Monitoring Performance on a Hardware Level With LIKWID.jl

Symbolic picture for the article. The link opens the image in a large view.

At JuliaCon 2022 last week, Carsten Bauer from PC2 Paderborn gave a great talk on using hardware performance counters in Julia programs with LIKWID.jl, the Julia wrapper for the LIKWID library. With LIKWID.jl, you can leverage the full power of the LIKWID library, first and foremost counting performance events. Carsten demonstrated how LIKWID can be used from Julia programs and how hardware performance events can show you how well your code performs via empirical Roofline modeling. His talk is available on the YouTube channel of the Juila programming language.

Enjoy:

 

Thomas Gruber

Development LIKWID, ClusterCockpit & MachineState

Erlangen National High Performance Computing Center
Software & Tools Division