F90 compiler

Double Precision

In F90, double precision can be handled in various ways.

Using Fortran KIND command

One of them is using the KIND command in F90.

Example Program:

A simple example is given below:

PROGRAM TEST_DOUBLE
IMPLICIT NONE
!
INTEGER, PARAMETER :: num = KIND(1.E0)
INTEGER, PARAMETER :: dbl = KIND(1.D0)
!
REAL(num), PARAMETER :: pi1=0.12345678901234567890_num
REAL(dbl), PARAMETER :: pi2=0.12345678901234567890_dbl
REAL(num), PARAMETER :: pi3=0.12345678901234567890
REAL(dbl), PARAMETER :: pi4=0.12345678901234567890
REAL*8, PARAMETER :: pi5=0.12345678901234567890
REAL*4, PARAMETER :: pi6=0.12345678901234567890
!
write(*,*)'num = KIND(1.E0)',num
write(*,*)'dbl = KIND(1.D0)',dbl
write(*,*)'REAL(num) :: pi=0.12345678901234567890_num',pi1
write(*,*)'REAL(dbl) :: pi=0.12345678901234567890_dbl',pi2
write(*,*)'REAL(num) :: pi=0.12345678901234567890',pi3
write(*,*)'REAL(dbl) :: pi=0.12345678901234567890',pi4
write(*,*)'REAL :: pi5=0.12345678901234567890',pi5
write(*,*)'REAL :: pi6=0.12345678901234567890',pi6
!
STOP
END PROGRAM TEST_DOUBLE

You can compile the program wthout any compiler options.

pgf90 program.f90

Then the output is:

Sample Output:

A sample output is given below:

  num = KIND(1.E0)
  dbl = KIND(1.D0)
  REAL(num) :: pi=0.12345678901234567890_num  0.1234568
  REAL(dbl) :: pi=0.12345678901234567890_dbl  0.123456789012346
  REAL(num) :: pi=0.12345678901234567890  0.1234568
  REAL(dbl) :: pi=0.12345678901234567890  0.123456791043282
  REAL*8 :: pi5=0.12345678901234567890  0.123456791043282
  REAL :: pi6=0.12345678901234567890  0.1234568

Using compile option

Perhaps easier way is using the -r8 compiler option. pgf90 -r8 program.f90