Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!wuarchive!udel!gatech!hubcap!gropp-bill From: gropp-bill@YALE.EDU (Bill Gropp) Newsgroups: comp.parallel Subject: Re: N-body problem on parallel machines Message-ID: <6603@hubcap.clemson.edu> Date: 27 Sep 89 12:50:48 GMT Sender: fpst@hubcap.clemson.edu Lines: 21 Approved: parallel@hubcap.clemson.edu In article <6588@hubcap.clemson.edu> hhd0@gte.com (Horace Dediu) writes: >Could someone mail or post references to work on solving the n-body problem >on parallel machines. I'm especially interested in implementation of >algorithms for hypercubes or the Connection Machine. Here are two that were implemented on an Encore multimax (in a TeX format) \refbody{L.~Greengard and W.~Gropp, {\it A Parallel Version of the Fast Multipole Method,} Technical Report YALE/DCS/RR-640, Yale University, Department of Computer Science, August\ 1988. } \refbody{L.~Greengard and W.~Gropp, {\it A Parallel Implementation of the Fast Multipole Method,} in "Parallel Processing for Scientific Computing," Ed.~G.~Rodrigue. SIAM, Philadelphia, 1989.} A hypercube implementation of these is fairly straightforward. I also have a vector version of the 2-d FMM. A version of a 3-d fast n-body algorithm has been done on the Connection Machine by F. Zhao. Both the 2-d and 3-d n-body algorithms are O((n/p)log n) on a machine with p processors, and are preferable to the obvious O(n*n/p) algorithm. Bill Gropp