Exploiting FPGAs and GPUs for Electromagnetics Applications: Interferometric Imaging in Random Media Case Study

摘要

There is a growing need for reliable and efficient numerical methods for electromagnetic applications. This is important for addressing the complex designs with fine features on electrically large platforms. As designs become more complex, a good prediction of overall system performance becomes essential for cost reduction especially in the conceptualization stage. Researchers have attempted to address this issue by developing hybrid methods based on asymptotic techniques that can avoid the numerical inefficiency while maintaining high degrees of accuracy. Another approach is to implement fast computational methods that utilize parallel computing platforms. This paper focuses on the latter; i.e. by investigating the use of field programmable gate arrays (FPGA) and general purpose graphics processing units (GPGPU) as coprocessors to parallelize numerically challenging problems. The weaknesses and strengths of both platforms will be investigated in the context of their ease of use, efficiency, and potential for accelerated computations.