NAMD: Scalable Molecular Dynamics Based on the Charm++ Parallel Runtime System
Book Chapter 2017
Publication Type: Paper
Repository URL:
Download:
[BIB]
Abstract
NAMD is a scalable molecular dynamics application designed for high performance atomic-level simulation of large bio-molecular systems at a femtosecond time step resolution [1]. Used by tens of thousands of scientists - 85,000 users since 2000 - on everything from laptops and desktops to supercomputers and GPUs (even iPads!), NAMD has enabled significant breakthroughs in understanding the structure and the behavior of viruses and cellular organelles.
The fixed size nature of the molecular systems requires fine-grained parallelization techniques and strong scaling to achieve efficient simulation of long timescales. NAMD is built on top of the parallel framework Charm++ [2] [3], which provides a robust infrastructure that enables simulations of 100 million atoms size to scale up to hundreds of thousands of cores, while providing portability with performance across different supercomputing architectures. The NAMD and Charm++ collaboration was started in 1992 by principal investigators Klaus Schulten, Laxmikant V. Kale and Robert Skeel. NAMD’s performance needs have motivated Charm++ to develop new methods and abstractions resulting in successful co-development and interdisciplinary collaboration. In 2012, the IEEE Computer Society’s Sidney Fernbach Award was jointly awarded to Kale and Schulten “for outstanding contributions to the development of widely used parallel software for large biomolecular systems simulation”. One example of this is the petascale simulation of the HIV capsid with 64 million atoms on the Blue Waters Cray XE6 System at NCSA, enabling the precise determination of the chemical structure of the HIV capsid for the first time. This groundbreaking study was published and featured in Nature in 2013 and “petascale simulations are now being used to explore the interactions of the full capsid with drugs and with host cell factors critical to the infective cycle” [4]. Meanwhile, our efforts continue to prepare NAMD for exascale simulations not only in terms of performance and portability, but also in terms of reliability and energy efficiency as well.
People
- Bilge Acun
- Ronak Buch
- Laxmikant Kale
- James Phillips
Research Areas