When applying a parallel finite element solution method using non-structured grids and adaptive grid refinement on a multiprocessor based computer the load on individual processors can become unbalanced. In cases like this it is necessary to consider the problem of dynamic load balancing. When applying load balancing methods it is necessary, not only to calculate how many elements need into be moved, but also which elements need to be moved to preserve a minimum boundary between the domains allocated to the processors. The diffusion algorithm is described as one method by which the number of elements to be transferred can be calculated. Two new algorithms are also described. The edge optimisation algorithm can be used, in conjunction with other load balancing methods, to minimise communication costs. The move optimisation algorithm combines a load balancing algorithm with the minimisation of communications.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £66 +P&P)