Accelerating Biomolecular Simulations on Reconfigurable Computing Hardware

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Accelerating biomolecular simulations will have a direct impact on investigations in many areas of health related research. Simulations of biomolecules are widely used for fundamental understanding of their structure, folding, dynamics and function. The underlying calculations in these simulations are computationally intensive and have benefited considerably from more than an order of magnitude increase in computer processor speeds in last decade alone. There is widespread interest in alternate hardware and software solutions that can speed-up these simulations, as physical challenges in the computing technology are currently placing the limits on future speed increase of processors. Here we propose the development of biomolecular simulations software for adaptive computing that includes Reconfigurable Computing (RC) hardware and General Purpose Graphical Processing Units (GPGPUs) devices. The RC hardware, including Field Programmable Gate Arrays (FPGAs), and GPGPUs provide a tremendous amount of raw computing power even at the desktop level at a fraction of power requirements and cost of multi-processors parallel systems. PMEMD and LAMMPS, widely used biomolecular simulation engines, will be ported and optimized for popular RC/GPGPU devices. Moreover a molecular dynamics (MD) kernel specially designed to efficiently exploit the computational power of current and future RC/GPGPU devices will be developed. The proposed work will benefit the wide community of biochemists, biophysicists and computational chemists. The availability of these codes optimized on adaptive computing hardware will allow the non-expert user to benefit without worrying about the porting and optimizing details. Moreover, the availability of performance profiling utilities and the optimized MD kernel will enable other groups of application code developers to extend our implementation to exploit future FPGA and GPGPU devices enabled platforms. PUBLIC HEALTH RELEVANCE: The development of proposed optimized biomolecular simulations software will have direct impact on health and medical related research in many different areas including biochemical/biophysical characterization of cellular processes, drug-discovery and protein engineering. Biomolecular simulation software is used to investigate biological complexes and activities including protein folding, enzyme catalysis, conformational changes associated with bimolecular function, molecular recognition of proteins, DNA, and biological membrane complexes as well as docking/binding of small compounds to biomolecules.