Intrinsic Procedures

Developer Guide and Reference

Download PDF

ID 767251

Date 10/31/2024

Version

Public

Visible to Intel only — GUID: GUID-23BF81DC-14D6-4E5B-80E7-2A01199B3468

View Details

Intrinsic Procedures

Intrinsic procedures are functions and subroutines that are included in the Fortran library. The following are classes of intrinsic procedures:

Elemental procedures

These procedures have scalar dummy arguments that can be called with scalar or array actual arguments. There are many elemental intrinsic functions and one elemental intrinsic subroutine (MVBITS). All elemental intrinsic procedures are pure.

If the arguments are all scalar, the result is scalar. If an actual argument is array-valued, the intrinsic procedure is applied to each element of the actual argument, resulting in an array that has the same shape as the actual argument.

If there is more than one array-valued argument, they must all have the same shape.

Many algorithms involving arrays can now be written conveniently as a series of computations with whole arrays. For example, consider the following:
```
  a = b + c
  ...               ! a, b, c, and s are all arrays of similar shape
  s = sum(a)
```
The above statements can replace entire DO loops.

Consider the following:
```
  real, dimension (5,5) :: x,y
  . . .      ! Assign values to x
  y = sin(x) ! Pass the entire array as an argument
```
In this example, since the SIN(X) function is an elemental procedure, it operates element-by-element on the array x when you pass it the name of the whole array.
Inquiry functions

These functions have results that depend on the properties of their principal argument, not the value of the argument (the argument value can be undefined).
Transformational functions

These functions have one or more array-valued dummy or actual arguments, an array result, or both. The intrinsic function is not applied elementally to an array-valued actual argument; instead it changes (transforms) the argument array into another array, or in some cases, a scalar.
Nonelemental procedures

These procedures must be called with only scalar arguments; they return scalar results. All intrinsic subroutines (except MVBITS) are nonelemental.
Atomic subroutines

These subroutines perform an action on a variable (its atom argument) atomically. When an atomic subroutine is executed, it is as if the subroutine were executed instantaneously without overlapping other atomic actions that might occur asynchronously. For information on the semantics of atomic subroutines, see Overview of Atomic Subroutines.
Collective subroutines

These subroutines perform a cooperative calculation on a team of images and require no synchronization. For information on the semantics of collective subroutines, see Overview of Collective Subroutines.

The intrinsic subroutine MVBITS, and the subroutine MOVE_ALLOC with a noncoarray argument FROM, are pure. All other intrinsic subroutines are impure.

Intrinsic procedures are invoked the same way as other procedures, and follow the same rules of argument association.

The intrinsic procedures have generic (or common) names, and many of the intrinsic functions have specific names. Some intrinsic function names are both generic and specific. The interfaces of intrinsic procedures are explicit.

In general, generic functions accept arguments of more than one data type; the data type of the result is the same as that of the arguments in the function reference. For elemental functions with more than one argument, all arguments must be of the same type (except for the function MERGE).

When an intrinsic function is passed as an actual argument to a procedure, its specific name must be used, and when called, its arguments must be scalar. Some specific intrinsic functions are not allowed as actual arguments in all circumstances. The following table lists specific functions that cannot be passed as actual arguments or as targets in procedure pointer assignment statements.

Starting with Fortran 2018, specific names of intrinsic functions that also have generic names are obsolescent.

Specific Intrinsic Functions Not Allowed as Actual Arguments
AIMAX0	FLOATJ	JFIX	MAX0
AIMIN0	FLOATK	JIDINT	MAX1
AJMAX0	FP_CLASS	JIFIX	MIN0
AJMIN0	HFIX	JINT	MIN1
AKMAX0	IADDR	JIQINT	NARGS
AKMIN0	IARGC	JMAX0	QCMPLX
AMAX0	ICHAR	JMAX1	QEXT
AMAX1	IDINT	JMIN0	QEXTD
AMIN0	IFIX	JMIN1	QMAX1
AMIN1	IIDINT	JNUM	QMIN1
CHAR	IIFIX	JZEXT	QNUM
CMPLX	IINT	KIDINT	QREAL
DBLE	IIQINT	KIFIX	RAN
DBLEQ	IJINT	KINT	RANF
DCMPLX	IMAX0	KIQINT	REAL
DFLOTI	IMAX1	KMAX0	RNUM
DFLOTJ	IMIN0	KMAX1	SECNDS
DFLOTK	IMIN1	KMIN0	SHIFTL
DMAX1	INT	KMIN1	SHIFTR
DMIN1	INT1	KNUM	SNGL
DNUM	INT2	KZEXT	SNGLQ
DREAL	INT4	LGE	ZEXT
DSHIFTL	INT8	LGT
DSHIFTR	INUM	LLE
FLOAT	IQINT	LLT
FLOATI	IZEXT	LOC

Note that none of the intrinsic subroutines can be passed as actual arguments or as targets in procedure pointer assignment statements.

The A to Z Reference contains the descriptions of all intrinsics listed in alphabetical order. Each reference entry indicates whether the procedure is inquiry, elemental, transformational, or nonelemental, and whether it is a function or a subroutine.

Parent topic: Language Reference

Select Your Language

Using Intel.com Search

Quick Links

Recent Searches

Advanced Search

Only search in

Developer Guide and Reference

Intrinsic Procedures

See Also