Update llvm, clang and lldb to 3.7.0 release.

This commit is contained in:
Dimitry Andric 2015-09-06 19:58:48 +00:00
commit b6c25e0ef3
Notes: svn2git 2020-12-20 02:59:44 +00:00
svn path=/projects/clang370-import/; revision=287521
183 changed files with 12571 additions and 1554 deletions

View File

@ -115,7 +115,7 @@ char *LLVMGetTargetMachineCPU(LLVMTargetMachineRef T);
LLVMDisposeMessage. */
char *LLVMGetTargetMachineFeatureString(LLVMTargetMachineRef T);
/** Returns the llvm::DataLayout used for this llvm:TargetMachine. */
/** Deprecated: use LLVMGetDataLayout(LLVMModuleRef M) instead. */
LLVMTargetDataRef LLVMGetTargetMachineData(LLVMTargetMachineRef T);
/** Set the target machine's ASM verbosity. */

View File

@ -315,8 +315,10 @@ class SmallVectorTemplateBase<T, true> : public SmallVectorTemplateCommon<T> {
T2>::value>::type * = nullptr) {
// Use memcpy for PODs iterated by pointers (which includes SmallVector
// iterators): std::uninitialized_copy optimizes to memmove, but we can
// use memcpy here.
memcpy(Dest, I, (E-I)*sizeof(T));
// use memcpy here. Note that I and E are iterators and thus might be
// invalid for memcpy if they are equal.
if (I != E)
memcpy(Dest, I, (E - I) * sizeof(T));
}
/// Double the size of the allocated memory, guaranteeing space for at

View File

@ -158,7 +158,8 @@ class StringMapEntry : public StringMapEntryBase {
// Copy the string information.
char *StrBuffer = const_cast<char*>(NewItem->getKeyData());
memcpy(StrBuffer, Key.data(), KeyLength);
if (KeyLength > 0)
memcpy(StrBuffer, Key.data(), KeyLength);
StrBuffer[KeyLength] = 0; // Null terminate for convenience of clients.
return NewItem;
}

View File

@ -32,11 +32,13 @@ namespace llvm {
class LiveInterval;
class LiveIntervalAnalysis;
class MachineRegisterInfo;
class TargetRegisterInfo;
class VirtRegMap;
class LiveRegMatrix : public MachineFunctionPass {
const TargetRegisterInfo *TRI;
MachineRegisterInfo *MRI;
LiveIntervals *LIS;
VirtRegMap *VRM;

View File

@ -95,8 +95,20 @@ class MachineRegisterInfo {
return MO->Contents.Reg.Next;
}
/// UsedRegUnits - This is a bit vector that is computed and set by the
/// register allocator, and must be kept up to date by passes that run after
/// register allocation (though most don't modify this). This is used
/// so that the code generator knows which callee save registers to save and
/// for other target specific uses.
/// This vector has bits set for register units that are modified in the
/// current function. It doesn't include registers clobbered by function
/// calls with register mask operands.
BitVector UsedRegUnits;
/// UsedPhysRegMask - Additional used physregs including aliases.
/// This bit vector represents all the registers clobbered by function calls.
/// It can model things that UsedRegUnits can't, such as function calls that
/// clobber ymm7 but preserve the low half in xmm7.
BitVector UsedPhysRegMask;
/// ReservedRegs - This is a bit vector of reserved registers. The target
@ -641,12 +653,55 @@ class MachineRegisterInfo {
/// ignored.
bool isPhysRegModified(unsigned PhysReg) const;
//===--------------------------------------------------------------------===//
// Physical Register Use Info
//===--------------------------------------------------------------------===//
/// isPhysRegUsed - Return true if the specified register is used in this
/// function. Also check for clobbered aliases and registers clobbered by
/// function calls with register mask operands.
///
/// This only works after register allocation.
bool isPhysRegUsed(unsigned Reg) const {
if (UsedPhysRegMask.test(Reg))
return true;
for (MCRegUnitIterator Units(Reg, getTargetRegisterInfo());
Units.isValid(); ++Units)
if (UsedRegUnits.test(*Units))
return true;
return false;
}
/// Mark the specified register unit as used in this function.
/// This should only be called during and after register allocation.
void setRegUnitUsed(unsigned RegUnit) {
UsedRegUnits.set(RegUnit);
}
/// setPhysRegUsed - Mark the specified register used in this function.
/// This should only be called during and after register allocation.
void setPhysRegUsed(unsigned Reg) {
for (MCRegUnitIterator Units(Reg, getTargetRegisterInfo());
Units.isValid(); ++Units)
UsedRegUnits.set(*Units);
}
/// addPhysRegsUsedFromRegMask - Mark any registers not in RegMask as used.
/// This corresponds to the bit mask attached to register mask operands.
void addPhysRegsUsedFromRegMask(const uint32_t *RegMask) {
UsedPhysRegMask.setBitsNotInMask(RegMask);
}
/// setPhysRegUnused - Mark the specified register unused in this function.
/// This should only be called during and after register allocation.
void setPhysRegUnused(unsigned Reg) {
UsedPhysRegMask.reset(Reg);
for (MCRegUnitIterator Units(Reg, getTargetRegisterInfo());
Units.isValid(); ++Units)
UsedRegUnits.reset(*Units);
}
//===--------------------------------------------------------------------===//
// Reserved Register Info
//===--------------------------------------------------------------------===//

View File

@ -125,10 +125,15 @@ class TargetMachine {
return *static_cast<const STC*>(getSubtargetImpl(F));
}
/// Deprecated in 3.7, will be removed in 3.8. Use createDataLayout() instead.
///
/// This method returns a pointer to the DataLayout for the target. It should
/// be unchanging for every subtarget.
const DataLayout *getDataLayout() const { return &DL; }
/// Create a DataLayout.
const DataLayout createDataLayout() const { return DL; }
/// \brief Reset the target options based on the function's attributes.
// FIXME: Remove TargetOptions that affect per-function code generation
// from TargetMachine.

View File

@ -206,14 +206,6 @@ static Value *GetLinearExpression(Value *V, APInt &Scale, APInt &Offset,
return V;
}
if (ConstantInt *Const = dyn_cast<ConstantInt>(V)) {
// if it's a constant, just convert it to an offset
// and remove the variable.
Offset += Const->getValue();
assert(Scale == 0 && "Constant values don't have a scale");
return V;
}
if (BinaryOperator *BOp = dyn_cast<BinaryOperator>(V)) {
if (ConstantInt *RHSC = dyn_cast<ConstantInt>(BOp->getOperand(1))) {
switch (BOp->getOpcode()) {
@ -261,10 +253,7 @@ static Value *GetLinearExpression(Value *V, APInt &Scale, APInt &Offset,
Value *Result = GetLinearExpression(CastOp, Scale, Offset, Extension, DL,
Depth + 1, AC, DT);
Scale = Scale.zext(OldWidth);
// We have to sign-extend even if Extension == EK_ZeroExt as we can't
// decompose a sign extension (i.e. zext(x - 1) != zext(x) - zext(-1)).
Offset = Offset.sext(OldWidth);
Offset = Offset.zext(OldWidth);
return Result;
}
@ -1135,43 +1124,12 @@ AliasResult BasicAliasAnalysis::aliasGEP(
}
}
// Try to distinguish something like &A[i][1] against &A[42][0].
// Grab the least significant bit set in any of the scales.
if (!GEP1VariableIndices.empty()) {
uint64_t Modulo = 0;
bool AllPositive = true;
for (unsigned i = 0, e = GEP1VariableIndices.size(); i != e; ++i) {
// Try to distinguish something like &A[i][1] against &A[42][0].
// Grab the least significant bit set in any of the scales. We
// don't need std::abs here (even if the scale's negative) as we'll
// be ^'ing Modulo with itself later.
for (unsigned i = 0, e = GEP1VariableIndices.size(); i != e; ++i)
Modulo |= (uint64_t) GEP1VariableIndices[i].Scale;
if (AllPositive) {
// If the Value could change between cycles, then any reasoning about
// the Value this cycle may not hold in the next cycle. We'll just
// give up if we can't determine conditions that hold for every cycle:
const Value *V = GEP1VariableIndices[i].V;
bool SignKnownZero, SignKnownOne;
ComputeSignBit(const_cast<Value *>(V), SignKnownZero, SignKnownOne, *DL,
0, AC1, nullptr, DT);
// Zero-extension widens the variable, and so forces the sign
// bit to zero.
bool IsZExt = GEP1VariableIndices[i].Extension == EK_ZeroExt;
SignKnownZero |= IsZExt;
SignKnownOne &= !IsZExt;
// If the variable begins with a zero then we know it's
// positive, regardless of whether the value is signed or
// unsigned.
int64_t Scale = GEP1VariableIndices[i].Scale;
AllPositive =
(SignKnownZero && Scale >= 0) ||
(SignKnownOne && Scale < 0);
}
}
Modulo = Modulo ^ (Modulo & (Modulo - 1));
// We can compute the difference between the two addresses
@ -1182,12 +1140,6 @@ AliasResult BasicAliasAnalysis::aliasGEP(
V2Size != MemoryLocation::UnknownSize && ModOffset >= V2Size &&
V1Size <= Modulo - ModOffset)
return NoAlias;
// If we know all the variables are positive, then GEP1 >= GEP1BasePtr.
// If GEP1BasePtr > V2 (GEP1BaseOffset > 0) then we know the pointers
// don't alias if V2Size can fit in the gap between V2 and GEP1BasePtr.
if (AllPositive && GEP1BaseOffset > 0 && V2Size <= (uint64_t) GEP1BaseOffset)
return NoAlias;
}
// Statically, we can see that the base objects are the same, but the

View File

@ -440,31 +440,40 @@ void GlobalsModRef::AnalyzeCallGraph(CallGraph &CG, Module &M) {
}
// Scan the function bodies for explicit loads or stores.
for (unsigned i = 0, e = SCC.size(); i != e && FunctionEffect != ModRef;
++i)
for (inst_iterator II = inst_begin(SCC[i]->getFunction()),
E = inst_end(SCC[i]->getFunction());
II != E && FunctionEffect != ModRef; ++II)
if (LoadInst *LI = dyn_cast<LoadInst>(&*II)) {
FunctionEffect |= Ref;
if (LI->isVolatile())
// Volatile loads may have side-effects, so mark them as writing
// memory (for example, a flag inside the processor).
FunctionEffect |= Mod;
} else if (StoreInst *SI = dyn_cast<StoreInst>(&*II)) {
FunctionEffect |= Mod;
if (SI->isVolatile())
// Treat volatile stores as reading memory somewhere.
FunctionEffect |= Ref;
} else if (isAllocationFn(&*II, TLI) || isFreeCall(&*II, TLI)) {
FunctionEffect |= ModRef;
} else if (IntrinsicInst *Intrinsic = dyn_cast<IntrinsicInst>(&*II)) {
// The callgraph doesn't include intrinsic calls.
Function *Callee = Intrinsic->getCalledFunction();
ModRefBehavior Behaviour = AliasAnalysis::getModRefBehavior(Callee);
FunctionEffect |= (Behaviour & ModRef);
for (auto *Node : SCC) {
if (FunctionEffect == ModRef)
break; // The mod/ref lattice saturates here.
for (Instruction &I : inst_range(Node->getFunction())) {
if (FunctionEffect == ModRef)
break; // The mod/ref lattice saturates here.
// We handle calls specially because the graph-relevant aspects are
// handled above.
if (auto CS = CallSite(&I)) {
if (isAllocationFn(&I, TLI) || isFreeCall(&I, TLI)) {
// FIXME: It is completely unclear why this is necessary and not
// handled by the above graph code.
FunctionEffect |= ModRef;
} else if (Function *Callee = CS.getCalledFunction()) {
// The callgraph doesn't include intrinsic calls.
if (Callee->isIntrinsic()) {
ModRefBehavior Behaviour =
AliasAnalysis::getModRefBehavior(Callee);
FunctionEffect |= (Behaviour & ModRef);
}
}
continue;
}
// All non-call instructions we use the primary predicates for whether
// thay read or write memory.
if (I.mayReadFromMemory())
FunctionEffect |= Ref;
if (I.mayWriteToMemory())
FunctionEffect |= Mod;
}
}
if ((FunctionEffect & Mod) == 0)
++NumReadMemFunctions;
if (FunctionEffect == 0)

View File

@ -3574,18 +3574,9 @@ static Value *SimplifyExtractElementInst(Value *Vec, Value *Idx, const Query &,
// If extracting a specified index from the vector, see if we can recursively
// find a previously computed scalar that was inserted into the vector.
if (auto *IdxC = dyn_cast<ConstantInt>(Idx)) {
unsigned IndexVal = IdxC->getZExtValue();
unsigned VectorWidth = Vec->getType()->getVectorNumElements();
// If this is extracting an invalid index, turn this into undef, to avoid
// crashing the code below.
if (IndexVal >= VectorWidth)
return UndefValue::get(Vec->getType()->getVectorElementType());
if (Value *Elt = findScalarElement(Vec, IndexVal))
if (auto *IdxC = dyn_cast<ConstantInt>(Idx))
if (Value *Elt = findScalarElement(Vec, IdxC->getZExtValue()))
return Elt;
}
return nullptr;
}

View File

@ -374,9 +374,10 @@ InsertPHITranslatedSubExpr(Value *InVal, BasicBlock *CurBB,
if (!Tmp.PHITranslateValue(CurBB, PredBB, &DT, /*MustDominate=*/true))
return Tmp.getAddr();
// If we don't have an available version of this value, it must be an
// instruction.
Instruction *Inst = cast<Instruction>(InVal);
// We don't need to PHI translate values which aren't instructions.
auto *Inst = dyn_cast<Instruction>(InVal);
if (!Inst)
return nullptr;
// Handle cast of PHI translatable value.
if (CastInst *Cast = dyn_cast<CastInst>(Inst)) {

View File

@ -402,8 +402,9 @@ llvm::Value *llvm::findScalarElement(llvm::Value *V, unsigned EltNo) {
if (match(V,
llvm::PatternMatch::m_Add(llvm::PatternMatch::m_Value(Val),
llvm::PatternMatch::m_Constant(Con)))) {
if (Con->getAggregateElement(EltNo)->isNullValue())
return findScalarElement(Val, EltNo);
if (Constant *Elt = Con->getAggregateElement(EltNo))
if (Elt->isNullValue())
return findScalarElement(Val, EltNo);
}
// Otherwise, we don't know.

View File

@ -733,14 +733,12 @@ bool ExeDepsFix::runOnMachineFunction(MachineFunction &mf) {
// If no relevant registers are used in the function, we can skip it
// completely.
bool anyregs = false;
const MachineRegisterInfo &MRI = mf.getRegInfo();
for (TargetRegisterClass::const_iterator I = RC->begin(), E = RC->end();
I != E && !anyregs; ++I)
for (MCRegAliasIterator AI(*I, TRI, true); AI.isValid(); ++AI)
if (!MRI.reg_nodbg_empty(*AI)) {
anyregs = true;
break;
}
I != E; ++I)
if (MF->getRegInfo().isPhysRegUsed(*I)) {
anyregs = true;
break;
}
if (!anyregs) return false;
// Initialize the AliasMap on the first use.

View File

@ -15,12 +15,12 @@
#include "RegisterCoalescer.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/CodeGen/LiveIntervalAnalysis.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/VirtRegMap.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/Format.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetRegisterInfo.h"
#include "llvm/Target/TargetSubtargetInfo.h"
using namespace llvm;
@ -49,6 +49,7 @@ void LiveRegMatrix::getAnalysisUsage(AnalysisUsage &AU) const {
bool LiveRegMatrix::runOnMachineFunction(MachineFunction &MF) {
TRI = MF.getSubtarget().getRegisterInfo();
MRI = &MF.getRegInfo();
LIS = &getAnalysis<LiveIntervals>();
VRM = &getAnalysis<VirtRegMap>();
@ -100,6 +101,7 @@ void LiveRegMatrix::assign(LiveInterval &VirtReg, unsigned PhysReg) {
<< " to " << PrintReg(PhysReg, TRI) << ':');
assert(!VRM->hasPhys(VirtReg.reg) && "Duplicate VirtReg assignment");
VRM->assignVirt2Phys(VirtReg.reg, PhysReg);
MRI->setPhysRegUsed(PhysReg);
foreachUnit(TRI, VirtReg, PhysReg, [&](unsigned Unit,
const LiveRange &Range) {

View File

@ -29,6 +29,7 @@ MachineRegisterInfo::MachineRegisterInfo(const MachineFunction *MF)
TracksSubRegLiveness(false) {
VRegInfo.reserve(256);
RegAllocHints.reserve(256);
UsedRegUnits.resize(getTargetRegisterInfo()->getNumRegUnits());
UsedPhysRegMask.resize(getTargetRegisterInfo()->getNumRegs());
// Create the physreg use/def lists.

View File

@ -624,6 +624,10 @@ struct DataDep {
static bool getDataDeps(const MachineInstr *UseMI,
SmallVectorImpl<DataDep> &Deps,
const MachineRegisterInfo *MRI) {
// Debug values should not be included in any calculations.
if (UseMI->isDebugValue())
return false;
bool HasPhysRegs = false;
for (MachineInstr::const_mop_iterator I = UseMI->operands_begin(),
E = UseMI->operands_end(); I != E; ++I) {

View File

@ -1026,8 +1026,12 @@ PEI::scavengeFrameVirtualRegs(MachineFunction &Fn) {
// Replace this reference to the virtual register with the
// scratch register.
assert (ScratchReg && "Missing scratch register!");
MachineRegisterInfo &MRI = Fn.getRegInfo();
Fn.getRegInfo().replaceRegWith(Reg, ScratchReg);
// Make sure MRI now accounts this register as used.
MRI.setPhysRegUsed(ScratchReg);
// Because this instruction was processed by the RS before this
// register was allocated, make sure that the RS now records the
// register as being used.

View File

@ -986,6 +986,10 @@ void RAFast::AllocateBasicBlock() {
}
}
for (UsedInInstrSet::iterator
I = UsedInInstr.begin(), E = UsedInInstr.end(); I != E; ++I)
MRI->setRegUnitUsed(*I);
// Track registers defined by instruction - early clobbers and tied uses at
// this point.
UsedInInstr.clear();
@ -1046,6 +1050,10 @@ void RAFast::AllocateBasicBlock() {
killVirtReg(VirtDead[i]);
VirtDead.clear();
for (UsedInInstrSet::iterator
I = UsedInInstr.begin(), E = UsedInInstr.end(); I != E; ++I)
MRI->setRegUnitUsed(*I);
if (CopyDst && CopyDst == CopySrc && CopyDstSub == CopySrcSub) {
DEBUG(dbgs() << "-- coalescing: " << *MI);
Coalesced.push_back(MI);
@ -1095,6 +1103,12 @@ bool RAFast::runOnMachineFunction(MachineFunction &Fn) {
AllocateBasicBlock();
}
// Add the clobber lists for all the instructions we skipped earlier.
for (const MCInstrDesc *Desc : SkippedInstrs)
if (const uint16_t *Defs = Desc->getImplicitDefs())
while (*Defs)
MRI->setPhysRegUsed(*Defs++);
// All machine operands and other references to virtual registers have been
// replaced. Remove the virtual registers.
MRI->clearVirtRegs();

View File

@ -1531,6 +1531,14 @@ bool RegisterCoalescer::joinReservedPhysReg(CoalescerPair &CP) {
DEBUG(dbgs() << "\t\tInterference (read): " << *MI);
return false;
}
// We must also check for clobbers caused by regmasks.
for (const auto &MO : MI->operands()) {
if (MO.isRegMask() && MO.clobbersPhysReg(DstReg)) {
DEBUG(dbgs() << "\t\tInterference (regmask clobber): " << *MI);
return false;
}
}
}
// We're going to remove the copy which defines a physical reserved

View File

@ -8365,12 +8365,12 @@ SDValue DAGCombiner::visitFDIV(SDNode *N) {
if (N0CFP && N0CFP->isExactlyValue(1.0))
return SDValue();
SmallVector<SDNode *, 4> Users;
// Find all FDIV users of the same divisor.
for (auto *U : N1->uses()) {
// Use a set because duplicates may be present in the user list.
SetVector<SDNode *> Users;
for (auto *U : N1->uses())
if (U->getOpcode() == ISD::FDIV && U->getOperand(1) == N1)
Users.push_back(U);
}
Users.insert(U);
if (TLI.combineRepeatedFPDivisors(Users.size())) {
SDValue FPOne = DAG.getConstantFP(1.0, DL, VT);

View File

@ -163,6 +163,7 @@ class VirtRegRewriter : public MachineFunctionPass {
SlotIndexes *Indexes;
LiveIntervals *LIS;
VirtRegMap *VRM;
SparseSet<unsigned> PhysRegs;
void rewrite();
void addMBBLiveIns();
@ -318,15 +319,54 @@ void VirtRegRewriter::rewrite() {
SmallVector<unsigned, 8> SuperDeads;
SmallVector<unsigned, 8> SuperDefs;
SmallVector<unsigned, 8> SuperKills;
SmallPtrSet<const MachineInstr *, 4> NoReturnInsts;
// Here we have a SparseSet to hold which PhysRegs are actually encountered
// in the MF we are about to iterate over so that later when we call
// setPhysRegUsed, we are only doing it for physRegs that were actually found
// in the program and not for all of the possible physRegs for the given
// target architecture. If the target has a lot of physRegs, then for a small
// program there will be a significant compile time reduction here.
PhysRegs.clear();
PhysRegs.setUniverse(TRI->getNumRegs());
// The function with uwtable should guarantee that the stack unwinder
// can unwind the stack to the previous frame. Thus, we can't apply the
// noreturn optimization if the caller function has uwtable attribute.
bool HasUWTable = MF->getFunction()->hasFnAttribute(Attribute::UWTable);
for (MachineFunction::iterator MBBI = MF->begin(), MBBE = MF->end();
MBBI != MBBE; ++MBBI) {
DEBUG(MBBI->print(dbgs(), Indexes));
bool IsExitBB = MBBI->succ_empty();
for (MachineBasicBlock::instr_iterator
MII = MBBI->instr_begin(), MIE = MBBI->instr_end(); MII != MIE;) {
MachineInstr *MI = MII;
++MII;
// Check if this instruction is a call to a noreturn function. If this
// is a call to noreturn function and we don't need the stack unwinding
// functionality (i.e. this function does not have uwtable attribute and
// the callee function has the nounwind attribute), then we can ignore
// the definitions set by this instruction.
if (!HasUWTable && IsExitBB && MI->isCall()) {
for (MachineInstr::mop_iterator MOI = MI->operands_begin(),
MOE = MI->operands_end(); MOI != MOE; ++MOI) {
MachineOperand &MO = *MOI;
if (!MO.isGlobal())
continue;
const Function *Func = dyn_cast<Function>(MO.getGlobal());
if (!Func || !Func->hasFnAttribute(Attribute::NoReturn) ||
// We need to keep correct unwind information
// even if the function will not return, since the
// runtime may need it.
!Func->hasFnAttribute(Attribute::NoUnwind))
continue;
NoReturnInsts.insert(MI);
break;
}
}
for (MachineInstr::mop_iterator MOI = MI->operands_begin(),
MOE = MI->operands_end(); MOI != MOE; ++MOI) {
MachineOperand &MO = *MOI;
@ -335,6 +375,15 @@ void VirtRegRewriter::rewrite() {
if (MO.isRegMask())
MRI->addPhysRegsUsedFromRegMask(MO.getRegMask());
// If we encounter a VirtReg or PhysReg then get at the PhysReg and add
// it to the physreg bitset. Later we use only the PhysRegs that were
// actually encountered in the MF to populate the MRI's used physregs.
if (MO.isReg() && MO.getReg())
PhysRegs.insert(
TargetRegisterInfo::isVirtualRegister(MO.getReg()) ?
VRM->getPhys(MO.getReg()) :
MO.getReg());
if (!MO.isReg() || !TargetRegisterInfo::isVirtualRegister(MO.getReg()))
continue;
unsigned VirtReg = MO.getReg();
@ -421,5 +470,29 @@ void VirtRegRewriter::rewrite() {
}
}
}
// Tell MRI about physical registers in use.
if (NoReturnInsts.empty()) {
for (SparseSet<unsigned>::iterator
RegI = PhysRegs.begin(), E = PhysRegs.end(); RegI != E; ++RegI)
if (!MRI->reg_nodbg_empty(*RegI))
MRI->setPhysRegUsed(*RegI);
} else {
for (SparseSet<unsigned>::iterator
I = PhysRegs.begin(), E = PhysRegs.end(); I != E; ++I) {
unsigned Reg = *I;
if (MRI->reg_nodbg_empty(Reg))
continue;
// Check if this register has a use that will impact the rest of the
// code. Uses in debug and noreturn instructions do not impact the
// generated code.
for (MachineInstr &It : MRI->reg_nodbg_instructions(Reg)) {
if (!NoReturnInsts.count(&It)) {
MRI->setPhysRegUsed(Reg);
break;
}
}
}
}
}

View File

@ -180,10 +180,17 @@ uint64_t ExecutionEngineState::RemoveMapping(StringRef Name) {
}
std::string ExecutionEngine::getMangledName(const GlobalValue *GV) {
assert(GV->hasName() && "Global must have name.");
MutexGuard locked(lock);
Mangler Mang;
SmallString<128> FullName;
Mang.getNameWithPrefix(FullName, GV, false);
const DataLayout &DL =
GV->getParent()->getDataLayout().isDefault()
? *getDataLayout()
: GV->getParent()->getDataLayout();
Mangler::getNameWithPrefix(FullName, GV->getName(), DL);
return FullName.str();
}

View File

@ -266,6 +266,12 @@ void MCJIT::finalizeModule(Module *M) {
RuntimeDyld::SymbolInfo MCJIT::findExistingSymbol(const std::string &Name) {
SmallString<128> FullName;
Mangler::getNameWithPrefix(FullName, Name, *TM->getDataLayout());
if (void *Addr = getPointerToGlobalIfAvailable(FullName))
return RuntimeDyld::SymbolInfo(static_cast<uint64_t>(
reinterpret_cast<uintptr_t>(Addr)),
JITSymbolFlags::Exported);
return Dyld.getSymbol(FullName);
}

View File

@ -98,7 +98,7 @@ void RTDyldMemoryManager::registerEHFrames(uint8_t *Addr,
uint64_t LoadAddr,
size_t Size) {
// On OS X OS X __register_frame takes a single FDE as an argument.
// See http://lists.cs.uiuc.edu/pipermail/llvmdev/2013-April/061768.html
// See http://lists.llvm.org/pipermail/llvm-dev/2013-April/061768.html
const char *P = (const char *)Addr;
const char *End = P + Size;
do {

View File

@ -613,6 +613,9 @@ bool StructType::isLayoutIdentical(StructType *Other) const {
if (isPacked() != Other->isPacked() ||
getNumElements() != Other->getNumElements())
return false;
if (!getNumElements())
return true;
return std::equal(element_begin(), element_end(), Other->element_begin());
}

View File

@ -57,7 +57,8 @@ void MemoryBuffer::init(const char *BufStart, const char *BufEnd,
/// CopyStringRef - Copies contents of a StringRef into a block of memory and
/// null-terminates it.
static void CopyStringRef(char *Memory, StringRef Data) {
memcpy(Memory, Data.data(), Data.size());
if (!Data.empty())
memcpy(Memory, Data.data(), Data.size());
Memory[Data.size()] = 0; // Null terminate string.
}

View File

@ -593,6 +593,7 @@ bool AArch64A57FPLoadBalancing::colorChain(Chain *G, Color C,
if (Change) {
Substs[MO.getReg()] = Reg;
MO.setReg(Reg);
MRI->setPhysRegUsed(Reg);
Changed = true;
}

View File

@ -354,6 +354,7 @@ void AArch64FrameLowering::emitPrologue(MachineFunction &MF,
if (NumBytes && NeedsRealignment) {
// Use the first callee-saved register as a scratch register.
scratchSPReg = AArch64::X9;
MF.getRegInfo().setPhysRegUsed(scratchSPReg);
}
// If we're a leaf function, try using the red zone.

View File

@ -123,6 +123,11 @@ def FeatureSGPRInitBug : SubtargetFeature<"sgpr-init-bug",
"true",
"VI SGPR initilization bug requiring a fixed SGPR allocation size">;
def FeatureEnableHugeScratchBuffer : SubtargetFeature<"huge-scratch-buffer",
"EnableHugeScratchBuffer",
"true",
"Enable scratch buffer sizes greater than 128 GB">;
class SubtargetFeatureFetchLimit <string Value> :
SubtargetFeature <"fetch"#Value,
"TexVTXClauseSize",

View File

@ -1029,6 +1029,10 @@ bool AMDGPUDAGToDAGISel::SelectMUBUFAddr64(SDValue Addr, SDValue &SRsrc,
SDValue &SLC, SDValue &TFE) const {
SDValue Ptr, Offen, Idxen, Addr64;
// addr64 bit was removed for volcanic islands.
if (Subtarget->getGeneration() >= AMDGPUSubtarget::VOLCANIC_ISLANDS)
return false;
SelectMUBUF(Addr, Ptr, VAddr, SOffset, Offset, Offen, Idxen, Addr64,
GLC, SLC, TFE);
@ -1095,13 +1099,16 @@ bool AMDGPUDAGToDAGISel::SelectMUBUFScratch(SDValue Addr, SDValue &Rsrc,
// (add n0, c1)
if (CurDAG->isBaseWithConstantOffset(Addr)) {
SDValue N0 = Addr.getOperand(0);
SDValue N1 = Addr.getOperand(1);
ConstantSDNode *C1 = cast<ConstantSDNode>(N1);
if (isLegalMUBUFImmOffset(C1)) {
VAddr = Addr.getOperand(0);
ImmOffset = CurDAG->getTargetConstant(C1->getZExtValue(), DL, MVT::i16);
return true;
// Offsets in vaddr must be positive.
if (CurDAG->SignBitIsZero(N0)) {
ConstantSDNode *C1 = cast<ConstantSDNode>(N1);
if (isLegalMUBUFImmOffset(C1)) {
VAddr = N0;
ImmOffset = CurDAG->getTargetConstant(C1->getZExtValue(), DL, MVT::i16);
return true;
}
}
}

View File

@ -73,7 +73,7 @@ AMDGPUSubtarget::AMDGPUSubtarget(const Triple &TT, StringRef GPU, StringRef FS,
WavefrontSize(0), CFALUBug(false), LocalMemorySize(0),
EnableVGPRSpilling(false), SGPRInitBug(false), IsGCN(false),
GCN1Encoding(false), GCN3Encoding(false), CIInsts(false), LDSBankCount(0),
IsaVersion(ISAVersion0_0_0),
IsaVersion(ISAVersion0_0_0), EnableHugeScratchBuffer(false),
FrameLowering(TargetFrameLowering::StackGrowsUp,
64 * 16, // Maximum stack alignment (long16)
0),

View File

@ -89,6 +89,7 @@ class AMDGPUSubtarget : public AMDGPUGenSubtargetInfo {
bool FeatureDisable;
int LDSBankCount;
unsigned IsaVersion;
bool EnableHugeScratchBuffer;
AMDGPUFrameLowering FrameLowering;
std::unique_ptr<AMDGPUTargetLowering> TLInfo;
@ -271,6 +272,10 @@ class AMDGPUSubtarget : public AMDGPUGenSubtargetInfo {
return DevName;
}
bool enableHugeScratchBuffer() const {
return EnableHugeScratchBuffer;
}
bool dumpCode() const {
return DumpCode;
}

View File

@ -1719,7 +1719,6 @@ MachineBasicBlock *
AMDGPUCFGStructurizer::normalizeInfiniteLoopExit(MachineLoop* LoopRep) {
MachineBasicBlock *LoopHeader = LoopRep->getHeader();
MachineBasicBlock *LoopLatch = LoopRep->getLoopLatch();
const TargetRegisterClass * I32RC = TRI->getCFGStructurizerRegClass(MVT::i32);
if (!LoopHeader || !LoopLatch)
return nullptr;
@ -1732,18 +1731,9 @@ AMDGPUCFGStructurizer::normalizeInfiniteLoopExit(MachineLoop* LoopRep) {
FuncRep->push_back(DummyExitBlk); //insert to function
SHOWNEWBLK(DummyExitBlk, "DummyExitBlock to normalize infiniteLoop: ");
DEBUG(dbgs() << "Old branch instr: " << *BranchMI << "\n";);
MachineBasicBlock::iterator I = BranchMI;
unsigned ImmReg = FuncRep->getRegInfo().createVirtualRegister(I32RC);
llvm_unreachable("Extra register needed to handle CFG");
MachineInstr *NewMI = insertInstrBefore(I, AMDGPU::BRANCH_COND_i32);
MachineInstrBuilder MIB(*FuncRep, NewMI);
MIB.addMBB(LoopHeader);
MIB.addReg(ImmReg, false);
SHOWNEWINSTR(NewMI);
BranchMI->eraseFromParent();
LoopLatch->addSuccessor(DummyExitBlk);
return DummyExitBlk;
LLVMContext &Ctx = LoopHeader->getParent()->getFunction()->getContext();
Ctx.emitError("Extra register needed to handle CFG");
return nullptr;
}
void AMDGPUCFGStructurizer::removeUnconditionalBranch(MachineBasicBlock *MBB) {

View File

@ -138,3 +138,7 @@ def : ProcessorModel<"iceland", SIQuarterSpeedModel,
def : ProcessorModel<"carrizo", SIQuarterSpeedModel,
[FeatureVolcanicIslands, FeatureISAVersion8_0_1]
>;
def : ProcessorModel<"fiji", SIQuarterSpeedModel,
[FeatureVolcanicIslands, FeatureISAVersion8_0_1]
>;

View File

@ -254,6 +254,12 @@ bool SITargetLowering::isShuffleMaskLegal(const SmallVectorImpl<int> &,
return false;
}
bool SITargetLowering::isLegalFlatAddressingMode(const AddrMode &AM) const {
// Flat instructions do not have offsets, and only have the register
// address.
return AM.BaseOffs == 0 && (AM.Scale == 0 || AM.Scale == 1);
}
bool SITargetLowering::isLegalAddressingMode(const DataLayout &DL,
const AddrMode &AM, Type *Ty,
unsigned AS) const {
@ -263,8 +269,21 @@ bool SITargetLowering::isLegalAddressingMode(const DataLayout &DL,
switch (AS) {
case AMDGPUAS::GLOBAL_ADDRESS:
case AMDGPUAS::CONSTANT_ADDRESS: // XXX - Should we assume SMRD instructions?
if (Subtarget->getGeneration() >= AMDGPUSubtarget::VOLCANIC_ISLANDS) {
// Assume the we will use FLAT for all global memory accesses
// on VI.
// FIXME: This assumption is currently wrong. On VI we still use
// MUBUF instructions for the r + i addressing mode. As currently
// implemented, the MUBUF instructions only work on buffer < 4GB.
// It may be possible to support > 4GB buffers with MUBUF instructions,
// by setting the stride value in the resource descriptor which would
// increase the size limit to (stride * 4GB). However, this is risky,
// because it has never been validated.
return isLegalFlatAddressingMode(AM);
}
// fall-through
case AMDGPUAS::PRIVATE_ADDRESS:
case AMDGPUAS::CONSTANT_ADDRESS: // XXX - Should we assume SMRD instructions?
case AMDGPUAS::UNKNOWN_ADDRESS_SPACE: {
// MUBUF / MTBUF instructions have a 12-bit unsigned byte offset, and
// additionally can do r + r + i with addr64. 32-bit has more addressing
@ -324,11 +343,9 @@ bool SITargetLowering::isLegalAddressingMode(const DataLayout &DL,
return false;
}
case AMDGPUAS::FLAT_ADDRESS: {
// Flat instructions do not have offsets, and only have the register
// address.
return AM.BaseOffs == 0 && (AM.Scale == 0 || AM.Scale == 1);
}
case AMDGPUAS::FLAT_ADDRESS:
return isLegalFlatAddressingMode(AM);
default:
llvm_unreachable("unhandled address space");
}
@ -812,10 +829,29 @@ static SDNode *findUser(SDValue Value, unsigned Opcode) {
SDValue SITargetLowering::LowerFrameIndex(SDValue Op, SelectionDAG &DAG) const {
SDLoc SL(Op);
FrameIndexSDNode *FINode = cast<FrameIndexSDNode>(Op);
unsigned FrameIndex = FINode->getIndex();
return DAG.getTargetFrameIndex(FrameIndex, MVT::i32);
// A FrameIndex node represents a 32-bit offset into scratch memory. If
// the high bit of a frame index offset were to be set, this would mean
// that it represented an offset of ~2GB * 64 = ~128GB from the start of the
// scratch buffer, with 64 being the number of threads per wave.
//
// If we know the machine uses less than 128GB of scratch, then we can
// amrk the high bit of the FrameIndex node as known zero,
// which is important, because it means in most situations we can
// prove that values derived from FrameIndex nodes are non-negative.
// This enables us to take advantage of more addressing modes when
// accessing scratch buffers, since for scratch reads/writes, the register
// offset must always be positive.
SDValue TFI = DAG.getTargetFrameIndex(FrameIndex, MVT::i32);
if (Subtarget->enableHugeScratchBuffer())
return TFI;
return DAG.getNode(ISD::AssertZext, SL, MVT::i32, TFI,
DAG.getValueType(EVT::getIntegerVT(*DAG.getContext(), 31)));
}
/// This transforms the control flow intrinsics to get the branch destination as
@ -2034,6 +2070,13 @@ void SITargetLowering::adjustWritemask(MachineSDNode *&Node,
}
}
static bool isFrameIndexOp(SDValue Op) {
if (Op.getOpcode() == ISD::AssertZext)
Op = Op.getOperand(0);
return isa<FrameIndexSDNode>(Op);
}
/// \brief Legalize target independent instructions (e.g. INSERT_SUBREG)
/// with frame index operands.
/// LLVM assumes that inputs are to these instructions are registers.
@ -2042,7 +2085,7 @@ void SITargetLowering::legalizeTargetIndependentNode(SDNode *Node,
SmallVector<SDValue, 8> Ops;
for (unsigned i = 0; i < Node->getNumOperands(); ++i) {
if (!isa<FrameIndexSDNode>(Node->getOperand(i))) {
if (!isFrameIndexOp(Node->getOperand(i))) {
Ops.push_back(Node->getOperand(i));
continue;
}

View File

@ -56,6 +56,7 @@ class SITargetLowering : public AMDGPUTargetLowering {
SDValue performMin3Max3Combine(SDNode *N, DAGCombinerInfo &DCI) const;
SDValue performSetCCCombine(SDNode *N, DAGCombinerInfo &DCI) const;
bool isLegalFlatAddressingMode(const AddrMode &AM) const;
public:
SITargetLowering(TargetMachine &tm, const AMDGPUSubtarget &STI);

View File

@ -1600,12 +1600,14 @@ multiclass VOPC_m <vopc op, dag outs, dag ins, string asm, list<dag> pattern,
SIMCInstr <opName#"_e32", SISubtarget.SI> {
let Defs = !if(DefExec, [EXEC], []);
let hasSideEffects = DefExec;
let AssemblerPredicates = [isSICI];
}
def _vi : VOPC<op.VI, ins, asm, []>,
SIMCInstr <opName#"_e32", SISubtarget.VI> {
let Defs = !if(DefExec, [EXEC], []);
let hasSideEffects = DefExec;
let AssemblerPredicates = [isVI];
}
}

View File

@ -2910,9 +2910,6 @@ defm : MUBUFLoad_Pattern <BUFFER_LOAD_SBYTE_ADDR64, i32, sextloadi8_constant>;
defm : MUBUFLoad_Pattern <BUFFER_LOAD_UBYTE_ADDR64, i32, az_extloadi8_constant>;
defm : MUBUFLoad_Pattern <BUFFER_LOAD_SSHORT_ADDR64, i32, sextloadi16_constant>;
defm : MUBUFLoad_Pattern <BUFFER_LOAD_USHORT_ADDR64, i32, az_extloadi16_constant>;
defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORD_ADDR64, i32, constant_load>;
defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORDX2_ADDR64, v2i32, constant_load>;
defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORDX4_ADDR64, v4i32, constant_load>;
} // End Predicates = [isSICI]
class MUBUFScratchLoadPat <MUBUF Instr, ValueType vt, PatFrag ld> : Pat <
@ -3273,13 +3270,13 @@ def : Pat <
(f64 (fadd (f64 (VOP3Mods f64:$x, i32:$mods)),
(f64 (fneg (f64 (ffloor (f64 (VOP3Mods f64:$x, i32:$mods)))))))),
(V_CNDMASK_B64_PSEUDO
$x,
(V_MIN_F64
SRCMODS.NONE,
(V_FRACT_F64_e64 $mods, $x, DSTCLAMP.NONE, DSTOMOD.NONE),
SRCMODS.NONE,
(V_MOV_B64_PSEUDO 0x3fefffffffffffff),
DSTCLAMP.NONE, DSTOMOD.NONE),
$x,
(V_CMP_CLASS_F64_e64 SRCMODS.NONE, $x, 3/*NaN*/))
>;
@ -3291,13 +3288,13 @@ def : Pat <
$x,
SRCMODS.NEG,
(V_CNDMASK_B64_PSEUDO
$x,
(V_MIN_F64
SRCMODS.NONE,
(V_FRACT_F64_e64 $mods, $x, DSTCLAMP.NONE, DSTOMOD.NONE),
SRCMODS.NONE,
(V_MOV_B64_PSEUDO 0x3fefffffffffffff),
DSTCLAMP.NONE, DSTOMOD.NONE),
$x,
(V_CMP_CLASS_F64_e64 SRCMODS.NONE, $x, 3/*NaN*/)),
DSTCLAMP.NONE, DSTOMOD.NONE)
>;

View File

@ -53,6 +53,7 @@ SIMachineFunctionInfo::SpilledReg SIMachineFunctionInfo::getSpilledReg(
if (!LaneVGPRs.count(LaneVGPRIdx)) {
unsigned LaneVGPR = TRI->findUnusedRegister(MRI, &AMDGPU::VGPR_32RegClass);
LaneVGPRs[LaneVGPRIdx] = LaneVGPR;
MRI.setPhysRegUsed(LaneVGPR);
// Add this register as live-in to all blocks to avoid machine verifer
// complaining about use of an undefined physical register.

View File

@ -91,6 +91,7 @@ bool SIPrepareScratchRegs::runOnMachineFunction(MachineFunction &MF) {
if (ScratchOffsetReg != AMDGPU::NoRegister) {
// Found an SGPR to use
MRI.setPhysRegUsed(ScratchOffsetReg);
BuildMI(*Entry, I, DL, TII->get(AMDGPU::S_MOV_B32), ScratchOffsetReg)
.addReg(ScratchOffsetPreloadReg);
} else {

View File

@ -348,7 +348,8 @@ const TargetRegisterClass *SIRegisterInfo::getPhysRegClass(unsigned Reg) const {
&AMDGPU::SReg_128RegClass,
&AMDGPU::VReg_256RegClass,
&AMDGPU::SReg_256RegClass,
&AMDGPU::VReg_512RegClass
&AMDGPU::VReg_512RegClass,
&AMDGPU::SReg_512RegClass
};
for (const TargetRegisterClass *BaseClass : BaseClasses) {
@ -499,7 +500,7 @@ unsigned SIRegisterInfo::findUnusedRegister(const MachineRegisterInfo &MRI,
for (TargetRegisterClass::iterator I = RC->begin(), E = RC->end();
I != E; ++I) {
if (MRI.reg_nodbg_empty(*I))
if (!MRI.isPhysRegUsed(*I))
return *I;
}
return AMDGPU::NoRegister;

View File

@ -103,4 +103,46 @@ def : Pat <
(S_BUFFER_LOAD_DWORD_IMM $sbase, (as_i32imm $offset))
>;
// Patterns for global loads with no offset
class FlatLoadPat <FLAT inst, SDPatternOperator node, ValueType vt> : Pat <
(vt (node i64:$addr)),
(inst $addr, 0, 0, 0)
>;
def : FlatLoadPat <FLAT_LOAD_UBYTE, az_extloadi8_global, i32>;
def : FlatLoadPat <FLAT_LOAD_SBYTE, sextloadi8_global, i32>;
def : FlatLoadPat <FLAT_LOAD_USHORT, az_extloadi16_global, i32>;
def : FlatLoadPat <FLAT_LOAD_SSHORT, sextloadi16_global, i32>;
def : FlatLoadPat <FLAT_LOAD_DWORD, global_load, i32>;
def : FlatLoadPat <FLAT_LOAD_DWORDX2, global_load, v2i32>;
def : FlatLoadPat <FLAT_LOAD_DWORDX4, global_load, v4i32>;
class FlatStorePat <FLAT inst, SDPatternOperator node, ValueType vt> : Pat <
(node vt:$data, i64:$addr),
(inst $data, $addr, 0, 0, 0)
>;
def : FlatStorePat <FLAT_STORE_BYTE, truncstorei8_global, i32>;
def : FlatStorePat <FLAT_STORE_SHORT, truncstorei16_global, i32>;
def : FlatStorePat <FLAT_STORE_DWORD, global_store, i32>;
def : FlatStorePat <FLAT_STORE_DWORDX2, global_store, v2i32>;
def : FlatStorePat <FLAT_STORE_DWORDX4, global_store, v4i32>;
class FlatAtomicPat <FLAT inst, SDPatternOperator node, ValueType vt> : Pat <
(vt (node i64:$addr, vt:$data)),
(inst $addr, $data, 0, 0)
>;
def : FlatAtomicPat <FLAT_ATOMIC_ADD_RTN, atomic_add_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_AND_RTN, atomic_and_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_SUB_RTN, atomic_sub_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_SMAX_RTN, atomic_max_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_UMAX_RTN, atomic_umax_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_SMIN_RTN, atomic_min_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_UMIN_RTN, atomic_umin_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_OR_RTN, atomic_or_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_SWAP_RTN, atomic_swap_global, i32>;
def : FlatAtomicPat <FLAT_ATOMIC_XOR_RTN, atomic_xor_global, i32>;
} // End Predicates = [isVI]

View File

@ -4583,6 +4583,12 @@ static SDValue LowerVSETCC(SDValue Op, SelectionDAG &DAG) {
ISD::CondCode SetCCOpcode = cast<CondCodeSDNode>(CC)->get();
SDLoc dl(Op);
if (CmpVT.getVectorElementType() == MVT::i64)
// 64-bit comparisons are not legal. We've marked SETCC as non-Custom,
// but it's possible that our operands are 64-bit but our result is 32-bit.
// Bail in this case.
return SDValue();
if (Op1.getValueType().isFloatingPoint()) {
switch (SetCCOpcode) {
default: llvm_unreachable("Illegal FP comparison");

View File

@ -118,7 +118,6 @@ namespace {
};
SpecificBumpPtrAllocator<MergeCandidate> Allocator;
SmallVector<const MergeCandidate*,4> Candidates;
SmallVector<MachineInstr*,4> MergeBaseCandidates;
void moveLiveRegsBefore(const MachineBasicBlock &MBB,
MachineBasicBlock::const_iterator Before);
@ -141,7 +140,6 @@ namespace {
MachineBasicBlock::iterator &MBBI);
bool MergeBaseUpdateLoadStore(MachineInstr *MI);
bool MergeBaseUpdateLSMultiple(MachineInstr *MI);
bool MergeBaseUpdateLSDouble(MachineInstr &MI) const;
bool LoadStoreMultipleOpti(MachineBasicBlock &MBB);
bool MergeReturnIntoLDM(MachineBasicBlock &MBB);
};
@ -933,6 +931,11 @@ void ARMLoadStoreOpt::FormCandidates(const MemOpQueue &MemOps) {
if (STI->isSwift() && !isNotVFP && (PRegNum % 2) == 1)
CanMergeToLSMulti = false;
// LDRD/STRD do not allow SP/PC. LDM/STM do not support it or have it
// deprecated; LDM to PC is fine but cannot happen here.
if (PReg == ARM::SP || PReg == ARM::PC)
CanMergeToLSMulti = CanMergeToLSDouble = false;
// Merge following instructions where possible.
for (unsigned I = SIndex+1; I < EIndex; ++I, ++Count) {
int NewOffset = MemOps[I].Offset;
@ -940,16 +943,15 @@ void ARMLoadStoreOpt::FormCandidates(const MemOpQueue &MemOps) {
break;
const MachineOperand &MO = getLoadStoreRegOp(*MemOps[I].MI);
unsigned Reg = MO.getReg();
unsigned RegNum = MO.isUndef() ? UINT_MAX : TRI->getEncodingValue(Reg);
if (Reg == ARM::SP || Reg == ARM::PC)
break;
// See if the current load/store may be part of a multi load/store.
unsigned RegNum = MO.isUndef() ? UINT_MAX : TRI->getEncodingValue(Reg);
bool PartOfLSMulti = CanMergeToLSMulti;
if (PartOfLSMulti) {
// Cannot load from SP
if (Reg == ARM::SP)
PartOfLSMulti = false;
// Register numbers must be in ascending order.
else if (RegNum <= PRegNum)
if (RegNum <= PRegNum)
PartOfLSMulti = false;
// For VFP / NEON load/store multiples, the registers must be
// consecutive and within the limit on the number of registers per
@ -993,6 +995,76 @@ void ARMLoadStoreOpt::FormCandidates(const MemOpQueue &MemOps) {
} while (SIndex < EIndex);
}
static bool isMatchingDecrement(MachineInstr *MI, unsigned Base,
unsigned Bytes, unsigned Limit,
ARMCC::CondCodes Pred, unsigned PredReg) {
unsigned MyPredReg = 0;
if (!MI)
return false;
bool CheckCPSRDef = false;
switch (MI->getOpcode()) {
default: return false;
case ARM::tSUBi8:
case ARM::t2SUBri:
case ARM::SUBri:
CheckCPSRDef = true;
break;
case ARM::tSUBspi:
break;
}
// Make sure the offset fits in 8 bits.
if (Bytes == 0 || (Limit && Bytes >= Limit))
return false;
unsigned Scale = (MI->getOpcode() == ARM::tSUBspi ||
MI->getOpcode() == ARM::tSUBi8) ? 4 : 1; // FIXME
if (!(MI->getOperand(0).getReg() == Base &&
MI->getOperand(1).getReg() == Base &&
(MI->getOperand(2).getImm() * Scale) == Bytes &&
getInstrPredicate(MI, MyPredReg) == Pred &&
MyPredReg == PredReg))
return false;
return CheckCPSRDef ? !definesCPSR(MI) : true;
}
static bool isMatchingIncrement(MachineInstr *MI, unsigned Base,
unsigned Bytes, unsigned Limit,
ARMCC::CondCodes Pred, unsigned PredReg) {
unsigned MyPredReg = 0;
if (!MI)
return false;
bool CheckCPSRDef = false;
switch (MI->getOpcode()) {
default: return false;
case ARM::tADDi8:
case ARM::t2ADDri:
case ARM::ADDri:
CheckCPSRDef = true;
break;
case ARM::tADDspi:
break;
}
if (Bytes == 0 || (Limit && Bytes >= Limit))
// Make sure the offset fits in 8 bits.
return false;
unsigned Scale = (MI->getOpcode() == ARM::tADDspi ||
MI->getOpcode() == ARM::tADDi8) ? 4 : 1; // FIXME
if (!(MI->getOperand(0).getReg() == Base &&
MI->getOperand(1).getReg() == Base &&
(MI->getOperand(2).getImm() * Scale) == Bytes &&
getInstrPredicate(MI, MyPredReg) == Pred &&
MyPredReg == PredReg))
return false;
return CheckCPSRDef ? !definesCPSR(MI) : true;
}
static unsigned getUpdatingLSMultipleOpcode(unsigned Opc,
ARM_AM::AMSubMode Mode) {
switch (Opc) {
@ -1060,75 +1132,6 @@ static unsigned getUpdatingLSMultipleOpcode(unsigned Opc,
}
}
/// Check if the given instruction increments or decrements a register and
/// return the amount it is incremented/decremented. Returns 0 if the CPSR flags
/// generated by the instruction are possibly read as well.
static int isIncrementOrDecrement(const MachineInstr &MI, unsigned Reg,
ARMCC::CondCodes Pred, unsigned PredReg) {
bool CheckCPSRDef;
int Scale;
switch (MI.getOpcode()) {
case ARM::tADDi8: Scale = 4; CheckCPSRDef = true; break;
case ARM::tSUBi8: Scale = -4; CheckCPSRDef = true; break;
case ARM::t2SUBri:
case ARM::SUBri: Scale = -1; CheckCPSRDef = true; break;
case ARM::t2ADDri:
case ARM::ADDri: Scale = 1; CheckCPSRDef = true; break;
case ARM::tADDspi: Scale = 4; CheckCPSRDef = false; break;
case ARM::tSUBspi: Scale = -4; CheckCPSRDef = false; break;
default: return 0;
}
unsigned MIPredReg;
if (MI.getOperand(0).getReg() != Reg ||
MI.getOperand(1).getReg() != Reg ||
getInstrPredicate(&MI, MIPredReg) != Pred ||
MIPredReg != PredReg)
return 0;
if (CheckCPSRDef && definesCPSR(&MI))
return 0;
return MI.getOperand(2).getImm() * Scale;
}
/// Searches for an increment or decrement of \p Reg before \p MBBI.
static MachineBasicBlock::iterator
findIncDecBefore(MachineBasicBlock::iterator MBBI, unsigned Reg,
ARMCC::CondCodes Pred, unsigned PredReg, int &Offset) {
Offset = 0;
MachineBasicBlock &MBB = *MBBI->getParent();
MachineBasicBlock::iterator BeginMBBI = MBB.begin();
MachineBasicBlock::iterator EndMBBI = MBB.end();
if (MBBI == BeginMBBI)
return EndMBBI;
// Skip debug values.
MachineBasicBlock::iterator PrevMBBI = std::prev(MBBI);
while (PrevMBBI->isDebugValue() && PrevMBBI != BeginMBBI)
--PrevMBBI;
Offset = isIncrementOrDecrement(*PrevMBBI, Reg, Pred, PredReg);
return Offset == 0 ? EndMBBI : PrevMBBI;
}
/// Searches for a increment or decrement of \p Reg after \p MBBI.
static MachineBasicBlock::iterator
findIncDecAfter(MachineBasicBlock::iterator MBBI, unsigned Reg,
ARMCC::CondCodes Pred, unsigned PredReg, int &Offset) {
Offset = 0;
MachineBasicBlock &MBB = *MBBI->getParent();
MachineBasicBlock::iterator EndMBBI = MBB.end();
MachineBasicBlock::iterator NextMBBI = std::next(MBBI);
// Skip debug values.
while (NextMBBI != EndMBBI && NextMBBI->isDebugValue())
++NextMBBI;
if (NextMBBI == EndMBBI)
return EndMBBI;
Offset = isIncrementOrDecrement(*NextMBBI, Reg, Pred, PredReg);
return Offset == 0 ? EndMBBI : NextMBBI;
}
/// Fold proceeding/trailing inc/dec of base register into the
/// LDM/STM/VLDM{D|S}/VSTM{D|S} op when possible:
///
@ -1148,6 +1151,7 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLSMultiple(MachineInstr *MI) {
const MachineOperand &BaseOP = MI->getOperand(0);
unsigned Base = BaseOP.getReg();
bool BaseKill = BaseOP.isKill();
unsigned Bytes = getLSMultipleTransferSize(MI);
unsigned PredReg = 0;
ARMCC::CondCodes Pred = getInstrPredicate(MI, PredReg);
unsigned Opcode = MI->getOpcode();
@ -1159,24 +1163,49 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLSMultiple(MachineInstr *MI) {
if (MI->getOperand(i).getReg() == Base)
return false;
int Bytes = getLSMultipleTransferSize(MI);
MachineBasicBlock &MBB = *MI->getParent();
MachineBasicBlock::iterator MBBI(MI);
int Offset;
MachineBasicBlock::iterator MergeInstr
= findIncDecBefore(MBBI, Base, Pred, PredReg, Offset);
bool DoMerge = false;
ARM_AM::AMSubMode Mode = getLoadStoreMultipleSubMode(Opcode);
if (Mode == ARM_AM::ia && Offset == -Bytes) {
Mode = ARM_AM::db;
} else if (Mode == ARM_AM::ib && Offset == -Bytes) {
Mode = ARM_AM::da;
} else {
MergeInstr = findIncDecAfter(MBBI, Base, Pred, PredReg, Offset);
if (((Mode != ARM_AM::ia && Mode != ARM_AM::ib) || Offset != Bytes) &&
((Mode != ARM_AM::da && Mode != ARM_AM::db) || Offset != -Bytes))
return false;
// Try merging with the previous instruction.
MachineBasicBlock &MBB = *MI->getParent();
MachineBasicBlock::iterator BeginMBBI = MBB.begin();
MachineBasicBlock::iterator MBBI(MI);
if (MBBI != BeginMBBI) {
MachineBasicBlock::iterator PrevMBBI = std::prev(MBBI);
while (PrevMBBI != BeginMBBI && PrevMBBI->isDebugValue())
--PrevMBBI;
if (Mode == ARM_AM::ia &&
isMatchingDecrement(PrevMBBI, Base, Bytes, 0, Pred, PredReg)) {
Mode = ARM_AM::db;
DoMerge = true;
} else if (Mode == ARM_AM::ib &&
isMatchingDecrement(PrevMBBI, Base, Bytes, 0, Pred, PredReg)) {
Mode = ARM_AM::da;
DoMerge = true;
}
if (DoMerge)
MBB.erase(PrevMBBI);
}
MBB.erase(MergeInstr);
// Try merging with the next instruction.
MachineBasicBlock::iterator EndMBBI = MBB.end();
if (!DoMerge && MBBI != EndMBBI) {
MachineBasicBlock::iterator NextMBBI = std::next(MBBI);
while (NextMBBI != EndMBBI && NextMBBI->isDebugValue())
++NextMBBI;
if ((Mode == ARM_AM::ia || Mode == ARM_AM::ib) &&
isMatchingIncrement(NextMBBI, Base, Bytes, 0, Pred, PredReg)) {
DoMerge = true;
} else if ((Mode == ARM_AM::da || Mode == ARM_AM::db) &&
isMatchingDecrement(NextMBBI, Base, Bytes, 0, Pred, PredReg)) {
DoMerge = true;
}
if (DoMerge)
MBB.erase(NextMBBI);
}
if (!DoMerge)
return false;
unsigned NewOpc = getUpdatingLSMultipleOpcode(Opcode, Mode);
MachineInstrBuilder MIB = BuildMI(MBB, MBBI, DL, TII->get(NewOpc))
@ -1254,6 +1283,7 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLoadStore(MachineInstr *MI) {
unsigned Base = getLoadStoreBaseOp(*MI).getReg();
bool BaseKill = getLoadStoreBaseOp(*MI).isKill();
unsigned Bytes = getLSMultipleTransferSize(MI);
unsigned Opcode = MI->getOpcode();
DebugLoc DL = MI->getDebugLoc();
bool isAM5 = (Opcode == ARM::VLDRD || Opcode == ARM::VLDRS ||
@ -1265,6 +1295,7 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLoadStore(MachineInstr *MI) {
if (isAM5 && ARM_AM::getAM5Offset(MI->getOperand(2).getImm()) != 0)
return false;
bool isLd = isLoadSingle(Opcode);
// Can't do the merge if the destination register is the same as the would-be
// writeback register.
if (MI->getOperand(0).getReg() == Base)
@ -1272,31 +1303,55 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLoadStore(MachineInstr *MI) {
unsigned PredReg = 0;
ARMCC::CondCodes Pred = getInstrPredicate(MI, PredReg);
int Bytes = getLSMultipleTransferSize(MI);
bool DoMerge = false;
ARM_AM::AddrOpc AddSub = ARM_AM::add;
unsigned NewOpc = 0;
// AM2 - 12 bits, thumb2 - 8 bits.
unsigned Limit = isAM5 ? 0 : (isAM2 ? 0x1000 : 0x100);
// Try merging with the previous instruction.
MachineBasicBlock &MBB = *MI->getParent();
MachineBasicBlock::iterator BeginMBBI = MBB.begin();
MachineBasicBlock::iterator MBBI(MI);
int Offset;
MachineBasicBlock::iterator MergeInstr
= findIncDecBefore(MBBI, Base, Pred, PredReg, Offset);
unsigned NewOpc;
if (!isAM5 && Offset == Bytes) {
NewOpc = getPreIndexedLoadStoreOpcode(Opcode, ARM_AM::add);
} else if (Offset == -Bytes) {
NewOpc = getPreIndexedLoadStoreOpcode(Opcode, ARM_AM::sub);
} else {
MergeInstr = findIncDecAfter(MBBI, Base, Pred, PredReg, Offset);
if (Offset == Bytes) {
NewOpc = getPostIndexedLoadStoreOpcode(Opcode, ARM_AM::add);
} else if (!isAM5 && Offset == -Bytes) {
NewOpc = getPostIndexedLoadStoreOpcode(Opcode, ARM_AM::sub);
} else
return false;
if (MBBI != BeginMBBI) {
MachineBasicBlock::iterator PrevMBBI = std::prev(MBBI);
while (PrevMBBI != BeginMBBI && PrevMBBI->isDebugValue())
--PrevMBBI;
if (isMatchingDecrement(PrevMBBI, Base, Bytes, Limit, Pred, PredReg)) {
DoMerge = true;
AddSub = ARM_AM::sub;
} else if (!isAM5 &&
isMatchingIncrement(PrevMBBI, Base, Bytes, Limit,Pred,PredReg)) {
DoMerge = true;
}
if (DoMerge) {
NewOpc = getPreIndexedLoadStoreOpcode(Opcode, AddSub);
MBB.erase(PrevMBBI);
}
}
MBB.erase(MergeInstr);
ARM_AM::AddrOpc AddSub = Offset < 0 ? ARM_AM::sub : ARM_AM::add;
// Try merging with the next instruction.
MachineBasicBlock::iterator EndMBBI = MBB.end();
if (!DoMerge && MBBI != EndMBBI) {
MachineBasicBlock::iterator NextMBBI = std::next(MBBI);
while (NextMBBI != EndMBBI && NextMBBI->isDebugValue())
++NextMBBI;
if (!isAM5 &&
isMatchingDecrement(NextMBBI, Base, Bytes, Limit, Pred, PredReg)) {
DoMerge = true;
AddSub = ARM_AM::sub;
} else if (isMatchingIncrement(NextMBBI, Base, Bytes, Limit,Pred,PredReg)) {
DoMerge = true;
}
if (DoMerge) {
NewOpc = getPostIndexedLoadStoreOpcode(Opcode, AddSub);
MBB.erase(NextMBBI);
}
}
if (!DoMerge)
return false;
bool isLd = isLoadSingle(Opcode);
if (isAM5) {
// VLDM[SD]_UPD, VSTM[SD]_UPD
// (There are no base-updating versions of VLDR/VSTR instructions, but the
@ -1313,16 +1368,18 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLoadStore(MachineInstr *MI) {
if (isAM2) {
// LDR_PRE, LDR_POST
if (NewOpc == ARM::LDR_PRE_IMM || NewOpc == ARM::LDRB_PRE_IMM) {
int Offset = AddSub == ARM_AM::sub ? -Bytes : Bytes;
BuildMI(MBB, MBBI, DL, TII->get(NewOpc), MI->getOperand(0).getReg())
.addReg(Base, RegState::Define)
.addReg(Base).addImm(Offset).addImm(Pred).addReg(PredReg);
} else {
int Imm = ARM_AM::getAM2Opc(AddSub, Bytes, ARM_AM::no_shift);
int Offset = ARM_AM::getAM2Opc(AddSub, Bytes, ARM_AM::no_shift);
BuildMI(MBB, MBBI, DL, TII->get(NewOpc), MI->getOperand(0).getReg())
.addReg(Base, RegState::Define)
.addReg(Base).addReg(0).addImm(Imm).addImm(Pred).addReg(PredReg);
.addReg(Base).addReg(0).addImm(Offset).addImm(Pred).addReg(PredReg);
}
} else {
int Offset = AddSub == ARM_AM::sub ? -Bytes : Bytes;
// t2LDR_PRE, t2LDR_POST
BuildMI(MBB, MBBI, DL, TII->get(NewOpc), MI->getOperand(0).getReg())
.addReg(Base, RegState::Define)
@ -1334,12 +1391,13 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLoadStore(MachineInstr *MI) {
// the vestigal zero-reg offset register. When that's fixed, this clause
// can be removed entirely.
if (isAM2 && NewOpc == ARM::STR_POST_IMM) {
int Imm = ARM_AM::getAM2Opc(AddSub, Bytes, ARM_AM::no_shift);
int Offset = ARM_AM::getAM2Opc(AddSub, Bytes, ARM_AM::no_shift);
// STR_PRE, STR_POST
BuildMI(MBB, MBBI, DL, TII->get(NewOpc), Base)
.addReg(MO.getReg(), getKillRegState(MO.isKill()))
.addReg(Base).addReg(0).addImm(Imm).addImm(Pred).addReg(PredReg);
.addReg(Base).addReg(0).addImm(Offset).addImm(Pred).addReg(PredReg);
} else {
int Offset = AddSub == ARM_AM::sub ? -Bytes : Bytes;
// t2STR_PRE, t2STR_POST
BuildMI(MBB, MBBI, DL, TII->get(NewOpc), Base)
.addReg(MO.getReg(), getKillRegState(MO.isKill()))
@ -1351,66 +1409,6 @@ bool ARMLoadStoreOpt::MergeBaseUpdateLoadStore(MachineInstr *MI) {
return true;
}
bool ARMLoadStoreOpt::MergeBaseUpdateLSDouble(MachineInstr &MI) const {
unsigned Opcode = MI.getOpcode();
assert((Opcode == ARM::t2LDRDi8 || Opcode == ARM::t2STRDi8) &&
"Must have t2STRDi8 or t2LDRDi8");
if (MI.getOperand(3).getImm() != 0)
return false;
// Behaviour for writeback is undefined if base register is the same as one
// of the others.
const MachineOperand &BaseOp = MI.getOperand(2);
unsigned Base = BaseOp.getReg();
const MachineOperand &Reg0Op = MI.getOperand(0);
const MachineOperand &Reg1Op = MI.getOperand(1);
if (Reg0Op.getReg() == Base || Reg1Op.getReg() == Base)
return false;
unsigned PredReg;
ARMCC::CondCodes Pred = getInstrPredicate(&MI, PredReg);
MachineBasicBlock::iterator MBBI(MI);
MachineBasicBlock &MBB = *MI.getParent();
int Offset;
MachineBasicBlock::iterator MergeInstr = findIncDecBefore(MBBI, Base, Pred,
PredReg, Offset);
unsigned NewOpc;
if (Offset == 8 || Offset == -8) {
NewOpc = Opcode == ARM::t2LDRDi8 ? ARM::t2LDRD_PRE : ARM::t2STRD_PRE;
} else {
MergeInstr = findIncDecAfter(MBBI, Base, Pred, PredReg, Offset);
if (Offset == 8 || Offset == -8) {
NewOpc = Opcode == ARM::t2LDRDi8 ? ARM::t2LDRD_POST : ARM::t2STRD_POST;
} else
return false;
}
MBB.erase(MergeInstr);
DebugLoc DL = MI.getDebugLoc();
MachineInstrBuilder MIB = BuildMI(MBB, MBBI, DL, TII->get(NewOpc));
if (NewOpc == ARM::t2LDRD_PRE || NewOpc == ARM::t2LDRD_POST) {
MIB.addOperand(Reg0Op).addOperand(Reg1Op)
.addReg(BaseOp.getReg(), RegState::Define);
} else {
assert(NewOpc == ARM::t2STRD_PRE || NewOpc == ARM::t2STRD_POST);
MIB.addReg(BaseOp.getReg(), RegState::Define)
.addOperand(Reg0Op).addOperand(Reg1Op);
}
MIB.addReg(BaseOp.getReg(), RegState::Kill)
.addImm(Offset).addImm(Pred).addReg(PredReg);
assert(TII->get(Opcode).getNumOperands() == 6 &&
TII->get(NewOpc).getNumOperands() == 7 &&
"Unexpected number of operands in Opcode specification.");
// Transfer implicit operands.
for (const MachineOperand &MO : MI.implicit_operands())
MIB.addOperand(MO);
MIB->setMemRefs(MI.memoperands_begin(), MI.memoperands_end());
MBB.erase(MBBI);
return true;
}
/// Returns true if instruction is a memory operation that this pass is capable
/// of operating on.
static bool isMemoryOp(const MachineInstr *MI) {
@ -1618,7 +1616,6 @@ bool ARMLoadStoreOpt::LoadStoreMultipleOpti(MachineBasicBlock &MBB) {
ARMCC::CondCodes CurrPred = ARMCC::AL;
unsigned Position = 0;
assert(Candidates.size() == 0);
assert(MergeBaseCandidates.size() == 0);
LiveRegsValid = false;
for (MachineBasicBlock::iterator I = MBB.end(), MBBI; I != MBB.begin();
@ -1697,15 +1694,8 @@ bool ARMLoadStoreOpt::LoadStoreMultipleOpti(MachineBasicBlock &MBB) {
MBBI = I;
--Position;
// Fallthrough to look into existing chain.
} else if (MBBI->isDebugValue()) {
} else if (MBBI->isDebugValue())
continue;
} else if (MBBI->getOpcode() == ARM::t2LDRDi8 ||
MBBI->getOpcode() == ARM::t2STRDi8) {
// ARMPreAllocLoadStoreOpt has already formed some LDRD/STRD instructions
// remember them because we may still be able to merge add/sub into them.
MergeBaseCandidates.push_back(MBBI);
}
// If we are here then the chain is broken; Extract candidates for a merge.
if (MemOps.size() > 0) {
@ -1736,9 +1726,7 @@ bool ARMLoadStoreOpt::LoadStoreMultipleOpti(MachineBasicBlock &MBB) {
if (Merged) {
Changed = true;
unsigned Opcode = Merged->getOpcode();
if (Opcode == ARM::t2STRDi8 || Opcode == ARM::t2LDRDi8)
MergeBaseUpdateLSDouble(*Merged);
else
if (Opcode != ARM::t2STRDi8 && Opcode != ARM::t2LDRDi8)
MergeBaseUpdateLSMultiple(Merged);
} else {
for (MachineInstr *MI : Candidate->Instrs) {
@ -1753,10 +1741,6 @@ bool ARMLoadStoreOpt::LoadStoreMultipleOpti(MachineBasicBlock &MBB) {
}
}
Candidates.clear();
// Try to fold add/sub into the LDRD/STRD formed by ARMPreAllocLoadStoreOpt.
for (MachineInstr *MI : MergeBaseCandidates)
MergeBaseUpdateLSDouble(*MI);
MergeBaseCandidates.clear();
return Changed;
}

View File

@ -57,7 +57,7 @@ void Thumb1InstrInfo::copyPhysReg(MachineBasicBlock &MBB,
// Some things to try that should be better:
// * 'mov hi, $src; mov $dst, hi', with hi as either r10 or r11
// * 'movs $dst, $src' if cpsr isn't live
// See: http://lists.cs.uiuc.edu/pipermail/llvmdev/2014-August/075998.html
// See: http://lists.llvm.org/pipermail/llvm-dev/2014-August/075998.html
// 'MOV lo, lo' is unpredictable on < v6, so use the stack to do it
AddDefaultPred(BuildMI(MBB, I, DL, get(ARM::tPUSH)))

View File

@ -864,13 +864,13 @@ static bool needToReserveScavengingSpillSlots(MachineFunction &MF,
// Check for an unused caller-saved register.
for ( ; *CallerSavedRegs; ++CallerSavedRegs) {
MCPhysReg FreeReg = *CallerSavedRegs;
if (!MRI.reg_nodbg_empty(FreeReg))
if (MRI.isPhysRegUsed(FreeReg))
continue;
// Check aliased register usage.
bool IsCurrentRegUsed = false;
for (MCRegAliasIterator AI(FreeReg, &HRI, false); AI.isValid(); ++AI)
if (!MRI.reg_nodbg_empty(*AI)) {
if (MRI.isPhysRegUsed(*AI)) {
IsCurrentRegUsed = true;
break;
}

View File

@ -500,14 +500,6 @@ def : MipsPat<(trunc (assertzext GPR64:$src)),
def : MipsPat<(i32 (trunc GPR64:$src)),
(SLL (EXTRACT_SUBREG GPR64:$src, sub_32), 0)>;
// Bypass trunc nodes for bitwise ops.
def : MipsPat<(i32 (trunc (and GPR64:$lhs, GPR64:$rhs))),
(EXTRACT_SUBREG (AND64 GPR64:$lhs, GPR64:$rhs), sub_32)>;
def : MipsPat<(i32 (trunc (or GPR64:$lhs, GPR64:$rhs))),
(EXTRACT_SUBREG (OR64 GPR64:$lhs, GPR64:$rhs), sub_32)>;
def : MipsPat<(i32 (trunc (xor GPR64:$lhs, GPR64:$rhs))),
(EXTRACT_SUBREG (XOR64 GPR64:$lhs, GPR64:$rhs), sub_32)>;
// variable shift instructions patterns
def : MipsPat<(shl GPR64:$rt, (i32 (trunc GPR64:$rs))),
(DSLLV GPR64:$rt, (EXTRACT_SUBREG GPR64:$rs, sub_32))>;

View File

@ -267,6 +267,9 @@ unsigned MipsFastISel::emitLogicalOp(unsigned ISDOpc, MVT RetVT,
}
unsigned MipsFastISel::fastMaterializeAlloca(const AllocaInst *AI) {
if (!TargetSupported)
return 0;
assert(TLI.getValueType(DL, AI->getType(), true) == MVT::i32 &&
"Alloca should always return a pointer.");
@ -290,12 +293,7 @@ unsigned MipsFastISel::materializeInt(const Constant *C, MVT VT) {
return 0;
const TargetRegisterClass *RC = &Mips::GPR32RegClass;
const ConstantInt *CI = cast<ConstantInt>(C);
int64_t Imm;
if ((VT != MVT::i1) && CI->isNegative())
Imm = CI->getSExtValue();
else
Imm = CI->getZExtValue();
return materialize32BitInt(Imm, RC);
return materialize32BitInt(CI->getZExtValue(), RC);
}
unsigned MipsFastISel::materialize32BitInt(int64_t Imm,
@ -382,6 +380,9 @@ unsigned MipsFastISel::materializeExternalCallSym(MCSymbol *Sym) {
// Materialize a constant into a register, and return the register
// number (or zero if we failed to handle it).
unsigned MipsFastISel::fastMaterializeConstant(const Constant *C) {
if (!TargetSupported)
return 0;
EVT CEVT = TLI.getValueType(DL, C->getType(), true);
// Only handle simple types.
@ -981,6 +982,13 @@ bool MipsFastISel::selectSelect(const Instruction *I) {
if (!Src1Reg || !Src2Reg || !CondReg)
return false;
unsigned ZExtCondReg = createResultReg(&Mips::GPR32RegClass);
if (!ZExtCondReg)
return false;
if (!emitIntExt(MVT::i1, CondReg, MVT::i32, ZExtCondReg, true))
return false;
unsigned ResultReg = createResultReg(RC);
unsigned TempReg = createResultReg(RC);
@ -989,7 +997,7 @@ bool MipsFastISel::selectSelect(const Instruction *I) {
emitInst(TargetOpcode::COPY, TempReg).addReg(Src2Reg);
emitInst(CondMovOpc, ResultReg)
.addReg(Src1Reg).addReg(CondReg).addReg(TempReg);
.addReg(Src1Reg).addReg(ZExtCondReg).addReg(TempReg);
updateValueMap(I, ResultReg);
return true;
}
@ -1232,12 +1240,19 @@ bool MipsFastISel::finishCall(CallLoweringInfo &CLI, MVT RetVT,
}
bool MipsFastISel::fastLowerCall(CallLoweringInfo &CLI) {
if (!TargetSupported)
return false;
CallingConv::ID CC = CLI.CallConv;
bool IsTailCall = CLI.IsTailCall;
bool IsVarArg = CLI.IsVarArg;
const Value *Callee = CLI.Callee;
MCSymbol *Symbol = CLI.Symbol;
// Do not handle FastCC.
if (CC == CallingConv::Fast)
return false;
// Allow SelectionDAG isel to handle tail calls.
if (IsTailCall)
return false;
@ -1312,6 +1327,9 @@ bool MipsFastISel::fastLowerCall(CallLoweringInfo &CLI) {
}
bool MipsFastISel::fastLowerIntrinsicCall(const IntrinsicInst *II) {
if (!TargetSupported)
return false;
switch (II->getIntrinsicID()) {
default:
return false;
@ -1415,6 +1433,11 @@ bool MipsFastISel::selectRet(const Instruction *I) {
if (Ret->getNumOperands() > 0) {
CallingConv::ID CC = F.getCallingConv();
// Do not handle FastCC.
if (CC == CallingConv::Fast)
return false;
SmallVector<ISD::OutputArg, 4> Outs;
GetReturnInfo(F.getReturnType(), F.getAttributes(), Outs, TLI, DL);

View File

@ -27,6 +27,7 @@
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineJumpTableInfo.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/FunctionLoweringInfo.h"
#include "llvm/CodeGen/SelectionDAGISel.h"
#include "llvm/CodeGen/ValueTypes.h"
#include "llvm/IR/CallingConv.h"
@ -53,11 +54,6 @@ NoZeroDivCheck("mno-check-zero-division", cl::Hidden,
cl::desc("MIPS: Don't trap on integer division by zero."),
cl::init(false));
cl::opt<bool>
EnableMipsFastISel("mips-fast-isel", cl::Hidden,
cl::desc("Allow mips-fast-isel to be used"),
cl::init(false));
static const MCPhysReg Mips64DPRegs[8] = {
Mips::D12_64, Mips::D13_64, Mips::D14_64, Mips::D15_64,
Mips::D16_64, Mips::D17_64, Mips::D18_64, Mips::D19_64
@ -461,7 +457,7 @@ const MipsTargetLowering *MipsTargetLowering::create(const MipsTargetMachine &TM
FastISel *
MipsTargetLowering::createFastISel(FunctionLoweringInfo &funcInfo,
const TargetLibraryInfo *libInfo) const {
if (!EnableMipsFastISel)
if (!funcInfo.MF->getTarget().Options.EnableFastISel)
return TargetLowering::createFastISel(funcInfo, libInfo);
return Mips::createFastISel(funcInfo, libInfo);
}

View File

@ -12,6 +12,7 @@
#include "llvm/MC/MCFixedLenDisassembler.h"
#include "llvm/MC/MCInst.h"
#include "llvm/MC/MCSubtargetInfo.h"
#include "llvm/Support/Endian.h"
#include "llvm/Support/TargetRegistry.h"
using namespace llvm;
@ -22,10 +23,12 @@ typedef MCDisassembler::DecodeStatus DecodeStatus;
namespace {
class PPCDisassembler : public MCDisassembler {
bool IsLittleEndian;
public:
PPCDisassembler(const MCSubtargetInfo &STI, MCContext &Ctx)
: MCDisassembler(STI, Ctx) {}
~PPCDisassembler() override {}
PPCDisassembler(const MCSubtargetInfo &STI, MCContext &Ctx,
bool IsLittleEndian)
: MCDisassembler(STI, Ctx), IsLittleEndian(IsLittleEndian) {}
DecodeStatus getInstruction(MCInst &Instr, uint64_t &Size,
ArrayRef<uint8_t> Bytes, uint64_t Address,
@ -37,7 +40,13 @@ class PPCDisassembler : public MCDisassembler {
static MCDisassembler *createPPCDisassembler(const Target &T,
const MCSubtargetInfo &STI,
MCContext &Ctx) {
return new PPCDisassembler(STI, Ctx);
return new PPCDisassembler(STI, Ctx, /*IsLittleEndian=*/false);
}
static MCDisassembler *createPPCLEDisassembler(const Target &T,
const MCSubtargetInfo &STI,
MCContext &Ctx) {
return new PPCDisassembler(STI, Ctx, /*IsLittleEndian=*/true);
}
extern "C" void LLVMInitializePowerPCDisassembler() {
@ -47,7 +56,7 @@ extern "C" void LLVMInitializePowerPCDisassembler() {
TargetRegistry::RegisterMCDisassembler(ThePPC64Target,
createPPCDisassembler);
TargetRegistry::RegisterMCDisassembler(ThePPC64LETarget,
createPPCDisassembler);
createPPCLEDisassembler);
}
// FIXME: These can be generated by TableGen from the existing register
@ -383,9 +392,9 @@ DecodeStatus PPCDisassembler::getInstruction(MCInst &MI, uint64_t &Size,
return MCDisassembler::Fail;
}
// The instruction is big-endian encoded.
uint32_t Inst =
(Bytes[0] << 24) | (Bytes[1] << 16) | (Bytes[2] << 8) | (Bytes[3] << 0);
// Read the instruction in the proper endianness.
uint32_t Inst = IsLittleEndian ? support::endian::read32le(Bytes.data())
: support::endian::read32be(Bytes.data());
if (STI.getFeatureBits()[PPC::FeatureQPX]) {
DecodeStatus result =

View File

@ -363,71 +363,85 @@ void PPCAsmPrinter::LowerPATCHPOINT(MCStreamer &OutStreamer, StackMaps &SM,
SM.recordPatchPoint(MI);
PatchPointOpers Opers(&MI);
int64_t CallTarget = Opers.getMetaOper(PatchPointOpers::TargetPos).getImm();
unsigned EncodedBytes = 0;
if (CallTarget) {
assert((CallTarget & 0xFFFFFFFFFFFF) == CallTarget &&
"High 16 bits of call target should be zero.");
unsigned ScratchReg = MI.getOperand(Opers.getNextScratchIdx()).getReg();
EncodedBytes = 0;
// Materialize the jump address:
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::LI8)
.addReg(ScratchReg)
.addImm((CallTarget >> 32) & 0xFFFF));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::RLDIC)
.addReg(ScratchReg)
.addReg(ScratchReg)
.addImm(32).addImm(16));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::ORIS8)
.addReg(ScratchReg)
.addReg(ScratchReg)
.addImm((CallTarget >> 16) & 0xFFFF));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::ORI8)
.addReg(ScratchReg)
.addReg(ScratchReg)
.addImm(CallTarget & 0xFFFF));
const MachineOperand &CalleeMO =
Opers.getMetaOper(PatchPointOpers::TargetPos);
// Save the current TOC pointer before the remote call.
int TOCSaveOffset = Subtarget->isELFv2ABI() ? 24 : 40;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::STD)
.addReg(PPC::X2)
.addImm(TOCSaveOffset)
.addReg(PPC::X1));
++EncodedBytes;
if (CalleeMO.isImm()) {
int64_t CallTarget = Opers.getMetaOper(PatchPointOpers::TargetPos).getImm();
if (CallTarget) {
assert((CallTarget & 0xFFFFFFFFFFFF) == CallTarget &&
"High 16 bits of call target should be zero.");
unsigned ScratchReg = MI.getOperand(Opers.getNextScratchIdx()).getReg();
EncodedBytes = 0;
// Materialize the jump address:
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::LI8)
.addReg(ScratchReg)
.addImm((CallTarget >> 32) & 0xFFFF));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::RLDIC)
.addReg(ScratchReg)
.addReg(ScratchReg)
.addImm(32).addImm(16));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::ORIS8)
.addReg(ScratchReg)
.addReg(ScratchReg)
.addImm((CallTarget >> 16) & 0xFFFF));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::ORI8)
.addReg(ScratchReg)
.addReg(ScratchReg)
.addImm(CallTarget & 0xFFFF));
// Save the current TOC pointer before the remote call.
int TOCSaveOffset = Subtarget->isELFv2ABI() ? 24 : 40;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::STD)
.addReg(PPC::X2)
.addImm(TOCSaveOffset)
.addReg(PPC::X1));
++EncodedBytes;
// If we're on ELFv1, then we need to load the actual function pointer from
// the function descriptor.
if (!Subtarget->isELFv2ABI()) {
// Load the new TOC pointer and the function address, but not r11
// (needing this is rare, and loading it here would prevent passing it
// via a 'nest' parameter.
// If we're on ELFv1, then we need to load the actual function pointer
// from the function descriptor.
if (!Subtarget->isELFv2ABI()) {
// Load the new TOC pointer and the function address, but not r11
// (needing this is rare, and loading it here would prevent passing it
// via a 'nest' parameter.
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::LD)
.addReg(PPC::X2)
.addImm(8)
.addReg(ScratchReg));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::LD)
.addReg(ScratchReg)
.addImm(0)
.addReg(ScratchReg));
++EncodedBytes;
}
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::MTCTR8)
.addReg(ScratchReg));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::BCTRL8));
++EncodedBytes;
// Restore the TOC pointer after the call.
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::LD)
.addReg(PPC::X2)
.addImm(8)
.addReg(ScratchReg));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::LD)
.addReg(ScratchReg)
.addImm(0)
.addReg(ScratchReg));
.addImm(TOCSaveOffset)
.addReg(PPC::X1));
++EncodedBytes;
}
} else if (CalleeMO.isGlobal()) {
const GlobalValue *GValue = CalleeMO.getGlobal();
MCSymbol *MOSymbol = getSymbol(GValue);
const MCExpr *SymVar = MCSymbolRefExpr::create(MOSymbol, OutContext);
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::MTCTR8).addReg(ScratchReg));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::BCTRL8));
++EncodedBytes;
// Restore the TOC pointer after the call.
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::LD)
.addReg(PPC::X2)
.addImm(TOCSaveOffset)
.addReg(PPC::X1));
++EncodedBytes;
EmitToStreamer(OutStreamer, MCInstBuilder(PPC::BL8_NOP)
.addExpr(SymVar));
EncodedBytes += 2;
}
// Each instruction is 4 bytes.

View File

@ -306,10 +306,9 @@ static void HandleVRSaveUpdate(MachineInstr *MI, const TargetInstrInfo &TII) {
const TargetRegisterInfo *TRI = MF->getSubtarget().getRegisterInfo();
DebugLoc dl = MI->getDebugLoc();
const MachineRegisterInfo &MRI = MF->getRegInfo();
unsigned UsedRegMask = 0;
for (unsigned i = 0; i != 32; ++i)
if (MRI.isPhysRegModified(VRRegNo[i]))
if (MF->getRegInfo().isPhysRegUsed(VRRegNo[i]))
UsedRegMask |= 1 << (31-i);
// Live in and live out values already must be in the mask, so don't bother

View File

@ -2305,14 +2305,15 @@ SDNode *PPCDAGToDAGISel::SelectSETCC(SDNode *N) {
if (Swap)
std::swap(LHS, RHS);
EVT ResVT = VecVT.changeVectorElementTypeToInteger();
if (Negate) {
SDValue VCmp(CurDAG->getMachineNode(VCmpInst, dl, VecVT, LHS, RHS), 0);
SDValue VCmp(CurDAG->getMachineNode(VCmpInst, dl, ResVT, LHS, RHS), 0);
return CurDAG->SelectNodeTo(N, PPCSubTarget->hasVSX() ? PPC::XXLNOR :
PPC::VNOR,
VecVT, VCmp, VCmp);
ResVT, VCmp, VCmp);
}
return CurDAG->SelectNodeTo(N, VCmpInst, VecVT, LHS, RHS);
return CurDAG->SelectNodeTo(N, VCmpInst, ResVT, LHS, RHS);
}
if (PPCSubTarget->useCRBits())

View File

@ -580,6 +580,7 @@ PPCTargetLowering::PPCTargetLowering(const PPCTargetMachine &TM,
addRegisterClass(MVT::f64, &PPC::VSFRCRegClass);
addRegisterClass(MVT::v4i32, &PPC::VSRCRegClass);
addRegisterClass(MVT::v4f32, &PPC::VSRCRegClass);
addRegisterClass(MVT::v2f64, &PPC::VSRCRegClass);
@ -1416,7 +1417,7 @@ int PPC::isVSLDOIShuffleMask(SDNode *N, unsigned ShuffleKind,
} else
return -1;
if (ShuffleKind == 2 && isLE)
if (isLE)
ShiftAmt = 16 - ShiftAmt;
return ShiftAmt;
@ -1429,6 +1430,11 @@ bool PPC::isSplatShuffleMask(ShuffleVectorSDNode *N, unsigned EltSize) {
assert(N->getValueType(0) == MVT::v16i8 &&
(EltSize == 1 || EltSize == 2 || EltSize == 4));
// The consecutive indices need to specify an element, not part of two
// different elements. So abandon ship early if this isn't the case.
if (N->getMaskElt(0) % EltSize != 0)
return false;
// This is a splat operation if each element of the permute is the same, and
// if the value doesn't reference the second vector.
unsigned ElementBase = N->getMaskElt(0);
@ -7011,17 +7017,20 @@ SDValue PPCTargetLowering::LowerBUILD_VECTOR(SDValue Op,
// t = vsplti c, result = vsldoi t, t, 1
if (SextVal == (int)(((unsigned)i << 8) | (i < 0 ? 0xFF : 0))) {
SDValue T = BuildSplatI(i, SplatSize, MVT::v16i8, DAG, dl);
return BuildVSLDOI(T, T, 1, Op.getValueType(), DAG, dl);
unsigned Amt = Subtarget.isLittleEndian() ? 15 : 1;
return BuildVSLDOI(T, T, Amt, Op.getValueType(), DAG, dl);
}
// t = vsplti c, result = vsldoi t, t, 2
if (SextVal == (int)(((unsigned)i << 16) | (i < 0 ? 0xFFFF : 0))) {
SDValue T = BuildSplatI(i, SplatSize, MVT::v16i8, DAG, dl);
return BuildVSLDOI(T, T, 2, Op.getValueType(), DAG, dl);
unsigned Amt = Subtarget.isLittleEndian() ? 14 : 2;
return BuildVSLDOI(T, T, Amt, Op.getValueType(), DAG, dl);
}
// t = vsplti c, result = vsldoi t, t, 3
if (SextVal == (int)(((unsigned)i << 24) | (i < 0 ? 0xFFFFFF : 0))) {
SDValue T = BuildSplatI(i, SplatSize, MVT::v16i8, DAG, dl);
return BuildVSLDOI(T, T, 3, Op.getValueType(), DAG, dl);
unsigned Amt = Subtarget.isLittleEndian() ? 13 : 3;
return BuildVSLDOI(T, T, Amt, Op.getValueType(), DAG, dl);
}
}
@ -9957,6 +9966,9 @@ SDValue PPCTargetLowering::combineFPToIntToFP(SDNode *N,
if (Src.getValueType() == MVT::f32) {
Src = DAG.getNode(ISD::FP_EXTEND, dl, MVT::f64, Src);
DCI.AddToWorklist(Src.getNode());
} else if (Src.getValueType() != MVT::f64) {
// Make sure that we don't pick up a ppc_fp128 source value.
return SDValue();
}
unsigned FCTOp =

View File

@ -190,11 +190,11 @@ static bool LLVM_ATTRIBUTE_UNUSED verifyLeafProcRegUse(MachineRegisterInfo *MRI)
{
for (unsigned reg = SP::I0; reg <= SP::I7; ++reg)
if (!MRI->reg_nodbg_empty(reg))
if (MRI->isPhysRegUsed(reg))
return false;
for (unsigned reg = SP::L0; reg <= SP::L7; ++reg)
if (!MRI->reg_nodbg_empty(reg))
if (MRI->isPhysRegUsed(reg))
return false;
return true;
@ -206,10 +206,10 @@ bool SparcFrameLowering::isLeafProc(MachineFunction &MF) const
MachineRegisterInfo &MRI = MF.getRegInfo();
MachineFrameInfo *MFI = MF.getFrameInfo();
return !(MFI->hasCalls() // has calls
|| !MRI.reg_nodbg_empty(SP::L0) // Too many registers needed
|| !MRI.reg_nodbg_empty(SP::O6) // %SP is used
|| hasFP(MF)); // need %FP
return !(MFI->hasCalls() // has calls
|| MRI.isPhysRegUsed(SP::L0) // Too many registers needed
|| MRI.isPhysRegUsed(SP::O6) // %SP is used
|| hasFP(MF)); // need %FP
}
void SparcFrameLowering::remapRegsForLeafProc(MachineFunction &MF) const {
@ -218,13 +218,16 @@ void SparcFrameLowering::remapRegsForLeafProc(MachineFunction &MF) const {
// Remap %i[0-7] to %o[0-7].
for (unsigned reg = SP::I0; reg <= SP::I7; ++reg) {
if (MRI.reg_nodbg_empty(reg))
if (!MRI.isPhysRegUsed(reg))
continue;
unsigned mapped_reg = (reg - SP::I0 + SP::O0);
assert(MRI.reg_nodbg_empty(mapped_reg));
assert(!MRI.isPhysRegUsed(mapped_reg));
// Replace I register with O register.
MRI.replaceRegWith(reg, mapped_reg);
// Mark the reg unused.
MRI.setPhysRegUnused(reg);
}
// Rewrite MBB's Live-ins.

View File

@ -53,10 +53,6 @@ def RetCC_SystemZ : CallingConv<[
CCIfSubtarget<"hasVector()",
CCIfType<[v16i8, v8i16, v4i32, v2i64, v4f32, v2f64],
CCAssignToReg<[V24, V26, V28, V30, V25, V27, V29, V31]>>>
// ABI-compliant code returns long double by reference, but that conversion
// is left to higher-level code. Perhaps we could add an f128 definition
// here for code that doesn't care about the ABI?
]>;
//===----------------------------------------------------------------------===//

View File

@ -1175,6 +1175,20 @@ SystemZTargetLowering::LowerCall(CallLoweringInfo &CLI,
return Chain;
}
bool SystemZTargetLowering::
CanLowerReturn(CallingConv::ID CallConv,
MachineFunction &MF, bool isVarArg,
const SmallVectorImpl<ISD::OutputArg> &Outs,
LLVMContext &Context) const {
// Detect unsupported vector return types.
if (Subtarget.hasVector())
VerifyVectorTypes(Outs);
SmallVector<CCValAssign, 16> RetLocs;
CCState RetCCInfo(CallConv, isVarArg, MF, RetLocs, Context);
return RetCCInfo.CheckReturn(Outs, RetCC_SystemZ);
}
SDValue
SystemZTargetLowering::LowerReturn(SDValue Chain,
CallingConv::ID CallConv, bool IsVarArg,

View File

@ -423,6 +423,10 @@ class SystemZTargetLowering : public TargetLowering {
SDValue LowerCall(CallLoweringInfo &CLI,
SmallVectorImpl<SDValue> &InVals) const override;
bool CanLowerReturn(CallingConv::ID CallConv, MachineFunction &MF,
bool isVarArg,
const SmallVectorImpl<ISD::OutputArg> &Outs,
LLVMContext &Context) const override;
SDValue LowerReturn(SDValue Chain, CallingConv::ID CallConv, bool IsVarArg,
const SmallVectorImpl<ISD::OutputArg> &Outs,
const SmallVectorImpl<SDValue> &OutVals,

View File

@ -681,6 +681,9 @@ class X86AsmParser : public MCTargetAsmParser {
std::unique_ptr<X86Operand> DefaultMemSIOperand(SMLoc Loc);
std::unique_ptr<X86Operand> DefaultMemDIOperand(SMLoc Loc);
void AddDefaultSrcDestOperands(
OperandVector& Operands, std::unique_ptr<llvm::MCParsedAsmOperand> &&Src,
std::unique_ptr<llvm::MCParsedAsmOperand> &&Dst);
std::unique_ptr<X86Operand> ParseOperand();
std::unique_ptr<X86Operand> ParseATTOperand();
std::unique_ptr<X86Operand> ParseIntelOperand();
@ -1014,6 +1017,19 @@ std::unique_ptr<X86Operand> X86AsmParser::DefaultMemDIOperand(SMLoc Loc) {
Loc, Loc, 0);
}
void X86AsmParser::AddDefaultSrcDestOperands(
OperandVector& Operands, std::unique_ptr<llvm::MCParsedAsmOperand> &&Src,
std::unique_ptr<llvm::MCParsedAsmOperand> &&Dst) {
if (isParsingIntelSyntax()) {
Operands.push_back(std::move(Dst));
Operands.push_back(std::move(Src));
}
else {
Operands.push_back(std::move(Src));
Operands.push_back(std::move(Dst));
}
}
std::unique_ptr<X86Operand> X86AsmParser::ParseOperand() {
if (isParsingIntelSyntax())
return ParseIntelOperand();
@ -2228,26 +2244,18 @@ bool X86AsmParser::ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
if (Name.startswith("ins") && Operands.size() == 1 &&
(Name == "insb" || Name == "insw" || Name == "insl" ||
Name == "insd" )) {
if (isParsingIntelSyntax()) {
Operands.push_back(X86Operand::CreateReg(X86::DX, NameLoc, NameLoc));
Operands.push_back(DefaultMemDIOperand(NameLoc));
} else {
Operands.push_back(X86Operand::CreateReg(X86::DX, NameLoc, NameLoc));
Operands.push_back(DefaultMemDIOperand(NameLoc));
}
AddDefaultSrcDestOperands(Operands,
X86Operand::CreateReg(X86::DX, NameLoc, NameLoc),
DefaultMemDIOperand(NameLoc));
}
// Append default arguments to "outs[bwld]"
if (Name.startswith("outs") && Operands.size() == 1 &&
(Name == "outsb" || Name == "outsw" || Name == "outsl" ||
Name == "outsd" )) {
if (isParsingIntelSyntax()) {
Operands.push_back(DefaultMemSIOperand(NameLoc));
Operands.push_back(X86Operand::CreateReg(X86::DX, NameLoc, NameLoc));
} else {
Operands.push_back(DefaultMemSIOperand(NameLoc));
Operands.push_back(X86Operand::CreateReg(X86::DX, NameLoc, NameLoc));
}
AddDefaultSrcDestOperands(Operands,
DefaultMemSIOperand(NameLoc),
X86Operand::CreateReg(X86::DX, NameLoc, NameLoc));
}
// Transform "lods[bwlq]" into "lods[bwlq] ($SIREG)" for appropriate
@ -2279,13 +2287,9 @@ bool X86AsmParser::ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
(Name == "cmps" || Name == "cmpsb" || Name == "cmpsw" ||
Name == "cmpsl" || Name == "cmpsd" || Name == "cmpsq")) {
if (Operands.size() == 1) {
if (isParsingIntelSyntax()) {
Operands.push_back(DefaultMemSIOperand(NameLoc));
Operands.push_back(DefaultMemDIOperand(NameLoc));
} else {
Operands.push_back(DefaultMemDIOperand(NameLoc));
Operands.push_back(DefaultMemSIOperand(NameLoc));
}
AddDefaultSrcDestOperands(Operands,
DefaultMemDIOperand(NameLoc),
DefaultMemSIOperand(NameLoc));
} else if (Operands.size() == 3) {
X86Operand &Op = (X86Operand &)*Operands[1];
X86Operand &Op2 = (X86Operand &)*Operands[2];
@ -2305,13 +2309,9 @@ bool X86AsmParser::ParseInstruction(ParseInstructionInfo &Info, StringRef Name,
if (Operands.size() == 1) {
if (Name == "movsd")
Operands.back() = X86Operand::CreateToken("movsl", NameLoc);
if (isParsingIntelSyntax()) {
Operands.push_back(DefaultMemDIOperand(NameLoc));
Operands.push_back(DefaultMemSIOperand(NameLoc));
} else {
Operands.push_back(DefaultMemSIOperand(NameLoc));
Operands.push_back(DefaultMemDIOperand(NameLoc));
}
AddDefaultSrcDestOperands(Operands,
DefaultMemSIOperand(NameLoc),
DefaultMemDIOperand(NameLoc));
} else if (Operands.size() == 3) {
X86Operand &Op = (X86Operand &)*Operands[1];
X86Operand &Op2 = (X86Operand &)*Operands[2];

View File

@ -301,9 +301,8 @@ bool FPS::runOnMachineFunction(MachineFunction &MF) {
bool FPIsUsed = false;
static_assert(X86::FP6 == X86::FP0+6, "Register enums aren't sorted right!");
const MachineRegisterInfo &MRI = MF.getRegInfo();
for (unsigned i = 0; i <= 6; ++i)
if (!MRI.reg_nodbg_empty(X86::FP0 + i)) {
if (MF.getRegInfo().isPhysRegUsed(X86::FP0+i)) {
FPIsUsed = true;
break;
}

View File

@ -1682,6 +1682,8 @@ void X86FrameLowering::adjustForSegmentedStacks(
.addImm(StackSize);
BuildMI(allocMBB, DL, TII.get(MOVri), Reg11)
.addImm(X86FI->getArgumentStackSize());
MF.getRegInfo().setPhysRegUsed(Reg10);
MF.getRegInfo().setPhysRegUsed(Reg11);
} else {
BuildMI(allocMBB, DL, TII.get(X86::PUSHi32))
.addImm(X86FI->getArgumentStackSize());

View File

@ -12640,24 +12640,29 @@ static SDValue LowerFABSorFNEG(SDValue Op, SelectionDAG &DAG) {
if (User->getOpcode() == ISD::FNEG)
return Op;
SDValue Op0 = Op.getOperand(0);
bool IsFNABS = !IsFABS && (Op0.getOpcode() == ISD::FABS);
SDLoc dl(Op);
MVT VT = Op.getSimpleValueType();
// Assume scalar op for initialization; update for vector if needed.
// Note that there are no scalar bitwise logical SSE/AVX instructions, so we
// generate a 16-byte vector constant and logic op even for the scalar case.
// Using a 16-byte mask allows folding the load of the mask with
// the logic op, so it can save (~4 bytes) on code size.
MVT EltVT = VT;
unsigned NumElts = VT == MVT::f64 ? 2 : 4;
// FIXME: Use function attribute "OptimizeForSize" and/or CodeGenOpt::Level to
// decide if we should generate a 16-byte constant mask when we only need 4 or
// 8 bytes for the scalar case.
MVT LogicVT;
MVT EltVT;
unsigned NumElts;
if (VT.isVector()) {
LogicVT = VT;
EltVT = VT.getVectorElementType();
NumElts = VT.getVectorNumElements();
} else {
// There are no scalar bitwise logical SSE/AVX instructions, so we
// generate a 16-byte vector constant and logic op even for the scalar case.
// Using a 16-byte mask allows folding the load of the mask with
// the logic op, so it can save (~4 bytes) on code size.
LogicVT = (VT == MVT::f64) ? MVT::v2f64 : MVT::v4f32;
EltVT = VT;
NumElts = (VT == MVT::f64) ? 2 : 4;
}
unsigned EltBits = EltVT.getSizeInBits();
@ -12670,26 +12675,25 @@ static SDValue LowerFABSorFNEG(SDValue Op, SelectionDAG &DAG) {
const TargetLowering &TLI = DAG.getTargetLoweringInfo();
SDValue CPIdx = DAG.getConstantPool(C, TLI.getPointerTy(DAG.getDataLayout()));
unsigned Alignment = cast<ConstantPoolSDNode>(CPIdx)->getAlignment();
SDValue Mask = DAG.getLoad(VT, dl, DAG.getEntryNode(), CPIdx,
SDValue Mask = DAG.getLoad(LogicVT, dl, DAG.getEntryNode(), CPIdx,
MachinePointerInfo::getConstantPool(),
false, false, false, Alignment);
if (VT.isVector()) {
// For a vector, cast operands to a vector type, perform the logic op,
// and cast the result back to the original value type.
MVT VecVT = MVT::getVectorVT(MVT::i64, VT.getSizeInBits() / 64);
SDValue MaskCasted = DAG.getBitcast(VecVT, Mask);
SDValue Operand = IsFNABS ? DAG.getBitcast(VecVT, Op0.getOperand(0))
: DAG.getBitcast(VecVT, Op0);
unsigned BitOp = IsFABS ? ISD::AND : IsFNABS ? ISD::OR : ISD::XOR;
return DAG.getBitcast(VT,
DAG.getNode(BitOp, dl, VecVT, Operand, MaskCasted));
}
// If not vector, then scalar.
unsigned BitOp = IsFABS ? X86ISD::FAND : IsFNABS ? X86ISD::FOR : X86ISD::FXOR;
SDValue Op0 = Op.getOperand(0);
bool IsFNABS = !IsFABS && (Op0.getOpcode() == ISD::FABS);
unsigned LogicOp =
IsFABS ? X86ISD::FAND : IsFNABS ? X86ISD::FOR : X86ISD::FXOR;
SDValue Operand = IsFNABS ? Op0.getOperand(0) : Op0;
return DAG.getNode(BitOp, dl, VT, Operand, Mask);
if (VT.isVector())
return DAG.getNode(LogicOp, dl, LogicVT, Operand, Mask);
// For the scalar case extend to a 128-bit vector, perform the logic op,
// and extract the scalar result back out.
Operand = DAG.getNode(ISD::SCALAR_TO_VECTOR, dl, LogicVT, Operand);
SDValue LogicNode = DAG.getNode(LogicOp, dl, LogicVT, Operand, Mask);
return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, VT, LogicNode,
DAG.getIntPtrConstant(0, dl));
}
static SDValue LowerFCOPYSIGN(SDValue Op, SelectionDAG &DAG) {
@ -12729,10 +12733,16 @@ static SDValue LowerFCOPYSIGN(SDValue Op, SelectionDAG &DAG) {
Constant *C = ConstantVector::get(CV);
auto PtrVT = TLI.getPointerTy(DAG.getDataLayout());
SDValue CPIdx = DAG.getConstantPool(C, PtrVT, 16);
SDValue Mask1 = DAG.getLoad(SrcVT, dl, DAG.getEntryNode(), CPIdx,
// Perform all logic operations as 16-byte vectors because there are no
// scalar FP logic instructions in SSE. This allows load folding of the
// constants into the logic instructions.
MVT LogicVT = (VT == MVT::f64) ? MVT::v2f64 : MVT::v4f32;
SDValue Mask1 = DAG.getLoad(LogicVT, dl, DAG.getEntryNode(), CPIdx,
MachinePointerInfo::getConstantPool(),
false, false, false, 16);
SDValue SignBit = DAG.getNode(X86ISD::FAND, dl, SrcVT, Op1, Mask1);
Op1 = DAG.getNode(ISD::SCALAR_TO_VECTOR, dl, LogicVT, Op1);
SDValue SignBit = DAG.getNode(X86ISD::FAND, dl, LogicVT, Op1, Mask1);
// Next, clear the sign bit from the first operand (magnitude).
// If it's a constant, we can clear it here.
@ -12740,7 +12750,8 @@ static SDValue LowerFCOPYSIGN(SDValue Op, SelectionDAG &DAG) {
APFloat APF = Op0CN->getValueAPF();
// If the magnitude is a positive zero, the sign bit alone is enough.
if (APF.isPosZero())
return SignBit;
return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, SrcVT, SignBit,
DAG.getIntPtrConstant(0, dl));
APF.clearSign();
CV[0] = ConstantFP::get(*Context, APF);
} else {
@ -12750,15 +12761,18 @@ static SDValue LowerFCOPYSIGN(SDValue Op, SelectionDAG &DAG) {
}
C = ConstantVector::get(CV);
CPIdx = DAG.getConstantPool(C, PtrVT, 16);
SDValue Val = DAG.getLoad(VT, dl, DAG.getEntryNode(), CPIdx,
SDValue Val = DAG.getLoad(LogicVT, dl, DAG.getEntryNode(), CPIdx,
MachinePointerInfo::getConstantPool(),
false, false, false, 16);
// If the magnitude operand wasn't a constant, we need to AND out the sign.
if (!isa<ConstantFPSDNode>(Op0))
Val = DAG.getNode(X86ISD::FAND, dl, VT, Op0, Val);
if (!isa<ConstantFPSDNode>(Op0)) {
Op0 = DAG.getNode(ISD::SCALAR_TO_VECTOR, dl, LogicVT, Op0);
Val = DAG.getNode(X86ISD::FAND, dl, LogicVT, Op0, Val);
}
// OR the magnitude value with the sign bit.
return DAG.getNode(X86ISD::FOR, dl, VT, Val, SignBit);
Val = DAG.getNode(X86ISD::FOR, dl, LogicVT, Val, SignBit);
return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, SrcVT, Val,
DAG.getIntPtrConstant(0, dl));
}
static SDValue LowerFGETSIGN(SDValue Op, SelectionDAG &DAG) {

View File

@ -956,18 +956,9 @@ X86InstrInfo::X86InstrInfo(X86Subtarget &STI)
{ X86::DPPDrri, X86::DPPDrmi, TB_ALIGN_16 },
{ X86::DPPSrri, X86::DPPSrmi, TB_ALIGN_16 },
// FIXME: We should not be folding Fs* scalar loads into vector
// instructions because the vector instructions require vector-sized
// loads. Lowering should create vector-sized instructions (the Fv*
// variants below) to allow load folding.
{ X86::FsANDNPDrr, X86::FsANDNPDrm, TB_ALIGN_16 },
{ X86::FsANDNPSrr, X86::FsANDNPSrm, TB_ALIGN_16 },
{ X86::FsANDPDrr, X86::FsANDPDrm, TB_ALIGN_16 },
{ X86::FsANDPSrr, X86::FsANDPSrm, TB_ALIGN_16 },
{ X86::FsORPDrr, X86::FsORPDrm, TB_ALIGN_16 },
{ X86::FsORPSrr, X86::FsORPSrm, TB_ALIGN_16 },
{ X86::FsXORPDrr, X86::FsXORPDrm, TB_ALIGN_16 },
{ X86::FsXORPSrr, X86::FsXORPSrm, TB_ALIGN_16 },
// Do not fold Fs* scalar logical op loads because there are no scalar
// load variants for these instructions. When folded, the load is required
// to be 128-bits, so the load size would not match.
{ X86::FvANDNPDrr, X86::FvANDNPDrm, TB_ALIGN_16 },
{ X86::FvANDNPSrr, X86::FvANDNPSrm, TB_ALIGN_16 },

View File

@ -2919,6 +2919,14 @@ multiclass sse12_fp_packed_vector_logical_alias<
defm V#NAME#PD : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode,
VR128, v2f64, f128mem, loadv2f64, SSEPackedDouble, itins, 0>,
PD, VEX_4V;
defm V#NAME#PSY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "ps"), OpNode,
VR256, v8f32, f256mem, loadv8f32, SSEPackedSingle, itins, 0>,
PS, VEX_4V, VEX_L;
defm V#NAME#PDY : sse12_fp_packed<opc, !strconcat(OpcodeStr, "pd"), OpNode,
VR256, v4f64, f256mem, loadv4f64, SSEPackedDouble, itins, 0>,
PD, VEX_4V, VEX_L;
}
let Constraints = "$src1 = $dst" in {

View File

@ -93,7 +93,8 @@ static Value *getFCmpValue(bool isordered, unsigned code,
case 5: Pred = isordered ? FCmpInst::FCMP_ONE : FCmpInst::FCMP_UNE; break;
case 6: Pred = isordered ? FCmpInst::FCMP_OLE : FCmpInst::FCMP_ULE; break;
case 7:
if (!isordered) return ConstantInt::getTrue(LHS->getContext());
if (!isordered)
return ConstantInt::get(CmpInst::makeCmpResultType(LHS->getType()), 1);
Pred = FCmpInst::FCMP_ORD; break;
}
return Builder->CreateFCmp(Pred, LHS, RHS);

View File

@ -2112,9 +2112,8 @@ static Instruction *ProcessUGT_ADDCST_ADD(ICmpInst &I, Value *A, Value *B,
bool InstCombiner::OptimizeOverflowCheck(OverflowCheckFlavor OCF, Value *LHS,
Value *RHS, Instruction &OrigI,
Value *&Result, Constant *&Overflow) {
assert((!OrigI.isCommutative() ||
!(isa<Constant>(LHS) && !isa<Constant>(RHS))) &&
"call with a constant RHS if possible!");
if (OrigI.isCommutative() && isa<Constant>(LHS) && !isa<Constant>(RHS))
std::swap(LHS, RHS);
auto SetResult = [&](Value *OpResult, Constant *OverflowVal, bool ReuseName) {
Result = OpResult;

View File

@ -658,7 +658,7 @@ bool EarlyCSE::run() {
// gains over vector when the container becomes very large due to the
// specific access patterns. For more information see the mailing list
// discussion on this:
// http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20120116/135228.html
// http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20120116/135228.html
std::deque<StackNode *> nodesToProcess;
bool Changed = false;

View File

@ -1847,10 +1847,17 @@ static unsigned getAdjustedAlignment(Instruction *I, uint64_t Offset,
static bool canConvertValue(const DataLayout &DL, Type *OldTy, Type *NewTy) {
if (OldTy == NewTy)
return true;
if (IntegerType *OldITy = dyn_cast<IntegerType>(OldTy))
if (IntegerType *NewITy = dyn_cast<IntegerType>(NewTy))
if (NewITy->getBitWidth() >= OldITy->getBitWidth())
return true;
// For integer types, we can't handle any bit-width differences. This would
// break both vector conversions with extension and introduce endianness
// issues when in conjunction with loads and stores.
if (isa<IntegerType>(OldTy) && isa<IntegerType>(NewTy)) {
assert(cast<IntegerType>(OldTy)->getBitWidth() !=
cast<IntegerType>(NewTy)->getBitWidth() &&
"We can't have the same bitwidth for different int types");
return false;
}
if (DL.getTypeSizeInBits(NewTy) != DL.getTypeSizeInBits(OldTy))
return false;
if (!NewTy->isSingleValueType() || !OldTy->isSingleValueType())
@ -1885,10 +1892,8 @@ static Value *convertValue(const DataLayout &DL, IRBuilderTy &IRB, Value *V,
if (OldTy == NewTy)
return V;
if (IntegerType *OldITy = dyn_cast<IntegerType>(OldTy))
if (IntegerType *NewITy = dyn_cast<IntegerType>(NewTy))
if (NewITy->getBitWidth() > OldITy->getBitWidth())
return IRB.CreateZExt(V, NewITy);
assert(!(isa<IntegerType>(OldTy) && isa<IntegerType>(NewTy)) &&
"Integer types must be the exact same to convert.");
// See if we need inttoptr for this type pair. A cast involving both scalars
// and vectors requires and additional bitcast.
@ -2134,6 +2139,9 @@ static bool isIntegerWideningViableForSlice(const Slice &S,
if (LoadInst *LI = dyn_cast<LoadInst>(U->getUser())) {
if (LI->isVolatile())
return false;
// We can't handle loads that extend past the allocated memory.
if (DL.getTypeStoreSize(LI->getType()) > Size)
return false;
// Note that we don't count vector loads or stores as whole-alloca
// operations which enable integer widening because we would prefer to use
// vector widening instead.
@ -2152,6 +2160,9 @@ static bool isIntegerWideningViableForSlice(const Slice &S,
Type *ValueTy = SI->getValueOperand()->getType();
if (SI->isVolatile())
return false;
// We can't handle stores that extend past the allocated memory.
if (DL.getTypeStoreSize(ValueTy) > Size)
return false;
// Note that we don't count vector loads or stores as whole-alloca
// operations which enable integer widening because we would prefer to use
// vector widening instead.
@ -2585,6 +2596,7 @@ class AllocaSliceRewriter : public InstVisitor<AllocaSliceRewriter, bool> {
Type *TargetTy = IsSplit ? Type::getIntNTy(LI.getContext(), SliceSize * 8)
: LI.getType();
const bool IsLoadPastEnd = DL.getTypeStoreSize(TargetTy) > SliceSize;
bool IsPtrAdjusted = false;
Value *V;
if (VecTy) {
@ -2592,13 +2604,27 @@ class AllocaSliceRewriter : public InstVisitor<AllocaSliceRewriter, bool> {
} else if (IntTy && LI.getType()->isIntegerTy()) {
V = rewriteIntegerLoad(LI);
} else if (NewBeginOffset == NewAllocaBeginOffset &&
canConvertValue(DL, NewAllocaTy, LI.getType())) {
NewEndOffset == NewAllocaEndOffset &&
(canConvertValue(DL, NewAllocaTy, TargetTy) ||
(IsLoadPastEnd && NewAllocaTy->isIntegerTy() &&
TargetTy->isIntegerTy()))) {
LoadInst *NewLI = IRB.CreateAlignedLoad(&NewAI, NewAI.getAlignment(),
LI.isVolatile(), LI.getName());
if (LI.isVolatile())
NewLI->setAtomic(LI.getOrdering(), LI.getSynchScope());
V = NewLI;
// If this is an integer load past the end of the slice (which means the
// bytes outside the slice are undef or this load is dead) just forcibly
// fix the integer size with correct handling of endianness.
if (auto *AITy = dyn_cast<IntegerType>(NewAllocaTy))
if (auto *TITy = dyn_cast<IntegerType>(TargetTy))
if (AITy->getBitWidth() < TITy->getBitWidth()) {
V = IRB.CreateZExt(V, TITy, "load.ext");
if (DL.isBigEndian())
V = IRB.CreateShl(V, TITy->getBitWidth() - AITy->getBitWidth(),
"endian_shift");
}
} else {
Type *LTy = TargetTy->getPointerTo();
LoadInst *NewLI = IRB.CreateAlignedLoad(getNewAllocaSlicePtr(IRB, LTy),
@ -2718,10 +2744,25 @@ class AllocaSliceRewriter : public InstVisitor<AllocaSliceRewriter, bool> {
if (IntTy && V->getType()->isIntegerTy())
return rewriteIntegerStore(V, SI);
const bool IsStorePastEnd = DL.getTypeStoreSize(V->getType()) > SliceSize;
StoreInst *NewSI;
if (NewBeginOffset == NewAllocaBeginOffset &&
NewEndOffset == NewAllocaEndOffset &&
canConvertValue(DL, V->getType(), NewAllocaTy)) {
(canConvertValue(DL, V->getType(), NewAllocaTy) ||
(IsStorePastEnd && NewAllocaTy->isIntegerTy() &&
V->getType()->isIntegerTy()))) {
// If this is an integer store past the end of slice (and thus the bytes
// past that point are irrelevant or this is unreachable), truncate the
// value prior to storing.
if (auto *VITy = dyn_cast<IntegerType>(V->getType()))
if (auto *AITy = dyn_cast<IntegerType>(NewAllocaTy))
if (VITy->getBitWidth() > AITy->getBitWidth()) {
if (DL.isBigEndian())
V = IRB.CreateLShr(V, VITy->getBitWidth() - AITy->getBitWidth(),
"endian_shift");
V = IRB.CreateTrunc(V, AITy, "load.trunc");
}
V = convertValue(DL, IRB, V, NewAllocaTy);
NewSI = IRB.CreateAlignedStore(V, &NewAI, NewAI.getAlignment(),
SI.isVolatile());

View File

@ -227,10 +227,16 @@ Value *Scatterer::operator[](unsigned I) {
if (!Idx)
break;
unsigned J = Idx->getZExtValue();
CV[J] = Insert->getOperand(1);
V = Insert->getOperand(0);
if (I == J)
if (I == J) {
CV[J] = Insert->getOperand(1);
return CV[J];
} else if (!CV[J]) {
// Only cache the first entry we find for each index we're not actively
// searching for. This prevents us from going too far up the chain and
// caching incorrect entries.
CV[J] = Insert->getOperand(1);
}
}
CV[I] = Builder.CreateExtractElement(V, Builder.getInt32(I),
V->getName() + ".i" + Twine(I));

View File

@ -384,14 +384,15 @@ void ASTVector<T>::grow(const ASTContext &C, size_t MinSize) {
T *NewElts = new (C, llvm::alignOf<T>()) T[NewCapacity];
// Copy the elements over.
if (std::is_class<T>::value) {
std::uninitialized_copy(Begin, End, NewElts);
// Destroy the original elements.
destroy_range(Begin, End);
}
else {
// Use memcpy for PODs (std::uninitialized_copy optimizes to memmove).
memcpy(NewElts, Begin, CurSize * sizeof(T));
if (Begin != End) {
if (std::is_class<T>::value) {
std::uninitialized_copy(Begin, End, NewElts);
// Destroy the original elements.
destroy_range(Begin, End);
} else {
// Use memcpy for PODs (std::uninitialized_copy optimizes to memmove).
memcpy(NewElts, Begin, CurSize * sizeof(T));
}
}
// ASTContext never frees any memory.

View File

@ -16,6 +16,7 @@
namespace clang {
class ASTContext;
class ObjCInterfaceDecl;
class QualType;
class Expr;
@ -35,11 +36,10 @@ class NSAPI {
ClassId_NSMutableDictionary,
ClassId_NSNumber,
ClassId_NSMutableSet,
ClassId_NSCountedSet,
ClassId_NSMutableOrderedSet,
ClassId_NSValue
};
static const unsigned NumClassIds = 11;
static const unsigned NumClassIds = 10;
enum NSStringMethodKind {
NSStr_stringWithString,
@ -220,6 +220,10 @@ class NSAPI {
/// \brief Returns \c true if \p Id is currently defined as a macro.
bool isMacroDefined(StringRef Id) const;
/// \brief Returns \c true if \p InterfaceDecl is subclass of \p NSClassKind
bool isSubclassOfNSClass(ObjCInterfaceDecl *InterfaceDecl,
NSClassIdKindKind NSClassKind) const;
private:
bool isObjCTypedef(QualType T, StringRef name, IdentifierInfo *&II) const;
bool isObjCEnumerator(const Expr *E,

View File

@ -312,18 +312,26 @@ class OMPLoopDirective : public OMPExecutableDirective {
}
/// \brief Get the updates storage.
MutableArrayRef<Expr *> getUpdates() {
MutableArrayRef<Expr *> getInits() {
Expr **Storage = reinterpret_cast<Expr **>(
&*std::next(child_begin(),
getArraysOffset(getDirectiveKind()) + CollapsedNum));
return MutableArrayRef<Expr *>(Storage, CollapsedNum);
}
/// \brief Get the updates storage.
MutableArrayRef<Expr *> getUpdates() {
Expr **Storage = reinterpret_cast<Expr **>(
&*std::next(child_begin(),
getArraysOffset(getDirectiveKind()) + 2 * CollapsedNum));
return MutableArrayRef<Expr *>(Storage, CollapsedNum);
}
/// \brief Get the final counter updates storage.
MutableArrayRef<Expr *> getFinals() {
Expr **Storage = reinterpret_cast<Expr **>(
&*std::next(child_begin(),
getArraysOffset(getDirectiveKind()) + 2 * CollapsedNum));
getArraysOffset(getDirectiveKind()) + 3 * CollapsedNum));
return MutableArrayRef<Expr *>(Storage, CollapsedNum);
}
@ -358,7 +366,7 @@ class OMPLoopDirective : public OMPExecutableDirective {
static unsigned numLoopChildren(unsigned CollapsedNum,
OpenMPDirectiveKind Kind) {
return getArraysOffset(Kind) +
3 * CollapsedNum; // Counters, Updates and Finals
4 * CollapsedNum; // Counters, Inits, Updates and Finals
}
void setIterationVariable(Expr *IV) {
@ -414,6 +422,7 @@ class OMPLoopDirective : public OMPExecutableDirective {
*std::next(child_begin(), NextUpperBoundOffset) = NUB;
}
void setCounters(ArrayRef<Expr *> A);
void setInits(ArrayRef<Expr *> A);
void setUpdates(ArrayRef<Expr *> A);
void setFinals(ArrayRef<Expr *> A);
@ -453,6 +462,8 @@ class OMPLoopDirective : public OMPExecutableDirective {
Expr *NUB;
/// \brief Counters Loop counters.
SmallVector<Expr *, 4> Counters;
/// \brief Expressions for loop counters inits for CodeGen.
SmallVector<Expr *, 4> Inits;
/// \brief Expressions for loop counters update for CodeGen.
SmallVector<Expr *, 4> Updates;
/// \brief Final loop counter values for GodeGen.
@ -484,10 +495,12 @@ class OMPLoopDirective : public OMPExecutableDirective {
NLB = nullptr;
NUB = nullptr;
Counters.resize(Size);
Inits.resize(Size);
Updates.resize(Size);
Finals.resize(Size);
for (unsigned i = 0; i < Size; ++i) {
Counters[i] = nullptr;
Inits[i] = nullptr;
Updates[i] = nullptr;
Finals[i] = nullptr;
}
@ -584,6 +597,12 @@ class OMPLoopDirective : public OMPExecutableDirective {
return const_cast<OMPLoopDirective *>(this)->getCounters();
}
ArrayRef<Expr *> inits() { return getInits(); }
ArrayRef<Expr *> inits() const {
return const_cast<OMPLoopDirective *>(this)->getInits();
}
ArrayRef<Expr *> updates() { return getUpdates(); }
ArrayRef<Expr *> updates() const {

View File

@ -223,14 +223,15 @@ void BumpVector<T>::grow(BumpVectorContext &C, size_t MinSize) {
T *NewElts = C.getAllocator().template Allocate<T>(NewCapacity);
// Copy the elements over.
if (std::is_class<T>::value) {
std::uninitialized_copy(Begin, End, NewElts);
// Destroy the original elements.
destroy_range(Begin, End);
}
else {
// Use memcpy for PODs (std::uninitialized_copy optimizes to memmove).
memcpy(NewElts, Begin, CurSize * sizeof(T));
if (Begin != End) {
if (std::is_class<T>::value) {
std::uninitialized_copy(Begin, End, NewElts);
// Destroy the original elements.
destroy_range(Begin, End);
} else {
// Use memcpy for PODs (std::uninitialized_copy optimizes to memmove).
memcpy(NewElts, Begin, CurSize * sizeof(T));
}
}
// For now, leak 'Begin'. We can add it back to a freelist in

View File

@ -1133,7 +1133,7 @@ def ObjCRuntimeName : Attr {
def ObjCBoxable : Attr {
let Spellings = [GNU<"objc_boxable">];
let Subjects = SubjectList<[Record], ErrorDiag, "ExpectedStructOrUnion">;
let Documentation = [Undocumented];
let Documentation = [ObjCBoxableDocs];
}
def OptimizeNone : InheritableAttr {

View File

@ -492,6 +492,34 @@ can only be placed before an @protocol or @interface declaration:
}];
}
def ObjCBoxableDocs : Documentation {
let Category = DocCatFunction;
let Content = [{
Structs and unions marked with the ``objc_boxable`` attribute can be used
with the Objective-C boxed expression syntax, ``@(...)``.
**Usage**: ``__attribute__((objc_boxable))``. This attribute
can only be placed on a declaration of a trivially-copyable struct or union:
.. code-block:: objc
struct __attribute__((objc_boxable)) some_struct {
int i;
};
union __attribute__((objc_boxable)) some_union {
int i;
float f;
};
typedef struct __attribute__((objc_boxable)) _some_struct some_struct;
// ...
some_struct ss;
NSValue *boxed = @(ss);
}];
}
def AvailabilityDocs : Documentation {
let Category = DocCatFunction;
let Content = [{

View File

@ -195,6 +195,8 @@ def err_unable_to_make_temp : Error<
// Modules
def err_module_file_conflict : Error<"module '%0' found in both '%1' and '%2'">;
def err_module_format_unhandled : Error<
"no handler registered for module format '%0'">;
// TransformActions
// TODO: Use a custom category name to distinguish rewriter errors.

View File

@ -358,6 +358,10 @@ def err_invalid_pixel_decl_spec_combination : Error<
"'%0' declaration specifier not allowed here">;
def err_invalid_vector_bool_decl_spec : Error<
"cannot use '%0' with '__vector bool'">;
def err_invalid_vector_long_decl_spec : Error<
"cannot use 'long' with '__vector'">;
def err_invalid_vector_float_decl_spec : Error<
"cannot use 'float' with '__vector'">;
def err_invalid_vector_double_decl_spec : Error <
"use of 'double' with '__vector' requires VSX support to be enabled "
"(available on POWER7 or later)">;

View File

@ -5358,7 +5358,7 @@ def err_objc_object_catch : Error<
def err_incomplete_type_objc_at_encode : Error<
"'@encode' of incomplete type %0">;
def warn_objc_circular_container : Warning<
"adding '%0' to '%0' might cause circular dependency in container">,
"adding '%0' to '%1' might cause circular dependency in container">,
InGroup<DiagGroup<"objc-circular-container">>;
def note_objc_circular_container_declared_here : Note<"'%0' declared here">;

View File

@ -104,6 +104,7 @@ LANGOPT(WritableStrings , 1, 0, "writable string support")
LANGOPT(ConstStrings , 1, 0, "const-qualified string support")
LANGOPT(LaxVectorConversions , 1, 1, "lax vector conversions")
LANGOPT(AltiVec , 1, 0, "AltiVec-style vector initializers")
LANGOPT(ZVector , 1, 0, "System z vector extensions")
LANGOPT(Exceptions , 1, 0, "exception handling")
LANGOPT(ObjCExceptions , 1, 0, "Objective-C exceptions")
LANGOPT(CXXExceptions , 1, 0, "C++ exceptions")

View File

@ -239,6 +239,8 @@ PUNCTUATOR(greatergreatergreater, ">>>")
// KEYOPENCL - This is a keyword in OpenCL
// KEYNOOPENCL - This is a keyword that is not supported in OpenCL
// KEYALTIVEC - This is a keyword in AltiVec
// KEYZVECTOR - This is a keyword for the System z vector extensions,
// which are heavily based on AltiVec
// KEYBORLAND - This is a keyword if Borland extensions are enabled
// BOOLSUPPORT - This is a keyword if 'bool' is a built-in type
// HALFSUPPORT - This is a keyword if 'half' is a built-in type
@ -501,7 +503,7 @@ ALIAS("write_only", __write_only , KEYOPENCL)
ALIAS("read_write", __read_write , KEYOPENCL)
// OpenCL builtins
KEYWORD(__builtin_astype , KEYOPENCL)
KEYWORD(vec_step , KEYOPENCL|KEYALTIVEC)
KEYWORD(vec_step , KEYOPENCL|KEYALTIVEC|KEYZVECTOR)
// OpenMP Type Traits
KEYWORD(__builtin_omp_required_simd_align, KEYALL)
@ -510,9 +512,9 @@ KEYWORD(__builtin_omp_required_simd_align, KEYALL)
KEYWORD(__pascal , KEYALL)
// Altivec Extension.
KEYWORD(__vector , KEYALTIVEC)
KEYWORD(__vector , KEYALTIVEC|KEYZVECTOR)
KEYWORD(__pixel , KEYALTIVEC)
KEYWORD(__bool , KEYALTIVEC)
KEYWORD(__bool , KEYALTIVEC|KEYZVECTOR)
// ARM NEON extensions.
ALIAS("__fp16", half , KEYALL)

View File

@ -14,30 +14,32 @@
namespace clang {
/// \brief A PCHContainerOperations implementation that uses LLVM to
/// A PCHContainerWriter implementation that uses LLVM to
/// wraps Clang modules inside a COFF, ELF, or Mach-O container.
class ObjectFilePCHContainerOperations
: public PCHContainerOperations {
/// \brief Return an ASTConsumer that can be chained with a
class ObjectFilePCHContainerWriter : public PCHContainerWriter {
StringRef getFormat() const override { return "obj"; }
/// Return an ASTConsumer that can be chained with a
/// PCHGenerator that produces a wrapper file format
/// that also contains full debug info for the module.
std::unique_ptr<ASTConsumer>
CreatePCHContainerGenerator(
std::unique_ptr<ASTConsumer> CreatePCHContainerGenerator(
DiagnosticsEngine &Diags, const HeaderSearchOptions &HSO,
const PreprocessorOptions &PPO, const TargetOptions &TO,
const LangOptions &LO, const std::string &MainFileName,
const std::string &OutputFileName, llvm::raw_pwrite_stream *OS,
std::shared_ptr<PCHBuffer> Buffer) const override;
};
/// \brief Initialize an llvm::BitstreamReader with the serialized
/// A PCHContainerReader implementation that uses LLVM to
/// wraps Clang modules inside a COFF, ELF, or Mach-O container.
class ObjectFilePCHContainerReader : public PCHContainerReader {
StringRef getFormat() const override { return "obj"; }
/// Initialize an llvm::BitstreamReader with the serialized
/// AST inside the PCH container Buffer.
void ExtractPCH(llvm::MemoryBufferRef Buffer,
llvm::BitstreamReader &StreamFile) const override;
};
}
#endif

View File

@ -369,6 +369,9 @@ def fmodules_local_submodule_visibility :
Flag<["-"], "fmodules-local-submodule-visibility">,
HelpText<"Enforce name visibility rules across submodules of the same "
"top-level module.">;
def fmodule_format_EQ : Joined<["-"], "fmodule-format=">,
HelpText<"Select the container format for clang modules and PCH. "
"Supported options are 'raw' and 'obj'.">;
def fno_modules_hide_internal_linkage :
Flag<["-"], "fno-modules-hide-internal-linkage">,
HelpText<"Make all declarations visible to redeclaration lookup, "

View File

@ -1351,6 +1351,13 @@ def mno_altivec : Flag<["-"], "mno-altivec">, Alias<fno_altivec>;
def mvx : Flag<["-"], "mvx">, Group<m_Group>;
def mno_vx : Flag<["-"], "mno-vx">, Group<m_Group>;
def fzvector : Flag<["-"], "fzvector">, Group<f_Group>, Flags<[CC1Option]>,
HelpText<"Enable System z vector language extension">;
def fno_zvector : Flag<["-"], "fno-zvector">, Group<f_Group>,
Flags<[CC1Option]>;
def mzvector : Flag<["-"], "mzvector">, Alias<fzvector>;
def mno_zvector : Flag<["-"], "mno-zvector">, Alias<fno_zvector>;
def mno_warn_nonportable_cfstrings : Flag<["-"], "mno-warn-nonportable-cfstrings">, Group<m_Group>;
def mno_omit_leaf_frame_pointer : Flag<["-"], "mno-omit-leaf-frame-pointer">, Group<m_Group>;
def momit_leaf_frame_pointer : Flag<["-"], "momit-leaf-frame-pointer">, Group<m_Group>,

View File

@ -57,6 +57,7 @@ class FileManager;
class HeaderSearch;
class Preprocessor;
class PCHContainerOperations;
class PCHContainerReader;
class SourceManager;
class TargetInfo;
class ASTFrontendAction;
@ -725,8 +726,7 @@ class ASTUnit : public ModuleLoader {
///
/// \returns - The initialized ASTUnit or null if the AST failed to load.
static std::unique_ptr<ASTUnit> LoadFromASTFile(
const std::string &Filename,
std::shared_ptr<PCHContainerOperations> PCHContainerOps,
const std::string &Filename, const PCHContainerReader &PCHContainerRdr,
IntrusiveRefCntPtr<DiagnosticsEngine> Diags,
const FileSystemOptions &FileSystemOpts, bool OnlyLocalDecls = false,
ArrayRef<RemappedFile> RemappedFiles = None,

View File

@ -183,7 +183,7 @@ class CompilerInstance : public ModuleLoader {
public:
explicit CompilerInstance(
std::shared_ptr<PCHContainerOperations> PCHContainerOps =
std::make_shared<RawPCHContainerOperations>(),
std::make_shared<PCHContainerOperations>(),
bool BuildingModule = false);
~CompilerInstance() override;
@ -508,6 +508,34 @@ class CompilerInstance : public ModuleLoader {
return ThePCHContainerOperations;
}
/// Return the appropriate PCHContainerWriter depending on the
/// current CodeGenOptions.
const PCHContainerWriter &getPCHContainerWriter() const {
assert(Invocation && "cannot determine module format without invocation");
StringRef Format = getHeaderSearchOpts().ModuleFormat;
auto *Writer = ThePCHContainerOperations->getWriterOrNull(Format);
if (!Writer) {
if (Diagnostics)
Diagnostics->Report(diag::err_module_format_unhandled) << Format;
llvm::report_fatal_error("unknown module format");
}
return *Writer;
}
/// Return the appropriate PCHContainerReader depending on the
/// current CodeGenOptions.
const PCHContainerReader &getPCHContainerReader() const {
assert(Invocation && "cannot determine module format without invocation");
StringRef Format = getHeaderSearchOpts().ModuleFormat;
auto *Reader = ThePCHContainerOperations->getReaderOrNull(Format);
if (!Reader) {
if (Diagnostics)
Diagnostics->Report(diag::err_module_format_unhandled) << Format;
llvm::report_fatal_error("unknown module format");
}
return *Reader;
}
/// }
/// @name Code Completion
/// {
@ -621,7 +649,7 @@ class CompilerInstance : public ModuleLoader {
static IntrusiveRefCntPtr<ASTReader> createPCHExternalASTSource(
StringRef Path, StringRef Sysroot, bool DisablePCHValidation,
bool AllowPCHWithCompilerErrors, Preprocessor &PP, ASTContext &Context,
const PCHContainerOperations &PCHContainerOps,
const PCHContainerReader &PCHContainerRdr,
void *DeserializationListener, bool OwnDeserializationListener,
bool Preamble, bool UseGlobalModuleIndex);

View File

@ -11,6 +11,7 @@
#define LLVM_CLANG_PCH_CONTAINER_OPERATIONS_H
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/StringMap.h"
#include "llvm/Support/MemoryBuffer.h"
#include <memory>
@ -19,6 +20,8 @@ class raw_pwrite_stream;
class BitstreamReader;
}
using llvm::StringRef;
namespace clang {
class ASTConsumer;
@ -33,14 +36,16 @@ struct PCHBuffer {
bool IsComplete;
llvm::SmallVector<char, 0> Data;
};
/// This abstract interface provides operations for creating
/// containers for serialized ASTs (precompiled headers and clang
/// modules).
class PCHContainerWriter {
public:
virtual ~PCHContainerWriter() = 0;
virtual StringRef getFormat() const = 0;
/// \brief This abstract interface provides operations for creating
/// and unwrapping containers for serialized ASTs (precompiled headers
/// and clang modules).
class PCHContainerOperations {
public:
virtual ~PCHContainerOperations();
/// \brief Return an ASTConsumer that can be chained with a
/// Return an ASTConsumer that can be chained with a
/// PCHGenerator that produces a wrapper file format containing a
/// serialized AST bitstream.
virtual std::unique_ptr<ASTConsumer> CreatePCHContainerGenerator(
@ -49,16 +54,28 @@ class PCHContainerOperations {
const LangOptions &LO, const std::string &MainFileName,
const std::string &OutputFileName, llvm::raw_pwrite_stream *OS,
std::shared_ptr<PCHBuffer> Buffer) const = 0;
};
/// \brief Initialize an llvm::BitstreamReader with the serialized AST inside
/// This abstract interface provides operations for unwrapping
/// containers for serialized ASTs (precompiled headers and clang
/// modules).
class PCHContainerReader {
public:
virtual ~PCHContainerReader() = 0;
/// Equivalent to the format passed to -fmodule-format=
virtual StringRef getFormat() const = 0;
/// Initialize an llvm::BitstreamReader with the serialized AST inside
/// the PCH container Buffer.
virtual void ExtractPCH(llvm::MemoryBufferRef Buffer,
llvm::BitstreamReader &StreamFile) const = 0;
};
/// \brief Implements a raw pass-through PCH container.
class RawPCHContainerOperations : public PCHContainerOperations {
/// \brief Return an ASTConsumer that can be chained with a
/// Implements write operations for a raw pass-through PCH container.
class RawPCHContainerWriter : public PCHContainerWriter {
StringRef getFormat() const override { return "raw"; }
/// Return an ASTConsumer that can be chained with a
/// PCHGenerator that writes the module to a flat file.
std::unique_ptr<ASTConsumer> CreatePCHContainerGenerator(
DiagnosticsEngine &Diags, const HeaderSearchOptions &HSO,
@ -66,11 +83,42 @@ class RawPCHContainerOperations : public PCHContainerOperations {
const LangOptions &LO, const std::string &MainFileName,
const std::string &OutputFileName, llvm::raw_pwrite_stream *OS,
std::shared_ptr<PCHBuffer> Buffer) const override;
};
/// \brief Initialize an llvm::BitstreamReader with Buffer.
/// Implements read operations for a raw pass-through PCH container.
class RawPCHContainerReader : public PCHContainerReader {
StringRef getFormat() const override { return "raw"; }
/// Initialize an llvm::BitstreamReader with Buffer.
void ExtractPCH(llvm::MemoryBufferRef Buffer,
llvm::BitstreamReader &StreamFile) const override;
};
/// A registry of PCHContainerWriter and -Reader objects for different formats.
class PCHContainerOperations {
llvm::StringMap<std::unique_ptr<PCHContainerWriter>> Writers;
llvm::StringMap<std::unique_ptr<PCHContainerReader>> Readers;
public:
/// Automatically registers a RawPCHContainerWriter and
/// RawPCHContainerReader.
PCHContainerOperations();
void registerWriter(std::unique_ptr<PCHContainerWriter> Writer) {
Writers[Writer->getFormat()] = std::move(Writer);
}
void registerReader(std::unique_ptr<PCHContainerReader> Reader) {
Readers[Reader->getFormat()] = std::move(Reader);
}
const PCHContainerWriter *getWriterOrNull(StringRef Format) {
return Writers[Format].get();
}
const PCHContainerReader *getReaderOrNull(StringRef Format) {
return Readers[Format].get();
}
const PCHContainerReader &getRawReader() {
return *getReaderOrNull("raw");
}
};
}
#endif

View File

@ -45,7 +45,7 @@ class HeaderSearch;
class HeaderSearchOptions;
class IdentifierTable;
class LangOptions;
class PCHContainerOperations;
class PCHContainerReader;
class Preprocessor;
class PreprocessorOptions;
class PreprocessorOutputOptions;
@ -63,7 +63,7 @@ void ApplyHeaderSearchOptions(HeaderSearch &HS,
/// InitializePreprocessor - Initialize the preprocessor getting it and the
/// environment ready to process a single file.
void InitializePreprocessor(Preprocessor &PP, const PreprocessorOptions &PPOpts,
const PCHContainerOperations &PCHContainerOps,
const PCHContainerReader &PCHContainerRdr,
const FrontendOptions &FEOpts);
/// DoPrintPreprocessedInput - Implement -E mode.

View File

@ -92,6 +92,9 @@ class HeaderSearchOptions : public RefCountedBase<HeaderSearchOptions> {
/// \brief The directory used for a user build.
std::string ModuleUserBuildPath;
/// The module/pch container format.
std::string ModuleFormat;
/// \brief Whether we should disable the use of the hash string within the
/// module cache.
///
@ -167,16 +170,14 @@ class HeaderSearchOptions : public RefCountedBase<HeaderSearchOptions> {
public:
HeaderSearchOptions(StringRef _Sysroot = "/")
: Sysroot(_Sysroot), DisableModuleHash(0), ImplicitModuleMaps(0),
ModuleMapFileHomeIsCwd(0),
ModuleCachePruneInterval(7*24*60*60),
ModuleCachePruneAfter(31*24*60*60),
BuildSessionTimestamp(0),
UseBuiltinIncludes(true),
UseStandardSystemIncludes(true), UseStandardCXXIncludes(true),
UseLibcxx(false), Verbose(false),
ModulesValidateOncePerBuildSession(false),
ModulesValidateSystemHeaders(false) {}
: Sysroot(_Sysroot), ModuleFormat("raw"), DisableModuleHash(0),
ImplicitModuleMaps(0), ModuleMapFileHomeIsCwd(0),
ModuleCachePruneInterval(7 * 24 * 60 * 60),
ModuleCachePruneAfter(31 * 24 * 60 * 60), BuildSessionTimestamp(0),
UseBuiltinIncludes(true), UseStandardSystemIncludes(true),
UseStandardCXXIncludes(true), UseLibcxx(false), Verbose(false),
ModulesValidateOncePerBuildSession(false),
ModulesValidateSystemHeaders(false) {}
/// AddPath - Add the \p Path path to the specified \p Group list.
void AddPath(StringRef Path, frontend::IncludeDirGroup Group,

View File

@ -108,12 +108,13 @@ class Parser : public CodeCompletionHandler {
/// Ident_super - IdentifierInfo for "super", to support fast
/// comparison.
IdentifierInfo *Ident_super;
/// Ident_vector, Ident_pixel, Ident_bool - cached IdentifierInfo's
/// for "vector", "pixel", and "bool" fast comparison. Only present
/// if AltiVec enabled.
/// Ident_vector, Ident_bool - cached IdentifierInfos for "vector" and
/// "bool" fast comparison. Only present if AltiVec or ZVector are enabled.
IdentifierInfo *Ident_vector;
IdentifierInfo *Ident_pixel;
IdentifierInfo *Ident_bool;
/// Ident_pixel - cached IdentifierInfos for "pixel" fast comparison.
/// Only present if AltiVec enabled.
IdentifierInfo *Ident_pixel;
/// Objective-C contextual keywords.
mutable IdentifierInfo *Ident_instancetype;
@ -605,10 +606,12 @@ class Parser : public CodeCompletionHandler {
bool TryAltiVecToken(DeclSpec &DS, SourceLocation Loc,
const char *&PrevSpec, unsigned &DiagID,
bool &isInvalid) {
if (!getLangOpts().AltiVec ||
(Tok.getIdentifierInfo() != Ident_vector &&
Tok.getIdentifierInfo() != Ident_pixel &&
Tok.getIdentifierInfo() != Ident_bool))
if (!getLangOpts().AltiVec && !getLangOpts().ZVector)
return false;
if (Tok.getIdentifierInfo() != Ident_vector &&
Tok.getIdentifierInfo() != Ident_bool &&
(!getLangOpts().AltiVec || Tok.getIdentifierInfo() != Ident_pixel))
return false;
return TryAltiVecTokenOutOfLine(DS, Loc, PrevSpec, DiagID, isInvalid);
@ -618,7 +621,7 @@ class Parser : public CodeCompletionHandler {
/// identifier token, replacing it with the non-context-sensitive __vector.
/// This returns true if the token was replaced.
bool TryAltiVecVectorToken() {
if (!getLangOpts().AltiVec ||
if ((!getLangOpts().AltiVec && !getLangOpts().ZVector) ||
Tok.getIdentifierInfo() != Ident_vector) return false;
return TryAltiVecVectorTokenOutOfLine();
}

View File

@ -729,27 +729,12 @@ class Sema {
/// \brief The declaration of the Objective-C NSArray class.
ObjCInterfaceDecl *NSArrayDecl;
/// \brief Pointer to NSMutableArray type (NSMutableArray *).
QualType NSMutableArrayPointer;
/// \brief The declaration of the arrayWithObjects:count: method.
ObjCMethodDecl *ArrayWithObjectsMethod;
/// \brief The declaration of the Objective-C NSDictionary class.
ObjCInterfaceDecl *NSDictionaryDecl;
/// \brief Pointer to NSMutableDictionary type (NSMutableDictionary *).
QualType NSMutableDictionaryPointer;
/// \brief Pointer to NSMutableSet type (NSMutableSet *).
QualType NSMutableSetPointer;
/// \brief Pointer to NSCountedSet type (NSCountedSet *).
QualType NSCountedSetPointer;
/// \brief Pointer to NSMutableOrderedSet type (NSMutableOrderedSet *).
QualType NSMutableOrderedSetPointer;
/// \brief The declaration of the dictionaryWithObjects:forKeys:count: method.
ObjCMethodDecl *DictionaryWithObjectsMethod;
@ -8363,7 +8348,8 @@ class Sema {
/// type checking for vector binary operators.
QualType CheckVectorOperands(ExprResult &LHS, ExprResult &RHS,
SourceLocation Loc, bool IsCompAssign);
SourceLocation Loc, bool IsCompAssign,
bool AllowBothBool, bool AllowBoolConversion);
QualType GetSignedVectorType(QualType V);
QualType CheckVectorCompareOperands(ExprResult &LHS, ExprResult &RHS,
SourceLocation Loc, bool isRelational);

View File

@ -362,7 +362,7 @@ class ASTReader
SourceManager &SourceMgr;
FileManager &FileMgr;
const PCHContainerOperations &PCHContainerOps;
const PCHContainerReader &PCHContainerRdr;
DiagnosticsEngine &Diags;
/// \brief The semantic analysis object that will be processing the
@ -1289,7 +1289,7 @@ class ASTReader
/// \param ReadTimer If non-null, a timer used to track the time spent
/// deserializing.
ASTReader(Preprocessor &PP, ASTContext &Context,
const PCHContainerOperations &PCHContainerOps,
const PCHContainerReader &PCHContainerRdr,
StringRef isysroot = "", bool DisableValidation = false,
bool AllowASTWithCompilerErrors = false,
bool AllowConfigurationMismatch = false,
@ -1458,7 +1458,7 @@ class ASTReader
/// the AST file, without actually loading the AST file.
static std::string
getOriginalSourceFile(const std::string &ASTFileName, FileManager &FileMgr,
const PCHContainerOperations &PCHContainerOps,
const PCHContainerReader &PCHContainerRdr,
DiagnosticsEngine &Diags);
/// \brief Read the control block for the named AST file.
@ -1466,13 +1466,13 @@ class ASTReader
/// \returns true if an error occurred, false otherwise.
static bool
readASTFileControlBlock(StringRef Filename, FileManager &FileMgr,
const PCHContainerOperations &PCHContainerOps,
const PCHContainerReader &PCHContainerRdr,
ASTReaderListener &Listener);
/// \brief Determine whether the given AST file is acceptable to load into a
/// translation unit with the given language and target options.
static bool isAcceptableASTFile(StringRef Filename, FileManager &FileMgr,
const PCHContainerOperations &PCHContainerOps,
const PCHContainerReader &PCHContainerRdr,
const LangOptions &LangOpts,
const TargetOptions &TargetOpts,
const PreprocessorOptions &PPOpts,

View File

@ -198,10 +198,9 @@ class GlobalModuleIndex {
/// \param Path The path to the directory containing module files, into
/// which the global index will be written.
static ErrorCode writeIndex(FileManager &FileMgr,
const PCHContainerOperations &PCHContainerOps,
const PCHContainerReader &PCHContainerRdr,
StringRef Path);
};
}
#endif

View File

@ -24,7 +24,7 @@ namespace clang {
class GlobalModuleIndex;
class ModuleMap;
class PCHContainerOperations;
class PCHContainerReader;
namespace serialization {
@ -52,7 +52,7 @@ class ModuleManager {
FileManager &FileMgr;
/// \brief Knows how to unwrap module containers.
const PCHContainerOperations &PCHContainerOps;
const PCHContainerReader &PCHContainerRdr;
/// \brief A lookup of in-memory (virtual file) buffers
llvm::DenseMap<const FileEntry *, std::unique_ptr<llvm::MemoryBuffer>>
@ -118,9 +118,9 @@ class ModuleManager {
typedef std::pair<uint32_t, StringRef> ModuleOffset;
explicit ModuleManager(FileManager &FileMgr,
const PCHContainerOperations &PCHContainerOps);
const PCHContainerReader &PCHContainerRdr);
~ModuleManager();
/// \brief Forward iterator to traverse all loaded modules. This is reverse
/// source-order.
ModuleIterator begin() { return Chain.begin(); }

View File

@ -40,7 +40,7 @@ class RefactoringTool : public ClangTool {
RefactoringTool(const CompilationDatabase &Compilations,
ArrayRef<std::string> SourcePaths,
std::shared_ptr<PCHContainerOperations> PCHContainerOps =
std::make_shared<RawPCHContainerOperations>());
std::make_shared<PCHContainerOperations>());
/// \brief Returns the set of replacements to which replacements should
/// be added during the run of the tool.

View File

@ -150,7 +150,7 @@ inline std::unique_ptr<FrontendActionFactory> newFrontendActionFactory(
bool runToolOnCode(clang::FrontendAction *ToolAction, const Twine &Code,
const Twine &FileName = "input.cc",
std::shared_ptr<PCHContainerOperations> PCHContainerOps =
std::make_shared<RawPCHContainerOperations>());
std::make_shared<PCHContainerOperations>());
/// The first part of the pair is the filename, the second part the
/// file-content.
@ -171,7 +171,7 @@ bool runToolOnCodeWithArgs(
clang::FrontendAction *ToolAction, const Twine &Code,
const std::vector<std::string> &Args, const Twine &FileName = "input.cc",
std::shared_ptr<PCHContainerOperations> PCHContainerOps =
std::make_shared<RawPCHContainerOperations>(),
std::make_shared<PCHContainerOperations>(),
const FileContentMappings &VirtualMappedFiles = FileContentMappings());
/// \brief Builds an AST for 'Code'.
@ -185,7 +185,7 @@ bool runToolOnCodeWithArgs(
std::unique_ptr<ASTUnit>
buildASTFromCode(const Twine &Code, const Twine &FileName = "input.cc",
std::shared_ptr<PCHContainerOperations> PCHContainerOps =
std::make_shared<RawPCHContainerOperations>());
std::make_shared<PCHContainerOperations>());
/// \brief Builds an AST for 'Code' with additional flags.
///
@ -200,7 +200,7 @@ std::unique_ptr<ASTUnit> buildASTFromCodeWithArgs(
const Twine &Code, const std::vector<std::string> &Args,
const Twine &FileName = "input.cc",
std::shared_ptr<PCHContainerOperations> PCHContainerOps =
std::make_shared<RawPCHContainerOperations>());
std::make_shared<PCHContainerOperations>());
/// \brief Utility to run a FrontendAction in a single clang invocation.
class ToolInvocation {
@ -219,7 +219,7 @@ class ToolInvocation {
ToolInvocation(std::vector<std::string> CommandLine, FrontendAction *FAction,
FileManager *Files,
std::shared_ptr<PCHContainerOperations> PCHContainerOps =
std::make_shared<RawPCHContainerOperations>());
std::make_shared<PCHContainerOperations>());
/// \brief Create a tool invocation.
///
@ -288,7 +288,7 @@ class ClangTool {
ClangTool(const CompilationDatabase &Compilations,
ArrayRef<std::string> SourcePaths,
std::shared_ptr<PCHContainerOperations> PCHContainerOps =
std::make_shared<RawPCHContainerOperations>());
std::make_shared<PCHContainerOperations>());
~ClangTool();

View File

@ -167,7 +167,7 @@ static bool HasARCRuntime(CompilerInvocation &origCI) {
static CompilerInvocation *
createInvocationForMigration(CompilerInvocation &origCI,
const PCHContainerOperations &PCHContainerOps) {
const PCHContainerReader &PCHContainerRdr) {
std::unique_ptr<CompilerInvocation> CInvok;
CInvok.reset(new CompilerInvocation(origCI));
PreprocessorOptions &PPOpts = CInvok->getPreprocessorOpts();
@ -180,7 +180,7 @@ createInvocationForMigration(CompilerInvocation &origCI,
new DiagnosticsEngine(DiagID, &origCI.getDiagnosticOpts(),
new IgnoringDiagConsumer()));
std::string OriginalFile = ASTReader::getOriginalSourceFile(
PPOpts.ImplicitPCHInclude, FileMgr, PCHContainerOps, *Diags);
PPOpts.ImplicitPCHInclude, FileMgr, PCHContainerRdr, *Diags);
if (!OriginalFile.empty())
PPOpts.Includes.insert(PPOpts.Includes.begin(), OriginalFile);
PPOpts.ImplicitPCHInclude.clear();
@ -247,7 +247,8 @@ bool arcmt::checkForManualIssues(
assert(!transforms.empty());
std::unique_ptr<CompilerInvocation> CInvok;
CInvok.reset(createInvocationForMigration(origCI, *PCHContainerOps));
CInvok.reset(
createInvocationForMigration(origCI, PCHContainerOps->getRawReader()));
CInvok->getFrontendOpts().Inputs.clear();
CInvok->getFrontendOpts().Inputs.push_back(Input);
@ -517,7 +518,8 @@ MigrationProcess::MigrationProcess(
bool MigrationProcess::applyTransform(TransformFn trans,
RewriteListener *listener) {
std::unique_ptr<CompilerInvocation> CInvok;
CInvok.reset(createInvocationForMigration(OrigCI, *PCHContainerOps));
CInvok.reset(
createInvocationForMigration(OrigCI, PCHContainerOps->getRawReader()));
CInvok->getDiagnosticOpts().IgnoreWarnings = true;
Remapper.applyMappings(CInvok->getPreprocessorOpts());

View File

@ -9,6 +9,7 @@
#include "clang/AST/NSAPI.h"
#include "clang/AST/ASTContext.h"
#include "clang/AST/DeclObjC.h"
#include "clang/AST/Expr.h"
#include "llvm/ADT/StringSwitch.h"
@ -29,7 +30,6 @@ IdentifierInfo *NSAPI::getNSClassId(NSClassIdKindKind K) const {
"NSMutableDictionary",
"NSNumber",
"NSMutableSet",
"NSCountedSet",
"NSMutableOrderedSet",
"NSValue"
};
@ -511,6 +511,26 @@ bool NSAPI::isMacroDefined(StringRef Id) const {
return Ctx.Idents.get(Id).hasMacroDefinition();
}
bool NSAPI::isSubclassOfNSClass(ObjCInterfaceDecl *InterfaceDecl,
NSClassIdKindKind NSClassKind) const {
if (!InterfaceDecl) {
return false;
}
IdentifierInfo *NSClassID = getNSClassId(NSClassKind);
bool IsSubclass = false;
do {
IsSubclass = NSClassID == InterfaceDecl->getIdentifier();
if (IsSubclass) {
break;
}
} while ((InterfaceDecl = InterfaceDecl->getSuperClass()));
return IsSubclass;
}
bool NSAPI::isObjCTypedef(QualType T,
StringRef name, IdentifierInfo *&II) const {
if (!Ctx.getLangOpts().ObjC1)

View File

@ -435,17 +435,19 @@ TypeLoc NestedNameSpecifierLoc::getTypeLoc() const {
namespace {
void Append(char *Start, char *End, char *&Buffer, unsigned &BufferSize,
unsigned &BufferCapacity) {
if (Start == End)
return;
if (BufferSize + (End - Start) > BufferCapacity) {
// Reallocate the buffer.
unsigned NewCapacity
= std::max((unsigned)(BufferCapacity? BufferCapacity * 2
: sizeof(void*) * 2),
(unsigned)(BufferSize + (End - Start)));
unsigned NewCapacity = std::max(
(unsigned)(BufferCapacity ? BufferCapacity * 2 : sizeof(void *) * 2),
(unsigned)(BufferSize + (End - Start)));
char *NewBuffer = static_cast<char *>(malloc(NewCapacity));
memcpy(NewBuffer, Buffer, BufferSize);
if (BufferCapacity)
if (BufferCapacity) {
memcpy(NewBuffer, Buffer, BufferSize);
free(Buffer);
}
Buffer = NewBuffer;
BufferCapacity = NewCapacity;
}

View File

@ -2014,6 +2014,12 @@ static const CXXMethodDecl *computeKeyFunction(ASTContext &Context,
continue;
}
// If the key function is dllimport but the class isn't, then the class has
// no key function. The DLL that exports the key function won't export the
// vtable in this case.
if (MD->hasAttr<DLLImportAttr>() && !RD->hasAttr<DLLImportAttr>())
return nullptr;
// We found it.
return MD;
}

View File

@ -724,6 +724,8 @@ MSAsmStmt::MSAsmStmt(const ASTContext &C, SourceLocation asmloc,
}
static StringRef copyIntoContext(const ASTContext &C, StringRef str) {
if (str.empty())
return StringRef();
size_t size = str.size();
char *buffer = new (C) char[size];
memcpy(buffer, str.data(), size);
@ -1499,6 +1501,12 @@ void OMPLoopDirective::setCounters(ArrayRef<Expr *> A) {
std::copy(A.begin(), A.end(), getCounters().begin());
}
void OMPLoopDirective::setInits(ArrayRef<Expr *> A) {
assert(A.size() == getCollapsedNumber() &&
"Number of counter inits is not the same as the collapsed number");
std::copy(A.begin(), A.end(), getInits().begin());
}
void OMPLoopDirective::setUpdates(ArrayRef<Expr *> A) {
assert(A.size() == getCollapsedNumber() &&
"Number of counter updates is not the same as the collapsed number");
@ -1664,6 +1672,7 @@ OMPSimdDirective::Create(const ASTContext &C, SourceLocation StartLoc,
Dir->setInit(Exprs.Init);
Dir->setInc(Exprs.Inc);
Dir->setCounters(Exprs.Counters);
Dir->setInits(Exprs.Inits);
Dir->setUpdates(Exprs.Updates);
Dir->setFinals(Exprs.Finals);
return Dir;
@ -1710,6 +1719,7 @@ OMPForDirective::Create(const ASTContext &C, SourceLocation StartLoc,
Dir->setNextLowerBound(Exprs.NLB);
Dir->setNextUpperBound(Exprs.NUB);
Dir->setCounters(Exprs.Counters);
Dir->setInits(Exprs.Inits);
Dir->setUpdates(Exprs.Updates);
Dir->setFinals(Exprs.Finals);
return Dir;
@ -1756,6 +1766,7 @@ OMPForSimdDirective::Create(const ASTContext &C, SourceLocation StartLoc,
Dir->setNextLowerBound(Exprs.NLB);
Dir->setNextUpperBound(Exprs.NUB);
Dir->setCounters(Exprs.Counters);
Dir->setInits(Exprs.Inits);
Dir->setUpdates(Exprs.Updates);
Dir->setFinals(Exprs.Finals);
return Dir;
@ -1911,6 +1922,7 @@ OMPParallelForDirective *OMPParallelForDirective::Create(
Dir->setNextLowerBound(Exprs.NLB);
Dir->setNextUpperBound(Exprs.NUB);
Dir->setCounters(Exprs.Counters);
Dir->setInits(Exprs.Inits);
Dir->setUpdates(Exprs.Updates);
Dir->setFinals(Exprs.Finals);
return Dir;
@ -1955,6 +1967,7 @@ OMPParallelForSimdDirective *OMPParallelForSimdDirective::Create(
Dir->setNextLowerBound(Exprs.NLB);
Dir->setNextUpperBound(Exprs.NUB);
Dir->setCounters(Exprs.Counters);
Dir->setInits(Exprs.Inits);
Dir->setUpdates(Exprs.Updates);
Dir->setFinals(Exprs.Finals);
return Dir;

View File

@ -587,4 +587,6 @@ void FileManager::PrintStats() const {
//llvm::errs() << PagesMapped << BytesOfPagesMapped << FSLookups;
}
PCHContainerOperations::~PCHContainerOperations() {}
// Virtual destructors for abstract base classes that need live in Basic.
PCHContainerWriter::~PCHContainerWriter() {}
PCHContainerReader::~PCHContainerReader() {}

Some files were not shown because too many files have changed in this diff Show More