unsigned ARMBaseInstrInfo::RemoveBranch(MachineBasicBlock &MBB) const {
- MachineBasicBlock::iterator I = MBB.end();
- if (I == MBB.begin()) return 0;
- --I;
- while (I->isDebugValue()) {
- if (I == MBB.begin())
- return 0;
- --I;
- }
+ MachineBasicBlock::iterator I = MBB.getLastNonDebugInstr();
+ if (I == MBB.end())
+ return 0;
+
if (!isUncondBranchOpcode(I->getOpcode()) &&
!isCondBranchOpcode(I->getOpcode()))
return 0;
static bool isCPSRDefined(const MachineInstr *MI) {
for (const auto &MO : MI->operands())
- if (MO.isReg() && MO.getReg() == ARM::CPSR && MO.isDef())
+ if (MO.isReg() && MO.getReg() == ARM::CPSR && MO.isDef() && !MO.isDead())
return true;
return false;
}
MachineFrameInfo &MFI = *MF.getFrameInfo();
unsigned Align = MFI.getObjectAlignment(FI);
- MachineMemOperand *MMO =
- MF.getMachineMemOperand(MachinePointerInfo::getFixedStack(FI),
- MachineMemOperand::MOStore,
- MFI.getObjectSize(FI),
- Align);
+ MachineMemOperand *MMO = MF.getMachineMemOperand(
+ MachinePointerInfo::getFixedStack(MF, FI), MachineMemOperand::MOStore,
+ MFI.getObjectSize(FI), Align);
switch (RC->getSize()) {
case 4:
MachineFunction &MF = *MBB.getParent();
MachineFrameInfo &MFI = *MF.getFrameInfo();
unsigned Align = MFI.getObjectAlignment(FI);
- MachineMemOperand *MMO =
- MF.getMachineMemOperand(
- MachinePointerInfo::getFixedStack(FI),
- MachineMemOperand::MOLoad,
- MFI.getObjectSize(FI),
- Align);
+ MachineMemOperand *MMO = MF.getMachineMemOperand(
+ MachinePointerInfo::getFixedStack(MF, FI), MachineMemOperand::MOLoad,
+ MFI.getObjectSize(FI), Align);
switch (RC->getSize()) {
case 4:
return MI->mayLoad() && hasLoadFromStackSlot(MI, Dummy, FrameIndex);
}
+/// \brief Expands MEMCPY to either LDMIA/STMIA or LDMIA_UPD/STMID_UPD
+/// depending on whether the result is used.
+void ARMBaseInstrInfo::expandMEMCPY(MachineBasicBlock::iterator MBBI) const {
+ bool isThumb1 = Subtarget.isThumb1Only();
+ bool isThumb2 = Subtarget.isThumb2();
+ const ARMBaseInstrInfo *TII = Subtarget.getInstrInfo();
+
+ MachineInstr *MI = MBBI;
+ DebugLoc dl = MI->getDebugLoc();
+ MachineBasicBlock *BB = MI->getParent();
+
+ MachineInstrBuilder LDM, STM;
+ if (isThumb1 || !MI->getOperand(1).isDead()) {
+ LDM = BuildMI(*BB, MI, dl, TII->get(isThumb2 ? ARM::t2LDMIA_UPD
+ : isThumb1 ? ARM::tLDMIA_UPD
+ : ARM::LDMIA_UPD))
+ .addOperand(MI->getOperand(1));
+ } else {
+ LDM = BuildMI(*BB, MI, dl, TII->get(isThumb2 ? ARM::t2LDMIA : ARM::LDMIA));
+ }
+
+ if (isThumb1 || !MI->getOperand(0).isDead()) {
+ STM = BuildMI(*BB, MI, dl, TII->get(isThumb2 ? ARM::t2STMIA_UPD
+ : isThumb1 ? ARM::tSTMIA_UPD
+ : ARM::STMIA_UPD))
+ .addOperand(MI->getOperand(0));
+ } else {
+ STM = BuildMI(*BB, MI, dl, TII->get(isThumb2 ? ARM::t2STMIA : ARM::STMIA));
+ }
+
+ AddDefaultPred(LDM.addOperand(MI->getOperand(3)));
+ AddDefaultPred(STM.addOperand(MI->getOperand(2)));
+
+ // Sort the scratch registers into ascending order.
+ const TargetRegisterInfo &TRI = getRegisterInfo();
+ llvm::SmallVector<unsigned, 6> ScratchRegs;
+ for(unsigned I = 5; I < MI->getNumOperands(); ++I)
+ ScratchRegs.push_back(MI->getOperand(I).getReg());
+ std::sort(ScratchRegs.begin(), ScratchRegs.end(),
+ [&TRI](const unsigned &Reg1,
+ const unsigned &Reg2) -> bool {
+ return TRI.getEncodingValue(Reg1) <
+ TRI.getEncodingValue(Reg2);
+ });
+
+ for (const auto &Reg : ScratchRegs) {
+ LDM.addReg(Reg, RegState::Define);
+ STM.addReg(Reg, RegState::Kill);
+ }
+
+ BB->erase(MBBI);
+}
+
+
bool
ARMBaseInstrInfo::expandPostRAPseudo(MachineBasicBlock::iterator MI) const {
MachineFunction &MF = *MI->getParent()->getParent();
Reloc::Model RM = MF.getTarget().getRelocationModel();
if (MI->getOpcode() == TargetOpcode::LOAD_STACK_GUARD) {
- assert(getSubtarget().getTargetTriple().getObjectFormat() ==
- Triple::MachO &&
+ assert(getSubtarget().getTargetTriple().isOSBinFormatMachO() &&
"LOAD_STACK_GUARD currently supported only for MachO.");
expandLoadStackGuard(MI, RM);
MI->getParent()->erase(MI);
return true;
}
+ if (MI->getOpcode() == ARM::MEMCPY) {
+ expandMEMCPY(MI);
+ return true;
+ }
+
// This hook gets to expand COPY instructions before they become
// copyPhysReg() calls. Look for VMOVS instructions that can legally be
// widened to VMOVD. We prefer the VMOVD when possible because it may be
bool ARMBaseInstrInfo::
isProfitableToIfCvt(MachineBasicBlock &MBB,
unsigned NumCycles, unsigned ExtraPredCycles,
- const BranchProbability &Probability) const {
+ BranchProbability Probability) const {
if (!NumCycles)
return false;
// If we are optimizing for size, see if the branch in the predecessor can be
// lowered to cbn?z by the constant island lowering pass, and return false if
// so. This results in a shorter instruction sequence.
- const Function *F = MBB.getParent()->getFunction();
- if (F->hasFnAttribute(Attribute::OptimizeForSize) ||
- F->hasFnAttribute(Attribute::MinSize)) {
+ if (MBB.getParent()->getFunction()->optForSize()) {
MachineBasicBlock *Pred = *MBB.pred_begin();
if (!Pred->empty()) {
MachineInstr *LastMI = &*Pred->rbegin();
}
// Attempt to estimate the relative costs of predication versus branching.
- unsigned UnpredCost = Probability.getNumerator() * NumCycles;
- UnpredCost /= Probability.getDenominator();
- UnpredCost += 1; // The branch itself
- UnpredCost += Subtarget.getMispredictionPenalty() / 10;
-
- return (NumCycles + ExtraPredCycles) <= UnpredCost;
+ // Here we scale up each component of UnpredCost to avoid precision issue when
+ // scaling NumCycles by Probability.
+ const unsigned ScalingUpFactor = 1024;
+ unsigned UnpredCost = Probability.scale(NumCycles * ScalingUpFactor);
+ UnpredCost += ScalingUpFactor; // The branch itself
+ UnpredCost += Subtarget.getMispredictionPenalty() * ScalingUpFactor / 10;
+
+ return (NumCycles + ExtraPredCycles) * ScalingUpFactor <= UnpredCost;
}
bool ARMBaseInstrInfo::
unsigned TCycles, unsigned TExtra,
MachineBasicBlock &FMBB,
unsigned FCycles, unsigned FExtra,
- const BranchProbability &Probability) const {
+ BranchProbability Probability) const {
if (!TCycles || !FCycles)
return false;
// Attempt to estimate the relative costs of predication versus branching.
- unsigned TUnpredCost = Probability.getNumerator() * TCycles;
- TUnpredCost /= Probability.getDenominator();
-
- uint32_t Comp = Probability.getDenominator() - Probability.getNumerator();
- unsigned FUnpredCost = Comp * FCycles;
- FUnpredCost /= Probability.getDenominator();
-
+ // Here we scale up each component of UnpredCost to avoid precision issue when
+ // scaling TCycles/FCycles by Probability.
+ const unsigned ScalingUpFactor = 1024;
+ unsigned TUnpredCost = Probability.scale(TCycles * ScalingUpFactor);
+ unsigned FUnpredCost =
+ Probability.getCompl().scale(FCycles * ScalingUpFactor);
unsigned UnpredCost = TUnpredCost + FUnpredCost;
- UnpredCost += 1; // The branch itself
- UnpredCost += Subtarget.getMispredictionPenalty() / 10;
+ UnpredCost += 1 * ScalingUpFactor; // The branch itself
+ UnpredCost += Subtarget.getMispredictionPenalty() * ScalingUpFactor / 10;
- return (TCycles + FCycles + TExtra + FExtra) <= UnpredCost;
+ return (TCycles + FCycles + TExtra + FExtra) * ScalingUpFactor <= UnpredCost;
}
bool
llvm_unreachable("Unknown unconditional branch opcode!");
}
-/// commuteInstruction - Handle commutable instructions.
-MachineInstr *
-ARMBaseInstrInfo::commuteInstruction(MachineInstr *MI, bool NewMI) const {
+MachineInstr *ARMBaseInstrInfo::commuteInstructionImpl(MachineInstr *MI,
+ bool NewMI,
+ unsigned OpIdx1,
+ unsigned OpIdx2) const {
switch (MI->getOpcode()) {
case ARM::MOVCCr:
case ARM::t2MOVCCr: {
// MOVCC AL can't be inverted. Shouldn't happen.
if (CC == ARMCC::AL || PredReg != ARM::CPSR)
return nullptr;
- MI = TargetInstrInfo::commuteInstruction(MI, NewMI);
+ MI = TargetInstrInfo::commuteInstructionImpl(MI, NewMI, OpIdx1, OpIdx2);
if (!MI)
return nullptr;
// After swapping the MOVCC operands, also invert the condition.
return MI;
}
}
- return TargetInstrInfo::commuteInstruction(MI, NewMI);
+ return TargetInstrInfo::commuteInstructionImpl(MI, NewMI, OpIdx1, OpIdx2);
}
/// Identify instructions that can be folded into a MOVCC instruction, and
unsigned NumBytes) {
// This optimisation potentially adds lots of load and store
// micro-operations, it's only really a great benefit to code-size.
- if (!MF.getFunction()->hasFnAttribute(Attribute::MinSize))
+ if (!MF.getFunction()->optForMinSize())
return false;
// If only one register is pushed/popped, LLVM can use an LDR/STR
// instructions).
if (Latency > 0 && Subtarget.isThumb2()) {
const MachineFunction *MF = DefMI->getParent()->getParent();
+ // FIXME: Use Function::optForSize().
if (MF->getFunction()->hasFnAttribute(Attribute::OptimizeForSize))
--Latency;
}
MIB = BuildMI(MBB, MI, DL, get(LoadOpc), Reg);
MIB.addReg(Reg, RegState::Kill).addImm(0);
unsigned Flag = MachineMemOperand::MOLoad | MachineMemOperand::MOInvariant;
- MachineMemOperand *MMO = MBB.getParent()->
- getMachineMemOperand(MachinePointerInfo::getGOT(), Flag, 4, 4);
+ MachineMemOperand *MMO = MBB.getParent()->getMachineMemOperand(
+ MachinePointerInfo::getGOT(*MBB.getParent()), Flag, 4, 4);
MIB.addMemOperand(MMO);
AddDefaultPred(MIB);
}
projects / oota-llvm.git / blobdiff