//
//===----------------------------------------------------------------------===//
-#define DEBUG_TYPE "loop-idiom"
#include "llvm/Transforms/Scalar.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/LoopPass.h"
#include "llvm/Analysis/ScalarEvolutionExpander.h"
#include "llvm/Analysis/ScalarEvolutionExpressions.h"
+#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Analysis/ValueTracking.h"
-#include "llvm/DataLayout.h"
-#include "llvm/IRBuilder.h"
-#include "llvm/IntrinsicInst.h"
-#include "llvm/Module.h"
+#include "llvm/IR/DataLayout.h"
+#include "llvm/IR/Dominators.h"
+#include "llvm/IR/IRBuilder.h"
+#include "llvm/IR/IntrinsicInst.h"
+#include "llvm/IR/Module.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetLibraryInfo.h"
-#include "llvm/TargetTransformInfo.h"
#include "llvm/Transforms/Utils/Local.h"
using namespace llvm;
+#define DEBUG_TYPE "loop-idiom"
+
STATISTIC(NumMemSet, "Number of memset's formed from loop stores");
STATISTIC(NumMemCpy, "Number of memcpy's formed from loop load+stores");
return dyn_cast<BranchInst>(BB->getTerminator());
}
- /// Return the condition of the branch terminating the given basic block.
- static Value *getBrCondtion(BasicBlock *);
-
- /// Derive the precondition block (i.e the block that guards the loop
+ /// Derive the precondition block (i.e the block that guards the loop
/// preheader) from the given preheader.
static BasicBlock *getPrecondBb(BasicBlock *PreHead);
};
bool preliminaryScreen();
/// Check if the given conditional branch is based on the comparison
- /// beween a variable and zero, and if the variable is non-zero, the
- /// control yeilds to the loop entry. If the branch matches the behavior,
+ /// between a variable and zero, and if the variable is non-zero, the
+ /// control yields to the loop entry. If the branch matches the behavior,
/// the variable involved in the comparion is returned. This function will
- /// be called to see if the precondition and postcondition of the loop
+ /// be called to see if the precondition and postcondition of the loop
/// are in desirable form.
- Value *matchCondition (BranchInst *Br, BasicBlock *NonZeroTarget) const;
+ Value *matchCondition(BranchInst *Br, BasicBlock *NonZeroTarget) const;
/// Return true iff the idiom is detected in the loop. and 1) \p CntInst
- /// is set to the instruction counting the pupulation bit. 2) \p CntPhi
+ /// is set to the instruction counting the population bit. 2) \p CntPhi
/// is set to the corresponding phi node. 3) \p Var is set to the value
/// whose population bits are being counted.
bool detectIdiom
(Instruction *&CntInst, PHINode *&CntPhi, Value *&Var) const;
/// Insert ctpop intrinsic function and some obviously dead instructions.
- void transform (Instruction *CntInst, PHINode *CntPhi, Value *Var);
+ void transform(Instruction *CntInst, PHINode *CntPhi, Value *Var);
/// Create llvm.ctpop.* intrinsic function.
CallInst *createPopcntIntrinsic(IRBuilderTy &IRB, Value *Val, DebugLoc DL);
class LoopIdiomRecognize : public LoopPass {
Loop *CurLoop;
- const DataLayout *TD;
+ const DataLayout *DL;
DominatorTree *DT;
ScalarEvolution *SE;
TargetLibraryInfo *TLI;
- const ScalarTargetTransformInfo *STTI;
+ const TargetTransformInfo *TTI;
public:
static char ID;
explicit LoopIdiomRecognize() : LoopPass(ID) {
initializeLoopIdiomRecognizePass(*PassRegistry::getPassRegistry());
- TD = 0; DT = 0; SE = 0; TLI = 0; STTI = 0;
+ DL = nullptr; DT = nullptr; SE = nullptr; TLI = nullptr; TTI = nullptr;
}
- bool runOnLoop(Loop *L, LPPassManager &LPM);
+ bool runOnLoop(Loop *L, LPPassManager &LPM) override;
bool runOnLoopBlock(BasicBlock *BB, const SCEV *BECount,
SmallVectorImpl<BasicBlock*> &ExitBlocks);
/// This transformation requires natural loop information & requires that
/// loop preheaders be inserted into the CFG.
///
- virtual void getAnalysisUsage(AnalysisUsage &AU) const {
+ void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.addRequired<LoopInfo>();
AU.addPreserved<LoopInfo>();
AU.addRequiredID(LoopSimplifyID);
AU.addPreserved<AliasAnalysis>();
AU.addRequired<ScalarEvolution>();
AU.addPreserved<ScalarEvolution>();
- AU.addPreserved<DominatorTree>();
- AU.addRequired<DominatorTree>();
+ AU.addPreserved<DominatorTreeWrapperPass>();
+ AU.addRequired<DominatorTreeWrapperPass>();
AU.addRequired<TargetLibraryInfo>();
+ AU.addRequired<TargetTransformInfo>();
}
const DataLayout *getDataLayout() {
- return TD ? TD : TD=getAnalysisIfAvailable<DataLayout>();
+ if (DL)
+ return DL;
+ DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>();
+ DL = DLP ? &DLP->getDataLayout() : nullptr;
+ return DL;
}
DominatorTree *getDominatorTree() {
- return DT ? DT : (DT=&getAnalysis<DominatorTree>());
+ return DT ? DT
+ : (DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree());
}
ScalarEvolution *getScalarEvolution() {
return TLI ? TLI : (TLI = &getAnalysis<TargetLibraryInfo>());
}
- const ScalarTargetTransformInfo *getScalarTargetTransformInfo() {
- if (!STTI) {
- TargetTransformInfo *TTI = getAnalysisIfAvailable<TargetTransformInfo>();
- if (TTI) STTI = TTI->getScalarTargetTransformInfo();
- }
- return STTI;
+ const TargetTransformInfo *getTargetTransformInfo() {
+ return TTI ? TTI : (TTI = &getAnalysis<TargetTransformInfo>());
}
Loop *getLoop() const { return CurLoop; }
INITIALIZE_PASS_BEGIN(LoopIdiomRecognize, "loop-idiom", "Recognize loop idioms",
false, false)
INITIALIZE_PASS_DEPENDENCY(LoopInfo)
-INITIALIZE_PASS_DEPENDENCY(DominatorTree)
+INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
INITIALIZE_PASS_DEPENDENCY(LoopSimplify)
INITIALIZE_PASS_DEPENDENCY(LCSSA)
INITIALIZE_PASS_DEPENDENCY(ScalarEvolution)
INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfo)
INITIALIZE_AG_DEPENDENCY(AliasAnalysis)
+INITIALIZE_AG_DEPENDENCY(TargetTransformInfo)
INITIALIZE_PASS_END(LoopIdiomRecognize, "loop-idiom", "Recognize loop idioms",
false, false)
for (unsigned op = 0, e = DeadInst->getNumOperands(); op != e; ++op) {
Value *Op = DeadInst->getOperand(op);
- DeadInst->setOperand(op, 0);
+ DeadInst->setOperand(op, nullptr);
// If this operand just became dead, add it to the NowDeadInsts list.
if (!Op->use_empty()) continue;
//
//===----------------------------------------------------------------------===//
-// This fucntion will return true iff the given block contains nothing but goto.
-// A typical usage of this function is to check if the preheader fucntion is
-// "almost" empty such that generated intrinsic function can be moved across
-// preheader and to be placed at the end of the preconditiona block without
-// concerning of breaking data dependence.
+// This function will return true iff the given block contains nothing but goto.
+// A typical usage of this function is to check if the preheader function is
+// "almost" empty such that generated intrinsic functions can be moved across
+// the preheader and be placed at the end of the precondition block without
+// the concern of breaking data dependence.
bool LIRUtil::isAlmostEmpty(BasicBlock *BB) {
if (BranchInst *Br = getBranch(BB)) {
return Br->isUnconditional() && BB->size() == 1;
return false;
}
-Value *LIRUtil::getBrCondtion(BasicBlock *BB) {
- BranchInst *Br = getBranch(BB);
- return Br ? Br->getCondition() : 0;
-}
-
BasicBlock *LIRUtil::getPrecondBb(BasicBlock *PreHead) {
if (BasicBlock *BB = PreHead->getSinglePredecessor()) {
BranchInst *Br = getBranch(BB);
- return Br && Br->isConditional() ? BB : 0;
+ return Br && Br->isConditional() ? BB : nullptr;
}
- return 0;
+ return nullptr;
}
//===----------------------------------------------------------------------===//
//===----------------------------------------------------------------------===//
NclPopcountRecognize::NclPopcountRecognize(LoopIdiomRecognize &TheLIR):
- LIR(TheLIR), CurLoop(TheLIR.getLoop()), PreCondBB(0) {
+ LIR(TheLIR), CurLoop(TheLIR.getLoop()), PreCondBB(nullptr) {
}
bool NclPopcountRecognize::preliminaryScreen() {
- const ScalarTargetTransformInfo *STTI = LIR.getScalarTargetTransformInfo();
- if (STTI->getPopcntHwSupport(32) != ScalarTargetTransformInfo::Fast)
+ const TargetTransformInfo *TTI = LIR.getTargetTransformInfo();
+ if (TTI->getPopcntSupport(32) != TargetTransformInfo::PSK_FastHardware)
return false;
- // Counting population are usually conducted by few arithmetic instrutions.
+ // Counting population are usually conducted by few arithmetic instructions.
// Such instructions can be easilly "absorbed" by vacant slots in a
// non-compact loop. Therefore, recognizing popcount idiom only makes sense
// in a compact loop.
PreCondBB = LIRUtil::getPrecondBb(PreHead);
if (!PreCondBB)
return false;
-
+
return true;
}
-Value *NclPopcountRecognize::matchCondition (BranchInst *Br,
- BasicBlock *LoopEntry) const {
+Value *NclPopcountRecognize::matchCondition(BranchInst *Br,
+ BasicBlock *LoopEntry) const {
if (!Br || !Br->isConditional())
- return 0;
+ return nullptr;
ICmpInst *Cond = dyn_cast<ICmpInst>(Br->getCondition());
if (!Cond)
- return 0;
+ return nullptr;
ConstantInt *CmpZero = dyn_cast<ConstantInt>(Cond->getOperand(1));
if (!CmpZero || !CmpZero->isZero())
- return 0;
+ return nullptr;
ICmpInst::Predicate Pred = Cond->getPredicate();
if ((Pred == ICmpInst::ICMP_NE && Br->getSuccessor(0) == LoopEntry) ||
(Pred == ICmpInst::ICMP_EQ && Br->getSuccessor(1) == LoopEntry))
return Cond->getOperand(0);
- return 0;
+ return nullptr;
}
bool NclPopcountRecognize::detectIdiom(Instruction *&CntInst,
Value *VarX1, *VarX0;
PHINode *PhiX, *CountPhi;
- DefX2 = CountInst = 0;
- VarX1 = VarX0 = 0;
- PhiX = CountPhi = 0;
+ DefX2 = CountInst = nullptr;
+ VarX1 = VarX0 = nullptr;
+ PhiX = CountPhi = nullptr;
LoopEntry = *(CurLoop->block_begin());
// step 1: Check if the loop-back branch is in desirable form.
// step 2: detect instructions corresponding to "x2 = x1 & (x1 - 1)"
{
- if (DefX2->getOpcode() != Instruction::And)
+ if (!DefX2 || DefX2->getOpcode() != Instruction::And)
return false;
BinaryOperator *SubOneOp;
// step 4: Find the instruction which count the population: cnt2 = cnt1 + 1
{
- CountInst = NULL;
+ CountInst = nullptr;
for (BasicBlock::iterator Iter = LoopEntry->getFirstNonPHI(),
IterE = LoopEntry->end(); Iter != IterE; Iter++) {
Instruction *Inst = Iter;
continue;
PHINode *Phi = dyn_cast<PHINode>(Inst->getOperand(0));
- if (!Phi && Phi->getParent() != LoopEntry)
+ if (!Phi || Phi->getParent() != LoopEntry)
continue;
// Check if the result of the instruction is live of the loop.
bool LiveOutLoop = false;
- for (Value::use_iterator I = Inst->use_begin(), E = Inst->use_end();
- I != E; I++) {
- if ((cast<Instruction>(*I))->getParent() != LoopEntry) {
+ for (User *U : Inst->users()) {
+ if ((cast<Instruction>(U))->getParent() != LoopEntry) {
LiveOutLoop = true; break;
}
}
// Assuming before transformation, the loop is following:
// if (x) // the precondition
// do { cnt++; x &= x - 1; } while(x);
-
+
// Step 1: Insert the ctpop instruction at the end of the precondition block
IRBuilderTy Builder(PreCondBr);
- Value *PopCnt, *PopCntZext, *NewCount;
+ Value *PopCnt, *PopCntZext, *NewCount, *TripCnt;
{
PopCnt = createPopcntIntrinsic(Builder, Var, DL);
NewCount = PopCntZext =
if (NewCount != PopCnt)
(cast<Instruction>(NewCount))->setDebugLoc(DL);
- // If the popoulation counter's initial value is not zero, insert Add Inst.
+ // TripCnt is exactly the number of iterations the loop has
+ TripCnt = NewCount;
+
+ // If the population counter's initial value is not zero, insert Add Inst.
Value *CntInitVal = CntPhi->getIncomingValueForBlock(PreHead);
ConstantInt *InitConst = dyn_cast<ConstantInt>(CntInitVal);
if (!InitConst || !InitConst->isZero()) {
- NewCount = Builder.CreateAdd(PopCnt, InitConst);
+ NewCount = Builder.CreateAdd(NewCount, CntInitVal);
(cast<Instruction>(NewCount))->setDebugLoc(DL);
}
}
{
BranchInst *LbBr = LIRUtil::getBranch(Body);
ICmpInst *LbCond = cast<ICmpInst>(LbBr->getCondition());
- Type *Ty = NewCount->getType();
+ Type *Ty = TripCnt->getType();
PHINode *TcPhi = PHINode::Create(Ty, 2, "tcphi", Body->begin());
Instruction *TcDec =
cast<Instruction>(Builder.CreateSub(Opnd1, Opnd2, "tcdec", false, true));
- TcPhi->addIncoming(NewCount, PreHead);
+ TcPhi->addIncoming(TripCnt, PreHead);
TcPhi->addIncoming(TcDec, Body);
CmpInst::Predicate Pred = (LbBr->getSuccessor(0) == Body) ?
// __builtin_ctpop().
{
SmallVector<Value *, 4> CntUses;
- for (Value::use_iterator I = CntInst->use_begin(), E = CntInst->use_end();
- I != E; I++) {
- if (cast<Instruction>(*I)->getParent() != Body)
- CntUses.push_back(*I);
- }
+ for (User *U : CntInst->users())
+ if (cast<Instruction>(U)->getParent() != Body)
+ CntUses.push_back(U);
for (unsigned Idx = 0; Idx < CntUses.size(); Idx++) {
(cast<Instruction>(CntUses[Idx]))->replaceUsesOfWith(CntInst, NewCount);
}
SE->forgetLoop(CurLoop);
}
-CallInst *NclPopcountRecognize::createPopcntIntrinsic(IRBuilderTy &IRBuilder,
+CallInst *NclPopcountRecognize::createPopcntIntrinsic(IRBuilderTy &IRBuilder,
Value *Val, DebugLoc DL) {
Value *Ops[] = { Val };
Type *Tys[] = { Val->getType() };
/// call, and return true; otherwise, return false.
bool NclPopcountRecognize::recognize() {
- if (!LIR.getScalarTargetTransformInfo())
+ if (!LIR.getTargetTransformInfo())
return false;
LIR.getScalarEvolution();
if (!getDataLayout())
return false;
- getDominatorTree();
+ // set DT
+ (void)getDominatorTree();
LoopInfo &LI = getAnalysis<LoopInfo>();
TLI = &getAnalysis<TargetLibraryInfo>();
- getTargetLibraryInfo();
+ // set TLI
+ (void)getTargetLibraryInfo();
SmallVector<BasicBlock*, 8> ExitBlocks;
CurLoop->getUniqueExitBlocks(ExitBlocks);
}
bool LoopIdiomRecognize::runOnLoop(Loop *L, LPPassManager &LPM) {
+ if (skipOptnoneFunction(L))
+ return false;
+
CurLoop = L;
// If the loop could not be converted to canonical form, it must have an
// If processing the store invalidated our iterator, start over from the
// top of the block.
- if (InstPtr == 0)
+ if (!InstPtr)
I = BB->begin();
continue;
}
// If processing the memset invalidated our iterator, start over from the
// top of the block.
- if (InstPtr == 0)
+ if (!InstPtr)
I = BB->begin();
continue;
}
Value *StorePtr = SI->getPointerOperand();
// Reject stores that are so large that they overflow an unsigned.
- uint64_t SizeInBits = TD->getTypeSizeInBits(StoredVal->getType());
+ uint64_t SizeInBits = DL->getTypeSizeInBits(StoredVal->getType());
if ((SizeInBits & 7) || (SizeInBits >> 32) != 0)
return false;
// random store we can't handle.
const SCEVAddRecExpr *StoreEv =
dyn_cast<SCEVAddRecExpr>(SE->getSCEV(StorePtr));
- if (StoreEv == 0 || StoreEv->getLoop() != CurLoop || !StoreEv->isAffine())
+ if (!StoreEv || StoreEv->getLoop() != CurLoop || !StoreEv->isAffine())
return false;
// Check to see if the stride matches the size of the store. If so, then we
unsigned StoreSize = (unsigned)SizeInBits >> 3;
const SCEVConstant *Stride = dyn_cast<SCEVConstant>(StoreEv->getOperand(1));
- if (Stride == 0 || StoreSize != Stride->getValue()->getValue()) {
+ if (!Stride || StoreSize != Stride->getValue()->getValue()) {
// TODO: Could also handle negative stride here someday, that will require
// the validity check in mayLoopAccessLocation to be updated though.
// Enable this to print exact negative strides.
// loop, which indicates a strided store. If we have something else, it's a
// random store we can't handle.
const SCEVAddRecExpr *Ev = dyn_cast<SCEVAddRecExpr>(SE->getSCEV(Pointer));
- if (Ev == 0 || Ev->getLoop() != CurLoop || !Ev->isAffine())
+ if (!Ev || Ev->getLoop() != CurLoop || !Ev->isAffine())
return false;
// Reject memsets that are so large that they overflow an unsigned.
// TODO: Could also handle negative stride here someday, that will require the
// validity check in mayLoopAccessLocation to be updated though.
- if (Stride == 0 || MSI->getLength() != Stride->getValue())
+ if (!Stride || MSI->getLength() != Stride->getValue())
return false;
return processLoopStridedStore(Pointer, (unsigned)SizeInBytes,
///
/// Note that we don't ever attempt to use memset_pattern8 or 4, because these
/// just replicate their input array and then pass on to memset_pattern16.
-static Constant *getMemSetPatternValue(Value *V, const DataLayout &TD) {
+static Constant *getMemSetPatternValue(Value *V, const DataLayout &DL) {
// If the value isn't a constant, we can't promote it to being in a constant
// array. We could theoretically do a store to an alloca or something, but
// that doesn't seem worthwhile.
Constant *C = dyn_cast<Constant>(V);
- if (C == 0) return 0;
+ if (!C) return nullptr;
// Only handle simple values that are a power of two bytes in size.
- uint64_t Size = TD.getTypeSizeInBits(V->getType());
+ uint64_t Size = DL.getTypeSizeInBits(V->getType());
if (Size == 0 || (Size & 7) || (Size & (Size-1)))
- return 0;
+ return nullptr;
// Don't care enough about darwin/ppc to implement this.
- if (TD.isBigEndian())
- return 0;
+ if (DL.isBigEndian())
+ return nullptr;
// Convert to size in bytes.
Size /= 8;
// TODO: If CI is larger than 16-bytes, we can try slicing it in half to see
// if the top and bottom are the same (e.g. for vectors and large integers).
- if (Size > 16) return 0;
+ if (Size > 16) return nullptr;
// If the constant is exactly 16 bytes, just use it.
if (Size == 16) return C;
// are stored. A store of i32 0x01020304 can never be turned into a memset,
// but it can be turned into memset_pattern if the target supports it.
Value *SplatValue = isBytewiseValue(StoredVal);
- Constant *PatternValue = 0;
+ Constant *PatternValue = nullptr;
+
+ unsigned DestAS = DestPtr->getType()->getPointerAddressSpace();
// If we're allowed to form a memset, and the stored value would be acceptable
// for memset, use it.
// promote the memset.
CurLoop->isLoopInvariant(SplatValue)) {
// Keep and use SplatValue.
- PatternValue = 0;
- } else if (TLI->has(LibFunc::memset_pattern16) &&
- (PatternValue = getMemSetPatternValue(StoredVal, *TD))) {
+ PatternValue = nullptr;
+ } else if (DestAS == 0 &&
+ TLI->has(LibFunc::memset_pattern16) &&
+ (PatternValue = getMemSetPatternValue(StoredVal, *DL))) {
+ // Don't create memset_pattern16s with address spaces.
// It looks like we can use PatternValue!
- SplatValue = 0;
+ SplatValue = nullptr;
} else {
// Otherwise, this isn't an idiom we can transform. For example, we can't
// do anything with a 3-byte store.
IRBuilder<> Builder(Preheader->getTerminator());
SCEVExpander Expander(*SE, "loop-idiom");
+ Type *DestInt8PtrTy = Builder.getInt8PtrTy(DestAS);
+
// Okay, we have a strided store "p[i]" of a splattable value. We can turn
// this into a memset in the loop preheader now if we want. However, this
// would be unsafe to do if there is anything else in the loop that may read
// or write to the aliased location. Check for any overlap by generating the
// base pointer and checking the region.
- unsigned AddrSpace = cast<PointerType>(DestPtr->getType())->getAddressSpace();
Value *BasePtr =
- Expander.expandCodeFor(Ev->getStart(), Builder.getInt8PtrTy(AddrSpace),
+ Expander.expandCodeFor(Ev->getStart(), DestInt8PtrTy,
Preheader->getTerminator());
-
if (mayLoopAccessLocation(BasePtr, AliasAnalysis::ModRef,
CurLoop, BECount,
- StoreSize, getAnalysis<AliasAnalysis>(), TheStore)){
+ StoreSize, getAnalysis<AliasAnalysis>(), TheStore)) {
Expander.clear();
// If we generated new code for the base pointer, clean up.
deleteIfDeadInstruction(BasePtr, *SE, TLI);
// The # stored bytes is (BECount+1)*Size. Expand the trip count out to
// pointer size if it isn't already.
- Type *IntPtr = TD->getIntPtrType(DestPtr->getContext());
+ Type *IntPtr = Builder.getIntPtrTy(DL, DestAS);
BECount = SE->getTruncateOrZeroExtend(BECount, IntPtr);
const SCEV *NumBytesS = SE->getAddExpr(BECount, SE->getConstant(IntPtr, 1),
SCEV::FlagNUW);
- if (StoreSize != 1)
+ if (StoreSize != 1) {
NumBytesS = SE->getMulExpr(NumBytesS, SE->getConstant(IntPtr, StoreSize),
SCEV::FlagNUW);
+ }
Value *NumBytes =
Expander.expandCodeFor(NumBytesS, IntPtr, Preheader->getTerminator());
CallInst *NewCall;
- if (SplatValue)
- NewCall = Builder.CreateMemSet(BasePtr, SplatValue,NumBytes,StoreAlignment);
- else {
+ if (SplatValue) {
+ NewCall = Builder.CreateMemSet(BasePtr,
+ SplatValue,
+ NumBytes,
+ StoreAlignment);
+ } else {
+ // Everything is emitted in default address space
+ Type *Int8PtrTy = DestInt8PtrTy;
+
Module *M = TheStore->getParent()->getParent()->getParent();
Value *MSP = M->getOrInsertFunction("memset_pattern16",
Builder.getVoidTy(),
- Builder.getInt8PtrTy(),
- Builder.getInt8PtrTy(), IntPtr,
- (void*)0);
+ Int8PtrTy,
+ Int8PtrTy,
+ IntPtr,
+ (void*)nullptr);
// Otherwise we should form a memset_pattern16. PatternValue is known to be
// an constant array of 16-bytes. Plop the value into a mergable global.
PatternValue, ".memset_pattern");
GV->setUnnamedAddr(true); // Ok to merge these.
GV->setAlignment(16);
- Value *PatternPtr = ConstantExpr::getBitCast(GV, Builder.getInt8PtrTy());
+ Value *PatternPtr = ConstantExpr::getBitCast(GV, Int8PtrTy);
NewCall = Builder.CreateCall3(MSP, BasePtr, PatternPtr, NumBytes);
}
// The # stored bytes is (BECount+1)*Size. Expand the trip count out to
// pointer size if it isn't already.
- Type *IntPtr = TD->getIntPtrType(SI->getContext());
- BECount = SE->getTruncateOrZeroExtend(BECount, IntPtr);
+ Type *IntPtrTy = Builder.getIntPtrTy(DL, SI->getPointerAddressSpace());
+ BECount = SE->getTruncateOrZeroExtend(BECount, IntPtrTy);
- const SCEV *NumBytesS = SE->getAddExpr(BECount, SE->getConstant(IntPtr, 1),
+ const SCEV *NumBytesS = SE->getAddExpr(BECount, SE->getConstant(IntPtrTy, 1),
SCEV::FlagNUW);
if (StoreSize != 1)
- NumBytesS = SE->getMulExpr(NumBytesS, SE->getConstant(IntPtr, StoreSize),
+ NumBytesS = SE->getMulExpr(NumBytesS, SE->getConstant(IntPtrTy, StoreSize),
SCEV::FlagNUW);
Value *NumBytes =
- Expander.expandCodeFor(NumBytesS, IntPtr, Preheader->getTerminator());
+ Expander.expandCodeFor(NumBytesS, IntPtrTy, Preheader->getTerminator());
CallInst *NewCall =
Builder.CreateMemCpy(StoreBasePtr, LoadBasePtr, NumBytes,
projects / oota-llvm.git / blobdiff