Approach: Develop parallel codes from 'scratch' rather than porting existing serial codes so that parallelism inherent in the techniques can be more easily exploited.
Accomplishments: A code capable of simulating a large number of gravitationally interacting particles using a tree algorithm as well as a gaseous component using the SPH algorithm has been developed using the MasPar MP-2 at Goddard Space Flight Center (GSFC). This code reaches performance levels of roughly 1 GFLOPS on 16,384 processors of this machine. This code also has been ported to the Convex Exemplar SPP parallel computer at GSFC. Here, the code exceeds 1 GFLOPS on 15 processors. Message passing versions of the tree algorithm have also been developed but do not yet incorporate the SPH algorithm (CRAY T3D, Intel Paragon, Beowulf, and distributed Silicon Graphics workstations).
The code is written in Fortran 90 and has only been tested on the Maspar MP-2 at Goddard so its performance or correctness on other machines is unknown.
They also have a list of papers available online.