Question

In VS 2013这是代码，当使用Openmp在for循环中引入并行性时，它没有提高速度。我想把时间缩短到20ms。可能吗？

当有或没有Openmp时，当前执行时间约为150ms。

speedtest.pro

TEMPLATE = app
CONFIG += console c++11
CONFIG -= app_bundle
CONFIG -= qt
SOURCES += main.cpp
QMAKE_CXXFLAGS += -openmp

的main.cpp

#include <iostream>
#include <chrono>
#include <ctime>
#include <omp.h>
#include <conio.h>
using namespace std;

void fun(bool *c, int *a, int aSize, int *b, int bSize)
{
    #pragma omp parallel for
    for (int i = 0; i<aSize; i++){
        for (int j = 0; j<bSize; j++)
            if ((a[i*3] - b[j*3])*(a[i*3] - b[j*3]) + 
                (a[i*3 + 1] - b[j*3 + 1])*(a[i*3 + 1] - b[j*3 + 1]) 
              + (a[i*3 + 2] - b[j*3 + 2])*(a[i*3 + 2] - b[j*3 + 2])<30)
                c[i*bSize + j] = 1;
    }
}


int main()
{

    const int aSize = 16000;
    const int bSize = 6000;

    int *a = new int[aSize*3];
    int *b = new int[bSize*3];
    bool *c = new bool[aSize*bSize];

    for (unsigned int i = 0; i < aSize; i++){
        a[i*3] = (int)i;
        a[i*3 + 1] = (int)i;
        a[i*3 + 2] = (int)i;
    }
    for (unsigned int i = 0; i < bSize; i++){
        b[i*3] = (int)i;
        b[i*3 + 1] = (int)i;
        b[i*3 + 2] = (int)i;
    }

    for (unsigned int i = 0; i < aSize*bSize; i++){
        c[i] = 0;
    }

    std::chrono::time_point<std::chrono::system_clock> start, end;
    start = std::chrono::system_clock::now();



    fun(c, a, aSize, b, bSize);


    end = std::chrono::system_clock::now();


    int c1 = 0;
    for (unsigned int i = 0; i < aSize*bSize; i++){
        if (c[i] == 0){
            c1 = 1; break;
        }
    }

    if (c1 == 1)
        cout << "Collision" << endl;
    else
        cout << "No collision" << endl;

    std::chrono::duration<double> elapsed_seconds = end - start;
    std::time_t end_time = std::chrono::system_clock::to_time_t(end);

    std::cout << "finished computation at " << std::ctime(&end_time)
        << "elapsed time: " << 1000 * elapsed_seconds.count() << "ms\n";
    getch();
        delete [] a;
        delete [] b;
        delete [] c;


    return 0;
}

Answer 1

您需要在for #pragma omp parallel for

之前添加

#pragma omp parallel for
for(int i=0;i<N;++i)
{
   //....
}

使用Qt，您需要在.pro文件中添加：

QMAKE_CXXFLAGS += -fopenmp
QMAKE_LFLAGS += -fopenmp

Answer 2

不要用时钟测量时间！它测量时钟滴答。如果您使用的是CPU的许多内核，结果可能是错误的。

使用例如std::chrono

使用Openmp的Qt并没有在Windows中提供任何加速，我使用的是Visual Studio 2013和Qt 5.5

2 个答案: