Chapter 17: Using the Solver in Parallel

The parallel implementation of the CFX-Solver is based on the Single-Program-Multiple-Data (SPMD) model. This model runs identical versions of the code on one or more processors.

The overall parallel run procedure is divided into two steps:

  1. A partitioning step, where the mesh is divided into a number of different segments, or partitions.

  2. A running step, where the mesh partitions are solved using separate processes (a leader and one or more follower processes), often on different machines, with each process working on its own partition.

The CFX-Solver is designed so that all of the numerically intensive tasks can be performed in parallel. Administrative tasks, such as simulation control and user interaction, as well as the input/output phases of a parallel run, are performed serially by the leader process. This approach guarantees good parallel performance and scalability of the parallel code, as well as ensuring that the input/output files are like those of a sequential run.

Communication between processes during a parallel run is performed using the MPI message-passing libraries. Platform-specific versions of MPI are available. In all cases, the processes of a parallel run are distributed among the processors in the specified pool of hosts.

The following limitation exists for combining hardware in a parallel run:

  • Intel MPI can only be used on a homogeneous network of the supported platforms.

This chapter describes the ideas behind parallel processing in Ansys CFX and provides some advice on using it effectively:

More detailed instructions on how to set up a parallel run are available in Parallel Run in the CFX-Solver Manager User's Guide.