High Performance Computing On Boundary Element Simulations
Free (open access)
J.M. Cela & A. Julia
In this paper we describe the parallelisation of a boundary element package. We exploit different levels of parallelism using threads and processes. Moreover, the threads are used to overlap communications and I/O operations with computations. 1 Introduction The boundary element method (BEM) has several advantages over the classical finite element method in some specific simulations, like crack grow analysis. However, BEM has an important drawback, the linear systems generated by this method are dense. Then, the computational complexity grows with N* and the storage requirements with A^, where TV is the amount of boundary elements. Although, some tricks can be used to mitigate this problem, the complexity of the BEM limits seriously its use. Then, high performance computing is specially critical in BEM simulations.