Parallelization and Performance Evaluation of Method of Moment (MoM) Dense Matrix Direct Solution on Cray XC40 -Supercomputer (SAHASRAT)

The objective of this paper is to evaluate and analyze the performance of a suite of in-house developed parallel Integral Equation solvers using Method of Moments on a massively parallel-processing supercomputer system Cray XC40 named as SAHASRAT. The developed solvers can be used to accurately analyze composite structures, Radar Cross Section computation and antenna parameters for electrically large radar targets. With this methodology, large problems that were impossible to solve on a single-processor platform can be solved in a reasonable time on shared and distributed memory machines consisting of several thousands of cores. The size of the problem that can be extended to be as large as the amount of main memory on the available platform. In order to fill and perform dense linear algebra computations on the matrix, Message Passing Interface (MPI) with the state-of the art numerical library ScaLAPACK is used.