Help to a FORTRAN newbie with KIND, please

[balthamossa2b]

Hi, I'm starting with FORTRAN (I have Microsoft Powerstation) and I was experimenting with KIND. Specifically, what I wanted was:

integer (kind=8):: list of variables



But the compiler doesn't let me use KIND with integers if kind>4, and I really need to use long integers. Can someone point out what I'm doing wrong, please?

Thanks in advance.

 

[Mark44]

I don't know this for a fact, but I believe the KIND keyword applies only to REAL variables, and the choices are 1, 2, and 3. These represent 4-byte, 8-byte, and 10-byte reals.

For integers, I believe you can specify the size of storage as INTEGER*1, INTEGER*2, INTEGER*4, and INTEGER*8. You'll need to look at your compiler documentation to get a definitive answer.

 

[minger]

In a large CFD code that I use, we have a module that looks like this:

MODULE Precision_Def
  IMPLICIT NONE

  PRIVATE

  PUBLIC :: i_def,      r_def,      & ! precision for calculation
            i_def_sp,   r_def_sp,   & ! single precision (4 byte)
            i_def_dp,   r_def_dp,   & ! double precision (8 byte)
            intDef,     realDef

  INTEGER, PARAMETER :: i_def    = SELECTED_INT_KIND(8),   &
                        r_def    = SELECTED_REAL_KIND(12), &
                        i_def_sp = SELECTED_INT_KIND(9),   &
                        r_def_sp = SELECTED_REAL_KIND(6),  &
                        i_def_dp = SELECTED_INT_KIND(16),  &
                        r_def_dp = SELECTED_REAL_KIND(12), &
                        intDef   = i_def,                  &
                        realDef  = r_def


END MODULE Precision_Def

Then, when we creat variables in subroutines, etc, they look like this:

INTEGER(KIND=i_def) :: ...
REAL(KIND=r_def) :: ...

Again as mentioned, it amy be compiler specific, although we've compiled and ran using both g95 and ifort.

 

[silverfrost]

KIND values are compiler specific (which seems pretty silly but that is what the standard says). minger's answer seems like the way to go -- although I quite like the real*8 and integer*8 type declarations.

 

[minger]

The nice thing about a module defining it, is e.g. if you decide you want to port over to a 64-bit machine and want to run a higher precision, you can just change the one module, rather than every single subroutine where variables are declared.