#include "X86RegisterInfo.h"
#include "X86MachineFunctionInfo.h"
#include "llvm/Target/TargetLowering.h"
+#include "llvm/CodeGen/FastISel.h"
#include "llvm/CodeGen/SelectionDAG.h"
#include "llvm/CodeGen/CallingConvLower.h"
// X86 Specific DAG Nodes
enum NodeType {
// Start the numbering where the builtin ops leave off.
- FIRST_NUMBER = ISD::BUILTIN_OP_END+X86::INSTRUCTION_LIST_END,
+ FIRST_NUMBER = ISD::BUILTIN_OP_END,
/// BSF - Bit scan forward.
/// BSR - Bit scan reverse.
/// X86 compare and logical compare instructions.
CMP, COMI, UCOMI,
+ /// X86 bit-test instructions.
+ BT,
+
/// X86 SetCC. Operand 1 is condition code, and operand 2 is the flag
/// operand produced by a CMP instruction.
SETCC,
/// in order to obtain suitable precision.
FRSQRT, FRCP,
- // TLSADDR, THREAThread - Thread Local Storage.
+ // TLSADDR, THREAD_POINTER - Thread Local Storage.
TLSADDR, THREAD_POINTER,
// EH_RETURN - Exception Handling helpers.
LCMPXCHG_DAG,
LCMPXCHG8_DAG,
+ // ATOMADD64_DAG, ATOMSUB64_DAG, ATOMOR64_DAG, ATOMAND64_DAG,
+ // ATOMXOR64_DAG, ATOMNAND64_DAG, ATOMSWAP64_DAG -
+ // Atomic 64-bit binary operations.
+ ATOMADD64_DAG,
+ ATOMSUB64_DAG,
+ ATOMOR64_DAG,
+ ATOMXOR64_DAG,
+ ATOMAND64_DAG,
+ ATOMNAND64_DAG,
+ ATOMSWAP64_DAG,
+
// FNSTCW16m - Store FP control world into i16 memory.
FNSTCW16m,
// PCMP* - Vector integer comparisons.
PCMPEQB, PCMPEQW, PCMPEQD, PCMPEQQ,
- PCMPGTB, PCMPGTW, PCMPGTD, PCMPGTQ
+ PCMPGTB, PCMPGTW, PCMPGTD, PCMPGTQ,
+
+ // ADD, SUB, SMUL, UMUL - Arithmetic operations with overflow/carry
+ // intrinsics.
+ ADD, SUB, SMUL, UMUL
};
}
/// specifies a splat of zero element.
bool isSplatLoMask(SDNode *N);
+ /// isMOVDDUPMask - Return true if the specified VECTOR_SHUFFLE operand
+ /// specifies a shuffle of elements that is suitable for input to MOVDDUP.
+ bool isMOVDDUPMask(SDNode *N);
+
/// getShuffleSHUFImmediate - Return the appropriate immediate to shuffle
/// the specified isShuffleMask VECTOR_SHUFFLE mask with PSHUF* and SHUFP*
/// instructions.
///
virtual SDValue LowerOperation(SDValue Op, SelectionDAG &DAG);
- /// ReplaceNodeResults - Replace a node with an illegal result type
- /// with a new node built out of custom code.
+ /// ReplaceNodeResults - Replace the results of node with an illegal result
+ /// type with new values built out of custom code.
///
- virtual SDNode *ReplaceNodeResults(SDNode *N, SelectionDAG &DAG);
+ virtual void ReplaceNodeResults(SDNode *N, SmallVectorImpl<SDValue>&Results,
+ SelectionDAG &DAG);
virtual SDValue PerformDAGCombine(SDNode *N, DAGCombinerInfo &DCI) const;
virtual const char *getTargetNodeName(unsigned Opcode) const;
/// getSetCCResultType - Return the ISD::SETCC ValueType
- virtual MVT getSetCCResultType(const SDValue &) const;
+ virtual MVT getSetCCResultType(MVT VT) const;
/// computeMaskedBitsForTargetNode - Determine which of the bits specified
/// in Mask are known to be either zero or one and return them in the
virtual const char *LowerXConstraint(MVT ConstraintVT) const;
/// LowerAsmOperandForConstraint - Lower the specified operand into the Ops
- /// vector. If it is invalid, don't add anything to 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;
/// IsEligibleForTailCallOptimization - Check whether the call is eligible
/// for tail call optimization. Target which want to do tail call
/// optimization should implement this function.
- virtual bool IsEligibleForTailCallOptimization(SDValue Call,
+ virtual bool IsEligibleForTailCallOptimization(CallSDNode *TheCall,
SDValue Ret,
SelectionDAG &DAG) const;
return (VT == MVT::f64 && X86ScalarSSEf64) || // f64 is when SSE2
(VT == MVT::f32 && X86ScalarSSEf32); // f32 is when SSE1
}
+
+ /// getWidenVectorType: given a vector type, returns the type to widen
+ /// to (e.g., v7i8 to v8i8). If the vector type is legal, it returns itself.
+ /// If there is no vector type that we want to widen to, returns MVT::Other
+ /// When and were to widen is target dependent based on the cost of
+ /// scalarizing vs using the wider vector type.
+ virtual MVT getWidenVectorType(MVT VT) const;
+
+ /// createFastISel - This method returns a target specific FastISel object,
+ /// or null if the target does not support "fast" ISel.
+ virtual FastISel *
+ createFastISel(MachineFunction &mf,
+ MachineModuleInfo *mmi, DwarfWriter *dw,
+ DenseMap<const Value *, unsigned> &,
+ DenseMap<const BasicBlock *, MachineBasicBlock *> &,
+ DenseMap<const AllocaInst *, int> &
+#ifndef NDEBUG
+ , SmallSet<Instruction*, 8> &
+#endif
+ );
private:
/// Subtarget - Keep a pointer to the X86Subtarget around so that we can
/// make the right decision when generating code for different targets.
const X86Subtarget *Subtarget;
const X86RegisterInfo *RegInfo;
+ const TargetData *TD;
/// X86StackPtr - X86 physical register used as stack ptr.
unsigned X86StackPtr;
bool X86ScalarSSEf32;
bool X86ScalarSSEf64;
- SDNode *LowerCallResult(SDValue Chain, SDValue InFlag, SDNode*TheCall,
+ SDNode *LowerCallResult(SDValue Chain, SDValue InFlag, CallSDNode *TheCall,
unsigned CallingConv, SelectionDAG &DAG);
SDValue LowerMemArgument(SDValue Op, SelectionDAG &DAG,
const CCValAssign &VA, MachineFrameInfo *MFI,
unsigned CC, SDValue Root, unsigned i);
- SDValue LowerMemOpCallTo(SDValue Op, SelectionDAG &DAG,
+ SDValue LowerMemOpCallTo(CallSDNode *TheCall, SelectionDAG &DAG,
const SDValue &StackPtr,
const CCValAssign &VA, SDValue Chain,
- SDValue Arg);
+ SDValue Arg, ISD::ArgFlagsTy Flags);
// Call lowering helpers.
- bool IsCalleePop(SDValue Op);
+ bool IsCalleePop(bool isVarArg, unsigned CallingConv);
bool CallRequiresGOTPtrInReg(bool Is64Bit, bool IsTailCall);
bool CallRequiresFnAddressInReg(bool Is64Bit, bool IsTailCall);
SDValue EmitTailCallLoadRetAddr(SelectionDAG &DAG, SDValue &OutRetAddr,
SDValue Chain, bool IsTailCall, bool Is64Bit,
- int FPDiff);
+ int FPDiff, DebugLoc dl);
- CCAssignFn *CCAssignFnForNode(SDValue Op) const;
+ CCAssignFn *CCAssignFnForNode(unsigned CallingConv) const;
NameDecorationStyle NameDecorationForFORMAL_ARGUMENTS(SDValue Op);
unsigned GetAlignedArgumentStackSize(unsigned StackSize, SelectionDAG &DAG);
SDValue LowerINSERT_VECTOR_ELT_SSE4(SDValue Op, SelectionDAG &DAG);
SDValue LowerSCALAR_TO_VECTOR(SDValue Op, SelectionDAG &DAG);
SDValue LowerConstantPool(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerGlobalAddress(const GlobalValue *GV, int64_t Offset,
+ SelectionDAG &DAG) const;
SDValue LowerGlobalAddress(SDValue Op, SelectionDAG &DAG);
SDValue LowerGlobalTLSAddress(SDValue Op, SelectionDAG &DAG);
SDValue LowerExternalSymbol(SDValue Op, SelectionDAG &DAG);
SDValue LowerShift(SDValue Op, SelectionDAG &DAG);
SDValue LowerSINT_TO_FP(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerUINT_TO_FP(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerUINT_TO_FP_i64(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerUINT_TO_FP_i32(SDValue Op, SelectionDAG &DAG);
SDValue LowerFP_TO_SINT(SDValue Op, SelectionDAG &DAG);
SDValue LowerFABS(SDValue Op, SelectionDAG &DAG);
SDValue LowerFNEG(SDValue Op, SelectionDAG &DAG);
SDValue LowerFLT_ROUNDS_(SDValue Op, SelectionDAG &DAG);
SDValue LowerCTLZ(SDValue Op, SelectionDAG &DAG);
SDValue LowerCTTZ(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerMUL_V2I64(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerXALUO(SDValue Op, SelectionDAG &DAG);
+
SDValue LowerCMP_SWAP(SDValue Op, SelectionDAG &DAG);
- SDNode *ExpandFP_TO_SINT(SDNode *N, SelectionDAG &DAG);
- SDNode *ExpandREADCYCLECOUNTER(SDNode *N, SelectionDAG &DAG);
- SDNode *ExpandATOMIC_CMP_SWAP(SDNode *N, SelectionDAG &DAG);
- SDNode *ExpandATOMIC_LOAD_SUB(SDNode *N, SelectionDAG &DAG);
-
- SDValue EmitTargetCodeForMemset(SelectionDAG &DAG,
- SDValue Chain,
- SDValue Dst, SDValue Src,
- SDValue Size, unsigned Align,
- const Value *DstSV, uint64_t DstSVOff);
- SDValue EmitTargetCodeForMemcpy(SelectionDAG &DAG,
- SDValue Chain,
- SDValue Dst, SDValue Src,
- SDValue Size, unsigned Align,
- bool AlwaysInline,
- const Value *DstSV, uint64_t DstSVOff,
- const Value *SrcSV, uint64_t SrcSVOff);
+ SDValue LowerLOAD_SUB(SDValue Op, SelectionDAG &DAG);
+ SDValue LowerREADCYCLECOUNTER(SDValue Op, SelectionDAG &DAG);
+
+ void ReplaceATOMIC_BINARY_64(SDNode *N, SmallVectorImpl<SDValue> &Results,
+ SelectionDAG &DAG, unsigned NewOp);
+
+ SDValue EmitTargetCodeForMemset(SelectionDAG &DAG, DebugLoc dl,
+ SDValue Chain,
+ SDValue Dst, SDValue Src,
+ SDValue Size, unsigned Align,
+ const Value *DstSV, uint64_t DstSVOff);
+ SDValue EmitTargetCodeForMemcpy(SelectionDAG &DAG, DebugLoc dl,
+ SDValue Chain,
+ SDValue Dst, SDValue Src,
+ SDValue Size, unsigned Align,
+ bool AlwaysInline,
+ const Value *DstSV, uint64_t DstSVOff,
+ const Value *SrcSV, uint64_t SrcSVOff);
/// Utility function to emit atomic bitwise operations (and, or, xor).
// It takes the bitwise instruction to expand, the associated machine basic
unsigned EAXreg,
TargetRegisterClass *RC,
bool invSrc = false);
+
+ MachineBasicBlock *EmitAtomicBit6432WithCustomInserter(
+ MachineInstr *BInstr,
+ MachineBasicBlock *BB,
+ unsigned regOpcL,
+ unsigned regOpcH,
+ unsigned immOpcL,
+ unsigned immOpcH,
+ bool invSrc = false);
/// Utility function to emit atomic min and max. It takes the min/max
// instruction to expand, the associated basic block, and the associated
MachineBasicBlock *BB,
unsigned cmovOpc);
};
+
+ namespace X86 {
+ FastISel *createFastISel(MachineFunction &mf,
+ MachineModuleInfo *mmi, DwarfWriter *dw,
+ DenseMap<const Value *, unsigned> &,
+ DenseMap<const BasicBlock *, MachineBasicBlock *> &,
+ DenseMap<const AllocaInst *, int> &
+#ifndef NDEBUG
+ , SmallSet<Instruction*, 8> &
+#endif
+ );
+ }
}
#endif // X86ISELLOWERING_H