Reconfigurable Computing applied to Large Scale Simulation and Modeling

Abstract

Over the past decade, reconfigurable computing has found its way as a 
powerful paradigm for accelerating complex computational tasks in diverse 
domains. Due to growing simulation and modeling loads in both scale and 
complexity, reconfigurable architectures must be accordingly adapted to 
meet the demand for efficient execution of these workloads. In this work, 
we elaborate on the challenges and potentials of employing reconfigurable 
computing (RC) for large scale simulation and modeling, using key technolo
gy advances, architectural questions as well as prominent application areas 
to make our point. Reconfigurable hardware (e.g. field programmable gate 
array), in conjunction with high performance computing (HPC) infrastruc
tures, has enabled new approaches to approximately solve computationally 
intensive simulations and models. Researchers and engineers can potential
ly obtain dramatic speedups and energy efficiency as reconfigurable plat
forms offer the flexibility and parallelism necessary to leverage their power. 
Yet restructuring of reconfigurable computing for large scale simulation and 
modeling is not free of difficulties. In order to fully realize the potential 
of this approach there are issues to be addressed including programming 
complexity, scalability, and integration with existing software ecosystems. In 
this article we explore such challenges towards reconfigurable computing for 
simulation and modeling applications and explore brewing in what’s possible 
with reconfigurable computing.