[IR]消除Falltrhough现象

[IR]优化生成Ret指令逻辑
[README]添加TODO表

.gitignore

@@ -36,7 +36,7 @@
 doxygen
 
 !/testdata/functional/*.out
-!/testdata/performance/*.out
+!/testdata/h_functional/*.out
 build/
 .antlr
 .vscode/

README.md

@@ -38,3 +38,12 @@ 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

src/AddressCalculationExpansion.cpp

@@ -0,0 +1,160 @@
+#include "AddressCalculationExpansion.h"
+#include <iostream>
+#include <vector>
+#include "IR.h"
+#include "IRBuilder.h"
+
+extern int DEBUG;
+
+namespace sysy {
+
+bool AddressCalculationExpansion::run() {
+    bool changed = false;
+
+    for (auto& funcPair : pModule->getFunctions()) {
+        Function* func = funcPair.second.get();
+        for (auto& bb_ptr : func->getBasicBlocks()) {
+            BasicBlock* bb = bb_ptr.get();
+            for (auto it = bb->getInstructions().begin(); it != bb->getInstructions().end(); ) {
+                Instruction* inst = it->get();
+
+                Value* basePointer = nullptr;
+                Value* valueToStore = nullptr;
+                size_t firstIndexOperandIdx = 0;
+                size_t numBaseOperands = 0;
+
+                if (inst->isLoad()) {
+                    numBaseOperands = 1;
+                    basePointer = inst->getOperand(0);
+                    firstIndexOperandIdx = 1;
+                } else if (inst->isStore()) {
+                    numBaseOperands = 2;
+                    valueToStore = inst->getOperand(0);
+                    basePointer = inst->getOperand(1);
+                    firstIndexOperandIdx = 2;
+                } else {
+                    ++it;
+                    continue;
+                }
+
+                if (inst->getNumOperands() <= numBaseOperands) {
+                    ++it;
+                    continue;
+                }
+
+                std::vector<int> dims;
+                if (AllocaInst* allocaInst = dynamic_cast<AllocaInst*>(basePointer)) {
+                    for (const auto& use_ptr : allocaInst->getDims()) {
+                        Value* dimValue = use_ptr->getValue();
+                        if (ConstantValue* constVal = dynamic_cast<ConstantValue*>(dimValue)) {
+                            dims.push_back(constVal->getInt());
+                        } else {
+                            std::cerr << "Warning: AllocaInst dimension is not a constant integer. Skipping GEP expansion for: ";
+                            SysYPrinter::printValue(allocaInst);
+                            std::cerr << "\n";
+                            dims.clear();
+                            break;
+                        }
+                    }
+                } 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;
+                } 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);
+                    std::cerr << " in instruction ";
+                    SysYPrinter::printInst(inst);
+                    std::cerr << "\n";
+                    ++it;
+                    continue;
+                }
+
+                if (dims.empty() && (inst->getNumOperands() > numBaseOperands)) {
+                    if (DEBUG) {
+                        std::cerr << "ACE Warning: Could not get valid array dimensions for ";
+                        SysYPrinter::printValue(basePointer);
+                        std::cerr << " in instruction ";
+                        SysYPrinter::printInst(inst);
+                        std::cerr << " (expected dimensions for indices, but got none).\n";
+                    }
+                    ++it;
+                    continue;
+                }
+
+                std::vector<Value*> indexOperands;
+                for (size_t i = firstIndexOperandIdx; i < inst->getNumOperands(); ++i) {
+                    indexOperands.push_back(inst->getOperand(i));
+                }
+
+                if (AllocaInst* allocaInst = dynamic_cast<AllocaInst*>(basePointer)) {
+                    if (allocaInst->getNumDims() != indexOperands.size()) {
+                        if (DEBUG) {
+                            std::cerr << "ACE Warning: Index count (" << indexOperands.size() << ") does not match AllocaInst dimensions (" << allocaInst->getNumDims() << ") for instruction ";
+                            SysYPrinter::printInst(inst);
+                            std::cerr << "\n";
+                        }
+                        ++it;
+                        continue;
+                    }
+                }
+
+                Value* totalOffset = ConstantInteger::get(0);
+                pBuilder->setPosition(bb, it);
+
+                for (size_t i = 0; i < indexOperands.size(); ++i) {
+                    Value* index = indexOperands[i];
+                    int stride = calculateStride(dims, i);
+                    Value* strideConst = ConstantInteger::get(stride);
+                    Type* intType = Type::getIntType();
+                    BinaryInst* currentDimOffsetInst = pBuilder->createBinaryInst(Instruction::kMul, intType, index, strideConst);
+                    BinaryInst* newTotalOffsetInst = pBuilder->createBinaryInst(Instruction::kAdd, intType, totalOffset, currentDimOffsetInst);
+                    totalOffset = newTotalOffsetInst;
+                }
+
+                // 计算有效地址：effective_address = basePointer + totalOffset
+                Value* effective_address = pBuilder->createBinaryInst(Instruction::kAdd, basePointer->getType(), basePointer, totalOffset);
+
+                // 创建新的 LoadInst 或 StoreInst，indices 为空
+                Instruction* newInst = nullptr;
+                if (inst->isLoad()) {
+                    newInst = pBuilder->createLoadInst(effective_address, {});
+                    inst->replaceAllUsesWith(newInst);
+                } else { // StoreInst
+                    newInst = pBuilder->createStoreInst(valueToStore, effective_address, {});
+                }
+
+                Instruction* oldInst = it->get();
+                ++it;
+
+                for (size_t i = 0; i < oldInst->getNumOperands(); ++i) {
+                    Value* operandValue = oldInst->getOperand(i);
+                    if (operandValue) {
+                        for (auto use_it = operandValue->getUses().begin(); use_it != operandValue->getUses().end(); ++use_it) {
+                            if ((*use_it)->getUser() == oldInst && (*use_it)->getIndex() == i) {
+                                operandValue->removeUse(*use_it);
+                                break;
+                            }
+                        }
+                    }
+                }
+
+                bb->getInstructions().erase(std::prev(it));
+                changed = true;
+
+                if (DEBUG) {
+                    std::cerr << "ACE: Computed effective address:\n";
+                    SysYPrinter::printInst(dynamic_cast<Instruction*>(effective_address));
+                    std::cerr << "ACE: New Load/Store instruction:\n";
+                    SysYPrinter::printInst(newInst);
+                    std::cerr << "--------------------------------\n";
+                }
+            }
+        }
+    }
+    return changed;
+}
+
+} // namespace sysy

src/CMakeLists.txt

@@ -21,12 +21,17 @@ add_executable(sysyc
   IR.cpp
   SysYIRGenerator.cpp
   SysYIRPrinter.cpp
-  SysYIROptPre.cpp
+  SysYIRCFGOpt.cpp
-  SysYIRAnalyser.cpp
+  # SysYIRAnalyser.cpp
-  DeadCodeElimination.cpp
+  # DeadCodeElimination.cpp
-  Mem2Reg.cpp
+  AddressCalculationExpansion.cpp
-  Reg2Mem.cpp
+  # Mem2Reg.cpp
+  # Reg2Mem.cpp
   RISCv64Backend.cpp
+  RISCv64ISel.cpp
+  RISCv64RegAlloc.cpp
+  RISCv64AsmPrinter.cpp
+  RISCv64Passes.cpp
 )
 
 # 设置 include 路径，包含 ANTLR 运行时库和项目头文件

src/DeadCodeElimination.cpp

@@ -37,7 +37,7 @@ void DeadCodeElimination::eliminateDeadStores(Function* func, bool& changed) {
       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 (isGlobal(pointer) || (isArr(pointer) &&
+      if (SysYIROptUtils::isGlobal(pointer) || (SysYIROptUtils::isArr(pointer) &&
             std::find(func->getEntryBlock()->getArguments().begin(),
                     func->getEntryBlock()->getArguments().end(),
                     pointer) != func->getEntryBlock()->getArguments().end())) {
@@ -63,7 +63,7 @@ void DeadCodeElimination::eliminateDeadStores(Function* func, bool& changed) {
           std::cout << "=== Dead Store Found ===\n";
           SysYPrinter::printInst(storeInst);
         }
-        usedelete(storeInst);
+        SysYIROptUtils::usedelete(storeInst);
         iter = instrs.erase(iter);
       } else {
         ++iter;
@@ -85,7 +85,7 @@ void DeadCodeElimination::eliminateDeadLoads(Function* func, bool& changed) {
             std::cout << "=== Dead Load Binary Unary Found ===\n";
             SysYPrinter::printInst(inst);
           }
-          usedelete(inst);
+          SysYIROptUtils::usedelete(inst);
           iter = instrs.erase(iter);
           continue;
         }
@@ -114,7 +114,7 @@ void DeadCodeElimination::eliminateDeadAllocas(Function* func, bool& changed) {
             std::cout << "=== Dead Alloca Found ===\n";
             SysYPrinter::printInst(inst);
           }
-          usedelete(inst);
+          SysYIROptUtils::usedelete(inst);
           iter = instrs.erase(iter);
           continue;
         }
@@ -183,7 +183,7 @@ void DeadCodeElimination::eliminateDeadRedundantLoadStore(Function* func, bool&
         /// 如果 pointer 仅被该 phi 使用，可以删除 ph
         if (tag) {
           changed = true;
-          usedelete(inst);
+          SysYIROptUtils::usedelete(inst);
           iter = instrs.erase(iter);
           continue;
         }
@@ -193,7 +193,7 @@ void DeadCodeElimination::eliminateDeadRedundantLoadStore(Function* func, bool&
         auto pointer = memsetInst->getPointer();
         if (pointer->getUses().empty()) {
           changed = true;
-          usedelete(inst);
+          SysYIROptUtils::usedelete(inst);
           iter = instrs.erase(iter);
           continue;
         }
@@ -234,7 +234,7 @@ void DeadCodeElimination::eliminateDeadRedundantLoadStore(Function* func, bool&
                         SysYPrinter::printInst(loadInst);
                         SysYPrinter::printInst(nextStore);
                       }
-                      usedelete(loadInst);
+                      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()) {
@@ -256,21 +256,4 @@ void DeadCodeElimination::eliminateDeadRedundantLoadStore(Function* func, bool&
 }
 
 
-bool DeadCodeElimination::isGlobal(Value *val){
-  auto gval = dynamic_cast<GlobalValue *>(val);
-  return gval != nullptr;
-}
-
-bool DeadCodeElimination::isArr(Value *val){
-  auto aval = dynamic_cast<AllocaInst *>(val);
-  return aval != nullptr && aval->getNumDims() != 0;
-}
-
-void DeadCodeElimination::usedelete(Instruction *instr){
-  for (auto &use1 : instr->getOperands()) {
-    auto val1 = use1->getValue();
-    val1->removeUse(use1);
-  }
-}
-
 } // namespace sysy

src/IR.cpp

@@ -49,6 +49,11 @@ 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:
@@ -58,6 +63,10 @@ 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;
   }
@@ -95,6 +104,11 @@ 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);
@@ -102,30 +116,54 @@ void Value::replaceAllUsesWith(Value *value) {
   uses.clear();
 }
 
-ConstantValue* ConstantValue::get(int value) {
+
-  static std::map<int, std::unique_ptr<ConstantValue>> intConstants;
+// Implementations for static members
-  auto iter = intConstants.find(value);
+
-  if (iter != intConstants.end()) {
+std::unordered_map<ConstantValueKey, ConstantValue*, ConstantValueHash, ConstantValueEqual> ConstantValue::mConstantPool;
-    return iter->second.get();
+std::unordered_map<Type*, UndefinedValue*> UndefinedValue::UndefValues;
+
+ConstantValue* ConstantValue::get(Type* type, ConstantValVariant val) {
+  ConstantValueKey key = {type, val};
+  auto it = mConstantPool.find(key);
+  if (it != mConstantPool.end()) {
+    return it->second;
   }
-  auto inst = new ConstantValue(value);
+
-  assert(inst);
+  ConstantValue* newConstant = nullptr;
-  auto result = intConstants.emplace(value, inst);
+  if (std::holds_alternative<int>(val)) {
-  return result.first->second.get();
+    newConstant = new ConstantInteger(type, std::get<int>(val));
+  } else if (std::holds_alternative<float>(val)) {
+    newConstant = new ConstantFloating(type, std::get<float>(val));
+  } else {
+    assert(false && "Unsupported ConstantValVariant type");
+  }
+
+  mConstantPool[key] = newConstant;
+  return newConstant;
 }
 
-ConstantValue* ConstantValue::get(float value) {
+ConstantInteger* ConstantInteger::get(Type* type, int val) {
-  static std::map<float, std::unique_ptr<ConstantValue>> floatConstants;
+  return dynamic_cast<ConstantInteger*>(ConstantValue::get(type, val));
-  auto iter = floatConstants.find(value);
-  if (iter != floatConstants.end()) {
-    return iter->second.get();
-  }
-  auto inst = new ConstantValue(value);
-  assert(inst);
-  auto result = floatConstants.emplace(value, inst);
-  return result.first->second.get();
 }
 
+ConstantFloating* ConstantFloating::get(Type* type, float val) {
+  return dynamic_cast<ConstantFloating*>(ConstantValue::get(type, val));
+}
+
+UndefinedValue* UndefinedValue::get(Type* type) {
+  assert(!type->isVoid() && "Cannot get UndefinedValue of void type!");
+
+  auto it = UndefValues.find(type);
+  if (it != UndefValues.end()) {
+    return it->second;
+  }
+
+  UndefinedValue* newUndef = new UndefinedValue(type);
+  UndefValues[type] = newUndef;
+  return newUndef;
+}
+
+
 auto Function::getCalleesWithNoExternalAndSelf() -> std::set<Function *> {
   std::set<Function *> result;
   for (auto callee : callees) {
@@ -441,44 +479,7 @@ Function * Function::clone(const std::string &suffix) const  {
         break;
       }
 
-      case Instruction::kLa: {
+      // TODO：复制GEP指令
-        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);
@@ -545,6 +546,83 @@ void User::replaceOperand(unsigned index, Value *value) {
   value->addUse(use);
 }
 
+/**
+ * phi相关函数
+ */
+
+ Value* PhiInst::getvalfromBlk(BasicBlock* blk){
+  refreshB2VMap();
+  if( blk2val.find(blk) != blk2val.end()) {
+    return blk2val.at(blk);
+  }
+  return nullptr;
+}
+
+BasicBlock* PhiInst::getBlkfromVal(Value* val){
+  // 返回第一个值对应的基本块
+  for(unsigned i = 0; i < vsize; i++) {
+    if(getValue(i) == val) {
+      return getBlock(i);
+    }
+  }
+  return nullptr;
+}
+
+void PhiInst::delValue(Value* val){
+  //根据value删除对应的基本块和值
+  unsigned i = 0;
+  BasicBlock* blk = getBlkfromVal(val);
+  for(i = 0; i < vsize; i++) {
+    if(getValue(i) == val) {
+      break;
+    }
+  }
+  removeOperand(2 * i + 1); // 删除blk
+  removeOperand(2 * i); // 删除val
+  vsize--;
+  blk2val.erase(blk); // 删除blk2val映射
+}
+
+void PhiInst::delBlk(BasicBlock* blk){
+  //根据Blk删除对应的基本块和值
+  unsigned i = 0;
+  Value* val = getvalfromBlk(blk);
+  for(i = 0; i < vsize; i++) {
+    if(getBlock(i) == blk) {
+      break;
+    }
+  }
+  removeOperand(2 * i + 1); // 删除blk
+  removeOperand(2 * i); // 删除val
+  vsize--;
+  blk2val.erase(blk); // 删除blk2val映射
+}
+
+void PhiInst::replaceBlk(BasicBlock* newBlk, unsigned k){
+  refreshB2VMap();
+  Value* val = blk2val.at(getBlock(k));
+  // 替换基本块
+  setOperand(2 * k + 1, newBlk);
+  // 替换blk2val映射
+  blk2val.erase(getBlock(k));
+  blk2val.emplace(newBlk, val);
+}
+
+void PhiInst::replaceold2new(BasicBlock* oldBlk, BasicBlock* newBlk){
+  refreshB2VMap();
+  Value* val = blk2val.at(oldBlk);
+  // 替换基本块
+  delBlk(oldBlk);
+  addIncoming(val, newBlk);
+}
+
+void PhiInst::refreshB2VMap(){
+  blk2val.clear();
+  for(unsigned i = 0; i < vsize; i++) {
+    blk2val.emplace(getBlock(i), getValue(i));
+  }
+}
+
 CallInst::CallInst(Function *callee, const std::vector<Value *> &args, BasicBlock *parent, const std::string &name)
     : Instruction(kCall, callee->getReturnType(), parent, name) {
   addOperand(callee);

src/RISCv64AsmPrinter.cpp

@@ -0,0 +1,225 @@
+#include "RISCv64AsmPrinter.h"
+#include "RISCv64ISel.h"
+#include <stdexcept>
+
+namespace sysy {
+
+// 检查是否为内存加载/存储指令，以处理特殊的打印格式
+bool isMemoryOp(RVOpcodes opcode) {
+    switch (opcode) {
+        case RVOpcodes::LB: case RVOpcodes::LH: case RVOpcodes::LW: case RVOpcodes::LD:
+        case RVOpcodes::LBU: case RVOpcodes::LHU: case RVOpcodes::LWU:
+        case RVOpcodes::SB: case RVOpcodes::SH: case RVOpcodes::SW: case RVOpcodes::SD:
+            return true;
+        default:
+            return false;
+    }
+}
+
+RISCv64AsmPrinter::RISCv64AsmPrinter(MachineFunction* mfunc) : MFunc(mfunc) {}
+
+void RISCv64AsmPrinter::run(std::ostream& os) {
+    OS = &os;
+
+    *OS << ".globl " << MFunc->getName() << "\n";
+    *OS << MFunc->getName() << ":\n";
+
+    printPrologue();
+
+    for (auto& mbb : MFunc->getBlocks()) {
+        printBasicBlock(mbb.get());
+    }
+}
+
+void RISCv64AsmPrinter::printPrologue() {
+    StackFrameInfo& frame_info = MFunc->getFrameInfo();
+    // 序言需要为保存ra和s0预留16字节
+    int total_stack_size = frame_info.locals_size + frame_info.spill_size + 16;
+    int aligned_stack_size = (total_stack_size + 15) & ~15;
+    frame_info.total_size = aligned_stack_size;
+
+    if (aligned_stack_size > 0) {
+        *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";
+    }
+
+    // 忠实还原保存函数入口参数的逻辑
+    Function* F = MFunc->getFunc();
+    if (F && F->getEntryBlock()) {
+        int arg_idx = 0;
+        RISCv64ISel* isel = MFunc->getISel();
+        for (AllocaInst* alloca_for_param : F->getEntryBlock()->getArguments()) {
+            if (arg_idx >= 8) break;
+            
+            unsigned vreg = isel->getVReg(alloca_for_param);
+            if (frame_info.alloca_offsets.count(vreg)) {
+                int offset = frame_info.alloca_offsets.at(vreg);
+                auto arg_reg = static_cast<PhysicalReg>(static_cast<int>(PhysicalReg::A0) + arg_idx);
+                *OS << "    sw " << regToString(arg_reg) << ", " << offset << "(s0)\n";
+            }
+            arg_idx++;
+        }
+    }
+}
+
+void RISCv64AsmPrinter::printEpilogue() {
+    int aligned_stack_size = MFunc->getFrameInfo().total_size;
+    if (aligned_stack_size > 0) {
+        *OS << "    ld ra, " << (aligned_stack_size - 8) << "(sp)\n";
+        *OS << "    ld s0, " << (aligned_stack_size - 16) << "(sp)\n";
+        *OS << "    addi sp, sp, " << aligned_stack_size << "\n";
+    }
+}
+
+void RISCv64AsmPrinter::printBasicBlock(MachineBasicBlock* mbb) {
+    if (!mbb->getName().empty()) {
+        *OS << mbb->getName() << ":\n";
+    }
+    for (auto& instr : mbb->getInstructions()) {
+        printInstruction(instr.get());
+    }
+}
+
+void RISCv64AsmPrinter::printInstruction(MachineInstr* instr) {
+    auto opcode = instr->getOpcode();
+    if (opcode == RVOpcodes::RET) {
+        printEpilogue();
+    }
+    if (opcode != RVOpcodes::LABEL) {
+        *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;
+        case RVOpcodes::SUB:   *OS << "sub ";   break; case RVOpcodes::SUBW:  *OS << "subw ";  break;
+        case RVOpcodes::MUL:   *OS << "mul ";   break; case RVOpcodes::MULW:  *OS << "mulw ";  break;
+        case RVOpcodes::DIV:   *OS << "div ";   break; case RVOpcodes::DIVW:  *OS << "divw ";  break;
+        case RVOpcodes::REM:   *OS << "rem ";   break; case RVOpcodes::REMW:  *OS << "remw ";  break;
+        case RVOpcodes::XOR:   *OS << "xor ";   break; case RVOpcodes::XORI:  *OS << "xori ";  break;
+        case RVOpcodes::OR:    *OS << "or ";    break; case RVOpcodes::ORI:   *OS << "ori ";   break;
+        case RVOpcodes::AND:   *OS << "and ";   break; case RVOpcodes::ANDI:  *OS << "andi ";  break;
+        case RVOpcodes::SLL:   *OS << "sll ";   break; case RVOpcodes::SLLI:  *OS << "slli ";  break;
+        case RVOpcodes::SLLW:  *OS << "sllw ";  break; case RVOpcodes::SLLIW: *OS << "slliw "; break;
+        case RVOpcodes::SRL:   *OS << "srl ";   break; case RVOpcodes::SRLI:  *OS << "srli ";  break;
+        case RVOpcodes::SRLW:  *OS << "srlw ";  break; case RVOpcodes::SRLIW: *OS << "srliw "; break;
+        case RVOpcodes::SRA:   *OS << "sra ";   break; case RVOpcodes::SRAI:  *OS << "srai ";  break;
+        case RVOpcodes::SRAW:  *OS << "sraw ";  break; case RVOpcodes::SRAIW: *OS << "sraiw "; break;
+        case RVOpcodes::SLT:   *OS << "slt ";   break; case RVOpcodes::SLTI:  *OS << "slti ";  break;
+        case RVOpcodes::SLTU:  *OS << "sltu ";  break; case RVOpcodes::SLTIU: *OS << "sltiu "; break;
+        case RVOpcodes::LW:    *OS << "lw ";    break; case RVOpcodes::LH:    *OS << "lh ";    break;
+        case RVOpcodes::LB:    *OS << "lb ";    break; case RVOpcodes::LWU:   *OS << "lwu ";   break;
+        case RVOpcodes::LHU:   *OS << "lhu ";   break; case RVOpcodes::LBU:   *OS << "lbu ";   break;
+        case RVOpcodes::SW:    *OS << "sw ";    break; case RVOpcodes::SH:    *OS << "sh ";    break;
+        case RVOpcodes::SB:    *OS << "sb ";    break; case RVOpcodes::LD:    *OS << "ld ";    break;
+        case RVOpcodes::SD:    *OS << "sd ";    break;
+        case RVOpcodes::J:     *OS << "j ";     break; case RVOpcodes::JAL:   *OS << "jal ";   break;
+        case RVOpcodes::JALR:  *OS << "jalr ";  break; case RVOpcodes::RET:   *OS << "ret";    break;
+        case RVOpcodes::BEQ:   *OS << "beq ";   break; case RVOpcodes::BNE:   *OS << "bne ";   break;
+        case RVOpcodes::BLT:   *OS << "blt ";   break; case RVOpcodes::BGE:   *OS << "bge ";   break;
+        case RVOpcodes::BLTU:  *OS << "bltu ";  break; case RVOpcodes::BGEU:  *OS << "bgeu ";  break;
+        case RVOpcodes::LI:    *OS << "li ";    break; case RVOpcodes::LA:    *OS << "la ";    break;
+        case RVOpcodes::MV:    *OS << "mv ";    break; case RVOpcodes::NEG:   *OS << "neg ";   break;
+        case RVOpcodes::NEGW:  *OS << "negw ";  break; case RVOpcodes::SEQZ:  *OS << "seqz ";  break;
+        case RVOpcodes::SNEZ:  *OS << "snez ";  break;
+        case RVOpcodes::CALL:  *OS << "call ";  break;
+        case RVOpcodes::LABEL:
+            printOperand(instr->getOperands()[0].get());
+            *OS << ":";
+            break;
+        case RVOpcodes::FRAME_LOAD:
+        case RVOpcodes::FRAME_STORE:
+            // These should have been eliminated by RegAlloc
+            throw std::runtime_error("FRAME pseudo-instruction not eliminated before AsmPrinter");
+        default: 
+            throw std::runtime_error("Unknown opcode in AsmPrinter");
+    }
+
+    const auto& operands = instr->getOperands();
+    if (!operands.empty()) {
+        if (isMemoryOp(opcode)) {
+            printOperand(operands[0].get());
+            *OS << ", ";
+            printOperand(operands[1].get());
+        } else {
+            for (size_t i = 0; i < operands.size(); ++i) {
+                printOperand(operands[i].get());
+                if (i < operands.size() - 1) {
+                    *OS << ", ";
+                }
+            }
+        }
+    }
+    *OS << "\n";
+}
+
+void RISCv64AsmPrinter::printOperand(MachineOperand* op) {
+    if (!op) return;
+    switch(op->getKind()) {
+        case MachineOperand::KIND_REG: {
+            auto reg_op = static_cast<RegOperand*>(op);
+            if (reg_op->isVirtual()) {
+                *OS << "%vreg" << reg_op->getVRegNum();
+            } else {
+                *OS << regToString(reg_op->getPReg());
+            }
+            break;
+        }
+        case MachineOperand::KIND_IMM:
+            *OS << static_cast<ImmOperand*>(op)->getValue();
+            break;
+        case MachineOperand::KIND_LABEL:
+            *OS << static_cast<LabelOperand*>(op)->getName();
+            break;
+        case MachineOperand::KIND_MEM: {
+            auto mem_op = static_cast<MemOperand*>(op);
+            printOperand(mem_op->getOffset());
+            *OS << "(";
+            printOperand(mem_op->getBase());
+            *OS << ")";
+            break;
+        }
+    }
+}
+
+std::string RISCv64AsmPrinter::regToString(PhysicalReg reg) {
+    switch (reg) {
+        case PhysicalReg::ZERO: return "x0";  case PhysicalReg::RA: return "ra";
+        case PhysicalReg::SP: return "sp";    case PhysicalReg::GP: return "gp";
+        case PhysicalReg::TP: return "tp";    case PhysicalReg::T0: return "t0";
+        case PhysicalReg::T1: return "t1";    case PhysicalReg::T2: return "t2";
+        case PhysicalReg::S0: return "s0";    case PhysicalReg::S1: return "s1";
+        case PhysicalReg::A0: return "a0";    case PhysicalReg::A1: return "a1";
+        case PhysicalReg::A2: return "a2";    case PhysicalReg::A3: return "a3";
+        case PhysicalReg::A4: return "a4";    case PhysicalReg::A5: return "a5";
+        case PhysicalReg::A6: return "a6";    case PhysicalReg::A7: return "a7";
+        case PhysicalReg::S2: return "s2";    case PhysicalReg::S3: return "s3";
+        case PhysicalReg::S4: return "s4";    case PhysicalReg::S5: return "s5";
+        case PhysicalReg::S6: return "s6";    case PhysicalReg::S7: return "s7";
+        case PhysicalReg::S8: return "s8";    case PhysicalReg::S9: return "s9";
+        case PhysicalReg::S10: return "s10";  case PhysicalReg::S11: return "s11";
+        case PhysicalReg::T3: return "t3";    case PhysicalReg::T4: return "t4";
+        case PhysicalReg::T5: return "t5";    case PhysicalReg::T6: return "t6";
+        case PhysicalReg::F0: return "f0";    case PhysicalReg::F1: return "f1";
+        case PhysicalReg::F2: return "f2";    case PhysicalReg::F3: return "f3";
+        case PhysicalReg::F4: return "f4";    case PhysicalReg::F5: return "f5";
+        case PhysicalReg::F6: return "f6";    case PhysicalReg::F7: return "f7";
+        case PhysicalReg::F8: return "f8";    case PhysicalReg::F9: return "f9";
+        case PhysicalReg::F10: return "f10";  case PhysicalReg::F11: return "f11";
+        case PhysicalReg::F12: return "f12";  case PhysicalReg::F13: return "f13";
+        case PhysicalReg::F14: return "f14";  case PhysicalReg::F15: return "f15";
+        case PhysicalReg::F16: return "f16";  case PhysicalReg::F17: return "f17";
+        case PhysicalReg::F18: return "f18";  case PhysicalReg::F19: return "f19";
+        case PhysicalReg::F20: return "f20";  case PhysicalReg::F21: return "f21";
+        case PhysicalReg::F22: return "f22";  case PhysicalReg::F23: return "f23";
+        case PhysicalReg::F24: return "f24";  case PhysicalReg::F25: return "f25";
+        case PhysicalReg::F26: return "f26";  case PhysicalReg::F27: return "f27";
+        case PhysicalReg::F28: return "f28";  case PhysicalReg::F29: return "f29";
+        case PhysicalReg::F30: return "f30";  case PhysicalReg::F31: return "f31";
+        default: return "UNKNOWN_REG";
+    }
+}
+
+} // namespace sysy

src/RISCv64ISel.cpp

@@ -0,0 +1,635 @@
+#include "RISCv64ISel.h"
+#include <stdexcept>
+#include <set>
+#include <functional>
+#include <cmath> // For std::fabs
+#include <limits> // For std::numeric_limits
+
+namespace sysy {
+
+// DAG节点定义 (内部实现)
+struct RISCv64ISel::DAGNode {
+    enum NodeKind { CONSTANT, LOAD, STORE, BINARY, CALL, RETURN, BRANCH, ALLOCA_ADDR, UNARY, MEMSET };
+    NodeKind kind;
+    Value* value = nullptr;
+    std::vector<DAGNode*> operands;
+    std::vector<DAGNode*> users;
+    DAGNode(NodeKind k) : kind(k) {}
+};
+
+RISCv64ISel::RISCv64ISel() : vreg_counter(0), local_label_counter(0) {}
+
+// 为一个IR Value获取或分配一个新的虚拟寄存器
+unsigned RISCv64ISel::getVReg(Value* val) {
+    if (!val) {
+        throw std::runtime_error("Cannot get vreg for a null Value.");
+    }
+    if (vreg_map.find(val) == vreg_map.end()) {
+        if (vreg_counter == 0) {
+            vreg_counter = 1; // vreg 0 保留
+        }
+        vreg_map[val] = vreg_counter++;
+    }
+    return vreg_map.at(val);
+}
+
+// 主入口函数
+std::unique_ptr<MachineFunction> RISCv64ISel::runOnFunction(Function* func) {
+    F = func;
+    if (!F) return nullptr;
+    MFunc = std::make_unique<MachineFunction>(F, this);
+    vreg_map.clear();
+    bb_map.clear();
+    vreg_counter = 0;
+    local_label_counter = 0;
+
+    select();
+
+    return std::move(MFunc);
+}
+
+// 指令选择主流程
+void RISCv64ISel::select() {
+    for (const auto& bb_ptr : F->getBasicBlocks()) {
+        auto mbb = std::make_unique<MachineBasicBlock>(bb_ptr->getName(), MFunc.get());
+        bb_map[bb_ptr.get()] = mbb.get();
+        MFunc->addBlock(std::move(mbb));
+    }
+    
+    if (F->getEntryBlock()) {
+        for (auto* arg_alloca : F->getEntryBlock()->getArguments()) {
+            getVReg(arg_alloca);
+        }
+    }
+
+    for (const auto& bb_ptr : F->getBasicBlocks()) {
+        selectBasicBlock(bb_ptr.get());
+    }
+    
+    for (const auto& bb_ptr : F->getBasicBlocks()) {
+        CurMBB = bb_map.at(bb_ptr.get());
+        for (auto succ : bb_ptr->getSuccessors()) {
+            CurMBB->successors.push_back(bb_map.at(succ));
+        }
+        for (auto pred : bb_ptr->getPredecessors()) {
+            CurMBB->predecessors.push_back(bb_map.at(pred));
+        }
+    }
+}
+
+// 处理单个基本块
+void RISCv64ISel::selectBasicBlock(BasicBlock* bb) {
+    CurMBB = bb_map.at(bb);
+    auto dag = build_dag(bb);
+
+    std::map<Value*, DAGNode*> value_to_node;
+    for(const auto& node : dag) {
+        if (node->value) {
+            value_to_node[node->value] = node.get();
+        }
+    }
+    
+    std::set<DAGNode*> selected_nodes;
+    std::function<void(DAGNode*)> select_recursive = 
+        [&](DAGNode* node) {
+        if (!node || selected_nodes.count(node)) return;
+        for (auto operand : node->operands) {
+            select_recursive(operand);
+        }
+        selectNode(node);
+        selected_nodes.insert(node);
+    };
+    
+    for (const auto& inst_ptr : bb->getInstructions()) {
+        DAGNode* node_to_select = nullptr;
+        if (value_to_node.count(inst_ptr.get())) {
+            node_to_select = value_to_node.at(inst_ptr.get());
+        } else {
+            for(const auto& node : dag) {
+                if(node->value == inst_ptr.get()) {
+                    node_to_select = node.get();
+                    break;
+                }
+            }
+        }
+        if(node_to_select) {
+            select_recursive(node_to_select);
+        }
+    }
+}
+
+// 核心函数：为DAG节点选择并生成MachineInstr (忠实移植版)
+void RISCv64ISel::selectNode(DAGNode* node) {
+    switch (node->kind) {
+        case DAGNode::CONSTANT:
+        case DAGNode::ALLOCA_ADDR:
+            if (node->value) getVReg(node->value);
+            break;
+        
+        case DAGNode::LOAD: {
+            auto dest_vreg = getVReg(node->value);
+            Value* ptr_val = node->operands[0]->value;
+
+            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)) {
+                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()));
+                CurMBB->addInstruction(std::move(la));
+
+                auto lw = std::make_unique<MachineInstr>(RVOpcodes::LW);
+                lw->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                lw->addOperand(std::make_unique<MemOperand>(
+                    std::make_unique<RegOperand>(addr_vreg),
+                    std::make_unique<ImmOperand>(0)
+                ));
+                CurMBB->addInstruction(std::move(lw));
+            } else {
+                auto ptr_vreg = getVReg(ptr_val);
+                auto lw = std::make_unique<MachineInstr>(RVOpcodes::LW);
+                lw->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                lw->addOperand(std::make_unique<MemOperand>(
+                    std::make_unique<RegOperand>(ptr_vreg),
+                    std::make_unique<ImmOperand>(0)
+                ));
+                CurMBB->addInstruction(std::move(lw));
+            }
+            break;
+        }
+        
+        case DAGNode::STORE: {
+            Value* val_to_store = node->operands[0]->value;
+            Value* ptr_val = node->operands[1]->value;
+
+            if (auto val_const = dynamic_cast<ConstantValue*>(val_to_store)) {
+                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()));
+                CurMBB->addInstruction(std::move(li));
+            }
+            auto val_vreg = getVReg(val_to_store);
+
+            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();
+                auto la = std::make_unique<MachineInstr>(RVOpcodes::LA);
+                la->addOperand(std::make_unique<RegOperand>(addr_vreg));
+                la->addOperand(std::make_unique<LabelOperand>(global->getName()));
+                CurMBB->addInstruction(std::move(la));
+
+                auto sw = std::make_unique<MachineInstr>(RVOpcodes::SW);
+                sw->addOperand(std::make_unique<RegOperand>(val_vreg));
+                sw->addOperand(std::make_unique<MemOperand>(
+                    std::make_unique<RegOperand>(addr_vreg),
+                    std::make_unique<ImmOperand>(0)
+                ));
+                CurMBB->addInstruction(std::move(sw));
+            } else {
+                auto ptr_vreg = getVReg(ptr_val);
+                auto sw = std::make_unique<MachineInstr>(RVOpcodes::SW);
+                sw->addOperand(std::make_unique<RegOperand>(val_vreg));
+                sw->addOperand(std::make_unique<MemOperand>(
+                    std::make_unique<RegOperand>(ptr_vreg),
+                    std::make_unique<ImmOperand>(0)
+                ));
+                CurMBB->addInstruction(std::move(sw));
+            }
+            break;
+        }
+
+        case DAGNode::BINARY: {
+            auto bin = dynamic_cast<BinaryInst*>(node->value);
+            Value* lhs = bin->getLhs();
+            Value* rhs = bin->getRhs();
+
+            auto load_val_if_const = [&](Value* val) {
+                if (auto c = dynamic_cast<ConstantValue*>(val)) {
+                    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;
+                    }
+                }
+            }
+
+            switch (bin->getKind()) {
+                case BinaryInst::kAdd: {
+                    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));
+                    instr->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    CurMBB->addInstruction(std::move(instr));
+                    break;
+                }
+                case BinaryInst::kSub: {
+                    auto instr = std::make_unique<MachineInstr>(RVOpcodes::SUBW);
+                    instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    CurMBB->addInstruction(std::move(instr));
+                    break;
+                }
+                case BinaryInst::kMul: {
+                    auto instr = std::make_unique<MachineInstr>(RVOpcodes::MULW);
+                    instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    CurMBB->addInstruction(std::move(instr));
+                    break;
+                }
+                case Instruction::kDiv: {
+                    auto instr = std::make_unique<MachineInstr>(RVOpcodes::DIVW);
+                    instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    CurMBB->addInstruction(std::move(instr));
+                    break;
+                }
+                case Instruction::kRem: {
+                    auto instr = std::make_unique<MachineInstr>(RVOpcodes::REMW);
+                    instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    CurMBB->addInstruction(std::move(instr));
+                    break;
+                }
+                case BinaryInst::kICmpEQ: {
+                    auto sub = std::make_unique<MachineInstr>(RVOpcodes::SUBW);
+                    sub->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    sub->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    sub->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    CurMBB->addInstruction(std::move(sub));
+
+                    auto seqz = std::make_unique<MachineInstr>(RVOpcodes::SEQZ);
+                    seqz->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    seqz->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    CurMBB->addInstruction(std::move(seqz));
+                    break;
+                }
+                case BinaryInst::kICmpNE: {
+                    auto sub = std::make_unique<MachineInstr>(RVOpcodes::SUBW);
+                    sub->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    sub->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    sub->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    CurMBB->addInstruction(std::move(sub));
+
+                    auto snez = std::make_unique<MachineInstr>(RVOpcodes::SNEZ);
+                    snez->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    snez->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    CurMBB->addInstruction(std::move(snez));
+                    break;
+                }
+                case BinaryInst::kICmpLT: {
+                    auto instr = std::make_unique<MachineInstr>(RVOpcodes::SLT);
+                    instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    CurMBB->addInstruction(std::move(instr));
+                    break;
+                }
+                case BinaryInst::kICmpGT: {
+                    auto instr = std::make_unique<MachineInstr>(RVOpcodes::SLT);
+                    instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    CurMBB->addInstruction(std::move(instr));
+                    break;
+                }
+                case BinaryInst::kICmpLE: {
+                    auto slt = std::make_unique<MachineInstr>(RVOpcodes::SLT);
+                    slt->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    slt->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    slt->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    CurMBB->addInstruction(std::move(slt));
+
+                    auto xori = std::make_unique<MachineInstr>(RVOpcodes::XORI);
+                    xori->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    xori->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    xori->addOperand(std::make_unique<ImmOperand>(1));
+                    CurMBB->addInstruction(std::move(xori));
+                    break;
+                }
+                 case BinaryInst::kICmpGE: {
+                    auto slt = std::make_unique<MachineInstr>(RVOpcodes::SLT);
+                    slt->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    slt->addOperand(std::make_unique<RegOperand>(lhs_vreg));
+                    slt->addOperand(std::make_unique<RegOperand>(rhs_vreg));
+                    CurMBB->addInstruction(std::move(slt));
+
+                    auto xori = std::make_unique<MachineInstr>(RVOpcodes::XORI);
+                    xori->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    xori->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    xori->addOperand(std::make_unique<ImmOperand>(1));
+                    CurMBB->addInstruction(std::move(xori));
+                    break;
+                }
+                default:
+                    throw std::runtime_error("Unsupported binary instruction in ISel");
+            }
+            break;
+        }
+
+        case DAGNode::UNARY: {
+            auto unary = dynamic_cast<UnaryInst*>(node->value);
+            auto dest_vreg = getVReg(unary);
+            auto src_vreg = getVReg(unary->getOperand());
+
+            switch (unary->getKind()) {
+                case UnaryInst::kNeg: {
+                    auto instr = std::make_unique<MachineInstr>(RVOpcodes::SUBW);
+                    instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(PhysicalReg::ZERO));
+                    instr->addOperand(std::make_unique<RegOperand>(src_vreg));
+                    CurMBB->addInstruction(std::move(instr));
+                    break;
+                }
+                case UnaryInst::kNot: {
+                    auto instr = std::make_unique<MachineInstr>(RVOpcodes::SEQZ);
+                    instr->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                    instr->addOperand(std::make_unique<RegOperand>(src_vreg));
+                    CurMBB->addInstruction(std::move(instr));
+                    break;
+                }
+                default:
+                    throw std::runtime_error("Unsupported unary instruction in ISel");
+            }
+            break;
+        }
+
+        case DAGNode::CALL: {
+            auto call = dynamic_cast<CallInst*>(node->value);
+            for (size_t i = 0; i < node->operands.size() && i < 8; ++i) {
+                DAGNode* arg_node = node->operands[i];
+                auto arg_preg = static_cast<PhysicalReg>(static_cast<int>(PhysicalReg::A0) + i);
+
+                if (arg_node->kind == DAGNode::CONSTANT) {
+                    if (auto const_val = dynamic_cast<ConstantValue*>(arg_node->value)) {
+                        auto li = std::make_unique<MachineInstr>(RVOpcodes::LI);
+                        li->addOperand(std::make_unique<RegOperand>(arg_preg));
+                        li->addOperand(std::make_unique<ImmOperand>(const_val->getInt()));
+                        CurMBB->addInstruction(std::move(li));
+                    }
+                } else {
+                    auto src_vreg = getVReg(arg_node->value);
+                    auto mv = std::make_unique<MachineInstr>(RVOpcodes::MV);
+                    mv->addOperand(std::make_unique<RegOperand>(arg_preg));
+                    mv->addOperand(std::make_unique<RegOperand>(src_vreg));
+                    CurMBB->addInstruction(std::move(mv));
+                }
+            }
+
+            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 (!call->getType()->isVoid()) {
+                auto mv_instr = std::make_unique<MachineInstr>(RVOpcodes::MV);
+                mv_instr->addOperand(std::make_unique<RegOperand>(getVReg(call)));
+                mv_instr->addOperand(std::make_unique<RegOperand>(PhysicalReg::A0));
+                CurMBB->addInstruction(std::move(mv_instr));
+            }
+            break;
+        }
+
+        case DAGNode::RETURN: {
+            auto ret_inst_ir = dynamic_cast<ReturnInst*>(node->value);
+            if (ret_inst_ir && ret_inst_ir->hasReturnValue()) {
+                Value* ret_val = ret_inst_ir->getReturnValue();
+                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));
+                    li_instr->addOperand(std::make_unique<ImmOperand>(const_val->getInt()));
+                    CurMBB->addInstruction(std::move(li_instr));
+                } else {
+                    auto mv_instr = std::make_unique<MachineInstr>(RVOpcodes::MV);
+                    mv_instr->addOperand(std::make_unique<RegOperand>(PhysicalReg::A0));
+                    mv_instr->addOperand(std::make_unique<RegOperand>(getVReg(ret_val)));
+                    CurMBB->addInstruction(std::move(mv_instr));
+                }
+            }
+            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)) {
+                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>(PhysicalReg::ZERO));
+                br_instr->addOperand(std::make_unique<LabelOperand>(cond_br->getThenBlock()->getName()));
+                CurMBB->addInstruction(std::move(br_instr));
+            } 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: {
+            auto memset = dynamic_cast<MemsetInst*>(node->value);
+            auto r_dest_addr = getVReg(memset->getPointer());
+            auto r_num_bytes = getVReg(memset->getSize());
+            auto r_value_byte = getVReg(memset->getValue());
+            auto r_counter = getNewVReg();
+            auto r_end_addr = getNewVReg();
+            auto r_current_addr = getNewVReg();
+            auto r_temp_val = getNewVReg();
+
+            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));
+                i->addOperand(std::make_unique<RegOperand>(rs1));
+                i->addOperand(std::make_unique<RegOperand>(rs2));
+                CurMBB->addInstruction(std::move(i));
+            };
+            auto addi_instr = [&](RVOpcodes op, unsigned rd, unsigned rs1, int64_t imm) {
+                auto i = std::make_unique<MachineInstr>(op);
+                i->addOperand(std::make_unique<RegOperand>(rd));
+                i->addOperand(std::make_unique<RegOperand>(rs1));
+                i->addOperand(std::make_unique<ImmOperand>(imm));
+                CurMBB->addInstruction(std::move(i));
+            };
+            auto store_instr = [&](RVOpcodes op, unsigned src, unsigned base, int64_t off) {
+                auto i = std::make_unique<MachineInstr>(op);
+                i->addOperand(std::make_unique<RegOperand>(src));
+                i->addOperand(std::make_unique<MemOperand>(std::make_unique<RegOperand>(base), std::make_unique<ImmOperand>(off)));
+                CurMBB->addInstruction(std::move(i));
+            };
+             auto branch_instr = [&](RVOpcodes op, unsigned rs1, unsigned rs2, const std::string& label) {
+                auto i = std::make_unique<MachineInstr>(op);
+                i->addOperand(std::make_unique<RegOperand>(rs1));
+                i->addOperand(std::make_unique<RegOperand>(rs2));
+                i->addOperand(std::make_unique<LabelOperand>(label));
+                CurMBB->addInstruction(std::move(i));
+            };
+            auto jump_instr = [&](const std::string& label) {
+                auto i = std::make_unique<MachineInstr>(RVOpcodes::J);
+                i->addOperand(std::make_unique<LabelOperand>(label));
+                CurMBB->addInstruction(std::move(i));
+            };
+            auto label_instr = [&](const std::string& name) {
+                auto i = std::make_unique<MachineInstr>(RVOpcodes::LABEL);
+                i->addOperand(std::make_unique<LabelOperand>(name));
+                CurMBB->addInstruction(std::move(i));
+            };
+
+            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);
+            
+            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);
+            addi_instr(RVOpcodes::SLLI, r_value_byte, r_temp_val, 16);
+            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);
+            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));
+            mv->addOperand(std::make_unique<RegOperand>(r_dest_addr));
+            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);
+            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);
+            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;
+        }
+
+        default:
+            throw std::runtime_error("Unsupported DAGNode kind in ISel");
+    }
+}
+
+// 以下是忠实移植的DAG构建函数
+RISCv64ISel::DAGNode* RISCv64ISel::create_node(int kind_int, Value* val, std::map<Value*, DAGNode*>& value_to_node, std::vector<std::unique_ptr<DAGNode>>& nodes_storage) {
+    auto kind = static_cast<DAGNode::NodeKind>(kind_int);
+    if (val && value_to_node.count(val) && kind != DAGNode::STORE && kind != DAGNode::RETURN && kind != DAGNode::BRANCH && kind != DAGNode::MEMSET) {
+        return value_to_node[val];
+    }
+    auto node = std::make_unique<DAGNode>(kind);
+    node->value = val;
+    DAGNode* raw_node_ptr = node.get();
+    nodes_storage.push_back(std::move(node));
+    if (val && !val->getType()->isVoid() && (dynamic_cast<Instruction*>(val) || dynamic_cast<GlobalValue*>(val))) {
+        value_to_node[val] = raw_node_ptr;
+    }
+    return raw_node_ptr;
+}
+
+RISCv64ISel::DAGNode* RISCv64ISel::get_operand_node(Value* val_ir, std::map<Value*, DAGNode*>& value_to_node, std::vector<std::unique_ptr<DAGNode>>& nodes_storage) {
+    if (value_to_node.count(val_ir)) {
+        return value_to_node[val_ir];
+    } else if (dynamic_cast<ConstantValue*>(val_ir)) {
+        return create_node(DAGNode::CONSTANT, val_ir, value_to_node, nodes_storage);
+    } else if (dynamic_cast<GlobalValue*>(val_ir)) {
+        return create_node(DAGNode::CONSTANT, val_ir, value_to_node, nodes_storage);
+    } else if (dynamic_cast<AllocaInst*>(val_ir)) {
+        return create_node(DAGNode::ALLOCA_ADDR, val_ir, value_to_node, nodes_storage);
+    }
+    return create_node(DAGNode::LOAD, val_ir, value_to_node, nodes_storage);
+}
+
+std::vector<std::unique_ptr<RISCv64ISel::DAGNode>> RISCv64ISel::build_dag(BasicBlock* bb) {
+    std::vector<std::unique_ptr<DAGNode>> nodes_storage;
+    std::map<Value*, DAGNode*> value_to_node;
+
+    for (const auto& inst_ptr : bb->getInstructions()) {
+        Instruction* inst = inst_ptr.get();
+        if (auto alloca = dynamic_cast<AllocaInst*>(inst)) {
+            create_node(DAGNode::ALLOCA_ADDR, alloca, value_to_node, nodes_storage);
+        } else if (auto store = dynamic_cast<StoreInst*>(inst)) {
+            auto store_node = create_node(DAGNode::STORE, store, value_to_node, nodes_storage);
+            store_node->operands.push_back(get_operand_node(store->getValue(), value_to_node, nodes_storage));
+            store_node->operands.push_back(get_operand_node(store->getPointer(), value_to_node, nodes_storage));
+        } else if (auto memset = dynamic_cast<MemsetInst*>(inst)) {
+            auto memset_node = create_node(DAGNode::MEMSET, memset, value_to_node, nodes_storage);
+            memset_node->operands.push_back(get_operand_node(memset->getPointer(), value_to_node, nodes_storage));
+            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));
+        } 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));
+        } else if (auto bin = dynamic_cast<BinaryInst*>(inst)) {
+            if(value_to_node.count(bin)) continue;
+            if (bin->getKind() == BinaryInst::kSub) {
+                if (auto const_lhs = dynamic_cast<ConstantValue*>(bin->getLhs())) {
+                    if (const_lhs->getInt() == 0) {
+                        auto unary_node = create_node(DAGNode::UNARY, bin, value_to_node, nodes_storage);
+                        unary_node->operands.push_back(get_operand_node(bin->getRhs(), value_to_node, nodes_storage));
+                        continue;
+                    }
+                }
+            }
+            auto bin_node = create_node(DAGNode::BINARY, bin, value_to_node, nodes_storage);
+            bin_node->operands.push_back(get_operand_node(bin->getLhs(), value_to_node, nodes_storage));
+            bin_node->operands.push_back(get_operand_node(bin->getRhs(), value_to_node, nodes_storage));
+        } else if (auto un = dynamic_cast<UnaryInst*>(inst)) {
+            if(value_to_node.count(un)) continue;
+            auto unary_node = create_node(DAGNode::UNARY, un, value_to_node, nodes_storage);
+            unary_node->operands.push_back(get_operand_node(un->getOperand(), value_to_node, nodes_storage));
+        } else if (auto call = dynamic_cast<CallInst*>(inst)) {
+            if(value_to_node.count(call)) continue;
+            auto call_node = create_node(DAGNode::CALL, call, value_to_node, nodes_storage);
+            for (auto arg : call->getArguments()) {
+                call_node->operands.push_back(get_operand_node(arg->getValue(), value_to_node, nodes_storage));
+            }
+        } else if (auto ret = dynamic_cast<ReturnInst*>(inst)) {
+            auto ret_node = create_node(DAGNode::RETURN, ret, value_to_node, nodes_storage);
+            if (ret->hasReturnValue()) {
+                ret_node->operands.push_back(get_operand_node(ret->getReturnValue(), value_to_node, nodes_storage));
+            }
+        } else if (auto cond_br = dynamic_cast<CondBrInst*>(inst)) {
+            auto br_node = create_node(DAGNode::BRANCH, cond_br, value_to_node, nodes_storage);
+            br_node->operands.push_back(get_operand_node(cond_br->getCondition(), value_to_node, nodes_storage));
+        } else if (auto uncond_br = dynamic_cast<UncondBrInst*>(inst)) {
+            create_node(DAGNode::BRANCH, uncond_br, value_to_node, nodes_storage);
+        }
+    }
+    return nodes_storage;
+}
+
+} // namespace sysy

src/RISCv64Passes.cpp

@@ -0,0 +1,54 @@
+#include "RISCv64Passes.h"
+#include <iostream>
+
+namespace sysy {
+
+// --- 寄存器分配前优化 ---
+
+void PreRA_Scheduler::runOnMachineFunction(MachineFunction* mfunc) {
+    // TODO: 在此实现寄存器分配前的指令调度。
+    // 遍历mfunc中的每一个MachineBasicBlock。
+    // 对每个基本块内的MachineInstr列表进行重排。
+    //
+    // 实现思路:
+    // 1. 分析每个基本块内指令的数据依赖关系，构建依赖图(DAG)。
+    // 2. 根据目标处理器的流水线特性（指令延迟等），使用列表调度等算法对指令进行重排。
+    // 3. 此时操作的是虚拟寄存器，只存在真依赖，调度自由度最大。
+    //
+    // std::cout << "Running Pre-RA Instruction Scheduler..." << std::endl;
+}
+
+
+// --- 寄存器分配后优化 ---
+
+void PeepholeOptimizer::runOnMachineFunction(MachineFunction* mfunc) {
+    // TODO: 在此实现窥孔优化。
+    // 遍历mfunc中的每一个MachineBasicBlock。
+    // 对每个基本块内的MachineInstr列表进行扫描和替换。
+    //
+    // 实现思路:
+    // 1. 维护一个大小固定（例如3-5条指令）的滑动窗口。
+    // 2. 识别特定的冗余模式，例如:
+    //    - `mv a0, a1` 后紧跟 `mv a1, a0` (可消除的交换)
+    //    - `sw t0, 12(s0)` 后紧跟 `lw t1, 12(s0)` (冗余加载)
+    //    - 强度削减: `mul x, x, 2` -> `slli x, x, 1`
+    // 3. 识别后，直接修改MachineInstr列表（删除、替换或插入指令）。
+    //
+    // std::cout << "Running Post-RA Peephole Optimizer..." << std::endl;
+}
+
+void PostRA_Scheduler::runOnMachineFunction(MachineFunction* mfunc) {
+    // TODO: 在此实现寄存器分配后的局部指令调度。
+    // 遍历mfunc中的每一个MachineBasicBlock。
+    // 重点关注由寄存器分配器插入的spill/fill代码。
+    //
+    // 实现思路:
+    // 1. 识别出用于spill/fill的lw/sw指令。
+    // 2. 在不违反数据依赖（包括物理寄存器引入的伪依赖）的前提下，
+    //    尝试将lw指令向上移动，使其与使用它的指令之间有足够的距离，以隐藏访存延迟。
+    // 3. 同样，可以尝试将sw指令向下移动。
+    //
+    // std::cout << "Running Post-RA Local Scheduler..." << std::endl;
+}
+
+} // namespace sysy

src/RISCv64RegAlloc.cpp

@@ -0,0 +1,322 @@
+#include "RISCv64RegAlloc.h"
+#include "RISCv64ISel.h"
+#include <algorithm>
+#include <vector>
+
+namespace sysy {
+
+RISCv64RegAlloc::RISCv64RegAlloc(MachineFunction* mfunc) : MFunc(mfunc) {
+    allocable_int_regs = {
+        PhysicalReg::T0, PhysicalReg::T1, PhysicalReg::T2, PhysicalReg::T3,
+        PhysicalReg::T4, PhysicalReg::T5, PhysicalReg::T6,
+        PhysicalReg::A0, PhysicalReg::A1, PhysicalReg::A2, PhysicalReg::A3,
+        PhysicalReg::A4, PhysicalReg::A5, PhysicalReg::A6, PhysicalReg::A7,
+        PhysicalReg::S0, PhysicalReg::S1, PhysicalReg::S2, PhysicalReg::S3,
+        PhysicalReg::S4, PhysicalReg::S5, PhysicalReg::S6, PhysicalReg::S7,
+        PhysicalReg::S8, PhysicalReg::S9, PhysicalReg::S10, PhysicalReg::S11,
+    };
+}
+
+void RISCv64RegAlloc::run() {
+    eliminateFrameIndices();
+    analyzeLiveness();
+    buildInterferenceGraph();
+    colorGraph();
+    rewriteFunction();
+}
+
+void RISCv64RegAlloc::eliminateFrameIndices() {
+    StackFrameInfo& frame_info = MFunc->getFrameInfo();
+    int current_offset = 0;
+    Function* F = MFunc->getFunc();
+    RISCv64ISel* isel = MFunc->getISel();
+
+    for (auto& bb : F->getBasicBlocks()) {
+        for (auto& inst : bb->getInstructions()) {
+            if (auto alloca = dynamic_cast<AllocaInst*>(inst.get())) {
+                int size = 4;
+                if (!alloca->getDims().empty()) {
+                    int num_elements = 1;
+                    for (const auto& dim_use : alloca->getDims()) {
+                        if (auto const_dim = dynamic_cast<ConstantValue*>(dim_use->getValue())) {
+                            num_elements *= const_dim->getInt();
+                        }
+                    }
+                    size *= num_elements;
+                }
+                current_offset += size;
+                unsigned alloca_vreg = isel->getVReg(alloca);
+                frame_info.alloca_offsets[alloca_vreg] = -current_offset;
+            }
+        }
+    }
+    frame_info.locals_size = current_offset;
+
+    for (auto& mbb : MFunc->getBlocks()) {
+        std::vector<std::unique_ptr<MachineInstr>> new_instructions;
+        for (auto& instr_ptr : mbb->getInstructions()) {
+            if (instr_ptr->getOpcode() == RVOpcodes::FRAME_LOAD) {
+                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);
+                auto addr_vreg = isel->getNewVReg();
+
+                auto addi = std::make_unique<MachineInstr>(RVOpcodes::ADDI);
+                addi->addOperand(std::make_unique<RegOperand>(addr_vreg));
+                addi->addOperand(std::make_unique<RegOperand>(PhysicalReg::S0));
+                addi->addOperand(std::make_unique<ImmOperand>(offset));
+                new_instructions.push_back(std::move(addi));
+
+                auto lw = std::make_unique<MachineInstr>(RVOpcodes::LW);
+                lw->addOperand(std::make_unique<RegOperand>(dest_vreg));
+                lw->addOperand(std::make_unique<MemOperand>(
+                    std::make_unique<RegOperand>(addr_vreg),
+                    std::make_unique<ImmOperand>(0)));
+                new_instructions.push_back(std::move(lw));
+
+            } else if (instr_ptr->getOpcode() == RVOpcodes::FRAME_STORE) {
+                auto& operands = instr_ptr->getOperands();
+                unsigned src_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);
+                auto addr_vreg = isel->getNewVReg();
+                
+                auto addi = std::make_unique<MachineInstr>(RVOpcodes::ADDI);
+                addi->addOperand(std::make_unique<RegOperand>(addr_vreg));
+                addi->addOperand(std::make_unique<RegOperand>(PhysicalReg::S0));
+                addi->addOperand(std::make_unique<ImmOperand>(offset));
+                new_instructions.push_back(std::move(addi));
+
+                auto sw = std::make_unique<MachineInstr>(RVOpcodes::SW);
+                sw->addOperand(std::make_unique<RegOperand>(src_vreg));
+                sw->addOperand(std::make_unique<MemOperand>(
+                    std::make_unique<RegOperand>(addr_vreg),
+                    std::make_unique<ImmOperand>(0)));
+                new_instructions.push_back(std::move(sw));
+            } else {
+                new_instructions.push_back(std::move(instr_ptr));
+            }
+        }
+        mbb->getInstructions() = std::move(new_instructions);
+    }
+}
+
+void RISCv64RegAlloc::getInstrUseDef(MachineInstr* instr, LiveSet& use, LiveSet& def) {
+    bool is_def = true;
+    auto opcode = instr->getOpcode();
+
+    // 预定义def和use规则
+    if (opcode == RVOpcodes::SW || opcode == RVOpcodes::SD || 
+        opcode == RVOpcodes::BEQ || opcode == RVOpcodes::BNE || 
+        opcode == RVOpcodes::BLT || opcode == RVOpcodes::BGE ||
+        opcode == RVOpcodes::RET || opcode == RVOpcodes::J) {
+        is_def = false;
+    }
+    if (opcode == RVOpcodes::CALL) {
+        // CALL会杀死所有调用者保存寄存器，这是一个简化处理
+        // 同时也使用了传入a0-a7的参数
+    }
+    
+    for (const auto& op : instr->getOperands()) {
+        if (op->getKind() == MachineOperand::KIND_REG) {
+            auto reg_op = static_cast<RegOperand*>(op.get());
+            if (reg_op->isVirtual()) {
+                if (is_def) {
+                    def.insert(reg_op->getVRegNum());
+                    is_def = false; 
+                } else {
+                    use.insert(reg_op->getVRegNum());
+                }
+            }
+        } else if (op->getKind() == MachineOperand::KIND_MEM) {
+            auto mem_op = static_cast<MemOperand*>(op.get());
+            if (mem_op->getBase()->isVirtual()) {
+                use.insert(mem_op->getBase()->getVRegNum());
+            }
+        }
+    }
+}
+
+void RISCv64RegAlloc::analyzeLiveness() {
+    bool changed = true;
+    while (changed) {
+        changed = false;
+        for (auto it = MFunc->getBlocks().rbegin(); it != MFunc->getBlocks().rend(); ++it) {
+            auto& mbb = *it;
+            LiveSet live_out;
+            for (auto succ : mbb->successors) {
+                if (!succ->getInstructions().empty()) {
+                    auto first_instr = succ->getInstructions().front().get();
+                    if (live_in_map.count(first_instr)) {
+                        live_out.insert(live_in_map.at(first_instr).begin(), live_in_map.at(first_instr).end());
+                    }
+                }
+            }
+
+            for (auto instr_it = mbb->getInstructions().rbegin(); instr_it != mbb->getInstructions().rend(); ++instr_it) {
+                MachineInstr* instr = instr_it->get();
+                LiveSet old_live_in = live_in_map[instr];
+                live_out_map[instr] = live_out;
+                
+                LiveSet use, def;
+                getInstrUseDef(instr, use, def);
+
+                LiveSet live_in = use;
+                LiveSet diff = live_out;
+                for (auto vreg : def) {
+                    diff.erase(vreg);
+                }
+                live_in.insert(diff.begin(), diff.end());
+                live_in_map[instr] = live_in;
+                
+                live_out = live_in;
+
+                if (live_in_map[instr] != old_live_in) {
+                    changed = true;
+                }
+            }
+        }
+    }
+}
+
+void RISCv64RegAlloc::buildInterferenceGraph() {
+    std::set<unsigned> all_vregs;
+    for (auto& mbb : MFunc->getBlocks()) {
+        for(auto& instr : mbb->getInstructions()) {
+            LiveSet use, def;
+            getInstrUseDef(instr.get(), use, def);
+            for(auto u : use) all_vregs.insert(u);
+            for(auto d : def) all_vregs.insert(d);
+        }
+    }
+
+    for (auto vreg : all_vregs) { interference_graph[vreg] = {}; }
+
+    for (auto& mbb : MFunc->getBlocks()) {
+        for (auto& instr : mbb->getInstructions()) {
+            LiveSet def, use;
+            getInstrUseDef(instr.get(), use, def);
+            const LiveSet& live_out = live_out_map.at(instr.get());
+
+            for (unsigned d : def) {
+                for (unsigned l : live_out) {
+                    if (d != l) {
+                        interference_graph[d].insert(l);
+                        interference_graph[l].insert(d);
+                    }
+                }
+            }
+        }
+    }
+}
+
+void RISCv64RegAlloc::colorGraph() {
+    std::vector<unsigned> sorted_vregs;
+    for (auto const& [vreg, neighbors] : interference_graph) {
+        sorted_vregs.push_back(vreg);
+    }
+
+    std::sort(sorted_vregs.begin(), sorted_vregs.end(), [&](unsigned a, unsigned b) {
+        return interference_graph[a].size() > interference_graph[b].size();
+    });
+
+    for (unsigned vreg : sorted_vregs) {
+        std::set<PhysicalReg> used_colors;
+        for (unsigned neighbor : interference_graph.at(vreg)) {
+            if (color_map.count(neighbor)) {
+                used_colors.insert(color_map.at(neighbor));
+            }
+        }
+        
+        bool colored = false;
+        for (PhysicalReg preg : allocable_int_regs) {
+            if (used_colors.find(preg) == used_colors.end()) {
+                color_map[vreg] = preg;
+                colored = true;
+                break;
+            }
+        }
+        if (!colored) {
+            spilled_vregs.insert(vreg);
+        }
+    }
+}
+
+void RISCv64RegAlloc::rewriteFunction() {
+    StackFrameInfo& frame_info = MFunc->getFrameInfo();
+    int current_offset = frame_info.locals_size;
+    for (unsigned vreg : spilled_vregs) {
+        current_offset += 4;
+        frame_info.spill_offsets[vreg] = -current_offset;
+    }
+    frame_info.spill_size = current_offset - frame_info.locals_size;
+
+    for (auto& mbb : MFunc->getBlocks()) {
+        std::vector<std::unique_ptr<MachineInstr>> new_instructions;
+        for (auto& instr_ptr : mbb->getInstructions()) {
+            LiveSet use, def;
+            getInstrUseDef(instr_ptr.get(), use, def);
+
+            for (unsigned vreg : use) {
+                if (spilled_vregs.count(vreg)) {
+                    int offset = frame_info.spill_offsets.at(vreg);
+                    auto load = std::make_unique<MachineInstr>(RVOpcodes::LW);
+                    load->addOperand(std::make_unique<RegOperand>(vreg));
+                    load->addOperand(std::make_unique<MemOperand>(
+                        std::make_unique<RegOperand>(PhysicalReg::S0),
+                        std::make_unique<ImmOperand>(offset)
+                    ));
+                    new_instructions.push_back(std::move(load));
+                }
+            }
+
+            new_instructions.push_back(std::move(instr_ptr));
+
+            for (unsigned vreg : def) {
+                if (spilled_vregs.count(vreg)) {
+                    int offset = frame_info.spill_offsets.at(vreg);
+                    auto store = std::make_unique<MachineInstr>(RVOpcodes::SW);
+                    store->addOperand(std::make_unique<RegOperand>(vreg));
+                    store->addOperand(std::make_unique<MemOperand>(
+                        std::make_unique<RegOperand>(PhysicalReg::S0),
+                        std::make_unique<ImmOperand>(offset)
+                    ));
+                    new_instructions.push_back(std::move(store));
+                }
+            }
+        }
+        mbb->getInstructions() = std::move(new_instructions);
+    }
+    
+    for (auto& mbb : MFunc->getBlocks()) {
+        for (auto& instr_ptr : mbb->getInstructions()) {
+            for (auto& op_ptr : instr_ptr->getOperands()) {
+                if(op_ptr->getKind() == MachineOperand::KIND_REG) {
+                    auto reg_op = static_cast<RegOperand*>(op_ptr.get());
+                    if (reg_op->isVirtual()) {
+                        unsigned vreg = reg_op->getVRegNum();
+                        if (color_map.count(vreg)) {
+                            reg_op->setPReg(color_map.at(vreg));
+                        } else if (spilled_vregs.count(vreg)) {
+                            reg_op->setPReg(PhysicalReg::T6); // 溢出统一用t6
+                        }
+                    }
+                } else if (op_ptr->getKind() == MachineOperand::KIND_MEM) {
+                    auto mem_op = static_cast<MemOperand*>(op_ptr.get());
+                    auto base_reg_op = mem_op->getBase();
+                    if(base_reg_op->isVirtual()){
+                         unsigned vreg = base_reg_op->getVRegNum();
+                         if(color_map.count(vreg)) {
+                             base_reg_op->setPReg(color_map.at(vreg));
+                         } else if (spilled_vregs.count(vreg)) {
+                             base_reg_op->setPReg(PhysicalReg::T6);
+                         }
+                    }
+                }
+            }
+        }
+    }
+}
+
+} // namespace sysy

src/Reg2Mem.cpp

@@ -38,7 +38,7 @@ void Reg2Mem::DeletePhiInst(){
               // 创建一个basicblock
               auto newbasicBlock = function.second->addBasicBlock();
               std::stringstream ss;
-              ss << " phidel.L" << pBuilder->getLabelIndex();
+              ss << "phidel.L" << pBuilder->getLabelIndex();
               newbasicBlock->setName(ss.str());
               ss.str("");
               // // 修改前驱后继关系
@@ -103,7 +103,7 @@ void Reg2Mem::DeletePhiInst(){
           }
           // 删除phi指令
           auto &instructions = basicBlock->getInstructions();
-          usedelete(iter->get());
+          SysYIROptUtils::usedelete(iter->get());
           iter = instructions.erase(iter);
           if (basicBlock->getNumInstructions() == 0) {
             if (basicBlock->getNumSuccessors() == 1) {
@@ -119,11 +119,4 @@ void Reg2Mem::DeletePhiInst(){
   }
 }
 
-void Reg2Mem::usedelete(Instruction *instr) {
-  for (auto &use : instr->getOperands()) {
-    auto val = use->getValue();
-    val->removeUse(use);
-  }
-}
-
 }  // namespace sysy

src/SysYIRAnalyser.cpp

@@ -523,9 +523,6 @@ 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

@@ -0,0 +1,565 @@
+#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

@@ -10,12 +10,78 @@
 #include <sstream>
 #include <string>
 #include <vector>
-using namespace std;
 #include "SysYIRGenerator.h"
 
-
+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:
@@ -119,41 +185,111 @@ 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()) {
+
+      if (dims.empty()) { // 标量变量初始化
         builder.createStoreInst(values.getValue(0), alloca);
-      } else {
+      } else { // 数组变量初始化
-        // 对于多维数组，使用memset初始化
+        const std::vector<sysy::Value *> &counterValues = values.getValues();
-        // 计算每个维度的大小
+
-        // 这里的values.getNumbers()返回的是每个维度的大小
+        int numElements = 1;
-        // 这里的values.getValues()返回的是每个维度对应的值
+        std::vector<int> dimSizes;
-        // 例如：对于一个二维数组，values.getNumbers()可能是[3, 4]，表示3行4列
+        for (Value *dimVal : dims) {
-        // values.getValues()可能是[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12]
+          if (ConstantInteger *constInt = dynamic_cast<ConstantInteger *>(dimVal)) {
-        // 对于每个维度，使用memset将对应的值填充到数组中
+            int dimSize = constInt->getInt();
-        // 这里的alloca是一个指向数组的指针
+            numElements *= dimSize;
-        const std::vector<unsigned int> & counterNumbers = values.getNumbers();
+            dimSizes.push_back(dimSize);
-        const std::vector<sysy::Value *> & counterValues = values.getValues();
+          }
-        unsigned begin = 0;
+          // TODO else 错误处理：数组维度必须是常量（对于静态分配）
-        for (size_t i = 0; i < counterNumbers.size(); i++) {
+        }
+        unsigned int elementSizeInBytes = type->getSize();
+        unsigned int totalSizeInBytes = numElements * elementSizeInBytes;
+
+        bool allValuesAreZero = false;
+        if (counterValues.empty()) {
+          allValuesAreZero = true;
+        }
+        else {
+          allValuesAreZero = true;
+          for (Value *val : counterValues){
+            if (ConstantInteger *constInt = dynamic_cast<ConstantInteger *>(val)){
+              if (constInt->getInt() != 0){
+                allValuesAreZero = false;
+                break;
+              }
+            }
+            else{
+              allValuesAreZero = false;
+              break;
+            }
+          }
+        }
+
+        if (allValuesAreZero) {
+          builder.createMemsetInst(
+              alloca,
+              ConstantInteger::get(0),
+              ConstantInteger::get(totalSizeInBytes),
+              ConstantInteger::get(0));
+        }
+        else {
+          for (size_t k = 0; k < counterValues.size(); ++k) {
+            std::vector<Value *> currentIndices;
+            int tempLinearIndex = k;
+
+            // 将线性索引转换为多维索引
+            for (int dimIdx = dimSizes.size() - 1; dimIdx >= 0; --dimIdx)
+            {
+              currentIndices.insert(currentIndices.begin(),
+                                    ConstantInteger::get(static_cast<int>(tempLinearIndex % dimSizes[dimIdx])));
+              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);
+          }
+        }
+      }
+    }
+    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)) {
+            numElements *= constInt->getInt();
+          }
+        }
+        unsigned int elementSizeInBytes = type->getSize();
+        unsigned int totalSizeInBytes = numElements * elementSizeInBytes;
 
         builder.createMemsetInst(
-              alloca, ConstantValue::get(static_cast<int>(begin)),
+            alloca,
-              ConstantValue::get(static_cast<int>(counterNumbers[i])),
+            ConstantInteger::get(0),
-              counterValues[i]);
+            ConstantInteger::get(totalSizeInBytes),
-          begin += counterNumbers[i];
+            ConstantInteger::get(0)
-        }
+          );
       }
     }
+
     module->addVariable(name, alloca);
   }
+
   return std::any();
 }
 
@@ -204,7 +340,6 @@ 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;
@@ -218,7 +353,7 @@ std::any SysYIRGenerator::visitFuncDef(SysYParser::FuncDefContext *ctx){
       paramNames.push_back(param->Ident()->getText());
       std::vector<Value *> dims = {};
       if (!param->LBRACK().empty()) {
-        dims.push_back(ConstantValue::get(-1)); // 第一个维度不确定
+        dims.push_back(ConstantInteger::get(-1)); // 第一个维度不确定
         for (const auto &exp : param->exp()) {
           dims.push_back(std::any_cast<Value *>(visitExp(exp)));
         }
@@ -240,22 +375,34 @@ 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);
   }
 
-  if(HasReturnInst == false) {
+  // 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
-    // 如果没有return语句，则默认返回0
+  ReturnInst* retinst = nullptr;
-    if (returnType != Type::getVoidType()) {
+  retinst = dynamic_cast<ReturnInst*>(builder.getBasicBlock()->terminator()->get());
-      Value* returnValue = ConstantValue::get(0);
+
-      if (returnType == Type::getFloatType()) {
+  if (!retinst) {
-        returnValue = ConstantValue::get(0.0f);
+    if (returnType->isVoid()) {
-      }
-      builder.createReturnInst(returnValue);
-    } else {
         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
+    } else {
+        assert(false && "Function with no explicit return and non-void type should return a value.");
     }
   }
+
   module->leaveScope();
 
   return std::any();
@@ -277,28 +424,55 @@ std::any SysYIRGenerator::visitAssignStmt(SysYParser::AssignStmtContext *ctx) {
     dims.push_back(std::any_cast<Value *>(visitExp(exp)));
   }
   
-  auto variable = module->getVariable(name);
+  auto variable = module->getVariable(name); // 获取 AllocaInst 或 GlobalValue
-  Value* value = std::any_cast<Value *>(visitExp(ctx->exp()));
+  Value* value = std::any_cast<Value *>(visitExp(ctx->exp())); // 右值
-  Type* variableType = dynamic_cast<PointerType *>(variable->getType())->getBaseType();
 
-  // 左值右值类型不同处理
+  if (variable == nullptr) {
-  if (variableType != value->getType()) {
+      throw std::runtime_error("Variable " + name + " not found in assignment.");
+  }
+
+  // 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()) {
     ConstantValue * constValue = dynamic_cast<ConstantValue *>(value);
     if (constValue != nullptr) {
-      if (variableType == Type::getFloatType()) {
+      if (targetElementType == Type::getFloatType()) {
-        value = ConstantValue::get(static_cast<float>(constValue->getInt()));
+        value = ConstantFloating::get(static_cast<float>(constValue->getInt()));
-      } else {
+      } else { // 假设如果不是浮点型，就是整型
-        value = ConstantValue::get(static_cast<int>(constValue->getFloat()));
+        value = ConstantInteger::get(static_cast<int>(constValue->getFloat()));
       }
     } else {
-      if (variableType == Type::getFloatType()) {
+      if (targetElementType == Type::getFloatType()) {
         value = builder.createIToFInst(value);
-      } else {
+      } else { // 假设如果不是浮点型，就是整型
         value = builder.createFtoIInst(value);
       }
     }
   }
-  builder.createStoreInst(value, variable, dims, variable->getName());
+
+  // 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);
 
   return std::any();
 }
@@ -386,6 +560,7 @@ 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();
@@ -407,6 +582,7 @@ 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());
 
@@ -478,9 +654,9 @@ std::any SysYIRGenerator::visitReturnStmt(SysYParser::ReturnStmtContext *ctx) {
       ConstantValue * constValue = dynamic_cast<ConstantValue *>(returnValue);
       if (constValue != nullptr) {
         if (funcType == Type::getFloatType()) {
-          returnValue = ConstantValue::get(static_cast<float>(constValue->getInt()));
+          returnValue = ConstantInteger::get(static_cast<float>(constValue->getInt()));
         } else {
-          returnValue = ConstantValue::get(static_cast<int>(constValue->getFloat()));
+          returnValue = ConstantFloating::get(static_cast<int>(constValue->getFloat()));
         }
       } else {
         if (funcType == Type::getFloatType()) {
@@ -491,56 +667,94 @@ 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)));
   }
 
-  if (variable == nullptr) {
+  // 1. 获取变量的声明维度数量
-    throw std::runtime_error("Variable " + name + " not found.");
+  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();
   }
 
-  bool indicesConstant = true;
+  // 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) {
-      indicesConstant = false;
+        allIndicesConstant = false;
         break;
       }
     }
-
+    if (allIndicesConstant) {
-  ConstantVariable* constVar = dynamic_cast<ConstantVariable *>(variable);
+      // 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
-  GlobalValue* globalVar = dynamic_cast<GlobalValue *>(variable);
+      // 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)
-  AllocaInst* localVar = dynamic_cast<AllocaInst *>(variable);
+      return constVar->getByIndices(dims);
-  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 {
-    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 {
-      value = builder.createLoadInst(variable, 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);
+    }
+  } 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);
+    } 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);
+    }
+  }
   return value;
 }
 
@@ -560,10 +774,10 @@ std::any SysYIRGenerator::visitPrimaryExp(SysYParser::PrimaryExpContext *ctx) {
 std::any SysYIRGenerator::visitNumber(SysYParser::NumberContext *ctx) {
   if (ctx->ILITERAL() != nullptr) {
     int value = std::stol(ctx->ILITERAL()->getText(), nullptr, 0);
-    return static_cast<Value *>(ConstantValue::get(value));
+    return static_cast<Value *>(ConstantInteger::get(value));
   } else if (ctx->FLITERAL() != nullptr) {
     float value = std::stof(ctx->FLITERAL()->getText());
-    return static_cast<Value *>(ConstantValue::get(value));
+    return static_cast<Value *>(ConstantFloating::get(value));
   }
   throw std::runtime_error("Unknown number type.");
   return std::any(); // 不会到达这里
@@ -599,9 +813,9 @@ std::any SysYIRGenerator::visitCall(SysYParser::CallContext *ctx) {
         ConstantValue * constValue = dynamic_cast<ConstantValue *>(args[i]);
         if (constValue != nullptr) {
           if (params[i]->getType() == Type::getPointerType(Type::getFloatType())) {
-            args[i] = ConstantValue::get(static_cast<float>(constValue->getInt()));
+            args[i] = ConstantInteger::get(static_cast<float>(constValue->getInt()));
           } else {
-            args[i] = ConstantValue::get(static_cast<int>(constValue->getFloat()));
+            args[i] = ConstantFloating::get(static_cast<int>(constValue->getFloat()));
           }
         } else {
           if (params[i]->getType() == Type::getPointerType(Type::getFloatType())) {
@@ -629,9 +843,9 @@ std::any SysYIRGenerator::visitUnaryExp(SysYParser::UnaryExpContext *ctx) {
     ConstantValue * constValue = dynamic_cast<ConstantValue *>(value);
     if (constValue != nullptr) {
       if (constValue->isFloat()) {
-        result = ConstantValue::get(-constValue->getFloat());
+        result = ConstantFloating::get(-constValue->getFloat());
       } else {
-        result = ConstantValue::get(-constValue->getInt());
+        result = ConstantInteger::get(-constValue->getInt());
       }
     } else if (value != nullptr) {
       if (value->getType() == Type::getIntType()) {
@@ -648,9 +862,9 @@ std::any SysYIRGenerator::visitUnaryExp(SysYParser::UnaryExpContext *ctx) {
     if (constValue != nullptr) {
       if (constValue->isFloat()) {
         result =
-            ConstantValue::get(1 - (constValue->getFloat() != 0.0F ? 1 : 0));
+            ConstantFloating::get(1 - (constValue->getFloat() != 0.0F ? 1 : 0));
       } else {
-        result = ConstantValue::get(1 - (constValue->getInt() != 0 ? 1 : 0));
+        result = ConstantInteger::get(1 - (constValue->getInt() != 0 ? 1 : 0));
       }
     } else if (value != nullptr) {
       if (value->getType() == Type::getIntType()) {
@@ -692,13 +906,13 @@ std::any SysYIRGenerator::visitMulExp(SysYParser::MulExpContext *ctx) {
       if (operandType != floatType) {
         ConstantValue * constValue = dynamic_cast<ConstantValue *>(operand);
         if (constValue != nullptr) 
-          operand = ConstantValue::get(static_cast<float>(constValue->getInt()));
+          operand = ConstantFloating::get(static_cast<float>(constValue->getInt()));
         else 
           operand = builder.createIToFInst(operand);
       } else if (resultType != floatType) {
         ConstantValue* constResult = dynamic_cast<ConstantValue *>(result);
         if (constResult != nullptr) 
-          result = ConstantValue::get(static_cast<float>(constResult->getInt()));
+          result = ConstantFloating::get(static_cast<float>(constResult->getInt()));
          else 
           result = builder.createIToFInst(result);
       }
@@ -707,14 +921,14 @@ std::any SysYIRGenerator::visitMulExp(SysYParser::MulExpContext *ctx) {
       ConstantValue* constOperand = dynamic_cast<ConstantValue *>(operand);
       if (opType == SysYParser::MUL) {
         if ((constOperand != nullptr) && (constResult != nullptr)) {
-          result = ConstantValue::get(constResult->getFloat() *
+          result = ConstantFloating::get(constResult->getFloat() *
                                       constOperand->getFloat());
         } else {
           result = builder.createFMulInst(result, operand);
         }
       } else if (opType == SysYParser::DIV) {
         if ((constOperand != nullptr) && (constResult != nullptr)) {
-          result = ConstantValue::get(constResult->getFloat() /
+          result = ConstantFloating::get(constResult->getFloat() /
                                       constOperand->getFloat());
         } else {
           result = builder.createFDivInst(result, operand);
@@ -729,17 +943,17 @@ std::any SysYIRGenerator::visitMulExp(SysYParser::MulExpContext *ctx) {
       ConstantValue * constOperand = dynamic_cast<ConstantValue *>(operand);
       if (opType == SysYParser::MUL) {
         if ((constOperand != nullptr) && (constResult != nullptr)) 
-          result = ConstantValue::get(constResult->getInt() * constOperand->getInt());
+          result = ConstantInteger::get(constResult->getInt() * constOperand->getInt());
         else 
           result = builder.createMulInst(result, operand);
       } else if (opType == SysYParser::DIV) {
         if ((constOperand != nullptr) && (constResult != nullptr)) 
-          result = ConstantValue::get(constResult->getInt() / constOperand->getInt());
+          result = ConstantInteger::get(constResult->getInt() / constOperand->getInt());
          else 
           result = builder.createDivInst(result, operand);
       } else {
         if ((constOperand != nullptr) && (constResult != nullptr)) 
-          result = ConstantValue::get(constResult->getInt() % constOperand->getInt());
+          result = ConstantInteger::get(constResult->getInt() % constOperand->getInt());
          else 
           result = builder.createRemInst(result, operand);
       }
@@ -767,13 +981,13 @@ std::any SysYIRGenerator::visitAddExp(SysYParser::AddExpContext *ctx) {
       if (operandType != floatType) {
         ConstantValue * constOperand = dynamic_cast<ConstantValue *>(operand);
         if (constOperand != nullptr) 
-          operand = ConstantValue::get(static_cast<float>(constOperand->getInt()));
+          operand = ConstantFloating::get(static_cast<float>(constOperand->getInt()));
         else 
           operand = builder.createIToFInst(operand);
       } else if (resultType != floatType) {
         ConstantValue * constResult = dynamic_cast<ConstantValue *>(result);
         if (constResult != nullptr) 
-          result = ConstantValue::get(static_cast<float>(constResult->getInt()));
+          result = ConstantFloating::get(static_cast<float>(constResult->getInt()));
         else 
           result = builder.createIToFInst(result);
       }
@@ -782,12 +996,12 @@ std::any SysYIRGenerator::visitAddExp(SysYParser::AddExpContext *ctx) {
       ConstantValue *  constOperand = dynamic_cast<ConstantValue *>(operand);
       if (opType == SysYParser::ADD) {
         if ((constResult != nullptr) && (constOperand != nullptr))
-          result = ConstantValue::get(constResult->getFloat() + constOperand->getFloat());
+          result = ConstantFloating::get(constResult->getFloat() + constOperand->getFloat());
         else 
           result = builder.createFAddInst(result, operand);
       } else {
         if ((constResult != nullptr) && (constOperand != nullptr)) 
-          result = ConstantValue::get(constResult->getFloat() - constOperand->getFloat());
+          result = ConstantFloating::get(constResult->getFloat() - constOperand->getFloat());
         else
           result = builder.createFSubInst(result, operand);
       }
@@ -796,12 +1010,12 @@ std::any SysYIRGenerator::visitAddExp(SysYParser::AddExpContext *ctx) {
       ConstantValue *  constOperand = dynamic_cast<ConstantValue *>(operand);
       if (opType == SysYParser::ADD) {
         if ((constResult != nullptr) && (constOperand != nullptr))
-          result = ConstantValue::get(constResult->getInt() + constOperand->getInt());
+          result = ConstantInteger::get(constResult->getInt() + constOperand->getInt());
         else 
           result = builder.createAddInst(result, operand);
       } else {
         if ((constResult != nullptr) && (constOperand != nullptr))
-          result = ConstantValue::get(constResult->getInt() - constOperand->getInt());
+          result = ConstantInteger::get(constResult->getInt() - constOperand->getInt());
         else 
           result = builder.createSubInst(result, operand);
       }
@@ -833,10 +1047,10 @@ std::any SysYIRGenerator::visitRelExp(SysYParser::RelExpContext *ctx) {
       auto operand2 = constOperand->isFloat() ? constOperand->getFloat()
                                               : constOperand->getInt();
 
-      if (opType == SysYParser::LT)      result = ConstantValue::get(operand1 < operand2 ? 1 : 0);
+      if (opType == SysYParser::LT)      result = ConstantInteger::get(operand1 < operand2 ? 1 : 0);
-      else if (opType == SysYParser::GT) result = ConstantValue::get(operand1 > operand2 ? 1 : 0);
+      else if (opType == SysYParser::GT) result = ConstantInteger::get(operand1 > operand2 ? 1 : 0);
-      else if (opType == SysYParser::LE) result = ConstantValue::get(operand1 <= operand2 ? 1 : 0);
+      else if (opType == SysYParser::LE) result = ConstantInteger::get(operand1 <= operand2 ? 1 : 0);
-      else if (opType == SysYParser::GE) result = ConstantValue::get(operand1 >= operand2 ? 1 : 0);
+      else if (opType == SysYParser::GE) result = ConstantInteger::get(operand1 >= operand2 ? 1 : 0);
       else                          assert(false);
 
     } else {
@@ -848,14 +1062,14 @@ std::any SysYIRGenerator::visitRelExp(SysYParser::RelExpContext *ctx) {
       if (resultType == floatType || operandType == floatType) {
         if (resultType != floatType) {
           if (constResult != nullptr)
-            result = ConstantValue::get(static_cast<float>(constResult->getInt()));
+            result = ConstantFloating::get(static_cast<float>(constResult->getInt()));
           else 
             result = builder.createIToFInst(result);
           
         }
         if (operandType != floatType) {
           if (constOperand != nullptr) 
-            operand = ConstantValue::get(static_cast<float>(constOperand->getInt()));
+            operand = ConstantFloating::get(static_cast<float>(constOperand->getInt()));
           else 
             operand = builder.createIToFInst(operand);
           
@@ -901,8 +1115,8 @@ std::any SysYIRGenerator::visitEqExp(SysYParser::EqExpContext *ctx) {
       auto operand2 = constOperand->isFloat() ? constOperand->getFloat()
                                               : constOperand->getInt();
 
-      if (opType == SysYParser::EQ)      result = ConstantValue::get(operand1 == operand2 ? 1 : 0);
+      if (opType == SysYParser::EQ)      result = ConstantInteger::get(operand1 == operand2 ? 1 : 0);
-      else if (opType == SysYParser::NE) result = ConstantValue::get(operand1 != operand2 ? 1 : 0);
+      else if (opType == SysYParser::NE) result = ConstantInteger::get(operand1 != operand2 ? 1 : 0);
       else                          assert(false);
       
     } else {
@@ -913,13 +1127,13 @@ std::any SysYIRGenerator::visitEqExp(SysYParser::EqExpContext *ctx) {
       if (resultType == floatType || operandType == floatType) {
         if (resultType != floatType) {
           if (constResult != nullptr)
-            result = ConstantValue::get(static_cast<float>(constResult->getInt()));
+            result = ConstantFloating::get(static_cast<float>(constResult->getInt()));
           else 
             result = builder.createIToFInst(result);
         }
         if (operandType != floatType) {
           if (constOperand != nullptr) 
-            operand = ConstantValue::get(static_cast<float>(constOperand->getInt()));
+            operand = ConstantFloating::get(static_cast<float>(constOperand->getInt()));
           else
             operand = builder.createIToFInst(operand);
         }
@@ -943,9 +1157,9 @@ std::any SysYIRGenerator::visitEqExp(SysYParser::EqExpContext *ctx) {
     // 如果只有一个关系表达式，则将结果转换为0或1
     if (constResult != nullptr) {
       if (constResult->isFloat()) 
-        result = ConstantValue::get(constResult->getFloat() != 0.0F ? 1 : 0);
+        result = ConstantInteger::get(constResult->getFloat() != 0.0F ? 1 : 0);
       else
-        result = ConstantValue::get(constResult->getInt() != 0 ? 1 : 0);
+        result = ConstantInteger::get(constResult->getInt() != 0 ? 1 : 0);
     }
   }
 
@@ -1013,6 +1227,7 @@ void Utils::tree2Array(Type *type, ArrayValueTree *root,
                         ValueCounter &result, IRBuilder *builder) {
   Value* value = root->getValue();
   auto &children = root->getChildren();
+  // 类型转换
   if (value != nullptr) {
     if (type == value->getType()) {
       result.push_back(value);
@@ -1020,14 +1235,14 @@ void Utils::tree2Array(Type *type, ArrayValueTree *root,
       if (type == Type::getFloatType()) {
         ConstantValue* constValue = dynamic_cast<ConstantValue *>(value);
         if (constValue != nullptr)
-          result.push_back(ConstantValue::get(static_cast<float>(constValue->getInt())));
+          result.push_back(ConstantFloating::get(static_cast<float>(constValue->getInt())));
         else 
           result.push_back(builder->createIToFInst(value));
         
       } else {
         ConstantValue* constValue = dynamic_cast<ConstantValue *>(value);
         if (constValue != nullptr)
-          result.push_back(ConstantValue::get(static_cast<int>(constValue->getFloat())));
+          result.push_back(ConstantInteger::get(static_cast<int>(constValue->getFloat())));
         else 
           result.push_back(builder->createFtoIInst(value));
         
@@ -1061,9 +1276,9 @@ void Utils::tree2Array(Type *type, ArrayValueTree *root,
   int num = blockSize - afterSize + beforeSize;
   if (num > 0) {
     if (type == Type::getFloatType())
-      result.push_back(ConstantValue::get(0.0F), num);
+      result.push_back(ConstantFloating::get(0.0F), num);
     else 
-      result.push_back(ConstantValue::get(0), num);
+      result.push_back(ConstantInteger::get(0), num);
   }
 }
 
@@ -1101,7 +1316,7 @@ void Utils::initExternalFunction(Module *pModule, IRBuilder *pBuilder) {
                                  funcName, pModule, pBuilder);
   paramTypes.push_back(Type::getIntType());
   paramNames.emplace_back("x");
-  paramDims.push_back(std::vector<Value *>{ConstantValue::get(-1)});
+  paramDims.push_back(std::vector<Value *>{ConstantInteger::get(-1)});
   funcName = "getarray";
   Utils::createExternalFunction(paramTypes, paramNames, paramDims, returnType,
                                  funcName, pModule, pBuilder);
@@ -1117,7 +1332,7 @@ void Utils::initExternalFunction(Module *pModule, IRBuilder *pBuilder) {
   returnType = Type::getIntType();
   paramTypes.push_back(Type::getFloatType());
   paramNames.emplace_back("x");
-  paramDims.push_back(std::vector<Value *>{ConstantValue::get(-1)});
+  paramDims.push_back(std::vector<Value *>{ConstantInteger::get(-1)});
   funcName = "getfarray";
   Utils::createExternalFunction(paramTypes, paramNames, paramDims, returnType,
                                  funcName, pModule, pBuilder);
@@ -1141,7 +1356,7 @@ void Utils::initExternalFunction(Module *pModule, IRBuilder *pBuilder) {
   paramTypes.push_back(Type::getIntType());
   paramDims.clear();
   paramDims.emplace_back();
-  paramDims.push_back(std::vector<Value *>{ConstantValue::get(-1)});
+  paramDims.push_back(std::vector<Value *>{ConstantInteger::get(-1)});
   paramNames.clear();
   paramNames.emplace_back("n");
   paramNames.emplace_back("a");
@@ -1164,7 +1379,7 @@ void Utils::initExternalFunction(Module *pModule, IRBuilder *pBuilder) {
   paramTypes.push_back(Type::getFloatType());
   paramDims.clear();
   paramDims.emplace_back();
-  paramDims.push_back(std::vector<Value *>{ConstantValue::get(-1)});
+  paramDims.push_back(std::vector<Value *>{ConstantInteger::get(-1)});
   paramNames.clear();
   paramNames.emplace_back("n");
   paramNames.emplace_back("a");

src/SysYIROptPre.cpp

@@ -1,484 +0,0 @@
-#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(ConstantValue::get(0));
-          } else if (func->getReturnType()->isFloat()) {
-            pBuilder->createReturnInst(ConstantValue::get(0.0F));
-          } else {
-            pBuilder->createReturnInst();
-          }
-        }
-      }
-    }
-  }
-}
-
-}  // namespace sysy

src/SysYIRPassManager.cpp

@@ -0,0 +1,36 @@
+// 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

@@ -3,12 +3,11 @@
 #include <fstream>
 #include <iostream>
 #include <string>
-#include "IR.h"
+#include "IR.h" // 确保IR.h包含了ArrayType、GetElementPtrInst等的定义
 
 namespace sysy {
 
 void SysYPrinter::printIR() {
-
   const auto &functions = pModule->getFunctions();
 
   //TODO: Print target datalayout and triple (minimal required by LLVM)
@@ -36,11 +35,18 @@ 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 ptrType = dynamic_cast<FunctionType*>(type)) {
+  } else if (auto funcType = dynamic_cast<FunctionType*>(type)) {
-    return getTypeString(ptrType->getReturnType());
+    // 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()) + "]";
   }
   assert(false && "Unsupported type");
   return "";
@@ -51,15 +57,23 @@ std::string SysYPrinter::getValueName(Value *value) {
     return "@" + global->getName();
   } else if (auto inst = dynamic_cast<Instruction*>(value)) {
     return "%" + inst->getName();
-  } else if (auto constVal = dynamic_cast<ConstantValue*>(value)) {
+  } else if (auto constInt = dynamic_cast<ConstantInteger*>(value)) { // 优先匹配具体的常量类型
-    if (constVal->isFloat()) {
+    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());
     }
     return std::to_string(constVal->getInt());
   } else if (auto constVar = dynamic_cast<ConstantVariable*>(value)) {
-    return constVar->getName();
+    return constVar->getName(); // 假设ConstantVariable有自己的名字或通过getByIndices获取值
   }
-  assert(false && "Unknown value type");
+  assert(false && "Unknown value type or unable to get value name");
   return "";
 }
 
@@ -77,44 +91,35 @@ void SysYPrinter::printGlobalVariable() {
   for (const auto &global : globals) {
     std::cout << "@" << global->getName() << " = global ";
     
-    auto baseType = dynamic_cast<PointerType *>(global->getType())->getBaseType();
+    // 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)
-    printType(baseType);
+    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])
-    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 << " ";
     
-    if (global->getNumDims() > 0) {
+    // 检查是否是数组类型 (通过检查 globalVarBaseType 是否是 ArrayType)
-      // Array initializer
+    if (globalVarBaseType->isArray()) {
-      std::cout << "[";
+      // 数组初始化器
-      auto values = global->getInitValues();
+      std::cout << "["; // LLVM IR 数组初始化器格式: [type value, type value, ...]
-      auto counterValues = values.getValues();
+      auto values = global->getInitValues(); // 假设 getInitValues() 返回一个 ValueCounter
-      auto counterNumbers = values.getNumbers();
+      const std::vector<sysy::Value *> &counterValues = values.getValues(); // 获取所有值
       
-      for (size_t i = 0; i < counterNumbers.size(); i++) {
+      for (size_t i = 0; i < counterValues.size(); i++) {
         if (i > 0) std::cout << ", ";
-        if (baseType->isFloat()) {
+        // 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
-          std::cout << "float " << dynamic_cast<ConstantValue*>(counterValues[i])->getFloat();
+        // 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
-        } else {
+        printType(counterValues[i]->getType());
-          std::cout << "i32 " << dynamic_cast<ConstantValue*>(counterValues[i])->getInt();
+        std::cout << " ";
-        }
+        printValue(counterValues[i]);
       }
       std::cout << "]";
     } else {
-      // Scalar initializer
+      // 标量初始化器
-      if (baseType->isFloat()) {
+      // 假设标量全局变量的初始化值通过 getByIndex(0) 获取
-        std::cout << "float " << dynamic_cast<ConstantValue*>(global->getByIndex(0))->getFloat();
+      Value* initVal = global->getByIndex(0);
-      } else {
+      printType(initVal->getType()); // 打印标量值的类型
-        std::cout << "i32 " << dynamic_cast<ConstantValue*>(global->getByIndex(0))->getInt();
+      std::cout << " ";
-      }
+      printValue(initVal); // 打印标量值
     }
     
     std::cout << ", align 4" << std::endl;
@@ -209,19 +214,19 @@ 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;
+        case Kind::kICmpLT: std::cout << "icmp slt"; break; // LLVM uses slt/sgt for signed less/greater than
         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;
+        case Kind::kFCmpEQ: std::cout << "fcmp oeq"; break; // oeq for ordered equal
-        case Kind::kFCmpNE: std::cout << "fcmp one"; break;
+        case Kind::kFCmpNE: std::cout << "fcmp one"; break; // one for ordered not equal
-        case Kind::kFCmpLT: std::cout << "fcmp olt"; break;
+        case Kind::kFCmpLT: std::cout << "fcmp olt"; break; // olt for ordered less than
-        case Kind::kFCmpGT: std::cout << "fcmp ogt"; break;
+        case Kind::kFCmpGT: std::cout << "fcmp ogt"; break; // ogt for ordered greater than
-        case Kind::kFCmpLE: std::cout << "fcmp ole"; break;
+        case Kind::kFCmpLE: std::cout << "fcmp ole"; break; // ole for ordered less than or equal
-        case Kind::kFCmpGE: std::cout << "fcmp oge"; break;
+        case Kind::kFCmpGE: std::cout << "fcmp oge"; break; // oge for ordered greater than or equal
         case Kind::kAnd: std::cout << "and"; break;
         case Kind::kOr: std::cout << "or"; break;
-        default: break;
+        default: break; // Should not reach here
       }
       
       // Types and operands
@@ -238,7 +243,6 @@ 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:
@@ -250,31 +254,39 @@ void SysYPrinter::printInst(Instruction *pInst) {
       }
       
       switch (pInst->getKind()) {
-        case Kind::kNeg: std::cout << "sub "; break;
+        case Kind::kNeg: std::cout << "sub "; break; // integer negation is `sub i32 0, operand`
-        case Kind::kNot: std::cout << "not "; break;
+        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;
+        case Kind::kFNeg: std::cout << "fneg "; break; // float negation
-        case Kind::kFNot: std::cout << "fneg "; break; // FNot not standard, map to fneg
+        case Kind::kFtoI: std::cout << "fptosi "; break; // float to signed integer
-        case Kind::kFtoI: std::cout << "fptosi "; break;
+        case Kind::kBitFtoI: std::cout << "bitcast "; break; // bitcast float to int
-        case Kind::kBitFtoI: std::cout << "bitcast "; break;
+        case Kind::kItoF: std::cout << "sitofp "; break; // signed integer to float
-        case Kind::kItoF: std::cout << "sitofp "; break;
+        case Kind::kBitItoF: std::cout << "bitcast "; break; // bitcast int to float
-        case Kind::kBitItoF: std::cout << "bitcast "; break;
+        default: break; // Should not reach here
-        default: break;
       }
       
-      printType(unyInst->getType());
+      printType(unyInst->getOperand()->getType()); // Print operand type
       std::cout << " ";
       
-      // Special handling for negation
+      // Special handling for integer negation and logical NOT
-      if (pInst->getKind() == Kind::kNeg || pInst->getKind() == Kind::kNot) {
+      if (pInst->getKind() == Kind::kNeg) {
-        std::cout << "i32 0, ";
+        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, ";
+        }
       }
       
       printValue(pInst->getOperand(0));
       
-      // For bitcast, need to specify destination type
+      // For type conversions (fptosi, sitofp, bitcast), need to specify destination type
-      if (pInst->getKind() == Kind::kBitFtoI || pInst->getKind() == Kind::kBitItoF) {
+      if (pInst->getKind() == Kind::kFtoI || pInst->getKind() == Kind::kItoF ||
+          pInst->getKind() == Kind::kBitFtoI || pInst->getKind() == Kind::kBitItoF) {
         std::cout << " to ";
-        printType(unyInst->getType());
+        printType(unyInst->getType()); // Print result type
       }
       
       std::cout << std::endl;
@@ -289,7 +301,7 @@ void SysYPrinter::printInst(Instruction *pInst) {
       }
       
       std::cout << "call ";
-      printType(callInst->getType());
+      printType(callInst->getType()); // Return type of the call
       std::cout << " @" << function->getName() << "(";
       
       auto params = callInst->getArguments();
@@ -297,9 +309,9 @@ void SysYPrinter::printInst(Instruction *pInst) {
       for (auto &param : params) {
         if (!first) std::cout << ", ";
         first = false;
-        printType(param->getValue()->getType());
+        printType(param->getValue()->getType()); // Type of argument
         std::cout << " ";
-        printValue(param->getValue());
+        printValue(param->getValue()); // Value of argument
       }
       
       std::cout << ")" << std::endl;
@@ -307,7 +319,7 @@ void SysYPrinter::printInst(Instruction *pInst) {
     
     case Kind::kCondBr: {
       auto condBrInst = dynamic_cast<CondBrInst *>(pInst);
-      std::cout << "br i1 ";
+      std::cout << "br i1 "; // Condition type should be i1
       printValue(condBrInst->getCondition());
       std::cout << ", label %" << condBrInst->getThenBlock()->getName();
       std::cout << ", label %" << condBrInst->getElseBlock()->getName();
@@ -337,14 +349,17 @@ void SysYPrinter::printInst(Instruction *pInst) {
       auto allocaInst = dynamic_cast<AllocaInst *>(pInst);
       std::cout << "%" << allocaInst->getName() << " = alloca ";
       
-      auto baseType = dynamic_cast<PointerType *>(allocaInst->getType())->getBaseType();
+      // AllocaInst 的类型现在应该是一个 PointerType，指向正确的 ArrayType 或 ScalarType
-      printType(baseType);
+      // 例如：alloca i32, align 4  或者 alloca [10 x i32], align 4
+      auto allocatedType = dynamic_cast<PointerType *>(allocaInst->getType())->getBaseType();
+      printType(allocatedType);
       
+      // 仍然打印维度信息，如果存在的话
       if (allocaInst->getNumDims() > 0) { 
         std::cout << ", ";
         for (size_t i = 0; i < allocaInst->getNumDims(); i++) {
           if (i > 0) std::cout << ", ";
-          printType(Type::getIntType());
+          printType(Type::getIntType()); // 维度大小通常是 i32 类型
           std::cout << " ";
           printValue(allocaInst->getDim(i));
         }
@@ -356,17 +371,18 @@ void SysYPrinter::printInst(Instruction *pInst) {
     case Kind::kLoad: {
       auto loadInst = dynamic_cast<LoadInst *>(pInst);
       std::cout << "%" << loadInst->getName() << " = load ";
-      printType(loadInst->getType());
+      printType(loadInst->getType()); // 加载的结果类型
       std::cout << ", ";
-      printType(loadInst->getPointer()->getType());
+      printType(loadInst->getPointer()->getType()); // 指针类型
       std::cout << " ";
-      printValue(loadInst->getPointer());
+      printValue(loadInst->getPointer()); // 要加载的地址
       
+      // 仍然打印索引信息，如果存在的话
       if (loadInst->getNumIndices() > 0) {
-        std::cout << ", ";
+        std::cout << ", indices "; // 或者其他分隔符，取决于你期望的格式
         for (size_t i = 0; i < loadInst->getNumIndices(); i++) {
             if (i > 0) std::cout << ", ";
-          printType(Type::getIntType());
+            printType(loadInst->getIndex(i)->getType());
             std::cout << " ";
             printValue(loadInst->getIndex(i));
         }
@@ -375,44 +391,23 @@ void SysYPrinter::printInst(Instruction *pInst) {
       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());
+      printType(storeInst->getValue()->getType()); // 要存储的值的类型
       std::cout << " ";
-      printValue(storeInst->getValue());
+      printValue(storeInst->getValue()); // 要存储的值
       std::cout << ", ";
-      printType(storeInst->getPointer()->getType());
+      printType(storeInst->getPointer()->getType()); // 目标指针的类型
       std::cout << " ";
-      printValue(storeInst->getPointer());
+      printValue(storeInst->getPointer()); // 目标地址
       
+      // 仍然打印索引信息，如果存在的话
       if (storeInst->getNumIndices() > 0) {
-        std::cout << ", ";
+        std::cout << ", indices "; // 或者其他分隔符
         for (size_t i = 0; i < storeInst->getNumIndices(); i++) {
             if (i > 0) std::cout << ", ";
-          printType(Type::getIntType());
+            printType(storeInst->getIndex(i)->getType());
             std::cout << " ";
             printValue(storeInst->getIndex(i));
         }
@@ -421,6 +416,30 @@ void SysYPrinter::printInst(Instruction *pInst) {
       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);
       std::cout << "call void @llvm.memset.p0.";
@@ -433,49 +452,38 @@ void SysYPrinter::printInst(Instruction *pInst) {
       printValue(memsetInst->getValue());
       std::cout << ", i32 ";
       printValue(memsetInst->getSize());
-      std::cout << ", i1 false)" << std::endl;
+      std::cout << ", i1 false)" << std::endl; // alignment for memset is typically i1
     } break;
     
     case Kind::kPhi: {
       auto phiInst = dynamic_cast<PhiInst *>(pInst);
-      printValue(phiInst->getOperand(0));
+      // Phi 指令的名称通常是结果变量
-      std::cout << " = phi ";
+      std::cout << "%" << phiInst->getName() << " = phi ";
-      printType(phiInst->getType());
+      printType(phiInst->getType()); // Phi 结果类型
       
-      for (unsigned i = 1; i < phiInst->getNumOperands(); i++) {
+      // Phi 指令的操作数是成对的 [value, basic_block]
-        if (i > 0) std::cout << ", ";
+      // 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;
         std::cout << "[ ";
-        printValue(phiInst->getOperand(i));
+        printValue(phiInst->getOperand(i * 2)); // value
+        std::cout << ", %";
+        printValue(phiInst->getOperand(i * 2 + 1)); // block
         std::cout << " ]";
       }
       std::cout << std::endl;
     } break;
     
-    case Kind::kGetSubArray: {
+    // 以下两个 Kind 应该删除或替换为 kGEP
-      auto getSubArrayInst = dynamic_cast<GetSubArrayInst *>(pInst);
+    // case Kind::kLa: { /* REMOVED */ } break;
-      std::cout << "%" << getSubArrayInst->getName() << " = getelementptr inbounds ";
+    // case Kind::kGetSubArray: { /* REMOVED */ } break;
-      
-      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");
+      assert(false && "Unsupported instruction kind in SysYPrinter");
       break;
   }
 }

src/include/AddressCalculationExpansion.h

@@ -0,0 +1,59 @@
+#pragma once
+
+#include "IR.h" // 假设IR.h包含了Module, Function, BasicBlock, Instruction, Value, IRBuilder, Type等定义
+#include "IRBuilder.h" // 需要IRBuilder来创建新指令
+#include "SysYIRPrinter.h" // 新增: 用于调试输出
+#include <memory>
+#include <string>
+#include <unordered_map>
+#include <vector>
+#include <list> // 用于迭代和修改指令列表
+#include <algorithm> // for std::reverse (if needed, although not used in final version)
+#include <iostream> // MODIFICATION: 用于警告输出
+
+namespace sysy {
+
+/**
+ * @brief AddressCalculationExpansion Pass
+ *
+ * 这是一个IR优化Pass，用于将LoadInst和StoreInst中包含的多维数组索引
+ * 显式地转换为IR中的BinaryInst（乘法和加法）序列，并生成带有线性偏移量的
+ * LoadInst/StoreInst。
+ *
+ * 目的：确保在寄存器分配之前，所有中间地址计算的结果都有明确的IR指令和对应的虚拟寄存器，
+ * 从而避免在后端DAG构建时临时创建值而导致寄存器分配缺失的问题。
+ *
+ * SysY语言特性：
+ * - 无指针类型（所有数组访问的基地址是alloca或global的AllocaType/ArrayType）
+ * - 数据类型只有int和float，且都占用4字节。
+ * - LoadInst和StoreInst直接接受多个索引作为额外操作数。
+ */
+class AddressCalculationExpansion {
+private:
+    Module* pModule;
+    IRBuilder* pBuilder; // 用于在IR中插入新指令
+
+    // 数组元素的固定大小，根据SysY特性，int和float都是4字节
+    static const int ELEMENT_SIZE = 4;
+
+    // 辅助函数：根据数组的维度信息和当前索引的维度，计算该索引的步长（字节数）
+    // dims: 包含所有维度大小的vector，例如 {2, 3, 4}
+    // currentDimIndex: 当前正在处理的索引在 dims 中的位置 (0, 1, 2...)
+    int calculateStride(const std::vector<int>& dims, size_t currentDimIndex) {
+        int stride = ELEMENT_SIZE; // 最内层元素大小 (4字节)
+        // 乘以当前维度之后的所有维度的大小
+        for (size_t i = currentDimIndex + 1; i < dims.size(); ++i) {
+            stride *= dims[i];
+        }
+        return stride;
+    }
+
+public:
+    AddressCalculationExpansion(Module* module, IRBuilder* builder)
+        : pModule(module), pBuilder(builder) {}
+
+    // 运行此Pass
+    bool run();
+};
+
+} // namespace sysy

src/include/DeadCodeElimination.h

@@ -3,6 +3,7 @@
 #include "IR.h"
 #include "SysYIRAnalyser.h"
 #include "SysYIRPrinter.h"
+#include "SysYIROptUtils.h"
 
 namespace sysy {
 
@@ -31,9 +32,5 @@ 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

@@ -49,6 +49,7 @@ class Type {
     kLabel,
     kPointer,
     kFunction,
+    kArray,
   };
 
   Kind kind;  ///< 表示具体类型的变量
@@ -65,6 +66,7 @@ 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
@@ -74,6 +76,7 @@ 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>
@@ -115,6 +118,22 @@ 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
+};
+
 /*!
  * @}
  */
@@ -268,6 +287,51 @@ class ValueCounter {
   }  ///< 清空ValueCounter
 };
 
+
+// --- Refactored ConstantValue and related classes start here ---
+
+using ConstantValVariant = std::variant<int, float>;
+
+// Helper for hashing std::variant
+struct VariantHash {
+  template <class T>
+  std::size_t operator()(const T& val) const {
+    return std::hash<T>{}(val);
+  }
+  std::size_t operator()(const ConstantValVariant& v) const {
+    return std::visit(*this, v);
+  }
+};
+
+struct ConstantValueKey {
+  Type* type;
+  ConstantValVariant val;
+
+  bool operator==(const ConstantValueKey& other) const {
+    // Assuming Type objects are canonicalized, or add Type::isSame()
+    // If Type::isSame() is not available and Type objects are not canonicalized,
+    // this comparison might not be robust enough for structural equivalence of types.
+    return type == other.type && val == other.val;
+  }
+};
+
+struct ConstantValueHash {
+  std::size_t operator()(const ConstantValueKey& key) const {
+    std::size_t typeHash = std::hash<Type*>{}(key.type);
+    std::size_t valHash = VariantHash{}(key.val);
+    // A simple way to combine hashes
+    return typeHash ^ (valHash << 1);
+  }
+};
+
+struct ConstantValueEqual {
+  bool operator()(const ConstantValueKey& lhs, const ConstantValueKey& rhs) const {
+    // Assuming Type objects are canonicalized (e.g., Type::getIntType() always returns same pointer)
+    // If not, and Type::isSame() is intended, it should be added to Type class.
+    return lhs.type == rhs.type && lhs.val == rhs.val;
+  }
+};
+
 /*!
  * Static constants known at compile time.
  *
@@ -276,45 +340,135 @@ class ValueCounter {
  * `ConstantValue`并不由指令定义, 也不使用任何Value。它的类型为int/float。
  */
 
+template<class T> struct always_false : std::false_type {};
+template<class T> constexpr bool always_false_v = always_false<T>::value;
 
 class ConstantValue : public Value {
- protected:
+protected:
-  /// 定义字面量类型的聚合类型
+  static std::unordered_map<ConstantValueKey, ConstantValue*, ConstantValueHash, ConstantValueEqual> mConstantPool;
-  union {
-    int iScalar;
-    float fScalar;
-  };
 
- protected:
+public:
-  explicit ConstantValue(int value, const std::string &name = "") : Value(Type::getIntType(), name), iScalar(value) {}
+  explicit ConstantValue(Type* type, const std::string& name = "") : Value(type, name) {}
-  explicit ConstantValue(float value, const std::string &name = "")
+  virtual ~ConstantValue() = default;
-      : Value(Type::getFloatType(), name), fScalar(value) {}
 
- public:
+  virtual size_t hash() const = 0;
-  static ConstantValue* get(int value);    ///< 获取一个int类型的ConstValue *，其值为value
+  virtual ConstantValVariant getVal() const = 0;
-  static ConstantValue* get(float value);  ///< 获取一个float类型的ConstValue *，其值为value
 
- public:
+  // Static factory method to get a canonical ConstantValue from the pool
+  static ConstantValue* get(Type* type, ConstantValVariant val);
+
+  // Helper methods to access constant values with appropriate casting
   int getInt() const {
-    assert(isInt());
+    assert(getType()->isInt() && "Calling getInt() on non-integer type");
-    return iScalar;
+    return std::get<int>(getVal());
-  }  ///< 返回int类型的值
+  }
   float getFloat() const {
-    assert(isFloat());
+    assert(getType()->isFloat() && "Calling getFloat() on non-float type");
-    return fScalar;
+    return std::get<float>(getVal());
-  }  ///< 返回float类型的值
+  }
-  template <typename T>
+
-  T getValue() const {
+  template<typename T>
-    if (std::is_same<T, int>::value && isInt()) {
+  T getVal() const {
+    if constexpr (std::is_same_v<T, int>) {
       return getInt();
-    }
+    } else if constexpr (std::is_same_v<T, float>) {
-    if (std::is_same<T, float>::value && isFloat()) {
       return getFloat();
+    } else {
+      // This ensures a compilation error if an unsupported type is used
+      static_assert(always_false_v<T>, "Unsupported type for ConstantValue::getValue()");
     }
-    throw std::bad_cast();  // 或者其他适当的异常处理
+  }
-  }                         ///< 返回值，getInt和getFloat统一化，整数返回整形，浮点返回浮点型
+
+  virtual bool isZero() const = 0;
+  virtual bool isOne() const = 0;
 };
 
+class ConstantInteger : public ConstantValue {
+  int constVal;
+public:
+  explicit ConstantInteger(Type* type, int val, const std::string& name = "")
+      : ConstantValue(type, name), constVal(val) {}
+
+  size_t hash() const override {
+      std::size_t typeHash = std::hash<Type*>{}(getType());
+      std::size_t valHash = std::hash<int>{}(constVal);
+      return typeHash ^ (valHash << 1);
+  }
+  int getInt() const { return constVal; }
+  ConstantValVariant getVal() const override { return constVal; }
+
+  static ConstantInteger* get(Type* type, int val);
+  static ConstantInteger* get(int val) { return get(Type::getIntType(), val); }
+
+  ConstantInteger* getNeg() const {
+    assert(getType()->isInt() && "Cannot negate non-integer constant");
+    return ConstantInteger::get(-constVal);
+  }
+
+  bool isZero() const override { return constVal == 0; }
+  bool isOne() const override { return constVal == 1; }
+};
+
+class ConstantFloating : public ConstantValue {
+  float constFVal;
+public:
+  explicit ConstantFloating(Type* type, float val, const std::string& name = "")
+      : ConstantValue(type, name), constFVal(val) {}
+
+  size_t hash() const override {
+    std::size_t typeHash = std::hash<Type*>{}(getType());
+    std::size_t valHash = std::hash<float>{}(constFVal);
+    return typeHash ^ (valHash << 1);
+  }
+  float getFloat() const { return constFVal; }
+  ConstantValVariant getVal() const override { return constFVal; }
+
+  static ConstantFloating* get(Type* type, float val);
+  static ConstantFloating* get(float val) { return get(Type::getFloatType(), val); }
+
+  ConstantFloating* getNeg() const {
+    assert(getType()->isFloat() && "Cannot negate non-float constant");
+    return ConstantFloating::get(-constFVal);
+  }
+
+  bool isZero() const override { return constFVal == 0.0f; }
+  bool isOne() const override { return constFVal == 1.0f; }
+};
+
+class UndefinedValue : public ConstantValue {
+private:
+  static std::unordered_map<Type*, UndefinedValue*> UndefValues;
+
+protected:
+  explicit UndefinedValue(Type* type, const std::string& name = "")
+      : ConstantValue(type, name) {
+    assert(!type->isVoid() && "Cannot create UndefinedValue of void type!");
+  }
+
+public:
+  static UndefinedValue* get(Type* type);
+
+  size_t hash() const override {
+    return std::hash<Type*>{}(getType());
+  }
+
+  ConstantValVariant getVal() const override {
+    if (getType()->isInt()) {
+      return 0; // Return 0 for undefined integer
+    } else if (getType()->isFloat()) {
+      return 0.0f; // Return 0.0f for undefined float
+    }
+    assert(false && "UndefinedValue has unexpected type for getValue()");
+    return 0; // Should not be reached
+  }
+
+  bool isZero() const override { return false; }
+  bool isOne() const override { return false; }
+};
+
+// --- End of refactored ConstantValue and related classes ---
+
+
 class Instruction;
 class Function;
 class BasicBlock;
@@ -467,49 +621,6 @@ 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.
  */
@@ -559,15 +670,15 @@ class Instruction : public User {
     kAlloca = 0x1UL << 33,
     kLoad = 0x1UL << 34,
     kStore = 0x1UL << 35,
-    kLa = 0x1UL << 36,
+    kGetElementPtr = 0x1UL << 36,
     kMemset = 0x1UL << 37,
-    kGetSubArray = 0x1UL << 38,
+    // kGetSubArray = 0x1UL << 38,
-    // constant
+    // Constant Kind removed as Constants are now Values, not Instructions.
-    kConstant = 0x1UL << 37, 
+    // kConstant = 0x1UL << 37, // Conflicts with kMemset if kept as is
     // phi
     kPhi = 0x1UL << 39,
     kBitItoF = 0x1UL << 40,
-    kBitFtoI = 0x1UL << 41
+    kBitFtoI = 0x1UL << 41,
   };
 
 protected:
@@ -658,14 +769,12 @@ public:
         return "Load";
       case kStore:
         return "Store";
-      case kLa:
+      case kGetElementPtr:
-        return "La";
+        return "GetElementPtr";
       case kMemset:
         return "Memset";
       case kPhi:
         return "Phi";
-      case kGetSubArray:
-        return "GetSubArray";
       default:
         return "Unknown";
     }
@@ -718,9 +827,8 @@ public:
   bool isAlloca() const { return kind == kAlloca; }
   bool isLoad() const { return kind == kLoad; }
   bool isStore() const { return kind == kStore; }
-  bool isLa() const { return kind == kLa; }
+  bool isGetElementPtr() const { return kind == kGetElementPtr; }
   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 {
@@ -732,47 +840,54 @@ 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;
-  friend class SysySSA;
 
  protected:
-  Value *map_val;  // Phi的旧值
 
-  PhiInst(Type *type, Value *lhs, const std::vector<Value *> &rhs, Value *mval, BasicBlock *parent,
+  std::unordered_map<BasicBlock *, Value *> blk2val;  ///< 存储每个基本块对应的值
+  unsigned vsize;  ///< 存储值的数量
+
+  PhiInst(Type *type, 
+          const std::vector<Value *> &rhs = {}, 
+          const std::vector<BasicBlock*> &Blocks = {},
+          BasicBlock *parent = nullptr,
           const std::string &name = "")
-      : Instruction(Kind::kPhi, type, parent, name) {
+      : Instruction(Kind::kPhi, type, parent, name), vsize(rhs.size()) {
-    map_val = mval;
+    assert(rhs.size() == Blocks.size() && "PhiInst: rhs and Blocks must have the same size");
-    addOperand(lhs);
+    for(size_t i = 0; i < rhs.size(); ++i) {
-    addOperands(rhs);
+      addOperand(rhs[i]);
+      blk2val[Blocks[i]] = rhs[i];
+    }
   }
 
  public:
-  Value* getMapVal() { return map_val; }
+  Value* getValue(unsigned k) const {return getOperand(2 * k);}  ///< 获取位置为k的值
-  Value* getPointer() const { return getOperand(0); }
+  BasicBlock* getBlock(unsigned k) const {return dynamic_cast<BasicBlock*>(getOperand(2 * k + 1));}
+
+  auto& getincomings() const {return blk2val;}  ///< 获取所有的基本块和对应的值
+
+  Value* getvalfromBlk(BasicBlock* blk);
+  BasicBlock* getBlkfromVal(Value* val);
+
+  unsigned getNumIncomingValues() const { return vsize; }  ///< 获取传入值的数量
+  void addIncoming(Value *value, BasicBlock *block) {
+    assert(value && block && "PhiInst: value and block must not be null");
+    addOperand(value);
+    addOperand(block);
+    blk2val[block] = value;
+    vsize++;
+  }  ///< 添加传入值和对应的基本块
+
+  void delValue(Value* val);
+  void delBlk(BasicBlock* blk);
+
+  void replaceBlk(BasicBlock* newBlk, unsigned k);
+  void replaceold2new(BasicBlock* oldBlk, BasicBlock* newBlk);
+  void refreshB2VMap();
+
   auto getValues() { return make_range(std::next(operand_begin()), operand_end()); }
-  Value* getValue(unsigned index) const { return getOperand(index + 1); }
 };
 
 
@@ -884,7 +999,7 @@ public:
     }
   }  ///< 根据指令类型进行二元计算，eval template模板实现
   static BinaryInst* create(Kind kind, Type *type, Value *lhs, Value *rhs, BasicBlock *parent, const std::string &name = "") {
-        // 后端处理数组访存操作时需要创建计算地址的指令，需要在外部构造 BinaryInst 对象，所以写了个public的方法。
+        // 后端处理数组访存操作时需要创建计算地址的指令，需要在外部构造 BinaryInst 对象
         return new BinaryInst(kind, type, lhs, rhs, parent, name);
   }
 }; // class BinaryInst
@@ -972,7 +1087,7 @@ public:
 }; // class CondBrInst
 
 //! Allocate memory for stack variables, used for non-global variable declartion
-class AllocaInst : public Instruction , public LVal {
+class AllocaInst : public Instruction {
   friend class IRBuilder;
   friend class Function;
 protected:
@@ -983,14 +1098,6 @@ 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(); }
@@ -999,37 +1106,40 @@ public:
 }; // class AllocaInst
 
 
-class GetSubArrayInst : public Instruction {
+class GetElementPtrInst : public Instruction {
-  friend class IRBuilder;
+  friend class IRBuilder; // 如果您有IRBuilder来创建指令，需要friend
-  friend class Function;
 
- public:
+protected:
-  GetSubArrayInst(LVal *fatherArray, LVal *childArray, const std::vector<Value *> &indices,
+  // 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::kGetSubArray, Type::getVoidType(), parent, name) {
+      : Instruction(Kind::kGetElementPtr, basePointer->getType(), parent, name) {
-    auto predicate = [childArray](const std::unique_ptr<LVal> &child) -> bool { return child.get() == childArray; };
+    assert(basePointer && "GEP base pointer cannot be null!");
-    if (std::find_if(fatherArray->childrenLVals.begin(), fatherArray->childrenLVals.end(), predicate) ==
+    // TODO : 安全检查
-        fatherArray->childrenLVals.end()) {
+    assert(basePointer->getType()->isPointer() );
-      fatherArray->childrenLVals.emplace_back(childArray);
+    addOperand(basePointer); // 第一个操作数是基指针
-    }
+    addOperands(indices);    // 随后的操作数是索引
-    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:
+public:
-  Value* getFatherArray() const { return getOperand(0); }                              ///< 获取父数组
+  Value* getBasePointer() const { return getOperand(0); }
-  Value* getChildArray() const { return getOperand(1); }                               ///< 获取子数组
+  unsigned getNumIndices() const { return getNumOperands() - 1; }
-  LVal* getFatherLVal() const { return dynamic_cast<LVal *>(getOperand(0)); }          ///< 获取父左值
+  auto getIndices() const { return make_range(std::next(operand_begin()), operand_end());}
-  LVal* getChildLVal() const { return dynamic_cast<LVal *>(getOperand(1)); }           ///< 获取子左值
+  Value* getIndex(unsigned index) const {
-  auto getIndices() const { return make_range(std::next(operand_begin(), 2), operand_end()); }  ///< 获取索引
+    assert(index < getNumIndices() && "Index out of bounds for GEP!");
-  unsigned getNumIndices() const { return getNumOperands() - 2; }                       ///< 获取索引数量
+    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);
+  }
 };
 
 //! Load a value from memory address specified by a pointer value
@@ -1053,22 +1163,7 @@ 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
@@ -1094,22 +1189,6 @@ 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
 
@@ -1211,7 +1290,7 @@ protected:
 };
 
 //! Global value declared at file scope
-class GlobalValue : public User, public LVal {
+class GlobalValue : public User {
   friend class Module;
 
 protected:
@@ -1230,28 +1309,21 @@ protected:
     if (init.size() == 0) {
       unsigned num = 1;
       for (unsigned i = 0; i < numDims; i++) {
-        num *= dynamic_cast<ConstantValue *>(dims[i])->getInt();
+        // Assume dims elements are ConstantInteger and cast appropriately
+        auto dim_val = dynamic_cast<ConstantInteger*>(dims[i]);
+        assert(dim_val && "GlobalValue dims must be constant integers");
+        num *= dim_val->getInt();
       }
       if (dynamic_cast<PointerType *>(type)->getBaseType() == Type::getFloatType()) {
-        init.push_back(ConstantValue::get(0.0F), num);
+        init.push_back(ConstantFloating::get(0.0F), num); // Use new constant factory
       } else {
-        init.push_back(ConstantValue::get(0), num);
+        init.push_back(ConstantInteger::get(0), num); // Use new constant factory
       }
     }
     initValues = init;
   }
 
 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
@@ -1261,8 +1333,11 @@ public:
    Value* getByIndices(const std::vector<Value *> &indices) const  {
     int index = 0;
     for (size_t i = 0; i < indices.size(); i++) {
-      index = dynamic_cast<ConstantValue *>(getDim(i))->getInt() * index +
+      // Ensure dims[i] and indices[i] are ConstantInteger and retrieve their values correctly
-              dynamic_cast<ConstantValue *>(indices[i])->getInt();
+      auto dim_val = dynamic_cast<ConstantInteger*>(getDim(i));
+      auto idx_val = dynamic_cast<ConstantInteger*>(indices[i]);
+      assert(dim_val && idx_val && "Dims and indices must be constant integers");
+      index = dim_val->getInt() * index + idx_val->getInt();
     }
     return getByIndex(index);
   }  ///< 通过多维索引indices获取初始值
@@ -1270,7 +1345,7 @@ public:
 }; // class GlobalValue
 
 
-class ConstantVariable : public User, public LVal {
+class ConstantVariable : public User {
   friend class Module;
 
  protected:
@@ -1289,22 +1364,16 @@ class ConstantVariable : public User, public LVal {
   }
 
  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;
     // 计算偏移量
     for (size_t i = 0; i < indices.size(); i++) {
-      index = dynamic_cast<ConstantValue *>(getDim(i))->getInt() * index +
+      // Ensure dims[i] and indices[i] are ConstantInteger and retrieve their values correctly
-              dynamic_cast<ConstantValue *>(indices[i])->getInt();
+      auto dim_val = dynamic_cast<ConstantInteger*>(getDim(i));
+      auto idx_val = dynamic_cast<ConstantInteger*>(indices[i]);
+      assert(dim_val && idx_val && "Dims and indices must be constant integers");
+      index = dim_val->getInt() * index + idx_val->getInt();
     }
 
     return getByIndex(index);

src/include/IRBuilder.h

@@ -280,46 +280,6 @@ 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);
@@ -333,17 +293,31 @@ class IRBuilder {
     block->getInstructions().emplace(position, inst);
     return inst;
   }  ///< 创建store指令
-  PhiInst * createPhiInst(Type *type, Value *lhs, BasicBlock *parent, const std::string &name = "") {
+  PhiInst * createPhiInst(Type *type, const std::vector<Value*> &vals = {}, const std::vector<BasicBlock*> &blks = {}, const std::string &name = "") {
-    auto predNum = parent->getNumPredecessors();
+    auto predNum = block->getNumPredecessors();
-    std::vector<Value *> rhs;
+    auto inst = new PhiInst(type, vals, blks, block, name);
-    for (size_t i = 0; i < predNum; i++) {
-      rhs.push_back(lhs);
-    }
-    auto inst = new PhiInst(type, lhs, rhs, lhs, parent, name);
     assert(inst);
-    parent->getInstructions().emplace(parent->begin(), inst);
+    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

@@ -1,59 +1,79 @@
-#pragma once
+// 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
 
-#include <list>
+#include <vector>
-#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;
+  Module* pModule;
-  IRBuilder *pBuilder;
+  IRBuilder* pBuilder;
-  ControlFlowAnalysis *controlFlowAnalysis;  // 控制流分析
+  ControlFlowAnalysis* controlFlowAnalysis;
-  ActiveVarAnalysis *activeVarAnalysis;  // 活跃变量分析
+  ActiveVarAnalysis* activeVarAnalysis;
-  DataFlowAnalysisUtils dataFlowAnalysisUtils;
+  DataFlowAnalysisUtils dataFlowAnalysisUtils; // If this is part of Mem2Reg or an external helper
 
 public:
-  Mem2Reg(Module *pMoudle, IRBuilder *pBuilder, 
+  Mem2Reg(Module* module, IRBuilder* builder, ControlFlowAnalysis* cfa, ActiveVarAnalysis* ava)
-          ControlFlowAnalysis *pCFA = nullptr, ActiveVarAnalysis *pAVA = nullptr) :
+    : pModule(module), pBuilder(builder), controlFlowAnalysis(cfa), activeVarAnalysis(ava) {}
-    pModule(pMoudle), pBuilder(pBuilder), controlFlowAnalysis(pCFA), activeVarAnalysis(pAVA), dataFlowAnalysisUtils()
+    // Constructor initializes members
-   {}  // 初始化函数
+  void run();
 
-  void mem2regPipeline();  ///< mem2reg
+  // --- 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; 
 
-private:
+  // 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
 
-  // phi节点的插入需要计算IDF
+  // --- 核心 SSA 转换辅助函数 ---
-  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
+  // 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);
 
-  auto computeValue2Blocks() -> void;  ///< 计算value2block的映射(不包括数组和global)
+  // 分析一个 alloca 的所有 uses，填充 allocaDefsBlock 和 allocaUsesBlock
+  void allocaAnalysis(AllocaInst* alloca);
 
-  auto preOptimize1() -> void;  ///< llvm memtoreg预优化1: 删除不含load的alloc和store
+  // 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)
-  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
+  bool is_promoted(AllocaInst* alloca);
-  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
 
-  auto insertPhi() -> void;  ///< 为所有变量的迭代支配边界插入phi结点
+  // 在迭代支配边界处插入 Phi 指令
+  void insertPhiNodes(Function* func);
 
-  auto rename(BasicBlock *block, std::unordered_map<Value *, int> &count,
+  // 递归地重命名基本块中的变量并填充 Phi 指令
-              std::unordered_map<Value *, std::stack<Instruction *>> &stacks) -> void;  ///< 单个块的重命名
+  // 这里的 `count` 和 `stacks` 是临时的，用于 DFS 过程中传递状态
-  auto renameAll() -> void;                                                             ///< 重命名所有块
+  void renameBlock(BasicBlock* block, 
+                      std::unordered_map<AllocaInst*, Value*>& currentIncomings,
+                      std::unordered_set<BasicBlock*>& visitedBlocks); // 修改为传递 map 和 set
 
-  // private helper function.
+  // 简化冗余的 Phi 指令 (当所有输入都相同时)
-private:
+  void simplifyphi(PhiInst* phi);
-  auto getPredIndex(BasicBlock *n, BasicBlock *s) -> int;  ///< 获取前驱索引
+
-  auto cascade(Instruction *instr, bool &changed, Function *func, BasicBlock *block,
+  // 获取前驱块在后继块前驱列表中的索引，用于 Phi 指令入边
-               std::list<std::unique_ptr<Instruction>> &instrs) -> void;  ///< 消除级联关系
+  int getPredIndex(BasicBlock* pred, BasicBlock* succ);
-  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
+  // --- Mem2Reg 的主要工作流函数 ---
-  auto usedelete(Instruction *instr) -> void;                             ///< 删除指令相关的value-use-user关系
+  // 对单个函数执行内存到寄存器的提升
+  bool promoteMemoryToRegisters(Function* func);
 
 };
+
 } // namespace sysy
+
+#endif // SYSY_MEM2REG_H

src/include/RISCv64AsmPrinter.h

@@ -0,0 +1,32 @@
+#ifndef RISCV64_ASMPRINTER_H
+#define RISCV64_ASMPRINTER_H
+
+#include "RISCv64LLIR.h"
+#include <iostream>
+
+namespace sysy {
+
+class RISCv64AsmPrinter {
+public:
+    RISCv64AsmPrinter(MachineFunction* mfunc);
+    // 主入口
+    void run(std::ostream& os);
+
+private:
+    // 打印各个部分
+    void printPrologue();
+    void printEpilogue();
+    void printBasicBlock(MachineBasicBlock* mbb);
+    void printInstruction(MachineInstr* instr);
+    
+    // 辅助函数
+    std::string regToString(PhysicalReg reg);
+    void printOperand(MachineOperand* op);
+
+    MachineFunction* MFunc;
+    std::ostream* OS;
+};
+
+} // namespace sysy
+
+#endif // RISCV64_ASMPRINTER_H

src/include/RISCv64Backend.h

@@ -3,118 +3,23 @@
 
 #include "IR.h"
 #include <string>
-#include <vector>
-#include <map>
-#include <set>
-#include <memory>
-#include <iostream>
-#include <functional> // For std::function
-
-extern int DEBUG;
-extern int DEEPDEBUG;
 
 namespace sysy {
 
+// RISCv64CodeGen 现在是一个高层驱动器
 class RISCv64CodeGen {
 public:
-    enum class PhysicalReg {
-        ZERO, RA, SP, GP, TP, T0, T1, T2, S0, S1, A0, A1, A2, A3, A4, A5, A6, A7, S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, T3, T4, T5, T6,
-        F0, F1, F2, F3, F4, F5, F6, F7, F8, F9, F10, F11, F12, F13, F14, F15,F16, F17, F18, F19, F20, F21, F22, F23, F24, F25, F26, F27, F28, F29, F30, F31
-    };
-
-    // Move DAGNode and RegAllocResult to public section
-    struct DAGNode {
-        enum NodeKind { CONSTANT, LOAD, STORE, BINARY, CALL, RETURN, BRANCH, ALLOCA_ADDR, UNARY };
-        NodeKind kind;
-        Value* value = nullptr; // For IR Value
-        std::string inst;       // Generated RISC-V instruction(s) for this node
-        std::string result_vreg; // Virtual register assigned to this node's result
-        std::vector<DAGNode*> operands;
-        std::vector<DAGNode*> users; // For debugging and potentially optimizations
-        DAGNode(NodeKind k) : kind(k) {}
-
-        // Debugging / helper
-        std::string getNodeKindString() const {
-            switch (kind) {
-                case CONSTANT: return "CONSTANT";
-                case LOAD: return "LOAD";
-                case STORE: return "STORE";
-                case BINARY: return "BINARY";
-                case CALL: return "CALL";
-                case RETURN: return "RETURN";
-                case BRANCH: return "BRANCH";
-                case ALLOCA_ADDR: return "ALLOCA_ADDR";
-                case UNARY: return "UNARY";
-                default: return "UNKNOWN";
-            }
-        }
-    };
-
-    struct RegAllocResult {
-        std::map<std::string, PhysicalReg> vreg_to_preg; // Virtual register to Physical Register mapping
-        std::map<Value*, int> stack_map; // Value (AllocaInst) to stack offset
-        int stack_size = 0; // Total stack frame size for locals and spills
-    };
-
     RISCv64CodeGen(Module* mod) : module(mod) {}
-
+    // 唯一的公共入口点
     std::string code_gen();
-    std::string module_gen();
-    std::string function_gen(Function* func);
-    // 修改 basicBlock_gen 的声明，添加 int block_idx 参数
-    std::string basicBlock_gen(BasicBlock* bb, const RegAllocResult& alloc, int block_idx);
-
-    // DAG related
-    std::vector<std::unique_ptr<DAGNode>> build_dag(BasicBlock* bb);
-    void select_instructions(DAGNode* node, const RegAllocResult& alloc);
-    // 改变 emit_instructions 的参数，使其可以直接添加汇编指令到 main ss
-    void emit_instructions(DAGNode* node, std::stringstream& ss, const RegAllocResult& alloc, std::set<DAGNode*>& emitted_nodes);
-
-    // Register Allocation related
-    std::map<Instruction*, std::set<std::string>> liveness_analysis(Function* func);
-    std::map<std::string, std::set<std::string>> build_interference_graph(
-        const std::map<Instruction*, std::set<std::string>>& live_sets);
-    void color_graph(std::map<std::string, PhysicalReg>& vreg_to_preg,
-                     const std::map<std::string, std::set<std::string>>& interference_graph);
-    RegAllocResult register_allocation(Function* func);
-    void eliminate_phi(Function* func); // Phi elimination is typically done before DAG building
-
-    // Utility
-    std::string reg_to_string(PhysicalReg reg);
-    void print_dag(const std::vector<std::unique_ptr<DAGNode>>& dag, const std::string& bb_name);
 
 private:
-    static const std::vector<PhysicalReg> allocable_regs;
+    // 模块级代码生成
-    std::map<Value*, std::string> value_vreg_map; // Maps IR Value* to its virtual register name
+    std::string module_gen();
+    // 函数级代码生成 (实现新的流水线)
+    std::string function_gen(Function* func);
+
     Module* module;
-    int vreg_counter = 0; // Counter for unique virtual register names
-    int alloca_offset_counter = 0; // Counter for alloca offsets
-
-    // 新增一个成员变量来存储当前函数的所有 DAGNode，以确保其生命周期贯穿整个函数代码生成
-    // 这样可以在多个 BasicBlock_gen 调用中访问到完整的 DAG 节点
-    std::vector<std::unique_ptr<DAGNode>> current_function_dag_nodes;
-
-    // 为空标签定义一个伪名称前缀，加上块索引以确保唯一性
-    const std::string ENTRY_BLOCK_PSEUDO_NAME = "entry_block_";
-
-    // !!! 修改：get_operand_node 辅助函数现在需要传入 value_to_node 和 nodes_storage 的引用
-    // 因为它们是 build_dag 局部管理的
-    DAGNode* get_operand_node(
-        Value* val_ir,
-        std::map<Value*, DAGNode*>& value_to_node,
-        std::vector<std::unique_ptr<DAGNode>>& nodes_storage
-    );
-
-    // !!! 新增：create_node 辅助函数也需要传入 value_to_node 和 nodes_storage 的引用
-    // 并且它应该不再是 lambda，而是一个真正的成员函数
-    DAGNode* create_node(
-        DAGNode::NodeKind kind,
-        Value* val,
-        std::map<Value*, DAGNode*>& value_to_node,
-        std::vector<std::unique_ptr<DAGNode>>& nodes_storage
-    );
-
-    std::vector<std::unique_ptr<Instruction>> temp_instructions_storage; // 用于存储 build_dag 中创建的临时 BinaryInst
 };
 
 } // namespace sysy

src/include/RISCv64ISel.h

@@ -0,0 +1,49 @@
+#ifndef RISCV64_ISEL_H
+#define RISCV64_ISEL_H
+
+#include "RISCv64LLIR.h"
+
+namespace sysy {
+
+class RISCv64ISel {
+public:
+    RISCv64ISel();
+    // 模块主入口：将一个高层IR函数转换为底层LLIR函数
+    std::unique_ptr<MachineFunction> runOnFunction(Function* func);
+
+    // 公开接口，以便后续模块（如RegAlloc）可以查询或创建vreg
+    unsigned getVReg(Value* val);
+    unsigned getNewVReg() { return vreg_counter++; }
+
+private:
+    // DAG节点定义，作为ISel的内部实现细节
+    struct DAGNode; 
+    
+    // 指令选择主流程
+    void select();
+    // 为单个基本块生成指令
+    void selectBasicBlock(BasicBlock* bb);
+    // 核心函数：为DAG节点选择并生成MachineInstr
+    void selectNode(DAGNode* node);
+
+    // DAG 构建相关函数 (从原RISCv64Backend迁移)
+    std::vector<std::unique_ptr<DAGNode>> build_dag(BasicBlock* bb);
+    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>>&);
+    
+    // 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
+    MachineBasicBlock* CurMBB; // 当前正在处理的机器基本块
+
+    // 映射关系
+    std::map<Value*, unsigned> vreg_map;
+    std::map<const BasicBlock*, MachineBasicBlock*> bb_map;
+
+    unsigned vreg_counter;
+    int local_label_counter;
+};
+
+} // namespace sysy
+
+#endif // RISCV64_ISEL_H

src/include/RISCv64LLIR.h

@@ -0,0 +1,200 @@
+#ifndef RISCV64_LLIR_H
+#define RISCV64_LLIR_H
+
+#include "IR.h" // 确保包含了您自己的IR头文件
+#include <string>
+#include <vector>
+#include <memory>
+#include <cstdint>
+#include <map>
+
+// 前向声明，避免循环引用
+namespace sysy {
+class Function;
+class RISCv64ISel;
+}
+
+namespace sysy {
+
+// 物理寄存器定义
+enum class PhysicalReg {
+    ZERO, RA, SP, GP, TP, T0, T1, T2, S0, S1, A0, A1, A2, A3, A4, A5, A6, A7, S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, T3, T4, T5, T6,
+    F0, F1, F2, F3, F4, F5, F6, F7, F8, F9, F10, F11, F12, F13, F14, F15,F16, F17, F18, F19, F20, F21, F22, F23, F24, F25, F26, F27, F28, F29, F30, F31
+};
+
+// RISC-V 指令操作码枚举
+enum class RVOpcodes {
+    // 算术指令
+    ADD, ADDI, ADDW, ADDIW, SUB, SUBW, MUL, MULW, DIV, DIVW, REM, REMW,
+    // 逻辑指令
+    XOR, XORI, OR, ORI, AND, ANDI,
+    // 移位指令
+    SLL, SLLI, SLLW, SLLIW, SRL, SRLI, SRLW, SRLIW, SRA, SRAI, SRAW, SRAIW,
+    // 比较指令
+    SLT, SLTI, SLTU, SLTIU,
+    // 内存访问指令
+    LW, LH, LB, LWU, LHU, LBU, SW, SH, SB, LD, SD,
+    // 控制流指令
+    J, JAL, JALR, RET,
+    BEQ, BNE, BLT, BGE, BLTU, BGEU,
+    // 伪指令
+    LI, LA, MV, NEG, NEGW, SEQZ, SNEZ,
+    // 函数调用
+    CALL,
+    // 特殊标记，非指令
+    LABEL,
+    // 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)
+};
+
+class MachineOperand;
+class RegOperand;
+class ImmOperand;
+class LabelOperand;
+class MemOperand;
+class MachineInstr;
+class MachineBasicBlock;
+class MachineFunction;
+
+// 操作数基类
+class MachineOperand {
+public:
+    enum OperandKind { KIND_REG, KIND_IMM, KIND_LABEL, KIND_MEM };
+    MachineOperand(OperandKind kind) : kind(kind) {}
+    virtual ~MachineOperand() = default;
+    OperandKind getKind() const { return kind; }
+private:
+    OperandKind kind;
+};
+
+// 寄存器操作数
+class RegOperand : public MachineOperand {
+public:
+    // 构造虚拟寄存器
+    RegOperand(unsigned vreg_num) 
+        : MachineOperand(KIND_REG), vreg_num(vreg_num), is_virtual(true) {}
+
+    // 构造物理寄存器
+    RegOperand(PhysicalReg preg)
+        : MachineOperand(KIND_REG), preg(preg), is_virtual(false) {}
+
+    bool isVirtual() const { return is_virtual; }
+    unsigned getVRegNum() const { return vreg_num; }
+    PhysicalReg getPReg() const { return preg; }
+
+    void setPReg(PhysicalReg new_preg) {
+        preg = new_preg;
+        is_virtual = false;
+    }
+private:
+    unsigned vreg_num = 0;
+    PhysicalReg preg = PhysicalReg::ZERO;
+    bool is_virtual;
+};
+
+// 立即数操作数
+class ImmOperand : public MachineOperand {
+public:
+    ImmOperand(int64_t value) : MachineOperand(KIND_IMM), value(value) {}
+    int64_t getValue() const { return value; }
+private:
+    int64_t value;
+};
+
+// 标签操作数
+class LabelOperand : public MachineOperand {
+public:
+    LabelOperand(const std::string& name) : MachineOperand(KIND_LABEL), name(name) {}
+    const std::string& getName() const { return name; }
+private:
+    std::string name;
+};
+
+// 内存操作数, 表示 offset(base_reg)
+class MemOperand : public MachineOperand {
+public:
+    MemOperand(std::unique_ptr<RegOperand> base, std::unique_ptr<ImmOperand> offset)
+        : MachineOperand(KIND_MEM), base(std::move(base)), offset(std::move(offset)) {}
+    RegOperand* getBase() const { return base.get(); }
+    ImmOperand* getOffset() const { return offset.get(); }
+private:
+    std::unique_ptr<RegOperand> base;
+    std::unique_ptr<ImmOperand> offset;
+};
+
+// 机器指令
+class MachineInstr {
+public:
+    MachineInstr(RVOpcodes opcode) : opcode(opcode) {}
+
+    RVOpcodes getOpcode() const { return opcode; }
+    const std::vector<std::unique_ptr<MachineOperand>>& getOperands() const { return operands; }
+    std::vector<std::unique_ptr<MachineOperand>>& getOperands() { return operands; }
+
+    void addOperand(std::unique_ptr<MachineOperand> operand) {
+        operands.push_back(std::move(operand));
+    }
+private:
+    RVOpcodes opcode;
+    std::vector<std::unique_ptr<MachineOperand>> operands;
+};
+
+// 机器基本块
+class MachineBasicBlock {
+public:
+    MachineBasicBlock(const std::string& name, MachineFunction* parent) 
+        : name(name), parent(parent) {}
+
+    const std::string& getName() const { return name; }
+    MachineFunction* getParent() const { return parent; }
+    const std::vector<std::unique_ptr<MachineInstr>>& getInstructions() const { return instructions; }
+    std::vector<std::unique_ptr<MachineInstr>>& getInstructions() { return instructions; }
+    
+    void addInstruction(std::unique_ptr<MachineInstr> instr) {
+        instructions.push_back(std::move(instr));
+    }
+    
+    std::vector<MachineBasicBlock*> successors;
+    std::vector<MachineBasicBlock*> predecessors;
+private:
+    std::string name;
+    std::vector<std::unique_ptr<MachineInstr>> instructions;
+    MachineFunction* parent;
+};
+
+// 栈帧信息
+struct StackFrameInfo {
+    int locals_size = 0; // 仅为AllocaInst分配的大小
+    int spill_size = 0; // 仅为溢出分配的大小
+    int total_size = 0; // 总大小
+    std::map<unsigned, int> alloca_offsets; // <AllocaInst的vreg, 栈偏移>
+    std::map<unsigned, int> spill_offsets;  // <溢出vreg, 栈偏移>
+};
+
+// 机器函数
+class MachineFunction {
+public:
+    MachineFunction(Function* func, RISCv64ISel* isel) : F(func), name(func->getName()), isel(isel) {}
+
+    Function* getFunc() const { return F; }
+    RISCv64ISel* getISel() const { return isel; }
+    const std::string& getName() const { return name; }
+    StackFrameInfo& getFrameInfo() { return frame_info; }
+    const std::vector<std::unique_ptr<MachineBasicBlock>>& getBlocks() const { return blocks; }
+    std::vector<std::unique_ptr<MachineBasicBlock>>& getBlocks() { return blocks; }
+
+    void addBlock(std::unique_ptr<MachineBasicBlock> block) {
+        blocks.push_back(std::move(block));
+    }
+private:
+    Function* F;
+    RISCv64ISel* isel; // 指向创建它的ISel，用于获取vreg映射等信息
+    std::string name;
+    std::vector<std::unique_ptr<MachineBasicBlock>> blocks;
+    StackFrameInfo frame_info;
+};
+
+} // namespace sysy
+
+#endif // RISCV64_LLIR_H

src/include/RISCv64Passes.h

@@ -0,0 +1,61 @@
+#ifndef RISCV64_PASSES_H
+#define RISCV64_PASSES_H
+
+#include "RISCv64LLIR.h"
+
+namespace sysy {
+
+/**
+ * @class Pass
+ * @brief 所有优化Pass的抽象基类 (可选，但推荐)
+ * * 定义一个通用的接口，所有优化都应该实现它。
+ */
+class Pass {
+public:
+    virtual ~Pass() = default;
+    virtual void runOnMachineFunction(MachineFunction* mfunc) = 0;
+};
+
+
+// --- 寄存器分配前优化 ---
+
+/**
+ * @class PreRA_Scheduler
+ * @brief 寄存器分配前的指令调度器
+ * * 在虚拟寄存器上进行操作，此时调度自由度最大，
+ * 主要目标是隐藏指令延迟，提高流水线效率。
+ */
+class PreRA_Scheduler : public Pass {
+public:
+    void runOnMachineFunction(MachineFunction* mfunc) override;
+};
+
+
+// --- 寄存器分配后优化 ---
+
+/**
+ * @class PeepholeOptimizer
+ * @brief 窥孔优化器
+ * * 在已分配物理寄存器的指令流上，通过一个小的滑动窗口来查找
+ * 并替换掉一些冗余或低效的指令模式。
+ */
+class PeepholeOptimizer : public Pass {
+public:
+    void runOnMachineFunction(MachineFunction* mfunc) override;
+};
+
+/**
+ * @class PostRA_Scheduler
+ * @brief 寄存器分配后的局部指令调度器
+ * * 主要目标是优化寄存器分配器插入的spill/fill代码(lw/sw)，
+ * 尝试将加载指令提前，以隐藏其访存延迟。
+ */
+class PostRA_Scheduler : public Pass {
+public:
+    void runOnMachineFunction(MachineFunction* mfunc) override;
+};
+
+
+} // namespace sysy
+
+#endif // RISCV64_PASSES_H

src/include/RISCv64RegAlloc.h

@@ -0,0 +1,56 @@
+#ifndef RISCV64_REGALLOC_H
+#define RISCV64_REGALLOC_H
+
+#include "RISCv64LLIR.h"
+
+namespace sysy {
+
+class RISCv64RegAlloc {
+public:
+    RISCv64RegAlloc(MachineFunction* mfunc);
+
+    // 模块主入口
+    void run();
+
+private:
+    using LiveSet = std::set<unsigned>; // 活跃虚拟寄存器集合
+    using InterferenceGraph = std::map<unsigned, std::set<unsigned>>;
+
+    // 栈帧管理
+    void eliminateFrameIndices();
+    
+    // 活跃性分析
+    void analyzeLiveness();
+
+    // 构建干扰图
+    void buildInterferenceGraph();
+
+    // 图着色分配寄存器
+    void colorGraph();
+
+    // 重写函数，替换vreg并插入溢出代码
+    void rewriteFunction();
+    
+    // 辅助函数，获取指令的Use/Def集合
+    void getInstrUseDef(MachineInstr* instr, LiveSet& use, LiveSet& def);
+
+    MachineFunction* MFunc;
+    
+    // 活跃性分析结果
+    std::map<const MachineInstr*, LiveSet> live_in_map;
+    std::map<const MachineInstr*, LiveSet> live_out_map;
+
+    // 干扰图
+    InterferenceGraph interference_graph;
+
+    // 图着色结果
+    std::map<unsigned, PhysicalReg> color_map; // vreg -> preg
+    std::set<unsigned> spilled_vregs;         // 被溢出的vreg集合
+
+    // 可用的物理寄存器池
+    std::vector<PhysicalReg> allocable_int_regs;
+};
+
+} // namespace sysy
+
+#endif // RISCV64_REGALLOC_H

src/include/Reg2Mem.h

@@ -2,6 +2,7 @@
 
 #include "IR.h"
 #include "IRBuilder.h"
+#include "SysYIROptUtils.h"
 
 namespace sysy {
 /**
@@ -16,8 +17,6 @@ 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

@@ -0,0 +1,196 @@
+#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

@@ -1,340 +0,0 @@
-#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

@@ -0,0 +1,60 @@
+#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

@@ -62,12 +62,11 @@ 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;
@@ -134,6 +133,11 @@ 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

@@ -1,37 +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
-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

@@ -0,0 +1,33 @@
+#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

@@ -0,0 +1,59 @@
+#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

@@ -0,0 +1,58 @@
+// 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

@@ -13,12 +13,13 @@ using namespace antlr4;
 
 #include "SysYIRGenerator.h"
 #include "SysYIRPrinter.h"
-#include "SysYIROptPre.h"
+#include "SysYIRCFGOpt.h"
 #include "RISCv64Backend.h"
-#include "SysYIRAnalyser.h"
+// #include "SysYIRAnalyser.h"
-#include "DeadCodeElimination.h"
+// #include "DeadCodeElimination.h"
-#include "Mem2Reg.h"
+#include "AddressCalculationExpansion.h"
-#include "Reg2Mem.h"
+// #include "Mem2Reg.h"
+// #include "Reg2Mem.h"
 
 using namespace sysy;
 
@@ -124,42 +125,41 @@ int main(int argc, char **argv) {
   // 无论最终输出是 IR 还是 ASM，只要不是停止在 AST 阶段，都会进入此优化流程。
   // optLevel = 0 时，执行默认优化。
   // optLevel >= 1 时，执行默认优化 + 额外的 -O1 优化。
-  cout << "Applying middle-end optimizations (level -O" << optLevel << ")...\n";
+  if (DEBUG) cout << "Applying middle-end optimizations (level -O" << optLevel << ")...\n";
 
   // 设置 DEBUG 模式（如果指定了 'ird'）
   if (argStopAfter == "ird") {
     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)
-  SysYOptPre optPre(moduleIR, builder);
+  SysYCFGOpt cfgopt(moduleIR, builder);
-  optPre.SysYOptimizateAfterIR(); 
+  cfgopt.SysYOptimizateAfterIR(); 
 
-  ControlFlowAnalysis cfa(moduleIR);
+  // ControlFlowAnalysis cfa(moduleIR);
-  cfa.init();
+  // cfa.init();
-  ActiveVarAnalysis ava;
+  // ActiveVarAnalysis ava;
-  ava.init(moduleIR);
+  // ava.init(moduleIR);
 
   if (DEBUG) {
     cout << "=== After CFA & AVA (Default) ===\n";
     SysYPrinter(moduleIR).printIR(); // 临时打印器用于调试
   }
-
+  AddressCalculationExpansion ace(moduleIR, builder);
-  DeadCodeElimination dce(moduleIR, &cfa, &ava);
+  if (ace.run()) {
-  dce.runDCEPipeline();
+      if (DEBUG) cout << "AddressCalculationExpansion made changes.\n";
+      // 如果 ACE 改变了IR，并且 DEBUG 模式开启，可以考虑打印IR
       if (DEBUG) {
-    cout << "=== After 1st DCE (Default) ===\n";
+          cout << "=== After AddressCalculationExpansion ===\n";
           SysYPrinter(moduleIR).printIR();
       }
-
+  } else {
-
+      if (DEBUG) cout << "AddressCalculationExpansion made no changes.\n";
-
+  }
-
-
 
 
   // 根据优化级别，执行额外的优化 pass
   if (optLevel >= 1) { 
-    cout << "Applying additional -O" << optLevel << " optimizations...\n";
+    if (DEBUG) cout << "Applying additional -O" << optLevel << " optimizations...\n";
     // 放置 -O1 及其以上级别要启用的额外优化 pass
     // 例如：
     // MyNewOptimizationPass newOpt(moduleIR, builder);
@@ -174,28 +174,34 @@ int main(int argc, char **argv) {
     // MyCustomOpt2 opt2_pass(moduleIR, builder, &cfa); // 假设需要CFA
     // opt2_pass.run();
     // ... 更多 -O1 特有的优化
+    // DeadCodeElimination dce(moduleIR, &cfa, &ava);
+    // dce.runDCEPipeline();
+    // if (DEBUG) {
+    //   cout << "=== After 1st DCE (Default) ===\n";
+    //   SysYPrinter(moduleIR).printIR();
+    // }
 
-    Mem2Reg mem2reg(moduleIR, builder, &cfa, &ava);
+    // Mem2Reg mem2reg(moduleIR, builder, &cfa, &ava);
-    mem2reg.mem2regPipeline();
+    // mem2reg.mem2regPipeline();
-    if (DEBUG) {
+    // if (DEBUG) {
-      cout << "=== After Mem2Reg (Default) ===\n";
+    //   cout << "=== After Mem2Reg (Default) ===\n";
-      SysYPrinter(moduleIR).printIR();
+    //   SysYPrinter(moduleIR).printIR();
-    }
+    // }
 
-    Reg2Mem reg2mem(moduleIR, builder);
+    // Reg2Mem reg2mem(moduleIR, builder);
-    reg2mem.DeletePhiInst();
+    // reg2mem.DeletePhiInst();
-    if (DEBUG) {
+    // if (DEBUG) {
-      cout << "=== After Reg2Mem (Default) ===\n";
+    //   cout << "=== After Reg2Mem (Default) ===\n";
-      SysYPrinter(moduleIR).printIR();
+    //   SysYPrinter(moduleIR).printIR();
-    }
+    // }
 
-    dce.runDCEPipeline(); // 第二次 DCE (默认)
+    // dce.runDCEPipeline(); // 第二次 DCE (默认)
-    if (DEBUG) {
+    // if (DEBUG) {
-      cout << "=== After 2nd DCE (Default) ===\n";
+    //   cout << "=== After 2nd DCE (Default) ===\n";
-      SysYPrinter(moduleIR).printIR();
+    //   SysYPrinter(moduleIR).printIR();
-    }
+    // }
   } else {
-    cout << "No additional middle-end optimizations applied for -O" << optLevel << ".\n";
+    if (DEBUG) cout << "No additional middle-end optimizations applied for -O" << optLevel << ".\n";
   }
 
   // 5. 根据 argStopAfter 决定后续操作
@@ -212,7 +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.sh

@@ -20,7 +20,12 @@ TESTDATA_DIR="${SCRIPT_DIR}/testdata"
 
 # 定义编译器 (这里假设 gcc 在 VM 内部是可用的)
 GCC_NATIVE="gcc" # VM 内部的 gcc
-# 不再需要 QEMU_RISCV64，因为直接执行
+
+# --- 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
+TOTAL_CASES=0
+PASSED_CASES=0
 
 # 显示帮助信息的函数
 show_help() {
@@ -30,30 +35,31 @@ show_help() {
     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 "  -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. 使用 VM 内部的 gcc 将 .s 文件汇编并链接为可执行文件 (链接 -L./lib -lsysy_riscv -static)。"
+    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. 直接运行编译后的可执行文件。"
+    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 testdata/*.out 文件内容（最后一行是否为整数）决定是进行返回值比较、标准输出比较，或两者都进行。"
+    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. 如果没有对应的 .in/.out 文件，则打印可执行文件的返回值。"
+    echo "5. 输出比较时会忽略行尾多余的换行符。"
-    echo "6. 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。"
 }
 
 # 清理临时文件的函数
 clean_tmp() {
     echo "正在清理临时目录: ${TMP_DIR}"
-    # 清理所有由本脚本和 runit.sh 生成的文件
     rm -rf "${TMP_DIR}"/*.s \
            "${TMP_DIR}"/*_sysyc_riscv64 \
            "${TMP_DIR}"/*_sysyc_riscv64.actual_out \
            "${TMP_DIR}"/*_sysyc_riscv64.expected_stdout \
-           "${TMP_DIR}"/*_sysyc_riscv64.o # 以防生成了 .o 文件
+           "${TMP_DIR}"/*_sysyc_riscv64.o
     echo "清理完成。"
 }
 
-# 如果临时目录不存在，则创建它 (尽管 runit.sh 应该已经创建了)
+# 如果临时目录不存在，则创建它
 mkdir -p "${TMP_DIR}"
 
 # 解析命令行参数
@@ -63,6 +69,24 @@ 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
+            ;;
+        -ct|--compile-timeout)
+            if [[ -n "$2" && "$2" =~ ^[0-9]+$ ]]; then
+                COMPILE_TIMEOUT_SECONDS="$2"
+                shift # 移过参数值
+            else
+                echo "错误: --compile-timeout 需要一个正整数参数。" >&2
+                exit 1
+            fi
+            ;;
         -h|--help)
             show_help
             exit 0
@@ -73,30 +97,33 @@ while [[ "$#" -gt 0 ]]; do
             exit 1
             ;;
     esac
+    shift # 移过参数名
 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 "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")
+TOTAL_CASES=$(echo "$s_files" | wc -w)
+
 # 遍历找到的每个 .s 文件
-find "${TMP_DIR}" -maxdepth 1 -name "*.s" | while read s_file; do
+echo "$s_files" | while read s_file; do
+    # --- 新增功能: 初始化用例通过状态 ---
+    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
-    # 例如：从 functional_21_if_test2_sysyc_riscv64.s 提取 functional_21_if_test2
     base_name_from_s_file=$(basename "$s_file" .s)
-    # 这一步得到的是 'functional_21_if_test2' 或 'performance_2024-2D0-22'
     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
-    # 例如：'functional_21_if_test2' 分割为 'functional' 和 '21_if_test2'
     category=$(echo "$original_test_name_underscored" | cut -d'_' -f1)
-    # cut -d'_' -f2- 会从第二个下划线开始获取所有剩余部分
     test_file_base=$(echo "$original_test_name_underscored" | cut -d'_' -f2-)
-
-    # 构建原始的相对路径，例如：'functional/21_if_test2'
     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
@@ -109,109 +136,112 @@ find "${TMP_DIR}" -maxdepth 1 -name "*.s" | 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
-    # 注意：这里假设 gcc 在 VM 环境中可用，且 ./lib 是相对于当前脚本运行目录
+    echo "  使用 gcc 汇编并链接 (超时 ${COMPILE_TIMEOUT_SECONDS}s)..."
-    echo "  使用 gcc 汇编并链接: ${GCC_NATIVE} \"${s_file}\" -o \"${executable_file}\" -L\"${LIB_DIR}\" -lsysy_riscv -static -g"
+    # --- 修改点: 为 gcc 增加 timeout ---
-    "${GCC_NATIVE}" "${s_file}" -o "${executable_file}" -L"${LIB_DIR}" -lsysy_riscv -static -g
+    timeout ${COMPILE_TIMEOUT_SECONDS} "${GCC_NATIVE}" "${s_file}" -o "${executable_file}" -L"${LIB_DIR}" -lsysy_riscv -static -g
-    if [ $? -ne 0 ]; then
+    GCC_STATUS=$?
-        echo -e "\e[31m错误: GCC 汇编/链接 ${s_file} 失败\e[0m"
+    if [ $GCC_STATUS -eq 124 ]; then
-        continue
+        echo -e "\e[31m错误: GCC 编译/链接 ${s_file} 超时 (超过 ${COMPILE_TIMEOUT_SECONDS} 秒)\e[0m"
-    fi
+        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
         echo "  生成的可执行文件: ${executable_file}"
+        echo "  正在执行 (超时 ${TIMEOUT_SECONDS}s): \"${executable_file}\""
 
         # 步骤 2: 执行编译后的文件并比较/报告结果
-    # 直接执行可执行文件，不再通过 qemu-riscv64
-    echo "  正在执行: \"${executable_file}\"" # 修改点：移除多余的 ./
-
-    # 检查是否存在 .out 文件
         if [ -f "${output_reference_file}" ]; then
-        # 尝试从 .out 文件中提取期望的返回码和期望的标准输出
-        # 获取 .out 文件的最后一行，去除空白字符
             LAST_LINE_TRIMMED=$(tail -n 1 "${output_reference_file}" | tr -d '[:space:]')
             
-        # 检查最后一行是否为纯整数 (允许正负号)
             if [[ "$LAST_LINE_TRIMMED" =~ ^[-+]?[0-9]+$ ]]; then
-            # 假设最后一行是期望的返回码
                 EXPECTED_RETURN_CODE="$LAST_LINE_TRIMMED"
-            
-            # 创建一个只包含期望标准输出的临时文件 (所有行除了最后一行)
                 EXPECTED_STDOUT_FILE="${TMP_DIR}/${base_name_from_s_file}.expected_stdout"
-            # 使用 head -n -1 来获取除了最后一行之外的所有行。如果文件只有一行，则生成一个空文件。
                 head -n -1 "${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: ${EXPECTED_RETURN_CODE}"
-            if [ -s "${EXPECTED_STDOUT_FILE}" ]; then # -s 检查文件是否非空
-                echo "  期望标准输出文件: ${EXPECTED_STDOUT_FILE}"
-            else
-                echo "  期望标准输出为空。"
-            fi
 
-            # 执行程序，捕获实际返回码和实际标准输出
                 if [ -f "${input_file}" ]; then
-                echo "  使用输入文件: ${input_file}"
+                    timeout ${TIMEOUT_SECONDS} "${executable_file}" < "${input_file}" > "${output_actual_file}"
-                "${executable_file}" < "${input_file}" > "${output_actual_file}" # 修改点：移除多余的 ./
                 else
-                "${executable_file}" > "${output_actual_file}" # 修改点：移除多余的 ./
+                    timeout ${TIMEOUT_SECONDS} "${executable_file}" > "${output_actual_file}"
                 fi
-            ACTUAL_RETURN_CODE=$? # 捕获执行状态
+                ACTUAL_RETURN_CODE=$?
 
-            # 比较实际返回码与期望返回码
+                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: ${original_relative_path}.sy 的返回码 (${ACTUAL_RETURN_CODE}) 与期望值 (${EXPECTED_RETURN_CODE}) 匹配\e[0m"
+                        echo -e "\e[32m  返回码测试成功: (${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: ${original_relative_path}.sy 的返回码不匹配。期望: ${EXPECTED_RETURN_CODE}, 实际: ${ACTUAL_RETURN_CODE}\e[0m"
+                        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
 
-            # 比较实际标准输出与期望标准输出，忽略文件末尾的换行符差异
                     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: 输出与 ${original_relative_path}.sy 的参考输出匹配 (忽略行尾换行符差异)\e[0m"
+                        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: ${original_relative_path}.sy 的输出不匹配\e[0m"
+                        echo -e "\e[31m  标准输出测试失败\e[0m"
-                echo "    差异 (可能包含行尾换行符差异):"
+                        echo "    差异:"
-                diff "${output_actual_file}" "${EXPECTED_STDOUT_FILE}" # 显示原始差异以便调试
+                        diff "${output_actual_file}" "${EXPECTED_STDOUT_FILE}"
+                        is_passed=0
+                    fi
                 fi
-            
             else
-            # 最后一行不是纯整数，将整个 .out 文件视为纯标准输出
+                echo "  检测到 .out 文件为纯标准输出参考。"
-            echo "  检测到 .out 文件为纯标准输出参考。正在与输出文件比较: ${output_reference_file}"
-            
-            # 执行程序，并将输出重定向到临时文件
                 if [ -f "${input_file}" ]; then
-                echo "  使用输入文件: ${input_file}"
+                    timeout ${TIMEOUT_SECONDS} "${executable_file}" < "${input_file}" > "${output_actual_file}"
-                "${executable_file}" < "${input_file}" > "${output_actual_file}" # 修改点：移除多余的 ./
                 else
-                "${executable_file}" > "${output_actual_file}" # 修改点：移除多余的 ./
+                    timeout ${TIMEOUT_SECONDS} "${executable_file}" > "${output_actual_file}"
                 fi
-            EXEC_STATUS=$? # 捕获执行状态
+                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: 可执行文件 ${original_relative_path}.sy 以非零状态 ${EXEC_STATUS} 退出 (纯输出比较模式)。请检查程序逻辑或其是否应返回此状态。\e[0m"
+                        echo -e "\e[33m警告: 程序以非零状态 ${EXEC_STATUS} 退出 (纯输出比较模式)。\e[0m"
+                    fi
+                    if diff -q <(sed ':a;N;$!ba;s/\n*$//' "${output_actual_file}") <(sed ':a;N;$!ba;s/\n*$//' "${output_reference_file}") >/dev/null 2>&1; then
+                        echo -e "\e[32m  成功: 输出与参考输出匹配\e[0m"
+                    else
+                        echo -e "\e[31m  失败: 输出不匹配\e[0m"
+                        echo "    差异:"
+                        diff "${output_actual_file}" "${output_reference_file}"
+                        is_passed=0
+                    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}"
+            fi
+        fi
     fi
 
-            # 比较实际输出与参考输出，忽略文件末尾的换行符差异
+    # --- 新增功能: 更新通过用例计数 ---
-            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 [ "$is_passed" -eq 1 ]; then
-                echo -e "\e[32m  成功: 输出与 ${original_relative_path}.sy 的参考输出匹配 (忽略行尾换行符差异)\e[0m"
+        ((PASSED_CASES++))
-            else
-                echo -e "\e[31m  失败: ${original_relative_path}.sy 的输出不匹配\e[0m"
-                echo "    差异 (可能包含行尾换行符差异):"
-                diff "${output_actual_file}" "${output_reference_file}" # 显示原始差异以便调试
     fi
-        fi
+    echo "" # 为测试用例之间添加一个空行
-    elif [ -f "${input_file}" ]; then
-        # 只有 .in 文件存在，使用输入运行并报告退出码（无参考输出）
-        echo "  使用输入文件: ${input_file}"
-        echo "  没有 .out 文件进行比较。正在运行并报告返回码。"
-        "${executable_file}" < "${input_file}" # 修改点：移除多余的 ./
-        EXEC_STATUS=$?
-        echo "  ${original_relative_path}.sy 的返回码: ${EXEC_STATUS}"
-    else
-        # .in 和 .out 文件都不存在，只运行并报告退出码
-        echo "  未找到 .in 或 .out 文件。正在运行并报告返回码。"
-        "${executable_file}" # 修改点：移除多余的 ./
-        EXEC_STATUS=$?
-        echo "  ${original_relative_path}.sy 的返回码: ${EXEC_STATUS}"
-    fi
-    echo "" # 为测试用例之间添加一个空行，以提高可读性
 done
 
-echo "脚本完成。"
+# --- 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}]"
+echo "========================================"
+
+if [ "$PASSED_CASES" -eq "$TOTAL_CASES" ]; then
+    exit 0
+else
+    exit 1
+fi

test_script/runit.sh

@@ -9,45 +9,40 @@ TESTDATA_DIR="${SCRIPT_DIR}/../testdata"
 BUILD_BIN_DIR="${SCRIPT_DIR}/../build/bin"
 LIB_DIR="${SCRIPT_DIR}/../lib"
 # TMP_DIR="${SCRIPT_DIR}/tmp"
-TMP_DIR="/home/ladev987/paraComp/debug/share_folder/tmp"
+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"
 
-# 标志，用于确定是否应该生成和运行可执行文件
+# --- 新增功能: 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 执行超时 (秒)
+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 "此脚本用于编译 .sy 文件，并可选择性地运行它们进行测试。"
+    echo "此脚本用于按文件名前缀数字升序编译和测试 .sy 文件。"
     echo ""
     echo "选项:"
-    echo "  -e, --executable    编译为可执行文件，运行可执行文件，并比较输出（如果存在 .in/.out 文件）。"
+    echo "  -e, --executable         编译为可执行文件并运行测试。"
-    echo "                      如果 .out 文件的最后一行是整数，则将其视为期望的返回值进行比较，其余内容视为期望的标准输出。"
-    echo "                      如果 .out 文件的最后一行不是整数，则将整个 .out 文件视为期望的标准输出进行比较。"
-    echo "                      输出比较时会忽略行尾多余的换行符。"
-    echo "                      如果不存在 .in/.out 文件，则打印返回码。"
     echo "  -c, --clean              清理 'tmp' 目录下的所有生成文件。"
+    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 "  -h, --help               显示此帮助信息并退出。"
-    echo ""
-    echo "编译步骤:"
-    echo "1. 调用 sysyc 将 .sy 编译为 .s (RISC-V 汇编)。"
-    echo "2. 调用 riscv64-linux-gnu-gcc 将 .s 编译为可执行文件，并链接 -L../lib/ -lsysy_riscv -static。"
-    echo "3. 调用 qemu-riscv64 执行编译后的文件。"
-    echo "4. 根据 .out 文件内容（最后一行是否为整数）决定是进行返回值比较、标准输出比较，或两者都进行。"
-    echo "5. 如果没有 .in/.out 文件，则打印可执行文件的返回值。"
 }
 
 # 清理临时文件的函数
 clean_tmp() {
     echo "正在清理临时目录: ${TMP_DIR}"
     rm -rf "${TMP_DIR}"/*
-    # 如果需要，也可以根据 clean.sh 示例清理其他特定文件
-    # rm -rf "${SCRIPT_DIR}"/*.s "${SCRIPT_DIR}"/*.ll "${SCRIPT_DIR}"/*clang "${SCRIPT_DIR}"/*sysyc
-    # rm -rf "${SCRIPT_DIR}"/*_riscv64
 }
 
 # 如果临时目录不存在，则创建它
@@ -58,12 +53,20 @@ while [[ "$#" -gt 0 ]]; do
     case "$1" in
         -e|--executable)
             EXECUTE_MODE=true
-            shift
             ;;
         -c|--clean)
             clean_tmp
             exit 0
             ;;
+        -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="$2"; shift; else echo "错误: -et 需要一个正整数参数。" >&2; exit 1; fi
+            ;;
         -h|--help)
             show_help
             exit 0
@@ -74,150 +77,175 @@ while [[ "$#" -gt 0 ]]; do
             exit 1
             ;;
     esac
+    shift
 done
 
 echo "SysY 测试运行器启动..."
 echo "输入目录: ${TESTDATA_DIR}"
 echo "临时目录: ${TMP_DIR}"
-echo "执行模式已启用: ${EXECUTE_MODE}"
+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"
+fi
 echo ""
 
-# 查找 testdata 目录及其子目录中的所有 .sy 文件
+# --- 修改点: 查找所有 .sy 文件并按文件名前缀数字排序 ---
-# 遍历找到的每个 .sy 文件
+sy_files=$(find "${TESTDATA_DIR}" -name "*.sy" | sort -V)
-find "${TESTDATA_DIR}" -name "*.sy" | while read sy_file; do
+TOTAL_CASES=$(echo "$sy_files" | wc -w)
-    # 获取 .sy 文件的基本名称（例如：21_if_test2）
+
-    # 这也处理了文件位于子目录中的情况（例如：functional/21_if_test2.sy）
+# --- 本次修复: 使用 here-string (<<<) 代替管道 (|) 来避免子 shell 问题 ---
+# 这样可以确保循环内的 PASSED_CASES 变量修改在循环结束后依然有效
+while IFS= read -r sy_file; do
+    is_passed=1 # 1 表示通过, 0 表示失败
+
     relative_path_no_ext=$(realpath --relative-to="${TESTDATA_DIR}" "${sy_file%.*}")
-    # 将斜杠替换为下划线，用于输出文件名，以避免冲突并保持结构
     output_base_name=$(echo "${relative_path_no_ext}" | tr '/' '_')
 
-    # 定义汇编文件、可执行文件、输入文件和输出文件的路径
     assembly_file="${TMP_DIR}/${output_base_name}_sysyc_riscv64.s"
     executable_file="${TMP_DIR}/${output_base_name}_sysyc_riscv64"
     input_file="${sy_file%.*}.in"
     output_reference_file="${sy_file%.*}.out"
     output_actual_file="${TMP_DIR}/${output_base_name}_sysyc_riscv64.actual_out"
 
-    echo "正在处理: $(basename "$sy_file")"
+    echo "正在处理: $(basename "$sy_file") (路径: ${relative_path_no_ext}.sy)"
-    echo "  SY 文件: ${sy_file}"
 
     # 步骤 1: 使用 sysyc 编译 .sy 到 .s
-    echo "  使用 sysyc 编译: ${SYSYC} -s asm \"${sy_file}\" > \"${assembly_file}\""
+    echo "  使用 sysyc 编译 (超时 ${SYSYC_TIMEOUT}s)..."
-    "${SYSYC}" -s asm "${sy_file}" > "${assembly_file}"
+    timeout ${SYSYC_TIMEOUT} "${SYSYC}" -S "${sy_file}" -o "${assembly_file}"
-    if [ $? -ne 0 ]; then
+    SYSYC_STATUS=$?
-        echo -e "\e[31m错误: SysY 编译 ${sy_file} 失败\e[0m"
+    if [ $SYSYC_STATUS -eq 124 ]; then
-        continue
+        echo -e "\e[31m错误: SysY 编译 ${sy_file} 超时\e[0m"
+        is_passed=0
+    elif [ $SYSYC_STATUS -ne 0 ]; then
+        echo -e "\e[31m错误: SysY 编译 ${sy_file} 失败，退出码: ${SYSYC_STATUS}\e[0m"
+        is_passed=0
     fi
-    echo "  生成的汇编文件: ${assembly_file}"
 
-    # 只有当 EXECUTE_MODE 为 true 时才继续生成和执行可执行文件
+    # 只有当 EXECUTE_MODE 为 true 且上一步成功时才继续
-    if ${EXECUTE_MODE}; then
+    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: ${GCC_RISCV64} \"${assembly_file}\" -o \"${executable_file}\" -L\"${LIB_DIR}\" -lsysy_riscv -static"
+        echo "  使用 gcc 编译 (超时 ${GCC_TIMEOUT}s)..."
-        "${GCC_RISCV64}" "${assembly_file}" -o "${executable_file}" -L"${LIB_DIR}" -lsysy_riscv -static
+        timeout ${GCC_TIMEOUT} "${GCC_RISCV64}" "${assembly_file}" -o "${executable_file}" -L"${LIB_DIR}" -lsysy_riscv -static
-        if [ $? -ne 0 ]; then
+        GCC_STATUS=$?
-            echo -e "\e[31m错误: GCC 编译 ${assembly_file} 失败\e[0m"
+        if [ $GCC_STATUS -eq 124 ]; then
+            echo -e "\e[31m错误: GCC 编译 ${assembly_file} 超时\e[0m"
+            is_passed=0
+        elif [ $GCC_STATUS -ne 0 ]; then
+            echo -e "\e[31m错误: GCC 编译 ${assembly_file} 失败，退出码: ${GCC_STATUS}\e[0m"
+            is_passed=0
+        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 ""
         continue
     fi
-        echo "  生成的可执行文件: ${executable_file}"
 
-        # 步骤 3, 4, 5: 执行编译后的文件并比较/报告结果
+    # 步骤 3, 4, 5: 只有当编译都成功时才执行
-        echo "  正在执行: ${QEMU_RISCV664} \"${executable_file}\""
+    if [ "$is_passed" -eq 1 ]; then
+        echo "  正在执行 (超时 ${EXEC_TIMEOUT}s)..."
         
-        # 检查是否存在 .out 文件
+        # 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} ${exec_cmd}"
+        ACTUAL_RETURN_CODE=$?
+
+        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
-            # 尝试从 .out 文件中提取期望的返回码和期望的标准输出
-            # 获取 .out 文件的最后一行，去除空白字符
                 LAST_LINE_TRIMMED=$(tail -n 1 "${output_reference_file}" | tr -d '[:space:]')
                 
-            # 检查最后一行是否为纯整数 (允许正负号)
                 if [[ "$LAST_LINE_TRIMMED" =~ ^[-+]?[0-9]+$ ]]; then
-                # 假设最后一行是期望的返回码
                     EXPECTED_RETURN_CODE="$LAST_LINE_TRIMMED"
-                
-                # 创建一个只包含期望标准输出的临时文件 (所有行除了最后一行)
                     EXPECTED_STDOUT_FILE="${TMP_DIR}/${output_base_name}_sysyc_riscv64.expected_stdout"
-                # 使用 head -n -1 来获取除了最后一行之外的所有行。如果文件只有一行，则生成一个空文件。
                     head -n -1 "${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。"
+                    # 比较返回码
-                echo "  期望返回码: ${EXPECTED_RETURN_CODE}"
-                if [ -s "${EXPECTED_STDOUT_FILE}" ]; then # -s 检查文件是否非空
-                    echo "  期望标准输出文件: ${EXPECTED_STDOUT_FILE}"
-                else
-                    echo "  期望标准输出为空。"
-                fi
-
-                # 执行程序，捕获实际返回码和实际标准输出
-                if [ -f "${input_file}" ]; then
-                    echo "  使用输入文件: ${input_file}"
-                    "${QEMU_RISCV64}" "${executable_file}" < "${input_file}" > "${output_actual_file}"
-                else
-                    "${QEMU_RISCV64}" "${executable_file}" > "${output_actual_file}"
-                fi
-                ACTUAL_RETURN_CODE=$? # 捕获执行状态
-
-                # 比较实际返回码与期望返回码
                     if [ "$ACTUAL_RETURN_CODE" -eq "$EXPECTED_RETURN_CODE" ]; then
-                    echo -e "\e[32m  返回码测试成功: ${sy_file} 的返回码 (${ACTUAL_RETURN_CODE}) 与期望值 (${EXPECTED_RETURN_CODE}) 匹配\e[0m"
+                        echo -e "\e[32m  返回码测试成功: (${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: ${sy_file} 的返回码不匹配。期望: ${EXPECTED_RETURN_CODE}, 实际: ${ACTUAL_RETURN_CODE}\e[0m"
+                        echo -e "\e[31m  返回码测试失败: 期望: ${EXPECTED_RETURN_CODE}, 实际: ${ACTUAL_RETURN_CODE}\e[0m"
+                        is_passed=0
                     fi
-
+                    # 比较标准输出
-                # 比较实际标准输出与期望标准输出，忽略文件末尾的换行符差异
-                # 使用 sed 命令去除文件末尾的所有换行符，再通过 diff 进行比较
                     if diff -q <(sed ':a;N;$!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: 输出与 ${sy_file} 的参考输出匹配 (忽略行尾换行符差异)\e[0m"
+                        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: ${sy_file} 的输出不匹配\e[0m"
+                        echo -e "\e[31m  标准输出测试失败\e[0m"
-                    echo "    差异 (可能包含行尾换行符差异):"
+                        is_passed=0
-                    diff "${output_actual_file}" "${EXPECTED_STDOUT_FILE}" # 显示原始差异以便调试
+                        echo -e "    \e[36m---------- 期望输出 ----------\e[0m"
+                        cat "${EXPECTED_STDOUT_FILE}"
+                        echo -e "    \e[36m---------- 实际输出 ----------\e[0m"
+                        cat "${output_actual_file}"
+                        echo -e "    \e[36m------------------------------\e[0m"
                     fi
-                
                 else
-                # 最后一行不是纯整数，将整个 .out 文件视为纯标准输出
+                    # 纯标准输出比较
-                echo "  检测到 .out 文件为纯标准输出参考。正在与输出文件比较: ${output_reference_file}"
+                    if [ $ACTUAL_RETURN_CODE -ne 0 ]; then
-                
+                        echo -e "\e[33m警告: 程序以非零状态 ${ACTUAL_RETURN_CODE} 退出 (纯输出比较模式)。\e[0m"
-                # 使用输入文件（如果存在）运行可执行文件，并将输出重定向到临时文件
-                if [ -f "${input_file}" ]; then
-                    echo "  使用输入文件: ${input_file}"
-                    "${QEMU_RISCV64}" "${executable_file}" < "${input_file}" > "${output_actual_file}"
-                else
-                    "${QEMU_RISCV64}" "${executable_file}" > "${output_actual_file}"
                     fi
-                EXEC_STATUS=$? # 捕获执行状态
-
-                if [ $EXEC_STATUS -ne 0 ]; then
-                    echo -e "\e[33m警告: 可执行文件 ${sy_file} 以非零状态 ${EXEC_STATUS} 退出 (纯输出比较模式)。请检查程序逻辑或其是否应返回此状态。\e[0m"
-                fi
-
-                # 比较实际输出与参考输出，忽略文件末尾的换行符差异
                     if diff -q <(sed ':a;N;$!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 ${sy_file} 的参考输出匹配 (忽略行尾换行符差异)\e[0m"
+                        echo -e "\e[32m  成功: 输出与参考输出匹配\e[0m"
                     else
-                    echo -e "\e[31m  失败: ${sy_file} 的输出不匹配\e[0m"
+                        echo -e "\e[31m  失败: 输出不匹配\e[0m"
-                    echo "    差异 (可能包含行尾换行符差异):"
+                        is_passed=0
-                    diff "${output_actual_file}" "${output_reference_file}" # 显示原始差异以便调试
+                        echo -e "    \e[36m---------- 期望输出 ----------\e[0m"
+                        cat "${output_reference_file}"
+                        echo -e "    \e[36m---------- 实际输出 ----------\e[0m"
+                        cat "${output_actual_file}"
+                        echo -e "    \e[36m------------------------------\e[0m"
                     fi
                 fi
-        elif [ -f "${input_file}" ]; then
-            # 只有 .in 文件存在，使用输入运行并报告退出码（无参考输出）
-            echo "  使用输入文件: ${input_file}"
-            echo "  没有 .out 文件进行比较。正在运行并报告返回码。"
-            "${QEMU_RISCV64}" "${executable_file}" < "${input_file}"
-            EXEC_STATUS=$?
-            echo "  ${sy_file} 的返回码: ${EXEC_STATUS}"
-        else
-            # .in 和 .out 文件都不存在，只运行并报告退出码
-            echo "  未找到 .in 或 .out 文件。正在运行并报告返回码。"
-            "${QEMU_RISCV64}" "${executable_file}"
-            EXEC_STATUS=$?
-            echo "  ${sy_file} 的返回码: ${EXEC_STATUS}"
-        fi
             else
-        echo "  跳过执行模式。仅生成汇编文件。"
+                # 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
-    echo "" # 为测试用例之间添加一个空行，以提高可读性
-done
 
-echo "脚本完成。"
+    # 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
+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 "${FAILED_CASES_LIST}"
+fi
+
+echo "========================================"
+
+if [ "$PASSED_CASES" -eq "$TOTAL_CASES" ]; then
+    exit 0
+else
+    exit 1
+fi

testdata/functional/00_main.out

@@ -0,0 +1 @@
+3

testdata/functional/00_main.sy

@@ -0,0 +1,3 @@
+int main(){
+    return 3;
+}

testdata/functional/01_var_defn2.out

@@ -1,2 +1 @@
 10
-0

testdata/functional/01_var_defn2.sy

@@ -0,0 +1,8 @@
+//test domain of global var define and local define
+int a = 3;
+int b = 5;
+
+int main(){
+    int a = 5;
+    return a + b;
+}

testdata/functional/02_var_defn3.out

@@ -0,0 +1 @@
+5

testdata/functional/02_var_defn3.sy

@@ -0,0 +1,8 @@
+//test local var define
+int main(){
+    int a, b0, _c;
+    a = 1;
+    b0 = 2;
+    _c = 3;
+    return b0 + _c;
+}

testdata/functional/03_arr_defn2.out

@@ -0,0 +1 @@
+0

testdata/functional/03_arr_defn2.sy

@@ -0,0 +1,4 @@
+int a[10][10];
+int main(){
+    return 0;
+}

testdata/functional/04_arr_defn3.out

@@ -0,0 +1 @@
+14

testdata/functional/04_arr_defn3.sy

@@ -0,0 +1,9 @@
+//test array define
+int main(){
+    int a[4][2] = {};
+    int b[4][2] = {1, 2, 3, 4, 5, 6, 7, 8};
+    int c[4][2] = {{1, 2}, {3, 4}, {5, 6}, {7, 8}};
+    int d[4][2] = {1, 2, {3}, {5}, 7 , 8};
+    int e[4][2] = {{d[2][1], c[2][1]}, {3, 4}, {5, 6}, {7, 8}};
+    return e[3][1] + e[0][0] + e[0][1] + a[2][0];
+}

testdata/functional/05_arr_defn4.out

@@ -0,0 +1 @@
+21

testdata/functional/05_arr_defn4.sy

@@ -0,0 +1,9 @@
+int main(){
+    const int a[4][2] = {{1, 2}, {3, 4}, {}, 7};
+
+    int b[4][2] = {};
+    int c[4][2] = {1, 2, 3, 4, 5, 6, 7, 8};
+    int d[3 + 1][2] = {1, 2, {3}, {5}, a[3][0], 8};
+    int e[4][2][1] = {{d[2][1], {c[2][1]}}, {3, 4}, {5, 6}, {7, 8}};
+    return e[3][1][0] + e[0][0][0] + e[0][1][0] + d[3][0];
+}

testdata/functional/06_const_var_defn2.out

@@ -0,0 +1 @@
+5

testdata/functional/06_const_var_defn2.sy

@@ -0,0 +1,6 @@
+//test const gloal var define
+const int a = 10, b = 5;
+
+int main(){
+    return b;
+}

testdata/functional/07_const_var_defn3.out

@@ -0,0 +1 @@
+5

testdata/functional/07_const_var_defn3.sy

@@ -0,0 +1,5 @@
+//test const local var define
+int main(){
+    const int a = 10, b = 5;
+    return b;
+}

testdata/functional/08_const_array_defn.out

@@ -0,0 +1 @@
+4

testdata/functional/08_const_array_defn.sy

@@ -0,0 +1,5 @@
+const int a[5]={0,1,2,3,4};
+
+int main(){
+    return a[4];
+}

testdata/functional/09_func_defn.out

@@ -0,0 +1 @@
+9

testdata/functional/09_func_defn.sy

@@ -0,0 +1,11 @@
+int a;
+int func(int p){
+	p = p - 1;
+	return p;
+}
+int main(){
+	int b;
+	a = 10;
+	b = func(a);
+	return b;
+}

testdata/functional/10_var_defn_func.out

@@ -0,0 +1 @@
+4

testdata/functional/10_var_defn_func.sy

@@ -0,0 +1,8 @@
+int defn(){
+    return 4;
+}
+
+int main(){
+    int a=defn();
+    return a;
+}

testdata/functional/11_add2.out

@@ -0,0 +1 @@
+9

testdata/functional/11_add2.sy

@@ -0,0 +1,7 @@
+//test add
+int main(){
+    int a, b;
+    a = 10;
+    b = -1;
+    return a + b;
+}

testdata/functional/12_addc.out

@@ -0,0 +1 @@
+15

testdata/functional/12_addc.sy

@@ -0,0 +1,5 @@
+//test addc
+const int a = 10;
+int main(){
+    return a + 5;
+}

testdata/functional/13_sub2.out

@@ -0,0 +1 @@
+248

testdata/functional/13_sub2.sy

@@ -0,0 +1,7 @@
+//test sub
+const int a = 10;
+int main(){
+    int b;
+    b = 2;
+    return b - a;
+}

testdata/functional/14_subc.out

@@ -0,0 +1 @@
+8

testdata/functional/14_subc.sy

@@ -0,0 +1,6 @@
+//test subc
+int main(){
+    int a;
+    a = 10;
+    return a - 2;
+}

testdata/functional/15_mul.out

@@ -0,0 +1 @@
+50

testdata/functional/15_mul.sy

@@ -0,0 +1,7 @@
+//test mul
+int main(){
+    int a, b;
+    a = 10;
+    b = 5;
+    return a * b;
+}

testdata/functional/16_mulc.out

@@ -0,0 +1 @@
+25

testdata/functional/16_mulc.sy

@@ -0,0 +1,5 @@
+//test mulc
+const int a = 5;
+int main(){
+    return a * 5;
+}

testdata/functional/17_div.out

@@ -0,0 +1 @@
+2

testdata/functional/17_div.sy

@@ -0,0 +1,7 @@
+//test div
+int main(){
+    int a, b;
+    a = 10;
+    b = 5;
+    return a / b;
+}

testdata/functional/18_divc.out

@@ -0,0 +1 @@
+2

testdata/functional/18_divc.sy

@@ -0,0 +1,5 @@
+//test divc
+const int a = 10;
+int main(){
+    return a / 5;
+}

testdata/functional/19_mod.out

@@ -0,0 +1 @@
+3

testdata/functional/19_mod.sy

@@ -0,0 +1,6 @@
+//test mod
+int main(){
+    int a;
+    a = 10;
+    return a / 3;
+}

testdata/functional/20_rem.out

@@ -0,0 +1 @@
+1

testdata/functional/20_rem.sy

@@ -0,0 +1,6 @@
+//test rem
+int main(){
+    int a;
+    a = 10;
+    return a % 3;
+}

testdata/functional/22_if_test3.out

@@ -0,0 +1 @@
+25

testdata/functional/22_if_test3.sy

@@ -0,0 +1,18 @@
+// test if-if-else
+int ififElse() {
+  int a;
+  a = 5;
+  int b;
+  b = 10;
+  if(a == 5)
+    if (b == 10) 
+      a = 25;
+    else 
+      a = a + 15;
+    
+  return (a);
+}
+
+int main(){
+  return (ififElse());
+}

testdata/functional/23_if_test4.out

@@ -0,0 +1 @@
+25

testdata/functional/23_if_test4.sy

@@ -0,0 +1,18 @@
+// test if-{if-else}
+int if_ifElse_() {
+  int a;
+  a = 5;
+  int b;
+  b = 10;
+  if(a == 5){
+    if (b == 10) 
+      a = 25;
+    else 
+      a = a + 15;
+  }
+  return (a);
+}
+
+int main(){
+  return (if_ifElse_());
+}

testdata/functional/24_if_test5.out

@@ -0,0 +1 @@
+25

testdata/functional/24_if_test5.sy

@@ -0,0 +1,18 @@
+// test if-{if}-else
+int if_if_Else() {
+  int a;
+  a = 5;
+  int b;
+  b = 10;
+  if(a == 5){
+    if (b == 10) 
+      a = 25;
+  }
+    else 
+      a = a + 15;
+  return (a);
+}
+
+int main(){
+  return (if_if_Else());
+}

testdata/functional/25_while_if.out

@@ -0,0 +1,2 @@
+88
+0

testdata/functional/25_while_if.sy

@@ -0,0 +1,31 @@
+int get_one(int a) {
+  return 1;
+}
+
+int deepWhileBr(int a, int b) {
+  int c;
+  c = a + b;
+  while (c < 75) {
+    int d;
+    d = 42;
+    if (c < 100) {
+      c = c + d;
+      if (c > 99) {
+        int e;
+        e = d * 2;
+        if (get_one(0) == 1) {
+          c = e * 2;
+        }
+      }
+    }
+  }
+  return (c);
+}
+
+int main() {
+  int p;
+  p = 2;
+  p = deepWhileBr(p, p);
+  putint(p);
+  return 0;
+}

testdata/functional/27_while_test2.out

@@ -0,0 +1 @@
+54

testdata/functional/27_while_test2.sy

@@ -0,0 +1,31 @@
+int FourWhile() {
+  int a;
+  a = 5;
+  int b;
+  int c;
+  b = 6;
+  c = 7;
+  int d;
+  d = 10;
+  while (a < 20) {
+    a = a + 3;
+    while(b < 10){
+      b = b + 1;
+      while(c == 7){
+        c = c - 1;
+        while(d < 20){
+          d = d + 3;
+        }
+        d = d - 1;
+      }
+      c = c + 1;
+    }
+    b = b - 2;
+  }
+  
+  return (a + (b + d) + c);
+}
+
+int main() {
+  return FourWhile();
+}

testdata/functional/28_while_test3.out

@@ -0,0 +1 @@
+23

testdata/functional/28_while_test3.sy

@@ -0,0 +1,55 @@
+int g;
+int h;
+int f;
+int e;
+int EightWhile() {
+  int a;
+  a = 5;
+  int b;
+  int c;
+  b = 6;
+  c = 7;
+  int d;
+  d = 10;
+  while (a < 20) {
+    a = a + 3;
+    while(b < 10){
+      b = b + 1;
+      while(c == 7){
+        c = c - 1;
+        while(d < 20){
+          d = d + 3;
+          while(e > 1){
+            e = e-1;
+            while(f > 2){
+              f = f -2;
+              while(g < 3){
+                g = g +10;
+                while(h < 10){
+                  h = h + 8;
+                }
+                h = h-1;
+              }
+              g = g- 8;
+            }
+            f = f + 1;
+          }
+          e = e + 1;
+        }
+        d = d - 1;
+      }
+      c = c + 1;
+    }
+    b = b - 2;
+  }
+  
+  return (a + (b + d) + c)-(e + d - g + h);
+}
+
+int main() {
+  g = 1;
+  h = 2;
+  e = 4;
+  f = 6;
+  return EightWhile();
+}