Hierarchical symbolic circuit analysis of large-scale networks on multi-processor systems

A parallel implementation of a hierarchical symbolic analysis algorithm (SCAPP) on a 128-node nCUBE hypercube class processor is presented. The basic operation involves exploiting the inherent parallelism in the symbolic analysis algorithm on two levels. The first level is scheduling the partitions of a circuit onto a subset of available processors. The second level is dividing and scheduling, for each partition, the tasks of reducing the sparse symbolic modified nodal analysis matrix using the rest of the available processors. The algorithm does not impose any constraints on the number of processors required and uses minimal data movement. This results in low communication overhead between the processors. The algorithm is applicable to any general message-passing multiprocessor architecture. A speedup on the order p (number of processors) is achieved