MPI_RACCUMULATE(3)

📦

version 5.0.11 · 16 versions

💅 MPI_RACCUMULATE(3) — the Open MPI glow-up

okay so bestie, welcome to the Open MPI manual page for MPI_Accumulate and MPI_Raccumulate — yes it's one page for two functions, they're sisters, they come as a set, we love that for them. This is the page that tells you how to smoosh your origin buffer into somebody else's target buffer with math. iconic behavior honestly ✨

🌸 NAME 🌸

MPI_Accumulate, MPI_Raccumulate — Combines the contents of the origin buffer with that of a target buffer.

Like, it's giving "I have data, you have data, let's do arithmetic about it." 🫶

✨ SYNTAX ✨

🎀 C Syntax

#include <mpi.h>

int MPI_Accumulate(const void *origin_addr, int origin_count,
     MPI_Datatype origin_datatype, int target_rank,
     MPI_Aint target_disp, int target_count,
     MPI_Datatype target_datatype, MPI_Op op, MPI_Win win)

int MPI_Raccumulate(const void *origin_addr, int origin_count,
     MPI_Datatype origin_datatype, int target_rank,
     MPI_Aint target_disp, int target_count,
     MPI_Datatype target_datatype, MPI_Op op, MPI_Win win,
     MPI_Request *request)

🦋 Fortran Syntax

USE MPI
! or the older form: INCLUDE 'mpif.h'
MPI_ACCUMULATE(ORIGIN_ADDR, ORIGIN_COUNT, ORIGIN_DATATYPE, TARGET_RANK,
     TARGET_DISP, TARGET_COUNT, TARGET_DATATYPE, OP, WIN, IERROR)
     <type> ORIGIN_ADDR(*)
     INTEGER(KIND=MPI_ADDRESS_KIND) TARGET_DISP
     INTEGER ORIGIN_COUNT, ORIGIN_DATATYPE, TARGET_RANK, TARGET_COUNT,
     TARGET_DATATYPE, OP, WIN, IERROR

MPI_RACCUMULATE(ORIGIN_ADDR, ORIGIN_COUNT, ORIGIN_DATATYPE, TARGET_RANK,
     TARGET_DISP, TARGET_COUNT, TARGET_DATATYPE, OP, WIN, REQUEST, IERROR)
     <type> ORIGIN_ADDR(*)
     INTEGER(KIND=MPI_ADDRESS_KIND) TARGET_DISP
     INTEGER ORIGIN_COUNT, ORIGIN_DATATYPE, TARGET_RANK, TARGET_COUNT,
     TARGET_DATATYPE, OP, WIN, REQUEST, IERROR

🧚 Fortran 2008 Syntax

USE mpi_f08
MPI_Accumulate(origin_addr, origin_count, origin_datatype, target_rank,
             target_disp, target_count, target_datatype, op, win, ierror)
     TYPE(*), DIMENSION(..), INTENT(IN), ASYNCHRONOUS :: origin_addr
     INTEGER, INTENT(IN) :: origin_count, target_rank, target_count
     TYPE(MPI_Datatype), INTENT(IN) :: origin_datatype, target_datatype
     INTEGER(KIND=MPI_ADDRESS_KIND), INTENT(IN) :: target_disp
     TYPE(MPI_Op), INTENT(IN) :: op
     TYPE(MPI_Win), INTENT(IN) :: win
     INTEGER, OPTIONAL, INTENT(OUT) :: ierror

MPI_Raccumulate(origin_addr, origin_count, origin_datatype, target_rank,
     target_disp, target_count, target_datatype, op, win, request,
             ierror)
     TYPE(*), DIMENSION(..), INTENT(IN), ASYNCHRONOUS :: origin_addr
     INTEGER, INTENT(IN) :: origin_count, target_rank, target_count
     TYPE(MPI_Datatype), INTENT(IN) :: origin_datatype, target_datatype
     INTEGER(KIND=MPI_ADDRESS_KIND), INTENT(IN) :: target_disp
     TYPE(MPI_Op), INTENT(IN) :: op
     TYPE(MPI_Win), INTENT(IN) :: win
     TYPE(MPI_Request), INTENT(OUT) :: request
     INTEGER, OPTIONAL, INTENT(OUT) :: ierror

💖 INPUT PARAMETERS 💖

okay bestie, here's the whole squad, every single one, no cap:

origin_addr — Initial address of buffer (choice). This is your stuff, the data you're bringing to the party 🎁
origin_count — Number of entries in buffer (nonnegative integer). Count 'em.
origin_datatype — Data type of each buffer entry (handle).
target_rank — Rank of target (nonnegative integer). Who's getting the goods 👀
target_disp — Displacement from start of window to beginning of target buffer (nonnegative integer).
target_count — Number of entries in target buffer (nonnegative integer).
target_datatype — Data type of each entry in target buffer (handle).
op — Reduce operation (handle). The math. The vibe. The whole point 🔥
win — Window object (handle).

🎀 OUTPUT PARAMETER 🎀

MPI_Raccumulate — RMA request
ierror — Fortran only: Error status (integer).

👑 DESCRIPTION 👑

So MPI_Accumulate is a one-sided MPI communication function — meaning only ONE side is doing the work, main character energy, no handshake drama — and what it does is: it adds the contents of the origin buffer (as defined by origin_addr, origin_count, and origin_datatype) to the buffer specified by the arguments target_count and target_datatype, at offset target_disp, in the target window specified by target_rank and win, using the operation op.

The target window can only be accessed by processes within the same node. (Yes that's a real constraint, read it again, it's not a suggestion, it's the rule.)

This is similar to MPI_Put, except that data is combined into the target area instead of overwriting it. So like, MPI_Put is "here's my data, replace yours," and MPI_Accumulate is "here's my data, let's merge." growth mindset tbh 💅

Any of the predefined operations for MPI_Reduce can be used. User-defined functions cannot be used. Not allowed. Not today. If op is MPI_SUM, each element of the origin buffer is added to the corresponding element in the target, replacing the former value in the target. Math happened, the old value is gone, that's just how it is 🫶

Each datatype argument must be a predefined data type or a derived data type, where all basic components are of the same predefined data type. Both datatype arguments must be constructed from the same predefined data type. The operation op applies to elements of that predefined type. The target_datatype argument must not specify overlapping entries, and the target buffer must fit in the target window. (Overlapping entries? ✋ Not on this page. Behave.)

A new predefined operation, MPI_REPLACE, is defined. It corresponds to the associative function f(a, b) =b; that is, the current value in the target memory is replaced by the value supplied by the origin. (Yes, that's literally "b wins," that's the whole function, and yes it's still associative, slay.)

MPI_Raccumulate is similar to MPI_Accumulate, except that it allocates a communication request object and associates it with the request handle (the argument request) that can be used to wait or test for completion.

⚠️ bestie listen to me: The completion of an MPI_Raccumulate operation indicates that the origin_addr buffer is free to be updated. It does not indicate that the operation has completed at the target window. Do not confuse these. Do not pass go. Do not touch that buffer thinking the target is done. 😭

🦋 NOTES 🦋

MPI_Put is a special case of MPI_Accumulate, with the operation MPI_REPLACE. Cute! Note, however, that MPI_Put and MPI_Accumulate have different constraints on concurrent updates. (Different rules, same energy — don't mix them up.)

It is the user's responsibility to guarantee that, when using the accumulate functions, the target displacement argument is such that accesses to the window are properly aligned according to the data type arguments in the call to the MPI_Accumulate function. That's on you, bestie. Align your displacements. 🙈

🌈 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. 💀 everybody's done.
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… don't bank on it. 🫠)

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_Put
MPI_Get_accumulate
MPI_Reduce

(yes, go read those too, they're all in the same friend group 💜)

🍓 COPYRIGHT 🍓

2003-2026, The Open MPI Community

and that's the whole page, bestie! fr fr you now understand accumulate better than half the internet. this was Open MPI's MPI_RACCUMULATE(3) page, dated Sep 11, 2026 — go forth and accumulate responsibly, ily, mwah 💋✨🎀

📦 where this page comes from

arch rolling: openmpi 5.0.11-1 · BSD-3-Clause AND LicenseRef-MPICH
💖 girlified by deepseek-ai/DeepSeek-V4.1-Flash-1017 on 2026-11-06: a language model's rewrite of the original, under its terms