Sponsor: National Science Foundation
The purpose of this project is to develop innovative computational methods and software for problems in electromagnetics arising in VLSI design. It focuses on developing robust, effective, and parallelizable solvers and preconditioners for linear systems arising from finite element and boundary element modeling of inductive and capacitive effects in high speed integrated circuits. The project investigates techniques based on a hierarchy of representations using block low-rank approximations of the inverse. It also studies the implications of these techniques for compression schemes used in approximate dense solvers, and multilevel schemes for sparse solvers. Portable parallel formulations of the techniques along with theoretical estimates for convergence and parallel performance are being developed.