Fortran: What exactly is .d0?

Question

I understand that .d0 denotes double precision however I wonder how it goes together with user defined precision. I use

module precision
implicit none
    integer, parameter :: DP = selected_real_kind(r=250,p=13)
end module precision

and declare floats as

real(KIND=DP) :: var1

Whenever I need to assign a value to such a float somewhere in the programme I do

var1 = 1_DP

Now the problem I have is that it does not work if I read the value from a namelist. The compiler insists on

&namelistName
    var1 = 1.d0
    !var1 = 1_DP   ! this does not work
&end

Two questions follow:

1. What happens if I assign value of 1.d0 to a variable declared as real(KIND=DP)?
2. Is there a way how to enforce 1_DP in namelists? If it matters of course.

Answer 1

The whole concept of kinds makes sense only in the source code, there is no way making it work for input data, because at runtime the system knows nothing about them.

In the source code

  1.d0

is a double precision real,

  1._dp 

is a real with kind dp,

  1_dp

is an integer (no decimal there) with kind dp.

In the input data you don't need to specify the precision at all, you just need a number. The runtime system then converts the string with a number to the right type of variable you are trying to read.

You can use 1.e0, 1.d0, as an extension maybe even 1.q0, just to be flexible, but it doesn't change the meaning. In the input file it is just a string representing a number.

What happens, if you in the source code assign

real(dp) :: x = 1.d0 ?

It is just simply assigned if the dp kind is the same kind as the double precision. Otherwise it is converted. If the real constant value is a small integer, the conversion is simple and no precision is lost as they are represented exactly.