X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=include%2Fllvm%2FTarget%2FTargetLowering.h;h=eaf2a843e2324ff1c930b6308d2a6d1a1f7e305e;hb=f29cc18dcb043a173f211f8f3e5caabc6f2e7b08;hp=9c0189782249d8d959077e9872db6f9b3b8ceb3a;hpb=fd350586f5047c093ea5c7861ea0d575fc539b28;p=oota-llvm.git diff --git a/include/llvm/Target/TargetLowering.h b/include/llvm/Target/TargetLowering.h index 9c018978224..eaf2a843e23 100644 --- a/include/llvm/Target/TargetLowering.h +++ b/include/llvm/Target/TargetLowering.h @@ -30,9 +30,9 @@ #include "llvm/IR/Attributes.h" #include "llvm/IR/CallSite.h" #include "llvm/IR/CallingConv.h" +#include "llvm/IR/IRBuilder.h" #include "llvm/IR/InlineAsm.h" #include "llvm/IR/Instructions.h" -#include "llvm/IR/IRBuilder.h" #include "llvm/MC/MCRegisterInfo.h" #include "llvm/Target/TargetCallingConv.h" #include "llvm/Target/TargetMachine.h" @@ -51,6 +51,7 @@ namespace llvm { class MachineFunction; class MachineInstr; class MachineJumpTableInfo; + class MachineLoop; class Mangler; class MCContext; class MCExpr; @@ -76,8 +77,8 @@ namespace llvm { /// This base class for TargetLowering contains the SelectionDAG-independent /// parts that can be used from the rest of CodeGen. class TargetLoweringBase { - TargetLoweringBase(const TargetLoweringBase&) LLVM_DELETED_FUNCTION; - void operator=(const TargetLoweringBase&) LLVM_DELETED_FUNCTION; + TargetLoweringBase(const TargetLoweringBase&) = delete; + void operator=(const TargetLoweringBase&) = delete; public: /// This enum indicates whether operations are valid for a target, and if not, @@ -99,7 +100,8 @@ public: TypeExpandFloat, // Split this float into two of half the size. TypeScalarizeVector, // Replace this one-element vector with its element. TypeSplitVector, // Split this vector into two of half the size. - TypeWidenVector // This vector should be widened into a larger vector. + TypeWidenVector, // This vector should be widened into a larger vector. + TypePromoteFloat // Replace this float with a larger one. }; /// LegalizeKind holds the legalization kind that needs to happen to EVT @@ -122,6 +124,18 @@ public: // mask (ex: x86 blends). }; + /// Enum that specifies what a AtomicRMWInst is expanded to, if at all. Exists + /// because different targets have different levels of support for these + /// atomic RMW instructions, and also have different options w.r.t. what they + /// should expand to. + enum class AtomicRMWExpansionKind { + None, // Don't expand the instruction. + LLSC, // Expand the instruction into loadlinked/storeconditional; used + // by ARM/AArch64. Implies `hasLoadLinkedStoreConditional` + // returns true. + CmpXChg, // Expand the instruction into cmpxchg; used by at least X86. + }; + static ISD::NodeType getExtendForContent(BooleanContent Content) { switch (Content) { case UndefinedBooleanContent: @@ -147,20 +161,16 @@ protected: public: const TargetMachine &getTargetMachine() const { return TM; } - const DataLayout *getDataLayout() const { return DL; } - const TargetLoweringObjectFile &getObjFileLowering() const { - return *TM.getObjFileLowering(); - } + const DataLayout *getDataLayout() const { return TM.getDataLayout(); } - bool isBigEndian() const { return !IsLittleEndian; } - bool isLittleEndian() const { return IsLittleEndian; } + virtual bool useSoftFloat() const { return false; } /// Return the pointer type for the given address space, defaults to /// the pointer type from the data layout. /// FIXME: The default needs to be removed once all the code is updated. - virtual MVT getPointerTy(uint32_t /*AS*/ = 0) const; - unsigned getPointerSizeInBits(uint32_t AS = 0) const; - unsigned getPointerTypeSizeInBits(Type *Ty) const; + MVT getPointerTy(const DataLayout &DL, uint32_t AS = 0) const { + return MVT::getIntegerVT(DL.getPointerSizeInBits(AS)); + } virtual MVT getScalarShiftAmountTy(EVT LHSTy) const; EVT getShiftAmountTy(EVT LHSTy) const; @@ -168,8 +178,8 @@ public: /// Returns the type to be used for the index operand of: /// ISD::INSERT_VECTOR_ELT, ISD::EXTRACT_VECTOR_ELT, /// ISD::INSERT_SUBVECTOR, and ISD::EXTRACT_SUBVECTOR - virtual MVT getVectorIdxTy() const { - return getPointerTy(); + virtual MVT getVectorIdxTy(const DataLayout &DL) const { + return getPointerTy(DL); } /// Return true if the select operation is expensive for this target. @@ -216,6 +226,11 @@ public: /// several shifts, adds, and multiplies for this target. bool isIntDivCheap() const { return IntDivIsCheap; } + /// Return true if sqrt(x) is as cheap or cheaper than 1 / rsqrt(x) + bool isFsqrtCheap() const { + return FsqrtIsCheap; + } + /// Returns true if target has indicated at least one type should be bypassed. bool isSlowDivBypassed() const { return !BypassSlowDivWidths.empty(); } @@ -241,14 +256,34 @@ public: /// isLoadBitCastBeneficial() - Return true if the following transform /// is beneficial. /// fold (conv (load x)) -> (load (conv*)x) - /// On architectures that don't natively support some vector loads efficiently, - /// casting the load to a smaller vector of larger types and loading - /// is more efficient, however, this can be undone by optimizations in + /// On architectures that don't natively support some vector loads + /// efficiently, casting the load to a smaller vector of larger types and + /// loading is more efficient, however, this can be undone by optimizations in /// dag combiner. - virtual bool isLoadBitCastBeneficial(EVT /* Load */, EVT /* Bitcast */) const { + virtual bool isLoadBitCastBeneficial(EVT /* Load */, + EVT /* Bitcast */) const { return true; } + /// Return true if it is expected to be cheaper to do a store of a non-zero + /// vector constant with the given size and type for the address space than to + /// store the individual scalar element constants. + virtual bool storeOfVectorConstantIsCheap(EVT MemVT, + unsigned NumElem, + unsigned AddrSpace) const { + return false; + } + + /// \brief Return true if it is cheap to speculate a call to intrinsic cttz. + virtual bool isCheapToSpeculateCttz() const { + return false; + } + + /// \brief Return true if it is cheap to speculate a call to intrinsic ctlz. + virtual bool isCheapToSpeculateCtlz() const { + return false; + } + /// \brief Return if the target supports combining a /// chain like: /// \code @@ -290,7 +325,8 @@ public: } /// Return the ValueType of the result of SETCC operations. - virtual EVT getSetCCResultType(LLVMContext &Context, EVT VT) const; + virtual EVT getSetCCResultType(const DataLayout &DL, LLVMContext &Context, + EVT VT) const; /// Return the ValueType for comparison libcalls. Comparions libcalls include /// floating point comparion calls, and Ordered/Unordered check calls on @@ -546,18 +582,28 @@ public: /// Return how this load with extension should be treated: either it is legal, /// needs to be promoted to a larger size, needs to be expanded to some other /// code sequence, or the target has a custom expander for it. - LegalizeAction getLoadExtAction(unsigned ExtType, EVT VT) const { - if (VT.isExtended()) return Expand; - unsigned I = (unsigned) VT.getSimpleVT().SimpleTy; - assert(ExtType < ISD::LAST_LOADEXT_TYPE && I < MVT::LAST_VALUETYPE && - "Table isn't big enough!"); - return (LegalizeAction)LoadExtActions[I][ExtType]; + LegalizeAction getLoadExtAction(unsigned ExtType, EVT ValVT, + EVT MemVT) const { + if (ValVT.isExtended() || MemVT.isExtended()) return Expand; + unsigned ValI = (unsigned) ValVT.getSimpleVT().SimpleTy; + unsigned MemI = (unsigned) MemVT.getSimpleVT().SimpleTy; + assert(ExtType < ISD::LAST_LOADEXT_TYPE && ValI < MVT::LAST_VALUETYPE && + MemI < MVT::LAST_VALUETYPE && "Table isn't big enough!"); + return (LegalizeAction)LoadExtActions[ValI][MemI][ExtType]; } /// Return true if the specified load with extension is legal on this target. - bool isLoadExtLegal(unsigned ExtType, EVT VT) const { - return VT.isSimple() && - getLoadExtAction(ExtType, VT.getSimpleVT()) == Legal; + bool isLoadExtLegal(unsigned ExtType, EVT ValVT, EVT MemVT) const { + return ValVT.isSimple() && MemVT.isSimple() && + getLoadExtAction(ExtType, ValVT, MemVT) == Legal; + } + + /// Return true if the specified load with extension is legal or custom + /// on this target. + bool isLoadExtLegalOrCustom(unsigned ExtType, EVT ValVT, EVT MemVT) const { + return ValVT.isSimple() && MemVT.isSimple() && + (getLoadExtAction(ExtType, ValVT, MemVT) == Legal || + getLoadExtAction(ExtType, ValVT, MemVT) == Custom); } /// Return how this store with truncation should be treated: either it is @@ -584,7 +630,7 @@ public: /// sequence, or the target has a custom expander for it. LegalizeAction getIndexedLoadAction(unsigned IdxMode, MVT VT) const { - assert(IdxMode < ISD::LAST_INDEXED_MODE && VT < MVT::LAST_VALUETYPE && + assert(IdxMode < ISD::LAST_INDEXED_MODE && VT.isValid() && "Table isn't big enough!"); unsigned Ty = (unsigned)VT.SimpleTy; return (LegalizeAction)((IndexedModeActions[Ty][IdxMode] & 0xf0) >> 4); @@ -602,7 +648,7 @@ public: /// sequence, or the target has a custom expander for it. LegalizeAction getIndexedStoreAction(unsigned IdxMode, MVT VT) const { - assert(IdxMode < ISD::LAST_INDEXED_MODE && VT < MVT::LAST_VALUETYPE && + assert(IdxMode < ISD::LAST_INDEXED_MODE && VT.isValid() && "Table isn't big enough!"); unsigned Ty = (unsigned)VT.SimpleTy; return (LegalizeAction)(IndexedModeActions[Ty][IdxMode] & 0x0f); @@ -668,17 +714,18 @@ public: /// operations except for the pointer size. If AllowUnknown is true, this /// will return MVT::Other for types with no EVT counterpart (e.g. structs), /// otherwise it will assert. - EVT getValueType(Type *Ty, bool AllowUnknown = false) const { + EVT getValueType(const DataLayout &DL, Type *Ty, + bool AllowUnknown = false) const { // Lower scalar pointers to native pointer types. if (PointerType *PTy = dyn_cast(Ty)) - return getPointerTy(PTy->getAddressSpace()); + return getPointerTy(DL, PTy->getAddressSpace()); if (Ty->isVectorTy()) { VectorType *VTy = cast(Ty); Type *Elm = VTy->getElementType(); // Lower vectors of pointers to native pointer types. if (PointerType *PT = dyn_cast(Elm)) { - EVT PointerTy(getPointerTy(PT->getAddressSpace())); + EVT PointerTy(getPointerTy(DL, PT->getAddressSpace())); Elm = PointerTy.getTypeForEVT(Ty->getContext()); } @@ -689,8 +736,9 @@ public: } /// Return the MVT corresponding to this LLVM type. See getValueType. - MVT getSimpleValueType(Type *Ty, bool AllowUnknown = false) const { - return getValueType(Ty, AllowUnknown).getSimpleVT(); + MVT getSimpleValueType(const DataLayout &DL, Type *Ty, + bool AllowUnknown = false) const { + return getValueType(DL, Ty, AllowUnknown).getSimpleVT(); } /// Return the desired alignment for ByVal or InAlloca aggregate function @@ -771,8 +819,8 @@ public: /// When splitting a value of the specified type into parts, does the Lo /// or Hi part come first? This usually follows the endianness, except /// for ppcf128, where the Hi part always comes first. - bool hasBigEndianPartOrdering(EVT VT) const { - return isBigEndian() || VT == MVT::ppcf128; + bool hasBigEndianPartOrdering(EVT VT, const DataLayout &DL) const { + return DL.isBigEndian() || VT == MVT::ppcf128; } /// If true, the target has custom DAG combine transformations that it can @@ -919,7 +967,7 @@ public: } /// Return the preferred loop alignment. - unsigned getPrefLoopAlignment() const { + virtual unsigned getPrefLoopAlignment(MachineLoop *ML = nullptr) const { return PrefLoopAlignment; } @@ -937,17 +985,20 @@ public: return false; } - /// Returns the maximal possible offset which can be used for loads / stores - /// from the global. - virtual unsigned getMaximalGlobalOffset() const { - return 0; - } - /// Returns true if a cast between SrcAS and DestAS is a noop. virtual bool isNoopAddrSpaceCast(unsigned SrcAS, unsigned DestAS) const { return false; } + /// Return true if the pointer arguments to CI should be aligned by aligning + /// the object whose address is being passed. If so then MinSize is set to the + /// minimum size the object must be to be aligned and PrefAlign is set to the + /// preferred alignment. + virtual bool shouldAlignPointerArgs(CallInst * /*CI*/, unsigned & /*MinSize*/, + unsigned & /*PrefAlign*/) const { + return false; + } + //===--------------------------------------------------------------------===// /// \name Helpers for TargetTransformInfo implementations /// @{ @@ -956,7 +1007,8 @@ public: int InstructionOpcodeToISD(unsigned Opcode) const; /// Estimate the cost of type-legalization and the legalized type. - std::pair getTypeLegalizationCost(Type *Ty) const; + std::pair getTypeLegalizationCost(const DataLayout &DL, + Type *Ty) const; /// @} @@ -1014,8 +1066,9 @@ public: /// seq_cst. But if they are lowered to monotonic accesses, no amount of /// IR-level fences can prevent it. /// @{ - virtual Instruction* emitLeadingFence(IRBuilder<> &Builder, AtomicOrdering Ord, - bool IsStore, bool IsLoad) const { + virtual Instruction *emitLeadingFence(IRBuilder<> &Builder, + AtomicOrdering Ord, bool IsStore, + bool IsLoad) const { if (!getInsertFencesForAtomic()) return nullptr; @@ -1025,8 +1078,9 @@ public: return nullptr; } - virtual Instruction* emitTrailingFence(IRBuilder<> &Builder, AtomicOrdering Ord, - bool IsStore, bool IsLoad) const { + virtual Instruction *emitTrailingFence(IRBuilder<> &Builder, + AtomicOrdering Ord, bool IsStore, + bool IsLoad) const { if (!getInsertFencesForAtomic()) return nullptr; @@ -1043,15 +1097,21 @@ public: return false; } + /// Returns true if arguments should be sign-extended in lib calls. + virtual bool shouldSignExtendTypeInLibCall(EVT Type, bool IsSigned) const { + return IsSigned; + } + /// Returns true if the given (atomic) load should be expanded by the /// IR-level AtomicExpand pass into a load-linked instruction /// (through emitLoadLinked()). virtual bool shouldExpandAtomicLoadInIR(LoadInst *LI) const { return false; } - /// Returns true if the given AtomicRMW should be expanded by the - /// IR-level AtomicExpand pass into a loop using LoadLinked/StoreConditional. - virtual bool shouldExpandAtomicRMWInIR(AtomicRMWInst *RMWI) const { - return false; + /// Returns how the IR-level AtomicExpand pass should expand the given + /// AtomicRMW, if at all. Default is to never expand. + virtual AtomicRMWExpansionKind + shouldExpandAtomicRMWInIR(AtomicRMWInst *) const { + return AtomicRMWExpansionKind::None; } /// On some platforms, an AtomicRMW that never actually modifies the value @@ -1065,17 +1125,33 @@ public: /// it succeeds, and nullptr otherwise. /// If shouldExpandAtomicLoadInIR returns true on that load, it will undergo /// another round of expansion. - virtual LoadInst *lowerIdempotentRMWIntoFencedLoad(AtomicRMWInst *RMWI) const { + virtual LoadInst * + lowerIdempotentRMWIntoFencedLoad(AtomicRMWInst *RMWI) const { return nullptr; } + + /// Returns true if we should normalize + /// select(N0&N1, X, Y) => select(N0, select(N1, X, Y), Y) and + /// select(N0|N1, X, Y) => select(N0, select(N1, X, Y, Y)) if it is likely + /// that it saves us from materializing N0 and N1 in an integer register. + /// Targets that are able to perform and/or on flags should return false here. + virtual bool shouldNormalizeToSelectSequence(LLVMContext &Context, + EVT VT) const { + // If a target has multiple condition registers, then it likely has logical + // operations on those registers. + if (hasMultipleConditionRegisters()) + return false; + // Only do the transform if the value won't be split into multiple + // registers. + LegalizeTypeAction Action = getTypeAction(Context, VT); + return Action != TypeExpandInteger && Action != TypeExpandFloat && + Action != TypeSplitVector; + } + //===--------------------------------------------------------------------===// // TargetLowering Configuration Methods - These methods should be invoked by // the derived class constructor to configure this object for the target. // - - /// \brief Reset the operation actions based on target options. - virtual void resetOperationActions() {} - protected: /// Specify how the target extends the result of integer and floating point /// boolean values from i1 to a wider type. See getBooleanContents. @@ -1161,17 +1237,20 @@ protected: HasExtractBitsInsn = hasExtractInsn; } - /// Tells the code generator not to expand sequence of operations into a - /// separate sequences that increases the amount of flow control. - void setJumpIsExpensive(bool isExpensive = true) { - JumpIsExpensive = isExpensive; - } + /// Tells the code generator not to expand logic operations on comparison + /// predicates into separate sequences that increase the amount of flow + /// control. + void setJumpIsExpensive(bool isExpensive = true); /// Tells the code generator that integer divide is expensive, and if /// possible, should be replaced by an alternate sequence of instructions not /// containing an integer divide. void setIntDivIsCheap(bool isCheap = true) { IntDivIsCheap = isCheap; } - + + /// Tells the code generator that fsqrt is cheap, and should not be replaced + /// with an alternative sequence of instructions. + void setFsqrtIsCheap(bool isCheap = true) { FsqrtIsCheap = isCheap; } + /// Tells the code generator that this target supports floating point /// exceptions and cares about preserving floating point exception behavior. void setHasFloatingPointExceptions(bool FPExceptions = true) { @@ -1209,12 +1288,12 @@ protected: /// Return the largest legal super-reg register class of the register class /// for the specified type and its associated "cost". - virtual std::pair - findRepresentativeClass(MVT VT) const; + virtual std::pair + findRepresentativeClass(const TargetRegisterInfo *TRI, MVT VT) const; /// Once all of the register classes are added, this allows us to compute /// derived properties we expose. - void computeRegisterProperties(); + void computeRegisterProperties(const TargetRegisterInfo *TRI); /// Indicate that the specified operation does not work with the specified /// type and indicate what to do about it. @@ -1226,19 +1305,18 @@ protected: /// Indicate that the specified load with extension does not work with the /// specified type and indicate what to do about it. - void setLoadExtAction(unsigned ExtType, MVT VT, + void setLoadExtAction(unsigned ExtType, MVT ValVT, MVT MemVT, LegalizeAction Action) { - assert(ExtType < ISD::LAST_LOADEXT_TYPE && VT < MVT::LAST_VALUETYPE && - "Table isn't big enough!"); - LoadExtActions[VT.SimpleTy][ExtType] = (uint8_t)Action; + assert(ExtType < ISD::LAST_LOADEXT_TYPE && ValVT.isValid() && + MemVT.isValid() && "Table isn't big enough!"); + LoadExtActions[ValVT.SimpleTy][MemVT.SimpleTy][ExtType] = (uint8_t)Action; } /// Indicate that the specified truncating store does not work with the /// specified type and indicate what to do about it. void setTruncStoreAction(MVT ValVT, MVT MemVT, LegalizeAction Action) { - assert(ValVT < MVT::LAST_VALUETYPE && MemVT < MVT::LAST_VALUETYPE && - "Table isn't big enough!"); + assert(ValVT.isValid() && MemVT.isValid() && "Table isn't big enough!"); TruncStoreActions[ValVT.SimpleTy][MemVT.SimpleTy] = (uint8_t)Action; } @@ -1249,7 +1327,7 @@ protected: /// TargetLowering.cpp void setIndexedLoadAction(unsigned IdxMode, MVT VT, LegalizeAction Action) { - assert(VT < MVT::LAST_VALUETYPE && IdxMode < ISD::LAST_INDEXED_MODE && + assert(VT.isValid() && IdxMode < ISD::LAST_INDEXED_MODE && (unsigned)Action < 0xf && "Table isn't big enough!"); // Load action are kept in the upper half. IndexedModeActions[(unsigned)VT.SimpleTy][IdxMode] &= ~0xf0; @@ -1263,7 +1341,7 @@ protected: /// TargetLowering.cpp void setIndexedStoreAction(unsigned IdxMode, MVT VT, LegalizeAction Action) { - assert(VT < MVT::LAST_VALUETYPE && IdxMode < ISD::LAST_INDEXED_MODE && + assert(VT.isValid() && IdxMode < ISD::LAST_INDEXED_MODE && (unsigned)Action < 0xf && "Table isn't big enough!"); // Store action are kept in the lower half. IndexedModeActions[(unsigned)VT.SimpleTy][IdxMode] &= ~0x0f; @@ -1274,8 +1352,7 @@ protected: /// target and indicate what to do about it. void setCondCodeAction(ISD::CondCode CC, MVT VT, LegalizeAction Action) { - assert(VT < MVT::LAST_VALUETYPE && - (unsigned)CC < array_lengthof(CondCodeActions) && + assert(VT.isValid() && (unsigned)CC < array_lengthof(CondCodeActions) && "Table isn't big enough!"); /// The lower 5 bits of the SimpleTy index into Nth 2bit set from the 32-bit /// value and the upper 27 bits index into the second dimension of the array @@ -1326,7 +1403,8 @@ protected: /// Set the target's preferred loop alignment. Default alignment is zero, it /// means the target does not care about loop alignment. The alignment is - /// specified in log2(bytes). + /// specified in log2(bytes). The target may also override + /// getPrefLoopAlignment to provide per-loop values. void setPrefLoopAlignment(unsigned Align) { PrefLoopAlignment = Align; } @@ -1354,7 +1432,8 @@ public: /// load/store. virtual bool GetAddrModeArguments(IntrinsicInst * /*I*/, SmallVectorImpl &/*Ops*/, - Type *&/*AccessTy*/) const { + Type *&/*AccessTy*/, + unsigned AddrSpace = 0) const { return false; } @@ -1379,7 +1458,12 @@ public: /// The type may be VoidTy, in which case only return true if the addressing /// mode is legal for a load/store of any legal type. TODO: Handle /// pre/postinc as well. - virtual bool isLegalAddressingMode(const AddrMode &AM, Type *Ty) const; + /// + /// If the address space cannot be determined, it will be -1. + /// + /// TODO: Remove default argument + virtual bool isLegalAddressingMode(const AddrMode &AM, Type *Ty, + unsigned AddrSpace) const; /// \brief Return the cost of the scaling factor used in the addressing mode /// represented by AM for this target, for a load/store of the specified type. @@ -1387,9 +1471,12 @@ public: /// If the AM is supported, the return value must be >= 0. /// If the AM is not supported, it returns a negative value. /// TODO: Handle pre/postinc as well. - virtual int getScalingFactorCost(const AddrMode &AM, Type *Ty) const { + /// TODO: Remove default argument + virtual int getScalingFactorCost(const AddrMode &AM, Type *Ty, + unsigned AS = 0) const { // Default: assume that any scaling factor used in a legal AM is free. - if (isLegalAddressingMode(AM, Ty)) return 0; + if (isLegalAddressingMode(AM, Ty, AS)) + return 0; return -1; } @@ -1435,6 +1522,35 @@ public: return false; } + virtual bool isProfitableToHoist(Instruction *I) const { return true; } + + /// Return true if the extension represented by \p I is free. + /// Unlikely the is[Z|FP]ExtFree family which is based on types, + /// this method can use the context provided by \p I to decide + /// whether or not \p I is free. + /// This method extends the behavior of the is[Z|FP]ExtFree family. + /// In other words, if is[Z|FP]Free returns true, then this method + /// returns true as well. The converse is not true. + /// The target can perform the adequate checks by overriding isExtFreeImpl. + /// \pre \p I must be a sign, zero, or fp extension. + bool isExtFree(const Instruction *I) const { + switch (I->getOpcode()) { + case Instruction::FPExt: + if (isFPExtFree(EVT::getEVT(I->getType()))) + return true; + break; + case Instruction::ZExt: + if (isZExtFree(I->getOperand(0)->getType(), I->getType())) + return true; + break; + case Instruction::SExt: + break; + default: + llvm_unreachable("Instruction is not an extension"); + } + return isExtFreeImpl(I); + } + /// Return true if any actual instruction that defines a value of type Ty1 /// implicitly zero-extends the value to Ty2 in the result register. /// @@ -1482,6 +1598,35 @@ public: return false; } + /// \brief Get the maximum supported factor for interleaved memory accesses. + /// Default to be the minimum interleave factor: 2. + virtual unsigned getMaxSupportedInterleaveFactor() const { return 2; } + + /// \brief Lower an interleaved load to target specific intrinsics. Return + /// true on success. + /// + /// \p LI is the vector load instruction. + /// \p Shuffles is the shufflevector list to DE-interleave the loaded vector. + /// \p Indices is the corresponding indices for each shufflevector. + /// \p Factor is the interleave factor. + virtual bool lowerInterleavedLoad(LoadInst *LI, + ArrayRef Shuffles, + ArrayRef Indices, + unsigned Factor) const { + return false; + } + + /// \brief Lower an interleaved store to target specific intrinsics. Return + /// true on success. + /// + /// \p SI is the vector store instruction. + /// \p SVI is the shufflevector to RE-interleave the stored vector. + /// \p Factor is the interleave factor. + virtual bool lowerInterleavedStore(StoreInst *SI, ShuffleVectorInst *SVI, + unsigned Factor) const { + return false; + } + /// Return true if zero-extending the specific node Val to type VT2 is free /// (either because it's implicitly zero-extended such as ARM ldrb / ldrh or /// because it's folded such as X86 zero-extending loads). @@ -1489,6 +1634,18 @@ public: return isZExtFree(Val.getValueType(), VT2); } + /// Return true if an fpext operation is free (for instance, because + /// single-precision floating-point numbers are implicitly extended to + /// double-precision). + virtual bool isFPExtFree(EVT VT) const { + assert(VT.isFloatingPoint()); + return false; + } + + /// Return true if folding a vector load into ExtVal (a sign, zero, or any + /// extend node) is profitable. + virtual bool isVectorLoadExtDesirable(SDValue ExtVal) const { return false; } + /// Return true if an fneg operation is free to the point where it is never /// worthwhile to replace it with a bitwise operation. virtual bool isFNegFree(EVT VT) const { @@ -1578,10 +1735,6 @@ public: private: const TargetMachine &TM; - const DataLayout *DL; - - /// True if this is a little endian target. - bool IsLittleEndian; /// Tells the code generator not to expand operations into sequences that use /// the select operations if possible. @@ -1606,6 +1759,9 @@ private: /// unconditionally. bool IntDivIsCheap; + // Don't expand fsqrt with an approximation based on the inverse sqrt. + bool FsqrtIsCheap; + /// Tells the code generator to bypass slow divide or remainder /// instructions. For example, BypassSlowDivWidths[32,8] tells the code /// generator to bypass 32-bit integer div/rem with an 8-bit unsigned integer @@ -1727,7 +1883,8 @@ private: /// For each load extension type and each value type, keep a LegalizeAction /// that indicates how instruction selection should deal with a load of a /// specific value type and extension type. - uint8_t LoadExtActions[MVT::LAST_VALUETYPE][ISD::LAST_LOADEXT_TYPE]; + uint8_t LoadExtActions[MVT::LAST_VALUETYPE][MVT::LAST_VALUETYPE] + [ISD::LAST_LOADEXT_TYPE]; /// For each value type pair keep a LegalizeAction that indicates whether a /// truncating store of a specific value type and truncating type is legal. @@ -1751,136 +1908,8 @@ private: ValueTypeActionImpl ValueTypeActions; -public: - LegalizeKind - getTypeConversion(LLVMContext &Context, EVT VT) const { - // If this is a simple type, use the ComputeRegisterProp mechanism. - if (VT.isSimple()) { - MVT SVT = VT.getSimpleVT(); - assert((unsigned)SVT.SimpleTy < array_lengthof(TransformToType)); - MVT NVT = TransformToType[SVT.SimpleTy]; - LegalizeTypeAction LA = ValueTypeActions.getTypeAction(SVT); - - assert( - (LA == TypeLegal || LA == TypeSoftenFloat || - ValueTypeActions.getTypeAction(NVT) != TypePromoteInteger) - && "Promote may not follow Expand or Promote"); - - if (LA == TypeSplitVector) - return LegalizeKind(LA, EVT::getVectorVT(Context, - SVT.getVectorElementType(), - SVT.getVectorNumElements()/2)); - if (LA == TypeScalarizeVector) - return LegalizeKind(LA, SVT.getVectorElementType()); - return LegalizeKind(LA, NVT); - } - - // Handle Extended Scalar Types. - if (!VT.isVector()) { - assert(VT.isInteger() && "Float types must be simple"); - unsigned BitSize = VT.getSizeInBits(); - // First promote to a power-of-two size, then expand if necessary. - if (BitSize < 8 || !isPowerOf2_32(BitSize)) { - EVT NVT = VT.getRoundIntegerType(Context); - assert(NVT != VT && "Unable to round integer VT"); - LegalizeKind NextStep = getTypeConversion(Context, NVT); - // Avoid multi-step promotion. - if (NextStep.first == TypePromoteInteger) return NextStep; - // Return rounded integer type. - return LegalizeKind(TypePromoteInteger, NVT); - } - - return LegalizeKind(TypeExpandInteger, - EVT::getIntegerVT(Context, VT.getSizeInBits()/2)); - } - - // Handle vector types. - unsigned NumElts = VT.getVectorNumElements(); - EVT EltVT = VT.getVectorElementType(); - - // Vectors with only one element are always scalarized. - if (NumElts == 1) - return LegalizeKind(TypeScalarizeVector, EltVT); - - // Try to widen vector elements until the element type is a power of two and - // promote it to a legal type later on, for example: - // <3 x i8> -> <4 x i8> -> <4 x i32> - if (EltVT.isInteger()) { - // Vectors with a number of elements that is not a power of two are always - // widened, for example <3 x i8> -> <4 x i8>. - if (!VT.isPow2VectorType()) { - NumElts = (unsigned)NextPowerOf2(NumElts); - EVT NVT = EVT::getVectorVT(Context, EltVT, NumElts); - return LegalizeKind(TypeWidenVector, NVT); - } - - // Examine the element type. - LegalizeKind LK = getTypeConversion(Context, EltVT); - - // If type is to be expanded, split the vector. - // <4 x i140> -> <2 x i140> - if (LK.first == TypeExpandInteger) - return LegalizeKind(TypeSplitVector, - EVT::getVectorVT(Context, EltVT, NumElts / 2)); - - // Promote the integer element types until a legal vector type is found - // or until the element integer type is too big. If a legal type was not - // found, fallback to the usual mechanism of widening/splitting the - // vector. - EVT OldEltVT = EltVT; - while (1) { - // Increase the bitwidth of the element to the next pow-of-two - // (which is greater than 8 bits). - EltVT = EVT::getIntegerVT(Context, 1 + EltVT.getSizeInBits() - ).getRoundIntegerType(Context); - - // Stop trying when getting a non-simple element type. - // Note that vector elements may be greater than legal vector element - // types. Example: X86 XMM registers hold 64bit element on 32bit - // systems. - if (!EltVT.isSimple()) break; - - // Build a new vector type and check if it is legal. - MVT NVT = MVT::getVectorVT(EltVT.getSimpleVT(), NumElts); - // Found a legal promoted vector type. - if (NVT != MVT() && ValueTypeActions.getTypeAction(NVT) == TypeLegal) - return LegalizeKind(TypePromoteInteger, - EVT::getVectorVT(Context, EltVT, NumElts)); - } - - // Reset the type to the unexpanded type if we did not find a legal vector - // type with a promoted vector element type. - EltVT = OldEltVT; - } - - // Try to widen the vector until a legal type is found. - // If there is no wider legal type, split the vector. - while (1) { - // Round up to the next power of 2. - NumElts = (unsigned)NextPowerOf2(NumElts); - - // If there is no simple vector type with this many elements then there - // cannot be a larger legal vector type. Note that this assumes that - // there are no skipped intermediate vector types in the simple types. - if (!EltVT.isSimple()) break; - MVT LargerVector = MVT::getVectorVT(EltVT.getSimpleVT(), NumElts); - if (LargerVector == MVT()) break; - - // If this type is legal then widen the vector. - if (ValueTypeActions.getTypeAction(LargerVector) == TypeLegal) - return LegalizeKind(TypeWidenVector, LargerVector); - } - - // Widen odd vectors to next power of two. - if (!VT.isPow2VectorType()) { - EVT NVT = VT.getPow2VectorType(Context); - return LegalizeKind(TypeWidenVector, NVT); - } - - // Vectors with illegal element types are expanded. - EVT NVT = EVT::getVectorVT(Context, EltVT, VT.getVectorNumElements() / 2); - return LegalizeKind(TypeSplitVector, NVT); - } +private: + LegalizeKind getTypeConversion(LLVMContext &Context, EVT VT) const; private: std::vector > AvailableRegClasses; @@ -1911,6 +1940,11 @@ private: CallingConv::ID LibcallCallingConvs[RTLIB::UNKNOWN_LIBCALL]; protected: + /// Return true if the extension represented by \p I is free. + /// \pre \p I is a sign, zero, or fp extension and + /// is[Z|FP]ExtFree of the related types is not true. + virtual bool isExtFreeImpl(const Instruction *I) const { return false; } + /// \brief Specify maximum number of store instructions per memset call. /// /// When lowering \@llvm.memset this field specifies the maximum number of @@ -1978,7 +2012,8 @@ protected: /// Replace/modify any TargetFrameIndex operands with a targte-dependent /// sequence of memory operands that is recognized by PrologEpilogInserter. - MachineBasicBlock *emitPatchPoint(MachineInstr *MI, MachineBasicBlock *MBB) const; + MachineBasicBlock *emitPatchPoint(MachineInstr *MI, + MachineBasicBlock *MBB) const; }; /// This class defines information used to lower LLVM code to legal SelectionDAG @@ -1987,8 +2022,8 @@ protected: /// This class also defines callbacks that targets must implement to lower /// target-specific constructs to SelectionDAG operators. class TargetLowering : public TargetLoweringBase { - TargetLowering(const TargetLowering&) LLVM_DELETED_FUNCTION; - void operator=(const TargetLowering&) LLVM_DELETED_FUNCTION; + TargetLowering(const TargetLowering&) = delete; + void operator=(const TargetLowering&) = delete; public: /// NOTE: The TargetMachine owns TLOF. @@ -2048,6 +2083,7 @@ public: ISD::CondCode &CCCode, SDLoc DL) const; /// Returns a pair of (return value, chain). + /// It is an error to pass RTLIB::UNKNOWN_LIBCALL as \p LC. std::pair makeLibCall(SelectionDAG &DAG, RTLIB::Libcall LC, EVT RetVT, const SDValue *Ops, unsigned NumOps, bool isSigned, @@ -2139,8 +2175,7 @@ public: void AddToWorklist(SDNode *N); void RemoveFromWorklist(SDNode *N); - SDValue CombineTo(SDNode *N, const std::vector &To, - bool AddTo = true); + SDValue CombineTo(SDNode *N, ArrayRef To, bool AddTo = true); SDValue CombineTo(SDNode *N, SDValue Res, bool AddTo = true); SDValue CombineTo(SDNode *N, SDValue Res0, SDValue Res1, bool AddTo = true); @@ -2279,6 +2314,7 @@ public: SelectionDAG &DAG; SDLoc DL; ImmutableCallSite *CS; + bool IsPatchPoint; SmallVector Outs; SmallVector OutVals; SmallVector Ins; @@ -2287,7 +2323,7 @@ public: : RetTy(nullptr), RetSExt(false), RetZExt(false), IsVarArg(false), IsInReg(false), DoesNotReturn(false), IsReturnValueUsed(true), IsTailCall(false), NumFixedArgs(-1), CallConv(CallingConv::C), - DAG(DAG), CS(nullptr) {} + DAG(DAG), CS(nullptr), IsPatchPoint(false) {} CallLoweringInfo &setDebugLoc(SDLoc dl) { DL = dl; @@ -2369,9 +2405,15 @@ public: return *this; } + CallLoweringInfo &setIsPatchPoint(bool Value = true) { + IsPatchPoint = Value; + return *this; + } + ArgListTy &getArgs() { return Args; } + }; /// This function lowers an abstract call to a function into an actual call. @@ -2380,7 +2422,7 @@ public: /// outgoing token chain. It calls LowerCall to do the actual lowering. std::pair LowerCallTo(CallLoweringInfo &CLI) const; - /// This hook must be implemented to lower calls into the the specified + /// This hook must be implemented to lower calls into the specified /// DAG. The outgoing arguments to the call are described by the Outs array, /// and the values to be returned by the call are described by the Ins /// array. The implementation should fill in the InVals array with legal-type @@ -2615,7 +2657,9 @@ public: /// specific constraints and their prefixes, and also tie in the associated /// operand values. If this returns an empty vector, and if the constraint /// string itself isn't empty, there was an error parsing. - virtual AsmOperandInfoVector ParseConstraints(ImmutableCallSite CS) const; + virtual AsmOperandInfoVector ParseConstraints(const DataLayout &DL, + const TargetRegisterInfo *TRI, + ImmutableCallSite CS) const; /// Examine constraint type and operand type and determine a weight value. /// The operand object must already have been set up with the operand type. @@ -2636,7 +2680,7 @@ public: SelectionDAG *DAG = nullptr) const; /// Given a constraint, return the type of constraint it is for this target. - virtual ConstraintType getConstraintType(const std::string &Constraint) const; + virtual ConstraintType getConstraintType(StringRef Constraint) const; /// Given a physical register constraint (e.g. {edx}), return the register /// number and the register class for the register. @@ -2646,10 +2690,18 @@ public: /// pointer. /// /// This should only be used for C_Register constraints. On error, this - /// returns a register number of 0 and a null register class pointer.. - virtual std::pair - getRegForInlineAsmConstraint(const std::string &Constraint, - MVT VT) const; + /// returns a register number of 0 and a null register class pointer. + virtual std::pair + getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI, + StringRef Constraint, MVT VT) const; + + virtual unsigned getInlineAsmMemConstraint(StringRef ConstraintCode) const { + if (ConstraintCode == "i") + return InlineAsm::Constraint_i; + else if (ConstraintCode == "m") + return InlineAsm::Constraint_m; + return InlineAsm::Constraint_Unknown; + } /// Try to replace an X constraint, which matches anything, with another that /// has more specific requirements based on the type of the corresponding @@ -2665,8 +2717,6 @@ public: //===--------------------------------------------------------------------===// // Div utility functions // - SDValue BuildExactSDIV(SDValue Op1, SDValue Op2, SDLoc dl, - SelectionDAG &DAG) const; SDValue BuildSDIV(SDNode *N, const APInt &Divisor, SelectionDAG &DAG, bool IsAfterLegalization, std::vector *Created) const; @@ -2687,7 +2737,7 @@ public: /// Hooks for building estimates in place of slower divisions and square /// roots. - + /// Return a reciprocal square root estimate value for the input operand. /// The RefinementSteps output is the number of Newton-Raphson refinement /// iterations required to generate a sufficient (though not necessarily @@ -2698,10 +2748,9 @@ public: /// If that's true, then return '0' as the number of RefinementSteps to avoid /// any further refinement of the estimate. /// An empty SDValue return means no estimate sequence can be created. - virtual SDValue getRsqrtEstimate(SDValue Operand, - DAGCombinerInfo &DCI, - unsigned &RefinementSteps, - bool &UseOneConstNR) const { + virtual SDValue getRsqrtEstimate(SDValue Operand, DAGCombinerInfo &DCI, + unsigned &RefinementSteps, + bool &UseOneConstNR) const { return SDValue(); } @@ -2713,8 +2762,7 @@ public: /// If that's true, then return '0' as the number of RefinementSteps to avoid /// any further refinement of the estimate. /// An empty SDValue return means no estimate sequence can be created. - virtual SDValue getRecipEstimate(SDValue Operand, - DAGCombinerInfo &DCI, + virtual SDValue getRecipEstimate(SDValue Operand, DAGCombinerInfo &DCI, unsigned &RefinementSteps) const { return SDValue(); } @@ -2753,6 +2801,8 @@ public: /// is created but not inserted into any basic blocks, and this method is /// called to expand it into a sequence of instructions, potentially also /// creating new basic blocks and control flow. + /// As long as the returned basic block is different (i.e., we created a new + /// one), the custom inserter is free to modify the rest of \p MBB. virtual MachineBasicBlock * EmitInstrWithCustomInserter(MachineInstr *MI, MachineBasicBlock *MBB) const; @@ -2773,9 +2823,9 @@ public: /// Given an LLVM IR type and return type attributes, compute the return value /// EVTs and flags, and optionally also the offsets, if the return value is /// being lowered to memory. -void GetReturnInfo(Type* ReturnType, AttributeSet attr, +void GetReturnInfo(Type *ReturnType, AttributeSet attr, SmallVectorImpl &Outs, - const TargetLowering &TLI); + const TargetLowering &TLI, const DataLayout &DL); } // end llvm namespace