MAN page from Old RedHat 5.X mpich-1.1.1-8.i386.rpm
Constants
Section: MPI (3)
Updated: 12/10/1997
Index NAME
Constants - Meaning of MPI
s defined constants DATA TYPES
Note that the Fortran types should only be used in Fortran programs,and the C types should only be used in C programs. For example,it is in error to use MPI_INT
for a Fortran INTEGER.Datatypes are of type MPI_Datatype
in C and of tyep INTEGER
in Fortran.
C DATATYPES
- MPI_CHAR
- - char
- MPI_BYTE
- - See standard; like unsigned char
- MPI_SHORT
- - short
- MPI_INT
- - int
- MPI_LONG
- - long
- MPI_FLOAT
- - float
- MPI_DOUBLE
- - double
- MPI_UNSIGNED_CHAR
- - unsigned char
- MPI_UNSIGNED_SHORT
- - unsigned short
- MPI_UNSIGNED
- - unsigned int
- MPI_UNSIGNED_LONG
- - unsigned long
- MPI_LONG_DOUBLE
- - long double (some systems may not implement)
The following are datatypes for the MPI functions MPI_MAXLOCandMPI_MINLOC.
- MPI_FLOAT_INT
- - struct { float, int }
- MPI_LONG_INT
- - struct { long, int }
- MPI_DOUBLE_INT
- - struct { double, int }
- MPI_SHORT_INT
- - struct { short, int }
- MPI_2INT
- - struct { int, int }
- MPI_LONG_DOUBLE_INT
- - struct { long double, int }; this is an OPTIONAL type, and may be set to NULL
- MPI_LONG_LONG_INT
- - struct { long long, int }; this is an OPTIONAL type,and may be set to NULL.
Special datatypes for C and Fortran
- MPI_PACKED
- - For MPI_Packand MPI_Unpack
- MPI_UB
- - For MPI_Type_struct; an upper-bound indicator
- MPI_LB
- - For MPI_Type_struct; a lower-bound indicator
FORTRAN DATATYPES
- MPI_REAL
- - REAL
- MPI_INTEGER
- - INTEGER
- MPI_LOGICAL
- - LOGICAL
- MPI_DOUBLE_PRECISION
- - DOUBLE PRECISION
- MPI_COMPLEX
- - COMPLEX
- MPI_DOUBLE_COMPLEX
- - complex*16(or complex*32) where supported.
The following datatypes are optional
- MPI_INTEGER1
- - integer*1if supported
- MPI_INTEGER2
- - integer*2if supported
- MPI_INTEGER4
- - integer*4if supported
- MPI_REAL4
- - real*4if supported
- MPI_REAL8
- - real*8if supported
The following are datatypes for the MPI functions MPI_MAXLOCandMPI_MINLOCtwo elements of the same Fortran type.
- MPI_2INTEGER
- - INTEGER,INTEGER
- MPI_2REAL
- - REAL, REAL
- MPI_2DOUBLE_PRECISION
- - DOUBLE PRECISION, DOUBLE PRECISION
- MPI_2COMPLEX
- - COMPLEX, COMPLEX
- MPI_2DOUBLE_COMPLEX
- - complex*16, complex*16
COMMUNICATORS
Communicators are of type
MPI_Commin C and
INTEGERin Fortran
- MPI_COMM_WORLD
- - Contains all of the processes
- MPI_COMM_SELF
- - Contains only the calling process
GROUPS
Groups are of type
MPI_Groupin C and
INTEGERin Fortran
- MPI_GROUP_EMPTY
- - A group containing no members.
RESULTS OF THE COMPARE OPERATIONS
- MPI_IDENT
- - Identical
- MPI_CONGRUENT
- - (only for MPI_COMM_COMPARE) The groups are identical
- MPI_SIMILAR
- - Same members, but in a different order
- MPI_UNEQUAL
- - Different
COLLECTIVE OPERATIONS
The collective combination operations (
MPI_REDUCE,
MPI_ALLREDUCE,
MPI_REDUCE_SCATTER, and
MPI_SCAN) take a combination operation.This operation is of type
MPI_Opin C and of type
INTEGERin Fortran.The predefined operations are
- MPI_MAX
- - return the maximum
- MPI_MIN
- - return the minumum
- MPI_SUM
- - return the sum
- MPI_PROD
- - return the product
- MPI_LAND
- - return the logical and
- MPI_BAND
- - return the bitwise and
- MPI_LOR
- - return the logical or
- MPI_BOR
- - return the bitwise of
- MPI_LXOR
- - return the logical exclusive or
- MPI_BXOR
- - return the bitwise exclusive or
- MPI_MINLOC
- - return the minimum and the location (actually, the value ofthe second element of the structure where the minimum ofthe first is found)
- MPI_MAXLOC
- - return the maximum and the location
NOTES ON COLLECTIVE OPERATIONS
The reduction functions (MPI_Op) do not return an error value. As a result,if the functions detect an error, all they can do is either call MPI_Abortor silently skip the problem. Thus, if you change the error handler fromMPI_ERRORS_ARE_FATALto something else, for example, MPI_ERRORS_RETURN,then no error may be indicated.
The reason for this is the performance problems in ensuring thatall collective routines return the same error value.
Note that not all datatypes are valid for these functions. For example,MPI_COMPLEXis not valid for MPI_MAXand MPI_MIN1.1 standard did not include the C types MPI_CHARand MPI_UNSIGNED_CHARamong the lists of arithmetic types for operations like MPI_SUMsince the C type charis an integer type (like short), it should have beenincluded. The MPI Forum will probably include charand unsigned charas a clarification to MPI 1.1; until then, users are advised that MPIimplementations may not accept MPI_CHARand MPI_UNSIGNED_CHARas validdatatypes for MPI_SUM, MPI_PROD, etc. MPICH does allow these datatypes.
PERMANENT KEY VALUES
These are the same in C and Fortran
- MPI_TAG_UB
- - Largest tag value
- MPI_HOST
- - Rank of process that is host, if any
- MPI_IO
- - Rank of process that can do I/O
- MPI_WTIME_IS_GLOBAL
- - Has value 1 if MPI_WTIMEis globally synchronized.
NULL OBJECTS
- MPI_COMM_NULL
- - Null communicator
- MPI_OP_NULL
- - Null operation
- MPI_GROUP_NULL
- - Null group
- MPI_DATATYPE_NULL
- - Null datatype
- MPI_REQUEST_NULL
- - Null request
- MPI_ERRHANDLER_NULL
- - Null error handler
PREDEFINED CONSTANTS
- MPI_MAX_PROCESSOR_NAME
- - Maximum length of name returned byMPI_GET_PROCESSOR_NAME
- MPI_MAX_ERROR_STRING
- - Maximum length of string return byMPI_ERROR_STRING
- MPI_UNDEFINED
- - Used by many routines to indicated undefined or unknowninteger value
- MPI_UNDEFINED_RANK
- - Unknown rank
- MPI_KEYVAL_INVALID
- - Special keyval that may be used to detect uninitializedkeyvals.
- MPI_BSEND_OVERHEAD
- - Add this to the size of a MPI_BSENDbuffer for eachoutstanding message
- MPI_PROC_NULL
- - This rank may be used to send or receive from no-one.
- MPI_ANY_SOURCE
- - In a receive, accept a message from anyone.
- MPI_ANY_TAG
- - In a receive, accept a message with any tag value.
- MPI_BOTTOM
- - May be used to indicate the bottom of the address space
TOPOLOGY TYPES
- MPI_GRAPH
- - General graph
- MPI_CART
- - Cartesian grid
MPI STATUS
The
MPI_Statusdatatype is a structure. The three elements for useby programmers are
- MPI_SOURCE
- - Who sent the message
- MPI_TAG
- - What tag the message was sent with
- MPI_ERROR
- - Any error return
SPECIAL MPI TYPES AND FUNCTIONS
- MPI_Aint
- - C type that holds any valid address.
- MPI_Handler_function
- - C function for handling errors (seeMPI_Errhandler_create) .
- MPI_User_function
- - C function to combine values (see collective operationsand MPI_Op_create)
- MPI_Copy_function
- - Function to copy attributes (see MPI_Keyval_create)
- MPI_NULL_COPY_FN
- - Predefined copy function
- MPI_Delete_function
- - Function to delete attributes (see MPI_Keyval_create)
- MPI_NULL_DELETE_FN
- - Predefined delete function
- MPI_DUP_FN
- - Predefined duplication function
- MPI_ERRORS_ARE_FATAL
- - Error handler that forces exit on error
- MPI_ERRORS_RETURN
- - Error handler that returns error codes (as value ofMPI routine in C and through last argument in Fortran)
MPI ERROR CLASSES
- MPI_SUCCESS
- - Successful return code
- MPI_ERR_BUFFER
- - Invalid buffer pointer
- MPI_ERR_COUNT
- - Invalid count argument
- MPI_ERR_TYPE
- - Invalid datatype argument
- MPI_ERR_TAG
- - Invalid tag argument
- MPI_ERR_COMM
- - Invalid communicator
- MPI_ERR_RANK
- - Invalid rank
- MPI_ERR_ROOT
- - Invalid root
- MPI_ERR_GROUP
- - Null group passed to function
- MPI_ERR_OP
- - Invalid operation
- MPI_ERR_TOPOLOGY
- - Invalid topology
- MPI_ERR_DIMS
- - Illegal dimension argument
- MPI_ERR_ARG
- - Invalid argument
- MPI_ERR_UNKNOWN
- - Unknown error
- MPI_ERR_TRUNCATE
- - message truncated on receive
- MPI_ERR_OTHER
- - Other error; use Error_string
- MPI_ERR_INTERN
- - internal error code
- MPI_ERR_IN_STATUS
- - Look in status for error value
- MPI_ERR_PENDING
- - Pending request
- MPI_ERR_REQUEST
- - illegal mpi_request handle
- MPI_ERR_LASTCODE
- - Last error code -- always at end
LOCATION
/home/MPI/mansrc/mpiconsts.txt
Index
- NAME
- DATA TYPES
- C DATATYPES
- FORTRAN DATATYPES
- COMMUNICATORS
- GROUPS
- RESULTS OF THE COMPARE OPERATIONS
- COLLECTIVE OPERATIONS
- NOTES ON COLLECTIVE OPERATIONS
- PERMANENT KEY VALUES
- NULL OBJECTS
- PREDEFINED CONSTANTS
- TOPOLOGY TYPES
- MPI STATUS
- SPECIAL MPI TYPES AND FUNCTIONS
- MPI ERROR CLASSES
- LOCATION
This document was created byman2html,using the manual pages.
