (unsigned char *)&int32;
aggBuffer->addBytes(ptr, 4, Bytes);
break;
} else if (const ConstantExpr *Cexpr = dyn_cast<ConstantExpr>(CPV)) {
if (const ConstantInt *constInt = dyn_cast<ConstantInt>(
- ConstantFoldConstantExpression(Cexpr, TD))) {
+ ConstantFoldConstantExpression(Cexpr, *TD))) {
long long int64 = (long long)(constInt->getZExtValue());
ptr = (unsigned char *)&int64;
aggBuffer->addBytes(ptr, 8, Bytes);
// selected branch.
MadeChange = true;
- SCEVExpander SCEVE(*SE, "loopcnt");
+ SCEVExpander SCEVE(*SE, Preheader->getModule()->getDataLayout(), "loopcnt");
LLVMContext &C = SE->getContext();
Type *CountType = TT.isArch64Bit() ? Type::getInt64Ty(C) :
Type::getInt32Ty(C);
PrefLoads.push_back(std::make_pair(MemI, LSCEVAddRec));
Type *I8Ptr = Type::getInt8PtrTy((*I)->getContext(), PtrAddrSpace);
- SCEVExpander SCEVE(*SE, "prefaddr");
+ SCEVExpander SCEVE(*SE, J->getModule()->getDataLayout(), "prefaddr");
Value *PrefPtrValue = SCEVE.expandCodeFor(NextLSCEV, I8Ptr, MemI);
IRBuilder<> Builder(MemI);
MemI->hasName() ? MemI->getName() + ".phi" : "",
Header->getFirstNonPHI());
- SCEVExpander SCEVE(*SE, "pistart");
+ SCEVExpander SCEVE(*SE, Header->getModule()->getDataLayout(), "pistart");
Value *BasePtrStart = SCEVE.expandCodeFor(BasePtrStartSCEV, I8PtrTy,
LoopPredecessor->getTerminator());
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/CodeGen/BasicTTIImpl.h"
+#include "llvm/IR/Module.h"
#include "llvm/Support/Debug.h"
#include "llvm/Target/CostTable.h"
#include "llvm/Target/TargetLowering.h"
// TODO: Do we want runtime unrolling?
for (const BasicBlock *BB : L->getBlocks()) {
+ const DataLayout &DL = BB->getModule()->getDataLayout();
for (const Instruction &I : *BB) {
const GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(&I);
if (!GEP || GEP->getAddressSpace() != AMDGPUAS::PRIVATE_ADDRESS)
continue;
const Value *Ptr = GEP->getPointerOperand();
- const AllocaInst *Alloca = dyn_cast<AllocaInst>(GetUnderlyingObject(Ptr));
+ const AllocaInst *Alloca =
+ dyn_cast<AllocaInst>(GetUnderlyingObject(Ptr, DL));
if (Alloca) {
// We want to do whatever we can to limit the number of alloca
// instructions that make it through to the code generator. allocas
// Lower loads constant address space global variable loads
if (LoadNode->getAddressSpace() == AMDGPUAS::CONSTANT_ADDRESS &&
- isa<GlobalVariable>(
- GetUnderlyingObject(LoadNode->getMemOperand()->getValue()))) {
+ isa<GlobalVariable>(GetUnderlyingObject(
+ LoadNode->getMemOperand()->getValue(), *getDataLayout()))) {
SDValue Ptr = DAG.getZExtOrTrunc(LoadNode->getBasePtr(), DL,
getPointerTy(AMDGPUAS::PRIVATE_ADDRESS));
bool runOnSCC(CallGraphSCC &SCC) override;
static char ID; // Pass identification, replacement for typeid
explicit ArgPromotion(unsigned maxElements = 3)
- : CallGraphSCCPass(ID), DL(nullptr), maxElements(maxElements) {
+ : CallGraphSCCPass(ID), maxElements(maxElements) {
initializeArgPromotionPass(*PassRegistry::getPassRegistry());
}
/// A vector used to hold the indices of a single GEP instruction
typedef std::vector<uint64_t> IndicesVector;
- const DataLayout *DL;
private:
- bool isDenselyPacked(Type *type);
+ bool isDenselyPacked(Type *type, const DataLayout &DL);
bool canPaddingBeAccessed(Argument *Arg);
CallGraphNode *PromoteArguments(CallGraphNode *CGN);
bool isSafeToPromoteArgument(Argument *Arg, bool isByVal) const;
}
/// \brief Checks if a type could have padding bytes.
-bool ArgPromotion::isDenselyPacked(Type *type) {
+bool ArgPromotion::isDenselyPacked(Type *type, const DataLayout &DL) {
// There is no size information, so be conservative.
if (!type->isSized())
// If the alloc size is not equal to the storage size, then there are padding
// bytes. For x86_fp80 on x86-64, size: 80 alloc size: 128.
- if (!DL || DL->getTypeSizeInBits(type) != DL->getTypeAllocSizeInBits(type))
+ if (DL.getTypeSizeInBits(type) != DL.getTypeAllocSizeInBits(type))
return false;
if (!isa<CompositeType>(type))
// For homogenous sequential types, check for padding within members.
if (SequentialType *seqTy = dyn_cast<SequentialType>(type))
- return isa<PointerType>(seqTy) || isDenselyPacked(seqTy->getElementType());
+ return isa<PointerType>(seqTy) ||
+ isDenselyPacked(seqTy->getElementType(), DL);
// Check for padding within and between elements of a struct.
StructType *StructTy = cast<StructType>(type);
- const StructLayout *Layout = DL->getStructLayout(StructTy);
+ const StructLayout *Layout = DL.getStructLayout(StructTy);
uint64_t StartPos = 0;
for (unsigned i = 0, E = StructTy->getNumElements(); i < E; ++i) {
Type *ElTy = StructTy->getElementType(i);
- if (!isDenselyPacked(ElTy))
+ if (!isDenselyPacked(ElTy, DL))
return false;
if (StartPos != Layout->getElementOffsetInBits(i))
return false;
- StartPos += DL->getTypeAllocSizeInBits(ElTy);
+ StartPos += DL.getTypeAllocSizeInBits(ElTy);
}
return true;
// Make sure that it is local to this module.
if (!F || !F->hasLocalLinkage()) return nullptr;
- DL = &F->getParent()->getDataLayout();
-
// First check: see if there are any pointer arguments! If not, quick exit.
SmallVector<Argument*, 16> PointerArgs;
for (Function::arg_iterator I = F->arg_begin(), E = F->arg_end(); I != E; ++I)
// IR, while in the callee the classification is determined dynamically based
// on the number of registers consumed so far.
if (F->isVarArg()) return nullptr;
+ const DataLayout &DL = F->getParent()->getDataLayout();
// Check to see which arguments are promotable. If an argument is promotable,
// add it to ArgsToPromote.
// packed or if we can prove the padding bytes are never accessed. This does
// not apply to inalloca.
bool isSafeToPromote =
- PtrArg->hasByValAttr() &&
- (isDenselyPacked(AgTy) || !canPaddingBeAccessed(PtrArg));
+ PtrArg->hasByValAttr() &&
+ (isDenselyPacked(AgTy, DL) || !canPaddingBeAccessed(PtrArg));
if (isSafeToPromote) {
if (StructType *STy = dyn_cast<StructType>(AgTy)) {
if (maxElements > 0 && STy->getNumElements() > maxElements) {
/// AllCallersPassInValidPointerForArgument - Return true if we can prove that
/// all callees pass in a valid pointer for the specified function argument.
-static bool AllCallersPassInValidPointerForArgument(Argument *Arg,
- const DataLayout *DL) {
+static bool AllCallersPassInValidPointerForArgument(Argument *Arg) {
Function *Callee = Arg->getParent();
+ const DataLayout &DL = Callee->getParent()->getDataLayout();
unsigned ArgNo = Arg->getArgNo();
GEPIndicesSet ToPromote;
// If the pointer is always valid, any load with first index 0 is valid.
- if (isByValOrInAlloca || AllCallersPassInValidPointerForArgument(Arg, DL))
+ if (isByValOrInAlloca || AllCallersPassInValidPointerForArgument(Arg))
SafeToUnconditionallyLoad.insert(IndicesVector(1, 0));
// First, iterate the entry block and mark loads of (geps of) arguments as
// alignment to a concrete value.
unsigned getAlignment(GlobalVariable *GV) const;
- const DataLayout *DL;
};
}
return A.hasUnnamedAddr();
}
-bool ConstantMerge::hasKnownAlignment(GlobalVariable *GV) const {
- return DL || GV->getAlignment() != 0;
-}
-
unsigned ConstantMerge::getAlignment(GlobalVariable *GV) const {
unsigned Align = GV->getAlignment();
if (Align)
return Align;
- if (DL)
- return DL->getPreferredAlignment(GV);
- return 0;
+ return GV->getParent()->getDataLayout().getPreferredAlignment(GV);
}
bool ConstantMerge::runOnModule(Module &M) {
- DL = &M.getDataLayout();
// Find all the globals that are marked "used". These cannot be merged.
SmallPtrSet<const GlobalValue*, 8> UsedGlobals;
FindUsedValues(M.getGlobalVariable("llvm.used"), UsedGlobals);
FindUsedValues(M.getGlobalVariable("llvm.compiler.used"), UsedGlobals);
-
- // Map unique <constants, has-unknown-alignment> pairs to globals. We don't
- // want to merge globals of unknown alignment with those of explicit
- // alignment. If we have DataLayout, we always know the alignment.
- DenseMap<PointerIntPair<Constant*, 1, bool>, GlobalVariable*> CMap;
+
+ // Map unique constants to globals.
+ DenseMap<Constant *, GlobalVariable *> CMap;
// Replacements - This vector contains a list of replacements to perform.
SmallVector<std::pair<GlobalVariable*, GlobalVariable*>, 32> Replacements;
Constant *Init = GV->getInitializer();
// Check to see if the initializer is already known.
- PointerIntPair<Constant*, 1, bool> Pair(Init, hasKnownAlignment(GV));
- GlobalVariable *&Slot = CMap[Pair];
+ GlobalVariable *&Slot = CMap[Init];
// If this is the first constant we find or if the old one is local,
// replace with the current one. If the current is externally visible
Constant *Init = GV->getInitializer();
// Check to see if the initializer is already known.
- PointerIntPair<Constant*, 1, bool> Pair(Init, hasKnownAlignment(GV));
- GlobalVariable *Slot = CMap[Pair];
+ GlobalVariable *Slot = CMap[Init];
if (!Slot || Slot == GV)
continue;
const GlobalStatus &GS);
bool OptimizeEmptyGlobalCXXDtors(Function *CXAAtExitFn);
- // const DataLayout *DL;
TargetLibraryInfo *TLI;
SmallSet<const Comdat *, 8> NotDiscardableComdats;
};
Constant *SubInit = nullptr;
if (!isa<ConstantExpr>(GEP->getOperand(0))) {
ConstantExpr *CE = dyn_cast_or_null<ConstantExpr>(
- ConstantFoldInstruction(GEP, &DL, TLI));
+ ConstantFoldInstruction(GEP, DL, TLI));
if (Init && CE && CE->getOpcode() == Instruction::GetElementPtr)
SubInit = ConstantFoldLoadThroughGEPConstantExpr(Init, CE);
TargetLibraryInfo *TLI) {
for (Value::user_iterator UI = V->user_begin(), E = V->user_end(); UI != E; )
if (Instruction *I = dyn_cast<Instruction>(*UI++))
- if (Constant *NewC = ConstantFoldInstruction(I, &DL, TLI)) {
+ if (Constant *NewC = ConstantFoldInstruction(I, DL, TLI)) {
I->replaceAllUsesWith(NewC);
// Advance UI to the next non-I use to avoid invalidating it!
// This eliminates dynamic allocation, avoids an indirection accessing the
// data, and exposes the resultant global to further GlobalOpt.
// We cannot optimize the malloc if we cannot determine malloc array size.
- Value *NElems = getMallocArraySize(CI, &DL, TLI, true);
+ Value *NElems = getMallocArraySize(CI, DL, TLI, true);
if (!NElems)
return false;
CI = cast<CallInst>(Malloc);
}
- GVI = PerformHeapAllocSRoA(GV, CI, getMallocArraySize(CI, &DL, TLI, true),
+ GVI = PerformHeapAllocSRoA(GV, CI, getMallocArraySize(CI, DL, TLI, true),
DL, TLI);
return true;
}
if (GV->hasInitializer())
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(GV->getInitializer())) {
auto &DL = M.getDataLayout();
- Constant *New = ConstantFoldConstantExpression(CE, &DL, TLI);
+ Constant *New = ConstantFoldConstantExpression(CE, DL, TLI);
if (New && New != CE)
GV->setInitializer(New);
}
Constant *Ptr = getVal(SI->getOperand(1));
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ptr)) {
DEBUG(dbgs() << "Folding constant ptr expression: " << *Ptr);
- Ptr = ConstantFoldConstantExpression(CE, &DL, TLI);
+ Ptr = ConstantFoldConstantExpression(CE, DL, TLI);
DEBUG(dbgs() << "; To: " << *Ptr << "\n");
}
if (!isSimpleEnoughPointerToCommit(Ptr)) {
Ptr = ConstantExpr::getGetElementPtr(Ptr, IdxList);
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ptr))
- Ptr = ConstantFoldConstantExpression(CE, &DL, TLI);
+ Ptr = ConstantFoldConstantExpression(CE, DL, TLI);
// If we can't improve the situation by introspecting NewTy,
// we have to give up.
Constant *Ptr = getVal(LI->getOperand(0));
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(Ptr)) {
- Ptr = ConstantFoldConstantExpression(CE, &DL, TLI);
+ Ptr = ConstantFoldConstantExpression(CE, DL, TLI);
DEBUG(dbgs() << "Found a constant pointer expression, constant "
"folding: " << *Ptr << "\n");
}
if (!CurInst->use_empty()) {
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(InstResult))
- InstResult = ConstantFoldConstantExpression(CE, &DL, TLI);
+ InstResult = ConstantFoldConstantExpression(CE, DL, TLI);
setVal(CurInst, InstResult);
}
}
bool BitSetInfo::containsValue(
- const DataLayout *DL,
+ const DataLayout &DL,
const DenseMap<GlobalVariable *, uint64_t> &GlobalLayout, Value *V,
uint64_t COffset) const {
if (auto GV = dyn_cast<GlobalVariable>(V)) {
}
if (auto GEP = dyn_cast<GEPOperator>(V)) {
- APInt APOffset(DL->getPointerSizeInBits(0), 0);
- bool Result = GEP->accumulateConstantOffset(*DL, APOffset);
+ APInt APOffset(DL.getPointerSizeInBits(0), 0);
+ bool Result = GEP->accumulateConstantOffset(DL, APOffset);
if (!Result)
return false;
COffset += APOffset.getZExtValue();
Module *M;
- const DataLayout *DL;
IntegerType *Int1Ty;
IntegerType *Int8Ty;
IntegerType *Int32Ty;
bool LowerBitSets::doInitialization(Module &Mod) {
M = &Mod;
- DL = &Mod.getDataLayout();
+ const DataLayout &DL = Mod.getDataLayout();
Int1Ty = Type::getInt1Ty(M->getContext());
Int8Ty = Type::getInt8Ty(M->getContext());
Int32Ty = Type::getInt32Ty(M->getContext());
Int32PtrTy = PointerType::getUnqual(Int32Ty);
Int64Ty = Type::getInt64Ty(M->getContext());
- IntPtrTy = DL->getIntPtrType(M->getContext(), 0);
+ IntPtrTy = DL.getIntPtrType(M->getContext(), 0);
BitSetNM = M->getNamedMetadata("llvm.bitsets");
GlobalVariable *CombinedGlobal,
const DenseMap<GlobalVariable *, uint64_t> &GlobalLayout) {
Value *Ptr = CI->getArgOperand(0);
+ const DataLayout &DL = M->getDataLayout();
if (BSI.containsValue(DL, GlobalLayout, Ptr))
return ConstantInt::getTrue(CombinedGlobal->getParent()->getContext());
Value *OffsetSHR =
B.CreateLShr(PtrOffset, ConstantInt::get(IntPtrTy, BSI.AlignLog2));
Value *OffsetSHL = B.CreateShl(
- PtrOffset, ConstantInt::get(IntPtrTy, DL->getPointerSizeInBits(0) -
- BSI.AlignLog2));
+ PtrOffset,
+ ConstantInt::get(IntPtrTy, DL.getPointerSizeInBits(0) - BSI.AlignLog2));
BitOffset = B.CreateOr(OffsetSHR, OffsetSHL);
}
const std::vector<GlobalVariable *> &Globals) {
// Build a new global with the combined contents of the referenced globals.
std::vector<Constant *> GlobalInits;
+ const DataLayout &DL = M->getDataLayout();
for (GlobalVariable *G : Globals) {
GlobalInits.push_back(G->getInitializer());
- uint64_t InitSize = DL->getTypeAllocSize(G->getInitializer()->getType());
+ uint64_t InitSize = DL.getTypeAllocSize(G->getInitializer()->getType());
// Compute the amount of padding required to align the next element to the
// next power of 2.
GlobalValue::PrivateLinkage, NewInit);
const StructLayout *CombinedGlobalLayout =
- DL->getStructLayout(cast<StructType>(NewInit->getType()));
+ DL.getStructLayout(cast<StructType>(NewInit->getType()));
// Compute the offsets of the original globals within the new global.
DenseMap<GlobalVariable *, uint64_t> GlobalLayout;
/// side of claiming that two functions are different).
class FunctionComparator {
public:
- FunctionComparator(const DataLayout *DL, const Function *F1,
- const Function *F2)
- : FnL(F1), FnR(F2), DL(DL) {}
+ FunctionComparator(const Function *F1, const Function *F2)
+ : FnL(F1), FnR(F2) {}
/// Test whether the two functions have equivalent behaviour.
int compare();
/// Parts to be compared for each comparison stage,
/// most significant stage first:
/// 1. Address space. As numbers.
- /// 2. Constant offset, (if "DataLayout *DL" field is not NULL,
- /// using GEPOperator::accumulateConstantOffset method).
+ /// 2. Constant offset, (using GEPOperator::accumulateConstantOffset method).
/// 3. Pointer operand type (using cmpType method).
/// 4. Number of operands.
/// 5. Compare operands, using cmpValues method.
// The two functions undergoing comparison.
const Function *FnL, *FnR;
- const DataLayout *DL;
-
/// Assign serial numbers to values from left function, and values from
/// right function.
/// Explanation:
class FunctionNode {
AssertingVH<Function> F;
- const DataLayout *DL;
public:
- FunctionNode(Function *F, const DataLayout *DL) : F(F), DL(DL) {}
+ FunctionNode(Function *F) : F(F) {}
Function *getFunc() const { return F; }
void release() { F = 0; }
bool operator<(const FunctionNode &RHS) const {
- return (FunctionComparator(DL, F, RHS.getFunc()).compare()) == -1;
+ return (FunctionComparator(F, RHS.getFunc()).compare()) == -1;
}
};
}
PointerType *PTyL = dyn_cast<PointerType>(TyL);
PointerType *PTyR = dyn_cast<PointerType>(TyR);
- if (DL) {
- if (PTyL && PTyL->getAddressSpace() == 0) TyL = DL->getIntPtrType(TyL);
- if (PTyR && PTyR->getAddressSpace() == 0) TyR = DL->getIntPtrType(TyR);
- }
+ const DataLayout &DL = FnL->getParent()->getDataLayout();
+ if (PTyL && PTyL->getAddressSpace() == 0)
+ TyL = DL.getIntPtrType(TyL);
+ if (PTyR && PTyR->getAddressSpace() == 0)
+ TyR = DL.getIntPtrType(TyR);
if (TyL == TyR)
return 0;
// When we have target data, we can reduce the GEP down to the value in bytes
// added to the address.
- if (DL) {
- unsigned BitWidth = DL->getPointerSizeInBits(ASL);
- APInt OffsetL(BitWidth, 0), OffsetR(BitWidth, 0);
- if (GEPL->accumulateConstantOffset(*DL, OffsetL) &&
- GEPR->accumulateConstantOffset(*DL, OffsetR))
- return cmpAPInts(OffsetL, OffsetR);
- }
+ const DataLayout &DL = FnL->getParent()->getDataLayout();
+ unsigned BitWidth = DL.getPointerSizeInBits(ASL);
+ APInt OffsetL(BitWidth, 0), OffsetR(BitWidth, 0);
+ if (GEPL->accumulateConstantOffset(DL, OffsetL) &&
+ GEPR->accumulateConstantOffset(DL, OffsetR))
+ return cmpAPInts(OffsetL, OffsetR);
if (int Res = cmpNumbers((uint64_t)GEPL->getPointerOperand()->getType(),
(uint64_t)GEPR->getPointerOperand()->getType()))
/// to modify it.
FnTreeType FnTree;
- /// DataLayout for more accurate GEP comparisons. May be NULL.
- const DataLayout *DL;
-
/// Whether or not the target supports global aliases.
bool HasGlobalAliases;
};
for (std::vector<WeakVH>::iterator J = I; J != E && j < Max; ++J, ++j) {
Function *F1 = cast<Function>(*I);
Function *F2 = cast<Function>(*J);
- int Res1 = FunctionComparator(DL, F1, F2).compare();
- int Res2 = FunctionComparator(DL, F2, F1).compare();
+ int Res1 = FunctionComparator(F1, F2).compare();
+ int Res2 = FunctionComparator(F2, F1).compare();
// If F1 <= F2, then F2 >= F1, otherwise report failure.
if (Res1 != -Res2) {
continue;
Function *F3 = cast<Function>(*K);
- int Res3 = FunctionComparator(DL, F1, F3).compare();
- int Res4 = FunctionComparator(DL, F2, F3).compare();
+ int Res3 = FunctionComparator(F1, F3).compare();
+ int Res4 = FunctionComparator(F2, F3).compare();
bool Transitive = true;
bool MergeFunctions::runOnModule(Module &M) {
bool Changed = false;
- DL = &M.getDataLayout();
for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I) {
if (!I->isDeclaration() && !I->hasAvailableExternallyLinkage())
// that was already inserted.
bool MergeFunctions::insert(Function *NewFunction) {
std::pair<FnTreeType::iterator, bool> Result =
- FnTree.insert(FunctionNode(NewFunction, DL));
+ FnTree.insert(FunctionNode(NewFunction));
if (Result.second) {
DEBUG(dbgs() << "Inserting as unique: " << NewFunction->getName() << '\n');
void MergeFunctions::remove(Function *F) {
// We need to make sure we remove F, not a function "equal" to F per the
// function equality comparator.
- FnTreeType::iterator found = FnTree.find(FunctionNode(F, DL));
+ FnTreeType::iterator found = FnTree.find(FunctionNode(F));
size_t Erased = 0;
if (found != FnTree.end() && found->getFunc() == F) {
Erased = 1;
/// This basically requires proving that the add in the original type would not
/// overflow to change the sign bit or have a carry out.
bool InstCombiner::WillNotOverflowSignedAdd(Value *LHS, Value *RHS,
- Instruction *CxtI) {
+ Instruction &CxtI) {
// There are different heuristics we can use for this. Here are some simple
// ones.
//
// Since the carry into the most significant position is always equal to
// the carry out of the addition, there is no signed overflow.
- if (ComputeNumSignBits(LHS, 0, CxtI) > 1 &&
- ComputeNumSignBits(RHS, 0, CxtI) > 1)
+ if (ComputeNumSignBits(LHS, 0, &CxtI) > 1 &&
+ ComputeNumSignBits(RHS, 0, &CxtI) > 1)
return true;
unsigned BitWidth = LHS->getType()->getScalarSizeInBits();
APInt LHSKnownZero(BitWidth, 0);
APInt LHSKnownOne(BitWidth, 0);
- computeKnownBits(LHS, LHSKnownZero, LHSKnownOne, 0, CxtI);
+ computeKnownBits(LHS, LHSKnownZero, LHSKnownOne, 0, &CxtI);
APInt RHSKnownZero(BitWidth, 0);
APInt RHSKnownOne(BitWidth, 0);
- computeKnownBits(RHS, RHSKnownZero, RHSKnownOne, 0, CxtI);
+ computeKnownBits(RHS, RHSKnownZero, RHSKnownOne, 0, &CxtI);
// Addition of two 2's compliment numbers having opposite signs will never
// overflow.
/// overflow to change the sign bit or have a carry out.
/// TODO: Handle this for Vectors.
bool InstCombiner::WillNotOverflowSignedSub(Value *LHS, Value *RHS,
- Instruction *CxtI) {
+ Instruction &CxtI) {
// If LHS and RHS each have at least two sign bits, the subtraction
// cannot overflow.
- if (ComputeNumSignBits(LHS, 0, CxtI) > 1 &&
- ComputeNumSignBits(RHS, 0, CxtI) > 1)
+ if (ComputeNumSignBits(LHS, 0, &CxtI) > 1 &&
+ ComputeNumSignBits(RHS, 0, &CxtI) > 1)
return true;
unsigned BitWidth = LHS->getType()->getScalarSizeInBits();
APInt LHSKnownZero(BitWidth, 0);
APInt LHSKnownOne(BitWidth, 0);
- computeKnownBits(LHS, LHSKnownZero, LHSKnownOne, 0, CxtI);
+ computeKnownBits(LHS, LHSKnownZero, LHSKnownOne, 0, &CxtI);
APInt RHSKnownZero(BitWidth, 0);
APInt RHSKnownOne(BitWidth, 0);
- computeKnownBits(RHS, RHSKnownZero, RHSKnownOne, 0, CxtI);
+ computeKnownBits(RHS, RHSKnownZero, RHSKnownOne, 0, &CxtI);
// Subtraction of two 2's compliment numbers having identical signs will
// never overflow.
/// \brief Return true if we can prove that:
/// (sub LHS, RHS) === (sub nuw LHS, RHS)
bool InstCombiner::WillNotOverflowUnsignedSub(Value *LHS, Value *RHS,
- Instruction *CxtI) {
+ Instruction &CxtI) {
// If the LHS is negative and the RHS is non-negative, no unsigned wrap.
bool LHSKnownNonNegative, LHSKnownNegative;
bool RHSKnownNonNegative, RHSKnownNegative;
- ComputeSignBit(LHS, LHSKnownNonNegative, LHSKnownNegative, /*Depth=*/0, CxtI);
- ComputeSignBit(RHS, RHSKnownNonNegative, RHSKnownNegative, /*Depth=*/0, CxtI);
+ ComputeSignBit(LHS, LHSKnownNonNegative, LHSKnownNegative, /*Depth=*/0,
+ &CxtI);
+ ComputeSignBit(RHS, RHSKnownNonNegative, RHSKnownNegative, /*Depth=*/0,
+ &CxtI);
if (LHSKnownNegative && RHSKnownNonNegative)
return true;
}
Instruction *InstCombiner::visitAdd(BinaryOperator &I) {
- bool Changed = SimplifyAssociativeOrCommutative(I);
- Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
+ bool Changed = SimplifyAssociativeOrCommutative(I);
+ Value *LHS = I.getOperand(0), *RHS = I.getOperand(1);
- if (Value *V = SimplifyVectorOp(I))
- return ReplaceInstUsesWith(I, V);
+ if (Value *V = SimplifyVectorOp(I))
+ return ReplaceInstUsesWith(I, V);
- if (Value *V = SimplifyAddInst(LHS, RHS, I.hasNoSignedWrap(),
- I.hasNoUnsignedWrap(), DL, TLI, DT, AC))
- return ReplaceInstUsesWith(I, V);
+ if (Value *V = SimplifyAddInst(LHS, RHS, I.hasNoSignedWrap(),
+ I.hasNoUnsignedWrap(), DL, TLI, DT, AC))
+ return ReplaceInstUsesWith(I, V);
// (A*B)+(A*C) -> A*(B+C) etc
if (Value *V = SimplifyUsingDistributiveLaws(I))
ConstantExpr::getTrunc(RHSC, LHSConv->getOperand(0)->getType());
if (LHSConv->hasOneUse() &&
ConstantExpr::getSExt(CI, I.getType()) == RHSC &&
- WillNotOverflowSignedAdd(LHSConv->getOperand(0), CI, &I)) {
+ WillNotOverflowSignedAdd(LHSConv->getOperand(0), CI, I)) {
// Insert the new, smaller add.
Value *NewAdd = Builder->CreateNSWAdd(LHSConv->getOperand(0),
CI, "addconv");
// Only do this if x/y have the same type, if at last one of them has a
// single use (so we don't increase the number of sexts), and if the
// integer add will not overflow.
- if (LHSConv->getOperand(0)->getType()==RHSConv->getOperand(0)->getType()&&
+ if (LHSConv->getOperand(0)->getType() ==
+ RHSConv->getOperand(0)->getType() &&
(LHSConv->hasOneUse() || RHSConv->hasOneUse()) &&
WillNotOverflowSignedAdd(LHSConv->getOperand(0),
- RHSConv->getOperand(0), &I)) {
+ RHSConv->getOperand(0), I)) {
// Insert the new integer add.
Value *NewAdd = Builder->CreateNSWAdd(LHSConv->getOperand(0),
RHSConv->getOperand(0), "addconv");
// TODO(jingyue): Consider WillNotOverflowSignedAdd and
// WillNotOverflowUnsignedAdd to reduce the number of invocations of
// computeKnownBits.
- if (!I.hasNoSignedWrap() && WillNotOverflowSignedAdd(LHS, RHS, &I)) {
+ if (!I.hasNoSignedWrap() && WillNotOverflowSignedAdd(LHS, RHS, I)) {
Changed = true;
I.setHasNoSignedWrap(true);
}
ConstantExpr::getFPToSI(CFP, LHSConv->getOperand(0)->getType());
if (LHSConv->hasOneUse() &&
ConstantExpr::getSIToFP(CI, I.getType()) == CFP &&
- WillNotOverflowSignedAdd(LHSConv->getOperand(0), CI, &I)) {
+ WillNotOverflowSignedAdd(LHSConv->getOperand(0), CI, I)) {
// Insert the new integer add.
Value *NewAdd = Builder->CreateNSWAdd(LHSConv->getOperand(0),
CI, "addconv");
// Only do this if x/y have the same type, if at last one of them has a
// single use (so we don't increase the number of int->fp conversions),
// and if the integer add will not overflow.
- if (LHSConv->getOperand(0)->getType()==RHSConv->getOperand(0)->getType()&&
+ if (LHSConv->getOperand(0)->getType() ==
+ RHSConv->getOperand(0)->getType() &&
(LHSConv->hasOneUse() || RHSConv->hasOneUse()) &&
WillNotOverflowSignedAdd(LHSConv->getOperand(0),
- RHSConv->getOperand(0), &I)) {
+ RHSConv->getOperand(0), I)) {
// Insert the new integer add.
Value *NewAdd = Builder->CreateNSWAdd(LHSConv->getOperand(0),
RHSConv->getOperand(0),"addconv");
///
Value *InstCombiner::OptimizePointerDifference(Value *LHS, Value *RHS,
Type *Ty) {
- assert(DL && "Must have target data info for this");
-
// If LHS is a gep based on RHS or RHS is a gep based on LHS, we can optimize
// this.
bool Swapped = false;
// Optimize pointer differences into the same array into a size. Consider:
// &A[10] - &A[0]: we should compile this to "10".
- if (DL) {
- Value *LHSOp, *RHSOp;
- if (match(Op0, m_PtrToInt(m_Value(LHSOp))) &&
- match(Op1, m_PtrToInt(m_Value(RHSOp))))
- if (Value *Res = OptimizePointerDifference(LHSOp, RHSOp, I.getType()))
- return ReplaceInstUsesWith(I, Res);
-
- // trunc(p)-trunc(q) -> trunc(p-q)
- if (match(Op0, m_Trunc(m_PtrToInt(m_Value(LHSOp)))) &&
- match(Op1, m_Trunc(m_PtrToInt(m_Value(RHSOp)))))
- if (Value *Res = OptimizePointerDifference(LHSOp, RHSOp, I.getType()))
- return ReplaceInstUsesWith(I, Res);
- }
+ Value *LHSOp, *RHSOp;
+ if (match(Op0, m_PtrToInt(m_Value(LHSOp))) &&
+ match(Op1, m_PtrToInt(m_Value(RHSOp))))
+ if (Value *Res = OptimizePointerDifference(LHSOp, RHSOp, I.getType()))
+ return ReplaceInstUsesWith(I, Res);
+
+ // trunc(p)-trunc(q) -> trunc(p-q)
+ if (match(Op0, m_Trunc(m_PtrToInt(m_Value(LHSOp)))) &&
+ match(Op1, m_Trunc(m_PtrToInt(m_Value(RHSOp)))))
+ if (Value *Res = OptimizePointerDifference(LHSOp, RHSOp, I.getType()))
+ return ReplaceInstUsesWith(I, Res);
bool Changed = false;
- if (!I.hasNoSignedWrap() && WillNotOverflowSignedSub(Op0, Op1, &I)) {
+ if (!I.hasNoSignedWrap() && WillNotOverflowSignedSub(Op0, Op1, I)) {
Changed = true;
I.setHasNoSignedWrap(true);
}
- if (!I.hasNoUnsignedWrap() && WillNotOverflowUnsignedSub(Op0, Op1, &I)) {
+ if (!I.hasNoUnsignedWrap() && WillNotOverflowUnsignedSub(Op0, Op1, I)) {
Changed = true;
I.setHasNoUnsignedWrap(true);
}
Value *Mask = nullptr;
Value *Masked = nullptr;
if (LAnd->getOperand(0) == RAnd->getOperand(0) &&
- isKnownToBeAPowerOfTwo(LAnd->getOperand(1), false, 0, AC, CxtI, DT) &&
- isKnownToBeAPowerOfTwo(RAnd->getOperand(1), false, 0, AC, CxtI, DT)) {
+ isKnownToBeAPowerOfTwo(LAnd->getOperand(1), DL, false, 0, AC, CxtI,
+ DT) &&
+ isKnownToBeAPowerOfTwo(RAnd->getOperand(1), DL, false, 0, AC, CxtI,
+ DT)) {
Mask = Builder->CreateOr(LAnd->getOperand(1), RAnd->getOperand(1));
Masked = Builder->CreateAnd(LAnd->getOperand(0), Mask);
} else if (LAnd->getOperand(1) == RAnd->getOperand(1) &&
- isKnownToBeAPowerOfTwo(LAnd->getOperand(0), false, 0, AC, CxtI,
- DT) &&
- isKnownToBeAPowerOfTwo(RAnd->getOperand(0), false, 0, AC, CxtI,
- DT)) {
+ isKnownToBeAPowerOfTwo(LAnd->getOperand(0), DL, false, 0, AC,
+ CxtI, DT) &&
+ isKnownToBeAPowerOfTwo(RAnd->getOperand(0), DL, false, 0, AC,
+ CxtI, DT)) {
Mask = Builder->CreateOr(LAnd->getOperand(0), RAnd->getOperand(0));
Masked = Builder->CreateAnd(LAnd->getOperand(1), Mask);
}
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/IR/CallSite.h"
-#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/IR/Statepoint.h"
}
Instruction *InstCombiner::SimplifyMemTransfer(MemIntrinsic *MI) {
- unsigned DstAlign = getKnownAlignment(MI->getArgOperand(0), DL, AC, MI, DT);
- unsigned SrcAlign = getKnownAlignment(MI->getArgOperand(1), DL, AC, MI, DT);
+ unsigned DstAlign = getKnownAlignment(MI->getArgOperand(0), DL, MI, AC, DT);
+ unsigned SrcAlign = getKnownAlignment(MI->getArgOperand(1), DL, MI, AC, DT);
unsigned MinAlign = std::min(DstAlign, SrcAlign);
unsigned CopyAlign = MI->getAlignment();
if (StrippedDest != MI->getArgOperand(0)) {
Type *SrcETy = cast<PointerType>(StrippedDest->getType())
->getElementType();
- if (DL && SrcETy->isSized() && DL->getTypeStoreSize(SrcETy) == Size) {
+ if (SrcETy->isSized() && DL.getTypeStoreSize(SrcETy) == Size) {
// The SrcETy might be something like {{{double}}} or [1 x double]. Rip
// down through these levels if so.
SrcETy = reduceToSingleValueType(SrcETy);
}
Instruction *InstCombiner::SimplifyMemSet(MemSetInst *MI) {
- unsigned Alignment = getKnownAlignment(MI->getDest(), DL, AC, MI, DT);
+ unsigned Alignment = getKnownAlignment(MI->getDest(), DL, MI, AC, DT);
if (MI->getAlignment() < Alignment) {
MI->setAlignment(ConstantInt::get(MI->getAlignmentType(),
Alignment, false));
// can prove that it will never overflow.
if (II->getIntrinsicID() == Intrinsic::sadd_with_overflow) {
Value *LHS = II->getArgOperand(0), *RHS = II->getArgOperand(1);
- if (WillNotOverflowSignedAdd(LHS, RHS, II)) {
+ if (WillNotOverflowSignedAdd(LHS, RHS, *II)) {
return CreateOverflowTuple(II, Builder->CreateNSWAdd(LHS, RHS), false);
}
}
}
}
if (II->getIntrinsicID() == Intrinsic::ssub_with_overflow) {
- if (WillNotOverflowSignedSub(LHS, RHS, II)) {
+ if (WillNotOverflowSignedSub(LHS, RHS, *II)) {
return CreateOverflowTuple(II, Builder->CreateNSWSub(LHS, RHS), false);
}
} else {
- if (WillNotOverflowUnsignedSub(LHS, RHS, II)) {
+ if (WillNotOverflowUnsignedSub(LHS, RHS, *II)) {
return CreateOverflowTuple(II, Builder->CreateNUWSub(LHS, RHS), false);
}
}
}
if (II->getIntrinsicID() == Intrinsic::smul_with_overflow) {
Value *LHS = II->getArgOperand(0), *RHS = II->getArgOperand(1);
- if (WillNotOverflowSignedMul(LHS, RHS, II)) {
+ if (WillNotOverflowSignedMul(LHS, RHS, *II)) {
return CreateOverflowTuple(II, Builder->CreateNSWMul(LHS, RHS), false);
}
}
case Intrinsic::ppc_altivec_lvx:
case Intrinsic::ppc_altivec_lvxl:
// Turn PPC lvx -> load if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, DL, AC, II, DT) >=
+ if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, DL, II, AC, DT) >=
16) {
Value *Ptr = Builder->CreateBitCast(II->getArgOperand(0),
PointerType::getUnqual(II->getType()));
case Intrinsic::ppc_altivec_stvx:
case Intrinsic::ppc_altivec_stvxl:
// Turn stvx -> store if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(1), 16, DL, AC, II, DT) >=
+ if (getOrEnforceKnownAlignment(II->getArgOperand(1), 16, DL, II, AC, DT) >=
16) {
Type *OpPtrTy =
PointerType::getUnqual(II->getArgOperand(0)->getType());
}
case Intrinsic::ppc_qpx_qvlfs:
// Turn PPC QPX qvlfs -> load if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, DL, AC, II, DT) >=
+ if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, DL, II, AC, DT) >=
16) {
Value *Ptr = Builder->CreateBitCast(II->getArgOperand(0),
PointerType::getUnqual(II->getType()));
break;
case Intrinsic::ppc_qpx_qvlfd:
// Turn PPC QPX qvlfd -> load if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(0), 32, DL, AC, II, DT) >=
+ if (getOrEnforceKnownAlignment(II->getArgOperand(0), 32, DL, II, AC, DT) >=
32) {
Value *Ptr = Builder->CreateBitCast(II->getArgOperand(0),
PointerType::getUnqual(II->getType()));
break;
case Intrinsic::ppc_qpx_qvstfs:
// Turn PPC QPX qvstfs -> store if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(1), 16, DL, AC, II, DT) >=
+ if (getOrEnforceKnownAlignment(II->getArgOperand(1), 16, DL, II, AC, DT) >=
16) {
Type *OpPtrTy =
PointerType::getUnqual(II->getArgOperand(0)->getType());
break;
case Intrinsic::ppc_qpx_qvstfd:
// Turn PPC QPX qvstfd -> store if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(1), 32, DL, AC, II, DT) >=
+ if (getOrEnforceKnownAlignment(II->getArgOperand(1), 32, DL, II, AC, DT) >=
32) {
Type *OpPtrTy =
PointerType::getUnqual(II->getArgOperand(0)->getType());
case Intrinsic::x86_sse2_storeu_pd:
case Intrinsic::x86_sse2_storeu_dq:
// Turn X86 storeu -> store if the pointer is known aligned.
- if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, DL, AC, II, DT) >=
+ if (getOrEnforceKnownAlignment(II->getArgOperand(0), 16, DL, II, AC, DT) >=
16) {
Type *OpPtrTy =
PointerType::getUnqual(II->getArgOperand(1)->getType());
unsigned LowHalfElts = VWidth / 2;
APInt InputDemandedElts(APInt::getBitsSet(VWidth, 0, LowHalfElts));
APInt UndefElts(VWidth, 0);
- if (Value *TmpV = SimplifyDemandedVectorElts(II->getArgOperand(0),
- InputDemandedElts,
- UndefElts)) {
+ if (Value *TmpV = SimplifyDemandedVectorElts(
+ II->getArgOperand(0), InputDemandedElts, UndefElts)) {
II->setArgOperand(0, TmpV);
return II;
}
unsigned Idx =
cast<ConstantInt>(Mask->getAggregateElement(i))->getZExtValue();
Idx &= 31; // Match the hardware behavior.
- if (DL && DL->isLittleEndian())
+ if (DL.isLittleEndian())
Idx = 31 - Idx;
if (!ExtractedElts[Idx]) {
- Value *Op0ToUse = (DL && DL->isLittleEndian()) ? Op1 : Op0;
- Value *Op1ToUse = (DL && DL->isLittleEndian()) ? Op0 : Op1;
+ Value *Op0ToUse = (DL.isLittleEndian()) ? Op1 : Op0;
+ Value *Op1ToUse = (DL.isLittleEndian()) ? Op0 : Op1;
ExtractedElts[Idx] =
Builder->CreateExtractElement(Idx < 16 ? Op0ToUse : Op1ToUse,
Builder->getInt32(Idx&15));
case Intrinsic::arm_neon_vst2lane:
case Intrinsic::arm_neon_vst3lane:
case Intrinsic::arm_neon_vst4lane: {
- unsigned MemAlign = getKnownAlignment(II->getArgOperand(0), DL, AC, II, DT);
+ unsigned MemAlign = getKnownAlignment(II->getArgOperand(0), DL, II, AC, DT);
unsigned AlignArg = II->getNumArgOperands() - 1;
ConstantInt *IntrAlign = dyn_cast<ConstantInt>(II->getArgOperand(AlignArg));
if (IntrAlign && IntrAlign->getZExtValue() < MemAlign) {
RHS->getType()->isPointerTy() &&
cast<Constant>(RHS)->isNullValue()) {
LoadInst* LI = cast<LoadInst>(LHS);
- if (isValidAssumeForContext(II, LI, DL, DT)) {
+ if (isValidAssumeForContext(II, LI, DT)) {
MDNode *MD = MDNode::get(II->getContext(), None);
LI->setMetadata(LLVMContext::MD_nonnull, MD);
return EraseInstFromFunction(*II);
/// isSafeToEliminateVarargsCast - If this cast does not affect the value
/// passed through the varargs area, we can eliminate the use of the cast.
static bool isSafeToEliminateVarargsCast(const CallSite CS,
- const CastInst * const CI,
- const DataLayout * const DL,
+ const DataLayout &DL,
+ const CastInst *const CI,
const int ix) {
if (!CI->isLosslessCast())
return false;
Type* DstTy = cast<PointerType>(CI->getType())->getElementType();
if (!SrcTy->isSized() || !DstTy->isSized())
return false;
- if (!DL || DL->getTypeAllocSize(SrcTy) != DL->getTypeAllocSize(DstTy))
+ if (DL.getTypeAllocSize(SrcTy) != DL.getTypeAllocSize(DstTy))
return false;
return true;
}
// Currently we're only working with the checking functions, memcpy_chk,
// mempcpy_chk, memmove_chk, memset_chk, strcpy_chk, stpcpy_chk, strncpy_chk,
// strcat_chk and strncat_chk.
-Instruction *InstCombiner::tryOptimizeCall(CallInst *CI, const DataLayout *DL) {
+Instruction *InstCombiner::tryOptimizeCall(CallInst *CI) {
if (!CI->getCalledFunction()) return nullptr;
auto InstCombineRAUW = [this](Instruction *From, Value *With) {
for (CallSite::arg_iterator I = CS.arg_begin() + FTy->getNumParams(),
E = CS.arg_end(); I != E; ++I, ++ix) {
CastInst *CI = dyn_cast<CastInst>(*I);
- if (CI && isSafeToEliminateVarargsCast(CS, CI, DL, ix)) {
+ if (CI && isSafeToEliminateVarargsCast(CS, DL, CI, ix)) {
*I = CI->getOperand(0);
Changed = true;
}
// this. None of these calls are seen as possibly dead so go ahead and
// delete the instruction now.
if (CallInst *CI = dyn_cast<CallInst>(CS.getInstruction())) {
- Instruction *I = tryOptimizeCall(CI, DL);
+ Instruction *I = tryOptimizeCall(CI);
// If we changed something return the result, etc. Otherwise let
// the fallthrough check.
if (I) return EraseInstFromFunction(*I);
CallerPAL.getParamAttributes(i + 1).hasAttribute(i + 1,
Attribute::ByVal)) {
PointerType *ParamPTy = dyn_cast<PointerType>(ParamTy);
- if (!ParamPTy || !ParamPTy->getElementType()->isSized() || !DL)
+ if (!ParamPTy || !ParamPTy->getElementType()->isSized())
return false;
Type *CurElTy = ActTy->getPointerElementType();
- if (DL->getTypeAllocSize(CurElTy) !=
- DL->getTypeAllocSize(ParamPTy->getElementType()))
+ if (DL.getTypeAllocSize(CurElTy) !=
+ DL.getTypeAllocSize(ParamPTy->getElementType()))
return false;
}
}
/// try to eliminate the cast by moving the type information into the alloc.
Instruction *InstCombiner::PromoteCastOfAllocation(BitCastInst &CI,
AllocaInst &AI) {
- // This requires DataLayout to get the alloca alignment and size information.
- if (!DL) return nullptr;
-
PointerType *PTy = cast<PointerType>(CI.getType());
BuilderTy AllocaBuilder(*Builder);
Type *CastElTy = PTy->getElementType();
if (!AllocElTy->isSized() || !CastElTy->isSized()) return nullptr;
- unsigned AllocElTyAlign = DL->getABITypeAlignment(AllocElTy);
- unsigned CastElTyAlign = DL->getABITypeAlignment(CastElTy);
+ unsigned AllocElTyAlign = DL.getABITypeAlignment(AllocElTy);
+ unsigned CastElTyAlign = DL.getABITypeAlignment(CastElTy);
if (CastElTyAlign < AllocElTyAlign) return nullptr;
// If the allocation has multiple uses, only promote it if we are strictly
// same, we open the door to infinite loops of various kinds.
if (!AI.hasOneUse() && CastElTyAlign == AllocElTyAlign) return nullptr;
- uint64_t AllocElTySize = DL->getTypeAllocSize(AllocElTy);
- uint64_t CastElTySize = DL->getTypeAllocSize(CastElTy);
+ uint64_t AllocElTySize = DL.getTypeAllocSize(AllocElTy);
+ uint64_t CastElTySize = DL.getTypeAllocSize(CastElTy);
if (CastElTySize == 0 || AllocElTySize == 0) return nullptr;
// If the allocation has multiple uses, only promote it if we're not
// shrinking the amount of memory being allocated.
- uint64_t AllocElTyStoreSize = DL->getTypeStoreSize(AllocElTy);
- uint64_t CastElTyStoreSize = DL->getTypeStoreSize(CastElTy);
+ uint64_t AllocElTyStoreSize = DL.getTypeStoreSize(AllocElTy);
+ uint64_t CastElTyStoreSize = DL.getTypeStoreSize(CastElTy);
if (!AI.hasOneUse() && CastElTyStoreSize < AllocElTyStoreSize) return nullptr;
// See if we can satisfy the modulus by pulling a scale out of the array
PHINode *OPN = cast<PHINode>(I);
PHINode *NPN = PHINode::Create(Ty, OPN->getNumIncomingValues());
for (unsigned i = 0, e = OPN->getNumIncomingValues(); i != e; ++i) {
- Value *V =EvaluateInDifferentType(OPN->getIncomingValue(i), Ty, isSigned);
+ Value *V =
+ EvaluateInDifferentType(OPN->getIncomingValue(i), Ty, isSigned);
NPN->addIncoming(V, OPN->getIncomingBlock(i));
}
Res = NPN;
/// This function is a wrapper around CastInst::isEliminableCastPair. It
/// simply extracts arguments and returns what that function returns.
static Instruction::CastOps
-isEliminableCastPair(
- const CastInst *CI, ///< The first cast instruction
- unsigned opcode, ///< The opcode of the second cast instruction
- Type *DstTy, ///< The target type for the second cast instruction
- const DataLayout *DL ///< The target data for pointer size
-) {
-
+isEliminableCastPair(const CastInst *CI, ///< First cast instruction
+ unsigned opcode, ///< Opcode for the second cast
+ Type *DstTy, ///< Target type for the second cast
+ const DataLayout &DL) {
Type *SrcTy = CI->getOperand(0)->getType(); // A from above
Type *MidTy = CI->getType(); // B from above
// Get the opcodes of the two Cast instructions
Instruction::CastOps firstOp = Instruction::CastOps(CI->getOpcode());
Instruction::CastOps secondOp = Instruction::CastOps(opcode);
- Type *SrcIntPtrTy = DL && SrcTy->isPtrOrPtrVectorTy() ?
- DL->getIntPtrType(SrcTy) : nullptr;
- Type *MidIntPtrTy = DL && MidTy->isPtrOrPtrVectorTy() ?
- DL->getIntPtrType(MidTy) : nullptr;
- Type *DstIntPtrTy = DL && DstTy->isPtrOrPtrVectorTy() ?
- DL->getIntPtrType(DstTy) : nullptr;
+ Type *SrcIntPtrTy =
+ SrcTy->isPtrOrPtrVectorTy() ? DL.getIntPtrType(SrcTy) : nullptr;
+ Type *MidIntPtrTy =
+ MidTy->isPtrOrPtrVectorTy() ? DL.getIntPtrType(MidTy) : nullptr;
+ Type *DstIntPtrTy =
+ DstTy->isPtrOrPtrVectorTy() ? DL.getIntPtrType(DstTy) : nullptr;
unsigned Res = CastInst::isEliminableCastPair(firstOp, secondOp, SrcTy, MidTy,
DstTy, SrcIntPtrTy, MidIntPtrTy,
DstIntPtrTy);
// eliminate it now.
if (CastInst *CSrc = dyn_cast<CastInst>(Src)) { // A->B->C cast
if (Instruction::CastOps opc =
- isEliminableCastPair(CSrc, CI.getOpcode(), CI.getType(), DL)) {
+ isEliminableCastPair(CSrc, CI.getOpcode(), CI.getType(), DL)) {
// The first cast (CSrc) is eliminable so we need to fix up or replace
// the second cast (CI). CSrc will then have a good chance of being dead.
return CastInst::Create(opc, CSrc->getOperand(0), CI.getType());
if (isa<PHINode>(Src)) {
// We don't do this if this would create a PHI node with an illegal type if
// it is currently legal.
- if (!Src->getType()->isIntegerTy() ||
- !CI.getType()->isIntegerTy() ||
+ if (!Src->getType()->isIntegerTy() || !CI.getType()->isIntegerTy() ||
ShouldChangeType(CI.getType(), Src->getType()))
if (Instruction *NV = FoldOpIntoPhi(CI))
return NV;
// If the source integer type is not the intptr_t type for this target, do a
// trunc or zext to the intptr_t type, then inttoptr of it. This allows the
// cast to be exposed to other transforms.
-
- if (DL) {
- unsigned AS = CI.getAddressSpace();
- if (CI.getOperand(0)->getType()->getScalarSizeInBits() !=
- DL->getPointerSizeInBits(AS)) {
- Type *Ty = DL->getIntPtrType(CI.getContext(), AS);
- if (CI.getType()->isVectorTy()) // Handle vectors of pointers.
- Ty = VectorType::get(Ty, CI.getType()->getVectorNumElements());
-
- Value *P = Builder->CreateZExtOrTrunc(CI.getOperand(0), Ty);
- return new IntToPtrInst(P, CI.getType());
- }
+ unsigned AS = CI.getAddressSpace();
+ if (CI.getOperand(0)->getType()->getScalarSizeInBits() !=
+ DL.getPointerSizeInBits(AS)) {
+ Type *Ty = DL.getIntPtrType(CI.getContext(), AS);
+ if (CI.getType()->isVectorTy()) // Handle vectors of pointers.
+ Ty = VectorType::get(Ty, CI.getType()->getVectorNumElements());
+
+ Value *P = Builder->CreateZExtOrTrunc(CI.getOperand(0), Ty);
+ return new IntToPtrInst(P, CI.getType());
}
if (Instruction *I = commonCastTransforms(CI))
return &CI;
}
- if (!DL)
- return commonCastTransforms(CI);
-
// If the GEP has a single use, and the base pointer is a bitcast, and the
// GEP computes a constant offset, see if we can convert these three
// instructions into fewer. This typically happens with unions and other
// non-type-safe code.
unsigned AS = GEP->getPointerAddressSpace();
- unsigned OffsetBits = DL->getPointerSizeInBits(AS);
+ unsigned OffsetBits = DL.getPointerSizeInBits(AS);
APInt Offset(OffsetBits, 0);
BitCastInst *BCI = dyn_cast<BitCastInst>(GEP->getOperand(0));
- if (GEP->hasOneUse() &&
- BCI &&
- GEP->accumulateConstantOffset(*DL, Offset)) {
+ if (GEP->hasOneUse() && BCI && GEP->accumulateConstantOffset(DL, Offset)) {
// Get the base pointer input of the bitcast, and the type it points to.
Value *OrigBase = BCI->getOperand(0);
SmallVector<Value*, 8> NewIndices;
- if (FindElementAtOffset(OrigBase->getType(),
- Offset.getSExtValue(),
+ if (FindElementAtOffset(OrigBase->getType(), Offset.getSExtValue(),
NewIndices)) {
// If we were able to index down into an element, create the GEP
// and bitcast the result. This eliminates one bitcast, potentially
// do a ptrtoint to intptr_t then do a trunc or zext. This allows the cast
// to be exposed to other transforms.
- if (!DL)
- return commonPointerCastTransforms(CI);
-
Type *Ty = CI.getType();
unsigned AS = CI.getPointerAddressSpace();
- if (Ty->getScalarSizeInBits() == DL->getPointerSizeInBits(AS))
+ if (Ty->getScalarSizeInBits() == DL.getPointerSizeInBits(AS))
return commonPointerCastTransforms(CI);
- Type *PtrTy = DL->getIntPtrType(CI.getContext(), AS);
+ Type *PtrTy = DL.getIntPtrType(CI.getContext(), AS);
if (Ty->isVectorTy()) // Handle vectors of pointers.
PtrTy = VectorType::get(PtrTy, Ty->getVectorNumElements());
/// This returns false if the pattern can't be matched or true if it can,
/// filling in Elements with the elements found here.
static bool CollectInsertionElements(Value *V, unsigned Shift,
- SmallVectorImpl<Value*> &Elements,
- Type *VecEltTy, InstCombiner &IC) {
+ SmallVectorImpl<Value *> &Elements,
+ Type *VecEltTy, bool isBigEndian) {
assert(isMultipleOfTypeSize(Shift, VecEltTy) &&
"Shift should be a multiple of the element type size");
return true;
unsigned ElementIndex = getTypeSizeIndex(Shift, VecEltTy);
- if (IC.getDataLayout()->isBigEndian())
+ if (isBigEndian)
ElementIndex = Elements.size() - ElementIndex - 1;
// Fail if multiple elements are inserted into this slot.
// it to the right type so it gets properly inserted.
if (NumElts == 1)
return CollectInsertionElements(ConstantExpr::getBitCast(C, VecEltTy),
- Shift, Elements, VecEltTy, IC);
+ Shift, Elements, VecEltTy, isBigEndian);
// Okay, this is a constant that covers multiple elements. Slice it up into
// pieces and insert each element-sized piece into the vector.
Constant *Piece = ConstantExpr::getLShr(C, ConstantInt::get(C->getType(),
ShiftI));
Piece = ConstantExpr::getTrunc(Piece, ElementIntTy);
- if (!CollectInsertionElements(Piece, ShiftI, Elements, VecEltTy, IC))
+ if (!CollectInsertionElements(Piece, ShiftI, Elements, VecEltTy,
+ isBigEndian))
return false;
}
return true;
switch (I->getOpcode()) {
default: return false; // Unhandled case.
case Instruction::BitCast:
- return CollectInsertionElements(I->getOperand(0), Shift,
- Elements, VecEltTy, IC);
+ return CollectInsertionElements(I->getOperand(0), Shift, Elements, VecEltTy,
+ isBigEndian);
case Instruction::ZExt:
if (!isMultipleOfTypeSize(
I->getOperand(0)->getType()->getPrimitiveSizeInBits(),
VecEltTy))
return false;
- return CollectInsertionElements(I->getOperand(0), Shift,
- Elements, VecEltTy, IC);
+ return CollectInsertionElements(I->getOperand(0), Shift, Elements, VecEltTy,
+ isBigEndian);
case Instruction::Or:
- return CollectInsertionElements(I->getOperand(0), Shift,
- Elements, VecEltTy, IC) &&
- CollectInsertionElements(I->getOperand(1), Shift,
- Elements, VecEltTy, IC);
+ return CollectInsertionElements(I->getOperand(0), Shift, Elements, VecEltTy,
+ isBigEndian) &&
+ CollectInsertionElements(I->getOperand(1), Shift, Elements, VecEltTy,
+ isBigEndian);
case Instruction::Shl: {
// Must be shifting by a constant that is a multiple of the element size.
ConstantInt *CI = dyn_cast<ConstantInt>(I->getOperand(1));
if (!CI) return false;
Shift += CI->getZExtValue();
if (!isMultipleOfTypeSize(Shift, VecEltTy)) return false;
- return CollectInsertionElements(I->getOperand(0), Shift,
- Elements, VecEltTy, IC);
+ return CollectInsertionElements(I->getOperand(0), Shift, Elements, VecEltTy,
+ isBigEndian);
}
}
/// Into two insertelements that do "buildvector{%inc, %inc5}".
static Value *OptimizeIntegerToVectorInsertions(BitCastInst &CI,
InstCombiner &IC) {
- // We need to know the target byte order to perform this optimization.
- if (!IC.getDataLayout()) return nullptr;
-
VectorType *DestVecTy = cast<VectorType>(CI.getType());
Value *IntInput = CI.getOperand(0);
SmallVector<Value*, 8> Elements(DestVecTy->getNumElements());
if (!CollectInsertionElements(IntInput, 0, Elements,
- DestVecTy->getElementType(), IC))
+ DestVecTy->getElementType(),
+ IC.getDataLayout().isBigEndian()))
return nullptr;
// If we succeeded, we know that all of the element are specified by Elements
/// OptimizeIntToFloatBitCast - See if we can optimize an integer->float/double
/// bitcast. The various long double bitcasts can't get in here.
-static Instruction *OptimizeIntToFloatBitCast(BitCastInst &CI,InstCombiner &IC){
- // We need to know the target byte order to perform this optimization.
- if (!IC.getDataLayout()) return nullptr;
-
+static Instruction *OptimizeIntToFloatBitCast(BitCastInst &CI, InstCombiner &IC,
+ const DataLayout &DL) {
Value *Src = CI.getOperand(0);
Type *DestTy = CI.getType();
}
unsigned Elt = 0;
- if (IC.getDataLayout()->isBigEndian())
+ if (DL.isBigEndian())
Elt = VecTy->getPrimitiveSizeInBits() / DestWidth - 1;
return ExtractElementInst::Create(VecInput, IC.Builder->getInt32(Elt));
}
}
unsigned Elt = ShAmt->getZExtValue() / DestWidth;
- if (IC.getDataLayout()->isBigEndian())
+ if (DL.isBigEndian())
Elt = VecTy->getPrimitiveSizeInBits() / DestWidth - 1 - Elt;
return ExtractElementInst::Create(VecInput, IC.Builder->getInt32(Elt));
}
// Try to optimize int -> float bitcasts.
if ((DestTy->isFloatTy() || DestTy->isDoubleTy()) && isa<IntegerType>(SrcTy))
- if (Instruction *I = OptimizeIntToFloatBitCast(CI, *this))
+ if (Instruction *I = OptimizeIntToFloatBitCast(CI, *this, DL))
return I;
if (VectorType *DestVTy = dyn_cast<VectorType>(DestTy)) {
Instruction *InstCombiner::
FoldCmpLoadFromIndexedGlobal(GetElementPtrInst *GEP, GlobalVariable *GV,
CmpInst &ICI, ConstantInt *AndCst) {
- // We need TD information to know the pointer size unless this is inbounds.
- if (!GEP->isInBounds() && !DL)
- return nullptr;
-
Constant *Init = GV->getInitializer();
if (!isa<ConstantArray>(Init) && !isa<ConstantDataArray>(Init))
return nullptr;
// the array, this will fully represent all the comparison results.
uint64_t MagicBitvector = 0;
-
// Scan the array and see if one of our patterns matches.
Constant *CompareRHS = cast<Constant>(ICI.getOperand(1));
for (unsigned i = 0, e = ArrayElementCount; i != e; ++i) {
// index down like the GEP would do implicitly. We don't have to do this for
// an inbounds GEP because the index can't be out of range.
if (!GEP->isInBounds()) {
- Type *IntPtrTy = DL->getIntPtrType(GEP->getType());
+ Type *IntPtrTy = DL.getIntPtrType(GEP->getType());
unsigned PtrSize = IntPtrTy->getIntegerBitWidth();
if (Idx->getType()->getPrimitiveSizeInBits() > PtrSize)
Idx = Builder->CreateTrunc(Idx, IntPtrTy);
// - Default to i32
if (ArrayElementCount <= Idx->getType()->getIntegerBitWidth())
Ty = Idx->getType();
- else if (DL)
- Ty = DL->getSmallestLegalIntType(Init->getContext(), ArrayElementCount);
- else if (ArrayElementCount <= 32)
- Ty = Type::getInt32Ty(Init->getContext());
+ else
+ Ty = DL.getSmallestLegalIntType(Init->getContext(), ArrayElementCount);
if (Ty) {
Value *V = Builder->CreateIntCast(Idx, Ty, false);
///
/// If we can't emit an optimized form for this expression, this returns null.
///
-static Value *EvaluateGEPOffsetExpression(User *GEP, InstCombiner &IC) {
- const DataLayout &DL = *IC.getDataLayout();
+static Value *EvaluateGEPOffsetExpression(User *GEP, InstCombiner &IC,
+ const DataLayout &DL) {
gep_type_iterator GTI = gep_type_begin(GEP);
// Check to see if this gep only has a single variable index. If so, and if
RHS = RHS->stripPointerCasts();
Value *PtrBase = GEPLHS->getOperand(0);
- if (DL && PtrBase == RHS && GEPLHS->isInBounds()) {
+ if (PtrBase == RHS && GEPLHS->isInBounds()) {
// ((gep Ptr, OFFSET) cmp Ptr) ---> (OFFSET cmp 0).
// This transformation (ignoring the base and scales) is valid because we
// know pointers can't overflow since the gep is inbounds. See if we can
// output an optimized form.
- Value *Offset = EvaluateGEPOffsetExpression(GEPLHS, *this);
+ Value *Offset = EvaluateGEPOffsetExpression(GEPLHS, *this, DL);
// If not, synthesize the offset the hard way.
if (!Offset)
// If we're comparing GEPs with two base pointers that only differ in type
// and both GEPs have only constant indices or just one use, then fold
// the compare with the adjusted indices.
- if (DL && GEPLHS->isInBounds() && GEPRHS->isInBounds() &&
+ if (GEPLHS->isInBounds() && GEPRHS->isInBounds() &&
(GEPLHS->hasAllConstantIndices() || GEPLHS->hasOneUse()) &&
(GEPRHS->hasAllConstantIndices() || GEPRHS->hasOneUse()) &&
PtrBase->stripPointerCasts() ==
- GEPRHS->getOperand(0)->stripPointerCasts()) {
+ GEPRHS->getOperand(0)->stripPointerCasts()) {
Value *LOffset = EmitGEPOffset(GEPLHS);
Value *ROffset = EmitGEPOffset(GEPRHS);
// Only lower this if the icmp is the only user of the GEP or if we expect
// the result to fold to a constant!
- if (DL &&
- GEPsInBounds &&
- (isa<ConstantExpr>(GEPLHS) || GEPLHS->hasOneUse()) &&
+ if (GEPsInBounds && (isa<ConstantExpr>(GEPLHS) || GEPLHS->hasOneUse()) &&
(isa<ConstantExpr>(GEPRHS) || GEPRHS->hasOneUse())) {
// ((gep Ptr, OFFSET1) cmp (gep Ptr, OFFSET2) ---> (OFFSET1 cmp OFFSET2)
Value *L = EmitGEPOffset(GEPLHS);
// Turn icmp (ptrtoint x), (ptrtoint/c) into a compare of the input if the
// integer type is the same size as the pointer type.
- if (DL && LHSCI->getOpcode() == Instruction::PtrToInt &&
- DL->getPointerTypeSizeInBits(SrcTy) == DestTy->getIntegerBitWidth()) {
+ if (LHSCI->getOpcode() == Instruction::PtrToInt &&
+ DL.getPointerTypeSizeInBits(SrcTy) == DestTy->getIntegerBitWidth()) {
Value *RHSOp = nullptr;
if (PtrToIntOperator *RHSC = dyn_cast<PtrToIntOperator>(ICI.getOperand(1))) {
Value *RHSCIOp = RHSC->getOperand(0);
unsigned BitWidth = 0;
if (Ty->isIntOrIntVectorTy())
BitWidth = Ty->getScalarSizeInBits();
- else if (DL) // Pointers require DL info to get their size.
- BitWidth = DL->getTypeSizeInBits(Ty->getScalarType());
+ else // Get pointer size.
+ BitWidth = DL.getTypeSizeInBits(Ty->getScalarType());
bool isSignBit = false;
Op0KnownZero, Op0KnownOne, 0))
return &I;
if (SimplifyDemandedBits(I.getOperandUse(1),
- APInt::getAllOnesValue(BitWidth),
- Op1KnownZero, Op1KnownOne, 0))
+ APInt::getAllOnesValue(BitWidth), Op1KnownZero,
+ Op1KnownOne, 0))
return &I;
// Given the known and unknown bits, compute a range that the LHS could be
}
case Instruction::IntToPtr:
// icmp pred inttoptr(X), null -> icmp pred X, 0
- if (RHSC->isNullValue() && DL &&
- DL->getIntPtrType(RHSC->getType()) ==
- LHSI->getOperand(0)->getType())
+ if (RHSC->isNullValue() &&
+ DL.getIntPtrType(RHSC->getType()) == LHSI->getOperand(0)->getType())
return new ICmpInst(I.getPredicate(), LHSI->getOperand(0),
Constant::getNullValue(LHSI->getOperand(0)->getType()));
break;
// if A is a power of 2.
if (match(Op0, m_And(m_Value(A), m_Not(m_Value(B)))) &&
match(Op1, m_Zero()) &&
- isKnownToBeAPowerOfTwo(A, false, 0, AC, &I, DT) && I.isEquality())
+ isKnownToBeAPowerOfTwo(A, DL, false, 0, AC, &I, DT) && I.isEquality())
return new ICmpInst(I.getInversePredicate(),
Builder->CreateAnd(A, B),
Op1);
AssumptionCache *AC;
TargetLibraryInfo *TLI;
DominatorTree *DT;
+ const DataLayout &DL;
// Optional analyses. When non-null, these can both be used to do better
// combining and will be updated to reflect any changes.
- const DataLayout *DL;
LoopInfo *LI;
bool MadeIRChange;
public:
InstCombiner(InstCombineWorklist &Worklist, BuilderTy *Builder,
bool MinimizeSize, AssumptionCache *AC, TargetLibraryInfo *TLI,
- DominatorTree *DT, const DataLayout *DL, LoopInfo *LI)
+ DominatorTree *DT, const DataLayout &DL, LoopInfo *LI)
: Worklist(Worklist), Builder(Builder), MinimizeSize(MinimizeSize),
AC(AC), TLI(TLI), DT(DT), DL(DL), LI(LI), MadeIRChange(false) {}
AssumptionCache *getAssumptionCache() const { return AC; }
- const DataLayout *getDataLayout() const { return DL; }
+ const DataLayout &getDataLayout() const { return DL; }
DominatorTree *getDominatorTree() const { return DT; }
Type *Ty);
Instruction *visitCallSite(CallSite CS);
- Instruction *tryOptimizeCall(CallInst *CI, const DataLayout *DL);
+ Instruction *tryOptimizeCall(CallInst *CI);
bool transformConstExprCastCall(CallSite CS);
Instruction *transformCallThroughTrampoline(CallSite CS,
IntrinsicInst *Tramp);
Instruction *transformZExtICmp(ICmpInst *ICI, Instruction &CI,
bool DoXform = true);
Instruction *transformSExtICmp(ICmpInst *ICI, Instruction &CI);
- bool WillNotOverflowSignedAdd(Value *LHS, Value *RHS, Instruction *CxtI);
- bool WillNotOverflowSignedSub(Value *LHS, Value *RHS, Instruction *CxtI);
- bool WillNotOverflowUnsignedSub(Value *LHS, Value *RHS, Instruction *CxtI);
- bool WillNotOverflowSignedMul(Value *LHS, Value *RHS, Instruction *CxtI);
+ bool WillNotOverflowSignedAdd(Value *LHS, Value *RHS, Instruction &CxtI);
+ bool WillNotOverflowSignedSub(Value *LHS, Value *RHS, Instruction &CxtI);
+ bool WillNotOverflowUnsignedSub(Value *LHS, Value *RHS, Instruction &CxtI);
+ bool WillNotOverflowSignedMul(Value *LHS, Value *RHS, Instruction &CxtI);
Value *EmitGEPOffset(User *GEP);
Instruction *scalarizePHI(ExtractElementInst &EI, PHINode *PN);
Value *EvaluateInDifferentElementOrder(Value *V, ArrayRef<int> Mask);
}
void computeKnownBits(Value *V, APInt &KnownZero, APInt &KnownOne,
- unsigned Depth = 0, Instruction *CxtI = nullptr) const {
+ unsigned Depth, Instruction *CxtI) const {
return llvm::computeKnownBits(V, KnownZero, KnownOne, DL, Depth, AC, CxtI,
DT);
}
/// bits.
Value *SimplifyDemandedUseBits(Value *V, APInt DemandedMask, APInt &KnownZero,
APInt &KnownOne, unsigned Depth,
- Instruction *CxtI = nullptr);
+ Instruction *CxtI);
bool SimplifyDemandedBits(Use &U, APInt DemandedMask, APInt &KnownZero,
APInt &KnownOne, unsigned Depth = 0);
/// Helper routine of SimplifyDemandedUseBits. It tries to simplify demanded
Instruction *InstCombiner::visitAllocaInst(AllocaInst &AI) {
// Ensure that the alloca array size argument has type intptr_t, so that
// any casting is exposed early.
- if (DL) {
- Type *IntPtrTy = DL->getIntPtrType(AI.getType());
- if (AI.getArraySize()->getType() != IntPtrTy) {
- Value *V = Builder->CreateIntCast(AI.getArraySize(),
- IntPtrTy, false);
- AI.setOperand(0, V);
- return &AI;
- }
+ Type *IntPtrTy = DL.getIntPtrType(AI.getType());
+ if (AI.getArraySize()->getType() != IntPtrTy) {
+ Value *V = Builder->CreateIntCast(AI.getArraySize(), IntPtrTy, false);
+ AI.setOperand(0, V);
+ return &AI;
}
// Convert: alloca Ty, C - where C is a constant != 1 into: alloca [C x Ty], 1
// Now that I is pointing to the first non-allocation-inst in the block,
// insert our getelementptr instruction...
//
- Type *IdxTy = DL
- ? DL->getIntPtrType(AI.getType())
- : Type::getInt64Ty(AI.getContext());
+ Type *IdxTy = DL.getIntPtrType(AI.getType());
Value *NullIdx = Constant::getNullValue(IdxTy);
Value *Idx[2] = { NullIdx, NullIdx };
Instruction *GEP =
}
}
- if (DL && AI.getAllocatedType()->isSized()) {
+ if (AI.getAllocatedType()->isSized()) {
// If the alignment is 0 (unspecified), assign it the preferred alignment.
if (AI.getAlignment() == 0)
- AI.setAlignment(DL->getPrefTypeAlignment(AI.getAllocatedType()));
+ AI.setAlignment(DL.getPrefTypeAlignment(AI.getAllocatedType()));
// Move all alloca's of zero byte objects to the entry block and merge them
// together. Note that we only do this for alloca's, because malloc should
// allocate and return a unique pointer, even for a zero byte allocation.
- if (DL->getTypeAllocSize(AI.getAllocatedType()) == 0) {
+ if (DL.getTypeAllocSize(AI.getAllocatedType()) == 0) {
// For a zero sized alloca there is no point in doing an array allocation.
// This is helpful if the array size is a complicated expression not used
// elsewhere.
// dominance as the array size was forced to a constant earlier already.
AllocaInst *EntryAI = dyn_cast<AllocaInst>(FirstInst);
if (!EntryAI || !EntryAI->getAllocatedType()->isSized() ||
- DL->getTypeAllocSize(EntryAI->getAllocatedType()) != 0) {
+ DL.getTypeAllocSize(EntryAI->getAllocatedType()) != 0) {
AI.moveBefore(FirstInst);
return &AI;
}
// assign it the preferred alignment.
if (EntryAI->getAlignment() == 0)
EntryAI->setAlignment(
- DL->getPrefTypeAlignment(EntryAI->getAllocatedType()));
+ DL.getPrefTypeAlignment(EntryAI->getAllocatedType()));
// Replace this zero-sized alloca with the one at the start of the entry
// block after ensuring that the address will be aligned enough for both
// types.
SmallVector<Instruction *, 4> ToDelete;
if (MemTransferInst *Copy = isOnlyCopiedFromConstantGlobal(&AI, ToDelete)) {
unsigned SourceAlign = getOrEnforceKnownAlignment(
- Copy->getSource(), AI.getAlignment(), DL, AC, &AI, DT);
+ Copy->getSource(), AI.getAlignment(), DL, &AI, AC, DT);
if (AI.getAlignment() <= SourceAlign) {
DEBUG(dbgs() << "Found alloca equal to global: " << AI << '\n');
DEBUG(dbgs() << " memcpy = " << *Copy << '\n');
return nullptr;
Type *Ty = LI.getType();
+ const DataLayout &DL = IC.getDataLayout();
// Try to canonicalize loads which are only ever stored to operate over
// integers instead of any other type. We only do this when the loaded type
// is sized and has a size exactly the same as its store size and the store
// size is a legal integer type.
- const DataLayout *DL = IC.getDataLayout();
- if (!Ty->isIntegerTy() && Ty->isSized() && DL &&
- DL->isLegalInteger(DL->getTypeStoreSizeInBits(Ty)) &&
- DL->getTypeStoreSizeInBits(Ty) == DL->getTypeSizeInBits(Ty)) {
+ if (!Ty->isIntegerTy() && Ty->isSized() &&
+ DL.isLegalInteger(DL.getTypeStoreSizeInBits(Ty)) &&
+ DL.getTypeStoreSizeInBits(Ty) == DL.getTypeSizeInBits(Ty)) {
if (std::all_of(LI.user_begin(), LI.user_end(), [&LI](User *U) {
auto *SI = dyn_cast<StoreInst>(U);
return SI && SI->getPointerOperand() != &LI;
})) {
LoadInst *NewLoad = combineLoadToNewType(
IC, LI,
- Type::getIntNTy(LI.getContext(), DL->getTypeStoreSizeInBits(Ty)));
+ Type::getIntNTy(LI.getContext(), DL.getTypeStoreSizeInBits(Ty)));
// Replace all the stores with stores of the newly loaded value.
for (auto UI = LI.user_begin(), UE = LI.user_end(); UI != UE;) {
auto *SI = cast<StoreInst>(*UI++);
//
// FIXME: This should probably live in ValueTracking (or similar).
static bool isObjectSizeLessThanOrEq(Value *V, uint64_t MaxSize,
- const DataLayout *DL) {
+ const DataLayout &DL) {
SmallPtrSet<Value *, 4> Visited;
SmallVector<Value *, 4> Worklist(1, V);
if (!CS)
return false;
- uint64_t TypeSize = DL->getTypeAllocSize(AI->getAllocatedType());
+ uint64_t TypeSize = DL.getTypeAllocSize(AI->getAllocatedType());
// Make sure that, even if the multiplication below would wrap as an
// uint64_t, we still do the right thing.
if ((CS->getValue().zextOrSelf(128)*APInt(128, TypeSize)).ugt(MaxSize))
if (!GV->hasDefinitiveInitializer() || !GV->isConstant())
return false;
- uint64_t InitSize = DL->getTypeAllocSize(GV->getType()->getElementType());
+ uint64_t InitSize = DL.getTypeAllocSize(GV->getType()->getElementType());
if (InitSize > MaxSize)
return false;
continue;
// offsets those indices implied.
static bool canReplaceGEPIdxWithZero(InstCombiner &IC, GetElementPtrInst *GEPI,
Instruction *MemI, unsigned &Idx) {
- const DataLayout *DL = IC.getDataLayout();
- if (GEPI->getNumOperands() < 2 || !DL)
+ if (GEPI->getNumOperands() < 2)
return false;
// Find the first non-zero index of a GEP. If all indices are zero, return
GetElementPtrInst::getIndexedType(GEPI->getOperand(0)->getType(), Ops);
if (!AllocTy || !AllocTy->isSized())
return false;
- uint64_t TyAllocSize = DL->getTypeAllocSize(AllocTy);
+ const DataLayout &DL = IC.getDataLayout();
+ uint64_t TyAllocSize = DL.getTypeAllocSize(AllocTy);
// If there are more indices after the one we might replace with a zero, make
// sure they're all non-negative. If any of them are negative, the overall
return Res;
// Attempt to improve the alignment.
- if (DL) {
- unsigned KnownAlign = getOrEnforceKnownAlignment(
- Op, DL->getPrefTypeAlignment(LI.getType()), DL, AC, &LI, DT);
- unsigned LoadAlign = LI.getAlignment();
- unsigned EffectiveLoadAlign = LoadAlign != 0 ? LoadAlign :
- DL->getABITypeAlignment(LI.getType());
-
- if (KnownAlign > EffectiveLoadAlign)
- LI.setAlignment(KnownAlign);
- else if (LoadAlign == 0)
- LI.setAlignment(EffectiveLoadAlign);
- }
+ unsigned KnownAlign = getOrEnforceKnownAlignment(
+ Op, DL.getPrefTypeAlignment(LI.getType()), DL, &LI, AC, DT);
+ unsigned LoadAlign = LI.getAlignment();
+ unsigned EffectiveLoadAlign =
+ LoadAlign != 0 ? LoadAlign : DL.getABITypeAlignment(LI.getType());
+
+ if (KnownAlign > EffectiveLoadAlign)
+ LI.setAlignment(KnownAlign);
+ else if (LoadAlign == 0)
+ LI.setAlignment(EffectiveLoadAlign);
// Replace GEP indices if possible.
if (Instruction *NewGEPI = replaceGEPIdxWithZero(*this, Op, LI)) {
if (SelectInst *SI = dyn_cast<SelectInst>(Op)) {
// load (select (Cond, &V1, &V2)) --> select(Cond, load &V1, load &V2).
unsigned Align = LI.getAlignment();
- if (isSafeToLoadUnconditionally(SI->getOperand(1), SI, Align, DL) &&
- isSafeToLoadUnconditionally(SI->getOperand(2), SI, Align, DL)) {
+ if (isSafeToLoadUnconditionally(SI->getOperand(1), SI, Align) &&
+ isSafeToLoadUnconditionally(SI->getOperand(2), SI, Align)) {
LoadInst *V1 = Builder->CreateLoad(SI->getOperand(1),
SI->getOperand(1)->getName()+".val");
LoadInst *V2 = Builder->CreateLoad(SI->getOperand(2),
return EraseInstFromFunction(SI);
// Attempt to improve the alignment.
- if (DL) {
- unsigned KnownAlign = getOrEnforceKnownAlignment(
- Ptr, DL->getPrefTypeAlignment(Val->getType()), DL, AC, &SI, DT);
- unsigned StoreAlign = SI.getAlignment();
- unsigned EffectiveStoreAlign = StoreAlign != 0 ? StoreAlign :
- DL->getABITypeAlignment(Val->getType());
-
- if (KnownAlign > EffectiveStoreAlign)
- SI.setAlignment(KnownAlign);
- else if (StoreAlign == 0)
- SI.setAlignment(EffectiveStoreAlign);
- }
+ unsigned KnownAlign = getOrEnforceKnownAlignment(
+ Ptr, DL.getPrefTypeAlignment(Val->getType()), DL, &SI, AC, DT);
+ unsigned StoreAlign = SI.getAlignment();
+ unsigned EffectiveStoreAlign =
+ StoreAlign != 0 ? StoreAlign : DL.getABITypeAlignment(Val->getType());
+
+ if (KnownAlign > EffectiveStoreAlign)
+ SI.setAlignment(KnownAlign);
+ else if (StoreAlign == 0)
+ SI.setAlignment(EffectiveStoreAlign);
// Replace GEP indices if possible.
if (Instruction *NewGEPI = replaceGEPIdxWithZero(*this, Ptr, SI)) {
/// where it is known to be non-zero. If this allows us to simplify the
/// computation, do so and return the new operand, otherwise return null.
static Value *simplifyValueKnownNonZero(Value *V, InstCombiner &IC,
- Instruction *CxtI) {
+ Instruction &CxtI) {
// If V has multiple uses, then we would have to do more analysis to determine
// if this is safe. For example, the use could be in dynamically unreached
// code.
// inexact. Similarly for <<.
if (BinaryOperator *I = dyn_cast<BinaryOperator>(V))
if (I->isLogicalShift() &&
- isKnownToBeAPowerOfTwo(I->getOperand(0), false, 0,
- IC.getAssumptionCache(), CxtI,
+ isKnownToBeAPowerOfTwo(I->getOperand(0), IC.getDataLayout(), false, 0,
+ IC.getAssumptionCache(), &CxtI,
IC.getDominatorTree())) {
// We know that this is an exact/nuw shift and that the input is a
// non-zero context as well.
/// \brief Return true if we can prove that:
/// (mul LHS, RHS) === (mul nsw LHS, RHS)
bool InstCombiner::WillNotOverflowSignedMul(Value *LHS, Value *RHS,
- Instruction *CxtI) {
+ Instruction &CxtI) {
// Multiplying n * m significant bits yields a result of n + m significant
// bits. If the total number of significant bits does not exceed the
// result bit width (minus 1), there is no overflow.
// Note that underestimating the number of sign bits gives a more
// conservative answer.
- unsigned SignBits = ComputeNumSignBits(LHS, 0, CxtI) +
- ComputeNumSignBits(RHS, 0, CxtI);
+ unsigned SignBits =
+ ComputeNumSignBits(LHS, 0, &CxtI) + ComputeNumSignBits(RHS, 0, &CxtI);
// First handle the easy case: if we have enough sign bits there's
// definitely no overflow.
// For simplicity we just check if at least one side is not negative.
bool LHSNonNegative, LHSNegative;
bool RHSNonNegative, RHSNegative;
- ComputeSignBit(LHS, LHSNonNegative, LHSNegative, /*Depth=*/0, CxtI);
- ComputeSignBit(RHS, RHSNonNegative, RHSNegative, /*Depth=*/0, CxtI);
+ ComputeSignBit(LHS, LHSNonNegative, LHSNegative, /*Depth=*/0, &CxtI);
+ ComputeSignBit(RHS, RHSNonNegative, RHSNegative, /*Depth=*/0, &CxtI);
if (LHSNonNegative || RHSNonNegative)
return true;
}
}
}
- if (!I.hasNoSignedWrap() && WillNotOverflowSignedMul(Op0, Op1, &I)) {
+ if (!I.hasNoSignedWrap() && WillNotOverflowSignedMul(Op0, Op1, I)) {
Changed = true;
I.setHasNoSignedWrap(true);
}
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
// The RHS is known non-zero.
- if (Value *V = simplifyValueKnownNonZero(I.getOperand(1), *this, &I)) {
+ if (Value *V = simplifyValueKnownNonZero(I.getOperand(1), *this, I)) {
I.setOperand(1, V);
return &I;
}
return BO;
}
- if (isKnownToBeAPowerOfTwo(Op1, /*OrZero*/ true, 0, AC, &I, DT)) {
+ if (isKnownToBeAPowerOfTwo(Op1, DL, /*OrZero*/ true, 0, AC, &I, DT)) {
// X sdiv (1 << Y) -> X udiv (1 << Y) ( -> X u>> Y)
// Safe because the only negative value (1 << Y) can take on is
// INT_MIN, and X sdiv INT_MIN == X udiv INT_MIN == 0 if X doesn't have
Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
// The RHS is known non-zero.
- if (Value *V = simplifyValueKnownNonZero(I.getOperand(1), *this, &I)) {
+ if (Value *V = simplifyValueKnownNonZero(I.getOperand(1), *this, I)) {
I.setOperand(1, V);
return &I;
}
I.getType());
// X urem Y -> X and Y-1, where Y is a power of 2,
- if (isKnownToBeAPowerOfTwo(Op1, /*OrZero*/ true, 0, AC, &I, DT)) {
+ if (isKnownToBeAPowerOfTwo(Op1, DL, /*OrZero*/ true, 0, AC, &I, DT)) {
Constant *N1 = Constant::getAllOnesValue(I.getType());
Value *Add = Builder->CreateAdd(Op1, N1);
return BinaryOperator::CreateAnd(Op0, Add);
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/Analysis/InstructionSimplify.h"
-#include "llvm/IR/DataLayout.h"
using namespace llvm;
#define DEBUG_TYPE "instcombine"
// it is only used by trunc or trunc(lshr) operations. If so, we split the
// PHI into the various pieces being extracted. This sort of thing is
// introduced when SROA promotes an aggregate to a single large integer type.
- if (PN.getType()->isIntegerTy() && DL &&
- !DL->isLegalInteger(PN.getType()->getPrimitiveSizeInBits()))
+ if (PN.getType()->isIntegerTy() &&
+ !DL.isLegalInteger(PN.getType()->getPrimitiveSizeInBits()))
if (Instruction *Res = SliceUpIllegalIntegerPHI(PN))
return Res;
/// SimplifyWithOpReplaced - See if V simplifies when its operand Op is
/// replaced with RepOp.
static Value *SimplifyWithOpReplaced(Value *V, Value *Op, Value *RepOp,
- const DataLayout *TD,
const TargetLibraryInfo *TLI,
- DominatorTree *DT, AssumptionCache *AC) {
+ const DataLayout &DL, DominatorTree *DT,
+ AssumptionCache *AC) {
// Trivial replacement.
if (V == Op)
return RepOp;
// If this is a binary operator, try to simplify it with the replaced op.
if (BinaryOperator *B = dyn_cast<BinaryOperator>(I)) {
if (B->getOperand(0) == Op)
- return SimplifyBinOp(B->getOpcode(), RepOp, B->getOperand(1), TD, TLI);
+ return SimplifyBinOp(B->getOpcode(), RepOp, B->getOperand(1), DL, TLI);
if (B->getOperand(1) == Op)
- return SimplifyBinOp(B->getOpcode(), B->getOperand(0), RepOp, TD, TLI);
+ return SimplifyBinOp(B->getOpcode(), B->getOperand(0), RepOp, DL, TLI);
}
// Same for CmpInsts.
if (CmpInst *C = dyn_cast<CmpInst>(I)) {
if (C->getOperand(0) == Op)
- return SimplifyCmpInst(C->getPredicate(), RepOp, C->getOperand(1), TD,
+ return SimplifyCmpInst(C->getPredicate(), RepOp, C->getOperand(1), DL,
TLI, DT, AC);
if (C->getOperand(1) == Op)
- return SimplifyCmpInst(C->getPredicate(), C->getOperand(0), RepOp, TD,
+ return SimplifyCmpInst(C->getPredicate(), C->getOperand(0), RepOp, DL,
TLI, DT, AC);
}
if (ConstOps.size() == I->getNumOperands()) {
if (CmpInst *C = dyn_cast<CmpInst>(I))
return ConstantFoldCompareInstOperands(C->getPredicate(), ConstOps[0],
- ConstOps[1], TD, TLI);
+ ConstOps[1], DL, TLI);
if (LoadInst *LI = dyn_cast<LoadInst>(I))
if (!LI->isVolatile())
- return ConstantFoldLoadFromConstPtr(ConstOps[0], TD);
+ return ConstantFoldLoadFromConstPtr(ConstOps[0], DL);
- return ConstantFoldInstOperands(I->getOpcode(), I->getType(),
- ConstOps, TD, TLI);
+ return ConstantFoldInstOperands(I->getOpcode(), I->getType(), ConstOps,
+ DL, TLI);
}
}
// arms of the select. See if substituting this value into the arm and
// simplifying the result yields the same value as the other arm.
if (Pred == ICmpInst::ICMP_EQ) {
- if (SimplifyWithOpReplaced(FalseVal, CmpLHS, CmpRHS, DL, TLI, DT, AC) ==
+ if (SimplifyWithOpReplaced(FalseVal, CmpLHS, CmpRHS, TLI, DL, DT, AC) ==
TrueVal ||
- SimplifyWithOpReplaced(FalseVal, CmpRHS, CmpLHS, DL, TLI, DT, AC) ==
+ SimplifyWithOpReplaced(FalseVal, CmpRHS, CmpLHS, TLI, DL, DT, AC) ==
TrueVal)
return ReplaceInstUsesWith(SI, FalseVal);
- if (SimplifyWithOpReplaced(TrueVal, CmpLHS, CmpRHS, DL, TLI, DT, AC) ==
+ if (SimplifyWithOpReplaced(TrueVal, CmpLHS, CmpRHS, TLI, DL, DT, AC) ==
FalseVal ||
- SimplifyWithOpReplaced(TrueVal, CmpRHS, CmpLHS, DL, TLI, DT, AC) ==
+ SimplifyWithOpReplaced(TrueVal, CmpRHS, CmpLHS, TLI, DL, DT, AC) ==
FalseVal)
return ReplaceInstUsesWith(SI, FalseVal);
} else if (Pred == ICmpInst::ICMP_NE) {
- if (SimplifyWithOpReplaced(TrueVal, CmpLHS, CmpRHS, DL, TLI, DT, AC) ==
+ if (SimplifyWithOpReplaced(TrueVal, CmpLHS, CmpRHS, TLI, DL, DT, AC) ==
FalseVal ||
- SimplifyWithOpReplaced(TrueVal, CmpRHS, CmpLHS, DL, TLI, DT, AC) ==
+ SimplifyWithOpReplaced(TrueVal, CmpRHS, CmpLHS, TLI, DL, DT, AC) ==
FalseVal)
return ReplaceInstUsesWith(SI, TrueVal);
- if (SimplifyWithOpReplaced(FalseVal, CmpLHS, CmpRHS, DL, TLI, DT, AC) ==
+ if (SimplifyWithOpReplaced(FalseVal, CmpLHS, CmpRHS, TLI, DL, DT, AC) ==
TrueVal ||
- SimplifyWithOpReplaced(FalseVal, CmpRHS, CmpLHS, DL, TLI, DT, AC) ==
+ SimplifyWithOpReplaced(FalseVal, CmpRHS, CmpLHS, TLI, DL, DT, AC) ==
TrueVal)
return ReplaceInstUsesWith(SI, TrueVal);
}
/// GetShiftedValue - When CanEvaluateShifted returned true for an expression,
/// this value inserts the new computation that produces the shifted value.
static Value *GetShiftedValue(Value *V, unsigned NumBits, bool isLeftShift,
- InstCombiner &IC) {
+ InstCombiner &IC, const DataLayout &DL) {
// We can always evaluate constants shifted.
if (Constant *C = dyn_cast<Constant>(V)) {
if (isLeftShift)
V = IC.Builder->CreateLShr(C, NumBits);
// If we got a constantexpr back, try to simplify it with TD info.
if (ConstantExpr *CE = dyn_cast<ConstantExpr>(V))
- V = ConstantFoldConstantExpression(CE, IC.getDataLayout(),
- IC.getTargetLibraryInfo());
+ V = ConstantFoldConstantExpression(CE, DL, IC.getTargetLibraryInfo());
return V;
}
case Instruction::Or:
case Instruction::Xor:
// Bitwise operators can all arbitrarily be arbitrarily evaluated shifted.
- I->setOperand(0, GetShiftedValue(I->getOperand(0), NumBits,isLeftShift,IC));
- I->setOperand(1, GetShiftedValue(I->getOperand(1), NumBits,isLeftShift,IC));
+ I->setOperand(
+ 0, GetShiftedValue(I->getOperand(0), NumBits, isLeftShift, IC, DL));
+ I->setOperand(
+ 1, GetShiftedValue(I->getOperand(1), NumBits, isLeftShift, IC, DL));
return I;
case Instruction::Shl: {
}
case Instruction::Select:
- I->setOperand(1, GetShiftedValue(I->getOperand(1), NumBits,isLeftShift,IC));
- I->setOperand(2, GetShiftedValue(I->getOperand(2), NumBits,isLeftShift,IC));
+ I->setOperand(
+ 1, GetShiftedValue(I->getOperand(1), NumBits, isLeftShift, IC, DL));
+ I->setOperand(
+ 2, GetShiftedValue(I->getOperand(2), NumBits, isLeftShift, IC, DL));
return I;
case Instruction::PHI: {
// We can change a phi if we can change all operands. Note that we never
// instructions with a single use.
PHINode *PN = cast<PHINode>(I);
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
- PN->setIncomingValue(i, GetShiftedValue(PN->getIncomingValue(i),
- NumBits, isLeftShift, IC));
+ PN->setIncomingValue(i, GetShiftedValue(PN->getIncomingValue(i), NumBits,
+ isLeftShift, IC, DL));
return PN;
}
}
DEBUG(dbgs() << "ICE: GetShiftedValue propagating shift through expression"
" to eliminate shift:\n IN: " << *Op0 << "\n SH: " << I <<"\n");
- return ReplaceInstUsesWith(I,
- GetShiftedValue(Op0, COp1->getZExtValue(), isLeftShift, *this));
+ return ReplaceInstUsesWith(
+ I, GetShiftedValue(Op0, COp1->getZExtValue(), isLeftShift, *this, DL));
}
// See if we can simplify any instructions used by the instruction whose sole
//===----------------------------------------------------------------------===//
#include "InstCombineInternal.h"
-#include "llvm/IR/DataLayout.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/PatternMatch.h"
APInt KnownZero(BitWidth, 0), KnownOne(BitWidth, 0);
APInt DemandedMask(APInt::getAllOnesValue(BitWidth));
- Value *V = SimplifyDemandedUseBits(&Inst, DemandedMask,
- KnownZero, KnownOne, 0, &Inst);
+ Value *V = SimplifyDemandedUseBits(&Inst, DemandedMask, KnownZero, KnownOne,
+ 0, &Inst);
if (!V) return false;
if (V == &Inst) return true;
ReplaceInstUsesWith(Inst, V);
bool InstCombiner::SimplifyDemandedBits(Use &U, APInt DemandedMask,
APInt &KnownZero, APInt &KnownOne,
unsigned Depth) {
- Value *NewVal = SimplifyDemandedUseBits(U.get(), DemandedMask,
- KnownZero, KnownOne, Depth,
- dyn_cast<Instruction>(U.getUser()));
+ Value *NewVal =
+ SimplifyDemandedUseBits(U.get(), DemandedMask, KnownZero, KnownOne, Depth,
+ dyn_cast<Instruction>(U.getUser()));
if (!NewVal) return false;
U = NewVal;
return true;
assert(Depth <= 6 && "Limit Search Depth");
uint32_t BitWidth = DemandedMask.getBitWidth();
Type *VTy = V->getType();
- assert((DL || !VTy->isPointerTy()) &&
- "SimplifyDemandedBits needs to know bit widths!");
- assert((!DL || DL->getTypeSizeInBits(VTy->getScalarType()) == BitWidth) &&
- (!VTy->isIntOrIntVectorTy() ||
- VTy->getScalarSizeInBits() == BitWidth) &&
- KnownZero.getBitWidth() == BitWidth &&
- KnownOne.getBitWidth() == BitWidth &&
- "Value *V, DemandedMask, KnownZero and KnownOne "
- "must have same BitWidth");
+ assert(
+ (!VTy->isIntOrIntVectorTy() || VTy->getScalarSizeInBits() == BitWidth) &&
+ KnownZero.getBitWidth() == BitWidth &&
+ KnownOne.getBitWidth() == BitWidth &&
+ "Value *V, DemandedMask, KnownZero and KnownOne "
+ "must have same BitWidth");
if (ConstantInt *CI = dyn_cast<ConstantInt>(V)) {
// We know all of the bits for a constant!
KnownOne = CI->getValue() & DemandedMask;
// this instruction has a simpler value in that context.
if (I->getOpcode() == Instruction::And) {
// If either the LHS or the RHS are Zero, the result is zero.
- computeKnownBits(I->getOperand(1), RHSKnownZero, RHSKnownOne, Depth+1,
+ computeKnownBits(I->getOperand(1), RHSKnownZero, RHSKnownOne, Depth + 1,
CxtI);
- computeKnownBits(I->getOperand(0), LHSKnownZero, LHSKnownOne, Depth+1,
+ computeKnownBits(I->getOperand(0), LHSKnownZero, LHSKnownOne, Depth + 1,
CxtI);
// If all of the demanded bits are known 1 on one side, return the other.
// only bits from X or Y are demanded.
// If either the LHS or the RHS are One, the result is One.
- computeKnownBits(I->getOperand(1), RHSKnownZero, RHSKnownOne, Depth+1,
+ computeKnownBits(I->getOperand(1), RHSKnownZero, RHSKnownOne, Depth + 1,
CxtI);
- computeKnownBits(I->getOperand(0), LHSKnownZero, LHSKnownOne, Depth+1,
+ computeKnownBits(I->getOperand(0), LHSKnownZero, LHSKnownOne, Depth + 1,
CxtI);
// If all of the demanded bits are known zero on one side, return the
// We can simplify (X^Y) -> X or Y in the user's context if we know that
// only bits from X or Y are demanded.
- computeKnownBits(I->getOperand(1), RHSKnownZero, RHSKnownOne, Depth+1,
+ computeKnownBits(I->getOperand(1), RHSKnownZero, RHSKnownOne, Depth + 1,
CxtI);
- computeKnownBits(I->getOperand(0), LHSKnownZero, LHSKnownOne, Depth+1,
+ computeKnownBits(I->getOperand(0), LHSKnownZero, LHSKnownOne, Depth + 1,
CxtI);
// If all of the demanded bits are known zero on one side, return the
break;
case Instruction::And:
// If either the LHS or the RHS are Zero, the result is zero.
- if (SimplifyDemandedBits(I->getOperandUse(1), DemandedMask,
- RHSKnownZero, RHSKnownOne, Depth+1) ||
+ if (SimplifyDemandedBits(I->getOperandUse(1), DemandedMask, RHSKnownZero,
+ RHSKnownOne, Depth + 1) ||
SimplifyDemandedBits(I->getOperandUse(0), DemandedMask & ~RHSKnownZero,
- LHSKnownZero, LHSKnownOne, Depth+1))
+ LHSKnownZero, LHSKnownOne, Depth + 1))
return I;
assert(!(RHSKnownZero & RHSKnownOne) && "Bits known to be one AND zero?");
assert(!(LHSKnownZero & LHSKnownOne) && "Bits known to be one AND zero?");
break;
case Instruction::Or:
// If either the LHS or the RHS are One, the result is One.
- if (SimplifyDemandedBits(I->getOperandUse(1), DemandedMask,
- RHSKnownZero, RHSKnownOne, Depth+1) ||
+ if (SimplifyDemandedBits(I->getOperandUse(1), DemandedMask, RHSKnownZero,
+ RHSKnownOne, Depth + 1) ||
SimplifyDemandedBits(I->getOperandUse(0), DemandedMask & ~RHSKnownOne,
- LHSKnownZero, LHSKnownOne, Depth+1))
+ LHSKnownZero, LHSKnownOne, Depth + 1))
return I;
assert(!(RHSKnownZero & RHSKnownOne) && "Bits known to be one AND zero?");
assert(!(LHSKnownZero & LHSKnownOne) && "Bits known to be one AND zero?");
KnownOne = RHSKnownOne | LHSKnownOne;
break;
case Instruction::Xor: {
- if (SimplifyDemandedBits(I->getOperandUse(1), DemandedMask,
- RHSKnownZero, RHSKnownOne, Depth+1) ||
- SimplifyDemandedBits(I->getOperandUse(0), DemandedMask,
- LHSKnownZero, LHSKnownOne, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(1), DemandedMask, RHSKnownZero,
+ RHSKnownOne, Depth + 1) ||
+ SimplifyDemandedBits(I->getOperandUse(0), DemandedMask, LHSKnownZero,
+ LHSKnownOne, Depth + 1))
return I;
assert(!(RHSKnownZero & RHSKnownOne) && "Bits known to be one AND zero?");
assert(!(LHSKnownZero & LHSKnownOne) && "Bits known to be one AND zero?");
break;
}
case Instruction::Select:
- if (SimplifyDemandedBits(I->getOperandUse(2), DemandedMask,
- RHSKnownZero, RHSKnownOne, Depth+1) ||
- SimplifyDemandedBits(I->getOperandUse(1), DemandedMask,
- LHSKnownZero, LHSKnownOne, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(2), DemandedMask, RHSKnownZero,
+ RHSKnownOne, Depth + 1) ||
+ SimplifyDemandedBits(I->getOperandUse(1), DemandedMask, LHSKnownZero,
+ LHSKnownOne, Depth + 1))
return I;
assert(!(RHSKnownZero & RHSKnownOne) && "Bits known to be one AND zero?");
assert(!(LHSKnownZero & LHSKnownOne) && "Bits known to be one AND zero?");
DemandedMask = DemandedMask.zext(truncBf);
KnownZero = KnownZero.zext(truncBf);
KnownOne = KnownOne.zext(truncBf);
- if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMask,
- KnownZero, KnownOne, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMask, KnownZero,
+ KnownOne, Depth + 1))
return I;
DemandedMask = DemandedMask.trunc(BitWidth);
KnownZero = KnownZero.trunc(BitWidth);
// Don't touch a vector-to-scalar bitcast.
return nullptr;
- if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMask,
- KnownZero, KnownOne, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMask, KnownZero,
+ KnownOne, Depth + 1))
return I;
assert(!(KnownZero & KnownOne) && "Bits known to be one AND zero?");
break;
DemandedMask = DemandedMask.trunc(SrcBitWidth);
KnownZero = KnownZero.trunc(SrcBitWidth);
KnownOne = KnownOne.trunc(SrcBitWidth);
- if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMask,
- KnownZero, KnownOne, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMask, KnownZero,
+ KnownOne, Depth + 1))
return I;
DemandedMask = DemandedMask.zext(BitWidth);
KnownZero = KnownZero.zext(BitWidth);
InputDemandedBits = InputDemandedBits.trunc(SrcBitWidth);
KnownZero = KnownZero.trunc(SrcBitWidth);
KnownOne = KnownOne.trunc(SrcBitWidth);
- if (SimplifyDemandedBits(I->getOperandUse(0), InputDemandedBits,
- KnownZero, KnownOne, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(0), InputDemandedBits, KnownZero,
+ KnownOne, Depth + 1))
return I;
InputDemandedBits = InputDemandedBits.zext(BitWidth);
KnownZero = KnownZero.zext(BitWidth);
// Find information about known zero/one bits in the input.
if (SimplifyDemandedBits(I->getOperandUse(0), InDemandedBits,
- LHSKnownZero, LHSKnownOne, Depth+1))
+ LHSKnownZero, LHSKnownOne, Depth + 1))
return I;
// If the RHS of the add has bits set that can't affect the input, reduce
// significant bit and all those below it.
APInt DemandedFromOps(APInt::getLowBitsSet(BitWidth, BitWidth-NLZ));
if (SimplifyDemandedBits(I->getOperandUse(0), DemandedFromOps,
- LHSKnownZero, LHSKnownOne, Depth+1) ||
+ LHSKnownZero, LHSKnownOne, Depth + 1) ||
SimplifyDemandedBits(I->getOperandUse(1), DemandedFromOps,
- LHSKnownZero, LHSKnownOne, Depth+1))
+ LHSKnownZero, LHSKnownOne, Depth + 1))
return I;
}
}
uint32_t NLZ = DemandedMask.countLeadingZeros();
APInt DemandedFromOps(APInt::getLowBitsSet(BitWidth, BitWidth-NLZ));
if (SimplifyDemandedBits(I->getOperandUse(0), DemandedFromOps,
- LHSKnownZero, LHSKnownOne, Depth+1) ||
+ LHSKnownZero, LHSKnownOne, Depth + 1) ||
SimplifyDemandedBits(I->getOperandUse(1), DemandedFromOps,
- LHSKnownZero, LHSKnownOne, Depth+1))
+ LHSKnownZero, LHSKnownOne, Depth + 1))
return I;
}
else if (IOp->hasNoUnsignedWrap())
DemandedMaskIn |= APInt::getHighBitsSet(BitWidth, ShiftAmt);
- if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMaskIn,
- KnownZero, KnownOne, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMaskIn, KnownZero,
+ KnownOne, Depth + 1))
return I;
assert(!(KnownZero & KnownOne) && "Bits known to be one AND zero?");
KnownZero <<= ShiftAmt;
if (cast<LShrOperator>(I)->isExact())
DemandedMaskIn |= APInt::getLowBitsSet(BitWidth, ShiftAmt);
- if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMaskIn,
- KnownZero, KnownOne, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMaskIn, KnownZero,
+ KnownOne, Depth + 1))
return I;
assert(!(KnownZero & KnownOne) && "Bits known to be one AND zero?");
KnownZero = APIntOps::lshr(KnownZero, ShiftAmt);
if (cast<AShrOperator>(I)->isExact())
DemandedMaskIn |= APInt::getLowBitsSet(BitWidth, ShiftAmt);
- if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMaskIn,
- KnownZero, KnownOne, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(0), DemandedMaskIn, KnownZero,
+ KnownOne, Depth + 1))
return I;
assert(!(KnownZero & KnownOne) && "Bits known to be one AND zero?");
// Compute the new bits that are at the top now.
APInt LowBits = RA - 1;
APInt Mask2 = LowBits | APInt::getSignBit(BitWidth);
- if (SimplifyDemandedBits(I->getOperandUse(0), Mask2,
- LHSKnownZero, LHSKnownOne, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(0), Mask2, LHSKnownZero,
+ LHSKnownOne, Depth + 1))
return I;
// The low bits of LHS are unchanged by the srem.
// remainder is zero.
if (DemandedMask.isNegative() && KnownZero.isNonNegative()) {
APInt LHSKnownZero(BitWidth, 0), LHSKnownOne(BitWidth, 0);
- computeKnownBits(I->getOperand(0), LHSKnownZero, LHSKnownOne, Depth+1,
+ computeKnownBits(I->getOperand(0), LHSKnownZero, LHSKnownOne, Depth + 1,
CxtI);
// If it's known zero, our sign bit is also zero.
if (LHSKnownZero.isNegative())
case Instruction::URem: {
APInt KnownZero2(BitWidth, 0), KnownOne2(BitWidth, 0);
APInt AllOnes = APInt::getAllOnesValue(BitWidth);
- if (SimplifyDemandedBits(I->getOperandUse(0), AllOnes,
- KnownZero2, KnownOne2, Depth+1) ||
- SimplifyDemandedBits(I->getOperandUse(1), AllOnes,
- KnownZero2, KnownOne2, Depth+1))
+ if (SimplifyDemandedBits(I->getOperandUse(0), AllOnes, KnownZero2,
+ KnownOne2, Depth + 1) ||
+ SimplifyDemandedBits(I->getOperandUse(1), AllOnes, KnownZero2,
+ KnownOne2, Depth + 1))
return I;
unsigned Leaders = KnownZero2.countLeadingOnes();
// Note that we can't propagate undef elt info, because we don't know
// which elt is getting updated.
TmpV = SimplifyDemandedVectorElts(I->getOperand(0), DemandedElts,
- UndefElts2, Depth+1);
+ UndefElts2, Depth + 1);
if (TmpV) { I->setOperand(0, TmpV); MadeChange = true; }
break;
}
APInt DemandedElts2 = DemandedElts;
DemandedElts2.clearBit(IdxNo);
TmpV = SimplifyDemandedVectorElts(I->getOperand(0), DemandedElts2,
- UndefElts, Depth+1);
+ UndefElts, Depth + 1);
if (TmpV) { I->setOperand(0, TmpV); MadeChange = true; }
// The inserted element is defined.
APInt UndefElts4(LHSVWidth, 0);
TmpV = SimplifyDemandedVectorElts(I->getOperand(0), LeftDemanded,
- UndefElts4, Depth+1);
+ UndefElts4, Depth + 1);
if (TmpV) { I->setOperand(0, TmpV); MadeChange = true; }
APInt UndefElts3(LHSVWidth, 0);
TmpV = SimplifyDemandedVectorElts(I->getOperand(1), RightDemanded,
- UndefElts3, Depth+1);
+ UndefElts3, Depth + 1);
if (TmpV) { I->setOperand(1, TmpV); MadeChange = true; }
bool NewUndefElts = false;
}
}
- TmpV = SimplifyDemandedVectorElts(I->getOperand(1), LeftDemanded,
- UndefElts, Depth+1);
+ TmpV = SimplifyDemandedVectorElts(I->getOperand(1), LeftDemanded, UndefElts,
+ Depth + 1);
if (TmpV) { I->setOperand(1, TmpV); MadeChange = true; }
TmpV = SimplifyDemandedVectorElts(I->getOperand(2), RightDemanded,
- UndefElts2, Depth+1);
+ UndefElts2, Depth + 1);
if (TmpV) { I->setOperand(2, TmpV); MadeChange = true; }
// Output elements are undefined if both are undefined.
// div/rem demand all inputs, because they don't want divide by zero.
TmpV = SimplifyDemandedVectorElts(I->getOperand(0), InputDemandedElts,
- UndefElts2, Depth+1);
+ UndefElts2, Depth + 1);
if (TmpV) {
I->setOperand(0, TmpV);
MadeChange = true;
case Instruction::Sub:
case Instruction::Mul:
// div/rem demand all inputs, because they don't want divide by zero.
- TmpV = SimplifyDemandedVectorElts(I->getOperand(0), DemandedElts,
- UndefElts, Depth+1);
+ TmpV = SimplifyDemandedVectorElts(I->getOperand(0), DemandedElts, UndefElts,
+ Depth + 1);
if (TmpV) { I->setOperand(0, TmpV); MadeChange = true; }
TmpV = SimplifyDemandedVectorElts(I->getOperand(1), DemandedElts,
- UndefElts2, Depth+1);
+ UndefElts2, Depth + 1);
if (TmpV) { I->setOperand(1, TmpV); MadeChange = true; }
// Output elements are undefined if both are undefined. Consider things
break;
case Instruction::FPTrunc:
case Instruction::FPExt:
- TmpV = SimplifyDemandedVectorElts(I->getOperand(0), DemandedElts,
- UndefElts, Depth+1);
+ TmpV = SimplifyDemandedVectorElts(I->getOperand(0), DemandedElts, UndefElts,
+ Depth + 1);
if (TmpV) { I->setOperand(0, TmpV); MadeChange = true; }
break;
case Intrinsic::x86_sse2_min_sd:
case Intrinsic::x86_sse2_max_sd:
TmpV = SimplifyDemandedVectorElts(II->getArgOperand(0), DemandedElts,
- UndefElts, Depth+1);
+ UndefElts, Depth + 1);
if (TmpV) { II->setArgOperand(0, TmpV); MadeChange = true; }
TmpV = SimplifyDemandedVectorElts(II->getArgOperand(1), DemandedElts,
- UndefElts2, Depth+1);
+ UndefElts2, Depth + 1);
if (TmpV) { II->setArgOperand(1, TmpV); MadeChange = true; }
// If only the low elt is demanded and this is a scalarizable intrinsic,
APInt UndefElts(VectorWidth, 0);
APInt DemandedMask(VectorWidth, 0);
DemandedMask.setBit(IndexVal);
- if (Value *V = SimplifyDemandedVectorElts(EI.getOperand(0),
- DemandedMask, UndefElts)) {
+ if (Value *V = SimplifyDemandedVectorElts(EI.getOperand(0), DemandedMask,
+ UndefElts)) {
EI.setOperand(0, V);
return &EI;
}
STATISTIC(NumReassoc , "Number of reassociations");
Value *InstCombiner::EmitGEPOffset(User *GEP) {
- return llvm::EmitGEPOffset(Builder, *getDataLayout(), GEP);
+ return llvm::EmitGEPOffset(Builder, DL, GEP);
}
/// ShouldChangeType - Return true if it is desirable to convert a computation
bool InstCombiner::ShouldChangeType(Type *From, Type *To) const {
assert(From->isIntegerTy() && To->isIntegerTy());
- // If we don't have DL, we don't know if the source/dest are legal.
- if (!DL) return false;
-
unsigned FromWidth = From->getPrimitiveSizeInBits();
unsigned ToWidth = To->getPrimitiveSizeInBits();
- bool FromLegal = DL->isLegalInteger(FromWidth);
- bool ToLegal = DL->isLegalInteger(ToWidth);
+ bool FromLegal = DL.isLegalInteger(FromWidth);
+ bool ToLegal = DL.isLegalInteger(ToWidth);
// If this is a legal integer from type, and the result would be an illegal
// type, don't do the transformation.
/// This tries to simplify binary operations by factorizing out common terms
/// (e. g. "(A*B)+(A*C)" -> "A*(B+C)").
static Value *tryFactorization(InstCombiner::BuilderTy *Builder,
- const DataLayout *DL, BinaryOperator &I,
+ const DataLayout &DL, BinaryOperator &I,
Instruction::BinaryOps InnerOpcode, Value *A,
Value *B, Value *C, Value *D) {
/// will land us at the specified offset. If so, fill them into NewIndices and
/// return the resultant element type, otherwise return null.
Type *InstCombiner::FindElementAtOffset(Type *PtrTy, int64_t Offset,
- SmallVectorImpl<Value*> &NewIndices) {
+ SmallVectorImpl<Value *> &NewIndices) {
assert(PtrTy->isPtrOrPtrVectorTy());
- if (!DL)
- return nullptr;
-
Type *Ty = PtrTy->getPointerElementType();
if (!Ty->isSized())
return nullptr;
// Start with the index over the outer type. Note that the type size
// might be zero (even if the offset isn't zero) if the indexed type
// is something like [0 x {int, int}]
- Type *IntPtrTy = DL->getIntPtrType(PtrTy);
+ Type *IntPtrTy = DL.getIntPtrType(PtrTy);
int64_t FirstIdx = 0;
- if (int64_t TySize = DL->getTypeAllocSize(Ty)) {
+ if (int64_t TySize = DL.getTypeAllocSize(Ty)) {
FirstIdx = Offset/TySize;
Offset -= FirstIdx*TySize;
// Index into the types. If we fail, set OrigBase to null.
while (Offset) {
// Indexing into tail padding between struct/array elements.
- if (uint64_t(Offset*8) >= DL->getTypeSizeInBits(Ty))
+ if (uint64_t(Offset * 8) >= DL.getTypeSizeInBits(Ty))
return nullptr;
if (StructType *STy = dyn_cast<StructType>(Ty)) {
- const StructLayout *SL = DL->getStructLayout(STy);
+ const StructLayout *SL = DL.getStructLayout(STy);
assert(Offset < (int64_t)SL->getSizeInBytes() &&
"Offset must stay within the indexed type");
Offset -= SL->getElementOffset(Elt);
Ty = STy->getElementType(Elt);
} else if (ArrayType *AT = dyn_cast<ArrayType>(Ty)) {
- uint64_t EltSize = DL->getTypeAllocSize(AT->getElementType());
+ uint64_t EltSize = DL.getTypeAllocSize(AT->getElementType());
assert(EltSize && "Cannot index into a zero-sized array");
NewIndices.push_back(ConstantInt::get(IntPtrTy,Offset/EltSize));
Offset %= EltSize;
// It may not be safe to reorder shuffles and things like div, urem, etc.
// because we may trap when executing those ops on unknown vector elements.
// See PR20059.
- if (!isSafeToSpeculativelyExecute(&Inst, DL)) return nullptr;
+ if (!isSafeToSpeculativelyExecute(&Inst))
+ return nullptr;
unsigned VWidth = cast<VectorType>(Inst.getType())->getNumElements();
Value *LHS = Inst.getOperand(0), *RHS = Inst.getOperand(1);
// Eliminate unneeded casts for indices, and replace indices which displace
// by multiples of a zero size type with zero.
- if (DL) {
- bool MadeChange = false;
- Type *IntPtrTy = DL->getIntPtrType(GEP.getPointerOperandType());
-
- gep_type_iterator GTI = gep_type_begin(GEP);
- for (User::op_iterator I = GEP.op_begin() + 1, E = GEP.op_end();
- I != E; ++I, ++GTI) {
- // Skip indices into struct types.
- SequentialType *SeqTy = dyn_cast<SequentialType>(*GTI);
- if (!SeqTy) continue;
-
- // If the element type has zero size then any index over it is equivalent
- // to an index of zero, so replace it with zero if it is not zero already.
- if (SeqTy->getElementType()->isSized() &&
- DL->getTypeAllocSize(SeqTy->getElementType()) == 0)
- if (!isa<Constant>(*I) || !cast<Constant>(*I)->isNullValue()) {
- *I = Constant::getNullValue(IntPtrTy);
- MadeChange = true;
- }
+ bool MadeChange = false;
+ Type *IntPtrTy = DL.getIntPtrType(GEP.getPointerOperandType());
+
+ gep_type_iterator GTI = gep_type_begin(GEP);
+ for (User::op_iterator I = GEP.op_begin() + 1, E = GEP.op_end(); I != E;
+ ++I, ++GTI) {
+ // Skip indices into struct types.
+ SequentialType *SeqTy = dyn_cast<SequentialType>(*GTI);
+ if (!SeqTy)
+ continue;
- Type *IndexTy = (*I)->getType();
- if (IndexTy != IntPtrTy) {
- // If we are using a wider index than needed for this platform, shrink
- // it to what we need. If narrower, sign-extend it to what we need.
- // This explicit cast can make subsequent optimizations more obvious.
- *I = Builder->CreateIntCast(*I, IntPtrTy, true);
+ // If the element type has zero size then any index over it is equivalent
+ // to an index of zero, so replace it with zero if it is not zero already.
+ if (SeqTy->getElementType()->isSized() &&
+ DL.getTypeAllocSize(SeqTy->getElementType()) == 0)
+ if (!isa<Constant>(*I) || !cast<Constant>(*I)->isNullValue()) {
+ *I = Constant::getNullValue(IntPtrTy);
MadeChange = true;
}
+
+ Type *IndexTy = (*I)->getType();
+ if (IndexTy != IntPtrTy) {
+ // If we are using a wider index than needed for this platform, shrink
+ // it to what we need. If narrower, sign-extend it to what we need.
+ // This explicit cast can make subsequent optimizations more obvious.
+ *I = Builder->CreateIntCast(*I, IntPtrTy, true);
+ MadeChange = true;
}
- if (MadeChange) return &GEP;
}
+ if (MadeChange)
+ return &GEP;
// Check to see if the inputs to the PHI node are getelementptr instructions.
if (PHINode *PN = dyn_cast<PHINode>(PtrOp)) {
GetElementPtrInst::Create(Src->getOperand(0), Indices, GEP.getName());
}
- if (DL && GEP.getNumIndices() == 1) {
+ if (GEP.getNumIndices() == 1) {
unsigned AS = GEP.getPointerAddressSpace();
if (GEP.getOperand(1)->getType()->getScalarSizeInBits() ==
- DL->getPointerSizeInBits(AS)) {
+ DL.getPointerSizeInBits(AS)) {
Type *PtrTy = GEP.getPointerOperandType();
Type *Ty = PtrTy->getPointerElementType();
- uint64_t TyAllocSize = DL->getTypeAllocSize(Ty);
+ uint64_t TyAllocSize = DL.getTypeAllocSize(Ty);
bool Matched = false;
uint64_t C;
// into: %t1 = getelementptr [2 x i32]* %str, i32 0, i32 %V; bitcast
Type *SrcElTy = StrippedPtrTy->getElementType();
Type *ResElTy = PtrOp->getType()->getPointerElementType();
- if (DL && SrcElTy->isArrayTy() &&
- DL->getTypeAllocSize(SrcElTy->getArrayElementType()) ==
- DL->getTypeAllocSize(ResElTy)) {
- Type *IdxType = DL->getIntPtrType(GEP.getType());
+ if (SrcElTy->isArrayTy() &&
+ DL.getTypeAllocSize(SrcElTy->getArrayElementType()) ==
+ DL.getTypeAllocSize(ResElTy)) {
+ Type *IdxType = DL.getIntPtrType(GEP.getType());
Value *Idx[2] = { Constant::getNullValue(IdxType), GEP.getOperand(1) };
Value *NewGEP = GEP.isInBounds() ?
Builder->CreateInBoundsGEP(StrippedPtr, Idx, GEP.getName()) :
// %V = mul i64 %N, 4
// %t = getelementptr i8* bitcast (i32* %arr to i8*), i32 %V
// into: %t1 = getelementptr i32* %arr, i32 %N; bitcast
- if (DL && ResElTy->isSized() && SrcElTy->isSized()) {
+ if (ResElTy->isSized() && SrcElTy->isSized()) {
// Check that changing the type amounts to dividing the index by a scale
// factor.
- uint64_t ResSize = DL->getTypeAllocSize(ResElTy);
- uint64_t SrcSize = DL->getTypeAllocSize(SrcElTy);
+ uint64_t ResSize = DL.getTypeAllocSize(ResElTy);
+ uint64_t SrcSize = DL.getTypeAllocSize(SrcElTy);
if (ResSize && SrcSize % ResSize == 0) {
Value *Idx = GEP.getOperand(1);
unsigned BitWidth = Idx->getType()->getPrimitiveSizeInBits();
// Earlier transforms ensure that the index has type IntPtrType, which
// considerably simplifies the logic by eliminating implicit casts.
- assert(Idx->getType() == DL->getIntPtrType(GEP.getType()) &&
+ assert(Idx->getType() == DL.getIntPtrType(GEP.getType()) &&
"Index not cast to pointer width?");
bool NSW;
// getelementptr i8* bitcast ([100 x double]* X to i8*), i32 %tmp
// (where tmp = 8*tmp2) into:
// getelementptr [100 x double]* %arr, i32 0, i32 %tmp2; bitcast
- if (DL && ResElTy->isSized() && SrcElTy->isSized() &&
- SrcElTy->isArrayTy()) {
+ if (ResElTy->isSized() && SrcElTy->isSized() && SrcElTy->isArrayTy()) {
// Check that changing to the array element type amounts to dividing the
// index by a scale factor.
- uint64_t ResSize = DL->getTypeAllocSize(ResElTy);
- uint64_t ArrayEltSize
- = DL->getTypeAllocSize(SrcElTy->getArrayElementType());
+ uint64_t ResSize = DL.getTypeAllocSize(ResElTy);
+ uint64_t ArrayEltSize =
+ DL.getTypeAllocSize(SrcElTy->getArrayElementType());
if (ResSize && ArrayEltSize % ResSize == 0) {
Value *Idx = GEP.getOperand(1);
unsigned BitWidth = Idx->getType()->getPrimitiveSizeInBits();
// Earlier transforms ensure that the index has type IntPtrType, which
// considerably simplifies the logic by eliminating implicit casts.
- assert(Idx->getType() == DL->getIntPtrType(GEP.getType()) &&
+ assert(Idx->getType() == DL.getIntPtrType(GEP.getType()) &&
"Index not cast to pointer width?");
bool NSW;
// If the multiplication NewIdx * Scale may overflow then the new
// GEP may not be "inbounds".
Value *Off[2] = {
- Constant::getNullValue(DL->getIntPtrType(GEP.getType())),
- NewIdx
- };
+ Constant::getNullValue(DL.getIntPtrType(GEP.getType())),
+ NewIdx};
Value *NewGEP = GEP.isInBounds() && NSW ?
Builder->CreateInBoundsGEP(StrippedPtr, Off, GEP.getName()) :
}
}
- if (!DL)
- return nullptr;
-
// addrspacecast between types is canonicalized as a bitcast, then an
// addrspacecast. To take advantage of the below bitcast + struct GEP, look
// through the addrspacecast.
if (BitCastInst *BCI = dyn_cast<BitCastInst>(PtrOp)) {
Value *Operand = BCI->getOperand(0);
PointerType *OpType = cast<PointerType>(Operand->getType());
- unsigned OffsetBits = DL->getPointerTypeSizeInBits(GEP.getType());
+ unsigned OffsetBits = DL.getPointerTypeSizeInBits(GEP.getType());
APInt Offset(OffsetBits, 0);
if (!isa<BitCastInst>(Operand) &&
- GEP.accumulateConstantOffset(*DL, Offset)) {
+ GEP.accumulateConstantOffset(DL, Offset)) {
// If this GEP instruction doesn't move the pointer, just replace the GEP
// with a bitcast of the real input to the dest type.
Value *Cond = SI.getCondition();
unsigned BitWidth = cast<IntegerType>(Cond->getType())->getBitWidth();
APInt KnownZero(BitWidth, 0), KnownOne(BitWidth, 0);
- computeKnownBits(Cond, KnownZero, KnownOne);
+ computeKnownBits(Cond, KnownZero, KnownOne, 0, &SI);
unsigned LeadingKnownZeros = KnownZero.countLeadingOnes();
unsigned LeadingKnownOnes = KnownOne.countLeadingOnes();
// x86 generates redundant zero-extenstion instructions if the operand is
// truncated to i8 or i16.
bool TruncCond = false;
- if (DL && BitWidth > NewWidth &&
- NewWidth >= DL->getLargestLegalIntTypeSize()) {
+ if (BitWidth > NewWidth && NewWidth >= DL.getLargestLegalIntTypeSize()) {
TruncCond = true;
IntegerType *Ty = IntegerType::get(SI.getContext(), NewWidth);
Builder->SetInsertPoint(&SI);
}
// Instruction isn't dead, see if we can constant propagate it.
- if (!I->use_empty() && isa<Constant>(I->getOperand(0)))
+ if (!I->use_empty() && isa<Constant>(I->getOperand(0))) {
if (Constant *C = ConstantFoldInstruction(I, DL, TLI)) {
DEBUG(dbgs() << "IC: ConstFold to: " << *C << " from: " << *I << '\n');
MadeIRChange = true;
continue;
}
+ }
// See if we can trivially sink this instruction to a successor basic block.
if (I->hasOneUse()) {
/// many instructions are dead or constant). Additionally, if we find a branch
/// whose condition is a known constant, we only visit the reachable successors.
///
-static bool AddReachableCodeToWorklist(BasicBlock *BB,
- SmallPtrSetImpl<BasicBlock*> &Visited,
+static bool AddReachableCodeToWorklist(BasicBlock *BB, const DataLayout &DL,
+ SmallPtrSetImpl<BasicBlock *> &Visited,
InstCombineWorklist &ICWorklist,
- const DataLayout *DL,
const TargetLibraryInfo *TLI) {
bool MadeIRChange = false;
SmallVector<BasicBlock*, 256> Worklist;
continue;
}
- if (DL) {
- // See if we can constant fold its operands.
- for (User::op_iterator i = Inst->op_begin(), e = Inst->op_end();
- i != e; ++i) {
- ConstantExpr *CE = dyn_cast<ConstantExpr>(i);
- if (CE == nullptr) continue;
+ // See if we can constant fold its operands.
+ for (User::op_iterator i = Inst->op_begin(), e = Inst->op_end(); i != e;
+ ++i) {
+ ConstantExpr *CE = dyn_cast<ConstantExpr>(i);
+ if (CE == nullptr)
+ continue;
- Constant*& FoldRes = FoldedConstants[CE];
- if (!FoldRes)
- FoldRes = ConstantFoldConstantExpression(CE, DL, TLI);
- if (!FoldRes)
- FoldRes = CE;
+ Constant *&FoldRes = FoldedConstants[CE];
+ if (!FoldRes)
+ FoldRes = ConstantFoldConstantExpression(CE, DL, TLI);
+ if (!FoldRes)
+ FoldRes = CE;
- if (FoldRes != CE) {
- *i = FoldRes;
- MadeIRChange = true;
- }
+ if (FoldRes != CE) {
+ *i = FoldRes;
+ MadeIRChange = true;
}
}
///
/// This also does basic constant propagation and other forward fixing to make
/// the combiner itself run much faster.
-static bool prepareICWorklistFromFunction(Function &F, const DataLayout *DL,
+static bool prepareICWorklistFromFunction(Function &F, const DataLayout &DL,
TargetLibraryInfo *TLI,
InstCombineWorklist &ICWorklist) {
bool MadeIRChange = false;
// track of which blocks we visit.
SmallPtrSet<BasicBlock *, 64> Visited;
MadeIRChange |=
- AddReachableCodeToWorklist(F.begin(), Visited, ICWorklist, DL, TLI);
+ AddReachableCodeToWorklist(F.begin(), DL, Visited, ICWorklist, TLI);
// Do a quick scan over the function. If we find any blocks that are
// unreachable, remove any instructions inside of them. This prevents
DominatorTree &DT, LoopInfo *LI = nullptr) {
// Minimizing size?
bool MinimizeSize = F.hasFnAttribute(Attribute::MinSize);
- const DataLayout &DL = F.getParent()->getDataLayout();
+ auto &DL = F.getParent()->getDataLayout();
/// Builder - This is an IRBuilder that automatically inserts new
/// instructions into the worklist when they are created.
IRBuilder<true, TargetFolder, InstCombineIRInserter> Builder(
- F.getContext(), TargetFolder(&DL), InstCombineIRInserter(Worklist, &AC));
+ F.getContext(), TargetFolder(DL), InstCombineIRInserter(Worklist, &AC));
// Lower dbg.declare intrinsics otherwise their value may be clobbered
// by instcombiner.
<< F.getName() << "\n");
bool Changed = false;
- if (prepareICWorklistFromFunction(F, &DL, &TLI, Worklist))
+ if (prepareICWorklistFromFunction(F, DL, &TLI, Worklist))
Changed = true;
- InstCombiner IC(Worklist, &Builder, MinimizeSize, &AC, &TLI, &DT, &DL, LI);
+ InstCombiner IC(Worklist, &Builder, MinimizeSize, &AC, &TLI, &DT, DL, LI);
if (IC.run())
Changed = true;
}
uint64_t getAllocaSizeInBytes(AllocaInst *AI) const {
Type *Ty = AI->getAllocatedType();
- uint64_t SizeInBytes = DL->getTypeAllocSize(Ty);
+ uint64_t SizeInBytes =
+ AI->getModule()->getDataLayout().getTypeAllocSize(Ty);
return SizeInBytes;
}
/// Check if we want (and can) handle this alloca.
uint64_t *TypeSize,
unsigned *Alignment) const;
void instrumentMop(ObjectSizeOffsetVisitor &ObjSizeVis, Instruction *I,
- bool UseCalls);
+ bool UseCalls, const DataLayout &DL);
void instrumentPointerComparisonOrSubtraction(Instruction *I);
void instrumentAddress(Instruction *OrigIns, Instruction *InsertBefore,
Value *Addr, uint32_t TypeSize, bool IsWrite,
uint64_t TypeSize) const;
LLVMContext *C;
- const DataLayout *DL;
Triple TargetTriple;
int LongSize;
Type *IntptrTy;
GlobalsMetadata GlobalsMD;
Type *IntptrTy;
LLVMContext *C;
- const DataLayout *DL;
Triple TargetTriple;
ShadowMapping Mapping;
Function *AsanPoisonGlobals;
// Right shift for BigEndian and left shift for LittleEndian.
Value *shiftAllocaMagic(Value *Val, IRBuilder<> &IRB, Value *Shift) {
- return ASan.DL->isLittleEndian() ? IRB.CreateShl(Val, Shift)
- : IRB.CreateLShr(Val, Shift);
+ auto &DL = F.getParent()->getDataLayout();
+ return DL.isLittleEndian() ? IRB.CreateShl(Val, Shift)
+ : IRB.CreateLShr(Val, Shift);
}
// Compute PartialRzMagic for dynamic alloca call. Since we don't know the
if (I->getMetadata("nosanitize")) return nullptr;
Value *PtrOperand = nullptr;
+ const DataLayout &DL = I->getModule()->getDataLayout();
if (LoadInst *LI = dyn_cast<LoadInst>(I)) {
if (!ClInstrumentReads) return nullptr;
*IsWrite = false;
- *TypeSize = DL->getTypeStoreSizeInBits(LI->getType());
+ *TypeSize = DL.getTypeStoreSizeInBits(LI->getType());
*Alignment = LI->getAlignment();
PtrOperand = LI->getPointerOperand();
} else if (StoreInst *SI = dyn_cast<StoreInst>(I)) {
if (!ClInstrumentWrites) return nullptr;
*IsWrite = true;
- *TypeSize = DL->getTypeStoreSizeInBits(SI->getValueOperand()->getType());
+ *TypeSize = DL.getTypeStoreSizeInBits(SI->getValueOperand()->getType());
*Alignment = SI->getAlignment();
PtrOperand = SI->getPointerOperand();
} else if (AtomicRMWInst *RMW = dyn_cast<AtomicRMWInst>(I)) {
if (!ClInstrumentAtomics) return nullptr;
*IsWrite = true;
- *TypeSize = DL->getTypeStoreSizeInBits(RMW->getValOperand()->getType());
+ *TypeSize = DL.getTypeStoreSizeInBits(RMW->getValOperand()->getType());
*Alignment = 0;
PtrOperand = RMW->getPointerOperand();
} else if (AtomicCmpXchgInst *XCHG = dyn_cast<AtomicCmpXchgInst>(I)) {
if (!ClInstrumentAtomics) return nullptr;
*IsWrite = true;
- *TypeSize =
- DL->getTypeStoreSizeInBits(XCHG->getCompareOperand()->getType());
+ *TypeSize = DL.getTypeStoreSizeInBits(XCHG->getCompareOperand()->getType());
*Alignment = 0;
PtrOperand = XCHG->getPointerOperand();
}
}
void AddressSanitizer::instrumentMop(ObjectSizeOffsetVisitor &ObjSizeVis,
- Instruction *I, bool UseCalls) {
+ Instruction *I, bool UseCalls,
+ const DataLayout &DL) {
bool IsWrite = false;
unsigned Alignment = 0;
uint64_t TypeSize = 0;
if (ClOpt && ClOptGlobals) {
// If initialization order checking is disabled, a simple access to a
// dynamically initialized global is always valid.
- GlobalVariable *G =
- dyn_cast<GlobalVariable>(GetUnderlyingObject(Addr, nullptr));
+ GlobalVariable *G = dyn_cast<GlobalVariable>(GetUnderlyingObject(Addr, DL));
if (G != NULL && (!ClInitializers || GlobalIsLinkerInitialized(G)) &&
isSafeAccess(ObjSizeVis, Addr, TypeSize)) {
NumOptimizedAccessesToGlobalVar++;
if (ClOpt && ClOptStack) {
// A direct inbounds access to a stack variable is always valid.
- if (isa<AllocaInst>(GetUnderlyingObject(Addr, nullptr)) &&
+ if (isa<AllocaInst>(GetUnderlyingObject(Addr, DL)) &&
isSafeAccess(ObjSizeVis, Addr, TypeSize)) {
NumOptimizedAccessesToStackVar++;
return;
GlobalVariable *ModuleName = createPrivateGlobalForString(
M, M.getModuleIdentifier(), /*AllowMerging*/ false);
+ auto &DL = M.getDataLayout();
for (size_t i = 0; i < n; i++) {
static const uint64_t kMaxGlobalRedzone = 1 << 18;
GlobalVariable *G = GlobalsToChange[i];
PointerType *PtrTy = cast<PointerType>(G->getType());
Type *Ty = PtrTy->getElementType();
- uint64_t SizeInBytes = DL->getTypeAllocSize(Ty);
+ uint64_t SizeInBytes = DL.getTypeAllocSize(Ty);
uint64_t MinRZ = MinRedzoneSizeForGlobal();
// MinRZ <= RZ <= kMaxGlobalRedzone
// and trying to make RZ to be ~ 1/4 of SizeInBytes.
}
bool AddressSanitizerModule::runOnModule(Module &M) {
- DL = &M.getDataLayout();
C = &(M.getContext());
- int LongSize = DL->getPointerSizeInBits();
+ int LongSize = M.getDataLayout().getPointerSizeInBits();
IntptrTy = Type::getIntNTy(*C, LongSize);
TargetTriple = Triple(M.getTargetTriple());
Mapping = getShadowMapping(TargetTriple, LongSize);
// virtual
bool AddressSanitizer::doInitialization(Module &M) {
// Initialize the private fields. No one has accessed them before.
- DL = &M.getDataLayout();
GlobalsMD.init(M);
C = &(M.getContext());
- LongSize = DL->getPointerSizeInBits();
+ LongSize = M.getDataLayout().getPointerSizeInBits();
IntptrTy = Type::getIntNTy(*C, LongSize);
TargetTriple = Triple(M.getTargetTriple());
const TargetLibraryInfo *TLI =
&getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
+ const DataLayout &DL = F.getParent()->getDataLayout();
ObjectSizeOffsetVisitor ObjSizeVis(DL, TLI, F.getContext(),
/*RoundToAlign=*/true);
if (ClDebugMin < 0 || ClDebugMax < 0 ||
(NumInstrumented >= ClDebugMin && NumInstrumented <= ClDebugMax)) {
if (isInterestingMemoryAccess(Inst, &IsWrite, &TypeSize, &Alignment))
- instrumentMop(ObjSizeVis, Inst, UseCalls);
+ instrumentMop(ObjSizeVis, Inst, UseCalls,
+ F.getParent()->getDataLayout());
else
instrumentMemIntrinsic(cast<MemIntrinsic>(Inst));
}
for (; i + LargeStoreSizeInBytes - 1 < n; i += LargeStoreSizeInBytes) {
uint64_t Val = 0;
for (size_t j = 0; j < LargeStoreSizeInBytes; j++) {
- if (ASan.DL->isLittleEndian())
+ if (F.getParent()->getDataLayout().isLittleEndian())
Val |= (uint64_t)ShadowBytes[i + j] << (8 * j);
else
Val = (Val << 8) | ShadowBytes[i + j];
Value *Shift = IRB.CreateAnd(PartialSize, IRB.getInt32(~7));
unsigned Val1Int = kAsanAllocaPartialVal1;
unsigned Val2Int = kAsanAllocaPartialVal2;
- if (!ASan.DL->isLittleEndian()) {
+ if (!F.getParent()->getDataLayout().isLittleEndian()) {
Val1Int = sys::getSwappedBytes(Val1Int);
Val2Int = sys::getSwappedBytes(Val2Int);
}
Value *Val1 = shiftAllocaMagic(IRB.getInt32(Val1Int), IRB, Shift);
Value *PartialBits = IRB.CreateAnd(PartialSize, IRB.getInt32(7));
// For BigEndian get 0x000000YZ -> 0xYZ000000.
- if (ASan.DL->isBigEndian())
+ if (F.getParent()->getDataLayout().isBigEndian())
PartialBits = IRB.CreateShl(PartialBits, IRB.getInt32(24));
Value *Val2 = IRB.getInt32(Val2Int);
Value *Cond =
// redzones, and OldSize is number of allocated blocks with
// ElementSize size, get allocated memory size in bytes by
// OldSize * ElementSize.
- unsigned ElementSize = ASan.DL->getTypeAllocSize(AI->getAllocatedType());
+ unsigned ElementSize =
+ F.getParent()->getDataLayout().getTypeAllocSize(AI->getAllocatedType());
Value *OldSize = IRB.CreateMul(AI->getArraySize(),
ConstantInt::get(IntptrTy, ElementSize));
}
private:
- const DataLayout *DL;
const TargetLibraryInfo *TLI;
ObjectSizeOffsetEvaluator *ObjSizeEval;
BuilderTy *Builder;
BasicBlock *getTrapBB();
void emitBranchToTrap(Value *Cmp = nullptr);
- bool instrument(Value *Ptr, Value *Val);
+ bool instrument(Value *Ptr, Value *Val, const DataLayout &DL);
};
}
/// result from the load or the value being stored. It is used to determine the
/// size of memory block that is touched.
/// Returns true if any change was made to the IR, false otherwise.
-bool BoundsChecking::instrument(Value *Ptr, Value *InstVal) {
- uint64_t NeededSize = DL->getTypeStoreSize(InstVal->getType());
+bool BoundsChecking::instrument(Value *Ptr, Value *InstVal,
+ const DataLayout &DL) {
+ uint64_t NeededSize = DL.getTypeStoreSize(InstVal->getType());
DEBUG(dbgs() << "Instrument " << *Ptr << " for " << Twine(NeededSize)
<< " bytes\n");
Value *Offset = SizeOffset.second;
ConstantInt *SizeCI = dyn_cast<ConstantInt>(Size);
- Type *IntTy = DL->getIntPtrType(Ptr->getType());
+ Type *IntTy = DL.getIntPtrType(Ptr->getType());
Value *NeededSizeVal = ConstantInt::get(IntTy, NeededSize);
// three checks are required to ensure safety:
}
bool BoundsChecking::runOnFunction(Function &F) {
- DL = &F.getParent()->getDataLayout();
+ const DataLayout &DL = F.getParent()->getDataLayout();
TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
TrapBB = nullptr;
Builder->SetInsertPoint(Inst);
if (LoadInst *LI = dyn_cast<LoadInst>(Inst)) {
- MadeChange |= instrument(LI->getPointerOperand(), LI);
+ MadeChange |= instrument(LI->getPointerOperand(), LI, DL);
} else if (StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
- MadeChange |= instrument(SI->getPointerOperand(), SI->getValueOperand());
+ MadeChange |=
+ instrument(SI->getPointerOperand(), SI->getValueOperand(), DL);
} else if (AtomicCmpXchgInst *AI = dyn_cast<AtomicCmpXchgInst>(Inst)) {
- MadeChange |= instrument(AI->getPointerOperand(),AI->getCompareOperand());
+ MadeChange |=
+ instrument(AI->getPointerOperand(), AI->getCompareOperand(), DL);
} else if (AtomicRMWInst *AI = dyn_cast<AtomicRMWInst>(Inst)) {
- MadeChange |= instrument(AI->getPointerOperand(), AI->getValOperand());
+ MadeChange |=
+ instrument(AI->getPointerOperand(), AI->getValOperand(), DL);
} else {
llvm_unreachable("unknown Instruction type");
}
WK_Custom
};
- const DataLayout *DL;
Module *Mod;
LLVMContext *Ctx;
IntegerType *ShadowTy;
bool IsMIPS64 = TargetTriple.getArch() == llvm::Triple::mips64 ||
TargetTriple.getArch() == llvm::Triple::mips64el;
- DL = &M.getDataLayout();
+ const DataLayout &DL = M.getDataLayout();
Mod = &M;
Ctx = &M.getContext();
ShadowTy = IntegerType::get(*Ctx, ShadowWidth);
ShadowPtrTy = PointerType::getUnqual(ShadowTy);
- IntptrTy = DL->getIntPtrType(*Ctx);
+ IntptrTy = DL.getIntPtrType(*Ctx);
ZeroShadow = ConstantInt::getSigned(ShadowTy, 0);
ShadowPtrMul = ConstantInt::getSigned(IntptrTy, ShadowWidth / 8);
if (IsX86_64)
uint64_t ShadowAlign = Align * DFS.ShadowWidth / 8;
SmallVector<Value *, 2> Objs;
- GetUnderlyingObjects(Addr, Objs, DFS.DL);
+ GetUnderlyingObjects(Addr, Objs, Pos->getModule()->getDataLayout());
bool AllConstants = true;
for (SmallVector<Value *, 2>::iterator i = Objs.begin(), e = Objs.end();
i != e; ++i) {
}
void DFSanVisitor::visitLoadInst(LoadInst &LI) {
- uint64_t Size = DFSF.DFS.DL->getTypeStoreSize(LI.getType());
+ auto &DL = LI.getModule()->getDataLayout();
+ uint64_t Size = DL.getTypeStoreSize(LI.getType());
if (Size == 0) {
DFSF.setShadow(&LI, DFSF.DFS.ZeroShadow);
return;
if (ClPreserveAlignment) {
Align = LI.getAlignment();
if (Align == 0)
- Align = DFSF.DFS.DL->getABITypeAlignment(LI.getType());
+ Align = DL.getABITypeAlignment(LI.getType());
} else {
Align = 1;
}
}
void DFSanVisitor::visitStoreInst(StoreInst &SI) {
- uint64_t Size =
- DFSF.DFS.DL->getTypeStoreSize(SI.getValueOperand()->getType());
+ auto &DL = SI.getModule()->getDataLayout();
+ uint64_t Size = DL.getTypeStoreSize(SI.getValueOperand()->getType());
if (Size == 0)
return;
if (ClPreserveAlignment) {
Align = SI.getAlignment();
if (Align == 0)
- Align = DFSF.DFS.DL->getABITypeAlignment(SI.getValueOperand()->getType());
+ Align = DL.getABITypeAlignment(SI.getValueOperand()->getType());
} else {
Align = 1;
}
MemorySanitizer(int TrackOrigins = 0)
: FunctionPass(ID),
TrackOrigins(std::max(TrackOrigins, (int)ClTrackOrigins)),
- DL(nullptr),
WarningFn(nullptr) {}
const char *getPassName() const override { return "MemorySanitizer"; }
bool runOnFunction(Function &F) override;
/// \brief Track origins (allocation points) of uninitialized values.
int TrackOrigins;
- const DataLayout *DL;
LLVMContext *C;
Type *IntptrTy;
Type *OriginTy;
///
/// inserts a call to __msan_init to the module's constructor list.
bool MemorySanitizer::doInitialization(Module &M) {
- DL = &M.getDataLayout();
+ auto &DL = M.getDataLayout();
Triple TargetTriple(M.getTargetTriple());
switch (TargetTriple.getOS()) {
}
Value *originToIntptr(IRBuilder<> &IRB, Value *Origin) {
- unsigned IntptrSize = MS.DL->getTypeStoreSize(MS.IntptrTy);
+ const DataLayout &DL = F.getParent()->getDataLayout();
+ unsigned IntptrSize = DL.getTypeStoreSize(MS.IntptrTy);
if (IntptrSize == kOriginSize) return Origin;
assert(IntptrSize == kOriginSize * 2);
Origin = IRB.CreateIntCast(Origin, MS.IntptrTy, /* isSigned */ false);
/// \brief Fill memory range with the given origin value.
void paintOrigin(IRBuilder<> &IRB, Value *Origin, Value *OriginPtr,
unsigned Size, unsigned Alignment) {
- unsigned IntptrAlignment = MS.DL->getABITypeAlignment(MS.IntptrTy);
- unsigned IntptrSize = MS.DL->getTypeStoreSize(MS.IntptrTy);
+ const DataLayout &DL = F.getParent()->getDataLayout();
+ unsigned IntptrAlignment = DL.getABITypeAlignment(MS.IntptrTy);
+ unsigned IntptrSize = DL.getTypeStoreSize(MS.IntptrTy);
assert(IntptrAlignment >= kMinOriginAlignment);
assert(IntptrSize >= kOriginSize);
void storeOrigin(IRBuilder<> &IRB, Value *Addr, Value *Shadow, Value *Origin,
unsigned Alignment, bool AsCall) {
+ const DataLayout &DL = F.getParent()->getDataLayout();
unsigned OriginAlignment = std::max(kMinOriginAlignment, Alignment);
- unsigned StoreSize = MS.DL->getTypeStoreSize(Shadow->getType());
+ unsigned StoreSize = DL.getTypeStoreSize(Shadow->getType());
if (isa<StructType>(Shadow->getType())) {
paintOrigin(IRB, updateOrigin(Origin, IRB),
getOriginPtr(Addr, IRB, Alignment), StoreSize,
}
unsigned TypeSizeInBits =
- MS.DL->getTypeSizeInBits(ConvertedShadow->getType());
+ DL.getTypeSizeInBits(ConvertedShadow->getType());
unsigned SizeIndex = TypeSizeToSizeIndex(TypeSizeInBits);
if (AsCall && SizeIndex < kNumberOfAccessSizes) {
Value *Fn = MS.MaybeStoreOriginFn[SizeIndex];
return;
}
- unsigned TypeSizeInBits =
- MS.DL->getTypeSizeInBits(ConvertedShadow->getType());
+ const DataLayout &DL = OrigIns->getModule()->getDataLayout();
+
+ unsigned TypeSizeInBits = DL.getTypeSizeInBits(ConvertedShadow->getType());
unsigned SizeIndex = TypeSizeToSizeIndex(TypeSizeInBits);
if (AsCall && SizeIndex < kNumberOfAccessSizes) {
Value *Fn = MS.MaybeWarningFn[SizeIndex];
/// \brief Add MemorySanitizer instrumentation to a function.
bool runOnFunction() {
MS.initializeCallbacks(*F.getParent());
- if (!MS.DL) return false;
// In the presence of unreachable blocks, we may see Phi nodes with
// incoming nodes from such blocks. Since InstVisitor skips unreachable
// This may return weird-sized types like i1.
if (IntegerType *IT = dyn_cast<IntegerType>(OrigTy))
return IT;
+ const DataLayout &DL = F.getParent()->getDataLayout();
if (VectorType *VT = dyn_cast<VectorType>(OrigTy)) {
- uint32_t EltSize = MS.DL->getTypeSizeInBits(VT->getElementType());
+ uint32_t EltSize = DL.getTypeSizeInBits(VT->getElementType());
return VectorType::get(IntegerType::get(*MS.C, EltSize),
VT->getNumElements());
}
DEBUG(dbgs() << "getShadowTy: " << *ST << " ===> " << *Res << "\n");
return Res;
}
- uint32_t TypeSize = MS.DL->getTypeSizeInBits(OrigTy);
+ uint32_t TypeSize = DL.getTypeSizeInBits(OrigTy);
return IntegerType::get(*MS.C, TypeSize);
}
Function *F = A->getParent();
IRBuilder<> EntryIRB(F->getEntryBlock().getFirstNonPHI());
unsigned ArgOffset = 0;
+ const DataLayout &DL = F->getParent()->getDataLayout();
for (auto &FArg : F->args()) {
if (!FArg.getType()->isSized()) {
DEBUG(dbgs() << "Arg is not sized\n");
continue;
}
- unsigned Size = FArg.hasByValAttr()
- ? MS.DL->getTypeAllocSize(FArg.getType()->getPointerElementType())
- : MS.DL->getTypeAllocSize(FArg.getType());
+ unsigned Size =
+ FArg.hasByValAttr()
+ ? DL.getTypeAllocSize(FArg.getType()->getPointerElementType())
+ : DL.getTypeAllocSize(FArg.getType());
if (A == &FArg) {
bool Overflow = ArgOffset + Size > kParamTLSSize;
Value *Base = getShadowPtrForArgument(&FArg, EntryIRB, ArgOffset);
unsigned ArgAlign = FArg.getParamAlignment();
if (ArgAlign == 0) {
Type *EltType = A->getType()->getPointerElementType();
- ArgAlign = MS.DL->getABITypeAlignment(EltType);
+ ArgAlign = DL.getABITypeAlignment(EltType);
}
if (Overflow) {
// ParamTLS overflow.
DEBUG(dbgs() << " Arg#" << i << ": " << *A <<
" Shadow: " << *ArgShadow << "\n");
bool ArgIsInitialized = false;
+ const DataLayout &DL = F.getParent()->getDataLayout();
if (CS.paramHasAttr(i + 1, Attribute::ByVal)) {
assert(A->getType()->isPointerTy() &&
"ByVal argument is not a pointer!");
- Size = MS.DL->getTypeAllocSize(A->getType()->getPointerElementType());
+ Size = DL.getTypeAllocSize(A->getType()->getPointerElementType());
if (ArgOffset + Size > kParamTLSSize) break;
unsigned ParamAlignment = CS.getParamAlignment(i + 1);
unsigned Alignment = std::min(ParamAlignment, kShadowTLSAlignment);
getShadowPtr(A, Type::getInt8Ty(*MS.C), IRB),
Size, Alignment);
} else {
- Size = MS.DL->getTypeAllocSize(A->getType());
+ Size = DL.getTypeAllocSize(A->getType());
if (ArgOffset + Size > kParamTLSSize) break;
Store = IRB.CreateAlignedStore(ArgShadow, ArgShadowBase,
kShadowTLSAlignment);
setShadow(&I, getCleanShadow(&I));
setOrigin(&I, getCleanOrigin());
IRBuilder<> IRB(I.getNextNode());
- uint64_t Size = MS.DL->getTypeAllocSize(I.getAllocatedType());
+ const DataLayout &DL = F.getParent()->getDataLayout();
+ uint64_t Size = DL.getTypeAllocSize(I.getAllocatedType());
if (PoisonStack && ClPoisonStackWithCall) {
IRB.CreateCall2(MS.MsanPoisonStackFn,
IRB.CreatePointerCast(&I, IRB.getInt8PtrTy()),
unsigned GpOffset = 0;
unsigned FpOffset = AMD64GpEndOffset;
unsigned OverflowOffset = AMD64FpEndOffset;
+ const DataLayout &DL = F.getParent()->getDataLayout();
for (CallSite::arg_iterator ArgIt = CS.arg_begin(), End = CS.arg_end();
ArgIt != End; ++ArgIt) {
Value *A = *ArgIt;
// ByVal arguments always go to the overflow area.
assert(A->getType()->isPointerTy());
Type *RealTy = A->getType()->getPointerElementType();
- uint64_t ArgSize = MS.DL->getTypeAllocSize(RealTy);
+ uint64_t ArgSize = DL.getTypeAllocSize(RealTy);
Value *Base = getShadowPtrForVAArgument(RealTy, IRB, OverflowOffset);
OverflowOffset += RoundUpToAlignment(ArgSize, 8);
IRB.CreateMemCpy(Base, MSV.getShadowPtr(A, IRB.getInt8Ty(), IRB),
FpOffset += 16;
break;
case AK_Memory:
- uint64_t ArgSize = MS.DL->getTypeAllocSize(A->getType());
+ uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
Base = getShadowPtrForVAArgument(A->getType(), IRB, OverflowOffset);
OverflowOffset += RoundUpToAlignment(ArgSize, 8);
}
void visitCallSite(CallSite &CS, IRBuilder<> &IRB) override {
unsigned VAArgOffset = 0;
+ const DataLayout &DL = F.getParent()->getDataLayout();
for (CallSite::arg_iterator ArgIt = CS.arg_begin() + 1, End = CS.arg_end();
ArgIt != End; ++ArgIt) {
Value *A = *ArgIt;
Value *Base;
- uint64_t ArgSize = MS.DL->getTypeAllocSize(A->getType());
+ uint64_t ArgSize = DL.getTypeAllocSize(A->getType());
#if defined(__MIPSEB__) || defined(MIPSEB)
// Adjusting the shadow for argument with size < 8 to match the placement
// of bits in big endian system
/// ThreadSanitizer: instrument the code in module to find races.
struct ThreadSanitizer : public FunctionPass {
- ThreadSanitizer() : FunctionPass(ID), DL(nullptr) {}
+ ThreadSanitizer() : FunctionPass(ID) {}
const char *getPassName() const override;
bool runOnFunction(Function &F) override;
bool doInitialization(Module &M) override;
private:
void initializeCallbacks(Module &M);
- bool instrumentLoadOrStore(Instruction *I);
- bool instrumentAtomic(Instruction *I);
+ bool instrumentLoadOrStore(Instruction *I, const DataLayout &DL);
+ bool instrumentAtomic(Instruction *I, const DataLayout &DL);
bool instrumentMemIntrinsic(Instruction *I);
- void chooseInstructionsToInstrument(SmallVectorImpl<Instruction*> &Local,
- SmallVectorImpl<Instruction*> &All);
+ void chooseInstructionsToInstrument(SmallVectorImpl<Instruction *> &Local,
+ SmallVectorImpl<Instruction *> &All,
+ const DataLayout &DL);
bool addrPointsToConstantData(Value *Addr);
- int getMemoryAccessFuncIndex(Value *Addr);
+ int getMemoryAccessFuncIndex(Value *Addr, const DataLayout &DL);
- const DataLayout *DL;
Type *IntptrTy;
IntegerType *OrdTy;
// Callbacks to run-time library are computed in doInitialization.
}
bool ThreadSanitizer::doInitialization(Module &M) {
- DL = &M.getDataLayout();
+ const DataLayout &DL = M.getDataLayout();
// Always insert a call to __tsan_init into the module's CTORs.
IRBuilder<> IRB(M.getContext());
// 'Local' is a vector of insns within the same BB (no calls between).
// 'All' is a vector of insns that will be instrumented.
void ThreadSanitizer::chooseInstructionsToInstrument(
- SmallVectorImpl<Instruction*> &Local,
- SmallVectorImpl<Instruction*> &All) {
+ SmallVectorImpl<Instruction *> &Local, SmallVectorImpl<Instruction *> &All,
+ const DataLayout &DL) {
SmallSet<Value*, 8> WriteTargets;
// Iterate from the end.
for (SmallVectorImpl<Instruction*>::reverse_iterator It = Local.rbegin(),
Value *Addr = isa<StoreInst>(*I)
? cast<StoreInst>(I)->getPointerOperand()
: cast<LoadInst>(I)->getPointerOperand();
- if (isa<AllocaInst>(GetUnderlyingObject(Addr, nullptr)) &&
+ if (isa<AllocaInst>(GetUnderlyingObject(Addr, DL)) &&
!PointerMayBeCaptured(Addr, true, true)) {
// The variable is addressable but not captured, so it cannot be
// referenced from a different thread and participate in a data race
}
bool ThreadSanitizer::runOnFunction(Function &F) {
- if (!DL) return false;
initializeCallbacks(*F.getParent());
SmallVector<Instruction*, 8> RetVec;
SmallVector<Instruction*, 8> AllLoadsAndStores;
bool Res = false;
bool HasCalls = false;
bool SanitizeFunction = F.hasFnAttribute(Attribute::SanitizeThread);
+ const DataLayout &DL = F.getParent()->getDataLayout();
// Traverse all instructions, collect loads/stores/returns, check for calls.
for (auto &BB : F) {
if (isa<MemIntrinsic>(Inst))
MemIntrinCalls.push_back(&Inst);
HasCalls = true;
- chooseInstructionsToInstrument(LocalLoadsAndStores, AllLoadsAndStores);
+ chooseInstructionsToInstrument(LocalLoadsAndStores, AllLoadsAndStores,
+ DL);
}
}
- chooseInstructionsToInstrument(LocalLoadsAndStores, AllLoadsAndStores);
+ chooseInstructionsToInstrument(LocalLoadsAndStores, AllLoadsAndStores, DL);
}
// We have collected all loads and stores.
// Instrument memory accesses only if we want to report bugs in the function.
if (ClInstrumentMemoryAccesses && SanitizeFunction)
for (auto Inst : AllLoadsAndStores) {
- Res |= instrumentLoadOrStore(Inst);
+ Res |= instrumentLoadOrStore(Inst, DL);
}
// Instrument atomic memory accesses in any case (they can be used to
// implement synchronization).
if (ClInstrumentAtomics)
for (auto Inst : AtomicAccesses) {
- Res |= instrumentAtomic(Inst);
+ Res |= instrumentAtomic(Inst, DL);
}
if (ClInstrumentMemIntrinsics && SanitizeFunction)
return Res;
}
-bool ThreadSanitizer::instrumentLoadOrStore(Instruction *I) {
+bool ThreadSanitizer::instrumentLoadOrStore(Instruction *I,
+ const DataLayout &DL) {
IRBuilder<> IRB(I);
bool IsWrite = isa<StoreInst>(*I);
Value *Addr = IsWrite
? cast<StoreInst>(I)->getPointerOperand()
: cast<LoadInst>(I)->getPointerOperand();
- int Idx = getMemoryAccessFuncIndex(Addr);
+ int Idx = getMemoryAccessFuncIndex(Addr, DL);
if (Idx < 0)
return false;
if (IsWrite && isVtableAccess(I)) {
? cast<StoreInst>(I)->getAlignment()
: cast<LoadInst>(I)->getAlignment();
Type *OrigTy = cast<PointerType>(Addr->getType())->getElementType();
- const uint32_t TypeSize = DL->getTypeStoreSizeInBits(OrigTy);
+ const uint32_t TypeSize = DL.getTypeStoreSizeInBits(OrigTy);
Value *OnAccessFunc = nullptr;
if (Alignment == 0 || Alignment >= 8 || (Alignment % (TypeSize / 8)) == 0)
OnAccessFunc = IsWrite ? TsanWrite[Idx] : TsanRead[Idx];
// The following page contains more background information:
// http://www.hpl.hp.com/personal/Hans_Boehm/c++mm/
-bool ThreadSanitizer::instrumentAtomic(Instruction *I) {
+bool ThreadSanitizer::instrumentAtomic(Instruction *I, const DataLayout &DL) {
IRBuilder<> IRB(I);
if (LoadInst *LI = dyn_cast<LoadInst>(I)) {
Value *Addr = LI->getPointerOperand();
- int Idx = getMemoryAccessFuncIndex(Addr);
+ int Idx = getMemoryAccessFuncIndex(Addr, DL);
if (Idx < 0)
return false;
const size_t ByteSize = 1 << Idx;
} else if (StoreInst *SI = dyn_cast<StoreInst>(I)) {
Value *Addr = SI->getPointerOperand();
- int Idx = getMemoryAccessFuncIndex(Addr);
+ int Idx = getMemoryAccessFuncIndex(Addr, DL);
if (Idx < 0)
return false;
const size_t ByteSize = 1 << Idx;
ReplaceInstWithInst(I, C);
} else if (AtomicRMWInst *RMWI = dyn_cast<AtomicRMWInst>(I)) {
Value *Addr = RMWI->getPointerOperand();
- int Idx = getMemoryAccessFuncIndex(Addr);
+ int Idx = getMemoryAccessFuncIndex(Addr, DL);
if (Idx < 0)
return false;
Function *F = TsanAtomicRMW[RMWI->getOperation()][Idx];
ReplaceInstWithInst(I, C);
} else if (AtomicCmpXchgInst *CASI = dyn_cast<AtomicCmpXchgInst>(I)) {
Value *Addr = CASI->getPointerOperand();
- int Idx = getMemoryAccessFuncIndex(Addr);
+ int Idx = getMemoryAccessFuncIndex(Addr, DL);
if (Idx < 0)
return false;
const size_t ByteSize = 1 << Idx;
return true;
}
-int ThreadSanitizer::getMemoryAccessFuncIndex(Value *Addr) {
+int ThreadSanitizer::getMemoryAccessFuncIndex(Value *Addr,
+ const DataLayout &DL) {
Type *OrigPtrTy = Addr->getType();
Type *OrigTy = cast<PointerType>(OrigPtrTy)->getElementType();
assert(OrigTy->isSized());
- uint32_t TypeSize = DL->getTypeStoreSizeInBits(OrigTy);
+ uint32_t TypeSize = DL.getTypeStoreSizeInBits(OrigTy);
if (TypeSize != 8 && TypeSize != 16 &&
TypeSize != 32 && TypeSize != 64 && TypeSize != 128) {
NumAccessesWithBadSize++;
if (AliasAnalysis::onlyReadsMemory(MRB))
return false;
if (AliasAnalysis::onlyAccessesArgPointees(MRB)) {
+ const DataLayout &DL = Inst->getModule()->getDataLayout();
for (ImmutableCallSite::arg_iterator I = CS.arg_begin(), E = CS.arg_end();
I != E; ++I) {
const Value *Op = *I;
- if (IsPotentialRetainableObjPtr(Op, *PA.getAA()) && PA.related(Ptr, Op))
+ if (IsPotentialRetainableObjPtr(Op, *PA.getAA()) &&
+ PA.related(Ptr, Op, DL))
return true;
}
return false;
if (Class == ARCInstKind::Call)
return false;
+ const DataLayout &DL = Inst->getModule()->getDataLayout();
+
// Consider various instructions which may have pointer arguments which are
// not "uses".
if (const ICmpInst *ICI = dyn_cast<ICmpInst>(Inst)) {
for (ImmutableCallSite::arg_iterator OI = CS.arg_begin(),
OE = CS.arg_end(); OI != OE; ++OI) {
const Value *Op = *OI;
- if (IsPotentialRetainableObjPtr(Op, *PA.getAA()) && PA.related(Ptr, Op))
+ if (IsPotentialRetainableObjPtr(Op, *PA.getAA()) &&
+ PA.related(Ptr, Op, DL))
return true;
}
return false;
} else if (const StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
// Special-case stores, because we don't care about the stored value, just
// the store address.
- const Value *Op = GetUnderlyingObjCPtr(SI->getPointerOperand());
+ const Value *Op = GetUnderlyingObjCPtr(SI->getPointerOperand(), DL);
// If we can't tell what the underlying object was, assume there is a
// dependence.
- return IsPotentialRetainableObjPtr(Op, *PA.getAA()) && PA.related(Op, Ptr);
+ return IsPotentialRetainableObjPtr(Op, *PA.getAA()) &&
+ PA.related(Op, Ptr, DL);
}
// Check each operand for a match.
for (User::const_op_iterator OI = Inst->op_begin(), OE = Inst->op_end();
OI != OE; ++OI) {
const Value *Op = *OI;
- if (IsPotentialRetainableObjPtr(Op, *PA.getAA()) && PA.related(Ptr, Op))
+ if (IsPotentialRetainableObjPtr(Op, *PA.getAA()) && PA.related(Ptr, Op, DL))
return true;
}
return false;
/// \brief This is a wrapper around getUnderlyingObject which also knows how to
/// look through objc_retain and objc_autorelease calls, which we know to return
/// their argument verbatim.
-static inline const Value *GetUnderlyingObjCPtr(const Value *V) {
+static inline const Value *GetUnderlyingObjCPtr(const Value *V,
+ const DataLayout &DL) {
for (;;) {
- V = GetUnderlyingObject(V);
+ V = GetUnderlyingObject(V, DL);
if (!IsForwarding(GetBasicARCInstKind(V)))
break;
V = cast<CallInst>(V)->getArgOperand(0);
// If that failed, climb to the underlying object, including climbing through
// ObjC-specific no-ops, and try making an imprecise alias query.
- const Value *UA = GetUnderlyingObjCPtr(SA);
- const Value *UB = GetUnderlyingObjCPtr(SB);
+ const Value *UA = GetUnderlyingObjCPtr(SA, *DL);
+ const Value *UB = GetUnderlyingObjCPtr(SB, *DL);
if (UA != SA || UB != SB) {
Result = AliasAnalysis::alias(Location(UA), Location(UB));
// We can't use MustAlias or PartialAlias results here because
// If that failed, climb to the underlying object, including climbing through
// ObjC-specific no-ops, and try making an imprecise alias query.
- const Value *U = GetUnderlyingObjCPtr(S);
+ const Value *U = GetUnderlyingObjCPtr(S, *DL);
if (U != S)
return AliasAnalysis::pointsToConstantMemory(Location(U), OrLocal);
/// This is a wrapper around getUnderlyingObjCPtr along the lines of
/// GetUnderlyingObjects except that it returns early when it sees the first
/// alloca.
-static inline bool AreAnyUnderlyingObjectsAnAlloca(const Value *V) {
+static inline bool AreAnyUnderlyingObjectsAnAlloca(const Value *V,
+ const DataLayout &DL) {
SmallPtrSet<const Value *, 4> Visited;
SmallVector<const Value *, 4> Worklist;
Worklist.push_back(V);
do {
const Value *P = Worklist.pop_back_val();
- P = GetUnderlyingObjCPtr(P);
+ P = GetUnderlyingObjCPtr(P, DL);
if (isa<AllocaInst>(P))
return true;
// in the presence of allocas we only unconditionally remove pointers if
// both our retain and our release are KnownSafe.
if (StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
- if (AreAnyUnderlyingObjectsAnAlloca(SI->getPointerOperand())) {
+ const DataLayout &DL = BB->getModule()->getDataLayout();
+ if (AreAnyUnderlyingObjectsAnAlloca(SI->getPointerOperand(), DL)) {
auto I = MyStates.findPtrBottomUpState(
GetRCIdentityRoot(SI->getValueOperand()));
if (I != MyStates.bottom_up_ptr_end())
bool ProvenanceAnalysis::relatedSelect(const SelectInst *A,
const Value *B) {
+ const DataLayout &DL = A->getModule()->getDataLayout();
// If the values are Selects with the same condition, we can do a more precise
// check: just check for relations between the values on corresponding arms.
if (const SelectInst *SB = dyn_cast<SelectInst>(B))
if (A->getCondition() == SB->getCondition())
- return related(A->getTrueValue(), SB->getTrueValue()) ||
- related(A->getFalseValue(), SB->getFalseValue());
+ return related(A->getTrueValue(), SB->getTrueValue(), DL) ||
+ related(A->getFalseValue(), SB->getFalseValue(), DL);
// Check both arms of the Select node individually.
- return related(A->getTrueValue(), B) ||
- related(A->getFalseValue(), B);
+ return related(A->getTrueValue(), B, DL) ||
+ related(A->getFalseValue(), B, DL);
}
bool ProvenanceAnalysis::relatedPHI(const PHINode *A,
const Value *B) {
+ const DataLayout &DL = A->getModule()->getDataLayout();
// If the values are PHIs in the same block, we can do a more precise as well
// as efficient check: just check for relations between the values on
// corresponding edges.
if (PNB->getParent() == A->getParent()) {
for (unsigned i = 0, e = A->getNumIncomingValues(); i != e; ++i)
if (related(A->getIncomingValue(i),
- PNB->getIncomingValueForBlock(A->getIncomingBlock(i))))
+ PNB->getIncomingValueForBlock(A->getIncomingBlock(i)), DL))
return true;
return false;
}
SmallPtrSet<const Value *, 4> UniqueSrc;
for (unsigned i = 0, e = A->getNumIncomingValues(); i != e; ++i) {
const Value *PV1 = A->getIncomingValue(i);
- if (UniqueSrc.insert(PV1).second && related(PV1, B))
+ if (UniqueSrc.insert(PV1).second && related(PV1, B, DL))
return true;
}
return false;
}
-bool ProvenanceAnalysis::relatedCheck(const Value *A,
- const Value *B) {
+bool ProvenanceAnalysis::relatedCheck(const Value *A, const Value *B,
+ const DataLayout &DL) {
// Skip past provenance pass-throughs.
- A = GetUnderlyingObjCPtr(A);
- B = GetUnderlyingObjCPtr(B);
+ A = GetUnderlyingObjCPtr(A, DL);
+ B = GetUnderlyingObjCPtr(B, DL);
// Quick check.
if (A == B)
return true;
}
-bool ProvenanceAnalysis::related(const Value *A,
- const Value *B) {
+bool ProvenanceAnalysis::related(const Value *A, const Value *B,
+ const DataLayout &DL) {
// Begin by inserting a conservative value into the map. If the insertion
// fails, we have the answer already. If it succeeds, leave it there until we
// compute the real answer to guard against recursive queries.
if (!Pair.second)
return Pair.first->second;
- bool Result = relatedCheck(A, B);
+ bool Result = relatedCheck(A, B, DL);
CachedResults[ValuePairTy(A, B)] = Result;
return Result;
}
namespace llvm {
class Value;
class AliasAnalysis;
+ class DataLayout;
class PHINode;
class SelectInst;
}
typedef DenseMap<ValuePairTy, bool> CachedResultsTy;
CachedResultsTy CachedResults;
- bool relatedCheck(const Value *A, const Value *B);
+ bool relatedCheck(const Value *A, const Value *B, const DataLayout &DL);
bool relatedSelect(const SelectInst *A, const Value *B);
bool relatedPHI(const PHINode *A, const Value *B);
AliasAnalysis *getAA() const { return AA; }
- bool related(const Value *A, const Value *B);
+ bool related(const Value *A, const Value *B, const DataLayout &DL);
void clear() {
CachedResults.clear();
#include "llvm/Analysis/Passes.h"
#include "llvm/IR/InstIterator.h"
#include "llvm/IR/Function.h"
+#include "llvm/IR/Module.h"
#include "llvm/Support/raw_ostream.h"
using namespace llvm;
ProvenanceAnalysis PA;
PA.setAA(&getAnalysis<AliasAnalysis>());
+ const DataLayout &DL = F.getParent()->getDataLayout();
for (Value *V1 : Values) {
StringRef NameV1 = getName(V1);
if (NameV1 >= NameV2)
continue;
errs() << NameV1 << " and " << NameV2;
- if (PA.related(V1, V2))
+ if (PA.related(V1, V2, DL))
errs() << " are related.\n";
else
errs() << " are not related.\n";
#include "llvm/IR/Instruction.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Intrinsics.h"
-#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Module.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
ScalarEvolution *SE;
DominatorTree *DT;
- const DataLayout *DL;
bool extractAlignmentInfo(CallInst *I, Value *&AAPtr, const SCEV *&AlignSCEV,
const SCEV *&OffSCEV);
continue;
if (Instruction *K = dyn_cast<Instruction>(J))
- if (isValidAssumeForContext(ACall, K, DL, DT))
+ if (isValidAssumeForContext(ACall, K, DT))
WorkList.push_back(K);
}
Visited.insert(J);
for (User *UJ : J->users()) {
Instruction *K = cast<Instruction>(UJ);
- if (!Visited.count(K) && isValidAssumeForContext(ACall, K, DL, DT))
+ if (!Visited.count(K) && isValidAssumeForContext(ACall, K, DT))
WorkList.push_back(K);
}
}
auto &AC = getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
SE = &getAnalysis<ScalarEvolution>();
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
- DL = &F.getParent()->getDataLayout();
NewDestAlignments.clear();
NewSrcAlignments.clear();
APInt &KnownZero2, APInt &KnownOne2);
AssumptionCache *AC;
- const DataLayout *DL;
DominatorTree *DT;
};
}
// however, want to do this twice, so we cache the result in APInts that live
// in the caller. For the two-relevant-operands case, both operand values are
// provided here.
- auto ComputeKnownBits = [&](unsigned BitWidth, const Value *V1,
- const Value *V2) {
- KnownZero = APInt(BitWidth, 0);
- KnownOne = APInt(BitWidth, 0);
- computeKnownBits(const_cast<Value*>(V1), KnownZero, KnownOne, DL, 0, AC,
- UserI, DT);
-
- if (V2) {
- KnownZero2 = APInt(BitWidth, 0);
- KnownOne2 = APInt(BitWidth, 0);
- computeKnownBits(const_cast<Value*>(V2), KnownZero2, KnownOne2, DL, 0, AC,
- UserI, DT);
- }
- };
+ auto ComputeKnownBits =
+ [&](unsigned BitWidth, const Value *V1, const Value *V2) {
+ const DataLayout &DL = I->getModule()->getDataLayout();
+ KnownZero = APInt(BitWidth, 0);
+ KnownOne = APInt(BitWidth, 0);
+ computeKnownBits(const_cast<Value *>(V1), KnownZero, KnownOne, DL, 0,
+ AC, UserI, DT);
+
+ if (V2) {
+ KnownZero2 = APInt(BitWidth, 0);
+ KnownOne2 = APInt(BitWidth, 0);
+ computeKnownBits(const_cast<Value *>(V2), KnownZero2, KnownOne2, DL,
+ 0, AC, UserI, DT);
+ }
+ };
switch (UserI->getOpcode()) {
default: break;
return false;
AC = &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
- DL = &F.getParent()->getDataLayout();
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
DenseMap<Instruction *, APInt> AliveBits;
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/IR/Constant.h"
-#include "llvm/IR/DataLayout.h"
#include "llvm/IR/InstIterator.h"
#include "llvm/IR/Instruction.h"
#include "llvm/Pass.h"
WorkList.erase(WorkList.begin()); // Get an element from the worklist...
if (!I->use_empty()) // Don't muck with dead instructions...
- if (Constant *C = ConstantFoldInstruction(I, &DL, TLI)) {
+ if (Constant *C = ConstantFoldInstruction(I, DL, TLI)) {
// Add all of the users of this instruction to the worklist, they might
// be constant propagatable now...
for (User *U : I->users())
#include "llvm/IR/Constants.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/Instructions.h"
+#include "llvm/IR/Module.h"
#include "llvm/Pass.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
Changed = true;
}
- // FIXME: Provide DL, TLI, DT, AT to SimplifyInstruction.
- if (Value *V = SimplifyInstruction(P)) {
+ // FIXME: Provide TLI, DT, AT to SimplifyInstruction.
+ const DataLayout &DL = BB->getModule()->getDataLayout();
+ if (Value *V = SimplifyInstruction(P, DL)) {
P->replaceAllUsesWith(V);
P->eraseFromParent();
Changed = true;
bool HandleFree(CallInst *F);
bool handleEndBlock(BasicBlock &BB);
void RemoveAccessedObjects(const AliasAnalysis::Location &LoadedLoc,
- SmallSetVector<Value*, 16> &DeadStackObjects);
+ SmallSetVector<Value *, 16> &DeadStackObjects,
+ const DataLayout &DL);
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.setPreservesCFG();
/// describe the memory operations for this instruction.
static AliasAnalysis::Location
getLocForWrite(Instruction *Inst, AliasAnalysis &AA) {
- const DataLayout *DL = AA.getDataLayout();
if (StoreInst *SI = dyn_cast<StoreInst>(Inst))
return AA.getLocation(SI);
if (MemIntrinsic *MI = dyn_cast<MemIntrinsic>(Inst)) {
// memcpy/memmove/memset.
AliasAnalysis::Location Loc = AA.getLocationForDest(MI);
- // If we don't have target data around, an unknown size in Location means
- // that we should use the size of the pointee type. This isn't valid for
- // memset/memcpy, which writes more than an i8.
- if (Loc.Size == AliasAnalysis::UnknownSize && DL == nullptr)
- return AliasAnalysis::Location();
return Loc;
}
switch (II->getIntrinsicID()) {
default: return AliasAnalysis::Location(); // Unhandled intrinsic.
case Intrinsic::init_trampoline:
- // If we don't have target data around, an unknown size in Location means
- // that we should use the size of the pointee type. This isn't valid for
- // init.trampoline, which writes more than an i8.
- if (!DL) return AliasAnalysis::Location();
-
// FIXME: We don't know the size of the trampoline, so we can't really
// handle it here.
return AliasAnalysis::Location(II->getArgOperand(0));
return CS.getArgument(0);
}
-static uint64_t getPointerSize(const Value *V, AliasAnalysis &AA) {
+static uint64_t getPointerSize(const Value *V, const DataLayout &DL,
+ const TargetLibraryInfo *TLI) {
uint64_t Size;
- if (getObjectSize(V, Size, AA.getDataLayout(), AA.getTargetLibraryInfo()))
+ if (getObjectSize(V, Size, DL, TLI))
return Size;
return AliasAnalysis::UnknownSize;
}
/// overwritten by 'Later', or 'OverwriteUnknown' if nothing can be determined
static OverwriteResult isOverwrite(const AliasAnalysis::Location &Later,
const AliasAnalysis::Location &Earlier,
- AliasAnalysis &AA,
- int64_t &EarlierOff,
- int64_t &LaterOff) {
- const DataLayout *DL = AA.getDataLayout();
+ const DataLayout &DL,
+ const TargetLibraryInfo *TLI,
+ int64_t &EarlierOff, int64_t &LaterOff) {
const Value *P1 = Earlier.Ptr->stripPointerCasts();
const Value *P2 = Later.Ptr->stripPointerCasts();
// Otherwise, we have to have size information, and the later store has to be
// larger than the earlier one.
if (Later.Size == AliasAnalysis::UnknownSize ||
- Earlier.Size == AliasAnalysis::UnknownSize || DL == nullptr)
+ Earlier.Size == AliasAnalysis::UnknownSize)
return OverwriteUnknown;
// Check to see if the later store is to the entire object (either a global,
return OverwriteUnknown;
// If the "Later" store is to a recognizable object, get its size.
- uint64_t ObjectSize = getPointerSize(UO2, AA);
+ uint64_t ObjectSize = getPointerSize(UO2, DL, TLI);
if (ObjectSize != AliasAnalysis::UnknownSize)
if (ObjectSize == Later.Size && ObjectSize >= Earlier.Size)
return OverwriteComplete;
if (isRemovable(DepWrite) &&
!isPossibleSelfRead(Inst, Loc, DepWrite, *AA)) {
int64_t InstWriteOffset, DepWriteOffset;
- OverwriteResult OR = isOverwrite(Loc, DepLoc, *AA,
- DepWriteOffset, InstWriteOffset);
+ const DataLayout &DL = BB.getModule()->getDataLayout();
+ OverwriteResult OR =
+ isOverwrite(Loc, DepLoc, DL, AA->getTargetLibraryInfo(),
+ DepWriteOffset, InstWriteOffset);
if (OR == OverwriteComplete) {
DEBUG(dbgs() << "DSE: Remove Dead Store:\n DEAD: "
<< *DepWrite << "\n KILLER: " << *Inst << '\n');
AliasAnalysis::Location Loc = AliasAnalysis::Location(F->getOperand(0));
SmallVector<BasicBlock *, 16> Blocks;
Blocks.push_back(F->getParent());
+ const DataLayout &DL = F->getModule()->getDataLayout();
while (!Blocks.empty()) {
BasicBlock *BB = Blocks.pop_back_val();
break;
Value *DepPointer =
- GetUnderlyingObject(getStoredPointerOperand(Dependency));
+ GetUnderlyingObject(getStoredPointerOperand(Dependency), DL);
// Check for aliasing.
if (!AA->isMustAlias(F->getArgOperand(0), DepPointer))
if (AI->hasByValOrInAllocaAttr())
DeadStackObjects.insert(AI);
+ const DataLayout &DL = BB.getModule()->getDataLayout();
+
// Scan the basic block backwards
for (BasicBlock::iterator BBI = BB.end(); BBI != BB.begin(); ){
--BBI;
if (hasMemoryWrite(BBI, TLI) && isRemovable(BBI)) {
// See through pointer-to-pointer bitcasts
SmallVector<Value *, 4> Pointers;
- GetUnderlyingObjects(getStoredPointerOperand(BBI), Pointers);
+ GetUnderlyingObjects(getStoredPointerOperand(BBI), Pointers, DL);
// Stores to stack values are valid candidates for removal.
bool AllDead = true;
// the call is live.
DeadStackObjects.remove_if([&](Value *I) {
// See if the call site touches the value.
- AliasAnalysis::ModRefResult A =
- AA->getModRefInfo(CS, I, getPointerSize(I, *AA));
+ AliasAnalysis::ModRefResult A = AA->getModRefInfo(
+ CS, I, getPointerSize(I, DL, AA->getTargetLibraryInfo()));
return A == AliasAnalysis::ModRef || A == AliasAnalysis::Ref;
});
// Remove any allocas from the DeadPointer set that are loaded, as this
// makes any stores above the access live.
- RemoveAccessedObjects(LoadedLoc, DeadStackObjects);
+ RemoveAccessedObjects(LoadedLoc, DeadStackObjects, DL);
// If all of the allocas were clobbered by the access then we're not going
// to find anything else to process.
/// of the stack objects in the DeadStackObjects set. If so, they become live
/// because the location is being loaded.
void DSE::RemoveAccessedObjects(const AliasAnalysis::Location &LoadedLoc,
- SmallSetVector<Value*, 16> &DeadStackObjects) {
- const Value *UnderlyingPointer = GetUnderlyingObject(LoadedLoc.Ptr);
+ SmallSetVector<Value *, 16> &DeadStackObjects,
+ const DataLayout &DL) {
+ const Value *UnderlyingPointer = GetUnderlyingObject(LoadedLoc.Ptr, DL);
// A constant can't be in the dead pointer set.
if (isa<Constant>(UnderlyingPointer))
// Remove objects that could alias LoadedLoc.
DeadStackObjects.remove_if([&](Value *I) {
// See if the loaded location could alias the stack location.
- AliasAnalysis::Location StackLoc(I, getPointerSize(I, *AA));
+ AliasAnalysis::Location StackLoc(
+ I, getPointerSize(I, DL, AA->getTargetLibraryInfo()));
return !AA->isNoAlias(StackLoc, LoadedLoc);
});
}
class EarlyCSE {
public:
Function &F;
- const DataLayout *DL;
const TargetLibraryInfo &TLI;
const TargetTransformInfo &TTI;
DominatorTree &DT;
unsigned CurrentGeneration;
/// \brief Set up the EarlyCSE runner for a particular function.
- EarlyCSE(Function &F, const DataLayout *DL, const TargetLibraryInfo &TLI,
+ EarlyCSE(Function &F, const TargetLibraryInfo &TLI,
const TargetTransformInfo &TTI, DominatorTree &DT,
AssumptionCache &AC)
- : F(F), DL(DL), TLI(TLI), TTI(TTI), DT(DT), AC(AC), CurrentGeneration(0) {
- }
+ : F(F), TLI(TLI), TTI(TTI), DT(DT), AC(AC), CurrentGeneration(0) {}
bool run();
Instruction *LastStore = nullptr;
bool Changed = false;
+ const DataLayout &DL = BB->getModule()->getDataLayout();
// See if any instructions in the block can be eliminated. If so, do it. If
// not, add them to AvailableValues.
PreservedAnalyses EarlyCSEPass::run(Function &F,
AnalysisManager<Function> *AM) {
- const DataLayout &DL = F.getParent()->getDataLayout();
-
auto &TLI = AM->getResult<TargetLibraryAnalysis>(F);
auto &TTI = AM->getResult<TargetIRAnalysis>(F);
auto &DT = AM->getResult<DominatorTreeAnalysis>(F);
auto &AC = AM->getResult<AssumptionAnalysis>(F);
- EarlyCSE CSE(F, &DL, TLI, TTI, DT, AC);
+ EarlyCSE CSE(F, TLI, TTI, DT, AC);
if (!CSE.run())
return PreservedAnalyses::all();
if (skipOptnoneFunction(F))
return false;
- auto &DL = F.getParent()->getDataLayout();
auto &TLI = getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
auto &TTI = getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
auto &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
auto &AC = getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
- EarlyCSE CSE(F, &DL, TLI, TTI, DT, AC);
+ EarlyCSE CSE(F, TLI, TTI, DT, AC);
return CSE.run();
}
/// Emit code into this block to adjust the value defined here to the
/// specified type. This handles various coercion cases.
- Value *MaterializeAdjustedValue(Type *LoadTy, GVN &gvn) const;
+ Value *MaterializeAdjustedValue(LoadInst *LI, GVN &gvn) const;
};
class GVN : public FunctionPass {
bool NoLoads;
MemoryDependenceAnalysis *MD;
DominatorTree *DT;
- const DataLayout *DL;
const TargetLibraryInfo *TLI;
AssumptionCache *AC;
SetVector<BasicBlock *> DeadBlocks;
InstrsToErase.push_back(I);
}
- const DataLayout *getDataLayout() const { return DL; }
DominatorTree &getDominatorTree() const { return *DT; }
AliasAnalysis *getAliasAnalysis() const { return VN.getAliasAnalysis(); }
MemoryDependenceAnalysis &getMemDep() const { return *MD; }
return -1;
int64_t StoreOffset = 0, LoadOffset = 0;
- Value *StoreBase = GetPointerBaseWithConstantOffset(WritePtr,StoreOffset,&DL);
- Value *LoadBase = GetPointerBaseWithConstantOffset(LoadPtr, LoadOffset, &DL);
+ Value *StoreBase =
+ GetPointerBaseWithConstantOffset(WritePtr, StoreOffset, DL);
+ Value *LoadBase = GetPointerBaseWithConstantOffset(LoadPtr, LoadOffset, DL);
if (StoreBase != LoadBase)
return -1;
/// This function is called when we have a
/// memdep query of a load that ends up being a clobbering store.
static int AnalyzeLoadFromClobberingStore(Type *LoadTy, Value *LoadPtr,
- StoreInst *DepSI,
- const DataLayout &DL) {
+ StoreInst *DepSI) {
// Cannot handle reading from store of first-class aggregate yet.
if (DepSI->getValueOperand()->getType()->isStructTy() ||
DepSI->getValueOperand()->getType()->isArrayTy())
return -1;
+ const DataLayout &DL = DepSI->getModule()->getDataLayout();
Value *StorePtr = DepSI->getPointerOperand();
uint64_t StoreSize =DL.getTypeSizeInBits(DepSI->getValueOperand()->getType());
return AnalyzeLoadFromClobberingWrite(LoadTy, LoadPtr,
// then we should widen it!
int64_t LoadOffs = 0;
const Value *LoadBase =
- GetPointerBaseWithConstantOffset(LoadPtr, LoadOffs, &DL);
+ GetPointerBaseWithConstantOffset(LoadPtr, LoadOffs, DL);
unsigned LoadSize = DL.getTypeStoreSize(LoadTy);
- unsigned Size = MemoryDependenceAnalysis::
- getLoadLoadClobberFullWidthSize(LoadBase, LoadOffs, LoadSize, DepLI, DL);
+ unsigned Size = MemoryDependenceAnalysis::getLoadLoadClobberFullWidthSize(
+ LoadBase, LoadOffs, LoadSize, DepLI);
if (Size == 0) return -1;
return AnalyzeLoadFromClobberingWrite(LoadTy, LoadPtr, DepPtr, Size*8, DL);
Constant *Src = dyn_cast<Constant>(MTI->getSource());
if (!Src) return -1;
- GlobalVariable *GV = dyn_cast<GlobalVariable>(GetUnderlyingObject(Src, &DL));
+ GlobalVariable *GV = dyn_cast<GlobalVariable>(GetUnderlyingObject(Src, DL));
if (!GV || !GV->isConstant()) return -1;
// See if the access is within the bounds of the transfer.
ConstantInt::get(Type::getInt64Ty(Src->getContext()), (unsigned)Offset);
Src = ConstantExpr::getGetElementPtr(Src, OffsetCst);
Src = ConstantExpr::getBitCast(Src, PointerType::get(LoadTy, AS));
- if (ConstantFoldLoadFromConstPtr(Src, &DL))
+ if (ConstantFoldLoadFromConstPtr(Src, DL))
return Offset;
return -1;
}
static Value *GetLoadValueForLoad(LoadInst *SrcVal, unsigned Offset,
Type *LoadTy, Instruction *InsertPt,
GVN &gvn) {
- const DataLayout &DL = *gvn.getDataLayout();
+ const DataLayout &DL = SrcVal->getModule()->getDataLayout();
// If Offset+LoadTy exceeds the size of SrcVal, then we must be wanting to
// widen SrcVal out to a larger load.
unsigned SrcValSize = DL.getTypeStoreSize(SrcVal->getType());
ConstantInt::get(Type::getInt64Ty(Src->getContext()), (unsigned)Offset);
Src = ConstantExpr::getGetElementPtr(Src, OffsetCst);
Src = ConstantExpr::getBitCast(Src, PointerType::get(LoadTy, AS));
- return ConstantFoldLoadFromConstPtr(Src, &DL);
+ return ConstantFoldLoadFromConstPtr(Src, DL);
}
gvn.getDominatorTree().properlyDominates(ValuesPerBlock[0].BB,
LI->getParent())) {
assert(!ValuesPerBlock[0].isUndefValue() && "Dead BB dominate this block");
- return ValuesPerBlock[0].MaterializeAdjustedValue(LI->getType(), gvn);
+ return ValuesPerBlock[0].MaterializeAdjustedValue(LI, gvn);
}
// Otherwise, we have to construct SSA form.
SSAUpdater SSAUpdate(&NewPHIs);
SSAUpdate.Initialize(LI->getType(), LI->getName());
- Type *LoadTy = LI->getType();
-
for (unsigned i = 0, e = ValuesPerBlock.size(); i != e; ++i) {
const AvailableValueInBlock &AV = ValuesPerBlock[i];
BasicBlock *BB = AV.BB;
if (SSAUpdate.HasValueForBlock(BB))
continue;
- SSAUpdate.AddAvailableValue(BB, AV.MaterializeAdjustedValue(LoadTy, gvn));
+ SSAUpdate.AddAvailableValue(BB, AV.MaterializeAdjustedValue(LI, gvn));
}
// Perform PHI construction.
return V;
}
-Value *AvailableValueInBlock::MaterializeAdjustedValue(Type *LoadTy, GVN &gvn) const {
+Value *AvailableValueInBlock::MaterializeAdjustedValue(LoadInst *LI,
+ GVN &gvn) const {
Value *Res;
+ Type *LoadTy = LI->getType();
+ const DataLayout &DL = LI->getModule()->getDataLayout();
if (isSimpleValue()) {
Res = getSimpleValue();
if (Res->getType() != LoadTy) {
- const DataLayout *DL = gvn.getDataLayout();
- assert(DL && "Need target data to handle type mismatch case");
- Res = GetStoreValueForLoad(Res, Offset, LoadTy, BB->getTerminator(),
- *DL);
-
+ Res = GetStoreValueForLoad(Res, Offset, LoadTy, BB->getTerminator(), DL);
+
DEBUG(dbgs() << "GVN COERCED NONLOCAL VAL:\nOffset: " << Offset << " "
<< *getSimpleValue() << '\n'
<< *Res << '\n' << "\n\n\n");
<< *Res << '\n' << "\n\n\n");
}
} else if (isMemIntrinValue()) {
- const DataLayout *DL = gvn.getDataLayout();
- assert(DL && "Need target data to handle type mismatch case");
- Res = GetMemInstValueForLoad(getMemIntrinValue(), Offset,
- LoadTy, BB->getTerminator(), *DL);
+ Res = GetMemInstValueForLoad(getMemIntrinValue(), Offset, LoadTy,
+ BB->getTerminator(), DL);
DEBUG(dbgs() << "GVN COERCED NONLOCAL MEM INTRIN:\nOffset: " << Offset
<< " " << *getMemIntrinValue() << '\n'
<< *Res << '\n' << "\n\n\n");
// dependencies that produce an unknown value for the load (such as a call
// that could potentially clobber the load).
unsigned NumDeps = Deps.size();
+ const DataLayout &DL = LI->getModule()->getDataLayout();
for (unsigned i = 0, e = NumDeps; i != e; ++i) {
BasicBlock *DepBB = Deps[i].getBB();
MemDepResult DepInfo = Deps[i].getResult();
// read by the load, we can extract the bits we need for the load from the
// stored value.
if (StoreInst *DepSI = dyn_cast<StoreInst>(DepInfo.getInst())) {
- if (DL && Address) {
- int Offset = AnalyzeLoadFromClobberingStore(LI->getType(), Address,
- DepSI, *DL);
+ if (Address) {
+ int Offset =
+ AnalyzeLoadFromClobberingStore(LI->getType(), Address, DepSI);
if (Offset != -1) {
ValuesPerBlock.push_back(AvailableValueInBlock::get(DepBB,
DepSI->getValueOperand(),
if (LoadInst *DepLI = dyn_cast<LoadInst>(DepInfo.getInst())) {
// If this is a clobber and L is the first instruction in its block, then
// we have the first instruction in the entry block.
- if (DepLI != LI && Address && DL) {
- int Offset = AnalyzeLoadFromClobberingLoad(LI->getType(), Address,
- DepLI, *DL);
+ if (DepLI != LI && Address) {
+ int Offset =
+ AnalyzeLoadFromClobberingLoad(LI->getType(), Address, DepLI, DL);
if (Offset != -1) {
ValuesPerBlock.push_back(AvailableValueInBlock::getLoad(DepBB,DepLI,
// If the clobbering value is a memset/memcpy/memmove, see if we can
// forward a value on from it.
if (MemIntrinsic *DepMI = dyn_cast<MemIntrinsic>(DepInfo.getInst())) {
- if (DL && Address) {
+ if (Address) {
int Offset = AnalyzeLoadFromClobberingMemInst(LI->getType(), Address,
- DepMI, *DL);
+ DepMI, DL);
if (Offset != -1) {
ValuesPerBlock.push_back(AvailableValueInBlock::getMI(DepBB, DepMI,
Offset));
if (S->getValueOperand()->getType() != LI->getType()) {
// If the stored value is larger or equal to the loaded value, we can
// reuse it.
- if (!DL || !CanCoerceMustAliasedValueToLoad(S->getValueOperand(),
- LI->getType(), *DL)) {
+ if (!CanCoerceMustAliasedValueToLoad(S->getValueOperand(),
+ LI->getType(), DL)) {
UnavailableBlocks.push_back(DepBB);
continue;
}
if (LD->getType() != LI->getType()) {
// If the stored value is larger or equal to the loaded value, we can
// reuse it.
- if (!DL || !CanCoerceMustAliasedValueToLoad(LD, LI->getType(),*DL)) {
+ if (!CanCoerceMustAliasedValueToLoad(LD, LI->getType(), DL)) {
UnavailableBlocks.push_back(DepBB);
continue;
}
// Check if the load can safely be moved to all the unavailable predecessors.
bool CanDoPRE = true;
+ const DataLayout &DL = LI->getModule()->getDataLayout();
SmallVector<Instruction*, 8> NewInsts;
for (auto &PredLoad : PredLoads) {
BasicBlock *UnavailablePred = PredLoad.first;
// ... to a pointer that has been loaded from before...
MemDepResult Dep = MD->getDependency(L);
+ const DataLayout &DL = L->getModule()->getDataLayout();
// If we have a clobber and target data is around, see if this is a clobber
// that we can fix up through code synthesis.
- if (Dep.isClobber() && DL) {
+ if (Dep.isClobber()) {
// Check to see if we have something like this:
// store i32 123, i32* %P
// %A = bitcast i32* %P to i8*
// access code.
Value *AvailVal = nullptr;
if (StoreInst *DepSI = dyn_cast<StoreInst>(Dep.getInst())) {
- int Offset = AnalyzeLoadFromClobberingStore(L->getType(),
- L->getPointerOperand(),
- DepSI, *DL);
+ int Offset = AnalyzeLoadFromClobberingStore(
+ L->getType(), L->getPointerOperand(), DepSI);
if (Offset != -1)
AvailVal = GetStoreValueForLoad(DepSI->getValueOperand(), Offset,
- L->getType(), L, *DL);
+ L->getType(), L, DL);
}
// Check to see if we have something like this:
if (DepLI == L)
return false;
- int Offset = AnalyzeLoadFromClobberingLoad(L->getType(),
- L->getPointerOperand(),
- DepLI, *DL);
+ int Offset = AnalyzeLoadFromClobberingLoad(
+ L->getType(), L->getPointerOperand(), DepLI, DL);
if (Offset != -1)
AvailVal = GetLoadValueForLoad(DepLI, Offset, L->getType(), L, *this);
}
// If the clobbering value is a memset/memcpy/memmove, see if we can forward
// a value on from it.
if (MemIntrinsic *DepMI = dyn_cast<MemIntrinsic>(Dep.getInst())) {
- int Offset = AnalyzeLoadFromClobberingMemInst(L->getType(),
- L->getPointerOperand(),
- DepMI, *DL);
+ int Offset = AnalyzeLoadFromClobberingMemInst(
+ L->getType(), L->getPointerOperand(), DepMI, DL);
if (Offset != -1)
- AvailVal = GetMemInstValueForLoad(DepMI, Offset, L->getType(), L, *DL);
+ AvailVal = GetMemInstValueForLoad(DepMI, Offset, L->getType(), L, DL);
}
if (AvailVal) {
// actually have the same type. See if we know how to reuse the stored
// value (depending on its type).
if (StoredVal->getType() != L->getType()) {
- if (DL) {
- StoredVal = CoerceAvailableValueToLoadType(StoredVal, L->getType(),
- L, *DL);
- if (!StoredVal)
- return false;
-
- DEBUG(dbgs() << "GVN COERCED STORE:\n" << *DepSI << '\n' << *StoredVal
- << '\n' << *L << "\n\n\n");
- }
- else
+ StoredVal =
+ CoerceAvailableValueToLoadType(StoredVal, L->getType(), L, DL);
+ if (!StoredVal)
return false;
+
+ DEBUG(dbgs() << "GVN COERCED STORE:\n" << *DepSI << '\n' << *StoredVal
+ << '\n' << *L << "\n\n\n");
}
// Remove it!
// the same type. See if we know how to reuse the previously loaded value
// (depending on its type).
if (DepLI->getType() != L->getType()) {
- if (DL) {
- AvailableVal = CoerceAvailableValueToLoadType(DepLI, L->getType(),
- L, *DL);
- if (!AvailableVal)
- return false;
-
- DEBUG(dbgs() << "GVN COERCED LOAD:\n" << *DepLI << "\n" << *AvailableVal
- << "\n" << *L << "\n\n\n");
- }
- else
+ AvailableVal = CoerceAvailableValueToLoadType(DepLI, L->getType(), L, DL);
+ if (!AvailableVal)
return false;
+
+ DEBUG(dbgs() << "GVN COERCED LOAD:\n" << *DepLI << "\n" << *AvailableVal
+ << "\n" << *L << "\n\n\n");
}
// Remove it!
// to value numbering it. Value numbering often exposes redundancies, for
// example if it determines that %y is equal to %x then the instruction
// "%z = and i32 %x, %y" becomes "%z = and i32 %x, %x" which we now simplify.
+ const DataLayout &DL = I->getModule()->getDataLayout();
if (Value *V = SimplifyInstruction(I, DL, TLI, DT, AC)) {
I->replaceAllUsesWith(V);
if (MD && V->getType()->getScalarType()->isPointerTy())
if (!NoLoads)
MD = &getAnalysis<MemoryDependenceAnalysis>();
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
- DL = &F.getParent()->getDataLayout();
AC = &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
VN.setAliasAnalysis(&getAnalysis<AliasAnalysis>());
LoopInfo *LI;
ScalarEvolution *SE;
DominatorTree *DT;
- const DataLayout *DL;
TargetLibraryInfo *TLI;
const TargetTransformInfo *TTI;
public:
static char ID; // Pass identification, replacement for typeid
- IndVarSimplify() : LoopPass(ID), LI(nullptr), SE(nullptr), DT(nullptr),
- DL(nullptr), Changed(false) {
+ IndVarSimplify()
+ : LoopPass(ID), LI(nullptr), SE(nullptr), DT(nullptr), Changed(false) {
initializeIndVarSimplifyPass(*PassRegistry::getPassRegistry());
}
/// extended by this sign or zero extend operation. This is used to determine
/// the final width of the IV before actually widening it.
static void visitIVCast(CastInst *Cast, WideIVInfo &WI, ScalarEvolution *SE,
- const DataLayout *DL, const TargetTransformInfo *TTI) {
+ const TargetTransformInfo *TTI) {
bool IsSigned = Cast->getOpcode() == Instruction::SExt;
if (!IsSigned && Cast->getOpcode() != Instruction::ZExt)
return;
Type *Ty = Cast->getType();
uint64_t Width = SE->getTypeSizeInBits(Ty);
- if (DL && !DL->isLegalInteger(Width))
+ if (!Cast->getModule()->getDataLayout().isLegalInteger(Width))
return;
// Cast is either an sext or zext up to this point.
namespace {
class IndVarSimplifyVisitor : public IVVisitor {
ScalarEvolution *SE;
- const DataLayout *DL;
const TargetTransformInfo *TTI;
PHINode *IVPhi;
WideIVInfo WI;
IndVarSimplifyVisitor(PHINode *IV, ScalarEvolution *SCEV,
- const DataLayout *DL, const TargetTransformInfo *TTI,
+ const TargetTransformInfo *TTI,
const DominatorTree *DTree)
- : SE(SCEV), DL(DL), TTI(TTI), IVPhi(IV) {
+ : SE(SCEV), TTI(TTI), IVPhi(IV) {
DT = DTree;
WI.NarrowIV = IVPhi;
if (ReduceLiveIVs)
}
// Implement the interface used by simplifyUsersOfIV.
- void visitCast(CastInst *Cast) override {
- visitIVCast(Cast, WI, SE, DL, TTI);
- }
+ void visitCast(CastInst *Cast) override { visitIVCast(Cast, WI, SE, TTI); }
};
}
PHINode *CurrIV = LoopPhis.pop_back_val();
// Information about sign/zero extensions of CurrIV.
- IndVarSimplifyVisitor Visitor(CurrIV, SE, DL, TTI, DT);
+ IndVarSimplifyVisitor Visitor(CurrIV, SE, TTI, DT);
Changed |= simplifyUsersOfIV(CurrIV, SE, &LPM, DeadInsts, &Visitor);
/// FIXME: Accept non-unit stride as long as SCEV can reduce BECount * Stride.
/// This is difficult in general for SCEV because of potential overflow. But we
/// could at least handle constant BECounts.
-static PHINode *
-FindLoopCounter(Loop *L, const SCEV *BECount,
- ScalarEvolution *SE, DominatorTree *DT, const DataLayout *DL) {
+static PHINode *FindLoopCounter(Loop *L, const SCEV *BECount,
+ ScalarEvolution *SE, DominatorTree *DT) {
uint64_t BCWidth = SE->getTypeSizeInBits(BECount->getType());
Value *Cond =
// AR may be wider than BECount. With eq/ne tests overflow is immaterial.
// AR may not be a narrower type, or we may never exit.
uint64_t PhiWidth = SE->getTypeSizeInBits(AR->getType());
- if (PhiWidth < BCWidth || (DL && !DL->isLegalInteger(PhiWidth)))
+ if (PhiWidth < BCWidth ||
+ !L->getHeader()->getModule()->getDataLayout().isLegalInteger(PhiWidth))
continue;
const SCEV *Step = dyn_cast<SCEVConstant>(AR->getStepRecurrence(*SE));
LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
SE = &getAnalysis<ScalarEvolution>();
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
- DL = &L->getHeader()->getModule()->getDataLayout();
auto *TLIP = getAnalysisIfAvailable<TargetLibraryInfoWrapperPass>();
TLI = TLIP ? &TLIP->getTLI() : nullptr;
auto *TTIP = getAnalysisIfAvailable<TargetTransformInfoWrapperPass>();
TTI = TTIP ? &TTIP->getTTI(*L->getHeader()->getParent()) : nullptr;
+ const DataLayout &DL = L->getHeader()->getModule()->getDataLayout();
DeadInsts.clear();
Changed = false;
const SCEV *BackedgeTakenCount = SE->getBackedgeTakenCount(L);
// Create a rewriter object which we'll use to transform the code with.
- SCEVExpander Rewriter(*SE, "indvars");
+ SCEVExpander Rewriter(*SE, DL, "indvars");
#ifndef NDEBUG
Rewriter.setDebugType(DEBUG_TYPE);
#endif
// If we have a trip count expression, rewrite the loop's exit condition
// using it. We can currently only handle loops with a single exit.
if (canExpandBackedgeTakenCount(L, SE) && needsLFTR(L, DT)) {
- PHINode *IndVar = FindLoopCounter(L, BackedgeTakenCount, SE, DT, DL);
+ PHINode *IndVar = FindLoopCounter(L, BackedgeTakenCount, SE, DT);
if (IndVar) {
// Check preconditions for proper SCEVExpander operation. SCEV does not
// express SCEVExpander's dependencies, such as LoopSimplify. Instead any
"loop variant exit count doesn't make sense!");
assert(!L.contains(LatchExit) && "expected an exit block!");
-
- Value *IndVarStartV = SCEVExpander(SE, "irce").expandCodeFor(
- IndVarStart, IndVarTy, &*Preheader->rbegin());
+ const DataLayout &DL = Preheader->getModule()->getDataLayout();
+ Value *IndVarStartV =
+ SCEVExpander(SE, DL, "irce")
+ .expandCodeFor(IndVarStart, IndVarTy, &*Preheader->rbegin());
IndVarStartV->setName("indvar.start");
LoopStructure Result;
IntegerType *IVTy =
cast<IntegerType>(MainLoopStructure.IndVarNext->getType());
- SCEVExpander Expander(SE, "irce");
+ SCEVExpander Expander(SE, F.getParent()->getDataLayout(), "irce");
Instruction *InsertPt = OriginalPreheader->getTerminator();
// It would have been better to make `PreLoop' and `PostLoop'
/// revectored to the false side of the second if.
///
class JumpThreading : public FunctionPass {
- const DataLayout *DL;
TargetLibraryInfo *TLI;
LazyValueInfo *LVI;
#ifdef NDEBUG
return false;
DEBUG(dbgs() << "Jump threading on function '" << F.getName() << "'\n");
- DL = &F.getParent()->getDataLayout();
TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
LVI = &getAnalysis<LazyValueInfo>();
assert(Preference == WantInteger && "Compares only produce integers");
PHINode *PN = dyn_cast<PHINode>(Cmp->getOperand(0));
if (PN && PN->getParent() == BB) {
+ const DataLayout &DL = PN->getModule()->getDataLayout();
// We can do this simplification if any comparisons fold to true or false.
// See if any do.
for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) {
// Run constant folding to see if we can reduce the condition to a simple
// constant.
if (Instruction *I = dyn_cast<Instruction>(Condition)) {
- Value *SimpleVal = ConstantFoldInstruction(I, DL, TLI);
+ Value *SimpleVal =
+ ConstantFoldInstruction(I, BB->getModule()->getDataLayout(), TLI);
if (SimpleVal) {
I->replaceAllUsesWith(SimpleVal);
I->eraseFromParent();
// At this point, the IR is fully up to date and consistent. Do a quick scan
// over the new instructions and zap any that are constants or dead. This
// frequently happens because of phi translation.
- SimplifyInstructionsInBlock(NewBB, DL, TLI);
+ SimplifyInstructionsInBlock(NewBB, TLI);
// Threaded an edge!
++NumThreads;
BasicBlock::iterator BI = BB->begin();
for (; PHINode *PN = dyn_cast<PHINode>(BI); ++BI)
ValueMapping[PN] = PN->getIncomingValueForBlock(PredBB);
-
// Clone the non-phi instructions of BB into PredBB, keeping track of the
// mapping and using it to remap operands in the cloned instructions.
for (; BI != BB->end(); ++BI) {
// If this instruction can be simplified after the operands are updated,
// just use the simplified value instead. This frequently happens due to
// phi translation.
- if (Value *IV = SimplifyInstruction(New, DL)) {
+ if (Value *IV =
+ SimplifyInstruction(New, BB->getModule()->getDataLayout())) {
delete New;
ValueMapping[BI] = IV;
} else {
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/Metadata.h"
-#include "llvm/IR/Module.h"
#include "llvm/IR/PredIteratorCache.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
static bool hoist(Instruction &I, BasicBlock *Preheader);
static bool sink(Instruction &I, LoopInfo *LI, DominatorTree *DT,
Loop *CurLoop, AliasSetTracker *CurAST );
-static bool isGuaranteedToExecute(Instruction &Inst, DominatorTree *DT,
- Loop *CurLoop, LICMSafetyInfo * SafetyInfo);
-static bool isSafeToExecuteUnconditionally(Instruction &Inst,DominatorTree *DT,
- const DataLayout *DL, Loop *CurLoop,
- LICMSafetyInfo * SafetyInfo);
+static bool isGuaranteedToExecute(Instruction &Inst, DominatorTree *DT,
+ Loop *CurLoop, LICMSafetyInfo *SafetyInfo);
+static bool isSafeToExecuteUnconditionally(Instruction &Inst, DominatorTree *DT,
+ Loop *CurLoop,
+ LICMSafetyInfo *SafetyInfo);
static bool pointerInvalidatedByLoop(Value *V, uint64_t Size,
const AAMDNodes &AAInfo,
AliasSetTracker *CurAST);
static Instruction *CloneInstructionInExitBlock(Instruction &I,
BasicBlock &ExitBlock,
PHINode &PN, LoopInfo *LI);
-static bool canSinkOrHoistInst(Instruction &I, AliasAnalysis *AA,
- DominatorTree *DT, const DataLayout *DL,
- Loop *CurLoop, AliasSetTracker *CurAST,
- LICMSafetyInfo * SafetyInfo);
+static bool canSinkOrHoistInst(Instruction &I, AliasAnalysis *AA,
+ DominatorTree *DT, Loop *CurLoop,
+ AliasSetTracker *CurAST,
+ LICMSafetyInfo *SafetyInfo);
namespace {
struct LICM : public LoopPass {
LoopInfo *LI; // Current LoopInfo
DominatorTree *DT; // Dominator Tree for the current Loop.
- const DataLayout *DL; // DataLayout for constant folding.
TargetLibraryInfo *TLI; // TargetLibraryInfo for constant folding.
// State that is updated as we process loops.
AA = &getAnalysis<AliasAnalysis>();
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
- DL = &L->getHeader()->getModule()->getDataLayout();
TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
assert(L->isLCSSAForm(*DT) && "Loop is not in LCSSA form.");
// instructions, we perform another pass to hoist them out of the loop.
//
if (L->hasDedicatedExits())
- Changed |= sinkRegion(DT->getNode(L->getHeader()), AA, LI, DT, DL, TLI,
- CurLoop, CurAST, &SafetyInfo);
+ Changed |= sinkRegion(DT->getNode(L->getHeader()), AA, LI, DT, TLI, CurLoop,
+ CurAST, &SafetyInfo);
if (Preheader)
- Changed |= hoistRegion(DT->getNode(L->getHeader()), AA, LI, DT, DL, TLI,
+ Changed |= hoistRegion(DT->getNode(L->getHeader()), AA, LI, DT, TLI,
CurLoop, CurAST, &SafetyInfo);
// Now that all loop invariants have been removed from the loop, promote any
/// first order w.r.t the DominatorTree. This allows us to visit uses before
/// definitions, allowing us to sink a loop body in one pass without iteration.
///
-bool llvm::sinkRegion(DomTreeNode *N, AliasAnalysis *AA, LoopInfo *LI,
- DominatorTree *DT, const DataLayout *DL,
- TargetLibraryInfo *TLI, Loop *CurLoop,
- AliasSetTracker *CurAST, LICMSafetyInfo * SafetyInfo) {
+bool llvm::sinkRegion(DomTreeNode *N, AliasAnalysis *AA, LoopInfo *LI,
+ DominatorTree *DT, TargetLibraryInfo *TLI, Loop *CurLoop,
+ AliasSetTracker *CurAST, LICMSafetyInfo *SafetyInfo) {
// Verify inputs.
assert(N != nullptr && AA != nullptr && LI != nullptr &&
// We are processing blocks in reverse dfo, so process children first.
const std::vector<DomTreeNode*> &Children = N->getChildren();
for (unsigned i = 0, e = Children.size(); i != e; ++i)
- Changed |= sinkRegion(Children[i], AA, LI, DT, DL, TLI, CurLoop,
- CurAST, SafetyInfo);
+ Changed |=
+ sinkRegion(Children[i], AA, LI, DT, TLI, CurLoop, CurAST, SafetyInfo);
// Only need to process the contents of this block if it is not part of a
// subloop (which would already have been processed).
if (inSubLoop(BB,CurLoop,LI)) return Changed;
// outside of the loop. In this case, it doesn't even matter if the
// operands of the instruction are loop invariant.
//
- if (isNotUsedInLoop(I, CurLoop) &&
- canSinkOrHoistInst(I, AA, DT, DL, CurLoop, CurAST, SafetyInfo)) {
+ if (isNotUsedInLoop(I, CurLoop) &&
+ canSinkOrHoistInst(I, AA, DT, CurLoop, CurAST, SafetyInfo)) {
++II;
Changed |= sink(I, LI, DT, CurLoop, CurAST);
}
/// order w.r.t the DominatorTree. This allows us to visit definitions before
/// uses, allowing us to hoist a loop body in one pass without iteration.
///
-bool llvm::hoistRegion(DomTreeNode *N, AliasAnalysis *AA, LoopInfo *LI,
- DominatorTree *DT, const DataLayout *DL,
- TargetLibraryInfo *TLI, Loop *CurLoop,
- AliasSetTracker *CurAST, LICMSafetyInfo *SafetyInfo) {
+bool llvm::hoistRegion(DomTreeNode *N, AliasAnalysis *AA, LoopInfo *LI,
+ DominatorTree *DT, TargetLibraryInfo *TLI, Loop *CurLoop,
+ AliasSetTracker *CurAST, LICMSafetyInfo *SafetyInfo) {
// Verify inputs.
assert(N != nullptr && AA != nullptr && LI != nullptr &&
DT != nullptr && CurLoop != nullptr && CurAST != nullptr &&
// Try constant folding this instruction. If all the operands are
// constants, it is technically hoistable, but it would be better to just
// fold it.
- if (Constant *C = ConstantFoldInstruction(&I, DL, TLI)) {
+ if (Constant *C = ConstantFoldInstruction(
+ &I, I.getModule()->getDataLayout(), TLI)) {
DEBUG(dbgs() << "LICM folding inst: " << I << " --> " << *C << '\n');
CurAST->copyValue(&I, C);
CurAST->deleteValue(&I);
// if all of the operands of the instruction are loop invariant and if it
// is safe to hoist the instruction.
//
- if (CurLoop->hasLoopInvariantOperands(&I) &&
- canSinkOrHoistInst(I, AA, DT, DL, CurLoop, CurAST, SafetyInfo) &&
- isSafeToExecuteUnconditionally(I, DT, DL, CurLoop, SafetyInfo))
+ if (CurLoop->hasLoopInvariantOperands(&I) &&
+ canSinkOrHoistInst(I, AA, DT, CurLoop, CurAST, SafetyInfo) &&
+ isSafeToExecuteUnconditionally(I, DT, CurLoop, SafetyInfo))
Changed |= hoist(I, CurLoop->getLoopPreheader());
}
const std::vector<DomTreeNode*> &Children = N->getChildren();
for (unsigned i = 0, e = Children.size(); i != e; ++i)
- Changed |= hoistRegion(Children[i], AA, LI, DT, DL, TLI, CurLoop,
- CurAST, SafetyInfo);
+ Changed |=
+ hoistRegion(Children[i], AA, LI, DT, TLI, CurLoop, CurAST, SafetyInfo);
return Changed;
}
/// canSinkOrHoistInst - Return true if the hoister and sinker can handle this
/// instruction.
///
-bool canSinkOrHoistInst(Instruction &I, AliasAnalysis *AA,
- DominatorTree *DT, const DataLayout *DL,
- Loop *CurLoop, AliasSetTracker *CurAST,
- LICMSafetyInfo * SafetyInfo) {
+bool canSinkOrHoistInst(Instruction &I, AliasAnalysis *AA, DominatorTree *DT,
+ Loop *CurLoop, AliasSetTracker *CurAST,
+ LICMSafetyInfo *SafetyInfo) {
// Loads have extra constraints we have to verify before we can hoist them.
if (LoadInst *LI = dyn_cast<LoadInst>(&I)) {
if (!LI->isUnordered())
!isa<InsertValueInst>(I))
return false;
- return isSafeToExecuteUnconditionally(I, DT, DL, CurLoop, SafetyInfo);
+ return isSafeToExecuteUnconditionally(I, DT, CurLoop, SafetyInfo);
}
/// Returns true if a PHINode is a trivially replaceable with an
/// or if it is a trapping instruction and is guaranteed to execute.
///
static bool isSafeToExecuteUnconditionally(Instruction &Inst, DominatorTree *DT,
- const DataLayout *DL, Loop *CurLoop,
- LICMSafetyInfo * SafetyInfo) {
+ Loop *CurLoop,
+ LICMSafetyInfo *SafetyInfo) {
// If it is not a trapping instruction, it is always safe to hoist.
- if (isSafeToSpeculativelyExecute(&Inst, DL))
+ if (isSafeToSpeculativelyExecute(&Inst))
return true;
return isGuaranteedToExecute(Inst, DT, CurLoop, SafetyInfo);
class LoadCombine : public BasicBlockPass {
LLVMContext *C;
- const DataLayout *DL;
AliasAnalysis *AA;
public:
- LoadCombine()
- : BasicBlockPass(ID),
- C(nullptr), DL(nullptr), AA(nullptr) {
+ LoadCombine() : BasicBlockPass(ID), C(nullptr), AA(nullptr) {
initializeSROAPass(*PassRegistry::getPassRegistry());
}
bool LoadCombine::doInitialization(Function &F) {
DEBUG(dbgs() << "LoadCombine function: " << F.getName() << "\n");
C = &F.getContext();
- DL = &F.getParent()->getDataLayout();
- if (!DL) {
- DEBUG(dbgs() << " Skipping LoadCombine -- no target data!\n");
- return false;
- }
return true;
}
POP.Offset = 0;
while (isa<BitCastInst>(POP.Pointer) || isa<GetElementPtrInst>(POP.Pointer)) {
if (auto *GEP = dyn_cast<GetElementPtrInst>(POP.Pointer)) {
- unsigned BitWidth = DL->getPointerTypeSizeInBits(GEP->getType());
+ auto &DL = LI.getModule()->getDataLayout();
+ unsigned BitWidth = DL.getPointerTypeSizeInBits(GEP->getType());
APInt Offset(BitWidth, 0);
- if (GEP->accumulateConstantOffset(*DL, Offset))
+ if (GEP->accumulateConstantOffset(DL, Offset))
POP.Offset += Offset.getZExtValue();
else
// Can't handle GEPs with variable indices.
if (PrevOffset == -1ull) {
BaseLoad = L.Load;
PrevOffset = L.POP.Offset;
- PrevSize = DL->getTypeStoreSize(L.Load->getType());
+ PrevSize = L.Load->getModule()->getDataLayout().getTypeStoreSize(
+ L.Load->getType());
AggregateLoads.push_back(L);
continue;
}
// FIXME: We may want to handle this case.
continue;
PrevOffset = L.POP.Offset;
- PrevSize = DL->getTypeStoreSize(L.Load->getType());
+ PrevSize = L.Load->getModule()->getDataLayout().getTypeStoreSize(
+ L.Load->getType());
AggregateLoads.push_back(L);
}
if (combineLoads(AggregateLoads))
for (const auto &L : Loads) {
Builder->SetInsertPoint(L.Load);
Value *V = Builder->CreateExtractInteger(
- *DL, NewLoad, cast<IntegerType>(L.Load->getType()),
+ L.Load->getModule()->getDataLayout(), NewLoad,
+ cast<IntegerType>(L.Load->getType()),
L.POP.Offset - Loads[0].POP.Offset, "combine.extract");
L.Load->replaceAllUsesWith(V);
}
}
bool LoadCombine::runOnBasicBlock(BasicBlock &BB) {
- if (skipOptnoneFunction(BB) || !DL)
+ if (skipOptnoneFunction(BB))
return false;
AA = &getAnalysis<AliasAnalysis>();
- IRBuilder<true, TargetFolder>
- TheBuilder(BB.getContext(), TargetFolder(DL));
+ IRBuilder<true, TargetFolder> TheBuilder(
+ BB.getContext(), TargetFolder(BB.getModule()->getDataLayout()));
Builder = &TheBuilder;
DenseMap<const Value *, SmallVector<LoadPOPPair, 8>> LoadMap;
// header. This allows us to insert code for it in the preheader.
BasicBlock *Preheader = CurLoop->getLoopPreheader();
IRBuilder<> Builder(Preheader->getTerminator());
- SCEVExpander Expander(*SE, "loop-idiom");
+ SCEVExpander Expander(*SE, DL, "loop-idiom");
Type *DestInt8PtrTy = Builder.getInt8PtrTy(DestAS);
// The # stored bytes is (BECount+1)*Size. Expand the trip count out to
// pointer size if it isn't already.
- Type *IntPtr = Builder.getIntPtrTy(&DL, DestAS);
+ Type *IntPtr = Builder.getIntPtrTy(DL, DestAS);
BECount = SE->getTruncateOrZeroExtend(BECount, IntPtr);
const SCEV *NumBytesS = SE->getAddExpr(BECount, SE->getConstant(IntPtr, 1),
// header. This allows us to insert code for it in the preheader.
BasicBlock *Preheader = CurLoop->getLoopPreheader();
IRBuilder<> Builder(Preheader->getTerminator());
- SCEVExpander Expander(*SE, "loop-idiom");
+ const DataLayout &DL = Preheader->getModule()->getDataLayout();
+ SCEVExpander Expander(*SE, DL, "loop-idiom");
// Okay, we have a strided store "p[i]" of a loaded value. We can turn
// this into a memcpy in the loop preheader now if we want. However, this
// The # stored bytes is (BECount+1)*Size. Expand the trip count out to
// pointer size if it isn't already.
- auto &DL = CurLoop->getHeader()->getModule()->getDataLayout();
- Type *IntPtrTy = Builder.getIntPtrTy(&DL, SI->getPointerAddressSpace());
+ Type *IntPtrTy = Builder.getIntPtrTy(DL, SI->getPointerAddressSpace());
BECount = SE->getTruncateOrZeroExtend(BECount, IntPtrTy);
const SCEV *NumBytesS = SE->getAddExpr(BECount, SE->getConstant(IntPtrTy, 1),
getAnalysisIfAvailable<DominatorTreeWrapperPass>();
DominatorTree *DT = DTWP ? &DTWP->getDomTree() : nullptr;
LoopInfo *LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
- const DataLayout *DL = &L->getHeader()->getModule()->getDataLayout();
const TargetLibraryInfo *TLI =
&getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
auto &AC = getAnalysis<AssumptionCacheTracker>().getAssumptionCache(
WorklistItem Item = VisitStack.pop_back_val();
BasicBlock *BB = Item.getPointer();
bool IsSubloopHeader = Item.getInt();
+ const DataLayout &DL = L->getHeader()->getModule()->getDataLayout();
// Simplify instructions in the current basic block.
for (BasicBlock::iterator BI = BB->begin(), BE = BB->end(); BI != BE;) {
AliasAnalysis *AA;
LoopInfo *LI;
ScalarEvolution *SE;
- const DataLayout *DL;
TargetLibraryInfo *TLI;
DominatorTree *DT;
struct DAGRootTracker {
DAGRootTracker(LoopReroll *Parent, Loop *L, Instruction *IV,
ScalarEvolution *SE, AliasAnalysis *AA,
- TargetLibraryInfo *TLI, const DataLayout *DL)
- : Parent(Parent), L(L), SE(SE), AA(AA), TLI(TLI),
- DL(DL), IV(IV) {
- }
+ TargetLibraryInfo *TLI)
+ : Parent(Parent), L(L), SE(SE), AA(AA), TLI(TLI), IV(IV) {}
/// Stage 1: Find all the DAG roots for the induction variable.
bool findRoots();
ScalarEvolution *SE;
AliasAnalysis *AA;
TargetLibraryInfo *TLI;
- const DataLayout *DL;
// The loop induction variable.
Instruction *IV;
// needed because otherwise isSafeToSpeculativelyExecute returns
// false on PHI nodes.
if (!isa<PHINode>(I) && !isSimpleLoadStore(I) &&
- !isSafeToSpeculativelyExecute(I, DL))
+ !isSafeToSpeculativelyExecute(I))
// Intervening instructions cause side effects.
FutureSideEffects = true;
}
// side effects, and this instruction might also, then we can't reorder
// them, and this matching fails. As an exception, we allow the alias
// set tracker to handle regular (simple) load/store dependencies.
- if (FutureSideEffects &&
- ((!isSimpleLoadStore(BaseInst) &&
- !isSafeToSpeculativelyExecute(BaseInst, DL)) ||
- (!isSimpleLoadStore(RootInst) &&
- !isSafeToSpeculativelyExecute(RootInst, DL)))) {
+ if (FutureSideEffects && ((!isSimpleLoadStore(BaseInst) &&
+ !isSafeToSpeculativelyExecute(BaseInst)) ||
+ (!isSimpleLoadStore(RootInst) &&
+ !isSafeToSpeculativelyExecute(RootInst)))) {
DEBUG(dbgs() << "LRR: iteration root match failed at " << *BaseInst <<
" vs. " << *RootInst <<
" (side effects prevent reordering)\n");
++J;
}
+ const DataLayout &DL = Header->getModule()->getDataLayout();
// We need to create a new induction variable for each different BaseInst.
for (auto &DRS : RootSets) {
SE->getConstant(RealIVSCEV->getType(), 1),
L, SCEV::FlagAnyWrap));
{ // Limit the lifetime of SCEVExpander.
- SCEVExpander Expander(*SE, "reroll");
+ SCEVExpander Expander(*SE, DL, "reroll");
Value *NewIV = Expander.expandCodeFor(H, IV->getType(), Header->begin());
for (auto &KV : Uses) {
}
}
- SimplifyInstructionsInBlock(Header, DL, TLI);
+ SimplifyInstructionsInBlock(Header, TLI);
DeleteDeadPHIs(Header, TLI);
}
bool LoopReroll::reroll(Instruction *IV, Loop *L, BasicBlock *Header,
const SCEV *IterCount,
ReductionTracker &Reductions) {
- DAGRootTracker DAGRoots(this, L, IV, SE, AA, TLI, DL);
+ DAGRootTracker DAGRoots(this, L, IV, SE, AA, TLI);
if (!DAGRoots.findRoots())
return false;
LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
SE = &getAnalysis<ScalarEvolution>();
TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
- DL = &L->getHeader()->getModule()->getDataLayout();
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
BasicBlock *Header = L->getHeader();
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/IntrinsicInst.h"
+#include "llvm/IR/Module.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
for (; PHINode *PN = dyn_cast<PHINode>(I); ++I)
ValueMap[PN] = PN->getIncomingValueForBlock(OrigPreheader);
+ const DataLayout &DL = L->getHeader()->getModule()->getDataLayout();
+
// For the rest of the instructions, either hoist to the OrigPreheader if
// possible or create a clone in the OldPreHeader if not.
TerminatorInst *LoopEntryBranch = OrigPreheader->getTerminator();
// With the operands remapped, see if the instruction constant folds or is
// otherwise simplifyable. This commonly occurs because the entry from PHI
// nodes allows icmps and other instructions to fold.
- // FIXME: Provide DL, TLI, DT, AC to SimplifyInstruction.
- Value *V = SimplifyInstruction(C);
+ // FIXME: Provide TLI, DT, AC to SimplifyInstruction.
+ Value *V = SimplifyInstruction(C, DL);
if (V && LI->replacementPreservesLCSSAForm(C, V)) {
// If so, then delete the temporary instruction and stick the folded value
// in the map.
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
+#include "llvm/IR/Module.h"
#include "llvm/IR/ValueHandle.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
// we can remove them after we are done working.
SmallVector<WeakVH, 16> DeadInsts;
- SCEVExpander Rewriter(SE, "lsr");
+ SCEVExpander Rewriter(SE, L->getHeader()->getModule()->getDataLayout(),
+ "lsr");
#ifndef NDEBUG
Rewriter.setDebugType(DEBUG_TYPE);
#endif
Changed |= DeleteDeadPHIs(L->getHeader());
if (EnablePhiElim && L->isLoopSimplifyForm()) {
SmallVector<WeakVH, 16> DeadInsts;
- SCEVExpander Rewriter(getAnalysis<ScalarEvolution>(), "lsr");
+ const DataLayout &DL = L->getHeader()->getModule()->getDataLayout();
+ SCEVExpander Rewriter(getAnalysis<ScalarEvolution>(), DL, "lsr");
#ifndef NDEBUG
Rewriter.setDebugType(DEBUG_TYPE);
#endif
if (Constant *SimpleRHS = SimplifiedValues.lookup(RHS))
RHS = SimpleRHS;
Value *SimpleV = nullptr;
+ const DataLayout &DL = I.getModule()->getDataLayout();
if (auto FI = dyn_cast<FPMathOperator>(&I))
SimpleV =
- SimplifyFPBinOp(I.getOpcode(), LHS, RHS, FI->getFastMathFlags());
+ SimplifyFPBinOp(I.getOpcode(), LHS, RHS, FI->getFastMathFlags(), DL);
else
- SimpleV = SimplifyBinOp(I.getOpcode(), LHS, RHS);
+ SimpleV = SimplifyBinOp(I.getOpcode(), LHS, RHS, DL);
if (SimpleV && CountedInstructions.insert(&I).second)
NumberOfOptimizedInstructions += TTI.getUserCost(&I);
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/Instructions.h"
+#include "llvm/IR/Module.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
/// pass.
///
void LoopUnswitch::SimplifyCode(std::vector<Instruction*> &Worklist, Loop *L) {
+ const DataLayout &DL = L->getHeader()->getModule()->getDataLayout();
while (!Worklist.empty()) {
Instruction *I = Worklist.back();
Worklist.pop_back();
// See if instruction simplification can hack this up. This is common for
// things like "select false, X, Y" after unswitching made the condition be
// 'false'. TODO: update the domtree properly so we can pass it here.
- if (Value *V = SimplifyInstruction(I))
+ if (Value *V = SimplifyInstruction(I, DL))
if (LI->replacementPreservesLCSSAForm(I, V)) {
ReplaceUsesOfWith(I, V, Worklist, L, LPM);
continue;
STATISTIC(NumCpyToSet, "Number of memcpys converted to memset");
static int64_t GetOffsetFromIndex(const GEPOperator *GEP, unsigned Idx,
- bool &VariableIdxFound, const DataLayout &TD){
+ bool &VariableIdxFound,
+ const DataLayout &DL) {
// Skip over the first indices.
gep_type_iterator GTI = gep_type_begin(GEP);
for (unsigned i = 1; i != Idx; ++i, ++GTI)
// Handle struct indices, which add their field offset to the pointer.
if (StructType *STy = dyn_cast<StructType>(*GTI)) {
- Offset += TD.getStructLayout(STy)->getElementOffset(OpC->getZExtValue());
+ Offset += DL.getStructLayout(STy)->getElementOffset(OpC->getZExtValue());
continue;
}
// Otherwise, we have a sequential type like an array or vector. Multiply
// the index by the ElementSize.
- uint64_t Size = TD.getTypeAllocSize(GTI.getIndexedType());
+ uint64_t Size = DL.getTypeAllocSize(GTI.getIndexedType());
Offset += Size*OpC->getSExtValue();
}
/// constant offset, and return that constant offset. For example, Ptr1 might
/// be &A[42], and Ptr2 might be &A[40]. In this case offset would be -8.
static bool IsPointerOffset(Value *Ptr1, Value *Ptr2, int64_t &Offset,
- const DataLayout &TD) {
+ const DataLayout &DL) {
Ptr1 = Ptr1->stripPointerCasts();
Ptr2 = Ptr2->stripPointerCasts();
// If one pointer is a GEP and the other isn't, then see if the GEP is a
// constant offset from the base, as in "P" and "gep P, 1".
if (GEP1 && !GEP2 && GEP1->getOperand(0)->stripPointerCasts() == Ptr2) {
- Offset = -GetOffsetFromIndex(GEP1, 1, VariableIdxFound, TD);
+ Offset = -GetOffsetFromIndex(GEP1, 1, VariableIdxFound, DL);
return !VariableIdxFound;
}
if (GEP2 && !GEP1 && GEP2->getOperand(0)->stripPointerCasts() == Ptr1) {
- Offset = GetOffsetFromIndex(GEP2, 1, VariableIdxFound, TD);
+ Offset = GetOffsetFromIndex(GEP2, 1, VariableIdxFound, DL);
return !VariableIdxFound;
}
if (GEP1->getOperand(Idx) != GEP2->getOperand(Idx))
break;
- int64_t Offset1 = GetOffsetFromIndex(GEP1, Idx, VariableIdxFound, TD);
- int64_t Offset2 = GetOffsetFromIndex(GEP2, Idx, VariableIdxFound, TD);
+ int64_t Offset1 = GetOffsetFromIndex(GEP1, Idx, VariableIdxFound, DL);
+ int64_t Offset2 = GetOffsetFromIndex(GEP2, Idx, VariableIdxFound, DL);
if (VariableIdxFound) return false;
Offset = Offset2-Offset1;
/// TheStores - The actual stores that make up this range.
SmallVector<Instruction*, 16> TheStores;
- bool isProfitableToUseMemset(const DataLayout &TD) const;
-
+ bool isProfitableToUseMemset(const DataLayout &DL) const;
};
} // end anon namespace
-bool MemsetRange::isProfitableToUseMemset(const DataLayout &TD) const {
+bool MemsetRange::isProfitableToUseMemset(const DataLayout &DL) const {
// If we found more than 4 stores to merge or 16 bytes, use memset.
if (TheStores.size() >= 4 || End-Start >= 16) return true;
// size. If so, check to see whether we will end up actually reducing the
// number of stores used.
unsigned Bytes = unsigned(End-Start);
- unsigned MaxIntSize = TD.getLargestLegalIntTypeSize();
+ unsigned MaxIntSize = DL.getLargestLegalIntTypeSize();
if (MaxIntSize == 0)
MaxIntSize = 1;
unsigned NumPointerStores = Bytes / MaxIntSize;
class MemCpyOpt : public FunctionPass {
MemoryDependenceAnalysis *MD;
TargetLibraryInfo *TLI;
- const DataLayout *DL;
public:
static char ID; // Pass identification, replacement for typeid
MemCpyOpt() : FunctionPass(ID) {
initializeMemCpyOptPass(*PassRegistry::getPassRegistry());
MD = nullptr;
TLI = nullptr;
- DL = nullptr;
}
bool runOnFunction(Function &F) override;
/// attempts to merge them together into a memcpy/memset.
Instruction *MemCpyOpt::tryMergingIntoMemset(Instruction *StartInst,
Value *StartPtr, Value *ByteVal) {
- if (!DL) return nullptr;
+ const DataLayout &DL = StartInst->getModule()->getDataLayout();
// Okay, so we now have a single store that can be splatable. Scan to find
// all subsequent stores of the same value to offset from the same pointer.
// Join these together into ranges, so we can decide whether contiguous blocks
// are stored.
- MemsetRanges Ranges(*DL);
+ MemsetRanges Ranges(DL);
BasicBlock::iterator BI = StartInst;
for (++BI; !isa<TerminatorInst>(BI); ++BI) {
// Check to see if this store is to a constant offset from the start ptr.
int64_t Offset;
- if (!IsPointerOffset(StartPtr, NextStore->getPointerOperand(),
- Offset, *DL))
+ if (!IsPointerOffset(StartPtr, NextStore->getPointerOperand(), Offset,
+ DL))
break;
Ranges.addStore(Offset, NextStore);
// Check to see if this store is to a constant offset from the start ptr.
int64_t Offset;
- if (!IsPointerOffset(StartPtr, MSI->getDest(), Offset, *DL))
+ if (!IsPointerOffset(StartPtr, MSI->getDest(), Offset, DL))
break;
Ranges.addMemSet(Offset, MSI);
if (Range.TheStores.size() == 1) continue;
// If it is profitable to lower this range to memset, do so now.
- if (!Range.isProfitableToUseMemset(*DL))
+ if (!Range.isProfitableToUseMemset(DL))
continue;
// Otherwise, we do want to transform this! Create a new memset.
if (Alignment == 0) {
Type *EltType =
cast<PointerType>(StartPtr->getType())->getElementType();
- Alignment = DL->getABITypeAlignment(EltType);
+ Alignment = DL.getABITypeAlignment(EltType);
}
AMemSet =
bool MemCpyOpt::processStore(StoreInst *SI, BasicBlock::iterator &BBI) {
if (!SI->isSimple()) return false;
-
- if (!DL) return false;
+ const DataLayout &DL = SI->getModule()->getDataLayout();
// Detect cases where we're performing call slot forwarding, but
// happen to be using a load-store pair to implement it, rather than
if (C) {
unsigned storeAlign = SI->getAlignment();
if (!storeAlign)
- storeAlign = DL->getABITypeAlignment(SI->getOperand(0)->getType());
+ storeAlign = DL.getABITypeAlignment(SI->getOperand(0)->getType());
unsigned loadAlign = LI->getAlignment();
if (!loadAlign)
- loadAlign = DL->getABITypeAlignment(LI->getType());
+ loadAlign = DL.getABITypeAlignment(LI->getType());
- bool changed = performCallSlotOptzn(LI,
- SI->getPointerOperand()->stripPointerCasts(),
- LI->getPointerOperand()->stripPointerCasts(),
- DL->getTypeStoreSize(SI->getOperand(0)->getType()),
- std::min(storeAlign, loadAlign), C);
+ bool changed = performCallSlotOptzn(
+ LI, SI->getPointerOperand()->stripPointerCasts(),
+ LI->getPointerOperand()->stripPointerCasts(),
+ DL.getTypeStoreSize(SI->getOperand(0)->getType()),
+ std::min(storeAlign, loadAlign), C);
if (changed) {
MD->removeInstruction(SI);
SI->eraseFromParent();
if (!srcAlloca)
return false;
- // Check that all of src is copied to dest.
- if (!DL) return false;
-
ConstantInt *srcArraySize = dyn_cast<ConstantInt>(srcAlloca->getArraySize());
if (!srcArraySize)
return false;
- uint64_t srcSize = DL->getTypeAllocSize(srcAlloca->getAllocatedType()) *
- srcArraySize->getZExtValue();
+ const DataLayout &DL = cpy->getModule()->getDataLayout();
+ uint64_t srcSize = DL.getTypeAllocSize(srcAlloca->getAllocatedType()) *
+ srcArraySize->getZExtValue();
if (cpyLen < srcSize)
return false;
if (!destArraySize)
return false;
- uint64_t destSize = DL->getTypeAllocSize(A->getAllocatedType()) *
- destArraySize->getZExtValue();
+ uint64_t destSize = DL.getTypeAllocSize(A->getAllocatedType()) *
+ destArraySize->getZExtValue();
if (destSize < srcSize)
return false;
return false;
}
- uint64_t destSize = DL->getTypeAllocSize(StructTy);
+ uint64_t destSize = DL.getTypeAllocSize(StructTy);
if (destSize < srcSize)
return false;
}
// Check that dest points to memory that is at least as aligned as src.
unsigned srcAlign = srcAlloca->getAlignment();
if (!srcAlign)
- srcAlign = DL->getABITypeAlignment(srcAlloca->getAllocatedType());
+ srcAlign = DL.getABITypeAlignment(srcAlloca->getAllocatedType());
bool isDestSufficientlyAligned = srcAlign <= cpyAlign;
// If dest is not aligned enough and we can't increase its alignment then
// bail out.
/// processByValArgument - This is called on every byval argument in call sites.
bool MemCpyOpt::processByValArgument(CallSite CS, unsigned ArgNo) {
- if (!DL) return false;
-
+ const DataLayout &DL = CS.getCaller()->getParent()->getDataLayout();
// Find out what feeds this byval argument.
Value *ByValArg = CS.getArgument(ArgNo);
Type *ByValTy = cast<PointerType>(ByValArg->getType())->getElementType();
- uint64_t ByValSize = DL->getTypeAllocSize(ByValTy);
+ uint64_t ByValSize = DL.getTypeAllocSize(ByValTy);
MemDepResult DepInfo =
MD->getPointerDependencyFrom(AliasAnalysis::Location(ByValArg, ByValSize),
true, CS.getInstruction(),
*CS->getParent()->getParent());
DominatorTree &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
if (MDep->getAlignment() < ByValAlign &&
- getOrEnforceKnownAlignment(MDep->getSource(), ByValAlign, DL, &AC,
- CS.getInstruction(), &DT) < ByValAlign)
+ getOrEnforceKnownAlignment(MDep->getSource(), ByValAlign, DL,
+ CS.getInstruction(), &AC, &DT) < ByValAlign)
return false;
// Verify that the copied-from memory doesn't change in between the memcpy and
bool MadeChange = false;
MD = &getAnalysis<MemoryDependenceAnalysis>();
- DL = &F.getParent()->getDataLayout();
TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
// If we don't have at least memset and memcpy, there is little point of doing
}
// Transform load from a constant into a constant if possible.
- if (Constant *C = ConstantFoldLoadFromConstPtr(Ptr, &DL))
+ if (Constant *C = ConstantFoldLoadFromConstPtr(Ptr, DL))
return markConstant(IV, &I, C);
// Otherwise we cannot say for certain what value this load will produce.
// by writing out the code here where we have tho underlying allocation
// size readily available.
APInt GEPOffset = Offset;
+ const DataLayout &DL = GEPI.getModule()->getDataLayout();
for (gep_type_iterator GTI = gep_type_begin(GEPI),
GTE = gep_type_end(GEPI);
GTI != GTE; ++GTI) {
if (!IsOffsetKnown)
return PI.setAborted(&LI);
+ const DataLayout &DL = LI.getModule()->getDataLayout();
uint64_t Size = DL.getTypeStoreSize(LI.getType());
return handleLoadOrStore(LI.getType(), LI, Offset, Size, LI.isVolatile());
}
if (!IsOffsetKnown)
return PI.setAborted(&SI);
+ const DataLayout &DL = SI.getModule()->getDataLayout();
uint64_t Size = DL.getTypeStoreSize(ValOp->getType());
// If this memory access can be shown to *statically* extend outside the
SmallVector<std::pair<Instruction *, Instruction *>, 4> Uses;
Visited.insert(Root);
Uses.push_back(std::make_pair(cast<Instruction>(*U), Root));
+ const DataLayout &DL = Root->getModule()->getDataLayout();
// If there are no loads or stores, the access is dead. We mark that as
// a size zero access.
Size = 0;
const bool RequiresDomTree;
LLVMContext *C;
- const DataLayout *DL;
DominatorTree *DT;
AssumptionCache *AC;
public:
SROA(bool RequiresDomTree = true)
: FunctionPass(ID), RequiresDomTree(RequiresDomTree), C(nullptr),
- DL(nullptr), DT(nullptr) {
+ DT(nullptr) {
initializeSROAPass(*PassRegistry::getPassRegistry());
}
bool runOnFunction(Function &F) override;
///
/// FIXME: This should be hoisted into a generic utility, likely in
/// Transforms/Util/Local.h
-static bool isSafePHIToSpeculate(PHINode &PN, const DataLayout *DL = nullptr) {
+static bool isSafePHIToSpeculate(PHINode &PN) {
// For now, we can only do this promotion if the load is in the same block
// as the PHI, and if there are no stores between the phi and load.
// TODO: Allow recursive phi users.
if (!HaveLoad)
return false;
+ const DataLayout &DL = PN.getModule()->getDataLayout();
+
// We can only transform this if it is safe to push the loads into the
// predecessor blocks. The only thing to watch out for is that we can't put
// a possibly trapping load in the predecessor if it is a critical edge.
// is already a load in the block, then we can move the load to the pred
// block.
if (InVal->isDereferenceablePointer(DL) ||
- isSafeToLoadUnconditionally(InVal, TI, MaxAlign, DL))
+ isSafeToLoadUnconditionally(InVal, TI, MaxAlign))
continue;
return false;
///
/// We can do this to a select if its only uses are loads and if the operand
/// to the select can be loaded unconditionally.
-static bool isSafeSelectToSpeculate(SelectInst &SI,
- const DataLayout *DL = nullptr) {
+static bool isSafeSelectToSpeculate(SelectInst &SI) {
Value *TValue = SI.getTrueValue();
Value *FValue = SI.getFalseValue();
+ const DataLayout &DL = SI.getModule()->getDataLayout();
bool TDerefable = TValue->isDereferenceablePointer(DL);
bool FDerefable = FValue->isDereferenceablePointer(DL);
// absolutely (e.g. allocas) or at this point because we can see other
// accesses to it.
if (!TDerefable &&
- !isSafeToLoadUnconditionally(TValue, LI, LI->getAlignment(), DL))
+ !isSafeToLoadUnconditionally(TValue, LI, LI->getAlignment()))
return false;
if (!FDerefable &&
- !isSafeToLoadUnconditionally(FValue, LI, LI->getAlignment(), DL))
+ !isSafeToLoadUnconditionally(FValue, LI, LI->getAlignment()))
return false;
}
// them to the alloca slices.
SmallDenseMap<LoadInst *, std::vector<LoadInst *>, 1> SplitLoadsMap;
std::vector<LoadInst *> SplitLoads;
+ const DataLayout &DL = AI.getModule()->getDataLayout();
for (LoadInst *LI : Loads) {
SplitLoads.clear();
auto *PartTy = Type::getIntNTy(Ty->getContext(), PartSize * 8);
auto *PartPtrTy = PartTy->getPointerTo(LI->getPointerAddressSpace());
LoadInst *PLoad = IRB.CreateAlignedLoad(
- getAdjustedPtr(IRB, *DL, BasePtr,
- APInt(DL->getPointerSizeInBits(), PartOffset),
+ getAdjustedPtr(IRB, DL, BasePtr,
+ APInt(DL.getPointerSizeInBits(), PartOffset),
PartPtrTy, BasePtr->getName() + "."),
- getAdjustedAlignment(LI, PartOffset, *DL), /*IsVolatile*/ false,
+ getAdjustedAlignment(LI, PartOffset, DL), /*IsVolatile*/ false,
LI->getName());
// Append this load onto the list of split loads so we can find it later
PLoad->getType()->getPointerTo(SI->getPointerAddressSpace());
StoreInst *PStore = IRB.CreateAlignedStore(
- PLoad, getAdjustedPtr(IRB, *DL, StoreBasePtr,
- APInt(DL->getPointerSizeInBits(), PartOffset),
+ PLoad, getAdjustedPtr(IRB, DL, StoreBasePtr,
+ APInt(DL.getPointerSizeInBits(), PartOffset),
PartPtrTy, StoreBasePtr->getName() + "."),
- getAdjustedAlignment(SI, PartOffset, *DL), /*IsVolatile*/ false);
+ getAdjustedAlignment(SI, PartOffset, DL), /*IsVolatile*/ false);
(void)PStore;
DEBUG(dbgs() << " +" << PartOffset << ":" << *PStore << "\n");
}
} else {
IRB.SetInsertPoint(BasicBlock::iterator(LI));
PLoad = IRB.CreateAlignedLoad(
- getAdjustedPtr(IRB, *DL, LoadBasePtr,
- APInt(DL->getPointerSizeInBits(), PartOffset),
+ getAdjustedPtr(IRB, DL, LoadBasePtr,
+ APInt(DL.getPointerSizeInBits(), PartOffset),
PartPtrTy, LoadBasePtr->getName() + "."),
- getAdjustedAlignment(LI, PartOffset, *DL), /*IsVolatile*/ false,
+ getAdjustedAlignment(LI, PartOffset, DL), /*IsVolatile*/ false,
LI->getName());
}
// And store this partition.
IRB.SetInsertPoint(BasicBlock::iterator(SI));
StoreInst *PStore = IRB.CreateAlignedStore(
- PLoad, getAdjustedPtr(IRB, *DL, StoreBasePtr,
- APInt(DL->getPointerSizeInBits(), PartOffset),
+ PLoad, getAdjustedPtr(IRB, DL, StoreBasePtr,
+ APInt(DL.getPointerSizeInBits(), PartOffset),
PartPtrTy, StoreBasePtr->getName() + "."),
- getAdjustedAlignment(SI, PartOffset, *DL), /*IsVolatile*/ false);
+ getAdjustedAlignment(SI, PartOffset, DL), /*IsVolatile*/ false);
// Now build a new slice for the alloca.
NewSlices.push_back(
// won't always succeed, in which case we fall back to a legal integer type
// or an i8 array of an appropriate size.
Type *SliceTy = nullptr;
+ const DataLayout &DL = AI.getModule()->getDataLayout();
if (Type *CommonUseTy = findCommonType(P.begin(), P.end(), P.endOffset()))
- if (DL->getTypeAllocSize(CommonUseTy) >= P.size())
+ if (DL.getTypeAllocSize(CommonUseTy) >= P.size())
SliceTy = CommonUseTy;
if (!SliceTy)
- if (Type *TypePartitionTy = getTypePartition(*DL, AI.getAllocatedType(),
+ if (Type *TypePartitionTy = getTypePartition(DL, AI.getAllocatedType(),
P.beginOffset(), P.size()))
SliceTy = TypePartitionTy;
if ((!SliceTy || (SliceTy->isArrayTy() &&
SliceTy->getArrayElementType()->isIntegerTy())) &&
- DL->isLegalInteger(P.size() * 8))
+ DL.isLegalInteger(P.size() * 8))
SliceTy = Type::getIntNTy(*C, P.size() * 8);
if (!SliceTy)
SliceTy = ArrayType::get(Type::getInt8Ty(*C), P.size());
- assert(DL->getTypeAllocSize(SliceTy) >= P.size());
+ assert(DL.getTypeAllocSize(SliceTy) >= P.size());
- bool IsIntegerPromotable = isIntegerWideningViable(P, SliceTy, *DL);
+ bool IsIntegerPromotable = isIntegerWideningViable(P, SliceTy, DL);
VectorType *VecTy =
- IsIntegerPromotable ? nullptr : isVectorPromotionViable(P, *DL);
+ IsIntegerPromotable ? nullptr : isVectorPromotionViable(P, DL);
if (VecTy)
SliceTy = VecTy;
// The minimum alignment which users can rely on when the explicit
// alignment is omitted or zero is that required by the ABI for this
// type.
- Alignment = DL->getABITypeAlignment(AI.getAllocatedType());
+ Alignment = DL.getABITypeAlignment(AI.getAllocatedType());
}
Alignment = MinAlign(Alignment, P.beginOffset());
// If we will get at least this much alignment from the type alone, leave
// the alloca's alignment unconstrained.
- if (Alignment <= DL->getABITypeAlignment(SliceTy))
+ if (Alignment <= DL.getABITypeAlignment(SliceTy))
Alignment = 0;
NewAI = new AllocaInst(
SliceTy, nullptr, Alignment,
SmallPtrSet<PHINode *, 8> PHIUsers;
SmallPtrSet<SelectInst *, 8> SelectUsers;
- AllocaSliceRewriter Rewriter(*DL, AS, *this, AI, *NewAI, P.beginOffset(),
+ AllocaSliceRewriter Rewriter(DL, AS, *this, AI, *NewAI, P.beginOffset(),
P.endOffset(), IsIntegerPromotable, VecTy,
PHIUsers, SelectUsers);
bool Promotable = true;
for (SmallPtrSetImpl<PHINode *>::iterator I = PHIUsers.begin(),
E = PHIUsers.end();
I != E; ++I)
- if (!isSafePHIToSpeculate(**I, DL)) {
+ if (!isSafePHIToSpeculate(**I)) {
Promotable = false;
PHIUsers.clear();
SelectUsers.clear();
for (SmallPtrSetImpl<SelectInst *>::iterator I = SelectUsers.begin(),
E = SelectUsers.end();
I != E; ++I)
- if (!isSafeSelectToSpeculate(**I, DL)) {
+ if (!isSafeSelectToSpeculate(**I)) {
Promotable = false;
PHIUsers.clear();
SelectUsers.clear();
unsigned NumPartitions = 0;
bool Changed = false;
+ const DataLayout &DL = AI.getModule()->getDataLayout();
// First try to pre-split loads and stores.
Changed |= presplitLoadsAndStores(AI, AS);
// confident that the above handling of splittable loads and stores is
// completely sufficient before we forcibly disable the remaining handling.
if (S.beginOffset() == 0 &&
- S.endOffset() >= DL->getTypeAllocSize(AI.getAllocatedType()))
+ S.endOffset() >= DL.getTypeAllocSize(AI.getAllocatedType()))
continue;
if (isa<LoadInst>(S.getUse()->getUser()) ||
isa<StoreInst>(S.getUse()->getUser())) {
Changed = true;
if (NewAI != &AI) {
uint64_t SizeOfByte = 8;
- uint64_t AllocaSize = DL->getTypeSizeInBits(NewAI->getAllocatedType());
+ uint64_t AllocaSize = DL.getTypeSizeInBits(NewAI->getAllocatedType());
// Don't include any padding.
uint64_t Size = std::min(AllocaSize, P.size() * SizeOfByte);
Pieces.push_back(Piece(NewAI, P.beginOffset() * SizeOfByte, Size));
AI.eraseFromParent();
return true;
}
+ const DataLayout &DL = AI.getModule()->getDataLayout();
// Skip alloca forms that this analysis can't handle.
if (AI.isArrayAllocation() || !AI.getAllocatedType()->isSized() ||
- DL->getTypeAllocSize(AI.getAllocatedType()) == 0)
+ DL.getTypeAllocSize(AI.getAllocatedType()) == 0)
return false;
bool Changed = false;
// First, split any FCA loads and stores touching this alloca to promote
// better splitting and promotion opportunities.
- AggLoadStoreRewriter AggRewriter(*DL);
+ AggLoadStoreRewriter AggRewriter(DL);
Changed |= AggRewriter.rewrite(AI);
// Build the slices using a recursive instruction-visiting builder.
- AllocaSlices AS(*DL, AI);
+ AllocaSlices AS(DL, AI);
DEBUG(AS.print(dbgs()));
if (AS.isEscaped())
return Changed;
DEBUG(dbgs() << "SROA function: " << F.getName() << "\n");
C = &F.getContext();
- DL = &F.getParent()->getDataLayout();
DominatorTreeWrapperPass *DTWP =
getAnalysisIfAvailable<DominatorTreeWrapperPass>();
DT = DTWP ? &DTWP->getDomTree() : nullptr;
private:
bool HasDomTree;
- const DataLayout *DL;
/// DeadInsts - Keep track of instructions we have made dead, so that
/// we can remove them after we are done working.
void isSafeMemAccess(uint64_t Offset, uint64_t MemSize,
Type *MemOpType, bool isStore, AllocaInfo &Info,
Instruction *TheAccess, bool AllowWholeAccess);
- bool TypeHasComponent(Type *T, uint64_t Offset, uint64_t Size);
- uint64_t FindElementAndOffset(Type *&T, uint64_t &Offset,
- Type *&IdxTy);
+ bool TypeHasComponent(Type *T, uint64_t Offset, uint64_t Size,
+ const DataLayout &DL);
+ uint64_t FindElementAndOffset(Type *&T, uint64_t &Offset, Type *&IdxTy,
+ const DataLayout &DL);
void DoScalarReplacement(AllocaInst *AI,
std::vector<AllocaInst*> &WorkList);
// If the source and destination are both to the same alloca, then this is
// a noop copy-to-self, just delete it. Otherwise, emit a load and store
// as appropriate.
- AllocaInst *OrigAI = cast<AllocaInst>(GetUnderlyingObject(Ptr, &DL, 0));
+ AllocaInst *OrigAI = cast<AllocaInst>(GetUnderlyingObject(Ptr, DL, 0));
- if (GetUnderlyingObject(MTI->getSource(), &DL, 0) != OrigAI) {
+ if (GetUnderlyingObject(MTI->getSource(), DL, 0) != OrigAI) {
// Dest must be OrigAI, change this to be a load from the original
// pointer (bitcasted), then a store to our new alloca.
assert(MTI->getRawDest() == Ptr && "Neither use is of pointer?");
LoadInst *SrcVal = Builder.CreateLoad(SrcPtr, "srcval");
SrcVal->setAlignment(MTI->getAlignment());
Builder.CreateStore(SrcVal, NewAI);
- } else if (GetUnderlyingObject(MTI->getDest(), &DL, 0) != OrigAI) {
+ } else if (GetUnderlyingObject(MTI->getDest(), DL, 0) != OrigAI) {
// Src must be OrigAI, change this to be a load from NewAI then a store
// through the original dest pointer (bitcasted).
assert(MTI->getRawSource() == Ptr && "Neither use is of pointer?");
if (skipOptnoneFunction(F))
return false;
- DL = &F.getParent()->getDataLayout();
-
bool Changed = performPromotion(F);
- // FIXME: ScalarRepl currently depends on DataLayout more than it
- // theoretically needs to. It should be refactored in order to support
- // target-independent IR. Until this is done, just skip the actual
- // scalar-replacement portion of this pass.
- if (!DL) return Changed;
-
while (1) {
bool LocalChange = performScalarRepl(F);
if (!LocalChange) break; // No need to repromote if no scalarrepl
///
/// We can do this to a select if its only uses are loads and if the operand to
/// the select can be loaded unconditionally.
-static bool isSafeSelectToSpeculate(SelectInst *SI, const DataLayout *DL) {
+static bool isSafeSelectToSpeculate(SelectInst *SI) {
+ const DataLayout &DL = SI->getModule()->getDataLayout();
bool TDerefable = SI->getTrueValue()->isDereferenceablePointer(DL);
bool FDerefable = SI->getFalseValue()->isDereferenceablePointer(DL);
// Both operands to the select need to be dereferencable, either absolutely
// (e.g. allocas) or at this point because we can see other accesses to it.
- if (!TDerefable && !isSafeToLoadUnconditionally(SI->getTrueValue(), LI,
- LI->getAlignment(), DL))
+ if (!TDerefable &&
+ !isSafeToLoadUnconditionally(SI->getTrueValue(), LI,
+ LI->getAlignment()))
return false;
- if (!FDerefable && !isSafeToLoadUnconditionally(SI->getFalseValue(), LI,
- LI->getAlignment(), DL))
+ if (!FDerefable &&
+ !isSafeToLoadUnconditionally(SI->getFalseValue(), LI,
+ LI->getAlignment()))
return false;
}
///
/// We can do this to a select if its only uses are loads and if the operand to
/// the select can be loaded unconditionally.
-static bool isSafePHIToSpeculate(PHINode *PN, const DataLayout *DL) {
+static bool isSafePHIToSpeculate(PHINode *PN) {
// For now, we can only do this promotion if the load is in the same block as
// the PHI, and if there are no stores between the phi and load.
// TODO: Allow recursive phi users.
MaxAlign = std::max(MaxAlign, LI->getAlignment());
}
+ const DataLayout &DL = PN->getModule()->getDataLayout();
+
// Okay, we know that we have one or more loads in the same block as the PHI.
// We can transform this if it is safe to push the loads into the predecessor
// blocks. The only thing to watch out for is that we can't put a possibly
// If this pointer is always safe to load, or if we can prove that there is
// already a load in the block, then we can move the load to the pred block.
if (InVal->isDereferenceablePointer(DL) ||
- isSafeToLoadUnconditionally(InVal, Pred->getTerminator(), MaxAlign, DL))
+ isSafeToLoadUnconditionally(InVal, Pred->getTerminator(), MaxAlign))
continue;
return false;
/// direct (non-volatile) loads and stores to it. If the alloca is close but
/// not quite there, this will transform the code to allow promotion. As such,
/// it is a non-pure predicate.
-static bool tryToMakeAllocaBePromotable(AllocaInst *AI, const DataLayout *DL) {
+static bool tryToMakeAllocaBePromotable(AllocaInst *AI, const DataLayout &DL) {
SetVector<Instruction*, SmallVector<Instruction*, 4>,
SmallPtrSet<Instruction*, 4> > InstsToRewrite;
for (User *U : AI->users()) {
// If it is safe to turn "load (select c, AI, ptr)" into a select of two
// loads, then we can transform this by rewriting the select.
- if (!isSafeSelectToSpeculate(SI, DL))
+ if (!isSafeSelectToSpeculate(SI))
return false;
InstsToRewrite.insert(SI);
// If it is safe to turn "load (phi [AI, ptr, ...])" into a PHI of loads
// in the pred blocks, then we can transform this by rewriting the PHI.
- if (!isSafePHIToSpeculate(PN, DL))
+ if (!isSafePHIToSpeculate(PN))
return false;
InstsToRewrite.insert(PN);
bool SROA::performPromotion(Function &F) {
std::vector<AllocaInst*> Allocas;
+ const DataLayout &DL = F.getParent()->getDataLayout();
DominatorTree *DT = nullptr;
if (HasDomTree)
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
//
bool SROA::performScalarRepl(Function &F) {
std::vector<AllocaInst*> WorkList;
+ const DataLayout &DL = F.getParent()->getDataLayout();
// Scan the entry basic block, adding allocas to the worklist.
BasicBlock &BB = F.getEntryBlock();
// transform the allocation instruction if it is an array allocation
// (allocations OF arrays are ok though), and an allocation of a scalar
// value cannot be decomposed at all.
- uint64_t AllocaSize = DL->getTypeAllocSize(AI->getAllocatedType());
+ uint64_t AllocaSize = DL.getTypeAllocSize(AI->getAllocatedType());
// Do not promote [0 x %struct].
if (AllocaSize == 0) continue;
// promoted itself. If so, we don't want to transform it needlessly. Note
// that we can't just check based on the type: the alloca may be of an i32
// but that has pointer arithmetic to set byte 3 of it or something.
- if (AllocaInst *NewAI = ConvertToScalarInfo(
- (unsigned)AllocaSize, *DL, ScalarLoadThreshold).TryConvert(AI)) {
+ if (AllocaInst *NewAI =
+ ConvertToScalarInfo((unsigned)AllocaSize, DL, ScalarLoadThreshold)
+ .TryConvert(AI)) {
NewAI->takeName(AI);
AI->eraseFromParent();
++NumConverted;
/// referenced by this instruction.
void SROA::isSafeForScalarRepl(Instruction *I, uint64_t Offset,
AllocaInfo &Info) {
+ const DataLayout &DL = I->getModule()->getDataLayout();
for (Use &U : I->uses()) {
Instruction *User = cast<Instruction>(U.getUser());
if (!LI->isSimple())
return MarkUnsafe(Info, User);
Type *LIType = LI->getType();
- isSafeMemAccess(Offset, DL->getTypeAllocSize(LIType),
- LIType, false, Info, LI, true /*AllowWholeAccess*/);
+ isSafeMemAccess(Offset, DL.getTypeAllocSize(LIType), LIType, false, Info,
+ LI, true /*AllowWholeAccess*/);
Info.hasALoadOrStore = true;
} else if (StoreInst *SI = dyn_cast<StoreInst>(User)) {
return MarkUnsafe(Info, User);
Type *SIType = SI->getOperand(0)->getType();
- isSafeMemAccess(Offset, DL->getTypeAllocSize(SIType),
- SIType, true, Info, SI, true /*AllowWholeAccess*/);
+ isSafeMemAccess(Offset, DL.getTypeAllocSize(SIType), SIType, true, Info,
+ SI, true /*AllowWholeAccess*/);
Info.hasALoadOrStore = true;
} else if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(User)) {
if (II->getIntrinsicID() != Intrinsic::lifetime_start &&
if (!Info.CheckedPHIs.insert(PN).second)
return;
+ const DataLayout &DL = I->getModule()->getDataLayout();
for (User *U : I->users()) {
Instruction *UI = cast<Instruction>(U);
if (!LI->isSimple())
return MarkUnsafe(Info, UI);
Type *LIType = LI->getType();
- isSafeMemAccess(Offset, DL->getTypeAllocSize(LIType),
- LIType, false, Info, LI, false /*AllowWholeAccess*/);
+ isSafeMemAccess(Offset, DL.getTypeAllocSize(LIType), LIType, false, Info,
+ LI, false /*AllowWholeAccess*/);
Info.hasALoadOrStore = true;
} else if (StoreInst *SI = dyn_cast<StoreInst>(UI)) {
return MarkUnsafe(Info, UI);
Type *SIType = SI->getOperand(0)->getType();
- isSafeMemAccess(Offset, DL->getTypeAllocSize(SIType),
- SIType, true, Info, SI, false /*AllowWholeAccess*/);
+ isSafeMemAccess(Offset, DL.getTypeAllocSize(SIType), SIType, true, Info,
+ SI, false /*AllowWholeAccess*/);
Info.hasALoadOrStore = true;
} else if (isa<PHINode>(UI) || isa<SelectInst>(UI)) {
isSafePHISelectUseForScalarRepl(UI, Offset, Info);
// constant part of the offset.
if (NonConstant)
Indices.pop_back();
- Offset += DL->getIndexedOffset(GEPI->getPointerOperandType(), Indices);
- if (!TypeHasComponent(Info.AI->getAllocatedType(), Offset,
- NonConstantIdxSize))
+
+ const DataLayout &DL = GEPI->getModule()->getDataLayout();
+ Offset += DL.getIndexedOffset(GEPI->getPointerOperandType(), Indices);
+ if (!TypeHasComponent(Info.AI->getAllocatedType(), Offset, NonConstantIdxSize,
+ DL))
MarkUnsafe(Info, GEPI);
}
Type *MemOpType, bool isStore,
AllocaInfo &Info, Instruction *TheAccess,
bool AllowWholeAccess) {
+ const DataLayout &DL = TheAccess->getModule()->getDataLayout();
// Check if this is a load/store of the entire alloca.
if (Offset == 0 && AllowWholeAccess &&
- MemSize == DL->getTypeAllocSize(Info.AI->getAllocatedType())) {
+ MemSize == DL.getTypeAllocSize(Info.AI->getAllocatedType())) {
// This can be safe for MemIntrinsics (where MemOpType is 0) and integer
// loads/stores (which are essentially the same as the MemIntrinsics with
// regard to copying padding between elements). But, if an alloca is
}
// Check if the offset/size correspond to a component within the alloca type.
Type *T = Info.AI->getAllocatedType();
- if (TypeHasComponent(T, Offset, MemSize)) {
+ if (TypeHasComponent(T, Offset, MemSize, DL)) {
Info.hasSubelementAccess = true;
return;
}
/// TypeHasComponent - Return true if T has a component type with the
/// specified offset and size. If Size is zero, do not check the size.
-bool SROA::TypeHasComponent(Type *T, uint64_t Offset, uint64_t Size) {
+bool SROA::TypeHasComponent(Type *T, uint64_t Offset, uint64_t Size,
+ const DataLayout &DL) {
Type *EltTy;
uint64_t EltSize;
if (StructType *ST = dyn_cast<StructType>(T)) {
- const StructLayout *Layout = DL->getStructLayout(ST);
+ const StructLayout *Layout = DL.getStructLayout(ST);
unsigned EltIdx = Layout->getElementContainingOffset(Offset);
EltTy = ST->getContainedType(EltIdx);
- EltSize = DL->getTypeAllocSize(EltTy);
+ EltSize = DL.getTypeAllocSize(EltTy);
Offset -= Layout->getElementOffset(EltIdx);
} else if (ArrayType *AT = dyn_cast<ArrayType>(T)) {
EltTy = AT->getElementType();
- EltSize = DL->getTypeAllocSize(EltTy);
+ EltSize = DL.getTypeAllocSize(EltTy);
if (Offset >= AT->getNumElements() * EltSize)
return false;
Offset %= EltSize;
} else if (VectorType *VT = dyn_cast<VectorType>(T)) {
EltTy = VT->getElementType();
- EltSize = DL->getTypeAllocSize(EltTy);
+ EltSize = DL.getTypeAllocSize(EltTy);
if (Offset >= VT->getNumElements() * EltSize)
return false;
Offset %= EltSize;
// Check if the component spans multiple elements.
if (Offset + Size > EltSize)
return false;
- return TypeHasComponent(EltTy, Offset, Size);
+ return TypeHasComponent(EltTy, Offset, Size, DL);
}
/// RewriteForScalarRepl - Alloca AI is being split into NewElts, so rewrite
/// instruction.
void SROA::RewriteForScalarRepl(Instruction *I, AllocaInst *AI, uint64_t Offset,
SmallVectorImpl<AllocaInst *> &NewElts) {
+ const DataLayout &DL = I->getModule()->getDataLayout();
for (Value::use_iterator UI = I->use_begin(), E = I->use_end(); UI!=E;) {
Use &TheUse = *UI++;
Instruction *User = cast<Instruction>(TheUse.getUser());
if (MemIntrinsic *MI = dyn_cast<MemIntrinsic>(User)) {
ConstantInt *Length = dyn_cast<ConstantInt>(MI->getLength());
uint64_t MemSize = Length->getZExtValue();
- if (Offset == 0 &&
- MemSize == DL->getTypeAllocSize(AI->getAllocatedType()))
+ if (Offset == 0 && MemSize == DL.getTypeAllocSize(AI->getAllocatedType()))
RewriteMemIntrinUserOfAlloca(MI, I, AI, NewElts);
// Otherwise the intrinsic can only touch a single element and the
// address operand will be updated, so nothing else needs to be done.
LI->replaceAllUsesWith(Insert);
DeadInsts.push_back(LI);
} else if (LIType->isIntegerTy() &&
- DL->getTypeAllocSize(LIType) ==
- DL->getTypeAllocSize(AI->getAllocatedType())) {
+ DL.getTypeAllocSize(LIType) ==
+ DL.getTypeAllocSize(AI->getAllocatedType())) {
// If this is a load of the entire alloca to an integer, rewrite it.
RewriteLoadUserOfWholeAlloca(LI, AI, NewElts);
}
}
DeadInsts.push_back(SI);
} else if (SIType->isIntegerTy() &&
- DL->getTypeAllocSize(SIType) ==
- DL->getTypeAllocSize(AI->getAllocatedType())) {
+ DL.getTypeAllocSize(SIType) ==
+ DL.getTypeAllocSize(AI->getAllocatedType())) {
// If this is a store of the entire alloca from an integer, rewrite it.
RewriteStoreUserOfWholeAlloca(SI, AI, NewElts);
}
Type *T = AI->getAllocatedType();
uint64_t EltOffset = 0;
Type *IdxTy;
- uint64_t Idx = FindElementAndOffset(T, EltOffset, IdxTy);
+ uint64_t Idx = FindElementAndOffset(T, EltOffset, IdxTy,
+ BC->getModule()->getDataLayout());
Instruction *Val = NewElts[Idx];
if (Val->getType() != BC->getDestTy()) {
Val = new BitCastInst(Val, BC->getDestTy(), "", BC);
/// Sets T to the type of the element and Offset to the offset within that
/// element. IdxTy is set to the type of the index result to be used in a
/// GEP instruction.
-uint64_t SROA::FindElementAndOffset(Type *&T, uint64_t &Offset,
- Type *&IdxTy) {
+uint64_t SROA::FindElementAndOffset(Type *&T, uint64_t &Offset, Type *&IdxTy,
+ const DataLayout &DL) {
uint64_t Idx = 0;
+
if (StructType *ST = dyn_cast<StructType>(T)) {
- const StructLayout *Layout = DL->getStructLayout(ST);
+ const StructLayout *Layout = DL.getStructLayout(ST);
Idx = Layout->getElementContainingOffset(Offset);
T = ST->getContainedType(Idx);
Offset -= Layout->getElementOffset(Idx);
return Idx;
} else if (ArrayType *AT = dyn_cast<ArrayType>(T)) {
T = AT->getElementType();
- uint64_t EltSize = DL->getTypeAllocSize(T);
+ uint64_t EltSize = DL.getTypeAllocSize(T);
Idx = Offset / EltSize;
Offset -= Idx * EltSize;
IdxTy = Type::getInt64Ty(T->getContext());
}
VectorType *VT = cast<VectorType>(T);
T = VT->getElementType();
- uint64_t EltSize = DL->getTypeAllocSize(T);
+ uint64_t EltSize = DL.getTypeAllocSize(T);
Idx = Offset / EltSize;
Offset -= Idx * EltSize;
IdxTy = Type::getInt64Ty(T->getContext());
void SROA::RewriteGEP(GetElementPtrInst *GEPI, AllocaInst *AI, uint64_t Offset,
SmallVectorImpl<AllocaInst *> &NewElts) {
uint64_t OldOffset = Offset;
+ const DataLayout &DL = GEPI->getModule()->getDataLayout();
SmallVector<Value*, 8> Indices(GEPI->op_begin() + 1, GEPI->op_end());
// If the GEP was dynamic then it must have been a dynamic vector lookup.
// In this case, it must be the last GEP operand which is dynamic so keep that
Value* NonConstantIdx = nullptr;
if (!GEPI->hasAllConstantIndices())
NonConstantIdx = Indices.pop_back_val();
- Offset += DL->getIndexedOffset(GEPI->getPointerOperandType(), Indices);
+ Offset += DL.getIndexedOffset(GEPI->getPointerOperandType(), Indices);
RewriteForScalarRepl(GEPI, AI, Offset, NewElts);
Type *T = AI->getAllocatedType();
Type *IdxTy;
- uint64_t OldIdx = FindElementAndOffset(T, OldOffset, IdxTy);
+ uint64_t OldIdx = FindElementAndOffset(T, OldOffset, IdxTy, DL);
if (GEPI->getOperand(0) == AI)
OldIdx = ~0ULL; // Force the GEP to be rewritten.
T = AI->getAllocatedType();
uint64_t EltOffset = Offset;
- uint64_t Idx = FindElementAndOffset(T, EltOffset, IdxTy);
+ uint64_t Idx = FindElementAndOffset(T, EltOffset, IdxTy, DL);
// If this GEP does not move the pointer across elements of the alloca
// being split, then it does not needs to be rewritten.
SmallVector<Value*, 8> NewArgs;
NewArgs.push_back(Constant::getNullValue(i32Ty));
while (EltOffset != 0) {
- uint64_t EltIdx = FindElementAndOffset(T, EltOffset, IdxTy);
+ uint64_t EltIdx = FindElementAndOffset(T, EltOffset, IdxTy, DL);
NewArgs.push_back(ConstantInt::get(IdxTy, EltIdx));
}
if (NonConstantIdx) {
// Put matching lifetime markers on everything from Offset up to
// Offset+OldSize.
Type *AIType = AI->getAllocatedType();
+ const DataLayout &DL = II->getModule()->getDataLayout();
uint64_t NewOffset = Offset;
Type *IdxTy;
- uint64_t Idx = FindElementAndOffset(AIType, NewOffset, IdxTy);
+ uint64_t Idx = FindElementAndOffset(AIType, NewOffset, IdxTy, DL);
IRBuilder<> Builder(II);
uint64_t Size = OldSize->getLimitedValue();
V = Builder.CreateGEP(V, Builder.getInt64(NewOffset));
IdxTy = NewElts[Idx]->getAllocatedType();
- uint64_t EltSize = DL->getTypeAllocSize(IdxTy) - NewOffset;
+ uint64_t EltSize = DL.getTypeAllocSize(IdxTy) - NewOffset;
if (EltSize > Size) {
EltSize = Size;
Size = 0;
for (; Idx != NewElts.size() && Size; ++Idx) {
IdxTy = NewElts[Idx]->getAllocatedType();
- uint64_t EltSize = DL->getTypeAllocSize(IdxTy);
+ uint64_t EltSize = DL.getTypeAllocSize(IdxTy);
if (EltSize > Size) {
EltSize = Size;
Size = 0;
bool SROADest = MI->getRawDest() == Inst;
Constant *Zero = Constant::getNullValue(Type::getInt32Ty(MI->getContext()));
+ const DataLayout &DL = MI->getModule()->getDataLayout();
for (unsigned i = 0, e = NewElts.size(); i != e; ++i) {
// If this is a memcpy/memmove, emit a GEP of the other element address.
PointerType *OtherPtrTy = cast<PointerType>(OtherPtr->getType());
Type *OtherTy = OtherPtrTy->getElementType();
if (StructType *ST = dyn_cast<StructType>(OtherTy)) {
- EltOffset = DL->getStructLayout(ST)->getElementOffset(i);
+ EltOffset = DL.getStructLayout(ST)->getElementOffset(i);
} else {
Type *EltTy = cast<SequentialType>(OtherTy)->getElementType();
- EltOffset = DL->getTypeAllocSize(EltTy)*i;
+ EltOffset = DL.getTypeAllocSize(EltTy) * i;
}
// The alignment of the other pointer is the guaranteed alignment of the
Type *ValTy = EltTy->getScalarType();
// Construct an integer with the right value.
- unsigned EltSize = DL->getTypeSizeInBits(ValTy);
+ unsigned EltSize = DL.getTypeSizeInBits(ValTy);
APInt OneVal(EltSize, CI->getZExtValue());
APInt TotalVal(OneVal);
// Set each byte.
// this element.
}
- unsigned EltSize = DL->getTypeAllocSize(EltTy);
+ unsigned EltSize = DL.getTypeAllocSize(EltTy);
if (!EltSize)
continue;
// and store the element value to the individual alloca.
Value *SrcVal = SI->getOperand(0);
Type *AllocaEltTy = AI->getAllocatedType();
- uint64_t AllocaSizeBits = DL->getTypeAllocSizeInBits(AllocaEltTy);
+ const DataLayout &DL = SI->getModule()->getDataLayout();
+ uint64_t AllocaSizeBits = DL.getTypeAllocSizeInBits(AllocaEltTy);
IRBuilder<> Builder(SI);
// Handle tail padding by extending the operand
- if (DL->getTypeSizeInBits(SrcVal->getType()) != AllocaSizeBits)
+ if (DL.getTypeSizeInBits(SrcVal->getType()) != AllocaSizeBits)
SrcVal = Builder.CreateZExt(SrcVal,
IntegerType::get(SI->getContext(), AllocaSizeBits));
// There are two forms here: AI could be an array or struct. Both cases
// have different ways to compute the element offset.
if (StructType *EltSTy = dyn_cast<StructType>(AllocaEltTy)) {
- const StructLayout *Layout = DL->getStructLayout(EltSTy);
+ const StructLayout *Layout = DL.getStructLayout(EltSTy);
for (unsigned i = 0, e = NewElts.size(); i != e; ++i) {
// Get the number of bits to shift SrcVal to get the value.
Type *FieldTy = EltSTy->getElementType(i);
uint64_t Shift = Layout->getElementOffsetInBits(i);
- if (DL->isBigEndian())
- Shift = AllocaSizeBits-Shift-DL->getTypeAllocSizeInBits(FieldTy);
+ if (DL.isBigEndian())
+ Shift = AllocaSizeBits - Shift - DL.getTypeAllocSizeInBits(FieldTy);
Value *EltVal = SrcVal;
if (Shift) {
}
// Truncate down to an integer of the right size.
- uint64_t FieldSizeBits = DL->getTypeSizeInBits(FieldTy);
+ uint64_t FieldSizeBits = DL.getTypeSizeInBits(FieldTy);
// Ignore zero sized fields like {}, they obviously contain no data.
if (FieldSizeBits == 0) continue;
} else {
ArrayType *ATy = cast<ArrayType>(AllocaEltTy);
Type *ArrayEltTy = ATy->getElementType();
- uint64_t ElementOffset = DL->getTypeAllocSizeInBits(ArrayEltTy);
- uint64_t ElementSizeBits = DL->getTypeSizeInBits(ArrayEltTy);
+ uint64_t ElementOffset = DL.getTypeAllocSizeInBits(ArrayEltTy);
+ uint64_t ElementSizeBits = DL.getTypeSizeInBits(ArrayEltTy);
uint64_t Shift;
- if (DL->isBigEndian())
+ if (DL.isBigEndian())
Shift = AllocaSizeBits-ElementOffset;
else
Shift = 0;
}
new StoreInst(EltVal, DestField, SI);
- if (DL->isBigEndian())
+ if (DL.isBigEndian())
Shift -= ElementOffset;
else
Shift += ElementOffset;
// Extract each element out of the NewElts according to its structure offset
// and form the result value.
Type *AllocaEltTy = AI->getAllocatedType();
- uint64_t AllocaSizeBits = DL->getTypeAllocSizeInBits(AllocaEltTy);
+ const DataLayout &DL = LI->getModule()->getDataLayout();
+ uint64_t AllocaSizeBits = DL.getTypeAllocSizeInBits(AllocaEltTy);
DEBUG(dbgs() << "PROMOTING LOAD OF WHOLE ALLOCA: " << *AI << '\n' << *LI
<< '\n');
const StructLayout *Layout = nullptr;
uint64_t ArrayEltBitOffset = 0;
if (StructType *EltSTy = dyn_cast<StructType>(AllocaEltTy)) {
- Layout = DL->getStructLayout(EltSTy);
+ Layout = DL.getStructLayout(EltSTy);
} else {
Type *ArrayEltTy = cast<ArrayType>(AllocaEltTy)->getElementType();
- ArrayEltBitOffset = DL->getTypeAllocSizeInBits(ArrayEltTy);
+ ArrayEltBitOffset = DL.getTypeAllocSizeInBits(ArrayEltTy);
}
Value *ResultVal =
Value *SrcField = NewElts[i];
Type *FieldTy =
cast<PointerType>(SrcField->getType())->getElementType();
- uint64_t FieldSizeBits = DL->getTypeSizeInBits(FieldTy);
+ uint64_t FieldSizeBits = DL.getTypeSizeInBits(FieldTy);
// Ignore zero sized fields like {}, they obviously contain no data.
if (FieldSizeBits == 0) continue;
else // Array case.
Shift = i*ArrayEltBitOffset;
- if (DL->isBigEndian())
+ if (DL.isBigEndian())
Shift = AllocaSizeBits-Shift-FieldIntTy->getBitWidth();
if (Shift) {
}
// Handle tail padding by truncating the result
- if (DL->getTypeSizeInBits(LI->getType()) != AllocaSizeBits)
+ if (DL.getTypeSizeInBits(LI->getType()) != AllocaSizeBits)
ResultVal = new TruncInst(ResultVal, LI->getType(), "", LI);
LI->replaceAllUsesWith(ResultVal);
return false;
}
+ const DataLayout &DL = AI->getModule()->getDataLayout();
+
// Okay, we know all the users are promotable. If the aggregate is a memcpy
// source and destination, we have to be careful. In particular, the memcpy
// could be moving around elements that live in structure padding of the LLVM
// types, but may actually be used. In these cases, we refuse to promote the
// struct.
if (Info.isMemCpySrc && Info.isMemCpyDst &&
- HasPadding(AI->getAllocatedType(), *DL))
+ HasPadding(AI->getAllocatedType(), DL))
return false;
// If the alloca never has an access to just *part* of it, but is accessed
void gather(Instruction *, const ValueVector &);
bool canTransferMetadata(unsigned Kind);
void transferMetadata(Instruction *, const ValueVector &);
- bool getVectorLayout(Type *, unsigned, VectorLayout &);
+ bool getVectorLayout(Type *, unsigned, VectorLayout &, const DataLayout &);
bool finish();
template<typename T> bool splitBinary(Instruction &, const T &);
ScatterMap Scattered;
GatherList Gathered;
unsigned ParallelLoopAccessMDKind;
- const DataLayout *DL;
bool ScalarizeLoadStore;
};
}
bool Scalarizer::runOnFunction(Function &F) {
- DL = &F.getParent()->getDataLayout();
for (Function::iterator BBI = F.begin(), BBE = F.end(); BBI != BBE; ++BBI) {
BasicBlock *BB = BBI;
for (BasicBlock::iterator II = BB->begin(), IE = BB->end(); II != IE;) {
// Try to fill in Layout from Ty, returning true on success. Alignment is
// the alignment of the vector, or 0 if the ABI default should be used.
bool Scalarizer::getVectorLayout(Type *Ty, unsigned Alignment,
- VectorLayout &Layout) {
- if (!DL)
- return false;
-
+ VectorLayout &Layout, const DataLayout &DL) {
// Make sure we're dealing with a vector.
Layout.VecTy = dyn_cast<VectorType>(Ty);
if (!Layout.VecTy)
// Check that we're dealing with full-byte elements.
Layout.ElemTy = Layout.VecTy->getElementType();
- if (DL->getTypeSizeInBits(Layout.ElemTy) !=
- DL->getTypeStoreSizeInBits(Layout.ElemTy))
+ if (DL.getTypeSizeInBits(Layout.ElemTy) !=
+ DL.getTypeStoreSizeInBits(Layout.ElemTy))
return false;
if (Alignment)
Layout.VecAlign = Alignment;
else
- Layout.VecAlign = DL->getABITypeAlignment(Layout.VecTy);
- Layout.ElemSize = DL->getTypeStoreSize(Layout.ElemTy);
+ Layout.VecAlign = DL.getABITypeAlignment(Layout.VecTy);
+ Layout.ElemSize = DL.getTypeStoreSize(Layout.ElemTy);
return true;
}
return false;
VectorLayout Layout;
- if (!getVectorLayout(LI.getType(), LI.getAlignment(), Layout))
+ if (!getVectorLayout(LI.getType(), LI.getAlignment(), Layout,
+ LI.getModule()->getDataLayout()))
return false;
unsigned NumElems = Layout.VecTy->getNumElements();
VectorLayout Layout;
Value *FullValue = SI.getValueOperand();
- if (!getVectorLayout(FullValue->getType(), SI.getAlignment(), Layout))
+ if (!getVectorLayout(FullValue->getType(), SI.getAlignment(), Layout,
+ SI.getModule()->getDataLayout()))
return false;
unsigned NumElems = Layout.VecTy->getNumElements();
/// new index representing the remainder (equal to the original index minus
/// the constant offset), or nullptr if we cannot extract a constant offset.
/// \p Idx The given GEP index
- /// \p DL The datalayout of the module
/// \p GEP The given GEP
- static Value *Extract(Value *Idx, const DataLayout *DL,
- GetElementPtrInst *GEP);
+ static Value *Extract(Value *Idx, GetElementPtrInst *GEP);
/// Looks for a constant offset from the given GEP index without extracting
/// it. It returns the numeric value of the extracted constant offset (0 if
/// failed). The meaning of the arguments are the same as Extract.
- static int64_t Find(Value *Idx, const DataLayout *DL, GetElementPtrInst *GEP);
+ static int64_t Find(Value *Idx, GetElementPtrInst *GEP);
private:
- ConstantOffsetExtractor(const DataLayout *Layout, Instruction *InsertionPt)
- : DL(Layout), IP(InsertionPt) {}
+ ConstantOffsetExtractor(Instruction *InsertionPt) : IP(InsertionPt) {}
/// Searches the expression that computes V for a non-zero constant C s.t.
/// V can be reassociated into the form V' + C. If the searching is
/// successful, returns C and update UserChain as a def-use chain from C to V;
/// A data structure used in rebuildWithoutConstOffset. Contains all
/// sext/zext instructions along UserChain.
SmallVector<CastInst *, 16> ExtInsts;
- /// The data layout of the module. Used in ComputeKnownBits.
- const DataLayout *DL;
Instruction *IP; /// Insertion position of cloned instructions.
};
AU.setPreservesCFG();
}
- bool doInitialization(Module &M) override {
- DL = &M.getDataLayout();
- return false;
- }
-
bool runOnFunction(Function &F) override;
private:
/// Verified in @i32_add in split-gep.ll
bool canonicalizeArrayIndicesToPointerSize(GetElementPtrInst *GEP);
- const DataLayout *DL;
const TargetMachine *TM;
/// Whether to lower a GEP with multiple indices into arithmetic operations or
/// multiple GEPs with a single index.
return BO;
}
-Value *ConstantOffsetExtractor::Extract(Value *Idx, const DataLayout *DL,
- GetElementPtrInst *GEP) {
- ConstantOffsetExtractor Extractor(DL, GEP);
+Value *ConstantOffsetExtractor::Extract(Value *Idx, GetElementPtrInst *GEP) {
+ ConstantOffsetExtractor Extractor(GEP);
// Find a non-zero constant offset first.
APInt ConstantOffset =
Extractor.find(Idx, /* SignExtended */ false, /* ZeroExtended */ false,
return Extractor.rebuildWithoutConstOffset();
}
-int64_t ConstantOffsetExtractor::Find(Value *Idx, const DataLayout *DL,
- GetElementPtrInst *GEP) {
+int64_t ConstantOffsetExtractor::Find(Value *Idx, GetElementPtrInst *GEP) {
// If Idx is an index of an inbound GEP, Idx is guaranteed to be non-negative.
- return ConstantOffsetExtractor(DL, GEP)
+ return ConstantOffsetExtractor(GEP)
.find(Idx, /* SignExtended */ false, /* ZeroExtended */ false,
GEP->isInBounds())
.getSExtValue();
IntegerType *IT = cast<IntegerType>(V->getType());
KnownOne = APInt(IT->getBitWidth(), 0);
KnownZero = APInt(IT->getBitWidth(), 0);
+ const DataLayout &DL = IP->getModule()->getDataLayout();
llvm::computeKnownBits(V, KnownZero, KnownOne, DL, 0);
}
bool SeparateConstOffsetFromGEP::canonicalizeArrayIndicesToPointerSize(
GetElementPtrInst *GEP) {
bool Changed = false;
- Type *IntPtrTy = DL->getIntPtrType(GEP->getType());
+ const DataLayout &DL = GEP->getModule()->getDataLayout();
+ Type *IntPtrTy = DL.getIntPtrType(GEP->getType());
gep_type_iterator GTI = gep_type_begin(*GEP);
for (User::op_iterator I = GEP->op_begin() + 1, E = GEP->op_end();
I != E; ++I, ++GTI) {
NeedsExtraction = false;
int64_t AccumulativeByteOffset = 0;
gep_type_iterator GTI = gep_type_begin(*GEP);
+ const DataLayout &DL = GEP->getModule()->getDataLayout();
for (unsigned I = 1, E = GEP->getNumOperands(); I != E; ++I, ++GTI) {
if (isa<SequentialType>(*GTI)) {
// Tries to extract a constant offset from this GEP index.
int64_t ConstantOffset =
- ConstantOffsetExtractor::Find(GEP->getOperand(I), DL, GEP);
+ ConstantOffsetExtractor::Find(GEP->getOperand(I), GEP);
if (ConstantOffset != 0) {
NeedsExtraction = true;
// A GEP may have multiple indices. We accumulate the extracted
// constant offset to a byte offset, and later offset the remainder of
// the original GEP with this byte offset.
AccumulativeByteOffset +=
- ConstantOffset * DL->getTypeAllocSize(GTI.getIndexedType());
+ ConstantOffset * DL.getTypeAllocSize(GTI.getIndexedType());
}
} else if (LowerGEP) {
StructType *StTy = cast<StructType>(*GTI);
if (Field != 0) {
NeedsExtraction = true;
AccumulativeByteOffset +=
- DL->getStructLayout(StTy)->getElementOffset(Field);
+ DL.getStructLayout(StTy)->getElementOffset(Field);
}
}
}
void SeparateConstOffsetFromGEP::lowerToSingleIndexGEPs(
GetElementPtrInst *Variadic, int64_t AccumulativeByteOffset) {
IRBuilder<> Builder(Variadic);
- Type *IntPtrTy = DL->getIntPtrType(Variadic->getType());
+ const DataLayout &DL = Variadic->getModule()->getDataLayout();
+ Type *IntPtrTy = DL.getIntPtrType(Variadic->getType());
Type *I8PtrTy =
Builder.getInt8PtrTy(Variadic->getType()->getPointerAddressSpace());
continue;
APInt ElementSize = APInt(IntPtrTy->getIntegerBitWidth(),
- DL->getTypeAllocSize(GTI.getIndexedType()));
+ DL.getTypeAllocSize(GTI.getIndexedType()));
// Scale the index by element size.
if (ElementSize != 1) {
if (ElementSize.isPowerOf2()) {
SeparateConstOffsetFromGEP::lowerToArithmetics(GetElementPtrInst *Variadic,
int64_t AccumulativeByteOffset) {
IRBuilder<> Builder(Variadic);
- Type *IntPtrTy = DL->getIntPtrType(Variadic->getType());
+ const DataLayout &DL = Variadic->getModule()->getDataLayout();
+ Type *IntPtrTy = DL.getIntPtrType(Variadic->getType());
Value *ResultPtr = Builder.CreatePtrToInt(Variadic->getOperand(0), IntPtrTy);
gep_type_iterator GTI = gep_type_begin(*Variadic);
continue;
APInt ElementSize = APInt(IntPtrTy->getIntegerBitWidth(),
- DL->getTypeAllocSize(GTI.getIndexedType()));
+ DL.getTypeAllocSize(GTI.getIndexedType()));
// Scale the index by element size.
if (ElementSize != 1) {
if (ElementSize.isPowerOf2()) {
if (isa<SequentialType>(*GTI)) {
// Splits this GEP index into a variadic part and a constant offset, and
// uses the variadic part as the new index.
- Value *NewIdx =
- ConstantOffsetExtractor::Extract(GEP->getOperand(I), DL, GEP);
+ Value *NewIdx = ConstantOffsetExtractor::Extract(GEP->getOperand(I), GEP);
if (NewIdx != nullptr) {
GEP->setOperand(I, NewIdx);
}
// Per ANSI C standard, signed / unsigned = unsigned and signed % unsigned =
// unsigned.. Therefore, we cast ElementTypeSizeOfGEP to signed because it is
// used with unsigned integers later.
+ const DataLayout &DL = GEP->getModule()->getDataLayout();
int64_t ElementTypeSizeOfGEP = static_cast<int64_t>(
- DL->getTypeAllocSize(GEP->getType()->getElementType()));
- Type *IntPtrTy = DL->getIntPtrType(GEP->getType());
+ DL.getTypeAllocSize(GEP->getType()->getElementType()));
+ Type *IntPtrTy = DL.getIntPtrType(GEP->getType());
if (AccumulativeByteOffset % ElementTypeSizeOfGEP == 0) {
// Very likely. As long as %gep is natually aligned, the byte offset we
// extracted should be a multiple of sizeof(*%gep).
/// iterativelySimplifyCFG - Call SimplifyCFG on all the blocks in the function,
/// iterating until no more changes are made.
static bool iterativelySimplifyCFG(Function &F, const TargetTransformInfo &TTI,
- const DataLayout *DL, AssumptionCache *AC,
+ AssumptionCache *AC,
unsigned BonusInstThreshold) {
bool Changed = false;
bool LocalChange = true;
// Loop over all of the basic blocks and remove them if they are unneeded...
//
for (Function::iterator BBIt = F.begin(); BBIt != F.end(); ) {
- if (SimplifyCFG(BBIt++, TTI, BonusInstThreshold, DL, AC)) {
+ if (SimplifyCFG(BBIt++, TTI, BonusInstThreshold, AC)) {
LocalChange = true;
++NumSimpl;
}
}
static bool simplifyFunctionCFG(Function &F, const TargetTransformInfo &TTI,
- const DataLayout *DL, AssumptionCache *AC,
- int BonusInstThreshold) {
+ AssumptionCache *AC, int BonusInstThreshold) {
bool EverChanged = removeUnreachableBlocks(F);
EverChanged |= mergeEmptyReturnBlocks(F);
- EverChanged |= iterativelySimplifyCFG(F, TTI, DL, AC, BonusInstThreshold);
+ EverChanged |= iterativelySimplifyCFG(F, TTI, AC, BonusInstThreshold);
// If neither pass changed anything, we're done.
if (!EverChanged) return false;
return true;
do {
- EverChanged = iterativelySimplifyCFG(F, TTI, DL, AC, BonusInstThreshold);
+ EverChanged = iterativelySimplifyCFG(F, TTI, AC, BonusInstThreshold);
EverChanged |= removeUnreachableBlocks(F);
} while (EverChanged);
PreservedAnalyses SimplifyCFGPass::run(Function &F,
AnalysisManager<Function> *AM) {
- auto &DL = F.getParent()->getDataLayout();
auto &TTI = AM->getResult<TargetIRAnalysis>(F);
auto &AC = AM->getResult<AssumptionAnalysis>(F);
- if (!simplifyFunctionCFG(F, TTI, &DL, &AC, BonusInstThreshold))
+ if (!simplifyFunctionCFG(F, TTI, &AC, BonusInstThreshold))
return PreservedAnalyses::none();
return PreservedAnalyses::all();
&getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
const TargetTransformInfo &TTI =
getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
- const DataLayout &DL = F.getParent()->getDataLayout();
- return simplifyFunctionCFG(F, TTI, &DL, AC, BonusInstThreshold);
+ return simplifyFunctionCFG(F, TTI, AC, BonusInstThreshold);
}
void getAnalysisUsage(AnalysisUsage &AU) const override {
DominatorTree *DT;
LoopInfo *LI;
AliasAnalysis *AA;
- const DataLayout *DL;
public:
static char ID; // Pass identification
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
AA = &getAnalysis<AliasAnalysis>();
- DL = &F.getParent()->getDataLayout();
bool MadeChange, EverMadeChange = false;
if (SuccToSinkTo->getUniquePredecessor() != Inst->getParent()) {
// We cannot sink a load across a critical edge - there may be stores in
// other code paths.
- if (!isSafeToSpeculativelyExecute(Inst, DL))
+ if (!isSafeToSpeculativelyExecute(Inst))
return false;
// We don't want to sink across a critical edge if we don't dominate the
namespace {
struct TailCallElim : public FunctionPass {
const TargetTransformInfo *TTI;
- const DataLayout *DL;
static char ID; // Pass identification, replacement for typeid
TailCallElim() : FunctionPass(ID) {
if (skipOptnoneFunction(F))
return false;
- DL = &F.getParent()->getDataLayout();
-
bool AllCallsAreTailCalls = false;
bool Modified = markTails(F, AllCallsAreTailCalls);
if (AllCallsAreTailCalls)
PHINode *PN = ArgumentPHIs[i];
// If the PHI Node is a dynamic constant, replace it with the value it is.
- if (Value *PNV = SimplifyInstruction(PN)) {
+ if (Value *PNV = SimplifyInstruction(PN, F.getParent()->getDataLayout())) {
PN->replaceAllUsesWith(PNV);
PN->eraseFromParent();
}
// being loaded from.
if (CI->mayWriteToMemory() ||
!isSafeToLoadUnconditionally(L->getPointerOperand(), L,
- L->getAlignment(), DL))
+ L->getAlignment()))
return false;
}
}
/// EmitStrLen - Emit a call to the strlen function to the builder, for the
/// specified pointer. This always returns an integer value of size intptr_t.
-Value *llvm::EmitStrLen(Value *Ptr, IRBuilder<> &B, const DataLayout *TD,
+Value *llvm::EmitStrLen(Value *Ptr, IRBuilder<> &B, const DataLayout &DL,
const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::strlen))
return nullptr;
AS[1] = AttributeSet::get(M->getContext(), AttributeSet::FunctionIndex, AVs);
LLVMContext &Context = B.GetInsertBlock()->getContext();
- Constant *StrLen = M->getOrInsertFunction("strlen",
- AttributeSet::get(M->getContext(),
- AS),
- TD->getIntPtrType(Context),
- B.getInt8PtrTy(),
- nullptr);
+ Constant *StrLen = M->getOrInsertFunction(
+ "strlen", AttributeSet::get(M->getContext(), AS),
+ DL.getIntPtrType(Context), B.getInt8PtrTy(), nullptr);
CallInst *CI = B.CreateCall(StrLen, CastToCStr(Ptr, B), "strlen");
if (const Function *F = dyn_cast<Function>(StrLen->stripPointerCasts()))
CI->setCallingConv(F->getCallingConv());
/// specified pointer. Ptr is required to be some pointer type, MaxLen must
/// be of size_t type, and the return value has 'intptr_t' type.
Value *llvm::EmitStrNLen(Value *Ptr, Value *MaxLen, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI) {
+ const DataLayout &DL, const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::strnlen))
return nullptr;
AS[1] = AttributeSet::get(M->getContext(), AttributeSet::FunctionIndex, AVs);
LLVMContext &Context = B.GetInsertBlock()->getContext();
- Constant *StrNLen = M->getOrInsertFunction("strnlen",
- AttributeSet::get(M->getContext(),
- AS),
- TD->getIntPtrType(Context),
- B.getInt8PtrTy(),
- TD->getIntPtrType(Context),
- nullptr);
+ Constant *StrNLen =
+ M->getOrInsertFunction("strnlen", AttributeSet::get(M->getContext(), AS),
+ DL.getIntPtrType(Context), B.getInt8PtrTy(),
+ DL.getIntPtrType(Context), nullptr);
CallInst *CI = B.CreateCall2(StrNLen, CastToCStr(Ptr, B), MaxLen, "strnlen");
if (const Function *F = dyn_cast<Function>(StrNLen->stripPointerCasts()))
CI->setCallingConv(F->getCallingConv());
/// specified pointer and character. Ptr is required to be some pointer type,
/// and the return value has 'i8*' type.
Value *llvm::EmitStrChr(Value *Ptr, char C, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI) {
+ const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::strchr))
return nullptr;
}
/// EmitStrNCmp - Emit a call to the strncmp function to the builder.
-Value *llvm::EmitStrNCmp(Value *Ptr1, Value *Ptr2, Value *Len,
- IRBuilder<> &B, const DataLayout *TD,
- const TargetLibraryInfo *TLI) {
+Value *llvm::EmitStrNCmp(Value *Ptr1, Value *Ptr2, Value *Len, IRBuilder<> &B,
+ const DataLayout &DL, const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::strncmp))
return nullptr;
AS[2] = AttributeSet::get(M->getContext(), AttributeSet::FunctionIndex, AVs);
LLVMContext &Context = B.GetInsertBlock()->getContext();
- Value *StrNCmp = M->getOrInsertFunction("strncmp",
- AttributeSet::get(M->getContext(),
- AS),
- B.getInt32Ty(),
- B.getInt8PtrTy(),
- B.getInt8PtrTy(),
- TD->getIntPtrType(Context), nullptr);
+ Value *StrNCmp = M->getOrInsertFunction(
+ "strncmp", AttributeSet::get(M->getContext(), AS), B.getInt32Ty(),
+ B.getInt8PtrTy(), B.getInt8PtrTy(), DL.getIntPtrType(Context), nullptr);
CallInst *CI = B.CreateCall3(StrNCmp, CastToCStr(Ptr1, B),
CastToCStr(Ptr2, B), Len, "strncmp");
/// EmitStrCpy - Emit a call to the strcpy function to the builder, for the
/// specified pointer arguments.
Value *llvm::EmitStrCpy(Value *Dst, Value *Src, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI,
- StringRef Name) {
+ const TargetLibraryInfo *TLI, StringRef Name) {
if (!TLI->has(LibFunc::strcpy))
return nullptr;
/// EmitStrNCpy - Emit a call to the strncpy function to the builder, for the
/// specified pointer arguments.
-Value *llvm::EmitStrNCpy(Value *Dst, Value *Src, Value *Len,
- IRBuilder<> &B, const DataLayout *TD,
+Value *llvm::EmitStrNCpy(Value *Dst, Value *Src, Value *Len, IRBuilder<> &B,
const TargetLibraryInfo *TLI, StringRef Name) {
if (!TLI->has(LibFunc::strncpy))
return nullptr;
/// This expects that the Len and ObjSize have type 'intptr_t' and Dst/Src
/// are pointers.
Value *llvm::EmitMemCpyChk(Value *Dst, Value *Src, Value *Len, Value *ObjSize,
- IRBuilder<> &B, const DataLayout *TD,
+ IRBuilder<> &B, const DataLayout &DL,
const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::memcpy_chk))
return nullptr;
AS = AttributeSet::get(M->getContext(), AttributeSet::FunctionIndex,
Attribute::NoUnwind);
LLVMContext &Context = B.GetInsertBlock()->getContext();
- Value *MemCpy = M->getOrInsertFunction("__memcpy_chk",
- AttributeSet::get(M->getContext(), AS),
- B.getInt8PtrTy(),
- B.getInt8PtrTy(),
- B.getInt8PtrTy(),
- TD->getIntPtrType(Context),
- TD->getIntPtrType(Context), nullptr);
+ Value *MemCpy = M->getOrInsertFunction(
+ "__memcpy_chk", AttributeSet::get(M->getContext(), AS), B.getInt8PtrTy(),
+ B.getInt8PtrTy(), B.getInt8PtrTy(), DL.getIntPtrType(Context),
+ DL.getIntPtrType(Context), nullptr);
Dst = CastToCStr(Dst, B);
Src = CastToCStr(Src, B);
CallInst *CI = B.CreateCall4(MemCpy, Dst, Src, Len, ObjSize);
/// EmitMemChr - Emit a call to the memchr function. This assumes that Ptr is
/// a pointer, Val is an i32 value, and Len is an 'intptr_t' value.
-Value *llvm::EmitMemChr(Value *Ptr, Value *Val,
- Value *Len, IRBuilder<> &B, const DataLayout *TD,
- const TargetLibraryInfo *TLI) {
+Value *llvm::EmitMemChr(Value *Ptr, Value *Val, Value *Len, IRBuilder<> &B,
+ const DataLayout &DL, const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::memchr))
return nullptr;
Attribute::AttrKind AVs[2] = { Attribute::ReadOnly, Attribute::NoUnwind };
AS = AttributeSet::get(M->getContext(), AttributeSet::FunctionIndex, AVs);
LLVMContext &Context = B.GetInsertBlock()->getContext();
- Value *MemChr = M->getOrInsertFunction("memchr",
- AttributeSet::get(M->getContext(), AS),
- B.getInt8PtrTy(),
- B.getInt8PtrTy(),
- B.getInt32Ty(),
- TD->getIntPtrType(Context),
- nullptr);
+ Value *MemChr = M->getOrInsertFunction(
+ "memchr", AttributeSet::get(M->getContext(), AS), B.getInt8PtrTy(),
+ B.getInt8PtrTy(), B.getInt32Ty(), DL.getIntPtrType(Context), nullptr);
CallInst *CI = B.CreateCall3(MemChr, CastToCStr(Ptr, B), Val, Len, "memchr");
if (const Function *F = dyn_cast<Function>(MemChr->stripPointerCasts()))
}
/// EmitMemCmp - Emit a call to the memcmp function.
-Value *llvm::EmitMemCmp(Value *Ptr1, Value *Ptr2,
- Value *Len, IRBuilder<> &B, const DataLayout *TD,
- const TargetLibraryInfo *TLI) {
+Value *llvm::EmitMemCmp(Value *Ptr1, Value *Ptr2, Value *Len, IRBuilder<> &B,
+ const DataLayout &DL, const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::memcmp))
return nullptr;
AS[2] = AttributeSet::get(M->getContext(), AttributeSet::FunctionIndex, AVs);
LLVMContext &Context = B.GetInsertBlock()->getContext();
- Value *MemCmp = M->getOrInsertFunction("memcmp",
- AttributeSet::get(M->getContext(), AS),
- B.getInt32Ty(),
- B.getInt8PtrTy(),
- B.getInt8PtrTy(),
- TD->getIntPtrType(Context), nullptr);
+ Value *MemCmp = M->getOrInsertFunction(
+ "memcmp", AttributeSet::get(M->getContext(), AS), B.getInt32Ty(),
+ B.getInt8PtrTy(), B.getInt8PtrTy(), DL.getIntPtrType(Context), nullptr);
CallInst *CI = B.CreateCall3(MemCmp, CastToCStr(Ptr1, B), CastToCStr(Ptr2, B),
Len, "memcmp");
/// EmitPutChar - Emit a call to the putchar function. This assumes that Char
/// is an integer.
-Value *llvm::EmitPutChar(Value *Char, IRBuilder<> &B, const DataLayout *TD,
+Value *llvm::EmitPutChar(Value *Char, IRBuilder<> &B,
const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::putchar))
return nullptr;
/// EmitPutS - Emit a call to the puts function. This assumes that Str is
/// some pointer.
-Value *llvm::EmitPutS(Value *Str, IRBuilder<> &B, const DataLayout *TD,
+Value *llvm::EmitPutS(Value *Str, IRBuilder<> &B,
const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::puts))
return nullptr;
/// EmitFPutC - Emit a call to the fputc function. This assumes that Char is
/// an integer and File is a pointer to FILE.
Value *llvm::EmitFPutC(Value *Char, Value *File, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI) {
+ const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::fputc))
return nullptr;
/// EmitFPutS - Emit a call to the puts function. Str is required to be a
/// pointer and File is a pointer to FILE.
Value *llvm::EmitFPutS(Value *Str, Value *File, IRBuilder<> &B,
- const DataLayout *TD, const TargetLibraryInfo *TLI) {
+ const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::fputs))
return nullptr;
/// EmitFWrite - Emit a call to the fwrite function. This assumes that Ptr is
/// a pointer, Size is an 'intptr_t', and File is a pointer to FILE.
-Value *llvm::EmitFWrite(Value *Ptr, Value *Size, Value *File,
- IRBuilder<> &B, const DataLayout *TD,
- const TargetLibraryInfo *TLI) {
+Value *llvm::EmitFWrite(Value *Ptr, Value *Size, Value *File, IRBuilder<> &B,
+ const DataLayout &DL, const TargetLibraryInfo *TLI) {
if (!TLI->has(LibFunc::fwrite))
return nullptr;
StringRef FWriteName = TLI->getName(LibFunc::fwrite);
Constant *F;
if (File->getType()->isPointerTy())
- F = M->getOrInsertFunction(FWriteName,
- AttributeSet::get(M->getContext(), AS),
- TD->getIntPtrType(Context),
- B.getInt8PtrTy(),
- TD->getIntPtrType(Context),
- TD->getIntPtrType(Context),
- File->getType(), nullptr);
+ F = M->getOrInsertFunction(
+ FWriteName, AttributeSet::get(M->getContext(), AS),
+ DL.getIntPtrType(Context), B.getInt8PtrTy(), DL.getIntPtrType(Context),
+ DL.getIntPtrType(Context), File->getType(), nullptr);
else
- F = M->getOrInsertFunction(FWriteName, TD->getIntPtrType(Context),
- B.getInt8PtrTy(),
- TD->getIntPtrType(Context),
- TD->getIntPtrType(Context),
- File->getType(), nullptr);
- CallInst *CI = B.CreateCall4(F, CastToCStr(Ptr, B), Size,
- ConstantInt::get(TD->getIntPtrType(Context), 1), File);
+ F = M->getOrInsertFunction(FWriteName, DL.getIntPtrType(Context),
+ B.getInt8PtrTy(), DL.getIntPtrType(Context),
+ DL.getIntPtrType(Context), File->getType(),
+ nullptr);
+ CallInst *CI =
+ B.CreateCall4(F, CastToCStr(Ptr, B), Size,
+ ConstantInt::get(DL.getIntPtrType(Context), 1), File);
if (const Function *Fn = dyn_cast<Function>(F->stripPointerCasts()))
CI->setCallingConv(Fn->getCallingConv());
bool ModuleLevelChanges;
const char *NameSuffix;
ClonedCodeInfo *CodeInfo;
- const DataLayout *DL;
CloningDirector *Director;
ValueMapTypeRemapper *TypeMapper;
ValueMaterializer *Materializer;
public:
PruningFunctionCloner(Function *newFunc, const Function *oldFunc,
- ValueToValueMapTy &valueMap,
- bool moduleLevelChanges,
- const char *nameSuffix,
- ClonedCodeInfo *codeInfo,
- const DataLayout *DL,
+ ValueToValueMapTy &valueMap, bool moduleLevelChanges,
+ const char *nameSuffix, ClonedCodeInfo *codeInfo,
CloningDirector *Director)
- : NewFunc(newFunc), OldFunc(oldFunc),
- VMap(valueMap), ModuleLevelChanges(moduleLevelChanges),
- NameSuffix(nameSuffix), CodeInfo(codeInfo), DL(DL),
- Director(Director) {
+ : NewFunc(newFunc), OldFunc(oldFunc), VMap(valueMap),
+ ModuleLevelChanges(moduleLevelChanges), NameSuffix(nameSuffix),
+ CodeInfo(codeInfo), Director(Director) {
// These are optional components. The Director may return null.
if (Director) {
TypeMapper = Director->getTypeRemapper();
// If we can simplify this instruction to some other value, simply add
// a mapping to that value rather than inserting a new instruction into
// the basic block.
- if (Value *V = SimplifyInstruction(NewInst, DL)) {
+ if (Value *V =
+ SimplifyInstruction(NewInst, BB->getModule()->getDataLayout())) {
// On the off-chance that this simplifies to an instruction in the old
// function, map it back into the new function.
if (Value *MappedV = VMap.lookup(V))
SmallVectorImpl<ReturnInst *> &Returns,
const char *NameSuffix,
ClonedCodeInfo *CodeInfo,
- const DataLayout *DL,
CloningDirector *Director) {
assert(NameSuffix && "NameSuffix cannot be null!");
#endif
PruningFunctionCloner PFC(NewFunc, OldFunc, VMap, ModuleLevelChanges,
- NameSuffix, CodeInfo, DL, Director);
+ NameSuffix, CodeInfo, Director);
const BasicBlock *StartingBB;
if (StartingInst)
StartingBB = StartingInst->getParent();
// node).
for (unsigned Idx = 0, Size = PHIToResolve.size(); Idx != Size; ++Idx)
if (PHINode *PN = dyn_cast<PHINode>(VMap[PHIToResolve[Idx]]))
- recursivelySimplifyInstruction(PN, DL);
+ recursivelySimplifyInstruction(PN);
// Now that the inlined function body has been fully constructed, go through
// and zap unconditional fall-through branches. This happen all the time when
SmallVectorImpl<ReturnInst*> &Returns,
const char *NameSuffix,
ClonedCodeInfo *CodeInfo,
- const DataLayout *DL,
Instruction *TheCall) {
- CloneAndPruneIntoFromInst(NewFunc, OldFunc, OldFunc->front().begin(),
- VMap, ModuleLevelChanges, Returns, NameSuffix,
- CodeInfo, DL, nullptr);
+ CloneAndPruneIntoFromInst(NewFunc, OldFunc, OldFunc->front().begin(), VMap,
+ ModuleLevelChanges, Returns, NameSuffix, CodeInfo,
+ nullptr);
}
/// parameters with noalias metadata specifying the new scope, and tag all
/// non-derived loads, stores and memory intrinsics with the new alias scopes.
static void AddAliasScopeMetadata(CallSite CS, ValueToValueMapTy &VMap,
- const DataLayout *DL, AliasAnalysis *AA) {
+ const DataLayout &DL, AliasAnalysis *AA) {
if (!EnableNoAliasConversion)
return;
// If we can already prove the asserted alignment in the context of the
// caller, then don't bother inserting the assumption.
Value *Arg = CS.getArgument(I->getArgNo());
- if (getKnownAlignment(Arg, &DL, &IFI.ACT->getAssumptionCache(*CalledFunc),
- CS.getInstruction(), &DT) >= Align)
+ if (getKnownAlignment(Arg, DL, CS.getInstruction(),
+ &IFI.ACT->getAssumptionCache(*CalledFunc),
+ &DT) >= Align)
continue;
IRBuilder<>(CS.getInstruction())
if (ByValAlignment <= 1) // 0 = unspecified, 1 = no particular alignment.
return Arg;
+ const DataLayout &DL = Caller->getParent()->getDataLayout();
+
// If the pointer is already known to be sufficiently aligned, or if we can
// round it up to a larger alignment, then we don't need a temporary.
- auto &DL = Caller->getParent()->getDataLayout();
- if (getOrEnforceKnownAlignment(Arg, ByValAlignment, &DL,
- &IFI.ACT->getAssumptionCache(*Caller),
- TheCall) >= ByValAlignment)
+ if (getOrEnforceKnownAlignment(Arg, ByValAlignment, DL, TheCall,
+ &IFI.ACT->getAssumptionCache(*Caller)) >=
+ ByValAlignment)
return Arg;
// Otherwise, we have to make a memcpy to get a safe alignment. This is bad
// happy with whatever the cloner can do.
CloneAndPruneFunctionInto(Caller, CalledFunc, VMap,
/*ModuleLevelChanges=*/false, Returns, ".i",
- &InlinedFunctionInfo, &DL, TheCall);
+ &InlinedFunctionInfo, TheCall);
// Remember the first block that is newly cloned over.
FirstNewBlock = LastBlock; ++FirstNewBlock;
CloneAliasScopeMetadata(CS, VMap);
// Add noalias metadata if necessary.
- AddAliasScopeMetadata(CS, VMap, &DL, IFI.AA);
+ AddAliasScopeMetadata(CS, VMap, DL, IFI.AA);
// FIXME: We could register any cloned assumptions instead of clearing the
// whole function's cache.
// block other optimizations.
if (PHI) {
auto &DL = Caller->getParent()->getDataLayout();
- if (Value *V = SimplifyInstruction(PHI, &DL, nullptr, nullptr,
+ if (Value *V = SimplifyInstruction(PHI, DL, nullptr, nullptr,
&IFI.ACT->getAssumptionCache(*Caller))) {
PHI->replaceAllUsesWith(V);
PHI->eraseFromParent();
///
/// This returns true if it changed the code, note that it can delete
/// instructions in other blocks as well in this block.
-bool llvm::SimplifyInstructionsInBlock(BasicBlock *BB, const DataLayout *TD,
+bool llvm::SimplifyInstructionsInBlock(BasicBlock *BB,
const TargetLibraryInfo *TLI) {
bool MadeChange = false;
Instruction *Inst = BI++;
WeakVH BIHandle(BI);
- if (recursivelySimplifyInstruction(Inst, TD, TLI)) {
+ if (recursivelySimplifyInstruction(Inst, TLI)) {
MadeChange = true;
if (BIHandle != BI)
BI = BB->begin();
///
/// .. and delete the predecessor corresponding to the '1', this will attempt to
/// recursively fold the and to 0.
-void llvm::RemovePredecessorAndSimplify(BasicBlock *BB, BasicBlock *Pred,
- DataLayout *TD) {
+void llvm::RemovePredecessorAndSimplify(BasicBlock *BB, BasicBlock *Pred) {
// This only adjusts blocks with PHI nodes.
if (!isa<PHINode>(BB->begin()))
return;
PhiIt = &*++BasicBlock::iterator(cast<Instruction>(PhiIt));
Value *OldPhiIt = PhiIt;
- if (!recursivelySimplifyInstruction(PN, TD))
+ if (!recursivelySimplifyInstruction(PN))
continue;
// If recursive simplification ended up deleting the next PHI node we would
/// their preferred alignment from the beginning.
///
static unsigned enforceKnownAlignment(Value *V, unsigned Align,
- unsigned PrefAlign, const DataLayout *TD) {
+ unsigned PrefAlign,
+ const DataLayout &DL) {
V = V->stripPointerCasts();
if (AllocaInst *AI = dyn_cast<AllocaInst>(V)) {
// If the preferred alignment is greater than the natural stack alignment
// then don't round up. This avoids dynamic stack realignment.
- if (TD && TD->exceedsNaturalStackAlignment(PrefAlign))
+ if (DL.exceedsNaturalStackAlignment(PrefAlign))
return Align;
// If there is a requested alignment and if this is an alloca, round up.
if (AI->getAlignment() >= PrefAlign)
/// and it is more than the alignment of the ultimate object, see if we can
/// increase the alignment of the ultimate object, making this check succeed.
unsigned llvm::getOrEnforceKnownAlignment(Value *V, unsigned PrefAlign,
- const DataLayout *DL,
- AssumptionCache *AC,
+ const DataLayout &DL,
const Instruction *CxtI,
+ AssumptionCache *AC,
const DominatorTree *DT) {
assert(V->getType()->isPointerTy() &&
"getOrEnforceKnownAlignment expects a pointer!");
- unsigned BitWidth = DL ? DL->getPointerTypeSizeInBits(V->getType()) : 64;
+ unsigned BitWidth = DL.getPointerTypeSizeInBits(V->getType());
APInt KnownZero(BitWidth, 0), KnownOne(BitWidth, 0);
computeKnownBits(V, KnownZero, KnownOne, DL, 0, AC, CxtI, DT);
static PHINode *findPHIToPartitionLoops(Loop *L, AliasAnalysis *AA,
DominatorTree *DT,
AssumptionCache *AC) {
+ const DataLayout &DL = L->getHeader()->getModule()->getDataLayout();
for (BasicBlock::iterator I = L->getHeader()->begin(); isa<PHINode>(I); ) {
PHINode *PN = cast<PHINode>(I);
++I;
- if (Value *V = SimplifyInstruction(PN, nullptr, nullptr, DT, AC)) {
+ if (Value *V = SimplifyInstruction(PN, DL, nullptr, DT, AC)) {
// This is a degenerate PHI already, don't modify it!
PN->replaceAllUsesWith(V);
if (AA) AA->deleteValue(PN);
/// explicit if they accepted the analysis directly and then updated it.
static bool simplifyOneLoop(Loop *L, SmallVectorImpl<Loop *> &Worklist,
AliasAnalysis *AA, DominatorTree *DT, LoopInfo *LI,
- ScalarEvolution *SE, Pass *PP, const DataLayout *DL,
+ ScalarEvolution *SE, Pass *PP,
AssumptionCache *AC) {
bool Changed = false;
ReprocessLoop:
}
}
+ const DataLayout &DL = L->getHeader()->getModule()->getDataLayout();
+
// Scan over the PHI nodes in the loop header. Since they now have only two
// incoming values (the loop is canonicalized), we may have simplified the PHI
// down to 'X = phi [X, Y]', which should be replaced with 'Y'.
PHINode *PN;
for (BasicBlock::iterator I = L->getHeader()->begin();
(PN = dyn_cast<PHINode>(I++)); )
- if (Value *V = SimplifyInstruction(PN, nullptr, nullptr, DT, AC)) {
+ if (Value *V = SimplifyInstruction(PN, DL, nullptr, DT, AC)) {
if (AA) AA->deleteValue(PN);
if (SE) SE->forgetValue(PN);
PN->replaceAllUsesWith(V);
// The block has now been cleared of all instructions except for
// a comparison and a conditional branch. SimplifyCFG may be able
// to fold it now.
- if (!FoldBranchToCommonDest(BI, DL)) continue;
+ if (!FoldBranchToCommonDest(BI))
+ continue;
// Success. The block is now dead, so remove it from the loop,
// update the dominator tree and delete it.
bool llvm::simplifyLoop(Loop *L, DominatorTree *DT, LoopInfo *LI, Pass *PP,
AliasAnalysis *AA, ScalarEvolution *SE,
- const DataLayout *DL, AssumptionCache *AC) {
+ AssumptionCache *AC) {
bool Changed = false;
// Worklist maintains our depth-first queue of loops in this nest to process.
while (!Worklist.empty())
Changed |= simplifyOneLoop(Worklist.pop_back_val(), Worklist, AA, DT, LI,
- SE, PP, DL, AC);
+ SE, PP, AC);
return Changed;
}
DominatorTree *DT;
LoopInfo *LI;
ScalarEvolution *SE;
- const DataLayout *DL;
AssumptionCache *AC;
bool runOnFunction(Function &F) override;
LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
SE = getAnalysisIfAvailable<ScalarEvolution>();
- DL = &F.getParent()->getDataLayout();
AC = &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
// Simplify each loop nest in the function.
for (LoopInfo::iterator I = LI->begin(), E = LI->end(); I != E; ++I)
- Changed |= simplifyLoop(*I, DT, LI, this, AA, SE, DL, AC);
+ Changed |= simplifyLoop(*I, DT, LI, this, AA, SE, AC);
return Changed;
}
// At this point, the code is well formed. We now do a quick sweep over the
// inserted code, doing constant propagation and dead code elimination as we
// go.
+ const DataLayout &DL = Header->getModule()->getDataLayout();
const std::vector<BasicBlock*> &NewLoopBlocks = L->getBlocks();
for (std::vector<BasicBlock*>::const_iterator BB = NewLoopBlocks.begin(),
BBE = NewLoopBlocks.end(); BB != BBE; ++BB)
if (isInstructionTriviallyDead(Inst))
(*BB)->getInstList().erase(Inst);
- else if (Value *V = SimplifyInstruction(Inst))
+ else if (Value *V = SimplifyInstruction(Inst, DL))
if (LI->replacementPreservesLCSSAForm(Inst, V)) {
Inst->replaceAllUsesWith(V);
(*BB)->getInstList().erase(Inst);
if (!OuterL && !CompletelyUnroll)
OuterL = L;
if (OuterL) {
- const DataLayout &DL = F->getParent()->getDataLayout();
- simplifyLoop(OuterL, DT, LI, PP, /*AliasAnalysis*/ nullptr, SE, &DL, AC);
+ simplifyLoop(OuterL, DT, LI, PP, /*AliasAnalysis*/ nullptr, SE, AC);
// LCSSA must be performed on the outermost affected loop. The unrolled
// loop's last loop latch is guaranteed to be in the outermost loop after
#include "llvm/IR/BasicBlock.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Metadata.h"
+#include "llvm/IR/Module.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Scalar.h"
BasicBlock *PEnd = SplitEdge(PH, Header, DT, LI);
BasicBlock *NewPH = SplitBlock(PEnd, PEnd->getTerminator(), DT, LI);
BranchInst *PreHeaderBR = cast<BranchInst>(PH->getTerminator());
+ const DataLayout &DL = Header->getModule()->getDataLayout();
// Compute the number of extra iterations required, which is:
// extra iterations = run-time trip count % (loop unroll factor + 1)
- SCEVExpander Expander(*SE, "loop-unroll");
+ SCEVExpander Expander(*SE, DL, "loop-unroll");
Value *TripCount = Expander.expandCodeFor(TripCountSC, TripCountSC->getType(),
PreHeaderBR);
Value *BECount = Expander.expandCodeFor(BECountSC, BECountSC->getType(),
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Metadata.h"
+#include "llvm/IR/Module.h"
#include "llvm/Transforms/Utils/Local.h"
#include <algorithm>
#include <queue>
A->eraseFromParent();
}
+ const DataLayout &DL = F.getParent()->getDataLayout();
+
// Remove alloca's dbg.declare instrinsics from the function.
for (unsigned i = 0, e = AllocaDbgDeclares.size(); i != e; ++i)
if (DbgDeclareInst *DDI = AllocaDbgDeclares[i])
PHINode *PN = I->second;
// If this PHI node merges one value and/or undefs, get the value.
- if (Value *V = SimplifyInstruction(PN, nullptr, nullptr, &DT, AC)) {
+ if (Value *V = SimplifyInstruction(PN, DL, nullptr, &DT, AC)) {
if (AST && PN->getType()->isPointerTy())
AST->deleteValue(PN);
PN->replaceAllUsesWith(V);
#include "llvm/IR/Constants.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
+#include "llvm/IR/Module.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
// See if the PHI node can be merged to a single value. This can happen in
// loop cases when we get a PHI of itself and one other value.
- if (Value *V = SimplifyInstruction(InsertedPHI)) {
+ if (Value *V =
+ SimplifyInstruction(InsertedPHI, BB->getModule()->getDataLayout())) {
InsertedPHI->eraseFromParent();
return V;
}
class SimplifyCFGOpt {
const TargetTransformInfo &TTI;
+ const DataLayout &DL;
unsigned BonusInstThreshold;
- const DataLayout *const DL;
AssumptionCache *AC;
Value *isValueEqualityComparison(TerminatorInst *TI);
BasicBlock *GetValueEqualityComparisonCases(TerminatorInst *TI,
bool SimplifyCondBranch(BranchInst *BI, IRBuilder <>&Builder);
public:
- SimplifyCFGOpt(const TargetTransformInfo &TTI, unsigned BonusInstThreshold,
- const DataLayout *DL, AssumptionCache *AC)
- : TTI(TTI), BonusInstThreshold(BonusInstThreshold), DL(DL), AC(AC) {}
+ SimplifyCFGOpt(const TargetTransformInfo &TTI, const DataLayout &DL,
+ unsigned BonusInstThreshold, AssumptionCache *AC)
+ : TTI(TTI), DL(DL), BonusInstThreshold(BonusInstThreshold), AC(AC) {}
bool run(BasicBlock *BB);
};
}
/// given instruction, which is assumed to be safe to speculate. TCC_Free means
/// cheap, TCC_Basic means less cheap, and TCC_Expensive means prohibitively
/// expensive.
-static unsigned ComputeSpeculationCost(const User *I, const DataLayout *DL,
+static unsigned ComputeSpeculationCost(const User *I,
const TargetTransformInfo &TTI) {
- assert(isSafeToSpeculativelyExecute(I, DL) &&
+ assert(isSafeToSpeculativelyExecute(I) &&
"Instruction is not safe to speculatively execute!");
return TTI.getUserCost(I);
}
static bool DominatesMergePoint(Value *V, BasicBlock *BB,
SmallPtrSetImpl<Instruction*> *AggressiveInsts,
unsigned &CostRemaining,
- const DataLayout *DL,
const TargetTransformInfo &TTI) {
Instruction *I = dyn_cast<Instruction>(V);
if (!I) {
// Okay, it looks like the instruction IS in the "condition". Check to
// see if it's a cheap instruction to unconditionally compute, and if it
// only uses stuff defined outside of the condition. If so, hoist it out.
- if (!isSafeToSpeculativelyExecute(I, DL))
+ if (!isSafeToSpeculativelyExecute(I))
return false;
- unsigned Cost = ComputeSpeculationCost(I, DL, TTI);
+ unsigned Cost = ComputeSpeculationCost(I, TTI);
if (Cost > CostRemaining)
return false;
// Okay, we can only really hoist these out if their operands do
// not take us over the cost threshold.
for (User::op_iterator i = I->op_begin(), e = I->op_end(); i != e; ++i)
- if (!DominatesMergePoint(*i, BB, AggressiveInsts, CostRemaining, DL, TTI))
+ if (!DominatesMergePoint(*i, BB, AggressiveInsts, CostRemaining, TTI))
return false;
// Okay, it's safe to do this! Remember this instruction.
AggressiveInsts->insert(I);
/// GetConstantInt - Extract ConstantInt from value, looking through IntToPtr
/// and PointerNullValue. Return NULL if value is not a constant int.
-static ConstantInt *GetConstantInt(Value *V, const DataLayout *DL) {
+static ConstantInt *GetConstantInt(Value *V, const DataLayout &DL) {
// Normal constant int.
ConstantInt *CI = dyn_cast<ConstantInt>(V);
- if (CI || !DL || !isa<Constant>(V) || !V->getType()->isPointerTy())
+ if (CI || !isa<Constant>(V) || !V->getType()->isPointerTy())
return CI;
// This is some kind of pointer constant. Turn it into a pointer-sized
// ConstantInt if possible.
- IntegerType *PtrTy = cast<IntegerType>(DL->getIntPtrType(V->getType()));
+ IntegerType *PtrTy = cast<IntegerType>(DL.getIntPtrType(V->getType()));
// Null pointer means 0, see SelectionDAGBuilder::getValue(const Value*).
if (isa<ConstantPointerNull>(V))
/// while for a chain of '&&' it will build the set elements that make the test
/// fail.
struct ConstantComparesGatherer {
-
+ const DataLayout &DL;
Value *CompValue; /// Value found for the switch comparison
Value *Extra; /// Extra clause to be checked before the switch
SmallVector<ConstantInt *, 8> Vals; /// Set of integers to match in switch
unsigned UsedICmps; /// Number of comparisons matched in the and/or chain
/// Construct and compute the result for the comparison instruction Cond
- ConstantComparesGatherer(Instruction *Cond, const DataLayout *DL)
- : CompValue(nullptr), Extra(nullptr), UsedICmps(0) {
- gather(Cond, DL);
+ ConstantComparesGatherer(Instruction *Cond, const DataLayout &DL)
+ : DL(DL), CompValue(nullptr), Extra(nullptr), UsedICmps(0) {
+ gather(Cond);
}
/// Prevent copy
/// against is placed in CompValue.
/// If CompValue is already set, the function is expected to fail if a match
/// is found but the value compared to is different.
- bool matchInstruction(Instruction *I, const DataLayout *DL, bool isEQ) {
+ bool matchInstruction(Instruction *I, bool isEQ) {
// If this is an icmp against a constant, handle this as one of the cases.
ICmpInst *ICI;
ConstantInt *C;
/// the value being compared, and stick the list constants into the Vals
/// vector.
/// One "Extra" case is allowed to differ from the other.
- void gather(Value *V, const DataLayout *DL) {
+ void gather(Value *V) {
Instruction *I = dyn_cast<Instruction>(V);
bool isEQ = (I->getOpcode() == Instruction::Or);
}
// Try to match the current instruction
- if (matchInstruction(I, DL, isEQ))
+ if (matchInstruction(I, isEQ))
// Match succeed, continue the loop
continue;
}
CV = SI->getCondition();
} else if (BranchInst *BI = dyn_cast<BranchInst>(TI))
if (BI->isConditional() && BI->getCondition()->hasOneUse())
- if (ICmpInst *ICI = dyn_cast<ICmpInst>(BI->getCondition()))
+ if (ICmpInst *ICI = dyn_cast<ICmpInst>(BI->getCondition())) {
if (ICI->isEquality() && GetConstantInt(ICI->getOperand(1), DL))
CV = ICI->getOperand(0);
+ }
// Unwrap any lossless ptrtoint cast.
- if (DL && CV) {
+ if (CV) {
if (PtrToIntInst *PTII = dyn_cast<PtrToIntInst>(CV)) {
Value *Ptr = PTII->getPointerOperand();
- if (PTII->getType() == DL->getIntPtrType(Ptr->getType()))
+ if (PTII->getType() == DL.getIntPtrType(Ptr->getType()))
CV = Ptr;
}
}
Builder.SetInsertPoint(PTI);
// Convert pointer to int before we switch.
if (CV->getType()->isPointerTy()) {
- assert(DL && "Cannot switch on pointer without DataLayout");
- CV = Builder.CreatePtrToInt(CV, DL->getIntPtrType(CV->getType()),
+ CV = Builder.CreatePtrToInt(CV, DL.getIntPtrType(CV->getType()),
"magicptr");
}
/// HoistThenElseCodeToIf - Given a conditional branch that goes to BB1 and
/// BB2, hoist any common code in the two blocks up into the branch block. The
/// caller of this function guarantees that BI's block dominates BB1 and BB2.
-static bool HoistThenElseCodeToIf(BranchInst *BI, const DataLayout *DL,
+static bool HoistThenElseCodeToIf(BranchInst *BI,
const TargetTransformInfo &TTI) {
// This does very trivial matching, with limited scanning, to find identical
// instructions in the two blocks. In particular, we don't want to get into
passingValueIsAlwaysUndefined(BB2V, PN))
return Changed;
- if (isa<ConstantExpr>(BB1V) && !isSafeToSpeculativelyExecute(BB1V, DL))
+ if (isa<ConstantExpr>(BB1V) && !isSafeToSpeculativelyExecute(BB1V))
return Changed;
- if (isa<ConstantExpr>(BB2V) && !isSafeToSpeculativelyExecute(BB2V, DL))
+ if (isa<ConstantExpr>(BB2V) && !isSafeToSpeculativelyExecute(BB2V))
return Changed;
}
}
///
/// \returns true if the conditional block is removed.
static bool SpeculativelyExecuteBB(BranchInst *BI, BasicBlock *ThenBB,
- const DataLayout *DL,
const TargetTransformInfo &TTI) {
// Be conservative for now. FP select instruction can often be expensive.
Value *BrCond = BI->getCondition();
return false;
// Don't hoist the instruction if it's unsafe or expensive.
- if (!isSafeToSpeculativelyExecute(I, DL) &&
- !(HoistCondStores &&
- (SpeculatedStoreValue = isSafeToSpeculateStore(I, BB, ThenBB,
- EndBB))))
+ if (!isSafeToSpeculativelyExecute(I) &&
+ !(HoistCondStores && (SpeculatedStoreValue = isSafeToSpeculateStore(
+ I, BB, ThenBB, EndBB))))
return false;
if (!SpeculatedStoreValue &&
- ComputeSpeculationCost(I, DL, TTI) > PHINodeFoldingThreshold *
- TargetTransformInfo::TCC_Basic)
+ ComputeSpeculationCost(I, TTI) >
+ PHINodeFoldingThreshold * TargetTransformInfo::TCC_Basic)
return false;
// Store the store speculation candidate.
if (!OrigCE && !ThenCE)
continue; // Known safe and cheap.
- if ((ThenCE && !isSafeToSpeculativelyExecute(ThenCE, DL)) ||
- (OrigCE && !isSafeToSpeculativelyExecute(OrigCE, DL)))
+ if ((ThenCE && !isSafeToSpeculativelyExecute(ThenCE)) ||
+ (OrigCE && !isSafeToSpeculativelyExecute(OrigCE)))
return false;
- unsigned OrigCost = OrigCE ? ComputeSpeculationCost(OrigCE, DL, TTI) : 0;
- unsigned ThenCost = ThenCE ? ComputeSpeculationCost(ThenCE, DL, TTI) : 0;
+ unsigned OrigCost = OrigCE ? ComputeSpeculationCost(OrigCE, TTI) : 0;
+ unsigned ThenCost = ThenCE ? ComputeSpeculationCost(ThenCE, TTI) : 0;
unsigned MaxCost = 2 * PHINodeFoldingThreshold *
TargetTransformInfo::TCC_Basic;
if (OrigCost + ThenCost > MaxCost)
/// that is defined in the same block as the branch and if any PHI entries are
/// constants, thread edges corresponding to that entry to be branches to their
/// ultimate destination.
-static bool FoldCondBranchOnPHI(BranchInst *BI, const DataLayout *DL) {
+static bool FoldCondBranchOnPHI(BranchInst *BI, const DataLayout &DL) {
BasicBlock *BB = BI->getParent();
PHINode *PN = dyn_cast<PHINode>(BI->getCondition());
// NOTE: we currently cannot transform this case if the PHI node is used
/// FoldTwoEntryPHINode - Given a BB that starts with the specified two-entry
/// PHI node, see if we can eliminate it.
-static bool FoldTwoEntryPHINode(PHINode *PN, const DataLayout *DL,
- const TargetTransformInfo &TTI) {
+static bool FoldTwoEntryPHINode(PHINode *PN, const TargetTransformInfo &TTI,
+ const DataLayout &DL) {
// Ok, this is a two entry PHI node. Check to see if this is a simple "if
// statement", which has a very simple dominance structure. Basically, we
// are trying to find the condition that is being branched on, which
}
if (!DominatesMergePoint(PN->getIncomingValue(0), BB, &AggressiveInsts,
- MaxCostVal0, DL, TTI) ||
+ MaxCostVal0, TTI) ||
!DominatesMergePoint(PN->getIncomingValue(1), BB, &AggressiveInsts,
- MaxCostVal1, DL, TTI))
+ MaxCostVal1, TTI))
return false;
}
/// FoldBranchToCommonDest - If this basic block is simple enough, and if a
/// predecessor branches to us and one of our successors, fold the block into
/// the predecessor and use logical operations to pick the right destination.
-bool llvm::FoldBranchToCommonDest(BranchInst *BI, const DataLayout *DL,
- unsigned BonusInstThreshold) {
+bool llvm::FoldBranchToCommonDest(BranchInst *BI, unsigned BonusInstThreshold) {
BasicBlock *BB = BI->getParent();
Instruction *Cond = nullptr;
// Ignore dbg intrinsics.
if (isa<DbgInfoIntrinsic>(I))
continue;
- if (!I->hasOneUse() || !isSafeToSpeculativelyExecute(I, DL))
+ if (!I->hasOneUse() || !isSafeToSpeculativelyExecute(I))
return false;
// I has only one use and can be executed unconditionally.
Instruction *User = dyn_cast<Instruction>(I->user_back());
/// We prefer to split the edge to 'end' so that there is a true/false entry to
/// the PHI, merging the third icmp into the switch.
static bool TryToSimplifyUncondBranchWithICmpInIt(
- ICmpInst *ICI, IRBuilder<> &Builder, const TargetTransformInfo &TTI,
- unsigned BonusInstThreshold, const DataLayout *DL, AssumptionCache *AC) {
+ ICmpInst *ICI, IRBuilder<> &Builder, const DataLayout &DL,
+ const TargetTransformInfo &TTI, unsigned BonusInstThreshold,
+ AssumptionCache *AC) {
BasicBlock *BB = ICI->getParent();
// If the block has any PHIs in it or the icmp has multiple uses, it is too
ICI->eraseFromParent();
}
// BB is now empty, so it is likely to simplify away.
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
}
// Ok, the block is reachable from the default dest. If the constant we're
ICI->replaceAllUsesWith(V);
ICI->eraseFromParent();
// BB is now empty, so it is likely to simplify away.
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
}
// The use of the icmp has to be in the 'end' block, by the only PHI node in
/// SimplifyBranchOnICmpChain - The specified branch is a conditional branch.
/// Check to see if it is branching on an or/and chain of icmp instructions, and
/// fold it into a switch instruction if so.
-static bool SimplifyBranchOnICmpChain(BranchInst *BI, const DataLayout *DL,
- IRBuilder<> &Builder) {
+static bool SimplifyBranchOnICmpChain(BranchInst *BI, IRBuilder<> &Builder,
+ const DataLayout &DL) {
Instruction *Cond = dyn_cast<Instruction>(BI->getCondition());
if (!Cond) return false;
Builder.SetInsertPoint(BI);
// Convert pointer to int before we switch.
if (CompVal->getType()->isPointerTy()) {
- assert(DL && "Cannot switch on pointer without DataLayout");
- CompVal = Builder.CreatePtrToInt(CompVal,
- DL->getIntPtrType(CompVal->getType()),
- "magicptr");
+ CompVal = Builder.CreatePtrToInt(
+ CompVal, DL.getIntPtrType(CompVal->getType()), "magicptr");
}
// Create the new switch instruction now.
/// EliminateDeadSwitchCases - Compute masked bits for the condition of a switch
/// and use it to remove dead cases.
-static bool EliminateDeadSwitchCases(SwitchInst *SI, const DataLayout *DL,
- AssumptionCache *AC) {
+static bool EliminateDeadSwitchCases(SwitchInst *SI, AssumptionCache *AC,
+ const DataLayout &DL) {
Value *Cond = SI->getCondition();
unsigned Bits = Cond->getType()->getIntegerBitWidth();
APInt KnownZero(Bits, 0), KnownOne(Bits, 0);
/// constant or can be replaced by constants from the ConstantPool. Returns the
/// resulting constant on success, 0 otherwise.
static Constant *
-ConstantFold(Instruction *I,
- const SmallDenseMap<Value *, Constant *> &ConstantPool,
- const DataLayout *DL) {
+ConstantFold(Instruction *I, const DataLayout &DL,
+ const SmallDenseMap<Value *, Constant *> &ConstantPool) {
if (SelectInst *Select = dyn_cast<SelectInst>(I)) {
Constant *A = LookupConstant(Select->getCondition(), ConstantPool);
if (!A)
return nullptr;
}
- if (CmpInst *Cmp = dyn_cast<CmpInst>(I))
+ if (CmpInst *Cmp = dyn_cast<CmpInst>(I)) {
return ConstantFoldCompareInstOperands(Cmp->getPredicate(), COps[0],
COps[1], DL);
+ }
return ConstantFoldInstOperands(I->getOpcode(), I->getType(), COps, DL);
}
/// destionations CaseDest corresponding to value CaseVal (0 for the default
/// case), of a switch instruction SI.
static bool
-GetCaseResults(SwitchInst *SI,
- ConstantInt *CaseVal,
- BasicBlock *CaseDest,
+GetCaseResults(SwitchInst *SI, ConstantInt *CaseVal, BasicBlock *CaseDest,
BasicBlock **CommonDest,
- SmallVectorImpl<std::pair<PHINode *, Constant *> > &Res,
- const DataLayout *DL) {
+ SmallVectorImpl<std::pair<PHINode *, Constant *>> &Res,
+ const DataLayout &DL) {
// The block from which we enter the common destination.
BasicBlock *Pred = SI->getParent();
} else if (isa<DbgInfoIntrinsic>(I)) {
// Skip debug intrinsic.
continue;
- } else if (Constant *C = ConstantFold(I, ConstantPool, DL)) {
+ } else if (Constant *C = ConstantFold(I, DL, ConstantPool)) {
// Instruction is side-effect free and constant.
// If the instruction has uses outside this block or a phi node slot for
// results for the PHI node of the common destination block for a switch
// instruction. Returns false if multiple PHI nodes have been found or if
// there is not a common destination block for the switch.
-static bool InitializeUniqueCases(
- SwitchInst *SI, const DataLayout *DL, PHINode *&PHI,
- BasicBlock *&CommonDest,
- SwitchCaseResultVectorTy &UniqueResults,
- Constant *&DefaultResult) {
+static bool InitializeUniqueCases(SwitchInst *SI, PHINode *&PHI,
+ BasicBlock *&CommonDest,
+ SwitchCaseResultVectorTy &UniqueResults,
+ Constant *&DefaultResult,
+ const DataLayout &DL) {
for (auto &I : SI->cases()) {
ConstantInt *CaseVal = I.getCaseValue();
/// phi nodes in a common successor block with only two different
/// constant values, replace the switch with select.
static bool SwitchToSelect(SwitchInst *SI, IRBuilder<> &Builder,
- const DataLayout *DL, AssumptionCache *AC) {
+ AssumptionCache *AC, const DataLayout &DL) {
Value *const Cond = SI->getCondition();
PHINode *PHI = nullptr;
BasicBlock *CommonDest = nullptr;
Constant *DefaultResult;
SwitchCaseResultVectorTy UniqueResults;
// Collect all the cases that will deliver the same value from the switch.
- if (!InitializeUniqueCases(SI, DL, PHI, CommonDest, UniqueResults,
- DefaultResult))
+ if (!InitializeUniqueCases(SI, PHI, CommonDest, UniqueResults, DefaultResult,
+ DL))
return false;
// Selects choose between maximum two values.
if (UniqueResults.size() != 2)
/// SwitchLookupTable - Create a lookup table to use as a switch replacement
/// with the contents of Values, using DefaultValue to fill any holes in the
/// table.
- SwitchLookupTable(Module &M,
- uint64_t TableSize,
- ConstantInt *Offset,
- const SmallVectorImpl<std::pair<ConstantInt*, Constant*> >& Values,
- Constant *DefaultValue,
- const DataLayout *DL);
+ SwitchLookupTable(
+ Module &M, uint64_t TableSize, ConstantInt *Offset,
+ const SmallVectorImpl<std::pair<ConstantInt *, Constant *>> &Values,
+ Constant *DefaultValue, const DataLayout &DL);
/// BuildLookup - Build instructions with Builder to retrieve the value at
/// the position given by Index in the lookup table.
/// WouldFitInRegister - Return true if a table with TableSize elements of
/// type ElementType would fit in a target-legal register.
- static bool WouldFitInRegister(const DataLayout *DL,
- uint64_t TableSize,
+ static bool WouldFitInRegister(const DataLayout &DL, uint64_t TableSize,
const Type *ElementType);
private:
};
}
-SwitchLookupTable::SwitchLookupTable(Module &M,
- uint64_t TableSize,
- ConstantInt *Offset,
- const SmallVectorImpl<std::pair<ConstantInt*, Constant*> >& Values,
- Constant *DefaultValue,
- const DataLayout *DL)
+SwitchLookupTable::SwitchLookupTable(
+ Module &M, uint64_t TableSize, ConstantInt *Offset,
+ const SmallVectorImpl<std::pair<ConstantInt *, Constant *>> &Values,
+ Constant *DefaultValue, const DataLayout &DL)
: SingleValue(nullptr), BitMap(nullptr), BitMapElementTy(nullptr),
LinearOffset(nullptr), LinearMultiplier(nullptr), Array(nullptr) {
assert(Values.size() && "Can't build lookup table without values!");
llvm_unreachable("Unknown lookup table kind!");
}
-bool SwitchLookupTable::WouldFitInRegister(const DataLayout *DL,
+bool SwitchLookupTable::WouldFitInRegister(const DataLayout &DL,
uint64_t TableSize,
const Type *ElementType) {
- if (!DL)
- return false;
const IntegerType *IT = dyn_cast<IntegerType>(ElementType);
if (!IT)
return false;
// Avoid overflow, fitsInLegalInteger uses unsigned int for the width.
if (TableSize >= UINT_MAX/IT->getBitWidth())
return false;
- return DL->fitsInLegalInteger(TableSize * IT->getBitWidth());
+ return DL.fitsInLegalInteger(TableSize * IT->getBitWidth());
}
/// ShouldBuildLookupTable - Determine whether a lookup table should be built
/// for this switch, based on the number of cases, size of the table and the
/// types of the results.
-static bool ShouldBuildLookupTable(SwitchInst *SI,
- uint64_t TableSize,
- const TargetTransformInfo &TTI,
- const DataLayout *DL,
- const SmallDenseMap<PHINode*, Type*>& ResultTypes) {
+static bool
+ShouldBuildLookupTable(SwitchInst *SI, uint64_t TableSize,
+ const TargetTransformInfo &TTI, const DataLayout &DL,
+ const SmallDenseMap<PHINode *, Type *> &ResultTypes) {
if (SI->getNumCases() > TableSize || TableSize >= UINT64_MAX / 10)
return false; // TableSize overflowed, or mul below might overflow.
/// SwitchToLookupTable - If the switch is only used to initialize one or more
/// phi nodes in a common successor block with different constant values,
/// replace the switch with lookup tables.
-static bool SwitchToLookupTable(SwitchInst *SI,
- IRBuilder<> &Builder,
- const TargetTransformInfo &TTI,
- const DataLayout* DL) {
+static bool SwitchToLookupTable(SwitchInst *SI, IRBuilder<> &Builder,
+ const DataLayout &DL,
+ const TargetTransformInfo &TTI) {
assert(SI->getNumCases() > 1 && "Degenerate switch?");
// Only build lookup table when we have a target that supports it.
// or a bitmask that fits in a register.
SmallVector<std::pair<PHINode*, Constant*>, 4> DefaultResultsList;
bool HasDefaultResults = GetCaseResults(SI, nullptr, SI->getDefaultDest(),
- &CommonDest, DefaultResultsList, DL);
+ &CommonDest, DefaultResultsList, DL);
bool NeedMask = (TableHasHoles && !HasDefaultResults);
if (NeedMask) {
// As an extra penalty for the validity test we require more cases.
if (SI->getNumCases() < 4) // FIXME: Find best threshold value (benchmark).
return false;
- if (!(DL && DL->fitsInLegalInteger(TableSize)))
+ if (!DL.fitsInLegalInteger(TableSize))
return false;
}
// see if that predecessor totally determines the outcome of this switch.
if (BasicBlock *OnlyPred = BB->getSinglePredecessor())
if (SimplifyEqualityComparisonWithOnlyPredecessor(SI, OnlyPred, Builder))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
Value *Cond = SI->getCondition();
if (SelectInst *Select = dyn_cast<SelectInst>(Cond))
if (SimplifySwitchOnSelect(SI, Select))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
// If the block only contains the switch, see if we can fold the block
// away into any preds.
++BBI;
if (SI == &*BBI)
if (FoldValueComparisonIntoPredecessors(SI, Builder))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
}
// Try to transform the switch into an icmp and a branch.
if (TurnSwitchRangeIntoICmp(SI, Builder))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
// Remove unreachable cases.
- if (EliminateDeadSwitchCases(SI, DL, AC))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ if (EliminateDeadSwitchCases(SI, AC, DL))
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
- if (SwitchToSelect(SI, Builder, DL, AC))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ if (SwitchToSelect(SI, Builder, AC, DL))
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
if (ForwardSwitchConditionToPHI(SI))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
- if (SwitchToLookupTable(SI, Builder, TTI, DL))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ if (SwitchToLookupTable(SI, Builder, DL, TTI))
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
return false;
}
if (SelectInst *SI = dyn_cast<SelectInst>(IBI->getAddress())) {
if (SimplifyIndirectBrOnSelect(IBI, SI))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
}
return Changed;
}
for (++I; isa<DbgInfoIntrinsic>(I); ++I)
;
if (I->isTerminator() &&
- TryToSimplifyUncondBranchWithICmpInIt(ICI, Builder, TTI,
- BonusInstThreshold, DL, AC))
+ TryToSimplifyUncondBranchWithICmpInIt(ICI, Builder, DL, TTI,
+ BonusInstThreshold, AC))
return true;
}
// branches to us and our successor, fold the comparison into the
// predecessor and use logical operations to update the incoming value
// for PHI nodes in common successor.
- if (FoldBranchToCommonDest(BI, DL, BonusInstThreshold))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ if (FoldBranchToCommonDest(BI, BonusInstThreshold))
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
return false;
}
// switch.
if (BasicBlock *OnlyPred = BB->getSinglePredecessor())
if (SimplifyEqualityComparisonWithOnlyPredecessor(BI, OnlyPred, Builder))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
// This block must be empty, except for the setcond inst, if it exists.
// Ignore dbg intrinsics.
++I;
if (&*I == BI) {
if (FoldValueComparisonIntoPredecessors(BI, Builder))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
} else if (&*I == cast<Instruction>(BI->getCondition())){
++I;
// Ignore dbg intrinsics.
while (isa<DbgInfoIntrinsic>(I))
++I;
if (&*I == BI && FoldValueComparisonIntoPredecessors(BI, Builder))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
}
}
// Try to turn "br (X == 0 | X == 1), T, F" into a switch instruction.
- if (SimplifyBranchOnICmpChain(BI, DL, Builder))
+ if (SimplifyBranchOnICmpChain(BI, Builder, DL))
return true;
// If this basic block is ONLY a compare and a branch, and if a predecessor
// branches to us and one of our successors, fold the comparison into the
// predecessor and use logical operations to pick the right destination.
- if (FoldBranchToCommonDest(BI, DL, BonusInstThreshold))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ if (FoldBranchToCommonDest(BI, BonusInstThreshold))
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
// We have a conditional branch to two blocks that are only reachable
// from BI. We know that the condbr dominates the two blocks, so see if
// can hoist it up to the branching block.
if (BI->getSuccessor(0)->getSinglePredecessor()) {
if (BI->getSuccessor(1)->getSinglePredecessor()) {
- if (HoistThenElseCodeToIf(BI, DL, TTI))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ if (HoistThenElseCodeToIf(BI, TTI))
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
} else {
// If Successor #1 has multiple preds, we may be able to conditionally
// execute Successor #0 if it branches to Successor #1.
TerminatorInst *Succ0TI = BI->getSuccessor(0)->getTerminator();
if (Succ0TI->getNumSuccessors() == 1 &&
Succ0TI->getSuccessor(0) == BI->getSuccessor(1))
- if (SpeculativelyExecuteBB(BI, BI->getSuccessor(0), DL, TTI))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ if (SpeculativelyExecuteBB(BI, BI->getSuccessor(0), TTI))
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
}
} else if (BI->getSuccessor(1)->getSinglePredecessor()) {
// If Successor #0 has multiple preds, we may be able to conditionally
TerminatorInst *Succ1TI = BI->getSuccessor(1)->getTerminator();
if (Succ1TI->getNumSuccessors() == 1 &&
Succ1TI->getSuccessor(0) == BI->getSuccessor(0))
- if (SpeculativelyExecuteBB(BI, BI->getSuccessor(1), DL, TTI))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ if (SpeculativelyExecuteBB(BI, BI->getSuccessor(1), TTI))
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
}
// If this is a branch on a phi node in the current block, thread control
if (PHINode *PN = dyn_cast<PHINode>(BI->getCondition()))
if (PN->getParent() == BI->getParent())
if (FoldCondBranchOnPHI(BI, DL))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
// Scan predecessor blocks for conditional branches.
for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI)
if (BranchInst *PBI = dyn_cast<BranchInst>((*PI)->getTerminator()))
if (PBI != BI && PBI->isConditional())
if (SimplifyCondBranchToCondBranch(PBI, BI))
- return SimplifyCFG(BB, TTI, BonusInstThreshold, DL, AC) | true;
+ return SimplifyCFG(BB, TTI, BonusInstThreshold, AC) | true;
return false;
}
// eliminate it, do so now.
if (PHINode *PN = dyn_cast<PHINode>(BB->begin()))
if (PN->getNumIncomingValues() == 2)
- Changed |= FoldTwoEntryPHINode(PN, DL, TTI);
+ Changed |= FoldTwoEntryPHINode(PN, TTI, DL);
Builder.SetInsertPoint(BB->getTerminator());
if (BranchInst *BI = dyn_cast<BranchInst>(BB->getTerminator())) {
/// of the CFG. It returns true if a modification was made.
///
bool llvm::SimplifyCFG(BasicBlock *BB, const TargetTransformInfo &TTI,
- unsigned BonusInstThreshold, const DataLayout *DL,
- AssumptionCache *AC) {
- return SimplifyCFGOpt(TTI, BonusInstThreshold, DL, AC).run(BB);
+ unsigned BonusInstThreshold, AssumptionCache *AC) {
+ return SimplifyCFGOpt(TTI, BB->getModule()->getDataLayout(),
+ BonusInstThreshold, AC).run(BB);
}
continue;
// Don't waste time simplifying unused instructions.
if (!I->use_empty())
- if (Value *V = SimplifyInstruction(I, &DL, TLI, DT, AC)) {
+ if (Value *V = SimplifyInstruction(I, DL, TLI, DT, AC)) {
// Mark all uses for resimplification next time round the loop.
for (User *U : I->users())
Next->insert(cast<Instruction>(U));
/// string/memory copying library function \p Func.
/// Acceptable functions are st[rp][n]?cpy, memove, memcpy, and memset.
/// Their fortified (_chk) counterparts are also accepted.
-static bool checkStringCopyLibFuncSignature(Function *F, LibFunc::Func Func,
- const DataLayout *DL) {
+static bool checkStringCopyLibFuncSignature(Function *F, LibFunc::Func Func) {
+ const DataLayout &DL = F->getParent()->getDataLayout();
FunctionType *FT = F->getFunctionType();
LLVMContext &Context = F->getContext();
Type *PCharTy = Type::getInt8PtrTy(Context);
- Type *SizeTTy = DL ? DL->getIntPtrType(Context) : nullptr;
+ Type *SizeTTy = DL.getIntPtrType(Context);
unsigned NumParams = FT->getNumParams();
// All string libfuncs return the same type as the first parameter.
if (Len == 0)
return Dst;
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
-
return emitStrLenMemCpy(Src, Dst, Len, B);
}
// We have enough information to now generate the memcpy call to do the
// concatenation for us. Make a memcpy to copy the nul byte with align = 1.
- B.CreateMemCpy(
- CpyDst, Src,
- ConstantInt::get(DL->getIntPtrType(Src->getContext()), Len + 1), 1);
+ B.CreateMemCpy(CpyDst, Src,
+ ConstantInt::get(DL.getIntPtrType(Src->getContext()), Len + 1),
+ 1);
return Dst;
}
if (SrcLen == 0 || Len == 0)
return Dst;
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
-
// We don't optimize this case
if (Len < SrcLen)
return nullptr;
// of the input string and turn this into memchr.
ConstantInt *CharC = dyn_cast<ConstantInt>(CI->getArgOperand(1));
if (!CharC) {
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
-
uint64_t Len = GetStringLength(SrcStr);
if (Len == 0 || !FT->getParamType(1)->isIntegerTy(32)) // memchr needs i32.
return nullptr;
- return EmitMemChr(
- SrcStr, CI->getArgOperand(1), // include nul.
- ConstantInt::get(DL->getIntPtrType(CI->getContext()), Len), B, DL, TLI);
+ return EmitMemChr(SrcStr, CI->getArgOperand(1), // include nul.
+ ConstantInt::get(DL.getIntPtrType(CI->getContext()), Len),
+ B, DL, TLI);
}
// Otherwise, the character is a constant, see if the first argument is
// a string literal. If so, we can constant fold.
StringRef Str;
if (!getConstantStringInfo(SrcStr, Str)) {
- if (DL && CharC->isZero()) // strchr(p, 0) -> p + strlen(p)
+ if (CharC->isZero()) // strchr(p, 0) -> p + strlen(p)
return B.CreateGEP(SrcStr, EmitStrLen(SrcStr, B, DL, TLI), "strchr");
return nullptr;
}
StringRef Str;
if (!getConstantStringInfo(SrcStr, Str)) {
// strrchr(s, 0) -> strchr(s, 0)
- if (DL && CharC->isZero())
- return EmitStrChr(SrcStr, '\0', B, DL, TLI);
+ if (CharC->isZero())
+ return EmitStrChr(SrcStr, '\0', B, TLI);
return nullptr;
}
uint64_t Len1 = GetStringLength(Str1P);
uint64_t Len2 = GetStringLength(Str2P);
if (Len1 && Len2) {
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
-
return EmitMemCmp(Str1P, Str2P,
- ConstantInt::get(DL->getIntPtrType(CI->getContext()),
+ ConstantInt::get(DL.getIntPtrType(CI->getContext()),
std::min(Len1, Len2)),
B, DL, TLI);
}
if (Length == 0) // strncmp(x,y,0) -> 0
return ConstantInt::get(CI->getType(), 0);
- if (DL && Length == 1) // strncmp(x,y,1) -> memcmp(x,y,1)
+ if (Length == 1) // strncmp(x,y,1) -> memcmp(x,y,1)
return EmitMemCmp(Str1P, Str2P, CI->getArgOperand(2), B, DL, TLI);
StringRef Str1, Str2;
Value *LibCallSimplifier::optimizeStrCpy(CallInst *CI, IRBuilder<> &B) {
Function *Callee = CI->getCalledFunction();
- if (!checkStringCopyLibFuncSignature(Callee, LibFunc::strcpy, DL))
+ if (!checkStringCopyLibFuncSignature(Callee, LibFunc::strcpy))
return nullptr;
Value *Dst = CI->getArgOperand(0), *Src = CI->getArgOperand(1);
if (Dst == Src) // strcpy(x,x) -> x
return Src;
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
-
// See if we can get the length of the input string.
uint64_t Len = GetStringLength(Src);
if (Len == 0)
// We have enough information to now generate the memcpy call to do the
// copy for us. Make a memcpy to copy the nul byte with align = 1.
B.CreateMemCpy(Dst, Src,
- ConstantInt::get(DL->getIntPtrType(CI->getContext()), Len), 1);
+ ConstantInt::get(DL.getIntPtrType(CI->getContext()), Len), 1);
return Dst;
}
// Verify the "stpcpy" function prototype.
FunctionType *FT = Callee->getFunctionType();
- if (!checkStringCopyLibFuncSignature(Callee, LibFunc::stpcpy, DL))
- return nullptr;
-
- // These optimizations require DataLayout.
- if (!DL)
+ if (!checkStringCopyLibFuncSignature(Callee, LibFunc::stpcpy))
return nullptr;
Value *Dst = CI->getArgOperand(0), *Src = CI->getArgOperand(1);
return nullptr;
Type *PT = FT->getParamType(0);
- Value *LenV = ConstantInt::get(DL->getIntPtrType(PT), Len);
+ Value *LenV = ConstantInt::get(DL.getIntPtrType(PT), Len);
Value *DstEnd =
- B.CreateGEP(Dst, ConstantInt::get(DL->getIntPtrType(PT), Len - 1));
+ B.CreateGEP(Dst, ConstantInt::get(DL.getIntPtrType(PT), Len - 1));
// We have enough information to now generate the memcpy call to do the
// copy for us. Make a memcpy to copy the nul byte with align = 1.
Function *Callee = CI->getCalledFunction();
FunctionType *FT = Callee->getFunctionType();
- if (!checkStringCopyLibFuncSignature(Callee, LibFunc::strncpy, DL))
+ if (!checkStringCopyLibFuncSignature(Callee, LibFunc::strncpy))
return nullptr;
Value *Dst = CI->getArgOperand(0);
if (Len == 0)
return Dst; // strncpy(x, y, 0) -> x
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
-
// Let strncpy handle the zero padding
if (Len > SrcLen + 1)
return nullptr;
Type *PT = FT->getParamType(0);
// strncpy(x, s, c) -> memcpy(x, s, c, 1) [s and c are constant]
- B.CreateMemCpy(Dst, Src, ConstantInt::get(DL->getIntPtrType(PT), Len), 1);
+ B.CreateMemCpy(Dst, Src, ConstantInt::get(DL.getIntPtrType(PT), Len), 1);
return Dst;
}
}
// strpbrk(s, "a") -> strchr(s, 'a')
- if (DL && HasS2 && S2.size() == 1)
- return EmitStrChr(CI->getArgOperand(0), S2[0], B, DL, TLI);
+ if (HasS2 && S2.size() == 1)
+ return EmitStrChr(CI->getArgOperand(0), S2[0], B, TLI);
return nullptr;
}
}
// strcspn(s, "") -> strlen(s)
- if (DL && HasS2 && S2.empty())
+ if (HasS2 && S2.empty())
return EmitStrLen(CI->getArgOperand(0), B, DL, TLI);
return nullptr;
return B.CreateBitCast(CI->getArgOperand(0), CI->getType());
// fold strstr(a, b) == a -> strncmp(a, b, strlen(b)) == 0
- if (DL && isOnlyUsedInEqualityComparison(CI, CI->getArgOperand(0))) {
+ if (isOnlyUsedInEqualityComparison(CI, CI->getArgOperand(0))) {
Value *StrLen = EmitStrLen(CI->getArgOperand(1), B, DL, TLI);
if (!StrLen)
return nullptr;
// fold strstr(x, "y") -> strchr(x, 'y').
if (HasStr2 && ToFindStr.size() == 1) {
- Value *StrChr = EmitStrChr(CI->getArgOperand(0), ToFindStr[0], B, DL, TLI);
+ Value *StrChr = EmitStrChr(CI->getArgOperand(0), ToFindStr[0], B, TLI);
return StrChr ? B.CreateBitCast(StrChr, CI->getType()) : nullptr;
}
return nullptr;
Value *LibCallSimplifier::optimizeMemCpy(CallInst *CI, IRBuilder<> &B) {
Function *Callee = CI->getCalledFunction();
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
- if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memcpy, DL))
+ if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memcpy))
return nullptr;
// memcpy(x, y, n) -> llvm.memcpy(x, y, n, 1)
Value *LibCallSimplifier::optimizeMemMove(CallInst *CI, IRBuilder<> &B) {
Function *Callee = CI->getCalledFunction();
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
- if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memmove, DL))
+ if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memmove))
return nullptr;
// memmove(x, y, n) -> llvm.memmove(x, y, n, 1)
Value *LibCallSimplifier::optimizeMemSet(CallInst *CI, IRBuilder<> &B) {
Function *Callee = CI->getCalledFunction();
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
- if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memset, DL))
+ if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memset))
return nullptr;
// memset(p, v, n) -> llvm.memset(p, v, n, 1)
// printf("x") -> putchar('x'), even for '%'.
if (FormatStr.size() == 1) {
- Value *Res = EmitPutChar(B.getInt32(FormatStr[0]), B, DL, TLI);
+ Value *Res = EmitPutChar(B.getInt32(FormatStr[0]), B, TLI);
if (CI->use_empty() || !Res)
return Res;
return B.CreateIntCast(Res, CI->getType(), true);
// pass to be run after this pass, to merge duplicate strings.
FormatStr = FormatStr.drop_back();
Value *GV = B.CreateGlobalString(FormatStr, "str");
- Value *NewCI = EmitPutS(GV, B, DL, TLI);
+ Value *NewCI = EmitPutS(GV, B, TLI);
return (CI->use_empty() || !NewCI)
? NewCI
: ConstantInt::get(CI->getType(), FormatStr.size() + 1);
// printf("%c", chr) --> putchar(chr)
if (FormatStr == "%c" && CI->getNumArgOperands() > 1 &&
CI->getArgOperand(1)->getType()->isIntegerTy()) {
- Value *Res = EmitPutChar(CI->getArgOperand(1), B, DL, TLI);
+ Value *Res = EmitPutChar(CI->getArgOperand(1), B, TLI);
if (CI->use_empty() || !Res)
return Res;
// printf("%s\n", str) --> puts(str)
if (FormatStr == "%s\n" && CI->getNumArgOperands() > 1 &&
CI->getArgOperand(1)->getType()->isPointerTy()) {
- return EmitPutS(CI->getArgOperand(1), B, DL, TLI);
+ return EmitPutS(CI->getArgOperand(1), B, TLI);
}
return nullptr;
}
if (FormatStr[i] == '%')
return nullptr; // we found a format specifier, bail out.
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
-
// sprintf(str, fmt) -> llvm.memcpy(str, fmt, strlen(fmt)+1, 1)
- B.CreateMemCpy(
- CI->getArgOperand(0), CI->getArgOperand(1),
- ConstantInt::get(DL->getIntPtrType(CI->getContext()),
- FormatStr.size() + 1),
- 1); // Copy the null byte.
+ B.CreateMemCpy(CI->getArgOperand(0), CI->getArgOperand(1),
+ ConstantInt::get(DL.getIntPtrType(CI->getContext()),
+ FormatStr.size() + 1),
+ 1); // Copy the null byte.
return ConstantInt::get(CI->getType(), FormatStr.size());
}
}
if (FormatStr[1] == 's') {
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
-
// sprintf(dest, "%s", str) -> llvm.memcpy(dest, str, strlen(str)+1, 1)
if (!CI->getArgOperand(2)->getType()->isPointerTy())
return nullptr;
if (FormatStr[i] == '%') // Could handle %% -> % if we cared.
return nullptr; // We found a format specifier.
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
-
return EmitFWrite(
CI->getArgOperand(1),
- ConstantInt::get(DL->getIntPtrType(CI->getContext()), FormatStr.size()),
+ ConstantInt::get(DL.getIntPtrType(CI->getContext()), FormatStr.size()),
CI->getArgOperand(0), B, DL, TLI);
}
// fprintf(F, "%c", chr) --> fputc(chr, F)
if (!CI->getArgOperand(2)->getType()->isIntegerTy())
return nullptr;
- return EmitFPutC(CI->getArgOperand(2), CI->getArgOperand(0), B, DL, TLI);
+ return EmitFPutC(CI->getArgOperand(2), CI->getArgOperand(0), B, TLI);
}
if (FormatStr[1] == 's') {
// fprintf(F, "%s", str) --> fputs(str, F)
if (!CI->getArgOperand(2)->getType()->isPointerTy())
return nullptr;
- return EmitFPutS(CI->getArgOperand(2), CI->getArgOperand(0), B, DL, TLI);
+ return EmitFPutS(CI->getArgOperand(2), CI->getArgOperand(0), B, TLI);
}
return nullptr;
}
// This optimisation is only valid, if the return value is unused.
if (Bytes == 1 && CI->use_empty()) { // fwrite(S,1,1,F) -> fputc(S[0],F)
Value *Char = B.CreateLoad(CastToCStr(CI->getArgOperand(0), B), "char");
- Value *NewCI = EmitFPutC(Char, CI->getArgOperand(3), B, DL, TLI);
+ Value *NewCI = EmitFPutC(Char, CI->getArgOperand(3), B, TLI);
return NewCI ? ConstantInt::get(CI->getType(), 1) : nullptr;
}
Function *Callee = CI->getCalledFunction();
- // These optimizations require DataLayout.
- if (!DL)
- return nullptr;
-
// Require two pointers. Also, we can't optimize if return value is used.
FunctionType *FT = Callee->getFunctionType();
if (FT->getNumParams() != 2 || !FT->getParamType(0)->isPointerTy() ||
// Known to have no uses (see above).
return EmitFWrite(
CI->getArgOperand(0),
- ConstantInt::get(DL->getIntPtrType(CI->getContext()), Len - 1),
+ ConstantInt::get(DL.getIntPtrType(CI->getContext()), Len - 1),
CI->getArgOperand(1), B, DL, TLI);
}
if (Str.empty() && CI->use_empty()) {
// puts("") -> putchar('\n')
- Value *Res = EmitPutChar(B.getInt32('\n'), B, DL, TLI);
+ Value *Res = EmitPutChar(B.getInt32('\n'), B, TLI);
if (CI->use_empty() || !Res)
return Res;
return B.CreateIntCast(Res, CI->getType(), true);
}
LibCallSimplifier::LibCallSimplifier(
- const DataLayout *DL, const TargetLibraryInfo *TLI,
+ const DataLayout &DL, const TargetLibraryInfo *TLI,
function_ref<void(Instruction *, Value *)> Replacer)
- : FortifiedSimplifier(DL, TLI), DL(DL), TLI(TLI), UnsafeFPShrink(false),
+ : FortifiedSimplifier(TLI), DL(DL), TLI(TLI), UnsafeFPShrink(false),
Replacer(Replacer) {}
void LibCallSimplifier::replaceAllUsesWith(Instruction *I, Value *With) {
Value *FortifiedLibCallSimplifier::optimizeMemCpyChk(CallInst *CI, IRBuilder<> &B) {
Function *Callee = CI->getCalledFunction();
- if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memcpy_chk, DL))
+ if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memcpy_chk))
return nullptr;
if (isFortifiedCallFoldable(CI, 3, 2, false)) {
Value *FortifiedLibCallSimplifier::optimizeMemMoveChk(CallInst *CI, IRBuilder<> &B) {
Function *Callee = CI->getCalledFunction();
- if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memmove_chk, DL))
+ if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memmove_chk))
return nullptr;
if (isFortifiedCallFoldable(CI, 3, 2, false)) {
Value *FortifiedLibCallSimplifier::optimizeMemSetChk(CallInst *CI, IRBuilder<> &B) {
Function *Callee = CI->getCalledFunction();
- if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memset_chk, DL))
+ if (!checkStringCopyLibFuncSignature(Callee, LibFunc::memset_chk))
return nullptr;
if (isFortifiedCallFoldable(CI, 3, 2, false)) {
LibFunc::Func Func) {
Function *Callee = CI->getCalledFunction();
StringRef Name = Callee->getName();
+ const DataLayout &DL = CI->getModule()->getDataLayout();
- if (!checkStringCopyLibFuncSignature(Callee, Func, DL))
+ if (!checkStringCopyLibFuncSignature(Callee, Func))
return nullptr;
Value *Dst = CI->getArgOperand(0), *Src = CI->getArgOperand(1),
// TODO: It might be nice to get a maximum length out of the possible
// string lengths for varying.
if (isFortifiedCallFoldable(CI, 2, 1, true)) {
- Value *Ret = EmitStrCpy(Dst, Src, B, DL, TLI, Name.substr(2, 6));
+ Value *Ret = EmitStrCpy(Dst, Src, B, TLI, Name.substr(2, 6));
return Ret;
} else if (!OnlyLowerUnknownSize) {
// Maybe we can stil fold __st[rp]cpy_chk to __memcpy_chk.
if (Len == 0)
return nullptr;
- // This optimization requires DataLayout.
- if (!DL)
- return nullptr;
-
- Type *SizeTTy = DL->getIntPtrType(CI->getContext());
+ Type *SizeTTy = DL.getIntPtrType(CI->getContext());
Value *LenV = ConstantInt::get(SizeTTy, Len);
Value *Ret = EmitMemCpyChk(Dst, Src, LenV, ObjSize, B, DL, TLI);
// If the function was an __stpcpy_chk, and we were able to fold it into
Function *Callee = CI->getCalledFunction();
StringRef Name = Callee->getName();
- if (!checkStringCopyLibFuncSignature(Callee, Func, DL))
+ if (!checkStringCopyLibFuncSignature(Callee, Func))
return nullptr;
if (isFortifiedCallFoldable(CI, 3, 2, false)) {
- Value *Ret =
- EmitStrNCpy(CI->getArgOperand(0), CI->getArgOperand(1),
- CI->getArgOperand(2), B, DL, TLI, Name.substr(2, 7));
+ Value *Ret = EmitStrNCpy(CI->getArgOperand(0), CI->getArgOperand(1),
+ CI->getArgOperand(2), B, TLI, Name.substr(2, 7));
return Ret;
}
return nullptr;
return nullptr;
}
-FortifiedLibCallSimplifier::
-FortifiedLibCallSimplifier(const DataLayout *DL, const TargetLibraryInfo *TLI,
- bool OnlyLowerUnknownSize)
- : DL(DL), TLI(TLI), OnlyLowerUnknownSize(OnlyLowerUnknownSize) {
-}
+FortifiedLibCallSimplifier::FortifiedLibCallSimplifier(
+ const TargetLibraryInfo *TLI, bool OnlyLowerUnknownSize)
+ : TLI(TLI), OnlyLowerUnknownSize(OnlyLowerUnknownSize) {}
AA = &P->getAnalysis<AliasAnalysis>();
DT = &P->getAnalysis<DominatorTreeWrapperPass>().getDomTree();
SE = &P->getAnalysis<ScalarEvolution>();
- DL = &F.getParent()->getDataLayout();
TTI = IgnoreTargetInfo
? nullptr
: &P->getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
AliasAnalysis *AA;
DominatorTree *DT;
ScalarEvolution *SE;
- const DataLayout *DL;
const TargetTransformInfo *TTI;
// FIXME: const correct?
AA = &getAnalysis<AliasAnalysis>();
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
SE = &getAnalysis<ScalarEvolution>();
- DL = &BB.getModule()->getDataLayout();
TTI = IgnoreTargetInfo
? nullptr
: &getAnalysis<TargetTransformInfoWrapperPass>().getTTI(
dyn_cast<SCEVConstant>(OffsetSCEV)) {
ConstantInt *IntOff = ConstOffSCEV->getValue();
int64_t Offset = IntOff->getSExtValue();
-
+ const DataLayout &DL = I->getModule()->getDataLayout();
Type *VTy = IPtr->getType()->getPointerElementType();
- int64_t VTyTSS = (int64_t) DL->getTypeStoreSize(VTy);
+ int64_t VTyTSS = (int64_t)DL.getTypeStoreSize(VTy);
Type *VTy2 = JPtr->getType()->getPointerElementType();
if (VTy != VTy2 && Offset < 0) {
- int64_t VTy2TSS = (int64_t) DL->getTypeStoreSize(VTy2);
+ int64_t VTy2TSS = (int64_t)DL.getTypeStoreSize(VTy2);
OffsetInElmts = Offset/VTy2TSS;
return (std::abs(Offset) % VTy2TSS) == 0;
}
// It is important to cleanup here so that future iterations of this
// function have less work to do.
- (void) SimplifyInstructionsInBlock(&BB, DL, AA->getTargetLibraryInfo());
+ (void)SimplifyInstructionsInBlock(&BB, AA->getTargetLibraryInfo());
return true;
}
return false;
}
- // We can't vectorize memory operations without target data
- if (!DL && IsSimpleLoadStore)
- return false;
-
Type *T1, *T2;
getInstructionTypes(I, T1, T2);
if (T2->isX86_FP80Ty() || T2->isPPC_FP128Ty() || T2->isX86_MMXTy())
return false;
- if ((!Config.VectorizePointers || !DL) &&
- (T1->getScalarType()->isPointerTy() ||
- T2->getScalarType()->isPointerTy()))
+ if (!Config.VectorizePointers && (T1->getScalarType()->isPointerTy() ||
+ T2->getScalarType()->isPointerTy()))
return false;
if (!TTI && (T1->getPrimitiveSizeInBits() >= Config.VectorBits ||
// An aligned load or store is possible only if the instruction
// with the lower offset has an alignment suitable for the
// vector type.
-
- unsigned VecAlignment = DL->getPrefTypeAlignment(VType);
+ const DataLayout &DL = I->getModule()->getDataLayout();
+ unsigned VecAlignment = DL.getPrefTypeAlignment(VType);
if (BottomAlignment < VecAlignment)
return false;
}
class InnerLoopVectorizer {
public:
InnerLoopVectorizer(Loop *OrigLoop, ScalarEvolution *SE, LoopInfo *LI,
- DominatorTree *DT, const DataLayout *DL,
- const TargetLibraryInfo *TLI, unsigned VecWidth,
- unsigned UnrollFactor)
- : OrigLoop(OrigLoop), SE(SE), LI(LI), DT(DT), DL(DL), TLI(TLI),
- VF(VecWidth), UF(UnrollFactor), Builder(SE->getContext()),
- Induction(nullptr), OldInduction(nullptr), WidenMap(UnrollFactor),
- Legal(nullptr), AddedSafetyChecks(false) {}
+ DominatorTree *DT, const TargetLibraryInfo *TLI,
+ unsigned VecWidth, unsigned UnrollFactor)
+ : OrigLoop(OrigLoop), SE(SE), LI(LI), DT(DT), TLI(TLI), VF(VecWidth),
+ UF(UnrollFactor), Builder(SE->getContext()), Induction(nullptr),
+ OldInduction(nullptr), WidenMap(UnrollFactor), Legal(nullptr),
+ AddedSafetyChecks(false) {}
// Perform the actual loop widening (vectorization).
void vectorize(LoopVectorizationLegality *L) {
DominatorTree *DT;
/// Alias Analysis.
AliasAnalysis *AA;
- /// Data Layout.
- const DataLayout *DL;
/// Target Library Info.
const TargetLibraryInfo *TLI;
class InnerLoopUnroller : public InnerLoopVectorizer {
public:
InnerLoopUnroller(Loop *OrigLoop, ScalarEvolution *SE, LoopInfo *LI,
- DominatorTree *DT, const DataLayout *DL,
- const TargetLibraryInfo *TLI, unsigned UnrollFactor) :
- InnerLoopVectorizer(OrigLoop, SE, LI, DT, DL, TLI, 1, UnrollFactor) { }
+ DominatorTree *DT, const TargetLibraryInfo *TLI,
+ unsigned UnrollFactor)
+ : InnerLoopVectorizer(OrigLoop, SE, LI, DT, TLI, 1, UnrollFactor) {}
private:
void scalarizeInstruction(Instruction *Instr,
/// induction variable and the different reduction variables.
class LoopVectorizationLegality {
public:
- LoopVectorizationLegality(Loop *L, ScalarEvolution *SE, const DataLayout *DL,
- DominatorTree *DT, TargetLibraryInfo *TLI,
- AliasAnalysis *AA, Function *F,
- const TargetTransformInfo *TTI,
+ LoopVectorizationLegality(Loop *L, ScalarEvolution *SE, DominatorTree *DT,
+ TargetLibraryInfo *TLI, AliasAnalysis *AA,
+ Function *F, const TargetTransformInfo *TTI,
LoopAccessAnalysis *LAA)
- : NumPredStores(0), TheLoop(L), SE(SE), DL(DL),
- TLI(TLI), TheFunction(F), TTI(TTI), DT(DT), LAA(LAA), LAI(nullptr),
- Induction(nullptr), WidestIndTy(nullptr), HasFunNoNaNAttr(false) {}
+ : NumPredStores(0), TheLoop(L), SE(SE), TLI(TLI), TheFunction(F),
+ TTI(TTI), DT(DT), LAA(LAA), LAI(nullptr), Induction(nullptr),
+ WidestIndTy(nullptr), HasFunNoNaNAttr(false) {}
/// This enum represents the kinds of reductions that we support.
enum ReductionKind {
Loop *TheLoop;
/// Scev analysis.
ScalarEvolution *SE;
- /// DataLayout analysis.
- const DataLayout *DL;
/// Target Library Info.
TargetLibraryInfo *TLI;
/// Parent function
LoopVectorizationCostModel(Loop *L, ScalarEvolution *SE, LoopInfo *LI,
LoopVectorizationLegality *Legal,
const TargetTransformInfo &TTI,
- const DataLayout *DL, const TargetLibraryInfo *TLI,
- AssumptionCache *AC, const Function *F,
- const LoopVectorizeHints *Hints)
- : TheLoop(L), SE(SE), LI(LI), Legal(Legal), TTI(TTI), DL(DL), TLI(TLI),
+ const TargetLibraryInfo *TLI, AssumptionCache *AC,
+ const Function *F, const LoopVectorizeHints *Hints)
+ : TheLoop(L), SE(SE), LI(LI), Legal(Legal), TTI(TTI), TLI(TLI),
TheFunction(F), Hints(Hints) {
CodeMetrics::collectEphemeralValues(L, AC, EphValues);
}
LoopVectorizationLegality *Legal;
/// Vector target information.
const TargetTransformInfo &TTI;
- /// Target data layout information.
- const DataLayout *DL;
/// Target Library Info.
const TargetLibraryInfo *TLI;
const Function *TheFunction;
}
ScalarEvolution *SE;
- const DataLayout *DL;
LoopInfo *LI;
TargetTransformInfo *TTI;
DominatorTree *DT;
bool runOnFunction(Function &F) override {
SE = &getAnalysis<ScalarEvolution>();
- DL = &F.getParent()->getDataLayout();
LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
TTI = &getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
if (!TTI->getNumberOfRegisters(true))
return false;
- if (!DL) {
- DEBUG(dbgs() << "\nLV: Not vectorizing " << F.getName()
- << ": Missing data layout\n");
- return false;
- }
-
// Build up a worklist of inner-loops to vectorize. This is necessary as
// the act of vectorizing or partially unrolling a loop creates new loops
// and can invalidate iterators across the loops.
}
// Check if it is legal to vectorize the loop.
- LoopVectorizationLegality LVL(L, SE, DL, DT, TLI, AA, F, TTI, LAA);
+ LoopVectorizationLegality LVL(L, SE, DT, TLI, AA, F, TTI, LAA);
if (!LVL.canVectorize()) {
DEBUG(dbgs() << "LV: Not vectorizing: Cannot prove legality.\n");
emitMissedWarning(F, L, Hints);
}
// Use the cost model.
- LoopVectorizationCostModel CM(L, SE, LI, &LVL, *TTI, DL, TLI, AC, F,
- &Hints);
+ LoopVectorizationCostModel CM(L, SE, LI, &LVL, *TTI, TLI, AC, F, &Hints);
// Check the function attributes to find out if this function should be
// optimized for size.
// We decided not to vectorize, but we may want to unroll.
- InnerLoopUnroller Unroller(L, SE, LI, DT, DL, TLI, UF);
+ InnerLoopUnroller Unroller(L, SE, LI, DT, TLI, UF);
Unroller.vectorize(&LVL);
} else {
// If we decided that it is *legal* to vectorize the loop then do it.
- InnerLoopVectorizer LB(L, SE, LI, DT, DL, TLI, VF.Width, UF);
+ InnerLoopVectorizer LB(L, SE, LI, DT, TLI, VF.Width, UF);
LB.vectorize(&LVL);
++LoopsVectorized;
/// \brief Find the operand of the GEP that should be checked for consecutive
/// stores. This ignores trailing indices that have no effect on the final
/// pointer.
-static unsigned getGEPInductionOperand(const DataLayout *DL,
- const GetElementPtrInst *Gep) {
+static unsigned getGEPInductionOperand(const GetElementPtrInst *Gep) {
+ const DataLayout &DL = Gep->getModule()->getDataLayout();
unsigned LastOperand = Gep->getNumOperands() - 1;
- unsigned GEPAllocSize = DL->getTypeAllocSize(
+ unsigned GEPAllocSize = DL.getTypeAllocSize(
cast<PointerType>(Gep->getType()->getScalarType())->getElementType());
// Walk backwards and try to peel off zeros.
// If it's a type with the same allocation size as the result of the GEP we
// can peel off the zero index.
- if (DL->getTypeAllocSize(*GEPTI) != GEPAllocSize)
+ if (DL.getTypeAllocSize(*GEPTI) != GEPAllocSize)
break;
--LastOperand;
}
return II.getConsecutiveDirection();
}
- unsigned InductionOperand = getGEPInductionOperand(DL, Gep);
+ unsigned InductionOperand = getGEPInductionOperand(Gep);
// Check that all of the gep indices are uniform except for our induction
// operand.
unsigned Alignment = LI ? LI->getAlignment() : SI->getAlignment();
// An alignment of 0 means target abi alignment. We need to use the scalar's
// target abi alignment in such a case.
+ const DataLayout &DL = Instr->getModule()->getDataLayout();
if (!Alignment)
- Alignment = DL->getABITypeAlignment(ScalarDataTy);
+ Alignment = DL.getABITypeAlignment(ScalarDataTy);
unsigned AddressSpace = Ptr->getType()->getPointerAddressSpace();
- unsigned ScalarAllocatedSize = DL->getTypeAllocSize(ScalarDataTy);
- unsigned VectorElementSize = DL->getTypeStoreSize(DataTy)/VF;
+ unsigned ScalarAllocatedSize = DL.getTypeAllocSize(ScalarDataTy);
+ unsigned VectorElementSize = DL.getTypeStoreSize(DataTy) / VF;
if (SI && Legal->blockNeedsPredication(SI->getParent()) &&
!Legal->isMaskRequired(SI))
// The last index does not have to be the induction. It can be
// consecutive and be a function of the index. For example A[I+1];
unsigned NumOperands = Gep->getNumOperands();
- unsigned InductionOperand = getGEPInductionOperand(DL, Gep);
+ unsigned InductionOperand = getGEPInductionOperand(Gep);
// Create the new GEP with the new induction variable.
GetElementPtrInst *Gep2 = cast<GetElementPtrInst>(Gep->clone());
ExitCount = SE->getAddExpr(BackedgeTakeCount,
SE->getConstant(BackedgeTakeCount->getType(), 1));
+ const DataLayout &DL = OldBasicBlock->getModule()->getDataLayout();
+
// Expand the trip count and place the new instructions in the preheader.
// Notice that the pre-header does not change, only the loop body.
- SCEVExpander Exp(*SE, "induction");
+ SCEVExpander Exp(*SE, DL, "induction");
// We need to test whether the backedge-taken count is uint##_max. Adding one
// to it will cause overflow and an incorrect loop trip count in the vector
// Look for the attribute signaling the absence of NaNs.
Function &F = *Header->getParent();
+ const DataLayout &DL = F.getParent()->getDataLayout();
if (F.hasFnAttribute("no-nans-fp-math"))
HasFunNoNaNAttr =
F.getFnAttribute("no-nans-fp-math").getValueAsString() == "true";
if (IK_NoInduction != IK) {
// Get the widest type.
if (!WidestIndTy)
- WidestIndTy = convertPointerToIntegerType(*DL, PhiTy);
+ WidestIndTy = convertPointerToIntegerType(DL, PhiTy);
else
- WidestIndTy = getWiderType(*DL, PhiTy, WidestIndTy);
+ WidestIndTy = getWiderType(DL, PhiTy, WidestIndTy);
// Int inductions are special because we only allow one IV.
if (IK == IK_IntInduction && StepValue->isOne()) {
///\brief Remove GEPs whose indices but the last one are loop invariant and
/// return the induction operand of the gep pointer.
-static Value *stripGetElementPtr(Value *Ptr, ScalarEvolution *SE,
- const DataLayout *DL, Loop *Lp) {
+static Value *stripGetElementPtr(Value *Ptr, ScalarEvolution *SE, Loop *Lp) {
GetElementPtrInst *GEP = dyn_cast<GetElementPtrInst>(Ptr);
if (!GEP)
return Ptr;
- unsigned InductionOperand = getGEPInductionOperand(DL, GEP);
+ unsigned InductionOperand = getGEPInductionOperand(GEP);
// Check that all of the gep indices are uniform except for our induction
// operand.
///\brief Get the stride of a pointer access in a loop.
/// Looks for symbolic strides "a[i*stride]". Returns the symbolic stride as a
/// pointer to the Value, or null otherwise.
-static Value *getStrideFromPointer(Value *Ptr, ScalarEvolution *SE,
- const DataLayout *DL, Loop *Lp) {
+static Value *getStrideFromPointer(Value *Ptr, ScalarEvolution *SE, Loop *Lp) {
const PointerType *PtrTy = dyn_cast<PointerType>(Ptr->getType());
if (!PtrTy || PtrTy->isAggregateType())
return nullptr;
// The size of the pointer access.
int64_t PtrAccessSize = 1;
- Ptr = stripGetElementPtr(Ptr, SE, DL, Lp);
+ Ptr = stripGetElementPtr(Ptr, SE, Lp);
const SCEV *V = SE->getSCEV(Ptr);
if (Ptr != OrigPtr)
// Strip off the size of access multiplication if we are still analyzing the
// pointer.
if (OrigPtr == Ptr) {
- DL->getTypeAllocSize(PtrTy->getElementType());
+ const DataLayout &DL = Lp->getHeader()->getModule()->getDataLayout();
+ DL.getTypeAllocSize(PtrTy->getElementType());
if (const SCEVMulExpr *M = dyn_cast<SCEVMulExpr>(V)) {
if (M->getOperand(0)->getSCEVType() != scConstant)
return nullptr;
else
return;
- Value *Stride = getStrideFromPointer(Ptr, SE, DL, TheLoop);
+ Value *Stride = getStrideFromPointer(Ptr, SE, TheLoop);
if (!Stride)
return;
if (!PointerElementType->isSized())
return IK_NoInduction;
- int64_t Size = static_cast<int64_t>(DL->getTypeAllocSize(PointerElementType));
+ const DataLayout &DL = Phi->getModule()->getDataLayout();
+ int64_t Size = static_cast<int64_t>(DL.getTypeAllocSize(PointerElementType));
int64_t CVSize = CV->getSExtValue();
if (CVSize % Size)
return IK_NoInduction;
unsigned LoopVectorizationCostModel::getWidestType() {
unsigned MaxWidth = 8;
+ const DataLayout &DL = TheFunction->getParent()->getDataLayout();
// For each block.
for (Loop::block_iterator bb = TheLoop->block_begin(),
continue;
MaxWidth = std::max(MaxWidth,
- (unsigned)DL->getTypeSizeInBits(T->getScalarType()));
+ (unsigned)DL.getTypeSizeInBits(T->getScalarType()));
}
}
// Scalarized loads/stores.
int ConsecutiveStride = Legal->isConsecutivePtr(Ptr);
bool Reverse = ConsecutiveStride < 0;
- unsigned ScalarAllocatedSize = DL->getTypeAllocSize(ValTy);
- unsigned VectorElementSize = DL->getTypeStoreSize(VectorTy)/VF;
+ const DataLayout &DL = I->getModule()->getDataLayout();
+ unsigned ScalarAllocatedSize = DL.getTypeAllocSize(ValTy);
+ unsigned VectorElementSize = DL.getTypeStoreSize(VectorTy) / VF;
if (!ConsecutiveStride || ScalarAllocatedSize != VectorElementSize) {
bool IsComplexComputation =
isLikelyComplexAddressComputation(Ptr, Legal, SE, TheLoop);
typedef SmallPtrSet<Value *, 16> ValueSet;
typedef SmallVector<StoreInst *, 8> StoreList;
- BoUpSLP(Function *Func, ScalarEvolution *Se, const DataLayout *Dl,
- TargetTransformInfo *Tti, TargetLibraryInfo *TLi, AliasAnalysis *Aa,
- LoopInfo *Li, DominatorTree *Dt, AssumptionCache *AC)
+ BoUpSLP(Function *Func, ScalarEvolution *Se, TargetTransformInfo *Tti,
+ TargetLibraryInfo *TLi, AliasAnalysis *Aa, LoopInfo *Li,
+ DominatorTree *Dt, AssumptionCache *AC)
: NumLoadsWantToKeepOrder(0), NumLoadsWantToChangeOrder(0), F(Func),
- SE(Se), DL(Dl), TTI(Tti), TLI(TLi), AA(Aa), LI(Li), DT(Dt),
+ SE(Se), TTI(Tti), TLI(TLi), AA(Aa), LI(Li), DT(Dt),
Builder(Se->getContext()) {
CodeMetrics::collectEphemeralValues(F, AC, EphValues);
}
}
/// \returns true if the memory operations A and B are consecutive.
- bool isConsecutiveAccess(Value *A, Value *B);
+ bool isConsecutiveAccess(Value *A, Value *B, const DataLayout &DL);
/// \brief Perform LICM and CSE on the newly generated gather sequences.
void optimizeGatherSequence();
// Analysis and block reference.
Function *F;
ScalarEvolution *SE;
- const DataLayout *DL;
TargetTransformInfo *TTI;
TargetLibraryInfo *TLI;
AliasAnalysis *AA;
DEBUG(dbgs() << "SLP: Gathering non-simple loads.\n");
return;
}
- if (!isConsecutiveAccess(VL[i], VL[i + 1])) {
- if (VL.size() == 2 && isConsecutiveAccess(VL[1], VL[0])) {
+ const DataLayout &DL = F->getParent()->getDataLayout();
+ if (!isConsecutiveAccess(VL[i], VL[i + 1], DL)) {
+ if (VL.size() == 2 && isConsecutiveAccess(VL[1], VL[0], DL)) {
++NumLoadsWantToChangeOrder;
}
BS.cancelScheduling(VL);
return;
}
case Instruction::Store: {
+ const DataLayout &DL = F->getParent()->getDataLayout();
// Check if the stores are consecutive or of we need to swizzle them.
for (unsigned i = 0, e = VL.size() - 1; i < e; ++i)
- if (!isConsecutiveAccess(VL[i], VL[i + 1])) {
+ if (!isConsecutiveAccess(VL[i], VL[i + 1], DL)) {
BS.cancelScheduling(VL);
newTreeEntry(VL, false);
DEBUG(dbgs() << "SLP: Non-consecutive store.\n");
return -1;
}
-bool BoUpSLP::isConsecutiveAccess(Value *A, Value *B) {
+bool BoUpSLP::isConsecutiveAccess(Value *A, Value *B, const DataLayout &DL) {
Value *PtrA = getPointerOperand(A);
Value *PtrB = getPointerOperand(B);
unsigned ASA = getAddressSpaceOperand(A);
if (PtrA == PtrB || PtrA->getType() != PtrB->getType())
return false;
- unsigned PtrBitWidth = DL->getPointerSizeInBits(ASA);
+ unsigned PtrBitWidth = DL.getPointerSizeInBits(ASA);
Type *Ty = cast<PointerType>(PtrA->getType())->getElementType();
- APInt Size(PtrBitWidth, DL->getTypeStoreSize(Ty));
+ APInt Size(PtrBitWidth, DL.getTypeStoreSize(Ty));
APInt OffsetA(PtrBitWidth, 0), OffsetB(PtrBitWidth, 0);
- PtrA = PtrA->stripAndAccumulateInBoundsConstantOffsets(*DL, OffsetA);
- PtrB = PtrB->stripAndAccumulateInBoundsConstantOffsets(*DL, OffsetB);
+ PtrA = PtrA->stripAndAccumulateInBoundsConstantOffsets(DL, OffsetA);
+ PtrB = PtrB->stripAndAccumulateInBoundsConstantOffsets(DL, OffsetB);
APInt OffsetDelta = OffsetB - OffsetA;
void BoUpSLP::reorderAltShuffleOperands(ArrayRef<Value *> VL,
SmallVectorImpl<Value *> &Left,
SmallVectorImpl<Value *> &Right) {
+ const DataLayout &DL = F->getParent()->getDataLayout();
// Push left and right operands of binary operation into Left and Right
for (unsigned i = 0, e = VL.size(); i < e; ++i) {
if (LoadInst *L1 = dyn_cast<LoadInst>(Right[j + 1])) {
Instruction *VL1 = cast<Instruction>(VL[j]);
Instruction *VL2 = cast<Instruction>(VL[j + 1]);
- if (isConsecutiveAccess(L, L1) && VL1->isCommutative()) {
+ if (isConsecutiveAccess(L, L1, DL) && VL1->isCommutative()) {
std::swap(Left[j], Right[j]);
continue;
- } else if (isConsecutiveAccess(L, L1) && VL2->isCommutative()) {
+ } else if (isConsecutiveAccess(L, L1, DL) && VL2->isCommutative()) {
std::swap(Left[j + 1], Right[j + 1]);
continue;
}
if (LoadInst *L1 = dyn_cast<LoadInst>(Left[j + 1])) {
Instruction *VL1 = cast<Instruction>(VL[j]);
Instruction *VL2 = cast<Instruction>(VL[j + 1]);
- if (isConsecutiveAccess(L, L1) && VL1->isCommutative()) {
+ if (isConsecutiveAccess(L, L1, DL) && VL1->isCommutative()) {
std::swap(Left[j], Right[j]);
continue;
- } else if (isConsecutiveAccess(L, L1) && VL2->isCommutative()) {
+ } else if (isConsecutiveAccess(L, L1, DL) && VL2->isCommutative()) {
std::swap(Left[j + 1], Right[j + 1]);
continue;
}
Right = OrigRight;
}
+ const DataLayout &DL = F->getParent()->getDataLayout();
+
// Finally check if we can get longer vectorizable chain by reordering
// without breaking the good operand order detected above.
// E.g. If we have something like-
for (unsigned j = 0; j < VL.size() - 1; ++j) {
if (LoadInst *L = dyn_cast<LoadInst>(Left[j])) {
if (LoadInst *L1 = dyn_cast<LoadInst>(Right[j + 1])) {
- if (isConsecutiveAccess(L, L1)) {
+ if (isConsecutiveAccess(L, L1, DL)) {
std::swap(Left[j + 1], Right[j + 1]);
continue;
}
}
if (LoadInst *L = dyn_cast<LoadInst>(Right[j])) {
if (LoadInst *L1 = dyn_cast<LoadInst>(Left[j + 1])) {
- if (isConsecutiveAccess(L, L1)) {
+ if (isConsecutiveAccess(L, L1, DL)) {
std::swap(Left[j + 1], Right[j + 1]);
continue;
}
return Gather(E->Scalars, VecTy);
}
+ const DataLayout &DL = F->getParent()->getDataLayout();
unsigned Opcode = getSameOpcode(E->Scalars);
switch (Opcode) {
unsigned Alignment = LI->getAlignment();
LI = Builder.CreateLoad(VecPtr);
- if (!Alignment)
- Alignment = DL->getABITypeAlignment(ScalarLoadTy);
+ if (!Alignment) {
+ Alignment = DL.getABITypeAlignment(ScalarLoadTy);
+ }
LI->setAlignment(Alignment);
E->VectorizedValue = LI;
++NumVectorInstructions;
ExternalUses.push_back(
ExternalUser(SI->getPointerOperand(), cast<User>(VecPtr), 0));
- if (!Alignment)
- Alignment = DL->getABITypeAlignment(SI->getValueOperand()->getType());
+ if (!Alignment) {
+ Alignment = DL.getABITypeAlignment(SI->getValueOperand()->getType());
+ }
S->setAlignment(Alignment);
E->VectorizedValue = S;
++NumVectorInstructions;
}
ScalarEvolution *SE;
- const DataLayout *DL;
TargetTransformInfo *TTI;
TargetLibraryInfo *TLI;
AliasAnalysis *AA;
return false;
SE = &getAnalysis<ScalarEvolution>();
- DL = &F.getParent()->getDataLayout();
TTI = &getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
auto *TLIP = getAnalysisIfAvailable<TargetLibraryInfoWrapperPass>();
TLI = TLIP ? &TLIP->getTLI() : nullptr;
if (!TTI->getNumberOfRegisters(true))
return false;
- // Must have DataLayout. We can't require it because some tests run w/o
- // triple.
- if (!DL)
- return false;
-
// Don't vectorize when the attribute NoImplicitFloat is used.
if (F.hasFnAttribute(Attribute::NoImplicitFloat))
return false;
// Use the bottom up slp vectorizer to construct chains that start with
// store instructions.
- BoUpSLP R(&F, SE, DL, TTI, TLI, AA, LI, DT, AC);
+ BoUpSLP R(&F, SE, TTI, TLI, AA, LI, DT, AC);
// A general note: the vectorizer must use BoUpSLP::eraseInstruction() to
// delete instructions.
DEBUG(dbgs() << "SLP: Analyzing a store chain of length " << ChainLen
<< "\n");
Type *StoreTy = cast<StoreInst>(Chain[0])->getValueOperand()->getType();
- unsigned Sz = DL->getTypeSizeInBits(StoreTy);
+ auto &DL = cast<StoreInst>(Chain[0])->getModule()->getDataLayout();
+ unsigned Sz = DL.getTypeSizeInBits(StoreTy);
unsigned VF = MinVecRegSize / Sz;
if (!isPowerOf2_32(Sz) || VF < 2)
bool SLPVectorizer::vectorizeStores(ArrayRef<StoreInst *> Stores,
int costThreshold, BoUpSLP &R) {
- SetVector<Value *> Heads, Tails;
- SmallDenseMap<Value *, Value *> ConsecutiveChain;
+ SetVector<StoreInst *> Heads, Tails;
+ SmallDenseMap<StoreInst *, StoreInst *> ConsecutiveChain;
// We may run into multiple chains that merge into a single chain. We mark the
// stores that we vectorized so that we don't visit the same store twice.
for (unsigned j = 0; j < e; ++j) {
if (i == j)
continue;
-
- if (R.isConsecutiveAccess(Stores[i], Stores[j])) {
+ const DataLayout &DL = Stores[i]->getModule()->getDataLayout();
+ if (R.isConsecutiveAccess(Stores[i], Stores[j], DL)) {
Tails.insert(Stores[j]);
Heads.insert(Stores[i]);
ConsecutiveChain[Stores[i]] = Stores[j];
}
// For stores that start but don't end a link in the chain:
- for (SetVector<Value *>::iterator it = Heads.begin(), e = Heads.end();
+ for (SetVector<StoreInst *>::iterator it = Heads.begin(), e = Heads.end();
it != e; ++it) {
if (Tails.count(*it))
continue;
// We found a store instr that starts a chain. Now follow the chain and try
// to vectorize it.
BoUpSLP::ValueList Operands;
- Value *I = *it;
+ StoreInst *I = *it;
// Collect the chain into a list.
while (Tails.count(I) || Heads.count(I)) {
if (VectorizedStores.count(I))
unsigned SLPVectorizer::collectStores(BasicBlock *BB, BoUpSLP &R) {
unsigned count = 0;
StoreRefs.clear();
+ const DataLayout &DL = BB->getModule()->getDataLayout();
for (BasicBlock::iterator it = BB->begin(), e = BB->end(); it != e; ++it) {
StoreInst *SI = dyn_cast<StoreInst>(it);
if (!SI)
return false;
unsigned Opcode0 = I0->getOpcode();
+ const DataLayout &DL = I0->getModule()->getDataLayout();
Type *Ty0 = I0->getType();
- unsigned Sz = DL->getTypeSizeInBits(Ty0);
+ unsigned Sz = DL.getTypeSizeInBits(Ty0);
unsigned VF = MinVecRegSize / Sz;
for (int i = 0, e = VL.size(); i < e; ++i) {
ReducedValueOpcode(0), ReduxWidth(0), IsPairwiseReduction(false) {}
/// \brief Try to find a reduction tree.
- bool matchAssociativeReduction(PHINode *Phi, BinaryOperator *B,
- const DataLayout *DL) {
+ bool matchAssociativeReduction(PHINode *Phi, BinaryOperator *B) {
assert((!Phi ||
std::find(Phi->op_begin(), Phi->op_end(), B) != Phi->op_end()) &&
"Thi phi needs to use the binary operator");
if (!isValidElementType(Ty))
return false;
+ const DataLayout &DL = B->getModule()->getDataLayout();
ReductionOpcode = B->getOpcode();
ReducedValueOpcode = 0;
- ReduxWidth = MinVecRegSize / DL->getTypeSizeInBits(Ty);
+ ReduxWidth = MinVecRegSize / DL.getTypeSizeInBits(Ty);
ReductionRoot = B;
ReductionPHI = Phi;
// Try to match and vectorize a horizontal reduction.
HorizontalReduction HorRdx;
- if (ShouldVectorizeHor &&
- HorRdx.matchAssociativeReduction(P, BI, DL) &&
+ if (ShouldVectorizeHor && HorRdx.matchAssociativeReduction(P, BI) &&
HorRdx.tryToReduce(R, TTI)) {
Changed = true;
it = BB->begin();
if (BinaryOperator *BinOp =
dyn_cast<BinaryOperator>(SI->getValueOperand())) {
HorizontalReduction HorRdx;
- if (((HorRdx.matchAssociativeReduction(nullptr, BinOp, DL) &&
+ if (((HorRdx.matchAssociativeReduction(nullptr, BinOp) &&
HorRdx.tryToReduce(R, TTI)) ||
tryToVectorize(BinOp, R))) {
Changed = true;
EXPECT_EQ(Ty1, IntegerType::get(Ctx, 1));
DataLayout* DL = new DataLayout(M.get());
- IntegerType *IntPtrTy = Builder.getIntPtrTy(DL);
+ IntegerType *IntPtrTy = Builder.getIntPtrTy(*DL);
unsigned IntPtrBitSize = DL->getPointerSizeInBits(0);
EXPECT_EQ(IntPtrTy, IntegerType::get(Ctx, IntPtrBitSize));
delete DL;
"2:32:32-f64:64:64-v64:64:64-v128:128:128-a:0:64-s:64:64-f80"
":128:128-n8:16:32:64-S128");
// Make sure we don't crash
- GetPointerBaseWithConstantOffset(Gep0, Offset, &TD);
- GetPointerBaseWithConstantOffset(Gep1, Offset, &TD);
- GetPointerBaseWithConstantOffset(Gep2, Offset, &TD);
- GetPointerBaseWithConstantOffset(Gep3, Offset, &TD);
+ GetPointerBaseWithConstantOffset(Gep0, Offset, TD);
+ GetPointerBaseWithConstantOffset(Gep1, Offset, TD);
+ GetPointerBaseWithConstantOffset(Gep2, Offset, TD);
+ GetPointerBaseWithConstantOffset(Gep3, Offset, TD);
// Gep of Geps
GetElementPtrInst *GepII0 = GetElementPtrInst::Create(Gep0, C2xi32b);
projects / oota-llvm.git / commitdiff