doc/html/rnddbl_8f90_source.html

 c NLSL Version 1.9.0 beta 2/13/15

 c*********************************************************************

 c                   =========================

 c                         module RNDDBL

 c                   =========================

 c

 c       Declarations of the machine epsilon parameter, and the

 c       highest and lowest representable double precision values

 c

 c       Notes:

 c               1) The machine epsilon, eps, is defined to be the

 c                  largest number such that the result of the

 c                  expression

 c

 c                       delta=(1.0D0+eps)-1.0D0

 c

 c                  returns the value delta=0.0D0.  In other words,

 c                  1.0D0+eps is smallest double precision floating

 c                  point number greater than 1.0D0.

 c

 c               2) A FORTRAN 77 program to determine the machine

 c                  epsilon is included below for completeness.

 c                  This program should be compiled with all

 c                  compiler optimization disabled.  The parameter

 c                  rndoff declared in this file should assigned to

 c                  be a small multiple of the larger of the two

 c                  numbers printed by this program.  This program

 c                  is an adaptation of the function subroutine epslon

 c                  supplied with EISPACK.

 c

 c               3) On most DEC machines, the two methods of

 c                  calculating the machine epsilon give the nearly

 c                  same the result, eps=2.77555756D-17.  An

 c                  appropriate value of rndoff is then 1.0D-16.  An

 c                  example of a machine which gives grossly

 c                  different results is the Prime 9955.

 c

 c               4) ANSI-compliant C compilers provide these values

 c                  in the header file "float.h".

 c

 c       written by DJS 11-SEP-87

 c       updated for IEEE754 by WH 13-FEB-2002

 c

 c*********************************************************************

 c

 c       program meps

 c

 c       double precision a,b,c,eps

 c

 c#######################################################################

 c

 c       a=1.0D0

 c10     b=a+1.0D0

 c       eps=c

 c       c=b-1.0D0

 c       if (c.eq.0.0D0) go to 20

 c       a=a/2.0D0

 c       go to 10

 c

 c20     write (*,*) 'first epsilon = ',eps

 c

 c       a=4.0D0/3.0D0

 c30     b=a-1.0D0

 c       c=b+b+b

 c       eps=dabs(c-1.0D0)

 c       if (d.eq.0.0D0) go to 30

 c

 c       write (*,*) 'second epsilon = ',eps

 c

 c       stop

 c       end

 c

 c***********************************************************************

 c

 c The values provided below are for IEEE754 double precision.

 c

       module rnddbl

       implicit none

 c

       double precision, parameter ::

      #           dbl_epsilon=2.2204460492503131d-16,

      #           dbl_min=2.2250738585072014d-308,

      #           dbl_max=1.7976931348623157d+308


 c The following is an alternative name.

       double precision, parameter :: rndoff=dbl_epsilon

 c

       end module rnddbl

rnddbl::dbl_max
double precision, parameter dbl_max
Definition: rnddbl.f90:80

rnddbl::dbl_min
double precision, parameter dbl_min
Definition: rnddbl.f90:80

rnddbl::dbl_epsilon
double precision, parameter dbl_epsilon
Definition: rnddbl.f90:80

rnddbl
Definition: rnddbl.f90:77

rnddbl::rndoff
double precision, parameter rndoff
Definition: rnddbl.f90:86