1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-22 02:33:06 +01:00
llvm-mirror/include/llvm/CodeGen/FastISel.h
Nikita Popov 5794a3620e [FastISel] Remove kill tracking
This is a followup to D98145: As far as I know, tracking of kill
flags in FastISel is just a compile-time optimization. However,
I'm not actually seeing any compile-time regression when removing
the tracking. This probably used to be more important in the past,
before FastRA was switched to allocate instructions in reverse
order, which means that it discovers kills as a matter of course.

As such, the kill tracking doesn't really seem to serve a purpose
anymore, and just adds additional complexity and potential for
errors. This patch removes it entirely. The primary changes are
dropping the hasTrivialKill() method and removing the kill
arguments from the emitFast methods. The rest is mechanical fixup.

Differential Revision: https://reviews.llvm.org/D98294
2021-04-03 15:50:13 +02:00

562 lines
21 KiB
C++

//===- FastISel.h - Definition of the FastISel class ------------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
///
/// \file
/// This file defines the FastISel class.
///
//===----------------------------------------------------------------------===//
#ifndef LLVM_CODEGEN_FASTISEL_H
#define LLVM_CODEGEN_FASTISEL_H
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/TargetLowering.h"
#include "llvm/IR/Attributes.h"
#include "llvm/IR/CallingConv.h"
#include "llvm/IR/DebugLoc.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/InstrTypes.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/Support/MachineValueType.h"
#include <algorithm>
#include <cstdint>
#include <utility>
namespace llvm {
class AllocaInst;
class BasicBlock;
class CallInst;
class Constant;
class ConstantFP;
class DataLayout;
class FunctionLoweringInfo;
class LoadInst;
class MachineConstantPool;
class MachineFrameInfo;
class MachineFunction;
class MachineInstr;
class MachineMemOperand;
class MachineOperand;
class MachineRegisterInfo;
class MCContext;
class MCInstrDesc;
class MCSymbol;
class TargetInstrInfo;
class TargetLibraryInfo;
class TargetMachine;
class TargetRegisterClass;
class TargetRegisterInfo;
class Type;
class User;
class Value;
/// This is a fast-path instruction selection class that generates poor
/// code and doesn't support illegal types or non-trivial lowering, but runs
/// quickly.
class FastISel {
public:
using ArgListEntry = TargetLoweringBase::ArgListEntry;
using ArgListTy = TargetLoweringBase::ArgListTy;
struct CallLoweringInfo {
Type *RetTy = nullptr;
bool RetSExt : 1;
bool RetZExt : 1;
bool IsVarArg : 1;
bool IsInReg : 1;
bool DoesNotReturn : 1;
bool IsReturnValueUsed : 1;
bool IsPatchPoint : 1;
// IsTailCall Should be modified by implementations of FastLowerCall
// that perform tail call conversions.
bool IsTailCall = false;
unsigned NumFixedArgs = -1;
CallingConv::ID CallConv = CallingConv::C;
const Value *Callee = nullptr;
MCSymbol *Symbol = nullptr;
ArgListTy Args;
const CallBase *CB = nullptr;
MachineInstr *Call = nullptr;
Register ResultReg;
unsigned NumResultRegs = 0;
SmallVector<Value *, 16> OutVals;
SmallVector<ISD::ArgFlagsTy, 16> OutFlags;
SmallVector<Register, 16> OutRegs;
SmallVector<ISD::InputArg, 4> Ins;
SmallVector<Register, 4> InRegs;
CallLoweringInfo()
: RetSExt(false), RetZExt(false), IsVarArg(false), IsInReg(false),
DoesNotReturn(false), IsReturnValueUsed(true), IsPatchPoint(false) {}
CallLoweringInfo &setCallee(Type *ResultTy, FunctionType *FuncTy,
const Value *Target, ArgListTy &&ArgsList,
const CallBase &Call) {
RetTy = ResultTy;
Callee = Target;
IsInReg = Call.hasRetAttr(Attribute::InReg);
DoesNotReturn = Call.doesNotReturn();
IsVarArg = FuncTy->isVarArg();
IsReturnValueUsed = !Call.use_empty();
RetSExt = Call.hasRetAttr(Attribute::SExt);
RetZExt = Call.hasRetAttr(Attribute::ZExt);
CallConv = Call.getCallingConv();
Args = std::move(ArgsList);
NumFixedArgs = FuncTy->getNumParams();
CB = &Call;
return *this;
}
CallLoweringInfo &setCallee(Type *ResultTy, FunctionType *FuncTy,
MCSymbol *Target, ArgListTy &&ArgsList,
const CallBase &Call,
unsigned FixedArgs = ~0U) {
RetTy = ResultTy;
Callee = Call.getCalledOperand();
Symbol = Target;
IsInReg = Call.hasRetAttr(Attribute::InReg);
DoesNotReturn = Call.doesNotReturn();
IsVarArg = FuncTy->isVarArg();
IsReturnValueUsed = !Call.use_empty();
RetSExt = Call.hasRetAttr(Attribute::SExt);
RetZExt = Call.hasRetAttr(Attribute::ZExt);
CallConv = Call.getCallingConv();
Args = std::move(ArgsList);
NumFixedArgs = (FixedArgs == ~0U) ? FuncTy->getNumParams() : FixedArgs;
CB = &Call;
return *this;
}
CallLoweringInfo &setCallee(CallingConv::ID CC, Type *ResultTy,
const Value *Target, ArgListTy &&ArgsList,
unsigned FixedArgs = ~0U) {
RetTy = ResultTy;
Callee = Target;
CallConv = CC;
Args = std::move(ArgsList);
NumFixedArgs = (FixedArgs == ~0U) ? Args.size() : FixedArgs;
return *this;
}
CallLoweringInfo &setCallee(const DataLayout &DL, MCContext &Ctx,
CallingConv::ID CC, Type *ResultTy,
StringRef Target, ArgListTy &&ArgsList,
unsigned FixedArgs = ~0U);
CallLoweringInfo &setCallee(CallingConv::ID CC, Type *ResultTy,
MCSymbol *Target, ArgListTy &&ArgsList,
unsigned FixedArgs = ~0U) {
RetTy = ResultTy;
Symbol = Target;
CallConv = CC;
Args = std::move(ArgsList);
NumFixedArgs = (FixedArgs == ~0U) ? Args.size() : FixedArgs;
return *this;
}
CallLoweringInfo &setTailCall(bool Value = true) {
IsTailCall = Value;
return *this;
}
CallLoweringInfo &setIsPatchPoint(bool Value = true) {
IsPatchPoint = Value;
return *this;
}
ArgListTy &getArgs() { return Args; }
void clearOuts() {
OutVals.clear();
OutFlags.clear();
OutRegs.clear();
}
void clearIns() {
Ins.clear();
InRegs.clear();
}
};
protected:
DenseMap<const Value *, Register> LocalValueMap;
FunctionLoweringInfo &FuncInfo;
MachineFunction *MF;
MachineRegisterInfo &MRI;
MachineFrameInfo &MFI;
MachineConstantPool &MCP;
DebugLoc DbgLoc;
const TargetMachine &TM;
const DataLayout &DL;
const TargetInstrInfo &TII;
const TargetLowering &TLI;
const TargetRegisterInfo &TRI;
const TargetLibraryInfo *LibInfo;
bool SkipTargetIndependentISel;
/// The position of the last instruction for materializing constants
/// for use in the current block. It resets to EmitStartPt when it makes sense
/// (for example, it's usually profitable to avoid function calls between the
/// definition and the use)
MachineInstr *LastLocalValue;
/// The top most instruction in the current block that is allowed for
/// emitting local variables. LastLocalValue resets to EmitStartPt when it
/// makes sense (for example, on function calls)
MachineInstr *EmitStartPt;
public:
virtual ~FastISel();
/// Return the position of the last instruction emitted for
/// materializing constants for use in the current block.
MachineInstr *getLastLocalValue() { return LastLocalValue; }
/// Update the position of the last instruction emitted for
/// materializing constants for use in the current block.
void setLastLocalValue(MachineInstr *I) {
EmitStartPt = I;
LastLocalValue = I;
}
/// Set the current block to which generated machine instructions will
/// be appended.
void startNewBlock();
/// Flush the local value map.
void finishBasicBlock();
/// Return current debug location information.
DebugLoc getCurDebugLoc() const { return DbgLoc; }
/// Do "fast" instruction selection for function arguments and append
/// the machine instructions to the current block. Returns true when
/// successful.
bool lowerArguments();
/// Do "fast" instruction selection for the given LLVM IR instruction
/// and append the generated machine instructions to the current block.
/// Returns true if selection was successful.
bool selectInstruction(const Instruction *I);
/// Do "fast" instruction selection for the given LLVM IR operator
/// (Instruction or ConstantExpr), and append generated machine instructions
/// to the current block. Return true if selection was successful.
bool selectOperator(const User *I, unsigned Opcode);
/// Create a virtual register and arrange for it to be assigned the
/// value for the given LLVM value.
Register getRegForValue(const Value *V);
/// Look up the value to see if its value is already cached in a
/// register. It may be defined by instructions across blocks or defined
/// locally.
Register lookUpRegForValue(const Value *V);
/// This is a wrapper around getRegForValue that also takes care of
/// truncating or sign-extending the given getelementptr index value.
Register getRegForGEPIndex(const Value *Idx);
/// We're checking to see if we can fold \p LI into \p FoldInst. Note
/// that we could have a sequence where multiple LLVM IR instructions are
/// folded into the same machineinstr. For example we could have:
///
/// A: x = load i32 *P
/// B: y = icmp A, 42
/// C: br y, ...
///
/// In this scenario, \p LI is "A", and \p FoldInst is "C". We know about "B"
/// (and any other folded instructions) because it is between A and C.
///
/// If we succeed folding, return true.
bool tryToFoldLoad(const LoadInst *LI, const Instruction *FoldInst);
/// The specified machine instr operand is a vreg, and that vreg is
/// being provided by the specified load instruction. If possible, try to
/// fold the load as an operand to the instruction, returning true if
/// possible.
///
/// This method should be implemented by targets.
virtual bool tryToFoldLoadIntoMI(MachineInstr * /*MI*/, unsigned /*OpNo*/,
const LoadInst * /*LI*/) {
return false;
}
/// Reset InsertPt to prepare for inserting instructions into the
/// current block.
void recomputeInsertPt();
/// Remove all dead instructions between the I and E.
void removeDeadCode(MachineBasicBlock::iterator I,
MachineBasicBlock::iterator E);
using SavePoint = MachineBasicBlock::iterator;
/// Prepare InsertPt to begin inserting instructions into the local
/// value area and return the old insert position.
SavePoint enterLocalValueArea();
/// Reset InsertPt to the given old insert position.
void leaveLocalValueArea(SavePoint Old);
protected:
explicit FastISel(FunctionLoweringInfo &FuncInfo,
const TargetLibraryInfo *LibInfo,
bool SkipTargetIndependentISel = false);
/// This method is called by target-independent code when the normal
/// FastISel process fails to select an instruction. This gives targets a
/// chance to emit code for anything that doesn't fit into FastISel's
/// framework. It returns true if it was successful.
virtual bool fastSelectInstruction(const Instruction *I) = 0;
/// This method is called by target-independent code to do target-
/// specific argument lowering. It returns true if it was successful.
virtual bool fastLowerArguments();
/// This method is called by target-independent code to do target-
/// specific call lowering. It returns true if it was successful.
virtual bool fastLowerCall(CallLoweringInfo &CLI);
/// This method is called by target-independent code to do target-
/// specific intrinsic lowering. It returns true if it was successful.
virtual bool fastLowerIntrinsicCall(const IntrinsicInst *II);
/// This method is called by target-independent code to request that an
/// instruction with the given type and opcode be emitted.
virtual unsigned fastEmit_(MVT VT, MVT RetVT, unsigned Opcode);
/// This method is called by target-independent code to request that an
/// instruction with the given type, opcode, and register operand be emitted.
virtual unsigned fastEmit_r(MVT VT, MVT RetVT, unsigned Opcode, unsigned Op0);
/// This method is called by target-independent code to request that an
/// instruction with the given type, opcode, and register operands be emitted.
virtual unsigned fastEmit_rr(MVT VT, MVT RetVT, unsigned Opcode, unsigned Op0,
unsigned Op1);
/// This method is called by target-independent code to request that an
/// instruction with the given type, opcode, and register and immediate
/// operands be emitted.
virtual unsigned fastEmit_ri(MVT VT, MVT RetVT, unsigned Opcode, unsigned Op0,
uint64_t Imm);
/// This method is a wrapper of fastEmit_ri.
///
/// It first tries to emit an instruction with an immediate operand using
/// fastEmit_ri. If that fails, it materializes the immediate into a register
/// and try fastEmit_rr instead.
Register fastEmit_ri_(MVT VT, unsigned Opcode, unsigned Op0, uint64_t Imm,
MVT ImmType);
/// This method is called by target-independent code to request that an
/// instruction with the given type, opcode, and immediate operand be emitted.
virtual unsigned fastEmit_i(MVT VT, MVT RetVT, unsigned Opcode, uint64_t Imm);
/// This method is called by target-independent code to request that an
/// instruction with the given type, opcode, and floating-point immediate
/// operand be emitted.
virtual unsigned fastEmit_f(MVT VT, MVT RetVT, unsigned Opcode,
const ConstantFP *FPImm);
/// Emit a MachineInstr with no operands and a result register in the
/// given register class.
Register fastEmitInst_(unsigned MachineInstOpcode,
const TargetRegisterClass *RC);
/// Emit a MachineInstr with one register operand and a result register
/// in the given register class.
Register fastEmitInst_r(unsigned MachineInstOpcode,
const TargetRegisterClass *RC, unsigned Op0);
/// Emit a MachineInstr with two register operands and a result
/// register in the given register class.
Register fastEmitInst_rr(unsigned MachineInstOpcode,
const TargetRegisterClass *RC, unsigned Op0,
unsigned Op1);
/// Emit a MachineInstr with three register operands and a result
/// register in the given register class.
Register fastEmitInst_rrr(unsigned MachineInstOpcode,
const TargetRegisterClass *RC, unsigned Op0,
unsigned Op1, unsigned Op2);
/// Emit a MachineInstr with a register operand, an immediate, and a
/// result register in the given register class.
Register fastEmitInst_ri(unsigned MachineInstOpcode,
const TargetRegisterClass *RC, unsigned Op0,
uint64_t Imm);
/// Emit a MachineInstr with one register operand and two immediate
/// operands.
Register fastEmitInst_rii(unsigned MachineInstOpcode,
const TargetRegisterClass *RC, unsigned Op0,
uint64_t Imm1, uint64_t Imm2);
/// Emit a MachineInstr with a floating point immediate, and a result
/// register in the given register class.
Register fastEmitInst_f(unsigned MachineInstOpcode,
const TargetRegisterClass *RC,
const ConstantFP *FPImm);
/// Emit a MachineInstr with two register operands, an immediate, and a
/// result register in the given register class.
Register fastEmitInst_rri(unsigned MachineInstOpcode,
const TargetRegisterClass *RC, unsigned Op0,
unsigned Op1, uint64_t Imm);
/// Emit a MachineInstr with a single immediate operand, and a result
/// register in the given register class.
Register fastEmitInst_i(unsigned MachineInstOpcode,
const TargetRegisterClass *RC, uint64_t Imm);
/// Emit a MachineInstr for an extract_subreg from a specified index of
/// a superregister to a specified type.
Register fastEmitInst_extractsubreg(MVT RetVT, unsigned Op0, uint32_t Idx);
/// Emit MachineInstrs to compute the value of Op with all but the
/// least significant bit set to zero.
Register fastEmitZExtFromI1(MVT VT, unsigned Op0);
/// Emit an unconditional branch to the given block, unless it is the
/// immediate (fall-through) successor, and update the CFG.
void fastEmitBranch(MachineBasicBlock *MSucc, const DebugLoc &DbgLoc);
/// Emit an unconditional branch to \p FalseMBB, obtains the branch weight
/// and adds TrueMBB and FalseMBB to the successor list.
void finishCondBranch(const BasicBlock *BranchBB, MachineBasicBlock *TrueMBB,
MachineBasicBlock *FalseMBB);
/// Update the value map to include the new mapping for this
/// instruction, or insert an extra copy to get the result in a previous
/// determined register.
///
/// NOTE: This is only necessary because we might select a block that uses a
/// value before we select the block that defines the value. It might be
/// possible to fix this by selecting blocks in reverse postorder.
void updateValueMap(const Value *I, Register Reg, unsigned NumRegs = 1);
Register createResultReg(const TargetRegisterClass *RC);
/// Try to constrain Op so that it is usable by argument OpNum of the
/// provided MCInstrDesc. If this fails, create a new virtual register in the
/// correct class and COPY the value there.
Register constrainOperandRegClass(const MCInstrDesc &II, Register Op,
unsigned OpNum);
/// Emit a constant in a register using target-specific logic, such as
/// constant pool loads.
virtual unsigned fastMaterializeConstant(const Constant *C) { return 0; }
/// Emit an alloca address in a register using target-specific logic.
virtual unsigned fastMaterializeAlloca(const AllocaInst *C) { return 0; }
/// Emit the floating-point constant +0.0 in a register using target-
/// specific logic.
virtual unsigned fastMaterializeFloatZero(const ConstantFP *CF) {
return 0;
}
/// Check if \c Add is an add that can be safely folded into \c GEP.
///
/// \c Add can be folded into \c GEP if:
/// - \c Add is an add,
/// - \c Add's size matches \c GEP's,
/// - \c Add is in the same basic block as \c GEP, and
/// - \c Add has a constant operand.
bool canFoldAddIntoGEP(const User *GEP, const Value *Add);
/// Create a machine mem operand from the given instruction.
MachineMemOperand *createMachineMemOperandFor(const Instruction *I) const;
CmpInst::Predicate optimizeCmpPredicate(const CmpInst *CI) const;
bool lowerCallTo(const CallInst *CI, MCSymbol *Symbol, unsigned NumArgs);
bool lowerCallTo(const CallInst *CI, const char *SymName,
unsigned NumArgs);
bool lowerCallTo(CallLoweringInfo &CLI);
bool lowerCall(const CallInst *I);
/// Select and emit code for a binary operator instruction, which has
/// an opcode which directly corresponds to the given ISD opcode.
bool selectBinaryOp(const User *I, unsigned ISDOpcode);
bool selectFNeg(const User *I, const Value *In);
bool selectGetElementPtr(const User *I);
bool selectStackmap(const CallInst *I);
bool selectPatchpoint(const CallInst *I);
bool selectCall(const User *I);
bool selectIntrinsicCall(const IntrinsicInst *II);
bool selectBitCast(const User *I);
bool selectFreeze(const User *I);
bool selectCast(const User *I, unsigned Opcode);
bool selectExtractValue(const User *U);
bool selectXRayCustomEvent(const CallInst *II);
bool selectXRayTypedEvent(const CallInst *II);
bool shouldOptForSize(const MachineFunction *MF) const {
// TODO: Implement PGSO.
return MF->getFunction().hasOptSize();
}
private:
/// Handle PHI nodes in successor blocks.
///
/// Emit code to ensure constants are copied into registers when needed.
/// Remember the virtual registers that need to be added to the Machine PHI
/// nodes as input. We cannot just directly add them, because expansion might
/// result in multiple MBB's for one BB. As such, the start of the BB might
/// correspond to a different MBB than the end.
bool handlePHINodesInSuccessorBlocks(const BasicBlock *LLVMBB);
/// Helper for materializeRegForValue to materialize a constant in a
/// target-independent way.
Register materializeConstant(const Value *V, MVT VT);
/// Helper for getRegForVale. This function is called when the value
/// isn't already available in a register and must be materialized with new
/// instructions.
Register materializeRegForValue(const Value *V, MVT VT);
/// Clears LocalValueMap and moves the area for the new local variables
/// to the beginning of the block. It helps to avoid spilling cached variables
/// across heavy instructions like calls.
void flushLocalValueMap();
/// Removes dead local value instructions after SavedLastLocalvalue.
void removeDeadLocalValueCode(MachineInstr *SavedLastLocalValue);
/// Insertion point before trying to select the current instruction.
MachineBasicBlock::iterator SavedInsertPt;
/// Add a stackmap or patchpoint intrinsic call's live variable
/// operands to a stackmap or patchpoint machine instruction.
bool addStackMapLiveVars(SmallVectorImpl<MachineOperand> &Ops,
const CallInst *CI, unsigned StartIdx);
bool lowerCallOperands(const CallInst *CI, unsigned ArgIdx, unsigned NumArgs,
const Value *Callee, bool ForceRetVoidTy,
CallLoweringInfo &CLI);
};
} // end namespace llvm
#endif // LLVM_CODEGEN_FASTISEL_H