X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTarget%2FXCore%2FXCoreISelLowering.cpp;h=5af2c9c0cc32a8873584c4a7dc1ec954a9d0c4c1;hb=5b3fca50a08865f0db55fc92ad1c037a04e12177;hp=8231d1baef2e876957b1026c92e9fc0d35ae4234;hpb=7e2c793a2b5c746344652b6579e958ee42fafdcc;p=oota-llvm.git diff --git a/lib/Target/XCore/XCoreISelLowering.cpp b/lib/Target/XCore/XCoreISelLowering.cpp index 8231d1baef2..5af2c9c0cc3 100644 --- a/lib/Target/XCore/XCoreISelLowering.cpp +++ b/lib/Target/XCore/XCoreISelLowering.cpp @@ -14,17 +14,11 @@ #define DEBUG_TYPE "xcore-lower" #include "XCoreISelLowering.h" -#include "XCoreMachineFunctionInfo.h" #include "XCore.h" -#include "XCoreTargetObjectFile.h" -#include "XCoreTargetMachine.h" +#include "XCoreMachineFunctionInfo.h" #include "XCoreSubtarget.h" -#include "llvm/DerivedTypes.h" -#include "llvm/Function.h" -#include "llvm/Intrinsics.h" -#include "llvm/CallingConv.h" -#include "llvm/GlobalVariable.h" -#include "llvm/GlobalAlias.h" +#include "XCoreTargetMachine.h" +#include "XCoreTargetObjectFile.h" #include "llvm/CodeGen/CallingConvLower.h" #include "llvm/CodeGen/MachineFrameInfo.h" #include "llvm/CodeGen/MachineFunction.h" @@ -33,9 +27,17 @@ #include "llvm/CodeGen/MachineRegisterInfo.h" #include "llvm/CodeGen/SelectionDAGISel.h" #include "llvm/CodeGen/ValueTypes.h" +#include "llvm/IR/CallingConv.h" +#include "llvm/IR/DerivedTypes.h" +#include "llvm/IR/Function.h" +#include "llvm/IR/GlobalAlias.h" +#include "llvm/IR/GlobalVariable.h" +#include "llvm/IR/Intrinsics.h" #include "llvm/Support/Debug.h" #include "llvm/Support/ErrorHandling.h" #include "llvm/Support/raw_ostream.h" +#include + using namespace llvm; const char *XCoreTargetLowering:: @@ -54,6 +56,7 @@ getTargetNodeName(unsigned Opcode) const case XCoreISD::LMUL : return "XCoreISD::LMUL"; case XCoreISD::MACCU : return "XCoreISD::MACCU"; case XCoreISD::MACCS : return "XCoreISD::MACCS"; + case XCoreISD::CRC8 : return "XCoreISD::CRC8"; case XCoreISD::BR_JT : return "XCoreISD::BR_JT"; case XCoreISD::BR_JT32 : return "XCoreISD::BR_JT32"; default : return NULL; @@ -83,7 +86,7 @@ XCoreTargetLowering::XCoreTargetLowering(XCoreTargetMachine &XTM) setBooleanVectorContents(ZeroOrOneBooleanContent); // FIXME: Is this correct? // XCore does not have the NodeTypes below. - setOperationAction(ISD::BR_CC, MVT::Other, Expand); + setOperationAction(ISD::BR_CC, MVT::i32, Expand); setOperationAction(ISD::SELECT_CC, MVT::i32, Custom); setOperationAction(ISD::ADDC, MVT::i32, Expand); setOperationAction(ISD::ADDE, MVT::i32, Expand); @@ -119,9 +122,6 @@ XCoreTargetLowering::XCoreTargetLowering(XCoreTargetMachine &XTM) setOperationAction(ISD::GlobalAddress, MVT::i32, Custom); setOperationAction(ISD::BlockAddress, MVT::i32 , Custom); - // Thread Local Storage - setOperationAction(ISD::GlobalTLSAddress, MVT::i32, Custom); - // Conversion of i64 -> double produces constantpool nodes setOperationAction(ISD::ConstantPool, MVT::i32, Custom); @@ -152,9 +152,12 @@ XCoreTargetLowering::XCoreTargetLowering(XCoreTargetMachine &XTM) setOperationAction(ISD::INIT_TRAMPOLINE, MVT::Other, Custom); setOperationAction(ISD::ADJUST_TRAMPOLINE, MVT::Other, Custom); - maxStoresPerMemset = maxStoresPerMemsetOptSize = 4; - maxStoresPerMemmove = maxStoresPerMemmoveOptSize - = maxStoresPerMemcpy = maxStoresPerMemcpyOptSize = 2; + // We want to custom lower some of our intrinsics. + setOperationAction(ISD::INTRINSIC_WO_CHAIN, MVT::Other, Custom); + + MaxStoresPerMemset = MaxStoresPerMemsetOptSize = 4; + MaxStoresPerMemmove = MaxStoresPerMemmoveOptSize + = MaxStoresPerMemcpy = MaxStoresPerMemcpyOptSize = 2; // We have target-specific dag combine patterns for the following nodes: setTargetDAGCombine(ISD::STORE); @@ -167,24 +170,24 @@ SDValue XCoreTargetLowering:: LowerOperation(SDValue Op, SelectionDAG &DAG) const { switch (Op.getOpcode()) { - case ISD::GlobalAddress: return LowerGlobalAddress(Op, DAG); - case ISD::GlobalTLSAddress: return LowerGlobalTLSAddress(Op, DAG); - case ISD::BlockAddress: return LowerBlockAddress(Op, DAG); - case ISD::ConstantPool: return LowerConstantPool(Op, DAG); - case ISD::BR_JT: return LowerBR_JT(Op, DAG); - case ISD::LOAD: return LowerLOAD(Op, DAG); - case ISD::STORE: return LowerSTORE(Op, DAG); - case ISD::SELECT_CC: return LowerSELECT_CC(Op, DAG); - case ISD::VAARG: return LowerVAARG(Op, DAG); - case ISD::VASTART: return LowerVASTART(Op, DAG); - case ISD::SMUL_LOHI: return LowerSMUL_LOHI(Op, DAG); - case ISD::UMUL_LOHI: return LowerUMUL_LOHI(Op, DAG); + case ISD::GlobalAddress: return LowerGlobalAddress(Op, DAG); + case ISD::BlockAddress: return LowerBlockAddress(Op, DAG); + case ISD::ConstantPool: return LowerConstantPool(Op, DAG); + case ISD::BR_JT: return LowerBR_JT(Op, DAG); + case ISD::LOAD: return LowerLOAD(Op, DAG); + case ISD::STORE: return LowerSTORE(Op, DAG); + case ISD::SELECT_CC: return LowerSELECT_CC(Op, DAG); + case ISD::VAARG: return LowerVAARG(Op, DAG); + case ISD::VASTART: return LowerVASTART(Op, DAG); + case ISD::SMUL_LOHI: return LowerSMUL_LOHI(Op, DAG); + case ISD::UMUL_LOHI: return LowerUMUL_LOHI(Op, DAG); // FIXME: Remove these when LegalizeDAGTypes lands. case ISD::ADD: - case ISD::SUB: return ExpandADDSUB(Op.getNode(), DAG); - case ISD::FRAMEADDR: return LowerFRAMEADDR(Op, DAG); - case ISD::INIT_TRAMPOLINE: return LowerINIT_TRAMPOLINE(Op, DAG); - case ISD::ADJUST_TRAMPOLINE: return LowerADJUST_TRAMPOLINE(Op, DAG); + case ISD::SUB: return ExpandADDSUB(Op.getNode(), DAG); + case ISD::FRAMEADDR: return LowerFRAMEADDR(Op, DAG); + case ISD::INIT_TRAMPOLINE: return LowerINIT_TRAMPOLINE(Op, DAG); + case ISD::ADJUST_TRAMPOLINE: return LowerADJUST_TRAMPOLINE(Op, DAG); + case ISD::INTRINSIC_WO_CHAIN: return LowerINTRINSIC_WO_CHAIN(Op, DAG); default: llvm_unreachable("unimplemented operand"); } @@ -212,7 +215,7 @@ void XCoreTargetLowering::ReplaceNodeResults(SDNode *N, SDValue XCoreTargetLowering:: LowerSELECT_CC(SDValue Op, SelectionDAG &DAG) const { - DebugLoc dl = Op.getDebugLoc(); + SDLoc dl(Op); SDValue Cond = DAG.getNode(ISD::SETCC, dl, MVT::i32, Op.getOperand(2), Op.getOperand(3), Op.getOperand(4)); return DAG.getNode(ISD::SELECT, dl, MVT::i32, Cond, Op.getOperand(0), @@ -224,78 +227,47 @@ getGlobalAddressWrapper(SDValue GA, const GlobalValue *GV, SelectionDAG &DAG) const { // FIXME there is no actual debug info here - DebugLoc dl = GA.getDebugLoc(); - if (isa(GV)) { - return DAG.getNode(XCoreISD::PCRelativeWrapper, dl, MVT::i32, GA); - } - const GlobalVariable *GVar = dyn_cast(GV); - if (!GVar) { - // If GV is an alias then use the aliasee to determine constness - if (const GlobalAlias *GA = dyn_cast(GV)) - GVar = dyn_cast_or_null(GA->resolveAliasedGlobal()); + SDLoc dl(GA); + const GlobalValue *UnderlyingGV = GV; + // If GV is an alias then use the aliasee to determine the wrapper type + if (const GlobalAlias *GA = dyn_cast(GV)) + UnderlyingGV = GA->resolveAliasedGlobal(); + if (const GlobalVariable *GVar = dyn_cast(UnderlyingGV)) { + if (GVar->isConstant()) + return DAG.getNode(XCoreISD::CPRelativeWrapper, dl, MVT::i32, GA); + return DAG.getNode(XCoreISD::DPRelativeWrapper, dl, MVT::i32, GA); } - bool isConst = GVar && GVar->isConstant(); - if (isConst) { - return DAG.getNode(XCoreISD::CPRelativeWrapper, dl, MVT::i32, GA); - } - return DAG.getNode(XCoreISD::DPRelativeWrapper, dl, MVT::i32, GA); + return DAG.getNode(XCoreISD::PCRelativeWrapper, dl, MVT::i32, GA); } SDValue XCoreTargetLowering:: LowerGlobalAddress(SDValue Op, SelectionDAG &DAG) const { - const GlobalValue *GV = cast(Op)->getGlobal(); - SDValue GA = DAG.getTargetGlobalAddress(GV, Op.getDebugLoc(), MVT::i32); - return getGlobalAddressWrapper(GA, GV, DAG); + SDLoc DL(Op); + const GlobalAddressSDNode *GN = cast(Op); + const GlobalValue *GV = GN->getGlobal(); + int64_t Offset = GN->getOffset(); + // We can only fold positive offsets that are a multiple of the word size. + int64_t FoldedOffset = std::max(Offset & ~3, (int64_t)0); + SDValue GA = DAG.getTargetGlobalAddress(GV, DL, MVT::i32, FoldedOffset); + GA = getGlobalAddressWrapper(GA, GV, DAG); + // Handle the rest of the offset. + if (Offset != FoldedOffset) { + SDValue Remaining = DAG.getConstant(Offset - FoldedOffset, MVT::i32); + GA = DAG.getNode(ISD::ADD, DL, MVT::i32, GA, Remaining); + } + return GA; } -static inline SDValue BuildGetId(SelectionDAG &DAG, DebugLoc dl) { +static inline SDValue BuildGetId(SelectionDAG &DAG, SDLoc dl) { return DAG.getNode(ISD::INTRINSIC_WO_CHAIN, dl, MVT::i32, DAG.getConstant(Intrinsic::xcore_getid, MVT::i32)); } -static inline bool isZeroLengthArray(Type *Ty) { - ArrayType *AT = dyn_cast_or_null(Ty); - return AT && (AT->getNumElements() == 0); -} - -SDValue XCoreTargetLowering:: -LowerGlobalTLSAddress(SDValue Op, SelectionDAG &DAG) const -{ - // FIXME there isn't really debug info here - DebugLoc dl = Op.getDebugLoc(); - // transform to label + getid() * size - const GlobalValue *GV = cast(Op)->getGlobal(); - SDValue GA = DAG.getTargetGlobalAddress(GV, dl, MVT::i32); - const GlobalVariable *GVar = dyn_cast(GV); - if (!GVar) { - // If GV is an alias then use the aliasee to determine size - if (const GlobalAlias *GA = dyn_cast(GV)) - GVar = dyn_cast_or_null(GA->resolveAliasedGlobal()); - } - if (!GVar) { - llvm_unreachable("Thread local object not a GlobalVariable?"); - } - Type *Ty = cast(GV->getType())->getElementType(); - if (!Ty->isSized() || isZeroLengthArray(Ty)) { -#ifndef NDEBUG - errs() << "Size of thread local object " << GVar->getName() - << " is unknown\n"; -#endif - llvm_unreachable(0); - } - SDValue base = getGlobalAddressWrapper(GA, GV, DAG); - const TargetData *TD = TM.getTargetData(); - unsigned Size = TD->getTypeAllocSize(Ty); - SDValue offset = DAG.getNode(ISD::MUL, dl, MVT::i32, BuildGetId(DAG, dl), - DAG.getConstant(Size, MVT::i32)); - return DAG.getNode(ISD::ADD, dl, MVT::i32, base, offset); -} - SDValue XCoreTargetLowering:: LowerBlockAddress(SDValue Op, SelectionDAG &DAG) const { - DebugLoc DL = Op.getDebugLoc(); + SDLoc DL(Op); const BlockAddress *BA = cast(Op)->getBlockAddress(); SDValue Result = DAG.getTargetBlockAddress(BA, getPointerTy()); @@ -308,7 +280,7 @@ LowerConstantPool(SDValue Op, SelectionDAG &DAG) const { ConstantPoolSDNode *CP = cast(Op); // FIXME there isn't really debug info here - DebugLoc dl = CP->getDebugLoc(); + SDLoc dl(CP); EVT PtrVT = Op.getValueType(); SDValue Res; if (CP->isMachineConstantPoolEntry()) { @@ -331,7 +303,7 @@ LowerBR_JT(SDValue Op, SelectionDAG &DAG) const SDValue Chain = Op.getOperand(0); SDValue Table = Op.getOperand(1); SDValue Index = Op.getOperand(2); - DebugLoc dl = Op.getDebugLoc(); + SDLoc dl(Op); JumpTableSDNode *JT = cast(Table); unsigned JTI = JT->getIndex(); MachineFunction &MF = DAG.getMachineFunction(); @@ -349,55 +321,58 @@ LowerBR_JT(SDValue Op, SelectionDAG &DAG) const ScaledIndex); } -static bool -IsWordAlignedBasePlusConstantOffset(SDValue Addr, SDValue &AlignedBase, - int64_t &Offset) +SDValue XCoreTargetLowering:: +lowerLoadWordFromAlignedBasePlusOffset(SDLoc DL, SDValue Chain, SDValue Base, + int64_t Offset, SelectionDAG &DAG) const { - if (Addr.getOpcode() != ISD::ADD) { - return false; - } - ConstantSDNode *CN = 0; - if (!(CN = dyn_cast(Addr.getOperand(1)))) { - return false; - } - int64_t off = CN->getSExtValue(); - const SDValue &Base = Addr.getOperand(0); - const SDValue *Root = &Base; - if (Base.getOpcode() == ISD::ADD && - Base.getOperand(1).getOpcode() == ISD::SHL) { - ConstantSDNode *CN = dyn_cast(Base.getOperand(1) - .getOperand(1)); - if (CN && (CN->getSExtValue() >= 2)) { - Root = &Base.getOperand(0); - } - } - if (isa(*Root)) { - // All frame indicies are word aligned - AlignedBase = Base; - Offset = off; - return true; - } - if (Root->getOpcode() == XCoreISD::DPRelativeWrapper || - Root->getOpcode() == XCoreISD::CPRelativeWrapper) { - // All dp / cp relative addresses are word aligned - AlignedBase = Base; - Offset = off; - return true; + if ((Offset & 0x3) == 0) { + return DAG.getLoad(getPointerTy(), DL, Chain, Base, MachinePointerInfo(), + false, false, false, 0); } - // Check for an aligned global variable. - if (GlobalAddressSDNode *GA = dyn_cast(*Root)) { - const GlobalValue *GV = GA->getGlobal(); - if (GA->getOffset() == 0 && GV->getAlignment() >= 4) { - AlignedBase = Base; - Offset = off; - return true; - } + // Lower to pair of consecutive word aligned loads plus some bit shifting. + int32_t HighOffset = RoundUpToAlignment(Offset, 4); + int32_t LowOffset = HighOffset - 4; + SDValue LowAddr, HighAddr; + if (GlobalAddressSDNode *GASD = + dyn_cast(Base.getNode())) { + LowAddr = DAG.getGlobalAddress(GASD->getGlobal(), DL, Base.getValueType(), + LowOffset); + HighAddr = DAG.getGlobalAddress(GASD->getGlobal(), DL, Base.getValueType(), + HighOffset); + } else { + LowAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, Base, + DAG.getConstant(LowOffset, MVT::i32)); + HighAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, Base, + DAG.getConstant(HighOffset, MVT::i32)); } - return false; + SDValue LowShift = DAG.getConstant((Offset - LowOffset) * 8, MVT::i32); + SDValue HighShift = DAG.getConstant((HighOffset - Offset) * 8, MVT::i32); + + SDValue Low = DAG.getLoad(getPointerTy(), DL, Chain, + LowAddr, MachinePointerInfo(), + false, false, false, 0); + SDValue High = DAG.getLoad(getPointerTy(), DL, Chain, + HighAddr, MachinePointerInfo(), + false, false, false, 0); + SDValue LowShifted = DAG.getNode(ISD::SRL, DL, MVT::i32, Low, LowShift); + SDValue HighShifted = DAG.getNode(ISD::SHL, DL, MVT::i32, High, HighShift); + SDValue Result = DAG.getNode(ISD::OR, DL, MVT::i32, LowShifted, HighShifted); + Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Low.getValue(1), + High.getValue(1)); + SDValue Ops[] = { Result, Chain }; + return DAG.getMergeValues(Ops, 2, DL); +} + +static bool isWordAligned(SDValue Value, SelectionDAG &DAG) +{ + APInt KnownZero, KnownOne; + DAG.ComputeMaskedBits(Value, KnownZero, KnownOne); + return KnownZero.countTrailingOnes() >= 2; } SDValue XCoreTargetLowering:: LowerLOAD(SDValue Op, SelectionDAG &DAG) const { + const TargetLowering &TLI = DAG.getTargetLoweringInfo(); LoadSDNode *LD = cast(Op); assert(LD->getExtensionType() == ISD::NON_EXTLOAD && "Unexpected extension type"); @@ -405,7 +380,7 @@ LowerLOAD(SDValue Op, SelectionDAG &DAG) const { if (allowsUnalignedMemoryAccesses(LD->getMemoryVT())) return SDValue(); - unsigned ABIAlignment = getTargetData()-> + unsigned ABIAlignment = getDataLayout()-> getABITypeAlignment(LD->getMemoryVT().getTypeForEVT(*DAG.getContext())); // Leave aligned load alone. if (LD->getAlignment() >= ABIAlignment) @@ -413,47 +388,25 @@ LowerLOAD(SDValue Op, SelectionDAG &DAG) const { SDValue Chain = LD->getChain(); SDValue BasePtr = LD->getBasePtr(); - DebugLoc DL = Op.getDebugLoc(); - - SDValue Base; - int64_t Offset; - if (!LD->isVolatile() && - IsWordAlignedBasePlusConstantOffset(BasePtr, Base, Offset)) { - if (Offset % 4 == 0) { - // We've managed to infer better alignment information than the load - // already has. Use an aligned load. - // - return DAG.getLoad(getPointerTy(), DL, Chain, BasePtr, - MachinePointerInfo(), - false, false, false, 0); + SDLoc DL(Op); + + if (!LD->isVolatile()) { + const GlobalValue *GV; + int64_t Offset = 0; + if (DAG.isBaseWithConstantOffset(BasePtr) && + isWordAligned(BasePtr->getOperand(0), DAG)) { + SDValue NewBasePtr = BasePtr->getOperand(0); + Offset = cast(BasePtr->getOperand(1))->getSExtValue(); + return lowerLoadWordFromAlignedBasePlusOffset(DL, Chain, NewBasePtr, + Offset, DAG); + } + if (TLI.isGAPlusOffset(BasePtr.getNode(), GV, Offset) && + MinAlign(GV->getAlignment(), 4) == 4) { + SDValue NewBasePtr = DAG.getGlobalAddress(GV, DL, + BasePtr->getValueType(0)); + return lowerLoadWordFromAlignedBasePlusOffset(DL, Chain, NewBasePtr, + Offset, DAG); } - // Lower to - // ldw low, base[offset >> 2] - // ldw high, base[(offset >> 2) + 1] - // shr low_shifted, low, (offset & 0x3) * 8 - // shl high_shifted, high, 32 - (offset & 0x3) * 8 - // or result, low_shifted, high_shifted - SDValue LowOffset = DAG.getConstant(Offset & ~0x3, MVT::i32); - SDValue HighOffset = DAG.getConstant((Offset & ~0x3) + 4, MVT::i32); - SDValue LowShift = DAG.getConstant((Offset & 0x3) * 8, MVT::i32); - SDValue HighShift = DAG.getConstant(32 - (Offset & 0x3) * 8, MVT::i32); - - SDValue LowAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, Base, LowOffset); - SDValue HighAddr = DAG.getNode(ISD::ADD, DL, MVT::i32, Base, HighOffset); - - SDValue Low = DAG.getLoad(getPointerTy(), DL, Chain, - LowAddr, MachinePointerInfo(), - false, false, false, 0); - SDValue High = DAG.getLoad(getPointerTy(), DL, Chain, - HighAddr, MachinePointerInfo(), - false, false, false, 0); - SDValue LowShifted = DAG.getNode(ISD::SRL, DL, MVT::i32, Low, LowShift); - SDValue HighShifted = DAG.getNode(ISD::SHL, DL, MVT::i32, High, HighShift); - SDValue Result = DAG.getNode(ISD::OR, DL, MVT::i32, LowShifted, HighShifted); - Chain = DAG.getNode(ISD::TokenFactor, DL, MVT::Other, Low.getValue(1), - High.getValue(1)); - SDValue Ops[] = { Result, Chain }; - return DAG.getMergeValues(Ops, 2, DL); } if (LD->getAlignment() == 2) { @@ -477,7 +430,7 @@ LowerLOAD(SDValue Op, SelectionDAG &DAG) const { } // Lower to a call to __misaligned_load(BasePtr). - Type *IntPtrTy = getTargetData()->getIntPtrType(*DAG.getContext()); + Type *IntPtrTy = getDataLayout()->getIntPtrType(*DAG.getContext()); TargetLowering::ArgListTy Args; TargetLowering::ArgListEntry Entry; @@ -507,7 +460,7 @@ LowerSTORE(SDValue Op, SelectionDAG &DAG) const if (allowsUnalignedMemoryAccesses(ST->getMemoryVT())) { return SDValue(); } - unsigned ABIAlignment = getTargetData()-> + unsigned ABIAlignment = getDataLayout()-> getABITypeAlignment(ST->getMemoryVT().getTypeForEVT(*DAG.getContext())); // Leave aligned store alone. if (ST->getAlignment() >= ABIAlignment) { @@ -516,7 +469,7 @@ LowerSTORE(SDValue Op, SelectionDAG &DAG) const SDValue Chain = ST->getChain(); SDValue BasePtr = ST->getBasePtr(); SDValue Value = ST->getValue(); - DebugLoc dl = Op.getDebugLoc(); + SDLoc dl(Op); if (ST->getAlignment() == 2) { SDValue Low = Value; @@ -536,7 +489,7 @@ LowerSTORE(SDValue Op, SelectionDAG &DAG) const } // Lower to a call to __misaligned_store(BasePtr, Value). - Type *IntPtrTy = getTargetData()->getIntPtrType(*DAG.getContext()); + Type *IntPtrTy = getDataLayout()->getIntPtrType(*DAG.getContext()); TargetLowering::ArgListTy Args; TargetLowering::ArgListEntry Entry; @@ -563,7 +516,7 @@ LowerSMUL_LOHI(SDValue Op, SelectionDAG &DAG) const { assert(Op.getValueType() == MVT::i32 && Op.getOpcode() == ISD::SMUL_LOHI && "Unexpected operand to lower!"); - DebugLoc dl = Op.getDebugLoc(); + SDLoc dl(Op); SDValue LHS = Op.getOperand(0); SDValue RHS = Op.getOperand(1); SDValue Zero = DAG.getConstant(0, MVT::i32); @@ -580,7 +533,7 @@ LowerUMUL_LOHI(SDValue Op, SelectionDAG &DAG) const { assert(Op.getValueType() == MVT::i32 && Op.getOpcode() == ISD::UMUL_LOHI && "Unexpected operand to lower!"); - DebugLoc dl = Op.getDebugLoc(); + SDLoc dl(Op); SDValue LHS = Op.getOperand(0); SDValue RHS = Op.getOperand(1); SDValue Zero = DAG.getConstant(0, MVT::i32); @@ -665,7 +618,7 @@ TryExpandADDWithMul(SDNode *N, SelectionDAG &DAG) const } else { return SDValue(); } - DebugLoc dl = N->getDebugLoc(); + SDLoc dl(N); SDValue LL, RL, AddendL, AddendH; LL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32, Mul.getOperand(0), DAG.getConstant(0, MVT::i32)); @@ -724,7 +677,7 @@ ExpandADDSUB(SDNode *N, SelectionDAG &DAG) const return Result; } - DebugLoc dl = N->getDebugLoc(); + SDLoc dl(N); // Extract components SDValue LHSL = DAG.getNode(ISD::EXTRACT_ELEMENT, dl, MVT::i32, @@ -740,13 +693,13 @@ ExpandADDSUB(SDNode *N, SelectionDAG &DAG) const unsigned Opcode = (N->getOpcode() == ISD::ADD) ? XCoreISD::LADD : XCoreISD::LSUB; SDValue Zero = DAG.getConstant(0, MVT::i32); - SDValue Carry = DAG.getNode(Opcode, dl, DAG.getVTList(MVT::i32, MVT::i32), - LHSL, RHSL, Zero); - SDValue Lo(Carry.getNode(), 1); + SDValue Lo = DAG.getNode(Opcode, dl, DAG.getVTList(MVT::i32, MVT::i32), + LHSL, RHSL, Zero); + SDValue Carry(Lo.getNode(), 1); - SDValue Ignored = DAG.getNode(Opcode, dl, DAG.getVTList(MVT::i32, MVT::i32), - LHSH, RHSH, Carry); - SDValue Hi(Ignored.getNode(), 1); + SDValue Hi = DAG.getNode(Opcode, dl, DAG.getVTList(MVT::i32, MVT::i32), + LHSH, RHSH, Carry); + SDValue Ignored(Hi.getNode(), 1); // Merge the pieces return DAG.getNode(ISD::BUILD_PAIR, dl, MVT::i64, Lo, Hi); } @@ -757,7 +710,7 @@ LowerVAARG(SDValue Op, SelectionDAG &DAG) const llvm_unreachable("unimplemented"); // FIXME Arguments passed by reference need a extra dereference. SDNode *Node = Op.getNode(); - DebugLoc dl = Node->getDebugLoc(); + SDLoc dl(Node); const Value *V = cast(Node->getOperand(2))->getValue(); EVT VT = Node->getValueType(0); SDValue VAList = DAG.getLoad(getPointerTy(), dl, Node->getOperand(0), @@ -778,7 +731,7 @@ LowerVAARG(SDValue Op, SelectionDAG &DAG) const SDValue XCoreTargetLowering:: LowerVASTART(SDValue Op, SelectionDAG &DAG) const { - DebugLoc dl = Op.getDebugLoc(); + SDLoc dl(Op); // vastart stores the address of the VarArgsFrameIndex slot into the // memory location argument MachineFunction &MF = DAG.getMachineFunction(); @@ -790,7 +743,7 @@ LowerVASTART(SDValue Op, SelectionDAG &DAG) const SDValue XCoreTargetLowering::LowerFRAMEADDR(SDValue Op, SelectionDAG &DAG) const { - DebugLoc dl = Op.getDebugLoc(); + SDLoc dl(Op); // Depths > 0 not supported yet! if (cast(Op.getOperand(0))->getZExtValue() > 0) return SDValue(); @@ -830,7 +783,7 @@ LowerINIT_TRAMPOLINE(SDValue Op, SelectionDAG &DAG) const { SDValue Addr = Trmp; - DebugLoc dl = Op.getDebugLoc(); + SDLoc dl(Op); OutChains[0] = DAG.getStore(Chain, dl, DAG.getConstant(0x0a3cd805, MVT::i32), Addr, MachinePointerInfo(TrmpAddr), false, false, 0); @@ -862,6 +815,23 @@ LowerINIT_TRAMPOLINE(SDValue Op, SelectionDAG &DAG) const { return DAG.getNode(ISD::TokenFactor, dl, MVT::Other, OutChains, 5); } +SDValue XCoreTargetLowering:: +LowerINTRINSIC_WO_CHAIN(SDValue Op, SelectionDAG &DAG) const { + SDLoc DL(Op); + unsigned IntNo = cast(Op.getOperand(0))->getZExtValue(); + switch (IntNo) { + case Intrinsic::xcore_crc8: + EVT VT = Op.getValueType(); + SDValue Data = + DAG.getNode(XCoreISD::CRC8, DL, DAG.getVTList(VT, VT), + Op.getOperand(1), Op.getOperand(2) , Op.getOperand(3)); + SDValue Crc(Data.getNode(), 1); + SDValue Results[] = { Crc, Data }; + return DAG.getMergeValues(Results, 2, DL); + } + return SDValue(); +} + //===----------------------------------------------------------------------===// // Calling Convention Implementation //===----------------------------------------------------------------------===// @@ -877,7 +847,7 @@ SDValue XCoreTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI, SmallVectorImpl &InVals) const { SelectionDAG &DAG = CLI.DAG; - DebugLoc &dl = CLI.DL; + SDLoc &dl = CLI.DL; SmallVector &Outs = CLI.Outs; SmallVector &OutVals = CLI.OutVals; SmallVector &Ins = CLI.Ins; @@ -913,7 +883,7 @@ XCoreTargetLowering::LowerCCCCallTo(SDValue Chain, SDValue Callee, const SmallVectorImpl &Outs, const SmallVectorImpl &OutVals, const SmallVectorImpl &Ins, - DebugLoc dl, SelectionDAG &DAG, + SDLoc dl, SelectionDAG &DAG, SmallVectorImpl &InVals) const { // Analyze operands of the call, assigning locations to each operand. @@ -931,7 +901,7 @@ XCoreTargetLowering::LowerCCCCallTo(SDValue Chain, SDValue Callee, unsigned NumBytes = CCInfo.getNextStackOffset(); Chain = DAG.getCALLSEQ_START(Chain,DAG.getConstant(NumBytes, - getPointerTy(), true)); + getPointerTy(), true), dl); SmallVector, 4> RegsToPass; SmallVector MemOpChains; @@ -1021,7 +991,7 @@ XCoreTargetLowering::LowerCCCCallTo(SDValue Chain, SDValue Callee, Chain = DAG.getCALLSEQ_END(Chain, DAG.getConstant(NumBytes, getPointerTy(), true), DAG.getConstant(0, getPointerTy(), true), - InFlag); + InFlag, dl); InFlag = Chain.getValue(1); // Handle result values, copying them out of physregs into vregs that we @@ -1036,7 +1006,7 @@ SDValue XCoreTargetLowering::LowerCallResult(SDValue Chain, SDValue InFlag, CallingConv::ID CallConv, bool isVarArg, const SmallVectorImpl &Ins, - DebugLoc dl, SelectionDAG &DAG, + SDLoc dl, SelectionDAG &DAG, SmallVectorImpl &InVals) const { // Assign locations to each value returned by this call. @@ -1067,7 +1037,7 @@ XCoreTargetLowering::LowerFormalArguments(SDValue Chain, CallingConv::ID CallConv, bool isVarArg, const SmallVectorImpl &Ins, - DebugLoc dl, + SDLoc dl, SelectionDAG &DAG, SmallVectorImpl &InVals) const { @@ -1092,7 +1062,7 @@ XCoreTargetLowering::LowerCCCArguments(SDValue Chain, bool isVarArg, const SmallVectorImpl &Ins, - DebugLoc dl, + SDLoc dl, SelectionDAG &DAG, SmallVectorImpl &InVals) const { MachineFunction &MF = DAG.getMachineFunction(); @@ -1218,7 +1188,7 @@ XCoreTargetLowering::LowerReturn(SDValue Chain, CallingConv::ID CallConv, bool isVarArg, const SmallVectorImpl &Outs, const SmallVectorImpl &OutVals, - DebugLoc dl, SelectionDAG &DAG) const { + SDLoc dl, SelectionDAG &DAG) const { // CCValAssign - represent the assignment of // the return value to a location @@ -1231,15 +1201,11 @@ XCoreTargetLowering::LowerReturn(SDValue Chain, // Analyze return values. CCInfo.AnalyzeReturn(Outs, RetCC_XCore); - // If this is the first return lowered for this function, add - // the regs to the liveout set for the function. - if (DAG.getMachineFunction().getRegInfo().liveout_empty()) { - for (unsigned i = 0; i != RVLocs.size(); ++i) - if (RVLocs[i].isRegLoc()) - DAG.getMachineFunction().getRegInfo().addLiveOut(RVLocs[i].getLocReg()); - } - SDValue Flag; + SmallVector RetOps(1, Chain); + + // Return on XCore is always a "retsp 0" + RetOps.push_back(DAG.getConstant(0, MVT::i32)); // Copy the result values into the output registers. for (unsigned i = 0; i != RVLocs.size(); ++i) { @@ -1252,15 +1218,17 @@ XCoreTargetLowering::LowerReturn(SDValue Chain, // guarantee that all emitted copies are // stuck together, avoiding something bad Flag = Chain.getValue(1); + RetOps.push_back(DAG.getRegister(VA.getLocReg(), VA.getLocVT())); } - // Return on XCore is always a "retsp 0" + RetOps[0] = Chain; // Update chain. + + // Add the flag if we have it. if (Flag.getNode()) - return DAG.getNode(XCoreISD::RETSP, dl, MVT::Other, - Chain, DAG.getConstant(0, MVT::i32), Flag); - else // Return Void - return DAG.getNode(XCoreISD::RETSP, dl, MVT::Other, - Chain, DAG.getConstant(0, MVT::i32)); + RetOps.push_back(Flag); + + return DAG.getNode(XCoreISD::RETSP, dl, MVT::Other, + &RetOps[0], RetOps.size()); } //===----------------------------------------------------------------------===// @@ -1337,7 +1305,7 @@ XCoreTargetLowering::EmitInstrWithCustomInserter(MachineInstr *MI, SDValue XCoreTargetLowering::PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const { SelectionDAG &DAG = DCI.DAG; - DebugLoc dl = N->getDebugLoc(); + SDLoc dl(N); switch (N->getOpcode()) { default: break; case XCoreISD::LADD: { @@ -1357,13 +1325,13 @@ SDValue XCoreTargetLowering::PerformDAGCombine(SDNode *N, SDValue Carry = DAG.getConstant(0, VT); SDValue Result = DAG.getNode(ISD::AND, dl, VT, N2, DAG.getConstant(1, VT)); - SDValue Ops [] = { Carry, Result }; + SDValue Ops[] = { Result, Carry }; return DAG.getMergeValues(Ops, 2, dl); } - // fold (ladd x, 0, y) -> 0, add x, y if carry is unused and y has only the + // fold (ladd x, 0, y) -> 0, add x, y iff carry is unused and y has only the // low bit set - if (N1C && N1C->isNullValue() && N->hasNUsesOfValue(0, 0)) { + if (N1C && N1C->isNullValue() && N->hasNUsesOfValue(0, 1)) { APInt KnownZero, KnownOne; APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(), VT.getSizeInBits() - 1); @@ -1371,7 +1339,7 @@ SDValue XCoreTargetLowering::PerformDAGCombine(SDNode *N, if ((KnownZero & Mask) == Mask) { SDValue Carry = DAG.getConstant(0, VT); SDValue Result = DAG.getNode(ISD::ADD, dl, VT, N0, N2); - SDValue Ops [] = { Carry, Result }; + SDValue Ops[] = { Result, Carry }; return DAG.getMergeValues(Ops, 2, dl); } } @@ -1385,7 +1353,7 @@ SDValue XCoreTargetLowering::PerformDAGCombine(SDNode *N, ConstantSDNode *N1C = dyn_cast(N1); EVT VT = N0.getValueType(); - // fold (lsub 0, 0, x) -> x, -x if x has only the low bit set + // fold (lsub 0, 0, x) -> x, -x iff x has only the low bit set if (N0C && N0C->isNullValue() && N1C && N1C->isNullValue()) { APInt KnownZero, KnownOne; APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(), @@ -1395,14 +1363,14 @@ SDValue XCoreTargetLowering::PerformDAGCombine(SDNode *N, SDValue Borrow = N2; SDValue Result = DAG.getNode(ISD::SUB, dl, VT, DAG.getConstant(0, VT), N2); - SDValue Ops [] = { Borrow, Result }; + SDValue Ops[] = { Result, Borrow }; return DAG.getMergeValues(Ops, 2, dl); } } - // fold (lsub x, 0, y) -> 0, sub x, y if borrow is unused and y has only the + // fold (lsub x, 0, y) -> 0, sub x, y iff borrow is unused and y has only the // low bit set - if (N1C && N1C->isNullValue() && N->hasNUsesOfValue(0, 0)) { + if (N1C && N1C->isNullValue() && N->hasNUsesOfValue(0, 1)) { APInt KnownZero, KnownOne; APInt Mask = APInt::getHighBitsSet(VT.getSizeInBits(), VT.getSizeInBits() - 1); @@ -1410,7 +1378,7 @@ SDValue XCoreTargetLowering::PerformDAGCombine(SDNode *N, if ((KnownZero & Mask) == Mask) { SDValue Borrow = DAG.getConstant(0, VT); SDValue Result = DAG.getNode(ISD::SUB, dl, VT, N0, N2); - SDValue Ops [] = { Borrow, Result }; + SDValue Ops[] = { Result, Borrow }; return DAG.getMergeValues(Ops, 2, dl); } } @@ -1436,11 +1404,15 @@ SDValue XCoreTargetLowering::PerformDAGCombine(SDNode *N, // If the high result is unused fold to add(a, b) if (N->hasNUsesOfValue(0, 0)) { SDValue Lo = DAG.getNode(ISD::ADD, dl, VT, N2, N3); - SDValue Ops [] = { Lo, Lo }; + SDValue Ops[] = { Lo, Lo }; return DAG.getMergeValues(Ops, 2, dl); } // Otherwise fold to ladd(a, b, 0) - return DAG.getNode(XCoreISD::LADD, dl, DAG.getVTList(VT, VT), N2, N3, N1); + SDValue Result = + DAG.getNode(XCoreISD::LADD, dl, DAG.getVTList(VT, VT), N2, N3, N1); + SDValue Carry(Result.getNode(), 1); + SDValue Ops[] = { Carry, Result }; + return DAG.getMergeValues(Ops, 2, dl); } } break; @@ -1499,7 +1471,7 @@ SDValue XCoreTargetLowering::PerformDAGCombine(SDNode *N, if (StoreBits % 8) { break; } - unsigned ABIAlignment = getTargetData()->getABITypeAlignment( + unsigned ABIAlignment = getDataLayout()->getABITypeAlignment( ST->getMemoryVT().getTypeForEVT(*DCI.DAG.getContext())); unsigned Alignment = ST->getAlignment(); if (Alignment >= ABIAlignment) { @@ -1534,7 +1506,7 @@ void XCoreTargetLowering::computeMaskedBitsForTargetNode(const SDValue Op, default: break; case XCoreISD::LADD: case XCoreISD::LSUB: - if (Op.getResNo() == 0) { + if (Op.getResNo() == 1) { // Top bits of carry / borrow are clear. KnownZero = APInt::getHighBitsSet(KnownZero.getBitWidth(), KnownZero.getBitWidth() - 1); @@ -1570,7 +1542,7 @@ XCoreTargetLowering::isLegalAddressingMode(const AddrMode &AM, if (Ty->getTypeID() == Type::VoidTyID) return AM.Scale == 0 && isImmUs(AM.BaseOffs) && isImmUs4(AM.BaseOffs); - const TargetData *TD = TM.getTargetData(); + const DataLayout *TD = TM.getDataLayout(); unsigned Size = TD->getTypeAllocSize(Ty); if (AM.BaseGV) { return Size >= 4 && !AM.HasBaseReg && AM.Scale == 0 && @@ -1610,7 +1582,7 @@ XCoreTargetLowering::isLegalAddressingMode(const AddrMode &AM, std::pair XCoreTargetLowering:: getRegForInlineAsmConstraint(const std::string &Constraint, - EVT VT) const { + MVT VT) const { if (Constraint.size() == 1) { switch (Constraint[0]) { default : break;