Tips on modernizing old fortran code?

Help
5

This kind of task can also be done by ChatGPT4.

To demonstrate it, I gave this code as input:

         SUBROUTINE RCKELI(N,A,Y,X)
         PARAMETER (IA=50)
         REAL A(IA,IA), Y(IA), X(IA)
         INTEGER N

         X(N)=Y(N)/A(N,N)
         DO 10 K=N-1,1,-1
            S=0.
            DO 20 J=K+1,N
      20       S=S+A(K,J)*X(J)
      10    X(K)=(Y(K)-S)/A(K,K)
         RETURN
         END

After a few (not many) tries, ChatGPT made this:

module back_substitution_module
  implicit none
  integer, parameter :: dp = kind(1.d0)
contains
  subroutine back_substitution(a, y, x)
    integer :: n
    real(dp), intent(in) :: a(:,:), y(:)
    real(dp), intent(out) :: x(size(y))
    real(dp) :: s
    integer :: k

    n = size(y)
    x(n) = y(n) / a(n, n)
  
    do k = n-1, 1, -1
       s = dot_product(a(k, k+1:n), x(k+1:n))
       x(k) = (y(k) - s) / a(k, k)
    end do

  end subroutine back_substitution

end module back_substitution_module

Unfortunately, ChatGPT isn’t as good in Fortran programming compared to Python programming, but giving some hints it’s possible to get good results.

I used ChatGPT to modernize an old FORTRAN routine (horizontal diffusion) at work and AFAIK ChatGPT makes fewer mistakes than a human translating the code manually would make.

Whole Chat