DeinoMPI

The Great and Terrible implementation of MPI-2

MPI_Testsome

Tests for some given requests to complete

int MPI_Testsome(
  int incount,
  MPI_Request array_of_requests[],
  int *outcount,
  int array_of_indices[],
  MPI_Status array_of_statuses[]
);

Parameters

incount: [in] length of array_of_requests (integer)
array_of_requests: [in] array of requests (array of handles)
outcount: [out] number of completed requests (integer)
array_of_indices: [out] array of indices of operations that completed (array of integers)
array_of_statuses: [out] array of status objects for operations that completed (array of Status). May be MPI_STATUSES_IGNORE.

Remarks

Behaves like MPI_WAITSOME, except that it returns immediately. If no operation has completed it returns outcount = 0. If there is no active handle in the list it returns outcount = MPI_UNDEFINED.

MPI_TESTSOME is a local operation, which returns immediately, whereas MPI_WAITSOME willblock until a communication completes, if it was passed a list that contains at least one active handle. Both calls fulfill a fairness requirement: If a request for a receive repeatedly appears in a list of requests passed to MPI_WAITSOME or MPI_TESTSOME, and a matching send has been posted, then the receive will eventually succeed, unless the send is satisfied by another receive; and similarly for send requests.

Errors that occur during the execution of MPI_TESTSOME are handled as for MPI_WAITSOME.

Advice to users.

The use of MPI_TESTSOME is likely to be more efficient than the use of MPI_TESTANY. The former returns information on all completed communications, with the latter, a new call is required for each communication that completes.

A server with multiple clients can use MPI_WAITSOME so as not to starve any client. Clients send messages to the server with service requests. The server calls MPI_WAITSOME with one receive request for each client, and then handles all receives that completed. If a call to MPI_WAITANY is used instead, then one client could starve while requests from another client always sneak in first.

While it is possible to list a request handle more than once in the array_of_requests, such an action is considered erroneous and may cause the program to unexecpectedly terminate or produce incorrect results.

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 mallocor other non-MPICH runtime routines that are themselves not interrupt-safe.

Note on status for send operations

For send operations, the only use of status is for MPI_Test_cancelled or in the case that there is an error, in which case the MPI_ERROR field of status will be set.

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. ierris 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 INTEGERin 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_SUCCESS: No error; MPI routine completed successfully.

MPI_ERR_IN_STATUS: The actual error value is in the MPI_Status argument. This error class is returned only from the multiple-completion routines (MPI_Testall, MPI_Testany, MPI_Testsome, MPI_Waitall, MPI_Waitany, and MPI_Waitsome). The field MPI_ERROR in the status argument contains the error value or MPI_SUCCESS (no error and complete) or MPI_ERR_PENDING to indicate that the request has not completed.

The MPI Standard does not specify what the result of the multiple completion routines is when an error occurs. For example, in an MPI_WAITALL, does the routine wait for all requests to either fail or complete, or does it return immediately (with the MPI definition of immediately, which means independent of actions of other MPI processes)? MPICH has chosen to make the return immediate (alternately, local in MPI terms), and to use the error class MPI_ERR_PENDING (introduced in MPI 1.1) to indicate which requests have not completed. In most cases, only one request with an error will be detected in each call to an MPI routine that tests multiple requests. The requests that have not been processed (because an error occured in one of the requests) will have their MPI_ERROR field marked with MPI_ERR_PENDING.

Example Code

The following sample code illustrates MPI_Testsome.

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

int main(int argc, char *argv[])
{
    int rank, size, flag, i, count;
    int buffer[100];
    MPI_Request r[4];
    MPI_Status status[4];
    int index[4];

    MPI_Init(&argc, &argv);
    MPI_Comm_size(MPI_COMM_WORLD, &size);
    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
    if (size != 4)
    {
        printf("Please run with 4 processes.\n");fflush(stdout);
        MPI_Finalize();
        return 1;
    }

    if (rank == 0)
    {
        int remaining = 3;
        for (i=1; i<size; i++)
        {
            MPI_Irecv(&buffer[i], 1, MPI_INT, i, 123, MPI_COMM_WORLD, &r[i-1]);
        }
        while (remaining > 0)
        {
            MPI_Testsome(size-1, r, &count, index, status);
            if (count > 0)
            {
                printf("%d finished\n", count);
                remaining -= count;
            }
            else
            {
                Sleep(1);
            }
        }
    }
    else
    {
        MPI_Send(buffer, 1, MPI_INT, 0, 123, MPI_COMM_WORLD);
    }
    MPI_Finalize();
    return 0;
}

DOWNLOAD

Win32 DeinoMPI.2.0.1.msi
Win64 DeinoMPI.x64.2.0.1.msi