by Rio YokotaKing Abdullah University of Science and Technology (formerly at Boston University)
Fast multipole methods (FMM) were originally developed for accelerating N-body problems in astrophysics and other particle based methods. A recent trend in HPC has been to use FMM in unconventional application areas. For example, the 2010 ACM Gordon Bell prize peak-performance paper applied the FMM to the deformation of red blood cells, while the 2009 price/performance paper used it for a turbulence simulation. The fact that the FMM is O(N), compute bound, and requires very little synchronization, makes it a favorable algorithm for next-generation architectures.
At the same time it has a wide range of applications and plenty of room for mathematical intervention, which makes it an interesting algorithm to study.
Rio Yokota obtained his PhD in Mechanical Engineering from Keio University, Japan, in 2009, and was a postdoctoral researcher at the Department of Mathematics at Univeristy of Bristol from 2009-2010, and also at Mechanical Engineering Department at Boston University from 2010-2011. During his PhD, he worked on the implementation of fast multipole methods on special purpose machines such as MDGRAPE-3, and then on GPUs after CUDA was released. During his post-doc he has continued to work on fast multipole methods, and has recently developed a massively parallel auto-tuning FMM library, ExaFMM. His applications of interest range from turbulence to molecular dynamics. You can contact him at email@example.com.