測試雙層loop

test1：

compiler directive in outer layer

#include <stdio.h>  
#include <omp.h>  
  
int main(){  
  
int i,j,k, tid;  


#pragma omp parallel for private(j,k,tid)   

for(i = 0; i < 300; i++){  
  
  for(j = 0; j< 200; j++){  

     
     tid = omp_get_thread_num();  
      printf("i is %d, j = %d, tid = %d\n", i, j,  tid);  
        
   
  
}  

}  

  
return 0;  
}  

results:

real	0m3.153s
user	0m0.827s
sys	0m1.033s

compiler directives in inner loop:

#include <stdio.h>  
#include <omp.h>  
  
int main(){  
  
int i,j,k, tid;  

 

for(i = 0; i < 300; i++){  

#pragma omp parallel for private(k,tid)  
  
  for(j = 0; j< 200; j++){  

     
     tid = omp_get_thread_num();  
      printf("i is %d, j = %d, tid = %d\n", i, j,  tid);  
        
   
  
}  

}  

  
return 0;  
}  

results:

real	0m4.313s
user	0m10.508s
sys	0m0.496s

結論：

在外層平行計算要比內層快 如果循環次數很多的時候