Last generation computing architectures have evolved from traditional standalone Embedded Systems to become complex environments where computational elements tightly interact with physical entities such as sensors networks and I/O devices. These systems, usually referred as Cyber-physical Systems (CPS), enabled a flourishing ecosystem of architectures and platforms where smart objects, users and communication infrastructures interact to support intelligent context-aware services and applications. Smart grids, medical monitoring, smart cities, distributed pollution and tracking are just a few examples of concrete applications that are gaining attraction among industries and institutions.

However, the mobility and pervasivity requirements of such environments impose energy consumption constraints that must be met in a context of increasing computational needs, due the processing of large amount of data coming from sensing and input devices. The conventional approach of providing such computational resources by means of cloud computing is becoming the limiting factor in the design of the future CPS, since the increased communication effort required to perform the data off-loading to external resources represents the major contribution to the overall energy consumption of the smart device. Due to the power hungry nature of the communication infrastructure, it can be envisaged a trend in which the smartness of the "things" will be even more shifted toward the things themselves rather than toward the cloud. Based on this, improving the computational capabilities of the smart objects in a even more limited energy envelope, becomes a key issue.