-//===- ARMBaseInstrInfo.h - ARM Base Instruction Information ----*- C++ -*-===//
+//===-- ARMBaseInstrInfo.h - ARM Base Instruction Information ---*- C++ -*-===//
//
// The LLVM Compiler Infrastructure
//
#define ARMBASEINSTRUCTIONINFO_H
#include "ARM.h"
+#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/SmallSet.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/Target/TargetInstrInfo.h"
+#define GET_INSTRINFO_HEADER
+#include "ARMGenInstrInfo.inc"
+
namespace llvm {
class ARMSubtarget;
class ARMBaseRegisterInfo;
-/// ARMII - This namespace holds all of the target specific flags that
-/// instruction info tracks.
-///
-namespace ARMII {
- enum {
- //===------------------------------------------------------------------===//
- // Instruction Flags.
-
- //===------------------------------------------------------------------===//
- // This four-bit field describes the addressing mode used.
-
- AddrModeMask = 0x1f,
- AddrModeNone = 0,
- AddrMode1 = 1,
- AddrMode2 = 2,
- AddrMode3 = 3,
- AddrMode4 = 4,
- AddrMode5 = 5,
- AddrMode6 = 6,
- AddrModeT1_1 = 7,
- AddrModeT1_2 = 8,
- AddrModeT1_4 = 9,
- AddrModeT1_s = 10, // i8 * 4 for pc and sp relative data
- AddrModeT2_i12 = 11,
- AddrModeT2_i8 = 12,
- AddrModeT2_so = 13,
- AddrModeT2_pc = 14, // +/- i12 for pc relative data
- AddrModeT2_i8s4 = 15, // i8 * 4
-
- // Size* - Flags to keep track of the size of an instruction.
- SizeShift = 5,
- SizeMask = 7 << SizeShift,
- SizeSpecial = 1, // 0 byte pseudo or special case.
- Size8Bytes = 2,
- Size4Bytes = 3,
- Size2Bytes = 4,
-
- // IndexMode - Unindex, pre-indexed, or post-indexed are valid for load
- // and store ops only. Generic "updating" flag is used for ld/st multiple.
- IndexModeShift = 8,
- IndexModeMask = 3 << IndexModeShift,
- IndexModePre = 1,
- IndexModePost = 2,
- IndexModeUpd = 3,
-
- //===------------------------------------------------------------------===//
- // Instruction encoding formats.
- //
- FormShift = 10,
- FormMask = 0x3f << FormShift,
-
- // Pseudo instructions
- Pseudo = 0 << FormShift,
-
- // Multiply instructions
- MulFrm = 1 << FormShift,
-
- // Branch instructions
- BrFrm = 2 << FormShift,
- BrMiscFrm = 3 << FormShift,
-
- // Data Processing instructions
- DPFrm = 4 << FormShift,
- DPSoRegFrm = 5 << FormShift,
-
- // Load and Store
- LdFrm = 6 << FormShift,
- StFrm = 7 << FormShift,
- LdMiscFrm = 8 << FormShift,
- StMiscFrm = 9 << FormShift,
- LdStMulFrm = 10 << FormShift,
-
- LdStExFrm = 11 << FormShift,
-
- // Miscellaneous arithmetic instructions
- ArithMiscFrm = 12 << FormShift,
- SatFrm = 13 << FormShift,
-
- // Extend instructions
- ExtFrm = 14 << FormShift,
-
- // VFP formats
- VFPUnaryFrm = 15 << FormShift,
- VFPBinaryFrm = 16 << FormShift,
- VFPConv1Frm = 17 << FormShift,
- VFPConv2Frm = 18 << FormShift,
- VFPConv3Frm = 19 << FormShift,
- VFPConv4Frm = 20 << FormShift,
- VFPConv5Frm = 21 << FormShift,
- VFPLdStFrm = 22 << FormShift,
- VFPLdStMulFrm = 23 << FormShift,
- VFPMiscFrm = 24 << FormShift,
-
- // Thumb format
- ThumbFrm = 25 << FormShift,
-
- // Miscelleaneous format
- MiscFrm = 26 << FormShift,
-
- // NEON formats
- NGetLnFrm = 27 << FormShift,
- NSetLnFrm = 28 << FormShift,
- NDupFrm = 29 << FormShift,
- NLdStFrm = 30 << FormShift,
- N1RegModImmFrm= 31 << FormShift,
- N2RegFrm = 32 << FormShift,
- NVCVTFrm = 33 << FormShift,
- NVDupLnFrm = 34 << FormShift,
- N2RegVShLFrm = 35 << FormShift,
- N2RegVShRFrm = 36 << FormShift,
- N3RegFrm = 37 << FormShift,
- N3RegVShFrm = 38 << FormShift,
- NVExtFrm = 39 << FormShift,
- NVMulSLFrm = 40 << FormShift,
- NVTBLFrm = 41 << FormShift,
-
- //===------------------------------------------------------------------===//
- // Misc flags.
-
- // UnaryDP - Indicates this is a unary data processing instruction, i.e.
- // it doesn't have a Rn operand.
- UnaryDP = 1 << 16,
-
- // Xform16Bit - Indicates this Thumb2 instruction may be transformed into
- // a 16-bit Thumb instruction if certain conditions are met.
- Xform16Bit = 1 << 17,
-
- //===------------------------------------------------------------------===//
- // Code domain.
- DomainShift = 18,
- DomainMask = 3 << DomainShift,
- DomainGeneral = 0 << DomainShift,
- DomainVFP = 1 << DomainShift,
- DomainNEON = 2 << DomainShift,
-
- //===------------------------------------------------------------------===//
- // Field shifts - such shifts are used to set field while generating
- // machine instructions.
- //
- // FIXME: This list will need adjusting/fixing as the MC code emitter
- // takes shape and the ARMCodeEmitter.cpp bits go away.
- ShiftTypeShift = 4,
-
- M_BitShift = 5,
- ShiftImmShift = 5,
- ShiftShift = 7,
- N_BitShift = 7,
- ImmHiShift = 8,
- SoRotImmShift = 8,
- RegRsShift = 8,
- ExtRotImmShift = 10,
- RegRdLoShift = 12,
- RegRdShift = 12,
- RegRdHiShift = 16,
- RegRnShift = 16,
- S_BitShift = 20,
- W_BitShift = 21,
- AM3_I_BitShift = 22,
- D_BitShift = 22,
- U_BitShift = 23,
- P_BitShift = 24,
- I_BitShift = 25,
- CondShift = 28
- };
-}
-
-class ARMBaseInstrInfo : public TargetInstrInfoImpl {
+class ARMBaseInstrInfo : public ARMGenInstrInfo {
const ARMSubtarget &Subtarget;
+
protected:
// Can be only subclassed.
explicit ARMBaseInstrInfo(const ARMSubtarget &STI);
+
public:
+ // Return whether the target has an explicit NOP encoding.
+ bool hasNOP() const;
+
// Return the non-pre/post incrementing version of 'Opc'. Return 0
// if there is not such an opcode.
virtual unsigned getUnindexedOpcode(unsigned Opc) const =0;
virtual const ARMBaseRegisterInfo &getRegisterInfo() const =0;
const ARMSubtarget &getSubtarget() const { return Subtarget; }
- bool spillCalleeSavedRegisters(MachineBasicBlock &MBB,
- MachineBasicBlock::iterator MI,
- const std::vector<CalleeSavedInfo> &CSI,
- const TargetRegisterInfo *TRI) const;
+ ScheduleHazardRecognizer *
+ CreateTargetHazardRecognizer(const TargetMachine *TM,
+ const ScheduleDAG *DAG) const;
+
+ ScheduleHazardRecognizer *
+ CreateTargetPostRAHazardRecognizer(const InstrItineraryData *II,
+ const ScheduleDAG *DAG) const;
// Branch analysis.
virtual bool AnalyzeBranch(MachineBasicBlock &MBB, MachineBasicBlock *&TBB,
bool ReverseBranchCondition(SmallVectorImpl<MachineOperand> &Cond) const;
// Predication support.
- bool isPredicated(const MachineInstr *MI) const {
- int PIdx = MI->findFirstPredOperandIdx();
- return PIdx != -1 && MI->getOperand(PIdx).getImm() != ARMCC::AL;
- }
+ bool isPredicated(const MachineInstr *MI) const;
ARMCC::CondCodes getPredicate(const MachineInstr *MI) const {
int PIdx = MI->findFirstPredOperandIdx();
int &FrameIndex) const;
virtual unsigned isStoreToStackSlot(const MachineInstr *MI,
int &FrameIndex) const;
+ virtual unsigned isLoadFromStackSlotPostFE(const MachineInstr *MI,
+ int &FrameIndex) const;
+ virtual unsigned isStoreToStackSlotPostFE(const MachineInstr *MI,
+ int &FrameIndex) const;
virtual void copyPhysReg(MachineBasicBlock &MBB,
MachineBasicBlock::iterator I, DebugLoc DL,
const TargetRegisterClass *RC,
const TargetRegisterInfo *TRI) const;
+ virtual bool expandPostRAPseudo(MachineBasicBlock::iterator MI) const;
+
virtual MachineInstr *emitFrameIndexDebugValue(MachineFunction &MF,
int FrameIx,
uint64_t Offset,
MachineInstr *duplicate(MachineInstr *Orig, MachineFunction &MF) const;
+ MachineInstr *commuteInstruction(MachineInstr*, bool=false) const;
+
virtual bool produceSameValue(const MachineInstr *MI0,
- const MachineInstr *MI1) const;
+ const MachineInstr *MI1,
+ const MachineRegisterInfo *MRI) const;
/// areLoadsFromSameBasePtr - This is used by the pre-regalloc scheduler to
/// determine if two loads are loading from the same base address. It should
int64_t &Offset1, int64_t &Offset2)const;
/// shouldScheduleLoadsNear - This is a used by the pre-regalloc scheduler to
- /// determine (in conjuction with areLoadsFromSameBasePtr) if two loads should
- /// be scheduled togther. On some targets if two loads are loading from
+ /// determine (in conjunction with areLoadsFromSameBasePtr) if two loads
+ /// should be scheduled togther. On some targets if two loads are loading from
/// addresses in the same cache line, it's better if they are scheduled
/// together. This function takes two integers that represent the load offsets
/// from the common base address. It returns true if it decides it's desirable
const MachineFunction &MF) const;
virtual bool isProfitableToIfCvt(MachineBasicBlock &MBB,
- unsigned NumInstrs,
- float Prob, float Confidence) const;
+ unsigned NumCycles, unsigned ExtraPredCycles,
+ const BranchProbability &Probability) const;
- virtual bool isProfitableToIfCvt(MachineBasicBlock &TMBB,unsigned NumT,
- MachineBasicBlock &FMBB,unsigned NumF,
- float Probability, float Confidence) const;
+ virtual bool isProfitableToIfCvt(MachineBasicBlock &TMBB,
+ unsigned NumT, unsigned ExtraT,
+ MachineBasicBlock &FMBB,
+ unsigned NumF, unsigned ExtraF,
+ const BranchProbability &Probability) const;
virtual bool isProfitableToDupForIfCvt(MachineBasicBlock &MBB,
- unsigned NumInstrs,
- float Probability,
- float Confidence) const {
- return NumInstrs && NumInstrs == 1;
+ unsigned NumCycles,
+ const BranchProbability
+ &Probability) const {
+ return NumCycles == 1;
}
- /// AnalyzeCompare - For a comparison instruction, return the source register
- /// in SrcReg and the value it compares against in CmpValue. Return true if
- /// the comparison instruction can be analyzed.
- virtual bool AnalyzeCompare(const MachineInstr *MI, unsigned &SrcReg,
- int &CmpMask, int &CmpValue) const;
+ virtual bool isProfitableToUnpredicate(MachineBasicBlock &TMBB,
+ MachineBasicBlock &FMBB) const;
+
+ /// analyzeCompare - For a comparison instruction, return the source registers
+ /// in SrcReg and SrcReg2 if having two register operands, and the value it
+ /// compares against in CmpValue. Return true if the comparison instruction
+ /// can be analyzed.
+ virtual bool analyzeCompare(const MachineInstr *MI, unsigned &SrcReg,
+ unsigned &SrcReg2, int &CmpMask,
+ int &CmpValue) const;
+
+ /// optimizeCompareInstr - Convert the instruction to set the zero flag so
+ /// that we can remove a "comparison with zero"; Remove a redundant CMP
+ /// instruction if the flags can be updated in the same way by an earlier
+ /// instruction such as SUB.
+ virtual bool optimizeCompareInstr(MachineInstr *CmpInstr, unsigned SrcReg,
+ unsigned SrcReg2, int CmpMask, int CmpValue,
+ const MachineRegisterInfo *MRI) const;
+
+ virtual bool analyzeSelect(const MachineInstr *MI,
+ SmallVectorImpl<MachineOperand> &Cond,
+ unsigned &TrueOp, unsigned &FalseOp,
+ bool &Optimizable) const;
+
+ virtual MachineInstr *optimizeSelect(MachineInstr *MI, bool) const;
- /// OptimizeCompareInstr - Convert the instruction to set the zero flag so
- /// that we can remove a "comparison with zero".
- virtual bool OptimizeCompareInstr(MachineInstr *CmpInstr, unsigned SrcReg,
- int CmpMask, int CmpValue,
- MachineBasicBlock::iterator &MII) const;
+ /// FoldImmediate - 'Reg' is known to be defined by a move immediate
+ /// instruction, try to fold the immediate into the use instruction.
+ virtual bool FoldImmediate(MachineInstr *UseMI, MachineInstr *DefMI,
+ unsigned Reg, MachineRegisterInfo *MRI) const;
- virtual unsigned getNumMicroOps(const MachineInstr *MI,
- const InstrItineraryData *ItinData) const;
+ virtual unsigned getNumMicroOps(const InstrItineraryData *ItinData,
+ const MachineInstr *MI) const;
virtual
int getOperandLatency(const InstrItineraryData *ItinData,
int getOperandLatency(const InstrItineraryData *ItinData,
SDNode *DefNode, unsigned DefIdx,
SDNode *UseNode, unsigned UseIdx) const;
+
+ /// VFP/NEON execution domains.
+ std::pair<uint16_t, uint16_t>
+ getExecutionDomain(const MachineInstr *MI) const;
+ void setExecutionDomain(MachineInstr *MI, unsigned Domain) const;
+
+ unsigned getPartialRegUpdateClearance(const MachineInstr*, unsigned,
+ const TargetRegisterInfo*) const;
+ void breakPartialRegDependency(MachineBasicBlock::iterator, unsigned,
+ const TargetRegisterInfo *TRI) const;
+ /// Get the number of addresses by LDM or VLDM or zero for unknown.
+ unsigned getNumLDMAddresses(const MachineInstr *MI) const;
+
private:
+ unsigned getInstBundleLength(const MachineInstr *MI) const;
+
int getVLDMDefCycle(const InstrItineraryData *ItinData,
- const TargetInstrDesc &DefTID,
+ const MCInstrDesc &DefMCID,
unsigned DefClass,
unsigned DefIdx, unsigned DefAlign) const;
int getLDMDefCycle(const InstrItineraryData *ItinData,
- const TargetInstrDesc &DefTID,
+ const MCInstrDesc &DefMCID,
unsigned DefClass,
unsigned DefIdx, unsigned DefAlign) const;
int getVSTMUseCycle(const InstrItineraryData *ItinData,
- const TargetInstrDesc &UseTID,
+ const MCInstrDesc &UseMCID,
unsigned UseClass,
unsigned UseIdx, unsigned UseAlign) const;
int getSTMUseCycle(const InstrItineraryData *ItinData,
- const TargetInstrDesc &UseTID,
+ const MCInstrDesc &UseMCID,
unsigned UseClass,
unsigned UseIdx, unsigned UseAlign) const;
int getOperandLatency(const InstrItineraryData *ItinData,
- const TargetInstrDesc &DefTID,
+ const MCInstrDesc &DefMCID,
unsigned DefIdx, unsigned DefAlign,
- const TargetInstrDesc &UseTID,
+ const MCInstrDesc &UseMCID,
unsigned UseIdx, unsigned UseAlign) const;
+
+ unsigned getInstrLatency(const InstrItineraryData *ItinData,
+ const MachineInstr *MI,
+ unsigned *PredCost = 0) const;
+
+ int getInstrLatency(const InstrItineraryData *ItinData,
+ SDNode *Node) const;
+
+ bool hasHighOperandLatency(const InstrItineraryData *ItinData,
+ const MachineRegisterInfo *MRI,
+ const MachineInstr *DefMI, unsigned DefIdx,
+ const MachineInstr *UseMI, unsigned UseIdx) const;
+ bool hasLowDefLatency(const InstrItineraryData *ItinData,
+ const MachineInstr *DefMI, unsigned DefIdx) const;
+
+ /// verifyInstruction - Perform target specific instruction verification.
+ bool verifyInstruction(const MachineInstr *MI, StringRef &ErrInfo) const;
+
+private:
+ /// Modeling special VFP / NEON fp MLA / MLS hazards.
+
+ /// MLxEntryMap - Map fp MLA / MLS to the corresponding entry in the internal
+ /// MLx table.
+ DenseMap<unsigned, unsigned> MLxEntryMap;
+
+ /// MLxHazardOpcodes - Set of add / sub and multiply opcodes that would cause
+ /// stalls when scheduled together with fp MLA / MLS opcodes.
+ SmallSet<unsigned, 16> MLxHazardOpcodes;
+
+public:
+ /// isFpMLxInstruction - Return true if the specified opcode is a fp MLA / MLS
+ /// instruction.
+ bool isFpMLxInstruction(unsigned Opcode) const {
+ return MLxEntryMap.count(Opcode);
+ }
+
+ /// isFpMLxInstruction - This version also returns the multiply opcode and the
+ /// addition / subtraction opcode to expand to. Return true for 'HasLane' for
+ /// the MLX instructions with an extra lane operand.
+ bool isFpMLxInstruction(unsigned Opcode, unsigned &MulOpc,
+ unsigned &AddSubOpc, bool &NegAcc,
+ bool &HasLane) const;
+
+ /// canCauseFpMLxStall - Return true if an instruction of the specified opcode
+ /// will cause stalls when scheduled after (within 4-cycle window) a fp
+ /// MLA / MLS instruction.
+ bool canCauseFpMLxStall(unsigned Opcode) const {
+ return MLxHazardOpcodes.count(Opcode);
+ }
};
static inline
static inline
bool isIndirectBranchOpcode(int Opc) {
- return Opc == ARM::BRIND || Opc == ARM::MOVPCRX || Opc == ARM::tBRIND;
+ return Opc == ARM::BX || Opc == ARM::MOVPCRX || Opc == ARM::tBRIND;
}
/// getInstrPredicate - If instruction is predicated, returns its predicate
int getMatchingCondBranchOpcode(int Opc);
+/// Determine if MI can be folded into an ARM MOVCC instruction, and return the
+/// opcode of the SSA instruction representing the conditional MI.
+unsigned canFoldARMInstrIntoMOVCC(unsigned Reg,
+ MachineInstr *&MI,
+ const MachineRegisterInfo &MRI);
+
+/// Map pseudo instructions that imply an 'S' bit onto real opcodes. Whether
+/// the instruction is encoded with an 'S' bit is determined by the optional
+/// CPSR def operand.
+unsigned convertAddSubFlagsOpcode(unsigned OldOpc);
+
/// emitARMRegPlusImmediate / emitT2RegPlusImmediate - Emits a series of
/// instructions to materializea destreg = basereg + immediate in ARM / Thumb2
/// code.
MachineBasicBlock::iterator &MBBI, DebugLoc dl,
unsigned DestReg, unsigned BaseReg, int NumBytes,
ARMCC::CondCodes Pred, unsigned PredReg,
- const ARMBaseInstrInfo &TII);
+ const ARMBaseInstrInfo &TII, unsigned MIFlags = 0);
void emitT2RegPlusImmediate(MachineBasicBlock &MBB,
MachineBasicBlock::iterator &MBBI, DebugLoc dl,
unsigned DestReg, unsigned BaseReg, int NumBytes,
ARMCC::CondCodes Pred, unsigned PredReg,
- const ARMBaseInstrInfo &TII);
+ const ARMBaseInstrInfo &TII, unsigned MIFlags = 0);
+void emitThumbRegPlusImmediate(MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator &MBBI, DebugLoc dl,
+ unsigned DestReg, unsigned BaseReg,
+ int NumBytes, const TargetInstrInfo &TII,
+ const ARMBaseRegisterInfo& MRI,
+ unsigned MIFlags = 0);
/// rewriteARMFrameIndex / rewriteT2FrameIndex -