STATISTIC(NumTailCalls, "Number of tail calls");
STATISTIC(NumMovwMovt, "Number of GAs materialized with movw + movt");
+STATISTIC(NumLoopByVals, "Number of loops generated for byval arguments");
// This option should go away when tail calls fully work.
static cl::opt<bool>
ARM::R0, ARM::R1, ARM::R2, ARM::R3
};
-void ARMTargetLowering::addTypeForNEON(EVT VT, EVT PromotedLdStVT,
- EVT PromotedBitwiseVT) {
+void ARMTargetLowering::addTypeForNEON(MVT VT, MVT PromotedLdStVT,
+ MVT PromotedBitwiseVT) {
if (VT != PromotedLdStVT) {
- setOperationAction(ISD::LOAD, VT.getSimpleVT(), Promote);
- AddPromotedToType (ISD::LOAD, VT.getSimpleVT(),
- PromotedLdStVT.getSimpleVT());
+ setOperationAction(ISD::LOAD, VT, Promote);
+ AddPromotedToType (ISD::LOAD, VT, PromotedLdStVT);
- setOperationAction(ISD::STORE, VT.getSimpleVT(), Promote);
- AddPromotedToType (ISD::STORE, VT.getSimpleVT(),
- PromotedLdStVT.getSimpleVT());
+ setOperationAction(ISD::STORE, VT, Promote);
+ AddPromotedToType (ISD::STORE, VT, PromotedLdStVT);
}
- EVT ElemTy = VT.getVectorElementType();
+ MVT ElemTy = VT.getVectorElementType();
if (ElemTy != MVT::i64 && ElemTy != MVT::f64)
- setOperationAction(ISD::SETCC, VT.getSimpleVT(), Custom);
- setOperationAction(ISD::INSERT_VECTOR_ELT, VT.getSimpleVT(), Custom);
- setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT.getSimpleVT(), Custom);
+ setOperationAction(ISD::SETCC, VT, Custom);
+ setOperationAction(ISD::INSERT_VECTOR_ELT, VT, Custom);
+ setOperationAction(ISD::EXTRACT_VECTOR_ELT, VT, Custom);
if (ElemTy == MVT::i32) {
- setOperationAction(ISD::SINT_TO_FP, VT.getSimpleVT(), Custom);
- setOperationAction(ISD::UINT_TO_FP, VT.getSimpleVT(), Custom);
- setOperationAction(ISD::FP_TO_SINT, VT.getSimpleVT(), Custom);
- setOperationAction(ISD::FP_TO_UINT, VT.getSimpleVT(), Custom);
+ setOperationAction(ISD::SINT_TO_FP, VT, Custom);
+ setOperationAction(ISD::UINT_TO_FP, VT, Custom);
+ setOperationAction(ISD::FP_TO_SINT, VT, Custom);
+ setOperationAction(ISD::FP_TO_UINT, VT, Custom);
} else {
- setOperationAction(ISD::SINT_TO_FP, VT.getSimpleVT(), Expand);
- setOperationAction(ISD::UINT_TO_FP, VT.getSimpleVT(), Expand);
- setOperationAction(ISD::FP_TO_SINT, VT.getSimpleVT(), Expand);
- setOperationAction(ISD::FP_TO_UINT, VT.getSimpleVT(), Expand);
- }
- setOperationAction(ISD::BUILD_VECTOR, VT.getSimpleVT(), Custom);
- setOperationAction(ISD::VECTOR_SHUFFLE, VT.getSimpleVT(), Custom);
- setOperationAction(ISD::CONCAT_VECTORS, VT.getSimpleVT(), Legal);
- setOperationAction(ISD::EXTRACT_SUBVECTOR, VT.getSimpleVT(), Legal);
- setOperationAction(ISD::SELECT, VT.getSimpleVT(), Expand);
- setOperationAction(ISD::SELECT_CC, VT.getSimpleVT(), Expand);
- setOperationAction(ISD::SIGN_EXTEND_INREG, VT.getSimpleVT(), Expand);
+ setOperationAction(ISD::SINT_TO_FP, VT, Expand);
+ setOperationAction(ISD::UINT_TO_FP, VT, Expand);
+ setOperationAction(ISD::FP_TO_SINT, VT, Expand);
+ setOperationAction(ISD::FP_TO_UINT, VT, Expand);
+ }
+ setOperationAction(ISD::BUILD_VECTOR, VT, Custom);
+ setOperationAction(ISD::VECTOR_SHUFFLE, VT, Custom);
+ setOperationAction(ISD::CONCAT_VECTORS, VT, Legal);
+ setOperationAction(ISD::EXTRACT_SUBVECTOR, VT, Legal);
+ setOperationAction(ISD::SELECT, VT, Expand);
+ setOperationAction(ISD::SELECT_CC, VT, Expand);
+ setOperationAction(ISD::SIGN_EXTEND_INREG, VT, Expand);
if (VT.isInteger()) {
- setOperationAction(ISD::SHL, VT.getSimpleVT(), Custom);
- setOperationAction(ISD::SRA, VT.getSimpleVT(), Custom);
- setOperationAction(ISD::SRL, VT.getSimpleVT(), Custom);
+ setOperationAction(ISD::SHL, VT, Custom);
+ setOperationAction(ISD::SRA, VT, Custom);
+ setOperationAction(ISD::SRL, VT, Custom);
}
// Promote all bit-wise operations.
if (VT.isInteger() && VT != PromotedBitwiseVT) {
- setOperationAction(ISD::AND, VT.getSimpleVT(), Promote);
- AddPromotedToType (ISD::AND, VT.getSimpleVT(),
- PromotedBitwiseVT.getSimpleVT());
- setOperationAction(ISD::OR, VT.getSimpleVT(), Promote);
- AddPromotedToType (ISD::OR, VT.getSimpleVT(),
- PromotedBitwiseVT.getSimpleVT());
- setOperationAction(ISD::XOR, VT.getSimpleVT(), Promote);
- AddPromotedToType (ISD::XOR, VT.getSimpleVT(),
- PromotedBitwiseVT.getSimpleVT());
+ setOperationAction(ISD::AND, VT, Promote);
+ AddPromotedToType (ISD::AND, VT, PromotedBitwiseVT);
+ setOperationAction(ISD::OR, VT, Promote);
+ AddPromotedToType (ISD::OR, VT, PromotedBitwiseVT);
+ setOperationAction(ISD::XOR, VT, Promote);
+ AddPromotedToType (ISD::XOR, VT, PromotedBitwiseVT);
}
// Neon does not support vector divide/remainder operations.
- setOperationAction(ISD::SDIV, VT.getSimpleVT(), Expand);
- setOperationAction(ISD::UDIV, VT.getSimpleVT(), Expand);
- setOperationAction(ISD::FDIV, VT.getSimpleVT(), Expand);
- setOperationAction(ISD::SREM, VT.getSimpleVT(), Expand);
- setOperationAction(ISD::UREM, VT.getSimpleVT(), Expand);
- setOperationAction(ISD::FREM, VT.getSimpleVT(), Expand);
+ setOperationAction(ISD::SDIV, VT, Expand);
+ setOperationAction(ISD::UDIV, VT, Expand);
+ setOperationAction(ISD::FDIV, VT, Expand);
+ setOperationAction(ISD::SREM, VT, Expand);
+ setOperationAction(ISD::UREM, VT, Expand);
+ setOperationAction(ISD::FREM, VT, Expand);
}
-void ARMTargetLowering::addDRTypeForNEON(EVT VT) {
+void ARMTargetLowering::addDRTypeForNEON(MVT VT) {
addRegisterClass(VT, &ARM::DPRRegClass);
addTypeForNEON(VT, MVT::f64, MVT::v2i32);
}
-void ARMTargetLowering::addQRTypeForNEON(EVT VT) {
+void ARMTargetLowering::addQRTypeForNEON(MVT VT) {
addRegisterClass(VT, &ARM::QPRRegClass);
addTypeForNEON(VT, MVT::v2f64, MVT::v4i32);
}
}
}
+ // ARM and Thumb2 support UMLAL/SMLAL.
+ if (!Subtarget->isThumb1Only())
+ setTargetDAGCombine(ISD::ADDC);
+
+
computeRegisterProperties();
// ARM does not have f32 extending load.
benefitFromCodePlacementOpt = true;
+ // Prefer likely predicted branches to selects on out-of-order cores.
+ predictableSelectIsExpensive = Subtarget->isCortexA9();
+
setMinFunctionAlignment(Subtarget->isThumb() ? 1 : 2);
}
case ARMISD::RET_FLAG: return "ARMISD::RET_FLAG";
case ARMISD::PIC_ADD: return "ARMISD::PIC_ADD";
case ARMISD::CMP: return "ARMISD::CMP";
+ case ARMISD::CMN: return "ARMISD::CMN";
case ARMISD::CMPZ: return "ARMISD::CMPZ";
case ARMISD::CMPFP: return "ARMISD::CMPFP";
case ARMISD::CMPFPw0: return "ARMISD::CMPFPw0";
case ARMISD::FMSTAT: return "ARMISD::FMSTAT";
case ARMISD::CMOV: return "ARMISD::CMOV";
- case ARMISD::CAND: return "ARMISD::CAND";
- case ARMISD::COR: return "ARMISD::COR";
- case ARMISD::CXOR: return "ARMISD::CXOR";
case ARMISD::RBIT: return "ARMISD::RBIT";
case ARMISD::VTBL2: return "ARMISD::VTBL2";
case ARMISD::VMULLs: return "ARMISD::VMULLs";
case ARMISD::VMULLu: return "ARMISD::VMULLu";
+ case ARMISD::UMLAL: return "ARMISD::UMLAL";
+ case ARMISD::SMLAL: return "ARMISD::SMLAL";
case ARMISD::BUILD_VECTOR: return "ARMISD::BUILD_VECTOR";
case ARMISD::FMAX: return "ARMISD::FMAX";
case ARMISD::FMIN: return "ARMISD::FMIN";
// Create a fast isel object.
FastISel *
-ARMTargetLowering::createFastISel(FunctionLoweringInfo &funcInfo) const {
- return ARM::createFastISel(funcInfo);
+ARMTargetLowering::createFastISel(FunctionLoweringInfo &funcInfo,
+ const TargetLibraryInfo *libInfo) const {
+ return ARM::createFastISel(funcInfo, libInfo);
}
/// getMaximalGlobalOffset - Returns the maximal possible offset which can
return (Return ? RetCC_ARM_AAPCS : CC_ARM_AAPCS);
case CallingConv::ARM_APCS:
return (Return ? RetCC_ARM_APCS : CC_ARM_APCS);
+ case CallingConv::GHC:
+ return (Return ? RetCC_ARM_APCS : CC_ARM_APCS_GHC);
}
}
/// ARMISD:CALL <- callseq_end chain. Also add input and output parameter
/// nodes.
SDValue
-ARMTargetLowering::LowerCall(SDValue Chain, SDValue Callee,
- CallingConv::ID CallConv, bool isVarArg,
- bool doesNotRet, bool &isTailCall,
- const SmallVectorImpl<ISD::OutputArg> &Outs,
- const SmallVectorImpl<SDValue> &OutVals,
- const SmallVectorImpl<ISD::InputArg> &Ins,
- DebugLoc dl, SelectionDAG &DAG,
+ARMTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
SmallVectorImpl<SDValue> &InVals) const {
+ SelectionDAG &DAG = CLI.DAG;
+ DebugLoc &dl = CLI.DL;
+ SmallVector<ISD::OutputArg, 32> &Outs = CLI.Outs;
+ SmallVector<SDValue, 32> &OutVals = CLI.OutVals;
+ SmallVector<ISD::InputArg, 32> &Ins = CLI.Ins;
+ SDValue Chain = CLI.Chain;
+ SDValue Callee = CLI.Callee;
+ bool &isTailCall = CLI.IsTailCall;
+ CallingConv::ID CallConv = CLI.CallConv;
+ bool doesNotRet = CLI.DoesNotReturn;
+ bool isVarArg = CLI.IsVarArg;
+
MachineFunction &MF = DAG.getMachineFunction();
bool IsStructRet = (Outs.empty()) ? false : Outs[0].Flags.isSRet();
bool IsSibCall = false;
CCInfo.clearFirstByValReg();
}
- unsigned LocMemOffset = VA.getLocMemOffset();
- SDValue StkPtrOff = DAG.getIntPtrConstant(LocMemOffset);
- SDValue Dst = DAG.getNode(ISD::ADD, dl, getPointerTy(), StackPtr,
- StkPtrOff);
- SDValue SrcOffset = DAG.getIntPtrConstant(4*offset);
- SDValue Src = DAG.getNode(ISD::ADD, dl, getPointerTy(), Arg, SrcOffset);
- SDValue SizeNode = DAG.getConstant(Flags.getByValSize() - 4*offset,
- MVT::i32);
- MemOpChains.push_back(DAG.getMemcpy(Chain, dl, Dst, Src, SizeNode,
- Flags.getByValAlign(),
- /*isVolatile=*/false,
- /*AlwaysInline=*/false,
- MachinePointerInfo(0),
- MachinePointerInfo(0)));
-
+ if (Flags.getByValSize() - 4*offset > 0) {
+ unsigned LocMemOffset = VA.getLocMemOffset();
+ SDValue StkPtrOff = DAG.getIntPtrConstant(LocMemOffset);
+ SDValue Dst = DAG.getNode(ISD::ADD, dl, getPointerTy(), StackPtr,
+ StkPtrOff);
+ SDValue SrcOffset = DAG.getIntPtrConstant(4*offset);
+ SDValue Src = DAG.getNode(ISD::ADD, dl, getPointerTy(), Arg, SrcOffset);
+ SDValue SizeNode = DAG.getConstant(Flags.getByValSize() - 4*offset,
+ MVT::i32);
+ SDValue AlignNode = DAG.getConstant(Flags.getByValAlign(), MVT::i32);
+
+ SDVTList VTs = DAG.getVTList(MVT::Other, MVT::Glue);
+ SDValue Ops[] = { Chain, Dst, Src, SizeNode, AlignNode};
+ MemOpChains.push_back(DAG.getNode(ARMISD::COPY_STRUCT_BYVAL, dl, VTs,
+ Ops, array_lengthof(Ops)));
+ }
} else if (!IsSibCall) {
assert(VA.isMemLoc());
Entry.Ty = (Type *) Type::getInt32Ty(*DAG.getContext());
Args.push_back(Entry);
// FIXME: is there useful debug info available here?
- std::pair<SDValue, SDValue> CallResult =
- LowerCallTo(Chain, (Type *) Type::getInt32Ty(*DAG.getContext()),
+ TargetLowering::CallLoweringInfo CLI(Chain,
+ (Type *) Type::getInt32Ty(*DAG.getContext()),
false, false, false, false,
0, CallingConv::C, /*isTailCall=*/false,
/*doesNotRet=*/false, /*isReturnValueUsed=*/true,
DAG.getExternalSymbol("__tls_get_addr", PtrVT), Args, DAG, dl);
+ std::pair<SDValue, SDValue> CallResult = LowerCallTo(CLI);
return CallResult.first;
}
// Record this extraction against the appropriate vector if possible...
SDValue SourceVec = V.getOperand(0);
+ // If the element number isn't a constant, we can't effectively
+ // analyze what's going on.
+ if (!isa<ConstantSDNode>(V.getOperand(1)))
+ return SDValue();
unsigned EltNo = cast<ConstantSDNode>(V.getOperand(1))->getZExtValue();
bool FoundSource = false;
for (unsigned j = 0; j < SourceVecs.size(); ++j) {
exitMBB->transferSuccessorsAndUpdatePHIs(BB);
const TargetRegisterClass *TRC = isThumb2 ?
- (const TargetRegisterClass*)&ARM::tGPRRegClass :
+ (const TargetRegisterClass*)&ARM::rGPRRegClass :
(const TargetRegisterClass*)&ARM::GPRRegClass;
unsigned scratch = MRI.createVirtualRegister(TRC);
unsigned scratch2 = (!BinOpcode) ? incr : MRI.createVirtualRegister(TRC);
exitMBB->transferSuccessorsAndUpdatePHIs(BB);
const TargetRegisterClass *TRC = isThumb2 ?
- (const TargetRegisterClass*)&ARM::tGPRRegClass :
+ (const TargetRegisterClass*)&ARM::rGPRRegClass :
(const TargetRegisterClass*)&ARM::GPRRegClass;
unsigned scratch = MRI.createVirtualRegister(TRC);
unsigned scratch2 = MRI.createVirtualRegister(TRC);
const TargetRegisterClass *TRC = Subtarget->isThumb() ?
(const TargetRegisterClass*)&ARM::tGPRRegClass :
- (const TargetRegisterClass*)&ARM::GPRRegClass;
+ (const TargetRegisterClass*)&ARM::GPRnopcRegClass;
// Get a mapping of the call site numbers to all of the landing pads they're
// associated with.
}
// Add the jump table entries as successors to the MBB.
- MachineBasicBlock *PrevMBB = 0;
+ SmallPtrSet<MachineBasicBlock*, 8> SeenMBBs;
for (std::vector<MachineBasicBlock*>::iterator
I = LPadList.begin(), E = LPadList.end(); I != E; ++I) {
MachineBasicBlock *CurMBB = *I;
- if (PrevMBB != CurMBB)
+ if (SeenMBBs.insert(CurMBB))
DispContBB->addSuccessor(CurMBB);
- PrevMBB = CurMBB;
}
// N.B. the order the invoke BBs are processed in doesn't matter here.
llvm_unreachable("Expecting a BB with two successors!");
}
+MachineBasicBlock *ARMTargetLowering::
+EmitStructByval(MachineInstr *MI, MachineBasicBlock *BB) const {
+ // This pseudo instruction has 3 operands: dst, src, size
+ // We expand it to a loop if size > Subtarget->getMaxInlineSizeThreshold().
+ // Otherwise, we will generate unrolled scalar copies.
+ const TargetInstrInfo *TII = getTargetMachine().getInstrInfo();
+ const BasicBlock *LLVM_BB = BB->getBasicBlock();
+ MachineFunction::iterator It = BB;
+ ++It;
+
+ unsigned dest = MI->getOperand(0).getReg();
+ unsigned src = MI->getOperand(1).getReg();
+ unsigned SizeVal = MI->getOperand(2).getImm();
+ unsigned Align = MI->getOperand(3).getImm();
+ DebugLoc dl = MI->getDebugLoc();
+
+ bool isThumb2 = Subtarget->isThumb2();
+ MachineFunction *MF = BB->getParent();
+ MachineRegisterInfo &MRI = MF->getRegInfo();
+ unsigned ldrOpc, strOpc, UnitSize = 0;
+
+ const TargetRegisterClass *TRC = isThumb2 ?
+ (const TargetRegisterClass*)&ARM::tGPRRegClass :
+ (const TargetRegisterClass*)&ARM::GPRRegClass;
+ const TargetRegisterClass *TRC_Vec = 0;
+
+ if (Align & 1) {
+ ldrOpc = isThumb2 ? ARM::t2LDRB_POST : ARM::LDRB_POST_IMM;
+ strOpc = isThumb2 ? ARM::t2STRB_POST : ARM::STRB_POST_IMM;
+ UnitSize = 1;
+ } else if (Align & 2) {
+ ldrOpc = isThumb2 ? ARM::t2LDRH_POST : ARM::LDRH_POST;
+ strOpc = isThumb2 ? ARM::t2STRH_POST : ARM::STRH_POST;
+ UnitSize = 2;
+ } else {
+ // Check whether we can use NEON instructions.
+ if (!MF->getFunction()->hasFnAttr(Attribute::NoImplicitFloat) &&
+ Subtarget->hasNEON()) {
+ if ((Align % 16 == 0) && SizeVal >= 16) {
+ ldrOpc = ARM::VLD1q32wb_fixed;
+ strOpc = ARM::VST1q32wb_fixed;
+ UnitSize = 16;
+ TRC_Vec = (const TargetRegisterClass*)&ARM::DPairRegClass;
+ }
+ else if ((Align % 8 == 0) && SizeVal >= 8) {
+ ldrOpc = ARM::VLD1d32wb_fixed;
+ strOpc = ARM::VST1d32wb_fixed;
+ UnitSize = 8;
+ TRC_Vec = (const TargetRegisterClass*)&ARM::DPRRegClass;
+ }
+ }
+ // Can't use NEON instructions.
+ if (UnitSize == 0) {
+ ldrOpc = isThumb2 ? ARM::t2LDR_POST : ARM::LDR_POST_IMM;
+ strOpc = isThumb2 ? ARM::t2STR_POST : ARM::STR_POST_IMM;
+ UnitSize = 4;
+ }
+ }
+
+ unsigned BytesLeft = SizeVal % UnitSize;
+ unsigned LoopSize = SizeVal - BytesLeft;
+
+ if (SizeVal <= Subtarget->getMaxInlineSizeThreshold()) {
+ // Use LDR and STR to copy.
+ // [scratch, srcOut] = LDR_POST(srcIn, UnitSize)
+ // [destOut] = STR_POST(scratch, destIn, UnitSize)
+ unsigned srcIn = src;
+ unsigned destIn = dest;
+ for (unsigned i = 0; i < LoopSize; i+=UnitSize) {
+ unsigned scratch = MRI.createVirtualRegister(UnitSize >= 8 ? TRC_Vec:TRC);
+ unsigned srcOut = MRI.createVirtualRegister(TRC);
+ unsigned destOut = MRI.createVirtualRegister(TRC);
+ if (UnitSize >= 8) {
+ AddDefaultPred(BuildMI(*BB, MI, dl,
+ TII->get(ldrOpc), scratch)
+ .addReg(srcOut, RegState::Define).addReg(srcIn).addImm(0));
+
+ AddDefaultPred(BuildMI(*BB, MI, dl, TII->get(strOpc), destOut)
+ .addReg(destIn).addImm(0).addReg(scratch));
+ } else if (isThumb2) {
+ AddDefaultPred(BuildMI(*BB, MI, dl,
+ TII->get(ldrOpc), scratch)
+ .addReg(srcOut, RegState::Define).addReg(srcIn).addImm(UnitSize));
+
+ AddDefaultPred(BuildMI(*BB, MI, dl, TII->get(strOpc), destOut)
+ .addReg(scratch).addReg(destIn)
+ .addImm(UnitSize));
+ } else {
+ AddDefaultPred(BuildMI(*BB, MI, dl,
+ TII->get(ldrOpc), scratch)
+ .addReg(srcOut, RegState::Define).addReg(srcIn).addReg(0)
+ .addImm(UnitSize));
+
+ AddDefaultPred(BuildMI(*BB, MI, dl, TII->get(strOpc), destOut)
+ .addReg(scratch).addReg(destIn)
+ .addReg(0).addImm(UnitSize));
+ }
+ srcIn = srcOut;
+ destIn = destOut;
+ }
+
+ // Handle the leftover bytes with LDRB and STRB.
+ // [scratch, srcOut] = LDRB_POST(srcIn, 1)
+ // [destOut] = STRB_POST(scratch, destIn, 1)
+ ldrOpc = isThumb2 ? ARM::t2LDRB_POST : ARM::LDRB_POST_IMM;
+ strOpc = isThumb2 ? ARM::t2STRB_POST : ARM::STRB_POST_IMM;
+ for (unsigned i = 0; i < BytesLeft; i++) {
+ unsigned scratch = MRI.createVirtualRegister(TRC);
+ unsigned srcOut = MRI.createVirtualRegister(TRC);
+ unsigned destOut = MRI.createVirtualRegister(TRC);
+ if (isThumb2) {
+ AddDefaultPred(BuildMI(*BB, MI, dl,
+ TII->get(ldrOpc),scratch)
+ .addReg(srcOut, RegState::Define).addReg(srcIn).addImm(1));
+
+ AddDefaultPred(BuildMI(*BB, MI, dl, TII->get(strOpc), destOut)
+ .addReg(scratch).addReg(destIn)
+ .addReg(0).addImm(1));
+ } else {
+ AddDefaultPred(BuildMI(*BB, MI, dl,
+ TII->get(ldrOpc),scratch)
+ .addReg(srcOut, RegState::Define).addReg(srcIn).addImm(1));
+
+ AddDefaultPred(BuildMI(*BB, MI, dl, TII->get(strOpc), destOut)
+ .addReg(scratch).addReg(destIn)
+ .addReg(0).addImm(1));
+ }
+ srcIn = srcOut;
+ destIn = destOut;
+ }
+ MI->eraseFromParent(); // The instruction is gone now.
+ return BB;
+ }
+
+ // Expand the pseudo op to a loop.
+ // thisMBB:
+ // ...
+ // movw varEnd, # --> with thumb2
+ // movt varEnd, #
+ // ldrcp varEnd, idx --> without thumb2
+ // fallthrough --> loopMBB
+ // loopMBB:
+ // PHI varPhi, varEnd, varLoop
+ // PHI srcPhi, src, srcLoop
+ // PHI destPhi, dst, destLoop
+ // [scratch, srcLoop] = LDR_POST(srcPhi, UnitSize)
+ // [destLoop] = STR_POST(scratch, destPhi, UnitSize)
+ // subs varLoop, varPhi, #UnitSize
+ // bne loopMBB
+ // fallthrough --> exitMBB
+ // exitMBB:
+ // epilogue to handle left-over bytes
+ // [scratch, srcOut] = LDRB_POST(srcLoop, 1)
+ // [destOut] = STRB_POST(scratch, destLoop, 1)
+ MachineBasicBlock *loopMBB = MF->CreateMachineBasicBlock(LLVM_BB);
+ MachineBasicBlock *exitMBB = MF->CreateMachineBasicBlock(LLVM_BB);
+ MF->insert(It, loopMBB);
+ MF->insert(It, exitMBB);
+
+ // Transfer the remainder of BB and its successor edges to exitMBB.
+ exitMBB->splice(exitMBB->begin(), BB,
+ llvm::next(MachineBasicBlock::iterator(MI)),
+ BB->end());
+ exitMBB->transferSuccessorsAndUpdatePHIs(BB);
+
+ // Load an immediate to varEnd.
+ unsigned varEnd = MRI.createVirtualRegister(TRC);
+ if (isThumb2) {
+ unsigned VReg1 = varEnd;
+ if ((LoopSize & 0xFFFF0000) != 0)
+ VReg1 = MRI.createVirtualRegister(TRC);
+ AddDefaultPred(BuildMI(BB, dl, TII->get(ARM::t2MOVi16), VReg1)
+ .addImm(LoopSize & 0xFFFF));
+
+ if ((LoopSize & 0xFFFF0000) != 0)
+ AddDefaultPred(BuildMI(BB, dl, TII->get(ARM::t2MOVTi16), varEnd)
+ .addReg(VReg1)
+ .addImm(LoopSize >> 16));
+ } else {
+ MachineConstantPool *ConstantPool = MF->getConstantPool();
+ Type *Int32Ty = Type::getInt32Ty(MF->getFunction()->getContext());
+ const Constant *C = ConstantInt::get(Int32Ty, LoopSize);
+
+ // MachineConstantPool wants an explicit alignment.
+ unsigned Align = getTargetData()->getPrefTypeAlignment(Int32Ty);
+ if (Align == 0)
+ Align = getTargetData()->getTypeAllocSize(C->getType());
+ unsigned Idx = ConstantPool->getConstantPoolIndex(C, Align);
+
+ AddDefaultPred(BuildMI(BB, dl, TII->get(ARM::LDRcp))
+ .addReg(varEnd, RegState::Define)
+ .addConstantPoolIndex(Idx)
+ .addImm(0));
+ }
+ BB->addSuccessor(loopMBB);
+
+ // Generate the loop body:
+ // varPhi = PHI(varLoop, varEnd)
+ // srcPhi = PHI(srcLoop, src)
+ // destPhi = PHI(destLoop, dst)
+ MachineBasicBlock *entryBB = BB;
+ BB = loopMBB;
+ unsigned varLoop = MRI.createVirtualRegister(TRC);
+ unsigned varPhi = MRI.createVirtualRegister(TRC);
+ unsigned srcLoop = MRI.createVirtualRegister(TRC);
+ unsigned srcPhi = MRI.createVirtualRegister(TRC);
+ unsigned destLoop = MRI.createVirtualRegister(TRC);
+ unsigned destPhi = MRI.createVirtualRegister(TRC);
+
+ BuildMI(*BB, BB->begin(), dl, TII->get(ARM::PHI), varPhi)
+ .addReg(varLoop).addMBB(loopMBB)
+ .addReg(varEnd).addMBB(entryBB);
+ BuildMI(BB, dl, TII->get(ARM::PHI), srcPhi)
+ .addReg(srcLoop).addMBB(loopMBB)
+ .addReg(src).addMBB(entryBB);
+ BuildMI(BB, dl, TII->get(ARM::PHI), destPhi)
+ .addReg(destLoop).addMBB(loopMBB)
+ .addReg(dest).addMBB(entryBB);
+
+ // [scratch, srcLoop] = LDR_POST(srcPhi, UnitSize)
+ // [destLoop] = STR_POST(scratch, destPhi, UnitSiz)
+ unsigned scratch = MRI.createVirtualRegister(UnitSize >= 8 ? TRC_Vec:TRC);
+ if (UnitSize >= 8) {
+ AddDefaultPred(BuildMI(BB, dl, TII->get(ldrOpc), scratch)
+ .addReg(srcLoop, RegState::Define).addReg(srcPhi).addImm(0));
+
+ AddDefaultPred(BuildMI(BB, dl, TII->get(strOpc), destLoop)
+ .addReg(destPhi).addImm(0).addReg(scratch));
+ } else if (isThumb2) {
+ AddDefaultPred(BuildMI(BB, dl, TII->get(ldrOpc), scratch)
+ .addReg(srcLoop, RegState::Define).addReg(srcPhi).addImm(UnitSize));
+
+ AddDefaultPred(BuildMI(BB, dl, TII->get(strOpc), destLoop)
+ .addReg(scratch).addReg(destPhi)
+ .addImm(UnitSize));
+ } else {
+ AddDefaultPred(BuildMI(BB, dl, TII->get(ldrOpc), scratch)
+ .addReg(srcLoop, RegState::Define).addReg(srcPhi).addReg(0)
+ .addImm(UnitSize));
+
+ AddDefaultPred(BuildMI(BB, dl, TII->get(strOpc), destLoop)
+ .addReg(scratch).addReg(destPhi)
+ .addReg(0).addImm(UnitSize));
+ }
+
+ // Decrement loop variable by UnitSize.
+ MachineInstrBuilder MIB = BuildMI(BB, dl,
+ TII->get(isThumb2 ? ARM::t2SUBri : ARM::SUBri), varLoop);
+ AddDefaultCC(AddDefaultPred(MIB.addReg(varPhi).addImm(UnitSize)));
+ MIB->getOperand(5).setReg(ARM::CPSR);
+ MIB->getOperand(5).setIsDef(true);
+
+ BuildMI(BB, dl, TII->get(isThumb2 ? ARM::t2Bcc : ARM::Bcc))
+ .addMBB(loopMBB).addImm(ARMCC::NE).addReg(ARM::CPSR);
+
+ // loopMBB can loop back to loopMBB or fall through to exitMBB.
+ BB->addSuccessor(loopMBB);
+ BB->addSuccessor(exitMBB);
+
+ // Add epilogue to handle BytesLeft.
+ BB = exitMBB;
+ MachineInstr *StartOfExit = exitMBB->begin();
+ ldrOpc = isThumb2 ? ARM::t2LDRB_POST : ARM::LDRB_POST_IMM;
+ strOpc = isThumb2 ? ARM::t2STRB_POST : ARM::STRB_POST_IMM;
+
+ // [scratch, srcOut] = LDRB_POST(srcLoop, 1)
+ // [destOut] = STRB_POST(scratch, destLoop, 1)
+ unsigned srcIn = srcLoop;
+ unsigned destIn = destLoop;
+ for (unsigned i = 0; i < BytesLeft; i++) {
+ unsigned scratch = MRI.createVirtualRegister(TRC);
+ unsigned srcOut = MRI.createVirtualRegister(TRC);
+ unsigned destOut = MRI.createVirtualRegister(TRC);
+ if (isThumb2) {
+ AddDefaultPred(BuildMI(*BB, StartOfExit, dl,
+ TII->get(ldrOpc),scratch)
+ .addReg(srcOut, RegState::Define).addReg(srcIn).addImm(1));
+
+ AddDefaultPred(BuildMI(*BB, StartOfExit, dl, TII->get(strOpc), destOut)
+ .addReg(scratch).addReg(destIn)
+ .addImm(1));
+ } else {
+ AddDefaultPred(BuildMI(*BB, StartOfExit, dl,
+ TII->get(ldrOpc),scratch)
+ .addReg(srcOut, RegState::Define).addReg(srcIn).addReg(0).addImm(1));
+
+ AddDefaultPred(BuildMI(*BB, StartOfExit, dl, TII->get(strOpc), destOut)
+ .addReg(scratch).addReg(destIn)
+ .addReg(0).addImm(1));
+ }
+ srcIn = srcOut;
+ destIn = destOut;
+ }
+
+ MI->eraseFromParent(); // The instruction is gone now.
+ return BB;
+}
+
MachineBasicBlock *
ARMTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI,
MachineBasicBlock *BB) const {
MachineRegisterInfo &MRI = Fn->getRegInfo();
// In Thumb mode S must not be specified if source register is the SP or
// PC and if destination register is the SP, so restrict register class
- unsigned NewMovDstReg = MRI.createVirtualRegister(isThumb2 ?
- (const TargetRegisterClass*)&ARM::rGPRRegClass :
- (const TargetRegisterClass*)&ARM::GPRRegClass);
unsigned NewRsbDstReg = MRI.createVirtualRegister(isThumb2 ?
(const TargetRegisterClass*)&ARM::rGPRRegClass :
(const TargetRegisterClass*)&ARM::GPRRegClass);
// fall through to SinkMBB
RSBBB->addSuccessor(SinkBB);
- // insert a movs at the end of BB
- BuildMI(BB, dl, TII->get(isThumb2 ? ARM::t2MOVr : ARM::MOVr),
- NewMovDstReg)
- .addReg(ABSSrcReg, RegState::Kill)
- .addImm((unsigned)ARMCC::AL).addReg(0)
- .addReg(ARM::CPSR, RegState::Define);
+ // insert a cmp at the end of BB
+ AddDefaultPred(BuildMI(BB, dl,
+ TII->get(isThumb2 ? ARM::t2CMPri : ARM::CMPri))
+ .addReg(ABSSrcReg).addImm(0));
// insert a bcc with opposite CC to ARMCC::MI at the end of BB
BuildMI(BB, dl,
// by if-conversion pass
BuildMI(*RSBBB, RSBBB->begin(), dl,
TII->get(isThumb2 ? ARM::t2RSBri : ARM::RSBri), NewRsbDstReg)
- .addReg(NewMovDstReg, RegState::Kill)
+ .addReg(ABSSrcReg, RegState::Kill)
.addImm(0).addImm((unsigned)ARMCC::AL).addReg(0).addReg(0);
// insert PHI in SinkBB,
BuildMI(*SinkBB, SinkBB->begin(), dl,
TII->get(ARM::PHI), ABSDstReg)
.addReg(NewRsbDstReg).addMBB(RSBBB)
- .addReg(NewMovDstReg).addMBB(BB);
+ .addReg(ABSSrcReg).addMBB(BB);
// remove ABS instruction
MI->eraseFromParent();
// return last added BB
return SinkBB;
}
+ case ARM::COPY_STRUCT_BYVAL_I32:
+ ++NumLoopByVals;
+ return EmitStructByval(MI, BB);
}
}
// ARM Optimization Hooks
//===----------------------------------------------------------------------===//
+// Helper function that checks if N is a null or all ones constant.
+static inline bool isZeroOrAllOnes(SDValue N, bool AllOnes) {
+ ConstantSDNode *C = dyn_cast<ConstantSDNode>(N);
+ if (!C)
+ return false;
+ return AllOnes ? C->isAllOnesValue() : C->isNullValue();
+}
+
+// Return true if N is conditionally 0 or all ones.
+// Detects these expressions where cc is an i1 value:
+//
+// (select cc 0, y) [AllOnes=0]
+// (select cc y, 0) [AllOnes=0]
+// (zext cc) [AllOnes=0]
+// (sext cc) [AllOnes=0/1]
+// (select cc -1, y) [AllOnes=1]
+// (select cc y, -1) [AllOnes=1]
+//
+// Invert is set when N is the null/all ones constant when CC is false.
+// OtherOp is set to the alternative value of N.
+static bool isConditionalZeroOrAllOnes(SDNode *N, bool AllOnes,
+ SDValue &CC, bool &Invert,
+ SDValue &OtherOp,
+ SelectionDAG &DAG) {
+ switch (N->getOpcode()) {
+ default: return false;
+ case ISD::SELECT: {
+ CC = N->getOperand(0);
+ SDValue N1 = N->getOperand(1);
+ SDValue N2 = N->getOperand(2);
+ if (isZeroOrAllOnes(N1, AllOnes)) {
+ Invert = false;
+ OtherOp = N2;
+ return true;
+ }
+ if (isZeroOrAllOnes(N2, AllOnes)) {
+ Invert = true;
+ OtherOp = N1;
+ return true;
+ }
+ return false;
+ }
+ case ISD::ZERO_EXTEND:
+ // (zext cc) can never be the all ones value.
+ if (AllOnes)
+ return false;
+ // Fall through.
+ case ISD::SIGN_EXTEND: {
+ EVT VT = N->getValueType(0);
+ CC = N->getOperand(0);
+ if (CC.getValueType() != MVT::i1)
+ return false;
+ Invert = !AllOnes;
+ if (AllOnes)
+ // When looking for an AllOnes constant, N is an sext, and the 'other'
+ // value is 0.
+ OtherOp = DAG.getConstant(0, VT);
+ else if (N->getOpcode() == ISD::ZERO_EXTEND)
+ // When looking for a 0 constant, N can be zext or sext.
+ OtherOp = DAG.getConstant(1, VT);
+ else
+ OtherOp = DAG.getConstant(APInt::getAllOnesValue(VT.getSizeInBits()), VT);
+ return true;
+ }
+ }
+}
+
+// Combine a constant select operand into its use:
+//
+// (add (select cc, 0, c), x) -> (select cc, x, (add, x, c))
+// (sub x, (select cc, 0, c)) -> (select cc, x, (sub, x, c))
+// (and (select cc, -1, c), x) -> (select cc, x, (and, x, c)) [AllOnes=1]
+// (or (select cc, 0, c), x) -> (select cc, x, (or, x, c))
+// (xor (select cc, 0, c), x) -> (select cc, x, (xor, x, c))
+//
+// The transform is rejected if the select doesn't have a constant operand that
+// is null, or all ones when AllOnes is set.
+//
+// Also recognize sext/zext from i1:
+//
+// (add (zext cc), x) -> (select cc (add x, 1), x)
+// (add (sext cc), x) -> (select cc (add x, -1), x)
+//
+// These transformations eventually create predicated instructions.
+//
+// @param N The node to transform.
+// @param Slct The N operand that is a select.
+// @param OtherOp The other N operand (x above).
+// @param DCI Context.
+// @param AllOnes Require the select constant to be all ones instead of null.
+// @returns The new node, or SDValue() on failure.
static
SDValue combineSelectAndUse(SDNode *N, SDValue Slct, SDValue OtherOp,
- TargetLowering::DAGCombinerInfo &DCI) {
+ TargetLowering::DAGCombinerInfo &DCI,
+ bool AllOnes = false) {
SelectionDAG &DAG = DCI.DAG;
- const TargetLowering &TLI = DAG.getTargetLoweringInfo();
EVT VT = N->getValueType(0);
- unsigned Opc = N->getOpcode();
- bool isSlctCC = Slct.getOpcode() == ISD::SELECT_CC;
- SDValue LHS = isSlctCC ? Slct.getOperand(2) : Slct.getOperand(1);
- SDValue RHS = isSlctCC ? Slct.getOperand(3) : Slct.getOperand(2);
- ISD::CondCode CC = ISD::SETCC_INVALID;
-
- if (isSlctCC) {
- CC = cast<CondCodeSDNode>(Slct.getOperand(4))->get();
- } else {
- SDValue CCOp = Slct.getOperand(0);
- if (CCOp.getOpcode() == ISD::SETCC)
- CC = cast<CondCodeSDNode>(CCOp.getOperand(2))->get();
- }
-
- bool DoXform = false;
- bool InvCC = false;
- assert ((Opc == ISD::ADD || (Opc == ISD::SUB && Slct == N->getOperand(1))) &&
- "Bad input!");
-
- if (LHS.getOpcode() == ISD::Constant &&
- cast<ConstantSDNode>(LHS)->isNullValue()) {
- DoXform = true;
- } else if (CC != ISD::SETCC_INVALID &&
- RHS.getOpcode() == ISD::Constant &&
- cast<ConstantSDNode>(RHS)->isNullValue()) {
- std::swap(LHS, RHS);
- SDValue Op0 = Slct.getOperand(0);
- EVT OpVT = isSlctCC ? Op0.getValueType() :
- Op0.getOperand(0).getValueType();
- bool isInt = OpVT.isInteger();
- CC = ISD::getSetCCInverse(CC, isInt);
-
- if (!TLI.isCondCodeLegal(CC, OpVT))
- return SDValue(); // Inverse operator isn't legal.
-
- DoXform = true;
- InvCC = true;
- }
-
- if (DoXform) {
- SDValue Result = DAG.getNode(Opc, RHS.getDebugLoc(), VT, OtherOp, RHS);
- if (isSlctCC)
- return DAG.getSelectCC(N->getDebugLoc(), OtherOp, Result,
- Slct.getOperand(0), Slct.getOperand(1), CC);
- SDValue CCOp = Slct.getOperand(0);
- if (InvCC)
- CCOp = DAG.getSetCC(Slct.getDebugLoc(), CCOp.getValueType(),
- CCOp.getOperand(0), CCOp.getOperand(1), CC);
- return DAG.getNode(ISD::SELECT, N->getDebugLoc(), VT,
- CCOp, OtherOp, Result);
+ SDValue NonConstantVal;
+ SDValue CCOp;
+ bool SwapSelectOps;
+ if (!isConditionalZeroOrAllOnes(Slct.getNode(), AllOnes, CCOp, SwapSelectOps,
+ NonConstantVal, DAG))
+ return SDValue();
+
+ // Slct is now know to be the desired identity constant when CC is true.
+ SDValue TrueVal = OtherOp;
+ SDValue FalseVal = DAG.getNode(N->getOpcode(), N->getDebugLoc(), VT,
+ OtherOp, NonConstantVal);
+ // Unless SwapSelectOps says CC should be false.
+ if (SwapSelectOps)
+ std::swap(TrueVal, FalseVal);
+
+ return DAG.getNode(ISD::SELECT, N->getDebugLoc(), VT,
+ CCOp, TrueVal, FalseVal);
+}
+
+// Attempt combineSelectAndUse on each operand of a commutative operator N.
+static
+SDValue combineSelectAndUseCommutative(SDNode *N, bool AllOnes,
+ TargetLowering::DAGCombinerInfo &DCI) {
+ SDValue N0 = N->getOperand(0);
+ SDValue N1 = N->getOperand(1);
+ if (N0.getNode()->hasOneUse()) {
+ SDValue Result = combineSelectAndUse(N, N0, N1, DCI, AllOnes);
+ if (Result.getNode())
+ return Result;
+ }
+ if (N1.getNode()->hasOneUse()) {
+ SDValue Result = combineSelectAndUse(N, N1, N0, DCI, AllOnes);
+ if (Result.getNode())
+ return Result;
}
return SDValue();
}
return DAG.getNode(ISD::TRUNCATE, N->getDebugLoc(), VT, tmp);
}
+static SDValue findMUL_LOHI(SDValue V) {
+ if (V->getOpcode() == ISD::UMUL_LOHI ||
+ V->getOpcode() == ISD::SMUL_LOHI)
+ return V;
+ return SDValue();
+}
+
+static SDValue AddCombineTo64bitMLAL(SDNode *AddcNode,
+ TargetLowering::DAGCombinerInfo &DCI,
+ const ARMSubtarget *Subtarget) {
+
+ if (Subtarget->isThumb1Only()) return SDValue();
+
+ // Only perform the checks after legalize when the pattern is available.
+ if (DCI.isBeforeLegalize()) return SDValue();
+
+ // Look for multiply add opportunities.
+ // The pattern is a ISD::UMUL_LOHI followed by two add nodes, where
+ // each add nodes consumes a value from ISD::UMUL_LOHI and there is
+ // a glue link from the first add to the second add.
+ // If we find this pattern, we can replace the U/SMUL_LOHI, ADDC, and ADDE by
+ // a S/UMLAL instruction.
+ // loAdd UMUL_LOHI
+ // \ / :lo \ :hi
+ // \ / \ [no multiline comment]
+ // ADDC | hiAdd
+ // \ :glue / /
+ // \ / /
+ // ADDE
+ //
+ assert(AddcNode->getOpcode() == ISD::ADDC && "Expect an ADDC");
+ SDValue AddcOp0 = AddcNode->getOperand(0);
+ SDValue AddcOp1 = AddcNode->getOperand(1);
+
+ // Check if the two operands are from the same mul_lohi node.
+ if (AddcOp0.getNode() == AddcOp1.getNode())
+ return SDValue();
+
+ assert(AddcNode->getNumValues() == 2 &&
+ AddcNode->getValueType(0) == MVT::i32 &&
+ AddcNode->getValueType(1) == MVT::Glue &&
+ "Expect ADDC with two result values: i32, glue");
+
+ // Check that the ADDC adds the low result of the S/UMUL_LOHI.
+ if (AddcOp0->getOpcode() != ISD::UMUL_LOHI &&
+ AddcOp0->getOpcode() != ISD::SMUL_LOHI &&
+ AddcOp1->getOpcode() != ISD::UMUL_LOHI &&
+ AddcOp1->getOpcode() != ISD::SMUL_LOHI)
+ return SDValue();
+
+ // Look for the glued ADDE.
+ SDNode* AddeNode = AddcNode->getGluedUser();
+ if (AddeNode == NULL)
+ return SDValue();
+
+ // Make sure it is really an ADDE.
+ if (AddeNode->getOpcode() != ISD::ADDE)
+ return SDValue();
+
+ assert(AddeNode->getNumOperands() == 3 &&
+ AddeNode->getOperand(2).getValueType() == MVT::Glue &&
+ "ADDE node has the wrong inputs");
+
+ // Check for the triangle shape.
+ SDValue AddeOp0 = AddeNode->getOperand(0);
+ SDValue AddeOp1 = AddeNode->getOperand(1);
+
+ // Make sure that the ADDE operands are not coming from the same node.
+ if (AddeOp0.getNode() == AddeOp1.getNode())
+ return SDValue();
+
+ // Find the MUL_LOHI node walking up ADDE's operands.
+ bool IsLeftOperandMUL = false;
+ SDValue MULOp = findMUL_LOHI(AddeOp0);
+ if (MULOp == SDValue())
+ MULOp = findMUL_LOHI(AddeOp1);
+ else
+ IsLeftOperandMUL = true;
+ if (MULOp == SDValue())
+ return SDValue();
+
+ // Figure out the right opcode.
+ unsigned Opc = MULOp->getOpcode();
+ unsigned FinalOpc = (Opc == ISD::SMUL_LOHI) ? ARMISD::SMLAL : ARMISD::UMLAL;
+
+ // Figure out the high and low input values to the MLAL node.
+ SDValue* HiMul = &MULOp;
+ SDValue* HiAdd = NULL;
+ SDValue* LoMul = NULL;
+ SDValue* LowAdd = NULL;
+
+ if (IsLeftOperandMUL)
+ HiAdd = &AddeOp1;
+ else
+ HiAdd = &AddeOp0;
+
+
+ if (AddcOp0->getOpcode() == Opc) {
+ LoMul = &AddcOp0;
+ LowAdd = &AddcOp1;
+ }
+ if (AddcOp1->getOpcode() == Opc) {
+ LoMul = &AddcOp1;
+ LowAdd = &AddcOp0;
+ }
+
+ if (LoMul == NULL)
+ return SDValue();
+
+ if (LoMul->getNode() != HiMul->getNode())
+ return SDValue();
+
+ // Create the merged node.
+ SelectionDAG &DAG = DCI.DAG;
+
+ // Build operand list.
+ SmallVector<SDValue, 8> Ops;
+ Ops.push_back(LoMul->getOperand(0));
+ Ops.push_back(LoMul->getOperand(1));
+ Ops.push_back(*LowAdd);
+ Ops.push_back(*HiAdd);
+
+ SDValue MLALNode = DAG.getNode(FinalOpc, AddcNode->getDebugLoc(),
+ DAG.getVTList(MVT::i32, MVT::i32),
+ &Ops[0], Ops.size());
+
+ // Replace the ADDs' nodes uses by the MLA node's values.
+ SDValue HiMLALResult(MLALNode.getNode(), 1);
+ DAG.ReplaceAllUsesOfValueWith(SDValue(AddeNode, 0), HiMLALResult);
+
+ SDValue LoMLALResult(MLALNode.getNode(), 0);
+ DAG.ReplaceAllUsesOfValueWith(SDValue(AddcNode, 0), LoMLALResult);
+
+ // Return original node to notify the driver to stop replacing.
+ SDValue resNode(AddcNode, 0);
+ return resNode;
+}
+
+/// PerformADDCCombine - Target-specific dag combine transform from
+/// ISD::ADDC, ISD::ADDE, and ISD::MUL_LOHI to MLAL.
+static SDValue PerformADDCCombine(SDNode *N,
+ TargetLowering::DAGCombinerInfo &DCI,
+ const ARMSubtarget *Subtarget) {
+
+ return AddCombineTo64bitMLAL(N, DCI, Subtarget);
+
+}
+
/// PerformADDCombineWithOperands - Try DAG combinations for an ADD with
/// operands N0 and N1. This is a helper for PerformADDCombine that is
/// called with the default operands, and if that fails, with commuted
return Result;
// fold (add (select cc, 0, c), x) -> (select cc, x, (add, x, c))
- if (N0.getOpcode() == ISD::SELECT && N0.getNode()->hasOneUse()) {
+ if (N0.getNode()->hasOneUse()) {
SDValue Result = combineSelectAndUse(N, N0, N1, DCI);
if (Result.getNode()) return Result;
}
SDValue N1 = N->getOperand(1);
// fold (sub x, (select cc, 0, c)) -> (select cc, x, (sub, x, c))
- if (N1.getOpcode() == ISD::SELECT && N1.getNode()->hasOneUse()) {
+ if (N1.getNode()->hasOneUse()) {
SDValue Result = combineSelectAndUse(N, N1, N0, DCI);
if (Result.getNode()) return Result;
}
return SDValue();
}
-static bool isCMOVWithZeroOrAllOnesLHS(SDValue N, bool AllOnes) {
- if (N.getOpcode() != ARMISD::CMOV || !N.getNode()->hasOneUse())
- return false;
-
- SDValue FalseVal = N.getOperand(0);
- ConstantSDNode *C = dyn_cast<ConstantSDNode>(FalseVal);
- if (!C)
- return false;
- if (AllOnes)
- return C->isAllOnesValue();
- return C->isNullValue();
-}
-
-/// formConditionalOp - Combine an operation with a conditional move operand
-/// to form a conditional op. e.g. (or x, (cmov 0, y, cond)) => (or.cond x, y)
-/// (and x, (cmov -1, y, cond)) => (and.cond, x, y)
-static SDValue formConditionalOp(SDNode *N, SelectionDAG &DAG,
- bool Commutable) {
- SDValue N0 = N->getOperand(0);
- SDValue N1 = N->getOperand(1);
-
- bool isAND = N->getOpcode() == ISD::AND;
- bool isCand = isCMOVWithZeroOrAllOnesLHS(N1, isAND);
- if (!isCand && Commutable) {
- isCand = isCMOVWithZeroOrAllOnesLHS(N0, isAND);
- if (isCand)
- std::swap(N0, N1);
- }
- if (!isCand)
- return SDValue();
-
- unsigned Opc = 0;
- switch (N->getOpcode()) {
- default: llvm_unreachable("Unexpected node");
- case ISD::AND: Opc = ARMISD::CAND; break;
- case ISD::OR: Opc = ARMISD::COR; break;
- case ISD::XOR: Opc = ARMISD::CXOR; break;
- }
- return DAG.getNode(Opc, N->getDebugLoc(), N->getValueType(0), N0,
- N1.getOperand(1), N1.getOperand(2), N1.getOperand(3),
- N1.getOperand(4));
-}
-
static SDValue PerformANDCombine(SDNode *N,
TargetLowering::DAGCombinerInfo &DCI,
const ARMSubtarget *Subtarget) {
}
if (!Subtarget->isThumb1Only()) {
- // (and x, (cmov -1, y, cond)) => (and.cond x, y)
- SDValue CAND = formConditionalOp(N, DAG, true);
- if (CAND.getNode())
- return CAND;
+ // fold (and (select cc, -1, c), x) -> (select cc, x, (and, x, c))
+ SDValue Result = combineSelectAndUseCommutative(N, true, DCI);
+ if (Result.getNode())
+ return Result;
}
return SDValue();
}
if (!Subtarget->isThumb1Only()) {
- // (or x, (cmov 0, y, cond)) => (or.cond x, y)
- SDValue COR = formConditionalOp(N, DAG, true);
- if (COR.getNode())
- return COR;
+ // fold (or (select cc, 0, c), x) -> (select cc, x, (or, x, c))
+ SDValue Result = combineSelectAndUseCommutative(N, false, DCI);
+ if (Result.getNode())
+ return Result;
}
+ // The code below optimizes (or (and X, Y), Z).
+ // The AND operand needs to have a single user to make these optimizations
+ // profitable.
SDValue N0 = N->getOperand(0);
- if (N0.getOpcode() != ISD::AND)
+ if (N0.getOpcode() != ISD::AND || !N0.hasOneUse())
return SDValue();
SDValue N1 = N->getOperand(1);
return SDValue();
if (!Subtarget->isThumb1Only()) {
- // (xor x, (cmov 0, y, cond)) => (xor.cond x, y)
- SDValue CXOR = formConditionalOp(N, DAG, true);
- if (CXOR.getNode())
- return CXOR;
+ // fold (xor (select cc, 0, c), x) -> (select cc, x, (xor, x, c))
+ SDValue Result = combineSelectAndUseCommutative(N, false, DCI);
+ if (Result.getNode())
+ return Result;
}
return SDValue();
if (St->isVolatile())
return SDValue();
- // Optimize trunc store (of multiple scalars) to shuffle and store. First,
+ // Optimize trunc store (of multiple scalars) to shuffle and store. First,
// pack all of the elements in one place. Next, store to memory in fewer
// chunks.
SDValue StVal = St->getValue();
DAGCombinerInfo &DCI) const {
switch (N->getOpcode()) {
default: break;
+ case ISD::ADDC: return PerformADDCCombine(N, DCI, Subtarget);
case ISD::ADD: return PerformADDCombine(N, DCI, Subtarget);
case ISD::SUB: return PerformSUBCombine(N, DCI);
case ISD::MUL: return PerformMULCombine(N, DCI, Subtarget);
case MVT::i16:
case MVT::i32:
return true;
+ case MVT::f64:
+ return Subtarget->hasNEON();
// FIXME: VLD1 etc with standard alignment is legal.
}
}
return Imm >= 0 && Imm <= 255;
}
-/// isLegalAddImmediate - Return true if the specified immediate is legal
-/// add immediate, that is the target has add instructions which can add
-/// a register with the immediate without having to materialize the
+/// isLegalAddImmediate - Return true if the specified immediate is a legal add
+/// *or sub* immediate, that is the target has add or sub instructions which can
+/// add a register with the immediate without having to materialize the
/// immediate into a register.
bool ARMTargetLowering::isLegalAddImmediate(int64_t Imm) const {
- return ARM_AM::getSOImmVal(Imm) != -1;
+ // Same encoding for add/sub, just flip the sign.
+ int64_t AbsImm = llvm::abs64(Imm);
+ if (!Subtarget->isThumb())
+ return ARM_AM::getSOImmVal(AbsImm) != -1;
+ if (Subtarget->isThumb2())
+ return ARM_AM::getT2SOImmVal(AbsImm) != -1;
+ // Thumb1 only has 8-bit unsigned immediate.
+ return AbsImm >= 0 && AbsImm <= 255;
}
static bool getARMIndexedAddressParts(SDNode *Ptr, EVT VT,