high performance computing on graphics processing units: hgpu.org

Due to the variety and importance of applications of treecodes and FMM, the combination of algorithmic acceleration with hardware acceleration can have tremendous impact. Alas, programming these algorithms efficiently is no piece of cake. In this contribution, we aim to present GPU kernels for treecode and FMM in, as much as possible, an uncomplicated, accessible way. The interested reader should consult some of the copious literature on the subject for a deeper understanding of the algorithms themselves. Here, we will offer the briefest of summaries. We will focus our attention on achieving a GPU implementation that is efficient in its utilization of the architecture, but without applying the most advanced techniques known in the field (which would complicate the presentation).