Link Copied
Hi,
I am using my single PC which is having the following specification and I have attached it. I am using my PC which is having 16 CPUS and I want to print that " Hellow Word" on all those processor. I have attached the specifications. Also how in which portion of I can select the libraries such as Open MP or MPI, coz I did not fine them. Can you please attach some snap from visual studion. I would be very thankful. Thank you so much for your help.
Kind Regards
Zeeshan Khan
1) Launch MS Visual Studio
2) Select Fortran Main Code (then click Next)'
3) Optional: change name of program and/or location of program (default is OK for this simple test), click Create
You will see a single threaded Hello World.
4) Test it to confirm it builds and works
4.1) Place break point at end of program by clicking on the grey (colored) bar immediately to the left of the end program line. A red dot (or colored dot) will appear in the bar to indicate break.
4.2) Click the Tool Bar Start button. This should compile and run to the break point. The output will appear in the (black) pop-up console window. Note, had you not inserted a break point this console window would have popped-up, displayed, then closed when program terminated. IOW before you have a chance to see its output.
4.3) Click on the toolbar "Stop Debugging" button (usually red square button)
5) Edit the program such that it reads:
program Console18
use omp_lib
implicit none
! Variables
! Body of Console18
!$omp parallel
print *, 'Hello World from thread:', omp_get_thread_num()
!$omp end parallel
end program Console18Note, your program name may differ - leave your program name alone.
6) "But wait, there's more..."
7) Right click on the Project Name in the Solution Explorer pane (On my system this is the [Fo] Console18 line)
9) Expand: |> Fortran
10) Select Language
11) Select "Process OpenMP Directives" (currently Disabled), then click Pull-Down button, select Generate Parallel Code
12) Click Apply, Click OK, Click Start button on tool bar
13) make the following edits to make a simple example of doing some work:
program Console18
use omp_lib
implicit none
! Variables
integer, parameter :: nX=100, nY=100, nZ=100
integer :: iX, iY, iZ
real, allocatable, dimension(:,:,:) :: pos
! Body of Console18
allocate(pos(nX,nY,nZ))
call RANDOM_NUMBER(pos)
print *,"sum(pos):)", sum(pos)
!$omp parallel do
do iZ = 1, nZ
do iY = 1, nY
do iX = 1, nZ
pos(iX, iY, iZ) = sqrt(pos(iX, iY, iZ))
end do
end do
end do
!$omp end parallel do
print *,"sum(pos):)", sum(pos)
end program Console18Note, with Fortran, the last dimension has the largest stride. Do your parallel partitioning using the last dimention .AND. have the first dimension be the inner most DO loop. As to provide best cache line usage as well as potential for vectorization optimizations.
Jim Dempsey
