Abstract

Creation of Software-Defined Radio (SDR) solutions is fraught with challenges. In the design phase, communication algorithms must be proven and prototyped. In the next phase, implementations are validated on specific hardware. In the deployment phase, veri- fied radios must reliably inter-operate with other software. We ad- dress the challenges by threading the formal Data-Flow (DF) meta- model throughout this process, supported by high-level languages, tools and diversity mechanisms. The meta-model is stratified: ker- nel algorithms are expressed in plain C and separated from depen- dency graphs expressed in a Domain-Specific Language (DSL). The separation is transparent: kernels are used “as-is” in the fi- nal application. We apply generative techniques to recover elided communication & synchronization code from abstract models and illustrate the benefits of our approach for streaming applications such as the Digital Audio Broadcast (DAB).