使用boost singleton_pool的自定义分配比默认值慢

Question

我为MyOrder类编写了自定义运算符new和operator delete.我使用boost :: singleton pool分配内存.这是测试性能的程序,

#include <boost/pool/singleton_pool.hpp>
#include <boost/progress.hpp>
#include <iostream>
#include <new>
#include <vector>


class MyOrder{
    std::vector<int> v1_;
    std::vector<double> v2_;

    std::string s1_;
    std::string s2_;

public:
    MyOrder(std::string s1, std::string s2): s1_(s1), s2_(s2) {}

    ~MyOrder(){}

    static void * operator new(size_t size); 
    static void operator delete(void * rawMemory) throw();
};

struct MyOrderTag{};
typedef boost::singleton_pool<MyOrderTag, sizeof(MyOrder)> MyOrderPool; 

void* MyOrder:: operator new(size_t size)
{
    if (size != sizeof(MyOrder)) 
        return ::operator new(size);

    while(true){
        void * ptr = MyOrderPool::malloc();
        if (ptr != NULL) return ptr;

        std::new_handler globalNewHandler = std::set_new_handler(0);
        std::set_new_handler(globalNewHandler);

        if(globalNewHandler)  globalNewHandler();
        else throw std::bad_alloc();

    }
}

void MyOrder::operator delete(void * rawMemory) throw()
{
    if(rawMemory == 0) return; 
    MyOrderPool::free(rawMemory);
}

int main()
{
    MyOrder* mo = NULL; 
    std::vector<MyOrder*> v;
    v.reserve(100000);

    boost::progress_timer howlong;
    for(int i = 0; i< 100000; ++i)
    {
        mo = new MyOrder("Sanket", "Sharma");
        v.push_back(mo);
    }

    for (std::vector<MyOrder*>::const_iterator it = v.begin(); it != v.end(); ++it)
    {
        delete *it;
    }
    return 0;
}

我使用-O2标志编译了上述程序,并在我的Macbook上运行了2.26 GHz Intel Core 2 Duo,耗时0.16秒.然后我评论了我声明的行并定义了自定义运算符new和operator delete,使用-O2标志重新编译并在同一台机器上运行花了0.13秒.

使用singleton_pool为相同大小的对象分配和释放内存应加快速度.为什么它变慢？或者是创建池的开销是否会使这个小程序中获得的性能优势无效？

更新:

我用int和double替换了两个std :: string变量,这次在3.0 GHZ AMD Phenom(tm)II X4 945处理器上运行两个程序,每次迭代100000000(即1000次)迭代.使用自定义内存分配的一个需要3.2秒,而使用默认内存分配的那个需要8.26秒.所以这次自定义内存分配获胜.

Answer 1

我认为你的数字毫无意义.如果您只检查了一次运行时,并且您发现0.13vs 0.16秒比完全没有意义,并且由开销支配.

您必须运行要测试数千次的代码段,然后比较数据以排除开销.

不是真的,0.03你的流程可以很容易地解释出秒差异等等.