//
// The SCEV for %i is {0,+,1}<%L>. The SCEV for %i.next is {1,+,1}<%L>, however
// it's useful to think about these as the same register, with some uses using
-// the value of the register before the add and some using // it after. In this
+// the value of the register before the add and some using it after. In this
// example, the icmp is a post-increment user, since it uses %i.next, which is
// the value of the induction variable after the increment. The other common
// case of post-increment users is users outside the loop.
//
//===----------------------------------------------------------------------===//
-#define DEBUG_TYPE "loop-reduce"
#include "llvm/Transforms/Scalar.h"
#include "llvm/ADT/DenseSet.h"
+#include "llvm/ADT/Hashing.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/SmallBitVector.h"
#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"
#include <algorithm>
using namespace llvm;
+#define DEBUG_TYPE "loop-reduce"
+
/// MaxIVUsers is an arbitrary threshold that provides an early opportunitiy for
/// bail out. This threshold is far beyond the number of users that LSR can
/// conceivably solve, so it should not affect generated code, but catches the
/// a particular register.
SmallBitVector UsedByIndices;
- RegSortData() {}
-
void print(raw_ostream &OS) const;
void dump() const;
};
// Update RegUses. The data structure is not optimized for this purpose;
// we must iterate through it and update each of the bit vectors.
- for (RegUsesTy::iterator I = RegUsesMap.begin(), E = RegUsesMap.end();
- I != E; ++I) {
- SmallBitVector &UsedByIndices = I->second.UsedByIndices;
+ for (auto &Pair : RegUsesMap) {
+ SmallBitVector &UsedByIndices = Pair.second.UsedByIndices;
if (LUIdx < UsedByIndices.size())
UsedByIndices[LUIdx] =
LastLUIdx < UsedByIndices.size() ? UsedByIndices[LastLUIdx] : 0;
int64_t Scale;
/// BaseRegs - The list of "base" registers for this use. When this is
- /// non-empty,
+ /// non-empty. The canonical representation of a formula is
+ /// 1. BaseRegs.size > 1 implies ScaledReg != NULL and
+ /// 2. ScaledReg != NULL implies Scale != 1 || !BaseRegs.empty().
+ /// #1 enforces that the scaled register is always used when at least two
+ /// registers are needed by the formula: e.g., reg1 + reg2 is reg1 + 1 * reg2.
+ /// #2 enforces that 1 * reg is reg.
+ /// This invariant can be temporarly broken while building a formula.
+ /// However, every formula inserted into the LSRInstance must be in canonical
+ /// form.
SmallVector<const SCEV *, 4> BaseRegs;
/// ScaledReg - The 'scaled' register for this use. This should be non-null
int64_t UnfoldedOffset;
Formula()
- : BaseGV(0), BaseOffset(0), HasBaseReg(false), Scale(0), ScaledReg(0),
- UnfoldedOffset(0) {}
+ : BaseGV(nullptr), BaseOffset(0), HasBaseReg(false), Scale(0),
+ ScaledReg(nullptr), UnfoldedOffset(0) {}
void InitialMatch(const SCEV *S, Loop *L, ScalarEvolution &SE);
- unsigned getNumRegs() const;
+ bool isCanonical() const;
+
+ void Canonicalize();
+
+ bool Unscale();
+
+ size_t getNumRegs() const;
Type *getType() const;
void DeleteBaseReg(const SCEV *&S);
// Look at add operands.
if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(S)) {
- for (SCEVAddExpr::op_iterator I = Add->op_begin(), E = Add->op_end();
- I != E; ++I)
- DoInitialMatch(*I, L, Good, Bad, SE);
+ for (const SCEV *S : Add->operands())
+ DoInitialMatch(S, L, Good, Bad, SE);
return;
}
DoInitialMatch(NewMul, L, MyGood, MyBad, SE);
const SCEV *NegOne = SE.getSCEV(ConstantInt::getAllOnesValue(
SE.getEffectiveSCEVType(NewMul->getType())));
- for (SmallVectorImpl<const SCEV *>::const_iterator I = MyGood.begin(),
- E = MyGood.end(); I != E; ++I)
- Good.push_back(SE.getMulExpr(NegOne, *I));
- for (SmallVectorImpl<const SCEV *>::const_iterator I = MyBad.begin(),
- E = MyBad.end(); I != E; ++I)
- Bad.push_back(SE.getMulExpr(NegOne, *I));
+ for (const SCEV *S : MyGood)
+ Good.push_back(SE.getMulExpr(NegOne, S));
+ for (const SCEV *S : MyBad)
+ Bad.push_back(SE.getMulExpr(NegOne, S));
return;
}
BaseRegs.push_back(Sum);
HasBaseReg = true;
}
+ Canonicalize();
+}
+
+/// \brief Check whether or not this formula statisfies the canonical
+/// representation.
+/// \see Formula::BaseRegs.
+bool Formula::isCanonical() const {
+ if (ScaledReg)
+ return Scale != 1 || !BaseRegs.empty();
+ return BaseRegs.size() <= 1;
+}
+
+/// \brief Helper method to morph a formula into its canonical representation.
+/// \see Formula::BaseRegs.
+/// Every formula having more than one base register, must use the ScaledReg
+/// field. Otherwise, we would have to do special cases everywhere in LSR
+/// to treat reg1 + reg2 + ... the same way as reg1 + 1*reg2 + ...
+/// On the other hand, 1*reg should be canonicalized into reg.
+void Formula::Canonicalize() {
+ if (isCanonical())
+ return;
+ // So far we did not need this case. This is easy to implement but it is
+ // useless to maintain dead code. Beside it could hurt compile time.
+ assert(!BaseRegs.empty() && "1*reg => reg, should not be needed.");
+ // Keep the invariant sum in BaseRegs and one of the variant sum in ScaledReg.
+ ScaledReg = BaseRegs.back();
+ BaseRegs.pop_back();
+ Scale = 1;
+ size_t BaseRegsSize = BaseRegs.size();
+ size_t Try = 0;
+ // If ScaledReg is an invariant, try to find a variant expression.
+ while (Try < BaseRegsSize && !isa<SCEVAddRecExpr>(ScaledReg))
+ std::swap(ScaledReg, BaseRegs[Try++]);
+}
+
+/// \brief Get rid of the scale in the formula.
+/// In other words, this method morphes reg1 + 1*reg2 into reg1 + reg2.
+/// \return true if it was possible to get rid of the scale, false otherwise.
+/// \note After this operation the formula may not be in the canonical form.
+bool Formula::Unscale() {
+ if (Scale != 1)
+ return false;
+ Scale = 0;
+ BaseRegs.push_back(ScaledReg);
+ ScaledReg = nullptr;
+ return true;
}
/// getNumRegs - Return the total number of register operands used by this
/// formula. This does not include register uses implied by non-constant
/// addrec strides.
-unsigned Formula::getNumRegs() const {
+size_t Formula::getNumRegs() const {
return !!ScaledReg + BaseRegs.size();
}
return !BaseRegs.empty() ? BaseRegs.front()->getType() :
ScaledReg ? ScaledReg->getType() :
BaseGV ? BaseGV->getType() :
- 0;
+ nullptr;
}
/// DeleteBaseReg - Delete the given base reg from the BaseRegs list.
if (ScaledReg)
if (RegUses.isRegUsedByUsesOtherThan(ScaledReg, LUIdx))
return true;
- for (SmallVectorImpl<const SCEV *>::const_iterator I = BaseRegs.begin(),
- E = BaseRegs.end(); I != E; ++I)
- if (RegUses.isRegUsedByUsesOtherThan(*I, LUIdx))
+ for (const SCEV *BaseReg : BaseRegs)
+ if (RegUses.isRegUsedByUsesOtherThan(BaseReg, LUIdx))
return true;
return false;
}
if (!First) OS << " + "; else First = false;
OS << BaseOffset;
}
- for (SmallVectorImpl<const SCEV *>::const_iterator I = BaseRegs.begin(),
- E = BaseRegs.end(); I != E; ++I) {
+ for (const SCEV *BaseReg : BaseRegs) {
if (!First) OS << " + "; else First = false;
- OS << "reg(" << **I << ')';
+ OS << "reg(" << *BaseReg << ')';
}
if (HasBaseReg && BaseRegs.empty()) {
if (!First) OS << " + "; else First = false;
OS << ')';
}
if (UnfoldedOffset != 0) {
- if (!First) OS << " + "; else First = false;
+ if (!First) OS << " + ";
OS << "imm(" << UnfoldedOffset << ')';
}
}
// Check for a division of a constant by a constant.
if (const SCEVConstant *C = dyn_cast<SCEVConstant>(LHS)) {
if (!RC)
- return 0;
+ return nullptr;
const APInt &LA = C->getValue()->getValue();
const APInt &RA = RC->getValue()->getValue();
if (LA.srem(RA) != 0)
- return 0;
+ return nullptr;
return SE.getConstant(LA.sdiv(RA));
}
if (IgnoreSignificantBits || isAddRecSExtable(AR, SE)) {
const SCEV *Step = getExactSDiv(AR->getStepRecurrence(SE), RHS, SE,
IgnoreSignificantBits);
- if (!Step) return 0;
+ if (!Step) return nullptr;
const SCEV *Start = getExactSDiv(AR->getStart(), RHS, SE,
IgnoreSignificantBits);
- if (!Start) return 0;
+ if (!Start) return nullptr;
// FlagNW is independent of the start value, step direction, and is
// preserved with smaller magnitude steps.
// FIXME: AR->getNoWrapFlags(SCEV::FlagNW)
return SE.getAddRecExpr(Start, Step, AR->getLoop(), SCEV::FlagAnyWrap);
}
- return 0;
+ return nullptr;
}
// Distribute the sdiv over add operands, if the add doesn't overflow.
if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(LHS)) {
if (IgnoreSignificantBits || isAddSExtable(Add, SE)) {
SmallVector<const SCEV *, 8> Ops;
- for (SCEVAddExpr::op_iterator I = Add->op_begin(), E = Add->op_end();
- I != E; ++I) {
- const SCEV *Op = getExactSDiv(*I, RHS, SE,
- IgnoreSignificantBits);
- if (!Op) return 0;
+ for (const SCEV *S : Add->operands()) {
+ const SCEV *Op = getExactSDiv(S, RHS, SE, IgnoreSignificantBits);
+ if (!Op) return nullptr;
Ops.push_back(Op);
}
return SE.getAddExpr(Ops);
}
- return 0;
+ return nullptr;
}
// Check for a multiply operand that we can pull RHS out of.
if (IgnoreSignificantBits || isMulSExtable(Mul, SE)) {
SmallVector<const SCEV *, 4> Ops;
bool Found = false;
- for (SCEVMulExpr::op_iterator I = Mul->op_begin(), E = Mul->op_end();
- I != E; ++I) {
- const SCEV *S = *I;
+ for (const SCEV *S : Mul->operands()) {
if (!Found)
if (const SCEV *Q = getExactSDiv(S, RHS, SE,
IgnoreSignificantBits)) {
}
Ops.push_back(S);
}
- return Found ? SE.getMulExpr(Ops) : 0;
+ return Found ? SE.getMulExpr(Ops) : nullptr;
}
- return 0;
+ return nullptr;
}
// Otherwise we don't know.
- return 0;
+ return nullptr;
}
/// ExtractImmediate - If S involves the addition of a constant integer value,
SCEV::FlagAnyWrap);
return Result;
}
- return 0;
+ return nullptr;
}
/// isAddressUse - Returns true if the specified instruction is using the
/// TODO: Allow UDivExpr if we can find an existing IV increment that is an
/// obvious multiple of the UDivExpr.
static bool isHighCostExpansion(const SCEV *S,
- SmallPtrSet<const SCEV*, 8> &Processed,
+ SmallPtrSetImpl<const SCEV*> &Processed,
ScalarEvolution &SE) {
// Zero/One operand expressions
switch (S->getSCEVType()) {
Processed, SE);
}
- if (!Processed.insert(S))
+ if (!Processed.insert(S).second)
return false;
if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(S)) {
- for (SCEVAddExpr::op_iterator I = Add->op_begin(), E = Add->op_end();
- I != E; ++I) {
- if (isHighCostExpansion(*I, Processed, SE))
+ for (const SCEV *S : Add->operands()) {
+ if (isHighCostExpansion(S, Processed, SE))
return true;
}
return false;
Value *V = DeadInsts.pop_back_val();
Instruction *I = dyn_cast_or_null<Instruction>(V);
- if (I == 0 || !isInstructionTriviallyDead(I))
+ if (!I || !isInstructionTriviallyDead(I))
continue;
- for (User::op_iterator OI = I->op_begin(), E = I->op_end(); OI != E; ++OI)
- if (Instruction *U = dyn_cast<Instruction>(*OI)) {
- *OI = 0;
+ for (Use &O : I->operands())
+ if (Instruction *U = dyn_cast<Instruction>(O)) {
+ O = nullptr;
if (U->use_empty())
- DeadInsts.push_back(U);
+ DeadInsts.emplace_back(U);
}
I->eraseFromParent();
namespace {
class LSRUse;
}
-// Check if it is legal to fold 2 base registers.
-static bool isLegal2RegAMUse(const TargetTransformInfo &TTI, const LSRUse &LU,
- const Formula &F);
+
+/// \brief Check if the addressing mode defined by \p F is completely
+/// folded in \p LU at isel time.
+/// This includes address-mode folding and special icmp tricks.
+/// This function returns true if \p LU can accommodate what \p F
+/// defines and up to 1 base + 1 scaled + offset.
+/// In other words, if \p F has several base registers, this function may
+/// still return true. Therefore, users still need to account for
+/// additional base registers and/or unfolded offsets to derive an
+/// accurate cost model.
+static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
+ const LSRUse &LU, const Formula &F);
// Get the cost of the scaling factor used in F for LU.
static unsigned getScalingFactorCost(const TargetTransformInfo &TTI,
const LSRUse &LU, const Formula &F);
void RateFormula(const TargetTransformInfo &TTI,
const Formula &F,
- SmallPtrSet<const SCEV *, 16> &Regs,
+ SmallPtrSetImpl<const SCEV *> &Regs,
const DenseSet<const SCEV *> &VisitedRegs,
const Loop *L,
const SmallVectorImpl<int64_t> &Offsets,
ScalarEvolution &SE, DominatorTree &DT,
const LSRUse &LU,
- SmallPtrSet<const SCEV *, 16> *LoserRegs = 0);
+ SmallPtrSetImpl<const SCEV *> *LoserRegs = nullptr);
void print(raw_ostream &OS) const;
void dump() const;
private:
void RateRegister(const SCEV *Reg,
- SmallPtrSet<const SCEV *, 16> &Regs,
+ SmallPtrSetImpl<const SCEV *> &Regs,
const Loop *L,
ScalarEvolution &SE, DominatorTree &DT);
void RatePrimaryRegister(const SCEV *Reg,
- SmallPtrSet<const SCEV *, 16> &Regs,
+ SmallPtrSetImpl<const SCEV *> &Regs,
const Loop *L,
ScalarEvolution &SE, DominatorTree &DT,
- SmallPtrSet<const SCEV *, 16> *LoserRegs);
+ SmallPtrSetImpl<const SCEV *> *LoserRegs);
};
}
/// RateRegister - Tally up interesting quantities from the given register.
void Cost::RateRegister(const SCEV *Reg,
- SmallPtrSet<const SCEV *, 16> &Regs,
+ SmallPtrSetImpl<const SCEV *> &Regs,
const Loop *L,
ScalarEvolution &SE, DominatorTree &DT) {
if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(Reg)) {
/// before, rate it. Optional LoserRegs provides a way to declare any formula
/// that refers to one of those regs an instant loser.
void Cost::RatePrimaryRegister(const SCEV *Reg,
- SmallPtrSet<const SCEV *, 16> &Regs,
+ SmallPtrSetImpl<const SCEV *> &Regs,
const Loop *L,
ScalarEvolution &SE, DominatorTree &DT,
- SmallPtrSet<const SCEV *, 16> *LoserRegs) {
+ SmallPtrSetImpl<const SCEV *> *LoserRegs) {
if (LoserRegs && LoserRegs->count(Reg)) {
Lose();
return;
}
- if (Regs.insert(Reg)) {
+ if (Regs.insert(Reg).second) {
RateRegister(Reg, Regs, L, SE, DT);
if (LoserRegs && isLoser())
LoserRegs->insert(Reg);
void Cost::RateFormula(const TargetTransformInfo &TTI,
const Formula &F,
- SmallPtrSet<const SCEV *, 16> &Regs,
+ SmallPtrSetImpl<const SCEV *> &Regs,
const DenseSet<const SCEV *> &VisitedRegs,
const Loop *L,
const SmallVectorImpl<int64_t> &Offsets,
ScalarEvolution &SE, DominatorTree &DT,
const LSRUse &LU,
- SmallPtrSet<const SCEV *, 16> *LoserRegs) {
+ SmallPtrSetImpl<const SCEV *> *LoserRegs) {
+ assert(F.isCanonical() && "Cost is accurate only for canonical formula");
// Tally up the registers.
if (const SCEV *ScaledReg = F.ScaledReg) {
if (VisitedRegs.count(ScaledReg)) {
if (isLoser())
return;
}
- for (SmallVectorImpl<const SCEV *>::const_iterator I = F.BaseRegs.begin(),
- E = F.BaseRegs.end(); I != E; ++I) {
- const SCEV *BaseReg = *I;
+ for (const SCEV *BaseReg : F.BaseRegs) {
if (VisitedRegs.count(BaseReg)) {
Lose();
return;
}
// Determine how many (unfolded) adds we'll need inside the loop.
- size_t NumBaseParts = F.BaseRegs.size() + (F.UnfoldedOffset != 0);
+ size_t NumBaseParts = F.getNumRegs();
if (NumBaseParts > 1)
// Do not count the base and a possible second register if the target
// allows to fold 2 registers.
- NumBaseAdds += NumBaseParts - (1 + isLegal2RegAMUse(TTI, LU, F));
+ NumBaseAdds +=
+ NumBaseParts - (1 + (F.Scale && isAMCompletelyFolded(TTI, LU, F)));
+ NumBaseAdds += (F.UnfoldedOffset != 0);
// Accumulate non-free scaling amounts.
ScaleCost += getScalingFactorCost(TTI, LU, F);
// Tally up the non-zero immediates.
- for (SmallVectorImpl<int64_t>::const_iterator I = Offsets.begin(),
- E = Offsets.end(); I != E; ++I) {
- int64_t Offset = (uint64_t)*I + F.BaseOffset;
+ for (int64_t O : Offsets) {
+ int64_t Offset = (uint64_t)O + F.BaseOffset;
if (F.BaseGV)
ImmCost += 64; // Handle symbolic values conservatively.
// TODO: This should probably be the pointer size.
}
LSRFixup::LSRFixup()
- : UserInst(0), OperandValToReplace(0), LUIdx(~size_t(0)), Offset(0) {}
+ : UserInst(nullptr), OperandValToReplace(nullptr), LUIdx(~size_t(0)),
+ Offset(0) {}
/// isUseFullyOutsideLoop - Test whether this fixup always uses its
/// value outside of the given loop.
OS << ", OperandValToReplace=";
OperandValToReplace->printAsOperand(OS, /*PrintType=*/false);
- for (PostIncLoopSet::const_iterator I = PostIncLoops.begin(),
- E = PostIncLoops.end(); I != E; ++I) {
+ for (const Loop *PIL : PostIncLoops) {
OS << ", PostIncLoop=";
- (*I)->getHeader()->printAsOperand(OS, /*PrintType=*/false);
+ PIL->getHeader()->printAsOperand(OS, /*PrintType=*/false);
}
if (LUIdx != ~size_t(0))
}
static unsigned getHashValue(const SmallVector<const SCEV *, 4> &V) {
- unsigned Result = 0;
- for (SmallVectorImpl<const SCEV *>::const_iterator I = V.begin(),
- E = V.end(); I != E; ++I)
- Result ^= DenseMapInfo<const SCEV *>::getHashValue(*I);
- return Result;
+ return static_cast<unsigned>(hash_combine_range(V.begin(), V.end()));
}
static bool isEqual(const SmallVector<const SCEV *, 4> &LHS,
// TODO: Add a generic icmp too?
};
+ typedef PointerIntPair<const SCEV *, 2, KindType> SCEVUseKindPair;
+
KindType Kind;
Type *AccessTy;
MaxOffset(INT64_MIN),
AllFixupsOutsideLoop(true),
RigidFormula(false),
- WidestFixupType(0) {}
+ WidestFixupType(nullptr) {}
bool HasFormulaWithSameRegs(const Formula &F) const;
bool InsertFormula(const Formula &F);
/// InsertFormula - If the given formula has not yet been inserted, add it to
/// the list, and return true. Return false otherwise.
+/// The formula must be in canonical form.
bool LSRUse::InsertFormula(const Formula &F) {
+ assert(F.isCanonical() && "Invalid canonical representation");
+
if (!Formulae.empty() && RigidFormula)
return false;
assert((!F.ScaledReg || !F.ScaledReg->isZero()) &&
"Zero allocated in a scaled register!");
#ifndef NDEBUG
- for (SmallVectorImpl<const SCEV *>::const_iterator I =
- F.BaseRegs.begin(), E = F.BaseRegs.end(); I != E; ++I)
- assert(!(*I)->isZero() && "Zero allocated in a base register!");
+ for (const SCEV *BaseReg : F.BaseRegs)
+ assert(!BaseReg->isZero() && "Zero allocated in a base register!");
#endif
// Add the formula to the list.
// Record registers now being used by this use.
Regs.insert(F.BaseRegs.begin(), F.BaseRegs.end());
+ if (F.ScaledReg)
+ Regs.insert(F.ScaledReg);
return true;
}
/// RecomputeRegs - Recompute the Regs field, and update RegUses.
void LSRUse::RecomputeRegs(size_t LUIdx, RegUseTracker &RegUses) {
// Now that we've filtered out some formulae, recompute the Regs set.
- SmallPtrSet<const SCEV *, 4> OldRegs = Regs;
+ SmallPtrSet<const SCEV *, 4> OldRegs = std::move(Regs);
Regs.clear();
- for (SmallVectorImpl<Formula>::const_iterator I = Formulae.begin(),
- E = Formulae.end(); I != E; ++I) {
- const Formula &F = *I;
+ for (const Formula &F : Formulae) {
if (F.ScaledReg) Regs.insert(F.ScaledReg);
Regs.insert(F.BaseRegs.begin(), F.BaseRegs.end());
}
// Update the RegTracker.
- for (SmallPtrSet<const SCEV *, 4>::iterator I = OldRegs.begin(),
- E = OldRegs.end(); I != E; ++I)
- if (!Regs.count(*I))
- RegUses.DropRegister(*I, LUIdx);
+ for (const SCEV *S : OldRegs)
+ if (!Regs.count(S))
+ RegUses.DropRegister(S, LUIdx);
}
void LSRUse::print(raw_ostream &OS) const {
}
OS << ", Offsets={";
- for (SmallVectorImpl<int64_t>::const_iterator I = Offsets.begin(),
- E = Offsets.end(); I != E; ++I) {
- OS << *I;
- if (std::next(I) != E)
- OS << ',';
+ bool NeedComma = false;
+ for (int64_t O : Offsets) {
+ if (NeedComma) OS << ',';
+ OS << O;
+ NeedComma = true;
}
OS << '}';
}
#endif
-/// isLegalUse - Test whether the use described by AM is "legal", meaning it can
-/// be completely folded into the user instruction at isel time. This includes
-/// address-mode folding and special icmp tricks.
-static bool isLegalUse(const TargetTransformInfo &TTI, LSRUse::KindType Kind,
- Type *AccessTy, GlobalValue *BaseGV, int64_t BaseOffset,
- bool HasBaseReg, int64_t Scale) {
+static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
+ LSRUse::KindType Kind, Type *AccessTy,
+ GlobalValue *BaseGV, int64_t BaseOffset,
+ bool HasBaseReg, int64_t Scale) {
switch (Kind) {
case LSRUse::Address:
return TTI.isLegalAddressingMode(AccessTy, BaseGV, BaseOffset, HasBaseReg, Scale);
- // Otherwise, just guess that reg+reg addressing is legal.
- //return ;
-
case LSRUse::ICmpZero:
// There's not even a target hook for querying whether it would be legal to
// fold a GV into an ICmp.
llvm_unreachable("Invalid LSRUse Kind!");
}
-static bool isLegalUse(const TargetTransformInfo &TTI, int64_t MinOffset,
- int64_t MaxOffset, LSRUse::KindType Kind, Type *AccessTy,
- GlobalValue *BaseGV, int64_t BaseOffset, bool HasBaseReg,
- int64_t Scale) {
+static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
+ int64_t MinOffset, int64_t MaxOffset,
+ LSRUse::KindType Kind, Type *AccessTy,
+ GlobalValue *BaseGV, int64_t BaseOffset,
+ bool HasBaseReg, int64_t Scale) {
// Check for overflow.
if (((int64_t)((uint64_t)BaseOffset + MinOffset) > BaseOffset) !=
(MinOffset > 0))
return false;
MaxOffset = (uint64_t)BaseOffset + MaxOffset;
- return isLegalUse(TTI, Kind, AccessTy, BaseGV, MinOffset, HasBaseReg,
- Scale) &&
- isLegalUse(TTI, Kind, AccessTy, BaseGV, MaxOffset, HasBaseReg, Scale);
+ return isAMCompletelyFolded(TTI, Kind, AccessTy, BaseGV, MinOffset,
+ HasBaseReg, Scale) &&
+ isAMCompletelyFolded(TTI, Kind, AccessTy, BaseGV, MaxOffset,
+ HasBaseReg, Scale);
+}
+
+static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
+ int64_t MinOffset, int64_t MaxOffset,
+ LSRUse::KindType Kind, Type *AccessTy,
+ const Formula &F) {
+ // For the purpose of isAMCompletelyFolded either having a canonical formula
+ // or a scale not equal to zero is correct.
+ // Problems may arise from non canonical formulae having a scale == 0.
+ // Strictly speaking it would best to just rely on canonical formulae.
+ // However, when we generate the scaled formulae, we first check that the
+ // scaling factor is profitable before computing the actual ScaledReg for
+ // compile time sake.
+ assert((F.isCanonical() || F.Scale != 0));
+ return isAMCompletelyFolded(TTI, MinOffset, MaxOffset, Kind, AccessTy,
+ F.BaseGV, F.BaseOffset, F.HasBaseReg, F.Scale);
+}
+
+/// isLegalUse - Test whether we know how to expand the current formula.
+static bool isLegalUse(const TargetTransformInfo &TTI, int64_t MinOffset,
+ int64_t MaxOffset, LSRUse::KindType Kind, Type *AccessTy,
+ GlobalValue *BaseGV, int64_t BaseOffset, bool HasBaseReg,
+ int64_t Scale) {
+ // We know how to expand completely foldable formulae.
+ return isAMCompletelyFolded(TTI, MinOffset, MaxOffset, Kind, AccessTy, BaseGV,
+ BaseOffset, HasBaseReg, Scale) ||
+ // Or formulae that use a base register produced by a sum of base
+ // registers.
+ (Scale == 1 &&
+ isAMCompletelyFolded(TTI, MinOffset, MaxOffset, Kind, AccessTy,
+ BaseGV, BaseOffset, true, 0));
}
static bool isLegalUse(const TargetTransformInfo &TTI, int64_t MinOffset,
F.BaseOffset, F.HasBaseReg, F.Scale);
}
-static bool isLegal2RegAMUse(const TargetTransformInfo &TTI, const LSRUse &LU,
- const Formula &F) {
- // If F is used as an Addressing Mode, it may fold one Base plus one
- // scaled register. If the scaled register is nil, do as if another
- // element of the base regs is a 1-scaled register.
- // This is possible if BaseRegs has at least 2 registers.
-
- // If this is not an address calculation, this is not an addressing mode
- // use.
- if (LU.Kind != LSRUse::Address)
- return false;
-
- // F is already scaled.
- if (F.Scale != 0)
- return false;
-
- // We need to keep one register for the base and one to scale.
- if (F.BaseRegs.size() < 2)
- return false;
-
- return isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy,
- F.BaseGV, F.BaseOffset, F.HasBaseReg, 1);
- }
+static bool isAMCompletelyFolded(const TargetTransformInfo &TTI,
+ const LSRUse &LU, const Formula &F) {
+ return isAMCompletelyFolded(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind,
+ LU.AccessTy, F.BaseGV, F.BaseOffset, F.HasBaseReg,
+ F.Scale);
+}
static unsigned getScalingFactorCost(const TargetTransformInfo &TTI,
const LSRUse &LU, const Formula &F) {
if (!F.Scale)
return 0;
- assert(isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind,
- LU.AccessTy, F) && "Illegal formula in use.");
+
+ // If the use is not completely folded in that instruction, we will have to
+ // pay an extra cost only for scale != 1.
+ if (!isAMCompletelyFolded(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind,
+ LU.AccessTy, F))
+ return F.Scale != 1;
switch (LU.Kind) {
case LSRUse::Address: {
return std::max(ScaleCostMinOffset, ScaleCostMaxOffset);
}
case LSRUse::ICmpZero:
- // ICmpZero BaseReg + -1*ScaleReg => ICmp BaseReg, ScaleReg.
- // Therefore, return 0 in case F.Scale == -1.
- return F.Scale != -1;
-
case LSRUse::Basic:
case LSRUse::Special:
+ // The use is completely folded, i.e., everything is folded into the
+ // instruction.
return 0;
}
HasBaseReg = true;
}
- return isLegalUse(TTI, Kind, AccessTy, BaseGV, BaseOffset, HasBaseReg, Scale);
+ return isAMCompletelyFolded(TTI, Kind, AccessTy, BaseGV, BaseOffset,
+ HasBaseReg, Scale);
}
static bool isAlwaysFoldable(const TargetTransformInfo &TTI,
// base and a scale.
int64_t Scale = Kind == LSRUse::ICmpZero ? -1 : 1;
- return isLegalUse(TTI, MinOffset, MaxOffset, Kind, AccessTy, BaseGV,
- BaseOffset, HasBaseReg, Scale);
+ return isAMCompletelyFolded(TTI, MinOffset, MaxOffset, Kind, AccessTy, BaseGV,
+ BaseOffset, HasBaseReg, Scale);
}
namespace {
-/// UseMapDenseMapInfo - A DenseMapInfo implementation for holding
-/// DenseMaps and DenseSets of pairs of const SCEV* and LSRUse::Kind.
-struct UseMapDenseMapInfo {
- static std::pair<const SCEV *, LSRUse::KindType> getEmptyKey() {
- return std::make_pair(reinterpret_cast<const SCEV *>(-1), LSRUse::Basic);
- }
-
- static std::pair<const SCEV *, LSRUse::KindType> getTombstoneKey() {
- return std::make_pair(reinterpret_cast<const SCEV *>(-2), LSRUse::Basic);
- }
-
- static unsigned
- getHashValue(const std::pair<const SCEV *, LSRUse::KindType> &V) {
- unsigned Result = DenseMapInfo<const SCEV *>::getHashValue(V.first);
- Result ^= DenseMapInfo<unsigned>::getHashValue(unsigned(V.second));
- return Result;
- }
-
- static bool isEqual(const std::pair<const SCEV *, LSRUse::KindType> &LHS,
- const std::pair<const SCEV *, LSRUse::KindType> &RHS) {
- return LHS == RHS;
- }
-};
-
/// IVInc - An individual increment in a Chain of IV increments.
/// Relate an IV user to an expression that computes the IV it uses from the IV
/// used by the previous link in the Chain.
SmallVector<IVInc,1> Incs;
const SCEV *ExprBase;
- IVChain() : ExprBase(0) {}
+ IVChain() : ExprBase(nullptr) {}
IVChain(const IVInc &Head, const SCEV *Base)
: Incs(1, Head), ExprBase(Base) {}
}
// Support for sharing of LSRUses between LSRFixups.
- typedef DenseMap<std::pair<const SCEV *, LSRUse::KindType>,
- size_t,
- UseMapDenseMapInfo> UseMapTy;
+ typedef DenseMap<LSRUse::SCEVUseKindPair, size_t> UseMapTy;
UseMapTy UseMap;
bool reconcileNewOffset(LSRUse &LU, int64_t NewOffset, bool HasBaseReg,
void GenerateReassociations(LSRUse &LU, unsigned LUIdx, Formula Base,
unsigned Depth = 0);
+
+ void GenerateReassociationsImpl(LSRUse &LU, unsigned LUIdx,
+ const Formula &Base, unsigned Depth,
+ size_t Idx, bool IsScaledReg = false);
void GenerateCombinations(LSRUse &LU, unsigned LUIdx, Formula Base);
+ void GenerateSymbolicOffsetsImpl(LSRUse &LU, unsigned LUIdx,
+ const Formula &Base, size_t Idx,
+ bool IsScaledReg = false);
void GenerateSymbolicOffsets(LSRUse &LU, unsigned LUIdx, Formula Base);
+ void GenerateConstantOffsetsImpl(LSRUse &LU, unsigned LUIdx,
+ const Formula &Base,
+ const SmallVectorImpl<int64_t> &Worklist,
+ size_t Idx, bool IsScaledReg = false);
void GenerateConstantOffsets(LSRUse &LU, unsigned LUIdx, Formula Base);
void GenerateICmpZeroScales(LSRUse &LU, unsigned LUIdx, Formula Base);
void GenerateScales(LSRUse &LU, unsigned LUIdx, Formula Base);
IVUsers::const_iterator CandidateUI = UI;
++UI;
Instruction *ShadowUse = CandidateUI->getUser();
- Type *DestTy = 0;
+ Type *DestTy = nullptr;
bool IsSigned = false;
/* If shadow use is a int->float cast then insert a second IV
continue;
/* Initialize new IV, double d = 0.0 in above example. */
- ConstantInt *C = 0;
+ ConstantInt *C = nullptr;
if (Incr->getOperand(0) == PH)
C = dyn_cast<ConstantInt>(Incr->getOperand(1));
else if (Incr->getOperand(1) == PH)
/// set the IV user and stride information and return true, otherwise return
/// false.
bool LSRInstance::FindIVUserForCond(ICmpInst *Cond, IVStrideUse *&CondUse) {
- for (IVUsers::iterator UI = IU.begin(), E = IU.end(); UI != E; ++UI)
- if (UI->getUser() == Cond) {
+ for (IVStrideUse &U : IU)
+ if (U.getUser() == Cond) {
// NOTE: we could handle setcc instructions with multiple uses here, but
// InstCombine does it as well for simple uses, it's not clear that it
// occurs enough in real life to handle.
- CondUse = UI;
+ CondUse = &U;
return true;
}
return false;
// for ICMP_ULE here because the comparison would be with zero, which
// isn't interesting.
CmpInst::Predicate Pred = ICmpInst::BAD_ICMP_PREDICATE;
- const SCEVNAryExpr *Max = 0;
+ const SCEVNAryExpr *Max = nullptr;
if (const SCEVSMaxExpr *S = dyn_cast<SCEVSMaxExpr>(BackedgeTakenCount)) {
Pred = ICmpInst::ICMP_SLE;
Max = S;
// Check the right operand of the select, and remember it, as it will
// be used in the new comparison instruction.
- Value *NewRHS = 0;
+ Value *NewRHS = nullptr;
if (ICmpInst::isTrueWhenEqual(Pred)) {
// Look for n+1, and grab n.
if (AddOperator *BO = dyn_cast<AddOperator>(Sel->getOperand(1)))
SmallVector<BasicBlock*, 8> ExitingBlocks;
L->getExitingBlocks(ExitingBlocks);
- for (unsigned i = 0, e = ExitingBlocks.size(); i != e; ++i) {
- BasicBlock *ExitingBlock = ExitingBlocks[i];
+ for (BasicBlock *ExitingBlock : ExitingBlocks) {
// Get the terminating condition for the loop if possible. If we
// can, we want to change it to use a post-incremented version of its
continue;
// Search IVUsesByStride to find Cond's IVUse if there is one.
- IVStrideUse *CondUse = 0;
+ IVStrideUse *CondUse = nullptr;
ICmpInst *Cond = cast<ICmpInst>(TermBr->getCondition());
if (!FindIVUserForCond(Cond, CondUse))
continue;
// Check for possible scaled-address reuse.
Type *AccessTy = getAccessType(UI->getUser());
int64_t Scale = C->getSExtValue();
- if (TTI.isLegalAddressingMode(AccessTy, /*BaseGV=*/ 0,
+ if (TTI.isLegalAddressingMode(AccessTy, /*BaseGV=*/ nullptr,
/*BaseOffset=*/ 0,
/*HasBaseReg=*/ false, Scale))
goto decline_post_inc;
Scale = -Scale;
- if (TTI.isLegalAddressingMode(AccessTy, /*BaseGV=*/ 0,
+ if (TTI.isLegalAddressingMode(AccessTy, /*BaseGV=*/ nullptr,
/*BaseOffset=*/ 0,
/*HasBaseReg=*/ false, Scale))
goto decline_post_inc;
// must dominate all the post-inc comparisons we just set up, and it must
// dominate the loop latch edge.
IVIncInsertPos = L->getLoopLatch()->getTerminator();
- for (SmallPtrSet<Instruction *, 4>::const_iterator I = PostIncs.begin(),
- E = PostIncs.end(); I != E; ++I) {
+ for (Instruction *Inst : PostIncs) {
BasicBlock *BB =
DT.findNearestCommonDominator(IVIncInsertPos->getParent(),
- (*I)->getParent());
- if (BB == (*I)->getParent())
- IVIncInsertPos = *I;
+ Inst->getParent());
+ if (BB == Inst->getParent())
+ IVIncInsertPos = Inst;
else if (BB != IVIncInsertPos->getParent())
IVIncInsertPos = BB->getTerminator();
}
// the uses will have all its uses outside the loop, for example.
if (LU.Kind != Kind)
return false;
+
+ // Check for a mismatched access type, and fall back conservatively as needed.
+ // TODO: Be less conservative when the type is similar and can use the same
+ // addressing modes.
+ if (Kind == LSRUse::Address && AccessTy != LU.AccessTy)
+ NewAccessTy = Type::getVoidTy(AccessTy->getContext());
+
// Conservatively assume HasBaseReg is true for now.
if (NewOffset < LU.MinOffset) {
- if (!isAlwaysFoldable(TTI, Kind, AccessTy, /*BaseGV=*/ 0,
+ if (!isAlwaysFoldable(TTI, Kind, NewAccessTy, /*BaseGV=*/nullptr,
LU.MaxOffset - NewOffset, HasBaseReg))
return false;
NewMinOffset = NewOffset;
} else if (NewOffset > LU.MaxOffset) {
- if (!isAlwaysFoldable(TTI, Kind, AccessTy, /*BaseGV=*/ 0,
+ if (!isAlwaysFoldable(TTI, Kind, NewAccessTy, /*BaseGV=*/nullptr,
NewOffset - LU.MinOffset, HasBaseReg))
return false;
NewMaxOffset = NewOffset;
}
- // Check for a mismatched access type, and fall back conservatively as needed.
- // TODO: Be less conservative when the type is similar and can use the same
- // addressing modes.
- if (Kind == LSRUse::Address && AccessTy != LU.AccessTy)
- NewAccessTy = Type::getVoidTy(AccessTy->getContext());
// Update the use.
LU.MinOffset = NewMinOffset;
int64_t Offset = ExtractImmediate(Expr, SE);
// Basic uses can't accept any offset, for example.
- if (!isAlwaysFoldable(TTI, Kind, AccessTy, /*BaseGV=*/ 0,
+ if (!isAlwaysFoldable(TTI, Kind, AccessTy, /*BaseGV=*/ nullptr,
Offset, /*HasBaseReg=*/ true)) {
Expr = Copy;
Offset = 0;
}
std::pair<UseMapTy::iterator, bool> P =
- UseMap.insert(std::make_pair(std::make_pair(Expr, Kind), 0));
+ UseMap.insert(std::make_pair(LSRUse::SCEVUseKindPair(Expr, Kind), 0));
if (!P.second) {
// A use already existed with this base.
size_t LUIdx = P.first->second;
LU.WidestFixupType == OrigLU.WidestFixupType &&
LU.HasFormulaWithSameRegs(OrigF)) {
// Scan through this use's formulae.
- for (SmallVectorImpl<Formula>::const_iterator I = LU.Formulae.begin(),
- E = LU.Formulae.end(); I != E; ++I) {
- const Formula &F = *I;
+ for (const Formula &F : LU.Formulae) {
// Check to see if this formula has the same registers and symbols
// as OrigF.
if (F.BaseRegs == OrigF.BaseRegs &&
}
// Nothing looked good.
- return 0;
+ return nullptr;
}
void LSRInstance::CollectInterestingTypesAndFactors() {
// Collect interesting types and strides.
SmallVector<const SCEV *, 4> Worklist;
- for (IVUsers::const_iterator UI = IU.begin(), E = IU.end(); UI != E; ++UI) {
- const SCEV *Expr = IU.getExpr(*UI);
+ for (const IVStrideUse &U : IU) {
+ const SCEV *Expr = IU.getExpr(U);
// Collect interesting types.
Types.insert(SE.getEffectiveSCEVType(Expr->getType()));
default: // uncluding scUnknown.
return S;
case scConstant:
- return 0;
+ return nullptr;
case scTruncate:
return getExprBase(cast<SCEVTruncateExpr>(S)->getOperand());
case scZeroExtend:
///
/// TODO: Consider IVInc free if it's already used in another chains.
static bool
-isProfitableChain(IVChain &Chain, SmallPtrSet<Instruction*, 4> &Users,
+isProfitableChain(IVChain &Chain, SmallPtrSetImpl<Instruction*> &Users,
ScalarEvolution &SE, const TargetTransformInfo &TTI) {
if (StressIVChain)
return true;
if (!Users.empty()) {
DEBUG(dbgs() << "Chain: " << *Chain.Incs[0].UserInst << " users:\n";
- for (SmallPtrSet<Instruction*, 4>::const_iterator I = Users.begin(),
- E = Users.end(); I != E; ++I) {
- dbgs() << " " << **I << "\n";
+ for (Instruction *Inst : Users) {
+ dbgs() << " " << *Inst << "\n";
});
return false;
}
&& SE.getSCEV(Chain.tailUserInst()) == Chain.Incs[0].IncExpr) {
--cost;
}
- const SCEV *LastIncExpr = 0;
+ const SCEV *LastIncExpr = nullptr;
unsigned NumConstIncrements = 0;
unsigned NumVarIncrements = 0;
unsigned NumReusedIncrements = 0;
- for (IVChain::const_iterator I = Chain.begin(), E = Chain.end();
- I != E; ++I) {
-
- if (I->IncExpr->isZero())
+ for (const IVInc &Inc : Chain) {
+ if (Inc.IncExpr->isZero())
continue;
// Incrementing by zero or some constant is neutral. We assume constants can
// be folded into an addressing mode or an add's immediate operand.
- if (isa<SCEVConstant>(I->IncExpr)) {
+ if (isa<SCEVConstant>(Inc.IncExpr)) {
++NumConstIncrements;
continue;
}
- if (I->IncExpr == LastIncExpr)
+ if (Inc.IncExpr == LastIncExpr)
++NumReusedIncrements;
else
++NumVarIncrements;
- LastIncExpr = I->IncExpr;
+ LastIncExpr = Inc.IncExpr;
}
// An IV chain with a single increment is handled by LSR's postinc
// uses. However, a chain with multiple increments requires keeping the IV's
// Visit all existing chains. Check if its IVOper can be computed as a
// profitable loop invariant increment from the last link in the Chain.
unsigned ChainIdx = 0, NChains = IVChainVec.size();
- const SCEV *LastIncExpr = 0;
+ const SCEV *LastIncExpr = nullptr;
for (; ChainIdx < NChains; ++ChainIdx) {
IVChain &Chain = IVChainVec[ChainIdx];
User::op_iterator IVOpIter = findIVOperand(I->op_begin(), IVOpEnd, L, SE);
while (IVOpIter != IVOpEnd) {
Instruction *IVOpInst = cast<Instruction>(*IVOpIter);
- if (UniqueOperands.insert(IVOpInst))
+ if (UniqueOperands.insert(IVOpInst).second)
ChainInstruction(I, IVOpInst, ChainUsersVec);
IVOpIter = findIVOperand(std::next(IVOpIter), IVOpEnd, L, SE);
}
assert(!Chain.Incs.empty() && "empty IV chains are not allowed");
DEBUG(dbgs() << "Final Chain: " << *Chain.Incs[0].UserInst << "\n");
- for (IVChain::const_iterator I = Chain.begin(), E = Chain.end();
- I != E; ++I) {
- DEBUG(dbgs() << " Inc: " << *I->UserInst << "\n");
- User::op_iterator UseI =
- std::find(I->UserInst->op_begin(), I->UserInst->op_end(), I->IVOperand);
- assert(UseI != I->UserInst->op_end() && "cannot find IV operand");
+ for (const IVInc &Inc : Chain) {
+ DEBUG(dbgs() << " Inc: " << Inc.UserInst << "\n");
+ auto UseI = std::find(Inc.UserInst->op_begin(), Inc.UserInst->op_end(),
+ Inc.IVOperand);
+ assert(UseI != Inc.UserInst->op_end() && "cannot find IV operand");
IVIncSet.insert(UseI);
}
}
int64_t IncOffset = IncConst->getValue()->getSExtValue();
if (!isAlwaysFoldable(TTI, LSRUse::Address,
- getAccessType(UserInst), /*BaseGV=*/ 0,
+ getAccessType(UserInst), /*BaseGV=*/ nullptr,
IncOffset, /*HaseBaseReg=*/ false))
return false;
// findIVOperand returns IVOpEnd if it can no longer find a valid IV user.
User::op_iterator IVOpIter = findIVOperand(Head.UserInst->op_begin(),
IVOpEnd, L, SE);
- Value *IVSrc = 0;
+ Value *IVSrc = nullptr;
while (IVOpIter != IVOpEnd) {
IVSrc = getWideOperand(*IVOpIter);
DEBUG(dbgs() << "Generate chain at: " << *IVSrc << "\n");
Type *IVTy = IVSrc->getType();
Type *IntTy = SE.getEffectiveSCEVType(IVTy);
- const SCEV *LeftOverExpr = 0;
- for (IVChain::const_iterator IncI = Chain.begin(),
- IncE = Chain.end(); IncI != IncE; ++IncI) {
-
- Instruction *InsertPt = IncI->UserInst;
+ const SCEV *LeftOverExpr = nullptr;
+ for (const IVInc &Inc : Chain) {
+ Instruction *InsertPt = Inc.UserInst;
if (isa<PHINode>(InsertPt))
InsertPt = L->getLoopLatch()->getTerminator();
// IVOper will replace the current IV User's operand. IVSrc is the IV
// value currently held in a register.
Value *IVOper = IVSrc;
- if (!IncI->IncExpr->isZero()) {
+ if (!Inc.IncExpr->isZero()) {
// IncExpr was the result of subtraction of two narrow values, so must
// be signed.
- const SCEV *IncExpr = SE.getNoopOrSignExtend(IncI->IncExpr, IntTy);
+ const SCEV *IncExpr = SE.getNoopOrSignExtend(Inc.IncExpr, IntTy);
LeftOverExpr = LeftOverExpr ?
SE.getAddExpr(LeftOverExpr, IncExpr) : IncExpr;
}
IVOper = Rewriter.expandCodeFor(IVOperExpr, IVTy, InsertPt);
// If an IV increment can't be folded, use it as the next IV value.
- if (!canFoldIVIncExpr(LeftOverExpr, IncI->UserInst, IncI->IVOperand,
- TTI)) {
+ if (!canFoldIVIncExpr(LeftOverExpr, Inc.UserInst, Inc.IVOperand, TTI)) {
assert(IVTy == IVOper->getType() && "inconsistent IV increment type");
IVSrc = IVOper;
- LeftOverExpr = 0;
+ LeftOverExpr = nullptr;
}
}
- Type *OperTy = IncI->IVOperand->getType();
+ Type *OperTy = Inc.IVOperand->getType();
if (IVTy != OperTy) {
assert(SE.getTypeSizeInBits(IVTy) >= SE.getTypeSizeInBits(OperTy) &&
"cannot extend a chained IV");
IRBuilder<> Builder(InsertPt);
IVOper = Builder.CreateTruncOrBitCast(IVOper, OperTy, "lsr.chain");
}
- IncI->UserInst->replaceUsesOfWith(IncI->IVOperand, IVOper);
- DeadInsts.push_back(IncI->IVOperand);
+ Inc.UserInst->replaceUsesOfWith(Inc.IVOperand, IVOper);
+ DeadInsts.emplace_back(Inc.IVOperand);
}
// If LSR created a new, wider phi, we may also replace its postinc. We only
// do this if we also found a wide value for the head of the chain.
IVOper = Builder.CreatePointerCast(IVSrc, PostIncTy, "lsr.chain");
}
Phi->replaceUsesOfWith(PostIncV, IVOper);
- DeadInsts.push_back(PostIncV);
+ DeadInsts.emplace_back(PostIncV);
}
}
}
void LSRInstance::CollectFixupsAndInitialFormulae() {
- for (IVUsers::const_iterator UI = IU.begin(), E = IU.end(); UI != E; ++UI) {
- Instruction *UserInst = UI->getUser();
+ for (const IVStrideUse &U : IU) {
+ Instruction *UserInst = U.getUser();
// Skip IV users that are part of profitable IV Chains.
User::op_iterator UseI = std::find(UserInst->op_begin(), UserInst->op_end(),
- UI->getOperandValToReplace());
+ U.getOperandValToReplace());
assert(UseI != UserInst->op_end() && "cannot find IV operand");
if (IVIncSet.count(UseI))
continue;
// Record the uses.
LSRFixup &LF = getNewFixup();
LF.UserInst = UserInst;
- LF.OperandValToReplace = UI->getOperandValToReplace();
- LF.PostIncLoops = UI->getPostIncLoops();
+ LF.OperandValToReplace = U.getOperandValToReplace();
+ LF.PostIncLoops = U.getPostIncLoops();
LSRUse::KindType Kind = LSRUse::Basic;
- Type *AccessTy = 0;
+ Type *AccessTy = nullptr;
if (isAddressUse(LF.UserInst, LF.OperandValToReplace)) {
Kind = LSRUse::Address;
AccessTy = getAccessType(LF.UserInst);
}
- const SCEV *S = IU.getExpr(*UI);
+ const SCEV *S = IU.getExpr(U);
// Equality (== and !=) ICmps are special. We can rewrite (i == N) as
// (N - i == 0), and this allows (N - i) to be the expression that we work
if (SE.isLoopInvariant(N, L) && isSafeToExpand(N, SE)) {
// S is normalized, so normalize N before folding it into S
// to keep the result normalized.
- N = TransformForPostIncUse(Normalize, N, CI, 0,
+ N = TransformForPostIncUse(Normalize, N, CI, nullptr,
LF.PostIncLoops, SE, DT);
Kind = LSRUse::ICmpZero;
S = SE.getMinusSCEV(N, S);
void LSRInstance::CountRegisters(const Formula &F, size_t LUIdx) {
if (F.ScaledReg)
RegUses.CountRegister(F.ScaledReg, LUIdx);
- for (SmallVectorImpl<const SCEV *>::const_iterator I = F.BaseRegs.begin(),
- E = F.BaseRegs.end(); I != E; ++I)
- RegUses.CountRegister(*I, LUIdx);
+ for (const SCEV *BaseReg : F.BaseRegs)
+ RegUses.CountRegister(BaseReg, LUIdx);
}
/// InsertFormula - If the given formula has not yet been inserted, add it to
/// the list, and return true. Return false otherwise.
bool LSRInstance::InsertFormula(LSRUse &LU, unsigned LUIdx, const Formula &F) {
+ // Do not insert formula that we will not be able to expand.
+ assert(isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy, F) &&
+ "Formula is illegal");
if (!LU.InsertFormula(F))
return false;
void
LSRInstance::CollectLoopInvariantFixupsAndFormulae() {
SmallVector<const SCEV *, 8> Worklist(RegUses.begin(), RegUses.end());
- SmallPtrSet<const SCEV *, 8> Inserted;
+ SmallPtrSet<const SCEV *, 32> Visited;
while (!Worklist.empty()) {
const SCEV *S = Worklist.pop_back_val();
+ // Don't process the same SCEV twice
+ if (!Visited.insert(S).second)
+ continue;
+
if (const SCEVNAryExpr *N = dyn_cast<SCEVNAryExpr>(S))
Worklist.append(N->op_begin(), N->op_end());
else if (const SCEVCastExpr *C = dyn_cast<SCEVCastExpr>(S))
Worklist.push_back(D->getLHS());
Worklist.push_back(D->getRHS());
} else if (const SCEVUnknown *US = dyn_cast<SCEVUnknown>(S)) {
- if (!Inserted.insert(US)) continue;
const Value *V = US->getValue();
if (const Instruction *Inst = dyn_cast<Instruction>(V)) {
// Look for instructions defined outside the loop.
LSRFixup &LF = getNewFixup();
LF.UserInst = const_cast<Instruction *>(UserInst);
LF.OperandValToReplace = U;
- std::pair<size_t, int64_t> P = getUse(S, LSRUse::Basic, 0);
+ std::pair<size_t, int64_t> P = getUse(S, LSRUse::Basic, nullptr);
LF.LUIdx = P.first;
LF.Offset = P.second;
LSRUse &LU = Uses[LF.LUIdx];
if (const SCEVAddExpr *Add = dyn_cast<SCEVAddExpr>(S)) {
// Break out add operands.
- for (SCEVAddExpr::op_iterator I = Add->op_begin(), E = Add->op_end();
- I != E; ++I) {
- const SCEV *Remainder = CollectSubexprs(*I, C, Ops, L, SE, Depth+1);
+ for (const SCEV *S : Add->operands()) {
+ const SCEV *Remainder = CollectSubexprs(S, C, Ops, L, SE, Depth+1);
if (Remainder)
Ops.push_back(C ? SE.getMulExpr(C, Remainder) : Remainder);
}
- return 0;
+ return nullptr;
} else if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(S)) {
// Split a non-zero base out of an addrec.
if (AR->getStart()->isZero())
// does not pertain to this loop.
if (Remainder && (AR->getLoop() == L || !isa<SCEVAddRecExpr>(Remainder))) {
Ops.push_back(C ? SE.getMulExpr(C, Remainder) : Remainder);
- Remainder = 0;
+ Remainder = nullptr;
}
if (Remainder != AR->getStart()) {
if (!Remainder)
CollectSubexprs(Mul->getOperand(1), C, Ops, L, SE, Depth+1);
if (Remainder)
Ops.push_back(SE.getMulExpr(C, Remainder));
- return 0;
+ return nullptr;
}
}
return S;
}
-/// GenerateReassociations - Split out subexpressions from adds and the bases of
-/// addrecs.
-void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx,
- Formula Base,
- unsigned Depth) {
- // Arbitrarily cap recursion to protect compile time.
- if (Depth >= 3) return;
+/// \brief Helper function for LSRInstance::GenerateReassociations.
+void LSRInstance::GenerateReassociationsImpl(LSRUse &LU, unsigned LUIdx,
+ const Formula &Base,
+ unsigned Depth, size_t Idx,
+ bool IsScaledReg) {
+ const SCEV *BaseReg = IsScaledReg ? Base.ScaledReg : Base.BaseRegs[Idx];
+ SmallVector<const SCEV *, 8> AddOps;
+ const SCEV *Remainder = CollectSubexprs(BaseReg, nullptr, AddOps, L, SE);
+ if (Remainder)
+ AddOps.push_back(Remainder);
+
+ if (AddOps.size() == 1)
+ return;
- for (size_t i = 0, e = Base.BaseRegs.size(); i != e; ++i) {
- const SCEV *BaseReg = Base.BaseRegs[i];
+ for (SmallVectorImpl<const SCEV *>::const_iterator J = AddOps.begin(),
+ JE = AddOps.end();
+ J != JE; ++J) {
- SmallVector<const SCEV *, 8> AddOps;
- const SCEV *Remainder = CollectSubexprs(BaseReg, 0, AddOps, L, SE);
- if (Remainder)
- AddOps.push_back(Remainder);
+ // Loop-variant "unknown" values are uninteresting; we won't be able to
+ // do anything meaningful with them.
+ if (isa<SCEVUnknown>(*J) && !SE.isLoopInvariant(*J, L))
+ continue;
- if (AddOps.size() == 1) continue;
+ // Don't pull a constant into a register if the constant could be folded
+ // into an immediate field.
+ if (isAlwaysFoldable(TTI, SE, LU.MinOffset, LU.MaxOffset, LU.Kind,
+ LU.AccessTy, *J, Base.getNumRegs() > 1))
+ continue;
- for (SmallVectorImpl<const SCEV *>::const_iterator J = AddOps.begin(),
- JE = AddOps.end(); J != JE; ++J) {
+ // Collect all operands except *J.
+ SmallVector<const SCEV *, 8> InnerAddOps(
+ ((const SmallVector<const SCEV *, 8> &)AddOps).begin(), J);
+ InnerAddOps.append(std::next(J),
+ ((const SmallVector<const SCEV *, 8> &)AddOps).end());
+
+ // Don't leave just a constant behind in a register if the constant could
+ // be folded into an immediate field.
+ if (InnerAddOps.size() == 1 &&
+ isAlwaysFoldable(TTI, SE, LU.MinOffset, LU.MaxOffset, LU.Kind,
+ LU.AccessTy, InnerAddOps[0], Base.getNumRegs() > 1))
+ continue;
- // Loop-variant "unknown" values are uninteresting; we won't be able to
- // do anything meaningful with them.
- if (isa<SCEVUnknown>(*J) && !SE.isLoopInvariant(*J, L))
- continue;
+ const SCEV *InnerSum = SE.getAddExpr(InnerAddOps);
+ if (InnerSum->isZero())
+ continue;
+ Formula F = Base;
- // Don't pull a constant into a register if the constant could be folded
- // into an immediate field.
- if (isAlwaysFoldable(TTI, SE, LU.MinOffset, LU.MaxOffset, LU.Kind,
- LU.AccessTy, *J, Base.getNumRegs() > 1))
- continue;
+ // Add the remaining pieces of the add back into the new formula.
+ const SCEVConstant *InnerSumSC = dyn_cast<SCEVConstant>(InnerSum);
+ if (InnerSumSC && SE.getTypeSizeInBits(InnerSumSC->getType()) <= 64 &&
+ TTI.isLegalAddImmediate((uint64_t)F.UnfoldedOffset +
+ InnerSumSC->getValue()->getZExtValue())) {
+ F.UnfoldedOffset =
+ (uint64_t)F.UnfoldedOffset + InnerSumSC->getValue()->getZExtValue();
+ if (IsScaledReg)
+ F.ScaledReg = nullptr;
+ else
+ F.BaseRegs.erase(F.BaseRegs.begin() + Idx);
+ } else if (IsScaledReg)
+ F.ScaledReg = InnerSum;
+ else
+ F.BaseRegs[Idx] = InnerSum;
+
+ // Add J as its own register, or an unfolded immediate.
+ const SCEVConstant *SC = dyn_cast<SCEVConstant>(*J);
+ if (SC && SE.getTypeSizeInBits(SC->getType()) <= 64 &&
+ TTI.isLegalAddImmediate((uint64_t)F.UnfoldedOffset +
+ SC->getValue()->getZExtValue()))
+ F.UnfoldedOffset =
+ (uint64_t)F.UnfoldedOffset + SC->getValue()->getZExtValue();
+ else
+ F.BaseRegs.push_back(*J);
+ // We may have changed the number of register in base regs, adjust the
+ // formula accordingly.
+ F.Canonicalize();
- // Collect all operands except *J.
- SmallVector<const SCEV *, 8> InnerAddOps(
- ((const SmallVector<const SCEV *, 8> &)AddOps).begin(), J);
- InnerAddOps.append(std::next(J),
- ((const SmallVector<const SCEV *, 8> &)AddOps).end());
-
- // Don't leave just a constant behind in a register if the constant could
- // be folded into an immediate field.
- if (InnerAddOps.size() == 1 &&
- isAlwaysFoldable(TTI, SE, LU.MinOffset, LU.MaxOffset, LU.Kind,
- LU.AccessTy, InnerAddOps[0], Base.getNumRegs() > 1))
- continue;
+ if (InsertFormula(LU, LUIdx, F))
+ // If that formula hadn't been seen before, recurse to find more like
+ // it.
+ GenerateReassociations(LU, LUIdx, LU.Formulae.back(), Depth + 1);
+ }
+}
- const SCEV *InnerSum = SE.getAddExpr(InnerAddOps);
- if (InnerSum->isZero())
- continue;
- Formula F = Base;
+/// GenerateReassociations - Split out subexpressions from adds and the bases of
+/// addrecs.
+void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx,
+ Formula Base, unsigned Depth) {
+ assert(Base.isCanonical() && "Input must be in the canonical form");
+ // Arbitrarily cap recursion to protect compile time.
+ if (Depth >= 3)
+ return;
- // Add the remaining pieces of the add back into the new formula.
- const SCEVConstant *InnerSumSC = dyn_cast<SCEVConstant>(InnerSum);
- if (InnerSumSC &&
- SE.getTypeSizeInBits(InnerSumSC->getType()) <= 64 &&
- TTI.isLegalAddImmediate((uint64_t)F.UnfoldedOffset +
- InnerSumSC->getValue()->getZExtValue())) {
- F.UnfoldedOffset = (uint64_t)F.UnfoldedOffset +
- InnerSumSC->getValue()->getZExtValue();
- F.BaseRegs.erase(F.BaseRegs.begin() + i);
- } else
- F.BaseRegs[i] = InnerSum;
-
- // Add J as its own register, or an unfolded immediate.
- const SCEVConstant *SC = dyn_cast<SCEVConstant>(*J);
- if (SC && SE.getTypeSizeInBits(SC->getType()) <= 64 &&
- TTI.isLegalAddImmediate((uint64_t)F.UnfoldedOffset +
- SC->getValue()->getZExtValue()))
- F.UnfoldedOffset = (uint64_t)F.UnfoldedOffset +
- SC->getValue()->getZExtValue();
- else
- F.BaseRegs.push_back(*J);
+ for (size_t i = 0, e = Base.BaseRegs.size(); i != e; ++i)
+ GenerateReassociationsImpl(LU, LUIdx, Base, Depth, i);
- if (InsertFormula(LU, LUIdx, F))
- // If that formula hadn't been seen before, recurse to find more like
- // it.
- GenerateReassociations(LU, LUIdx, LU.Formulae.back(), Depth+1);
- }
- }
+ if (Base.Scale == 1)
+ GenerateReassociationsImpl(LU, LUIdx, Base, Depth,
+ /* Idx */ -1, /* IsScaledReg */ true);
}
/// GenerateCombinations - Generate a formula consisting of all of the
void LSRInstance::GenerateCombinations(LSRUse &LU, unsigned LUIdx,
Formula Base) {
// This method is only interesting on a plurality of registers.
- if (Base.BaseRegs.size() <= 1) return;
+ if (Base.BaseRegs.size() + (Base.Scale == 1) <= 1)
+ return;
+ // Flatten the representation, i.e., reg1 + 1*reg2 => reg1 + reg2, before
+ // processing the formula.
+ Base.Unscale();
Formula F = Base;
F.BaseRegs.clear();
SmallVector<const SCEV *, 4> Ops;
- for (SmallVectorImpl<const SCEV *>::const_iterator
- I = Base.BaseRegs.begin(), E = Base.BaseRegs.end(); I != E; ++I) {
- const SCEV *BaseReg = *I;
+ for (const SCEV *BaseReg : Base.BaseRegs) {
if (SE.properlyDominates(BaseReg, L->getHeader()) &&
!SE.hasComputableLoopEvolution(BaseReg, L))
Ops.push_back(BaseReg);
// rather than proceed with zero in a register.
if (!Sum->isZero()) {
F.BaseRegs.push_back(Sum);
+ F.Canonicalize();
(void)InsertFormula(LU, LUIdx, F);
}
}
}
+/// \brief Helper function for LSRInstance::GenerateSymbolicOffsets.
+void LSRInstance::GenerateSymbolicOffsetsImpl(LSRUse &LU, unsigned LUIdx,
+ const Formula &Base, size_t Idx,
+ bool IsScaledReg) {
+ const SCEV *G = IsScaledReg ? Base.ScaledReg : Base.BaseRegs[Idx];
+ GlobalValue *GV = ExtractSymbol(G, SE);
+ if (G->isZero() || !GV)
+ return;
+ Formula F = Base;
+ F.BaseGV = GV;
+ if (!isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy, F))
+ return;
+ if (IsScaledReg)
+ F.ScaledReg = G;
+ else
+ F.BaseRegs[Idx] = G;
+ (void)InsertFormula(LU, LUIdx, F);
+}
+
/// GenerateSymbolicOffsets - Generate reuse formulae using symbolic offsets.
void LSRInstance::GenerateSymbolicOffsets(LSRUse &LU, unsigned LUIdx,
Formula Base) {
// We can't add a symbolic offset if the address already contains one.
if (Base.BaseGV) return;
- for (size_t i = 0, e = Base.BaseRegs.size(); i != e; ++i) {
- const SCEV *G = Base.BaseRegs[i];
- GlobalValue *GV = ExtractSymbol(G, SE);
- if (G->isZero() || !GV)
- continue;
+ for (size_t i = 0, e = Base.BaseRegs.size(); i != e; ++i)
+ GenerateSymbolicOffsetsImpl(LU, LUIdx, Base, i);
+ if (Base.Scale == 1)
+ GenerateSymbolicOffsetsImpl(LU, LUIdx, Base, /* Idx */ -1,
+ /* IsScaledReg */ true);
+}
+
+/// \brief Helper function for LSRInstance::GenerateConstantOffsets.
+void LSRInstance::GenerateConstantOffsetsImpl(
+ LSRUse &LU, unsigned LUIdx, const Formula &Base,
+ const SmallVectorImpl<int64_t> &Worklist, size_t Idx, bool IsScaledReg) {
+ const SCEV *G = IsScaledReg ? Base.ScaledReg : Base.BaseRegs[Idx];
+ for (int64_t Offset : Worklist) {
Formula F = Base;
- F.BaseGV = GV;
- if (!isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy, F))
- continue;
- F.BaseRegs[i] = G;
- (void)InsertFormula(LU, LUIdx, F);
+ F.BaseOffset = (uint64_t)Base.BaseOffset - Offset;
+ if (isLegalUse(TTI, LU.MinOffset - Offset, LU.MaxOffset - Offset, LU.Kind,
+ LU.AccessTy, F)) {
+ // Add the offset to the base register.
+ const SCEV *NewG = SE.getAddExpr(SE.getConstant(G->getType(), Offset), G);
+ // If it cancelled out, drop the base register, otherwise update it.
+ if (NewG->isZero()) {
+ if (IsScaledReg) {
+ F.Scale = 0;
+ F.ScaledReg = nullptr;
+ } else
+ F.DeleteBaseReg(F.BaseRegs[Idx]);
+ F.Canonicalize();
+ } else if (IsScaledReg)
+ F.ScaledReg = NewG;
+ else
+ F.BaseRegs[Idx] = NewG;
+
+ (void)InsertFormula(LU, LUIdx, F);
+ }
}
+
+ int64_t Imm = ExtractImmediate(G, SE);
+ if (G->isZero() || Imm == 0)
+ return;
+ Formula F = Base;
+ F.BaseOffset = (uint64_t)F.BaseOffset + Imm;
+ if (!isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy, F))
+ return;
+ if (IsScaledReg)
+ F.ScaledReg = G;
+ else
+ F.BaseRegs[Idx] = G;
+ (void)InsertFormula(LU, LUIdx, F);
}
/// GenerateConstantOffsets - Generate reuse formulae using symbolic offsets.
if (LU.MaxOffset != LU.MinOffset)
Worklist.push_back(LU.MaxOffset);
- for (size_t i = 0, e = Base.BaseRegs.size(); i != e; ++i) {
- const SCEV *G = Base.BaseRegs[i];
-
- for (SmallVectorImpl<int64_t>::const_iterator I = Worklist.begin(),
- E = Worklist.end(); I != E; ++I) {
- Formula F = Base;
- F.BaseOffset = (uint64_t)Base.BaseOffset - *I;
- if (isLegalUse(TTI, LU.MinOffset - *I, LU.MaxOffset - *I, LU.Kind,
- LU.AccessTy, F)) {
- // Add the offset to the base register.
- const SCEV *NewG = SE.getAddExpr(SE.getConstant(G->getType(), *I), G);
- // If it cancelled out, drop the base register, otherwise update it.
- if (NewG->isZero()) {
- std::swap(F.BaseRegs[i], F.BaseRegs.back());
- F.BaseRegs.pop_back();
- } else
- F.BaseRegs[i] = NewG;
-
- (void)InsertFormula(LU, LUIdx, F);
- }
- }
-
- int64_t Imm = ExtractImmediate(G, SE);
- if (G->isZero() || Imm == 0)
- continue;
- Formula F = Base;
- F.BaseOffset = (uint64_t)F.BaseOffset + Imm;
- if (!isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy, F))
- continue;
- F.BaseRegs[i] = G;
- (void)InsertFormula(LU, LUIdx, F);
- }
+ for (size_t i = 0, e = Base.BaseRegs.size(); i != e; ++i)
+ GenerateConstantOffsetsImpl(LU, LUIdx, Base, Worklist, i);
+ if (Base.Scale == 1)
+ GenerateConstantOffsetsImpl(LU, LUIdx, Base, Worklist, /* Idx */ -1,
+ /* IsScaledReg */ true);
}
/// GenerateICmpZeroScales - For ICmpZero, check to see if we can scale up
assert(!Base.BaseGV && "ICmpZero use is not legal!");
// Check each interesting stride.
- for (SmallSetVector<int64_t, 8>::const_iterator
- I = Factors.begin(), E = Factors.end(); I != E; ++I) {
- int64_t Factor = *I;
-
+ for (int64_t Factor : Factors) {
// Check that the multiplication doesn't overflow.
if (Base.BaseOffset == INT64_MIN && Factor == -1)
continue;
if (!IntTy) return;
// If this Formula already has a scaled register, we can't add another one.
- if (Base.Scale != 0) return;
+ // Try to unscale the formula to generate a better scale.
+ if (Base.Scale != 0 && !Base.Unscale())
+ return;
- // Check each interesting stride.
- for (SmallSetVector<int64_t, 8>::const_iterator
- I = Factors.begin(), E = Factors.end(); I != E; ++I) {
- int64_t Factor = *I;
+ assert(Base.Scale == 0 && "Unscale did not did its job!");
+ // Check each interesting stride.
+ for (int64_t Factor : Factors) {
Base.Scale = Factor;
Base.HasBaseReg = Base.BaseRegs.size() > 1;
// Check whether this scale is going to be legal.
Formula F = Base;
F.ScaledReg = Quotient;
F.DeleteBaseReg(F.BaseRegs[i]);
+ // The canonical representation of 1*reg is reg, which is already in
+ // Base. In that case, do not try to insert the formula, it will be
+ // rejected anyway.
+ if (F.Scale == 1 && F.BaseRegs.empty())
+ continue;
(void)InsertFormula(LU, LUIdx, F);
}
}
if (!DstTy) return;
DstTy = SE.getEffectiveSCEVType(DstTy);
- for (SmallSetVector<Type *, 4>::const_iterator
- I = Types.begin(), E = Types.end(); I != E; ++I) {
- Type *SrcTy = *I;
+ for (Type *SrcTy : Types) {
if (SrcTy != DstTy && TTI.isTruncateFree(SrcTy, DstTy)) {
Formula F = Base;
- if (F.ScaledReg) F.ScaledReg = SE.getAnyExtendExpr(F.ScaledReg, *I);
- for (SmallVectorImpl<const SCEV *>::iterator J = F.BaseRegs.begin(),
- JE = F.BaseRegs.end(); J != JE; ++J)
- *J = SE.getAnyExtendExpr(*J, SrcTy);
+ if (F.ScaledReg) F.ScaledReg = SE.getAnyExtendExpr(F.ScaledReg, SrcTy);
+ for (const SCEV *&BaseReg : F.BaseRegs)
+ BaseReg = SE.getAnyExtendExpr(BaseReg, SrcTy);
// TODO: This assumes we've done basic processing on all uses and
// have an idea what the register usage is.
void LSRInstance::GenerateCrossUseConstantOffsets() {
// Group the registers by their value without any added constant offset.
typedef std::map<int64_t, const SCEV *> ImmMapTy;
- typedef DenseMap<const SCEV *, ImmMapTy> RegMapTy;
- RegMapTy Map;
+ DenseMap<const SCEV *, ImmMapTy> Map;
DenseMap<const SCEV *, SmallBitVector> UsedByIndicesMap;
SmallVector<const SCEV *, 8> Sequence;
- for (RegUseTracker::const_iterator I = RegUses.begin(), E = RegUses.end();
- I != E; ++I) {
- const SCEV *Reg = *I;
+ for (const SCEV *Use : RegUses) {
+ const SCEV *Reg = Use; // Make a copy for ExtractImmediate to modify.
int64_t Imm = ExtractImmediate(Reg, SE);
- std::pair<RegMapTy::iterator, bool> Pair =
- Map.insert(std::make_pair(Reg, ImmMapTy()));
+ auto Pair = Map.insert(std::make_pair(Reg, ImmMapTy()));
if (Pair.second)
Sequence.push_back(Reg);
- Pair.first->second.insert(std::make_pair(Imm, *I));
- UsedByIndicesMap[Reg] |= RegUses.getUsedByIndices(*I);
+ Pair.first->second.insert(std::make_pair(Imm, Use));
+ UsedByIndicesMap[Reg] |= RegUses.getUsedByIndices(Use);
}
// Now examine each set of registers with the same base value. Build up
// not adding formulae and register counts while we're searching.
SmallVector<WorkItem, 32> WorkItems;
SmallSet<std::pair<size_t, int64_t>, 32> UniqueItems;
- for (SmallVectorImpl<const SCEV *>::const_iterator I = Sequence.begin(),
- E = Sequence.end(); I != E; ++I) {
- const SCEV *Reg = *I;
+ for (const SCEV *Reg : Sequence) {
const ImmMapTy &Imms = Map.find(Reg)->second;
// It's not worthwhile looking for reuse if there's only one offset.
continue;
DEBUG(dbgs() << "Generating cross-use offsets for " << *Reg << ':';
- for (ImmMapTy::const_iterator J = Imms.begin(), JE = Imms.end();
- J != JE; ++J)
- dbgs() << ' ' << J->first;
+ for (const auto &Entry : Imms)
+ dbgs() << ' ' << Entry.first;
dbgs() << '\n');
// Examine each offset.
for (int LUIdx = UsedByIndices.find_first(); LUIdx != -1;
LUIdx = UsedByIndices.find_next(LUIdx))
// Make a memo of this use, offset, and register tuple.
- if (UniqueItems.insert(std::make_pair(LUIdx, Imm)))
+ if (UniqueItems.insert(std::make_pair(LUIdx, Imm)).second)
WorkItems.push_back(WorkItem(LUIdx, Imm, OrigReg));
}
}
UniqueItems.clear();
// Now iterate through the worklist and add new formulae.
- for (SmallVectorImpl<WorkItem>::const_iterator I = WorkItems.begin(),
- E = WorkItems.end(); I != E; ++I) {
- const WorkItem &WI = *I;
+ for (const WorkItem &WI : WorkItems) {
size_t LUIdx = WI.LUIdx;
LSRUse &LU = Uses[LUIdx];
int64_t Imm = WI.Imm;
// TODO: Use a more targeted data structure.
for (size_t L = 0, LE = LU.Formulae.size(); L != LE; ++L) {
- const Formula &F = LU.Formulae[L];
+ Formula F = LU.Formulae[L];
+ // FIXME: The code for the scaled and unscaled registers looks
+ // very similar but slightly different. Investigate if they
+ // could be merged. That way, we would not have to unscale the
+ // Formula.
+ F.Unscale();
// Use the immediate in the scaled register.
if (F.ScaledReg == OrigReg) {
int64_t Offset = (uint64_t)F.BaseOffset + Imm * (uint64_t)F.Scale;
if (C->getValue()->isNegative() !=
(NewF.BaseOffset < 0) &&
(C->getValue()->getValue().abs() * APInt(BitWidth, F.Scale))
- .ule(abs64(NewF.BaseOffset)))
+ .ule(std::abs(NewF.BaseOffset)))
continue;
// OK, looks good.
+ NewF.Canonicalize();
(void)InsertFormula(LU, LUIdx, NewF);
} else {
// Use the immediate in a base register.
// If the new formula has a constant in a register, and adding the
// constant value to the immediate would produce a value closer to
// zero than the immediate itself, then the formula isn't worthwhile.
- for (SmallVectorImpl<const SCEV *>::const_iterator
- J = NewF.BaseRegs.begin(), JE = NewF.BaseRegs.end();
- J != JE; ++J)
- if (const SCEVConstant *C = dyn_cast<SCEVConstant>(*J))
+ for (const SCEV *NewReg : NewF.BaseRegs)
+ if (const SCEVConstant *C = dyn_cast<SCEVConstant>(NewReg))
if ((C->getValue()->getValue() + NewF.BaseOffset).abs().slt(
- abs64(NewF.BaseOffset)) &&
+ std::abs(NewF.BaseOffset)) &&
(C->getValue()->getValue() +
NewF.BaseOffset).countTrailingZeros() >=
countTrailingZeros<uint64_t>(NewF.BaseOffset))
goto skip_formula;
// Ok, looks good.
+ NewF.Canonicalize();
(void)InsertFormula(LU, LUIdx, NewF);
break;
skip_formula:;
}
else {
SmallVector<const SCEV *, 4> Key;
- for (SmallVectorImpl<const SCEV *>::const_iterator J = F.BaseRegs.begin(),
- JE = F.BaseRegs.end(); J != JE; ++J) {
- const SCEV *Reg = *J;
+ for (const SCEV *Reg : F.BaseRegs) {
if (RegUses.isRegUsedByUsesOtherThan(Reg, LUIdx))
Key.push_back(Reg);
}
/// isn't always sufficient.
size_t LSRInstance::EstimateSearchSpaceComplexity() const {
size_t Power = 1;
- for (SmallVectorImpl<LSRUse>::const_iterator I = Uses.begin(),
- E = Uses.end(); I != E; ++I) {
- size_t FSize = I->Formulae.size();
+ for (const LSRUse &LU : Uses) {
+ size_t FSize = LU.Formulae.size();
if (FSize >= ComplexityLimit) {
Power = ComplexityLimit;
break;
for (size_t LUIdx = 0, NumUses = Uses.size(); LUIdx != NumUses; ++LUIdx) {
LSRUse &LU = Uses[LUIdx];
- for (SmallVectorImpl<Formula>::const_iterator I = LU.Formulae.begin(),
- E = LU.Formulae.end(); I != E; ++I) {
- const Formula &F = *I;
- if (F.BaseOffset == 0 || F.Scale != 0)
+ for (const Formula &F : LU.Formulae) {
+ if (F.BaseOffset == 0 || (F.Scale != 0 && F.Scale != 1))
continue;
LSRUse *LUThatHas = FindUseWithSimilarFormula(F, LU);
LUThatHas->AllFixupsOutsideLoop &= LU.AllFixupsOutsideLoop;
// Update the relocs to reference the new use.
- for (SmallVectorImpl<LSRFixup>::iterator I = Fixups.begin(),
- E = Fixups.end(); I != E; ++I) {
- LSRFixup &Fixup = *I;
+ for (LSRFixup &Fixup : Fixups) {
if (Fixup.LUIdx == LUIdx) {
Fixup.LUIdx = LUThatHas - &Uses.front();
Fixup.Offset += F.BaseOffset;
// Pick the register which is used by the most LSRUses, which is likely
// to be a good reuse register candidate.
- const SCEV *Best = 0;
+ const SCEV *Best = nullptr;
unsigned BestNum = 0;
- for (RegUseTracker::const_iterator I = RegUses.begin(), E = RegUses.end();
- I != E; ++I) {
- const SCEV *Reg = *I;
+ for (const SCEV *Reg : RegUses) {
if (Taken.count(Reg))
continue;
if (!Best)
// reference that register in order to be considered. This prunes out
// unprofitable searching.
SmallSetVector<const SCEV *, 4> ReqRegs;
- for (SmallPtrSet<const SCEV *, 16>::const_iterator I = CurRegs.begin(),
- E = CurRegs.end(); I != E; ++I)
- if (LU.Regs.count(*I))
- ReqRegs.insert(*I);
+ for (const SCEV *S : CurRegs)
+ if (LU.Regs.count(S))
+ ReqRegs.insert(S);
SmallPtrSet<const SCEV *, 16> NewRegs;
Cost NewCost;
- for (SmallVectorImpl<Formula>::const_iterator I = LU.Formulae.begin(),
- E = LU.Formulae.end(); I != E; ++I) {
- const Formula &F = *I;
-
- // Ignore formulae which do not use any of the required registers.
- bool SatisfiedReqReg = true;
- for (SmallSetVector<const SCEV *, 4>::const_iterator J = ReqRegs.begin(),
- JE = ReqRegs.end(); J != JE; ++J) {
- const SCEV *Reg = *J;
- if ((!F.ScaledReg || F.ScaledReg != Reg) &&
- std::find(F.BaseRegs.begin(), F.BaseRegs.end(), Reg) ==
+ for (const Formula &F : LU.Formulae) {
+ // Ignore formulae which may not be ideal in terms of register reuse of
+ // ReqRegs. The formula should use all required registers before
+ // introducing new ones.
+ int NumReqRegsToFind = std::min(F.getNumRegs(), ReqRegs.size());
+ for (const SCEV *Reg : ReqRegs) {
+ if ((F.ScaledReg && F.ScaledReg == Reg) ||
+ std::find(F.BaseRegs.begin(), F.BaseRegs.end(), Reg) !=
F.BaseRegs.end()) {
- SatisfiedReqReg = false;
- break;
+ --NumReqRegsToFind;
+ if (NumReqRegsToFind == 0)
+ break;
}
}
- if (!SatisfiedReqReg) {
+ if (NumReqRegsToFind != 0) {
// If none of the formulae satisfied the required registers, then we could
// clear ReqRegs and try again. Currently, we simply give up in this case.
continue;
} else {
DEBUG(dbgs() << "New best at "; NewCost.print(dbgs());
dbgs() << ".\n Regs:";
- for (SmallPtrSet<const SCEV *, 16>::const_iterator
- I = NewRegs.begin(), E = NewRegs.end(); I != E; ++I)
- dbgs() << ' ' << **I;
+ for (const SCEV *S : NewRegs)
+ dbgs() << ' ' << *S;
dbgs() << '\n');
SolutionCost = NewCost;
}
bool AllDominate = true;
- Instruction *BetterPos = 0;
+ Instruction *BetterPos = nullptr;
Instruction *Tentative = IDom->getTerminator();
- for (SmallVectorImpl<Instruction *>::const_iterator I = Inputs.begin(),
- E = Inputs.end(); I != E; ++I) {
- Instruction *Inst = *I;
+ for (Instruction *Inst : Inputs) {
if (Inst == Tentative || !DT.dominates(Inst, Tentative)) {
AllDominate = false;
break;
}
// The expansion must also be dominated by the increment positions of any
// loops it for which it is using post-inc mode.
- for (PostIncLoopSet::const_iterator I = LF.PostIncLoops.begin(),
- E = LF.PostIncLoops.end(); I != E; ++I) {
- const Loop *PIL = *I;
+ for (const Loop *PIL : LF.PostIncLoops) {
if (PIL == L) continue;
// Be dominated by the loop exit.
SmallVector<const SCEV *, 8> Ops;
// Expand the BaseRegs portion.
- for (SmallVectorImpl<const SCEV *>::const_iterator I = F.BaseRegs.begin(),
- E = F.BaseRegs.end(); I != E; ++I) {
- const SCEV *Reg = *I;
+ for (const SCEV *Reg : F.BaseRegs) {
assert(!Reg->isZero() && "Zero allocated in a base register!");
// If we're expanding for a post-inc user, make the post-inc adjustment.
LF.UserInst, LF.OperandValToReplace,
Loops, SE, DT);
- Ops.push_back(SE.getUnknown(Rewriter.expandCodeFor(Reg, 0, IP)));
+ Ops.push_back(SE.getUnknown(Rewriter.expandCodeFor(Reg, nullptr, IP)));
}
// Expand the ScaledReg portion.
- Value *ICmpScaledV = 0;
+ Value *ICmpScaledV = nullptr;
if (F.Scale != 0) {
const SCEV *ScaledS = F.ScaledReg;
Loops, SE, DT);
if (LU.Kind == LSRUse::ICmpZero) {
- // An interesting way of "folding" with an icmp is to use a negated
- // scale, which we'll implement by inserting it into the other operand
- // of the icmp.
- assert(F.Scale == -1 &&
- "The only scale supported by ICmpZero uses is -1!");
- ICmpScaledV = Rewriter.expandCodeFor(ScaledS, 0, IP);
+ // Expand ScaleReg as if it was part of the base regs.
+ if (F.Scale == 1)
+ Ops.push_back(
+ SE.getUnknown(Rewriter.expandCodeFor(ScaledS, nullptr, IP)));
+ else {
+ // An interesting way of "folding" with an icmp is to use a negated
+ // scale, which we'll implement by inserting it into the other operand
+ // of the icmp.
+ assert(F.Scale == -1 &&
+ "The only scale supported by ICmpZero uses is -1!");
+ ICmpScaledV = Rewriter.expandCodeFor(ScaledS, nullptr, IP);
+ }
} else {
// Otherwise just expand the scaled register and an explicit scale,
// which is expected to be matched as part of the address.
// Flush the operand list to suppress SCEVExpander hoisting address modes.
- if (!Ops.empty() && LU.Kind == LSRUse::Address) {
+ // Unless the addressing mode will not be folded.
+ if (!Ops.empty() && LU.Kind == LSRUse::Address &&
+ isAMCompletelyFolded(TTI, LU, F)) {
Value *FullV = Rewriter.expandCodeFor(SE.getAddExpr(Ops), Ty, IP);
Ops.clear();
Ops.push_back(SE.getUnknown(FullV));
}
- ScaledS = SE.getUnknown(Rewriter.expandCodeFor(ScaledS, 0, IP));
- ScaledS = SE.getMulExpr(ScaledS,
- SE.getConstant(ScaledS->getType(), F.Scale));
+ ScaledS = SE.getUnknown(Rewriter.expandCodeFor(ScaledS, nullptr, IP));
+ if (F.Scale != 1)
+ ScaledS =
+ SE.getMulExpr(ScaledS, SE.getConstant(ScaledS->getType(), F.Scale));
Ops.push_back(ScaledS);
}
}
// form, update the ICmp's other operand.
if (LU.Kind == LSRUse::ICmpZero) {
ICmpInst *CI = cast<ICmpInst>(LF.UserInst);
- DeadInsts.push_back(CI->getOperand(1));
+ DeadInsts.emplace_back(CI->getOperand(1));
assert(!F.BaseGV && "ICmp does not support folding a global value and "
"a scale at the same time!");
if (F.Scale == -1) {
}
CI->setOperand(1, ICmpScaledV);
} else {
- assert(F.Scale == 0 &&
+ // A scale of 1 means that the scale has been expanded as part of the
+ // base regs.
+ assert((F.Scale == 0 || F.Scale == 1) &&
"ICmp does not support folding a global value and "
"a scale at the same time!");
Constant *C = ConstantInt::getSigned(SE.getEffectiveSCEVType(OpTy),
Loop *PNLoop = LI.getLoopFor(Parent);
if (!PNLoop || Parent != PNLoop->getHeader()) {
// Split the critical edge.
- BasicBlock *NewBB = 0;
+ BasicBlock *NewBB = nullptr;
if (!Parent->isLandingPad()) {
- NewBB = SplitCriticalEdge(BB, Parent, P,
- /*MergeIdenticalEdges=*/true,
- /*DontDeleteUselessPhis=*/true);
+ NewBB = SplitCriticalEdge(BB, Parent,
+ CriticalEdgeSplittingOptions(&DT, &LI)
+ .setMergeIdenticalEdges()
+ .setDontDeleteUselessPHIs());
} else {
SmallVector<BasicBlock*, 2> NewBBs;
- SplitLandingPadPredecessors(Parent, BB, "", "", P, NewBBs);
+ SplitLandingPadPredecessors(Parent, BB, "", "", NewBBs, &DT, &LI);
NewBB = NewBBs[0];
}
// If NewBB==NULL, then SplitCriticalEdge refused to split because all
}
std::pair<DenseMap<BasicBlock *, Value *>::iterator, bool> Pair =
- Inserted.insert(std::make_pair(BB, static_cast<Value *>(0)));
+ Inserted.insert(std::make_pair(BB, static_cast<Value *>(nullptr)));
if (!Pair.second)
PN->setIncomingValue(i, Pair.first->second);
else {
LF.UserInst->replaceUsesOfWith(LF.OperandValToReplace, FullV);
}
- DeadInsts.push_back(LF.OperandValToReplace);
+ DeadInsts.emplace_back(LF.OperandValToReplace);
}
/// ImplementSolution - Rewrite all the fixup locations with new values,
// 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
Rewriter.setIVIncInsertPos(L, IVIncInsertPos);
// Mark phi nodes that terminate chains so the expander tries to reuse them.
- for (SmallVectorImpl<IVChain>::const_iterator ChainI = IVChainVec.begin(),
- ChainE = IVChainVec.end(); ChainI != ChainE; ++ChainI) {
- if (PHINode *PN = dyn_cast<PHINode>(ChainI->tailUserInst()))
+ for (const IVChain &Chain : IVChainVec) {
+ if (PHINode *PN = dyn_cast<PHINode>(Chain.tailUserInst()))
Rewriter.setChainedPhi(PN);
}
// Expand the new value definitions and update the users.
- for (SmallVectorImpl<LSRFixup>::const_iterator I = Fixups.begin(),
- E = Fixups.end(); I != E; ++I) {
- const LSRFixup &Fixup = *I;
-
+ for (const LSRFixup &Fixup : Fixups) {
Rewrite(Fixup, *Solution[Fixup.LUIdx], Rewriter, DeadInsts, P);
Changed = true;
}
- for (SmallVectorImpl<IVChain>::const_iterator ChainI = IVChainVec.begin(),
- ChainE = IVChainVec.end(); ChainI != ChainE; ++ChainI) {
- GenerateIVChain(*ChainI, Rewriter, DeadInsts);
+ for (const IVChain &Chain : IVChainVec) {
+ GenerateIVChain(Chain, Rewriter, DeadInsts);
Changed = true;
}
// Clean up after ourselves. This must be done before deleting any
LSRInstance::LSRInstance(Loop *L, Pass *P)
: IU(P->getAnalysis<IVUsers>()), SE(P->getAnalysis<ScalarEvolution>()),
DT(P->getAnalysis<DominatorTreeWrapperPass>().getDomTree()),
- LI(P->getAnalysis<LoopInfo>()),
- TTI(P->getAnalysis<TargetTransformInfo>()), L(L), Changed(false),
- IVIncInsertPos(0) {
+ LI(P->getAnalysis<LoopInfoWrapperPass>().getLoopInfo()),
+ TTI(P->getAnalysis<TargetTransformInfoWrapperPass>().getTTI(
+ *L->getHeader()->getParent())),
+ L(L), Changed(false), IVIncInsertPos(nullptr) {
// If LoopSimplify form is not available, stay out of trouble.
if (!L->isLoopSimplifyForm())
return;
// If there's too much analysis to be done, bail early. We won't be able to
// model the problem anyway.
unsigned NumUsers = 0;
- for (IVUsers::const_iterator UI = IU.begin(), E = IU.end(); UI != E; ++UI) {
+ for (const IVStrideUse &U : IU) {
if (++NumUsers > MaxIVUsers) {
- DEBUG(dbgs() << "LSR skipping loop, too many IV Users in " << *L
- << "\n");
+ (void)U;
+ DEBUG(dbgs() << "LSR skipping loop, too many IV Users in " << U << "\n");
return;
}
}
#ifndef NDEBUG
// Formulae should be legal.
- for (SmallVectorImpl<LSRUse>::const_iterator I = Uses.begin(), E = Uses.end();
- I != E; ++I) {
- const LSRUse &LU = *I;
- for (SmallVectorImpl<Formula>::const_iterator J = LU.Formulae.begin(),
- JE = LU.Formulae.end();
- J != JE; ++J)
+ for (const LSRUse &LU : Uses) {
+ for (const Formula &F : LU.Formulae)
assert(isLegalUse(TTI, LU.MinOffset, LU.MaxOffset, LU.Kind, LU.AccessTy,
- *J) && "Illegal formula generated!");
+ F) && "Illegal formula generated!");
};
#endif
OS << "LSR has identified the following interesting factors and types: ";
bool First = true;
- for (SmallSetVector<int64_t, 8>::const_iterator
- I = Factors.begin(), E = Factors.end(); I != E; ++I) {
+ for (int64_t Factor : Factors) {
if (!First) OS << ", ";
First = false;
- OS << '*' << *I;
+ OS << '*' << Factor;
}
- for (SmallSetVector<Type *, 4>::const_iterator
- I = Types.begin(), E = Types.end(); I != E; ++I) {
+ for (Type *Ty : Types) {
if (!First) OS << ", ";
First = false;
- OS << '(' << **I << ')';
+ OS << '(' << *Ty << ')';
}
OS << '\n';
}
void LSRInstance::print_fixups(raw_ostream &OS) const {
OS << "LSR is examining the following fixup sites:\n";
- for (SmallVectorImpl<LSRFixup>::const_iterator I = Fixups.begin(),
- E = Fixups.end(); I != E; ++I) {
+ for (const LSRFixup &LF : Fixups) {
dbgs() << " ";
- I->print(OS);
+ LF.print(OS);
OS << '\n';
}
}
void LSRInstance::print_uses(raw_ostream &OS) const {
OS << "LSR is examining the following uses:\n";
- for (SmallVectorImpl<LSRUse>::const_iterator I = Uses.begin(),
- E = Uses.end(); I != E; ++I) {
- const LSRUse &LU = *I;
+ for (const LSRUse &LU : Uses) {
dbgs() << " ";
LU.print(OS);
OS << '\n';
- for (SmallVectorImpl<Formula>::const_iterator J = LU.Formulae.begin(),
- JE = LU.Formulae.end(); J != JE; ++J) {
+ for (const Formula &F : LU.Formulae) {
OS << " ";
- J->print(OS);
+ F.print(OS);
OS << '\n';
}
}
char LoopStrengthReduce::ID = 0;
INITIALIZE_PASS_BEGIN(LoopStrengthReduce, "loop-reduce",
"Loop Strength Reduction", false, false)
-INITIALIZE_AG_DEPENDENCY(TargetTransformInfo)
+INITIALIZE_PASS_DEPENDENCY(TargetTransformInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
INITIALIZE_PASS_DEPENDENCY(ScalarEvolution)
INITIALIZE_PASS_DEPENDENCY(IVUsers)
-INITIALIZE_PASS_DEPENDENCY(LoopInfo)
+INITIALIZE_PASS_DEPENDENCY(LoopInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(LoopSimplify)
INITIALIZE_PASS_END(LoopStrengthReduce, "loop-reduce",
"Loop Strength Reduction", false, false)
// many analyses if they are around.
AU.addPreservedID(LoopSimplifyID);
- AU.addRequired<LoopInfo>();
- AU.addPreserved<LoopInfo>();
+ AU.addRequired<LoopInfoWrapperPass>();
+ AU.addPreserved<LoopInfoWrapperPass>();
AU.addRequiredID(LoopSimplifyID);
AU.addRequired<DominatorTreeWrapperPass>();
AU.addPreserved<DominatorTreeWrapperPass>();
AU.addRequiredID(LoopSimplifyID);
AU.addRequired<IVUsers>();
AU.addPreserved<IVUsers>();
- AU.addRequired<TargetTransformInfo>();
+ AU.addRequired<TargetTransformInfoWrapperPass>();
}
bool LoopStrengthReduce::runOnLoop(Loop *L, LPPassManager & /*LPM*/) {
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
unsigned numFolded = Rewriter.replaceCongruentIVs(
L, &getAnalysis<DominatorTreeWrapperPass>().getDomTree(), DeadInsts,
- &getAnalysis<TargetTransformInfo>());
+ &getAnalysis<TargetTransformInfoWrapperPass>().getTTI(
+ *L->getHeader()->getParent()));
if (numFolded) {
Changed = true;
DeleteTriviallyDeadInstructions(DeadInsts);
projects / oota-llvm.git / blobdiff