6

Using scientific notation for floating point literals is easy enough in Fortran:

1.5d-10

would mean a double precision (whatever that means under current Fortran compiler settings) floating point value that approximates 1.5*10^-15.

However, the fusion of the exponent notation and the floating point kind specifier is a bit of an issue. How would one declare this floating point literal when one wants it to have a type of C_DOUBLE?

I know that this is a bit of a nitpicking issue, but there can be circumstances when double precision will not be the same as C_DOUBLE.

uLoop
  • 225
  • 1
  • 8

1 Answers1

6

A real literal may be specified by any of the following forms:

  • 1.2
  • 1.2e0
  • 1.2d0
  • 1.2_kind
  • 1.2e0_kind

This final one is an example of using a kind specifier and an exponent. So, specific to the question: 1.5e-15_C_DOUBLE.

There certainly can be cases where 1.5d-15 is not the same as 1.5e-15_C_DOUBLE. The kind of a double precision and real(C_DOUBLE) are choices by the Fortran and companion C compilers respectively.

Compilers which allow single and double precision literal constants to be promoted to higher kinds by a compiler flag won't touch real(C_DOUBLE).

For real values which have no decimal part, there are additional available forms:

  • 1.
  • 1e0
  • 1.e0
  • 1d0
  • 1.d0
  • 1.e0_kind
  • 1e0_kind

That is, the decimal part is optional, and if an exponent is given the decimal separator is also optional. Note however that 1 is an integer literal, while 1e0 and similar are real.

francescalus
  • 30,576
  • 16
  • 61
  • 96