问为什么valarray这么慢？
EN

Stack Overflow用户

提问于 2011-07-28 04:27:57

回答 6查看 11.4K关注 0票数 21

我尝试使用valarray，因为它在运算向量和矩阵时很像MATLAB。我首先做了一些性能检查，发现valarray无法达到Stroustrup在一书中所宣称的性能。

这个测试程序实际上做了500万次双精度乘法。我认为c= a*b至少可以与for循环的双类型元素乘法相媲美，但我完全错了。2008年，我在几台电脑和微软的Visual C++ 6.0和Visual Studio上进行了尝试。

顺便说一下，我在MATLAB上使用以下代码进行了测试：

len = 5*1024*1024;
a = rand(len, 1);
b = rand(len, 1);
c = zeros(len, 1);
tic;
c = a.*b;
toc;

结果是46ms。这个时间不是很精确；它只作为一个参考。

代码是：

#include <iostream>
#include <valarray>
#include <iostream>
#include "windows.h"

using namespace std;
SYSTEMTIME stime;
LARGE_INTEGER sys_freq;

double gettime_hp();

int main()
{
    enum { N = 5*1024*1024 };
    valarray<double> a(N), b(N), c(N);
    QueryPerformanceFrequency(&sys_freq);
    int i, j;
    for (j=0 ; j<8 ; ++j)
    {
        for (i=0 ; i<N ; ++i)
        {
            a[i] = rand();
            b[i] = rand();
        }

        double* a1 = &a[0], *b1 = &b[0], *c1 = &c[0];
        double dtime = gettime_hp();
        for (i=0 ; i<N ; ++i)
            c1[i] = a1[i] * b1[i];
        dtime = gettime_hp()-dtime;
        cout << "double operator* " << dtime << " ms\n";

        dtime = gettime_hp();
        c = a*b ;
        dtime = gettime_hp() - dtime;
        cout << "valarray operator* " << dtime << " ms\n";

        dtime = gettime_hp();
        for (i=0 ; i<N ; ++i)
            c[i] = a[i] * b[i];
        dtime = gettime_hp() - dtime;
        cout << "valarray[i] operator* " << dtime<< " ms\n";

        cout << "------------------------------------------------------\n";
    }
}

double gettime_hp()
{
    LARGE_INTEGER tick;
    extern LARGE_INTEGER sys_freq;
    QueryPerformanceCounter(&tick);
    return (double)tick.QuadPart * 1000.0 / sys_freq.QuadPart;
}

运行结果：(最大速度优化的发布模式)

double operator* 52.3019 ms
valarray operator* 128.338 ms
valarray[i] operator* 43.1801 ms
------------------------------------------------------
double operator* 43.4036 ms
valarray operator* 145.533 ms
valarray[i] operator* 44.9121 ms
------------------------------------------------------
double operator* 43.2619 ms
valarray operator* 158.681 ms
valarray[i] operator* 43.4871 ms
------------------------------------------------------
double operator* 42.7317 ms
valarray operator* 173.164 ms
valarray[i] operator* 80.1004 ms
------------------------------------------------------
double operator* 43.2236 ms
valarray operator* 158.004 ms
valarray[i] operator* 44.3813 ms
------------------------------------------------------

同样优化的调试模式：

double operator* 41.8123 ms
valarray operator* 201.484 ms
valarray[i] operator* 41.5452 ms
------------------------------------------------------
double operator* 40.2238 ms
valarray operator* 215.351 ms
valarray[i] operator* 40.2076 ms
------------------------------------------------------
double operator* 40.5859 ms
valarray operator* 232.007 ms
valarray[i] operator* 40.8803 ms
------------------------------------------------------
double operator* 40.9734 ms
valarray operator* 234.325 ms
valarray[i] operator* 40.9711 ms
------------------------------------------------------
double operator* 41.1977 ms
valarray operator* 234.409 ms
valarray[i] operator* 41.1429 ms
------------------------------------------------------
double operator* 39.7754 ms
valarray operator* 234.26 ms
valarray[i] operator* 39.6338 ms
------------------------------------------------------

c++

valarray

回答 6

Stack Overflow用户

发布于 2011-07-28 05:24:26

我刚刚在Linux x86-64系统(Sandy Bridge CPU)上尝试过：

gcc 4.5.0：

double operator* 9.64185 ms
valarray operator* 9.36987 ms
valarray[i] operator* 9.35815 ms

英特尔ICC 12.0.2：

double operator* 7.76757 ms
valarray operator* 9.60208 ms
valarray[i] operator* 7.51409 ms

在这两种情况下，我只使用了-O3，没有其他与优化相关的标志。

看起来MS C++编译器和/或valarray实现很烂。

下面是为Linux修改的OP代码：

#include <iostream>
#include <valarray>
#include <iostream>
#include <ctime>

using namespace std ;

double gettime_hp();

int main()
{
    enum { N = 5*1024*1024 };
    valarray<double> a(N), b(N), c(N) ;
    int i,j;
    for(  j=0 ; j<8 ; ++j )
    {
        for(  i=0 ; i<N ; ++i )
        {
            a[i]=rand();
            b[i]=rand();
        }

        double* a1 = &a[0], *b1 = &b[0], *c1 = &c[0] ;
        double dtime=gettime_hp();
        for(  i=0 ; i<N ; ++i ) c1[i] = a1[i] * b1[i] ;
        dtime=gettime_hp()-dtime;
        cout << "double operator* " << dtime << " ms\n" ;

        dtime=gettime_hp();
        c = a*b ;
        dtime=gettime_hp()-dtime;
        cout << "valarray operator* " << dtime << " ms\n" ;

        dtime=gettime_hp();
        for(  i=0 ; i<N ; ++i ) c[i] = a[i] * b[i] ;
        dtime=gettime_hp()-dtime;
        cout << "valarray[i] operator* " << dtime<< " ms\n" ;

        cout << "------------------------------------------------------\n" ;
    }
}

double gettime_hp()
{
    struct timespec timestamp;

    clock_gettime(CLOCK_REALTIME, &timestamp);
    return timestamp.tv_sec * 1000.0 + timestamp.tv_nsec * 1.0e-6;
}

票数 24

Stack Overflow用户

发布于 2011-09-07 05:12:11

valarray的全部要点是在向量机上速度快，这是x86机器所不具备的。

在非向量机上一个好的实现应该能够与您获得的性能相匹配，例如

for (i=0; i < N; ++i) 
    c1[i] = a1[i] * b1[i];

一个坏的当然不会，除非硬件中有一些东西可以加速并行处理，否则这将是你能做的最好的事情。

票数 6

Stack Overflow用户

发布于 2011-07-28 14:18:17

我最终通过使用延迟求值得到了这个结果。代码可能很难看，因为我刚刚开始学习这些C++高级概念。

代码如下：

#include <iostream>
#include <valarray>
#include <iostream>
#include "windows.h"

using namespace std;
SYSTEMTIME stime;
LARGE_INTEGER sys_freq;

double gettime_hp();

// To improve the c = a*b (it will generate a temporary first, assigned to 'c' and delete the temporary.
// Which causes the program really slow
// The solution is the expression template and let the compiler to decide when all the expression is known.


// Delayed evaluation
//typedef valarray<double> Vector;
class Vector;

class VecMul
{
    public:
        const Vector& va;
        const Vector& vb;
        //Vector& vc;
        VecMul(const Vector& v1, const Vector& v2): va(v1), vb(v2) {}
        operator Vector();
};

class Vector:public valarray<double>
{
    valarray<double> *p;

    public:
        explicit Vector(int n)
        {
            p = new valarray<double>(n);
        }
        Vector& operator = (const VecMul &m)
        {
            for(int i=0; i<m.va.size(); i++)
                (*p)[i] = (m.va)[i]*(m.vb)[i]; // Ambiguous
            return *this;
        }
        double& operator[](int i) const {return (*p)[i];} //const vector_type[i]
        int size()const {return (*p).size();}
};


inline VecMul operator*(const Vector& v1, const Vector& v2)
{
    return VecMul(v1, v2);
}


int main()
{
    enum {N = 5*1024*1024};
    Vector a(N), b(N), c(N);
    QueryPerformanceFrequency(&sys_freq);
    int i, j;
    for (j=0 ; j<8 ; ++j)
    {
        for (i=0 ; i<N ; ++i)
        {
            a[i] = rand();
            b[i] = rand();
        }

        double* a1 = &a[0], *b1 = &b[0], *c1 = &c[0];
        double dtime = gettime_hp();
        for (i=0 ; i<N ; ++i)
            c1[i] = a1[i] * b1[i];
        dtime = gettime_hp()-dtime;
        cout << "double operator* " << dtime << " ms\n";

        dtime = gettime_hp();
        c = a*b;
        dtime = gettime_hp()-dtime;
        cout << "valarray operator* " << dtime << " ms\n";

        dtime = gettime_hp();
        for (i=0 ; i<N ; ++i)
            c[i] = a[i] * b[i];
        dtime = gettime_hp() - dtime;
        cout << "valarray[i] operator* " << dtime << " ms\n";

        cout << "------------------------------------------------------\n";
    }
}

double gettime_hp()
{
    LARGE_INTEGER tick;
    extern LARGE_INTEGER sys_freq;
    QueryPerformanceCounter(&tick);
    return (double)tick.QuadPart*1000.0/sys_freq.QuadPart;
}

在Visual studio上运行的结果是：

double operator* 41.2031 ms
valarray operator* 43.8407 ms
valarray[i] operator* 42.49 ms

票数 3

页面原文内容由Stack Overflow提供。腾讯云小微IT领域专用引擎提供翻译支持

原文链接：

https://stackoverflow.com/questions/6850807

复制

相似问题

问为什么valarray这么慢？
EN

回答 6

Stack Overflow用户

Stack Overflow用户

Stack Overflow用户

社区

活动

资源

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问为什么valarray这么慢？EN

回答 6

Stack Overflow用户

Stack Overflow用户

Stack Overflow用户

社区

活动

资源

关于

腾讯云开发者

热门产品

热门推荐

更多推荐

问为什么valarray这么慢？
EN