MPIComm

MPIComm, in the context of the Message Passing Interface (MPI), refers to the concept of a communicator: an abstraction that defines a group of processes that can exchange messages with each other. In formal MPI terminology the handle is called MPI_Comm, but many bindings and libraries expose it as MPIComm. The key purpose of a communicator is to provide a distinct communication context so that messages from different groups do not interfere.

A communicator represents a set of processes and provides two essential properties: a rank for each process within the group and a size indicating how many processes are in the group. The rank is a locally unique identifier used in addressing messages and in performing collective operations. Communicators can be created, split, duplicated, or freed, and they may be intra-communicators (within a single group) or inter-communicators (between two groups).

Common operations involve creating and manipulating communicators, such as MPI_Comm_size and MPI_Comm_rank to query a communicator’s properties, MPI_Comm_dup to duplicate a communicator, MPI_Comm_split to form sub-communicators, and MPI_Comm_free to release resources when a communicator is no longer needed. Inter-communicators enable communication between distinct groups, enabling scalable hybrid algorithms.

All point-to-point and collective communication in MPI is performed within a communicator. For example, MPI_Send and MPI_Recv use a specified MPIComm to route messages to the appropriate processes, while collective operations like MPI_Bcast, MPI_Reduce, and MPI_Allgather operate within a communicator, ensuring coordinated participation of its members.

In practice, MPIComm is a central construct for organizing communication patterns, enabling modular and scalable parallel programs.