#ifndef X86ISELLOWERING_H
#define X86ISELLOWERING_H
-#include "X86Subtarget.h"
-#include "X86RegisterInfo.h"
#include "X86MachineFunctionInfo.h"
-#include "llvm/Target/TargetLowering.h"
-#include "llvm/Target/TargetOptions.h"
+#include "X86RegisterInfo.h"
+#include "X86Subtarget.h"
+#include "llvm/CodeGen/CallingConvLower.h"
#include "llvm/CodeGen/FastISel.h"
#include "llvm/CodeGen/SelectionDAG.h"
-#include "llvm/CodeGen/CallingConvLower.h"
+#include "llvm/Target/TargetLowering.h"
+#include "llvm/Target/TargetOptions.h"
namespace llvm {
namespace X86ISD {
/// to X86::XORPS or X86::XORPD.
FXOR,
+ /// FANDN - Bitwise logical ANDNOT of floating point values. This
+ /// corresponds to X86::ANDNPS or X86::ANDNPD.
+ FANDN,
+
/// FSRL - Bitwise logical right shift of floating point values. These
/// corresponds to X86::PSRLDQ.
FSRL,
/// mnemonic, so do I; blame Intel.
MOVDQ2Q,
+ /// MMX_MOVD2W - Copies a 32-bit value from the low word of a MMX
+ /// vector to a GPR.
+ MMX_MOVD2W,
+
/// PEXTRB - Extract an 8-bit value from a vector and zero extend it to
/// i32, corresponds to X86::PEXTRB.
PEXTRB,
/// PSIGN - Copy integer sign.
PSIGN,
- /// BLENDV - Blend where the selector is an XMM.
+ /// BLENDV - Blend where the selector is a register.
BLENDV,
- /// BLENDxx - Blend where the selector is an immediate.
- BLENDPW,
- BLENDPS,
- BLENDPD,
+ /// BLENDI - Blend where the selector is an immediate.
+ BLENDI,
+
+ // SUBUS - Integer sub with unsigned saturation.
+ SUBUS,
/// HADD - Integer horizontal add.
HADD,
/// FHSUB - Floating point horizontal sub.
FHSUB,
+ /// UMAX, UMIN - Unsigned integer max and min.
+ UMAX, UMIN,
+
+ /// SMAX, SMIN - Signed integer max and min.
+ SMAX, SMIN,
+
/// FMAX, FMIN - Floating point max and min.
///
FMAX, FMIN,
// EH_SJLJ_LONGJMP - SjLj exception handling longjmp.
EH_SJLJ_LONGJMP,
- /// TC_RETURN - Tail call return.
- /// operand #0 chain
- /// operand #1 callee (register or absolute)
- /// operand #2 stack adjustment
- /// operand #3 optional in flag
+ /// TC_RETURN - Tail call return. See X86TargetLowering::LowerCall for
+ /// the list of operands.
TC_RETURN,
// VZEXT_MOVL - Vector move low and zero extend.
// VSEXT - Vector integer signed-extend.
VSEXT,
+ // VTRUNC - Vector integer truncate.
+ VTRUNC,
+
+ // VTRUNC - Vector integer truncate with mask.
+ VTRUNCM,
+
// VFPEXT - Vector FP extend.
VFPEXT,
// PCMP* - Vector integer comparisons.
PCMPEQ, PCMPGT,
+ // PCMP*M - Vector integer comparisons, the result is in a mask vector.
+ PCMPEQM, PCMPGTM,
+
+ /// CMPM, CMPMU - Vector comparison generating mask bits for fp and
+ /// integer signed and unsigned data types.
+ CMPM,
+ CMPMU,
// ADD, SUB, SMUL, etc. - Arithmetic operations with FLAGS results.
ADD, SUB, ADC, SBB, SMUL,
INC, DEC, OR, XOR, AND,
- ANDN, // ANDN - Bitwise AND NOT with FLAGS results.
-
BLSI, // BLSI - Extract lowest set isolated bit
BLSMSK, // BLSMSK - Get mask up to lowest set bit
BLSR, // BLSR - Reset lowest set bit
+ BZHI, // BZHI - Zero high bits
+ BEXTR, // BEXTR - Bit field extract
UMUL, // LOW, HI, FLAGS = umul LHS, RHS
// MUL_IMM - X86 specific multiply by immediate.
MUL_IMM,
- // PTEST - Vector bitwise comparisons
+ // PTEST - Vector bitwise comparisons.
PTEST,
- // TESTP - Vector packed fp sign bitwise comparisons
+ // TESTP - Vector packed fp sign bitwise comparisons.
TESTP,
+ // TESTM - Vector "test" in AVX-512, the result is in a mask vector.
+ TESTM,
+
+ // OR/AND test for masks
+ KORTEST,
+ KTEST,
+
// Several flavors of instructions with vector shuffle behaviors.
- PALIGN,
+ PALIGNR,
PSHUFD,
PSHUFHW,
PSHUFLW,
UNPCKH,
VPERMILP,
VPERMV,
+ VPERMV3,
VPERMI,
VPERM2X128,
VBROADCAST,
+ // masked broadcast
+ VBROADCASTM,
+ VINSERT,
// PMULUDQ - Vector multiply packed unsigned doubleword integers
PMULUDQ,
// RDRAND - Get a random integer and indicate whether it is valid in CF.
RDRAND,
+ // RDSEED - Get a NIST SP800-90B & C compliant random integer and
+ // indicate whether it is valid in CF.
+ RDSEED,
+
// PCMP*STRI
PCMPISTRI,
PCMPESTRI,
+ // XTEST - Test if in transactional execution.
+ XTEST,
+
// ATOMADD64_DAG, ATOMSUB64_DAG, ATOMOR64_DAG, ATOMAND64_DAG,
// ATOMXOR64_DAG, ATOMNAND64_DAG, ATOMSWAP64_DAG -
// Atomic 64-bit binary operations.
/// Define some predicates that are used for node matching.
namespace X86 {
- /// isVEXTRACTF128Index - Return true if the specified
+ /// isVEXTRACT128Index - Return true if the specified
+ /// EXTRACT_SUBVECTOR operand specifies a vector extract that is
+ /// suitable for input to VEXTRACTF128, VEXTRACTI128 instructions.
+ bool isVEXTRACT128Index(SDNode *N);
+
+ /// isVINSERT128Index - Return true if the specified
+ /// INSERT_SUBVECTOR operand specifies a subvector insert that is
+ /// suitable for input to VINSERTF128, VINSERTI128 instructions.
+ bool isVINSERT128Index(SDNode *N);
+
+ /// isVEXTRACT256Index - Return true if the specified
/// EXTRACT_SUBVECTOR operand specifies a vector extract that is
- /// suitable for input to VEXTRACTF128.
- bool isVEXTRACTF128Index(SDNode *N);
+ /// suitable for input to VEXTRACTF64X4, VEXTRACTI64X4 instructions.
+ bool isVEXTRACT256Index(SDNode *N);
- /// isVINSERTF128Index - Return true if the specified
+ /// isVINSERT256Index - Return true if the specified
/// INSERT_SUBVECTOR operand specifies a subvector insert that is
- /// suitable for input to VINSERTF128.
- bool isVINSERTF128Index(SDNode *N);
+ /// suitable for input to VINSERTF64X4, VINSERTI64X4 instructions.
+ bool isVINSERT256Index(SDNode *N);
- /// getExtractVEXTRACTF128Immediate - Return the appropriate
+ /// getExtractVEXTRACT128Immediate - Return the appropriate
/// immediate to extract the specified EXTRACT_SUBVECTOR index
- /// with VEXTRACTF128 instructions.
- unsigned getExtractVEXTRACTF128Immediate(SDNode *N);
+ /// with VEXTRACTF128, VEXTRACTI128 instructions.
+ unsigned getExtractVEXTRACT128Immediate(SDNode *N);
- /// getInsertVINSERTF128Immediate - Return the appropriate
+ /// getInsertVINSERT128Immediate - Return the appropriate
/// immediate to insert at the specified INSERT_SUBVECTOR index
- /// with VINSERTF128 instructions.
- unsigned getInsertVINSERTF128Immediate(SDNode *N);
+ /// with VINSERTF128, VINSERT128 instructions.
+ unsigned getInsertVINSERT128Immediate(SDNode *N);
+
+ /// getExtractVEXTRACT256Immediate - Return the appropriate
+ /// immediate to extract the specified EXTRACT_SUBVECTOR index
+ /// with VEXTRACTF64X4, VEXTRACTI64x4 instructions.
+ unsigned getExtractVEXTRACT256Immediate(SDNode *N);
+
+ /// getInsertVINSERT256Immediate - Return the appropriate
+ /// immediate to insert at the specified INSERT_SUBVECTOR index
+ /// with VINSERTF64x4, VINSERTI64x4 instructions.
+ unsigned getInsertVINSERT256Immediate(SDNode *N);
/// isZeroNode - Returns true if Elt is a constant zero or a floating point
/// constant +0.0.
virtual unsigned getJumpTableEncoding() const;
- virtual MVT getShiftAmountTy(EVT LHSTy) const { return MVT::i8; }
+ virtual MVT getScalarShiftAmountTy(EVT LHSTy) const { return MVT::i8; }
virtual const MCExpr *
LowerCustomJumpTableEntry(const MachineJumpTableInfo *MJTI,
getPICJumpTableRelocBaseExpr(const MachineFunction *MF,
unsigned JTI, MCContext &Ctx) const;
- /// getStackPtrReg - Return the stack pointer register we are using: either
- /// ESP or RSP.
- unsigned getStackPtrReg() const { return X86StackPtr; }
-
/// getByValTypeAlignment - Return the desired alignment for ByVal aggregate
/// function arguments in the caller parameter area. For X86, aggregates
/// that contains are placed at 16-byte boundaries while the rest are at
/// lowering. If DstAlign is zero that means it's safe to destination
/// alignment can satisfy any constraint. Similarly if SrcAlign is zero it
/// means there isn't a need to check it against alignment requirement,
- /// probably because the source does not need to be loaded. If
- /// 'IsZeroVal' is true, that means it's safe to return a
- /// non-scalar-integer type, e.g. empty string source, constant, or loaded
- /// from memory. 'MemcpyStrSrc' indicates whether the memcpy source is
- /// constant so it does not need to be loaded.
+ /// probably because the source does not need to be loaded. If 'IsMemset' is
+ /// true, that means it's expanding a memset. If 'ZeroMemset' is true, that
+ /// means it's a memset of zero. 'MemcpyStrSrc' indicates whether the memcpy
+ /// source is constant so it does not need to be loaded.
/// It returns EVT::Other if the type should be determined using generic
/// target-independent logic.
virtual EVT
getOptimalMemOpType(uint64_t Size, unsigned DstAlign, unsigned SrcAlign,
- bool IsZeroVal, bool MemcpyStrSrc,
+ bool IsMemset, bool ZeroMemset, bool MemcpyStrSrc,
MachineFunction &MF) const;
+ /// isSafeMemOpType - Returns true if it's safe to use load / store of the
+ /// specified type to expand memcpy / memset inline. This is mostly true
+ /// for all types except for some special cases. For example, on X86
+ /// targets without SSE2 f64 load / store are done with fldl / fstpl which
+ /// also does type conversion. Note the specified type doesn't have to be
+ /// legal as the hook is used before type legalization.
+ virtual bool isSafeMemOpType(MVT VT) const;
+
/// allowsUnalignedMemoryAccesses - Returns true if the target allows
- /// unaligned memory accesses. of the specified type.
- virtual bool allowsUnalignedMemoryAccesses(EVT VT) const {
- return true;
- }
+ /// unaligned memory accesses. of the specified type. Returns whether it
+ /// is "fast" by reference in the second argument.
+ virtual bool allowsUnalignedMemoryAccesses(EVT VT, bool *Fast) const;
/// LowerOperation - Provide custom lowering hooks for some operations.
///
virtual const char *getTargetNodeName(unsigned Opcode) const;
/// getSetCCResultType - Return the value type to use for ISD::SETCC.
- virtual EVT getSetCCResultType(EVT VT) const;
+ virtual EVT getSetCCResultType(LLVMContext &Context, EVT VT) const;
/// computeMaskedBitsForTargetNode - Determine which of the bits specified
/// in Mask are known to be either zero or one and return them in the
/// error, this returns a register number of 0.
std::pair<unsigned, const TargetRegisterClass*>
getRegForInlineAsmConstraint(const std::string &Constraint,
- EVT VT) const;
+ MVT VT) 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 isTruncateFree(Type *Ty1, Type *Ty2) const;
virtual bool isTruncateFree(EVT VT1, EVT VT2) const;
+ virtual bool allowTruncateForTailCall(Type *Ty1, Type *Ty2) const;
+
/// isZExtFree - Return true if any actual instruction that defines a
/// value of type Ty1 implicit zero-extends the value to Ty2 in the result
/// register. This does not necessarily include registers defined in
/// result out to 64 bits.
virtual bool isZExtFree(Type *Ty1, Type *Ty2) const;
virtual bool isZExtFree(EVT VT1, EVT VT2) const;
+ virtual bool isZExtFree(SDValue Val, EVT VT2) const;
- /// isFMAFasterThanMulAndAdd - Return true if an FMA operation is faster than
- /// a pair of mul and add instructions. fmuladd intrinsics will be expanded to
- /// FMAs when this method returns true (and FMAs are legal), otherwise fmuladd
- /// is expanded to mul + add.
- virtual bool isFMAFasterThanMulAndAdd(EVT) const { return true; }
+ /// isFMAFasterThanFMulAndFAdd - Return true if an FMA operation is faster
+ /// than a pair of fmul and fadd instructions. fmuladd intrinsics will be
+ /// expanded to FMAs when this method returns true, otherwise fmuladd is
+ /// expanded to fmul + fadd.
+ virtual bool isFMAFasterThanFMulAndFAdd(EVT VT) const;
/// isNarrowingProfitable - Return true if it's profitable to narrow
/// operations of type VT1 to VT2. e.g. on x86, it's profitable to narrow
SDValue BuildFILD(SDValue Op, EVT SrcVT, SDValue Chain, SDValue StackSlot,
SelectionDAG &DAG) const;
+ virtual bool isNoopAddrSpaceCast(unsigned SrcAS, unsigned DestAS) const LLVM_OVERRIDE;
+
+ /// \brief Reset the operation actions based on target options.
+ virtual void resetOperationActions();
+
protected:
std::pair<const TargetRegisterClass*, uint8_t>
- findRepresentativeClass(EVT VT) const;
+ findRepresentativeClass(MVT VT) const;
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 DataLayout *TD;
- /// X86StackPtr - X86 physical register used as stack ptr.
- unsigned X86StackPtr;
+ /// Used to store the TargetOptions so that we don't waste time resetting
+ /// the operation actions unless we have to.
+ TargetOptions TO;
/// X86ScalarSSEf32, X86ScalarSSEf64 - Select between SSE or x87
/// floating point ops.
SDValue LowerCallResult(SDValue Chain, SDValue InFlag,
CallingConv::ID CallConv, bool isVarArg,
const SmallVectorImpl<ISD::InputArg> &Ins,
- DebugLoc dl, SelectionDAG &DAG,
+ SDLoc dl, SelectionDAG &DAG,
SmallVectorImpl<SDValue> &InVals) const;
SDValue LowerMemArgument(SDValue Chain,
CallingConv::ID CallConv,
const SmallVectorImpl<ISD::InputArg> &ArgInfo,
- DebugLoc dl, SelectionDAG &DAG,
+ SDLoc dl, SelectionDAG &DAG,
const CCValAssign &VA, MachineFrameInfo *MFI,
unsigned i) const;
SDValue LowerMemOpCallTo(SDValue Chain, SDValue StackPtr, SDValue Arg,
- DebugLoc dl, SelectionDAG &DAG,
+ SDLoc dl, SelectionDAG &DAG,
const CCValAssign &VA,
ISD::ArgFlagsTy Flags) const;
bool IsCalleePop(bool isVarArg, CallingConv::ID CallConv) const;
SDValue EmitTailCallLoadRetAddr(SelectionDAG &DAG, SDValue &OutRetAddr,
SDValue Chain, bool IsTailCall, bool Is64Bit,
- int FPDiff, DebugLoc dl) const;
+ int FPDiff, SDLoc dl) const;
unsigned GetAlignedArgumentStackSize(unsigned StackSize,
SelectionDAG &DAG) const;
bool isSigned,
bool isReplace) const;
- SDValue LowerAsSplatVectorLoad(SDValue SrcOp, EVT VT, DebugLoc dl,
- SelectionDAG &DAG) const;
SDValue LowerBUILD_VECTOR(SDValue Op, SelectionDAG &DAG) const;
+ SDValue LowerBUILD_VECTORvXi1(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerVECTOR_SHUFFLE(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerEXTRACT_VECTOR_ELT(SDValue Op, SelectionDAG &DAG) const;
- SDValue LowerEXTRACT_VECTOR_ELT_SSE4(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerINSERT_VECTOR_ELT(SDValue Op, SelectionDAG &DAG) const;
- SDValue LowerINSERT_VECTOR_ELT_SSE4(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerConstantPool(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerBlockAddress(SDValue Op, SelectionDAG &DAG) const;
- SDValue LowerGlobalAddress(const GlobalValue *GV, DebugLoc dl,
+ SDValue LowerGlobalAddress(const GlobalValue *GV, SDLoc dl,
int64_t Offset, SelectionDAG &DAG) const;
SDValue LowerGlobalAddress(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerGlobalTLSAddress(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerExternalSymbol(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerShiftParts(SDValue Op, SelectionDAG &DAG) const;
- SDValue LowerBITCAST(SDValue op, SelectionDAG &DAG) const;
SDValue LowerSINT_TO_FP(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerUINT_TO_FP(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerUINT_TO_FP_i64(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerUINT_TO_FP_i32(SDValue Op, SelectionDAG &DAG) const;
- SDValue lowerTRUNCATE(SDValue Op, SelectionDAG &DAG) const;
+ SDValue lowerUINT_TO_FP_vec(SDValue Op, SelectionDAG &DAG) const;
+ SDValue LowerTRUNCATE(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerFP_TO_SINT(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerFP_TO_UINT(SDValue Op, SelectionDAG &DAG) const;
- SDValue lowerFP_EXTEND(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerFABS(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerFNEG(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerFCOPYSIGN(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerToBT(SDValue And, ISD::CondCode CC,
- DebugLoc dl, SelectionDAG &DAG) const;
+ SDLoc dl, SelectionDAG &DAG) const;
SDValue LowerSETCC(SDValue Op, SelectionDAG &DAG) const;
- SDValue LowerVSETCC(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerSELECT(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerBRCOND(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerMEMSET(SDValue Op, SelectionDAG &DAG) const;
SDValue lowerEH_SJLJ_LONGJMP(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerINIT_TRAMPOLINE(SDValue Op, SelectionDAG &DAG) const;
SDValue LowerFLT_ROUNDS_(SDValue Op, SelectionDAG &DAG) const;
- SDValue LowerShift(SDValue Op, SelectionDAG &DAG) const;
-
SDValue LowerSIGN_EXTEND_INREG(SDValue Op, SelectionDAG &DAG) const;
- // Utility functions to help LowerVECTOR_SHUFFLE & LowerBUILD_VECTOR
- SDValue LowerVectorBroadcast(SDValue Op, SelectionDAG &DAG) const;
- SDValue NormalizeVectorShuffle(SDValue Op, SelectionDAG &DAG) const;
- SDValue buildFromShuffleMostly(SDValue Op, SelectionDAG &DAG) const;
-
- SDValue LowerVectorAllZeroTest(SDValue Op, SelectionDAG &DAG) const;
-
- SDValue lowerVectorIntExtend(SDValue Op, SelectionDAG &DAG) const;
-
virtual SDValue
LowerFormalArguments(SDValue Chain,
CallingConv::ID CallConv, bool isVarArg,
const SmallVectorImpl<ISD::InputArg> &Ins,
- DebugLoc dl, SelectionDAG &DAG,
+ SDLoc dl, SelectionDAG &DAG,
SmallVectorImpl<SDValue> &InVals) const;
virtual SDValue
LowerCall(CallLoweringInfo &CLI,
CallingConv::ID CallConv, bool isVarArg,
const SmallVectorImpl<ISD::OutputArg> &Outs,
const SmallVectorImpl<SDValue> &OutVals,
- DebugLoc dl, SelectionDAG &DAG) const;
+ SDLoc dl, SelectionDAG &DAG) const;
virtual bool isUsedByReturnOnly(SDNode *N, SDValue &Chain) const;
virtual bool mayBeEmittedAsTailCall(CallInst *CI) const;
- virtual EVT
- getTypeForExtArgOrReturn(LLVMContext &Context, EVT VT,
- ISD::NodeType ExtendKind) const;
+ virtual MVT
+ getTypeForExtArgOrReturn(MVT VT, ISD::NodeType ExtendKind) const;
virtual bool
CanLowerReturn(CallingConv::ID CallConv, MachineFunction &MF,
const SmallVectorImpl<ISD::OutputArg> &Outs,
LLVMContext &Context) const;
- /// Utility function to emit string processing sse4.2 instructions
- /// that return in xmm0.
- /// This takes the instruction to expand, the associated machine basic
- /// block, the number of args, and whether or not the second arg is
- /// in memory or not.
- MachineBasicBlock *EmitPCMP(MachineInstr *BInstr, MachineBasicBlock *BB,
- unsigned argNum, bool inMem) const;
-
- /// Utility functions to emit monitor and mwait instructions. These
- /// need to make sure that the arguments to the intrinsic are in the
- /// correct registers.
- MachineBasicBlock *EmitMonitor(MachineInstr *MI,
- MachineBasicBlock *BB) const;
- MachineBasicBlock *EmitMwait(MachineInstr *MI, MachineBasicBlock *BB) const;
+ virtual const uint16_t *getScratchRegisters(CallingConv::ID CC) const;
/// Utility function to emit atomic-load-arith operations (and, or, xor,
/// nand, max, min, umax, umin). It takes the corresponding instruction to
projects / oota-llvm.git / blobdiff