Study preparation

The preparation phase is similar to the standard usage of Simcenter Amesim, up to the point when the simulation needs to be started. The model and its simulation settings must be saved in a pack file (.pck), to ensure a safe transfer to the HPC platform. The Pack and Unpack facility is used for this purpose. This facility makes it possible to pack models with all their dependencies (user libraries, tables and any other needed data files or resources), which ensures that models are ready to run after their transfer to another machine. You can take a look at the Pack and Unpack facility section in the Simcenter Amesim Reference manual for more details. This feature guarantees the ability of models to be compiled and the integrity of their simulations on any remote machine, especially on your HPC platform. During the packing process, user libraries and user data are embedded.

Note

Using standard (i.e. not packed .ame model files) is also possible but requires special care, as the presence of all necessary files will not be automatically checked.

Model preparation overview

Procedure

1. Build the sketch.

2. Prepare the simulation (batch, DoE, single run, ...).

3. When the simulation is ready ... pack the model.

Transfer of the model to the head node machine

The input file (.ame or .pck) must be transferred to the head node or at least placed somewhere accessible for the head node.

Contact your IT service for more information on how to transfer files to your HPC platforms.

Filling in the submission template

Simcenter Amesim currently provides generic templates for the supported schedulers listed below:

templates

• Slurm template ($AME/scripting/python/hpc/templates/SLURM.template.sh)

• LSF/OpenLava template ($AME/scripting/python/hpc/templates/LSF.template.sh)

• PBS template ($AME/scripting/python/hpc/templates/PBS.template.sh)

• UGE template ($AME/scripting/python/hpc/templates/UGE.template.sh)

• SGE template ($AME/scripting/python/hpc/templates/SGE.template.sh)

Results

These templates are unified self-executable files that perform job submissions based on the description of an Simcenter Amesim study and of the associated resource requirements. You should first create a local copy of your selected scheduler template. Replace the TO_BE_DEFINED fields with simulation and model settings, and define the scheduler requirements in the AME_SCHED_ARGS environment variable.

Note

UGE and SGE templates use the same driver: UGE, the only difference in the template being the content of the AME_SCHED_ARGS environment variable.

The preparation of the job consists in describing the Simcenter Amesim study in the script.

Filling the template with correct settings might be a matter of minutes.

Details of the settings to specify in the submission scripts will be given in the following sections. Hereafter is a typical example of such a submission script:

Copy

...
    # Define the input model directory.
    DIRMODEL=$HOME/hpcdir/inputs
    # Define the model name.
    MODEL="DesignOptimization.ame"
    # Define the study type: single run, batch run and Design Exploration.
    # Use code: "single", "batch" and "DE"
    SIMTYPE="batch"
    ...