We have developed a PC-based parallel tree code, powerful and easy to use. It requires only the message passing library PVM installed, and a ANSI C compiler. The structure of the code was kept as simple as possible. Its power lies not only in its easy implementation and modularity --which allows the incorporation of, e.g., SPH-like hydrodynamics, but, more important, in that it was designed to work fine under hard conditions, i.e., on non-dedicated machines and through non-dedicated nets.
Thus, our program is able to determine the number of particles each processor should integrate based on its actual load. The differing loads may be due either to the features of the system being integrated, or, in the case of non-dedicated processors, to the sharing of CPU time with other processes. Thus, this optimization is of utmost importance for those places which cannot afford dedicated machines.
The program was tested by performing a number of standard experiments and simulations of collisions of galaxies, yielding satisfactory results in all cases. The complete ANSI C code, as well as an animation of the Antennæ experiment, is freely available via anonymous ftp to ftp.fcaglp.unlp.edu.ar in the subdirectory /pub/hviturro/poctgrav.
Acknowledgements
We would like to thank Drs. F.C. Wachlin and J.C. Muzzio for their useful advices.
Copyright The European Southern Observatory (ESO)