3.5.4 Parallel execution in Abaqus/CFD
Products: Abaqus/CFD Abaqus/CAE
References
Overview
Parallel execution in Abaqus/CFD:
reduces run time for analyses that require a large number of increments;
reduces run time for analyses that contain a large number of nodes and elements;
produces analysis results that are independent of the number of processors used for the analysis; and
is available for both shared memory computers and computer clusters using an MPI-based domain decomposition parallel implementation.
Invoking parallel processing
Abaqus/CFD uses domain-based parallelism implemented with explicit message passing for both shared memory and distributed memory computers. All procedures provided by Abaqus/CFD and their associated features are fully parallel (“Parallel execution: overview,” Section 3.5.1). Parallel execution is invoked by specifying the number of processors to be used.
| Input File Usage: | Enter the following input on the command line: |
abaqus job=job-name cpus=n For example, the following input will run the job “manifold” on two processors: abaqus job=manifold cpus=2 |
| Abaqus/CAE Usage: | Job module: job editor: Parallelization: toggle on Use multiple processors, and specify the number of processors, n |
Domain-based parallelism
Abaqus/CFD uses a domain-decomposition message passing paradigm for its parallel implementation. An element-based decomposition strategy is used that minimizes the number of communications required between subdomains while providing a nearly uniform computational work distribution among the processors. The number of domains maps exactly to the number of user-specified processors for a given calculation. The load-balancing procedures are implemented in parallel as well, so that you can avoid time consuming serial load-balancing procedures at the start of a calculation. Every attempt has been made to ensure that Abaqus/CFD provides scalable parallel solutions for a broad range of applications. All procedures and features in Abaqus/CFD are provided with a fully parallel implementation. All output is serialized automatically for the user so that there is no translation between parallel domains and the original user input. In addition, this permits Abaqus/CFD to restart seamlessly on any number of processors, regardless of how many were used for the original computation.
Co-simulation
The co-simulation technique (“Co-simulation: overview,” Section 17.1.1) for run-time coupling of Abaqus/CFD to Abaqus/Standard or to Abaqus/Explicit can be used with Abaqus/CFD running either in serial or parallel.