Stefan Marr

Use `cpupower` tool instead of custom implementation

1

See - https://wiki.archlinux.org/title/CPU_frequency_scaling#Scaling_governors - and it's in the linux-tools-common package on Ubuntu: https://packages.ubuntu.com/search?searchon=contents&keywords=cpupower&mode=exactfilename&suite=noble&arch=any

See - https://linux.die.net/man/1/ionice - https://stackoverflow.com/a/45970571/916546

Evaluate using `Popen(shell=False, ...)`

1

Should we avoid using the shell for running commands? Think this is a major change at the core of ReBench. So, we'd probably need to carefully test this. See https://github.com/smarr/ReBench/pull/248#issuecomment-2269272256

This is work in progress, all languages need to be adapted the same way

As noticed by @JensLincke, the List benchmark (for which the code originates in SOM https://github.com/SOM-st/som-java/commit/3a69207b7ae5672b7ef4986de649dd0da308f146), is a version derived from the classic Tak function: https://en.wikipedia.org/wiki/Tak_(function) Possibly via this version: https://www.khoury.northeastern.edu/home/will/Twobit/R6src/ntakl.sch...

Currently, the `-f` both means: 1. that we include data from failing runs, which can be caused either by benchmarks aborting or because the validation of results fails or output...

Make error on selecting non-existing scheduler more user friendly

6

`rebench` has an `-s` option, which allows to select a scheduler. If `-s` is given with something that's not a class name, or not an existing class, we should give...