numerical constant definition does not work ! 1

[tservas]

hello everybody,

during an intense programming session I stumbled accross a very strange problem. Have a look at the following simple program:

program test

implicit none

real(kind(1.0d0)), parameter :: real_const = 3.14159265358979
real(kind(1.0d0)) :: real_variable
real(kind(1.0d0)) :: real_variable_readin
character(128) :: real_char

real_char = '3.14159265358979'
real_variable = 3.14159265358979
read(real_char,*), real_variable_readin

write(*,*), real_variable
write(*,*), 2*acos(0.0d0)
write(*,*), real_variable_readin

end program test

The Ouptput should be three times the same value:

3.14159265358979
3.14159265358979
3.14159265358979

BUT IT IS NOT --> on my computer the output looks like this:

3.14159274101257
3.14159265358979
3.14159265358979

I'm using g95 (newest version 2005-07-14) on a dual opteron workstation running under linux ...

Can anybody tell me what's going on ...

thx

tservas

 

[xwb]

The * in the read statement assumes G format which probably defaults to real instead of double precision. You probably need to specify something like F17.15

 

[tservas]

... I'm happy with the behaviour of the read statement, the problem is the output of the real_variable = 3.14159265358979 which should give

3.14159265358979

but instead it gives

3.14159274101257


... ???

 

[ArkM]

Consult your compiler docs: what's a type of float literals. It seems you have real*4 value. Add D0 in

real_variable = 3.14159265358979D0

and try again.

It's in C/C++ languages float literals have a type of double. I don't remember what's its type in Fortran, but it does not depend on the left side of assignment, of course.

 

[tservas]

... thank you, this did the trick ... ))))))))))))