Simulated annealing: A trade-off between vector and parallel implementations