9–10 Jun 2026
UiT - The Arctic University of Norway in Tromsø
Europe/Oslo timezone

From For-Loop to Supercomputer: A Hands-On Introduction to Parallel Computing with OpenMP and MPI

10 Jun 2026, 13:50
40m
Auditorium Cerebrum (UiT - The Arctic University of Norway in Tromsø )

Auditorium Cerebrum

UiT - The Arctic University of Norway in Tromsø

UiT - The Arctic University of Norway Universitetsvegen 61 9019 Tromsø Norway
Tutorial (1h)

Speaker

Youssef Wally (The Arctic University of Norway - UiT)

Description

Most research code starts the same way: a for-loop over a dataset, processing one item at a time. This tutorial shows how to break that bottleneck; step by step, from a single CPU core all the way to a distributed HPC cluster.

Starting from a plain Python loop (SISD — Single Instruction, Single Data), we walk through Flynn's Taxonomy to build intuition about why parallelism works, then implement the same workload four different ways: Python multiprocessing, C++ (sequential), C++ with OpenMP, and C++ with MPI. Each version comes with live benchmarks so attendees can see the speedup for themselves.

By the end of the session, attendees will understand:
1) The difference between shared-memory (OpenMP) and distributed-memory (MPI) parallelism
2) When to reach for each tool and when not to
3) The key OpenMP pragma that can parallelise a loop in one line
4) How MPI coordinates work across multiple machines, and why it scales to thousands of nodes

All code is provided as ready-to-run scripts. No prior parallel programming experience is required, just basic familiarity with Python and/or C++.

This tutorial is particularly relevant to researchers who run computationally expensive pipelines (e.g simulations) and want to make better use of the computing resources available to them, like LUMI.

Author

Youssef Wally (The Arctic University of Norway - UiT)

Co-author

Elisabeth Wetzer (UiT The Arctic University of Norway)

Presentation materials

There are no materials yet.