When to use the 'allocatable' keyword when declaring dynamic arrays in Fortran

It can confuse when to use allocatable keyword when declaring dynamic arrays. We have two scenarios to declare the array here First scenario (pre allocation) real :: x(:,:)

Purpose:

Use this scenario inside the subroutine, or function when x was already allocated by the main program

program main
real, allocatable :: x(:,:)

  allocate(x(5,5))

  call process(x)

end program

subroutine process(x)

  real, intent(inout) :: x(:,:)  ! assumed-shape

  ! Can use x here (already allocated from main)

end subroutine

Second scenario (deferred allocation)

real, allocatable :: x(:,:)

Purpose:

Use this scenario inside the main or caller subroutine (function). This deferred array will be passed through and allocated inside the calling subroutine or function.

program main
 implicit none

  real, allocatable :: A(:,:)

  integer :: m, n

   m = 4
   n = 3

  call create_matrix(A, m, n)

  print *, 'Matrix A:'

  print *, A

end program

subroutine create_matrix(x, m, n)
 implicit none

  integer, intent(in) :: m, n

  real, allocatable, intent(out) :: x(:,:)  ! Note: allocatable with intent(out)

  integer :: i, j

  allocate(x(m, n))


  do i = 1, m

     do j = 1, n

       x(i,j) = real(i * 10 + j)   ! Just fill with dummy values like 11, 12, ...

     end do

  end do

end subroutine

It's important to master this concept because this will cause compile errors and be hard to debug.