MPI: The Complete Reference
Introduction to Algorithms
Hi-index | 0.00 |
This paper presents a methodology for calculating silicon nanowire (SiNW) bandstructures on parallel machines. A partition scheme is developed through domain decomposition and loading balance is considered for scheduling jobs on machines with different efficiencies. Using sp3d5s* tight-binding model, the Hamiltonian matrix of the SiNW is constructed and its eigenvalues are extracted with the Implicitly Restarted Arnoldi Method. Parallel calculation performance is tested on identical machines finally and a linear speedup is gained as the number of nodes increases.