DeinoMPI

The Great and Terrible implementation of MPI-2

function index

MPI_Allgather

Gathers data from all tasks and distribute the combined data to all tasks
int MPI_Allgather(
  void *sendbuf,
  int sendcount,
  MPI_Datatype sendtype,
  void *recvbuf,
  int recvcount,
  MPI_Datatype recvtype,
  MPI_Comm comm
);

Parameters

sendbuf
[in] starting address of send buffer (choice)
sendcount
[in] number of elements in send buffer (integer)
sendtype
[in] data type of send buffer elements (handle)
recvbuf
[out] address of receive buffer (choice)
recvcount
[in] number of elements received from any process (integer)
recvtype
[in] data type of receive buffer elements (handle)
comm
[in] communicator (handle)

Remarks

The MPI standard (1.0 and 1.1) says that

    The jth block of data sent from each proess is received by every process and placed in the jth block of the buffer recvbuf.

This is misleading; a better description is

    The block of data sent from the jth process is received by every process and placed in the jth block of the buffer recvbuf.

MPI_ALLGATHER can be thought of as MPI_GATHER, but where all processes receive the result, instead of just the root. The jth block of data sent from each process is received by every process and placed in the jth block of the buffer recvbuf.

The type signature associated with sendcount, sendtype, at a process must be equal to the type signature associated with recvcount, recvtype at any other process.

The outcome of a call to MPI_ALLGATHER(...) is as if all processes executed n calls to
 

MPI_GATHER(sendbuf,sendcount,sendtype,recvbuf,recvcount, recvtype,root,comm), 
for root = 0 , ..., n-1. The rules for correct usage of MPI_ALLGATHER are easily found from the corresponding rules for MPI_GATHER.

The "in place" option for intracommunicators is specified by passing the value MPI_IN_PLACE to the argument sendbuf at all processes. sendcount and sendtype are ignored. Then the input data of each process is assumed to be in the area where that process would receive its own contribution to the receive buffer. Specifically, the outcome of a call to MPI_ALLGATHER in the "in place" case is as if all processes executed n calls to
 

    MPI_GATHER( MPI_IN_PLACE, 0, MPI_DATATYPE_NULL, recvbuf, recvcount, recvtype, root, comm ) 
for root = 0, ..., n - 1.

If comm is an intercommunicator, then each process in group A contributes a data item; these items are concatenated and the result is stored at each process in group B. Conversely the concatenation of the contributions of the processes in group B is stored at each process in group A. The send buffer arguments in group A must be consistent with the receive buffer arguments in group B, and vice versa.

 


Advice to users.

The communication pattern of MPI_ALLGATHER executed on an intercommunication domain need not be symmetric. The number of items sent by processes in group A (as specified by the arguments sendcount, sendtype in group A and the arguments recvcount, recvtype in group B), need not equal the number of items sent by processes in group B (as specified by the arguments sendcount, sendtype in group B and the arguments recvcount, recvtype in group A). In particular, one can move data in only one direction by specifying sendcount = 0 for the communication in the reverse direction.

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.

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_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.

Example Code

The following sample code illustrates MPI_Allgather.
 

#include "mpi.h"
#include <stdio.h>

#define MAX_PROCESSES 10

int main( int argc, char **argv )
{
    int rank, size, i,j;
   
int table[MAX_PROCESSES][MAX_PROCESSES];
   
int errors=0;
   
int participants;

    MPI_Init( &argc, &argv );
    MPI_Comm_rank( MPI_COMM_WORLD, &rank );
    MPI_Comm_size( MPI_COMM_WORLD, &size );
   
/* Exactly MAX_PROCESSES processes can participate */
   
if ( size >= MAX_PROCESSES ) participants = MAX_PROCESSES;
   
else
   
{
        fprintf( stderr, "Number of processors must be at least %d\n", MAX_PROCESSES );fflush(stderr);
        MPI_Abort( MPI_COMM_WORLD, 1 );
    }
   
if ( (rank < participants) ) {
       
/* Determine what rows are my responsibility */
       
int block_size = MAX_PROCESSES / participants;
       
int begin_row = rank * block_size;
       
int end_row = (rank+1) * block_size;
       
int send_count = block_size * MAX_PROCESSES;
       
int recv_count = send_count;
       
/* Paint my rows my color */
       
for (i=begin_row; i<end_row ;i++)
           
for (j=0; j<MAX_PROCESSES; j++)
                table[i][j] = rank + 10;
       
/* Everybody gets the gathered table */
       
MPI_Allgather(&table[begin_row][0], send_count, MPI_INT,
                           &table[0][0], recv_count, MPI_INT, MPI_COMM_WORLD);
       
/* Everybody should have the same table now, */
       
for (i=0; i<MAX_PROCESSES;i++) {
           
if ( (table[i][0] - table[i][MAX_PROCESSES-1] !=0) )
                errors++;
        }
    }
    MPI_Finalize();
   
if (errors)
    {
        printf( "[%d] done with ERRORS(%d)!\n", rank, errors );fflush(stdout);
    }
   
return errors;
}