An Experimental Study of an Ant-colony Algorithm for the Mesh-partitioning Problem

Abstract

Parallel computer systems are suitable for solving computationally intensive numerical methods such as finite-volume and finite-element methods for solving partial di.erential equations. A program and associated data are distributed among the processors. The aim is to find a distribution where the communication is as low as possible. This leads to the mesh-partitioning optimization problem, which is known to be an NP-hard. Ant-colony algorithm (ACA) is a relatively new metaheuristic search technique for solving general optimization problems. We experimentally study ACA for solving the mesh-partitioning problem.