Dear NERSC users, Below are some upcoming training events in October that you are encouraged to attend: 1) In-Person ALCF Hands-On HPC Workshop, Oct 10-12 (Application deadline Sept 15) ALCF is hosting an in-person Hands-on HPC Workshop to provide an opportunity for hands-on time on Polaris and AI Testbeds focusing on porting applications to heterogeneous architectures (CPU + GPU), improving code performance, and exploring AI/ML applications development on ALCF systems. Workshop participants will: - Work directly with ALCF staff and industry experts during dedicated hands-on sessions - Learn how to use available tools and frameworks to improve productivity and performance - Hear from researchers running on early Aurora hardware - Tour Argonne facilities, including Aurora For more information please see https://www.nersc.gov/alcf-hands-on-hpc-workshop-oct2023/. 2) GPUs for Science Day, Oct 12 (in person, registration deadline Sept 28) GPUs have been instrumental in ground-breaking innovations, from scientific simulations to generative AI. This year, NERSC is proud to host the annual GPUs for Science event *in person*. Our goal is to celebrate recent GPU-enabled scientific achievements and inspire future roadmaps. The day will start with an introduction to three DOE compute facilities (NERSC, ALCF, and OLCF), followed by a series of talks on GPU-accelerated scientific applications and emerging software programming models. The day will wrap up with a panel of leading industry experts discussing their vision for upcoming GPU ecosystems (NVIDIA, AMD, and Intel). With increasing diversity in GPU hardware and software, users interested in performance and portability across DOE supercomputers are highly encouraged to join. The event is open to everyone interested in learning about the exciting science in action and registration is free. For additional details including the agenda and to register by Sept 28, please visit https://www.nersc.gov/gpus-for-science-day-2023. 3) HIP Training Series, Part 4 on Oct 2 and Part 5 on Oct 16 HIP® is a parallel computing platform and programming model that extends C++ to allow developers to program GPUs with a familiar programming language and simple APIs. AMD will present a multi-part HIP training series intended to help new and existing GPU programmers understand the main concepts of the HIP programming model. Each part will include a 1-hour presentation and example exercises. The exercises are meant to reinforce the material from the presentation and can be completed during a 1-hour hands-on session following each lecture on OLCF Frontier and NERSC Perlmutter. *Part 1* of the HIP training session was held on August 14, with the topic of Introduction to HIP and GPU. *Part 2* was held on August 28, with the topic of Porting Applications to GPU. Please see details, presentation materials, and exercises for *Part 1* <https://www.nersc.gov/intro-gpus-and-hip-part-1-of-hip-training-series-aug-14-2023/> and *Part 2* <https://www.nersc.gov/porting-applications-to-hip-part2-hip-training-series-aug2023/> . *Part 3 <https://www.nersc.gov/amd-memory-hierarchy-part3-hip-training-series-sep2023/>* will be held on Monday, September 18 with the topic of AMD Memory Hierarchy. With an understanding of GPU memory systems and in particular the AMD GPU memory system, this talk will explore how to improve the performance of applications. This understanding is crucial to designing code to perform well on AMD GPUs. *Part 4 <https://www.nersc.gov/gpu-profiling-performance-timelines-rocprof-omnitrace-part4-hip-series-oct2023/>* will be held on Monday, October 2, with the topic of GPU Profiling (Performance Timelines: Rocprof and Omnitrace), two tools for AMD GPUs to collect application performance timelines data. *Part 5 <https://www.nersc.gov/gpu-profiling-performance-profile-omniperf-part5-hip-series-oct2023/>* will be held on Monday, October 16, with the topic of GPU Profiling (Performance Profile: Omniperf), a tool for getting application performance profiles on AMD GPUs. For more information and to register, please see the event web pages linked above. 4) NERSC/Nvidia AI For Scientific Computing Bootcamp, Oct 18-20 (Application deadline Oct 4) NERSC, in collaboration with the OpenACC organization and NVIDIA, is hosting a virtual, three-day AI for Scientific Computing Bootcamp from Wednesday, October 18, through Friday, October 20, 2023. This bootcamp will provide a step-by-step overview of the fundamentals of deep neural networks and walk attendees through the hands-on experience of building and improving deep learning models for applications related to scientific computing and physical systems defined by differential equations. This event has limited capacity, and active NERSC/ALCF/OLCF users will be prioritized, but for best results *apply before Oct 4*. Acceptance will be confirmed via email close to the bootcamp start date. For more information and to register, please see https://www.nersc.gov/ai-for-scientific-computing-oct-2023/. 5) OpenMP Offload 2023 Training Series, Part 1 on Sept 29 and Part 2 on Oct 6 The OpenMP API is a scalable model that gives parallel programmers a simple and flexible interface for developing portable parallel applications in C/C++ and Fortran. This 4-part OpenMP Offload training will enable application teams and developers to accelerate their codes with the use of GPUs, as well as exploit the latest OpenMP functionality to program multi-core platforms like Perlmutter and Frontier. This training series, offered by OLCF and NERSC, is also part of the Performance Portability training series <https://www.olcf.ornl.gov/performance-portability-training-series/>. Part 1 <https://www.nersc.gov/openmp-offload-2023-training-part1-basics-of-offload/> on Sept 29 will cover a general overview of the OpenMP programming model and cover the basics of using OpenMP directives to offload work to GPUs. The hands-on sessions will be performed on OLCF Frontier and NERSC Perlmutter. Part 2 <https://www.nersc.gov/openmp-offload-training-part2-optimization-data-movement-oct2023/> on October 6 will cover optimization strategies and show how efficient data movement and a better understanding of the hierarchy of parallelism available can lead to improved performance. We will also cover best practices for OpenMP Offload. The hands-on sessions will be performed on OLCF Frontier and NERSC Perlmutter. For more information and to register for each part, please refer to the event web pages linked above. 6) Training on the RAJA C++ Portability Suite, Oct 10 As part of our Performance Portability training series <https://www.olcf.ornl.gov/performance-portability-training-series/>, NERSC and OLCF are hosting a training session on the RAJA C++ Portability Suite on October 10. RAJA is a C++ library offering software abstractions that enable architecture and programming model portability for HPC application codes. RAJA offers portable, parallel loop execution by providing building blocks that extend the generally-accepted parallel for idiom. This is a 1-part session that will allow participants to learn from and interact directly with RAJA team members. The session will give a general overview of RAJA and cover the basics of using RAJA abstractions to offload work to the GPUs. Throughout the session, a variety of quiz-like puzzles will be used to engage the audience and reinforce concepts. For more information and to register, please see https://www.nersc.gov/performance-portability-series-raja-oct2023/. 7) OLCF AI for Science at Scale Training, Oct 12 OLCF is holding a series of training events on the topic of AI for Science at Scale, which are open to NERSC users. The second training in the series is scheduled for October 12, and will focus on how to train a model on multiple GPUs and model parallelism techniques and frameworks, such as DeepSpeed, FSDP, and Megatron For more information and to register, please see https://www.nersc.gov/olcf-ai-training-series-ai-for-science-at-scale-part2-oct2023 . 8) Use Spin to Build Science Gateways at NERSC, Oct 18, 25-26 Spin is a service platform at NERSC based on Docker container technology. It can be used to deploy science gateways, workflow managers, databases, and all sorts of other services that can access NERSC systems and storage on the back end. New large-memory nodes have been added to the platform, increasing the potential of the platform for new memory-constrained applications. To learn more about how Spin works and what it can do, please listen to the NERSC User News podcast on Spin: https://anchor.fm/nersc-news/episodes/Spin --Interview-with-Cory-Snavely-and-Val-Hendrix-e1pa7p and see more information at the Spin documentation <https://docs.nersc.gov/services/spin/> page.Attend an upcoming SpinUp workshop to learn to use Spin for your own science gateway projects! Applications for sessions that begin on Wednesday, Oct 18 are now open <https://www.nersc.gov/users/training/spin>. SpinUp is hands-on and interactive, so space is limited. Participants will attend an instructional session and a hack-a-thon to learn about the platform, create running services, and learn maintenance and troubleshooting techniques. Local and remote participants are welcome. Best Regards, Helen He NERSC Training _______________________________________________ Users mailing list Users@nersc.gov