LRU是Least Recently Used的缩写,意思是最近最少使用,它是一种Cache替换算法。
实现思路: hashtable + 双向链表
时间复杂度: 插入,查找,删除:O(1)
空间使用情况: O(N) :一个链表存储K个数据(stl的hash_map实际占的空间比较大).
运行环境:
linux:redhat , fedora ,centos等(理论上ubuntu , debian,mac os等也可以运行)
代码:
#ifndef __LRUCACHE_H__ #define __LRUCACHE_H__ #include <vector> #include <ext/hash_map> #include <pthread.h> #include <assert.h> using namespace __gnu_cxx; template <class K, class D> struct Node{ K key; D data; Node *prev, *next; }; template <class K, class D> class LRUCache{ public: LRUCache(size_t size , bool is_pthread_safe = false){ if(size <= 0) size = 1024; pthread_safe = is_pthread_safe; if(pthread_safe) pthread_mutex_init(&cached_mutex , NULL); entries = new Node<K,D>[size]; for(size_t i = 0; i < size; ++i) cached_entries.push_back(entries + i); head = new Node<K,D>; tail = new Node<K,D>; head->prev = NULL; head->next = tail; tail->prev = head; tail->next = NULL; } ~LRUCache(){ if(pthread_safe) pthread_mutex_destroy(&cached_mutex); delete head; delete tail; delete[] entries; } void Put(K key, D data); D Get(K key); private: void cached_lock(void){ if(pthread_safe) pthread_mutex_lock(&cached_mutex); } void cached_unlock(void){ if(pthread_safe) pthread_mutex_unlock(&cached_mutex); } void detach(Node<K,D>* node){ node->prev->next = node->next; node->next->prev = node->prev; } void attach(Node<K,D>* node){ node->prev = head; node->next = head->next; head->next = node; node->next->prev = node; } private: hash_map<K, Node<K,D>* > cached_map; vector<Node<K,D>* > cached_entries; Node<K,D> * head, *tail; Node<K,D> * entries; bool pthread_safe; pthread_mutex_t cached_mutex; }; template<class K , class D> void LRUCache<K,D>::Put(K key , D data){ cached_lock(); Node<K,D> *node = cached_map[key]; if(node){ detach(node); node->data = data; attach(node); } else{ if(cached_entries.empty()){ node = tail->prev; detach(node); cached_map.erase(node->key); } else{ node = cached_entries.back(); cached_entries.pop_back(); } node->key = key; node->data = data; cached_map[key] = node; attach(node); } cached_unlock(); } template<class K , class D> D LRUCache<K,D>::Get(K key){ cached_lock(); Node<K,D> *node = cached_map[key]; if(node){ detach(node); attach(node); cached_unlock(); return node->data; } else{ cached_unlock(); return D(); } } #endif
测试用例:
/* Compile: g++ -o app app.cpp LRUCache.cpp -lpthread Run: ./app */ #include <iostream> #include <string> #include "LRUCache.h" using namespace std; int main(void){ //int k = 10 , // max = 100; int k = 100000 , max = 1000000; LRUCache<int , int> * lru_cache = new LRUCache<int , int>(k , true); int tmp = 0; for(int i = 0 ; i < 2*k ; ++i){ tmp = rand() % max; lru_cache->Put(tmp, tmp + 1000000); cout<<tmp<<endl; } for(int i = 0 ; i < k ; ++i){ tmp = rand() % max; if(lru_cache->Get(tmp) == 0) cout<<"miss : "<<tmp<<endl; else cout<<"hit : "<<tmp<<" value : "<<lru_cache->Get(tmp)<<endl; } delete lru_cache; return 0; }
其实,上面的代码,有一些毛病的。改天我会继续改进。
例如:
1:冗余操作。cached_entries完全可以用一个counter代替。
2:过度抽象。
3:Get、Put的interface不合理。如果真的去实现一个磁盘block的LRU cache,就会发现之前的接口需要重写了。
不过对于大家理解LRU算法。应该有一定的帮助的。