Performance analysis and optimization of a paralell numerical code for micromagnetic calculus

Numerical simulations are used for solving significant scientific and engineering problems. High performance numerical simulations can be divided in two categories: parallel and distributed simulation. In this paper we make a statistical study of performances in parallel simulation using Message Passing Interface (MPI) and Magpar software for micromagnetic problems. The results of total time, initialization time, resolving time, memory and bandwidth used depending on the number of processors are described.