OpenMP and Fortran 95 array operations

beliavsky_at_aol.com
Date: 12/29/04 

Date: 29 Dec 2004 08:01:02 -0800

I am reading the book "Parallel Programming in OpenMP". Looking at the
code examples, I wonder how OpenMP directives can be used in a Fortran
95 code that uses the parallel constructs of that language, such as
array operations and FORALL.

For example, on page 47 there is code

subroutine sums(a,m,n)
integer m,n,a(0:m,n),i,j
!$omp parallel do
do j=1,n
a(0,j) = 0
do i=1,m
a(0,j) = a(0,j) + a(i,j)
end do
end do
end subroutine sums

In F95 I would write the body of the subroutine as

a(0,:) = sum(a(1:,:),dim=1)

or

forall (j=1:n) a(0,j) = sum(a(1:,j))

Most of the F77-style examples in the book could be written in a
parallel manner in F95. Can OpenMP directives be used to speed up an
F95 style code on a multi-CPU machine? Or does code need to be
rewritten in F77 style with loops, with OpenMP directives then added? I
would not like that approach, since the code becomes larger and a bit
less readable.