Emerging Applications on Reconfigurable Computing Platforms

Security and high-performance computing have become two of the most critical demands in modern technology. The increasing complexity of digital systems, the need for real-time processing, and the emergence of sophisticated cyber threats require computing solutions that balance computational power with adaptability and security. Reconfigurable computing platforms, particularly Field-Programmable Gate Arrays (FPGAs), offer a promising solution to these challenges by combining flexibility with hardware acceleration. Many new applications have emerged with the developing of reconfigurable platforms. FPGAs can be utilized in two primary directions: as control and high-precision measurement units for security-sensitive applications, and as accelerators capable of outperforming GPUs in specific computational workloads. This dissertation explores both of these directions by developing FPGA-based solutions for time-domain reflectometry (TDR), finite-difference time-domain (FDTD) simulations, SPICE-based power electronics simulation, and recurrent neural network (RNN) training. The TDR implementation provides a high-precision diagnostic tool that has been further adapted to enhance hardware security by detecting unauthorized probes in communication lines. The FPGA-based FDTD accelerator enables efficient simulations for designing ultra-compact photonic devices, demonstrating the advantages of hardware acceleration in scientific computing. Additionally, a SPICE accelerator has been developed to optimize the design process of LLC converters in power electronics, reducing simulation time while maintaining accuracy. Finally, an FPGA-based RNN training accelerator is introduced to address the challenge of real-time learning, showcasing how FPGAs can be leveraged for emerging AI applications.