Hints, Tips, and Solutions

Q: How do I run a simulation on a cluster of computers using the distributed computing feature?

A new feature that is introduced in Silvaco’s TCAD applications allows the user to run a parallel simulation on a cluster of computers within a network.

Currently this feature is supported in the solution of linear systems using the PAM solver. The PAM solver is a domain decomposition type solver which is parallelized with MPI – message passing interface.

There are many advantages to using distributed computing especially for large computationally intensive simulations:

• Splitting a large problem into smaller ones and computing them in parallel is clearly a superior approach with regards to performance. Simulation time can decrease by orders of magnitude.
• More resources are available to the simulation – the user is not limited by the number of CPUs or the total amount of memory on a specific computer.
• This approach provides a way of using the resources within the network efficiently.
• For certain very large problems it might be the only feasible approach, especially in cases when the amount of memory required exceeds the limits of any one available computer. As problem size increases this becomes more and more relevant.