Parallel Algorithms for Fast Integral Equation Based Solvers

Abstract

Parallelized electromagnetic solvers are now becoming more prominent as computational architectures, both multicore and distributed clusters, enable multithreaded multi-tasking, and the need for larger capacity solvers is also growing with the exponential growth in complexity of microelectronic systems being analyzed with these solvers. Several techniques are proposed for improved parallel performance of fast integral equation solvers, including a 2-step balancing scheme, and restricted and hierarchical communication.