GTC Silicon Valley-2019: Using Nsight Tools to Optimize the NAMD Molecular Dynamics Simulation Program
GTC Silicon Valley-2019 ID:S9503:Using Nsight Tools to Optimize the NAMD Molecular Dynamics Simulation Program
David Hardy(University of Illinois at UrbanaChampaign),Robert Knight(NVIDIA)
Learn how we used NVIDIA's Nsight tool suite to help focus the development effort and optimize NAMD, which is being rewritten to vastly increase its utilization of GPUs. We'll explain how sophisticated tools like NVIDIA's Nsight Systems and Nsight Compute help determine if an application is using compute resources efficiently and whether your GPUs are starving for work or completely overwhelmed. We'll discuss how the NAMD development team at the University of Illinois at Urbana-Champaign is using Nsight tools to solve real-life performance bugs and bottlenecks, such as streamlining host-to-GPU transfers and optimizing NAMD kernels. We'll also cover how we used a new direct data access API to develop a custom tool that parses Nsight Systems data to quickly find and debug performance outliers.