17.2.308. MPI_Scan

MPI_Scan, MPI_Iscan, MPI_Scan_init - Computes an inclusive scan (partial reduction)

17.2.308.1. SYNTAX

17.2.308.1.1. C Syntax

#include <mpi.h>

int MPI_Scan(const void *sendbuf, void *recvbuf, int count,
             MPI_Datatype datatype, MPI_Op op, MPI_Comm comm)

int MPI_Iscan(const void *sendbuf, void *recvbuf, int count,
              MPI_Datatype datatype, MPI_Op op, MPI_Comm comm,
              MPI_Request *request)

int MPI_Scan_init(const void *sendbuf, void *recvbuf, int count,
              MPI_Datatype datatype, MPI_Op op, MPI_Comm comm,
              MPI_Info info, MPI_Request *request)

17.2.308.1.2. Fortran Syntax

USE MPI
! or the older form: INCLUDE 'mpif.h'
MPI_SCAN(SENDBUF, RECVBUF, COUNT, DATATYPE, OP, COMM, IERROR)
     <type>  SENDBUF(*), RECVBUF(*)
     INTEGER COUNT, DATATYPE, OP, COMM, IERROR

MPI_ISCAN(SENDBUF, RECVBUF, COUNT, DATATYPE, OP, COMM, REQUEST, IERROR)
     <type>  SENDBUF(*), RECVBUF(*)
     INTEGER COUNT, DATATYPE, OP, COMM, REQUEST, IERROR

MPI_SCAN_INIT(SENDBUF, RECVBUF, COUNT, DATATYPE, OP, COMM, INFO, REQUEST, IERROR)
     <type>  SENDBUF(*), RECVBUF(*)
     INTEGER COUNT, DATATYPE, OP, COMM, INFO, REQUEST, IERROR

17.2.308.1.3. Fortran 2008 Syntax

USE mpi_f08
MPI_Scan(sendbuf, recvbuf, count, datatype, op, comm, ierror)
     TYPE(*), DIMENSION(..), INTENT(IN) :: sendbuf
     TYPE(*), DIMENSION(..) :: recvbuf
     INTEGER, INTENT(IN) :: count
     TYPE(MPI_Datatype), INTENT(IN) :: datatype
     TYPE(MPI_Op), INTENT(IN) :: op
     TYPE(MPI_Comm), INTENT(IN) :: comm
     INTEGER, OPTIONAL, INTENT(OUT) :: ierror

MPI_Iscan(sendbuf, recvbuf, count, datatype, op, comm, request, ierror)
     TYPE(*), DIMENSION(..), INTENT(IN), ASYNCHRONOUS :: sendbuf
     TYPE(*), DIMENSION(..), ASYNCHRONOUS :: recvbuf
     INTEGER, INTENT(IN) :: count
     TYPE(MPI_Datatype), INTENT(IN) :: datatype
     TYPE(MPI_Op), INTENT(IN) :: op
     TYPE(MPI_Comm), INTENT(IN) :: comm
     TYPE(MPI_Request), INTENT(OUT) :: request
     INTEGER, OPTIONAL, INTENT(OUT) :: ierror

MPI_Scan_init(sendbuf, recvbuf, count, datatype, op, comm, info, request, ierror)
     TYPE(*), DIMENSION(..), INTENT(IN), ASYNCHRONOUS :: sendbuf
     TYPE(*), DIMENSION(..), ASYNCHRONOUS :: recvbuf
     INTEGER, INTENT(IN) :: count
     TYPE(MPI_Datatype), INTENT(IN) :: datatype
     TYPE(MPI_Op), INTENT(IN) :: op
     TYPE(MPI_Comm), INTENT(IN) :: comm
     TYPE(MPI_Info), INTENT(IN) :: info
     TYPE(MPI_Request), INTENT(OUT) :: request
     INTEGER, OPTIONAL, INTENT(OUT) :: ierror

17.2.308.2. INPUT PARAMETERS

  • sendbuf: Send buffer (choice).

  • count: Number of elements in input buffer (integer).

  • datatype: Data type of elements of input buffer (handle).

  • op: Operation (handle).

  • comm: Communicator (handle).

  • info: Info (handle, persistent only)

17.2.308.3. OUTPUT PARAMETERS

  • recvbuf: Receive buffer (choice).

  • request: Request (handle, non-blocking only).

  • IERROR: Fortran only: Error status (integer).

17.2.308.4. DESCRIPTION

MPI_Scan is used to perform an inclusive prefix reduction on data distributed across the calling processes. The operation returns, in the recvbuf of the process with rank i, the reduction (calculated according to the function op) of the values in the sendbufs of processes with ranks 0, …, i (inclusive). The type of operations supported, their semantics, and the constraints on send and receive buffers are as for MPI_Reduce.

17.2.308.5. EXAMPLE

This example uses a user-defined operation to produce a segmented scan. A segmented scan takes, as input, a set of values and a set of logicals, where the logicals delineate the various segments of the scan. For example,

values     v1      v2      v3      v4      v5      v6      v7      v8
logicals   0       0       1       1       1       0       0       1
result     v1    v1+v2     v3    v3+v4  v3+v4+v5   v6    v6+v7     v8

The result for rank j is thus the sum v(i) + … + v(j), where i is the lowest rank such that for all ranks n, i <= n <= j, logical(n) = logical(j). The operator that produces this effect is

[ u ]     [ v ]     [ w ]
[   ]  o  [   ]  =  [   ]
[ i ]     [ j ]     [ j ]

where

( u + v if i = j w = ( ( v if i != j

Note that this is a noncommutative operator. C code that implements it is given below.

typedef struct {
        double val;
        int log;
} SegScanPair;

/*
 * the user-defined function
 */
void segScan(SegScanPair *in, SegScanPair *inout, int *len,
        MPI_Datatype *dptr)
{
        int i;
        SegScanPair c;

        for (i = 0; i < *len; ++i) {
                if (in->log == inout->log)
                        c.val = in->val + inout->val;
                else
                        c.val = inout->val;

                c.log = inout->log;
                *inout = c;
                in++;
                inout++;
        }
}

Note that the inout argument to the user-defined function corresponds to the right-hand operand of the operator. When using this operator, we must be careful to specify that it is noncommutative, as in the following:

int                     i, base;
SeqScanPair     a, answer;
MPI_Op          myOp;
MPI_Datatype    type[2] = {MPI_DOUBLE, MPI_INT};
MPI_Aint                disp[2];
int                     blocklen[2] = {1, 1};
MPI_Datatype    sspair;

/*
 * explain to MPI how type SegScanPair is defined
 */
MPI_Get_address(a, disp);
MPI_Get_address(a.log, disp + 1);
base = disp[0];
for (i = 0; i < 2; ++i)
        disp[i] -= base;
MPI_Type_struct(2, blocklen, disp, type, &sspair);
MPI_Type_commit(&sspair);

/*
 * create the segmented-scan user-op
 * noncommutative - set commute (arg 2) to 0
 */
MPI_Op_create((MPI_User_function *)segScan, 0, &myOp);
...
MPI_Scan(a, answer, 1, sspair, myOp, comm);

17.2.308.6. USE OF IN-PLACE OPTION

When the communicator is an intracommunicator, you can perform a scanning operation in place (the output buffer is used as the input buffer). Use the variable MPI_IN_PLACE as the value of the sendbuf argument. The input data is taken from the receive buffer and replaced by the output data.

17.2.308.7. NOTES ON COLLECTIVE OPERATIONS

The reduction functions of type 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_Abort or silently skip the problem. Thus, if the error handler is changed from MPI_ERRORS_ARE_FATAL to something else (e.g., MPI_ERRORS_RETURN), then no error may be indicated.

The reason for this is the performance problems in ensuring that all collective routines return the same error value.

17.2.308.8. ERRORS

Almost all MPI routines return an error value; C routines as the return result of the function and Fortran routines in the last argument.

Before the error value is returned, the current MPI error handler associated with the communication object (e.g., communicator, window, file) is called. If no communication object is associated with the MPI call, then the call is considered attached to MPI_COMM_SELF and will call the associated MPI error handler. When MPI_COMM_SELF is not initialized (i.e., before MPI_INIT / MPI_INIT_THREAD, after MPI_FINALIZE, or when using the Sessions Model exclusively) the error raises the initial error handler. The initial error handler can be changed by calling MPI_COMM_SET_ERRHANDLER on MPI_COMM_SELF when using the World model, or the mpi_initial_errhandler CLI argument to mpiexec or info key to MPI_COMM_SPAWN[_MULTIPLE]. If no other appropriate error handler has been set, then the MPI_ERRORS_RETURN error handler is called for MPI I/O functions and the MPI_ERRORS_ABORT error handler is called for all other MPI functions.

In the sessions model, the error handler can be set during MPI_Session_init.

Open MPI includes three predefined error handlers that can be used:

MPI_ERRORS_ARE_FATAL: Causes the program to abort all connected MPI processes.
MPI_ERRORS_ABORT:     An error handler that can be invoked on a communicator,
                      window, file, or session. When called on a communicator, it
                      acts as if MPI_ABORT was called on that communicator. If
                      called on a window or file, acts as if MPI_ABORT was called
                      on a communicator containing the group of processes in the
                      corresponding window or file. If called on a session,
                      aborts only the local process.
MPI_ERRORS_RETURN:    Returns an error code to the application.

MPI applications can also implement their own error handlers.

Custom MPI error handlers can be created by calling: MPI_Comm_create_errhandler(3) MPI_File_create_errhandler(3) MPI_Session_create_errhandler(3) MPI_Win_create_errhandler(3)

Predefined and custom error handlers can be set by calling: MPI_Comm_set_errhandler(3) MPI_File_set_errhandler(3) MPI_Session_set_errhandler(3) MPI_Win_set_errhandler(3)

Note that MPI does not guarantee that an MPI program can continue past an error.

See the MPI man page for a full list of MPI error codes.

See the Error Handling section of the MPI-3 standard for more information.