openmp代码（并行）与串行代码的性能分析

Question

如何比较并行代码（使用OpenMP）和串行代码的性能？我正在使用以下方法

int arr[1000] = {1, 6, 1, 3, 1, 9, 7, 3, 2, 0, 5, 0, 8, 9, 8, 4, 4, 4, 0, 9, 6, 5, 9, 5, 9, 2, 5, 8, 6, 1, 0, 7, 7, 3, 2, 8, 3, 2, 3, 7, 2, 0, 7, 2, 9, 5, 8, 6, 2, 8, 5, 8, 5, 6, 3, 5, 8, 1, 3, 7, 2, 6, 6, 2, 1, 9, 0, 6, 1, 6, 3, 5, 6, 3, 0, 8, 0, 8, 4, 2, 7, 1, 0, 2, 7, 6, 9, 7, 7, 5, 4, 9, 3, 1, 1, 4, 2, 4, 1, 5, 2, 6, 0, 8, 9, 2, 6, 0, 1, 0, 2, 0, 3, 3, 4, 0, 1, 4, 8, 8, 1, 4, 9, 4, 7, 3, 8, 9, 9, 1, 4, 1, 8, 7, 9, 9, 9, 8, 9, 0, 0, 4, 2, 4, 9, 7, 6, 0, 3, 4, 8, 6, 1, 9, 0, 8, 2, 0, 8, 1, 2, 4, 2, 2, 1, 4, 1, 1, 4, 3, 3, 4, 9, 8, 0, 8, 7, 7, 8, 0, 3, 8, 8, 4, 7, 8, 5, 2, 0, 3, 3, 4, 9, 8, 6, 1, 4, 0, 4, 8, 5, 9, 4, 4, 7, 5, 2, 4, 2, 2, 6, 5, 2, 4, 2, 1, 4, 7, 3, 5, 2, 7, 9, 1, 7, 8, 4, 3, 0, 8, 1, 5, 8, 7, 1, 7, 2, 5, 2, 6, 9, 8, 2, 1, 5, 4, 2, 9, 1, 6, 6, 5, 5, 8, 6, 4, 6, 1, 7, 8, 1, 0, 3, 9, 7, 6, 7, 2, 1, 1, 8, 2, 9, 2, 3, 6, 8, 7, 8, 9, 5, 4, 4, 2, 2, 3, 6, 8, 4, 5, 6, 5, 7, 1, 7, 7, 9, 6, 9, 2, 7, 9, 4, 8, 2, 7, 5, 0, 7, 3, 2, 2, 9, 8, 7, 2, 3, 5, 2, 9, 1, 1, 5, 8, 4, 4, 5, 4, 0, 6, 6, 9, 8, 1, 7, 0, 0, 4, 2, 7, 9, 6, 2, 9, 7, 9, 1, 0, 4, 3, 0, 7, 6, 7, 8, 1, 1, 5, 5, 3, 4, 3, 2, 2, 4, 1, 2, 7, 6, 6, 4, 5, 3, 8, 4, 2, 9, 7, 2, 6, 3, 4, 3, 9, 1, 1, 0, 4, 9, 5, 7, 3, 9, 1, 5, 5, 5, 9, 2, 3, 5, 9, 8, 0, 9, 5, 2, 9, 4, 7, 5, 7, 1, 0, 7, 5, 4, 7, 9, 3, 5, 9, 8, 6, 2, 3, 1, 7, 2, 6, 0, 9, 7, 1, 2, 6, 8, 4, 5, 2, 3, 2, 2, 7, 3, 9, 2, 9, 6, 3, 2, 3, 2, 2, 9, 7, 5, 3, 4, 9, 9, 7, 8, 6, 0, 0, 4, 0, 7, 2, 4, 0, 4, 6, 9, 9, 5, 1, 0, 4, 5, 4, 7, 9, 6, 9, 6, 1, 2, 3, 0, 3, 2, 1, 1, 4, 1, 5, 4, 0, 7, 8, 3, 4, 5, 2, 5, 2, 6, 6, 6, 1, 0, 6, 2, 9, 5, 1, 0, 9, 6, 3, 4, 8, 4, 5, 2, 7, 2, 8, 8, 2, 6, 1, 6, 3, 5, 3, 6, 1, 1, 4, 4, 2, 0, 7, 1, 7, 0, 3, 8, 6, 6, 2, 6, 2, 7, 0, 0, 2, 8, 0, 4, 6, 3, 2, 0, 8, 5, 8, 2, 7, 2, 6, 1, 5, 5, 4, 4, 5, 9, 3, 3, 8, 7, 9, 0, 7, 1, 2, 9, 1, 2, 3, 8, 7, 5, 0, 8, 0, 8, 0, 9, 2, 6, 0, 7, 2, 6, 4, 9, 6, 7, 3, 4, 6, 4, 6, 3, 6, 9, 2, 7, 3, 5, 7, 1, 2, 7, 9, 5, 7, 1, 4, 0, 7, 7, 9, 1, 3, 3, 1, 1, 2, 4, 5, 9, 0, 4, 4, 6, 3, 7, 6, 8, 4, 3, 1, 7, 1, 2, 2, 8, 3, 6, 0, 1, 5, 0, 2, 1, 5, 5, 2, 0, 9, 0, 1, 0, 4, 5, 8, 7, 2, 4, 7, 7, 0, 9, 6, 1, 1, 8, 1, 5, 6, 4, 8, 2, 4, 0, 3, 1, 6, 5, 1, 7, 7, 4, 9, 1, 0, 0, 0, 4, 6, 8, 3, 6, 7, 9, 9, 0, 9, 3, 5, 6, 7, 3, 8, 3, 6, 3, 4, 4, 0, 8, 1, 8, 2, 3, 1, 4, 3, 2, 9, 1, 0, 4, 8, 9, 4, 9, 9, 3, 2, 7, 1, 9, 0, 1, 4, 8, 4, 9, 2, 7, 9, 6, 5, 1, 1, 6, 8, 4, 0, 9, 7, 2, 3, 5, 1, 9, 7, 3, 5, 9, 0, 6, 1, 2, 8, 5, 1, 4, 6, 5, 1, 5, 3, 8, 9, 4, 7, 7, 0, 9, 6, 8, 2, 9, 3, 5, 9, 2, 8, 4, 2, 0, 2, 5, 3, 2, 2, 6, 7, 9, 3, 0, 6, 7, 1, 5, 1, 0, 2, 2, 9, 0, 2, 1, 2, 7, 7, 3, 0, 7, 9, 4, 8, 1, 9, 3, 4, 1, 1, 3, 2, 6, 3, 9, 3, 6, 6, 7, 6, 1, 1, 6, 1, 3, 9, 3, 2, 6, 8, 2, 6, 7, 6, 4, 1, 5, 9, 5, 9, 2, 0, 3, 8, 5, 2, 4, 2, 9, 3, 8, 0, 6, 6, 3, 1, 6, 9, 3, 2, 7, 6, 0, 7, 2, 6, 8, 0, 5, 5, 9, 9, 5, 4, 8, 0, 7, 4, 2, 8, 9, 3, 0, 5, 9, 3, 6, 5, 4, 9, 0, 2, 7, 2, 9, 0, 9, 9, 2, 6, 4, 3, 6, 9, 7, 6, 1, 6, 0, 6, 4, 9, 9, 6, 6, 0, 2, 2, 6, 6, 3, 8, 8, 1, 0, 9, 3, 9, 8, 5, 6, 4, 8, 4, 3, 5, 0, 7, 2, 2, 3, 8, 3, 2, 5, 9, 2, 7, 1, 0, 5, 6, 0, 4}; 

clock_t begin, end; 
double time_spent; 

begin = clock(); 
/* here, do your time-consuming job */ 

    #pragma omp parallel for private(temp) 
    for(j=0;j<1000;j++){ 
     temp = arr[j]; 
     for(i=0;i<temp;temp--) 
     result[j]=result[j]*temp; 
    } 


end = clock(); 
time_spent = (double)(end - begin)/CLOCKS_PER_SEC; 
printf("\n\n%f",time_spent); 

但是每次运行代码时，都会得到不同的输出。我想看看代码的性能对于openmp和serial code是如何不同的。我应该用什么方法来实现相同？

Answer 1

由于计算机/服务器的使用情况，代码运行的时间会稍微有点变化;但是，如果您同时运行并行版本和串行版本，则应该在两者之间看到运行时间差异。另外，并行操作的大小非常小。但你应该看到和改进。

int arr[1000] = {1, 6, 1, 3, 1, 9, 7, 3, 2, 0, 5, 0, 8, 9, 8, 4, 4, 4, 0, 9, 6, 5, 9, 5, 9, 2, 5, 8, 6, 1, 0, 7, 7, 3, 2, 8, 3, 2, 3, 7, 2, 0, 7, 2, 9, 5, 8, 6, 2, 8, 5, 8, 5, 6, 3, 5, 8, 1, 3, 7, 2, 6, 6, 2, 1, 9, 0, 6, 1, 6, 3, 5, 6, 3, 0, 8, 0, 8, 4, 2, 7, 1, 0, 2, 7, 6, 9, 7, 7, 5, 4, 9, 3, 1, 1, 4, 2, 4, 1, 5, 2, 6, 0, 8, 9, 2, 6, 0, 1, 0, 2, 0, 3, 3, 4, 0, 1, 4, 8, 8, 1, 4, 9, 4, 7, 3, 8, 9, 9, 1, 4, 1, 8, 7, 9, 9, 9, 8, 9, 0, 0, 4, 2, 4, 9, 7, 6, 0, 3, 4, 8, 6, 1, 9, 0, 8, 2, 0, 8, 1, 2, 4, 2, 2, 1, 4, 1, 1, 4, 3, 3, 4, 9, 8, 0, 8, 7, 7, 8, 0, 3, 8, 8, 4, 7, 8, 5, 2, 0, 3, 3, 4, 9, 8, 6, 1, 4, 0, 4, 8, 5, 9, 4, 4, 7, 5, 2, 4, 2, 2, 6, 5, 2, 4, 2, 1, 4, 7, 3, 5, 2, 7, 9, 1, 7, 8, 4, 3, 0, 8, 1, 5, 8, 7, 1, 7, 2, 5, 2, 6, 9, 8, 2, 1, 5, 4, 2, 9, 1, 6, 6, 5, 5, 8, 6, 4, 6, 1, 7, 8, 1, 0, 3, 9, 7, 6, 7, 2, 1, 1, 8, 2, 9, 2, 3, 6, 8, 7, 8, 9, 5, 4, 4, 2, 2, 3, 6, 8, 4, 5, 6, 5, 7, 1, 7, 7, 9, 6, 9, 2, 7, 9, 4, 8, 2, 7, 5, 0, 7, 3, 2, 2, 9, 8, 7, 2, 3, 5, 2, 9, 1, 1, 5, 8, 4, 4, 5, 4, 0, 6, 6, 9, 8, 1, 7, 0, 0, 4, 2, 7, 9, 6, 2, 9, 7, 9, 1, 0, 4, 3, 0, 7, 6, 7, 8, 1, 1, 5, 5, 3, 4, 3, 2, 2, 4, 1, 2, 7, 6, 6, 4, 5, 3, 8, 4, 2, 9, 7, 2, 6, 3, 4, 3, 9, 1, 1, 0, 4, 9, 5, 7, 3, 9, 1, 5, 5, 5, 9, 2, 3, 5, 9, 8, 0, 9, 5, 2, 9, 4, 7, 5, 7, 1, 0, 7, 5, 4, 7, 9, 3, 5, 9, 8, 6, 2, 3, 1, 7, 2, 6, 0, 9, 7, 1, 2, 6, 8, 4, 5, 2, 3, 2, 2, 7, 3, 9, 2, 9, 6, 3, 2, 3, 2, 2, 9, 7, 5, 3, 4, 9, 9, 7, 8, 6, 0, 0, 4, 0, 7, 2, 4, 0, 4, 6, 9, 9, 5, 1, 0, 4, 5, 4, 7, 9, 6, 9, 6, 1, 2, 3, 0, 3, 2, 1, 1, 4, 1, 5, 4, 0, 7, 8, 3, 4, 5, 2, 5, 2, 6, 6, 6, 1, 0, 6, 2, 9, 5, 1, 0, 9, 6, 3, 4, 8, 4, 5, 2, 7, 2, 8, 8, 2, 6, 1, 6, 3, 5, 3, 6, 1, 1, 4, 4, 2, 0, 7, 1, 7, 0, 3, 8, 6, 6, 2, 6, 2, 7, 0, 0, 2, 8, 0, 4, 6, 3, 2, 0, 8, 5, 8, 2, 7, 2, 6, 1, 5, 5, 4, 4, 5, 9, 3, 3, 8, 7, 9, 0, 7, 1, 2, 9, 1, 2, 3, 8, 7, 5, 0, 8, 0, 8, 0, 9, 2, 6, 0, 7, 2, 6, 4, 9, 6, 7, 3, 4, 6, 4, 6, 3, 6, 9, 2, 7, 3, 5, 7, 1, 2, 7, 9, 5, 7, 1, 4, 0, 7, 7, 9, 1, 3, 3, 1, 1, 2, 4, 5, 9, 0, 4, 4, 6, 3, 7, 6, 8, 4, 3, 1, 7, 1, 2, 2, 8, 3, 6, 0, 1, 5, 0, 2, 1, 5, 5, 2, 0, 9, 0, 1, 0, 4, 5, 8, 7, 2, 4, 7, 7, 0, 9, 6, 1, 1, 8, 1, 5, 6, 4, 8, 2, 4, 0, 3, 1, 6, 5, 1, 7, 7, 4, 9, 1, 0, 0, 0, 4, 6, 8, 3, 6, 7, 9, 9, 0, 9, 3, 5, 6, 7, 3, 8, 3, 6, 3, 4, 4, 0, 8, 1, 8, 2, 3, 1, 4, 3, 2, 9, 1, 0, 4, 8, 9, 4, 9, 9, 3, 2, 7, 1, 9, 0, 1, 4, 8, 4, 9, 2, 7, 9, 6, 5, 1, 1, 6, 8, 4, 0, 9, 7, 2, 3, 5, 1, 9, 7, 3, 5, 9, 0, 6, 1, 2, 8, 5, 1, 4, 6, 5, 1, 5, 3, 8, 9, 4, 7, 7, 0, 9, 6, 8, 2, 9, 3, 5, 9, 2, 8, 4, 2, 0, 2, 5, 3, 2, 2, 6, 7, 9, 3, 0, 6, 7, 1, 5, 1, 0, 2, 2, 9, 0, 2, 1, 2, 7, 7, 3, 0, 7, 9, 4, 8, 1, 9, 3, 4, 1, 1, 3, 2, 6, 3, 9, 3, 6, 6, 7, 6, 1, 1, 6, 1, 3, 9, 3, 2, 6, 8, 2, 6, 7, 6, 4, 1, 5, 9, 5, 9, 2, 0, 3, 8, 5, 2, 4, 2, 9, 3, 8, 0, 6, 6, 3, 1, 6, 9, 3, 2, 7, 6, 0, 7, 2, 6, 8, 0, 5, 5, 9, 9, 5, 4, 8, 0, 7, 4, 2, 8, 9, 3, 0, 5, 9, 3, 6, 5, 4, 9, 0, 2, 7, 2, 9, 0, 9, 9, 2, 6, 4, 3, 6, 9, 7, 6, 1, 6, 0, 6, 4, 9, 9, 6, 6, 0, 2, 2, 6, 6, 3, 8, 8, 1, 0, 9, 3, 9, 8, 5, 6, 4, 8, 4, 3, 5, 0, 7, 2, 2, 3, 8, 3, 2, 5, 9, 2, 7, 1, 0, 5, 6, 0, 4}; 

clock_t begin, end; 
double time_spent_omp; 
double time_spent; 

begin = omp_get_wtime(); 
/* here, do your time-consuming job */ 

#pragma omp parallel for private(temp) 
for(j=0;j<1000;j++){ 
    temp = arr[j]; 
    for(i=0;i<temp;temp--) 
    result[j]=result[j]*temp; 
} 


end = omp_get_wtime(); 
time_spent_omp = (double)(end - begin)/CLOCKS_PER_SEC; 

begin = omp_get_wtime(); 
/* here, do your time-consuming job */ 

for(j=0;j<1000;j++){ 
    temp = arr[j]; 
    for(i=0;i<temp;temp--) 
    result[j]=result[j]*temp; 
} 


end = omp_get_wtime(); 
time_spent = (double)(end - begin)/CLOCKS_PER_SEC; 
printf("\n\n Time to process: %f --- Time to process with OPENMP %f",time_spent, time_spent_omp); 

这应该会让你更好地了解它是如何工作的。