Development of a highly parallelized micromagnetic simulator on graphics processors
This work introduces the development of a micromagnetic simulator running on highly parallel Graphics Processing Units (CPU's). We show how main components of the calculation (ODE solver and magnetic potential calculation) can be implemented to provide optimal speedup for various problem sizes. For smaller problems, direct potential calculation is the most efficient method, while large problems are best handled by a fast multipole algorithm.
Book / Congress title:
Computational Electronics (IWCE), 2010 14th International Workshop on