The Great and Terrible implementation of MPI-2

function index


Marks the communicator object for deallocation
int MPI_Comm_free(
  MPI_Comm *comm


[in] Communicator to be destroyed (handle)


This routine frees a communicator. Because the communicator may still be in use by other MPI routines, the actual communicator storage will not be freed until all references to this communicator are removed. For most users, the effect of this routine is the same as if it was in fact freed at this time of this call.

This collective operation marks the communication object for deallocation. The handle is set to MPI_COMM_NULL. Any pending operations that use this communicator will complete normally; the object is actually deallocated only if there are no other active references to it. This call applies to intra- and inter-communicators. The delete callback functions for all cached attributes are called in arbitrary order.

Null Handles
The MPI 1.1 specification, in the section on opaque objects, explicitly disallows freeing a null communicator. The text from the standard is

 A null handle argument is an erroneous IN argument in MPI calls, unless an
 exception is explicitly stated in the text that defines the function. Such
 exception is allowed for handles to request objects in Wait and Test calls
 (sections Communication Completion and Multiple Completions ). Otherwise, a
 null handle can only be passed to a function that allocates a new object and
 returns a reference to it in the handle.

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.


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.

No error; MPI routine completed successfully.
Invalid communicator. A common error is to use a null communicator in a call (not even allowed in MPI_Comm_rank).
Invalid argument. Some argument is invalid and is not identified by a specific error class (e.g., MPI_ERR_RANK).

Example Code

The following sample code illustrates MPI_Comm_free.

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

/* Test that communicators have reference count semantics */
#define NELM 128
#define NCOMM 1020

int main( int argc, char *argv[] )
int errs = 0;
    int rank, size, source, dest, i;
    MPI_Comm comm;
    MPI_Comm tmpComm[NCOMM];
    MPI_Status status;
    MPI_Request req;
int *buf=0;

    MPI_Init( &argc, &argv );
    MPI_Comm_dup( MPI_COMM_WORLD, &comm );
/* This is similar to the datatype test, except that we post
        an irecv on a simple data buffer but use a rank-reordered communicator.
        In this case, an error in handling the reference count will most
        likely cause the program to hang, so this should be run only
        if (a) you are confident that the code is correct or (b)
        a timeout is set for mpiexec
MPI_Comm_rank( comm, &rank );
    MPI_Comm_size( comm, &size );
    if (size < 2) {
        fprintf( stderr, "This test requires at least two processes." );fflush(stderr);
        MPI_Abort( MPI_COMM_WORLD, 1 );
    source = 0;
    dest = size - 1;
    if (rank == dest) {
        buf = (
int *)malloc( NELM * sizeof(int) );
for (i=0; i<NELM; i++) buf[i] = -i;
        MPI_Irecv( buf, NELM, MPI_INT, source, 0, comm, &req );
        MPI_Comm_free( &comm );
        if (comm != MPI_COMM_NULL) {
            printf( "Freed comm was not set to COMM_NULL\n" );
for (i=0; i<NCOMM; i++) {
            MPI_Comm_split( MPI_COMM_WORLD, 0, size - rank, &tmpComm[i] );
        MPI_Sendrecv( 0, 0, MPI_INT, source, 1, 0, 0, MPI_INT, source, 1, MPI_COMM_WORLD, &status );
        MPI_Wait( &req, &status );
for (i=0; i<NELM; i++) {
if (buf[i] != i) {
if (errs < 10) {
                    printf( "buf[%d] = %d, expected %d\n", i, buf[i], i );
        for (i=0; i<NCOMM; i++) {
            MPI_Comm_free( &tmpComm[i] );
        free( buf );
    else if (rank == source) {
        buf = (
int *)malloc( NELM * sizeof(int) );
for (i=0; i<NELM; i++) buf[i] = i;
        for (i=0; i<NCOMM; i++) {
            MPI_Comm_split( MPI_COMM_WORLD, 0, size - rank, &tmpComm[i] );
/* Synchronize with the receiver */
        MPI_Sendrecv( 0, 0, MPI_INT, dest, 1, 0, 0, MPI_INT, dest, 1, MPI_COMM_WORLD, &status );
        MPI_Send( buf, NELM, MPI_INT, dest, 0, comm );
        free( buf );
else {
for (i=0; i<NCOMM; i++) {
            MPI_Comm_split( MPI_COMM_WORLD, 0, size - rank, &tmpComm[i] );

    MPI_Barrier( MPI_COMM_WORLD );

    if (rank != dest) {
/* Clean up the communicators */
        for (i=0; i<NCOMM; i++) {
            MPI_Comm_free( &tmpComm[i] );
    return 0;