Skip to content

Application Guide: ANSYS

This page contains information on how to run a variety of ANSYS batch jobs on BlueBEAR.

See the ANSYS page on the BEAR Apps website for information on available versions.

Parallel Wrapper script

We provide a helper script to generate the node-file for parallel ANSYS jobs:

module load slurm-helpers
NODEFILE=$(make_nodelist)

The resulting ${NODEFILE} environment variable is then passed as follows:

  • CFX: -par-dist "${NODEFILE}"
  • Fluent: -cnf=${NODEFILE}

Example sbatch scripts

Please use the tabs below to see example batch scripts for ANSYS commands. For further general information on running BlueBEAR jobs, please see Jobs on BlueBEAR.

This example runs cfx5solve on the ANSYS-provided example file Benchmark.def. Please adjust the paths accordingly to use your own input files.

#!/bin/bash

#SBATCH --ntasks=10
#SBATCH --ntasks-per-node=5  # this forces the job to distribute evenly across two nodes
#SBATCH --account=_ACCOUNT_NAME_
#SBATCH --qos=bbdefault
#SBATCH --time=10

set -e

module purge
module load bluebear
module load slurm-helpers

module load bear-apps/2022a
module load ANSYS/2023R1

# Copy the example file into the working directory
cp "${EBROOTANSYS}/v2*/CFX/examples/Benchmark.def" .

# Assign the list of the job's nodes to an environment variable
NODELIST=$(make_nodelist)

cfx5solve -def Benchmark.def -parallel -par-dist "${NODELIST}" -start-method "Open MPI Distributed Parallel"

cfx5solve options explained

  • -def: Use specified file as the Solver Input File
  • -parallel: Run the ANSYS CFX Solver in parallel mode
  • -par-dist: Provide a comma-separated list of a job’s nodes, e.g. the one created by the make_nodelist script.
  • -start-method: Use the named start method (wrapper) to start the ANSYS CFX Solver. See the ${EBROOTANSYS}/v2*/CFX/etc/start-methods.ccl file for the possible methods.

Tip

For further information on the cfx5solve command, please run:

cfx5solve -help

N.B. Requires an input journal file and any related case or mesh files to be provided. In this example we are using aircraft_wing_2m.jou but please adjust the paths accordingly to use your own input files.

#!/bin/bash

#SBATCH --ntasks=10
#SBATCH --ntasks-per-node=5  # this forces the job to distribute evenly across two nodes
#SBATCH --account=_ACCOUNT_NAME_
#SBATCH --qos=bbdefault
#SBATCH --time=10

set -e

module purge
module load bluebear
module load slurm-helpers

module load bear-apps/2022a
module load ANSYS/2023R1

# Copy example input files
cp "${BB_APPS_DATA}/ANSYS/examples/fluent/"* .
JOURNAL_FILE="./aircraft_wing_2m.jou"

# Create the nodefile
NODEFILE=$(make_nodefile)

fluent 3ddp -t${SLURM_NTASKS} -mpi=intel -pib.ofed -cnf=${NODEFILE} -g -i "${JOURNAL_FILE}"

fluent options explained

  • -t: specify number of processors. Use the Slurm-provided ${SLURM_NTASKS} environment variable here.
  • -mpi: specify MPI implementation
  • -p: specify the MPI interconnect
  • -cnf: specify the hosts file
  • -g: run without GUI or graphics
  • -i: read the specified journal file

Tips