//===-- InstSelectSimple.cpp - A simple instruction selector for x86 ------===// // // This file defines a simple peephole instruction selector for the x86 platform // //===----------------------------------------------------------------------===// #include "X86.h" #include "X86InstrInfo.h" #include "X86InstrBuilder.h" #include "llvm/Function.h" #include "llvm/iTerminators.h" #include "llvm/iOperators.h" #include "llvm/iOther.h" #include "llvm/iPHINode.h" #include "llvm/iMemory.h" #include "llvm/Type.h" #include "llvm/DerivedTypes.h" #include "llvm/Constants.h" #include "llvm/Pass.h" #include "llvm/CodeGen/MachineFunction.h" #include "llvm/CodeGen/MachineInstrBuilder.h" #include "llvm/CodeGen/SSARegMap.h" #include "llvm/CodeGen/FunctionFrameInfo.h" #include "llvm/Target/TargetMachine.h" #include "llvm/Support/InstVisitor.h" #include "llvm/Target/MRegisterInfo.h" #include /// BMI - A special BuildMI variant that takes an iterator to insert the /// instruction at as well as a basic block. /// this is the version for when you have a destination register in mind. inline static MachineInstrBuilder BMI(MachineBasicBlock *MBB, MachineBasicBlock::iterator &I, MachineOpCode Opcode, unsigned NumOperands, unsigned DestReg) { assert(I >= MBB->begin() && I <= MBB->end() && "Bad iterator!"); MachineInstr *MI = new MachineInstr(Opcode, NumOperands+1, true, true); I = MBB->insert(I, MI)+1; return MachineInstrBuilder(MI).addReg(DestReg, MOTy::Def); } /// BMI - A special BuildMI variant that takes an iterator to insert the /// instruction at as well as a basic block. inline static MachineInstrBuilder BMI(MachineBasicBlock *MBB, MachineBasicBlock::iterator &I, MachineOpCode Opcode, unsigned NumOperands) { assert(I > MBB->begin() && I <= MBB->end() && "Bad iterator!"); MachineInstr *MI = new MachineInstr(Opcode, NumOperands, true, true); I = MBB->insert(I, MI)+1; return MachineInstrBuilder(MI); } namespace { struct ISel : public FunctionPass, InstVisitor { TargetMachine &TM; MachineFunction *F; // The function we are compiling into MachineBasicBlock *BB; // The current MBB we are compiling unsigned CurReg; std::map RegMap; // Mapping between Val's and SSA Regs // MBBMap - Mapping between LLVM BB -> Machine BB std::map MBBMap; ISel(TargetMachine &tm) : TM(tm), F(0), BB(0), CurReg(MRegisterInfo::FirstVirtualRegister) {} /// runOnFunction - Top level implementation of instruction selection for /// the entire function. /// bool runOnFunction(Function &Fn) { F = &MachineFunction::construct(&Fn, TM); // Create all of the machine basic blocks for the function... for (Function::iterator I = Fn.begin(), E = Fn.end(); I != E; ++I) F->getBasicBlockList().push_back(MBBMap[I] = new MachineBasicBlock(I)); BB = &F->front(); LoadArgumentsToVirtualRegs(Fn); // Instruction select everything except PHI nodes visit(Fn); // Select the PHI nodes SelectPHINodes(); RegMap.clear(); MBBMap.clear(); CurReg = MRegisterInfo::FirstVirtualRegister; F = 0; return false; // We never modify the LLVM itself. } virtual const char *getPassName() const { return "X86 Simple Instruction Selection"; } /// visitBasicBlock - This method is called when we are visiting a new basic /// block. This simply creates a new MachineBasicBlock to emit code into /// and adds it to the current MachineFunction. Subsequent visit* for /// instructions will be invoked for all instructions in the basic block. /// void visitBasicBlock(BasicBlock &LLVM_BB) { BB = MBBMap[&LLVM_BB]; } /// LoadArgumentsToVirtualRegs - Load all of the arguments to this function /// from the stack into virtual registers. /// void LoadArgumentsToVirtualRegs(Function &F); /// SelectPHINodes - Insert machine code to generate phis. This is tricky /// because we have to generate our sources into the source basic blocks, /// not the current one. /// void SelectPHINodes(); // Visitation methods for various instructions. These methods simply emit // fixed X86 code for each instruction. // // Control flow operators void visitReturnInst(ReturnInst &RI); void visitBranchInst(BranchInst &BI); void visitCallInst(CallInst &I); // Arithmetic operators void visitSimpleBinary(BinaryOperator &B, unsigned OpcodeClass); void visitAdd(BinaryOperator &B) { visitSimpleBinary(B, 0); } void visitSub(BinaryOperator &B) { visitSimpleBinary(B, 1); } void doMultiply(MachineBasicBlock *MBB, MachineBasicBlock::iterator &MBBI, unsigned destReg, const Type *resultType, unsigned op0Reg, unsigned op1Reg); void visitMul(BinaryOperator &B); void visitDiv(BinaryOperator &B) { visitDivRem(B); } void visitRem(BinaryOperator &B) { visitDivRem(B); } void visitDivRem(BinaryOperator &B); // Bitwise operators void visitAnd(BinaryOperator &B) { visitSimpleBinary(B, 2); } void visitOr (BinaryOperator &B) { visitSimpleBinary(B, 3); } void visitXor(BinaryOperator &B) { visitSimpleBinary(B, 4); } // Binary comparison operators void visitSetCCInst(SetCondInst &I, unsigned OpNum); void visitSetEQ(SetCondInst &I) { visitSetCCInst(I, 0); } void visitSetNE(SetCondInst &I) { visitSetCCInst(I, 1); } void visitSetLT(SetCondInst &I) { visitSetCCInst(I, 2); } void visitSetGT(SetCondInst &I) { visitSetCCInst(I, 3); } void visitSetLE(SetCondInst &I) { visitSetCCInst(I, 4); } void visitSetGE(SetCondInst &I) { visitSetCCInst(I, 5); } // Memory Instructions void visitLoadInst(LoadInst &I); void visitStoreInst(StoreInst &I); void visitGetElementPtrInst(GetElementPtrInst &I); void visitAllocaInst(AllocaInst &I); // We assume that by this point, malloc instructions have been // lowered to calls, and dlsym will magically find malloc for us. void visitMallocInst(MallocInst &I) { visitInstruction (I); } void visitFreeInst(FreeInst &I) { visitInstruction(I); } // Other operators void visitShiftInst(ShiftInst &I); void visitPHINode(PHINode &I) {} // PHI nodes handled by second pass void visitCastInst(CastInst &I); void visitInstruction(Instruction &I) { std::cerr << "Cannot instruction select: " << I; abort(); } /// promote32 - Make a value 32-bits wide, and put it somewhere. void promote32 (const unsigned targetReg, Value *v); // emitGEPOperation - Common code shared between visitGetElementPtrInst and // constant expression GEP support. // void emitGEPOperation(MachineBasicBlock *BB, MachineBasicBlock::iterator&IP, Value *Src, User::op_iterator IdxBegin, User::op_iterator IdxEnd, unsigned TargetReg); /// copyConstantToRegister - Output the instructions required to put the /// specified constant into the specified register. /// void copyConstantToRegister(MachineBasicBlock *MBB, MachineBasicBlock::iterator &MBBI, Constant *C, unsigned Reg); /// makeAnotherReg - This method returns the next register number /// we haven't yet used. unsigned makeAnotherReg(const Type *Ty) { // Add the mapping of regnumber => reg class to MachineFunction const TargetRegisterClass *RC = TM.getRegisterInfo()->getRegClassForType(Ty); F->getSSARegMap()->addRegMap(CurReg, RC); return CurReg++; } /// getReg - This method turns an LLVM value into a register number. This /// is guaranteed to produce the same register number for a particular value /// every time it is queried. /// unsigned getReg(Value &V) { return getReg(&V); } // Allow references unsigned getReg(Value *V) { // Just append to the end of the current bb. MachineBasicBlock::iterator It = BB->end(); return getReg(V, BB, It); } unsigned getReg(Value *V, MachineBasicBlock *MBB, MachineBasicBlock::iterator &IPt) { unsigned &Reg = RegMap[V]; if (Reg == 0) { Reg = makeAnotherReg(V->getType()); RegMap[V] = Reg; } // If this operand is a constant, emit the code to copy the constant into // the register here... // if (Constant *C = dyn_cast(V)) { copyConstantToRegister(MBB, IPt, C, Reg); RegMap.erase(V); // Assign a new name to this constant if ref'd again } else if (GlobalValue *GV = dyn_cast(V)) { // Move the address of the global into the register BMI(MBB, IPt, X86::MOVir32, 1, Reg).addReg(GV); RegMap.erase(V); // Assign a new name to this address if ref'd again } return Reg; } }; } /// TypeClass - Used by the X86 backend to group LLVM types by their basic X86 /// Representation. /// enum TypeClass { cByte, cShort, cInt, cFP, cLong }; /// getClass - Turn a primitive type into a "class" number which is based on the /// size of the type, and whether or not it is floating point. /// static inline TypeClass getClass(const Type *Ty) { switch (Ty->getPrimitiveID()) { case Type::SByteTyID: case Type::UByteTyID: return cByte; // Byte operands are class #0 case Type::ShortTyID: case Type::UShortTyID: return cShort; // Short operands are class #1 case Type::IntTyID: case Type::UIntTyID: case Type::PointerTyID: return cInt; // Int's and pointers are class #2 case Type::FloatTyID: case Type::DoubleTyID: return cFP; // Floating Point is #3 case Type::LongTyID: case Type::ULongTyID: //return cLong; // Longs are class #3 return cInt; // FIXME: LONGS ARE TREATED AS INTS! default: assert(0 && "Invalid type to getClass!"); return cByte; // not reached } } // getClassB - Just like getClass, but treat boolean values as bytes. static inline TypeClass getClassB(const Type *Ty) { if (Ty == Type::BoolTy) return cByte; return getClass(Ty); } /// copyConstantToRegister - Output the instructions required to put the /// specified constant into the specified register. /// void ISel::copyConstantToRegister(MachineBasicBlock *MBB, MachineBasicBlock::iterator &IP, Constant *C, unsigned R) { if (ConstantExpr *CE = dyn_cast(C)) { if (CE->getOpcode() == Instruction::GetElementPtr) { emitGEPOperation(MBB, IP, CE->getOperand(0), CE->op_begin()+1, CE->op_end(), R); return; } std::cerr << "Offending expr: " << C << "\n"; assert(0 && "Constant expressions not yet handled!\n"); } if (C->getType()->isIntegral()) { unsigned Class = getClassB(C->getType()); assert(Class <= cInt && "Type not handled yet!"); static const unsigned IntegralOpcodeTab[] = { X86::MOVir8, X86::MOVir16, X86::MOVir32 }; if (C->getType() == Type::BoolTy) { BMI(MBB, IP, X86::MOVir8, 1, R).addZImm(C == ConstantBool::True); } else if (C->getType()->isSigned()) { ConstantSInt *CSI = cast(C); BMI(MBB, IP, IntegralOpcodeTab[Class], 1, R).addSImm(CSI->getValue()); } else { ConstantUInt *CUI = cast(C); BMI(MBB, IP, IntegralOpcodeTab[Class], 1, R).addZImm(CUI->getValue()); } } else if (ConstantFP *CFP = dyn_cast(C)) { double Value = CFP->getValue(); if (Value == +0.0) BMI(MBB, IP, X86::FLD0, 0, R); else if (Value == +1.0) BMI(MBB, IP, X86::FLD1, 0, R); else { std::cerr << "Cannot load constant '" << Value << "'!\n"; assert(0); } } else if (isa(C)) { // Copy zero (null pointer) to the register. BMI(MBB, IP, X86::MOVir32, 1, R).addZImm(0); } else if (ConstantPointerRef *CPR = dyn_cast(C)) { unsigned SrcReg = getReg(CPR->getValue(), MBB, IP); BMI(MBB, IP, X86::MOVrr32, 1, R).addReg(SrcReg); } else { std::cerr << "Offending constant: " << C << "\n"; assert(0 && "Type not handled yet!"); } } /// LoadArgumentsToVirtualRegs - Load all of the arguments to this function from /// the stack into virtual registers. /// void ISel::LoadArgumentsToVirtualRegs(Function &Fn) { // Emit instructions to load the arguments... On entry to a function on the // X86, the stack frame looks like this: // // [ESP] -- return address // [ESP + 4] -- first argument (leftmost lexically) if four bytes in size // [ESP + 8] -- second argument, if four bytes in size // ... // unsigned ArgOffset = 0; FunctionFrameInfo *FFI = F->getFrameInfo(); for (Function::aiterator I = Fn.abegin(), E = Fn.aend(); I != E; ++I) { unsigned Reg = getReg(*I); ArgOffset += 4; // Each argument takes at least 4 bytes on the stack... int FI; // Frame object index switch (getClassB(I->getType())) { case cByte: FI = FFI->CreateFixedObject(1, ArgOffset); addFrameReference(BuildMI(BB, X86::MOVmr8, 4, Reg), FI); break; case cShort: FI = FFI->CreateFixedObject(2, ArgOffset); addFrameReference(BuildMI(BB, X86::MOVmr16, 4, Reg), FI); break; case cInt: FI = FFI->CreateFixedObject(4, ArgOffset); addFrameReference(BuildMI(BB, X86::MOVmr32, 4, Reg), FI); break; case cFP: unsigned Opcode; if (I->getType() == Type::FloatTy) { Opcode = X86::FLDr32; FI = FFI->CreateFixedObject(4, ArgOffset); } else { Opcode = X86::FLDr64; ArgOffset += 4; // doubles require 4 additional bytes FI = FFI->CreateFixedObject(8, ArgOffset); } addFrameReference(BuildMI(BB, Opcode, 4, Reg), FI); break; default: assert(0 && "Unhandled argument type!"); } } } /// SelectPHINodes - Insert machine code to generate phis. This is tricky /// because we have to generate our sources into the source basic blocks, not /// the current one. /// void ISel::SelectPHINodes() { const Function &LF = *F->getFunction(); // The LLVM function... for (Function::const_iterator I = LF.begin(), E = LF.end(); I != E; ++I) { const BasicBlock *BB = I; MachineBasicBlock *MBB = MBBMap[I]; // Loop over all of the PHI nodes in the LLVM basic block... unsigned NumPHIs = 0; for (BasicBlock::const_iterator I = BB->begin(); PHINode *PN = (PHINode*)dyn_cast(&*I); ++I) { // Create a new machine instr PHI node, and insert it. MachineInstr *MI = BuildMI(X86::PHI, PN->getNumOperands(), getReg(*PN)); MBB->insert(MBB->begin()+NumPHIs++, MI); // Insert it at the top of the BB for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i) { MachineBasicBlock *PredMBB = MBBMap[PN->getIncomingBlock(i)]; // Get the incoming value into a virtual register. If it is not already // available in a virtual register, insert the computation code into // PredMBB // // FIXME: This should insert the code into the BOTTOM of the block, not // the top of the block. This just makes for huge live ranges... MachineBasicBlock::iterator PI = PredMBB->begin(); while ((*PI)->getOpcode() == X86::PHI) ++PI; MI->addRegOperand(getReg(PN->getIncomingValue(i), PredMBB, PI)); MI->addMachineBasicBlockOperand(PredMBB); } } } } /// SetCC instructions - Here we just emit boilerplate code to set a byte-sized /// register, then move it to wherever the result should be. /// We handle FP setcc instructions by pushing them, doing a /// compare-and-pop-twice, and then copying the concodes to the main /// processor's concodes (I didn't make this up, it's in the Intel manual) /// void ISel::visitSetCCInst(SetCondInst &I, unsigned OpNum) { // The arguments are already supposed to be of the same type. const Type *CompTy = I.getOperand(0)->getType(); unsigned reg1 = getReg(I.getOperand(0)); unsigned reg2 = getReg(I.getOperand(1)); unsigned Class = getClass(CompTy); switch (Class) { // Emit: cmp , (do the comparison). We can // compare 8-bit with 8-bit, 16-bit with 16-bit, 32-bit with // 32-bit. case cByte: BuildMI (BB, X86::CMPrr8, 2).addReg (reg1).addReg (reg2); break; case cShort: BuildMI (BB, X86::CMPrr16, 2).addReg (reg1).addReg (reg2); break; case cInt: BuildMI (BB, X86::CMPrr32, 2).addReg (reg1).addReg (reg2); break; #if 0 // Push the variables on the stack with fldl opcodes. // FIXME: assuming var1, var2 are in memory, if not, spill to // stack first case cFP: // Floats BuildMI (BB, X86::FLDr32, 1).addReg (reg1); BuildMI (BB, X86::FLDr32, 1).addReg (reg2); break; case cFP (doubles): // Doubles BuildMI (BB, X86::FLDr64, 1).addReg (reg1); BuildMI (BB, X86::FLDr64, 1).addReg (reg2); break; #endif case cLong: default: visitInstruction(I); } #if 0 if (CompTy->isFloatingPoint()) { // (Non-trapping) compare and pop twice. BuildMI (BB, X86::FUCOMPP, 0); // Move fp status word (concodes) to ax. BuildMI (BB, X86::FNSTSWr8, 1, X86::AX); // Load real concodes from ax. BuildMI (BB, X86::SAHF, 1).addReg(X86::AH); } #endif // Emit setOp instruction (extract concode; clobbers ax), // using the following mapping: // LLVM -> X86 signed X86 unsigned // ----- ----- ----- // seteq -> sete sete // setne -> setne setne // setlt -> setl setb // setgt -> setg seta // setle -> setle setbe // setge -> setge setae static const unsigned OpcodeTab[2][6] = { {X86::SETEr, X86::SETNEr, X86::SETBr, X86::SETAr, X86::SETBEr, X86::SETAEr}, {X86::SETEr, X86::SETNEr, X86::SETLr, X86::SETGr, X86::SETLEr, X86::SETGEr}, }; BuildMI(BB, OpcodeTab[CompTy->isSigned()][OpNum], 0, getReg(I)); } /// promote32 - Emit instructions to turn a narrow operand into a 32-bit-wide /// operand, in the specified target register. void ISel::promote32 (unsigned targetReg, Value *v) { unsigned vReg = getReg(v); bool isUnsigned = v->getType()->isUnsigned(); switch (getClass(v->getType())) { case cByte: // Extend value into target register (8->32) if (isUnsigned) BuildMI(BB, X86::MOVZXr32r8, 1, targetReg).addReg(vReg); else BuildMI(BB, X86::MOVSXr32r8, 1, targetReg).addReg(vReg); break; case cShort: // Extend value into target register (16->32) if (isUnsigned) BuildMI(BB, X86::MOVZXr32r16, 1, targetReg).addReg(vReg); else BuildMI(BB, X86::MOVSXr32r16, 1, targetReg).addReg(vReg); break; case cInt: // Move value into target register (32->32) BuildMI(BB, X86::MOVrr32, 1, targetReg).addReg(vReg); break; default: assert(0 && "Unpromotable operand class in promote32"); } } /// 'ret' instruction - Here we are interested in meeting the x86 ABI. As such, /// we have the following possibilities: /// /// ret void: No return value, simply emit a 'ret' instruction /// ret sbyte, ubyte : Extend value into EAX and return /// ret short, ushort: Extend value into EAX and return /// ret int, uint : Move value into EAX and return /// ret pointer : Move value into EAX and return /// ret long, ulong : Move value into EAX/EDX and return /// ret float/double : Top of FP stack /// void ISel::visitReturnInst (ReturnInst &I) { if (I.getNumOperands() == 0) { BuildMI(BB, X86::RET, 0); // Just emit a 'ret' instruction return; } Value *RetVal = I.getOperand(0); switch (getClass(RetVal->getType())) { case cByte: // integral return values: extend or move into EAX and return case cShort: case cInt: promote32(X86::EAX, RetVal); break; case cFP: // Floats & Doubles: Return in ST(0) BuildMI(BB, X86::FpMOV, 1, X86::ST0).addReg(getReg(RetVal)); break; case cLong: // ret long: use EAX(least significant 32 bits)/EDX (most // significant 32)... default: visitInstruction (I); } // Emit a 'ret' instruction BuildMI(BB, X86::RET, 0); } /// visitBranchInst - Handle conditional and unconditional branches here. Note /// that since code layout is frozen at this point, that if we are trying to /// jump to a block that is the immediate successor of the current block, we can /// just make a fall-through. (but we don't currently). /// void ISel::visitBranchInst(BranchInst &BI) { if (BI.isConditional()) { BasicBlock *ifTrue = BI.getSuccessor(0); BasicBlock *ifFalse = BI.getSuccessor(1); // Compare condition with zero, followed by jump-if-equal to ifFalse, and // jump-if-nonequal to ifTrue unsigned condReg = getReg(BI.getCondition()); BuildMI(BB, X86::CMPri8, 2).addReg(condReg).addZImm(0); BuildMI(BB, X86::JNE, 1).addPCDisp(BI.getSuccessor(0)); BuildMI(BB, X86::JE, 1).addPCDisp(BI.getSuccessor(1)); } else { // unconditional branch BuildMI(BB, X86::JMP, 1).addPCDisp(BI.getSuccessor(0)); } } /// visitCallInst - Push args on stack and do a procedure call instruction. void ISel::visitCallInst(CallInst &CI) { // Count how many bytes are to be pushed on the stack... unsigned NumBytes = 0; if (CI.getNumOperands() > 1) { for (unsigned i = 1, e = CI.getNumOperands(); i != e; ++i) switch (getClass(CI.getOperand(i)->getType())) { case cByte: case cShort: case cInt: NumBytes += 4; break; case cLong: NumBytes += 8; break; case cFP: NumBytes += CI.getOperand(i)->getType() == Type::FloatTy ? 4 : 8; break; default: assert(0 && "Unknown class!"); } // Adjust the stack pointer for the new arguments... BuildMI(BB, X86::ADJCALLSTACKDOWN, 1).addZImm(NumBytes); // Arguments go on the stack in reverse order, as specified by the ABI. unsigned ArgOffset = 0; for (unsigned i = 1, e = CI.getNumOperands(); i != e; ++i) { Value *Arg = CI.getOperand(i); switch (getClass(Arg->getType())) { case cByte: case cShort: { // Promote arg to 32 bits wide into a temporary register... unsigned R = makeAnotherReg(Type::UIntTy); promote32(R, Arg); addRegOffset(BuildMI(BB, X86::MOVrm32, 5), X86::ESP, ArgOffset).addReg(R); break; } case cInt: addRegOffset(BuildMI(BB, X86::MOVrm32, 5), X86::ESP, ArgOffset).addReg(getReg(Arg)); break; case cFP: if (Arg->getType() == Type::FloatTy) { addRegOffset(BuildMI(BB, X86::FSTr32, 5), X86::ESP, ArgOffset).addReg(getReg(Arg)); } else { assert(Arg->getType() == Type::DoubleTy && "Unknown FP type!"); ArgOffset += 4; addRegOffset(BuildMI(BB, X86::FSTr32, 5), X86::ESP, ArgOffset).addReg(getReg(Arg)); } break; default: // FIXME: long/ulong/float/double args not handled. visitInstruction(CI); break; } ArgOffset += 4; } } if (Function *F = CI.getCalledFunction()) { // Emit a CALL instruction with PC-relative displacement. BuildMI(BB, X86::CALLpcrel32, 1).addPCDisp(F); } else { unsigned Reg = getReg(CI.getCalledValue()); BuildMI(BB, X86::CALLr32, 1).addReg(Reg); } BuildMI(BB, X86::ADJCALLSTACKUP, 1).addZImm(NumBytes); // If there is a return value, scavenge the result from the location the call // leaves it in... // if (CI.getType() != Type::VoidTy) { unsigned resultTypeClass = getClass(CI.getType()); switch (resultTypeClass) { case cByte: case cShort: case cInt: { // Integral results are in %eax, or the appropriate portion // thereof. static const unsigned regRegMove[] = { X86::MOVrr8, X86::MOVrr16, X86::MOVrr32 }; static const unsigned AReg[] = { X86::AL, X86::AX, X86::EAX }; BuildMI(BB, regRegMove[resultTypeClass], 1, getReg(CI)) .addReg(AReg[resultTypeClass]); break; } case cFP: // Floating-point return values live in %ST(0) BuildMI(BB, X86::FpMOV, 1, getReg(CI)).addReg(X86::ST0); break; default: std::cerr << "Cannot get return value for call of type '" << *CI.getType() << "'\n"; visitInstruction(CI); } } } /// visitSimpleBinary - Implement simple binary operators for integral types... /// OperatorClass is one of: 0 for Add, 1 for Sub, 2 for And, 3 for Or, /// 4 for Xor. /// void ISel::visitSimpleBinary(BinaryOperator &B, unsigned OperatorClass) { if (B.getType() == Type::BoolTy) // FIXME: Handle bools for logicals visitInstruction(B); unsigned Class = getClass(B.getType()); if (Class > cFP) // FIXME: Handle longs visitInstruction(B); static const unsigned OpcodeTab[][4] = { // Arithmetic operators { X86::ADDrr8, X86::ADDrr16, X86::ADDrr32, X86::FpADD }, // ADD { X86::SUBrr8, X86::SUBrr16, X86::SUBrr32, X86::FpSUB }, // SUB // Bitwise operators { X86::ANDrr8, X86::ANDrr16, X86::ANDrr32, 0 }, // AND { X86:: ORrr8, X86:: ORrr16, X86:: ORrr32, 0 }, // OR { X86::XORrr8, X86::XORrr16, X86::XORrr32, 0 }, // XOR }; unsigned Opcode = OpcodeTab[OperatorClass][Class]; assert(Opcode && "Floating point arguments to logical inst?"); unsigned Op0r = getReg(B.getOperand(0)); unsigned Op1r = getReg(B.getOperand(1)); BuildMI(BB, Opcode, 2, getReg(B)).addReg(Op0r).addReg(Op1r); } /// doMultiply - Emit appropriate instructions to multiply together /// the registers op0Reg and op1Reg, and put the result in destReg. /// The type of the result should be given as resultType. void ISel::doMultiply(MachineBasicBlock *MBB, MachineBasicBlock::iterator &MBBI, unsigned destReg, const Type *resultType, unsigned op0Reg, unsigned op1Reg) { unsigned Class = getClass(resultType); switch (Class) { case cFP: // Floating point multiply BuildMI(BB, X86::FpMUL, 2, destReg).addReg(op0Reg).addReg(op1Reg); return; default: case cLong: assert(0 && "doMultiply not implemented for this class yet!"); case cByte: case cShort: case cInt: // Small integerals, handled below... break; } static const unsigned Regs[] ={ X86::AL , X86::AX , X86::EAX }; static const unsigned MulOpcode[]={ X86::MULrr8, X86::MULrr16, X86::MULrr32 }; static const unsigned MovOpcode[]={ X86::MOVrr8, X86::MOVrr16, X86::MOVrr32 }; unsigned Reg = Regs[Class]; // Emit a MOV to put the first operand into the appropriately-sized // subreg of EAX. BMI(MBB, MBBI, MovOpcode[Class], 1, Reg).addReg (op0Reg); // Emit the appropriate multiply instruction. BMI(MBB, MBBI, MulOpcode[Class], 1).addReg (op1Reg); // Emit another MOV to put the result into the destination register. BMI(MBB, MBBI, MovOpcode[Class], 1, destReg).addReg (Reg); } /// visitMul - Multiplies are not simple binary operators because they must deal /// with the EAX register explicitly. /// void ISel::visitMul(BinaryOperator &I) { unsigned DestReg = getReg(I); unsigned Op0Reg = getReg(I.getOperand(0)); unsigned Op1Reg = getReg(I.getOperand(1)); MachineBasicBlock::iterator MBBI = BB->end(); doMultiply(BB, MBBI, DestReg, I.getType(), Op0Reg, Op1Reg); } /// visitDivRem - Handle division and remainder instructions... these /// instruction both require the same instructions to be generated, they just /// select the result from a different register. Note that both of these /// instructions work differently for signed and unsigned operands. /// void ISel::visitDivRem(BinaryOperator &I) { unsigned Class = getClass(I.getType()); unsigned Op0Reg = getReg(I.getOperand(0)); unsigned Op1Reg = getReg(I.getOperand(1)); unsigned ResultReg = getReg(I); switch (Class) { case cFP: // Floating point multiply if (I.getOpcode() == Instruction::Div) BuildMI(BB, X86::FpDIV, 2, ResultReg).addReg(Op0Reg).addReg(Op1Reg); else BuildMI(BB, X86::FpREM, 2, ResultReg).addReg(Op0Reg).addReg(Op1Reg); return; default: case cLong: assert(0 && "div/rem not implemented for this class yet!"); case cByte: case cShort: case cInt: // Small integerals, handled below... break; } static const unsigned Regs[] ={ X86::AL , X86::AX , X86::EAX }; static const unsigned MovOpcode[]={ X86::MOVrr8, X86::MOVrr16, X86::MOVrr32 }; static const unsigned ExtOpcode[]={ X86::CBW , X86::CWD , X86::CDQ }; static const unsigned ClrOpcode[]={ X86::XORrr8, X86::XORrr16, X86::XORrr32 }; static const unsigned ExtRegs[] ={ X86::AH , X86::DX , X86::EDX }; static const unsigned DivOpcode[][4] = { { X86::DIVrr8 , X86::DIVrr16 , X86::DIVrr32 , 0 }, // Unsigned division { X86::IDIVrr8, X86::IDIVrr16, X86::IDIVrr32, 0 }, // Signed division }; bool isSigned = I.getType()->isSigned(); unsigned Reg = Regs[Class]; unsigned ExtReg = ExtRegs[Class]; // Put the first operand into one of the A registers... BuildMI(BB, MovOpcode[Class], 1, Reg).addReg(Op0Reg); if (isSigned) { // Emit a sign extension instruction... BuildMI(BB, ExtOpcode[Class], 0); } else { // If unsigned, emit a zeroing instruction... (reg = xor reg, reg) BuildMI(BB, ClrOpcode[Class], 2, ExtReg).addReg(ExtReg).addReg(ExtReg); } // Emit the appropriate divide or remainder instruction... BuildMI(BB, DivOpcode[isSigned][Class], 1).addReg(Op1Reg); // Figure out which register we want to pick the result out of... unsigned DestReg = (I.getOpcode() == Instruction::Div) ? Reg : ExtReg; // Put the result into the destination register... BuildMI(BB, MovOpcode[Class], 1, ResultReg).addReg(DestReg); } /// Shift instructions: 'shl', 'sar', 'shr' - Some special cases here /// for constant immediate shift values, and for constant immediate /// shift values equal to 1. Even the general case is sort of special, /// because the shift amount has to be in CL, not just any old register. /// void ISel::visitShiftInst (ShiftInst &I) { unsigned Op0r = getReg (I.getOperand(0)); unsigned DestReg = getReg(I); bool isLeftShift = I.getOpcode() == Instruction::Shl; bool isOperandSigned = I.getType()->isUnsigned(); unsigned OperandClass = getClass(I.getType()); if (OperandClass > cInt) visitInstruction(I); // Can't handle longs yet! if (ConstantUInt *CUI = dyn_cast (I.getOperand (1))) { // The shift amount is constant, guaranteed to be a ubyte. Get its value. assert(CUI->getType() == Type::UByteTy && "Shift amount not a ubyte?"); unsigned char shAmt = CUI->getValue(); static const unsigned ConstantOperand[][4] = { { X86::SHRir8, X86::SHRir16, X86::SHRir32, 0 }, // SHR { X86::SARir8, X86::SARir16, X86::SARir32, 0 }, // SAR { X86::SHLir8, X86::SHLir16, X86::SHLir32, 0 }, // SHL { X86::SHLir8, X86::SHLir16, X86::SHLir32, 0 }, // SAL = SHL }; const unsigned *OpTab = // Figure out the operand table to use ConstantOperand[isLeftShift*2+isOperandSigned]; // Emit: reg, shamt (shift-by-immediate opcode "ir" form.) BuildMI(BB, OpTab[OperandClass], 2, DestReg).addReg(Op0r).addZImm(shAmt); } else { // The shift amount is non-constant. // // In fact, you can only shift with a variable shift amount if // that amount is already in the CL register, so we have to put it // there first. // // Emit: move cl, shiftAmount (put the shift amount in CL.) BuildMI(BB, X86::MOVrr8, 1, X86::CL).addReg(getReg(I.getOperand(1))); // This is a shift right (SHR). static const unsigned NonConstantOperand[][4] = { { X86::SHRrr8, X86::SHRrr16, X86::SHRrr32, 0 }, // SHR { X86::SARrr8, X86::SARrr16, X86::SARrr32, 0 }, // SAR { X86::SHLrr8, X86::SHLrr16, X86::SHLrr32, 0 }, // SHL { X86::SHLrr8, X86::SHLrr16, X86::SHLrr32, 0 }, // SAL = SHL }; const unsigned *OpTab = // Figure out the operand table to use NonConstantOperand[isLeftShift*2+isOperandSigned]; BuildMI(BB, OpTab[OperandClass], 1, DestReg).addReg(Op0r); } } /// visitLoadInst - Implement LLVM load instructions in terms of the x86 'mov' /// instruction. The load and store instructions are the only place where we /// need to worry about the memory layout of the target machine. /// void ISel::visitLoadInst(LoadInst &I) { bool isLittleEndian = TM.getTargetData().isLittleEndian(); bool hasLongPointers = TM.getTargetData().getPointerSize() == 8; unsigned SrcAddrReg = getReg(I.getOperand(0)); unsigned DestReg = getReg(I); unsigned Class = getClass(I.getType()); switch (Class) { default: visitInstruction(I); // FIXME: Handle longs... case cFP: { // FIXME: Handle endian swapping for FP values. unsigned Opcode = I.getType() == Type::FloatTy ? X86::FLDr32 : X86::FLDr64; addDirectMem(BuildMI(BB, Opcode, 4, DestReg), SrcAddrReg); return; } case cInt: // Integers of various sizes handled below case cShort: case cByte: break; } // We need to adjust the input pointer if we are emulating a big-endian // long-pointer target. On these systems, the pointer that we are interested // in is in the upper part of the eight byte memory image of the pointer. It // also happens to be byte-swapped, but this will be handled later. // if (!isLittleEndian && hasLongPointers && isa(I.getType())) { unsigned R = makeAnotherReg(Type::UIntTy); BuildMI(BB, X86::ADDri32, 2, R).addReg(SrcAddrReg).addZImm(4); SrcAddrReg = R; } unsigned IReg = DestReg; if (!isLittleEndian) { // If big endian we need an intermediate stage IReg = makeAnotherReg(I.getType()); std::swap(IReg, DestReg); } static const unsigned Opcode[] = { X86::MOVmr8, X86::MOVmr16, X86::MOVmr32 }; addDirectMem(BuildMI(BB, Opcode[Class], 4, DestReg), SrcAddrReg); if (!isLittleEndian) { // Emit the byte swap instruction... switch (Class) { case cByte: // No byteswap neccesary for 8 bit value... BuildMI(BB, X86::MOVrr8, 1, IReg).addReg(DestReg); break; case cInt: // Use the 32 bit bswap instruction to do a 32 bit swap... BuildMI(BB, X86::BSWAPr32, 1, IReg).addReg(DestReg); break; case cShort: // For 16 bit we have to use an xchg instruction, because there is no // 16-bit bswap. XCHG is neccesarily not in SSA form, so we force things // into AX to do the xchg. // BuildMI(BB, X86::MOVrr16, 1, X86::AX).addReg(DestReg); BuildMI(BB, X86::XCHGrr8, 2).addReg(X86::AL, MOTy::UseAndDef) .addReg(X86::AH, MOTy::UseAndDef); BuildMI(BB, X86::MOVrr16, 1, DestReg).addReg(X86::AX); break; default: assert(0 && "Class not handled yet!"); } } } /// visitStoreInst - Implement LLVM store instructions in terms of the x86 'mov' /// instruction. /// void ISel::visitStoreInst(StoreInst &I) { bool isLittleEndian = TM.getTargetData().isLittleEndian(); bool hasLongPointers = TM.getTargetData().getPointerSize() == 8; unsigned ValReg = getReg(I.getOperand(0)); unsigned AddressReg = getReg(I.getOperand(1)); unsigned Class = getClass(I.getOperand(0)->getType()); switch (Class) { default: visitInstruction(I); // FIXME: Handle longs... case cFP: { // FIXME: Handle endian swapping for FP values. unsigned Opcode = I.getOperand(0)->getType() == Type::FloatTy ? X86::FSTr32 : X86::FSTr64; addDirectMem(BuildMI(BB, Opcode, 1+4), AddressReg).addReg(ValReg); return; } case cInt: // Integers of various sizes handled below case cShort: case cByte: break; } if (!isLittleEndian && hasLongPointers && isa(I.getOperand(0)->getType())) { unsigned R = makeAnotherReg(Type::UIntTy); BuildMI(BB, X86::ADDri32, 2, R).addReg(AddressReg).addZImm(4); AddressReg = R; } if (!isLittleEndian && Class != cByte) { // Emit a byte swap instruction... switch (Class) { case cInt: { unsigned R = makeAnotherReg(I.getOperand(0)->getType()); BuildMI(BB, X86::BSWAPr32, 1, R).addReg(ValReg); ValReg = R; break; } case cShort: // For 16 bit we have to use an xchg instruction, because there is no // 16-bit bswap. XCHG is neccesarily not in SSA form, so we force things // into AX to do the xchg. // BuildMI(BB, X86::MOVrr16, 1, X86::AX).addReg(ValReg); BuildMI(BB, X86::XCHGrr8, 2).addReg(X86::AL, MOTy::UseAndDef) .addReg(X86::AH, MOTy::UseAndDef); ValReg = X86::AX; break; default: assert(0 && "Unknown class!"); } } static const unsigned Opcode[] = { X86::MOVrm8, X86::MOVrm16, X86::MOVrm32 }; addDirectMem(BuildMI(BB, Opcode[Class], 1+4), AddressReg).addReg(ValReg); } /// visitCastInst - Here we have various kinds of copying with or without /// sign extension going on. void ISel::visitCastInst (CastInst &CI) { const Type *targetType = CI.getType (); Value *operand = CI.getOperand (0); unsigned operandReg = getReg (operand); const Type *sourceType = operand->getType (); unsigned destReg = getReg (CI); // // Currently we handle: // // 1) cast * to bool // // 2) cast {sbyte, ubyte} to {sbyte, ubyte} // cast {short, ushort} to {ushort, short} // cast {int, uint, ptr} to {int, uint, ptr} // // 3) cast {sbyte, ubyte} to {ushort, short} // cast {sbyte, ubyte} to {int, uint, ptr} // cast {short, ushort} to {int, uint, ptr} // // 4) cast {int, uint, ptr} to {short, ushort} // cast {int, uint, ptr} to {sbyte, ubyte} // cast {short, ushort} to {sbyte, ubyte} // 1) Implement casts to bool by using compare on the operand followed // by set if not zero on the result. if (targetType == Type::BoolTy) { BuildMI (BB, X86::CMPri8, 2).addReg (operandReg).addZImm (0); BuildMI (BB, X86::SETNEr, 1, destReg); return; } // 2) Implement casts between values of the same type class (as determined // by getClass) by using a register-to-register move. unsigned srcClass = getClassB(sourceType); unsigned targClass = getClass(targetType); static const unsigned regRegMove[] = { X86::MOVrr8, X86::MOVrr16, X86::MOVrr32 }; if (srcClass <= cInt && targClass <= cInt && srcClass == targClass) { BuildMI(BB, regRegMove[srcClass], 1, destReg).addReg(operandReg); return; } // 3) Handle cast of SMALLER int to LARGER int using a move with sign // extension or zero extension, depending on whether the source type // was signed. if ((srcClass <= cInt) && (targClass <= cInt) && (srcClass < targClass)) { static const unsigned ops[] = { X86::MOVSXr16r8, X86::MOVSXr32r8, X86::MOVSXr32r16, X86::MOVZXr16r8, X86::MOVZXr32r8, X86::MOVZXr32r16 }; unsigned srcSigned = sourceType->isSigned (); BuildMI (BB, ops[3 * srcSigned + srcClass + targClass - 1], 1, destReg).addReg (operandReg); return; } // 4) Handle cast of LARGER int to SMALLER int using a move to EAX // followed by a move out of AX or AL. if ((srcClass <= cInt) && (targClass <= cInt) && (srcClass > targClass)) { static const unsigned AReg[] = { X86::AL, X86::AX, X86::EAX }; BuildMI (BB, regRegMove[srcClass], 1, AReg[srcClass]).addReg (operandReg); BuildMI (BB, regRegMove[targClass], 1, destReg).addReg (AReg[srcClass]); return; } // Anything we haven't handled already, we can't (yet) handle at all. // // FP to integral casts can be handled with FISTP to store onto the // stack while converting to integer, followed by a MOV to load from // the stack into the result register. Integral to FP casts can be // handled with MOV to store onto the stack, followed by a FILD to // load from the stack while converting to FP. For the moment, I // can't quite get straight in my head how to borrow myself some // stack space and write on it. Otherwise, this would be trivial. visitInstruction (CI); } // ExactLog2 - This function solves for (Val == 1 << (N-1)) and returns N. It // returns zero when the input is not exactly a power of two. static unsigned ExactLog2(unsigned Val) { if (Val == 0) return 0; unsigned Count = 0; while (Val != 1) { if (Val & 1) return 0; Val >>= 1; ++Count; } return Count+1; } /// visitGetElementPtrInst - I don't know, most programs don't have /// getelementptr instructions, right? That means we can put off /// implementing this, right? Right. This method emits machine /// instructions to perform type-safe pointer arithmetic. I am /// guessing this could be cleaned up somewhat to use fewer temporary /// registers. void ISel::visitGetElementPtrInst (GetElementPtrInst &I) { unsigned outputReg = getReg (I); MachineBasicBlock::iterator MI = BB->end(); emitGEPOperation(BB, MI, I.getOperand(0), I.op_begin()+1, I.op_end(), outputReg); } void ISel::emitGEPOperation(MachineBasicBlock *MBB, MachineBasicBlock::iterator &IP, Value *Src, User::op_iterator IdxBegin, User::op_iterator IdxEnd, unsigned TargetReg) { const TargetData &TD = TM.getTargetData(); const Type *Ty = Src->getType(); unsigned basePtrReg = getReg(Src, MBB, IP); // GEPs have zero or more indices; we must perform a struct access // or array access for each one. for (GetElementPtrInst::op_iterator oi = IdxBegin, oe = IdxEnd; oi != oe; ++oi) { Value *idx = *oi; unsigned nextBasePtrReg = makeAnotherReg(Type::UIntTy); if (const StructType *StTy = dyn_cast(Ty)) { // It's a struct access. idx is the index into the structure, // which names the field. This index must have ubyte type. const ConstantUInt *CUI = cast(idx); assert(CUI->getType() == Type::UByteTy && "Funny-looking structure index in GEP"); // Use the TargetData structure to pick out what the layout of // the structure is in memory. Since the structure index must // be constant, we can get its value and use it to find the // right byte offset from the StructLayout class's list of // structure member offsets. unsigned idxValue = CUI->getValue(); unsigned memberOffset = TD.getStructLayout(StTy)->MemberOffsets[idxValue]; // Emit an ADD to add memberOffset to the basePtr. BMI(MBB, IP, X86::ADDri32, 2, nextBasePtrReg).addReg(basePtrReg).addZImm(memberOffset); // The next type is the member of the structure selected by the // index. Ty = StTy->getElementTypes()[idxValue]; } else if (const SequentialType *SqTy = cast(Ty)) { // It's an array or pointer access: [ArraySize x ElementType]. // idx is the index into the array. Unlike with structure // indices, we may not know its actual value at code-generation // time. assert(idx->getType() == Type::LongTy && "Bad GEP array index!"); // We want to add basePtrReg to(idxReg * sizeof ElementType). First, we // must find the size of the pointed-to type (Not coincidentally, the next // type is the type of the elements in the array). Ty = SqTy->getElementType(); unsigned elementSize = TD.getTypeSize(Ty); // If idxReg is a constant, we don't need to perform the multiply! if (ConstantSInt *CSI = dyn_cast(idx)) { if (CSI->isNullValue()) { BMI(MBB, IP, X86::MOVrr32, 1, nextBasePtrReg).addReg(basePtrReg); } else { unsigned Offset = elementSize*CSI->getValue(); BMI(MBB, IP, X86::ADDri32, 2, nextBasePtrReg).addReg(basePtrReg).addZImm(Offset); } } else if (elementSize == 1) { // If the element size is 1, we don't have to multiply, just add unsigned idxReg = getReg(idx, MBB, IP); BMI(MBB, IP, X86::ADDrr32, 2, nextBasePtrReg).addReg(basePtrReg).addReg(idxReg); } else { unsigned idxReg = getReg(idx, MBB, IP); unsigned OffsetReg = makeAnotherReg(Type::UIntTy); if (unsigned Shift = ExactLog2(elementSize)) { // If the element size is exactly a power of 2, use a shift to get it. BMI(MBB, IP, X86::SHLir32, 2, OffsetReg).addReg(idxReg).addZImm(Shift-1); } else { // Most general case, emit a multiply... unsigned elementSizeReg = makeAnotherReg(Type::LongTy); BMI(MBB, IP, X86::MOVir32, 1, elementSizeReg).addZImm(elementSize); // Emit a MUL to multiply the register holding the index by // elementSize, putting the result in OffsetReg. doMultiply(MBB, IP, OffsetReg, Type::LongTy, idxReg, elementSizeReg); } // Emit an ADD to add OffsetReg to the basePtr. BMI(MBB, IP, X86::ADDrr32, 2, nextBasePtrReg).addReg(basePtrReg).addReg(OffsetReg); } } // Now that we are here, further indices refer to subtypes of this // one, so we don't need to worry about basePtrReg itself, anymore. basePtrReg = nextBasePtrReg; } // After we have processed all the indices, the result is left in // basePtrReg. Move it to the register where we were expected to // put the answer. A 32-bit move should do it, because we are in // ILP32 land. BMI(MBB, IP, X86::MOVrr32, 1, TargetReg).addReg(basePtrReg); } /// visitAllocaInst - If this is a fixed size alloca, allocate space from the /// frame manager, otherwise do it the hard way. /// void ISel::visitAllocaInst(AllocaInst &I) { // Find the data size of the alloca inst's getAllocatedType. const Type *Ty = I.getAllocatedType(); unsigned TySize = TM.getTargetData().getTypeSize(Ty); // If this is a fixed size alloca in the entry block for the function, // statically stack allocate the space. // if (ConstantUInt *CUI = dyn_cast(I.getArraySize())) { if (I.getParent() == I.getParent()->getParent()->begin()) { TySize *= CUI->getValue(); // Get total allocated size... unsigned Alignment = TM.getTargetData().getTypeAlignment(Ty); // Create a new stack object using the frame manager... int FrameIdx = F->getFrameInfo()->CreateStackObject(TySize, Alignment); addFrameReference(BuildMI(BB, X86::LEAr32, 5, getReg(I)), FrameIdx); return; } } // Create a register to hold the temporary result of multiplying the type size // constant by the variable amount. unsigned TotalSizeReg = makeAnotherReg(Type::UIntTy); unsigned SrcReg1 = getReg(I.getArraySize()); unsigned SizeReg = makeAnotherReg(Type::UIntTy); BuildMI(BB, X86::MOVir32, 1, SizeReg).addZImm(TySize); // TotalSizeReg = mul , MachineBasicBlock::iterator MBBI = BB->end(); doMultiply(BB, MBBI, TotalSizeReg, Type::UIntTy, SrcReg1, SizeReg); // AddedSize = add , 15 unsigned AddedSizeReg = makeAnotherReg(Type::UIntTy); BuildMI(BB, X86::ADDri32, 2, AddedSizeReg).addReg(TotalSizeReg).addZImm(15); // AlignedSize = and , ~15 unsigned AlignedSize = makeAnotherReg(Type::UIntTy); BuildMI(BB, X86::ANDri32, 2, AlignedSize).addReg(AddedSizeReg).addZImm(~15); // Subtract size from stack pointer, thereby allocating some space. BuildMI(BB, X86::SUBri32, 2, X86::ESP).addReg(X86::ESP).addZImm(AlignedSize); // Put a pointer to the space into the result register, by copying // the stack pointer. BuildMI(BB, X86::MOVrr32, 1, getReg(I)).addReg(X86::ESP); // Inform the Frame Information that we have just allocated a variable sized // object. F->getFrameInfo()->CreateVariableSizedObject(); } /// createSimpleX86InstructionSelector - This pass converts an LLVM function /// into a machine code representation is a very simple peep-hole fashion. The /// generated code sucks but the implementation is nice and simple. /// Pass *createSimpleX86InstructionSelector(TargetMachine &TM) { return new ISel(TM); }