nudt-compiler-cpp/include/ir/IR.h

// 当前只支撑 i32、i32*、void 以及最小的内存/算术指令，演示用。
//
// 当前已经实现：
// 1. 基础类型系统：void / i32 / i32* / float / float* / array / pointer
// 2. Value 体系：Value / ConstantValue / ConstantInt / ConstantFloat / ConstantArray / ConstantZero / Function / BasicBlock / User / GlobalValue / Instruction
// 3. 最小指令集：Add / Sub / Mul / Div / Mod / Neg / Alloca / Load / Store / Ret / Cmp / FCmp / Zext / Br / CondBr / Call / GEP / SIToFP / FPToSI
// 4. BasicBlock / Function / Module 三层组织结构
// 5. IRBuilder：便捷创建常量和各类指令
// 6. def-use 关系的轻量实现：
//    - Instruction 保存 operand 列表
//    - Value 保存 uses
//    - 支持 ReplaceAllUsesWith 的简化实现
//
// 当前尚未实现或只做了最小占位：
// 1. 完整类型系统：label 类型等
// 2. 更成熟的 Use 管理（例如 LLVM 风格的双向链式结构）
// 3. 更完整的 IR verifier 和优化基础设施
//
// 当前需要特别说明的两个简化点：
// 1. BasicBlock 虽然已经纳入 Value 体系，但其类型目前仍用 void 作为占位，
//    后续如果补 label type，可以再改成更合理的块标签类型。
//
// 建议的扩展顺序：
// 1. 先补更多指令和类型
// 2. 再补控制流相关 IR
// 3. 最后再考虑把 Value/User/Use 进一步抽象成更完整的框架

#pragma once

#include <iosfwd>
#include <memory>
#include <stdexcept>
#include <string>
#include <unordered_map>
#include <utility>
#include <vector>

namespace ir {

class Type;
class Value;
class User;
class ConstantValue;
class ConstantInt;
class ConstantFloat;
class ConstantArray;
class ConstantZero;
class GlobalValue;
class Instruction;
class BasicBlock;
class Function;

// --- Type System ---

class Type {
 public:
  enum class Kind { Void, Int1, Int32, PtrInt32, Float, PtrFloat, Array, Pointer };
  explicit Type(Kind k);
  Type(Kind k, std::shared_ptr<Type> elem_ty, int num_elems);
  Type(Kind k, std::shared_ptr<Type> pointed_ty);

  static const std::shared_ptr<Type>& GetVoidType();
  static const std::shared_ptr<Type>& GetInt1Type();
  static const std::shared_ptr<Type>& GetInt32Type();
  static const std::shared_ptr<Type>& GetPtrInt32Type();
  static const std::shared_ptr<Type>& GetFloatType();
  static const std::shared_ptr<Type>& GetPtrFloatType();
  static std::shared_ptr<Type> GetArrayType(std::shared_ptr<Type> elem_ty, int num_elems);
  static std::shared_ptr<Type> GetPointerType(std::shared_ptr<Type> pointed_ty);

  Kind GetKind() const;
  bool IsVoid() const;
  bool IsInt1() const;
  bool IsInt32() const;
  bool IsPtrInt32() const;
  bool IsFloat() const;
  bool IsPtrFloat() const;
  bool IsArray() const;
  bool IsPointer() const;

  std::shared_ptr<Type> GetElementType() const { return elem_ty_; }
  int GetNumElements() const { return num_elems_; }
  std::shared_ptr<Type> GetPointedType() const { return elem_ty_; }

 private:
  Kind kind_;
  std::shared_ptr<Type> elem_ty_;
  int num_elems_ = 0;
};

// --- Value & Use ---

class Use {
 public:
  Use() = default;
  Use(Value* value, User* user, size_t operand_index)
      : value_(value), user_(user), operand_index_(operand_index) {}

  Value* GetValue() const { return value_; }
  User* GetUser() const { return user_; }
  size_t GetOperandIndex() const { return operand_index_; }

  void SetValue(Value* value) { value_ = value; }
  void SetUser(User* user) { user_ = user; }
  void SetOperandIndex(size_t operand_index) { operand_index_ = operand_index; }

 private:
  Value* value_ = nullptr;
  User* user_ = nullptr;
  size_t operand_index_ = 0;
};

class Value {
 public:
  Value(std::shared_ptr<Type> ty, std::string name);
  virtual ~Value() = default;
  const std::shared_ptr<Type>& GetType() const;
  const std::string& GetName() const;
  void SetName(std::string n);
  bool IsVoid() const;
  bool IsInt1() const;
  bool IsInt32() const;
  bool IsPtrInt32() const;
  bool IsFloat() const;
  bool IsPtrFloat() const;
  bool IsConstant() const;
  bool IsInstruction() const;
  bool IsUser() const;
  bool IsFunction() const;
  bool IsArgument() const;
  void AddUse(User* user, size_t operand_index);
  void RemoveUse(User* user, size_t operand_index);
  const std::vector<Use>& GetUses() const;
  void ReplaceAllUsesWith(Value* new_value);

 protected:
  std::shared_ptr<Type> type_;
  std::string name_;
  std::vector<Use> uses_;
};

class Argument : public Value {
 public:
  Argument(std::shared_ptr<Type> ty, std::string name, Function* parent, size_t arg_no);
  Function* GetParent() const;
  size_t GetArgNo() const;
 private:
  Function* parent_;
  size_t arg_no_;
};

// --- Constants ---

class ConstantValue : public Value {
 public:
  ConstantValue(std::shared_ptr<Type> ty, std::string name = "");
};

class ConstantInt : public ConstantValue {
 public:
  ConstantInt(std::shared_ptr<Type> ty, int v);
  int GetValue() const { return value_; }

 private:
  int value_{};
};

class ConstantFloat : public ConstantValue {
 public:
  ConstantFloat(std::shared_ptr<Type> ty, float v);
  float GetValue() const { return value_; }

 private:
  float value_{};
};

class ConstantArray : public ConstantValue {
 public:
  ConstantArray(std::shared_ptr<Type> ty, std::vector<ConstantValue*> elements);
  const std::vector<ConstantValue*>& GetElements() const { return elements_; }

 private:
  std::vector<ConstantValue*> elements_;
};

class ConstantZero : public ConstantValue {
 public:
  explicit ConstantZero(std::shared_ptr<Type> ty);
};

// --- Context ---

class Context {
 public:
  Context() = default;
  ~Context();
  ConstantInt* GetConstInt(int v);
  ConstantFloat* GetConstFloat(float v);
  ConstantArray* GetConstArray(std::shared_ptr<Type> ty, std::vector<ConstantValue*> elements);
  ConstantZero* GetConstZero(std::shared_ptr<Type> ty);

  std::string NextTemp();

 private:
  std::unordered_map<int, std::unique_ptr<ConstantInt>> const_ints_;
  std::unordered_map<float, std::unique_ptr<ConstantFloat>> const_floats_;
  std::vector<std::unique_ptr<ConstantArray>> const_arrays_;
  std::vector<std::unique_ptr<ConstantZero>> const_zeros_;
  int temp_index_ = -1;
};

// --- Instructions ---

enum class Opcode { Add, Sub, Mul, Div, Mod, Neg, Alloca, Load, Store, Ret, Cmp, FCmp, Zext, Br, CondBr, Call, GEP, SIToFP, FPToSI };

enum class CmpOp { Eq, Ne, Lt, Gt, Le, Ge };

class User : public Value {
 public:
  User(std::shared_ptr<Type> ty, std::string name);
  size_t GetNumOperands() const;
  Value* GetOperand(size_t index) const;
  void SetOperand(size_t index, Value* value);

 protected:
  void AddOperand(Value* value);

 private:
  std::vector<Value*> operands_;
};

class GlobalValue : public User {
 public:
  GlobalValue(std::shared_ptr<Type> ty, std::string name);
};

class GlobalVariable : public GlobalValue {
 public:
  GlobalVariable(std::string name, std::shared_ptr<Type> type, ConstantValue* init);
  ConstantValue* GetInitializer() const { return init_; }
 private:
  ConstantValue* init_ = nullptr;
};

class Instruction : public User {
 public:
  Instruction(Opcode op, std::shared_ptr<Type> ty, std::string name = "");
  Opcode GetOpcode() const;
  bool IsTerminator() const;
  BasicBlock* GetParent() const;
  void SetParent(BasicBlock* parent);

 private:
  Opcode opcode_;
  BasicBlock* parent_ = nullptr;
};

class BinaryInst : public Instruction {
 public:
  BinaryInst(Opcode op, std::shared_ptr<Type> ty, Value* lhs, Value* rhs,
             std::string name);
  Value* GetLhs() const;
  Value* GetRhs() const;
};

class UnaryInst : public Instruction {
 public:
  UnaryInst(Opcode op, std::shared_ptr<Type> ty, Value* operand,
            std::string name);
  Value* GetUnaryOperand() const;
};

class ReturnInst : public Instruction {
 public:
  ReturnInst(std::shared_ptr<Type> void_ty, Value* val);
  Value* GetValue() const;
};

class AllocaInst : public Instruction {
 public:
  AllocaInst(std::shared_ptr<Type> ptr_ty, std::string name);
};

class LoadInst : public Instruction {
 public:
  LoadInst(std::shared_ptr<Type> val_ty, Value* ptr, std::string name);
  Value* GetPtr() const;
};

class StoreInst : public Instruction {
 public:
  StoreInst(std::shared_ptr<Type> void_ty, Value* val, Value* ptr);
  Value* GetValue() const;
  Value* GetPtr() const;
};

class CmpInst : public Instruction {
 public:
  CmpInst(CmpOp cmp_op, Value* lhs, Value* rhs, std::string name);
  CmpOp GetCmpOp() const;
  Value* GetLhs() const;
  Value* GetRhs() const;

 private:
  CmpOp cmp_op_;
};

class FCmpInst : public Instruction {
 public:
  FCmpInst(CmpOp cmp_op, Value* lhs, Value* rhs, std::string name);
  CmpOp GetCmpOp() const;
  Value* GetLhs() const;
  Value* GetRhs() const;

 private:
  CmpOp cmp_op_;
};

class ZextInst : public Instruction {
 public:
  ZextInst(std::shared_ptr<Type> dest_ty, Value* val, std::string name);
  Value* GetValue() const;
};

class BranchInst : public Instruction {
 public:
  BranchInst(BasicBlock* dest);
  BasicBlock* GetDest() const;
};

class CondBranchInst : public Instruction {
 public:
  CondBranchInst(Value* cond, BasicBlock* true_bb, BasicBlock* false_bb);
  Value* GetCond() const;
  BasicBlock* GetTrueBlock() const;
  BasicBlock* GetFalseBlock() const;
};

class CallInst : public Instruction {
 public:
  CallInst(Function* func, std::vector<Value*> args, std::string name = "");
  Function* GetFunc() const;
  const std::vector<Value*>& GetArgs() const;
  
 private:
  Function* func_;
  std::vector<Value*> args_;
};

class GEPInst : public Instruction {
 public:
  GEPInst(std::shared_ptr<Type> ty, Value* ptr, std::vector<Value*> indices, std::string name = "");
  Value* GetPtr() const;
  const std::vector<Value*>& GetIndices() const;
 private:
  std::vector<Value*> indices_;
};

class SIToFPInst : public Instruction {
 public:
  SIToFPInst(std::shared_ptr<Type> ty, Value* val, std::string name = "");
};

class FPToSIInst : public Instruction {
 public:
  FPToSIInst(std::shared_ptr<Type> ty, Value* val, std::string name = "");
};

// --- Structure ---

class BasicBlock : public Value {
 public:
  explicit BasicBlock(std::string name);
  Function* GetParent() const;
  void SetParent(Function* parent);
  bool HasTerminator() const;
  const std::vector<std::unique_ptr<Instruction>>& GetInstructions() const;
  const std::vector<BasicBlock*>& GetPredecessors() const;
  const std::vector<BasicBlock*>& GetSuccessors() const;
  template <typename T, typename... Args>
  T* Append(Args&&... args) {
    if (HasTerminator()) {
      throw std::runtime_error("BasicBlock 已有 terminator，不能继续追加指令: " +
                               name_);
    }
    auto inst = std::make_unique<T>(std::forward<Args>(args)...);
    auto* ptr = inst.get();
    ptr->SetParent(this);
    instructions_.push_back(std::move(inst));
    return ptr;
  }

 private:
  Function* parent_ = nullptr;
  std::vector<std::unique_ptr<Instruction>> instructions_;
  std::vector<BasicBlock*> predecessors_;
  std::vector<BasicBlock*> successors_;
};

class Function : public Value {
 public:
  Function(std::string name, std::shared_ptr<Type> ret_type);
  BasicBlock* CreateBlock(const std::string& name);
  BasicBlock* GetEntry();
  const BasicBlock* GetEntry() const;
  const std::vector<std::unique_ptr<BasicBlock>>& GetBlocks() const;

  Argument* AddArgument(std::shared_ptr<Type> ty, std::string name);
  const std::vector<std::unique_ptr<Argument>>& GetArgs() const;

 private:
  BasicBlock* entry_ = nullptr;
  std::vector<std::unique_ptr<BasicBlock>> blocks_;
  std::vector<std::unique_ptr<Argument>> args_;
};

class Module {
 public:
  Module() = default;
  Context& GetContext();
  const Context& GetContext() const;
  Function* CreateFunction(const std::string& name,
                           std::shared_ptr<Type> ret_type);
  const std::vector<std::unique_ptr<Function>>& GetFunctions() const;
  
  GlobalVariable* CreateGlobalVariable(const std::string& name, std::shared_ptr<Type> type, ConstantValue* init);
  const std::vector<std::unique_ptr<GlobalVariable>>& GetGlobalVariables() const;

 private:
  Context context_;
  std::vector<std::unique_ptr<Function>> functions_;
  std::vector<std::unique_ptr<GlobalVariable>> global_variables_;
};

class IRBuilder {
 public:
  IRBuilder(Context& ctx, BasicBlock* bb);
  void SetInsertPoint(BasicBlock* bb);
  BasicBlock* GetInsertBlock() const;

  ConstantInt* CreateConstInt(int v);
  BinaryInst* CreateBinary(Opcode op, Value* lhs, Value* rhs,
                           const std::string& name);
  BinaryInst* CreateAdd(Value* lhs, Value* rhs, const std::string& name);
  BinaryInst* CreateSub(Value* lhs, Value* rhs, const std::string& name);
  BinaryInst* CreateMul(Value* lhs, Value* rhs, const std::string& name);
  UnaryInst* CreateNeg(Value* operand, const std::string& name);
  AllocaInst* CreateAllocaI32(const std::string& name);
  AllocaInst* CreateAllocaFloat(const std::string& name);
  AllocaInst* CreateAlloca(std::shared_ptr<Type> ty, const std::string& name);
  LoadInst* CreateLoad(Value* ptr, const std::string& name);
  StoreInst* CreateStore(Value* val, Value* ptr);
  ReturnInst* CreateRet(Value* v);
  Instruction* CreateCmp(CmpOp op, Value* lhs, Value* rhs, const std::string& name);
  ZextInst* CreateZext(Value* val, const std::string& name);
  BranchInst* CreateBr(BasicBlock* dest);
  CondBranchInst* CreateCondBr(Value* cond, BasicBlock* true_bb, BasicBlock* false_bb);
  CallInst* CreateCall(Function* func, std::vector<Value*> args, const std::string& name);
  GEPInst* CreateGEP(std::shared_ptr<Type> ty, Value* ptr, std::vector<Value*> indices, const std::string& name);
  SIToFPInst* CreateSIToFP(Value* val, const std::string& name);
  FPToSIInst* CreateFPToSI(Value* val, const std::string& name);

 private:
  Context& ctx_;
  BasicBlock* insert_block_;
};

class IRPrinter {
 public:
  void Print(const Module& module, std::ostream& os);
};

}  // namespace ir