const AArch64Subtarget *Subtarget;
LLVMContext *Context;
+ bool FastLowerArguments() override;
bool FastLowerCall(CallLoweringInfo &CLI) override;
bool FastLowerIntrinsicCall(const IntrinsicInst *II) override;
bool SimplifyAddress(Address &Addr, MVT VT, int64_t ScaleFactor,
bool UseUnscaled);
void AddLoadStoreOperands(Address &Addr, const MachineInstrBuilder &MIB,
- unsigned Flags, bool UseUnscaled);
+ unsigned Flags, MachineMemOperand *MMO,
+ bool UseUnscaled);
bool IsMemCpySmall(uint64_t Len, unsigned Alignment);
bool TryEmitSmallMemCpy(Address Dest, Address Src, uint64_t Len,
unsigned Alignment);
// Emit functions.
bool EmitCmp(Value *Src1Value, Value *Src2Value, bool isZExt);
bool EmitLoad(MVT VT, unsigned &ResultReg, Address Addr,
- bool UseUnscaled = false);
+ MachineMemOperand *MMO = nullptr, bool UseUnscaled = false);
bool EmitStore(MVT VT, unsigned SrcReg, Address Addr,
- bool UseUnscaled = false);
+ MachineMemOperand *MMO = nullptr, bool UseUnscaled = false);
unsigned EmitIntExt(MVT SrcVT, unsigned SrcReg, MVT DestVT, bool isZExt);
unsigned Emiti1Ext(unsigned SrcReg, MVT DestVT, bool isZExt);
unsigned Emit_MUL_rr(MVT RetVT, unsigned Op0, bool Op0IsKill,
unsigned Emit_LSR_ri(MVT RetVT, unsigned Op0, bool Op0IsKill, uint64_t Imm);
unsigned Emit_ASR_ri(MVT RetVT, unsigned Op0, bool Op0IsKill, uint64_t Imm);
+ unsigned AArch64MaterializeInt(const ConstantInt *CI, MVT VT);
unsigned AArch64MaterializeFP(const ConstantFP *CFP, MVT VT);
unsigned AArch64MaterializeGV(const GlobalValue *GV);
return 0;
}
+unsigned AArch64FastISel::AArch64MaterializeInt(const ConstantInt *CI, MVT VT) {
+ if (VT > MVT::i64)
+ return 0;
+ return FastEmit_i(VT, VT, ISD::Constant, CI->getZExtValue());
+}
+
unsigned AArch64FastISel::AArch64MaterializeFP(const ConstantFP *CFP, MVT VT) {
if (VT != MVT::f32 && VT != MVT::f64)
return 0;
const APFloat Val = CFP->getValueAPF();
- bool is64bit = (VT == MVT::f64);
+ bool Is64Bit = (VT == MVT::f64);
// This checks to see if we can use FMOV instructions to materialize
// a constant, otherwise we have to materialize via the constant pool.
if (TLI.isFPImmLegal(Val, VT)) {
- int Imm;
- unsigned Opc;
- if (is64bit) {
- Imm = AArch64_AM::getFP64Imm(Val);
- Opc = AArch64::FMOVDi;
- } else {
- Imm = AArch64_AM::getFP32Imm(Val);
- Opc = AArch64::FMOVSi;
- }
unsigned ResultReg = createResultReg(TLI.getRegClassFor(VT));
+ // Positive zero (+0.0) has to be materialized with a fmov from the zero
+ // register, because the immediate version of fmov cannot encode zero.
+ if (Val.isPosZero()) {
+ unsigned ZReg = Is64Bit ? AArch64::XZR : AArch64::WZR;
+ unsigned Opc = Is64Bit ? AArch64::FMOVDr : AArch64::FMOVSr;
+ BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(Opc), ResultReg)
+ .addReg(ZReg, getKillRegState(true));
+ return ResultReg;
+ }
+ int Imm = Is64Bit ? AArch64_AM::getFP64Imm(Val)
+ : AArch64_AM::getFP32Imm(Val);
+ unsigned Opc = Is64Bit ? AArch64::FMOVDi : AArch64::FMOVSi;
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(Opc), ResultReg)
- .addImm(Imm);
+ .addImm(Imm);
return ResultReg;
}
if (Align == 0)
Align = DL.getTypeAllocSize(CFP->getType());
- unsigned Idx = MCP.getConstantPoolIndex(cast<Constant>(CFP), Align);
+ unsigned CPI = MCP.getConstantPoolIndex(cast<Constant>(CFP), Align);
unsigned ADRPReg = createResultReg(&AArch64::GPR64commonRegClass);
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::ADRP),
- ADRPReg).addConstantPoolIndex(Idx, 0, AArch64II::MO_PAGE);
+ ADRPReg)
+ .addConstantPoolIndex(CPI, 0, AArch64II::MO_PAGE);
- unsigned Opc = is64bit ? AArch64::LDRDui : AArch64::LDRSui;
+ unsigned Opc = Is64Bit ? AArch64::LDRDui : AArch64::LDRSui;
unsigned ResultReg = createResultReg(TLI.getRegClassFor(VT));
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(Opc), ResultReg)
- .addReg(ADRPReg)
- .addConstantPoolIndex(Idx, 0, AArch64II::MO_PAGEOFF | AArch64II::MO_NC);
+ .addReg(ADRPReg)
+ .addConstantPoolIndex(CPI, 0, AArch64II::MO_PAGEOFF | AArch64II::MO_NC);
return ResultReg;
}
// ADRP + LDRX
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::ADRP),
ADRPReg)
- .addGlobalAddress(GV, 0, AArch64II::MO_GOT | AArch64II::MO_PAGE);
+ .addGlobalAddress(GV, 0, AArch64II::MO_GOT | AArch64II::MO_PAGE);
ResultReg = createResultReg(&AArch64::GPR64RegClass);
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::LDRXui),
ResultReg)
- .addReg(ADRPReg)
- .addGlobalAddress(GV, 0, AArch64II::MO_GOT | AArch64II::MO_PAGEOFF |
- AArch64II::MO_NC);
+ .addReg(ADRPReg)
+ .addGlobalAddress(GV, 0, AArch64II::MO_GOT | AArch64II::MO_PAGEOFF |
+ AArch64II::MO_NC);
} else {
// ADRP + ADDX
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::ADRP),
- ADRPReg).addGlobalAddress(GV, 0, AArch64II::MO_PAGE);
+ ADRPReg)
+ .addGlobalAddress(GV, 0, AArch64II::MO_PAGE);
ResultReg = createResultReg(&AArch64::GPR64spRegClass);
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::ADDXri),
ResultReg)
- .addReg(ADRPReg)
- .addGlobalAddress(GV, 0, AArch64II::MO_PAGEOFF | AArch64II::MO_NC)
- .addImm(0);
+ .addReg(ADRPReg)
+ .addGlobalAddress(GV, 0, AArch64II::MO_PAGEOFF | AArch64II::MO_NC)
+ .addImm(0);
}
return ResultReg;
}
return 0;
MVT VT = CEVT.getSimpleVT();
- // FIXME: Handle ConstantInt.
- if (const ConstantFP *CFP = dyn_cast<ConstantFP>(C))
+ if (const auto *CI = dyn_cast<ConstantInt>(C))
+ return AArch64MaterializeInt(CI, VT);
+ else if (const ConstantFP *CFP = dyn_cast<ConstantFP>(C))
return AArch64MaterializeFP(CFP, VT);
else if (const GlobalValue *GV = dyn_cast<GlobalValue>(C))
return AArch64MaterializeGV(GV);
void AArch64FastISel::AddLoadStoreOperands(Address &Addr,
const MachineInstrBuilder &MIB,
- unsigned Flags, bool UseUnscaled) {
+ unsigned Flags,
+ MachineMemOperand *MMO,
+ bool UseUnscaled) {
int64_t Offset = Addr.getOffset();
// Frame base works a bit differently. Handle it separately.
if (Addr.getKind() == Address::FrameIndexBase) {
int FI = Addr.getFI();
// FIXME: We shouldn't be using getObjectSize/getObjectAlignment. The size
// and alignment should be based on the VT.
- MachineMemOperand *MMO = FuncInfo.MF->getMachineMemOperand(
- MachinePointerInfo::getFixedStack(FI, Offset), Flags,
- MFI.getObjectSize(FI), MFI.getObjectAlignment(FI));
+ MMO = FuncInfo.MF->getMachineMemOperand(
+ MachinePointerInfo::getFixedStack(FI, Offset), Flags,
+ MFI.getObjectSize(FI), MFI.getObjectAlignment(FI));
// Now add the rest of the operands.
- MIB.addFrameIndex(FI).addImm(Offset).addMemOperand(MMO);
+ MIB.addFrameIndex(FI).addImm(Offset);
} else {
// Now add the rest of the operands.
MIB.addReg(Addr.getReg());
MIB.addImm(Offset);
}
+
+ if (MMO)
+ MIB.addMemOperand(MMO);
}
bool AArch64FastISel::EmitLoad(MVT VT, unsigned &ResultReg, Address Addr,
- bool UseUnscaled) {
+ MachineMemOperand *MMO, bool UseUnscaled) {
// Negative offsets require unscaled, 9-bit, signed immediate offsets.
// Otherwise, we try using scaled, 12-bit, unsigned immediate offsets.
if (!UseUnscaled && Addr.getOffset() < 0)
int64_t Offset = Addr.getOffset();
if (Offset & (ScaleFactor - 1))
// Retry using an unscaled, 9-bit, signed immediate offset.
- return EmitLoad(VT, ResultReg, Addr, /*UseUnscaled*/ true);
+ return EmitLoad(VT, ResultReg, Addr, MMO, /*UseUnscaled*/ true);
Addr.setOffset(Offset / ScaleFactor);
}
ResultReg = createResultReg(RC);
MachineInstrBuilder MIB = BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
TII.get(Opc), ResultReg);
- AddLoadStoreOperands(Addr, MIB, MachineMemOperand::MOLoad, UseUnscaled);
+ AddLoadStoreOperands(Addr, MIB, MachineMemOperand::MOLoad, MMO, UseUnscaled);
// Loading an i1 requires special handling.
if (VTIsi1) {
return false;
unsigned ResultReg;
- if (!EmitLoad(VT, ResultReg, Addr))
+ if (!EmitLoad(VT, ResultReg, Addr, createMachineMemOperandFor(I)))
return false;
UpdateValueMap(I, ResultReg);
}
bool AArch64FastISel::EmitStore(MVT VT, unsigned SrcReg, Address Addr,
- bool UseUnscaled) {
+ MachineMemOperand *MMO, bool UseUnscaled) {
// Negative offsets require unscaled, 9-bit, signed immediate offsets.
// Otherwise, we try using scaled, 12-bit, unsigned immediate offsets.
if (!UseUnscaled && Addr.getOffset() < 0)
int64_t Offset = Addr.getOffset();
if (Offset & (ScaleFactor - 1))
// Retry using an unscaled, 9-bit, signed immediate offset.
- return EmitStore(VT, SrcReg, Addr, /*UseUnscaled*/ true);
+ return EmitStore(VT, SrcReg, Addr, MMO, /*UseUnscaled*/ true);
Addr.setOffset(Offset / ScaleFactor);
}
// Create the base instruction, then add the operands.
MachineInstrBuilder MIB = BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
TII.get(StrOpc)).addReg(SrcReg);
- AddLoadStoreOperands(Addr, MIB, MachineMemOperand::MOStore, UseUnscaled);
+ AddLoadStoreOperands(Addr, MIB, MachineMemOperand::MOStore, MMO, UseUnscaled);
+
return true;
}
if (!ComputeAddress(I->getOperand(1), Addr))
return false;
- if (!EmitStore(VT, SrcReg, Addr))
+ if (!EmitStore(VT, SrcReg, Addr, createMachineMemOperandFor(I)))
return false;
return true;
}
return true;
}
+bool AArch64FastISel::FastLowerArguments() {
+ if (!FuncInfo.CanLowerReturn)
+ return false;
+
+ const Function *F = FuncInfo.Fn;
+ if (F->isVarArg())
+ return false;
+
+ CallingConv::ID CC = F->getCallingConv();
+ if (CC != CallingConv::C)
+ return false;
+
+ // Only handle simple cases like i1/i8/i16/i32/i64/f32/f64 of up to 8 GPR and
+ // FPR each.
+ unsigned GPRCnt = 0;
+ unsigned FPRCnt = 0;
+ unsigned Idx = 0;
+ for (auto const &Arg : F->args()) {
+ // The first argument is at index 1.
+ ++Idx;
+ if (F->getAttributes().hasAttribute(Idx, Attribute::ByVal) ||
+ F->getAttributes().hasAttribute(Idx, Attribute::InReg) ||
+ F->getAttributes().hasAttribute(Idx, Attribute::StructRet) ||
+ F->getAttributes().hasAttribute(Idx, Attribute::Nest))
+ return false;
+
+ Type *ArgTy = Arg.getType();
+ if (ArgTy->isStructTy() || ArgTy->isArrayTy() || ArgTy->isVectorTy())
+ return false;
+
+ EVT ArgVT = TLI.getValueType(ArgTy);
+ if (!ArgVT.isSimple()) return false;
+ switch (ArgVT.getSimpleVT().SimpleTy) {
+ default: return false;
+ case MVT::i1:
+ case MVT::i8:
+ case MVT::i16:
+ case MVT::i32:
+ case MVT::i64:
+ ++GPRCnt;
+ break;
+ case MVT::f16:
+ case MVT::f32:
+ case MVT::f64:
+ ++FPRCnt;
+ break;
+ }
+
+ if (GPRCnt > 8 || FPRCnt > 8)
+ return false;
+ }
+
+ static const MCPhysReg Registers[5][8] = {
+ { AArch64::W0, AArch64::W1, AArch64::W2, AArch64::W3, AArch64::W4,
+ AArch64::W5, AArch64::W6, AArch64::W7 },
+ { AArch64::X0, AArch64::X1, AArch64::X2, AArch64::X3, AArch64::X4,
+ AArch64::X5, AArch64::X6, AArch64::X7 },
+ { AArch64::H0, AArch64::H1, AArch64::H2, AArch64::H3, AArch64::H4,
+ AArch64::H5, AArch64::H6, AArch64::H7 },
+ { AArch64::S0, AArch64::S1, AArch64::S2, AArch64::S3, AArch64::S4,
+ AArch64::S5, AArch64::S6, AArch64::S7 },
+ { AArch64::D0, AArch64::D1, AArch64::D2, AArch64::D3, AArch64::D4,
+ AArch64::D5, AArch64::D6, AArch64::D7 }
+ };
+
+ unsigned GPRIdx = 0;
+ unsigned FPRIdx = 0;
+ for (auto const &Arg : F->args()) {
+ MVT VT = TLI.getSimpleValueType(Arg.getType());
+ unsigned SrcReg;
+ switch (VT.SimpleTy) {
+ default: llvm_unreachable("Unexpected value type.");
+ case MVT::i1:
+ case MVT::i8:
+ case MVT::i16: VT = MVT::i32; // fall-through
+ case MVT::i32: SrcReg = Registers[0][GPRIdx++]; break;
+ case MVT::i64: SrcReg = Registers[1][GPRIdx++]; break;
+ case MVT::f16: SrcReg = Registers[2][FPRIdx++]; break;
+ case MVT::f32: SrcReg = Registers[3][FPRIdx++]; break;
+ case MVT::f64: SrcReg = Registers[4][FPRIdx++]; break;
+ }
+
+ // Skip unused arguments.
+ if (Arg.use_empty()) {
+ UpdateValueMap(&Arg, 0);
+ continue;
+ }
+
+ const TargetRegisterClass *RC = TLI.getRegClassFor(VT);
+ unsigned DstReg = FuncInfo.MF->addLiveIn(SrcReg, RC);
+ // FIXME: Unfortunately it's necessary to emit a copy from the livein copy.
+ // Without this, EmitLiveInCopies may eliminate the livein if its only
+ // use is a bitcast (which isn't turned into an instruction).
+ unsigned ResultReg = createResultReg(RC);
+ BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
+ TII.get(TargetOpcode::COPY), ResultReg)
+ .addReg(DstReg, getKillRegState(true));
+ UpdateValueMap(&Arg, ResultReg);
+ }
+ return true;
+}
+
bool AArch64FastISel::ProcessCallArgs(CallLoweringInfo &CLI,
SmallVectorImpl<MVT> &OutVTs,
unsigned &NumBytes) {
CallingConv::ID CC = CLI.CallConv;
SmallVector<CCValAssign, 16> ArgLocs;
- CCState CCInfo(CC, false, *FuncInfo.MF, TM, ArgLocs, *Context);
+ CCState CCInfo(CC, false, *FuncInfo.MF, ArgLocs, *Context);
CCInfo.AnalyzeCallOperands(OutVTs, CLI.OutFlags, CCAssignFnForCall(CC));
// Get a count of how many bytes are to be pushed on the stack.
Addr.setReg(AArch64::SP);
Addr.setOffset(VA.getLocMemOffset() + BEAlign);
- if (!EmitStore(ArgVT, ArgReg, Addr))
+ unsigned Alignment = DL.getABITypeAlignment(ArgVal->getType());
+ MachineMemOperand *MMO = FuncInfo.MF->getMachineMemOperand(
+ MachinePointerInfo::getStack(Addr.getOffset()),
+ MachineMemOperand::MOStore, ArgVT.getStoreSize(), Alignment);
+
+ if (!EmitStore(ArgVT, ArgReg, Addr, MMO))
return false;
}
}
// Now the return value.
if (RetVT != MVT::isVoid) {
SmallVector<CCValAssign, 16> RVLocs;
- CCState CCInfo(CC, false, *FuncInfo.MF, TM, RVLocs, *Context);
+ CCState CCInfo(CC, false, *FuncInfo.MF, RVLocs, *Context);
CCInfo.AnalyzeCallResult(RetVT, CCAssignFnForCall(CC));
// Only handle a single return value.
bool AArch64FastISel::FastLowerCall(CallLoweringInfo &CLI) {
CallingConv::ID CC = CLI.CallConv;
+ bool IsTailCall = CLI.IsTailCall;
bool IsVarArg = CLI.IsVarArg;
const Value *Callee = CLI.Callee;
const char *SymName = CLI.SymName;
+ // Allow SelectionDAG isel to handle tail calls.
+ if (IsTailCall)
+ return false;
+
CodeModel::Model CM = TM.getCodeModel();
// Only support the small and large code model.
if (CM != CodeModel::Small && CM != CodeModel::Large)
// Analyze operands of the call, assigning locations to each operand.
SmallVector<CCValAssign, 16> ValLocs;
- CCState CCInfo(CC, F.isVarArg(), *FuncInfo.MF, TM, ValLocs,
- I->getContext());
+ CCState CCInfo(CC, F.isVarArg(), *FuncInfo.MF, ValLocs, I->getContext());
CCAssignFn *RetCC = CC == CallingConv::WebKit_JS ? RetCC_AArch64_WebKit_JS
: RetCC_AArch64_AAPCS;
CCInfo.AnalyzeReturn(Outs, RetCC);
MVT SrcVT = SrcEVT.getSimpleVT();
MVT DestVT = DestEVT.getSimpleVT();
- unsigned ResultReg = EmitIntExt(SrcVT, SrcReg, DestVT, isZExt);
- if (ResultReg == 0)
+ unsigned ResultReg = 0;
+
+ // Check if it is an argument and if it is already zero/sign-extended.
+ if (const auto *Arg = dyn_cast<Argument>(Src)) {
+ if ((isZExt && Arg->hasZExtAttr()) || (!isZExt && Arg->hasSExtAttr())) {
+ if (DestVT == MVT::i64) {
+ ResultReg = createResultReg(TLI.getRegClassFor(DestVT));
+ BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
+ TII.get(AArch64::SUBREG_TO_REG), ResultReg)
+ .addImm(0)
+ .addReg(SrcReg)
+ .addImm(AArch64::sub_32);
+ } else
+ ResultReg = SrcReg;
+ }
+ }
+
+ if (!ResultReg)
+ ResultReg = EmitIntExt(SrcVT, SrcReg, DestVT, isZExt);
+
+ if (!ResultReg)
return false;
+
UpdateValueMap(I, ResultReg);
return true;
}
projects / oota-llvm.git / blobdiff