Paper 
Title 
Page 
MOPPC093 
Optimal Fast Multipole Method Data Structures 
352 

 S. Abeyratne, B. Erdelyi
Northern Illinois University, DeKalb, Illinois, USA
 S.L. Manikonda
ANL, Argonne, USA



The Fast Multipole Method (FMM) has been identified as one of the ten most significant numerical algorithms discovered in the 20th century. The FMM guarantees finding fast solutions to many problems in science, such as calculating Coulomb potentials among large number of particles by reducing memory footprint and run time while attaining very high accuracy levels. One important practical issue that we have to solve in implementing a FMM algorithm is organizing large amounts of data, also called data structuring. The nonadaptive FMM is appropriate when the particles are uniformly distributed while the adaptive FMM is most efficient when the distribution is nonuniform. In practice, we typically encounter highly nonuniform 3D particle distributions. This paper summarizes our implementation of a 3D adaptive FMM algorithm data structure setup for nonuniform particle distributions.


