This page describes different resource configurations for running Lumerical FDTD simulations for different use cases.
Default resource configuration
The default configuration is appropriate for running simulations on the local computer. The simulation job is run using multiple MPI processes with each process utilizing 1 core. It will run one simulation at a time using all logical cores on the machine.
Resource configuration without using MPI
It is possible to run a simulation job without using MPI. This is occasionally done when there are issues installing or running with MPI. Simulation speed should be very similar to the default configuration.
To run without MPI, 'Edit' the default resource line and select 'Local Computer' as the 'Job launching preset'. Set 'Threads' to the number of cores on your computer.
Configuration for parameter sweep and optimization
Parameter sweep or optimization jobs can run (1) sequentially on a local workstation or (2) concurrently on a single powerful machine or (3) concurrently across several machines. The latter option can be done with or without a job scheduler and using local workstations or cloud resources.
- The default resource configuration runs sweep/optimization jobs sequentially on a local machine.
- Depending on your simulation and machine's resources, you can run sweeps concurrently. Set 'threads' to 1, enter the number of 'processes' to use per-sweep and 'capacity' to the number of sweeps running concurrently. Ensure that:
threads * processes * capacity <= enabled CPU cores
- Try to test with different (processes * capacity) combination for optimum performance.
TIP: See concurrent computing for details on running simulations concurrently on multiple machines or remote cluster.
Distributed resource configuration
Simulations that require large amounts of memory can be distributed across several nodes of a cluster or cloud platform. See distributed computing for details.
Cluster job scheduler resource configuration integration
Running simulation from the design environment to a cluster through a job scheduler can be done from the resource configuration utility. See our job scheduler integration page for details.
Calculate solver license usage
The number of solver/accelerator/HPC licenses required to run your jobs depends on the number of cores used, the number of concurrent simulations, the type of job, and the type of license you have purchased. See the Understanding solver, accelerator, and HPC license consumption page for details.