The Great and Terrible implementation of MPI-2

function index


Initiates a communication with a persistent request handle
int MPI_Start(
  MPI_Request *request


[in] communication request (handle)


The argument, request, is a handle returned by one of the MPI_xsend/recv_init calls. The associated request should be inactive. The request becomes active once the call is made.

If the request is for a send with ready mode, then a matching receive should be posted before the call is made. The communication buffer should not be accessed after the call, and until the operation completes.

The call is local, with similar semantics to the nonblocking communication operations. That is, a call to MPI_START with a request created by MPI_SEND_INIT starts a communication in the same manner as a call to MPI_ISEND; a call to MPI_START with a request created by MPI_BSEND_INIT starts a communication in the same manner as a call to MPI_IBSEND; and so on.

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 MPI_Request. Either null or, in the case of a MPI_Start or MPI_Startall, not a persistent request.

Example Code

The following sample code illustrates MPI_Start.

#include "mpi.h"
#include <stdlib.h>
int main(int argc, char *argv[])
    MPI_Request r;
    MPI_Status s;
    int flag;
int buf[10];
    int rbuf[10];
    int tag = 27;
    int dest = 0;
    int rank, size, i;
    MPI_Init( &argc, &argv );
    MPI_Comm_size( MPI_COMM_WORLD, &size );
    MPI_Comm_rank( MPI_COMM_WORLD, &rank );
    /* Create a persistent send request */
    MPI_Send_init( buf, 10, MPI_INT, dest, tag, MPI_COMM_WORLD, &r );
    /* Use that request */
if (rank == 0) {
int i;
        MPI_Request *rr = (MPI_Request *)malloc(size *
for (i=0; i<size; i++) {
            MPI_Irecv( rbuf, 10, MPI_INT, i, tag, MPI_COMM_WORLD, &rr[i] );
        MPI_Start( &r );
        MPI_Wait( &r, &s );
        MPI_Waitall( size, rr, MPI_STATUSES_IGNORE );
    else {
        MPI_Start( &r );
        MPI_Wait( &r, &s );
    MPI_Request_free( &r );
    if (rank == 0)
        MPI_Request sr;
        /* Create a persistent receive request */
        MPI_Recv_init( rbuf, 10, MPI_INT, MPI_ANY_SOURCE, tag, MPI_COMM_WORLD, &r );
        MPI_Isend( buf, 10, MPI_INT, 0, tag, MPI_COMM_WORLD, &sr );
        for (i=0; i<size; i++) {
            MPI_Start( &r );
            MPI_Wait( &r, &s );
        MPI_Wait( &sr, &s );
        MPI_Request_free( &r );
else {
        MPI_Send( buf, 10, MPI_INT, 0, tag, MPI_COMM_WORLD );
return 0;