#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
-#include "llvm/Support/IntegersSubsetMapping.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetFrameLowering.h"
#include "llvm/Target/TargetLibraryInfo.h"
#include "llvm/Target/TargetLowering.h"
#include "llvm/Target/TargetOptions.h"
+#include "llvm/Target/TargetSelectionDAGInfo.h"
#include <algorithm>
using namespace llvm;
assert(PartEVT.getVectorNumElements() > ValueVT.getVectorNumElements() &&
"Cannot narrow, it would be a lossy transformation");
return DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL, ValueVT, Val,
- DAG.getIntPtrConstant(0));
+ DAG.getConstant(0, TLI.getVectorIdxTy()));
}
// Vector/Vector bitcast.
SmallVector<SDValue, 16> Ops;
for (unsigned i = 0, e = ValueVT.getVectorNumElements(); i != e; ++i)
Ops.push_back(DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL,
- ElementVT, Val, DAG.getIntPtrConstant(i)));
+ ElementVT, Val, DAG.getConstant(i,
+ TLI.getVectorIdxTy())));
for (unsigned i = ValueVT.getVectorNumElements(),
e = PartVT.getVectorNumElements(); i != e; ++i)
assert(ValueVT.getVectorNumElements() == 1 &&
"Only trivial vector-to-scalar conversions should get here!");
Val = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL,
- PartVT, Val, DAG.getIntPtrConstant(0));
+ PartVT, Val, DAG.getConstant(0, TLI.getVectorIdxTy()));
bool Smaller = ValueVT.bitsLE(PartVT);
Val = DAG.getNode((Smaller ? ISD::TRUNCATE : ISD::ANY_EXTEND),
if (IntermediateVT.isVector())
Ops[i] = DAG.getNode(ISD::EXTRACT_SUBVECTOR, DL,
IntermediateVT, Val,
- DAG.getIntPtrConstant(i * (NumElements / NumIntermediates)));
+ DAG.getConstant(i * (NumElements / NumIntermediates),
+ TLI.getVectorIdxTy()));
else
Ops[i] = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, DL,
- IntermediateVT, Val, DAG.getIntPtrConstant(i));
+ IntermediateVT, Val,
+ DAG.getConstant(i, TLI.getVectorIdxTy()));
}
// Split the intermediate operands into legal parts.
return DAG.getMergeValues(&Constants[0], Constants.size(),
getCurSDLoc());
}
-
+
if (const ConstantDataSequential *CDS =
dyn_cast<ConstantDataSequential>(C)) {
SmallVector<SDValue, 4> Ops;
} else
Cond = DAG.getSetCC(dl, MVT::i1, CondLHS, getValue(CB.CmpRHS), CB.CC);
} else {
- assert(CB.CC == ISD::SETCC_INVALID &&
- "Condition is undefined for to-the-range belonging check.");
+ assert(CB.CC == ISD::SETLE && "Can handle only LE ranges now");
const APInt& Low = cast<ConstantInt>(CB.CmpLHS)->getValue();
const APInt& High = cast<ConstantInt>(CB.CmpRHS)->getValue();
SDValue CmpOp = getValue(CB.CmpMHS);
EVT VT = CmpOp.getValueType();
-
- if (cast<ConstantInt>(CB.CmpLHS)->isMinValue(false)) {
+
+ if (cast<ConstantInt>(CB.CmpLHS)->isMinValue(true)) {
Cond = DAG.getSetCC(dl, MVT::i1, CmpOp, DAG.getConstant(High, VT),
- ISD::SETULE);
+ ISD::SETLE);
} else {
SDValue SUB = DAG.getNode(ISD::SUB, dl,
VT, CmpOp, DAG.getConstant(Low, VT));
DAG.setRoot(BrCond);
}
+/// Codegen a new tail for a stack protector check ParentMBB which has had its
+/// tail spliced into a stack protector check success bb.
+///
+/// For a high level explanation of how this fits into the stack protector
+/// generation see the comment on the declaration of class
+/// StackProtectorDescriptor.
+void SelectionDAGBuilder::visitSPDescriptorParent(StackProtectorDescriptor &SPD,
+ MachineBasicBlock *ParentBB) {
+
+ // First create the loads to the guard/stack slot for the comparison.
+ const TargetLowering *TLI = TM.getTargetLowering();
+ EVT PtrTy = TLI->getPointerTy();
+
+ MachineFrameInfo *MFI = ParentBB->getParent()->getFrameInfo();
+ int FI = MFI->getStackProtectorIndex();
+
+ const Value *IRGuard = SPD.getGuard();
+ SDValue GuardPtr = getValue(IRGuard);
+ SDValue StackSlotPtr = DAG.getFrameIndex(FI, PtrTy);
+
+ unsigned Align =
+ TLI->getDataLayout()->getPrefTypeAlignment(IRGuard->getType());
+ SDValue Guard = DAG.getLoad(PtrTy, getCurSDLoc(), DAG.getEntryNode(),
+ GuardPtr, MachinePointerInfo(IRGuard, 0),
+ true, false, false, Align);
+
+ SDValue StackSlot = DAG.getLoad(PtrTy, getCurSDLoc(), DAG.getEntryNode(),
+ StackSlotPtr,
+ MachinePointerInfo::getFixedStack(FI),
+ true, false, false, Align);
+
+ // Perform the comparison via a subtract/getsetcc.
+ EVT VT = Guard.getValueType();
+ SDValue Sub = DAG.getNode(ISD::SUB, getCurSDLoc(), VT, Guard, StackSlot);
+
+ SDValue Cmp = DAG.getSetCC(getCurSDLoc(),
+ TLI->getSetCCResultType(*DAG.getContext(),
+ Sub.getValueType()),
+ Sub, DAG.getConstant(0, VT),
+ ISD::SETNE);
+
+ // If the sub is not 0, then we know the guard/stackslot do not equal, so
+ // branch to failure MBB.
+ SDValue BrCond = DAG.getNode(ISD::BRCOND, getCurSDLoc(),
+ MVT::Other, StackSlot.getOperand(0),
+ Cmp, DAG.getBasicBlock(SPD.getFailureMBB()));
+ // Otherwise branch to success MBB.
+ SDValue Br = DAG.getNode(ISD::BR, getCurSDLoc(),
+ MVT::Other, BrCond,
+ DAG.getBasicBlock(SPD.getSuccessMBB()));
+
+ DAG.setRoot(Br);
+}
+
+/// Codegen the failure basic block for a stack protector check.
+///
+/// A failure stack protector machine basic block consists simply of a call to
+/// __stack_chk_fail().
+///
+/// For a high level explanation of how this fits into the stack protector
+/// generation see the comment on the declaration of class
+/// StackProtectorDescriptor.
+void
+SelectionDAGBuilder::visitSPDescriptorFailure(StackProtectorDescriptor &SPD) {
+ const TargetLowering *TLI = TM.getTargetLowering();
+ SDValue Chain = TLI->makeLibCall(DAG, RTLIB::STACKPROTECTOR_CHECK_FAIL,
+ MVT::isVoid, 0, 0, false, getCurSDLoc(),
+ false, false).second;
+ DAG.setRoot(Chain);
+}
+
/// visitBitTestHeader - This function emits necessary code to produce value
/// suitable for "bit tests"
void SelectionDAGBuilder::visitBitTestHeader(BitTestBlock &B,
CC = ISD::SETEQ;
LHS = SV; RHS = I->High; MHS = NULL;
} else {
- CC = ISD::SETCC_INVALID;
+ CC = ISD::SETLE;
LHS = I->Low; MHS = SV; RHS = I->High;
}
static APInt ComputeRange(const APInt &First, const APInt &Last) {
uint32_t BitWidth = std::max(Last.getBitWidth(), First.getBitWidth()) + 1;
- APInt LastExt = Last.zext(BitWidth), FirstExt = First.zext(BitWidth);
+ APInt LastExt = Last.sext(BitWidth), FirstExt = First.sext(BitWidth);
return (LastExt - FirstExt + 1ULL);
}
const APInt &Low = cast<ConstantInt>(I->Low)->getValue();
const APInt &High = cast<ConstantInt>(I->High)->getValue();
- if (Low.ule(TEI) && TEI.ule(High)) {
+ if (Low.sle(TEI) && TEI.sle(High)) {
DestBBs.push_back(I->BB);
if (TEI==High)
++I;
for (CaseItr I = CR.Range.first, E = CR.Range.second; I != E; ++I) {
DenseMap<MachineBasicBlock*, uint32_t>::iterator Itr =
DestWeights.find(I->BB);
- if (Itr != DestWeights.end())
+ if (Itr != DestWeights.end())
Itr->second += I->ExtraWeight;
else
DestWeights[I->BB] = I->ExtraWeight;
// Create a CaseBlock record representing a conditional branch to
// the LHS node if the value being switched on SV is less than C.
// Otherwise, branch to LHS.
- CaseBlock CB(ISD::SETULT, SV, C, NULL, TrueBB, FalseBB, CR.CaseBB);
+ CaseBlock CB(ISD::SETLT, SV, C, NULL, TrueBB, FalseBB, CR.CaseBB);
if (CR.CaseBB == SwitchBB)
visitSwitchCase(CB, SwitchBB);
MachineFunction *CurMF = FuncInfo.MF;
// If target does not have legal shift left, do not emit bit tests at all.
- if (!TLI->isOperationLegal(ISD::SHL, TLI->getPointerTy()))
+ if (!TLI->isOperationLegal(ISD::SHL, PTy))
return false;
size_t numCmps = 0;
// Optimize the case where all the case values fit in a
// word without having to subtract minValue. In this case,
// we can optimize away the subtraction.
- if (maxValue.ult(IntPtrBits)) {
+ if (minValue.isNonNegative() && maxValue.slt(IntPtrBits)) {
cmpRange = maxValue;
} else {
lowBound = minValue;
/// Clusterify - Transform simple list of Cases into list of CaseRange's
size_t SelectionDAGBuilder::Clusterify(CaseVector& Cases,
const SwitchInst& SI) {
-
- /// Use a shorter form of declaration, and also
- /// show the we want to use CRSBuilder as Clusterifier.
- typedef IntegersSubsetMapping<MachineBasicBlock> Clusterifier;
-
- Clusterifier TheClusterifier;
+ size_t numCmps = 0;
BranchProbabilityInfo *BPI = FuncInfo.BPI;
// Start with "simple" cases
const BasicBlock *SuccBB = i.getCaseSuccessor();
MachineBasicBlock *SMBB = FuncInfo.MBBMap[SuccBB];
- TheClusterifier.add(i.getCaseValueEx(), SMBB,
- BPI ? BPI->getEdgeWeight(SI.getParent(), i.getSuccessorIndex()) : 0);
- }
-
- TheClusterifier.optimize();
-
- size_t numCmps = 0;
- for (Clusterifier::RangeIterator i = TheClusterifier.begin(),
- e = TheClusterifier.end(); i != e; ++i, ++numCmps) {
- Clusterifier::Cluster &C = *i;
- // Update edge weight for the cluster.
- unsigned W = C.first.Weight;
-
- // FIXME: Currently work with ConstantInt based numbers.
- // Changing it to APInt based is a pretty heavy for this commit.
- Cases.push_back(Case(C.first.getLow().toConstantInt(),
- C.first.getHigh().toConstantInt(), C.second, W));
-
- if (C.first.getLow() != C.first.getHigh())
- // A range counts double, since it requires two compares.
- ++numCmps;
+ uint32_t ExtraWeight =
+ BPI ? BPI->getEdgeWeight(SI.getParent(), i.getSuccessorIndex()) : 0;
+
+ Cases.push_back(Case(i.getCaseValue(), i.getCaseValue(),
+ SMBB, ExtraWeight));
+ }
+ std::sort(Cases.begin(), Cases.end(), CaseCmp());
+
+ // Merge case into clusters
+ if (Cases.size() >= 2)
+ // Must recompute end() each iteration because it may be
+ // invalidated by erase if we hold on to it
+ for (CaseItr I = Cases.begin(), J = llvm::next(Cases.begin());
+ J != Cases.end(); ) {
+ const APInt& nextValue = cast<ConstantInt>(J->Low)->getValue();
+ const APInt& currentValue = cast<ConstantInt>(I->High)->getValue();
+ MachineBasicBlock* nextBB = J->BB;
+ MachineBasicBlock* currentBB = I->BB;
+
+ // If the two neighboring cases go to the same destination, merge them
+ // into a single case.
+ if ((nextValue - currentValue == 1) && (currentBB == nextBB)) {
+ I->High = J->High;
+ I->ExtraWeight += J->ExtraWeight;
+ J = Cases.erase(J);
+ } else {
+ I = J++;
+ }
+ }
+
+ for (CaseItr I=Cases.begin(), E=Cases.end(); I!=E; ++I, ++numCmps) {
+ if (I->Low != I->High)
+ // A range counts double, since it requires two compares.
+ ++numCmps;
}
return numCmps;
}
void SelectionDAGBuilder::visitInsertElement(const User &I) {
+ const TargetLowering &TLI = DAG.getTargetLoweringInfo();
SDValue InVec = getValue(I.getOperand(0));
SDValue InVal = getValue(I.getOperand(1));
- SDValue InIdx = DAG.getNode(ISD::ZERO_EXTEND, getCurSDLoc(),
- TM.getTargetLowering()->getPointerTy(),
- getValue(I.getOperand(2)));
+ SDValue InIdx = DAG.getSExtOrTrunc(getValue(I.getOperand(2)),
+ getCurSDLoc(), TLI.getVectorIdxTy());
setValue(&I, DAG.getNode(ISD::INSERT_VECTOR_ELT, getCurSDLoc(),
TM.getTargetLowering()->getValueType(I.getType()),
InVec, InVal, InIdx));
}
void SelectionDAGBuilder::visitExtractElement(const User &I) {
+ const TargetLowering &TLI = DAG.getTargetLoweringInfo();
SDValue InVec = getValue(I.getOperand(0));
- SDValue InIdx = DAG.getNode(ISD::ZERO_EXTEND, getCurSDLoc(),
- TM.getTargetLowering()->getPointerTy(),
- getValue(I.getOperand(1)));
+ SDValue InIdx = DAG.getSExtOrTrunc(getValue(I.getOperand(1)),
+ getCurSDLoc(), TLI.getVectorIdxTy());
setValue(&I, DAG.getNode(ISD::EXTRACT_VECTOR_ELT, getCurSDLoc(),
TM.getTargetLowering()->getValueType(I.getType()),
InVec, InIdx));
Src = DAG.getUNDEF(VT);
else
Src = DAG.getNode(ISD::EXTRACT_SUBVECTOR, getCurSDLoc(), VT,
- Src, DAG.getIntPtrConstant(StartIdx[Input]));
+ Src, DAG.getConstant(StartIdx[Input],
+ TLI->getVectorIdxTy()));
}
// Calculate new mask.
// replacing the shuffle with extract and build vector.
// to insert and build vector.
EVT EltVT = VT.getVectorElementType();
- EVT PtrVT = TLI->getPointerTy();
+ EVT IdxVT = TLI->getVectorIdxTy();
SmallVector<SDValue,8> Ops;
for (unsigned i = 0; i != MaskNumElts; ++i) {
int Idx = Mask[i];
if (Idx >= (int)SrcNumElts) Idx -= SrcNumElts;
Res = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, getCurSDLoc(),
- EltVT, Src, DAG.getConstant(Idx, PtrVT));
+ EltVT, Src, DAG.getConstant(Idx, IdxVT));
}
Ops.push_back(Res);
}
void SelectionDAGBuilder::visitGetElementPtr(const User &I) {
- SDValue N = getValue(I.getOperand(0));
+ Value *Op0 = I.getOperand(0);
// Note that the pointer operand may be a vector of pointers. Take the scalar
// element which holds a pointer.
- Type *Ty = I.getOperand(0)->getType()->getScalarType();
+ Type *Ty = Op0->getType()->getScalarType();
+ unsigned AS = Ty->getPointerAddressSpace();
+ SDValue N = getValue(Op0);
for (GetElementPtrInst::const_op_iterator OI = I.op_begin()+1, E = I.op_end();
OI != E; ++OI) {
uint64_t Offs =
TD->getTypeAllocSize(Ty)*cast<ConstantInt>(CI)->getSExtValue();
SDValue OffsVal;
- EVT PTy = TLI->getPointerTy();
+ EVT PTy = TLI->getPointerTy(AS);
unsigned PtrBits = PTy.getSizeInBits();
if (PtrBits < 64)
- OffsVal = DAG.getNode(ISD::TRUNCATE, getCurSDLoc(),
- TLI->getPointerTy(),
+ OffsVal = DAG.getNode(ISD::TRUNCATE, getCurSDLoc(), PTy,
DAG.getConstant(Offs, MVT::i64));
else
- OffsVal = DAG.getIntPtrConstant(Offs);
+ OffsVal = DAG.getConstant(Offs, PTy);
N = DAG.getNode(ISD::ADD, getCurSDLoc(), N.getValueType(), N,
OffsVal);
}
// N = N + Idx * ElementSize;
- APInt ElementSize = APInt(TLI->getPointerTy().getSizeInBits(),
+ APInt ElementSize = APInt(TLI->getPointerSizeInBits(AS),
TD->getTypeAllocSize(Ty));
SDValue IdxN = getValue(Idx);
SDValue L =
DAG.getAtomic(ISD::ATOMIC_CMP_SWAP, dl,
- getValue(I.getCompareOperand()).getValueType().getSimpleVT(),
+ getValue(I.getCompareOperand()).getSimpleValueType(),
InChain,
getValue(I.getPointerOperand()),
getValue(I.getCompareOperand()),
SDValue L =
DAG.getAtomic(NT, dl,
- getValue(I.getValOperand()).getValueType().getSimpleVT(),
+ getValue(I.getValOperand()).getSimpleValueType(),
InChain,
getValue(I.getPointerOperand()),
getValue(I.getValOperand()),
if (!Op)
return false;
+ // FIXME: This does not handle register-indirect values at offset 0.
+ bool IsIndirect = Offset != 0;
if (Op->isReg())
- Op->setIsDebug();
-
- FuncInfo.ArgDbgValues.push_back(
+ FuncInfo.ArgDbgValues.push_back(BuildMI(MF, getCurDebugLoc(),
+ TII->get(TargetOpcode::DBG_VALUE),
+ IsIndirect,
+ Op->getReg(), Offset, Variable));
+ else
+ FuncInfo.ArgDbgValues.push_back(
BuildMI(MF, getCurDebugLoc(), TII->get(TargetOpcode::DBG_VALUE))
.addOperand(*Op).addImm(Offset).addMetadata(Variable));
+
return true;
}
SDValue CfaArg = DAG.getSExtOrTrunc(getValue(I.getArgOperand(0)), sdl,
TLI->getPointerTy());
SDValue Offset = DAG.getNode(ISD::ADD, sdl,
- TLI->getPointerTy(),
+ CfaArg.getValueType(),
DAG.getNode(ISD::FRAME_TO_ARGS_OFFSET, sdl,
- TLI->getPointerTy()),
+ CfaArg.getValueType()),
CfaArg);
SDValue FA = DAG.getNode(ISD::FRAMEADDR, sdl,
TLI->getPointerTy(),
DAG.getConstant(0, TLI->getPointerTy()));
- setValue(&I, DAG.getNode(ISD::ADD, sdl, TLI->getPointerTy(),
+ setValue(&I, DAG.getNode(ISD::ADD, sdl, FA.getValueType(),
FA, Offset));
return 0;
}
Res = DAG.getNode(ISD::INSERT_SUBVECTOR, sdl, DestVT,
getValue(I.getArgOperand(0)),
getValue(I.getArgOperand(1)),
- DAG.getIntPtrConstant(Idx));
+ DAG.getConstant(Idx, TLI->getVectorIdxTy()));
setValue(&I, Res);
return 0;
}
DestVT.getVectorNumElements();
Res = DAG.getNode(ISD::EXTRACT_SUBVECTOR, sdl, DestVT,
getValue(I.getArgOperand(0)),
- DAG.getIntPtrConstant(Idx));
+ DAG.getConstant(Idx, TLI->getVectorIdxTy()));
setValue(&I, Res);
return 0;
}
case Intrinsic::ceil:
case Intrinsic::trunc:
case Intrinsic::rint:
- case Intrinsic::nearbyint: {
+ case Intrinsic::nearbyint:
+ case Intrinsic::round: {
unsigned Opcode;
switch (Intrinsic) {
default: llvm_unreachable("Impossible intrinsic"); // Can't reach here.
case Intrinsic::trunc: Opcode = ISD::FTRUNC; break;
case Intrinsic::rint: Opcode = ISD::FRINT; break;
case Intrinsic::nearbyint: Opcode = ISD::FNEARBYINT; break;
+ case Intrinsic::round: Opcode = ISD::FROUND; break;
}
setValue(&I, DAG.getNode(Opcode, sdl,
getValue(I.getArgOperand(0))));
return 0;
}
+ case Intrinsic::copysign:
+ setValue(&I, DAG.getNode(ISD::FCOPYSIGN, sdl,
+ getValue(I.getArgOperand(0)).getValueType(),
+ getValue(I.getArgOperand(0)),
+ getValue(I.getArgOperand(1))));
+ return 0;
case Intrinsic::fma:
setValue(&I, DAG.getNode(ISD::FMA, sdl,
getValue(I.getArgOperand(0)).getValueType(),
case Intrinsic::fmuladd: {
EVT VT = TLI->getValueType(I.getType());
if (TM.Options.AllowFPOpFusion != FPOpFusion::Strict &&
- TLI->isFMAFasterThanMulAndAdd(VT)) {
+ TLI->isFMAFasterThanFMulAndFAdd(VT)) {
setValue(&I, DAG.getNode(ISD::FMA, sdl,
getValue(I.getArgOperand(0)).getValueType(),
getValue(I.getArgOperand(0)),
case Intrinsic::trap: {
StringRef TrapFuncName = TM.Options.getTrapFunctionName();
if (TrapFuncName.empty()) {
- ISD::NodeType Op = (Intrinsic == Intrinsic::trap) ?
+ ISD::NodeType Op = (Intrinsic == Intrinsic::trap) ?
ISD::TRAP : ISD::DEBUGTRAP;
DAG.setRoot(DAG.getNode(Op, sdl,MVT::Other, getRoot()));
return 0;
case Intrinsic::invariant_end:
// Discard region information.
return 0;
+ case Intrinsic::stackprotectorcheck: {
+ // Do not actually emit anything for this basic block. Instead we initialize
+ // the stack protector descriptor and export the guard variable so we can
+ // access it in FinishBasicBlock.
+ const BasicBlock *BB = I.getParent();
+ SPDescriptor.initialize(BB, FuncInfo.MBBMap[BB], I);
+ ExportFromCurrentBlock(SPDescriptor.getGuard());
+
+ // Flush our exports since we are going to process a terminator.
+ (void)getControlRoot();
+ return 0;
+ }
case Intrinsic::donothing:
// ignore
return 0;
SDValue LoadVal = Builder.DAG.getLoad(LoadVT, Builder.getCurSDLoc(), Root,
Ptr, MachinePointerInfo(PtrVal),
false /*volatile*/,
- false /*nontemporal*/,
+ false /*nontemporal*/,
false /*isinvariant*/, 1 /* align=1 */);
if (!ConstantMemory)
return LoadVal;
}
+/// processIntegerCallValue - Record the value for an instruction that
+/// produces an integer result, converting the type where necessary.
+void SelectionDAGBuilder::processIntegerCallValue(const Instruction &I,
+ SDValue Value,
+ bool IsSigned) {
+ EVT VT = TM.getTargetLowering()->getValueType(I.getType(), true);
+ if (IsSigned)
+ Value = DAG.getSExtOrTrunc(Value, getCurSDLoc(), VT);
+ else
+ Value = DAG.getZExtOrTrunc(Value, getCurSDLoc(), VT);
+ setValue(&I, Value);
+}
/// visitMemCmpCall - See if we can lower a call to memcmp in an optimized form.
/// If so, return true and lower it, otherwise return false and it will be
!I.getType()->isIntegerTy())
return false;
- const ConstantInt *Size = dyn_cast<ConstantInt>(I.getArgOperand(2));
+ const Value *Size = I.getArgOperand(2);
+ const ConstantInt *CSize = dyn_cast<ConstantInt>(Size);
+ if (CSize && CSize->getZExtValue() == 0) {
+ EVT CallVT = TM.getTargetLowering()->getValueType(I.getType(), true);
+ setValue(&I, DAG.getConstant(0, CallVT));
+ return true;
+ }
+
+ const TargetSelectionDAGInfo &TSI = DAG.getSelectionDAGInfo();
+ std::pair<SDValue, SDValue> Res =
+ TSI.EmitTargetCodeForMemcmp(DAG, getCurSDLoc(), DAG.getRoot(),
+ getValue(LHS), getValue(RHS), getValue(Size),
+ MachinePointerInfo(LHS),
+ MachinePointerInfo(RHS));
+ if (Res.first.getNode()) {
+ processIntegerCallValue(I, Res.first, true);
+ PendingLoads.push_back(Res.second);
+ return true;
+ }
// memcmp(S1,S2,2) != 0 -> (*(short*)LHS != *(short*)RHS) != 0
// memcmp(S1,S2,4) != 0 -> (*(int*)LHS != *(int*)RHS) != 0
- if (Size && IsOnlyUsedInZeroEqualityComparison(&I)) {
+ if (CSize && IsOnlyUsedInZeroEqualityComparison(&I)) {
bool ActuallyDoIt = true;
MVT LoadVT;
Type *LoadTy;
- switch (Size->getZExtValue()) {
+ switch (CSize->getZExtValue()) {
default:
LoadVT = MVT::Other;
LoadTy = 0;
break;
case 2:
LoadVT = MVT::i16;
- LoadTy = Type::getInt16Ty(Size->getContext());
+ LoadTy = Type::getInt16Ty(CSize->getContext());
break;
case 4:
LoadVT = MVT::i32;
- LoadTy = Type::getInt32Ty(Size->getContext());
+ LoadTy = Type::getInt32Ty(CSize->getContext());
break;
case 8:
LoadVT = MVT::i64;
- LoadTy = Type::getInt64Ty(Size->getContext());
+ LoadTy = Type::getInt64Ty(CSize->getContext());
break;
/*
case 16:
LoadVT = MVT::v4i32;
- LoadTy = Type::getInt32Ty(Size->getContext());
+ LoadTy = Type::getInt32Ty(CSize->getContext());
LoadTy = VectorType::get(LoadTy, 4);
break;
*/
// supports unaligned loads of that type. Expanding into byte loads would
// bloat the code.
const TargetLowering *TLI = TM.getTargetLowering();
- if (ActuallyDoIt && Size->getZExtValue() > 4) {
+ if (ActuallyDoIt && CSize->getZExtValue() > 4) {
// TODO: Handle 5 byte compare as 4-byte + 1 byte.
// TODO: Handle 8 byte compare on x86-32 as two 32-bit loads.
if (!TLI->isTypeLegal(LoadVT) ||!TLI->allowsUnalignedMemoryAccesses(LoadVT))
SDValue Res = DAG.getSetCC(getCurSDLoc(), MVT::i1, LHSVal, RHSVal,
ISD::SETNE);
- EVT CallVT = TLI->getValueType(I.getType(), true);
- setValue(&I, DAG.getZExtOrTrunc(Res, getCurSDLoc(), CallVT));
+ processIntegerCallValue(I, Res, false);
return true;
}
}
return false;
}
+/// visitMemChrCall -- See if we can lower a memchr call into an optimized
+/// form. If so, return true and lower it, otherwise return false and it
+/// will be lowered like a normal call.
+bool SelectionDAGBuilder::visitMemChrCall(const CallInst &I) {
+ // Verify that the prototype makes sense. void *memchr(void *, int, size_t)
+ if (I.getNumArgOperands() != 3)
+ return false;
+
+ const Value *Src = I.getArgOperand(0);
+ const Value *Char = I.getArgOperand(1);
+ const Value *Length = I.getArgOperand(2);
+ if (!Src->getType()->isPointerTy() ||
+ !Char->getType()->isIntegerTy() ||
+ !Length->getType()->isIntegerTy() ||
+ !I.getType()->isPointerTy())
+ return false;
+
+ const TargetSelectionDAGInfo &TSI = DAG.getSelectionDAGInfo();
+ std::pair<SDValue, SDValue> Res =
+ TSI.EmitTargetCodeForMemchr(DAG, getCurSDLoc(), DAG.getRoot(),
+ getValue(Src), getValue(Char), getValue(Length),
+ MachinePointerInfo(Src));
+ if (Res.first.getNode()) {
+ setValue(&I, Res.first);
+ PendingLoads.push_back(Res.second);
+ return true;
+ }
+
+ return false;
+}
+
+/// visitStrCpyCall -- See if we can lower a strcpy or stpcpy call into an
+/// optimized form. If so, return true and lower it, otherwise return false
+/// and it will be lowered like a normal call.
+bool SelectionDAGBuilder::visitStrCpyCall(const CallInst &I, bool isStpcpy) {
+ // Verify that the prototype makes sense. char *strcpy(char *, char *)
+ if (I.getNumArgOperands() != 2)
+ return false;
+
+ const Value *Arg0 = I.getArgOperand(0), *Arg1 = I.getArgOperand(1);
+ if (!Arg0->getType()->isPointerTy() ||
+ !Arg1->getType()->isPointerTy() ||
+ !I.getType()->isPointerTy())
+ return false;
+
+ const TargetSelectionDAGInfo &TSI = DAG.getSelectionDAGInfo();
+ std::pair<SDValue, SDValue> Res =
+ TSI.EmitTargetCodeForStrcpy(DAG, getCurSDLoc(), getRoot(),
+ getValue(Arg0), getValue(Arg1),
+ MachinePointerInfo(Arg0),
+ MachinePointerInfo(Arg1), isStpcpy);
+ if (Res.first.getNode()) {
+ setValue(&I, Res.first);
+ DAG.setRoot(Res.second);
+ return true;
+ }
+
+ return false;
+}
+
+/// visitStrCmpCall - See if we can lower a call to strcmp in an optimized form.
+/// If so, return true and lower it, otherwise return false and it will be
+/// lowered like a normal call.
+bool SelectionDAGBuilder::visitStrCmpCall(const CallInst &I) {
+ // Verify that the prototype makes sense. int strcmp(void*,void*)
+ if (I.getNumArgOperands() != 2)
+ return false;
+
+ const Value *Arg0 = I.getArgOperand(0), *Arg1 = I.getArgOperand(1);
+ if (!Arg0->getType()->isPointerTy() ||
+ !Arg1->getType()->isPointerTy() ||
+ !I.getType()->isIntegerTy())
+ return false;
+
+ const TargetSelectionDAGInfo &TSI = DAG.getSelectionDAGInfo();
+ std::pair<SDValue, SDValue> Res =
+ TSI.EmitTargetCodeForStrcmp(DAG, getCurSDLoc(), DAG.getRoot(),
+ getValue(Arg0), getValue(Arg1),
+ MachinePointerInfo(Arg0),
+ MachinePointerInfo(Arg1));
+ if (Res.first.getNode()) {
+ processIntegerCallValue(I, Res.first, true);
+ PendingLoads.push_back(Res.second);
+ return true;
+ }
+
+ return false;
+}
+
+/// visitStrLenCall -- See if we can lower a strlen call into an optimized
+/// form. If so, return true and lower it, otherwise return false and it
+/// will be lowered like a normal call.
+bool SelectionDAGBuilder::visitStrLenCall(const CallInst &I) {
+ // Verify that the prototype makes sense. size_t strlen(char *)
+ if (I.getNumArgOperands() != 1)
+ return false;
+
+ const Value *Arg0 = I.getArgOperand(0);
+ if (!Arg0->getType()->isPointerTy() || !I.getType()->isIntegerTy())
+ return false;
+
+ const TargetSelectionDAGInfo &TSI = DAG.getSelectionDAGInfo();
+ std::pair<SDValue, SDValue> Res =
+ TSI.EmitTargetCodeForStrlen(DAG, getCurSDLoc(), DAG.getRoot(),
+ getValue(Arg0), MachinePointerInfo(Arg0));
+ if (Res.first.getNode()) {
+ processIntegerCallValue(I, Res.first, false);
+ PendingLoads.push_back(Res.second);
+ return true;
+ }
+
+ return false;
+}
+
+/// visitStrNLenCall -- See if we can lower a strnlen call into an optimized
+/// form. If so, return true and lower it, otherwise return false and it
+/// will be lowered like a normal call.
+bool SelectionDAGBuilder::visitStrNLenCall(const CallInst &I) {
+ // Verify that the prototype makes sense. size_t strnlen(char *, size_t)
+ if (I.getNumArgOperands() != 2)
+ return false;
+
+ const Value *Arg0 = I.getArgOperand(0), *Arg1 = I.getArgOperand(1);
+ if (!Arg0->getType()->isPointerTy() ||
+ !Arg1->getType()->isIntegerTy() ||
+ !I.getType()->isIntegerTy())
+ return false;
+
+ const TargetSelectionDAGInfo &TSI = DAG.getSelectionDAGInfo();
+ std::pair<SDValue, SDValue> Res =
+ TSI.EmitTargetCodeForStrnlen(DAG, getCurSDLoc(), DAG.getRoot(),
+ getValue(Arg0), getValue(Arg1),
+ MachinePointerInfo(Arg0));
+ if (Res.first.getNode()) {
+ processIntegerCallValue(I, Res.first, false);
+ PendingLoads.push_back(Res.second);
+ return true;
+ }
+
+ return false;
+}
+
/// visitUnaryFloatCall - If a call instruction is a unary floating-point
/// operation (as expected), translate it to an SDNode with the specified opcode
/// and return true.
if (visitUnaryFloatCall(I, ISD::FRINT))
return;
break;
+ case LibFunc::round:
+ case LibFunc::roundf:
+ case LibFunc::roundl:
+ if (visitUnaryFloatCall(I, ISD::FROUND))
+ return;
+ break;
case LibFunc::trunc:
case LibFunc::truncf:
case LibFunc::truncl:
if (visitMemCmpCall(I))
return;
break;
+ case LibFunc::memchr:
+ if (visitMemChrCall(I))
+ return;
+ break;
+ case LibFunc::strcpy:
+ if (visitStrCpyCall(I, false))
+ return;
+ break;
+ case LibFunc::stpcpy:
+ if (visitStrCpyCall(I, true))
+ return;
+ break;
+ case LibFunc::strcmp:
+ if (visitStrCmpCall(I))
+ return;
+ break;
+ case LibFunc::strlen:
+ if (visitStrLenCall(I))
+ return;
+ break;
+ case LibFunc::strnlen:
+ if (visitStrNLenCall(I))
+ return;
+ break;
}
}
}
// we can use.
if (OpInfo.AssignedRegs.Regs.empty()) {
LLVMContext &Ctx = *DAG.getContext();
- Ctx.emitError(CS.getInstruction(),
+ Ctx.emitError(CS.getInstruction(),
"couldn't allocate output register for constraint '" +
- Twine(OpInfo.ConstraintCode) + "'");
- break;
+ Twine(OpInfo.ConstraintCode) + "'");
+ return;
}
// If this is an indirect operand, store through the pointer after the
// Add information to the INLINEASM node to know that this register is
// set.
- OpInfo.AssignedRegs.AddInlineAsmOperands(OpInfo.isEarlyClobber ?
- InlineAsm::Kind_RegDefEarlyClobber :
- InlineAsm::Kind_RegDef,
- false,
- 0,
- DAG,
- AsmNodeOperands);
+ OpInfo.AssignedRegs
+ .AddInlineAsmOperands(OpInfo.isEarlyClobber
+ ? InlineAsm::Kind_RegDefEarlyClobber
+ : InlineAsm::Kind_RegDef,
+ false, 0, DAG, AsmNodeOperands);
break;
}
case InlineAsm::isInput: {
if (OpInfo.isIndirect) {
// This happens on gcc/testsuite/gcc.dg/pr8788-1.c
LLVMContext &Ctx = *DAG.getContext();
- Ctx.emitError(CS.getInstruction(), "inline asm not supported yet:"
- " don't know how to handle tied "
- "indirect register inputs");
- report_fatal_error("Cannot handle indirect register inputs!");
+ Ctx.emitError(CS.getInstruction(), "inline asm not supported yet:"
+ " don't know how to handle tied "
+ "indirect register inputs");
+ return;
}
RegsForValue MatchedRegs;
MatchedRegs.Regs.push_back(RegInfo.createVirtualRegister(RC));
else {
LLVMContext &Ctx = *DAG.getContext();
- Ctx.emitError(CS.getInstruction(), "inline asm error: This value"
+ Ctx.emitError(CS.getInstruction(),
+ "inline asm error: This value"
" type register class is not natively supported!");
- report_fatal_error("inline asm error: This value type register "
- "class is not natively supported!");
+ return;
}
}
// Use the produced MatchedRegs object to
LLVMContext &Ctx = *DAG.getContext();
Ctx.emitError(CS.getInstruction(),
"invalid operand for inline asm constraint '" +
- Twine(OpInfo.ConstraintCode) + "'");
- break;
+ Twine(OpInfo.ConstraintCode) + "'");
+ return;
}
// Add information to the INLINEASM node to know about this input.
LLVMContext &Ctx = *DAG.getContext();
Ctx.emitError(CS.getInstruction(),
"Don't know how to handle indirect register inputs yet "
- "for constraint '" + Twine(OpInfo.ConstraintCode) + "'");
- break;
+ "for constraint '" +
+ Twine(OpInfo.ConstraintCode) + "'");
+ return;
}
// Copy the input into the appropriate registers.
if (OpInfo.AssignedRegs.Regs.empty()) {
LLVMContext &Ctx = *DAG.getContext();
- Ctx.emitError(CS.getInstruction(),
+ Ctx.emitError(CS.getInstruction(),
"couldn't allocate input reg for constraint '" +
- Twine(OpInfo.ConstraintCode) + "'");
- break;
+ Twine(OpInfo.ConstraintCode) + "'");
+ return;
}
OpInfo.AssignedRegs.getCopyToRegs(InOperandVal, DAG, getCurSDLoc(),
SDB->setValue(I, Res);
if (!TM.Options.EnableFastISel && Res.getOpcode() == ISD::BUILD_PAIR) {
- if (LoadSDNode *LNode =
+ if (LoadSDNode *LNode =
dyn_cast<LoadSDNode>(Res.getOperand(0).getNode()))
if (FrameIndexSDNode *FI =
dyn_cast<FrameIndexSDNode>(LNode->getBasePtr().getNode()))
ConstantsOut.clear();
}
+
+/// Add a successor MBB to ParentMBB< creating a new MachineBB for BB if SuccMBB
+/// is 0.
+MachineBasicBlock *
+SelectionDAGBuilder::StackProtectorDescriptor::
+AddSuccessorMBB(const BasicBlock *BB,
+ MachineBasicBlock *ParentMBB,
+ MachineBasicBlock *SuccMBB) {
+ // If SuccBB has not been created yet, create it.
+ if (!SuccMBB) {
+ MachineFunction *MF = ParentMBB->getParent();
+ MachineFunction::iterator BBI = ParentMBB;
+ SuccMBB = MF->CreateMachineBasicBlock(BB);
+ MF->insert(++BBI, SuccMBB);
+ }
+ // Add it as a successor of ParentMBB.
+ ParentMBB->addSuccessor(SuccMBB);
+ return SuccMBB;
+}