MPI_ALLTOALL(3)

📦
version 5.0.10 · 7 versions
, in ubuntu stonking

🌸 NAME 🌸

MPI_Alltoall, MPI_Ialltoall, MPI_Alltoall_init — All processes send data to all processes 💌

okay so real talk, this is the MPI_ALLTOALL(3) page, straight out of the Open MPI manual, and it is genuinely one of my faves. it's giving everybody talks to everybody, and honestly? iconic behavior. let's get into it bestie 🫶

✨ SYNTAX ✨

🐣 C Syntax

#include <mpi.h>

int MPI_Alltoall(const void *sendbuf, int sendcount,
     MPI_Datatype sendtype, void *recvbuf, int recvcount,
     MPI_Datatype recvtype, MPI_Comm comm)

int MPI_Ialltoall(const void *sendbuf, int sendcount,
     MPI_Datatype sendtype, void *recvbuf, int recvcount,
     MPI_Datatype recvtype, MPI_Comm comm, MPI_Request *request)

int MPI_Alltoall_init(const void *sendbuf, int sendcount,
     MPI_Datatype sendtype, void *recvbuf, int recvcount,
     MPI_Datatype recvtype, MPI_Comm comm, MPI_Info info, MPI_Request *request)

🧚 Fortran Syntax

USE MPI
! or the older form: INCLUDE 'mpif.h'
MPI_ALLTOALL(SENDBUF, SENDCOUNT, SENDTYPE, RECVBUF, RECVCOUNT,
     RECVTYPE, COMM, IERROR)

     <type>  SENDBUF(*), RECVBUF(*)
     INTEGER SENDCOUNT, SENDTYPE, RECVCOUNT, RECVTYPE
     INTEGER COMM, IERROR

MPI_IALLTOALL(SENDBUF, SENDCOUNT, SENDTYPE, RECVBUF, RECVCOUNT,
     RECVTYPE, COMM, REQUEST, IERROR)

     <type>  SENDBUF(*), RECVBUF(*)
     INTEGER SENDCOUNT, SENDTYPE, RECVCOUNT, RECVTYPE
     INTEGER COMM, REQUEST, IERROR

MPI_ALLTOALL_INIT(SENDBUF, SENDCOUNT, SENDTYPE, RECVBUF, RECVCOUNT,
     RECVTYPE, COMM, INFO, REQUEST, IERROR)

     <type>  SENDBUF(*), RECVBUF(*)
     INTEGER SENDCOUNT, SENDTYPE, RECVCOUNT, RECVTYPE
     INTEGER COMM, INFO, REQUEST, IERROR

🦋 Fortran 2008 Syntax

USE mpi_f08
MPI_Alltoall(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype,
             comm, ierror)

     TYPE(*), DIMENSION(..), INTENT(IN) :: sendbuf
     TYPE(*), DIMENSION(..) :: recvbuf
     INTEGER, INTENT(IN) :: sendcount, recvcount
     TYPE(MPI_Datatype), INTENT(IN) :: sendtype, recvtype
     TYPE(MPI_Comm), INTENT(IN) :: comm
     INTEGER, OPTIONAL, INTENT(OUT) :: ierror

MPI_Ialltoall(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype,
             comm, request, ierror)

     TYPE(*), DIMENSION(..), INTENT(IN), ASYNCHRONOUS :: sendbuf
     TYPE(*), DIMENSION(..), ASYNCHRONOUS :: recvbuf
     INTEGER, INTENT(IN) :: sendcount, recvcount
     TYPE(MPI_Datatype), INTENT(IN) :: sendtype, recvtype
     TYPE(MPI_Comm), INTENT(IN) :: comm
     TYPE(MPI_Request), INTENT(OUT) :: request
     INTEGER, OPTIONAL, INTENT(OUT) :: ierror

MPI_Alltoall_init(sendbuf, sendcount, sendtype, recvbuf, recvcount, recvtype,
                     comm, info, request, ierror)

     TYPE(*), DIMENSION(..), INTENT(IN), ASYNCHRONOUS :: sendbuf
     TYPE(*), DIMENSION(..), ASYNCHRONOUS :: recvbuf
     INTEGER, INTENT(IN) :: sendcount, recvcount
     TYPE(MPI_Datatype), INTENT(IN) :: sendtype, recvtype
     TYPE(MPI_Comm), INTENT(IN) :: comm
     TYPE(MPI_Info), INTENT(IN) :: info
     TYPE(MPI_Request), INTENT(OUT) :: request
     INTEGER, OPTIONAL, INTENT(OUT) :: ierror

💜 INPUT PARAMETERS 💜

the stuff you hand to it, bestie:

  • sendbuf — Starting address of send buffer (choice).
  • sendcount — Number of elements to send to each process (integer).
  • sendtype — Datatype of send buffer elements (handle).
  • recvcount — Number of elements to receive from each process (integer).
  • recvtype — Datatype of receive buffer elements (handle).
  • comm — Communicator over which data is to be exchanged (handle).
  • info — Info (handle, persistent only)

🎀 OUTPUT PARAMETERS 🎀

and the stuff it hands back:

  • recvbuf — Starting address of receive buffer (choice).
  • request — Request (handle, non-blocking only).
  • ierror — Fortran only: Error status (integer).

🌈 DESCRIPTION 🌈

MPI_Alltoall is that collective operation where all processes send the same amount of data to each other, and receive the same amount of data from each other. no favorites, no main-character process, everybody eats 🍓. technically you can picture it as each process doing 2n independent point-to-point communications (where n = the number of processes in communicator comm), including talking to itself, because self-love is valid:

MPI_Comm_size(comm, &n);
for (i = 0, i < n; i++)
    MPI_Send(sendbuf + i * sendcount * extent(sendtype),
        sendcount, sendtype, i, ..., comm);
for (i = 0, i < n; i++)
    MPI_Recv(recvbuf + i * recvcount * extent(recvtype),
        recvcount, recvtype, i, ..., comm);

so here's the tea on how the buffers get chopped up: each process breaks its local sendbuf into n blocks — each one containing sendcount elements of type sendtype — and divides its recvbuf the same way, but according to recvcount and recvtype. Process j sends the k-th block of its local sendbuf to process k, which drops that data into the j-th block of its local recvbuf. and the amount of data sent has to equal the amount of data received, pairwise, between every pair of processes. it's a vibe check between every single duo. ✨

WHEN COMMUNICATOR IS AN INTER-COMMUNICATOR

bestie this is where it gets ~fancy~. when the communicator is an inter-communicator, the gather operation happens in two phases. first, data is gathered from all the members of the first group and received by all the members of the second group. then, data is gathered from all the members of the second group and received by all the members of the first. the whole thing exhibits a symmetric, full-duplex behavior. 💅

the first group defines the root process. the root process uses MPI_ROOT as the value of root. all other processes in the first group use MPI_PROC_NULL as the value of root. all processes in the second group use the rank of the root process in the first group as the value of root.

when the communicator is an intra-communicator, these groups are the same, and the operation occurs in a single phase.

🫶 USE OF IN-PLACE OPTION 🫶

when the communicator is an intracommunicator, you can do the all-to-all in-place — meaning the output buffer doubles as the input buffer, recycling era 🔥. just use the variable MPI_IN_PLACE as the value of sendbuf. in that case, sendcount and sendtype are ignored. 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.

👑 NOTES 👑

all arguments on all processes are significant. the comm argument especially has to describe the same communicator on all processes. no sneaky mismatches, we're all agreeing on the same thing here.

there are two MPI library functions that are more general than MPI_Alltoall. MPI_Alltoallv allows all-to-all communication to and from buffers that don't need to be contiguous, and different processes may send and receive different amounts of data. MPI_Alltoallw expands MPI_Alltoallv's functionality to allow the exchange of data with different datatypes. so if plain MPI_Alltoall is too rigid for your situation, you know where to go ✨

💀 ERRORS 💀

almost all MPI routines return an error value; C routines as the return result of the function and Fortran routines in the last argument.

before the error value is returned, the current MPI error handler associated with the communication object (e.g., communicator, window, file) is called. if no communication object is associated with the MPI call, then the call is considered attached to MPI_COMM_SELF and will call the associated MPI error handler. when MPI_COMM_SELF is not initialized (i.e., before MPI_Init/MPI_Init_thread, after MPI_Finalize, or when using the Sessions Model exclusively) the error raises the initial error handler. the initial error handler can be changed by calling MPI_Comm_set_errhandler on MPI_COMM_SELF when using the World model, or the mpi_initial_errhandler CLI argument to mpiexec or info key to MPI_Comm_spawn/MPI_Comm_spawn_multiple. if no other appropriate error handler has been set, then the MPI_ERRORS_RETURN error handler is called for MPI I/O functions and the MPI_ERRORS_ABORT error handler is called for all other MPI functions.

Open MPI includes three predefined error handlers that can be used:

  • MPI_ERRORS_ARE_FATAL — Causes the program to abort all connected MPI processes.
  • MPI_ERRORS_ABORT — An error handler that can be invoked on a communicator, window, file, or session. When called on a communicator, it acts as if MPI_Abort was called on that communicator. If called on a window or file, acts as if MPI_Abort was called on a communicator containing the group of processes in the corresponding window or file. If called on a session, aborts only the local process.
  • MPI_ERRORS_RETURN — Returns an error code to the application.

MPI applications can also implement their own error handlers by calling:

  • MPI_Comm_create_errhandler then MPI_Comm_set_errhandler
  • MPI_File_create_errhandler then MPI_File_set_errhandler
  • MPI_Session_create_errhandler then MPI_Session_set_errhandler or at MPI_Session_init
  • MPI_Win_create_errhandler then MPI_Win_set_errhandler

note that MPI does not guarantee that an MPI program can continue past an error. like, that's the real deal, don't just carry on and hope for the best 😭

see the MPI man page for a full list of MPI error codes.

see the Error Handling section of the MPI-3.1 standard for more information.

see also:

  • MPI_Alltoallv
  • MPI_Alltoallw

2003-2026, The Open MPI Community

and that's the whole page, bestie!! this MPI_ALLTOALL(3) entry is from the Open MPI manual, dated Sep 10, 2026, copyright 2003-2026, The Open MPI Community. go forth and all-to-all, you absolute legend 💖✨🦋💅

📦 where this page comes from