[deploy]部署版本2

README.md

@@ -37,13 +37,4 @@ mysysy/ $ bash setup.sh
     ```
 
 ### 配套脚本
-   （TODO: 需要完善）
-
-
-### TODO_list:
-
-除开注释中的TODO后续时间充足可以考虑的TODO:
-
-- store load指令由于gep指令的引入, 维度信息的记录是非必须的, 考虑删除
-
-- use def关系经过mem2reg和phi函数明确转换为ssa形式, 以及函数参数通过value数组明确定义, 使得基本块的args参数信息记录非必须, 考虑删除
+   （TODO: 需要完善）

src/AddressCalculationExpansion.cpp

@@ -57,11 +57,22 @@ bool AddressCalculationExpansion::run() {
                         }
                     }
                 } else if (GlobalValue* globalValue = dynamic_cast<GlobalValue*>(basePointer)) {
-                    std::cerr << "Warning: GlobalValue dimension handling needs explicit implementation for GEP expansion. Skipping GEP for: ";
-                    SysYPrinter::printValue(globalValue);
-                    std::cerr << "\n";
-                    ++it;
-                    continue;
+                    // 遍历 GlobalValue 的所有维度操作数
+                    for (const auto& use_ptr : globalValue->getDims()) {
+                        Value* dimValue = use_ptr->getValue();
+                        // 将维度值转换为常量整数
+                        if (ConstantInteger* constVal = dynamic_cast<ConstantInteger*>(dimValue)) {
+                            dims.push_back(constVal->getInt());
+                        } else {
+                            // 如果维度不是常量整数，则无法处理。
+                            // 根据 IR.h 中 GlobalValue 的构造函数，这种情况不应发生，但作为安全检查是好的。
+                            std::cerr << "Warning: GlobalValue dimension is not a constant integer. Skipping GEP expansion for: ";
+                            SysYPrinter::printValue(globalValue);
+                            std::cerr << "\n";
+                            dims.clear(); // 清空已收集的部分维度信息
+                            break;
+                        }
+                    }
                 } else {
                     std::cerr << "Warning: Base pointer is not AllocaInst/GlobalValue or its array dimensions cannot be determined for GEP expansion. Skipping GEP for: ";
                     SysYPrinter::printValue(basePointer);

src/CMakeLists.txt

@@ -21,8 +21,8 @@ add_executable(sysyc
   IR.cpp
   SysYIRGenerator.cpp
   SysYIRPrinter.cpp
-  SysYIRCFGOpt.cpp
-  # SysYIRAnalyser.cpp
+  SysYIROptPre.cpp
+  SysYIRAnalyser.cpp
   # DeadCodeElimination.cpp
   AddressCalculationExpansion.cpp
   # Mem2Reg.cpp

src/DeadCodeElimination.cpp

@@ -1,259 +0,0 @@
-#include "DeadCodeElimination.h"
-#include <iostream>
-
-extern int DEBUG;
-namespace sysy {
-
-void DeadCodeElimination::runDCEPipeline() {
-  const auto& functions = pModule->getFunctions();
-  for (const auto& function : functions) {
-    const auto& func = function.second;
-    bool changed = true;
-    while (changed) {
-      changed = false;
-      eliminateDeadStores(func.get(), changed);
-      eliminateDeadLoads(func.get(), changed);
-      eliminateDeadAllocas(func.get(), changed);
-      eliminateDeadRedundantLoadStore(func.get(), changed);
-      eliminateDeadGlobals(changed);
-    }
-  }
-}
-
-// 消除无用存储 消除条件：
-// 存储的目标指针（pointer）不是全局变量（!isGlobal(pointer)）。
-// 存储的目标指针不是数组参数（!isArr(pointer) 或不在函数参数列表里）。
-// 该指针的所有使用者（uses）仅限 alloca 或 store（即没有 load 或其他指令使用它）。
-void DeadCodeElimination::eliminateDeadStores(Function* func, bool& changed) {
-  for (const auto& block : func->getBasicBlocks()) {
-    auto& instrs = block->getInstructions();
-    for (auto iter = instrs.begin(); iter != instrs.end();) {
-      auto inst = iter->get();
-      if (!inst->isStore()) {
-        ++iter;
-        continue;
-      }
-
-      auto storeInst = dynamic_cast<StoreInst*>(inst);
-      auto pointer = storeInst->getPointer();
-      // 如果是全局变量或者是函数的数组参数
-      if (SysYIROptUtils::isGlobal(pointer) || (SysYIROptUtils::isArr(pointer) &&
-            std::find(func->getEntryBlock()->getArguments().begin(),
-                    func->getEntryBlock()->getArguments().end(),
-                    pointer) != func->getEntryBlock()->getArguments().end())) {
-        ++iter;
-        continue;
-      }
-
-      bool changetag = true;
-      for (auto& use : pointer->getUses()) {
-        // 依次判断store的指针是否被其他指令使用
-        auto user = use->getUser();
-        auto userInst = dynamic_cast<Instruction*>(user);
-        // 如果使用store的指针的指令不是Alloca或Store，则不删除
-        if (userInst != nullptr && !userInst->isAlloca() && !userInst->isStore()) {
-          changetag = false;
-          break;
-        }
-      }
-
-      if (changetag) {
-        changed = true;
-        if(DEBUG){
-          std::cout << "=== Dead Store Found ===\n";
-          SysYPrinter::printInst(storeInst);
-        }
-        SysYIROptUtils::usedelete(storeInst);
-        iter = instrs.erase(iter);
-      } else {
-        ++iter;
-      }
-    }
-  }
-}
-// 消除无用加载 消除条件：
-// 该指令的结果未被使用（inst->getUses().empty()）。
-void DeadCodeElimination::eliminateDeadLoads(Function* func, bool& changed) {
-  for (const auto& block : func->getBasicBlocks()) {
-    auto& instrs = block->getInstructions();
-    for (auto iter = instrs.begin(); iter != instrs.end();) {
-      auto inst = iter->get();
-      if (inst->isBinary() || inst->isUnary() || inst->isLoad()) {
-        if (inst->getUses().empty()) {
-          changed = true;
-          if(DEBUG){
-            std::cout << "=== Dead Load Binary Unary Found ===\n";
-            SysYPrinter::printInst(inst);
-          }
-          SysYIROptUtils::usedelete(inst);
-          iter = instrs.erase(iter);
-          continue;
-        }
-      }
-      ++iter;
-    }
-  }
-}
-
-// 消除无用加载 消除条件：
-// 该 alloca 未被任何指令使用（allocaInst->getUses().empty()）。
-// 该 alloca 不是函数的参数（不在 entry 块的参数列表里）。
-void DeadCodeElimination::eliminateDeadAllocas(Function* func, bool& changed) {
-  for (const auto& block : func->getBasicBlocks()) {
-    auto& instrs = block->getInstructions();
-    for (auto iter = instrs.begin(); iter != instrs.end();) {
-      auto inst = iter->get();
-      if (inst->isAlloca()) {
-        auto allocaInst = dynamic_cast<AllocaInst*>(inst);
-        if (allocaInst->getUses().empty() &&
-            std::find(func->getEntryBlock()->getArguments().begin(),
-                      func->getEntryBlock()->getArguments().end(),
-                      allocaInst) == func->getEntryBlock()->getArguments().end()) {
-          changed = true;
-          if(DEBUG){
-            std::cout << "=== Dead Alloca Found ===\n";
-            SysYPrinter::printInst(inst);
-          }
-          SysYIROptUtils::usedelete(inst);
-          iter = instrs.erase(iter);
-          continue;
-        }
-      }
-      ++iter;
-    }
-  }
-}
-
-void DeadCodeElimination::eliminateDeadIndirectiveAllocas(Function* func, bool& changed) {
-  // 删除mem2reg时引入的且现在已经没有value使用了的隐式alloca
-  FunctionAnalysisInfo* funcInfo = pCFA->getFunctionAnalysisInfo(func);
-  for (auto it = funcInfo->getIndirectAllocas().begin(); it != funcInfo->getIndirectAllocas().end();) {
-    auto &allocaInst = *it;
-    if (allocaInst->getUses().empty()) {  
-      changed = true;
-      if(DEBUG){
-        std::cout << "=== Dead Indirect Alloca Found ===\n";
-        SysYPrinter::printInst(allocaInst.get());
-      }
-      it = funcInfo->getIndirectAllocas().erase(it);
-    } else {
-      ++it;
-    }
-  }
-}
-
-// 该全局变量未被任何指令使用（global->getUses().empty()）。
-void DeadCodeElimination::eliminateDeadGlobals(bool& changed) {
-  auto& globals = pModule->getGlobals();
-  for (auto it = globals.begin(); it != globals.end();) {
-    auto& global = *it;
-    if (global->getUses().empty()) {
-      changed = true;
-      if(DEBUG){
-        std::cout << "=== Dead Global Found ===\n";
-        SysYPrinter::printValue(global.get());
-      }
-      it = globals.erase(it);
-    } else {
-      ++it;
-    }
-  }
-}
-
-// 消除冗余加载和存储 消除条件：
-// phi 指令的目标指针仅被该 phi 使用（无其他 store/load 使用）。
-// memset 指令的目标指针未被使用（pointer->getUses().empty()）
-// store -> load -> store 模式
-void DeadCodeElimination::eliminateDeadRedundantLoadStore(Function* func, bool& changed) {
-  for (const auto& block : func->getBasicBlocks()) {
-    auto& instrs = block->getInstructions();
-    for (auto iter = instrs.begin(); iter != instrs.end();) {
-      auto inst = iter->get();
-      if (inst->isPhi()) {
-        auto phiInst = dynamic_cast<PhiInst*>(inst);
-        auto pointer = phiInst->getPointer();
-        bool tag = true;
-        for (const auto& use : pointer->getUses()) {
-          auto user = use->getUser();
-          if (user != inst) {
-            tag = false;
-            break;
-          }
-        }
-        /// 如果 pointer 仅被该 phi 使用，可以删除 ph
-        if (tag) {
-          changed = true;
-          SysYIROptUtils::usedelete(inst);
-          iter = instrs.erase(iter);
-          continue;
-        }
-        // 数组指令还不完善，不保证memset优化效果
-      } else if (inst->isMemset()) {
-        auto memsetInst = dynamic_cast<MemsetInst*>(inst);
-        auto pointer = memsetInst->getPointer();
-        if (pointer->getUses().empty()) {
-          changed = true;
-          SysYIROptUtils::usedelete(inst);
-          iter = instrs.erase(iter);
-          continue;
-        }
-      }else if(inst->isLoad()) {
-        if (iter != instrs.begin()) {
-          auto loadInst = dynamic_cast<LoadInst*>(inst);
-          auto loadPointer = loadInst->getPointer();
-          // TODO:store -> load -> store 模式
-          auto prevIter = std::prev(iter);
-          auto prevInst = prevIter->get();
-          if (prevInst->isStore()) {
-            auto prevStore = dynamic_cast<StoreInst*>(prevInst);
-            auto prevStorePointer = prevStore->getPointer();
-            auto prevStoreValue = prevStore->getOperand(0);
-            // 确保前一个 store 不是数组操作
-            if (prevStore->getIndices().empty()) {
-              // 检查后一条指令是否是 store 同一个值
-              auto nextIter = std::next(iter);
-              if (nextIter != instrs.end()) {
-                auto nextInst = nextIter->get();
-                if (nextInst->isStore()) {
-                  auto nextStore = dynamic_cast<StoreInst*>(nextInst);
-                  auto nextStorePointer = nextStore->getPointer();
-                  auto nextStoreValue = nextStore->getOperand(0);
-                  // 确保后一个 store 不是数组操作
-                  if (nextStore->getIndices().empty()) {
-                    // 判断优化条件：
-                    // 1. prevStore 的指针操作数 == load 的指针操作数
-                    // 2. nextStore 的值操作数 == load 指令本身
-                    if (prevStorePointer == loadPointer &&
-                        nextStoreValue == loadInst) {
-                      // 可以优化直接把prevStorePointer的值存到nextStorePointer
-                      changed = true;
-                      nextStore->setOperand(0, prevStoreValue);
-                      if(DEBUG){
-                        std::cout << "=== Dead Store Load Store Found(now only del Load) ===\n";
-                        SysYPrinter::printInst(prevStore);
-                        SysYPrinter::printInst(loadInst);
-                        SysYPrinter::printInst(nextStore);
-                      }
-                      SysYIROptUtils::usedelete(loadInst);
-                      iter = instrs.erase(iter);
-                      // 删除 prevStore 这里是不是可以留给删除无用store处理？
-                      // if (prevStore->getUses().empty()) {
-                      //   usedelete(prevStore);
-                      //   instrs.erase(prevIter);  // 删除 prevStore
-                      // }
-                      continue;  // 跳过 ++iter，因为已经移动迭代器
-                    }
-                  }
-                }
-              }
-            }
-          }
-        }
-      }
-      ++iter;
-    }
-  }
-}
-
-
-} // namespace sysy

src/IR.cpp

@@ -49,11 +49,6 @@ auto Type::getFunctionType(Type *returnType, const std::vector<Type *> &paramTyp
   return FunctionType::get(returnType, paramTypes);
 }
 
-auto Type::getArrayType(Type *elementType, unsigned numElements) -> Type * {
-  // forward to ArrayType
-  return ArrayType::get(elementType, numElements);
-}
-
 auto Type::getSize() const -> unsigned {
   switch (kind) {
     case kInt:
@@ -63,10 +58,6 @@ auto Type::getSize() const -> unsigned {
     case kPointer:
     case kFunction:
       return 8;
-    case Kind::kArray: {
-      const ArrayType* arrType = static_cast<const ArrayType*>(this);
-      return arrType->getElementType()->getSize() * arrType->getNumElements();
-    }
     case kVoid:
       return 0;
   }
@@ -104,11 +95,6 @@ FunctionType*FunctionType::get(Type *returnType, const std::vector<Type *> &para
   return result.first->get();
 }
 
-ArrayType *ArrayType::get(Type *elementType, unsigned numElements) {
-  // TODO:可以考虑在这里添加缓存，避免重复创建相同的数组类型
-  return new ArrayType(elementType, numElements);
-}
-
 void Value::replaceAllUsesWith(Value *value) {
   for (auto &use : uses) {
     use->getUser()->setOperand(use->getIndex(), value);
@@ -479,7 +465,44 @@ Function * Function::clone(const std::string &suffix) const  {
         break;
       }
 
-      // TODO：复制GEP指令
+      case Instruction::kLa: {
+        auto oldLaInst = dynamic_cast<LaInst *>(inst);
+        auto oldPointer = oldLaInst->getPointer();
+        Value *newPointer;
+        std::vector<Value *> newIndices;
+        newPointer = oldNewValueMap.at(oldPointer);
+
+        for (const auto &index : oldLaInst->getIndices()) {
+          newIndices.emplace_back(oldNewValueMap.at(index->getValue()));
+        }
+        ss << oldLaInst->getName() << suffix;
+        auto newLaInst = new LaInst(newPointer, newIndices, oldNewBlockMap.at(oldLaInst->getParent()), ss.str());
+        ss.str("");
+        oldNewValueMap.emplace(oldLaInst, newLaInst);
+        break;
+      }
+
+      case Instruction::kGetSubArray: {
+        auto oldGetSubArrayInst = dynamic_cast<GetSubArrayInst *>(inst);
+        auto oldFather = oldGetSubArrayInst->getFatherArray();
+        auto oldChild = oldGetSubArrayInst->getChildArray();
+        Value *newFather;
+        Value *newChild;
+        std::vector<Value *> newIndices;
+        newFather = oldNewValueMap.at(oldFather);
+        newChild = oldNewValueMap.at(oldChild);
+
+        for (const auto &index : oldGetSubArrayInst->getIndices()) {
+          newIndices.emplace_back(oldNewValueMap.at(index->getValue()));
+        }
+        ss << oldGetSubArrayInst->getName() << suffix;
+        auto newGetSubArrayInst =
+            new GetSubArrayInst(dynamic_cast<LVal *>(newFather), dynamic_cast<LVal *>(newChild), newIndices,
+                                oldNewBlockMap.at(oldGetSubArrayInst->getParent()), ss.str());
+        ss.str("");
+        oldNewValueMap.emplace(oldGetSubArrayInst, newGetSubArrayInst);
+        break;
+      }
 
       case Instruction::kMemset: {
         auto oldMemsetInst = dynamic_cast<MemsetInst *>(inst);

src/Mem2Reg.cpp

@@ -1,515 +0,0 @@
-#include "Mem2Reg.h"
-#include "SysYIRPrinter.h"
-#include <algorithm>
-#include <cassert>
-#include <memory>
-#include <queue>
-#include <stack>
-#include <unordered_map>
-#include <unordered_set>
-#include <vector>
-
-namespace sysy {
-
-// --- 私有成员函数实现 ---
-
-// 计算给定定义块集合的迭代支配边界
-std::unordered_set<BasicBlock*> Mem2Reg::computeIteratedDomFrontiers(const std::unordered_set<BasicBlock*>& blocks) {
-  std::unordered_set<BasicBlock*> result;
-  std::queue<BasicBlock*> worklist; // 使用队列进行 BFS-like 遍历
-
-  for (auto* block : blocks)
-    worklist.push(block);
-
-  while (!worklist.empty()) {
-    auto* block = worklist.front();
-    worklist.pop();
-
-    auto* blockInfo = controlFlowAnalysis->getBlockAnalysisInfo(block);
-    if (!blockInfo) continue; 
-
-    for (auto* df : blockInfo->getDomFrontiers()) {
-      if (result.find(df) == result.end()) { // If not already in result
-        result.insert(df);
-        worklist.push(df);
-      }
-    }
-  }
-  return result;
-}
-
-// 分析一个 alloca 的所有 uses，填充 allocaDefsBlock 和 allocaUsesBlock
-void Mem2Reg::allocaAnalysis(AllocaInst* alloca) {
-    allocaDefsBlock[alloca].clear();
-    allocaUsesBlock[alloca].clear();
-
-    for (auto use : alloca->getUses()) {
-        Instruction* userInst = dynamic_cast<Instruction*>(use->getUser());
-        if (!userInst) continue; 
-
-        if (StoreInst* store = dynamic_cast<StoreInst*>(userInst)) {
-            if (store->getOperand(1) == alloca) { // Store's second operand is the pointer
-                allocaDefsBlock[alloca].insert(store->getParent()); // Store's parent is the defining block
-            }
-        } else if (LoadInst* load = dynamic_cast<LoadInst*>(userInst)) {
-            if (load->getOperand(0) == alloca) { // Load's first operand is the pointer
-                allocaUsesBlock[alloca].insert(load->getParent()); // Load's parent is the using block
-            }
-        }
-    }
-}
-
-// 判断一个 alloca 是否可以被提升为寄存器 (无地址逃逸，标量类型)
-bool Mem2Reg::is_promoted(AllocaInst* alloca) {
-  // 检查是否是标量类型 (非数组、非全局变量等)   
-  if(!(SysYIROptUtils::isArr(alloca) || SysYIROptUtils::isGlobal(alloca))){
-    return false; // 只有标量类型的 alloca 才能被提升
-  }
-
-    // 获取 alloca 指向的基类型
-  PointerType* ptrType = dynamic_cast<PointerType*>(alloca->getType());
-  if (!ptrType) return false; // Should always be a pointer type
-  Type* allocabaseType = ptrType->getBaseType();
-
-  for (const auto& use : alloca->getUses()) {
-    Instruction* userInst = dynamic_cast<Instruction*>(use->getUser());
-    if (!userInst) {
-      // 如果不是指令的 use，比如作为全局变量的初始值等，通常认为逃逸
-      return false; 
-    }
-
-    if (LoadInst* load = dynamic_cast<LoadInst*>(userInst)) {
-      // Load 指令结果的类型必须与 alloca 的基类型一致
-      if (load->getType() != allocabaseType) {
-        return false;
-      }
-    } else if (StoreInst* store = dynamic_cast<StoreInst*>(userInst)) {
-      // Store 指令的值操作数类型必须与 alloca 的基类型一致
-      // 且 store 的指针操作数必须是当前 alloca
-      if (store->getOperand(1) != alloca || store->getOperand(0)->getType() != allocabaseType) {
-        return false;
-      }
-    } else if (userInst->isGetSubArray()) {
-      // GSA 指令表示对数组的访问
-      // 这意味着地址逃逸，不能简单提升为单个寄存器
-      return false;
-    } else if (userInst->isCall()) {
-      // 如果 alloca 作为函数参数传递，通常认为地址逃逸
-      return false;
-    }
-    // 如果有其他类型的指令使用 alloca 的地址，也需要判断是否是逃逸
-    // 例如：BitCastInst, PtrToIntInst, 如果这些操作将地址暴露，则不能提升
-  }
-  return true;
-}
-
-// 在迭代支配边界处插入 Phi 指令
-void Mem2Reg::insertPhiNodes(Function* func) {
-  // 清空上次 Phi 插入的结果
-  phiMap.clear();
-  allPhiInstructions.clear();
-
-  std::unordered_set<BasicBlock*> phiPlacementBlocks; // 存放需要插入 Phi 的块
-  std::queue<BasicBlock*> workQueue; // BFS 队列，用于迭代支配边界计算
-
-  // 遍历所有可提升的 alloca
-  for (AllocaInst* alloca : currentFunctionAllocas) {
-    phiPlacementBlocks.clear(); // 为每个 alloca 重新计算 Phi 放置位置
-    
-    // 初始化工作队列，放入所有定义该 alloca 的基本块
-    for (BasicBlock* defBB : allocaDefsBlock[alloca]) {
-      workQueue.push(defBB);
-    }
-
-    while (!workQueue.empty()) {
-      BasicBlock* currentBB = workQueue.front();
-      workQueue.pop();
-
-      auto* blockInfo = controlFlowAnalysis->getBlockAnalysisInfo(currentBB);
-      if (!blockInfo) continue;
-
-      // 遍历当前块的支配边界
-      for (BasicBlock* domFrontierBB : blockInfo->getDomFrontiers()) {
-        // 如果这个支配边界块还没有为当前 alloca 插入 Phi 指令
-        if (phiPlacementBlocks.find(domFrontierBB) == phiPlacementBlocks.end()) {
-          // 获取 alloca 的基类型，作为 Phi 指令的结果类型
-          Type* phiType = dynamic_cast<PointerType*>(alloca->getType())->getBaseType();
-          
-          // 在支配边界块的开头插入 Phi 指令
-          pBuilder->setPosition(domFrontierBB->begin());
-          PhiInst* newPhi = pBuilder->createPhiInst(phiType, {}, {}); // 初始入边为空
-          
-          allPhiInstructions.push_back(newPhi); // 记录所有 Phi
-          phiPlacementBlocks.insert(domFrontierBB); // 标记已插入
-
-          // 将 Phi 指令映射到它所代表的原始 alloca
-          phiMap[domFrontierBB][newPhi] = alloca;
-
-          // 如果支配边界块本身没有定义该 alloca，则其支配边界也可能需要 Phi
-          // 只有当这个块不是当前alloca的定义块时，才将其加入workQueue，以计算其DF。
-          if (allocaDefsBlock[alloca].find(domFrontierBB) == allocaDefsBlock[alloca].end()) {
-            workQueue.push(domFrontierBB);
-          }
-        }
-      }
-    }
-  }
-}
-
-// 获取前驱块在后继块前驱列表中的索引
-int Mem2Reg::getPredIndex(BasicBlock* pred, BasicBlock* succ) {
-  int index = 0;
-  for (auto* elem : succ->getPredecessors()) {
-    if (elem == pred) {
-      return index;
-    }
-    ++index;
-  }
-  // 断言通常在你的 IR 框架中应该确保前驱是存在的
-  // assert(false && "Predecessor not found in successor's predecessor list"); 
-  return -1; // 应该不会发生
-}
-
-// 递归地重命名基本块中的变量并填充 Phi 指令
-void Mem2Reg::renameBlock(BasicBlock* block, 
-                          std::unordered_map<AllocaInst*, Value*>& currentIncomings,
-                          std::unordered_set<BasicBlock*>& visitedBlocks) {
-  
-  // 记录在此块中发生的定义，以便在退出时将它们从栈中弹出
-  std::unordered_map<AllocaInst*, int> definitionsInBlockCount; 
-
-  // 如果已经访问过这个块，直接返回（防止无限循环或重复处理，在DFS中尤其重要）
-  if (visitedBlocks.count(block)) {
-    return;
-  }
-  visitedBlocks.insert(block);
-
-  // --- 1. 处理当前基本块内的指令 ---
-  // 使用迭代器安全地遍历和删除指令
-  for (auto it = block->getInstructions().begin(); it != block->getInstructions().end(); ) {
-    Instruction* currentInst = it->get();
-
-    if (AllocaInst* alloca = dynamic_cast<AllocaInst*>(currentInst)) {
-      // 如果是可提升的 alloca，标记为删除
-      if (std::find(currentFunctionAllocas.begin(), currentFunctionAllocas.end(), alloca) != currentFunctionAllocas.end()) {
-        SysYIROptUtils::usedelete(currentInst); // 标记为删除（或直接删除取决于你的 IR 管理）
-        it = block->getInstructions().erase(it); // 从列表中移除
-        continue; // 继续下一个指令
-      }
-    } else if (LoadInst* load = dynamic_cast<LoadInst*>(currentInst)) {
-      AllocaInst* originalAlloca = dynamic_cast<AllocaInst*>(load->getOperand(0)); // load 的第一个操作数是指针
-      if (originalAlloca && std::find(currentFunctionAllocas.begin(), currentFunctionAllocas.end(), originalAlloca) != currentFunctionAllocas.end()) {
-        // 如果是可提升 alloca 的 load 指令
-        Value* incomingVal = nullptr;
-        if (currentIncomings.count(originalAlloca)) {
-          incomingVal = currentIncomings[originalAlloca];
-        } else {
-          // 如果在当前路径上没有找到定义，则使用 UndefinedValue
-          incomingVal = UndefinedValue::get(originalAlloca->getType()->isPointer() ?
-                                            dynamic_cast<PointerType*>(originalAlloca->getType())->getBaseType() :
-                                            originalAlloca->getType());
-        }
-        
-        load->replaceAllUsesWith(incomingVal); // 用最新值替换所有 load 的用途
-        SysYIROptUtils::usedelete(currentInst);
-        it = block->getInstructions().erase(it);
-        continue;
-      }
-    } else if (StoreInst* store = dynamic_cast<StoreInst*>(currentInst)) {
-      AllocaInst* originalAlloca = dynamic_cast<AllocaInst*>(store->getOperand(1)); // store 的第二个操作数是指针
-      if (originalAlloca && std::find(currentFunctionAllocas.begin(), currentFunctionAllocas.end(), originalAlloca) != currentFunctionAllocas.end()) {
-        // 如果是可提升 alloca 的 store 指令，更新当前值
-        currentIncomings[originalAlloca] = store->getOperand(0); // store 的第一个操作数是值
-        definitionsInBlockCount[originalAlloca]++; // 记录在该块中进行的定义数量
-        SysYIROptUtils::usedelete(currentInst);
-        it = block->getInstructions().erase(it);
-        continue;
-      }
-    } else if (PhiInst* phi = dynamic_cast<PhiInst*>(currentInst)) {
-      // 如果是 Mem2Reg 插入的 Phi 指令 (通过 phiMap 判断)
-      if (phiMap[block].count(phi)) {
-        AllocaInst* originalAlloca = phiMap[block][phi];
-        currentIncomings[originalAlloca] = phi; // Phi 指令本身成为该变量的新定义
-        definitionsInBlockCount[originalAlloca]++; // 记录该 Phi 的定义
-      }
-    }
-    ++it; // 移动到下一个指令
-  }
-
-  // --- 2. 填充后继基本块中 Phi 指令的入边 ---
-  for (BasicBlock* successorBB : block->getSuccessors()) {
-    int predIndex = getPredIndex(block, successorBB);
-    if (predIndex == -1) continue; 
-
-    // Phi 指令总是在基本块的开头
-    for (auto& inst_ptr : successorBB->getInstructions()) {
-      if (PhiInst* phi = dynamic_cast<PhiInst*>(inst_ptr.get())) {
-        if (phiMap[successorBB].count(phi)) { // 确保这是我们关心的 Phi 指令
-          AllocaInst* originalAlloca = phiMap[successorBB][phi];
-          Value* incomingValue = nullptr;
-          
-          if (currentIncomings.count(originalAlloca)) {
-            incomingValue = currentIncomings[originalAlloca];
-          } else {
-            // 如果在当前块没有找到对应的定义，使用 UndefinedValue
-            incomingValue = UndefinedValue::get(originalAlloca->getType()->isPointer() ?
-                                                dynamic_cast<PointerType*>(originalAlloca->getType())->getBaseType() :
-                                                originalAlloca->getType());
-          }
-          
-          if (incomingValue) {
-            phi->addIncoming(incomingValue, block); // 添加 (值, 前驱块) 对
-          }
-        }
-      } else {
-        // 遇到非 Phi 指令，说明已经处理完所有 Phi，可以跳出
-        break; 
-      }
-    }
-  }
-
-  // --- 3. 递归调用支配树的子节点 ---
-  auto* blockInfo = controlFlowAnalysis->getBlockAnalysisInfo(block);
-  if (blockInfo) {
-    for (BasicBlock* dominatedChildBB : blockInfo->getSdoms()) { // getSdoms 获取直接支配的子节点
-      // 递归调用，传递当前 Incomings 的副本（或通过值传递以实现回溯）
-      // 注意：这里是传递 `currentIncomings` 的拷贝，以便递归返回后可以恢复。
-      // 但如果 `currentIncomings` 是引用传递，则这里需要回溯逻辑。
-      // 鉴于它是值传递，此处的 `definitionsInBlockCount` 仅用于统计，无需实际操作 `currentIncomings`。
-      renameBlock(dominatedChildBB, currentIncomings, visitedBlocks); 
-    }
-  }
-
-  // --- 4. 回溯：从栈中弹出在此块中创建的所有定义 ---
-  for (auto const& [alloca, count] : definitionsInBlockCount) {
-    // 在我们的实现中，`currentIncomings` 是通过值传递的，每次递归都收到一个新的拷贝。
-    // 因此，不需要显式地 "pop" 栈。`currentIncomings` 在函数返回时会自动销毁。
-    // 这种方式模拟了 "SSA 栈" 的行为，每个函数调用帧有自己的局部定义环境。
-  }
-}
-
-// 简化冗余的 Phi 指令 (当所有输入都相同时)
-void Mem2Reg::simplifyphi(PhiInst* phi) {
-  BasicBlock* phifromblock = phi->getParent();
-  if (!phifromblock) return; // 指令可能已经被删除
-
-  Value* commonValue = nullptr;
-  bool allSame = true;
-
-  // Phi 指令的操作数是 Value, BasicBlock 交替出现，所以是 getOperandSize() / 2 个入边
-  if (phi->getNumOperands() == 0) { // 空 Phi，通常是无效的，直接删除
-    phi->replaceAllUsesWith(UndefinedValue::get(phi->getType())); // 用 UndefinedValue 替换所有用途
-    // phi->getParent()->delete_inst(phi);
-    // 删除 Phi 指令后直接返回
-    // phi指令在开头一个比较快
-    // TODO：后续可优化查找
-    auto tofind = std::find_if(phifromblock->getInstructions().begin(), phifromblock->getInstructions().end(),
-                                   [phi](const auto &instr) { return instr.get() == phi; });
-    SysYIROptUtils::usedelete(phi); // 使用 SysYIROptUtils 删除指令
-    phifromblock->getInstructions().erase(tofind);
-    // 从基本块中删除 Phi 指令
-    return;
-  }
-
-  for (size_t i = 0; i < phi->getNumIncomingValues(); ++i) {
-    Value* incomingVal = phi->getOperand(2 * i); // 值位于偶数索引
-
-    if (incomingVal == phi) { // 如果 Phi 指令引用自身 (循环变量)
-      // 这种情况下，Phi 暂时不能简化，除非所有入边都是它自己，这通常通过其他优化处理
-      // 为避免复杂性，我们在此处不处理自引用 Phi 的简化，除非它是唯一选择。
-      // 更好的做法是，如果所有入边都指向自身，则该Phi是冗余的，可以替换为undef或其第一个实际值
-      // 但这需要更复杂的分析来确定循环的初始值。目前简单返回。
-      // TODO：留到后续循环优化处理
-      return; 
-    }
-
-    if (commonValue == nullptr) {
-      commonValue = incomingVal;
-    } else if (commonValue != incomingVal) {
-      allSame = false;
-      break; // 发现不同的入边值
-    }
-  }
-
-  if (allSame && commonValue != nullptr) {
-    // 所有入边值都相同，用这个值替换 Phi 指令的所有用途
-    phi->replaceAllUsesWith(commonValue);
-    // 从基本块中删除 Phi 指令
-    auto tofind = std::find_if(phifromblock->getInstructions().begin(), phifromblock->getInstructions().end(),
-                                   [phi](const auto &instr) { return instr.get() == phi; });
-    SysYIROptUtils::usedelete(phi); // 使用 SysYIROptUtils 删除指令
-    phifromblock->getInstructions().erase(tofind);
-  }
-}
-
-// 对单个函数执行内存到寄存器的提升
-bool Mem2Reg::promoteMemoryToRegisters(Function* func) {
-  bool changed = false;
-
-  // 每次开始对一个函数进行 Mem2Reg 时，清空所有上下文信息
-  currentFunctionAllocas.clear();
-  allocaDefsBlock.clear();
-  allocaUsesBlock.clear();
-  phiMap.clear();
-  allPhiInstructions.clear();
-
-  // 1. 收集所有可提升的 AllocaInst，并进行初步分析
-  BasicBlock* entryBB = func->getEntryBlock();
-  if (!entryBB) return false;
-
-  // 逆序遍历入口块的指令，安全地识别 Alloca
-  for (auto it = entryBB->getInstructions().rbegin(); it != entryBB->getInstructions().rend(); ++it) {
-    if (AllocaInst* alloca = dynamic_cast<AllocaInst*>(it->get())) {
-      if (is_promoted(alloca)) {
-        currentFunctionAllocas.push_back(alloca);
-      }
-    }
-  }
-  // 收集后反转，使其按原始顺序排列 (如果需要的话，但对后续分析影响不大)
-  std::reverse(currentFunctionAllocas.begin(), currentFunctionAllocas.end());
-
-  // 对收集到的所有 alloca 进行 DefsBlock 和 UsesBlock 分析
-  for (AllocaInst* alloca : currentFunctionAllocas) {
-    allocaAnalysis(alloca);
-  }
-
-  // 2. 预处理：删除无用的 AllocaInst (没有 Load 和 Store)
-  // 迭代 currentFunctionAllocas，安全删除
-  for (unsigned int i = 0; i < currentFunctionAllocas.size(); ) {
-    AllocaInst* alloca = currentFunctionAllocas[i];
-    
-    bool hasRelevantUse = false;
-    // 检查 alloca 的 uses 列表，看是否有 Load 或 Store
-    // 只要有 Load/Store，就认为是"相关用途"，不删除
-    for (auto use_ptr : alloca->getUses()) {
-      Instruction* user_inst = dynamic_cast<Instruction*>(use_ptr->getUser());
-      if (user_inst && (dynamic_cast<LoadInst*>(user_inst) || dynamic_cast<StoreInst*>(user_inst))) {
-        hasRelevantUse = true;
-        break;
-      }
-    }
-
-    // 如果没有相关用途（没有 Load 和 Store），则 alloca 是死代码
-    if (!hasRelevantUse && allocaDefsBlock[alloca].empty() && allocaUsesBlock[alloca].empty()) {
-      if (alloca->getParent()) {
-        // alloca->getParent()->delete_inst(alloca); // 从其所在块删除 alloca 指令
-        auto tofind = std::find_if(alloca->getParent()->getInstructions().begin(), alloca->getParent()->getInstructions().end(),
-                                   [alloca](const auto &instr) { return instr.get() == alloca; });
-        SysYIROptUtils::usedelete(alloca);
-        alloca->getParent()->getInstructions().erase(tofind);
-      }
-      currentFunctionAllocas.erase(currentFunctionAllocas.begin() + i); // 从列表中移除
-      changed = true; // 发生了改变
-    } else {
-      i++; // 否则，移动到下一个 alloca
-    }
-  }
-  
-  // 如果没有可提升的 alloca 了，直接返回
-  if (currentFunctionAllocas.empty()) {
-    return changed;
-  }
-
-  // 3. 插入 Phi 指令
-  insertPhiNodes(func);
-  if (!allPhiInstructions.empty()) changed = true;
-
-  // 4. 重命名变量，转换为 SSA 形式并填充 Phi 指令
-  std::unordered_map<AllocaInst*, Value*> initialIncomings;
-  std::unordered_set<BasicBlock*> visitedBlocks; // 用于 DFS 遍历，防止循环
-
-  // 初始化 entry block 的 Incomings 状态
-  for (AllocaInst* alloca : currentFunctionAllocas) {
-    initialIncomings[alloca] = UndefinedValue::get(dynamic_cast<PointerType*>(alloca->getType())->getBaseType());
-  }
-
-  // 从入口块开始递归重命名
-  renameBlock(entryBB, initialIncomings, visitedBlocks); 
-
-  // 5. 简化 Phi 指令
-  // 由于 renameBlock 可能会删除 Phi，这里复制一份列表以安全迭代
-  std::vector<PhiInst*> phisToSimplify = allPhiInstructions; 
-  for (PhiInst* phi : phisToSimplify) {
-    // 检查 phi 是否还在 IR 中 (可能已被其他优化删除)
-    // 一个简单检查是看它是否有父块
-    if (phi->getParent()) {
-      simplifyphi(phi);
-      // simplifyphi 内部会删除 Phi，所以这里不需要再处理 allPhiInstructions
-      // 最终的 allPhiInstructions 清理将在 promoteMemoryToRegisters 结束后进行
-    }
-  }
-  
-  // 清理所有 Phi 的列表和映射
-  // 遍历 allPhiInstructions，删除那些在 simplifyphi 后可能仍然存在的、但已经没有 uses 的 Phi
-  std::vector<PhiInst*> remainingPhis;
-  for(PhiInst* phi : allPhiInstructions) {
-    if(phi->getParent() && phi->getUses().empty()){ // 如果还在IR中但没有用处
-
-      // phi->getParent()->delete_inst(phi);
-      // 找到phi节点对应的迭代器
-      auto tofind = std::find_if(phi->getParent()->getInstructions().begin(), phi->getParent()->getInstructions().end(),
-                                   [phi](const auto &instr) { return instr.get() == phi; });
-      SysYIROptUtils::usedelete(phi); // 使用 SysYIROptUtils 删除指令
-      phi->getParent()->getInstructions().erase(tofind);
-      
-      changed = true;
-    } else if (phi->getParent()) { // 仍在IR中且有uses
-      remainingPhis.push_back(phi);
-    }
-  }
-  allPhiInstructions = remainingPhis; // 更新为仅包含未被删除的 Phi
-
-  // 重新清理 phiMap 中已经删除的 Phi 指令项
-  for (auto& pairBBPhiMap : phiMap) {
-    std::vector<PhiInst*> phisToRemoveFromMap;
-    for (auto& pairPhiAlloca : pairBBPhiMap.second) {
-      if (!pairPhiAlloca.first->getParent()) { // 如果 Phi 已经被删除
-        phisToRemoveFromMap.push_back(pairPhiAlloca.first);
-      }
-    }
-    for (PhiInst* phi : phisToRemoveFromMap) {
-      pairBBPhiMap.second.erase(phi);
-    }
-  }
-
-
-  return changed;
-}
-
-// --- run函数实现 ---
-void Mem2Reg::run() {
-  // 每次运行整个 Mem2Reg Pass 时，重新进行分析
-  controlFlowAnalysis->clear();
-  controlFlowAnalysis->runControlFlowAnalysis();
-  activeVarAnalysis->clear();
-  // 假设 dataFlowAnalysisUtils 可以管理和运行各个分析器
-  dataFlowAnalysisUtils.addBackwardAnalyzer(activeVarAnalysis); 
-  dataFlowAnalysisUtils.backwardAnalyze(pModule); // 运行活跃变量分析
-
-  bool globalChanged = false;
-  // 循环直到没有更多的 alloca 可以被提升
-  // 每次 promoteMemoryToRegisters 会尝试在一个函数内完成所有 Mem2Reg 优化
-  do {
-    globalChanged = false;
-    for (const auto& [_, func] : pModule->getFunctions()) {
-      // 对每个函数执行 Mem2Reg
-      if (promoteMemoryToRegisters(func.get())) {
-        globalChanged = true;
-        // 如果一个函数发生改变，可能影响其他函数或需要重新分析
-        // 因此需要重新运行控制流和活跃变量分析，以备下一次循环
-        controlFlowAnalysis->clear();
-        controlFlowAnalysis->runControlFlowAnalysis();
-        activeVarAnalysis->clear();
-        dataFlowAnalysisUtils.backwardAnalyze(pModule); // 重新分析活跃变量
-      }
-    }
-  } while (globalChanged); // 如果全局有任何函数发生改变，则继续迭代
-
-  // 最终清理和重新分析
-  controlFlowAnalysis->clear();
-  controlFlowAnalysis->runControlFlowAnalysis();
-  activeVarAnalysis->clear();
-  dataFlowAnalysisUtils.backwardAnalyze(pModule);
-}
-
-} // namespace sysy

src/RISCv64AsmPrinter.cpp

@@ -18,7 +18,7 @@ bool isMemoryOp(RVOpcodes opcode) {
 
 RISCv64AsmPrinter::RISCv64AsmPrinter(MachineFunction* mfunc) : MFunc(mfunc) {}
 
-void RISCv64AsmPrinter::run(std::ostream& os) {
+void RISCv64AsmPrinter::run(std::ostream& os, bool debug) {
     OS = &os;
 
     *OS << ".globl " << MFunc->getName() << "\n";
@@ -27,7 +27,7 @@ void RISCv64AsmPrinter::run(std::ostream& os) {
     printPrologue();
 
     for (auto& mbb : MFunc->getBlocks()) {
-        printBasicBlock(mbb.get());
+        printBasicBlock(mbb.get(), debug);
     }
 }
 
@@ -42,7 +42,7 @@ void RISCv64AsmPrinter::printPrologue() {
         *OS << "    addi sp, sp, -" << aligned_stack_size << "\n";
         *OS << "    sd ra, " << (aligned_stack_size - 8) << "(sp)\n";
         *OS << "    sd s0, " << (aligned_stack_size - 16) << "(sp)\n";
-        *OS << "    mv s0, sp\n";
+        *OS << "    addi s0, sp, " << aligned_stack_size << "\n";
     }
 
     // 忠实还原保存函数入口参数的逻辑
@@ -73,24 +73,31 @@ void RISCv64AsmPrinter::printEpilogue() {
     }
 }
 
-void RISCv64AsmPrinter::printBasicBlock(MachineBasicBlock* mbb) {
+void RISCv64AsmPrinter::printBasicBlock(MachineBasicBlock* mbb, bool debug) {
     if (!mbb->getName().empty()) {
         *OS << mbb->getName() << ":\n";
     }
     for (auto& instr : mbb->getInstructions()) {
-        printInstruction(instr.get());
+        printInstruction(instr.get(), debug);
     }
 }
 
-void RISCv64AsmPrinter::printInstruction(MachineInstr* instr) {
+void RISCv64AsmPrinter::printInstruction(MachineInstr* instr, bool debug) {
     auto opcode = instr->getOpcode();
     if (opcode == RVOpcodes::RET) {
         printEpilogue();
     }
-    if (opcode != RVOpcodes::LABEL) {
-        *OS << "    ";
+    
+    if (opcode == RVOpcodes::LABEL) {
+        // 标签直接打印，不加缩进
+        printOperand(instr->getOperands()[0].get());
+        *OS << ":\n";
+        return; // 处理完毕，直接返回
     }
     
+    // 对于所有非标签指令，先打印缩进
+    *OS << "    ";
+    
     switch (opcode) {
         case RVOpcodes::ADD:   *OS << "add ";   break; case RVOpcodes::ADDI:  *OS << "addi ";  break;
         case RVOpcodes::ADDW:  *OS << "addw ";  break; case RVOpcodes::ADDIW: *OS << "addiw "; break;
@@ -126,13 +133,21 @@ void RISCv64AsmPrinter::printInstruction(MachineInstr* instr) {
         case RVOpcodes::SNEZ:  *OS << "snez ";  break;
         case RVOpcodes::CALL:  *OS << "call ";  break;
         case RVOpcodes::LABEL:
-            printOperand(instr->getOperands()[0].get());
-            *OS << ":";
+            // printOperand(instr->getOperands()[0].get());
+            // *OS << ":";
             break;
         case RVOpcodes::FRAME_LOAD:
+            // It should have been eliminated by RegAlloc
+            if (!debug) throw std::runtime_error("FRAME pseudo-instruction not eliminated before AsmPrinter");
+            *OS << "frame_load ";  break;
         case RVOpcodes::FRAME_STORE:
-            // These should have been eliminated by RegAlloc
-            throw std::runtime_error("FRAME pseudo-instruction not eliminated before AsmPrinter");
+            // It should have been eliminated by RegAlloc
+            if (!debug) throw std::runtime_error("FRAME pseudo-instruction not eliminated before AsmPrinter");
+            *OS << "frame_store ";  break;
+        case RVOpcodes::FRAME_ADDR:
+            // It should have been eliminated by RegAlloc
+            if (!debug) throw std::runtime_error("FRAME pseudo-instruction not eliminated before AsmPrinter");
+            *OS << "frame_addr "; break;
         default: 
             throw std::runtime_error("Unknown opcode in AsmPrinter");
     }

src/RISCv64Backend.cpp

@@ -61,6 +61,10 @@ std::string RISCv64CodeGen::function_gen(Function* func) {
     RISCv64ISel isel;
     std::unique_ptr<MachineFunction> mfunc = isel.runOnFunction(func);
 
+    std::stringstream ss1;
+    RISCv64AsmPrinter printer1(mfunc.get());
+    printer1.run(ss1, true);
+
     // 阶段 2: 指令调度 (Instruction Scheduling)
     PreRA_Scheduler scheduler;
     scheduler.runOnMachineFunction(mfunc.get());
@@ -81,7 +85,7 @@ std::string RISCv64CodeGen::function_gen(Function* func) {
     std::stringstream ss;
     RISCv64AsmPrinter printer(mfunc.get());
     printer.run(ss);
-
+    if (DEBUG) ss << ss1.str(); // 将指令选择阶段的结果也包含在最终输出中
     return ss.str();
 }
 

src/RISCv64ISel.cpp

@@ -4,6 +4,7 @@
 #include <functional>
 #include <cmath> // For std::fabs
 #include <limits> // For std::numeric_limits
+#include <iostream>
 
 namespace sysy {
 
@@ -82,6 +83,10 @@ void RISCv64ISel::selectBasicBlock(BasicBlock* bb) {
     CurMBB = bb_map.at(bb);
     auto dag = build_dag(bb);
 
+    if (DEBUG) { // 使用 DEBUG 宏或变量来控制是否打印
+        print_dag(dag, bb->getName());
+    }
+
     std::map<Value*, DAGNode*> value_to_node;
     for(const auto& node : dag) {
         if (node->value) {
@@ -92,6 +97,10 @@ void RISCv64ISel::selectBasicBlock(BasicBlock* bb) {
     std::set<DAGNode*> selected_nodes;
     std::function<void(DAGNode*)> select_recursive = 
         [&](DAGNode* node) {
+        if (DEEPDEBUG) {
+            std::cout << "[DEEPDEBUG] select_recursive: Visiting node with kind: " << node->kind 
+                    << " (Value: " << (node->value ? node->value->getName() : "null") << ")" << std::endl;
+        }
         if (!node || selected_nodes.count(node)) return;
         for (auto operand : node->operands) {
             select_recursive(operand);
@@ -118,24 +127,37 @@ void RISCv64ISel::selectBasicBlock(BasicBlock* bb) {
     }
 }
 
-// 核心函数：为DAG节点选择并生成MachineInstr (忠实移植版)
+// 核心函数：为DAG节点选择并生成MachineInstr (已修复和增强的完整版本)
 void RISCv64ISel::selectNode(DAGNode* node) {
+    // 调用者（select_recursive）已经保证了操作数节点会先于当前节点被选择。
+    // 因此，这里我们只处理当前节点。
+
     switch (node->kind) {
+        // [V2优点] 采纳“延迟物化”（Late Materialization）思想。
+        // 这两个节点仅作为标记，不直接生成指令。它们的目的是在DAG中保留类型信息。
+        // 加载其值的责任，被转移给了使用它们的父节点（如STORE, BINARY等）。
+        // 这修复了之前版本中“使用未初始化虚拟寄存器”的根本性bug。
         case DAGNode::CONSTANT:
         case DAGNode::ALLOCA_ADDR:
-            if (node->value) getVReg(node->value);
+            if (node->value) {
+                // 确保它有一个关联的虚拟寄存器即可，不生成代码。
+                getVReg(node->value);
+            }
             break;
         
         case DAGNode::LOAD: {
             auto dest_vreg = getVReg(node->value);
             Value* ptr_val = node->operands[0]->value;
 
+            // [V1设计保留] 对于从栈变量加载，继续使用伪指令 FRAME_LOAD。
+            // 这种设计将栈帧布局的具体计算推迟到后续的 `eliminateFrameIndices` 阶段，保持了模块化。
             if (auto alloca = dynamic_cast<AllocaInst*>(ptr_val)) {
                 auto instr = std::make_unique<MachineInstr>(RVOpcodes::FRAME_LOAD);
                 instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
                 instr->addOperand(std::make_unique<RegOperand>(getVReg(alloca)));
                 CurMBB->addInstruction(std::move(instr));
             } else if (auto global = dynamic_cast<GlobalValue*>(ptr_val)) {
+                // 对于全局变量，先用 la 加载其地址，再用 lw 加载其值。
                 auto addr_vreg = getNewVReg();
                 auto la = std::make_unique<MachineInstr>(RVOpcodes::LA);
                 la->addOperand(std::make_unique<RegOperand>(addr_vreg));
@@ -150,6 +172,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                 ));
                 CurMBB->addInstruction(std::move(lw));
             } else {
+                // 对于已经在虚拟寄存器中的指针地址，直接通过该地址加载。
                 auto ptr_vreg = getVReg(ptr_val);
                 auto lw = std::make_unique<MachineInstr>(RVOpcodes::LW);
                 lw->addOperand(std::make_unique<RegOperand>(dest_vreg));
@@ -166,7 +189,13 @@ void RISCv64ISel::selectNode(DAGNode* node) {
             Value* val_to_store = node->operands[0]->value;
             Value* ptr_val = node->operands[1]->value;
 
+            // [V2优点] 在STORE节点内部负责加载作为源的常量。
+            // 如果要存储的值是一个常量，就在这里生成 `li` 指令加载它。
             if (auto val_const = dynamic_cast<ConstantValue*>(val_to_store)) {
+                if (DEBUG) {
+                    std::cout << "[DEBUG] selectNode-BINARY: Found constant operand with value " << val_const->getInt()
+                            << ". Generating LI instruction." << std::endl;
+                }
                 auto li = std::make_unique<MachineInstr>(RVOpcodes::LI);
                 li->addOperand(std::make_unique<RegOperand>(getVReg(val_const)));
                 li->addOperand(std::make_unique<ImmOperand>(val_const->getInt()));
@@ -174,13 +203,15 @@ void RISCv64ISel::selectNode(DAGNode* node) {
             }
             auto val_vreg = getVReg(val_to_store);
 
+            // [V1设计保留] 同样，对于向栈变量的存储，使用 FRAME_STORE 伪指令。
             if (auto alloca = dynamic_cast<AllocaInst*>(ptr_val)) {
                 auto instr = std::make_unique<MachineInstr>(RVOpcodes::FRAME_STORE);
                 instr->addOperand(std::make_unique<RegOperand>(val_vreg));
                 instr->addOperand(std::make_unique<RegOperand>(getVReg(alloca)));
                 CurMBB->addInstruction(std::move(instr));
             } else if (auto global = dynamic_cast<GlobalValue*>(ptr_val)) {
-                 auto addr_vreg = getNewVReg();
+                // 向全局变量存储。
+                auto addr_vreg = getNewVReg();
                 auto la = std::make_unique<MachineInstr>(RVOpcodes::LA);
                 la->addOperand(std::make_unique<RegOperand>(addr_vreg));
                 la->addOperand(std::make_unique<LabelOperand>(global->getName()));
@@ -194,6 +225,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                 ));
                 CurMBB->addInstruction(std::move(sw));
             } else {
+                // 向一个指针（存储在虚拟寄存器中）指向的地址存储。
                 auto ptr_vreg = getVReg(ptr_val);
                 auto sw = std::make_unique<MachineInstr>(RVOpcodes::SW);
                 sw->addOperand(std::make_unique<RegOperand>(val_vreg));
@@ -210,37 +242,108 @@ void RISCv64ISel::selectNode(DAGNode* node) {
             auto bin = dynamic_cast<BinaryInst*>(node->value);
             Value* lhs = bin->getLhs();
             Value* rhs = bin->getRhs();
+            
+            if (bin->getKind() == BinaryInst::kAdd) {
+                Value* base = nullptr;
+                Value* offset = nullptr;
 
+                // [修改] 扩展基地址的判断，使其可以识别 AllocaInst 或 GlobalValue
+                if (dynamic_cast<AllocaInst*>(lhs) || dynamic_cast<GlobalValue*>(lhs)) {
+                    base = lhs;
+                    offset = rhs;
+                } else if (dynamic_cast<AllocaInst*>(rhs) || dynamic_cast<GlobalValue*>(rhs)) {
+                    base = rhs;
+                    offset = lhs;
+                }
+                
+                // 如果成功匹配到地址计算模式
+                if (base) {
+                    // 1. 先为偏移量加载常量（如果它是常量的话）
+                    if (auto const_offset = dynamic_cast<ConstantValue*>(offset)) {
+                        auto li = std::make_unique<MachineInstr>(RVOpcodes::LI);
+                        li->addOperand(std::make_unique<RegOperand>(getVReg(const_offset)));
+                        li->addOperand(std::make_unique<ImmOperand>(const_offset->getInt()));
+                        CurMBB->addInstruction(std::move(li));
+                    }
+                    
+                    // 2. [修改] 根据基地址的类型，生成不同的指令来获取基地址
+                    auto base_addr_vreg = getNewVReg(); // 创建一个新的临时vreg来存放基地址
+
+                    // 情况一：基地址是局部栈变量
+                    if (auto alloca_base = dynamic_cast<AllocaInst*>(base)) {
+                        auto frame_addr_instr = std::make_unique<MachineInstr>(RVOpcodes::FRAME_ADDR);
+                        frame_addr_instr->addOperand(std::make_unique<RegOperand>(base_addr_vreg));
+                        frame_addr_instr->addOperand(std::make_unique<RegOperand>(getVReg(alloca_base)));
+                        CurMBB->addInstruction(std::move(frame_addr_instr));
+                    } 
+                    // 情况二：基地址是全局变量
+                    else if (auto global_base = dynamic_cast<GlobalValue*>(base)) {
+                        auto la_instr = std::make_unique<MachineInstr>(RVOpcodes::LA);
+                        la_instr->addOperand(std::make_unique<RegOperand>(base_addr_vreg));
+                        la_instr->addOperand(std::make_unique<LabelOperand>(global_base->getName()));
+                        CurMBB->addInstruction(std::move(la_instr));
+                    }
+
+                    // 3. 生成真正的add指令，计算最终地址（这部分逻辑保持不变）
+                    auto final_addr_vreg = getVReg(bin); // 这是整个二元运算的结果vreg
+                    auto offset_vreg = getVReg(offset);
+                    auto add_instr = std::make_unique<MachineInstr>(RVOpcodes::ADD); // 指针运算是64位
+                    add_instr->addOperand(std::make_unique<RegOperand>(final_addr_vreg));
+                    add_instr->addOperand(std::make_unique<RegOperand>(base_addr_vreg));
+                    add_instr->addOperand(std::make_unique<RegOperand>(offset_vreg));
+                    CurMBB->addInstruction(std::move(add_instr));
+                    
+                    return; // 地址计算处理完毕，直接返回
+                }
+            }
+
+            // [V2优点] 在BINARY节点内部按需加载常量操作数。
             auto load_val_if_const = [&](Value* val) {
                 if (auto c = dynamic_cast<ConstantValue*>(val)) {
+                    if (DEBUG) {
+                        std::cout << "[DEBUG] selectNode-BINARY: Found constant operand with value " << c->getInt()
+                                << ". Generating LI instruction." << std::endl;
+                    }
                     auto li = std::make_unique<MachineInstr>(RVOpcodes::LI);
                     li->addOperand(std::make_unique<RegOperand>(getVReg(c)));
                     li->addOperand(std::make_unique<ImmOperand>(c->getInt()));
                     CurMBB->addInstruction(std::move(li));
                 }
             };
-            load_val_if_const(lhs);
-            load_val_if_const(rhs);
 
-            auto dest_vreg = getVReg(bin);
-            auto lhs_vreg = getVReg(lhs);
-            auto rhs_vreg = getVReg(rhs);
-
-            if (bin->getKind() == BinaryInst::kAdd) {
-                if (auto rhs_const = dynamic_cast<ConstantValue*>(rhs)) {
-                    if (rhs_const->getInt() >= -2048 && rhs_const->getInt() < 2048) {
-                        auto instr = std::make_unique<MachineInstr>(RVOpcodes::ADDIW);
-                        instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
-                        instr->addOperand(std::make_unique<RegOperand>(lhs_vreg));
-                        instr->addOperand(std::make_unique<ImmOperand>(rhs_const->getInt()));
-                        CurMBB->addInstruction(std::move(instr));
-                        return;
-                    }
+            // 检查是否能应用立即数优化。
+            bool rhs_is_imm_opt = false;
+            if (auto rhs_const = dynamic_cast<ConstantValue*>(rhs)) {
+                if (bin->getKind() == BinaryInst::kAdd && rhs_const->getInt() >= -2048 && rhs_const->getInt() < 2048) {
+                    rhs_is_imm_opt = true;
                 }
             }
 
+            // 仅在不能作为立即数操作数时才需要提前加载。
+            load_val_if_const(lhs);
+            if (!rhs_is_imm_opt) {
+                load_val_if_const(rhs);
+            }
+
+            auto dest_vreg = getVReg(bin);
+            auto lhs_vreg = getVReg(lhs);
+
+            // [V2优点] 融合 ADDIW 优化。
+            if (rhs_is_imm_opt) {
+                auto rhs_const = dynamic_cast<ConstantValue*>(rhs);
+                auto instr = std::make_unique<MachineInstr>(RVOpcodes::ADDIW);
+                instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                instr->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                instr->addOperand(std::make_unique<ImmOperand>(rhs_const->getInt()));
+                CurMBB->addInstruction(std::move(instr));
+                return; // 指令已生成，直接返回。
+            }
+            
+            auto rhs_vreg = getVReg(rhs);
+
             switch (bin->getKind()) {
                 case BinaryInst::kAdd: {
+                    // 区分指针运算（64位）和整数运算（32位）。
                     RVOpcodes opcode = (lhs->getType()->isPointer() || rhs->getType()->isPointer()) ? RVOpcodes::ADD : RVOpcodes::ADDW;
                     auto instr = std::make_unique<MachineInstr>(opcode);
                     instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
@@ -281,7 +384,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                     CurMBB->addInstruction(std::move(instr));
                     break;
                 }
-                case BinaryInst::kICmpEQ: {
+                case BinaryInst::kICmpEQ: { // 等于 (a == b) -> (subw; seqz)
                     auto sub = std::make_unique<MachineInstr>(RVOpcodes::SUBW);
                     sub->addOperand(std::make_unique<RegOperand>(dest_vreg));
                     sub->addOperand(std::make_unique<RegOperand>(lhs_vreg));
@@ -294,7 +397,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                     CurMBB->addInstruction(std::move(seqz));
                     break;
                 }
-                case BinaryInst::kICmpNE: {
+                case BinaryInst::kICmpNE: { // 不等于 (a != b) -> (subw; snez)
                     auto sub = std::make_unique<MachineInstr>(RVOpcodes::SUBW);
                     sub->addOperand(std::make_unique<RegOperand>(dest_vreg));
                     sub->addOperand(std::make_unique<RegOperand>(lhs_vreg));
@@ -307,7 +410,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                     CurMBB->addInstruction(std::move(snez));
                     break;
                 }
-                case BinaryInst::kICmpLT: {
+                case BinaryInst::kICmpLT: { // 小于 (a < b) -> slt
                     auto instr = std::make_unique<MachineInstr>(RVOpcodes::SLT);
                     instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
                     instr->addOperand(std::make_unique<RegOperand>(lhs_vreg));
@@ -315,7 +418,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                     CurMBB->addInstruction(std::move(instr));
                     break;
                 }
-                case BinaryInst::kICmpGT: {
+                case BinaryInst::kICmpGT: { // 大于 (a > b) -> (b < a) -> slt
                     auto instr = std::make_unique<MachineInstr>(RVOpcodes::SLT);
                     instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
                     instr->addOperand(std::make_unique<RegOperand>(rhs_vreg));
@@ -323,7 +426,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                     CurMBB->addInstruction(std::move(instr));
                     break;
                 }
-                case BinaryInst::kICmpLE: {
+                case BinaryInst::kICmpLE: { // 小于等于 (a <= b) -> !(b < a) -> (slt; xori)
                     auto slt = std::make_unique<MachineInstr>(RVOpcodes::SLT);
                     slt->addOperand(std::make_unique<RegOperand>(dest_vreg));
                     slt->addOperand(std::make_unique<RegOperand>(rhs_vreg));
@@ -337,7 +440,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                     CurMBB->addInstruction(std::move(xori));
                     break;
                 }
-                 case BinaryInst::kICmpGE: {
+                 case BinaryInst::kICmpGE: { // 大于等于 (a >= b) -> !(a < b) -> (slt; xori)
                     auto slt = std::make_unique<MachineInstr>(RVOpcodes::SLT);
                     slt->addOperand(std::make_unique<RegOperand>(dest_vreg));
                     slt->addOperand(std::make_unique<RegOperand>(lhs_vreg));
@@ -363,7 +466,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
             auto src_vreg = getVReg(unary->getOperand());
 
             switch (unary->getKind()) {
-                case UnaryInst::kNeg: {
+                case UnaryInst::kNeg: { // 取负: 0 - src
                     auto instr = std::make_unique<MachineInstr>(RVOpcodes::SUBW);
                     instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
                     instr->addOperand(std::make_unique<RegOperand>(PhysicalReg::ZERO));
@@ -371,7 +474,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                     CurMBB->addInstruction(std::move(instr));
                     break;
                 }
-                case UnaryInst::kNot: {
+                case UnaryInst::kNot: { // 逻辑非: src == 0 ? 1 : 0
                     auto instr = std::make_unique<MachineInstr>(RVOpcodes::SEQZ);
                     instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
                     instr->addOperand(std::make_unique<RegOperand>(src_vreg));
@@ -386,7 +489,10 @@ void RISCv64ISel::selectNode(DAGNode* node) {
 
         case DAGNode::CALL: {
             auto call = dynamic_cast<CallInst*>(node->value);
-            for (size_t i = 0; i < node->operands.size() && i < 8; ++i) {
+            // 处理函数参数，放入a0-a7物理寄存器
+            size_t num_operands = node->operands.size();
+            size_t reg_arg_count = std::min(num_operands, (size_t)8);
+            for (size_t i = 0; i < reg_arg_count; ++i) {
                 DAGNode* arg_node = node->operands[i];
                 auto arg_preg = static_cast<PhysicalReg>(static_cast<int>(PhysicalReg::A0) + i);
 
@@ -405,11 +511,64 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                     CurMBB->addInstruction(std::move(mv));
                 }
             }
+            if (num_operands > 8) {
+                size_t stack_arg_count = num_operands - 8;
+                int stack_space = stack_arg_count * 8; // RV64中每个参数槽位8字节
+
+                // 2a. 在栈上分配空间
+                auto alloc_instr = std::make_unique<MachineInstr>(RVOpcodes::ADDI);
+                alloc_instr->addOperand(std::make_unique<RegOperand>(PhysicalReg::SP));
+                alloc_instr->addOperand(std::make_unique<RegOperand>(PhysicalReg::SP));
+                alloc_instr->addOperand(std::make_unique<ImmOperand>(-stack_space));
+                CurMBB->addInstruction(std::move(alloc_instr));
+
+                // 2b. 存储每个栈参数
+                for (size_t i = 8; i < num_operands; ++i) {
+                    DAGNode* arg_node = node->operands[i];
+                    unsigned src_vreg;
+                    
+                    // 准备源寄存器
+                    if (arg_node->kind == DAGNode::CONSTANT) {
+                        // 如果是常量，先加载到临时寄存器
+                        src_vreg = getNewVReg();
+                        auto const_val = dynamic_cast<ConstantValue*>(arg_node->value);
+                        auto li = std::make_unique<MachineInstr>(RVOpcodes::LI);
+                        li->addOperand(std::make_unique<RegOperand>(src_vreg));
+                        li->addOperand(std::make_unique<ImmOperand>(const_val->getInt()));
+                        CurMBB->addInstruction(std::move(li));
+                    } else {
+                        src_vreg = getVReg(arg_node->value);
+                    }
+                    
+                    // 计算在栈上的偏移量
+                    int offset = (i - 8) * 8;
+                    
+                    // 生成 sd 指令
+                    auto sd_instr = std::make_unique<MachineInstr>(RVOpcodes::SD);
+                    sd_instr->addOperand(std::make_unique<RegOperand>(src_vreg));
+                    sd_instr->addOperand(std::make_unique<MemOperand>(
+                        std::make_unique<RegOperand>(PhysicalReg::SP),
+                        std::make_unique<ImmOperand>(offset)
+                    ));
+                    CurMBB->addInstruction(std::move(sd_instr));
+                }
+            }
 
             auto call_instr = std::make_unique<MachineInstr>(RVOpcodes::CALL);
             call_instr->addOperand(std::make_unique<LabelOperand>(call->getCallee()->getName()));
             CurMBB->addInstruction(std::move(call_instr));
+            
+            if (num_operands > 8) {
+                size_t stack_arg_count = num_operands - 8;
+                int stack_space = stack_arg_count * 8;
 
+                auto dealloc_instr = std::make_unique<MachineInstr>(RVOpcodes::ADDI);
+                dealloc_instr->addOperand(std::make_unique<RegOperand>(PhysicalReg::SP));
+                dealloc_instr->addOperand(std::make_unique<RegOperand>(PhysicalReg::SP));
+                dealloc_instr->addOperand(std::make_unique<ImmOperand>(stack_space));
+                CurMBB->addInstruction(std::move(dealloc_instr));
+            }
+            // 处理返回值，从a0移动到目标虚拟寄存器
             if (!call->getType()->isVoid()) {
                 auto mv_instr = std::make_unique<MachineInstr>(RVOpcodes::MV);
                 mv_instr->addOperand(std::make_unique<RegOperand>(getVReg(call)));
@@ -423,6 +582,7 @@ void RISCv64ISel::selectNode(DAGNode* node) {
             auto ret_inst_ir = dynamic_cast<ReturnInst*>(node->value);
             if (ret_inst_ir && ret_inst_ir->hasReturnValue()) {
                 Value* ret_val = ret_inst_ir->getReturnValue();
+                // [V2优点] 在RETURN节点内加载常量返回值
                 if (auto const_val = dynamic_cast<ConstantValue*>(ret_val)) {
                     auto li_instr = std::make_unique<MachineInstr>(RVOpcodes::LI);
                     li_instr->addOperand(std::make_unique<RegOperand>(PhysicalReg::A0));
@@ -435,36 +595,121 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                     CurMBB->addInstruction(std::move(mv_instr));
                 }
             }
+            // [V1设计保留] 函数尾声（epilogue）不由RETURN节点生成，
+            // 而是由后续的AsmPrinter或其它Pass统一处理，这是一种常见且有效的模块化设计。
             auto ret_mi = std::make_unique<MachineInstr>(RVOpcodes::RET);
             CurMBB->addInstruction(std::move(ret_mi));
             break;
         }
 
         case DAGNode::BRANCH: {
-            if (auto cond_br = dynamic_cast<CondBrInst*>(node->value)) {
+        // 处理条件分支
+        if (auto cond_br = dynamic_cast<CondBrInst*>(node->value)) {
+            Value* condition = cond_br->getCondition();
+            auto then_bb_name = cond_br->getThenBlock()->getName();
+            auto else_bb_name = cond_br->getElseBlock()->getName();
+
+            // [优化] 检查分支条件是否为编译期常量
+            if (auto const_cond = dynamic_cast<ConstantValue*>(condition)) {
+                // 如果条件是常量，直接生成一个无条件跳转J，而不是BNE
+                if (const_cond->getInt() != 0) { // 条件为 true
+                    auto j_instr = std::make_unique<MachineInstr>(RVOpcodes::J);
+                    j_instr->addOperand(std::make_unique<LabelOperand>(then_bb_name));
+                    CurMBB->addInstruction(std::move(j_instr));
+                } else { // 条件为 false
+                    auto j_instr = std::make_unique<MachineInstr>(RVOpcodes::J);
+                    j_instr->addOperand(std::make_unique<LabelOperand>(else_bb_name));
+                    CurMBB->addInstruction(std::move(j_instr));
+                }
+            } 
+            // 如果条件不是常量，则执行标准流程
+            else {
+                // [修复] 为条件变量生成加载指令（如果它是常量的话，尽管上面已经处理了）
+                // 这一步是为了逻辑完整，以防有其他类型的常量没有被捕获
+                if (auto const_val = dynamic_cast<ConstantValue*>(condition)) {
+                    auto li = std::make_unique<MachineInstr>(RVOpcodes::LI);
+                    li->addOperand(std::make_unique<RegOperand>(getVReg(const_val)));
+                    li->addOperand(std::make_unique<ImmOperand>(const_val->getInt()));
+                    CurMBB->addInstruction(std::move(li));
+                }
+
+                auto cond_vreg = getVReg(condition);
+                
+                // 生成 bne cond, zero, then_label (如果cond不为0，则跳转到then)
                 auto br_instr = std::make_unique<MachineInstr>(RVOpcodes::BNE);
-                br_instr->addOperand(std::make_unique<RegOperand>(getVReg(cond_br->getCondition())));
+                br_instr->addOperand(std::make_unique<RegOperand>(cond_vreg));
                 br_instr->addOperand(std::make_unique<RegOperand>(PhysicalReg::ZERO));
-                br_instr->addOperand(std::make_unique<LabelOperand>(cond_br->getThenBlock()->getName()));
+                br_instr->addOperand(std::make_unique<LabelOperand>(then_bb_name));
                 CurMBB->addInstruction(std::move(br_instr));
-            } else if (auto uncond_br = dynamic_cast<UncondBrInst*>(node->value)) {
+                
+                // 为else分支生成无条件跳转 (后续Pass可以优化掉不必要的跳转)
                 auto j_instr = std::make_unique<MachineInstr>(RVOpcodes::J);
-                j_instr->addOperand(std::make_unique<LabelOperand>(uncond_br->getBlock()->getName()));
+                j_instr->addOperand(std::make_unique<LabelOperand>(else_bb_name));
                 CurMBB->addInstruction(std::move(j_instr));
             }
-            break;
+        } 
+        // 处理无条件分支
+        else if (auto uncond_br = dynamic_cast<UncondBrInst*>(node->value)) {
+            auto j_instr = std::make_unique<MachineInstr>(RVOpcodes::J);
+            j_instr->addOperand(std::make_unique<LabelOperand>(uncond_br->getBlock()->getName()));
+            CurMBB->addInstruction(std::move(j_instr));
         }
+        break;
+    }
 
         case DAGNode::MEMSET: {
+            // [V1设计保留] Memset的核心展开逻辑在虚拟寄存器层面是正确的，无需修改。
+            // 之前的bug是由于其输入（地址、值、大小）的虚拟寄存器未被正确初始化。
+            // 在修复了CONSTANT/ALLOCA_ADDR的加载问题后，此处的逻辑现在可以正常工作。
+
+            if (DEBUG) {
+                std::cout << "[DEBUG] selectNode-MEMSET: Processing MEMSET node." << std::endl;
+            }
             auto memset = dynamic_cast<MemsetInst*>(node->value);
+            Value* val_to_set = memset->getValue();
+            Value* size_to_set = memset->getSize();
+            Value* ptr_val = memset->getPointer();
+            auto dest_addr_vreg = getVReg(ptr_val);
+
+            if (auto const_val = dynamic_cast<ConstantValue*>(val_to_set)) {
+                if (DEBUG) {
+                    std::cout << "[DEBUG] selectNode-MEMSET: Found constant 'value' operand (" << const_val->getInt() << "). Generating LI." << std::endl;
+                }
+                auto li = std::make_unique<MachineInstr>(RVOpcodes::LI);
+                li->addOperand(std::make_unique<RegOperand>(getVReg(const_val)));
+                li->addOperand(std::make_unique<ImmOperand>(const_val->getInt()));
+                CurMBB->addInstruction(std::move(li));
+            }
+            if (auto const_size = dynamic_cast<ConstantValue*>(size_to_set)) {
+                if (DEBUG) {
+                    std::cout << "[DEBUG] selectNode-MEMSET: Found constant 'size' operand (" << const_size->getInt() << "). Generating LI." << std::endl;
+                }
+                auto li = std::make_unique<MachineInstr>(RVOpcodes::LI);
+                li->addOperand(std::make_unique<RegOperand>(getVReg(const_size)));
+                li->addOperand(std::make_unique<ImmOperand>(const_size->getInt()));
+                CurMBB->addInstruction(std::move(li));
+            }
+            if (auto alloca = dynamic_cast<AllocaInst*>(ptr_val)) {
+                if (DEBUG) {
+                    std::cout << "[DEBUG] selectNode-MEMSET: Found 'pointer' operand is an AllocaInst. Generating FRAME_ADDR." << std::endl;
+                }
+                // 生成新的伪指令来获取栈地址
+                auto instr = std::make_unique<MachineInstr>(RVOpcodes::FRAME_ADDR);
+                instr->addOperand(std::make_unique<RegOperand>(dest_addr_vreg)); // 目标虚拟寄存器
+                instr->addOperand(std::make_unique<RegOperand>(getVReg(alloca)));   // 源AllocaInst
+                CurMBB->addInstruction(std::move(instr));
+            }
             auto r_dest_addr = getVReg(memset->getPointer());
             auto r_num_bytes = getVReg(memset->getSize());
             auto r_value_byte = getVReg(memset->getValue());
+            
+            // 为memset内部逻辑创建新的临时虚拟寄存器
             auto r_counter = getNewVReg();
             auto r_end_addr = getNewVReg();
             auto r_current_addr = getNewVReg();
             auto r_temp_val = getNewVReg();
 
+            // 定义一系列lambda表达式来简化指令创建
             auto add_instr = [&](RVOpcodes op, unsigned rd, unsigned rs1, unsigned rs2) {
                 auto i = std::make_unique<MachineInstr>(op);
                 i->addOperand(std::make_unique<RegOperand>(rd));
@@ -503,12 +748,14 @@ void RISCv64ISel::selectNode(DAGNode* node) {
                 CurMBB->addInstruction(std::move(i));
             };
 
+            // 生成唯一的循环标签
             int unique_id = this->local_label_counter++;
             std::string loop_start_label = MFunc->getName() + "_memset_loop_start_" + std::to_string(unique_id);
             std::string loop_end_label = MFunc->getName() + "_memset_loop_end_" + std::to_string(unique_id);
             std::string remainder_label = MFunc->getName() + "_memset_remainder_" + std::to_string(unique_id);
             std::string done_label = MFunc->getName() + "_memset_done_" + std::to_string(unique_id);
             
+            // 构造64位的填充值
             addi_instr(RVOpcodes::ANDI, r_temp_val, r_value_byte, 255);
             addi_instr(RVOpcodes::SLLI, r_value_byte, r_temp_val, 8);
             add_instr(RVOpcodes::OR, r_temp_val, r_temp_val, r_value_byte);
@@ -516,6 +763,8 @@ void RISCv64ISel::selectNode(DAGNode* node) {
             add_instr(RVOpcodes::OR, r_temp_val, r_temp_val, r_value_byte);
             addi_instr(RVOpcodes::SLLI, r_value_byte, r_temp_val, 32);
             add_instr(RVOpcodes::OR, r_temp_val, r_temp_val, r_value_byte);
+            
+            // 计算循环边界
             add_instr(RVOpcodes::ADD, r_end_addr, r_dest_addr, r_num_bytes);
             auto mv = std::make_unique<MachineInstr>(RVOpcodes::MV);
             mv->addOperand(std::make_unique<RegOperand>(r_current_addr));
@@ -523,17 +772,22 @@ void RISCv64ISel::selectNode(DAGNode* node) {
             CurMBB->addInstruction(std::move(mv));
             addi_instr(RVOpcodes::ANDI, r_counter, r_num_bytes, -8);
             add_instr(RVOpcodes::ADD, r_counter, r_dest_addr, r_counter);
+            
+            // 8字节主循环
             label_instr(loop_start_label);
             branch_instr(RVOpcodes::BGEU, r_current_addr, r_counter, loop_end_label);
             store_instr(RVOpcodes::SD, r_temp_val, r_current_addr, 0);
             addi_instr(RVOpcodes::ADDI, r_current_addr, r_current_addr, 8);
             jump_instr(loop_start_label);
+            
+            // 1字节收尾循环
             label_instr(loop_end_label);
             label_instr(remainder_label);
             branch_instr(RVOpcodes::BGEU, r_current_addr, r_end_addr, done_label);
             store_instr(RVOpcodes::SB, r_temp_val, r_current_addr, 0);
             addi_instr(RVOpcodes::ADDI, r_current_addr, r_current_addr, 1);
             jump_instr(remainder_label);
+            
             label_instr(done_label);
             break;
         }
@@ -590,6 +844,12 @@ std::vector<std::unique_ptr<RISCv64ISel::DAGNode>> RISCv64ISel::build_dag(BasicB
             memset_node->operands.push_back(get_operand_node(memset->getBegin(), value_to_node, nodes_storage));
             memset_node->operands.push_back(get_operand_node(memset->getSize(), value_to_node, nodes_storage));
             memset_node->operands.push_back(get_operand_node(memset->getValue(), value_to_node, nodes_storage));
+            if (DEBUG) {
+                std::cout << "[DEBUG] build_dag: Created MEMSET node for: " << memset->getName() << std::endl;
+                for (size_t i = 0; i < memset_node->operands.size(); ++i) {
+                    std::cout << "  -> Operand " << i << " has kind: " << memset_node->operands[i]->kind << std::endl;
+                }
+            }
         } else if (auto load = dynamic_cast<LoadInst*>(inst)) {
             auto load_node = create_node(DAGNode::LOAD, load, value_to_node, nodes_storage);
             load_node->operands.push_back(get_operand_node(load->getPointer(), value_to_node, nodes_storage));
@@ -632,4 +892,98 @@ std::vector<std::unique_ptr<RISCv64ISel::DAGNode>> RISCv64ISel::build_dag(BasicB
     return nodes_storage;
 }
 
+// [新] 打印DAG图以供调试的辅助函数
+void RISCv64ISel::print_dag(const std::vector<std::unique_ptr<DAGNode>>& dag, const std::string& bb_name) {
+    // 检查是否有DEBUG宏或者全局变量，避免在非调试模式下打印
+    // if (!DEBUG) return; 
+
+    std::cerr << "=== DAG for Basic Block: " << bb_name << " ===\n";
+    std::set<DAGNode*> visited;
+
+    // 为节点分配临时ID，方便阅读
+    std::map<DAGNode*, int> node_to_id;
+    int current_id = 0;
+    for (const auto& node_ptr : dag) {
+        node_to_id[node_ptr.get()] = current_id++;
+    }
+
+    // 将NodeKind枚举转换为字符串的辅助函数
+    auto get_kind_string = [](DAGNode::NodeKind kind) {
+        switch (kind) {
+            case DAGNode::CONSTANT: return "CONSTANT";
+            case DAGNode::LOAD: return "LOAD";
+            case DAGNode::STORE: return "STORE";
+            case DAGNode::BINARY: return "BINARY";
+            case DAGNode::CALL: return "CALL";
+            case DAGNode::RETURN: return "RETURN";
+            case DAGNode::BRANCH: return "BRANCH";
+            case DAGNode::ALLOCA_ADDR: return "ALLOCA_ADDR";
+            case DAGNode::UNARY: return "UNARY";
+            case DAGNode::MEMSET: return "MEMSET";
+            default: return "UNKNOWN";
+        }
+    };
+
+    // 递归打印节点的lambda表达式
+    std::function<void(DAGNode*, int)> print_node = 
+        [&](DAGNode* node, int indent) {
+        if (!node) return;
+
+        std::string current_indent(indent, ' ');
+        int node_id = node_to_id.count(node) ? node_to_id[node] : -1;
+
+        std::cerr << current_indent << "Node#" << node_id << ": " << get_kind_string(node->kind);
+        
+        // 尝试打印关联的虚拟寄存器
+        if (node->value && vreg_map.count(node->value)) {
+            std::cerr << " (vreg: %vreg" << vreg_map.at(node->value) << ")";
+        }
+
+        // 打印关联的IR Value信息
+        if (node->value) {
+            std::cerr << " [";
+            if (auto inst = dynamic_cast<Instruction*>(node->value)) {
+                std::cerr << inst->getKindString();
+                if (!inst->getName().empty()) {
+                    std::cerr << "(" << inst->getName() << ")";
+                }
+            } else if (auto constant = dynamic_cast<ConstantValue*>(node->value)) {
+                std::cerr << "Const(" << constant->getInt() << ")";
+            } else if (auto global = dynamic_cast<GlobalValue*>(node->value)) {
+                std::cerr << "Global(" << global->getName() << ")";
+            } else if (auto alloca = dynamic_cast<AllocaInst*>(node->value)) {
+                std::cerr << "Alloca(" << alloca->getName() << ")";
+            }
+            std::cerr << "]";
+        }
+        std::cerr << "\n";
+
+        if (visited.count(node)) {
+            std::cerr << current_indent << "  (已打印过子节点)\n";
+            return;
+        }
+        visited.insert(node);
+
+        if (!node->operands.empty()) {
+            std::cerr << current_indent << "  Operands:\n";
+            for (auto operand : node->operands) {
+                print_node(operand, indent + 4);
+            }
+        }
+    };
+
+    // 从根节点（没有用户的节点，或有副作用的节点）开始打印
+    for (const auto& node_ptr : dag) {
+        if (node_ptr->users.empty() || 
+            node_ptr->kind == DAGNode::STORE || 
+            node_ptr->kind == DAGNode::RETURN || 
+            node_ptr->kind == DAGNode::BRANCH ||
+            node_ptr->kind == DAGNode::MEMSET) 
+        {
+            print_node(node_ptr.get(), 0);
+        }
+    }
+    std::cerr << "======================================\n\n";
+}
+
 } // namespace sysy

src/RISCv64RegAlloc.cpp

@@ -27,7 +27,8 @@ void RISCv64RegAlloc::run() {
 
 void RISCv64RegAlloc::eliminateFrameIndices() {
     StackFrameInfo& frame_info = MFunc->getFrameInfo();
-    int current_offset = 0;
+    int current_offset = 20;    // 这里写20是为了在$s0和第一个变量之间留出20字节的安全区，
+                                // 以防止一些函数调用方面的恶性bug。
     Function* F = MFunc->getFunc();
     RISCv64ISel* isel = MFunc->getISel();
 
@@ -94,6 +95,18 @@ void RISCv64RegAlloc::eliminateFrameIndices() {
                     std::make_unique<RegOperand>(addr_vreg),
                     std::make_unique<ImmOperand>(0)));
                 new_instructions.push_back(std::move(sw));
+            } else if (instr_ptr->getOpcode() == RVOpcodes::FRAME_ADDR) { // [新] 处理FRAME_ADDR
+                auto& operands = instr_ptr->getOperands();
+                unsigned dest_vreg = static_cast<RegOperand*>(operands[0].get())->getVRegNum();
+                unsigned alloca_vreg = static_cast<RegOperand*>(operands[1].get())->getVRegNum();
+                int offset = frame_info.alloca_offsets.at(alloca_vreg);
+
+                // 将 `frame_addr rd, rs` 展开为 `addi rd, s0, offset`
+                auto addi = std::make_unique<MachineInstr>(RVOpcodes::ADDI);
+                addi->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                addi->addOperand(std::make_unique<RegOperand>(PhysicalReg::S0)); // 基地址是帧指针 s0
+                addi->addOperand(std::make_unique<ImmOperand>(offset));
+                new_instructions.push_back(std::move(addi));
             } else {
                 new_instructions.push_back(std::move(instr_ptr));
             }

src/Reg2Mem.cpp

@@ -1,122 +0,0 @@
-#include "Reg2Mem.h"
-#include <cstddef>
-#include <iostream>
-#include <list>
-#include <memory>
-
-namespace sysy {
-
-/**
- * 删除phi节点
- * 删除phi节点后可能会生成冗余存储代码
- */
-void Reg2Mem::DeletePhiInst(){
-  auto &functions = pModule->getFunctions();
-  for (auto &function : functions) {
-    auto basicBlocks = function.second->getBasicBlocks();
-    for (auto &basicBlock : basicBlocks) {
-
-      for (auto iter = basicBlock->begin(); iter != basicBlock->end();) {
-        auto &instruction = *iter;
-        if (instruction->isPhi()) {
-          auto predBlocks = basicBlock->getPredecessors();
-          // 寻找源和目的
-          // 目的就是phi指令的第一个操作数
-          // 源就是phi指令的后续操作数
-          auto destination = instruction->getOperand(0);
-          int predBlockindex = 0;
-          for (auto &predBlock : predBlocks) {
-            ++predBlockindex;
-            // 判断前驱块儿只有一个后继还是多个后继
-            // 如果有多个
-            auto source = instruction->getOperand(predBlockindex);
-            if (source == destination) {
-              continue;
-            }
-            // std::cout << predBlock->getNumSuccessors() << std::endl;
-            if (predBlock->getNumSuccessors() > 1) {
-              // 创建一个basicblock
-              auto newbasicBlock = function.second->addBasicBlock();
-              std::stringstream ss;
-              ss << "phidel.L" << pBuilder->getLabelIndex();
-              newbasicBlock->setName(ss.str());
-              ss.str("");
-              // // 修改前驱后继关系
-              basicBlock->replacePredecessor(predBlock, newbasicBlock);
-              // predBlock = newbasicBlock;
-              newbasicBlock->addPredecessor(predBlock);
-              newbasicBlock->addSuccessor(basicBlock.get());
-              predBlock->removeSuccessor(basicBlock.get());
-              predBlock->addSuccessor(newbasicBlock);
-              // std::cout << "the block name is " << basicBlock->getName() << std::endl;
-              // for (auto pb : basicBlock->getPredecessors()) {
-              //   // newbasicBlock->addPredecessor(pb);
-              //   std::cout << pb->getName() << std::endl;
-              // }
-              // sysy::BasicBlock::conectBlocks(newbasicBlock, static_cast<BasicBlock *>(basicBlock.get()));
-              // 若后为跳转指令，应该修改跳转指令所到达的位置
-              auto thelastinst = predBlock->end();
-              (--thelastinst);
-
-              if (thelastinst->get()->isConditional() || thelastinst->get()->isUnconditional()) {  // 如果是跳转指令
-                auto opnum = thelastinst->get()->getNumOperands();
-                for (size_t i = 0; i < opnum; i++) {
-                  if (thelastinst->get()->getOperand(i) == basicBlock.get()) {
-                    thelastinst->get()->replaceOperand(i, newbasicBlock);
-                  }
-                }
-              }
-              // 在新块中插入store指令
-              pBuilder->setPosition(newbasicBlock, newbasicBlock->end());
-              // pBuilder->createStoreInst(source, destination);
-              if (source->isInt() || source->isFloat()) {
-                pBuilder->createStoreInst(source, destination);
-              } else {
-                auto loadInst = pBuilder->createLoadInst(source);
-                pBuilder->createStoreInst(loadInst, destination);
-              }
-              // pBuilder->createMoveInst(Instruction::kMove, destination->getType(), destination, source,
-              // newbasicBlock);
-              pBuilder->setPosition(newbasicBlock, newbasicBlock->end());
-              pBuilder->createUncondBrInst(basicBlock.get(), {});
-            } else {
-              // 如果前驱块只有一个后继
-              auto thelastinst = predBlock->end();
-              (--thelastinst);
-              // std::cout << predBlock->getName() << std::endl;
-              // std::cout << thelastinst->get() << std::endl;
-              // std::cout << "First point  11 " << std::endl;
-              if (thelastinst->get()->isConditional() || thelastinst->get()->isUnconditional()) {
-                // 在跳转语句前insert st指令
-                pBuilder->setPosition(predBlock, thelastinst);
-              } else {
-                pBuilder->setPosition(predBlock, predBlock->end());
-              }
-
-              if (source->isInt() || source->isFloat()) {
-                pBuilder->createStoreInst(source, destination);
-              } else {
-                auto loadInst = pBuilder->createLoadInst(source);
-                pBuilder->createStoreInst(loadInst, destination);
-              }
-            }
-          }
-          // 删除phi指令
-          auto &instructions = basicBlock->getInstructions();
-          SysYIROptUtils::usedelete(iter->get());
-          iter = instructions.erase(iter);
-          if (basicBlock->getNumInstructions() == 0) {
-            if (basicBlock->getNumSuccessors() == 1) {
-              pBuilder->setPosition(basicBlock.get(), basicBlock->end());
-              pBuilder->createUncondBrInst(basicBlock->getSuccessors()[0], {});
-            }
-          }
-        } else {
-          break;
-        }
-      }
-    }
-  }
-}
-
-}  // namespace sysy

src/SysYIRAnalyser.cpp

@@ -523,6 +523,9 @@ bool ActiveVarAnalysis::analyze(Module *pModule, BasicBlock *block) {
 }
 
 
+auto ActiveVarAnalysis::getActiveTable() const -> const std::map<BasicBlock *, std::vector<std::set<User *>>> & {
+  return activeTable;
+}
 
 
 } // namespace sysy

src/SysYIRCFGOpt.cpp

@@ -1,565 +0,0 @@
-#include "SysYIRCFGOpt.h"
-#include "SysYIROptUtils.h"
-#include <cassert>
-#include <list>
-#include <map>
-#include <memory>
-#include <string>
-#include <iostream>
-#include "IR.h"
-#include "IRBuilder.h"
-
-namespace sysy {
-
-
-// 删除br后的无用指令
-bool SysYCFGOpt::SysYDelInstAfterBr(Function *func) {
-  bool changed = false;
-  
-  auto basicBlocks = func->getBasicBlocks();
-  for (auto &basicBlock : basicBlocks) {
-    bool Branch = false;
-    auto &instructions = basicBlock->getInstructions();
-    auto Branchiter = instructions.end();
-    for (auto iter = instructions.begin(); iter != instructions.end(); ++iter) {
-      if (Branch) 
-        SysYIROptUtils::usedelete(iter->get());
-      else if ((*iter)->isTerminator()){
-        Branch = true;
-        Branchiter = iter;
-      }
-    }
-    if (Branchiter != instructions.end()) ++Branchiter;
-    while (Branchiter != instructions.end()) {
-      changed = true;
-      Branchiter = instructions.erase(Branchiter);
-    }
-    
-    if (Branch) {  // 更新前驱后继关系
-      auto thelastinstinst = basicBlock->getInstructions().end();
-      --thelastinstinst;
-      auto &Successors = basicBlock->getSuccessors();
-      for (auto iterSucc = Successors.begin(); iterSucc != Successors.end();) {
-        (*iterSucc)->removePredecessor(basicBlock.get());
-        basicBlock->removeSuccessor(*iterSucc);
-      }
-      if (thelastinstinst->get()->isUnconditional()) {
-        BasicBlock* branchBlock = dynamic_cast<BasicBlock *>(thelastinstinst->get()->getOperand(0));
-        basicBlock->addSuccessor(branchBlock);
-        branchBlock->addPredecessor(basicBlock.get());
-      } else if (thelastinstinst->get()->isConditional()) {
-        BasicBlock* thenBlock = dynamic_cast<BasicBlock *>(thelastinstinst->get()->getOperand(1));
-        BasicBlock* elseBlock = dynamic_cast<BasicBlock *>(thelastinstinst->get()->getOperand(2));
-        basicBlock->addSuccessor(thenBlock);
-        basicBlock->addSuccessor(elseBlock);
-        thenBlock->addPredecessor(basicBlock.get());
-        elseBlock->addPredecessor(basicBlock.get());
-      }
-    }
-  }
-
-  return changed;
-}
-
-// 合并空基本块
-bool SysYCFGOpt::SysYBlockMerge(Function *func) {
-  bool changed = false;
-
-  for (auto blockiter = func->getBasicBlocks().begin();
-        blockiter != func->getBasicBlocks().end();) {
-    if (blockiter->get()->getNumSuccessors() == 1) {
-      // 如果当前块只有一个后继块
-      // 且后继块只有一个前驱块
-      // 则将当前块和后继块合并
-      if (((blockiter->get())->getSuccessors()[0])->getNumPredecessors() == 1) {
-        // std::cout << "merge block: " << blockiter->get()->getName() << std::endl;
-        BasicBlock* block = blockiter->get();
-        BasicBlock* nextBlock = blockiter->get()->getSuccessors()[0];
-        auto nextarguments = nextBlock->getArguments();
-        // 删除br指令
-        if (block->getNumInstructions() != 0) {
-          auto thelastinstinst = block->end();
-          (--thelastinstinst);
-          if (thelastinstinst->get()->isUnconditional()) {
-            SysYIROptUtils::usedelete(thelastinstinst->get());
-            block->getInstructions().erase(thelastinstinst);
-          } else if (thelastinstinst->get()->isConditional()) {
-            // 如果是条件分支，判断条件是否相同，主要优化相同布尔表达式
-            if (thelastinstinst->get()->getOperand(1)->getName() == thelastinstinst->get()->getOperand(1)->getName()) {
-              SysYIROptUtils::usedelete(thelastinstinst->get());
-              block->getInstructions().erase(thelastinstinst);
-            }
-          }
-        }
-        // 将后继块的指令移动到当前块
-        // 并将后继块的父指针改为当前块
-        for (auto institer = nextBlock->begin(); institer != nextBlock->end();) {
-          institer->get()->setParent(block);
-          block->getInstructions().emplace_back(institer->release());
-          institer = nextBlock->getInstructions().erase(institer);   
-        }
-        // 合并参数
-        // TODO：是否需要去重?
-        for (auto &argm : nextarguments) {
-          argm->setParent(block);
-          block->insertArgument(argm);
-        }
-        // 更新前驱后继关系，类似树节点操作
-        block->removeSuccessor(nextBlock);
-        nextBlock->removePredecessor(block);
-        std::list<BasicBlock *> succshoulddel;
-        for (auto &succ : nextBlock->getSuccessors()) {
-          block->addSuccessor(succ);
-          succ->replacePredecessor(nextBlock, block);
-          succshoulddel.push_back(succ);
-        }
-        for (auto del : succshoulddel) {
-          nextBlock->removeSuccessor(del);
-        }
-
-        func->removeBasicBlock(nextBlock);
-        changed = true;
-
-      } else {
-        blockiter++;
-      }
-    } else {
-      blockiter++;
-    }
-  }
-
-  return changed;
-}
-
-// 删除无前驱块，兼容SSA后的处理
-bool SysYCFGOpt::SysYDelNoPreBLock(Function *func) {
-  
-  bool changed = false;
-
-  for (auto &block : func->getBasicBlocks()) {
-    block->setreachableFalse();
-  }
-  // 对函数基本块做一个拓扑排序，排查不可达基本块
-  auto entryBlock = func->getEntryBlock();
-  entryBlock->setreachableTrue();
-  std::queue<BasicBlock *> blockqueue;
-  blockqueue.push(entryBlock);
-  while (!blockqueue.empty()) {
-    auto block = blockqueue.front();
-    blockqueue.pop();
-    for (auto &succ : block->getSuccessors()) {
-      if (!succ->getreachable()) {
-        succ->setreachableTrue();
-        blockqueue.push(succ);
-      }
-    }
-  }
-
-  // 删除不可达基本块指令
-  for (auto blockIter = func->getBasicBlocks().begin(); blockIter != func->getBasicBlocks().end();blockIter++) {
-    if (!blockIter->get()->getreachable()) 
-      for (auto &iterInst : blockIter->get()->getInstructions()) 
-        SysYIROptUtils::usedelete(iterInst.get());
-  
-  }
-
-  
-  for (auto blockIter = func->getBasicBlocks().begin(); blockIter != func->getBasicBlocks().end();) {
-    if (!blockIter->get()->getreachable()) {
-      for (auto succblock : blockIter->get()->getSuccessors()) {
-        int indexphi = 1;
-        for (auto pred : succblock->getPredecessors()) {
-          if (pred == blockIter->get()) {
-            break;
-          }
-          indexphi++;
-        }
-        for (auto &phiinst : succblock->getInstructions()) {
-          if (phiinst->getKind() != Instruction::kPhi) {
-            break;
-          }
-          phiinst->removeOperand(indexphi);
-        }
-      }
-      // 删除不可达基本块，注意迭代器不可达问题
-      func->removeBasicBlock((blockIter++)->get());
-      changed = true;
-    } else {
-      blockIter++;
-    }
-  }
-  
-  return changed;
-}
-
-// 删除空块
-bool SysYCFGOpt::SysYDelEmptyBlock(Function *func, IRBuilder* pBuilder) {
-  bool changed = false;
-
-  // 收集不可达基本块
-  // 这里的不可达基本块是指没有实际指令的基本块
-  // 当一个基本块没有实际指令例如只有phi指令和一个uncondbr指令时，也会被视作不可达
-  auto basicBlocks = func->getBasicBlocks();
-  std::map<sysy::BasicBlock *, BasicBlock *> EmptyBlocks;
-  // 空块儿和后继的基本块的映射
-  for (auto &basicBlock : basicBlocks) {
-    if (basicBlock->getNumInstructions() == 0) {
-      if (basicBlock->getNumSuccessors() == 1) {
-        EmptyBlocks[basicBlock.get()] = basicBlock->getSuccessors().front();
-      }
-    }
-    else{
-      // 如果只有phi指令和一个uncondbr。(phi)*(uncondbr)?
-        // 判断除了最后一个指令之外是不是只有phi指令
-      bool onlyPhi = true;
-      for (auto &inst : basicBlock->getInstructions()) {
-        if (!inst->isPhi() && !inst->isUnconditional()) {
-          onlyPhi = false;
-          break;
-        }
-      }
-      if(onlyPhi)
-        EmptyBlocks[basicBlock.get()] = basicBlock->getSuccessors().front();
-    }
-    
-    
-  }
-  // 更新基本块信息，增加必要指令
-  for (auto &basicBlock : basicBlocks) {
-    // 把空块转换成只有跳转指令的不可达块
-    if (distance(basicBlock->begin(), basicBlock->end()) == 0) {
-      if (basicBlock->getNumSuccessors() == 0) {
-        continue;
-      }
-      if (basicBlock->getNumSuccessors() > 1) {
-        assert("");
-      }
-      pBuilder->setPosition(basicBlock.get(), basicBlock->end());
-      pBuilder->createUncondBrInst(basicBlock->getSuccessors()[0], {});
-      continue;
-    }
-
-    auto thelastinst = basicBlock->getInstructions().end();
-    --thelastinst;
-
-    // 根据br指令传递的后继块信息，跳过空块链
-    if (thelastinst->get()->isUnconditional()) {
-      BasicBlock* OldBrBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0));
-      BasicBlock *thelastBlockOld = nullptr;
-      // 如果空块链表为多个块
-      while (EmptyBlocks.find(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))) !=
-              EmptyBlocks.end()) {
-        thelastBlockOld = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0));
-        thelastinst->get()->replaceOperand(0, EmptyBlocks[thelastBlockOld]);
-      }
-
-      basicBlock->removeSuccessor(OldBrBlock);
-      OldBrBlock->removePredecessor(basicBlock.get());
-      basicBlock->addSuccessor(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0)));
-      dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->addPredecessor(basicBlock.get());
-
-      if (thelastBlockOld != nullptr) {
-        int indexphi = 0;
-        for (auto &pred : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->getPredecessors()) {
-          if (pred == thelastBlockOld) {
-            break;
-          }
-          indexphi++;
-        }
-
-        // 更新phi指令的操作数
-        // 移除thelastBlockOld对应的phi操作数
-        for (auto &InstInNew : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->getInstructions()) {
-          if (InstInNew->isPhi()) {
-            dynamic_cast<PhiInst *>(InstInNew.get())->removeOperand(indexphi + 1);
-          } else {
-            break;
-          }
-        }
-      }
-
-    } else if (thelastinst->get()->getKind() == Instruction::kCondBr) {
-      auto OldThenBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1));
-      auto OldElseBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2));
-
-      BasicBlock *thelastBlockOld = nullptr;
-      while (EmptyBlocks.find(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1))) !=
-              EmptyBlocks.end()) {
-        thelastBlockOld = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1));
-        thelastinst->get()->replaceOperand(
-            1, EmptyBlocks[dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1))]);
-      }
-      basicBlock->removeSuccessor(OldThenBlock);
-      OldThenBlock->removePredecessor(basicBlock.get());
-      // 处理 then 和 else 分支合并的情况
-      if (dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1)) ==
-          dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))) {
-        auto thebrBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1));
-        SysYIROptUtils::usedelete(thelastinst->get());
-        thelastinst = basicBlock->getInstructions().erase(thelastinst);
-        pBuilder->setPosition(basicBlock.get(), basicBlock->end());
-        pBuilder->createUncondBrInst(thebrBlock, {});
-        continue;
-      }
-      basicBlock->addSuccessor(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1)));
-      dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1))->addPredecessor(basicBlock.get());
-      // auto indexInNew = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->getPredecessors().
-
-      if (thelastBlockOld != nullptr) {
-        int indexphi = 0;
-        for (auto &pred : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1))->getPredecessors()) {
-          if (pred == thelastBlockOld) {
-            break;
-          }
-          indexphi++;
-        }
-
-        for (auto &InstInNew : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1))->getInstructions()) {
-          if (InstInNew->isPhi()) {
-            dynamic_cast<PhiInst *>(InstInNew.get())->removeOperand(indexphi + 1);
-          } else {
-            break;
-          }
-        }
-      }
-
-      thelastBlockOld = nullptr;
-      while (EmptyBlocks.find(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))) !=
-              EmptyBlocks.end()) {
-        thelastBlockOld = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2));
-        thelastinst->get()->replaceOperand(
-            2, EmptyBlocks[dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))]);
-      }
-      basicBlock->removeSuccessor(OldElseBlock);
-      OldElseBlock->removePredecessor(basicBlock.get());
-      // 处理 then 和 else 分支合并的情况
-      if (dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1)) ==
-          dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))) {
-        auto thebrBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1));
-        SysYIROptUtils::usedelete(thelastinst->get());
-        thelastinst = basicBlock->getInstructions().erase(thelastinst);
-        pBuilder->setPosition(basicBlock.get(), basicBlock->end());
-        pBuilder->createUncondBrInst(thebrBlock, {});
-        continue;
-      }
-      basicBlock->addSuccessor(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2)));
-      dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))->addPredecessor(basicBlock.get());
-
-      if (thelastBlockOld != nullptr) {
-        int indexphi = 0;
-        for (auto &pred : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))->getPredecessors()) {
-          if (pred == thelastBlockOld) {
-            break;
-          }
-          indexphi++;
-        }
-        for (auto &InstInNew : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))->getInstructions()) {
-          if (InstInNew->isPhi()) {
-            dynamic_cast<PhiInst *>(InstInNew.get())->removeOperand(indexphi + 1);
-          } else {
-            break;
-          }
-        }
-      }
-    } else {
-      if (basicBlock->getNumSuccessors() == 1) {
-        pBuilder->setPosition(basicBlock.get(), basicBlock->end());
-        pBuilder->createUncondBrInst(basicBlock->getSuccessors()[0], {});
-        auto thelastinst = basicBlock->getInstructions().end();
-        (--thelastinst);
-        auto OldBrBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0));
-        sysy::BasicBlock *thelastBlockOld = nullptr;
-        while (EmptyBlocks.find(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))) !=
-                EmptyBlocks.end()) {
-          thelastBlockOld = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0));
-
-          thelastinst->get()->replaceOperand(
-              0, EmptyBlocks[dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))]);
-        }
-        
-        basicBlock->removeSuccessor(OldBrBlock);
-        OldBrBlock->removePredecessor(basicBlock.get());
-        basicBlock->addSuccessor(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0)));
-        dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->addPredecessor(basicBlock.get());
-        if (thelastBlockOld != nullptr) {
-          int indexphi = 0;
-          for (auto &pred : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->getPredecessors()) {
-            if (pred == thelastBlockOld) {
-              break;
-            }
-            indexphi++;
-          }
-
-          for (auto &InstInNew : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->getInstructions()) {
-            if (InstInNew->isPhi()) {
-              dynamic_cast<PhiInst *>(InstInNew.get())->removeOperand(indexphi + 1);
-            } else {
-              break;
-            }
-          }
-        }
-      }
-    }
-  }
-
-  for (auto iter = func->getBasicBlocks().begin(); iter != func->getBasicBlocks().end();) {
-    
-    if (EmptyBlocks.find(iter->get()) != EmptyBlocks.end()) {
-      // EntryBlock跳过
-      if (iter->get() == func->getEntryBlock()) {
-        ++iter;
-        continue;
-      }
-
-      for (auto &iterInst : iter->get()->getInstructions()) 
-        SysYIROptUtils::usedelete(iterInst.get());
-      // 删除不可达基本块的phi指令的操作数
-      for (auto &succ : iter->get()->getSuccessors()) {
-        int index = 0;
-        for (auto &pred : succ->getPredecessors()) {
-          if (pred == iter->get()) {
-            break;
-          }
-          index++;
-        }
-
-        for (auto &instinsucc : succ->getInstructions()) {
-          if (instinsucc->isPhi()) {
-            dynamic_cast<PhiInst *>(instinsucc.get())->removeOperand(index);
-          } else {
-            break;
-          }
-        }
-      }
-
-      func->removeBasicBlock((iter++)->get());
-      changed = true;
-    } else {
-      ++iter;
-    }
-  }
-  
-  return changed;
-
-}
-
-// 如果函数没有返回指令，则添加一个默认返回指令(主要解决void函数没有返回指令的问题)
-bool SysYCFGOpt::SysYAddReturn(Function *func, IRBuilder* pBuilder) {
-  bool changed = false;
-  auto basicBlocks = func->getBasicBlocks();
-  for (auto &block : basicBlocks) {
-    if (block->getNumSuccessors() == 0) {
-      changed = true;
-      // 如果基本块没有后继块，则添加一个返回指令
-      if (block->getNumInstructions() == 0) {
-        pBuilder->setPosition(block.get(), block->end());
-        pBuilder->createReturnInst();
-      }
-      auto thelastinst = block->getInstructions().end();
-      --thelastinst;
-      if (thelastinst->get()->getKind() != Instruction::kReturn) {
-        // std::cout << "Warning: Function " << func->getName() << " has no return instruction, adding default return." << std::endl;
-
-        pBuilder->setPosition(block.get(), block->end());
-        // TODO: 如果int float函数缺少返回值是否需要报错
-        if (func->getReturnType()->isInt()) {
-          pBuilder->createReturnInst(ConstantInteger::get(0));
-        } else if (func->getReturnType()->isFloat()) {
-          pBuilder->createReturnInst(ConstantFloating::get(0.0F));
-        } else {
-          pBuilder->createReturnInst();
-        }
-      }
-    }
-  }
-  
-  return changed;
-}
-
-// 条件分支转换为无条件分支
-// 主要针对已知条件值的分支转换为无条件分支
-// 例如 if (cond) { ... } else { ... } 中的 cond 已经
-// 确定为 true 或 false 的情况
-bool SysYCFGOpt::SysYCondBr2Br(Function *func, IRBuilder* pBuilder) {
-  bool changed = false;
-
-  for (auto &basicblock : func->getBasicBlocks()) {
-    if (basicblock->getNumInstructions() == 0)
-      continue;
-    
-    auto thelast = basicblock->getInstructions().end();
-    --thelast;
-
-    if (thelast->get()->isConditional()){
-      ConstantValue *constOperand = dynamic_cast<ConstantValue *>(thelast->get()->getOperand(0));
-      std::string opname;
-      int constint = 0;
-      float constfloat = 0.0F;
-      bool constint_Use = false;
-      bool constfloat_Use = false;
-      if (constOperand != nullptr) {
-        if (constOperand->isFloat()) {
-          constfloat = constOperand->getFloat();
-          constfloat_Use = true;
-        } else {
-          constint = constOperand->getInt();
-          constint_Use = true;
-        }
-      }
-      // 如果可以计算
-      if (constfloat_Use || constint_Use) {
-        changed = true;
-
-        auto thenBlock = dynamic_cast<BasicBlock *>(thelast->get()->getOperand(1));
-        auto elseBlock = dynamic_cast<BasicBlock *>(thelast->get()->getOperand(2));
-        SysYIROptUtils::usedelete(thelast->get());
-        thelast = basicblock->getInstructions().erase(thelast);
-        if ((constfloat_Use && constfloat == 1.0F) || (constint_Use && constint == 1)) {
-
-          pBuilder->setPosition(basicblock.get(), basicblock->end());
-          pBuilder->createUncondBrInst(thenBlock, {});
-          int phiindex = 0;
-          for (auto pred : elseBlock->getPredecessors()) {
-            phiindex++;
-            if (pred == basicblock.get()) {
-              break;
-            }
-          }
-
-          for (auto &phiinst : elseBlock->getInstructions()) {
-            if (phiinst->getKind() != Instruction::kPhi) {
-              break;
-            }
-            phiinst->removeOperand(phiindex);
-          }
-          basicblock->removeSuccessor(elseBlock);
-          elseBlock->removePredecessor(basicblock.get());
-        } else {
-
-          pBuilder->setPosition(basicblock.get(), basicblock->end());
-          pBuilder->createUncondBrInst(elseBlock, {});
-          int phiindex = 0;
-          for (auto pred : thenBlock->getPredecessors()) {
-            phiindex++;
-            if (pred == basicblock.get()) {
-              break;
-            }
-          }
-
-          for (auto &phiinst : thenBlock->getInstructions()) {
-            if (phiinst->getKind() != Instruction::kPhi) {
-              break;
-            }
-            phiinst->removeOperand(phiindex);
-          }
-          basicblock->removeSuccessor(thenBlock);
-          thenBlock->removePredecessor(basicblock.get());
-        }
-      }
-    }
-  }
-
-  return changed;
-}
-
-}  // namespace sysy

src/SysYIRGenerator.cpp

@@ -15,73 +15,6 @@
 using namespace std;
 namespace sysy {
 
-
-Type* SysYIRGenerator::buildArrayType(Type* baseType, const std::vector<Value*>& dims){
-  Type* currentType = baseType;
-    // 从最内层维度开始构建 ArrayType
-    // 例如对于 int arr[2][3]，先处理 [3]，再处理 [2]
-    // 注意：SysY 的 dims 是从最外层到最内层，所以我们需要反向迭代
-    // 或者调整逻辑，使得从内到外构建 ArrayType
-    // 假设 dims 列表是 [dim1, dim2, dim3...] (例如 [2, 3] for int[2][3])
-    // 我们需要从最内层维度开始向外构建 ArrayType
-    for (int i = dims.size() - 1; i >= 0; --i) {
-        // 维度大小必须是常量，否则无法构建 ArrayType
-        ConstantInteger* constDim = dynamic_cast<ConstantInteger*>(dims[i]);
-        if (constDim == nullptr) {
-            // 如果维度不是常量，可能需要特殊处理，例如将其视为指针
-            // 对于函数参数 int arr[] 这种，第一个维度可以为未知
-            // 在这里，我们假设所有声明的数组维度都是常量
-            assert(false && "Array dimension must be a constant integer!");
-            return nullptr;
-        }
-        unsigned dimSize = constDim->getInt();
-        currentType = Type::getArrayType(currentType, dimSize);
-    }
-    return currentType;
-}
-
-Value* SysYIRGenerator::getGEPAddressInst(Value* basePointer, const std::vector<Value*>& indices) {
-    // 检查 basePointer 是否为指针类型
-    if (!basePointer->getType()->isPointer()) {
-      assert(false && "GEP base pointer must be a pointer type!");
-    }
-
-    // 获取基指针所指向的实际类型 (例如 int* 指向 int, int[2][3]* 指向 int[2][3])
-    Type* currentElementType = basePointer->getType()->as<PointerType>()->getBaseType();
-
-    std::vector<Value*> actualGEPIndices;
-    // GEP 指令的第一个索引通常是0，用于“跳过”基指针指向的聚合类型本身，直接指向其第一个元素。
-    // 例如，对于 AllocaInst 返回的 `int[2][3]*`，第一个 `0` 索引表示从数组的开始而不是指针本身开始索引。
-    actualGEPIndices.push_back(ConstantInteger::get(0));
-
-    // 将用户提供的索引添加到 GEP 操作数中
-    for (Value* index : indices) {
-        actualGEPIndices.push_back(index);
-    }
-
-    // 根据索引链计算最终的元素类型
-    Type* finalTargetType = currentElementType;
-
-    // 遍历用户提供的索引（不包括我们添加的第一个0），逐步确定 GEP 的最终结果类型
-    // 每个索引都“深入”一个维度
-    for (size_t i = 0; i < indices.size(); ++i) { // 这里遍历的是用户提供的索引
-        if (finalTargetType && finalTargetType->isArray()) {
-            finalTargetType = finalTargetType->as<ArrayType>()->getElementType();
-        } else {
-            // 如果索引链还在继续，但当前类型已经不是数组或聚合类型，这通常是一个错误
-            // 或者表示访问的是标量，后续索引无效。此时，finalTargetType 已经是最终的标量类型，不能再深入。
-            // 例如，对 int arr[5]; 访问 arr[i][j] (j 是多余的)，这里会停止类型推断。
-            break;
-        }
-    }
-
-    // GEP 的结果总是指针类型，指向最终计算出的元素
-    Type* gepResultType = Type::getPointerType(finalTargetType);
-
-    // 创建 GEP 指令。假设 builder.createGetElementPtrInst 的签名为
-    // (Type* resultType, Value* basePointer, const std::vector<Value*>& indices)
-    return builder.createGetElementPtrInst(basePointer, actualGEPIndices);
-}
 /*
  * @brief: visit compUnit
  * @details:
@@ -185,28 +118,24 @@ std::any SysYIRGenerator::visitVarDecl(SysYParser::VarDeclContext *ctx) {
       }
     }
 
-    Type* variableType = type;
-    if (!dims.empty()) { // 如果有维度，说明是数组
-        variableType = buildArrayType(type, dims); // 构建完整的 ArrayType
-    }
-
-    // 对于数组，alloca 的类型将是指针指向数组类型，例如 `int[2][3]*`
-    // 对于标量，alloca 的类型将是指针指向标量类型，例如 `int*`
     AllocaInst* alloca =
         builder.createAllocaInst(Type::getPointerType(type), dims, name);
 
     if (varDef->initVal() != nullptr) {
       ValueCounter values;
+      // 这里的varDef->initVal()可能是ScalarInitValue或ArrayInitValue
       ArrayValueTree* root = std::any_cast<ArrayValueTree *>(varDef->initVal()->accept(this));
       Utils::tree2Array(type, root, dims, dims.size(), values, &builder);
       delete root;
-
-      if (dims.empty()) { // 标量变量初始化
+      if (dims.empty()) {
         builder.createStoreInst(values.getValue(0), alloca);
-      } else { // 数组变量初始化
+      } else{ 
+        // **数组变量初始化**
         const std::vector<sysy::Value *> &counterValues = values.getValues();
 
+        // 计算数组的**总元素数量**和**总字节大小**
         int numElements = 1;
+        // 存储每个维度的实际整数大小，用于索引计算
         std::vector<int> dimSizes;
         for (Value *dimVal : dims) {
           if (ConstantInteger *constInt = dynamic_cast<ConstantInteger *>(dimVal)) {
@@ -216,11 +145,12 @@ std::any SysYIRGenerator::visitVarDecl(SysYParser::VarDeclContext *ctx) {
           }
           // TODO else 错误处理：数组维度必须是常量（对于静态分配）
         }
-        unsigned int elementSizeInBytes = type->getSize();
+        unsigned int elementSizeInBytes = type->getSize(); // 获取单个元素的大小（字节）
         unsigned int totalSizeInBytes = numElements * elementSizeInBytes;
 
+        // **判断是否可以进行全零初始化优化**
         bool allValuesAreZero = false;
-        if (counterValues.empty()) {
+        if (counterValues.empty()) { // 例如 int arr[3] = {}; 或 int arr[3][4] = {};
           allValuesAreZero = true;
         }
         else {
@@ -233,6 +163,7 @@ std::any SysYIRGenerator::visitVarDecl(SysYParser::VarDeclContext *ctx) {
               }
             }
             else{
+              // 如果值不是常量，我们通常不能确定它是否为零，所以不进行 memset 优化
               allValuesAreZero = false;
               break;
             }
@@ -240,51 +171,64 @@ std::any SysYIRGenerator::visitVarDecl(SysYParser::VarDeclContext *ctx) {
         }
 
         if (allValuesAreZero) {
+          // 如果所有初始化值都是零（或没有明确初始化但语法允许），使用 memset 优化
           builder.createMemsetInst(
-              alloca,
-              ConstantInteger::get(0),
+              alloca,                   // 目标数组的起始地址
+              ConstantInteger::get(0),  // 偏移量（通常为0），后续删除
               ConstantInteger::get(totalSizeInBytes),
-              ConstantInteger::get(0));
+              ConstantInteger::get(0)); // 填充的总字节数
         }
         else {
+          // **逐元素存储：遍历所有初始值，并为每个值生成一个 store 指令**
           for (size_t k = 0; k < counterValues.size(); ++k) {
+            // 用于存储当前元素的索引列表
             std::vector<Value *> currentIndices;
-            int tempLinearIndex = k;
+            int tempLinearIndex = k; // 临时线性索引，用于计算多维索引
 
-            // 将线性索引转换为多维索引
+            // **将线性索引转换为多维索引**
+            // 这个循环从最内层维度开始倒推，计算每个维度的索引
+            // 假设是行主序（row-major order），这是 C/C++ 数组的标准存储方式
             for (int dimIdx = dimSizes.size() - 1; dimIdx >= 0; --dimIdx)
             {
+              // 计算当前维度的索引，并插入到列表的最前面
               currentIndices.insert(currentIndices.begin(),
                                     ConstantInteger::get(static_cast<int>(tempLinearIndex % dimSizes[dimIdx])));
+              // 更新线性索引，用于计算下一个更高维度的索引
               tempLinearIndex /= dimSizes[dimIdx];
             }
 
-            // 计算元素的地址
-            Value* elementAddress = getGEPAddressInst(alloca, currentIndices);
-            // 生成 store 指令 (假设 createStoreInst 接受 Value* value, Value* pointer)
-            builder.createStoreInst(counterValues[k], elementAddress);
+            // **生成 store 指令，传入值、基指针和计算出的索引列表**
+            // 你的 builder.createStoreInst 签名需要能够接受这些参数
+            // 假设你的 builder.createStoreInst(Value *val, Value *ptr, const std::vector<Value *> &indices, ...)
+            builder.createStoreInst(counterValues[k], alloca, currentIndices);
           }
         }
       }
     }
-    else { // 如果没有显式初始化值，默认对数组进行零初始化
-      if (!dims.empty()) { // 只有数组才需要默认的零初始化
+    else
+    { // **如果没有显式初始化值，默认对数组进行零初始化**
+      if (!dims.empty())
+      { // 只有数组才需要默认的零初始化
         int numElements = 1;
-        for (Value *dimVal : dims) {
-          if (ConstantInteger *constInt = dynamic_cast<ConstantInteger *>(dimVal)) {
+        for (Value *dimVal : dims)
+        {
+          if (ConstantInteger *constInt = dynamic_cast<ConstantInteger *>(dimVal))
+          {
             numElements *= constInt->getInt();
           }
         }
         unsigned int elementSizeInBytes = type->getSize();
         unsigned int totalSizeInBytes = numElements * elementSizeInBytes;
 
+        // 使用 memset 将整个数组清零
         builder.createMemsetInst(
             alloca,
             ConstantInteger::get(0),
             ConstantInteger::get(totalSizeInBytes),
             ConstantInteger::get(0)
-          );
+          ); // 填充的总字节数
       }
+      // 标量变量如果没有初始化值，通常不生成额外的初始化指令，因为其内存已分配但未赋值。
     }
 
     module->addVariable(name, alloca);
@@ -340,6 +284,7 @@ std::any SysYIRGenerator::visitFuncType(SysYParser::FuncTypeContext *ctx) {
 std::any SysYIRGenerator::visitFuncDef(SysYParser::FuncDefContext *ctx){
   // 更新作用域
   module->enterNewScope();
+  HasReturnInst = false;
 
   auto name = ctx->Ident()->getText();
   std::vector<Type *> paramTypes;
@@ -375,34 +320,22 @@ std::any SysYIRGenerator::visitFuncDef(SysYParser::FuncDefContext *ctx){
     module->addVariable(paramNames[i], alloca);
   }
 
-  // 在处理函数体之前，创建一个新的基本块作为函数体的实际入口
-  // 这样 entryBB 就可以在完成初始化后跳转到这里
-  BasicBlock* funcBodyEntry = function->addBasicBlock("funcBodyEntry");
-  
-  // 从 entryBB 无条件跳转到 funcBodyEntry
-  builder.createUncondBrInst(funcBodyEntry, {});
-  builder.setPosition(funcBodyEntry,funcBodyEntry->end()); // 将插入点设置到 funcBodyEntry
-
   for (auto item : ctx->blockStmt()->blockItem()) {
     visitBlockItem(item);
   }
 
-  // 如果函数没有显式的返回语句，且返回类型不是 void，则需要添加一个默认的返回值
-  ReturnInst* retinst = nullptr;
-  retinst = dynamic_cast<ReturnInst*>(builder.getBasicBlock()->terminator()->get());
-
-  if (!retinst) {
-    if (returnType->isVoid()) {
-        builder.createReturnInst();
-    } else if (returnType->isInt()) {
-        builder.createReturnInst(ConstantInteger::get(0)); // 默认返回 0
-    } else if (returnType->isFloat()) {
-        builder.createReturnInst(ConstantFloating::get(0.0f)); // 默认返回 0.0f
+  if(HasReturnInst == false) {
+    // 如果没有return语句，则默认返回0
+    if (returnType != Type::getVoidType()) {
+      Value* returnValue = ConstantInteger::get(0);
+      if (returnType == Type::getFloatType()) {
+        returnValue = ConstantFloating::get(0.0f);
+      }
+      builder.createReturnInst(returnValue);
     } else {
-        assert(false && "Function with no explicit return and non-void type should return a value.");
+      builder.createReturnInst();
     }
   }
-
   module->leaveScope();
 
   return std::any();
@@ -423,56 +356,29 @@ std::any SysYIRGenerator::visitAssignStmt(SysYParser::AssignStmtContext *ctx) {
   for (const auto &exp : lVal->exp()) {
     dims.push_back(std::any_cast<Value *>(visitExp(exp)));
   }
-  
-  auto variable = module->getVariable(name); // 获取 AllocaInst 或 GlobalValue
-  Value* value = std::any_cast<Value *>(visitExp(ctx->exp())); // 右值
 
-  if (variable == nullptr) {
-      throw std::runtime_error("Variable " + name + " not found in assignment.");
-  }
+  auto variable = module->getVariable(name);
+  Value* value = std::any_cast<Value *>(visitExp(ctx->exp()));
+  Type* variableType = dynamic_cast<PointerType *>(variable->getType())->getBaseType();
 
-  // 计算最终赋值目标元素的类型
-  // variable 本身应该是一个指针类型 (例如 int* 或 int[2][3]*)
-  if (!variable->getType()->isPointer()) {
-      assert(false && "Variable to be assigned must be a pointer type!");
-      return std::any();
-  }
-  Type* targetElementType = variable->getType()->as<PointerType>()->getBaseType(); // 从基指针指向的类型开始
-
-  // 模拟 GEP 路径，根据 dims 确定最终元素的类型
-  for (size_t i = 0; i < dims.size(); ++i) {
-      if (targetElementType && targetElementType->isArray()) {
-          targetElementType = targetElementType->as<ArrayType>()->getElementType();
-      } else {
-          break; // 如果不是数组类型但还有索引，或者索引超出维度，则停止推断
-      }
-  }
-
-  // 左值右值类型不同处理：根据最终元素类型进行转换
-  if (targetElementType != value->getType()) {
+  // 左值右值类型不同处理
+  if (variableType != value->getType()) {
     ConstantValue * constValue = dynamic_cast<ConstantValue *>(value);
     if (constValue != nullptr) {
-      if (targetElementType == Type::getFloatType()) {
-        value = ConstantFloating::get(static_cast<float>(constValue->getInt()));
-      } else { // 假设如果不是浮点型，就是整型
-        value = ConstantInteger::get(static_cast<int>(constValue->getFloat()));
+      if (variableType == Type::getFloatType()) {
+        value = ConstantInteger::get(static_cast<float>(constValue->getInt()));
+      } else {
+        value = ConstantFloating::get(static_cast<int>(constValue->getFloat()));
       }
     } else {
-      if (targetElementType == Type::getFloatType()) {
+      if (variableType == Type::getFloatType()) {
         value = builder.createIToFInst(value);
-      } else { // 假设如果不是浮点型，就是整型
+      } else {
         value = builder.createFtoIInst(value);
       }
     }
   }
-
-  // 计算目标地址：如果 dims 为空，就是变量本身地址；否则通过 GEP 计算
-  Value* targetAddress = variable;
-  if (!dims.empty()) {
-    targetAddress = getGEPAddressInst(variable, dims);
-  }
-
-  builder.createStoreInst(value, targetAddress);
+  builder.createStoreInst(value, variable, dims, variable->getName());
 
   return std::any();
 }
@@ -560,7 +466,6 @@ std::any SysYIRGenerator::visitIfStmt(SysYParser::IfStmtContext *ctx) {
       ctx->stmt(0)->accept(this);
       module->leaveScope();
     }
-    builder.createUncondBrInst(exitBlock, {});
     BasicBlock::conectBlocks(builder.getBasicBlock(), exitBlock);
 
     labelstring << "if_exit.L" << builder.getLabelIndex();
@@ -582,7 +487,6 @@ std::any SysYIRGenerator::visitWhileStmt(SysYParser::WhileStmtContext *ctx) {
   labelstring << "while_head.L" << builder.getLabelIndex();
   BasicBlock *headBlock = function->addBasicBlock(labelstring.str());
   labelstring.str("");
-  builder.createUncondBrInst(headBlock, {});
   BasicBlock::conectBlocks(curBlock, headBlock);
   builder.setPosition(headBlock, headBlock->end());
 
@@ -667,94 +571,56 @@ std::any SysYIRGenerator::visitReturnStmt(SysYParser::ReturnStmtContext *ctx) {
       }
     }
   builder.createReturnInst(returnValue);
-
+  HasReturnInst = true;
   return std::any();
 }
 
 
-// SysYIRGenerator.cpp (修改部分)
-
 std::any SysYIRGenerator::visitLValue(SysYParser::LValueContext *ctx) {
   std::string name = ctx->Ident()->getText();
   User* variable = module->getVariable(name);
 
   Value* value = nullptr;
-  if (variable == nullptr) {
-    throw std::runtime_error("Variable " + name + " not found.");
-  }
   std::vector<Value *> dims;
   for (const auto &exp : ctx->exp()) {
     dims.push_back(std::any_cast<Value *>(visitExp(exp)));
   }
 
-  // 1. 获取变量的声明维度数量
-  unsigned declaredNumDims = 0;
-  if (AllocaInst* alloc = dynamic_cast<AllocaInst*>(variable)) {
-    declaredNumDims = alloc->getNumDims();
-  } else if (GlobalValue* glob = dynamic_cast<GlobalValue*>(variable)) {
-    declaredNumDims = glob->getNumDims();
-  } else if (ConstantVariable* constV = dynamic_cast<ConstantVariable*>(variable)) {
-    declaredNumDims = constV->getNumDims();
+  if (variable == nullptr) {
+    throw std::runtime_error("Variable " + name + " not found.");
+  }
+  
+  bool indicesConstant = true;
+  for (const auto &dim : dims) {
+    if (dynamic_cast<ConstantValue *>(dim) == nullptr) {
+      indicesConstant = false;
+      break;
+    }
   }
 
-  // 2. 处理常量变量 (ConstantVariable) 且所有索引都是常量的情况
   ConstantVariable* constVar = dynamic_cast<ConstantVariable *>(variable);
-  if (constVar != nullptr) {
-    bool allIndicesConstant = true;
-    for (const auto &dim : dims) {
-      if (dynamic_cast<ConstantValue *>(dim) == nullptr) {
-        allIndicesConstant = false;
-        break;
-      }
-    }
-    if (allIndicesConstant) {
-      // 如果是常量变量且所有索引都是常量，直接通过 getByIndices 获取编译时值
-      // 这个方法会根据索引深度返回最终的标量值或指向子数组的指针 (作为 ConstantValue/Variable)
-      return constVar->getByIndices(dims);
-    }
-  }
-
-  // 3. 处理可变变量 (AllocaInst/GlobalValue) 或带非常量索引的常量变量
-  // 这里区分标量访问和数组元素/子数组访问
-
-  // 检查是否是访问标量变量本身（没有索引，且声明维度为0）
-  if (dims.empty() && declaredNumDims == 0) {
-    // 对于标量变量，直接加载其值。
-    // variable 本身就是指向标量的指针 (e.g., int* %a)
-    if (dynamic_cast<AllocaInst*>(variable) || dynamic_cast<GlobalValue*>(variable)) {
-        value = builder.createLoadInst(variable);
-    } else {
-        // 如果走到这里且不是AllocaInst/GlobalValue，但dims为空且declaredNumDims为0，
-        // 且又不是ConstantVariable (前面已处理)，则可能是错误情况。
-        assert(false && "Unhandled scalar variable type in LValue access.");
-        return static_cast<Value*>(nullptr);
-    }
+  GlobalValue* globalVar = dynamic_cast<GlobalValue *>(variable);
+  AllocaInst* localVar = dynamic_cast<AllocaInst *>(variable);
+  if (constVar != nullptr && indicesConstant) {
+    // 如果是常量变量，且索引是常量，则直接获取子数组
+    value = constVar->getByIndices(dims);
+  } else if (module->isInGlobalArea() && (globalVar != nullptr)) {
+    assert(indicesConstant);
+    value = globalVar->getByIndices(dims);
   } else {
-    // 访问数组元素或子数组（有索引，或变量本身是数组/多维指针）
-    Value* targetAddress = nullptr;
-
-    // GEP 的基指针就是变量本身（它是一个指向内存的指针）
-    if (dynamic_cast<AllocaInst*>(variable) || dynamic_cast<GlobalValue*>(variable) || (constVar != nullptr)) {
-        // 允许对 ConstantVariable (如果它代表全局数组常量) 进行 GEP
-        targetAddress = getGEPAddressInst(variable, dims);
+    if ((globalVar != nullptr && globalVar->getNumDims() > dims.size()) ||
+        (localVar != nullptr && localVar->getNumDims() > dims.size()) ||
+        (constVar != nullptr && constVar->getNumDims() > dims.size())) {
+      // value = builder.createLaInst(variable, indices);
+      // 如果变量是全局变量或局部变量，且索引数量小于维度数量，则创建createGetSubArray获取子数组
+      auto getArrayInst =
+          builder.createGetSubArray(dynamic_cast<LVal *>(variable), dims);
+      value = getArrayInst->getChildArray();
     } else {
-        // 其他情况（例如尝试对非指针类型或不支持的 LValue 进行 GEP）应报错
-        assert(false && "LValue variable type not supported for GEP or dynamic load.");
-        return static_cast<Value*>(nullptr);
-    }
-
-    // 现在 targetAddress 持有元素或子数组的地址。
-    // 需要判断是加载值，还是返回子数组的地址。
-
-    // 如果提供的索引数量少于声明的维度数量，则表示访问的是子数组，返回其地址
-    if (dims.size() < declaredNumDims) {
-        value = targetAddress;
-    } else {
-        // 否则，表示访问的是最终的标量元素，加载其值
-        // 假设 createLoadInst 接受 Value* pointer
-        value = builder.createLoadInst(targetAddress);
+      value = builder.createLoadInst(variable, dims);
     }
   }
+
   return value;
 }
 

src/SysYIROptPre.cpp

@@ -0,0 +1,484 @@
+#include "SysYIROptPre.h"
+#include <cassert>
+#include <list>
+#include <map>
+#include <memory>
+#include <string>
+#include <iostream>
+#include "IR.h"
+#include "IRBuilder.h"
+
+namespace sysy {
+
+/**
+ * use删除operand,以免扰乱后续分析
+ * instr: 要删除的指令
+ */
+void SysYOptPre::usedelete(Instruction *instr) {
+  for (auto &use : instr->getOperands()) {
+    Value* val = use->getValue();
+    // std::cout << delete << val->getName() << std::endl;
+    val->removeUse(use);
+  }
+}
+
+
+// 删除br后的无用指令
+void SysYOptPre::SysYDelInstAfterBr() {
+  auto &functions = pModule->getFunctions();
+  for (auto &function : functions) {
+    auto basicBlocks = function.second->getBasicBlocks();
+    for (auto &basicBlock : basicBlocks) {
+      bool Branch = false;
+      auto &instructions = basicBlock->getInstructions();
+      auto Branchiter = instructions.end();
+      for (auto iter = instructions.begin(); iter != instructions.end(); ++iter) {
+        if (Branch) 
+          usedelete(iter->get());
+        else if ((*iter)->isTerminator()){
+          Branch = true;
+          Branchiter = iter;
+        }
+      }
+      if (Branchiter != instructions.end()) ++Branchiter;
+      while (Branchiter != instructions.end()) 
+        Branchiter = instructions.erase(Branchiter);
+      
+      if (Branch) {  // 更新前驱后继关系
+        auto thelastinstinst = basicBlock->getInstructions().end();
+        --thelastinstinst;
+        auto &Successors = basicBlock->getSuccessors();
+        for (auto iterSucc = Successors.begin(); iterSucc != Successors.end();) {
+          (*iterSucc)->removePredecessor(basicBlock.get());
+          basicBlock->removeSuccessor(*iterSucc);
+        }
+        if (thelastinstinst->get()->isUnconditional()) {
+          BasicBlock* branchBlock = dynamic_cast<BasicBlock *>(thelastinstinst->get()->getOperand(0));
+          basicBlock->addSuccessor(branchBlock);
+          branchBlock->addPredecessor(basicBlock.get());
+        } else if (thelastinstinst->get()->isConditional()) {
+          BasicBlock* thenBlock = dynamic_cast<BasicBlock *>(thelastinstinst->get()->getOperand(1));
+          BasicBlock* elseBlock = dynamic_cast<BasicBlock *>(thelastinstinst->get()->getOperand(2));
+          basicBlock->addSuccessor(thenBlock);
+          basicBlock->addSuccessor(elseBlock);
+          thenBlock->addPredecessor(basicBlock.get());
+          elseBlock->addPredecessor(basicBlock.get());
+        }
+      }
+    }
+  }
+}
+
+
+void SysYOptPre::SysYBlockMerge() {
+  auto &functions = pModule->getFunctions(); //std::map<std::string, std::unique_ptr<Function>>
+  for (auto &function : functions) {
+    // auto basicBlocks = function.second->getBasicBlocks();
+    auto &func = function.second;
+    for (auto blockiter = func->getBasicBlocks().begin();
+         blockiter != func->getBasicBlocks().end();) {
+      if (blockiter->get()->getNumSuccessors() == 1) {
+        // 如果当前块只有一个后继块
+        // 且后继块只有一个前驱块
+        // 则将当前块和后继块合并
+        if (((blockiter->get())->getSuccessors()[0])->getNumPredecessors() == 1) {
+          // std::cout << "merge block: " << blockiter->get()->getName() << std::endl;
+          BasicBlock* block = blockiter->get();
+          BasicBlock* nextBlock = blockiter->get()->getSuccessors()[0];
+          auto nextarguments = nextBlock->getArguments();
+          // 删除br指令
+          if (block->getNumInstructions() != 0) {
+            auto thelastinstinst = block->end();
+            (--thelastinstinst);
+            if (thelastinstinst->get()->isUnconditional()) {
+              usedelete(thelastinstinst->get());
+              block->getInstructions().erase(thelastinstinst);
+            } else if (thelastinstinst->get()->isConditional()) {
+              // 如果是条件分支，判断条件是否相同，主要优化相同布尔表达式
+              if (thelastinstinst->get()->getOperand(1)->getName() == thelastinstinst->get()->getOperand(1)->getName()) {
+                usedelete(thelastinstinst->get());
+                block->getInstructions().erase(thelastinstinst);
+              }
+            }
+          }
+          // 将后继块的指令移动到当前块
+          // 并将后继块的父指针改为当前块
+          for (auto institer = nextBlock->begin(); institer != nextBlock->end();) {
+            institer->get()->setParent(block);
+            block->getInstructions().emplace_back(institer->release());
+            institer = nextBlock->getInstructions().erase(institer);   
+          }
+          // 合并参数
+          // TODO：是否需要去重?
+          for (auto &argm : nextarguments) {
+            argm->setParent(block);
+            block->insertArgument(argm);
+          }
+          // 更新前驱后继关系，类似树节点操作
+          block->removeSuccessor(nextBlock);
+          nextBlock->removePredecessor(block);
+          std::list<BasicBlock *> succshoulddel;
+          for (auto &succ : nextBlock->getSuccessors()) {
+            block->addSuccessor(succ);
+            succ->replacePredecessor(nextBlock, block);
+            succshoulddel.push_back(succ);
+          }
+          for (auto del : succshoulddel) {
+            nextBlock->removeSuccessor(del);
+          }
+
+          func->removeBasicBlock(nextBlock);
+
+        } else {
+          blockiter++;
+        }
+      } else {
+        blockiter++;
+      }
+    }
+  }
+}
+
+// 删除无前驱块，兼容SSA后的处理
+void SysYOptPre::SysYDelNoPreBLock() {
+  
+  auto &functions = pModule->getFunctions(); // std::map<std::string, std::unique_ptr<sysy::Function>>
+  for (auto &function : functions) {
+    auto &func = function.second;
+
+    for (auto &block : func->getBasicBlocks()) {
+      block->setreachableFalse();
+    }
+    // 对函数基本块做一个拓扑排序，排查不可达基本块
+    auto entryBlock = func->getEntryBlock();
+    entryBlock->setreachableTrue();
+    std::queue<BasicBlock *> blockqueue;
+    blockqueue.push(entryBlock);
+    while (!blockqueue.empty()) {
+      auto block = blockqueue.front();
+      blockqueue.pop();
+      for (auto &succ : block->getSuccessors()) {
+        if (!succ->getreachable()) {
+          succ->setreachableTrue();
+          blockqueue.push(succ);
+        }
+      }
+    }
+
+    // 删除不可达基本块指令
+    for (auto blockIter = func->getBasicBlocks().begin(); blockIter != func->getBasicBlocks().end();blockIter++) {
+      
+      if (!blockIter->get()->getreachable()) 
+        for (auto &iterInst : blockIter->get()->getInstructions()) 
+          usedelete(iterInst.get());
+    
+    }
+
+    
+    for (auto blockIter = func->getBasicBlocks().begin(); blockIter != func->getBasicBlocks().end();) {
+      if (!blockIter->get()->getreachable()) {
+        for (auto succblock : blockIter->get()->getSuccessors()) {
+          int indexphi = 1;
+          for (auto pred : succblock->getPredecessors()) {
+            if (pred == blockIter->get()) {
+              break;
+            }
+            indexphi++;
+          }
+          for (auto &phiinst : succblock->getInstructions()) {
+            if (phiinst->getKind() != Instruction::kPhi) {
+              break;
+            }
+            phiinst->removeOperand(indexphi);
+          }
+        }
+        // 删除不可达基本块，注意迭代器不可达问题
+        func->removeBasicBlock((blockIter++)->get());
+      } else {
+        blockIter++;
+      }
+    }
+  }
+}
+
+void SysYOptPre::SysYDelEmptyBlock() {
+  auto &functions = pModule->getFunctions();
+  for (auto &function : functions) {
+    // 收集不可达基本块
+    // 这里的不可达基本块是指没有实际指令的基本块
+    // 当一个基本块没有实际指令例如只有phi指令和一个uncondbr指令时，也会被视作不可达
+    auto basicBlocks = function.second->getBasicBlocks();
+    std::map<sysy::BasicBlock *, BasicBlock *> EmptyBlocks;
+    // 空块儿和后继的基本块的映射
+    for (auto &basicBlock : basicBlocks) {
+      if (basicBlock->getNumInstructions() == 0) {
+        if (basicBlock->getNumSuccessors() == 1) {
+          EmptyBlocks[basicBlock.get()] = basicBlock->getSuccessors().front();
+        }
+      }
+      else{
+        // 如果只有phi指令和一个uncondbr。(phi)*(uncondbr)?
+          // 判断除了最后一个指令之外是不是只有phi指令
+        bool onlyPhi = true;
+        for (auto &inst : basicBlock->getInstructions()) {
+          if (!inst->isPhi() && !inst->isUnconditional()) {
+            onlyPhi = false;
+            break;
+          }
+        }
+        if(onlyPhi)
+          EmptyBlocks[basicBlock.get()] = basicBlock->getSuccessors().front();
+      }
+      
+      
+    }
+    // 更新基本块信息，增加必要指令
+    for (auto &basicBlock : basicBlocks) {
+      // 把空块转换成只有跳转指令的不可达块
+      if (distance(basicBlock->begin(), basicBlock->end()) == 0) {
+        if (basicBlock->getNumSuccessors() == 0) {
+          continue;
+        }
+        if (basicBlock->getNumSuccessors() > 1) {
+          assert("");
+        }
+        pBuilder->setPosition(basicBlock.get(), basicBlock->end());
+        pBuilder->createUncondBrInst(basicBlock->getSuccessors()[0], {});
+        continue;
+      }
+
+      auto thelastinst = basicBlock->getInstructions().end();
+      --thelastinst;
+
+      // 根据br指令传递的后继块信息，跳过空块链
+      if (thelastinst->get()->isUnconditional()) {
+        BasicBlock* OldBrBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0));
+        BasicBlock *thelastBlockOld = nullptr;
+        // 如果空块链表为多个块
+        while (EmptyBlocks.find(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))) !=
+               EmptyBlocks.end()) {
+          thelastBlockOld = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0));
+          thelastinst->get()->replaceOperand(0, EmptyBlocks[thelastBlockOld]);
+        }
+
+        basicBlock->removeSuccessor(OldBrBlock);
+        OldBrBlock->removePredecessor(basicBlock.get());
+        basicBlock->addSuccessor(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0)));
+        dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->addPredecessor(basicBlock.get());
+
+        if (thelastBlockOld != nullptr) {
+          int indexphi = 0;
+          for (auto &pred : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->getPredecessors()) {
+            if (pred == thelastBlockOld) {
+              break;
+            }
+            indexphi++;
+          }
+
+          // 更新phi指令的操作数
+          // 移除thelastBlockOld对应的phi操作数
+          for (auto &InstInNew : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->getInstructions()) {
+            if (InstInNew->isPhi()) {
+              dynamic_cast<PhiInst *>(InstInNew.get())->removeOperand(indexphi + 1);
+            } else {
+              break;
+            }
+          }
+        }
+
+      } else if (thelastinst->get()->getKind() == Instruction::kCondBr) {
+        auto OldThenBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1));
+        auto OldElseBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2));
+
+        BasicBlock *thelastBlockOld = nullptr;
+        while (EmptyBlocks.find(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1))) !=
+               EmptyBlocks.end()) {
+          thelastBlockOld = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1));
+          thelastinst->get()->replaceOperand(
+              1, EmptyBlocks[dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1))]);
+        }
+        basicBlock->removeSuccessor(OldThenBlock);
+        OldThenBlock->removePredecessor(basicBlock.get());
+        // 处理 then 和 else 分支合并的情况
+        if (dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1)) ==
+            dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))) {
+          auto thebrBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1));
+          usedelete(thelastinst->get());
+          thelastinst = basicBlock->getInstructions().erase(thelastinst);
+          pBuilder->setPosition(basicBlock.get(), basicBlock->end());
+          pBuilder->createUncondBrInst(thebrBlock, {});
+          continue;
+        }
+        basicBlock->addSuccessor(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1)));
+        dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1))->addPredecessor(basicBlock.get());
+        // auto indexInNew = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->getPredecessors().
+
+        if (thelastBlockOld != nullptr) {
+          int indexphi = 0;
+          for (auto &pred : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1))->getPredecessors()) {
+            if (pred == thelastBlockOld) {
+              break;
+            }
+            indexphi++;
+          }
+
+          for (auto &InstInNew : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1))->getInstructions()) {
+            if (InstInNew->isPhi()) {
+              dynamic_cast<PhiInst *>(InstInNew.get())->removeOperand(indexphi + 1);
+            } else {
+              break;
+            }
+          }
+        }
+
+        thelastBlockOld = nullptr;
+        while (EmptyBlocks.find(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))) !=
+               EmptyBlocks.end()) {
+          thelastBlockOld = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2));
+          thelastinst->get()->replaceOperand(
+              2, EmptyBlocks[dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))]);
+        }
+        basicBlock->removeSuccessor(OldElseBlock);
+        OldElseBlock->removePredecessor(basicBlock.get());
+        // 处理 then 和 else 分支合并的情况
+        if (dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1)) ==
+            dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))) {
+          auto thebrBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(1));
+          usedelete(thelastinst->get());
+          thelastinst = basicBlock->getInstructions().erase(thelastinst);
+          pBuilder->setPosition(basicBlock.get(), basicBlock->end());
+          pBuilder->createUncondBrInst(thebrBlock, {});
+          continue;
+        }
+        basicBlock->addSuccessor(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2)));
+        dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))->addPredecessor(basicBlock.get());
+
+        if (thelastBlockOld != nullptr) {
+          int indexphi = 0;
+          for (auto &pred : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))->getPredecessors()) {
+            if (pred == thelastBlockOld) {
+              break;
+            }
+            indexphi++;
+          }
+          for (auto &InstInNew : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(2))->getInstructions()) {
+            if (InstInNew->isPhi()) {
+              dynamic_cast<PhiInst *>(InstInNew.get())->removeOperand(indexphi + 1);
+            } else {
+              break;
+            }
+          }
+        }
+      } else {
+        if (basicBlock->getNumSuccessors() == 1) {
+          pBuilder->setPosition(basicBlock.get(), basicBlock->end());
+          pBuilder->createUncondBrInst(basicBlock->getSuccessors()[0], {});
+          auto thelastinst = basicBlock->getInstructions().end();
+          (--thelastinst);
+          auto OldBrBlock = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0));
+          sysy::BasicBlock *thelastBlockOld = nullptr;
+          while (EmptyBlocks.find(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))) !=
+                 EmptyBlocks.end()) {
+            thelastBlockOld = dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0));
+
+            thelastinst->get()->replaceOperand(
+                0, EmptyBlocks[dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))]);
+          }
+          
+          basicBlock->removeSuccessor(OldBrBlock);
+          OldBrBlock->removePredecessor(basicBlock.get());
+          basicBlock->addSuccessor(dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0)));
+          dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->addPredecessor(basicBlock.get());
+          if (thelastBlockOld != nullptr) {
+            int indexphi = 0;
+            for (auto &pred : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->getPredecessors()) {
+              if (pred == thelastBlockOld) {
+                break;
+              }
+              indexphi++;
+            }
+
+            for (auto &InstInNew : dynamic_cast<BasicBlock *>(thelastinst->get()->getOperand(0))->getInstructions()) {
+              if (InstInNew->isPhi()) {
+                dynamic_cast<PhiInst *>(InstInNew.get())->removeOperand(indexphi + 1);
+              } else {
+                break;
+              }
+            }
+          }
+        }
+      }
+    }
+
+    for (auto iter = function.second->getBasicBlocks().begin(); iter != function.second->getBasicBlocks().end();) {
+      
+      if (EmptyBlocks.find(iter->get()) != EmptyBlocks.end()) {
+        // EntryBlock跳过
+        if (iter->get() == function.second->getEntryBlock()) {
+          ++iter;
+          continue;
+        }
+
+        for (auto &iterInst : iter->get()->getInstructions()) 
+          usedelete(iterInst.get());
+        // 删除不可达基本块的phi指令的操作数
+        for (auto &succ : iter->get()->getSuccessors()) {
+          int index = 0;
+          for (auto &pred : succ->getPredecessors()) {
+            if (pred == iter->get()) {
+              break;
+            }
+            index++;
+          }
+
+          for (auto &instinsucc : succ->getInstructions()) {
+            if (instinsucc->isPhi()) {
+              dynamic_cast<PhiInst *>(instinsucc.get())->removeOperand(index);
+            } else {
+              break;
+            }
+          }
+        }
+
+        function.second->removeBasicBlock((iter++)->get());
+      } else {
+        ++iter;
+      }
+    }
+  }
+}
+
+// 如果函数没有返回指令，则添加一个默认返回指令(主要解决void函数没有返回指令的问题)
+void SysYOptPre::SysYAddReturn() {
+  auto &functions = pModule->getFunctions();
+  for (auto &function : functions) {
+    auto &func = function.second;
+    auto basicBlocks = func->getBasicBlocks();
+    for (auto &block : basicBlocks) {
+      if (block->getNumSuccessors() == 0) {
+        // 如果基本块没有后继块，则添加一个返回指令
+        if (block->getNumInstructions() == 0) {
+          pBuilder->setPosition(block.get(), block->end());
+          pBuilder->createReturnInst();
+        }
+        auto thelastinst = block->getInstructions().end();
+        --thelastinst;
+        if (thelastinst->get()->getKind() != Instruction::kReturn) {
+          // std::cout << "Warning: Function " << func->getName() << " has no return instruction, adding default return." << std::endl;
+
+          pBuilder->setPosition(block.get(), block->end());
+          // TODO: 如果int float函数缺少返回值是否需要报错
+          if (func->getReturnType()->isInt()) {
+            pBuilder->createReturnInst(ConstantInteger::get(0));
+          } else if (func->getReturnType()->isFloat()) {
+            pBuilder->createReturnInst(ConstantFloating::get(0.0F));
+          } else {
+            pBuilder->createReturnInst();
+          }
+        }
+      }
+    }
+  }
+}
+
+}  // namespace sysy

src/SysYIRPassManager.cpp

@@ -1,36 +0,0 @@
-// PassManager.cpp
-#include "SysYIRPassManager.h"
-#include <iostream>
-
-namespace sysy {
-
-void PassManager::run(Module& M) {
-  // 首先运行Module级别的Pass
-  for (auto& pass : modulePasses) {
-    std::cout << "Running Module Pass: " << pass->getPassName() << std::endl;
-    pass->runOnModule(M);
-  }
-
-  // 然后对每个函数运行Function级别的Pass
-  auto& functions = M.getFunctions();
-  for (auto& pair : functions) {
-    Function& F = *(pair.second); // 获取Function的引用
-    std::cout << "  Processing Function: " << F.getName() << std::endl;
-
-    // 在每个函数上运行FunctionPasses
-    bool changedInFunction;
-    do {
-      changedInFunction = false;
-      for (auto& pass : functionPasses) {
-        // 对于FunctionPasses，可以考虑一个迭代执行的循环，直到稳定
-        std::cout << "    Running Function Pass: " << pass->getPassName() << std::endl;
-        changedInFunction |= pass->runOnFunction(F);
-      }
-    } while (changedInFunction); // 循环直到函数稳定，这模拟了您SysYCFGOpt的while(changed)逻辑
-  }
-
-  // 分析Pass的运行可以在其他Pass需要时触发，或者在特定的PassManager阶段触发
-  // 对于依赖于分析结果的Pass，可以在其run方法中通过PassManager::getAnalysis()来获取
-}
-
-} // namespace sysy

src/SysYIRPrinter.cpp

@@ -3,11 +3,12 @@
 #include <fstream>
 #include <iostream>
 #include <string>
-#include "IR.h" // 确保IR.h包含了ArrayType、GetElementPtrInst等的定义
+#include "IR.h"
 
 namespace sysy {
 
 void SysYPrinter::printIR() {
+
   const auto &functions = pModule->getFunctions();
 
   //TODO: Print target datalayout and triple (minimal required by LLVM)
@@ -35,18 +36,11 @@ std::string SysYPrinter::getTypeString(Type *type) {
     return "i32";
   } else if (type->isFloat()) {
     return "float";
+    
   } else if (auto ptrType = dynamic_cast<PointerType*>(type)) {
-    // 递归打印指针指向的类型，然后加上 '*'
     return getTypeString(ptrType->getBaseType()) + "*";
-  } else if (auto funcType = dynamic_cast<FunctionType*>(type)) {
-    // 对于函数类型，打印其返回类型
-    // 注意：这里可能需要更完整的函数签名打印，取决于你的IR表示方式
-    // 比如：`retType (paramType1, paramType2, ...)`
-    // 但为了简化和LLVM IR兼容性，通常在定义时完整打印
-    return getTypeString(funcType->getReturnType());
-  } else if (auto arrayType = dynamic_cast<ArrayType*>(type)) { // 新增：处理数组类型
-    // 打印格式为 [num_elements x element_type]
-    return "[" + std::to_string(arrayType->getNumElements()) + " x " + getTypeString(arrayType->getElementType()) + "]";
+  } else if (auto ptrType = dynamic_cast<FunctionType*>(type)) {
+    return getTypeString(ptrType->getReturnType());
   }
   assert(false && "Unsupported type");
   return "";
@@ -57,23 +51,15 @@ std::string SysYPrinter::getValueName(Value *value) {
     return "@" + global->getName();
   } else if (auto inst = dynamic_cast<Instruction*>(value)) {
     return "%" + inst->getName();
-  } else if (auto constInt = dynamic_cast<ConstantInteger*>(value)) { // 优先匹配具体的常量类型
-    return std::to_string(constInt->getInt());
-  } else if (auto constFloat = dynamic_cast<ConstantFloating*>(value)) { // 优先匹配具体的常量类型
-    return std::to_string(constFloat->getFloat());
-  } else if (auto constUndef = dynamic_cast<UndefinedValue*>(value)) { // 如果有Undef类型
-    return "undef";
-  } else if (auto constVal = dynamic_cast<ConstantValue*>(value)) { // fallback for generic ConstantValue
-    // 这里的逻辑可能需要根据你ConstantValue的实际设计调整
-    // 确保它能处理所有可能的ConstantValue
-    if (constVal->getType()->isFloat()) {
-        return std::to_string(constVal->getFloat());
+  } else if (auto constVal = dynamic_cast<ConstantValue*>(value)) {
+    if (constVal->isFloat()) {
+      return std::to_string(constVal->getFloat());
     }
     return std::to_string(constVal->getInt());
   } else if (auto constVar = dynamic_cast<ConstantVariable*>(value)) {
-    return constVar->getName(); // 假设ConstantVariable有自己的名字或通过getByIndices获取值
+    return constVar->getName();
   }
-  assert(false && "Unknown value type or unable to get value name");
+  assert(false && "Unknown value type");
   return "";
 }
 
@@ -91,35 +77,44 @@ void SysYPrinter::printGlobalVariable() {
   for (const auto &global : globals) {
     std::cout << "@" << global->getName() << " = global ";
     
-    // 全局变量的类型是一个指针，指向其基类型 (可能是 ArrayType 或 Integer/FloatType)
-    auto globalVarBaseType = dynamic_cast<PointerType *>(global->getType())->getBaseType();
-    printType(globalVarBaseType); // 打印全局变量的实际类型 (例如 i32 或 [10 x i32])
+    auto baseType = dynamic_cast<PointerType *>(global->getType())->getBaseType();
+    printType(baseType);
+    
+    if (global->getNumDims() > 0) {
+      // Array type
+      std::cout << " [";
+      for (unsigned i = 0; i < global->getNumDims(); i++) {
+        if (i > 0) std::cout << " x ";
+        std::cout << getValueName(global->getDim(i));
+      }
+      std::cout << "]";
+    }
     
     std::cout << " ";
     
-    // 检查是否是数组类型 (通过检查 globalVarBaseType 是否是 ArrayType)
-    if (globalVarBaseType->isArray()) {
-      // 数组初始化器
-      std::cout << "["; // LLVM IR 数组初始化器格式: [type value, type value, ...]
-      auto values = global->getInitValues(); // 假设 getInitValues() 返回一个 ValueCounter
-      const std::vector<sysy::Value *> &counterValues = values.getValues(); // 获取所有值
+    if (global->getNumDims() > 0) {
+      // Array initializer
+      std::cout << "[";
+      auto values = global->getInitValues();
+      auto counterValues = values.getValues();
+      auto counterNumbers = values.getNumbers();
       
-      for (size_t i = 0; i < counterValues.size(); i++) {
+      for (size_t i = 0; i < counterNumbers.size(); i++) {
         if (i > 0) std::cout << ", ";
-        // 打印元素类型，这个元素类型应该是数组的最终元素类型，例如 i32 或 float
-        // 可以从 globalVarBaseType 逐层剥离得到最终元素类型，但这里简化为直接从值获取
-        printType(counterValues[i]->getType());
-        std::cout << " ";
-        printValue(counterValues[i]);
+        if (baseType->isFloat()) {
+          std::cout << "float " << dynamic_cast<ConstantValue*>(counterValues[i])->getFloat();
+        } else {
+          std::cout << "i32 " << dynamic_cast<ConstantValue*>(counterValues[i])->getInt();
+        }
       }
       std::cout << "]";
     } else {
-      // 标量初始化器
-      // 假设标量全局变量的初始化值通过 getByIndex(0) 获取
-      Value* initVal = global->getByIndex(0);
-      printType(initVal->getType()); // 打印标量值的类型
-      std::cout << " ";
-      printValue(initVal); // 打印标量值
+      // Scalar initializer
+      if (baseType->isFloat()) {
+        std::cout << "float " << dynamic_cast<ConstantValue*>(global->getByIndex(0))->getFloat();
+      } else {
+        std::cout << "i32 " << dynamic_cast<ConstantValue*>(global->getByIndex(0))->getInt();
+      }
     }
     
     std::cout << ", align 4" << std::endl;
@@ -214,19 +209,19 @@ void SysYPrinter::printInst(Instruction *pInst) {
         case Kind::kFDiv: std::cout << "fdiv"; break;
         case Kind::kICmpEQ: std::cout << "icmp eq"; break;
         case Kind::kICmpNE: std::cout << "icmp ne"; break;
-        case Kind::kICmpLT: std::cout << "icmp slt"; break; // LLVM uses slt/sgt for signed less/greater than
+        case Kind::kICmpLT: std::cout << "icmp slt"; break;
         case Kind::kICmpGT: std::cout << "icmp sgt"; break;
         case Kind::kICmpLE: std::cout << "icmp sle"; break;
         case Kind::kICmpGE: std::cout << "icmp sge"; break;
-        case Kind::kFCmpEQ: std::cout << "fcmp oeq"; break; // oeq for ordered equal
-        case Kind::kFCmpNE: std::cout << "fcmp one"; break; // one for ordered not equal
-        case Kind::kFCmpLT: std::cout << "fcmp olt"; break; // olt for ordered less than
-        case Kind::kFCmpGT: std::cout << "fcmp ogt"; break; // ogt for ordered greater than
-        case Kind::kFCmpLE: std::cout << "fcmp ole"; break; // ole for ordered less than or equal
-        case Kind::kFCmpGE: std::cout << "fcmp oge"; break; // oge for ordered greater than or equal
+        case Kind::kFCmpEQ: std::cout << "fcmp oeq"; break;
+        case Kind::kFCmpNE: std::cout << "fcmp one"; break;
+        case Kind::kFCmpLT: std::cout << "fcmp olt"; break;
+        case Kind::kFCmpGT: std::cout << "fcmp ogt"; break;
+        case Kind::kFCmpLE: std::cout << "fcmp ole"; break;
+        case Kind::kFCmpGE: std::cout << "fcmp oge"; break;
         case Kind::kAnd: std::cout << "and"; break;
         case Kind::kOr: std::cout << "or"; break;
-        default: break; // Should not reach here
+        default: break;
       }
       
       // Types and operands
@@ -243,6 +238,7 @@ void SysYPrinter::printInst(Instruction *pInst) {
     case Kind::kNeg:
     case Kind::kNot:
     case Kind::kFNeg:
+    case Kind::kFNot:
     case Kind::kFtoI:
     case Kind::kBitFtoI:
     case Kind::kItoF:
@@ -254,39 +250,31 @@ void SysYPrinter::printInst(Instruction *pInst) {
       }
       
       switch (pInst->getKind()) {
-        case Kind::kNeg: std::cout << "sub "; break; // integer negation is `sub i32 0, operand`
-        case Kind::kNot: std::cout << "xor "; break; // logical/bitwise NOT is `xor i32 -1, operand` or `xor i1 true, operand`
-        case Kind::kFNeg: std::cout << "fneg "; break; // float negation
-        case Kind::kFtoI: std::cout << "fptosi "; break; // float to signed integer
-        case Kind::kBitFtoI: std::cout << "bitcast "; break; // bitcast float to int
-        case Kind::kItoF: std::cout << "sitofp "; break; // signed integer to float
-        case Kind::kBitItoF: std::cout << "bitcast "; break; // bitcast int to float
-        default: break; // Should not reach here
+        case Kind::kNeg: std::cout << "sub "; break;
+        case Kind::kNot: std::cout << "not "; break;
+        case Kind::kFNeg: std::cout << "fneg "; break;
+        case Kind::kFNot: std::cout << "fneg "; break; // FNot not standard, map to fneg
+        case Kind::kFtoI: std::cout << "fptosi "; break;
+        case Kind::kBitFtoI: std::cout << "bitcast "; break;
+        case Kind::kItoF: std::cout << "sitofp "; break;
+        case Kind::kBitItoF: std::cout << "bitcast "; break;
+        default: break;
       }
       
-      printType(unyInst->getOperand()->getType()); // Print operand type
+      printType(unyInst->getType());
       std::cout << " ";
       
-      // Special handling for integer negation and logical NOT
-      if (pInst->getKind() == Kind::kNeg) {
-        std::cout << "0, "; // for 'sub i32 0, operand'
-      } else if (pInst->getKind() == Kind::kNot) {
-        // For logical NOT (i1 -> i1), use 'xor i1 true, operand'
-        // For bitwise NOT (i32 -> i32), use 'xor i32 -1, operand'
-        if (unyInst->getOperand()->getType()->isInt()) { // Assuming i32 for bitwise NOT
-            std::cout << "NOT, "; // or specific bitmask for NOT
-        } else { // Assuming i1 for logical NOT
-            std::cout << "true, ";
-        }
+      // Special handling for negation
+      if (pInst->getKind() == Kind::kNeg || pInst->getKind() == Kind::kNot) {
+        std::cout << "i32 0, ";
       }
       
       printValue(pInst->getOperand(0));
       
-      // For type conversions (fptosi, sitofp, bitcast), need to specify destination type
-      if (pInst->getKind() == Kind::kFtoI || pInst->getKind() == Kind::kItoF ||
-          pInst->getKind() == Kind::kBitFtoI || pInst->getKind() == Kind::kBitItoF) {
+      // For bitcast, need to specify destination type
+      if (pInst->getKind() == Kind::kBitFtoI || pInst->getKind() == Kind::kBitItoF) {
         std::cout << " to ";
-        printType(unyInst->getType()); // Print result type
+        printType(unyInst->getType());
       }
       
       std::cout << std::endl;
@@ -301,7 +289,7 @@ void SysYPrinter::printInst(Instruction *pInst) {
       }
       
       std::cout << "call ";
-      printType(callInst->getType()); // Return type of the call
+      printType(callInst->getType());
       std::cout << " @" << function->getName() << "(";
       
       auto params = callInst->getArguments();
@@ -309,9 +297,9 @@ void SysYPrinter::printInst(Instruction *pInst) {
       for (auto &param : params) {
         if (!first) std::cout << ", ";
         first = false;
-        printType(param->getValue()->getType()); // Type of argument
+        printType(param->getValue()->getType());
         std::cout << " ";
-        printValue(param->getValue()); // Value of argument
+        printValue(param->getValue());
       }
       
       std::cout << ")" << std::endl;
@@ -319,7 +307,7 @@ void SysYPrinter::printInst(Instruction *pInst) {
     
     case Kind::kCondBr: {
       auto condBrInst = dynamic_cast<CondBrInst *>(pInst);
-      std::cout << "br i1 "; // Condition type should be i1
+      std::cout << "br i1 ";
       printValue(condBrInst->getCondition());
       std::cout << ", label %" << condBrInst->getThenBlock()->getName();
       std::cout << ", label %" << condBrInst->getElseBlock()->getName();
@@ -349,17 +337,14 @@ void SysYPrinter::printInst(Instruction *pInst) {
       auto allocaInst = dynamic_cast<AllocaInst *>(pInst);
       std::cout << "%" << allocaInst->getName() << " = alloca ";
       
-      // AllocaInst 的类型现在应该是一个 PointerType，指向正确的 ArrayType 或 ScalarType
-      // 例如：alloca i32, align 4  或者 alloca [10 x i32], align 4
-      auto allocatedType = dynamic_cast<PointerType *>(allocaInst->getType())->getBaseType();
-      printType(allocatedType);
+      auto baseType = dynamic_cast<PointerType *>(allocaInst->getType())->getBaseType();
+      printType(baseType);
       
-      // 仍然打印维度信息，如果存在的话
-      if (allocaInst->getNumDims() > 0) { 
+      if (allocaInst->getNumDims() > 0) {
         std::cout << ", ";
         for (size_t i = 0; i < allocaInst->getNumDims(); i++) {
           if (i > 0) std::cout << ", ";
-          printType(Type::getIntType()); // 维度大小通常是 i32 类型
+          printType(Type::getIntType());
           std::cout << " ";
           printValue(allocaInst->getDim(i));
         }
@@ -371,74 +356,70 @@ void SysYPrinter::printInst(Instruction *pInst) {
     case Kind::kLoad: {
       auto loadInst = dynamic_cast<LoadInst *>(pInst);
       std::cout << "%" << loadInst->getName() << " = load ";
-      printType(loadInst->getType()); // 加载的结果类型
+      printType(loadInst->getType());
       std::cout << ", ";
-      printType(loadInst->getPointer()->getType()); // 指针类型
+      printType(loadInst->getPointer()->getType());
       std::cout << " ";
-      printValue(loadInst->getPointer()); // 要加载的地址
+      printValue(loadInst->getPointer());
       
-      // 仍然打印索引信息，如果存在的话
       if (loadInst->getNumIndices() > 0) {
-        std::cout << ", indices "; // 或者其他分隔符，取决于你期望的格式
+        std::cout << ", ";
         for (size_t i = 0; i < loadInst->getNumIndices(); i++) {
-            if (i > 0) std::cout << ", ";
-            printType(loadInst->getIndex(i)->getType());
-            std::cout << " ";
-            printValue(loadInst->getIndex(i));
+          if (i > 0) std::cout << ", ";
+          printType(Type::getIntType());
+          std::cout << " ";
+          printValue(loadInst->getIndex(i));
         }
       }
       
       std::cout << ", align 4" << std::endl;
     } break;
     
+    case Kind::kLa: {
+      auto laInst = dynamic_cast<LaInst *>(pInst);
+      std::cout << "%" << laInst->getName() << " = getelementptr inbounds ";
+      
+      auto ptrType = dynamic_cast<PointerType*>(laInst->getPointer()->getType());
+      printType(ptrType->getBaseType());
+      std::cout << ", ";
+      printType(laInst->getPointer()->getType());
+      std::cout << " ";
+      printValue(laInst->getPointer());
+      std::cout << ", ";
+      
+      for (size_t i = 0; i < laInst->getNumIndices(); i++) {
+        if (i > 0) std::cout << ", ";
+        printType(Type::getIntType());
+        std::cout << " ";
+        printValue(laInst->getIndex(i));
+      }
+      
+      std::cout << std::endl;
+    } break;
+    
     case Kind::kStore: {
       auto storeInst = dynamic_cast<StoreInst *>(pInst);
       std::cout << "store ";
-      printType(storeInst->getValue()->getType()); // 要存储的值的类型
+      printType(storeInst->getValue()->getType());
       std::cout << " ";
-      printValue(storeInst->getValue()); // 要存储的值
+      printValue(storeInst->getValue());
       std::cout << ", ";
-      printType(storeInst->getPointer()->getType()); // 目标指针的类型
+      printType(storeInst->getPointer()->getType());
       std::cout << " ";
-      printValue(storeInst->getPointer()); // 目标地址
+      printValue(storeInst->getPointer());
       
-      // 仍然打印索引信息，如果存在的话
       if (storeInst->getNumIndices() > 0) {
-        std::cout << ", indices "; // 或者其他分隔符
+        std::cout << ", ";
         for (size_t i = 0; i < storeInst->getNumIndices(); i++) {
-            if (i > 0) std::cout << ", ";
-            printType(storeInst->getIndex(i)->getType());
-            std::cout << " ";
-            printValue(storeInst->getIndex(i));
+          if (i > 0) std::cout << ", ";
+          printType(Type::getIntType());
+          std::cout << " ";
+          printValue(storeInst->getIndex(i));
         }
       }
       
       std::cout << ", align 4" << std::endl;
     } break;
-
-    case Kind::kGetElementPtr: { // 新增：GetElementPtrInst 打印
-      auto gepInst = dynamic_cast<GetElementPtrInst*>(pInst);
-      std::cout << "%" << gepInst->getName() << " = getelementptr inbounds "; // 假设总是 inbounds
-      
-      // GEP 的第一个操作数是基指针，其类型是一个指向聚合类型的指针
-      // 第一个参数是基指针所指向的聚合类型的类型 (e.g., [10 x i32])
-      auto basePtrType = dynamic_cast<PointerType*>(gepInst->getBasePointer()->getType());
-      printType(basePtrType->getBaseType()); // 打印基指针指向的类型
-      
-      std::cout << ", ";
-      printType(gepInst->getBasePointer()->getType()); // 打印基指针自身的类型 (e.g., [10 x i32]*)
-      std::cout << " ";
-      printValue(gepInst->getBasePointer()); // 打印基指针
-      
-      // 打印所有索引
-      for (auto indexVal : gepInst->getIndices()) { // 使用 getIndices() 迭代器
-        std::cout << ", ";
-        printType(indexVal->getValue()->getType()); // 打印索引的类型 (通常是 i32)
-        std::cout << " ";
-        printValue(indexVal->getValue()); // 打印索引值
-      }
-      std::cout << std::endl;
-    } break;
     
     case Kind::kMemset: {
       auto memsetInst = dynamic_cast<MemsetInst *>(pInst);
@@ -452,40 +433,51 @@ void SysYPrinter::printInst(Instruction *pInst) {
       printValue(memsetInst->getValue());
       std::cout << ", i32 ";
       printValue(memsetInst->getSize());
-      std::cout << ", i1 false)" << std::endl; // alignment for memset is typically i1
+      std::cout << ", i1 false)" << std::endl;
     } break;
     
     case Kind::kPhi: {
       auto phiInst = dynamic_cast<PhiInst *>(pInst);
-      // Phi 指令的名称通常是结果变量
-      std::cout << "%" << phiInst->getName() << " = phi ";
-      printType(phiInst->getType()); // Phi 结果类型
+      printValue(phiInst->getOperand(0));
+      std::cout << " = phi ";
+      printType(phiInst->getType());
       
-      // Phi 指令的操作数是成对的 [value, basic_block]
-      // 这里假设 getOperands() 返回的是 (val1, block1, val2, block2...)
-      // 如果你的 PhiInst 存储方式是 getIncomingValues() 和 getIncomingBlocks()，请相应调整
-      // LLVM IR 格式: phi type [value1, block1], [value2, block2]
-      bool firstPair = true;
-      for (unsigned i = 0; i < phiInst->getNumOperands() / 2; ++i) { // 遍历成对的操作数
-        if (!firstPair) std::cout << ", ";
-        firstPair = false;
+      for (unsigned i = 1; i < phiInst->getNumOperands(); i++) {
+        if (i > 0) std::cout << ", ";
         std::cout << "[ ";
-        printValue(phiInst->getOperand(i * 2)); // value
-        std::cout << ", %";
-        printValue(phiInst->getOperand(i * 2 + 1)); // block
+        printValue(phiInst->getOperand(i));
         std::cout << " ]";
       }
       std::cout << std::endl;
     } break;
     
-    // 以下两个 Kind 应该删除或替换为 kGEP
-    // case Kind::kLa: { /* REMOVED */ } break;
-    // case Kind::kGetSubArray: { /* REMOVED */ } break;
+    case Kind::kGetSubArray: {
+      auto getSubArrayInst = dynamic_cast<GetSubArrayInst *>(pInst);
+      std::cout << "%" << getSubArrayInst->getName() << " = getelementptr inbounds ";
+      
+      auto ptrType = dynamic_cast<PointerType*>(getSubArrayInst->getFatherArray()->getType());
+      printType(ptrType->getBaseType());
+      std::cout << ", ";
+      printType(getSubArrayInst->getFatherArray()->getType());
+      std::cout << " ";
+      printValue(getSubArrayInst->getFatherArray());
+      std::cout << ", ";
+      bool firstIndex = true;
+      for (auto &index : getSubArrayInst->getIndices()) {
+        if (!firstIndex) std::cout << ", ";
+        firstIndex = false;
+        printType(Type::getIntType());
+        std::cout << " ";
+        printValue(index->getValue());
+      }
+      
+      std::cout << std::endl;
+    } break;
     
     default:
-      assert(false && "Unsupported instruction kind in SysYPrinter");
+      assert(false && "Unsupported instruction kind");
       break;
   }
 }
 
-}  // namespace sysy
+}  // namespace sysy

src/include/DeadCodeElimination.h

@@ -3,7 +3,6 @@
 #include "IR.h"
 #include "SysYIRAnalyser.h"
 #include "SysYIRPrinter.h"
-#include "SysYIROptUtils.h"
 
 namespace sysy {
 
@@ -32,5 +31,9 @@ class DeadCodeElimination {
   void eliminateDeadGlobals(bool& changed);   // 消除无用全局变量
   void eliminateDeadIndirectiveAllocas(Function* func, bool& changed);  // 消除无用间接内存分配(phi节点)
   void eliminateDeadRedundantLoadStore(Function* func, bool& changed);  // 消除冗余加载和存储
+  bool isGlobal(Value *val);
+  bool isArr(Value *val);
+  void usedelete(Instruction *instr);
+
 };
 }  // namespace sysy

src/include/IR.h

@@ -49,7 +49,6 @@ class Type {
     kLabel,
     kPointer,
     kFunction,
-    kArray,
   };
 
   Kind kind;  ///< 表示具体类型的变量
@@ -66,7 +65,6 @@ class Type {
   static Type* getPointerType(Type *baseType);  ///< 返回表示指向baseType类型的Pointer类型的Type指针
   static Type* getFunctionType(Type *returnType, const std::vector<Type *> &paramTypes = {});
   ///< 返回表示返回类型为returnType,形参类型列表为paramTypes的函数类型的Type指针
-  static Type* getArrayType(Type *elementType, unsigned numElements);
 
  public:
   Kind getKind() const { return kind; }                  ///< 返回Type对象代表原始标量类型
@@ -76,7 +74,6 @@ class Type {
   bool isLabel() const { return kind == kLabel; }        ///< 判定是否为Label类型
   bool isPointer() const { return kind == kPointer; }    ///< 判定是否为Pointer类型
   bool isFunction() const { return kind == kFunction; }  ///< 判定是否为Function类型
-  bool isArray() const { return kind == Kind::kArray; }
   unsigned getSize() const;                              ///< 返回类型所占的空间大小(字节)
   /// 尝试将一个变量转换为给定的Type及其派生类类型的变量
   template <typename T>
@@ -118,22 +115,6 @@ class FunctionType : public Type {
   unsigned getNumParams() const { return paramTypes.size(); }  ///< 获取形参数量
 };
 
-class ArrayType : public Type {
- public:
-  // elements：数组的元素类型 (例如，int[3] 的 elementType 是 int)
-  // numElements：该维度的大小 (例如，int[3] 的 numElements 是 3)
-  static ArrayType *get(Type *elementType, unsigned numElements);
-
-  Type *getElementType() const { return elementType; }
-  unsigned getNumElements() const { return numElements; }
-
- protected:
-  ArrayType(Type *elementType, unsigned numElements)
-      : Type(Kind::kArray), elementType(elementType), numElements(numElements) {}
-  Type *elementType;
-  unsigned numElements; // 当前维度的大小
-};
-
 /*!
  * @}
  */
@@ -621,6 +602,49 @@ class User : public Value {
   void setOperand(unsigned index, Value *value);      ///< 设置操作数
 };
 
+class GetSubArrayInst;
+/**
+ * 左值 具有地址的对象
+ */
+class LVal {
+  friend class GetSubArrayInst;
+
+ protected:
+  LVal *fatherLVal{};                              ///< 父左值
+  std::list<std::unique_ptr<LVal>> childrenLVals;  ///< 子左值
+  GetSubArrayInst *defineInst{};                   /// 定义该左值的GetSubArray指令
+
+ protected:
+  LVal() = default;
+
+ public:
+  virtual ~LVal() = default;
+  virtual std::vector<Value *> getLValDims() const = 0;  ///< 获取左值的维度
+  virtual unsigned getLValNumDims() const  = 0;           ///< 获取左值的维度数量
+
+ public:
+  LVal* getFatherLVal() const { return fatherLVal; }  ///< 获取父左值
+  const std::list<std::unique_ptr<LVal>>& getChildrenLVals() const {
+    return childrenLVals;
+  }  ///< 获取子左值列表
+  LVal* getAncestorLVal() const {
+    auto curLVal = const_cast<LVal *>(this);
+    while (curLVal->getFatherLVal() != nullptr) {
+      curLVal = curLVal->getFatherLVal();
+    }
+    return curLVal;
+  }                                                                  ///< 获取祖先左值
+  void setFatherLVal(LVal *father) { fatherLVal = father; }          ///< 设置父左值
+  void setDefineInst(GetSubArrayInst *inst) { defineInst = inst; }   ///< 设置定义指令
+  void addChild(LVal *child) { childrenLVals.emplace_back(child); }  ///< 添加子左值
+  void removeChild(LVal *child) {
+    auto iter = std::find_if(childrenLVals.begin(), childrenLVals.end(),
+                             [child](const std::unique_ptr<LVal> &ptr) { return ptr.get() == child; });
+    childrenLVals.erase(iter);
+  }                                                                       ///< 移除子左值
+  GetSubArrayInst* getDefineInst() const { return defineInst; }  ///< 获取定义指令
+};
+
 /*!
  * Base of all concrete instruction types.
  */
@@ -670,15 +694,15 @@ class Instruction : public User {
     kAlloca = 0x1UL << 33,
     kLoad = 0x1UL << 34,
     kStore = 0x1UL << 35,
-    kGetElementPtr = 0x1UL << 36,
+    kLa = 0x1UL << 36,
     kMemset = 0x1UL << 37,
-    // kGetSubArray = 0x1UL << 38,
+    kGetSubArray = 0x1UL << 38,
     // Constant Kind removed as Constants are now Values, not Instructions.
     // kConstant = 0x1UL << 37, // Conflicts with kMemset if kept as is
     // phi
     kPhi = 0x1UL << 39,
     kBitItoF = 0x1UL << 40,
-    kBitFtoI = 0x1UL << 41,
+    kBitFtoI = 0x1UL << 41
   };
 
 protected:
@@ -769,12 +793,14 @@ public:
         return "Load";
       case kStore:
         return "Store";
-      case kGetElementPtr:
-        return "GetElementPtr";
+      case kLa:
+        return "La";
       case kMemset:
         return "Memset";
       case kPhi:
         return "Phi";
+      case kGetSubArray:
+        return "GetSubArray";
       default:
         return "Unknown";
     }
@@ -827,8 +853,9 @@ public:
   bool isAlloca() const { return kind == kAlloca; }
   bool isLoad() const { return kind == kLoad; }
   bool isStore() const { return kind == kStore; }
-  bool isGetElementPtr() const { return kind == kGetElementPtr; }
+  bool isLa() const { return kind == kLa; }
   bool isMemset() const { return kind == kMemset; }
+  bool isGetSubArray() const { return kind == kGetSubArray; }
   bool isCall() const { return kind == kCall; }
   bool isReturn() const { return kind == kReturn; }
   bool isDefine() const {
@@ -840,6 +867,26 @@ public:
 class Function;
 //! Function call.
 
+class LaInst : public Instruction {
+  friend class Function;
+  friend class IRBuilder;
+
+ protected:
+  explicit LaInst(Value *pointer, const std::vector<Value *> &indices = {}, BasicBlock *parent = nullptr,
+                  const std::string &name = "")
+      : Instruction(Kind::kLa, pointer->getType(), parent, name) {
+    assert(pointer);
+    addOperand(pointer);
+    addOperands(indices);
+  }
+
+ public:
+  unsigned getNumIndices() const { return getNumOperands() - 1; }  ///< 获取索引长度
+  Value* getPointer() const { return getOperand(0); }             ///< 获取目标变量的Value指针
+  auto getIndices() const { return make_range(std::next(operand_begin()), operand_end()); }  ///< 获取索引列表
+  Value* getIndex(unsigned index) const { return getOperand(index + 1); }  ///< 获取位置为index的索引分量
+};
+
 class PhiInst : public Instruction {
   friend class IRBuilder;
   friend class Function;
@@ -1087,7 +1134,7 @@ public:
 }; // class CondBrInst
 
 //! Allocate memory for stack variables, used for non-global variable declartion
-class AllocaInst : public Instruction {
+class AllocaInst : public Instruction , public LVal {
   friend class IRBuilder;
   friend class Function;
 protected:
@@ -1098,6 +1145,14 @@ protected:
   }
 
 public:
+  std::vector<Value *> getLValDims() const override {
+    std::vector<Value *> dims;
+    for (const auto &dim : getOperands()) {
+      dims.emplace_back(dim->getValue());
+    }
+    return dims;
+  }  ///< 获取作为左值的维度数组
+  unsigned getLValNumDims() const override { return getNumOperands(); }
   
   int getNumDims() const { return getNumOperands(); }
   auto getDims() const { return getOperands(); }
@@ -1106,40 +1161,37 @@ public:
 }; // class AllocaInst
 
 
-class GetElementPtrInst : public Instruction {
-  friend class IRBuilder; // 如果您有IRBuilder来创建指令，需要friend
+class GetSubArrayInst : public Instruction {
+  friend class IRBuilder;
+  friend class Function;
 
-protected:
-  // GEP的构造函数：
-  // resultType: GEP计算出的地址的类型 (通常是指向目标元素类型的指针)
-  // basePointer: 基指针 (第一个操作数)
-  // indices: 索引列表 (后续操作数)
-  GetElementPtrInst(Value *basePointer,
-                      const std::vector<Value *> &indices = {},
-                      BasicBlock *parent = nullptr, const std::string &name = "")
-      : Instruction(Kind::kGetElementPtr, basePointer->getType(), parent, name) {
-    assert(basePointer && "GEP base pointer cannot be null!");
-    // TODO : 安全检查
-    assert(basePointer->getType()->isPointer() );
-    addOperand(basePointer); // 第一个操作数是基指针
-    addOperands(indices);    // 随后的操作数是索引
+ public:
+  GetSubArrayInst(LVal *fatherArray, LVal *childArray, const std::vector<Value *> &indices,
+                  BasicBlock *parent = nullptr, const std::string &name = "")
+      : Instruction(Kind::kGetSubArray, Type::getVoidType(), parent, name) {
+    auto predicate = [childArray](const std::unique_ptr<LVal> &child) -> bool { return child.get() == childArray; };
+    if (std::find_if(fatherArray->childrenLVals.begin(), fatherArray->childrenLVals.end(), predicate) ==
+        fatherArray->childrenLVals.end()) {
+      fatherArray->childrenLVals.emplace_back(childArray);
+    }
+    childArray->fatherLVal = fatherArray;
+    childArray->defineInst = this;
+    auto fatherArrayValue = dynamic_cast<Value *>(fatherArray);
+    auto childArrayValue = dynamic_cast<Value *>(childArray);
+    assert(fatherArrayValue);
+    assert(childArrayValue);
+    addOperand(fatherArrayValue);
+    addOperand(childArrayValue);
+    addOperands(indices);
   }
 
-public:
-  Value* getBasePointer() const { return getOperand(0); }
-  unsigned getNumIndices() const { return getNumOperands() - 1; }
-  auto getIndices() const { return make_range(std::next(operand_begin()), operand_end());}
-  Value* getIndex(unsigned index) const {
-    assert(index < getNumIndices() && "Index out of bounds for GEP!");
-    return getOperand(index + 1);
-  }
-
-  // 静态工厂方法，用于创建GEP指令 (如果需要外部直接创建而非通过IRBuilder)
-  static GetElementPtrInst* create(Type *resultType, Value *basePointer,
-                                    const std::vector<Value *> &indices = {},
-                                    BasicBlock *parent = nullptr, const std::string &name = "") {
-    return new GetElementPtrInst(basePointer, indices, parent, name);
-  }
+ public:
+  Value* getFatherArray() const { return getOperand(0); }                              ///< 获取父数组
+  Value* getChildArray() const { return getOperand(1); }                               ///< 获取子数组
+  LVal* getFatherLVal() const { return dynamic_cast<LVal *>(getOperand(0)); }          ///< 获取父左值
+  LVal* getChildLVal() const { return dynamic_cast<LVal *>(getOperand(1)); }           ///< 获取子左值
+  auto getIndices() const { return make_range(std::next(operand_begin(), 2), operand_end()); }  ///< 获取索引
+  unsigned getNumIndices() const { return getNumOperands() - 2; }                       ///< 获取索引数量
 };
 
 //! Load a value from memory address specified by a pointer value
@@ -1163,7 +1215,22 @@ public:
     return make_range(std::next(operand_begin()), operand_end());
   }
   Value* getIndex(int index) const { return getOperand(index + 1); }
-  
+  std::list<Value *> getAncestorIndices() const {
+    std::list<Value *> indices;
+    for (const auto &index : getIndices()) {
+      indices.emplace_back(index->getValue());
+    }
+    auto curPointer = dynamic_cast<LVal *>(getPointer());
+    while (curPointer->getFatherLVal() != nullptr) {
+      auto inserter = std::next(indices.begin());
+      for (const auto &index : curPointer->getDefineInst()->getIndices()) {
+        indices.insert(inserter, index->getValue());
+      }
+      curPointer = curPointer->getFatherLVal();
+    }
+
+    return indices;
+  }  ///< 获取相对于祖先数组的索引列表
 }; // class LoadInst
 
 //! Store a value to memory address specified by a pointer value
@@ -1189,6 +1256,22 @@ public:
     return make_range(std::next(operand_begin(), 2), operand_end());
   }
   Value* getIndex(int index) const { return getOperand(index + 2); }
+  std::list<Value *> getAncestorIndices() const {
+    std::list<Value *> indices;
+    for (const auto &index : getIndices()) {
+      indices.emplace_back(index->getValue());
+    }
+    auto curPointer = dynamic_cast<LVal *>(getPointer());
+    while (curPointer->getFatherLVal() != nullptr) {
+      auto inserter = std::next(indices.begin());
+      for (const auto &index : curPointer->getDefineInst()->getIndices()) {
+        indices.insert(inserter, index->getValue());
+      }
+      curPointer = curPointer->getFatherLVal();
+    }
+
+    return indices;
+  }  ///< 获取相对于祖先数组的索引列表
 
 }; // class StoreInst
 
@@ -1290,7 +1373,7 @@ protected:
 };
 
 //! Global value declared at file scope
-class GlobalValue : public User {
+class GlobalValue : public User, public LVal {
   friend class Module;
 
 protected:
@@ -1324,6 +1407,16 @@ protected:
   }
 
 public:
+  unsigned getLValNumDims() const override { return numDims; }  ///< 获取作为左值的维度数量
+  std::vector<Value *> getLValDims() const override {
+    std::vector<Value *> dims;
+    for (const auto &dim : getOperands()) {
+      dims.emplace_back(dim->getValue());
+    }
+
+    return dims;
+  }  ///< 获取作为左值的维度列表
+
   unsigned getNumDims() const  { return numDims; }                     ///< 获取维度数量
   Value* getDim(unsigned index) const { return getOperand(index); }  ///< 获取位置为index的维度
   auto getDims() const { return getOperands(); }                              ///< 获取维度列表
@@ -1345,7 +1438,7 @@ public:
 }; // class GlobalValue
 
 
-class ConstantVariable : public User {
+class ConstantVariable : public User, public LVal {
   friend class Module;
 
  protected:
@@ -1364,6 +1457,15 @@ class ConstantVariable : public User {
   }
 
  public:
+  unsigned getLValNumDims() const override { return numDims; }  ///< 获取作为左值的维度数量
+  std::vector<Value *> getLValDims() const override {
+    std::vector<Value *> dims;
+    for (const auto &dim : getOperands()) {
+      dims.emplace_back(dim->getValue());
+    }
+
+    return dims;
+  }  ///< 获取作为左值的维度列表
   Value* getByIndex(unsigned index) const { return initValues.getValue(index); }  ///< 通过一维位置index获取值
   Value* getByIndices(const std::vector<Value *> &indices) const {
     int index = 0;

src/include/IRBuilder.h

@@ -280,6 +280,46 @@ class IRBuilder {
     block->getInstructions().emplace(position, inst);
     return inst;
   }  ///< 创建load指令
+  LaInst * createLaInst(Value *pointer, const std::vector<Value *> &indices = {}, const std::string &name = "") {
+    std::string newName;
+    if (name.empty()) {
+      std::stringstream ss;
+      ss << tmpIndex;
+      newName = ss.str();
+      tmpIndex++;
+    } else {
+      newName = name;
+    }
+
+    auto inst = new LaInst(pointer, indices, block, newName);
+    assert(inst);
+    block->getInstructions().emplace(position, inst);
+    return inst;
+  }  ///< 创建la指令
+  GetSubArrayInst * createGetSubArray(LVal *fatherArray, const std::vector<Value *> &indices, const std::string &name = "") {
+    assert(fatherArray->getLValNumDims() > indices.size());
+    std::vector<Value *> subDims;
+    auto dims = fatherArray->getLValDims();
+    auto iter = std::next(dims.begin(), indices.size());
+    while (iter != dims.end()) {
+      subDims.emplace_back(*iter);
+      iter++;
+    }
+
+    std::string childArrayName;
+    std::stringstream ss;
+    ss << "A"
+       << "%" << tmpIndex;
+    childArrayName = ss.str();
+    tmpIndex++;
+
+    auto fatherArrayValue = dynamic_cast<Value *>(fatherArray);
+    auto childArray = new AllocaInst(fatherArrayValue->getType(), subDims, block, childArrayName);
+    auto inst = new GetSubArrayInst(fatherArray, childArray, indices, block, childArrayName);
+    assert(inst);
+    block->getInstructions().emplace(position, inst);
+    return inst;
+  }  ///< 创建获取部分数组指令
   MemsetInst * createMemsetInst(Value *pointer, Value *begin, Value *size, Value *value, const std::string &name = "") {
     auto inst = new MemsetInst(pointer, begin, size, value, block, name);
     assert(inst);
@@ -300,24 +340,6 @@ class IRBuilder {
     block->getInstructions().emplace(block->begin(), inst);
     return inst;
   }  ///< 创建Phi指令
-  GetElementPtrInst* createGetElementPtrInst(Value *basePointer,
-                                 const std::vector<Value *> &indices = {},
-                                 const std::string &name = "") {
-    std::string newName;
-    if (name.empty()) {
-      std::stringstream ss;
-      ss << tmpIndex;
-      newName = ss.str();
-      tmpIndex++;
-    } else {
-      newName = name;
-    }
-
-    auto inst = new GetElementPtrInst(basePointer, indices, block, newName);
-    assert(inst);
-    block->getInstructions().emplace(position, inst);
-    return inst;
-  }
 };
 
 }  // namespace sysy

src/include/Mem2Reg.h

@@ -1,79 +1,59 @@
-// 假设 Mem2Reg.h 看起来像这样 (你需要根据实际情况调整)
-#ifndef SYSY_MEM2REG_H
-#define SYSY_MEM2REG_H
+#pragma once
 
-#include <vector>
+#include <list>
+#include <memory>
+#include <stack>
 #include <unordered_map>
 #include <unordered_set>
-#include <stack>
-#include <queue> // For computeIteratedDomFrontiers
-
-// Include your IR and analysis headers
 #include "IR.h"
 #include "IRBuilder.h"
 #include "SysYIRAnalyser.h"
-#include "SysYIROptUtils.h"
 
 namespace sysy {
-
+/**
+ * 实现静态单变量赋值核心类 mem2reg
+ */
 class Mem2Reg {
 private:
-  Module* pModule;
-  IRBuilder* pBuilder;
-  ControlFlowAnalysis* controlFlowAnalysis;
-  ActiveVarAnalysis* activeVarAnalysis;
-  DataFlowAnalysisUtils dataFlowAnalysisUtils; // If this is part of Mem2Reg or an external helper
+  Module *pModule;
+  IRBuilder *pBuilder;
+  ControlFlowAnalysis *controlFlowAnalysis;  // 控制流分析
+  ActiveVarAnalysis *activeVarAnalysis;  // 活跃变量分析
+  DataFlowAnalysisUtils dataFlowAnalysisUtils;
 
 public:
-  Mem2Reg(Module* module, IRBuilder* builder, ControlFlowAnalysis* cfa, ActiveVarAnalysis* ava)
-    : pModule(module), pBuilder(builder), controlFlowAnalysis(cfa), activeVarAnalysis(ava) {}
-    // Constructor initializes members
-  void run();
+  Mem2Reg(Module *pMoudle, IRBuilder *pBuilder, 
+          ControlFlowAnalysis *pCFA = nullptr, ActiveVarAnalysis *pAVA = nullptr) :
+    pModule(pMoudle), pBuilder(pBuilder), controlFlowAnalysis(pCFA), activeVarAnalysis(pAVA), dataFlowAnalysisUtils()
+   {}  // 初始化函数
 
-  // --- 新增的私有成员变量和方法，用于SSA转换上下文 ---
-  // 这是核心，用于存储 SSA 转换过程中的状态
-  std::vector<AllocaInst*> currentFunctionAllocas; // 当前函数中所有可提升的 alloca
-  // alloca -> set of BasicBlocks where it's defined (stored into)
-  std::unordered_map<AllocaInst*, std::unordered_set<BasicBlock*>> allocaDefsBlock; 
-  // alloca -> set of BasicBlocks where it's used (loaded from)
-  std::unordered_map<AllocaInst*, std::unordered_set<BasicBlock*>> allocaUsesBlock; 
+  void mem2regPipeline();  ///< mem2reg
 
-  // BasicBlock -> Map of (PhiInst, Original AllocaInst)
-  // 用于在 rename 阶段通过 phi 指令找到它代表的原始 alloca
-  std::unordered_map<BasicBlock*, std::unordered_map<PhiInst*, AllocaInst*>> phiMap;
-  std::vector<PhiInst*> allPhiInstructions; // 收集所有创建的 Phi 指令以便后续简化和清理
+private:
 
-  // --- 核心 SSA 转换辅助函数 ---
-  // 计算给定定义块集合的迭代支配边界
-  std::unordered_set<BasicBlock*> computeIteratedDomFrontiers(const std::unordered_set<BasicBlock*>& blocks);
+  // phi节点的插入需要计算IDF
+  std::unordered_set<BasicBlock *> computeIterDf(const std::unordered_set<BasicBlock *> &blocks);  ///< 计算定义块集合的迭代支配边界
 
-  // 分析一个 alloca 的所有 uses，填充 allocaDefsBlock 和 allocaUsesBlock
-  void allocaAnalysis(AllocaInst* alloca);
+  auto computeValue2Blocks() -> void;  ///< 计算value2block的映射(不包括数组和global)
 
-  // 判断一个 alloca 是否可以被提升为寄存器 (无地址逃逸，标量类型)
-  bool is_promoted(AllocaInst* alloca);
+  auto preOptimize1() -> void;  ///< llvm memtoreg预优化1: 删除不含load的alloc和store
+  auto preOptimize2() -> void;  ///< llvm memtoreg预优化2: 针对某个变量的Defblocks只有一个块的情况
+  auto preOptimize3() -> void;  ///< llvm memtoreg预优化3: 针对某个变量的所有读写都在同一个块中的情况
 
-  // 在迭代支配边界处插入 Phi 指令
-  void insertPhiNodes(Function* func);
+  auto insertPhi() -> void;  ///< 为所有变量的迭代支配边界插入phi结点
 
-  // 递归地重命名基本块中的变量并填充 Phi 指令
-  // 这里的 `count` 和 `stacks` 是临时的，用于 DFS 过程中传递状态
-  void renameBlock(BasicBlock* block, 
-                      std::unordered_map<AllocaInst*, Value*>& currentIncomings,
-                      std::unordered_set<BasicBlock*>& visitedBlocks); // 修改为传递 map 和 set
+  auto rename(BasicBlock *block, std::unordered_map<Value *, int> &count,
+              std::unordered_map<Value *, std::stack<Instruction *>> &stacks) -> void;  ///< 单个块的重命名
+  auto renameAll() -> void;                                                             ///< 重命名所有块
 
-  // 简化冗余的 Phi 指令 (当所有输入都相同时)
-  void simplifyphi(PhiInst* phi);
-
-  // 获取前驱块在后继块前驱列表中的索引，用于 Phi 指令入边
-  int getPredIndex(BasicBlock* pred, BasicBlock* succ);
-
-  // --- Mem2Reg 的主要工作流函数 ---
-  // 对单个函数执行内存到寄存器的提升
-  bool promoteMemoryToRegisters(Function* func);
+  // private helper function.
+private:
+  auto getPredIndex(BasicBlock *n, BasicBlock *s) -> int;  ///< 获取前驱索引
+  auto cascade(Instruction *instr, bool &changed, Function *func, BasicBlock *block,
+               std::list<std::unique_ptr<Instruction>> &instrs) -> void;  ///< 消除级联关系
+  auto isGlobal(Value *val) -> bool;                                      ///< 判断是否是全局变量
+  auto isArr(Value *val) -> bool;                                         ///< 判断是否是数组
+  auto usedelete(Instruction *instr) -> void;                             ///< 删除指令相关的value-use-user关系
 
 };
-
-} // namespace sysy
-
-#endif // SYSY_MEM2REG_H
+}  // namespace sysy

src/include/RISCv64AsmPrinter.h

@@ -4,20 +4,23 @@
 #include "RISCv64LLIR.h"
 #include <iostream>
 
+extern int DEBUG;
+extern int DEEPDEBUG;
+
 namespace sysy {
 
 class RISCv64AsmPrinter {
 public:
     RISCv64AsmPrinter(MachineFunction* mfunc);
     // 主入口
-    void run(std::ostream& os);
+    void run(std::ostream& os, bool debug = false);
 
 private:
     // 打印各个部分
     void printPrologue();
     void printEpilogue();
-    void printBasicBlock(MachineBasicBlock* mbb);
-    void printInstruction(MachineInstr* instr);
+    void printBasicBlock(MachineBasicBlock* mbb, bool debug = false);
+    void printInstruction(MachineInstr* instr, bool debug = false);
     
     // 辅助函数
     std::string regToString(PhysicalReg reg);

src/include/RISCv64Backend.h

@@ -4,6 +4,9 @@
 #include "IR.h"
 #include <string>
 
+extern int DEBUG;
+extern int DEEPDEBUG;
+
 namespace sysy {
 
 // RISCv64CodeGen 现在是一个高层驱动器

src/include/RISCv64ISel.h

@@ -3,6 +3,9 @@
 
 #include "RISCv64LLIR.h"
 
+extern int DEBUG;
+extern int DEEPDEBUG;
+
 namespace sysy {
 
 class RISCv64ISel {
@@ -31,6 +34,8 @@ private:
     DAGNode* get_operand_node(Value* val_ir, std::map<Value*, DAGNode*>&, std::vector<std::unique_ptr<DAGNode>>&);
     DAGNode* create_node(int kind, Value* val, std::map<Value*, DAGNode*>&, std::vector<std::unique_ptr<DAGNode>>&);
     
+    void print_dag(const std::vector<std::unique_ptr<DAGNode>>& dag, const std::string& bb_name);
+    
     // 状态
     Function* F; // 当前处理的高层IR函数
     std::unique_ptr<MachineFunction> MFunc; // 正在构建的底层LLIR函数

src/include/RISCv64LLIR.h

@@ -46,6 +46,7 @@ enum class RVOpcodes {
     // 新增伪指令，用于解耦栈帧处理
     FRAME_LOAD,  // 从栈帧加载 (AllocaInst)
     FRAME_STORE, // 保存到栈帧 (AllocaInst)
+    FRAME_ADDR,  // [新] 获取栈帧变量的地址
 };
 
 class MachineOperand;

src/include/Reg2Mem.h

@@ -2,7 +2,6 @@
 
 #include "IR.h"
 #include "IRBuilder.h"
-#include "SysYIROptUtils.h"
 
 namespace sysy {
 /**
@@ -17,6 +16,8 @@ public:
   Reg2Mem(Module *pMoudle, IRBuilder *pBuilder) : pModule(pMoudle), pBuilder(pBuilder) {}
 
   void DeletePhiInst();
+  // 删除UD关系, 因为删除了phi指令会修改ud关系
+  void usedelete(Instruction *instr);
 };
 
 } // namespace sysy

src/include/SCCP.h

@@ -1,196 +0,0 @@
-#pragma once
-
-#include "IR.h"
-
-namespace sysy {
-
-// 稀疏条件常量传播类
-// Sparse Conditional Constant Propagation
-/*
-伪代码
-function SCCP_Optimization(Module):
-    for each Function in Module:
-        changed = true
-        while changed:
-            changed = false
-            // 阶段1: 常量传播与折叠
-            changed |= PropagateConstants(Function)
-            // 阶段2: 控制流简化
-            changed |= SimplifyControlFlow(Function)
-        end while
-    end for
-
-function PropagateConstants(Function):
-    // 初始化
-    executableBlocks = {entryBlock}
-    valueState = map<Value, State> // 值->状态映射
-    instWorkList = Queue()
-    edgeWorkList = Queue()
-    
-    // 初始化工作列表
-    for each inst in entryBlock:
-        instWorkList.push(inst)
-    
-    // 迭代处理
-    while !instWorkList.empty() || !edgeWorkList.empty():
-        // 处理指令工作列表
-        while !instWorkList.empty():
-            inst = instWorkList.pop()
-            // 如果指令是可执行基本块中的
-            if executableBlocks.contains(inst.parent):
-                ProcessInstruction(inst)
-        
-        // 处理边工作列表
-        while !edgeWorkList.empty():
-            edge = edgeWorkList.pop()
-            ProcessEdge(edge)
-    
-    // 应用常量替换
-    for each inst in Function:
-        if valueState[inst] == CONSTANT:
-            ReplaceWithConstant(inst, valueState[inst].constant)
-            changed = true
-    
-    return changed
-
-function ProcessInstruction(Instruction inst):
-    switch inst.type:
-        //二元操作
-        case BINARY_OP:
-            lhs = GetValueState(inst.operands[0])
-            rhs = GetValueState(inst.operands[1])
-            if lhs == CONSTANT && rhs == CONSTANT:
-                newState = ComputeConstant(inst.op, lhs.value, rhs.value)
-                UpdateState(inst, newState)
-            else if lhs == BOTTOM || rhs == BOTTOM:
-                UpdateState(inst, BOTTOM)
-        //phi
-        case PHI:
-            mergedState = ⊤
-            for each incoming in inst.incomings:
-            // 检查每个输入的状态
-                if executableBlocks.contains(incoming.block):
-                    incomingState = GetValueState(incoming.value)
-                    mergedState = Meet(mergedState, incomingState)
-            UpdateState(inst, mergedState)
-        // 条件分支
-        case COND_BRANCH:
-            cond = GetValueState(inst.condition)
-            if cond == CONSTANT:
-                // 判断条件分支
-                if cond.value == true:
-                    AddEdgeToWorkList(inst.parent, inst.trueTarget)
-                else:
-                    AddEdgeToWorkList(inst.parent, inst.falseTarget)
-            else if cond == BOTTOM:
-                AddEdgeToWorkList(inst.parent, inst.trueTarget)
-                AddEdgeToWorkList(inst.parent, inst.falseTarget)
-        
-        case UNCOND_BRANCH:
-            AddEdgeToWorkList(inst.parent, inst.target)
-        
-        // 其他指令处理...
-
-function ProcessEdge(Edge edge):
-    fromBB, toBB = edge
-    if !executableBlocks.contains(toBB):
-        executableBlocks.add(toBB)
-        for each inst in toBB:
-            if inst is PHI:
-                instWorkList.push(inst)
-            else:
-                instWorkList.push(inst) // 非PHI指令
-    
-    // 更新PHI节点的输入
-    for each phi in toBB.phis:
-        instWorkList.push(phi)
-
-function SimplifyControlFlow(Function):
-    changed = false
-    // 标记可达基本块
-    ReachableBBs = FindReachableBlocks(Function.entry)
-    
-    // 删除不可达块
-    for each bb in Function.blocks:
-        if !ReachableBBs.contains(bb):
-            RemoveDeadBlock(bb)
-            changed = true
-    
-    // 简化条件分支
-    for each bb in Function.blocks:
-        terminator = bb.terminator
-        if terminator is COND_BRANCH:
-            cond = GetValueState(terminator.condition)
-            if cond == CONSTANT:
-                SimplifyBranch(terminator, cond.value)
-                changed = true
-    
-    return changed
-
-function RemoveDeadBlock(BasicBlock bb):
-    // 1. 更新前驱块的分支指令
-    for each pred in bb.predecessors:
-        UpdateTerminator(pred, bb)
-    
-    // 2. 更新后继块的PHI节点
-    for each succ in bb.successors:
-        RemovePhiIncoming(succ, bb)
-    
-    // 3. 删除块内所有指令
-    for each inst in bb.instructions:
-        inst.remove()
-    
-    // 4. 从函数中移除基本块
-    Function.removeBlock(bb)
-
-function Meet(State a, State b):
-    if a == ⊤: return b
-    if b == ⊤: return a
-    if a == ⊥ || b == ⊥: return ⊥
-    if a.value == b.value: return a
-    return ⊥
-
-function UpdateState(Value v, State newState):
-    oldState = valueState.get(v, ⊤)
-    if newState != oldState:
-        valueState[v] = newState
-        for each user in v.users:
-            if user is Instruction:
-                instWorkList.push(user)
-
-*/
-
-enum class LatticeValue {
-  Top,       // ⊤ (Unknown)
-  Constant,  // c (Constant)
-  Bottom     // ⊥ (Undefined / Varying)
-};
-// LatticeValue: 用于表示值的状态，Top表示未知，Constant表示常量，Bottom表示未定义或变化的值。
-// 这里的LatticeValue用于跟踪每个SSA值（变量、指令结果）的状态，
-// 以便在SCCP过程中进行常量传播和控制流简化。
-
-//TODO: 下列数据结构考虑集成到类中，避免重命名问题
-static std::set<Instruction *> Worklist;
-static std::unordered_set<BasicBlock*> Executable_Blocks;
-static std::queue<std::pair<BasicBlock *, BasicBlock *> > Executable_Edges;
-static std::map<Value*, LatticeValue> valueState;
-
-class SCCP {
-private:
-  Module *pModule;
-
-public:
-  SCCP(Module *pMoudle) : pModule(pMoudle) {}
-
-  void run();
-  bool PropagateConstants(Function *function);
-  bool SimplifyControlFlow(Function *function);
-  void ProcessInstruction(Instruction *inst);
-  void ProcessEdge(const std::pair<BasicBlock *, BasicBlock *> &edge);
-  void RemoveDeadBlock(BasicBlock *bb);
-  void UpdateState(Value *v, LatticeValue newState);
-  LatticeValue Meet(LatticeValue a, LatticeValue b);
-  LatticeValue GetValueState(Value *v);
-};
-
-}  // namespace sysy

src/include/SysYFormatter.h

@@ -0,0 +1,340 @@
+#pragma once
+
+#include "SysYBaseVisitor.h"
+#include "SysYParser.h"
+#include <ostream>
+
+namespace sysy {
+
+class SysYFormatter : public SysYBaseVisitor {
+protected:
+  std::ostream &os;
+  int indent = 0;
+
+public:
+  SysYFormatter(std::ostream &os) : os(os), indent(0) {}
+
+protected:
+  struct Indentor {
+    static constexpr int TabSize = 2;
+    int &indent;
+    Indentor(int &indent) : indent(indent) { indent += TabSize; }
+    ~Indentor() { indent -= TabSize; }
+  };
+  std::ostream &space() { return os << std::string(indent, ' '); }
+  template <typename T>
+  std::ostream &interleave(const T &container, const std::string sep = ", ") {
+    auto b = container.begin(), e = container.end();
+    (*b)->accept(this);
+    for (b = std::next(b); b != e; b = std::next(b)) {
+      os << sep;
+      (*b)->accept(this);
+    }
+    return os;
+  }
+
+public:
+  //   virtual std::any visitModule(SysYParser::ModuleContext *ctx) override {
+  //     return visitChildren(ctx);
+  //   }
+
+  virtual std::any visitBtype(SysYParser::BtypeContext *ctx) override {
+    os << ctx->getText();
+    return 0;
+  }
+
+  virtual std::any visitDecl(SysYParser::DeclContext *ctx) override {
+    space();
+    if (ctx->CONST())
+      os << ctx->CONST()->getText() << ' ';
+    ctx->btype()->accept(this);
+    os << ' ';
+    interleave(ctx->varDef(), ", ") << ';' << '\n';
+    return 0;
+  }
+
+  virtual std::any visitVarDef(SysYParser::VarDefContext *ctx) override {
+    ctx->lValue()->accept(this);
+    if (ctx->initValue()) {
+      os << ' ' << '=' << ' ';
+      ctx->initValue()->accept(this);
+    }
+    return 0;
+  }
+
+  virtual std::any visitInitValue(SysYParser::InitValueContext *ctx) override {
+    if (not ctx->exp()) {
+      os << '{';
+      auto values = ctx->initValue();
+      if (values.size())
+        interleave(values, ", ");
+      os << '}';
+    }
+    return 0;
+  }
+
+  virtual std::any visitFunc(SysYParser::FuncContext *ctx) override {
+    ctx->funcType()->accept(this);
+    os << ' ' << ctx->ID()->getText() << '(';
+    if (ctx->funcFParams())
+      ctx->funcFParams()->accept(this);
+    os << ')' << ' ';
+    ctx->blockStmt()->accept(this);
+    os << '\n';
+    return 0;
+  }
+
+  virtual std::any visitFuncType(SysYParser::FuncTypeContext *ctx) override {
+    os << ctx->getText();
+    return 0;
+  }
+
+  virtual std::any
+  visitFuncFParams(SysYParser::FuncFParamsContext *ctx) override {
+    interleave(ctx->funcFParam(), ", ");
+    return 0;
+  }
+
+  virtual std::any
+  visitFuncFParam(SysYParser::FuncFParamContext *ctx) override {
+    ctx->btype()->accept(this);
+    os << ' ' << ctx->ID()->getText();
+    if (not ctx->LBRACKET().empty()) {
+      os << '[';
+      auto exp = ctx->exp();
+      if (not exp.empty()) {
+        os << '[';
+        interleave(exp, "][") << ']';
+      }
+    }
+    return 0;
+  }
+
+  virtual std::any visitBlockStmt(SysYParser::BlockStmtContext *ctx) override {
+    os << '{' << '\n';
+    {
+      Indentor indentor(indent);
+      auto items = ctx->blockItem();
+      if (not items.empty())
+        interleave(items, "");
+    }
+    space() << ctx->RBRACE()->getText() << '\n';
+    return 0;
+  }
+
+  //   virtual std::any visitBlockItem(SysYParser::BlockItemContext *ctx)
+  //   override {
+  //     return visitChildren(ctx);
+  //   }
+
+  //   virtual std::any visitStmt(SysYParser::StmtContext *ctx) override {
+  //     return visitChildren(ctx);
+  //   }
+
+  virtual std::any
+  visitAssignStmt(SysYParser::AssignStmtContext *ctx) override {
+    space();
+    ctx->lValue()->accept(this);
+    os << " = ";
+    ctx->exp()->accept(this);
+    os << ';' << '\n';
+    return 0;
+  }
+
+  virtual std::any visitExpStmt(SysYParser::ExpStmtContext *ctx) override {
+    space();
+    ctx->exp()->accept(this);
+    os << ';' << '\n';
+    return 0;
+  }
+
+  void wrapBlock(SysYParser::StmtContext *stmt) {
+    bool isBlock = stmt->blockStmt();
+    if (isBlock) {
+      stmt->accept(this);
+    } else {
+      os << "{\n";
+      {
+        Indentor indentor(indent);
+        stmt->accept(this);
+      }
+      space() << "}\n";
+    }
+  };
+  virtual std::any visitIfStmt(SysYParser::IfStmtContext *ctx) override {
+    space();
+    os << ctx->IF()->getText() << " (";
+    ctx->exp()->accept(this);
+    os << ") ";
+    auto stmt = ctx->stmt();
+    auto ifStmt = stmt[0];
+    wrapBlock(ifStmt);
+    if (stmt.size() == 2) {
+      auto elseStmt = stmt[1];
+      wrapBlock(elseStmt);
+    }
+    return 0;
+  }
+
+  virtual std::any visitWhileStmt(SysYParser::WhileStmtContext *ctx) override {
+    space();
+    os << ctx->WHILE()->getText() << " (";
+    ctx->exp()->accept(this);
+    os << ") ";
+    wrapBlock(ctx->stmt());
+    return 0;
+  }
+
+  virtual std::any visitBreakStmt(SysYParser::BreakStmtContext *ctx) override {
+    space() << ctx->BREAK()->getText() << ';' << '\n';
+    return 0;
+  }
+
+  virtual std::any
+  visitContinueStmt(SysYParser::ContinueStmtContext *ctx) override {
+    space() << ctx->CONTINUE()->getText() << ';' << '\n';
+    return 0;
+  }
+
+  virtual std::any
+  visitReturnStmt(SysYParser::ReturnStmtContext *ctx) override {
+    space() << ctx->RETURN()->getText();
+    if (ctx->exp()) {
+      os << ' ';
+      ctx->exp()->accept(this);
+    }
+    os << ';' << '\n';
+    return 0;
+  }
+
+  //   virtual std::any visitEmptyStmt(SysYParser::EmptyStmtContext *ctx)
+  //   override {
+  //     return visitChildren(ctx);
+  //   }
+
+  virtual std::any
+  visitRelationExp(SysYParser::RelationExpContext *ctx) override {
+    auto lhs = ctx->exp(0);
+    auto rhs = ctx->exp(1);
+    std::string op =
+        ctx->LT() ? "<" : (ctx->LE() ? "<=" : (ctx->GT() ? ">" : ">="));
+    lhs->accept(this);
+    os << ' ' << op << ' ';
+    rhs->accept(this);
+    return 0;
+  }
+
+  virtual std::any
+  visitMultiplicativeExp(SysYParser::MultiplicativeExpContext *ctx) override {
+    auto lhs = ctx->exp(0);
+    auto rhs = ctx->exp(1);
+    std::string op = ctx->MUL() ? "*" : (ctx->DIV() ? "/" : "%");
+    lhs->accept(this);
+    os << ' ' << op << ' ';
+    rhs->accept(this);
+    return 0;
+  }
+
+  //   virtual std::any visitLValueExp(SysYParser::LValueExpContext *ctx)
+  //   override {
+  //     return visitChildren(ctx);
+  //   }
+
+  //   virtual std::any visitNumberExp(SysYParser::NumberExpContext *ctx)
+  //   override {
+  //     return visitChildren(ctx);
+  //   }
+
+  virtual std::any visitAndExp(SysYParser::AndExpContext *ctx) override {
+    ctx->exp(0)->accept(this);
+    os << " && ";
+    ctx->exp(1)->accept(this);
+    return 0;
+  }
+
+  virtual std::any visitUnaryExp(SysYParser::UnaryExpContext *ctx) override {
+    std::string op = ctx->ADD() ? "+" : (ctx->SUB() ? "-" : "!");
+    os << op;
+    ctx->exp()->accept(this);
+    return 0;
+  }
+
+  virtual std::any visitParenExp(SysYParser::ParenExpContext *ctx) override {
+    os << '(';
+    ctx->exp()->accept(this);
+    os << ')';
+    return 0;
+  }
+
+  virtual std::any visitStringExp(SysYParser::StringExpContext *ctx) override {
+    return visitChildren(ctx);
+  }
+
+  virtual std::any visitOrExp(SysYParser::OrExpContext *ctx) override {
+    ctx->exp(0)->accept(this);
+    os << " || ";
+    ctx->exp(1)->accept(this);
+    return 0;
+  }
+
+  //   virtual std::any visitCallExp(SysYParser::CallExpContext *ctx) override {
+  //     return visitChildren(ctx);
+  //   }
+
+  virtual std::any
+  visitAdditiveExp(SysYParser::AdditiveExpContext *ctx) override {
+    auto lhs = ctx->exp(0);
+    auto rhs = ctx->exp(1);
+    std::string op = ctx->ADD() ? "+" : "-";
+    lhs->accept(this);
+    os << ' ' << op << ' ';
+    rhs->accept(this);
+    return 0;
+  }
+
+  virtual std::any visitEqualExp(SysYParser::EqualExpContext *ctx) override {
+    auto lhs = ctx->exp(0);
+    auto rhs = ctx->exp(1);
+    std::string op = ctx->EQ() ? "==" : "!=";
+    lhs->accept(this);
+    os << ' ' << op << ' ';
+    rhs->accept(this);
+    return 0;
+  }
+
+  virtual std::any visitCall(SysYParser::CallContext *ctx) override {
+    os << ctx->ID()->getText() << '(';
+    if (ctx->funcRParams())
+      ctx->funcRParams()->accept(this);
+    os << ')';
+    return 0;
+  }
+
+  virtual std::any visitLValue(SysYParser::LValueContext *ctx) override {
+    os << ctx->ID()->getText();
+    auto exp = ctx->exp();
+    if (not exp.empty()) {
+      os << '[';
+      interleave(exp, "][") << ']';
+    }
+    return 0;
+  }
+
+  virtual std::any visitNumber(SysYParser::NumberContext *ctx) override {
+    os << ctx->getText();
+    return 0;
+  }
+
+  virtual std::any visitString(SysYParser::StringContext *ctx) override {
+    os << ctx->getText();
+    return 0;
+  }
+
+  virtual std::any
+  visitFuncRParams(SysYParser::FuncRParamsContext *ctx) override {
+    interleave(ctx->exp(), ", ");
+    return 0;
+  }
+};
+
+} // namespace sysy

src/include/SysYIRCFGOpt.h

@@ -1,60 +0,0 @@
-#pragma once
-
-#include "IR.h"
-#include "IRBuilder.h"
-
-namespace sysy {
-
-// 优化前对SysY IR的预处理，也可以视作部分CFG优化
-// 主要包括删除无用指令、合并基本块、删除空块等
-// 这些操作可以在SysY IR生成时就完成，但为了简化IR生成过程，
-// 这里将其放在SysY IR生成后进行预处理
-// 同时兼容phi节点的处理，可以再mem2reg后再次调用优化
-
-//TODO: 可增加的CFG优化和方法
-// - 检查基本块跳转关系正确性
-// - 简化条件分支（Branch Simplification），如条件恒真/恒假转为直接跳转
-// - 合并连续的跳转指令（Jump Threading）在合并不可达块中似乎已经实现了
-// - 基本块重排序（Block Reordering），提升局部性
-
-class SysYCFGOpt {
- private:
-  Module *pModule;
-  IRBuilder *pBuilder;
-
- public:
-  SysYCFGOpt(Module *pMoudle, IRBuilder *pBuilder) : pModule(pMoudle), pBuilder(pBuilder) {}
-
-  void SysYOptimizateAfterIR(){
-    
-    auto &functions = pModule->getFunctions();
-    for (auto &function : functions) {
-      bool changed = false;
-      while(changed){
-        changed = false;
-        changed |= SysYCondBr2Br(function.second.get(), pBuilder);
-        // 删除br后面的无用指令
-        changed |= SysYDelInstAfterBr(function.second.get());
-        // 合并空基本块
-        changed |= SysYBlockMerge(function.second.get());
-        // 删除无前驱块
-        changed |= SysYDelNoPreBLock(function.second.get());
-        // 删除空块
-        changed |= SysYDelEmptyBlock(function.second.get(), pBuilder);
-        // 添加return指令
-        changed |= SysYAddReturn(function.second.get(), pBuilder);
-      }
-    }
-  }
-
-public:
-  static bool SysYDelInstAfterBr(Function *func); // 删除br后面的指令
-  static bool SysYDelEmptyBlock(Function *func, IRBuilder* pBuilder); // 空块删除
-  static bool SysYDelNoPreBLock(Function *func); // 删除无前驱块（不可达块）
-  static bool SysYBlockMerge(Function *func);    // 合并基本块(主要针对嵌套if while的exit块，
-                            // 也可以修改IR生成实现回填机制
-  static bool SysYAddReturn(Function *func, IRBuilder* pBuilder);     // 添加return指令(主要针对Void函数)
-  static bool SysYCondBr2Br(Function *func, IRBuilder* pBuilder);  // 条件分支(已知cond的值)转换为无条件分支
-};
-
-}  // namespace sysy

src/include/SysYIRGenerator.h

@@ -62,11 +62,12 @@ private:
 public:
   SysYIRGenerator() = default;
 
+  bool HasReturnInst;
+
 public:
   Module *get() const { return module.get(); }
   IRBuilder *getBuilder(){ return &builder; }
 public:
-
   std::any visitCompUnit(SysYParser::CompUnitContext *ctx) override;
 
   std::any visitGlobalConstDecl(SysYParser::GlobalConstDeclContext *ctx) override;
@@ -133,11 +134,6 @@ public:
   
   // std::any visitConstExp(SysYParser::ConstExpContext *ctx) override;
 
-public:
-  // 获取GEP指令的地址
-  Value* getGEPAddressInst(Value* basePointer, const std::vector<Value*>& indices);
-  // 构建数组类型
-  Type* buildArrayType(Type* baseType, const std::vector<Value*>& dims);
 
 }; // class SysYIRGenerator
 

src/include/SysYIROptPre.h

@@ -0,0 +1,37 @@
+#pragma once
+
+#include "IR.h"
+#include "IRBuilder.h"
+
+namespace sysy {
+
+// 优化前对SysY IR的预处理，也可以视作部分CFG优化
+// 主要包括删除无用指令、合并基本块、删除空块等
+// 这些操作可以在SysY IR生成时就完成，但为了简化IR生成过程，
+// 这里将其放在SysY IR生成后进行预处理
+// 同时兼容phi节点的处理，可以再mem2reg后再次调用优化
+class SysYOptPre {
+ private:
+  Module *pModule;
+  IRBuilder *pBuilder;
+
+ public:
+  SysYOptPre(Module *pMoudle, IRBuilder *pBuilder) : pModule(pMoudle), pBuilder(pBuilder) {}
+
+  void SysYOptimizateAfterIR(){
+    SysYDelInstAfterBr();
+    SysYBlockMerge();
+    SysYDelNoPreBLock();
+    SysYDelEmptyBlock();
+    SysYAddReturn();
+  }
+  void SysYDelInstAfterBr(); // 删除br后面的指令
+  void SysYDelEmptyBlock(); // 空块删除
+  void SysYDelNoPreBLock(); // 删除无前驱块
+  void SysYBlockMerge();    // 合并基本块(主要针对嵌套if while的exit块，
+                            // 也可以修改IR生成实现回填机制
+  void SysYAddReturn();     // 添加return指令(主要针对Void函数)
+  void usedelete(Instruction *instr); // use删除
+};
+
+}  // namespace sysy

src/include/SysYIROptUtils.h

@@ -1,33 +0,0 @@
-#pragma once
-
-#include "IR.h"
-
-namespace sysy {
-
-// 优化工具类，包含一些通用的优化方法
-// 这些方法可以在不同的优化 pass 中复用
-// 例如：删除use关系,判断是否是全局变量等
-class SysYIROptUtils{
-
-public:
-  // 删除use关系
-  static void usedelete(Instruction *instr) {
-    for (auto &use : instr->getOperands()) {
-      Value* val = use->getValue();
-      val->removeUse(use);
-    }
-  }
-
-  // 判断是否是全局变量
-  static bool isGlobal(Value *val) {
-    auto gval = dynamic_cast<GlobalValue *>(val);
-    return gval != nullptr;
-  }
-  // 判断是否是数组
-  static bool isArr(Value *val) {
-    auto aval = dynamic_cast<AllocaInst *>(val);
-  return aval != nullptr && aval->getNumDims() != 0;
-  }
-};
-
-}// namespace sysy

src/include/SysYIRPass.h

@@ -1,59 +0,0 @@
-#pragma once
-
-#include "IR.h"
-
-namespace sysy {
-
-// 前置声明
-class FunctionPass;
-class ModulePass;
-class AnalysisPass;
-class PassManager;
-
-// 抽象基类 Pass
-class Pass {
-public:
-    enum PassKind {
-        PK_Function,
-        PK_Module,
-        PK_Analysis
-    };
-
-    Pass(PassKind kind, const std::string& name) : Kind(kind), Name(name) {}
-    virtual ~Pass() = default;
-
-    PassKind getPassKind() const { return Kind; }
-    const std::string& getPassName() const { return Name; }
-
-    // 每个Pass需要实现此方法来执行其逻辑
-    // 具体的run方法将根据Pass类型在FunctionPass和ModulePass中定义
-protected:
-    PassKind Kind;
-    std::string Name;
-};
-
-// 针对函数的优化遍
-class FunctionPass : public Pass {
-public:
-    FunctionPass(const std::string& name) : Pass(PK_Function, name) {}
-    // 真正的优化逻辑将在此方法中实现
-    virtual bool runOnFunction(Function& F) = 0;
-};
-
-// 针对模块的优化遍
-class ModulePass : public Pass {
-public:
-    ModulePass(const std::string& name) : Pass(PK_Module, name) {}
-    // 真正的优化逻辑将在此方法中实现
-    virtual bool runOnModule(Module& M) = 0;
-};
-
-// 分析遍
-class AnalysisPass : public Pass {
-public:
-    AnalysisPass(const std::string& name) : Pass(PK_Analysis, name) {}
-    // 分析遍通常需要一个模块或函数作为输入，并计算出分析结果
-    // 具体分析结果的存储和访问方式需要设计
-};
-
-} // namespace sysy

src/include/SysYIRPassManager.h

@@ -1,58 +0,0 @@
-// PassManager.h
-#pragma once
-
-#include <vector>
-#include <memory>
-#include <typeindex> // For std::type_index
-#include <unordered_map>
-#include "SysYIRPass.h"
-#include "IR.h" // 假设你的IR.h定义了Module, Function等
-
-namespace sysy {
-
-class PassManager {
-public:
-	PassManager() = default;
-
-	// 添加一个FunctionPass
-	void addPass(std::unique_ptr<FunctionPass> pass) {
-		functionPasses.push_back(std::move(pass));
-	}
-
-	// 添加一个ModulePass
-	void addPass(std::unique_ptr<ModulePass> pass) {
-		modulePasses.push_back(std::move(pass));
-	}
-
-	// 添加一个AnalysisPass
-	template<typename T, typename... Args>
-	T* addAnalysisPass(Args&&... args) {
-		static_assert(std::is_base_of<AnalysisPass, T>::value, "T must derive from AnalysisPass");
-		auto analysis = std::make_unique<T>(std::forward<Args>(args)...);
-		T* rawPtr = analysis.get();
-		analysisPasses[std::type_index(typeid(T))] = std::move(analysis);
-		return rawPtr;
-	}
-
-	// 获取分析结果（用于其他Pass访问）
-	template<typename T>
-	T* getAnalysis() {
-		static_assert(std::is_base_of<AnalysisPass, T>::value, "T must derive from AnalysisPass");
-		auto it = analysisPasses.find(std::type_index(typeid(T)));
-		if (it != analysisPasses.end()) {
-			return static_cast<T*>(it->second.get());
-		}
-		return nullptr; // 或者抛出异常
-	}
-
-	// 运行所有注册的遍
-	void run(Module& M);
-
-private:
-	std::vector<std::unique_ptr<FunctionPass>> functionPasses;
-	std::vector<std::unique_ptr<ModulePass>> modulePasses;
-	std::unordered_map<std::type_index, std::unique_ptr<AnalysisPass>> analysisPasses;
-	// 未来可以添加AnalysisPass的缓存机制
-};
-
-} // namespace sysy

src/sysyc.cpp

@@ -13,9 +13,9 @@ using namespace antlr4;
 
 #include "SysYIRGenerator.h"
 #include "SysYIRPrinter.h"
-#include "SysYIRCFGOpt.h"
+#include "SysYIROptPre.h"
 #include "RISCv64Backend.h"
-// #include "SysYIRAnalyser.h"
+#include "SysYIRAnalyser.h"
 // #include "DeadCodeElimination.h"
 #include "AddressCalculationExpansion.h"
 // #include "Mem2Reg.h"
@@ -132,13 +132,13 @@ int main(int argc, char **argv) {
     DEBUG = 1; // 这里可能需要更精细地控制 DEBUG 的开启时机和范围
   }
   // 默认优化 pass (在所有优化级别都会执行)
-  SysYCFGOpt cfgopt(moduleIR, builder);
-  cfgopt.SysYOptimizateAfterIR(); 
+  SysYOptPre optPre(moduleIR, builder);
+  optPre.SysYOptimizateAfterIR(); 
 
-  // ControlFlowAnalysis cfa(moduleIR);
-  // cfa.init();
-  // ActiveVarAnalysis ava;
-  // ava.init(moduleIR);
+  ControlFlowAnalysis cfa(moduleIR);
+  cfa.init();
+  ActiveVarAnalysis ava;
+  ava.init(moduleIR);
 
   if (DEBUG) {
     cout << "=== After CFA & AVA (Default) ===\n";
@@ -218,7 +218,7 @@ int main(int argc, char **argv) {
   // 设置 DEBUG 模式（如果指定了 'asmd'）
   if (argStopAfter == "asmd") {
     DEBUG = 1;
-    // DEEPDEBUG = 1;
+    DEEPDEBUG = 1;
   }
   sysy::RISCv64CodeGen codegen(moduleIR); // 传入优化后的 moduleIR
   string asmCode = codegen.code_gen();

test_script/runit-riscv64-single.sh

@@ -0,0 +1,227 @@
+#!/bin/bash
+
+# runit-riscv64-single.sh - 用于在 RISC-V 虚拟机内部测试单个或少量 .s 文件的脚本
+# 模仿 runit-riscv64.sh 的功能，但以具体文件路径作为输入。
+
+# --- 配置区 ---
+# 假设此脚本位于项目根目录 (例如 /home/ubuntu/debug)
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" &>/dev/null && pwd)"
+LIB_DIR="${SCRIPT_DIR}/lib"
+TMP_DIR="${SCRIPT_DIR}/tmp" # 临时可执行文件将存放在这里
+TESTDATA_DIR="${SCRIPT_DIR}/testdata" # 用于查找 .in/.out 文件
+
+# 定义编译器
+GCC_NATIVE="gcc" # VM 内部的原生 gcc
+
+# --- 初始化变量 ---
+GCC_TIMEOUT=10        # gcc 编译超时 (秒)
+EXEC_TIMEOUT=5        # 程序自动化执行超时 (秒)
+MAX_OUTPUT_LINES=50   # 对比失败时显示的最大行数
+S_FILES=()            # 存储用户提供的 .s 文件列表
+PASSED_CASES=0
+FAILED_CASES_LIST=""
+
+# --- 函数定义 ---
+show_help() {
+    echo "用法: $0 [文件1.s] [文件2.s] ... [选项]"
+    echo "在 VM 内部编译并测试指定的 .s 文件。"
+    echo ""
+    echo "如果找到对应的 .in/.out 文件，则进行自动化测试。否则，进入交互模式。"
+    echo ""
+    echo "选项:"
+    echo "  -ct N                    设置 gcc 编译超时为 N 秒 (默认: 10)。"
+    echo "  -t N                     设置程序自动化执行超时为 N 秒 (默认: 5)。"
+    echo "  -ml N, --max-lines N     当输出对比失败时，最多显示 N 行内容 (默认: 50)。"
+    echo "  -h, --help               显示此帮助信息并退出。"
+}
+
+# 显示文件内容并根据行数截断的函数
+display_file_content() {
+    local file_path="$1"
+    local title="$2"
+    local max_lines="$3"
+
+    if [ ! -f "$file_path" ]; then
+        return
+    fi
+
+    echo -e "$title"
+    local line_count
+    line_count=$(wc -l < "$file_path")
+    
+    if [ "$line_count" -gt "$max_lines" ]; then
+        head -n "$max_lines" "$file_path"
+        echo -e "\e[33m[... 输出已截断，共 ${line_count} 行 ...]\e[0m"
+    else
+        cat "$file_path"
+    fi
+}
+
+# --- 参数解析 ---
+# 从参数中分离出 .s 文件和选项
+for arg in "$@"; do
+  case "$arg" in
+    -ct|-t|-ml|--max-lines)
+      # 选项和其值将在下一个循环中处理
+      ;;
+    -h|--help)
+      show_help
+      exit 0
+      ;;
+    -*)
+      # 检查是否是带值的选项
+      if ! [[ ${args_processed+x} ]]; then
+          args_processed=true # 标记已处理过参数
+          while [[ "$#" -gt 0 ]]; do
+              case "$1" in
+                  -ct) if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then GCC_TIMEOUT="$2"; shift; else echo "错误: -ct 需要一个正整数参数。" >&2; exit 1; fi ;;
+                  -t)  if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then EXEC_TIMEOUT="$2"; shift; else echo "错误: -t 需要一个正整数参数。" >&2; exit 1; fi ;;
+                  -ml|--max-lines) if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then MAX_OUTPUT_LINES="$2"; shift; else echo "错误: --max-lines 需要一个正整数参数。" >&2; exit 1; fi ;;
+                  *.s) S_FILES+=("$1") ;;
+                  *) if ! [[ "$1" =~ ^[0-9]+$ ]]; then echo "未知选项或无效文件: $1"; show_help; exit 1; fi ;;
+              esac
+              shift
+          done
+      fi
+      ;;
+    *.s)
+      if [[ -f "$arg" ]]; then
+        S_FILES+=("$arg")
+      else
+        echo "警告: 文件不存在，已忽略: $arg"
+      fi
+      ;;
+  esac
+done
+
+# --- 主逻辑开始 ---
+if [ ${#S_FILES[@]} -eq 0 ]; then
+    echo "错误: 未提供任何 .s 文件作为输入。"
+    show_help
+    exit 1
+fi
+
+mkdir -p "${TMP_DIR}"
+TOTAL_CASES=${#S_FILES[@]}
+
+echo "SysY VM 内单例测试运行器启动..."
+echo "超时设置: gcc=${GCC_TIMEOUT}s, 运行=${EXEC_TIMEOUT}s"
+echo "失败输出最大行数: ${MAX_OUTPUT_LINES}"
+echo ""
+
+for s_file in "${S_FILES[@]}"; do
+    is_passed=1
+    
+    # 从 .s 文件名反向推导原始测试用例路径
+    base_name_from_s_file=$(basename "$s_file" .s)
+    original_test_name_underscored=$(echo "$base_name_from_s_file" | sed 's/_sysyc_riscv64$//')
+    category=$(echo "$original_test_name_underscored" | cut -d'_' -f1)
+    test_file_base=$(echo "$original_test_name_underscored" | cut -d'_' -f2-)
+    original_relative_path="${category}/${test_file_base}"
+
+    executable_file="${TMP_DIR}/${base_name_from_s_file}"
+    input_file="${TESTDATA_DIR}/${original_relative_path}.in"
+    output_reference_file="${TESTDATA_DIR}/${original_relative_path}.out"
+    output_actual_file="${TMP_DIR}/${base_name_from_s_file}.actual_out"
+
+    echo "======================================================================"
+    echo "正在处理: ${s_file}"
+    echo "  (关联测试用例: ${original_relative_path}.sy)"
+
+    # 步骤 1: GCC 编译
+    echo "  使用 gcc 编译 (超时 ${GCC_TIMEOUT}s)..."
+    timeout -s KILL ${GCC_TIMEOUT} "${GCC_NATIVE}" "${s_file}" -o "${executable_file}" -L"${LIB_DIR}" -lsysy_riscv -static -g
+    if [ $? -ne 0 ]; then
+        echo -e "\e[31m错误: GCC 编译失败或超时。\e[0m"
+        is_passed=0
+    fi
+
+    # 步骤 2: 执行与测试
+    if [ "$is_passed" -eq 1 ]; then
+        # 检查是自动化测试还是交互模式
+        if [ -f "${input_file}" ] || [ -f "${output_reference_file}" ]; then
+            # --- 自动化测试模式 ---
+            echo "  检测到 .in/.out 文件，进入自动化测试模式..."
+            echo "  正在执行 (超时 ${EXEC_TIMEOUT}s)..."
+            
+            exec_cmd="\"${executable_file}\""
+            [ -f "${input_file}" ] && exec_cmd+=" < \"${input_file}\""
+            exec_cmd+=" > \"${output_actual_file}\""
+            
+            eval "timeout -s KILL ${EXEC_TIMEOUT} ${exec_cmd}"
+            ACTUAL_RETURN_CODE=$?
+
+            if [ "$ACTUAL_RETURN_CODE" -eq 124 ]; then
+                echo -e "\e[31m  执行超时。\e[0m"
+                is_passed=0
+            else
+                if [ -f "${output_reference_file}" ]; then
+                    LAST_LINE_TRIMMED=$(tail -n 1 "${output_reference_file}" | tr -d '[:space:]')
+                    if [[ "$LAST_LINE_TRIMMED" =~ ^[-+]?[0-9]+$ ]]; then
+                        EXPECTED_RETURN_CODE="$LAST_LINE_TRIMMED"
+                        EXPECTED_STDOUT_FILE="${TMP_DIR}/${base_name_from_s_file}.expected_stdout"
+                        head -n -1 "${output_reference_file}" > "${EXPECTED_STDOUT_FILE}"
+                        if [ "$ACTUAL_RETURN_CODE" -ne "$EXPECTED_RETURN_CODE" ]; then echo -e "\e[31m  返回码测试失败: 期望 ${EXPECTED_RETURN_CODE}, 实际 ${ACTUAL_RETURN_CODE}\e[0m"; is_passed=0; fi
+                        if ! diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${EXPECTED_STDOUT_FILE}") >/dev/null 2>&1; then
+                            echo -e "\e[31m  标准输出测试失败。\e[0m"; is_passed=0
+                            display_file_content "${EXPECTED_STDOUT_FILE}" "    \e[36m--- 期望输出 ---\e[0m" "${MAX_OUTPUT_LINES}"
+                            display_file_content "${output_actual_file}" "    \e[36m--- 实际输出 ---\e[0m" "${MAX_OUTPUT_LINES}"
+                            echo -e "    \e[36m----------------\e[0m"
+                        fi
+                    else
+                        if ! diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${output_reference_file}") >/dev/null 2>&1; then
+                            echo -e "\e[32m  标准输出测试成功。\e[0m"
+                        else
+                            echo -e "\e[31m  标准输出测试失败。\e[0m"; is_passed=0
+                            display_file_content "${output_reference_file}" "    \e[36m--- 期望输出 ---\e[0m" "${MAX_OUTPUT_LINES}"
+                            display_file_content "${output_actual_file}" "    \e[36m--- 实际输出 ---\e[0m" "${MAX_OUTPUT_LINES}"
+                            echo -e "    \e[36m----------------\e[0m"
+                        fi
+                    fi
+                else
+                    echo "  无参考输出文件。程序返回码: ${ACTUAL_RETURN_CODE}"
+                fi
+            fi
+        else
+            # --- 交互模式 ---
+            echo -e "\e[33m"
+            echo "  **********************************************************"
+            echo "  ** 未找到 .in 或 .out 文件，进入交互模式。             **"
+            echo "  ** 程序即将运行，你可以直接在终端中输入。             **"
+            echo "  ** 按下 Ctrl+D (EOF) 或以其他方式结束程序以继续。 **"
+            echo "  **********************************************************"
+            echo -e "\e[0m"
+            "${executable_file}"
+            INTERACTIVE_RET_CODE=$?
+            echo -e "\e[33m\n  交互模式执行完毕，程序返回码: ${INTERACTIVE_RET_CODE}\e[0m"
+            echo "  注意: 交互模式的结果未经验证。"
+        fi
+    fi
+
+    if [ "$is_passed" -eq 1 ]; then
+        echo -e "\e[32m状态: 通过\e[0m"
+        ((PASSED_CASES++))
+    else
+        echo -e "\e[31m状态: 失败\e[0m"
+        FAILED_CASES_LIST+="${original_relative_path}.sy\n"
+    fi
+done
+
+# --- 打印最终总结 ---
+echo "======================================================================"
+echo "所有测试完成"
+echo "测试通过率: [${PASSED_CASES}/${TOTAL_CASES}]"
+
+if [ -n "$FAILED_CASES_LIST" ]; then
+    echo ""
+    echo -e "\e[31m未通过的测例:\e[0m"
+    echo -e "${FAILED_CASES_LIST}"
+fi
+
+echo "======================================================================"
+
+if [ "$PASSED_CASES" -eq "$TOTAL_CASES" ]; then
+    exit 0
+else
+    exit 1
+fi

test_script/runit-riscv64.sh

@@ -1,16 +1,8 @@
 #!/bin/bash
 
-# run_vm_tests.sh - 用于在 RISC-V 虚拟机内部汇编、链接和测试 SysY 程序的脚本
+# runit-riscv64.sh - 用于在 RISC-V 虚拟机内部汇编、链接和测试 SysY 程序的脚本
 # 此脚本应位于您的项目根目录 (例如 /home/ubuntu/debug)
 # 假设当前运行环境已经是 RISC-V 64 位架构，可以直接执行编译后的程序。
-# 脚本的目录结构应该为：
-# .
-# ├── runit.sh
-# ├── lib
-# │   └── libsysy_riscv.a
-# └── testdata
-#     ├── functional
-#     └── performance
 
 # 定义相对于脚本位置的目录
 SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" &>/dev/null && pwd)"
@@ -22,30 +14,47 @@ TESTDATA_DIR="${SCRIPT_DIR}/testdata"
 GCC_NATIVE="gcc" # VM 内部的 gcc
 
 # --- 新增功能: 初始化变量 ---
-TIMEOUT_SECONDS=5 # 默认运行时超时时间为 5 秒
-COMPILE_TIMEOUT_SECONDS=10 # 默认编译超时时间为 10 秒
+GCC_TIMEOUT=10        # 默认 gcc 编译超时 (秒)
+EXEC_TIMEOUT=5        # 默认运行时超时 (秒)
+MAX_OUTPUT_LINES=50   # 对比失败时显示的最大行数
 TOTAL_CASES=0
 PASSED_CASES=0
+FAILED_CASES_LIST=""  # 用于存储未通过的测例列表
 
 # 显示帮助信息的函数
 show_help() {
     echo "用法: $0 [选项]"
     echo "此脚本用于在 RISC-V 虚拟机内部，对之前生成的 .s 汇编文件进行汇编、链接和测试。"
-    echo "假设当前运行环境已经是 RISC-V 64 位架构，可以直接执行编译后的程序。"
+    echo "测试会按文件名升序进行。"
     echo ""
     echo "选项:"
     echo "  -c, --clean              清理 'tmp' 目录下的所有生成文件。"
-    echo "  -t, --timeout N          设置每个测试用例的运行时超时为 N 秒 (默认: 5)。"
-    echo "  -ct, --compile-timeout M 设置 gcc 编译的超时时间为 M 秒 (默认: 10)。"
+    echo "  -ct M                    设置 gcc 编译的超时时间为 M 秒 (默认: 10)。"
+    echo "  -t N                     设置每个测试用例的运行时超时为 N 秒 (默认: 5)。"
+    echo "  -ml N, --max-lines N     当输出对比失败时，最多显示 N 行内容 (默认: 50)。"
     echo "  -h, --help               显示此帮助信息并退出。"
-    echo ""
-    echo "执行步骤:"
-    echo "1. 遍历 'tmp/' 目录下的所有 .s 汇编文件。"
-    echo "2. 在指定的超时时间内使用 VM 内部的 gcc 将 .s 文件汇编并链接为可执行文件。"
-    echo "3. 在指定的超时时间内运行编译后的可执行文件。"
-    echo "4. 根据对应的 .out 文件内容进行返回值和/或标准输出的比较。"
-    echo "5. 输出比较时会忽略行尾多余的换行符。"
-    echo "6. 所有测试结束后，报告总通过率。"
+}
+
+# 显示文件内容并根据行数截断的函数
+display_file_content() {
+    local file_path="$1"
+    local title="$2"
+    local max_lines="$3"
+
+    if [ ! -f "$file_path" ]; then
+        return
+    fi
+
+    echo -e "$title"
+    local line_count
+    line_count=$(wc -l < "$file_path")
+    
+    if [ "$line_count" -gt "$max_lines" ]; then
+        head -n "$max_lines" "$file_path"
+        echo -e "\e[33m[... 输出已截断，共 ${line_count} 行 ...]\e[0m"
+    else
+        cat "$file_path"
+    fi
 }
 
 # 清理临时文件的函数
@@ -69,23 +78,14 @@ while [[ "$#" -gt 0 ]]; do
             clean_tmp
             exit 0
             ;;
-        -t|--timeout)
-            if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then
-                TIMEOUT_SECONDS="$2"
-                shift # 移过参数值
-            else
-                echo "错误: --timeout 需要一个正整数参数。" >&2
-                exit 1
-            fi
+        -t)
+            if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then EXEC_TIMEOUT="$2"; shift; else echo "错误: -t 需要一个正整数参数。" >&2; exit 1; fi
             ;;
-        -ct|--compile-timeout)
-            if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then
-                COMPILE_TIMEOUT_SECONDS="$2"
-                shift # 移过参数值
-            else
-                echo "错误: --compile-timeout 需要一个正整数参数。" >&2
-                exit 1
-            fi
+        -ct)
+            if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then GCC_TIMEOUT="$2"; shift; else echo "错误: -ct 需要一个正整数参数。" >&2; exit 1; fi
+            ;;
+        -ml|--max-lines)
+            if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then MAX_OUTPUT_LINES="$2"; shift; else echo "错误: --max-lines 需要一个正整数参数。" >&2; exit 1; fi
             ;;
         -h|--help)
             show_help
@@ -101,32 +101,27 @@ while [[ "$#" -gt 0 ]]; do
 done
 
 echo "SysY VM 内部测试运行器启动..."
-echo "编译超时设置为: ${COMPILE_TIMEOUT_SECONDS} 秒"
-echo "运行时超时设置为: ${TIMEOUT_SECONDS} 秒"
+echo "GCC 编译超时设置为: ${GCC_TIMEOUT} 秒"
+echo "运行时超时设置为: ${EXEC_TIMEOUT} 秒"
+echo "失败输出最大行数: ${MAX_OUTPUT_LINES}"
 echo "汇编文件目录: ${TMP_DIR}"
-echo "库文件目录: ${LIB_DIR}"
-echo "测试数据目录: ${TESTDATA_DIR}"
 echo ""
 
-# 查找 tmp 目录下的所有 .s 汇编文件
-s_files=$(find "${TMP_DIR}" -maxdepth 1 -name "*.s")
+# 查找 tmp 目录下的所有 .s 汇编文件并排序
+s_files=$(find "${TMP_DIR}" -maxdepth 1 -name "*.s" | sort -V)
 TOTAL_CASES=$(echo "$s_files" | wc -w)
 
-# 遍历找到的每个 .s 文件
-echo "$s_files" | while read s_file; do
-    # --- 新增功能: 初始化用例通过状态 ---
+# 使用 here-string (<<<) 避免子 shell 问题
+while IFS= read -r s_file; do
     is_passed=1 # 1 表示通过, 0 表示失败
 
-    # 从 .s 文件名中提取原始的测试用例名称部分
     base_name_from_s_file=$(basename "$s_file" .s)
     original_test_name_underscored=$(echo "$base_name_from_s_file" | sed 's/_sysyc_riscv64$//')
 
-    # 将 `original_test_name_underscored` 分割成类别和文件名
     category=$(echo "$original_test_name_underscored" | cut -d'_' -f1)
     test_file_base=$(echo "$original_test_name_underscored" | cut -d'_' -f2-)
     original_relative_path="${category}/${test_file_base}"
 
-    # 定义可执行文件、输入文件、参考输出文件和实际输出文件的路径
     executable_file="${TMP_DIR}/${base_name_from_s_file}"
     input_file="${TESTDATA_DIR}/${original_relative_path}.in"
     output_reference_file="${TESTDATA_DIR}/${original_relative_path}.out"
@@ -136,42 +131,43 @@ echo "$s_files" | while read s_file; do
     echo "  对应的测试用例路径: ${original_relative_path}"
 
     # 步骤 1: 使用 VM 内部的 gcc 编译 .s 到可执行文件
-    echo "  使用 gcc 汇编并链接 (超时 ${COMPILE_TIMEOUT_SECONDS}s)..."
-    # --- 修改点: 为 gcc 增加 timeout ---
-    timeout ${COMPILE_TIMEOUT_SECONDS} "${GCC_NATIVE}" "${s_file}" -o "${executable_file}" -L"${LIB_DIR}" -lsysy_riscv -static -g
+    echo "  使用 gcc 汇编并链接 (超时 ${GCC_TIMEOUT}s)..."
+    timeout -s KILL ${GCC_TIMEOUT} "${GCC_NATIVE}" "${s_file}" -o "${executable_file}" -L"${LIB_DIR}" -lsysy_riscv -static -g
     GCC_STATUS=$?
     if [ $GCC_STATUS -eq 124 ]; then
-        echo -e "\e[31m错误: GCC 编译/链接 ${s_file} 超时 (超过 ${COMPILE_TIMEOUT_SECONDS} 秒)\e[0m"
+        echo -e "\e[31m错误: GCC 编译/链接 ${s_file} 超时\e[0m"
         is_passed=0
     elif [ $GCC_STATUS -ne 0 ]; then
         echo -e "\e[31m错误: GCC 汇编/链接 ${s_file} 失败，退出码: ${GCC_STATUS}\e[0m"
         is_passed=0
-    else
+    fi
+
+    # 步骤 2: 只有当编译成功时才执行
+    if [ "$is_passed" -eq 1 ]; then
         echo "  生成的可执行文件: ${executable_file}"
-        echo "  正在执行 (超时 ${TIMEOUT_SECONDS}s): \"${executable_file}\""
+        echo "  正在执行 (超时 ${EXEC_TIMEOUT}s)..."
 
-        # 步骤 2: 执行编译后的文件并比较/报告结果
-        if [ -f "${output_reference_file}" ]; then
-            LAST_LINE_TRIMMED=$(tail -n 1 "${output_reference_file}" | tr -d '[:space:]')
-            
-            if [[ "$LAST_LINE_TRIMMED" =~ ^[-+]?[0-9]+$ ]]; then
-                EXPECTED_RETURN_CODE="$LAST_LINE_TRIMMED"
-                EXPECTED_STDOUT_FILE="${TMP_DIR}/${base_name_from_s_file}.expected_stdout"
-                head -n -1 "${output_reference_file}" > "${EXPECTED_STDOUT_FILE}"
-                echo "  检测到 .out 文件同时包含标准输出和期望的返回码。"
-                echo "  期望返回码: ${EXPECTED_RETURN_CODE}"
+        exec_cmd="\"${executable_file}\""
+        if [ -f "${input_file}" ]; then
+            exec_cmd+=" < \"${input_file}\""
+        fi
+        exec_cmd+=" > \"${output_actual_file}\""
+        
+        eval "timeout -s KILL ${EXEC_TIMEOUT} ${exec_cmd}"
+        ACTUAL_RETURN_CODE=$?
 
-                if [ -f "${input_file}" ]; then
-                    timeout ${TIMEOUT_SECONDS} "${executable_file}" < "${input_file}" > "${output_actual_file}"
-                else
-                    timeout ${TIMEOUT_SECONDS} "${executable_file}" > "${output_actual_file}"
-                fi
-                ACTUAL_RETURN_CODE=$?
+        if [ "$ACTUAL_RETURN_CODE" -eq 124 ]; then
+            echo -e "\e[31m  执行超时: ${original_relative_path}.sy 运行超过 ${EXEC_TIMEOUT} 秒\e[0m"
+            is_passed=0
+        else
+            if [ -f "${output_reference_file}" ]; then
+                LAST_LINE_TRIMMED=$(tail -n 1 "${output_reference_file}" | tr -d '[:space:]')
+                
+                if [[ "$LAST_LINE_TRIMMED" =~ ^[-+]?[0-9]+$ ]]; then
+                    EXPECTED_RETURN_CODE="$LAST_LINE_TRIMMED"
+                    EXPECTED_STDOUT_FILE="${TMP_DIR}/${base_name_from_s_file}.expected_stdout"
+                    head -n -1 "${output_reference_file}" > "${EXPECTED_STDOUT_FILE}"
 
-                if [ "$ACTUAL_RETURN_CODE" -eq 124 ]; then
-                    echo -e "\e[31m  执行超时: ${original_relative_path}.sy 运行超过 ${TIMEOUT_SECONDS} 秒\e[0m"
-                    is_passed=0
-                else
                     if [ "$ACTUAL_RETURN_CODE" -eq "$EXPECTED_RETURN_CODE" ]; then
                         echo -e "\e[32m  返回码测试成功: (${ACTUAL_RETURN_CODE}) 与期望值 (${EXPECTED_RETURN_CODE}) 匹配\e[0m"
                     else
@@ -179,65 +175,51 @@ echo "$s_files" | while read s_file; do
                         is_passed=0
                     fi
 
-                    if diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${EXPECTED_STDOUT_FILE}") >/dev/null 2>&1; then
-                        echo -e "\e[32m  标准输出测试成功\e[0m"
-                    else
+                    if ! diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${EXPECTED_STDOUT_FILE}") >/dev/null 2>&1; then
                         echo -e "\e[31m  标准输出测试失败\e[0m"
-                        echo "    差异:"
-                        diff "${output_actual_file}" "${EXPECTED_STDOUT_FILE}"
                         is_passed=0
+                        display_file_content "${EXPECTED_STDOUT_FILE}" "    \e[36m---------- 期望输出 ----------\e[0m" "${MAX_OUTPUT_LINES}"
+                        display_file_content "${output_actual_file}" "    \e[36m---------- 实际输出 ----------\e[0m" "${MAX_OUTPUT_LINES}"
+                        echo -e "    \e[36m------------------------------\e[0m"
                     fi
-                fi
-            else
-                echo "  检测到 .out 文件为纯标准输出参考。"
-                if [ -f "${input_file}" ]; then
-                    timeout ${TIMEOUT_SECONDS} "${executable_file}" < "${input_file}" > "${output_actual_file}"
                 else
-                    timeout ${TIMEOUT_SECONDS} "${executable_file}" > "${output_actual_file}"
-                fi
-                EXEC_STATUS=$?
-
-                if [ $EXEC_STATUS -eq 124 ]; then
-                    echo -e "\e[31m  执行超时: ${original_relative_path}.sy 运行超过 ${TIMEOUT_SECONDS} 秒\e[0m"
-                    is_passed=0
-                else
-                    if [ $EXEC_STATUS -ne 0 ]; then
-                        echo -e "\e[33m警告: 程序以非零状态 ${EXEC_STATUS} 退出 (纯输出比较模式)。\e[0m"
+                    if [ $ACTUAL_RETURN_CODE -ne 0 ]; then
+                        echo -e "\e[33m警告: 程序以非零状态 ${ACTUAL_RETURN_CODE} 退出 (纯输出比较模式)。\e[0m"
                     fi
-                    if diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${output_reference_file}") >/dev/null 2>&1; then
+                    if ! diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${output_reference_file}") >/dev/null 2>&1; then
                         echo -e "\e[32m  成功: 输出与参考输出匹配\e[0m"
                     else
                         echo -e "\e[31m  失败: 输出不匹配\e[0m"
-                        echo "    差异:"
-                        diff "${output_actual_file}" "${output_reference_file}"
                         is_passed=0
+                        display_file_content "${output_reference_file}" "    \e[36m---------- 期望输出 ----------\e[0m" "${MAX_OUTPUT_LINES}"
+                        display_file_content "${output_actual_file}" "    \e[36m---------- 实际输出 ----------\e[0m" "${MAX_OUTPUT_LINES}"
+                        echo -e "    \e[36m------------------------------\e[0m"
                     fi
                 fi
-            fi
-        else
-            echo "  未找到 .out 文件。正在运行并报告返回码。"
-            timeout ${TIMEOUT_SECONDS} "${executable_file}"
-            EXEC_STATUS=$?
-            if [ $EXEC_STATUS -eq 124 ]; then
-                echo -e "\e[31m  执行超时: ${original_relative_path}.sy 运行超过 ${TIMEOUT_SECONDS} 秒\e[0m"
-                is_passed=0
             else
-                echo "  ${original_relative_path}.sy 的返回码: ${EXEC_STATUS}"
+                echo "  无参考输出文件。程序返回码: ${ACTUAL_RETURN_CODE}"
             fi
         fi
     fi
 
-    # --- 新增功能: 更新通过用例计数 ---
     if [ "$is_passed" -eq 1 ]; then
         ((PASSED_CASES++))
+    else
+        FAILED_CASES_LIST+="${original_relative_path}.sy\n"
     fi
-    echo "" # 为测试用例之间添加一个空行
-done
+    echo ""
+done <<< "$s_files"
 
-# --- 新增功能: 打印最终总结 ---
 echo "========================================"
 echo "测试完成"
 echo "测试通过率: [${PASSED_CASES}/${TOTAL_CASES}]"
+
+if [ -n "$FAILED_CASES_LIST" ]; then
+    echo ""
+    echo -e "\e[31m未通过的测例:\e[0m"
+    echo -e "${FAILED_CASES_LIST}"
+fi
+
 echo "========================================"
 
 if [ "$PASSED_CASES" -eq "$TOTAL_CASES" ]; then

test_script/runit-single.sh

@@ -0,0 +1,270 @@
+#!/bin/bash
+
+# runit-single.sh - 用于编译和测试单个或少量 SysY 程序的脚本
+# 模仿 runit.sh 的功能，但以具体文件路径作为输入。
+
+# --- 配置区 ---
+# 请根据你的环境修改这些路径
+# 假设此脚本位于你的项目根目录或一个脚本目录中
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" &>/dev/null && pwd)"
+# 默认寻找项目根目录下的 build 和 lib
+BUILD_BIN_DIR="${SCRIPT_DIR}/../build/bin"
+LIB_DIR="${SCRIPT_DIR}/../lib"
+# 临时文件会存储在脚本所在目录的 tmp 子目录中
+TMP_DIR="${SCRIPT_DIR}/tmp"
+
+# 定义编译器和模拟器
+SYSYC="${BUILD_BIN_DIR}/sysyc"
+GCC_RISCV64="riscv64-linux-gnu-gcc"
+QEMU_RISCV64="qemu-riscv64"
+
+# --- 初始化变量 ---
+EXECUTE_MODE=false
+SYSYC_TIMEOUT=10      # sysyc 编译超时 (秒)
+GCC_TIMEOUT=10        # gcc 编译超时 (秒)
+EXEC_TIMEOUT=5        # qemu 自动化执行超时 (秒)
+MAX_OUTPUT_LINES=50   # 对比失败时显示的最大行数
+SY_FILES=()           # 存储用户提供的 .sy 文件列表
+PASSED_CASES=0
+FAILED_CASES_LIST=""
+
+# --- 函数定义 ---
+show_help() {
+    echo "用法: $0 [文件1.sy] [文件2.sy] ... [选项]"
+    echo "编译并测试指定的 .sy 文件。"
+    echo ""
+    echo "如果找到对应的 .in/.out 文件，则进行自动化测试。否则，进入交互模式。"
+    echo ""
+    echo "选项:"
+    echo "  -e, --executable         编译为可执行文件并运行测试 (必须)。"
+    echo "  -sct N                   设置 sysyc 编译超时为 N 秒 (默认: 10)。"
+    echo "  -gct N                   设置 gcc 交叉编译超时为 N 秒 (默认: 10)。"
+    echo "  -et N                    设置 qemu 自动化执行超时为 N 秒 (默认: 5)。"
+    echo "  -ml N, --max-lines N     当输出对比失败时，最多显示 N 行内容 (默认: 50)。"
+    echo "  -h, --help               显示此帮助信息并退出。"
+}
+
+# --- 新增功能: 显示文件内容并根据行数截断 ---
+display_file_content() {
+    local file_path="$1"
+    local title="$2"
+    local max_lines="$3"
+
+    if [ ! -f "$file_path" ]; then
+        return
+    fi
+
+    echo -e "$title"
+    local line_count
+    line_count=$(wc -l < "$file_path")
+    
+    if [ "$line_count" -gt "$max_lines" ]; then
+        head -n "$max_lines" "$file_path"
+        echo -e "\e[33m[... 输出已截断，共 ${line_count} 行 ...]\e[0m"
+    else
+        cat "$file_path"
+    fi
+}
+
+
+# --- 参数解析 ---
+# 从参数中分离出 .sy 文件和选项
+for arg in "$@"; do
+  case "$arg" in
+    -e|--executable)
+      EXECUTE_MODE=true
+      ;;
+    -sct|-gct|-et|-ml|--max-lines)
+      # 选项和其值将在下一个循环中处理
+      ;;
+    -h|--help)
+      show_help
+      exit 0
+      ;;
+    -*)
+      # 检查是否是带值的选项
+      if ! [[ ${args_processed+x} ]]; then
+          args_processed=true # 标记已处理过参数
+          # 重新处理所有参数
+          while [[ "$#" -gt 0 ]]; do
+              case "$1" in
+                  -sct) if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then SYSYC_TIMEOUT="$2"; shift; else echo "错误: -sct 需要一个正整数参数。" >&2; exit 1; fi ;;
+                  -gct) if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then GCC_TIMEOUT="$2"; shift; else echo "错误: -gct 需要一个正整数参数。" >&2; exit 1; fi ;;
+                  -et)  if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then EXEC_TIMEOUT="$2"; shift; else echo "错误: -et 需要一个正整数参数。" >&2; exit 1; fi ;;
+                  -ml|--max-lines) if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then MAX_OUTPUT_LINES="$2"; shift; else echo "错误: --max-lines 需要一个正整数参数。" >&2; exit 1; fi ;;
+                  *.sy) SY_FILES+=("$1") ;;
+                  -e|--executable) ;; # 已在外部处理
+                  *) if ! [[ "$1" =~ ^[0-9]+$ ]]; then echo "未知选项或无效文件: $1"; show_help; exit 1; fi ;;
+              esac
+              shift
+          done
+      fi
+      ;;
+    *.sy)
+      if [[ -f "$arg" ]]; then
+        SY_FILES+=("$arg")
+      else
+        echo "警告: 文件不存在，已忽略: $arg"
+      fi
+      ;;
+  esac
+done
+
+# --- 主逻辑开始 ---
+if ! ${EXECUTE_MODE}; then
+    echo "错误: 请提供 -e 或 --executable 选项来运行测试。"
+    show_help
+    exit 1
+fi
+
+if [ ${#SY_FILES[@]} -eq 0 ]; then
+    echo "错误: 未提供任何 .sy 文件作为输入。"
+    show_help
+    exit 1
+fi
+
+mkdir -p "${TMP_DIR}"
+TOTAL_CASES=${#SY_FILES[@]}
+
+echo "SysY 单例测试运行器启动..."
+echo "超时设置: sysyc=${SYSYC_TIMEOUT}s, gcc=${GCC_TIMEOUT}s, qemu=${EXEC_TIMEOUT}s"
+echo "失败输出最大行数: ${MAX_OUTPUT_LINES}"
+echo ""
+
+for sy_file in "${SY_FILES[@]}"; do
+    is_passed=1
+    base_name=$(basename "${sy_file}" .sy)
+    source_dir=$(dirname "${sy_file}")
+
+    ir_file="${TMP_DIR}/${base_name}_sysyc_riscv64.ll"
+    assembly_file="${TMP_DIR}/${base_name}.s"
+    executable_file="${TMP_DIR}/${base_name}"
+    input_file="${source_dir}/${base_name}.in"
+    output_reference_file="${source_dir}/${base_name}.out"
+    output_actual_file="${TMP_DIR}/${base_name}.actual_out"
+
+    echo "======================================================================"
+    echo "正在处理: ${sy_file}"
+
+    # 步骤 1: sysyc 编译
+    echo "  使用 sysyc 编译 (超时 ${SYSYC_TIMEOUT}s)..."
+    timeout -s KILL ${SYSYC_TIMEOUT} "${SYSYC}" -s ir "${sy_file}" > "${ir_file}"
+    SYSYC_STATUS=$?
+    if [ $SYSYC_STATUS -eq 124 ]; then
+        echo -e "\e[31m错误: SysY 编译 ${sy_file} IR超时\e[0m"
+        is_passed=0
+    elif [ $SYSYC_STATUS -ne 0 ]; then
+        echo -e "\e[31m错误: SysY 编译 ${sy_file} IR失败，退出码: ${SYSYC_STATUS}\e[0m"
+        is_passed=0
+    fi
+    timeout -s KILL ${SYSYC_TIMEOUT} "${SYSYC}" -S "${sy_file}" -o "${assembly_file}"
+    if [ $? -ne 0 ]; then
+        echo -e "\e[31m错误: SysY 编译失败或超时。\e[0m"
+        is_passed=0
+    fi
+
+    # 步骤 2: GCC 编译
+    if [ "$is_passed" -eq 1 ]; then
+        echo "  使用 gcc 编译 (超时 ${GCC_TIMEOUT}s)..."
+        timeout -s KILL ${GCC_TIMEOUT} "${GCC_RISCV64}" "${assembly_file}" -o "${executable_file}" -L"${LIB_DIR}" -lsysy_riscv -static
+        if [ $? -ne 0 ]; then
+            echo -e "\e[31m错误: GCC 编译失败或超时。\e[0m"
+            is_passed=0
+        fi
+    fi
+
+    # 步骤 3: 执行与测试
+    if [ "$is_passed" -eq 1 ]; then
+        # 检查是自动化测试还是交互模式
+        if [ -f "${input_file}" ] || [ -f "${output_reference_file}" ]; then
+            # --- 自动化测试模式 ---
+            echo "  检测到 .in/.out 文件，进入自动化测试模式..."
+            echo "  正在执行 (超时 ${EXEC_TIMEOUT}s)..."
+            
+            exec_cmd="${QEMU_RISCV64} \"${executable_file}\""
+            [ -f "${input_file}" ] && exec_cmd+=" < \"${input_file}\""
+            exec_cmd+=" > \"${output_actual_file}\""
+            
+            eval "timeout -s KILL ${EXEC_TIMEOUT} ${exec_cmd}"
+            ACTUAL_RETURN_CODE=$?
+
+            if [ "$ACTUAL_RETURN_CODE" -eq 124 ]; then
+                echo -e "\e[31m  执行超时。\e[0m"
+                is_passed=0
+            else
+                if [ -f "${output_reference_file}" ]; then
+                    # 此处逻辑与 runit.sh 相同
+                    LAST_LINE_TRIMMED=$(tail -n 1 "${output_reference_file}" | tr -d '[:space:]')
+                    if [[ "$LAST_LINE_TRIMMED" =~ ^[-+]?[0-9]+$ ]]; then
+                        EXPECTED_RETURN_CODE="$LAST_LINE_TRIMMED"
+                        EXPECTED_STDOUT_FILE="${TMP_DIR}/${base_name}.expected_stdout"
+                        head -n -1 "${output_reference_file}" > "${EXPECTED_STDOUT_FILE}"
+                        if [ "$ACTUAL_RETURN_CODE" -ne "$EXPECTED_RETURN_CODE" ]; then echo -e "\e[31m  返回码测试失败: 期望 ${EXPECTED_RETURN_CODE}, 实际 ${ACTUAL_RETURN_CODE}\e[0m"; is_passed=0; fi
+                        if ! diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${EXPECTED_STDOUT_FILE}") >/dev/null 2>&1; then
+                            echo -e "\e[31m  标准输出测试失败。\e[0m"
+                            is_passed=0
+                            # --- 本次修改点: 使用新函数显示输出 ---
+                            display_file_content "${EXPECTED_STDOUT_FILE}" "    \e[36m--- 期望输出 ---\e[0m" "${MAX_OUTPUT_LINES}"
+                            display_file_content "${output_actual_file}" "    \e[36m--- 实际输出 ---\e[0m" "${MAX_OUTPUT_LINES}"
+                            echo -e "    \e[36m----------------\e[0m"
+                        fi
+                    else
+                        if ! diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${output_reference_file}") >/dev/null 2>&1; then
+                            echo -e "\e[32m  标准输出测试成功。\e[0m"
+                        else
+                            echo -e "\e[31m  标准输出测试失败。\e[0m"
+                            is_passed=0
+                            # --- 本次修改点: 使用新函数显示输出 ---
+                            display_file_content "${output_reference_file}" "    \e[36m--- 期望输出 ---\e[0m" "${MAX_OUTPUT_LINES}"
+                            display_file_content "${output_actual_file}" "    \e[36m--- 实际输出 ---\e[0m" "${MAX_OUTPUT_LINES}"
+                            echo -e "    \e[36m----------------\e[0m"
+                        fi
+                    fi
+                else
+                    echo "  无参考输出文件。程序返回码: ${ACTUAL_RETURN_CODE}"
+                fi
+            fi
+        else
+            # --- 交互模式 ---
+            echo -e "\e[33m"
+            echo "  **********************************************************"
+            echo "  ** 未找到 .in 或 .out 文件，进入交互模式。             **"
+            echo "  ** 程序即将运行，你可以直接在终端中输入。             **"
+            echo "  ** 按下 Ctrl+D (EOF) 或以其他方式结束程序以继续。 **"
+            echo "  **********************************************************"
+            echo -e "\e[0m"
+            "${QEMU_RISCV64}" "${executable_file}"
+            INTERACTIVE_RET_CODE=$?
+            echo -e "\e[33m\n  交互模式执行完毕，程序返回码: ${INTERACTIVE_RET_CODE}\e[0m"
+            # 交互模式无法自动判断对错，默认算通过，但会提示
+            echo "  注意: 交互模式的结果未经验证。"
+        fi
+    fi
+
+    if [ "$is_passed" -eq 1 ]; then
+        echo -e "\e[32m状态: 通过\e[0m"
+        ((PASSED_CASES++))
+    else
+        echo -e "\e[31m状态: 失败\e[0m"
+        FAILED_CASES_LIST+="${sy_file}\n"
+    fi
+done
+
+# --- 打印最终总结 ---
+echo "======================================================================"
+echo "所有测试完成"
+echo "测试通过率: [${PASSED_CASES}/${TOTAL_CASES}]"
+
+if [ -n "$FAILED_CASES_LIST" ]; then
+    echo ""
+    echo -e "\e[31m未通过的测例:\e[0m"
+    echo -e "${FAILED_CASES_LIST}"
+fi
+
+echo "======================================================================"
+
+if [ "$PASSED_CASES" -eq "$TOTAL_CASES" ]; then
+    exit 0
+else
+    exit 1
+fi

test_script/runit.sh

@@ -21,6 +21,7 @@ EXECUTE_MODE=false
 SYSYC_TIMEOUT=10      # sysyc 编译超时 (秒)
 GCC_TIMEOUT=10        # gcc 编译超时 (秒)
 EXEC_TIMEOUT=5        # qemu 执行超时 (秒)
+MAX_OUTPUT_LINES=50   # 对比失败时显示的最大行数
 TOTAL_CASES=0
 PASSED_CASES=0
 FAILED_CASES_LIST=""  # 用于存储未通过的测例列表
@@ -36,9 +37,32 @@ show_help() {
     echo "  -sct N                   设置 sysyc 编译超时为 N 秒 (默认: 10)。"
     echo "  -gct N                   设置 gcc 交叉编译超时为 N 秒 (默认: 10)。"
     echo "  -et N                    设置 qemu 执行超时为 N 秒 (默认: 5)。"
+    echo "  -ml N, --max-lines N     当输出对比失败时，最多显示 N 行内容 (默认: 50)。"
     echo "  -h, --help               显示此帮助信息并退出。"
 }
 
+# 显示文件内容并根据行数截断的函数
+display_file_content() {
+    local file_path="$1"
+    local title="$2"
+    local max_lines="$3"
+
+    if [ ! -f "$file_path" ]; then
+        return
+    fi
+
+    echo -e "$title"
+    local line_count
+    line_count=$(wc -l < "$file_path")
+    
+    if [ "$line_count" -gt "$max_lines" ]; then
+        head -n "$max_lines" "$file_path"
+        echo -e "\e[33m[... 输出已截断，共 ${line_count} 行 ...]\e[0m"
+    else
+        cat "$file_path"
+    fi
+}
+
 # 清理临时文件的函数
 clean_tmp() {
     echo "正在清理临时目录: ${TMP_DIR}"
@@ -67,6 +91,9 @@ while [[ "$#" -gt 0 ]]; do
         -et)
             if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then EXEC_TIMEOUT="$2"; shift; else echo "错误: -et 需要一个正整数参数。" >&2; exit 1; fi
             ;;
+        -ml|--max-lines)
+            if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then MAX_OUTPUT_LINES="$2"; shift; else echo "错误: --max-lines 需要一个正整数参数。" >&2; exit 1; fi
+            ;;
         -h|--help)
             show_help
             exit 0
@@ -86,6 +113,7 @@ echo "临时目录: ${TMP_DIR}"
 echo "执行模式: ${EXECUTE_MODE}"
 if ${EXECUTE_MODE}; then
     echo "超时设置: sysyc=${SYSYC_TIMEOUT}s, gcc=${GCC_TIMEOUT}s, qemu=${EXEC_TIMEOUT}s"
+    echo "失败输出最大行数: ${MAX_OUTPUT_LINES}"
 fi
 echo ""
 
@@ -93,8 +121,7 @@ echo ""
 sy_files=$(find "${TESTDATA_DIR}" -name "*.sy" | sort -V)
 TOTAL_CASES=$(echo "$sy_files" | wc -w)
 
-# --- 本次修复: 使用 here-string (<<<) 代替管道 (|) 来避免子 shell 问题 ---
-# 这样可以确保循环内的 PASSED_CASES 变量修改在循环结束后依然有效
+# --- 修复: 使用 here-string (<<<) 代替管道 (|) 来避免子 shell 问题 ---
 while IFS= read -r sy_file; do
     is_passed=1 # 1 表示通过, 0 表示失败
 
@@ -111,7 +138,7 @@ while IFS= read -r sy_file; do
 
     # 步骤 1: 使用 sysyc 编译 .sy 到 .s
     echo "  使用 sysyc 编译 (超时 ${SYSYC_TIMEOUT}s)..."
-    timeout ${SYSYC_TIMEOUT} "${SYSYC}" -S "${sy_file}" -o "${assembly_file}"
+    timeout -s KILL ${SYSYC_TIMEOUT} "${SYSYC}" -S "${sy_file}" -o "${assembly_file}"
     SYSYC_STATUS=$?
     if [ $SYSYC_STATUS -eq 124 ]; then
         echo -e "\e[31m错误: SysY 编译 ${sy_file} 超时\e[0m"
@@ -125,7 +152,7 @@ while IFS= read -r sy_file; do
     if ${EXECUTE_MODE} && [ "$is_passed" -eq 1 ]; then
         # 步骤 2: 使用 riscv64-linux-gnu-gcc 编译 .s 到可执行文件
         echo "  使用 gcc 编译 (超时 ${GCC_TIMEOUT}s)..."
-        timeout ${GCC_TIMEOUT} "${GCC_RISCV64}" "${assembly_file}" -o "${executable_file}" -L"${LIB_DIR}" -lsysy_riscv -static
+        timeout -s KILL ${GCC_TIMEOUT} "${GCC_RISCV64}" "${assembly_file}" -o "${executable_file}" -L"${LIB_DIR}" -lsysy_riscv -static
         GCC_STATUS=$?
         if [ $GCC_STATUS -eq 124 ]; then
             echo -e "\e[31m错误: GCC 编译 ${assembly_file} 超时\e[0m"
@@ -136,11 +163,9 @@ while IFS= read -r sy_file; do
         fi
     elif ! ${EXECUTE_MODE}; then
         echo "  跳过执行模式。仅生成汇编文件。"
-        # 如果只编译不执行，只要编译成功就算通过
         if [ "$is_passed" -eq 1 ]; then
              ((PASSED_CASES++))
         else
-            # --- 本次修改点 ---
             FAILED_CASES_LIST+="${relative_path_no_ext}.sy\n"
         fi
         echo ""
@@ -151,22 +176,19 @@ while IFS= read -r sy_file; do
     if [ "$is_passed" -eq 1 ]; then
         echo "  正在执行 (超时 ${EXEC_TIMEOUT}s)..."
         
-        # 准备执行命令
         exec_cmd="${QEMU_RISCV64} \"${executable_file}\""
         if [ -f "${input_file}" ]; then
             exec_cmd+=" < \"${input_file}\""
         fi
         exec_cmd+=" > \"${output_actual_file}\""
 
-        # 执行并捕获返回码
-        eval "timeout ${EXEC_TIMEOUT} ${exec_cmd}"
+        eval "timeout -s KILL ${EXEC_TIMEOUT} ${exec_cmd}"
         ACTUAL_RETURN_CODE=$?
 
         if [ "$ACTUAL_RETURN_CODE" -eq 124 ]; then
             echo -e "\e[31m  执行超时: ${sy_file} 运行超过 ${EXEC_TIMEOUT} 秒\e[0m"
             is_passed=0
         else
-            # 检查是否存在 .out 文件以进行比较
             if [ -f "${output_reference_file}" ]; then
                 LAST_LINE_TRIMMED=$(tail -n 1 "${output_reference_file}" | tr -d '[:space:]')
                 
@@ -175,70 +197,54 @@ while IFS= read -r sy_file; do
                     EXPECTED_STDOUT_FILE="${TMP_DIR}/${output_base_name}_sysyc_riscv64.expected_stdout"
                     head -n -1 "${output_reference_file}" > "${EXPECTED_STDOUT_FILE}"
 
-                    # 比较返回码
                     if [ "$ACTUAL_RETURN_CODE" -eq "$EXPECTED_RETURN_CODE" ]; then
                         echo -e "\e[32m  返回码测试成功: (${ACTUAL_RETURN_CODE}) 与期望值 (${EXPECTED_RETURN_CODE}) 匹配\e[0m"
                     else
                         echo -e "\e[31m  返回码测试失败: 期望: ${EXPECTED_RETURN_CODE}, 实际: ${ACTUAL_RETURN_CODE}\e[0m"
                         is_passed=0
                     fi
-                    # 比较标准输出
-                    if diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${EXPECTED_STDOUT_FILE}") >/dev/null 2>&1; then
-                        echo -e "\e[32m  标准输出测试成功\e[0m"
-                    else
+                    if ! diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${EXPECTED_STDOUT_FILE}") >/dev/null 2>&1; then
                         echo -e "\e[31m  标准输出测试失败\e[0m"
                         is_passed=0
-                        echo -e "    \e[36m---------- 期望输出 ----------\e[0m"
-                        cat "${EXPECTED_STDOUT_FILE}"
-                        echo -e "    \e[36m---------- 实际输出 ----------\e[0m"
-                        cat "${output_actual_file}"
+                        display_file_content "${EXPECTED_STDOUT_FILE}" "    \e[36m---------- 期望输出 ----------\e[0m" "${MAX_OUTPUT_LINES}"
+                        display_file_content "${output_actual_file}" "    \e[36m---------- 实际输出 ----------\e[0m" "${MAX_OUTPUT_LINES}"
                         echo -e "    \e[36m------------------------------\e[0m"
                     fi
                 else
-                    # 纯标准输出比较
                     if [ $ACTUAL_RETURN_CODE -ne 0 ]; then
                         echo -e "\e[33m警告: 程序以非零状态 ${ACTUAL_RETURN_CODE} 退出 (纯输出比较模式)。\e[0m"
                     fi
-                    if diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${output_reference_file}") >/dev/null 2>&1; then
+                    if ! diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${output_reference_file}") >/dev/null 2>&1; then
                         echo -e "\e[32m  成功: 输出与参考输出匹配\e[0m"
                     else
                         echo -e "\e[31m  失败: 输出不匹配\e[0m"
                         is_passed=0
-                        echo -e "    \e[36m---------- 期望输出 ----------\e[0m"
-                        cat "${output_reference_file}"
-                        echo -e "    \e[36m---------- 实际输出 ----------\e[0m"
-                        cat "${output_actual_file}"
+                        display_file_content "${output_reference_file}" "    \e[36m---------- 期望输出 ----------\e[0m" "${MAX_OUTPUT_LINES}"
+                        display_file_content "${output_actual_file}" "    \e[36m---------- 实际输出 ----------\e[0m" "${MAX_OUTPUT_LINES}"
                         echo -e "    \e[36m------------------------------\e[0m"
                     fi
                 fi
             else
-                # 没有 .out 文件，只报告返回码
                 echo "  无参考输出文件。程序返回码: ${ACTUAL_RETURN_CODE}"
             fi
         fi
     fi
 
-    # 更新通过用例计数
-    # --- 本次修改点 ---
     if [ "$is_passed" -eq 1 ]; then
         ((PASSED_CASES++))
     else
-        # 将失败的用例名称添加到列表中
         FAILED_CASES_LIST+="${relative_path_no_ext}.sy\n"
     fi
-    echo "" # 添加空行以提高可读性
+    echo ""
 done <<< "$sy_files"
 
-# --- 新增功能: 打印最终总结 ---
 echo "========================================"
 echo "测试完成"
 echo "测试通过率: [${PASSED_CASES}/${TOTAL_CASES}]"
 
-# --- 本次修改点: 打印未通过的测例列表 ---
 if [ -n "$FAILED_CASES_LIST" ]; then
     echo ""
     echo -e "\e[31m未通过的测例:\e[0m"
-    # 使用 -e 来解释换行符 \n
     echo -e "${FAILED_CASES_LIST}"
 fi