Abstract

We present an automatic, spatially local data distribution and load balancing schemeapplicable to many-body problems running on parallel architectures. The particledistribution is based on spatial decomposition of the simulation cell. A one-dimensionalHilbert curve is mapped onto the three-dimensional real space cell, which reduces thedimensionality of the problem and provides a way to assign different spatially local parts ofthe cell to each processor. The scheme is independent of the number of processors. Itcan be used for both ordered and disordered structures and does not dependon the dimensionality or shape of the system. Details of implementation in thelinear-scaling density functional code CONQUEST, as well as several case studiesof systems of various complexity, containing up to 55 755 particles, are given.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call