Issues in parallelizing multigrid-based substrate model extraction and analysis

Abstract

Accurate modeling of coupling effects via the substrate is an increasingly important concern in the design of mixed-signal systems such as communication, biomedical and analog signal processing circuits. Fast-switching digital blocks inject noise into the common substrate hindering the performance of high-precision sensible analog circuitry. Miniaturization effects on ICs complexity inevitably make the accuracy requirements for substrate coupling simulation increase. Due in part to the global nature of such couplings, model extraction and analysis is a computation-intensive task requiring the availability of fast and accurate substrate model extraction and analysis tools. One way to deal with this problem is to take further advantage of available computational technologies and distributed computing emerges as an interesting solution. In this paper we discuss several issues related to the parallelization of a multigrid-based substrate model extraction and analysis tool. This tool is used as a proxy for generic computations on a 3D discretized volume. The results presented indicate potential avenues for successfully exploiting parallelism as well as pitfalls to avoid in such a quest.