使用std :: vector分配的类比LOT分配指针的速度慢

Question

我有一个用于处理数组分配的类.我的课很简单,定义如下:

DimArray.hpp:

#ifndef DIMARRAY_HPP_INCLUDED
#define DIMARRAY_HPP_INCLUDED

#include <vector>

template<typename T>
class DimArray
{
    private:
        int Width, Height;
        std::vector<T> Data;

    public:
        DimArray(int Width, int Height);
        DimArray(T* Data, int Width, int Height);
        DimArray(T** Data, int Width, int Height);

        DimArray(const DimArray &da);
        DimArray(DimArray &&da);

        inline int size() {return Width * Height;}
        inline int size() const {return Width * Height;}

        inline int width() {return Width;}
        inline int width() const {return Width;}

        inline int height() {return Height;}
        inline int height() const {return Height;}

        inline T* operator [](int Index) {return const_cast<T*>(Data.data()) + Height * Index;}
        inline const T* operator [](int Index) const {return Data.data() + Height * Index;}

        DimArray& operator = (DimArray da);
};

template<typename T>
DimArray<T>::DimArray(int Width, int Height) : Width(Width), Height(Height), Data(Width * Height, 0) {}

template<typename T>
DimArray<T>::DimArray(T* Data, int Width, int Height) : Width(Width), Height(Height), Data(Width * Height, 0) {std::copy(Data, Data + Width * Height, const_cast<T*>(this->Data.data()));}

template<typename T>
DimArray<T>::DimArray(T** Data, int Width, int Height) : Width(Width), Height(Height), Data(Width * Height, 0) {std::copy(Data[0], Data[0] + Width * Height, const_cast<T*>(this->Data.data()));}

template<typename T>
DimArray<T>::DimArray(const DimArray &da) : Width(da.Width), Height(da.Height), Data(da.Data) {}

template<typename T>
DimArray<T>::DimArray(DimArray &&da) : Width(std::move(da.Width)), Height(std::move(da.Height)), Data(std::move(da.Data)) {}

template<typename T>
DimArray<T>& DimArray<T>::operator = (DimArray<T> da)
{
    this->Width = da.Width;
    this->Height = da.Height;
    this->Data.swap(da.Data);
    return *this;
}

#endif // DIMARRAY_HPP_INCLUDED

对于基准时序,我使用以下内容:

Timer.hpp:

#ifndef TIME_HPP_INCLUDED
#define TIME_HPP_INCLUDED

#include <chrono>

#if defined _WIN32 || defined _WIN64
#include <windows.h>

template<typename T>
class Timer
{
    private:
        typedef T duration;
        typedef typename T::rep rep;
        typedef typename T::period period;
        typedef std::chrono::time_point<Timer, duration> time_point;
        std::chrono::time_point<Timer, duration> Time;
        static const bool is_steady = true;

        const rep g_Frequency = []() -> rep
        {
            LARGE_INTEGER frequency;
            QueryPerformanceFrequency(&frequency);
            return frequency.QuadPart;
        }();

        inline std::chrono::time_point<Timer, duration> now()
        {
            LARGE_INTEGER count;
            QueryPerformanceCounter(&count);
            return time_point(duration(count.QuadPart * static_cast<rep>(period::den) / g_Frequency));
        }

    public:
        inline void Start() {this->Time = this->now();}
        inline rep End() {return std::chrono::duration_cast<T>(this->now() - this->Time).count();}
};
#else
template<typename T>
class Timer
{
    private:
        static const bool is_steady = true;
        std::chrono::high_resolution_clock Clock;
        std::chrono::time_point<std::chrono::high_resolution_clock> Time;

    public:
        inline void Start() {this->Time = this->Clock.now();}
        inline T::rep End() {return std::chrono::duration_cast<T>(this->Clock.now() - this->Time).count();}
};
#endif

#endif // TIME_HPP_INCLUDED

我的基准如下:

int main()
{
    Timer<std::chrono::nanoseconds> T;

    T.Start();
    for (int i = 0; i < 100; ++i)
    {
        int** T2DArray = new int*[10000];
        for (int i = 0; i < 10000; ++i)
        {
            T2DArray[i] = new int[10000];
        }

        for (int i = 0; i < 10000; ++i)
        {
            delete[] T2DArray[i];
        }
        delete[] T2DArray;
    }
    std::cout<<T.End()<<" us\n\n";

    T.Start();
    for (int i = 0; i < 100; ++i)
    {
        DimArray<int> TwoDArray(10000, 10000);
    }
    std::cout<<T.End()<<" us\n\n";
}

它打印的结果是:

4.9599256 seconds  //for int**
42.9303941 seconds //for DimArray<int>

这是一个巨大的差异!我弄不清楚为什么？!

所以我改成了:

int main()
{
    Timer<std::chrono::nanoseconds> T;

    T.Start();
    for (int i = 0; i < 100; ++i)
    {
        int** T2DArray = new int*[10000];
        for (int i = 0; i < 10000; ++i)
        {
            T2DArray[i] = new int[10000];
        }

        for (int i = 0; i < 10000; ++i)
        {
            delete[] T2DArray[i];
        }
        delete[] T2DArray;
    }
    std::cout<<T.End()<<" us\n\n";

    T.Start();
    for (int i = 0; i < 100; ++i)
    {
        int* TwoDArray = new int[10000 * 10000];
        delete[] TwoDArray;
    }
    std::cout<<T.End()<<" us\n\n";
}

结果如下:

4.6085725 seconds //for int**
0.1142958 seconds //for int*

std::vector与使用原始指针相比,我的类使用的任何想法都是如此之慢？

Answer 1

vector将为你分配的内存归零.你的代码new为你提供了"垃圾初始化"内存.因此,分配TwoDArray(10000, 10000)为您提供了一个充满零的数组,同时new int[10000 * 10000]为您提供了一个未定义内容的数组.(仅仅观察导致未定义的行为)

请注意,这意味着在向量的情况下,您的程序实际上写入所有100000000 ints,而在这种new情况下,您的程序只为那么多ints 留出地址空间.

对于可比较的测量,您必须先将新阵列归零; 例如,int* TwoDArray = new int[10000 * 10000]();而不是int* TwoDArray = new int[10000 * 10000];.