Abstract: For accelerating the convergence of numerical computation of sparse matrices, the classical approach is to partition sparse matrices into block-diagonal structures by graph partitioning ...