Modern embedded systems are becoming more and more complex thus demanding for new powerful development mechanisms. Model-driven engineering has been recognised as a promising paradigm for the development of complex systems especially for its capability of abstracting the problem through models and then manipulating them to automatically generate target code. In our previous works, we presented mechanisms for the generation of 100% of the target code from UML models to be run on singlecore platforms. In this work we provide possible solutions to enhance the generation process to entail a more complex set of platform configurations (i.e., multiprocess, multicore) as well as heterogeneous processing units (i.e., CPU, GPU).