Universität Rostock, 2022
Abstract: An Adaptive Voltage Scaling (AVS) system is presented, which is specifically tailored towards the properties of FPGAs. In particular, sensor calibration and placement are the focus of attention. Above, approximate computing is investigated as a possibility to implement applications more energy-efficiently. Therefore, different measures of quality are discussed and an evaluation methodology is developed. An approximate multiplier specifically optimized for modern FPGA architectures is proposed. It is pareto-optimal regarding multiple measures of the quality of the approximation.
doctoral thesis free access