P03: BEM4I: A Massively Parallel Boundary Element Solver
SessionPoster Reception
Event Type
ACM Student Research Competition
Poster
Reception
TimeTuesday, November 14th5:15pm -
7pm
LocationFour Seasons Ballroom
DescriptionIn this work we present a library of parallel solvers
based on the boundary element method (BEM). We provide a
brief description of BEM and its parallelization, focus
on SIMD vectorization and shared- and distributed-memory
parallelization by OpenMP and MPI, respectively. Two
approaches for distributed parallelization of BEM are
discussed - the first one based on a novel parallel
adaptive cross approximation (ACA) method, the second
one on the boundary element tearing and interconnecting
(BETI) domain decomposition method. To demonstrate the
efficiency of the library we provide results of
numerical experiments on the Xeon and Xeon Phi based
clusters.
