Using All Cores Efficiently with OpenMP: Scheduling and Load Balancing

Online hands-on workshop · Dec 14, 2026

CSHPC

Using All Cores Efficiently with OpenMP

Scheduling and Load Balancing

Monday, December 14, 2026 · 13:00–17:00 (CET)
Online · in English

by Jonas Korndörfer

Register by December 7 →

We cordially invite you to the online hands-on workshop Using All Cores Efficiently with OpenMP: Scheduling and Load Balancing, organized by Saarland University as part of NHR Süd-West. Modern compute nodes have dozens to hundreds of cores, but a parallel program is only as fast as its slowest thread: when the work is unevenly distributed, or the system slows down some threads, most cores end up waiting. OpenMP is the standard way to use all cores of a node from C, C++ and Fortran, and its scheduling options decide how the work is shared among threads. This workshop introduces the basics of OpenMP and the ideas behind scheduling and load balancing, and shows how to apply them in practice on an NHR system.

Who should attend?

Researchers and students who write or use C or C++ codes and want to use the cores of an HPC node efficiently. No prior OpenMP experience is required.

What you will learn

1

Load imbalance. What it is, what causes it (the algorithm, the system, communication and I/O delays) and how to measure it.

2

Scheduling ideas. Static and dynamic scheduling, self-scheduling and work stealing.

3

OpenMP basics. Parallel regions, data sharing, race conditions and synchronization.

4

Loop schedules. How static, dynamic and guided work, and what the OpenMP runtime actually does with them.

5

Tasks. Using OpenMP tasks for irregular work and keeping them balanced.

6

A practical recipe. Finding and fixing load imbalance in your own code.

Format

The workshop takes place online and combines a lecture with hands-on exercises in C++ on an NHR system. Participants parallelize small programs, see how different schedules distribute the work, and measure the effect on run time and load imbalance. Registered participants receive the meeting link by email closer to the workshop date.

Prerequisites

• Basic knowledge of C or C++ and the Linux command line
• No prior OpenMP knowledge required; some experience with Slurm is helpful
• A laptop with an internet connection and an SSH client

Registered participants receive information about the hands-on environment and any required setup by email before the workshop.

Registration

Register here → until Monday, December 7, 2026

Eligibility	Members of German universities, academic computing centers and research institutions
Registration fee	None
Organization	CSHPC, Saarland University, as part of NHR Süd-West
Contact	jonas.korndorfer@cs.uni-saarland.de

Please only register for the workshop if you are planning to attend.

The organization and coordination of this event are carried out by NHR Süd-West, a member of the National High Performance Computing (NHR) initiative in Germany. For more information, visit https://nhrsw.de/.