Parallel Sections - Output

runesha@origin1 [~/workshop/OpenMP-matrix-matrix] % f77 -mp -O3 psection_OpenMP.f
runesha@origin1 [~/workshop/OpenMP-matrix-matrix] % setenv OMP_NUM_THREADS 1
runesha@origin1 [~/workshop/OpenMP-matrix-matrix] % time a.out
 Done 
 d(1,1)=   1000.000000000000    
 d(NRA,NCB)=   1000.000000000000    
 e(1,1)=   2000.000000000000    
 e(NRA,NCC)=   2000.000000000000    
12.709u 0.332s 0:13.22 98.5% 0+0k 1+5io 1pf+0w
runesha@origin1 [~/workshop/OpenMP-matrix-matrix] % setenv OMP_NUM_THREADS 2
runesha@origin1 [~/workshop/OpenMP-matrix-matrix] % time a.out
 Done 
 d(1,1)=   1000.000000000000    
 d(NRA,NCB)=   1000.000000000000    
 e(1,1)=   2000.000000000000    
 e(NRA,NCC)=   2000.000000000000    
12.857u 0.410s 0:06.88 192.7% 0+0k 1+5io 1pf+0w
runesha@origin1 [~/workshop/OpenMP-matrix-matrix] % setenv OMP_NUM_THREADS 4
runesha@origin1 [~/workshop/OpenMP-matrix-matrix] % time a.out
 Done 
 d(1,1)=   1000.000000000000    
 d(NRA,NCB)=   1000.000000000000    
 e(1,1)=   2000.000000000000    
 e(NRA,NCC)=   2000.000000000000    
17.829u 4.686s 0:06.93 324.6% 0+0k 1+5io 1pf+0w

Timing for different number of processors

OMP_NUM_THREADS               time(sec)    CPU utilization
      1                        13.22             98.5%
      2                         6.88            192.7%
      4                         6.93            324.6%

The following timing were obtained running interactively. Better
results can be obtained by submitting jobs in the queue and increasing
the size of the problem.

Since the code had only two sections the speed up will not go beyond the
time achieved for 2 threads eventhough the CPU utilization  increased.      

1