LUMI Intensive: Comprehensive Intro and Advanced Workshop

Join us for an intensive five-day course designed to elevate your understanding and skills in utilizing the LUMI supercomputer. This course is divided into two parts: a two-day introductory segment followed by a three-day advanced workshop. Participants have the option to register for the three-day advanced workshop independently if they are already familiar with the introductory material.

Course Overview

Each day will run from 9:00 to 17:30 CET (Friday till 15:30 CET), with breaks scheduled throughout. A preliminary schedule is available on https://lumi-supercomputer.github.io/LUMI-training-materials/2p3day-20250303/schedule/.

Introductory Part (2 Days) – Monday & Tuesday (3.-4.3.2025)

This segment serves as an introduction to the LUMI architecture and setup. It includes lessons on hardware architecture, compiling, software usage, and efficient job execution. By the end of this part, you will be equipped to work effectively on both the CPU (LUMI-C) and GPU partition (LUMI-G).

This part consists of lectures and some hands-on exercises, covering the following topics:

• Connecting to LUMI and managing data transfers efficiently.
• Understanding LUMI’s hardware and compiling software effectively, using the module system and managing your software environment with EasyBuild.
• Submitting and managing jobs with Slurm, including job arrays and GPU/CPU binding.
• Creating Python environments and containers.

This segment is lectured from a lecture room with audience. Remote participants can ask written questions, and we’ll try to pick them up at the end of each section.

Advanced Workshop (3 Days, on-site workshop + streaming of lectures) – Wednesday-Friday (5.-7.3.2025)

Building on the introductory knowledge, this workshop offers a comprehensive exploration of the LUMI architecture and programming environment. It covers advanced topics such as programming models (HIP and OpenMP offload), porting, job execution, and application optimization for AMD MI250X. This part emphasizes code development and performance tuning. Some of the presentations present a tool that you can use to understand program behaviour, other talks discus a particular performance problem and how to solve it.

The workshop consists of a mix of short lectures, live demos and hands-on exercises, that cover the following key topics:

• Running CPU and GPU profilers and analyzing the output
• Identifying bottlenecks in applications
• Mitigation of the most common performance bottlenecks (for example serial code performance, application placement, MPI, data movement and I/O, GPU occupancy, registers occupancy, strided data access pattern)
• Expert advice on achieving optimal performance on the LUMI platform for your application (on-site only)

To maximize your learning experience and benefit from direct interactions with experts, we highly recommend attending the 3-day advanced workshop in person. This should not be considered a hybrid course, but an on-site course that is broadcast, where all available experts will focus first on interactions with the on-site participants rather than questions asked by an online audience. Experience has also shown that the discussions that sometimes unwind in the room, are hard to follow online.

Registration Options

• Full Course (5 Days): Ideal for those seeking a complete understanding of LUMI, from basics to advanced techniques.
• Advanced Workshop Only (3 Days): Suitable for participants who have already completed the introductory course or possess equivalent knowledge. (On-site attendance highly recommended as only the lectures but not the exercises will be streamed)
• Introductory part only(2 Days) – online only: For users who only need our introductory training. As we don’t want to take seats away from users taking the full course, there is no possibility to participate on-site for just the first two days. Note though that only written questions can be asked by online participants as the lecturers focus on interaction with the participants in the course room.

Course Prerequisites

Participants are expected to have:

• Introductory Part: Basic knowledge of HPC cluster computing and familiarity with Linux command line and programming (C, Fortran, or Python).
• Advanced Workshop: Completion of a recent LUST introductory course (May 2024 or later), or equivalent experience specifically covering:
  • How to connect to LUMI
  • Familiarity with the LUMI environment (directories, common commands)
  • How to run on LUMI
  • Binding of processes
• It is strongly advised that anyone intending to only attend the Advanced Workshop review the content of the previous LUMI introductory course as noted above in advance of attending.
  You can find slides and recordings of the last LUMI intro course on our training material page.

Please Note:
This course is focused on the specifics and peculiarities of LUMI and is not a general high performance computing (HPC) introduction. It is targeted at current and future users of LUMI.

Learning Outcomes

After completing the full course, participants will be able to:

• Connect to LUMI and transfer data efficiently.
• Understand LUMI’s hardware and compile software effectively.
• Utilize the module system and EasyBuild for software management.
• Submit and manage jobs with Slurm, including job arrays and GPU/CPU binding.
• Identify and mitigate I/O bottlenecks in the LUSTRE file system.
• Create Python environments and run containers on LUMI.
• Write job scripts for optimal performance, including efficient GPU binding.
• Compile programs using the Cray Programming Environment efficiently.
• Profile applications to understand performance bottlenecks.
• Use parallel debuggers to solve problems in MPI applications.

Registration

Register here: https://ssl.eventilla.com/event/LPNqg

Participants will receive confirmation shortly after the deadline. If your plans change, we kindly ask you to cancel your registration as soon as possible. The email acknowledging your registration will contain a link to manage it.

Users, who don’t have an account on LUMI yet, will receive temporary access for the purpose of the course. The compute time allocated to the course shall only be used for the purpose of doing the exercises of the course. Any abuse will lead to removal from the allocation for this and future courses.

Please, do not hesitate to contact the LUMI User Support Team https://lumi-supercomputer.eu/user-support/need-help/ if you need any assistance.

Event info

The course will be jointly conducted by the LUMI HPE Centre of Excellence (HPE CoE), AMD and the LUMI User Support Team (LUST). Several staff members from HPE, AMD and LUST will be on-site, so this is the perfect moment for interaction with them.

The course is intended for users with ongoing projects on LUMI, users with project proposals in one of the national or EuroHPC channels, and support staff of local organizations of the LUMI consortium members.

This course is not designed specifically for people running AI on LUMI although it does include some information relevant to execution of AI frameworks. Please consider the “Moving your AI training jobs to LUMI” workshop for an introduction to AI on LUMI.

The venue is at the offices of KTH in Stockholm, Sweden. Participants are themselves responsible for all travel bookings.

If registrations exceed the number of planned participants, we will maintain a waiting list. The registration link can be found at the bottom of this page. The registration deadline is 24. February 2025, at 12:00 CET.