Xuebin Chi is currently Professor in the Computing Scientific Application Center, Computer Network Information Center, Chinese Academy of Sciences. Dr. Chi is Director of Supercomputing Center and Vice Director of CNIC. His research focuses on Computational Mathematics. Since 1994, Dr. Chi has been rewarded three times for 2nd Prize of National Science and Technology Progress Award. As one of China's high performance computing and grid computing academic leaders, Dr. Chi presided over and assumed a lot of scientific research items, including the National High Technology Research and Development Program of China, the Major State Basic Research Development Program of China, General Program of National Natural Science Foundation of China, Knowledge Innovation Project of The Chinese Academy of Sciences, Etc.
Jian Zhang, PI. Ph. D., associate Professor, Supercomputing Center, Chinese Academy of Sciences. Dr. Zhang received his Ph. D. in Applied Mathematics from University of Minnesota, in May 2005. He worked as a postdoc fellow in the Pennsylvania State University from June 2005 to June 2009. From August 2009 to present, he worked as assistant Professor and associate Professor in the supercomputing Center, CAS. His research interests include high performance computing, scientific computing, and scientific visualization. He has participated in 7 national science and industry projects and published over 40 papers. His is currently the PI of one NSFC project.
Dr. Zhong Jin is currently Professor in the Computing Scientific Application Center, Computer Network Information Center, Chinese Academy of Sciences. He pursued his PhD in Department of Chemistry, Emory University, USA. Dr. Jin’s research focuses on software of high performance computational chemistry. He is running the Virtual Laboratory for Computational Chemistry which facilitates computational chemists to carry out their computing jobs, to visualize molecule and to analyze calculation results. He already had more than 30 publications published and more than 20 invited talks. In addition, Dr. Jin is in charge of international affairs of SCCAS.
Many interesting phenomena in biology, material science, and soft condensed matter physics that take place in the mesoscopic realm require performing long time simulations on large spatial scales, which are very computationally demanding and thus not easily accessible by current simulations on state-of-the-art CPU computers. Accordingly, to shed more light on this topic as well as to enhance our ability to access those slowly evolving large-scale complex processes, speeding up the simulation of large-scale systems is an important subjects in computational chemistry and computational material science. We focus on two mainstream mesoscopic simulation techniques, namely the Phase Field method and the Dissipative Particle Dynamics (DPD), and develop efficient algorithms and codes targeting Intel® Xeon® processor and Intel® Xeon Phi™ coprocessor.
We have devised fast and stable compact Exponential Time Difference (cETD) multistep methods for solving the Phase Field equations. Our time stepping methods are explicit in nature and thus free from the need to solve linear and nonlinear systems. They utilize accurate exponential time differences with multistep approximations to maintain accuracy and linear splitting to achieve stability. Compact representation of central differences are used for discretizations of spatial operators so that our method can deal with various boundary conditions and can be straightforwardly coupled with parallel domain decomposition methods. All these techniques are seamlessly coupled together to produce large time step, accurate and stable numerical algorithms.
We will develop cETD library for Intel® Xeon® processor and Intel® Xeon Phi™ coprocessor and it is expected to reach 30% of peak performance in production-level simulations. Besides the Phase Field models, the cETD methods are applicable to a wide variety of Partial Differential Equations including but not limited to the time-dependent advection-diffusion and Navier-Stokes equations for fluids dynamics, the Ginzburg-Landau equations for modeling superconductivity, the Schrödinger equations for quantum mechanics, and so on. The cETD library will also contribute to these areas.
We will also develop efficient DPD simulation code based on Intel® Many Integrated Core Architecture (Intel® MIC Architecture). This implementation is designed and optimized according to the nature of DPD simulationtechnique and will also fully take advantage of the computational power of MICs. This MIC-based implementation will provide a speedup of two comparied to that based on a single CPU.
The focus to modernize the software codes will have a positive impact to “LAMMPS”, which is the most widely used applications in molecular dynamics. LAMMPS is a classical molecular dynamics code that models an ensemble of particles in a liquid, solid, or gaseous state. Providing researchers and scientist an understanding of our basic biology and DNA, cures for disease, genetics, and so much more.
- May 11, 2018, Performance Tuning of LAMMPS Dissipative Particle Dynamics Simulation on Intel KNL, IPCC Asia Summit 2018
- Weipeng Jing, Danyu Tong, Yangang Wang, Jingyuan Wang, Yaqiu Liu, Peng Zhao, February 2017, MaMR: High-performance MapReduce programming model for material cloud applications, Science Direct
Product and Performance Information
Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.