Modern embedded systems present an ever increasing complexity and model-driven engineering has been shown to be helpful in mitigating it. In our previous works we exploited the power of model-driven engineering to develop a round-trip approach for aiding the evaluation and assessment of extra-functional properties preservation from models to code. In addition, we showed how the round-trip approach could be employed to evaluate different deployment strategies, and the focus was on homogeneous CPUbased platforms. Due to the fact that the assortment of target-platforms in the embedded domain is inevitably shifting to heterogeneous solutions, our goal is to broaden the scope of the round-trip approach towards mixed CPU-GPU configurations. In this work we focus on the modelling of heterogeneous deployment and the enhancement of the current automatic code generator to synthesize code targeting such heterogeneous configurations.