Programming techniques for efficient and interoperable software defined radios

Recently, Software-Dened Radios (SDRs) has became a hot research topic in wireless communications eld. This is jointly due to the increasing request of reconfigurable and interoperable multi-standard radio systems able to learn from their surrounding environment and efficiently exploit the available frequency spectrum resources, so realizing the cognitive radio paradigm, and to the availability of reprogrammable hardware architectures providing the computing power necessary to meet the tight real-time constraints typical of the state-of-art wideband communications standards. Most SDR implementations are based on mixed architectures in which Field Programmable Gate Arrays (FPGA), Digital Signal Processors (DSP) and General Purpose Processors (GPP) coexist. GPP-based solutions, even if providing the highest level of flexibility, are typically avoided because of their computational inefficiency and power consumption. Starting from these assumptions, this thesis tries to jointly face two of the main important issues in GPP-based SDR systems: the computational efficiency and the interoperability capacity. In the first part, this thesis presents the potential of a novel programming technique, named Memory Acceleration (MA), in which the memory resources typical of GPP-based systems are used to assist central processor in executing real-time signal processing operations. This technique, belonging to the classical computer-science optimization techniques known as Space-Time trade-offs, defines novel algorithmic methods to assist developers in designing their software-defined signal processing algorithms. In order to show its applicability some "real-world" case studies are presented together with the acceleration factor obtained. In the second part of the thesis, the interoperability issue in SDR systems is also considered. Existing software architectures, like the Software Communications Architecture (SCA), abstract the hardware/software components of a radio communications chain using a middleware like CORBA for providing full portability and interoperability to the implemented chain, called waveform in the SCA parlance. This feature is paid in terms of computational overhead introduced by the software communications middleware and this is one of the reasons why GPP-based architecture are generally discarded also for the implementation of narrow-band SCA-compliant communications standards. In this thesis we briefly analyse SCA architecture and an open-source SCA-compliant framework, ie. OSSIE, and provide guidelines to enable component-based multithreading programming and CPU affinity in that framework. We also detail the implementation of a real-time SCA-compliant waveform developed inside this modified framework, i.e. the VHF analogue aeronautical communications transceiver. Finally, we provide the proof of how it is possible to implement an efficient and interoperable real-time wideband SCA-compliant waveform, i.e. the AeroMACS waveform, on a GPP-based architecture by merging the acceleration factor provided by MA technique and the interoperability feature ensured by SCA architecture.