The traditional MoM formulation requires the solution of a linear
system of equations. Depending on the problem size, the memory for the
system matrix can easily exceed the available main memory on a computer
system. One option would then be to run a parallel version of
FEKO and combine the memory of different PCs or workstations, but FEKO
also includes a highly efficient out-of-core solver.
When not enough main memory for the matrix is found, FEKO switches automatically to this out-of-core solver, writes the MoM matrix to a file and performs the LU decomposition of the matrix (required for the solution) using a blocked algorithm where only small blocks are required which are read individually from the out-of-core file. Due to the blocked structure and by using low-level POSIX I/O routines, this algorithm is almost comparable in speed to an incore solution. Performance depends on the hard disk transfer rates, but typically run-times of the out-of-core solver are about 15% to 25% longer as compared to an incore solution.
This out-of-core solver is available in both sequential and parallel versions of FEKO.