C++11中对容器的各种循环遍历的效率比较

#include "CycleTimeTst.h"
#include <string>
#include <vector>
#include <list>
#include <limits>
#include <assert.h>
#include <QTime>
#include <QDebug>

class TimeUsedGuard
{
public:
    TimeUsedGuard(const std::string& msg)
        : m_msg(msg)
        , m_time(QTime::currentTime())
    {
    }
    ~TimeUsedGuard()
    {
        qDebug() << m_msg.c_str() << "use time:" << m_time.elapsed() << "ms";
    }

private:
    std::string     m_msg;
    QTime           m_time;
};

template <typename T>
T useData(T val)
{
    return std::sqrt(val);
}

template <typename _Container>
void normalForLoop(const _Container& datas)
{
    TimeUsedGuard timeUsedGuard(__FUNCTION__);
    size_t i = 0;
    for (; i < datas.size(); ++i)
    {
        useData(datas[i]);
    }
}

template <typename _Container>
void normalForLoopCallSizeOnce(const _Container& datas)
{
    TimeUsedGuard timeUsedGuard(__FUNCTION__);
    size_t i = 0;
    size_t size = datas.size();
    for (; i < size; ++i)
    {
        useData(datas[i]);
    }
}

template <typename _Container>
void iterator(const _Container& datas)
{
    TimeUsedGuard timeUsedGuard(__FUNCTION__);
    auto pos = datas.cbegin();
    for (; pos != datas.cend(); ++pos)
    {
        useData(*pos);
    }
}

template <typename _Container>
void iteratorCallCEndOnce(const _Container& datas)
{
    TimeUsedGuard timeUsedGuard(__FUNCTION__);
    auto pos = datas.cbegin();
    auto end = datas.cend();
    for (; pos != end; ++pos)
    {
        useData(*pos);
    }
}

template <typename _Container>
void qtForeach(const _Container& datas)
{
    TimeUsedGuard timeUsedGuard(__FUNCTION__);
    foreach (auto data, datas)
    {
        useData(data);
    }
}

template <typename _Container>
void stdForeach(const _Container& datas)
{
    TimeUsedGuard timeUsedGuard(__FUNCTION__);
    std::for_each(datas.cbegin(), datas.cend(), useData<typename _Container::value_type>);
}

template <typename _Container>
void rangeForLoop(const _Container& datas)
{
    TimeUsedGuard timeUsedGuard(__FUNCTION__);
    for (auto data : datas)
    {
        useData(data);
    }
}

template <typename _Container>
void rangeForLoopReference(const _Container& datas)
{
    TimeUsedGuard timeUsedGuard(__FUNCTION__);
    for (auto& data : datas)
    {
        useData(data);
    }
}

class CycleTimeTst
{
public:
    CycleTimeTst();
};

CycleTimeTst g_CycleTimeTst;
CycleTimeTst::CycleTimeTst()
{
    constexpr unsigned long MAX_AMOUNT = 100 * 1000 * 1000;

    // std::vector<int> datas;
    // for (unsigned long i = 1; i < MAX_AMOUNT; ++i)
    // {
    //     datas.push_back(i);
    // }
    std::list<double> datas;
    double d = 1.0;
    for (unsigned long i = 1; i < MAX_AMOUNT / 10; ++i)
    {
        datas.push_back(d);
    }

    // normalForLoop(datas);
    // normalForLoopCallSizeOnce(datas);
    iterator(datas);
    iteratorCallCEndOnce(datas);
    qtForeach(datas);
    stdForeach(datas);
    rangeForLoop(datas);
    rangeForLoopReference(datas);
    // vector<int> 100 * 1000 * 1000 times, debug & release:
    // normalForLoop use time: 1096 ms                   normalForLoop use time: 113 ms
    // normalForLoopCallSizeOnce use time: 926 ms        normalForLoopCallSizeOnce use time: 106 ms
    // iterator use time: 1941 ms                        iterator use time: 103 ms
    // iteratorCallCEndOnce use time: 1473 ms            iteratorCallCEndOnce use time: 101 ms
    // qtForeach use time: 1846 ms                       qtForeach use time: 262 ms
    // stdForeach use time: 1488 ms                      stdForeach use time: 120 ms
    // rangeForLoop use time: 1527 ms                    rangeForLoop use time: 101 ms
    // rangeForLoopReference use time: 1551 ms           rangeForLoopReference use time: 102 ms

    // list<double> 10 * 1000 * 1000 times, debug & release:
    // iterator use time: 206 ms                         iterator use time: 27 ms
    // iteratorCallCEndOnce use time: 183 ms             iteratorCallCEndOnce use time: 26 ms
    // qtForeach use time: 1493 ms                       qtForeach use time: 743 ms
    // stdForeach use time: 182 ms                       stdForeach use time: 27 ms
    // rangeForLoop use time: 186 ms                     rangeForLoop use time: 26 ms
    // rangeForLoopReference use time: 186 ms            rangeForLoopReference use time: 27 ms
}

个人总结:

vector支持随机访问,但无通用性,当采用list后,就不能再使用了,不可取.

迭代器与范围for循环所用时间无明显差异, qt提供的foreach显然是最差的方法, std算法for_each也还可以接受, 但范围for循环好看得多, 更好用啊!

在循环中调用调用cend方法,会频繁创建一个新的迭代器(不停的构造与析构),另外还有函数调用开销,在debug版本结果来看似乎确实有一定的影响,但在realease版本下所花时间几乎相等,也许编译器有优化吧!