The functionality in
bempp.core is almost exclusively for operator assembly and the multiplication
of discrete operators and vectors, as these is the most computationally-heavy components of BEM.
For larger problems, however, the use of dense matrices is expensive, both in terms of computation time and storage space. For such problems, Bempp can use the fast multipole method to speed up matrix assembly and the computation of matrix-vector products. This is done via interfaces to the external ExaFMM library.