Description: The NVIDIA HPC SDK provides a wide array of tools for developing HPC applications, especially those accelerated by NVIDIA GPUs. Day one of this training event will cover the HPC compilers (formerly known as the PGI compilers) and how to use them on a variety of HPCMP systems. Day two will focus specifically on how the compilers can be used to create GPU-accelerated code through a number of different programming paradigms. During both days, the wide variety of compiler options and associated libraries for optimizing, parallelizing, and offloading C, C++, and Fortran codes will be discussed.

Presenter(s): Ryan Montesi, GDIT/PET
Location: Webcast
Date & Time: August 25-26, 2021, 2:00p - 4:00p ET

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only, Administrative or Operational Use, 19 August 2021.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Distribution Statement C. Distribution a authorized to U.S. Government Agencies and their contractors, Administrative or Operational Use, 04 February 2020.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Description: This training discusses GPU memory management through compiling optimization, CUDA unified memory and advanced memory management technology (prefetch, memory hints, oversubscription etc.). In this seminar we will focus on memory serial operation, blocking, stream overlap technology, and on how to avoid excess communication between CPU and GPU, and the latest approaches such as prefetching and memory advise technique. Sample codes will be adopted to demonstrate global memory and shared memory optimization, pinned memory and pageable memory comparison, and CUDA unified memory management. Additionally, NVIDIA profiling tools such as Nsight Systems and Nsight Compute will be used to address performance of computing, for instance, GPU memory workload analysis.

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only, Administrative or Operational Use, 21 June 2022.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Description: This one day course is a preliminary introduction to performance optimization using TAU, the most widely available and used code profiler on HPCMP systems. This course covers use of Tau for novice to intermediate level users.

Presenter: Shawn Brown, Pittsburgh Supercomputing Center

Date/Location: July 24, 2014, ERDC, Vicksburg MS 

Distribution Statement D. Distribution limited to the Department of Defense and U.S. DoD contractors only, for administrative or operational use, 24 July 2014.  Other requests for this document shall be referred to the High Performance Modernization Office, 10501 Furnace Road, Suite 101, Lorton, VA 22079.

Description: Delivering solutions to address the complex problems encountered in science and engineering requires significant development and consideration of methods to automate and scale end-to-end analysis pipelines and workflows. Traditional methods of managing scientific workflows for High Performance Computing (HPC) environments have become increasingly difficult to implement in response to the growing number of heterogenous compute resources, availability of data, and ongoing advancements of simulation and data analysis tools. Workflow managers provide a high-level interface between system resource management and available programming models to abstract the process of defining computational tasks and how they should be executed. These interfaces and tools have become important components of recent scientific discoveries, demonstrating the ability to accelerate the time-to-solution while preserving data provenance and ensuring reproducible results. In this training event, we will discuss one such tool, Galaxy Simulation Builder, and how it can be used to accelerate the development, management, and execution of scientific workflows. To demonstrate the applicability of this tool, the development of three workflows related to modeling/simulation, optimization and machine learning will be discussed.

Description: Building on the HPCMP Storage and File Use (Mar 2021) seminar, this training discusses more advanced topics such as automating file transfers and improving performance of file copies and access.

File transfer topics include: using data transfer jobs before/after calculations; using rsync to keep files updated across various storage locations; and validating transfers via checksums.

Optimization methods include: speeding file copies via scampi and other tools, improving performance by using RAM disks, and optimizing the Lustre file system for a given application by adjusting striping.

UNIX file permissions to enable sharing is revisited, including setting defaults via umask and setting permissions used in a PBS job. Additional topics include using flock to lock files to allow updates from multiple processes or nodes and using split when necessary to ease file transfer.

Presenter(s): Dr. Robert Donnelly, GDIT / PET
Location: Webcast
Date & Time: May 18, 2021, 2:00p - 3:30p ET

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only, Administrative or Operational Use, 20 May 2021.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Description: KerasTuner is a library that assists in choosing the optimal set of hyperparameters for TensorFlow programs. It is basically a user-friendly interface for three types of hyperparameter search strategies: Random Search, Bayesian Optimization, and Hyperband. This webinar will examine what hyperparameters and hyperparameter tuning are. We will describe KerasTuner and how it works, outline the steps in using KerasTuner and show code examples and cover adjusting KerasTuner to meet specific needs. We will cover the three types of hyperparameter search strategies and reasons for using one approach over another.

Presenter(s): Sheila McGee-Rosser, GDIT/PET
Location: Webcast
Date & Time: April 21, 2022, 2:00p - 3:30p ET

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only, Administrative or Operational Use, 21 April.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Description: The Intel Streaming SIMD Extensions (SSE) are an extension to the x86 architecture that fosters explicit user programming of the SIMD hardware in Intel CPUs. This seminar will introduce these extensions to C and Fortran programmers and explain how to use them to manually optimize compute-intensive portions of their applications.

Presenter(s): Dr. George Gusciora, GDIT / PET
Location: Webcast
Date & Time: March 23, 2022, 2:00p - 3:30p ET

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only, Administrative or Operational Use, 23 March 2022.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Description: Problems involving parameter optimization are ubiquitous in science and engineering. We will present an introduction to how to perform parameter optimization with the program MOES, Multiple Objective Evolution Strategies. This includes an overview of evolutionary algorithms in general and a more detailed discussion of the self-adaptive algorithm, Evolution Strategies, along with its application to both unconstrained and constrained single-objective problems and to multiple objective problems. We will also discuss the concept of Pareto Dominance and give some background material on Linear Programming and Data Envelopment (or Frontier) Analysis, and show how these methods are combined with Evolution Strategies to solve problems in multiple objective optimizations. Day 2 and 3 we will look at examples using MOES.

Presenter(s): Dr. James Lill, GDIT/PET
Location: Webcast
Date & Time: August 17-19, 2021, 2:00p - 3:30p ET

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only, Administrative or Operational Use, 9 August 2021.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Description: Deal.II (Differential Equation Analysis Library II) is a state-of-the-art finite element Toolbox focused on generality, parallelism, and extensibility. It provides a powerful software library that is amongst the largest of the libraries supporting numerical PDE solvers and may serve as a potential alternative to currently supported commercial and in-house finite-element libraries and tools on the DoD HPC.

This seminar will offer HPC users the opportunity to gain experience with open-source software management using Spack and CMake, from installation to testing/debugging, with guidance on managing software dependencies, optimizing HPC environment, and identifying compiler requirement as well as libraries.

CUI

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only, Administrative or Operational Use, 25 June 2020.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Description:Computational methods developed in structural mechanics, heat transfer, fluid mechanics, shock physics, and many other fields of engineering can be an enormous aid to understanding the complex physical systems they simulate. Often, it is desired to use these simulations as virtual prototypes to obtain an optimized design for a particular system, or to develop confidence in performing predictions for systems that cannot be observed or tested directly. Dakota seeks to enhance the utility of these computational simulations by enabling their use for a variety of iterative analyses, so that simulations may be used not just for single-point solutions, but also achieve broader impact in the areas of credible prediction and optimal design.

Presenter: Zachary A. Wilson, Ph.D., PETTT Engility

Date/Location: June 20, 2018, DSRC Viewing Portal, APG, MD

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only; administrative or operational use; 20 June 2018. Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Description: Simple Linux Utility for Resource Management (Slurm) is a best-in-class job scheduler and resource manager for HPC clusters. This seminar provides DoD users with fundamental information about how to use the Slurm scheduler to manage job and task execution on allocated HPC systems. Also provided is an overview of how to transition from PBS to Slurm and leverage advanced Slurm features to optimize their workflows.

Description: In this seminar, you will be brought up-to-speed on the cutting edge in molecular dynamics and quantum chemistry on NVIDIA GPUs with a focus on newly ported features and performance optimizations. We will touch on the latest NVIDIA GPU programming models that enabled the optimization work, including acceleration with standard Fortran and C++. Finally, we will end with a demo session where you’ll learn how to pull pre-compiled, optimized applications from the NVIDIA NGC container registry and deploy them on GPU nodes.

Presenter(s): Dr. Kristopher Keipert, Nvidia
Location: Webcast
Date & Time: February 24, 2022, 2:00p - 3:30p ET

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only, Administrative or Operational Use, 24 February 2022.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Description: The pre-processing such as geometry cleanup, repair, defeaturing, and mesh generation usually takes 50 to 60% effort of a whole project in most large-scale engineering simulations with complicated geometries. This short training is to expand users’ knowledge of Ansys meshing capabilities. The lecture focuses on mesh optimization and control, mesh statistics and metrics, cell intersection and inflation collision remedy, smooth mesh transition, different approaches to multi-body parts and multiple-body parts, among other topics. 

Day 1 focuses on Ansys meshing fundamentals and method methods/algorithms. Day 2 focuses on global and control mesh controls and optimization following by an exercise of tutorials.

Presenter(s): Dr. Shaolin Mao, GDIT / PET
Location: Webcast
Date & Time: January 25 & 26, 2022, 2:00p - 4:00p ET

Distribution Statement D. Distribution authorized to the Department of Defense and U.S. DoD contractors only, Administrative or Operational Use, 26 January 2022.  Other requests for this document shall be referred to the High Performance Computing Modernization Office, 3909 Halls Ferry Road, Vicksburg, MS 39180.

Description: This seminar will introduce a new integration of MOES (Multiple Objective Evolution Strategies) with the molecular dynamics (MD) engine LAMMPS (Large-scale Atomic/Molecular Massively Parallel Simulator) that enables scalable optimization of MD potentials with advanced evolutionary algorithms. The MOES+LAMMPS interface preserves the full flexibility of LAMMPS while offering granular runtime control over construction of objective functions. In this training event, we will demonstrate the capabilities of MOES+LAMMPS and step through the process of optimizing an MD potential.

CUI