OpenMP - sections directive; Linux slower than Windows -

i have simple code prepared testing. important piece of code:

#pragma omp parallel sections {  #pragma omp section   {  (int j=0;j<100000;j++)   (int i=0;i<1000;i++) a1[i]=1;  }  #pragma omp section   {  (int j=0;j<100000;j++)   (int i=0;i<1000;i++) a2[i]=1;  } } 

i compiled program mingw compiler , results expected. going use computer linux only, compiled code on linux (using same machine). used gcc 4.7.2 , intel 12.1.0 compilers. efficiency of program decreased. slower sequential program (omp_set_num_threads(1))

i have tried private arrays in threads, effect similar.

can suggest explanation?

i don't understand mean achieve code difference in efficiency due compiler making use of not knowing how handle code has sections-within-sections.

first off, try different compiler. experience gcc-4.8.0 works better openmp maybe try start off.

secondly, use optimisation flags! if measuring performance fair use either -o1 -o2 or -o3. latter give best performance takes short-cuts mathematical functions make floating point operations slightly less accurate.

g++ -fopenmp name.cpp -o3 

you can read more on compiler flags on this page if interests you.

as end note, don't know how experienced openmp, when dealing loops in openmp use following:

#pragma omp parallel for(int i=0; i<n; ++i)    dosomething(); 

additionally, if using nested loops, can use collapse directive inform compiler turn nested loops single 1 (which can lead better performance)

#pragma omp parallel collapse(2) private(i, j) for(int i=0; i<n; ++i)    for(int j=0; j<n; ++j)        dosomething(); 

there things should aware of when using collapse can read here. prefer manually converting them single loop experience proves more efficient.