PROGRAM looptest
!
! Author:  Br. David Carlson
!
! Date:  Mar 2, 2004
!
! In computing partial sums for a series and in other applications
! we sometimes want to have a loop count from 1 to a very large number.
! If we use integers, we will eventually hit integer overflow.  If
! we use reals, overflow is possible, but a practical problem arises
! earlier:  Since we only get about 7 decimal digits of precision with
! ordinary reals, we quickly reach a point where r + 1 equals r, for
! some large real r.  Thus we can't use reals to control a counting
! loop beyond this point.
!
! The purpose of this program is to find the point where an integer
! loop control variable fails us as well as the point where a real
! as a loop control variable fails us.  The program simply prints its
! answers to these two items.
!
! On our machine we get these results:
! Largest value for a real loop control variable:         16777216.
! Largest value for an integer loop control variable:   2147483647
! Thus an integer loop control variable can handle larger values!

IMPLICIT NONE

REAL::n, m
INTEGER::j, k

! Test with reals:

n = 1.0

DO
   m = n + 1.0
   IF (m == n) THEN    ! Adding 1 has no effect at this point.
      EXIT 
   END IF
   n = m
END DO

WRITE (*, 100) n
100 FORMAT(1X, 'Largest value for a real loop control variable:', T60, F18.0)

! Test with integers:

j = 1

DO
   k = j + 1
   IF (k < 0) THEN    ! We have passed the largest positive integer
      EXIT            ! and wrapped around to a negative integer.
   END IF
   j = k
END DO

WRITE (*, 200) j
200 FORMAT(1X, 'Largest value for an integer loop control variable:', T60, I18)

END PROGRAM