Exploring heterogeneous computing platforms for molecular dynamics-based calculations

Abstract

Molecular dynamics (MD) simulations demand high computational power to model atomic interactions over time, often limiting their scalability on conventional processors. This thesis explores the role of FPGA-based heterogeneous computing in accelerating MD calculations by offloading key computational tasks such as force computation and integration. Through custom hardware design using high-level synthesis, the study demonstrates how FPGAs can offer significant performance gains with improved energy efficiency compared to CPUs and even GPUs in specific scenarios. By placing FPGAs at the heart of the MD simulation workflow, this work highlights their potential to drive scalable and efficient molecular simulations, paving the way for broader adoption in computational science.