MPI_NEIGHBOR_ALLTOALLV_INIT(3)

📦

version 5.0.10 · 4 versions

💖 MPI_NEIGHBOR_ALLTOALLV_INIT(3) 💖

okay bestie, this is the Open MPI manual page for MPI_NEIGHBOR_ALLTOALLV_INIT(3) and we are about to understand it fr fr. no cap, it's giving "all the neighborly data exchange you could ever want." 💅

✨ NAME ✨

MPI_Neighbor_alltoallv, MPI_Ineighbor_alltoallv, and MPI_Neighbor_alltoallv_init — all processes send different amounts of data to, and receive different amounts of data from, all neighbors. 👑🦋

🌸 SYNTAX 🌸

💜 C Syntax 💜

          #include <mpi.h>

          int MPI_Neighbor_alltoallv(const void *sendbuf, const int sendcounts[],
               const int sdispls[], MPI_Datatype sendtype,
               void *recvbuf, const int recvcounts[],
               const int rdispls[], MPI_Datatype recvtype, MPI_Comm comm)

          int MPI_Ineighbor_alltoallv(const void *sendbuf, const int sendcounts[],
               const int sdispls[], MPI_Datatype sendtype,
               void *recvbuf, const int recvcounts[],
               const int rdispls[], MPI_Datatype recvtype, MPI_Comm comm,
               MPI_Request *request)

          int MPI_Neighbor_alltoallv_init(const void *sendbuf, const int sendcounts[],
               const int sdispls[], MPI_Datatype sendtype,
               void *recvbuf, const int recvcounts[],
               const int rdispls[], MPI_Datatype recvtype, MPI_Comm comm,
               MPI_Info info, MPI_Request *request)

🧚 Fortran Syntax 🧚

          USE MPI
          ! or the older form: INCLUDE 'mpif.h'
          MPI_NEIGHBOR_ALLTOALLV(SENDBUF, SENDCOUNTS, SDISPLS, SENDTYPE,
               RECVBUF, RECVCOUNTS, RDISPLS, RECVTYPE, COMM, IERROR)

               <type>  SENDBUF(*), RECVBUF(*)
               INTEGER SENDCOUNTS(*), SDISPLS(*), SENDTYPE
               INTEGER RECVCOUNTS(*), RDISPLS(*), RECVTYPE
               INTEGER COMM, IERROR

          MPI_INEIGHBOR_ALLTOALLV(SENDBUF, SENDCOUNTS, SDISPLS, SENDTYPE,
               RECVBUF, RECVCOUNTS, RDISPLS, RECVTYPE, COMM, REQUEST, IERROR)

               <type>  SENDBUF(*), RECVBUF(*)
               INTEGER SENDCOUNTS(*), SDISPLS(*), SENDTYPE
               INTEGER RECVCOUNTS(*), RDISPLS(*), RECVTYPE
               INTEGER COMM, REQUEST, IERROR

          MPI_NEIGHBOR_ALLTOALLV_INIT(SENDBUF, SENDCOUNTS, SDISPLS, SENDTYPE,
               RECVBUF, RECVCOUNTS, RDISPLS, RECVTYPE, COMM, INFO, REQUEST, IERROR)

               <type>  SENDBUF(*), RECVBUF(*)
               INTEGER SENDCOUNTS(*), SDISPLS(*), SENDTYPE
               INTEGER RECVCOUNTS(*), RDISPLS(*), RECVTYPE
               INTEGER COMM, INFO, REQUEST, IERROR

🌈 Fortran 2008 Syntax 🌈

          USE mpi_f08
          MPI_Neighbor_alltoallv(sendbuf, sendcounts, sdispls, sendtype, recvbuf,
                       recvcounts, rdispls, recvtype, comm, ierror)

               TYPE(*), DIMENSION(..), INTENT(IN) :: sendbuf
               TYPE(*), DIMENSION(..) :: recvbuf
               INTEGER, INTENT(IN) :: sendcounts(*), sdispls(*), recvcounts(*),
               rdispls(*)
               TYPE(MPI_Datatype), INTENT(IN) :: sendtype, recvtype
               TYPE(MPI_Comm), INTENT(IN) :: comm
               INTEGER, OPTIONAL, INTENT(OUT) :: ierror

          MPI_Ineighbor_alltoallv(sendbuf, sendcounts, sdispls, sendtype, recvbuf,
                       recvcounts, rdispls, recvtype, comm, request, ierror)

               TYPE(*), DIMENSION(..), INTENT(IN), ASYNCHRONOUS :: sendbuf
               TYPE(*), DIMENSION(..), ASYNCHRONOUS :: recvbuf
               INTEGER, INTENT(IN), ASYNCHRONOUS :: sendcounts(*), sdispls(*),
               recvcounts(*), rdispls(*)
               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_Neighbor_alltoallv_init(sendbuf, sendcounts, sdispls, sendtype, recvbuf,
                       recvcounts, rdispls, recvtype, comm, info, request, ierror)

               TYPE(*), DIMENSION(..), INTENT(IN), ASYNCHRONOUS :: sendbuf
               TYPE(*), DIMENSION(..), ASYNCHRONOUS :: recvbuf
               INTEGER, INTENT(IN), ASYNCHRONOUS :: sendcounts(*), sdispls(*),
               recvcounts(*), rdispls(*)
               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 🫶

sendbuf — starting address of the send buffer. it's where the data you're sending lives. 📦✨
sendcounts — an integer array, where entry i specifies the number of elements to send to neighbor i. so many elements, so little time. 💅
sdispls — an integer array, where entry i specifies the displacement (offset from sendbuf, in units of sendtype) from which to send data to neighbor i. it's the little offset that could. 🦋
sendtype — datatype of send buffer elements. yes bestie, the flavor of your data. 🌸
recvcounts — an integer array, where entry j specifies the number of elements to receive from neighbor j. incoming! 📥
rdispls — an integer array, where entry j specifies the displacement (offset from recvbuf, in units of recvtype) to which data from neighbor j should be written. where the magic lands. ✨
recvtype — datatype of receive buffer elements. another flavor, another vibe. 🍓
comm — communicator over which data is to be exchanged. this is the group chat, basically. 💬
info — Info (handle, persistent only). only for the persistent version, iykyk. 👑

🎀 OUTPUT PARAMETERS 🎀

recvbuf — address of receive buffer. it's where the good stuff arrives. 📬
request — Request (handle, non-blocking only). only for the non-blocking version. ⏳
ierror — Fortran only: error status. for my Fortran besties. 🧚

👑 DESCRIPTION 👑

okay so MPI_Neighbor_alltoallv is a generalized collective operation in which all processes send data to and receive data from all neighbors. it adds flexibility to MPI_Neighbor_alltoall by letting you specify data to send and receive vector-style — via a displacement and element count. the way this works is each process performs 2n independent point-to-point communications, where n is the number of neighbors in the topology of communicator comm. the neighbors and buffer layout are determined by the topology of comm. here's the vibe, lowkey:

          MPI_Cart_get(comm, maxdims, dims, periods, coords);
          for (dim = 0, i = 0 ; dim < dims ; ++dim) {
              MPI_Cart_shift(comm, dim, 1, &r0, &r1);
              MPI_Isend(sendbuf + sdispls[i]  * extent(sendtype),
                        sendcount, sendtype, r0, ..., comm, ...);
              MPI_Irecv(recvbuf + rdispls[i] * extent(recvtype),
                        recvcount, recvtype, r0, ..., comm, ...);
              ++i;
              MPI_Isend(sendbuf + sdispls[i] * extent(sendtype),
                        sendcount, sendtype, r1, ..., comm, &req[i]);
              MPI_Irecv(recvbuf + rdispls[i] * extent(recvtype),
                        recvcount, recvtype, r1, ..., comm, ...);
              ++i;
          }

process j sends the k-th block of its local sendbuf to neighbor k, which places the data in the j-th block of its local recvbuf. periodt. 💅

when a pair of processes exchanges data, each may pass different element count and datatype arguments — so long as the sender specifies the same amount of data to send (in bytes) as the receiver expects to receive. no cap, it's giving flexible. ✨

note that process i may send a different amount of data to process j than it receives from process j. also, a process may send entirely different amounts of data to different processes in the communicator. main character energy tbh. 👑

🌈 NEIGHBOR ORDERING 🌈

for a distributed graph topology, created with MPI_Dist_graph_create, the sequence of neighbors in the send and receive buffers at each process is defined as the sequence returned by MPI_Dist_graph_neighbors for destinations and sources, respectively. for a general graph topology, created with MPI_Graph_create, the order of neighbors in the send and receive buffers is defined as the sequence of neighbors as returned by MPI_Graph_neighbors. note that general graph topologies should generally be replaced by the distributed graph topologies. bestie, upgrade your topology. 🦋

for a Cartesian topology, created with MPI_Cart_create, the sequence of neighbors in the send and receive buffers at each process is defined by order of the dimensions, first the neighbor in the negative direction and then in the positive direction with displacement 1. the numbers of sources and destinations in the communication routines are 2*ndims with ndims defined in MPI_Cart_create. if a neighbor does not exist, i.e., at the border of a Cartesian topology in the case of a non-periodic virtual grid dimension (i.e., periods[…]==false), then this neighbor is defined to be MPI_PROC_NULL. 💀

if a neighbor in any of the functions is MPI_PROC_NULL, then the neighborhood collective communication behaves like a point-to-point communication with MPI_PROC_NULL in this direction. that is, the buffer is still part of the sequence of neighbors but it is neither communicated nor updated. so it's like, present but not participating. 🙈

💀 NOTES 💀

the MPI_IN_PLACE option for sendbuf is not meaningful for this operation. so don't even try it, bestie. 🙅♀️

the specification of counts and displacements should not cause any location to be written more than once. so no double-dipping on memory locations, okay? ✋

all arguments on all processes are significant. the comm argument, in particular, must describe the same communicator on all processes. no cap. 💯

the offsets of sdispls and rdispls are measured in units of sendtype and recvtype, respectively. compare this to MPI_Neighbor_alltoallw, where these offsets are measured in bytes. iykyk — units matter. 💅

ERRORS 💀✨ (bestie, the error tea ☕)

Okay so literally almost all MPI routines hand you back an error value. For C routines, it's the return result of the function. For Fortran routines? It's in the last argument. No cap. 🫶

Before that error value even comes back, the current MPI error handler that's associated with the communication object (e.g., communicator, window, file) gets called. If there's no communication object associated with the MPI call, then the call is treated like it's attached to MPI_COMM_SELF, and it'll call the associated MPI error handler. But when MPI_COMM_SELF is not initialized — like before MPI_Init/MPI_Init_thread, after MPI_Finalize, or when using the Sessions Model exclusively — the error raises the initial error handler. That initial error handler can be changed by calling MPI_Comm_set_errhandler on MPI_COMM_SELF when using the World model, or by the mpi_initial_errhandler CLI argument to mpiexec or an 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. So like, bestie, be careful. ⚠️

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. 📚