Complete LIKWID support for Julia

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

We are happy to announce that the Julia wrapper LIKWID.jl now supports all features of the LIKWID library. Besides the initial CPU MarkerAPI support, this update enables full control over performance monitoring capabilities of various CPUs and Nvidia GPUs. In addition, “management tasks” like system topology queries, CPU frequency manipulation, or energy measurements can now be performed from within Julia applications.

The project had been initiated by Valentin Churavy (MIT CSAIL Julia Lab) and was continued and heavily extended by Carsten Bauer (Paderborn Center for Parallel Computing). The NHR@FAU team would like to thank both for their contributions.

An introduction to Julia was given by Valentin Churavy for the HPC Café organized by NHR@FAU. See HERE for the video.

Carsten Bauer has written some excellent tutorials about the use of LIKWID in Julia applications: Performance monitoring and MarkerAPI with CPUs & GPUs

[LIKWID.jl documentation]

The joint work on LIKWID.jl performed by NHR@FAU and PC2 has been supported by NHR fundings of federal and state authorities.

Thomas Gruber

Development LIKWID, ClusterCockpit & MachineState

Erlangen National High Performance Computing Center
Software & Tools Division