DeinoMPI

The Great and Terrible implementation of MPI-2

function index

MPI_Scan

Computes the scan (partial reductions) of data on a collection of processes
int MPI_Scan(
  void *sendbuf,
  void *recvbuf,
  int count,
  MPI_Datatype datatype,
  MPI_Op op,
  MPI_Comm comm
);

Parameters

sendbuf
[in] starting address of send buffer (choice)
recvbuf
[out] starting address of receive buffer (choice)
count
[in] number of elements in input buffer (integer)
datatype
[in] data type of elements of input buffer (handle)
op
[in] operation (handle)
comm
[in] communicator (handle)

Remarks

MPI_SCAN is used to perform a prefix reduction on data distributed across the group. The operation returns, in the receive buffer of the process with rank i, the reduction of the values in the send buffers 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.


Rationale.

We have defined an inclusive scan, that is, the prefix reduction on process i includes the data from process i. An alternative is to define scan in an exclusive manner, where the result on i only includes data up to i-1. Both definitions are useful. The latter has some advantages: the inclusive scan can always be computed from the exclusive scan with no additional communication; for non-invertible operations such as max and min, communication is required to compute the exclusive scan from the inclusive scan. There is, however, a complication with exclusive scan since one must define the "unit" element for the reduction in this case. That is, one must explicitly say what occurs for process 0. This was thought to be complex for user-defined operations and hence, the exclusive scan was dropped.

Thread and Interrupt Safety

This routine is thread-safe. This means that this routine may be safely used by multiple threads without the need for any user-provided thread locks. However, the routine is not interrupt safe. Typically, this is due to the use of memory allocation routines such as malloc or other non-MPICH runtime routines that are themselves not interrupt-safe.

Notes for Fortran

All MPI routines in Fortran (except for MPI_WTIME and MPI_WTICK) have an additional argument ierr at the end of the argument list. ierr is an integer and has the same meaning as the return value of the routine in C. In Fortran, MPI routines are subroutines, and are invoked with the call statement.

All MPI objects (e.g., MPI_Datatype, MPI_Comm) are of type INTEGER in Fortran.

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_Abort or silently skip the problem. Thus, if you change the error handler from MPI_ERRORS_ARE_FATAL to something else, for example, 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.

Errors

All MPI routines (except MPI_Wtime and MPI_Wtick) return an error value; C routines as the value of the function and Fortran routines in the last argument. Before the value is returned, the current MPI error handler is called. By default, this error handler aborts the MPI job. The error handler may be changed with MPI_Comm_set_errhandler (for communicators), MPI_File_set_errhandler (for files), and MPI_Win_set_errhandler (for RMA windows). The MPI-1 routine MPI_Errhandler_set may be used but its use is deprecated. The predefined error handler MPI_ERRORS_RETURN may be used to cause error values to be returned. Note that MPI does not guarentee that an MPI program can continue past an error; however, MPI implementations will attempt to continue whenever possible.

MPI_SUCCESS
No error; MPI routine completed successfully.
MPI_ERR_COMM
Invalid communicator. A common error is to use a null communicator in a call (not even allowed in MPI_Comm_rank).
MPI_ERR_COUNT
Invalid count argument. Count arguments must be non-negative; a count of zero is often valid.
MPI_ERR_TYPE
Invalid datatype argument. May be an uncommitted MPI_Datatype (see MPI_Type_commit).
MPI_ERR_BUFFER
Invalid buffer pointer. Usually a null buffer where one is not valid.
MPI_ERR_BUFFER
This error class is associcated with an error code that indicates that two buffer arguments are aliased; that is, the describe overlapping storage (often the exact same storage). This is prohibited in MPI (because it is prohibited by the Fortran standard, and rather than have a separate case for C and Fortran, the MPI Forum adopted the more restrictive requirements of Fortran).

Example Code

The following sample code illustrates MPI_Scan.

#include "mpi.h"
#include <stdio.h>
 
void addem ( int *, int *, int *, MPI_Datatype * );
void assoc ( int *, int *, int *, MPI_Datatype * );
 
void addem( int *invec, int *inoutvec, int *len, MPI_Datatype *dtype)
{
   
int i;
    for ( i=0; i<*len; i++ )
    inoutvec[i] += invec[i];
}
#define BAD_ANSWER 100000
/*
    The operation is inoutvec[i] = invec[i] op inoutvec[i]
    (see 4.9.4). The order is important.
    Note that the computation is in process rank (in the communicator) order, independant of the root.
*/
void assoc( int *invec, int *inoutvec, int *len, MPI_Datatype *dtype)
{
    int i;
    for ( i=0; i<*len; i++ ) {
        if (inoutvec[i] <= invec[i] ) {
            int rank;
            MPI_Comm_rank( MPI_COMM_WORLD, &rank );
            fprintf( stderr, "[%d] inout[0] = %d, in[0] = %d\n", rank, inoutvec[0], invec[0] );fflush(stderr);
            inoutvec[i] = BAD_ANSWER;
        }
        else
            inoutvec[i] = invec[i];
    }
}
 
int main( int argc, char **argv )
{
   
int rank, size, i;
    int data;
   
int errors=0;
    int result = -100;
   
int correct_result;
    MPI_Op op_assoc, op_addem;
    MPI_Comm comm=MPI_COMM_WORLD;
 
    MPI_Init( &argc, &argv );
    MPI_Op_create( (MPI_User_function *)assoc, 0, &op_assoc );
    MPI_Op_create( (MPI_User_function *)addem, 1, &op_addem );
 
    /* Run this for a variety of communicator sizes */
 
    MPI_Comm_rank( comm, &rank );
    MPI_Comm_size( comm, &size );
 
    data = rank;

    correct_result = 0;
   
for (i=0;i<=rank;i++)
        correct_result += i;

    MPI_Scan ( &data, &result, 1, MPI_INT, MPI_SUM, comm );
   
if (result != correct_result) {
        fprintf( stderr, "[%d] Error suming ints with scan\n", rank );fflush(stderr);
        errors++;
    }
    MPI_Scan ( &data, &result, 1, MPI_INT, MPI_SUM, comm );
   
if (result != correct_result) {
        fprintf( stderr, "[%d] Error summing ints with scan (2)\n", rank );fflush(stderr);
        errors++;
    }

    data = rank;
    result = -100;
    MPI_Scan ( &data, &result, 1, MPI_INT, op_addem, comm );
    if (result != correct_result) {
        fprintf( stderr, "[%d] Error summing ints with scan (userop)\n", rank );fflush(stderr);
        errors++;
    }

    MPI_Scan ( &data, &result, 1, MPI_INT, op_addem, comm );
    if (result != correct_result) {
        fprintf( stderr, "[%d] Error summing ints with scan (userop2)\n", rank );fflush(stderr);
        errors++;
    }
    result = -100;
    data = rank;
    MPI_Scan ( &data, &result, 1, MPI_INT, op_assoc, comm );
   
if (result == BAD_ANSWER) {
        fprintf( stderr, "[%d] Error scanning with non-commutative op\n", rank );fflush(stderr);
        errors++;
    }
 
    MPI_Op_free( &op_assoc );
    MPI_Op_free( &op_addem );

    MPI_Finalize();
   
return errors;
}