Next: DConjg Intrinsic, Previous: DbleQ Intrinsic, Up: Other Intrinsics

DCmplx(X,Y)

DCmplx: `COMPLEX(KIND=2)`

function.

`X`: `INTEGER`

, `REAL`

, or `COMPLEX`

; scalar; INTENT(IN).

`Y`: `INTEGER`

or `REAL`

; OPTIONAL (must be omitted if `X` is `COMPLEX`

); scalar; INTENT(IN).

Intrinsic groups: `f2c`

, `vxt`

.

Description:

If `X` is not type `COMPLEX`

,
constructs a value of type `COMPLEX(KIND=2)`

from the
real and imaginary values specified by `X` and
`Y`, respectively.
If `Y` is omitted, ``0D0`' is assumed.

If `X` is type `COMPLEX`

,
converts it to type `COMPLEX(KIND=2)`

.

Although this intrinsic is not standard Fortran,
it is a popular extension offered by many compilers
that support `DOUBLE COMPLEX`

, since it offers
the easiest way to convert to `DOUBLE COMPLEX`

without using Fortran 90 features (such as the ``KIND=`'
argument to the `CMPLX()`

intrinsic).

(``CMPLX(0D0, 0D0)`' returns a single-precision
`COMPLEX`

result, as required by standard FORTRAN 77.
That's why so many compilers provide `DCMPLX()`

, since
``DCMPLX(0D0, 0D0)`' returns a `DOUBLE COMPLEX`

result.
Still, `DCMPLX()`

converts even `REAL*16`

arguments
to their `REAL*8`

equivalents in most dialects of
Fortran, so neither it nor `CMPLX()`

allow easy
construction of arbitrary-precision values without
potentially forcing a conversion involving extending or
reducing precision.
GNU Fortran provides such an intrinsic, called `COMPLEX()`

.)

See Complex Intrinsic, for information on easily constructing
a `COMPLEX`

value of arbitrary precision from `REAL`

arguments.