FASBM: FPGA-specific Approximate Sum-based Booth multipliers for energy efficient Hardware Acceleration of Image Processing and Machine Learning Applications

In order to boost the performance of FPGA based resource-constrained embedded Image processing and Machine Learning applications, energy efficient approximate softcore Booth multipliers are designed using Adapative Logic modules (ALM) on Intel FPGAs. Two partial products of a multiplier partial product matrix are merged together by addition followed by a Karnaugh map reduction in a unique way. The two merged partial products are then mapped on a single six input ALM, in order to reduce the total resource utilization of the circuit.