Description: This episode will focus on the Modeling, Simulation and Analysis (MS&A) community’s computational needs and how HPC can be used to support mission outcomes. MS&A requires operational urgency outside traditional S&T and T&E expectations. Our top-notch panel composed of senior leaders, internal technical experts, community liaisons, resource management leaders, and users will share their experience and demonstrate how HPCMP is adapting to meet MS&A requirements. Our panelists include Mr. Josh Helms from the Center for Army Analysis, Mr. Jerry Smith, the N81 Campaign Modeling Lead for the Navy, Dr. Jed Richards, HPCMPs liaison to the MS&A community, Dr. Bob Donnelly, the HPCMP technical consultant for MS&A, and Ms. Sunita Batra, HPCMP’s Resource Management Associate Director. Please join us to hear first-hand how HPCMP is being utilized to support the MS&A community.

Each episode of HPCMP Talks will be a 30-45-minute live discussion guided by a moderator featuring a mix of program leaders, users, and invited guests, followed by a 15-minute question and answer session.

Description: This seminar will discuss Caliper, a performance analysis tool that provides lightweight always-on performance profiling, and Hatchet and Thicket, two tools designed to provide deeper insights into Caliper performance data for one or more profiles. This seminar will provide guidance on installing these tools, instrumenting applications and generating useful profiles with Caliper, and exploring the results to gain performance insights with Hatchet and Thicket. This will include an example demonstrating how to use these performance analysis tools with the MiniFE finite element mini-application.

CUI

Description: While MATLAB’s local parallel features perform well on a single workstation or HPC node, modern research often requires scaling beyond the core and memory limits of a single node. MATLAB Parallel Server allows users to transition their projects into larger, multi-node HPC jobs without needing to rewrite their core algorithms.

This seminar focuses on the practical setup and utilization of these tools within the MATLAB Live Editor in the Open OnDemand HPC environment. Attendees will review the fundamentals of MATLAB's parallel ecosystem before learning how to configure their environment for cluster execution, run multi-node interactive parallel sessions, and launch asynchronous batch jobs directly from the MATLAB Live Editor interface.

CUI

Description: This training will take attendees beyond the overview given in the HPCMP New Account Orientation and help them get started running jobs on HPC machines. The first part of the course will give users a better understanding of their HPCMP Account. They will learn more about pIE, subprojects, allocations, and how to get access to a machine. The next part of the course will go over how to access the machines by describing the portal, Kerberos/putty, and the SRD. Then the course will go over the linux environment that users encounter when they first log on to an HPC machine. Users will learn some basics for how to get started such as basic linux commands, details about the environment, shell scripting basics, modules, and basics on compiling codes. This will lead into the next part of the course that will show users how to submit jobs. They will learn about queueing systems, basics on how to write a batch script, and how to submit and monitor a job. The course will conclude by going over a range of other useful topics that would be helpful for new users to learn about. There will be demonstrations and hands-on examples throughout the course.

Description: This two-part seminar will introduce Xarray and Zarr and dive into the details of how these tools can be used to manipulate large multi-dimensional datasets. Zarr is becoming a common format for large multi-dimensional datasets in AI/ML workflows due to its ability to chunk large datasets and enable fast, scalable access. Xarray provides a high-level interface to multi-dimensional data that simplifies working with Zarr files. Part 1 of this seminar will cover the basics of Xarray for reading, writing, and interacting with multi-dimensional data and introduce the Zarr specification. Part 2 will cover a few advanced capabilities of Xarray and demonstrate an end-to-end workflow that pre-processes geophysical data, trains a model, and generates outputs.

CUI

Description: Scientists and engineers sometimes collect large volumes of video or imaging data, (e.g., schlieren imaging). To build machine learning (ML) models using this data, they need to label training data to feed to the ML model. This seminar will show how to harness the power of Meta Segment Anything Model (SAM) version 3 in the HPC environment so you can label your data faster and more accurately leading to stronger ML solutions to your research problems.

CUI

Description: This seminar will explore how to accelerate Fortran code using GPU offloading via OpenMP. This will be done by incrementally adapting a Fortran program to take advantage of the basic and advanced features provided by OpenMP for GPU offloading. Users with existing Fortran codes will gain the knowledge needed to incrementally adapt their own programs, and users writing new code will gain the knowledge needed to start with GPU offloading.

CUI

Description: Data visualization is a critical component of scientific workflows to effectively communicate results and data-driven insights. While Jupyter notebooks are commonly used to serve static visualizations, the full suite of interactive Python tools is often underutilized.

This seminar will teach users how to leverage existing libraries to create dynamic, interactive visualizations directly within Jupyter. We will then introduce Marimo, a next-generation reactive notebook that addresses common pitfalls of traditional notebooks, ensuring reproducibility and state consistency. Users will then learn how to deploy these interactive notebooks as standalone, web-based dashboards, making data presentation more concise and interactive.

CUI

Description: The HPCMP New Account Orientation webinar provides an introduction to the HPCMP and the resources available via the HPC Help Desk, PET, CREATE, and the DAAC. An overview of the following topics is also touched upon:

• The DoD Centers Website, Open OnDemand, and the HPC Training website
• User guides and SAMPLES_HOME
• The process for getting help (Help@helpdesk.hpc.mil)
• Highlights of popular online training curriculums on the HPC training website, such as New User training, DDA, and Singularity development

All followed up with a question and answer period.

CUI