• Directives-based for parallel computation on shared memory UNIX/WIndows/NT computers
• Available for fortran 77/90 and C/C++
• Platform independent
• Can be implemented incrementally one subroutine at a time
• Isnot instrusive to original serial code (instructions are comments)

A C example using the Monte Carlo PI calculation example

#pragma omp parallel default(none) \
private(i,x,y,z) shared(count)
#pragma omp for
   for ( i=0; i<niter; i++) {
      x = (double)rand()/RAND_MAX;
      y = (double)rand()/RAND_MAX;
      z = x*x+y*y;
      if (z<=1) count++; 
   }

A Fortran example using the Monte Carlo PI Calculation

c$omp parallel default(none)
c$omp&private(i,j,x,y,z,count,niter) shared(pi)
c$omp do
       do j= 1,1000
          niter = niter+100
          count =0
          do i=1,niter
             x=rand()
             y=rand()
             z= x*x +y*y
             if (z .le. 1) count =count+1
          end do
          pi(j)= count/niter*4.
          write(*,10) niter,pi(j)
10        format('Number of trials is: 'i5,'  estimate of pi is:',f8.5)
       end do
c$omp end do
c$omp end parallel
       end

OpenMP Resources:

• OpenMP Organization Web page
• OpenMP Tutorial at Boston University