We present an optimized approach for the calculation of the density of fully coupled vibrational states in high dimensional systems. This task is of paramount importance, because partition functions and several thermodynamic properties can be accurately estimated once the density of states is known. A new code, called paradensum, based on the implementation of the Wang-Landau Monte Carlo algorithm for parallel architectures is described and applied to real complex systems. We test the accuracy of paradensum on several molecular systems, including some benchmarks for which an exact evaluation of the vibrational density of states is doable by direct counting. In addition, we find a significant computational speedup with respect to standard approaches when applying our code to molecules up to 66 degrees of freedom. The new code can easily handle 150 degrees of freedom. These features make paradensum a very promising tool for future calculations of thermodynamic properties and thermal rate constants of complex systems.