Bulletin of the American Physical Society
APS March Meeting 2020
Volume 65, Number 1
Monday–Friday, March 2–6, 2020; Denver, Colorado
Session A40: Building the Bridge to Exascale: Applications and Opportunities for Materials, Chemistry, and Biology IFocus
|
Hide Abstracts |
Sponsoring Units: DCOMP DCMP DPOLY DBIO Chair: Jack Wells, Oak Ridge National Lab Room: 705 |
Monday, March 2, 2020 8:00AM - 8:36AM |
A40.00001: A Path to the Exascale for Atomistic Simulations with Improved Accuracy, Length and Time Scales Invited Speaker: Mitchell Wood With exascale super computers arriving in the near future, it is timely to ask whether our simulation software is capable of matching this unprecedented computing capability. While many research challenges in material physics, chemistry and biology lie just out of reach on peta-scale machines due to length and time restrictions inherent to Molecular Dynamics(MD), questions of the accuracy of our simulations will continue to linger. Simply running the same peta-scale simulations with more atoms on a larger computer (weak scaling) does not advance the accessible timescales, nor does it avoid the pitfalls of empirically developed constitutive models. This talk will overview the U.S. Department of Energy* EXAALT (EXascale Atomistics for Accuracy, Length and Time) project and our efforts to provide software tools for MD that not only scale efficiently to huge atom counts, but also enable efficient MD simulations for smaller systems too. New parallel time-acceleration methods such as sublattice-ParSplice and local hyperdynamics have been developed along with quantum accurate machine learned interatomic potentials to study damage accumulation in plasma facing materials. |
Monday, March 2, 2020 8:36AM - 8:48AM |
A40.00002: Accelerating Quantum Molecular Dynamics simulations: Can GPUs really help? Jean-Luc Fattebert, Christian F. A. Negre, Jamal Mohd-Yusof, Toks Adedoyin, Daniel Osei-Kuffuor, Susan Mniszewski GPU accelerators on the most powerful supercomputers give us an opportunity to speed up time-to-solution for large-scale Quantum Molecular Dynamics simulations that would otherwise be too slow for practical purposes. But using this type of hardware in an efficient manner presents serious challenges. Besides having to possibly rewrite large amounts of codes, algorithmic changes may be required for optimal efficiency. Even then, using GPUs at full capacity is not straightforward, in particular if there is not enough work for each GPU. In this talk we will present some software library solutions in development to facilitate porting electronic structure codes to new architectures, as well as parallel strategies and algorithms that can help speed up time-to-solution in real applications. |
Monday, March 2, 2020 8:48AM - 9:00AM |
A40.00003: Multibillion Atom Molecular Dynamics Simulations of Cellular Membranes Noah Trebesch, Emad Tajkhorshid Membranes are the basic organizational and defensive unit of the cell, and they thus play a vital role in biological function. Electron microscopy (EM) can provide 3D structures of these membranes, and, with the advent of exascale computing, there is a new opportunity to use molecular dynamics (MD) simulations to elucidate the intricate biophysical connection between the complex structure and function of these membranes. To take advantage of this opportunity, we have developed xMAS (Experimentally-Derived Membranes of Arbitrary Shape) Builder, software designed to turn low resolution EM-based structures of cellular membranes into atomistic models that are suitable for MD. Using xMAS Builder, we have built the first cell-scale (~4.5 billion atom) model of a representative cellular membrane (a helicoidal system from the endoplasmic reticulum), and we have also built several models of a smaller synthetic system with equivalent complexity. Preliminary simulations of these models have demonstrated their potential to reveal fundamental insights into the general behavior of cellular membranes, supporting the expectation that xMAS Builder will soon enable MD simulations that leverage exascale computing to provide detailed biophysical characterization of these key biological systems. |
Monday, March 2, 2020 9:00AM - 9:12AM |
A40.00004: Unveiling the structural properties of HIV-1 vesicle from atomistic molecular
dynamics simulations Fabio Gonzalez, Tyler Reddy, Juan Perilla The HIV-1 viral particle contains all the macromolecular components to infect host cells, |
Monday, March 2, 2020 9:12AM - 9:24AM |
A40.00005: Molecular Understanding of Membranes for the Water-Energy Nexus in the Exascale Realm Dvora Perahia, Gary Grest Polymeric membranes are used for a large variety of clean energy, medical, and environmental applications. Their use remains limited by tradeoffs of permeability, selectivity and longevity. A fundamental paradigm shift from a descriptive continuum approach to a predictive atomic and molecular level understanding of ion transport is essential for transformative progress to programable smart membranes. The need to control the interrelation of membrane structure, dynamics, transport and stability over a wide range of length and time scales is in the core of design of new membranes. Here we will present molecular dynamics simulation results that depict the structured motion and transport in polymeric membranes that consists of units, or blocks, with different chemical structures, tailored into a macromolecule with targeted roles. The impact of fundamentals that underline the structure/processing/properties relations that will be attained through exascale computing will be discussed. |
Monday, March 2, 2020 9:24AM - 9:36AM |
A40.00006: Multi-GPU parallelization of Deep Potential Molecular Dynamics for high-performance computing Denghui Lu, Weile Jia, Mohan Chen, Han Wang, Linfeng Zhang The recently developed Deep Potential Molecular Dynamics [1,2] (DPMD) builds many-body potentials for atomic systems based on the deep neural network and has been successfully applied to a variety of systems. The DPMD model owns the quantum mechanical accuracy and the linear growth computational complexity. In this work, we develop the multi-GPU parallelization for DeePMD-kit [3], an implementation of DPMD, and optimize the workflows when the package interfaces with LAMMPS and TensorFlow. We demonstrate that the resulting package is well-suited for high-performance computing with the aim of performing large-scale molecular simulations with quantum mechanical accuracy. |
Monday, March 2, 2020 9:36AM - 9:48AM |
A40.00007: Scalable Frameworks for Reinforcement Learning for Control of Self-Assembling Materials and for Chemistry Design Paul Welch, Christine Sweeney, Malachi Schram, Logan Ward The ExaLearn Exascale Computing Project has developed scalable frameworks for reinforcement learning (RL) to create policies to control scientific processes such as the self-assembly of block copolymers and chemical design. These policies could drastically reduce the time required to navigate large parameter spaces, optimizing experimental protocols. This accelerated search methodology may thus guide materials annealing experiments, exploration of candidate structures for battery materials, or evaluation of the configurational space for low-energy water clusters. The frameworks use various RL algorithms, environments and fast-running scientific simulations for the training process. RL training can be thought of as creating a sequence of moves in a game; at each move the player (agent) may decide to exploit previous knowledge (a policy) or explore new parameters (run a simulation). Scalability is achieved by running many RL training episodes on different nodes and aggregating models. Challenges include developing simulations, fully utilizing CPU and GPU resources on each node, and aggregating policies so as not to impede learning. We present results for full single node and preliminary multi-node computing resource utilization performance. |
Monday, March 2, 2020 9:48AM - 10:00AM |
A40.00008: Exascale-ready neural network interatomic potentials with CabanaMD Sam Reeve, Saaketh Desai, James Belak Computational predictions for materials require sufficiently accurate physics models which can be simulated in a reasonable amount of time. In the drive towards exascale computing, new hardware and software technologies are enabling more complex, accurate, and expensive models, but only with rethinking of algorithms, communication patterns, and data layouts. We exemplify this trend with a re-implementation of the Behler-style neural network potential (NNP), for classical molecular dynamics (MD) with near-quantum level accuracy. We use the Co-design center for Particle Applications (CoPA) Cabana particle library which, i) is built on Kokkos for on-node parallelism on various hardware, ii) provides performant particle-centric functionality, including MPI communication and neighbor lists, and iii) enables optimization of data structure for a given architecture through arrays-of-structs-of-arrays (AoSoA), intermediate between AoS and SoA. The NNP is added to the CabanaMD proxy app, where we show performance portability, including many-core CPU and GPU. |
Monday, March 2, 2020 10:00AM - 10:12AM |
A40.00009: DOE Software Center for Non-perturbative Studies of Functional Materials Under Non-equilibrium Conditions (NPNEQ) Tadashi Ogitsu, Xavier Andrade, Alfredo A. Correa, Liang Tan, David Prendergast, Sri Chaitanya Das Pemmaraju, Aaron Lindenberg Fundamental processes that underlie transformations of energy and information in material systems are intrinsically non-equilibrium in character as they involve additional coupling of electronic and ionic degrees of freedom to external time-dependent perturbations such as light or electrical voltages. Understanding and controlling the quantum non-equilibrium nature of functional materials on the characteristic Angstrom ( Å, 10-10 m) length scales and femtosecond (fs, 10-15 s) time scales of electronic motion is a crucial component in the rational design of energy-relevant materials and devices. |
Monday, March 2, 2020 10:12AM - 10:24AM |
A40.00010: Towards fast and accurate exascale density functional theory calculations using DFT-FE -- a massively parallel real-space code using adaptive finite-element discretization Sambit Das, Phani Motamarri, Vikram Gavini Kohn-Sham density functional theory (DFT) calculations have been instrumental in providing many crucial insights into materials behavior and occupy a sizable fraction of world’s computational resources today. However, the stringent accuracy requirements in DFT needed to compute meaningful material properties, in conjunction with the asymptotic cubic-scaling computational complexity with number of electrons, demand huge computational resources. Thus, these calculations are routinely limited to material systems with at most few thousands of electrons. In this talk, we present a significant advance in the state-of-the-art for accurate DFT calculations -via- the development of DFT-FE, that has enabled fast, scalable and accurate large-scale DFT calculations on material systems with tens of thousands of electrons. This has been facilitated by (i) the development of efficient and accurate spatially adaptive discretization strategies using higher-order finite-element discretization; (ii) developing efficient and scalable algorithms in conjunction with mixed-precision strategies; (iii) implementation innovations, both on many core and hybrid architectures, that significantly improve performance. |
Monday, March 2, 2020 10:24AM - 10:36AM |
A40.00011: Massively-Parallel Real-Time TDDFT Simulations of Electronic Stopping in Solvated DNA under Proton Irradiation Dillon C Yost, Yi Yao, Chris Shepard, Yosuke Kanai We discuss massively-parallel real-time, time-dependent density functional theory (RT-TDDFT) simulations for investigating electronic stopping in DNA solvated in water. We have developed RT-TDDFT capabilities within the Qbox/Qb@ll code, based on a planewave-pseudopotential formalism [1], which is applied to study the electronic excitation dynamics of solvated DNA under proton irradiation. In electronic stopping processes, massive electronic excitations are produced by fast energetic charged particles like protons. Electronic stopping is central to DNA damage by ion irradiation, which is central to ion beam cancer therapy. We discuss the scalable implementation and performance of the RT-TDDFT simulations and recent results for solvated DNA, a system which includes more than 13,000 electrons. |
Monday, March 2, 2020 10:36AM - 10:48AM |
A40.00012: Spatiotemporal Mapping of Polymer Dynamics Jihong Ma, Jan-Michael Carrillo, Bobby Sumpter, Yangyang Wang Polymers present unique challenges to computational science, as their structures and dynamics are characterized by a remarkably wide range of length scales and timescales. The recent advances in high performance computing systems have afforded new opportunities to investigate the slow dynamics in polymeric liquids. In this talk, we demonstrate a novel approach to resolve the fine spatiotemporal features of entangled polymer dynamics, by leveraging the computational resources at the Oak Ridge Leadership Computing Facility. Potential applications of this new method will also be discussed. |
Monday, March 2, 2020 10:48AM - 11:00AM |
A40.00013: Towards adaptive exascale workflows for simulating long timescales John Ossyra, Ada Sedova, Jeremy Smith Molecular dynamics simulations integrate trillions of short simulation steps and are reaching hardware-bound performance limits for clock time per step. Despite this, most biologically relevant processes occur over seconds to hours, orders of magnitude slower than feasible simulation timescales. Workflow-based methods harnessing complex statistical mechanics analyses are increasingly used with the hope of achieving kinetic estimates on experimentally relevant timescales from swarms of short parallel simulations. We present a highly scalable workflow software, AdaptiveMD, that implements a massively parallel adaptive-sampling algorithm to build a Markov-state model of long-timescale biomolecular kinetics. The software was ported to the Summit pre-exascale machine at the Oak Ridge Leadership Computing Facility (OLCF), after initial development on OLCF Titan, and demonstrated excellent scalability and fault tolerance. Summits software and hardware features allow for an unprecedented ability to simulate large biomolecular systems within this framework. These results suggest that this method at exascale may allow us to tackle grand challenges in biomolecular simulations. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700