CO_REDUCE

Intel® Fortran Compiler Classic and Intel® Fortran Compiler Developer Guide and Reference

Download PDF

ID 767251

Date 11/07/2023

Version

Public

A newer version of this document is available. Customers should click here to go to the newest version.

Visible to Intel only — GUID: GUID-26A4A8DA-B5FE-4DA8-AA5B-6C641B54B3E6

View Details

CO_REDUCE

Collective Intrinsic Subroutine (Generic): Performs generalized reduction across images.

CALL CO_REDUCE (a, operation [, result_image, stat, errmsg])

a	(Input; output) Must be non-polymorphic, non-coindexed, with the same shape, type and type parameters in corresponding references. It may not be a coindexed object. If `a` is scalar, the computed value is the result of the reduction of applying operation to the values of `a` in all corresponding references. If `a` is an array, each element of the computed value is equal to the result of the reduction of the reduction operation of applying operation to corresponding elements of `a` in all corresponding references. If no error occurs, the computed value is assigned to `a` on all images of the current team if `result_image` is not present, or on the executing image if the executing image is the image identified by `result_image`. Otherwise, `a` becomes undefined.
operation	Is a pure function with exactly two arguments whose result and each argument are a scalar, nonallocatable, nonpointer, nonpolymorphic data object with the same type and kind type parameters as `a`. Neither argument can be optional. If one argument has the attribute TARGET, VOLATILE, or ASYNCHRONOUS, the other argument must have that attribute. `operation` must implement a mathematical associative operation and be the same in each corresponding reference. If operation is not commutative, the computed value may depend on the order of evaluation. The computed value of a reduction operation over a set of values is the result of an iterative process. Each iteration evaluates operation (x, y) for x and y in that set, the removal of x and y from that set, and the addition of the result of operation (x, y) to that set. The process ends when the set has a single value, which is the computed value.
result_image	(Input; optional) Must be a scalar integer. If present, it must be present with the same value in all corresponding references and be a valid image index in the current team. If `result_image` is not present, it cannot be present in any corresponding reference.
stat	(Output; optional) Must be a non-coindexed integer scalar with a decimal exponent range of at least four (KIND=2 or greater). The value assigned to `stat` is specified in Overview of Collective Subroutines. If `stat` is not present and an error condition occurs, error termination is initiated.
errmsg	(Input; output; optional) Must be a non-coindexed default character scalar variable. The semantics of `errmsg` is described inOverview of Collective Subroutines.

Example

The following subroutine demonstrates how CO_REDUCE can be used to create a collective version of the intrinsic function ANY.

SUBROUTINE CO_ANY (VALUE)
  LOGICAL,INTENT(INOUT) :: VALUE
  CALL CO_REDUCE (VALUE, COMBINER)
  CONTAINS
    PURE FUNCTION COMBINER (OPND1, OPND2)RESULT = LOGICAL_SUM
      LOGICAL                  :: LOGICAL_SUM
      LOGICAL,INTENT(IN)       :: OPND1, OPND2
      LOGICAL_SUM = OPND1 .OR. OPND2
    END FUNCTION COMBINER
END SUBROUTINE CO_ANY

If the number of images is two, and R is a four-element logical array with the value [.T., .T., .F., .F.] on image one and [.T., .F., .T., .F.] on image two, CALL CO_REDUCE (R, CO_ANY) causes the value of R to become defined with the value of [.T., .T., .T., .F.] on both images if no error occurs during the reference.

Parent topic: C to D

Select Your Language

Using Intel.com Search

Quick Links

Recent Searches

Advanced Search

Only search in

Intel® Fortran Compiler Classic and Intel® Fortran Compiler Developer Guide and Reference

CO_REDUCE

Example

See Also