LUMI Performance Analysis and Optimization Workshop, Stockholm, 11-13 May 2026

Registration is via the LUMI web site.

Course organisation

• Location: KTH, Lindstedtsvägen 24, 114 28 Stockholm, Sweden, room 522 "Fantum".

  The building is just a 5 minute walk from a metro station with excellent connections to the central station (red metro line). See also the SL web site for more information on public transportation in Stockholm.

• Travel and local information

• Schedule

Course materials

Materials will follow as the course progresses

Due to copyright issues some of the materials are only available to current LUMI users and have to be downloaded from LUMI.

Presentation	slides	recording
Day 1
Welcome and introduction	/	recording
Architecture, Programming and Runtime Environment	slides	lumi
Exercises #1	/	/
Compilers, libraries, runtime, single-core optimisations	slides	lumi
Performance Analysis with Perftools	slides	lumi
Advanced Performance Analysis	slides	lumi
Exercises #2	/	/
Interactive demo: Strategies for code optimization	lumi	/
Day 2
MPI Optimizations	slides	lumi
Exercises #4	/	/
I/O Optimizations	slides	lumi
Exercises #5	/	/
AMD Profiling Tools and GPU optimisations 1	slides	recording
Exercises #6	/	/
AMD Profiling Tools and GPU optimisations 2	slides	recording
Exercises #7	/	/
Day 3
Best practices: GPU Optimization, tips & tricks	slides	recording
Exercises #8	/	/
Extras
Appendix: Links to documentation	/	/

Making the exercises after the course

HPE

The exercise material remains available in the course archive on LUMI:

• The PDF notes in /appl/local/training/paow-20260511/files/LUMI-paow-20260511-Exercises_HPE.pdf

• The other files for the exercises in either a bzip2-compressed tar file /appl/local/training/paow-20260511/files/LUMI-paow-20260511-Exercises_HPE.tar.bz2 or an uncompressed tar file /appl/local/training/paow-20260511/files/LUMI-paow-20260511-Exercises_HPE.tar.

To reconstruct the exercise material in your own home, project or scratch directory, all you need to do is run:

tar -xf /appl/local/training/paow-20260511/files/LUMI-paow-20260511-Exercises_HPE.tar.bz2

in the directory where you want to work on the exercises. This will create the Exercises/HPE subdirectory from the training project.

However, instead of running the lumi_c.sh or lumi_g.sh scripts that only work for the course as they set the course project as the active project for Slurm and also set a reservation, use the lumi_c_after.sh and lumi_g_after.sh scripts instead, but first edit them to use one of your projects.

AMD

See the notes at each session.

Files for the exercises can be downloaded as either a bzip2-compressed tar file or an uncompressed tar file and are also archived on LUMI as either the bzip2-compressed tar file /appl/local/training/paow-20260511/files/LUMI-paow-20260511-Exercises_AMD.tar.bz2 or an uncompressed tar file /appl/local/training/paow-20260511/files/LUMI-paow-20260511-Exercises_AMD.tar.

There is no guarantee though that the software that is referred to on the system, will be there forever or will still work after an update of the system.

Warning

The software and exercises were tested thoroughly at the time of the course. LUMI however is in continuous evolution and changes to the system may break exercises and software