Parallel Execution

MPI and ibrun: Parallel Execution on HPC Systems

OpenSees supports parallel execution through MPI (Message Passing Interface), a widely used protocol that allows multiple processes to work together across processors or nodes in a high-performance computing (HPC) environment. MPI enables large structural simulations to be split into smaller tasks and solved concurrently, greatly improving performance and scalability.

• On TACC systems like Stampede3, MPI programs must be launched using ibrun—a system-specific command that ensures MPI processes are correctly assigned to the compute resources allocated by SLURM.

• While mpiexec is suitable for interactive environments like JupyterHub, ibrun is required in batch jobs on Stampede3 to ensure proper task placement, resource binding, and compatibility with the system’s MPI stack.