C. Fortran Intrinsics

A lot of Fortran intrinsic functions can be translated with ADAPTOR but still remain Fortran intrinsic functions in the generated code. If they are FORTRAN 77 intrinsic functions they should be compiled by every FORTRAN 77 compiler. But Fortran 90 intrinsic functions might require a Fortran 90 compiler on the target system.

If an intrinsic function is available, it means that ADAPTOR translates it to corresponding FORTRAN 77 code with corresponding support in the runtime system. These functions can be used without any restrictions.

If an intrinsic is noted as restricted, it means that the functionality is not full available, e.g. a parameter value must be known at compile time or an optional parameter must not be available.

If an intrinsic is noted as not available, it means, that it cannot be used with ADAPTOR at all.

C..1 Numeric, mathematical, character, kind, logical and bit procedures

C..1.1 Numeric functions

The numeric functions are INT, REAL, DBLE, CMPLX, AIMAG, CONJG, AINT, ANINT, NINT, ABS, MOD, SIGN, DIM, DPROD, MODULO, FLOOR, CEILING, MAX and MIN. They are all elemental functions.

All numeric intrinsic functions are supported. Attention has to be paid to those functions that are not supported by the F77 compiler of the target machine.

Some of the Fortran intrinsic routines can have an optional KIND argument. ADAPTOR treats them in the following way:

• REAL (A,KIND) becomes DBLE(A) if the KIND argument implies double precision.
• CMPLX (X[,Y],KIND) becomes DCMPLX(X[,Y]) if the KIND argument implies double precision.
• AINT (A,KIND) becomes AINT (DBLE(A)) if A is single precision and the KIND argument implies double precision.
• AINT (A,KIND) becomes AINT (REAL(A)) if A is double precision and the KIND argument implies single precision.
• ANINT (A,KIND) is treated like AINT (A,KIND).
• CEILING (A,KIND) becomes CEILING (A) as ADAPTOR supports only default integer size.
• FLOOR (A,KIND) becomes DBLE (FLOOR (A)) as ADAPTOR supports only default integer size.
• INT (A,KIND) becomes INT (A)) as ADAPTOR supports only default integer size.
• NINT (A,KIND) becomes NINT (A)) as ADAPTOR supports only default integer size.

C..1.2 Mathematical functions

The elemental functions SQRT, EXP, LOG, LOG10, SIN, COS, TAN, ASIN, ACOS, ATAN, ATAN2, SINH, COSH and TANH evaluate elementary mathematical functions.

They are all supported without any restrictions.

C..1.3 Character functions

The elemental functions ICHAR, CHAR, LGE, LGT, LLE, LLT, IACHAR, ACHAR are handled.

The elemental functions LEN_TRIM, SCAN, ADJUSTL, ADJUSTR, and VERIFY, as well as the transformational functions REPEAT and TRIM are not supported (do not belong to FORTRAN 77).

C..1.4 Character inquiry functions

The inquiry function LEN is supported.

C..1.5 Kind functions

Kind parameters provide are way to parameterize the selection of different possible machine representations for each of the intrinsic data types. This feature is now supported in ADAPTOR. This includes the inquiry function KIND as well as the transformational functions SELECTED_REAL_KIND and SELECTED_INT_KIND.

C..1.6 Logical function

The elemental function LOGICAL is not available.

C..1.7 Bit manipulation and inquiry procedures

Logical operations on bits are provided by the functions IOR, IAND, NOT and IEOR, shift operations are given by the functions ISHFT and ISHFTC. Bit subfields may be referenced by the function IBITS and by the subroutine MVBITS. Single-bit processing is provided by the functions BTEST, IBSET, and IBCLR.

All of these functions are checked by ADAPTOR, but their realization is passed to the compiler of the generated code.

C..2 Transfer function

The function TRANSFER is not available.

C..3 Numeric manipulation and inquiry functions

All the corresponding numeric inquiry functions RADIX, DIGITS, MINEXPONENT, MAXEXPONENT, PRECISION, RANGE, HUGE, TINY and EPSILON, and the floating point manipulation functions EXPONENT, SCALE, NEAREST, FRACTION, SET_EXPONENT, SPACING and RRSPACING cannot be used as long as the available Fortran compiler does not support these functions.

C..4 Intrinsic subroutines

C..4.1 Date and time subroutines

The timing routines of Fortran 90 are available to guarantee also a rather good portability of codes that are timed. These routines are DATE_AND_TIME, SYSTEM_CLOCK and CPU_TIME.

    DATA_AND_TIME ([DATE] [,TIME] [,ZONE], [,VALUES])
    SYSTEM_CLOCK ([COUNT] [,COUNT_RATE] [,COUNT_MAX])
    CPU_TIME (TIME)

C..4.2 Pseudorandom numbers

The intrinsic subroutines RANDOM_NUMBER and RANDOM_SEED are supported by ADAPTOR and realized by own library implementations.

         RANDOM_NUMBER (harvey)
         RANDOM_SEED ([SIZE][,PUT],[,GET])

C..5 Array intrinsic functions

C..5.1 Array inquiry functions

The functions LBOUND, UBOUND, SHAPE, SIZE are realized by ADAPTOR and own library functions.

The function ALLOCATED can be used, but the use of allocation is currently restricted (see section A.2.1).

C..5.2 Vector and matrix multiply functions

The functions DOT_PRODUCT and MATMUL can be used without any restrictions. For both routines parallel code will be generated.

C..5.3 Array reduction functions

The functions ALL, ANY, COUNT, MAXVAL, MINVAL, PRODUCT and SUM are supported by ADAPTOR. There are efficient realizations for these reductions in the context of distributed arrays.

C..5.4 Array construction functions

The functions MERGE and SPREAD are supported.

The functions PACK and UNPACK can be used but are not parallelized.

C..5.5 Array reshape function

The function RESHAPE can be used but is not parallelized so far.

C..5.6 Array manipulation functions

The array manipulation functions CSHIFT, EOSHIFT and TRANSPOSE are implemented, also rather efficiently for distributed arrays.

C..5.7 Array location functions

The array location functions MINLOC and MAXLOC are available, but cannot be used with the DIM argument.

C..5.8 Pointer association status functions

The functions ASSOCIATED and NULL are not available as ADAPTOR does not support pointer at all.

Table 4: Fortran Intrinsics.

Intrinsic	Available	Class	Type
ABS	F77	Elemental	Numeric
ACHAR	F90	Elemental	Transfer
ACOS	F77	Elemental	Numeric
ADJUSTL	F90	Elemental	Character
ADJUSTR	F90	Elemental	Character
AIMAG	F77	Elemental	Transfer
AINT	F77	Elemental	Transfer
ALL	Yes	Transformational	Array reduction
ALLOCATED	No	Inquiry	Array inquiry
ANINT	F77	Elemental	Transfer
ANY	Yes	Transformational	Array reduction
ASIN	F77	Elemental	Numeric
ASSOCIATED	No	Inquiry	Pointer association
ATAN	F77	Elemental	Numeric
ATAN2	F77	Elemental	Numeric
BIT_SIZE	F90	Inquiry	Bit inquiry
BTEST	F90	Elemental	Bit computation
CEILING	F90	Elemental	Numeric
CHAR	F90	Elemental	Transfer
CMPLX	F77	Elemental	Transfer
CONJG	F77	Elemental	Transfer
COS	F77	Elemental	Numeric
COSH	F77	Elemental	Numeric
COUNT	Yes	Transformational	Array reduction
CPU_TIME	Yes	Subroutine	Subroutine
CSHIFT	Yes	Transformational	Array manipulation
DATE_AND_TIME	Yes	Subroutine	Subroutine
DBLE	F77	Elemental	Transfer
DIGITS	F90	Inquiry	Numeric inquiry
DIM	F77	Elemental	Numeric
DOT_PRODUCT	Yes	Transformational	Vector multiplication
DPROD	F90	Elemental	Numeric
EOSHIFT	Yes	Transformational	Array manipulation
EPSILON	F90	Inquiry	Numeric inquiry
EXP	F77	Elemental	Numeric
EXPONENT	F90	Elemental	Floating point man.

Table 5: Fortran Intrinsics (continued).

Intrinsic	Available	Class	Type
FLOOR	F90	Elemental	Numeric
FRACTION	F90	Elemental	Floating point man.
HUGE	F90	Inquiry	Numeric inquiry
IACHAR	F90	Elemental	Transfer
IAND	Mil. Std.	Elemental	Bit computation
IBCLR	Mil. Std.	Elemental	Bit computation
IBITS	F90	Elemental	Transfer
IBSET	Mil. Std.	Elemental	Bit computation
ICHAR	F90	Elemental	Transfer
IEOR	Mil. Std.	Elemental	Bit computation
INDEX	F90	Elemental	Character
INT	F90	Elemental	Transfer
IOR	Mil. Std.	Elemental	Bit computation
ISHFT	Mil. Std.	Elemental	Bit computation
ISHFTC	Mil. Std.	Elemental	Bit computation
KIND	Yes	Inquiry	Kind
LBOUND	Yes	Inquiry	Array inquiry
LEN	F90	Inquiry	Char inquiry
LEN_TRIM	F90	Elemental	Character
LGE	F77	Elemental	Character
LGT	F77	Elemental	Character
LLE	F77	Elemental	Character
LLT	F77	Elemental	Character
LOG	F77	Elemental	Numeric
LOG10	F77	Elemental	Numeric
LOGICAL	F90	Elemental	Transfer
MATMUL	Yes	Transformational	Array multiplication
MAX	F77	Elemental	Numeric
MAXEXPONENT	F90	Inquiry	Numeric inquiry
MAXLOC	restricted	Transformational	Array location
MAXVAL	Yes	Transformational	Array reduction
MERGE	Yes	Elemental	Array construction
MIN	F77	Elemental	Numeric
MINEXPONENT	F90	Inquiry	Numeric inquiry
MINLOC	Restricted	Transformational	Array location
MINVAL	Yes	Transformational	Array reduction
MOD	F77	Elemental	Numeric
MODULO	F90	Elemental	Numeric
MVBITS	F90	Subroutine	Subroutine
NEAREST	F90	Elemental	Floating point man.
NINT	F77	Elemental	Transfer
NOT	Mil. Std.	Elemental	Bit Computation

Table 6: Fortran Intrinsics (continued).

Intrinsic	Available	Class	Type
NULL	No	Transformational	Pointer association
PACK	serial	Transformational	Construction
PRECISION	F90	Inquiry	Numeric inquiry
PRESENT	Yes	Inquiry	Argument
PRODUCT	Yes	Transformational	Array reduction
RADIX	F90	Inquiry	Numeric inquiry
RANDOM_NUMBER	Yes	Subroutine	Subroutine
RANDOM_SEED	Yes	Subroutine	Subroutine
RANGE	F90	Inquiry	Numeric inquiry
REAL	F77	Elemental	Transfer
REPEAT	F90	Transformational	Character
RESHAPE	serial	Transformational	Array reshape
RRSPACING	F90	Elemental	Floating point man.
SCALE	F90	Elemental	Floating point man.
SCAN	F90	Elemental	Character
SELECTED_INT_KIND	Yes	Transformational	Kind
SELECTED_REAL_KIND	Yes	Transformational	Kind
SET_EXPONENT	F90	Elemental	Floating point man.
SHAPE	Yes	Inquiry	Array inquiry
SIGN	F77	Elemental	Numeric
SIN	F77	Elemental	Numeric
SINH	F77	Elemental	Numeric
SIZE	Yes	Inquiry	Array inquiry
SPACING	F90	Elemental	Floating point man.
SPREAD	Yes	Transformational	Array construction
SQRT	F77	Elemental	Numeric
SUM	Yes	Transformational	Array reduction
SYSTEM_CLOCK	Yes	Subroutine	Subroutine
TAN	F77	Elemental	Numeric
TANH	F77	Elemental	Numeric
TINY	F90	Inquiry	Numeric inquiry
TRANSFER	F90	Transformational	Transfer
TRANSPOSE	Yes	Transformational	Array manipulation
TRIM	F90	Transformational	Character
UBOUND	Yes	Inquiry	Array inquiry
UNPACK	serial	Transformational	Array construction
VERIFY	F90	Elemental	Character

Table 7: Fortran Specific Mathematic Intrinsics.

Generic	Number	Specific	Type of	Type of
Name	of Args	Name	Argument	Result
ACOS	1	ACOS	real	real
	1	DACOS	double precision	double precision
ASIN	1	ASIN	real	real
	1	DASIN	double precision	double precision
ATAN	1	ATAN	real	real
	1	DATAN	double precision	double precision
ATAN2	1	ATAN2	real	real
	1	DATAN2	double precision	double precision
COS	1	COS	real	real
	1	DCOS	double precision	double precision
	1	CCOS	complex	complex
	1	CDCOS	double complex	double complex
COSH	1	COSH	real	real
	1	DCOSH	double precision	double precision
EXP	1	EXP	real	real
	1	DEXP	double precision	double precision
	1	CEXP	complex	complex
	1	CDEXP	double complex	double complex
LOG	1	ALOG	real	real
	1	DLOG	double precision	double precision
	1	CLOG	complex	complex
	1	CDLOG	double complex	double complex
LOG10	1	ALOG10	real	real
	1	DLOG10	double precision	double precision
SIN	1	SIN	real	real
	1	DSIN	double precision	double precision
	1	CSIN	complex	complex
	1	CDSIN	double complex	double complex
SINH	1	SINH	real	real
	1	DSINH	double precision	double precision
SQRT	1	SQRT	real	real
	1	DSQRT	double precision	double precision
	1	CSQRT	complex	complex
	1	CDSQRT	double complex	double complex
TAN	1	TAN	real	real
	1	DTAN	double precision	double precision
TANH	1	TANH	real	real
	1	DTANH	double precision	double precision

Table 8: Fortran Specific Numeric Intrinsics.

Generic	Number	Specific	Type of	Type of
Name	of Args	Name	Argument	Result
AIMAG	1	AIMAG	complex	real
	1	DIMAG	double complex	double precision
CONJG	1	CONJG	complex	complex
	1	DCONJG	double complex	double complex
ABS	1	IABS	integer	integer
	1	ABS	real	real
	1	DABS	double precision	complex
	1	CABS	complex	real
	1	CDABS	double complex	double precision
AINT	1	AINT	real	real
	1	DINT	double precision	double precision
ANINT	1	ANINT	real	real
	1	DNINT	double precision	double precision
NINT	1	NINT	real	integer
	1	IDNINT	double precision	integer
MIN	$>$ 1	MIN0	integer	integer
		MIN1	real	integer
		AMIN0	integer	real
		AMIN1	real	real
		DMIN1	double precision	double precision
MAX	$>$ 1	MAX0	integer	integer
		MAX1	real	integer
		AMAX0	integer	real
		AMAX1	real	real
		DMAX1	double precision	double precision
MOD	2	MOD	integer	integer
		AMOD	real	real
		DMOD	double precision	double precision
SIGN	2	ISIGN	integer	integer
		SIGN	real	real
		DSIGN	double precision	double precision
DIM	2	IDIM	integer	integer
		DIM	real	real
		DDIM	double precision	double precision

Table 9: Fortran Specific Numeric Intrinsics (contd).

Generic	Number	Specific	Type of	Type of
Name	of Args	Name	Argument	Result
INT	1		integer	integer
	1	IFIX	real	integer
	1	IDINT	double precision	integer
	1		complex	integer
	1		double complex	integer
REAL	1	FLOAT	integer	real
	1		real	real
	1	SNGL	double precision	real
	1		complex	real
	1		double complex	real
DBLE	1		integer	double precision
	1		real	double precision
	1		double precision	double precision
	1		complex	double precision
	1		double complex	double precision
CMPLX	1		integer	complex
	1		real	complex
	1		double precision	complex
	1		complex	complex
	1		double complex	complex
DCMPLX	2		integer	double complex
	2		real	double complex
	2		double precision	double complex
DCMPLX	1		integer	double complex
	1		real	double complex
	1		double precision	double complex
	1		complex	double complex
	1		double complex	double complex