//
// The LLVM Compiler Infrastructure
//
-// This file was developed by Chris Lattner and is distributed under
-// the University of Illinois Open Source License. See LICENSE.TXT for details.
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
#include "llvm/Target/TargetLowering.h"
#include "llvm/CodeGen/SelectionDAG.h"
#include "PPC.h"
+#include "PPCSubtarget.h"
namespace llvm {
namespace PPCISD {
enum NodeType {
- // Start the numbering where the builting ops and target ops leave off.
- FIRST_NUMBER = ISD::BUILTIN_OP_END+PPC::INSTRUCTION_LIST_END,
+ // Start the numbering where the builtin ops and target ops leave off.
+ FIRST_NUMBER = ISD::BUILTIN_OP_END,
/// FSEL - Traditional three-operand fsel node.
///
FCTIDZ, FCTIWZ,
/// STFIWX - The STFIWX instruction. The first operand is an input token
- /// chain, then an f64 value to store, then an address to store it to,
- /// then a SRCVALUE for the address.
+ /// chain, then an f64 value to store, then an address to store it to.
STFIWX,
// VMADDFP, VNMSUBFP - The VMADDFP and VNMSUBFP instructions, taking
// three v4f32 operands and producing a v4f32 result.
VMADDFP, VNMSUBFP,
- /// LVE_X - The PPC LVE*X instructions. The size of the element loaded is
- /// the size of the element type of the vector result. The element loaded
- /// depends on the alignment of the input pointer.
- ///
- /// The first operand is a token chain, the second is the address to load
- /// the third is the SRCVALUE node.
- LVE_X,
-
/// VPERM - The PPC VPERM Instruction.
///
VPERM,
/// though these are usually folded into other nodes.
Hi, Lo,
+ TOC_ENTRY,
+
+ /// The following three target-specific nodes are used for calls through
+ /// function pointers in the 64-bit SVR4 ABI.
+
+ /// Restore the TOC from the TOC save area of the current stack frame.
+ /// This is basically a hard coded load instruction which additionally
+ /// takes/produces a flag.
+ TOC_RESTORE,
+
+ /// Like a regular LOAD but additionally taking/producing a flag.
+ LOAD,
+
+ /// LOAD into r2 (also taking/producing a flag). Like TOC_RESTORE, this is
+ /// a hard coded load instruction.
+ LOAD_TOC,
+
+ /// OPRC, CHAIN = DYNALLOC(CHAIN, NEGSIZE, FRAME_INDEX)
+ /// This instruction is lowered in PPCRegisterInfo::eliminateFrameIndex to
+ /// compute an allocation on the stack.
+ DYNALLOC,
+
/// GlobalBaseReg - On Darwin, this node represents the result of the mflr
/// at function entry, used for PIC code.
GlobalBaseReg,
/// registers.
EXTSW_32,
- /// STD_32 - This is the STD instruction for use with "32-bit" registers.
- STD_32,
+ /// CALL - A direct function call.
+ CALL_Darwin, CALL_SVR4,
+
+ /// NOP - Special NOP which follows 64-bit SVR4 calls.
+ NOP,
+
+ /// CHAIN,FLAG = MTCTR(VAL, CHAIN[, INFLAG]) - Directly corresponds to a
+ /// MTCTR instruction.
+ MTCTR,
- /// CALL - A function call.
- CALL,
+ /// CHAIN,FLAG = BCTRL(CHAIN, INFLAG) - Directly corresponds to a
+ /// BCTRL instruction.
+ BCTRL_Darwin, BCTRL_SVR4,
/// Return with a flag operand, matched by 'blr'
RET_FLAG,
+
+ /// R32 = MFCR(CRREG, INFLAG) - Represents the MFCR/MFOCRF instructions.
+ /// This copies the bits corresponding to the specified CRREG into the
+ /// resultant GPR. Bits corresponding to other CR regs are undefined.
+ MFCR,
+
+ /// RESVEC = VCMP(LHS, RHS, OPC) - Represents one of the altivec VCMP*
+ /// instructions. For lack of better number, we use the opcode number
+ /// encoding for the OPC field to identify the compare. For example, 838
+ /// is VCMPGTSH.
+ VCMP,
+
+ /// RESVEC, OUTFLAG = VCMPo(LHS, RHS, OPC) - Represents one of the
+ /// altivec VCMP*o instructions. For lack of better number, we use the
+ /// opcode number encoding for the OPC field to identify the compare. For
+ /// example, 838 is VCMPGTSH.
+ VCMPo,
+
+ /// CHAIN = COND_BRANCH CHAIN, CRRC, OPC, DESTBB [, INFLAG] - This
+ /// corresponds to the COND_BRANCH pseudo instruction. CRRC is the
+ /// condition register to branch on, OPC is the branch opcode to use (e.g.
+ /// PPC::BLE), DESTBB is the destination block to branch to, and INFLAG is
+ /// an optional input flag argument.
+ COND_BRANCH,
+
+ // The following 5 instructions are used only as part of the
+ // long double-to-int conversion sequence.
+
+ /// OUTFLAG = MFFS F8RC - This moves the FPSCR (not modelled) into the
+ /// register.
+ MFFS,
+
+ /// OUTFLAG = MTFSB0 INFLAG - This clears a bit in the FPSCR.
+ MTFSB0,
+
+ /// OUTFLAG = MTFSB1 INFLAG - This sets a bit in the FPSCR.
+ MTFSB1,
+
+ /// F8RC, OUTFLAG = FADDRTZ F8RC, F8RC, INFLAG - This is an FADD done with
+ /// rounding towards zero. It has flags added so it won't move past the
+ /// FPSCR-setting instructions.
+ FADDRTZ,
+
+ /// MTFSF = F8RC, INFLAG - This moves the register into the FPSCR.
+ MTFSF,
+
+ /// LARX = This corresponds to PPC l{w|d}arx instrcution: load and
+ /// reserve indexed. This is used to implement atomic operations.
+ LARX,
+
+ /// STCX = This corresponds to PPC stcx. instrcution: store conditional
+ /// indexed. This is used to implement atomic operations.
+ STCX,
+
+ /// TC_RETURN - A tail call return.
+ /// operand #0 chain
+ /// operand #1 callee (register or absolute)
+ /// operand #2 stack adjustment
+ /// operand #3 optional in flag
+ TC_RETURN,
+
+ /// STD_32 - This is the STD instruction for use with "32-bit" registers.
+ STD_32 = ISD::FIRST_TARGET_MEMORY_OPCODE,
+
+ /// CHAIN = STBRX CHAIN, GPRC, Ptr, Type - This is a
+ /// byte-swapping store instruction. It byte-swaps the low "Type" bits of
+ /// the GPRC input, then stores it through Ptr. Type can be either i16 or
+ /// i32.
+ STBRX,
+
+ /// GPRC, CHAIN = LBRX CHAIN, Ptr, Type - This is a
+ /// byte-swapping load instruction. It loads "Type" bits, byte swaps it,
+ /// then puts it in the bottom bits of the GPRC. TYPE can be either i16
+ /// or i32.
+ LBRX
};
}
/// Define some predicates that are used for node matching.
namespace PPC {
+ /// isVPKUHUMShuffleMask - Return true if this is the shuffle mask for a
+ /// VPKUHUM instruction.
+ bool isVPKUHUMShuffleMask(ShuffleVectorSDNode *N, bool isUnary);
+
+ /// isVPKUWUMShuffleMask - Return true if this is the shuffle mask for a
+ /// VPKUWUM instruction.
+ bool isVPKUWUMShuffleMask(ShuffleVectorSDNode *N, bool isUnary);
+
+ /// isVMRGLShuffleMask - Return true if this is a shuffle mask suitable for
+ /// a VRGL* instruction with the specified unit size (1,2 or 4 bytes).
+ bool isVMRGLShuffleMask(ShuffleVectorSDNode *N, unsigned UnitSize,
+ bool isUnary);
+
+ /// isVMRGHShuffleMask - Return true if this is a shuffle mask suitable for
+ /// a VRGH* instruction with the specified unit size (1,2 or 4 bytes).
+ bool isVMRGHShuffleMask(ShuffleVectorSDNode *N, unsigned UnitSize,
+ bool isUnary);
+
+ /// isVSLDOIShuffleMask - If this is a vsldoi shuffle mask, return the shift
+ /// amount, otherwise return -1.
+ int isVSLDOIShuffleMask(SDNode *N, bool isUnary);
+
/// isSplatShuffleMask - Return true if the specified VECTOR_SHUFFLE operand
/// specifies a splat of a single element that is suitable for input to
/// VSPLTB/VSPLTH/VSPLTW.
- bool isSplatShuffleMask(SDNode *N);
+ bool isSplatShuffleMask(ShuffleVectorSDNode *N, unsigned EltSize);
+ /// isAllNegativeZeroVector - Returns true if all elements of build_vector
+ /// are -0.0.
+ bool isAllNegativeZeroVector(SDNode *N);
+
/// getVSPLTImmediate - Return the appropriate VSPLT* immediate to splat the
/// specified isSplatShuffleMask VECTOR_SHUFFLE mask.
- unsigned getVSPLTImmediate(SDNode *N);
+ unsigned getVSPLTImmediate(SDNode *N, unsigned EltSize);
- /// isZeroVector - Return true if this build_vector is an all-zero vector.
- ///
- bool isZeroVector(SDNode *N);
+ /// get_VSPLTI_elt - If this is a build_vector of constants which can be
+ /// formed by using a vspltis[bhw] instruction of the specified element
+ /// size, return the constant being splatted. The ByteSize field indicates
+ /// the number of bytes of each element [124] -> [bhw].
+ SDValue get_VSPLTI_elt(SDNode *N, unsigned ByteSize, SelectionDAG &DAG);
}
class PPCTargetLowering : public TargetLowering {
int VarArgsFrameIndex; // FrameIndex for start of varargs area.
- int ReturnAddrIndex; // FrameIndex for return slot.
+ int VarArgsStackOffset; // StackOffset for start of stack
+ // arguments.
+ unsigned VarArgsNumGPR; // Index of the first unused integer
+ // register for parameter passing.
+ unsigned VarArgsNumFPR; // Index of the first unused double
+ // register for parameter passing.
+ const PPCSubtarget &PPCSubTarget;
public:
- PPCTargetLowering(TargetMachine &TM);
+ explicit PPCTargetLowering(PPCTargetMachine &TM);
/// getTargetNodeName() - This method returns the name of a target specific
/// DAG node.
virtual const char *getTargetNodeName(unsigned Opcode) const;
+
+ /// getSetCCResultType - Return the ISD::SETCC ValueType
+ virtual MVT::SimpleValueType getSetCCResultType(EVT VT) const;
+
+ /// getPreIndexedAddressParts - returns true by value, base pointer and
+ /// offset pointer and addressing mode by reference if the node's address
+ /// can be legally represented as pre-indexed load / store address.
+ virtual bool getPreIndexedAddressParts(SDNode *N, SDValue &Base,
+ SDValue &Offset,
+ ISD::MemIndexedMode &AM,
+ SelectionDAG &DAG) const;
+
+ /// SelectAddressRegReg - Given the specified addressed, check to see if it
+ /// can be represented as an indexed [r+r] operation. Returns false if it
+ /// can be more efficiently represented with [r+imm].
+ bool SelectAddressRegReg(SDValue N, SDValue &Base, SDValue &Index,
+ SelectionDAG &DAG) const;
+
+ /// SelectAddressRegImm - Returns true if the address N can be represented
+ /// by a base register plus a signed 16-bit displacement [r+imm], and if it
+ /// is not better represented as reg+reg.
+ bool SelectAddressRegImm(SDValue N, SDValue &Disp, SDValue &Base,
+ SelectionDAG &DAG) const;
+
+ /// SelectAddressRegRegOnly - Given the specified addressed, force it to be
+ /// represented as an indexed [r+r] operation.
+ bool SelectAddressRegRegOnly(SDValue N, SDValue &Base, SDValue &Index,
+ SelectionDAG &DAG) const;
+
+ /// SelectAddressRegImmShift - Returns true if the address N can be
+ /// represented by a base register plus a signed 14-bit displacement
+ /// [r+imm*4]. Suitable for use by STD and friends.
+ bool SelectAddressRegImmShift(SDValue N, SDValue &Disp, SDValue &Base,
+ SelectionDAG &DAG) const;
+
/// LowerOperation - Provide custom lowering hooks for some operations.
///
- virtual SDOperand LowerOperation(SDOperand Op, SelectionDAG &DAG);
-
- virtual SDOperand PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const;
+ virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG);
+
+ /// ReplaceNodeResults - Replace the results of node with an illegal result
+ /// type with new values built out of custom code.
+ ///
+ virtual void ReplaceNodeResults(SDNode *N, SmallVectorImpl<SDValue>&Results,
+ SelectionDAG &DAG);
- /// LowerArguments - This hook must be implemented to indicate how we should
- /// lower the arguments for the specified function, into the specified DAG.
- virtual std::vector<SDOperand>
- LowerArguments(Function &F, SelectionDAG &DAG);
+ virtual SDValue PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const;
- /// LowerCallTo - This hook lowers an abstract call to a function into an
- /// actual call.
- virtual std::pair<SDOperand, SDOperand>
- LowerCallTo(SDOperand Chain, const Type *RetTy, bool isVarArg,
- unsigned CC,
- bool isTailCall, SDOperand Callee, ArgListTy &Args,
- SelectionDAG &DAG);
-
- virtual MachineBasicBlock *InsertAtEndOfBasicBlock(MachineInstr *MI,
- MachineBasicBlock *MBB);
+ virtual void computeMaskedBitsForTargetNode(const SDValue Op,
+ const APInt &Mask,
+ APInt &KnownZero,
+ APInt &KnownOne,
+ const SelectionDAG &DAG,
+ unsigned Depth = 0) const;
+
+ virtual MachineBasicBlock *EmitInstrWithCustomInserter(MachineInstr *MI,
+ MachineBasicBlock *MBB,
+ DenseMap<MachineBasicBlock*, MachineBasicBlock*> *EM) const;
+ MachineBasicBlock *EmitAtomicBinary(MachineInstr *MI,
+ MachineBasicBlock *MBB, bool is64Bit,
+ unsigned BinOpcode) const;
+ MachineBasicBlock *EmitPartwordAtomicBinary(MachineInstr *MI,
+ MachineBasicBlock *MBB,
+ bool is8bit, unsigned Opcode) const;
- ConstraintType getConstraintType(char ConstraintLetter) const;
- std::vector<unsigned>
- getRegClassForInlineAsmConstraint(const std::string &Constraint,
- MVT::ValueType VT) const;
- bool isOperandValidForConstraint(SDOperand Op, char ConstraintLetter);
+ ConstraintType getConstraintType(const std::string &Constraint) const;
+ std::pair<unsigned, const TargetRegisterClass*>
+ getRegForInlineAsmConstraint(const std::string &Constraint,
+ EVT VT) const;
+ /// getByValTypeAlignment - Return the desired alignment for ByVal aggregate
+ /// function arguments in the caller parameter area. This is the actual
+ /// alignment, not its logarithm.
+ unsigned getByValTypeAlignment(const Type *Ty) const;
+
+ /// LowerAsmOperandForConstraint - Lower the specified operand into the Ops
+ /// vector. If it is invalid, don't add anything to Ops. If hasMemory is
+ /// true it means one of the asm constraint of the inline asm instruction
+ /// being processed is 'm'.
+ virtual void LowerAsmOperandForConstraint(SDValue Op,
+ char ConstraintLetter,
+ bool hasMemory,
+ std::vector<SDValue> &Ops,
+ SelectionDAG &DAG) const;
+
+ /// isLegalAddressingMode - Return true if the addressing mode represented
+ /// by AM is legal for this target, for a load/store of the specified type.
+ virtual bool isLegalAddressingMode(const AddrMode &AM, const Type *Ty)const;
+
/// isLegalAddressImmediate - Return true if the integer value can be used
- /// as the offset of the target addressing mode.
- virtual bool isLegalAddressImmediate(int64_t V) const;
+ /// as the offset of the target addressing mode for load / store of the
+ /// given type.
+ virtual bool isLegalAddressImmediate(int64_t V, const Type *Ty) const;
+
+ /// isLegalAddressImmediate - Return true if the GlobalValue can be used as
+ /// the offset of the target addressing mode.
+ virtual bool isLegalAddressImmediate(GlobalValue *GV) const;
+
+ virtual bool isOffsetFoldingLegal(const GlobalAddressSDNode *GA) const;
+
+ virtual EVT getOptimalMemOpType(uint64_t Size, unsigned Align,
+ bool isSrcConst, bool isSrcStr,
+ SelectionDAG &DAG) const;
+
+ /// getFunctionAlignment - Return the Log2 alignment of this function.
+ virtual unsigned getFunctionAlignment(const Function *F) const;
+
+ private:
+ SDValue getFramePointerFrameIndex(SelectionDAG & DAG) const;
+ SDValue getReturnAddrFrameIndex(SelectionDAG & DAG) const;
+
+ bool
+ IsEligibleForTailCallOptimization(SDValue Callee,
+ CallingConv::ID CalleeCC,
+ bool isVarArg,
+ const SmallVectorImpl<ISD::InputArg> &Ins,
+ SelectionDAG& DAG) const;
+
+ SDValue EmitTailCallLoadFPAndRetAddr(SelectionDAG & DAG,
+ int SPDiff,
+ SDValue Chain,
+ SDValue &LROpOut,
+ SDValue &FPOpOut,
+ bool isDarwinABI,
+ DebugLoc dl);
+
+ SDValue LowerRETURNADDR(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerFRAMEADDR(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerConstantPool(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerBlockAddress(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerGlobalAddress(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerGlobalTLSAddress(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerJumpTable(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerSETCC(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerTRAMPOLINE(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerVASTART(SDValue Op, SelectionDAG &DAG,
+ int VarArgsFrameIndex, int VarArgsStackOffset,
+ unsigned VarArgsNumGPR, unsigned VarArgsNumFPR,
+ const PPCSubtarget &Subtarget);
+ SDValue LowerVAARG(SDValue Op, SelectionDAG &DAG, int VarArgsFrameIndex,
+ int VarArgsStackOffset, unsigned VarArgsNumGPR,
+ unsigned VarArgsNumFPR, const PPCSubtarget &Subtarget);
+ SDValue LowerSTACKRESTORE(SDValue Op, SelectionDAG &DAG,
+ const PPCSubtarget &Subtarget);
+ SDValue LowerDYNAMIC_STACKALLOC(SDValue Op, SelectionDAG &DAG,
+ const PPCSubtarget &Subtarget);
+ SDValue LowerSELECT_CC(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerFP_TO_INT(SDValue Op, SelectionDAG &DAG, DebugLoc dl);
+ SDValue LowerSINT_TO_FP(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerFLT_ROUNDS_(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerSHL_PARTS(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerSRL_PARTS(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerSRA_PARTS(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerVECTOR_SHUFFLE(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerINTRINSIC_WO_CHAIN(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerSCALAR_TO_VECTOR(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerMUL(SDValue Op, SelectionDAG &DAG);
+
+ SDValue LowerCallResult(SDValue Chain, SDValue InFlag,
+ CallingConv::ID CallConv, bool isVarArg,
+ const SmallVectorImpl<ISD::InputArg> &Ins,
+ DebugLoc dl, SelectionDAG &DAG,
+ SmallVectorImpl<SDValue> &InVals);
+ SDValue FinishCall(CallingConv::ID CallConv, DebugLoc dl, bool isTailCall,
+ bool isVarArg,
+ SelectionDAG &DAG,
+ SmallVector<std::pair<unsigned, SDValue>, 8>
+ &RegsToPass,
+ SDValue InFlag, SDValue Chain,
+ SDValue &Callee,
+ int SPDiff, unsigned NumBytes,
+ const SmallVectorImpl<ISD::InputArg> &Ins,
+ SmallVectorImpl<SDValue> &InVals);
+
+ virtual SDValue
+ LowerFormalArguments(SDValue Chain,
+ CallingConv::ID CallConv, bool isVarArg,
+ const SmallVectorImpl<ISD::InputArg> &Ins,
+ DebugLoc dl, SelectionDAG &DAG,
+ SmallVectorImpl<SDValue> &InVals);
+
+ virtual SDValue
+ LowerCall(SDValue Chain, SDValue Callee, const Type *RetTy,
+ CallingConv::ID CallConv, bool isVarArg, bool &isTailCall,
+ const SmallVectorImpl<ISD::OutputArg> &Outs,
+ const SmallVectorImpl<ISD::InputArg> &Ins,
+ DebugLoc dl, SelectionDAG &DAG,
+ SmallVectorImpl<SDValue> &InVals);
+
+ virtual SDValue
+ LowerReturn(SDValue Chain,
+ CallingConv::ID CallConv, bool isVarArg,
+ const SmallVectorImpl<ISD::OutputArg> &Outs,
+ DebugLoc dl, SelectionDAG &DAG);
+
+ SDValue
+ LowerFormalArguments_Darwin(SDValue Chain,
+ CallingConv::ID CallConv, bool isVarArg,
+ const SmallVectorImpl<ISD::InputArg> &Ins,
+ DebugLoc dl, SelectionDAG &DAG,
+ SmallVectorImpl<SDValue> &InVals);
+ SDValue
+ LowerFormalArguments_SVR4(SDValue Chain,
+ CallingConv::ID CallConv, bool isVarArg,
+ const SmallVectorImpl<ISD::InputArg> &Ins,
+ DebugLoc dl, SelectionDAG &DAG,
+ SmallVectorImpl<SDValue> &InVals);
+
+ SDValue
+ LowerCall_Darwin(SDValue Chain, SDValue Callee,
+ CallingConv::ID CallConv, bool isVarArg, bool isTailCall,
+ const SmallVectorImpl<ISD::OutputArg> &Outs,
+ const SmallVectorImpl<ISD::InputArg> &Ins,
+ DebugLoc dl, SelectionDAG &DAG,
+ SmallVectorImpl<SDValue> &InVals);
+ SDValue
+ LowerCall_SVR4(SDValue Chain, SDValue Callee,
+ CallingConv::ID CallConv, bool isVarArg, bool isTailCall,
+ const SmallVectorImpl<ISD::OutputArg> &Outs,
+ const SmallVectorImpl<ISD::InputArg> &Ins,
+ DebugLoc dl, SelectionDAG &DAG,
+ SmallVectorImpl<SDValue> &InVals);
};
}