SimRunner enables users to execute simulations in parallel on one or more computers connected in a network. It accepts simulations from HyperLith, EM-Suite, and SOAPI, and dispatches them to a group of worker PSS (Panoramic Simulation Server) and HSS (Hardware Simulation Server) processes for execution.

SimRunner supports the following:

  • Hardware accelerated simulation: Hardware accelerated simulation uses graphics hardware to speed up TEMPESTpr2 and resist simulations. Depending on the simulation and the hardware, simulation times can be over 10x faster with hardware acceleration. Hardware accelerated simulation can only run with SimRunner controlling HSS worker processes.
  • Multiple users: SimRunner can be used to allow multiple users to share simulation computing resources.
  • Job control: Simulations are assigned a priority and are processed in priority order. SimRunner allows the user to reorder the execution of submitted simulations and cancel simulations.
  • Hetrogeneous clusters: The worker processes may run on machines with different processing configurations and operating systems. Windows and Linux 32 and 64 bit OSs are supported.
  • Cluster support: Worker processes may be launched on demand when a simulation is submitted which facilities integration with cluster managerment software.

Parallelization and Performance

Simulation created in HyperLith and EM-Suite can be parameterized with variables that take on multiple values. When this occurs, a separate simulation run is required for each case. When run inside of HyperLith, EM-Suite, or SOAPI, each simulation run is executed sequentially. When dispatched to SimRunner, multiple runs can be executed in parallel - one run per a worker process.

Depending on the complexity of each run, the total number of runs, and the number and types of worker processes, turn around time for simulations can be shortened dramatically. With hardware acceleration, performance is even more dramatic as shown in these benchmarks.

Distributed Computing for Large TEMPESTpr2 Simulations

With the shorter wavelengths and smaller feature sizes used in advanced lithography, TEMPESTpr2 simulations for 2D patterns can become extremely large. So large that the simulation domain may be too big to fit on a typical computer. This is especially true for computers still running 32 bit operation systems.

SimRunner allows the worker processes to be configured so that the single TEMPESTpr2 run is partitioned and run over the multiple machines. Thus if you have 28GB simulation, you could configure 4 8GB machines to run the simulation with each node working on one quarter of the domain. 

SimRunner Screen Shots