HyperFPGA: A Scalable Reconfigurable Platform for Computing Paradigms Research

In recent years, the limitations of von Neumann-based supercomputers in terms of energy efficiency and performance scalability have stressed the need for alternative computing paradigms. This paper introduces the HyperFPGA, an open SoC-FPGA cluster designed for experimental research on novel supercomputing architectures and paradigms. Unlike existing platforms, the HyperFPGA offers flexibility across a wide range of configurations, from electrical standards for data transmission to distributed computing models. Additionally, it is equipped with multiple power consumption monitors to independently measure different computational and hardware aspects, separating data movement from computation. The paper details the design and implementation of the HyperFPGA and demonstrates its basic functionality and scalability through a custom computational solution for the non-attacking n-queens problem, a well-known distributed computing problem. These preliminary tests demonstrate the ability of the HyperFPGA to efficiently handle computationally intensive and distributed tasks, showcasing its potential for fine-grained reconfigurable supercomputing solutions addressing a broad spectrum of scientific and engineering problems. The platform is available for open collaborative projects focused on energy-efficient supercomputing and computational physics.