Abstract
We generalize the microcanonical algorithm developed by Creutz et al. (1986), and make a detailed comparison with the exact solution in the case of a two-dimensional Ising model at finite volume. We present a new numerical method to compute the temperature in the microcanonical ensemble. This allows us to define a `thermalization` criterion to estimate the point where the differences between canonical and microcanonical results are the smallest. This criterion is shown to work well in the case of the two-dimensional Ising system.