An Optimized Dynamic Load Balancing Method for Parallel 3-D Mesh Refinement 
for Finite Element Electromagnetics with Tetrahedra

DAQI REN
(University of Tokyo)
Dennis Giannacopoulos
(McGill University)
Reiji Suda
(University of Tokyo)

We develop a promoted Dynamic Load Balancing (DLB) algorithm in parallel, dynamic, 
3-D mesh refinement with tetrahedra based on the characteristics of FEM mesh 
requirement. On the top of a traditional Random Polling DLB methodology, the prototype 
of the new DLB algorithm is illustrated and implemented with MPI for hierarchical 
tetrahedra and octahedra (HTO) meshing on a target multiprocessor architecture.  
Benchmarks for performance measures are derived.  By comparing with the performance 
results of other two existing approaches on the same HTO example geometric mesh 
refinement model, the benefits of our new design for achieving high performance 
parallel mesh refinement are demonstrated.