Abstract
A parallel dynamic load-balancing algorithm (LB) is proposed to deal with the load-imbalancing problem of a solution-adaptive finite element program on a 2D torus. The algorithm uses an iterative approach to achieve load-balancing. This algorithm is implemented along with two parallel mapping algorithms, parallel orthogonal recursive bisection (ORB) and parallel recursive mincut bipartitioning (MC), on a simulated torus. Three criteria, the execution time of load-balancing algorithm, the computation time of an application program under different load balancing algorithm, and the total execution time of an application program are used for performance evaluation. Simulation results show that 1) the execution of the LB is faster than those of the MC and ORB; 2) the mappings of the LB are better than those of the ORB and MC; and 3) the speedups of the LB are better than those of the ERB and MC.