tlyphed / copperbench Goto Github PK

When running a job with a large number of instances, all possible resources for the user are requested.
That might lead to problems on HPC environments with billing and a bug in the software as all resources are easily used up.
A parameter and updating the array description to:
--array=1-{number of jobs}%MAXJOBS
might to the trick.

Benchmarks take quite a lot of diskspace.
Hence, instances are oftentimes stored as compressed files (bz2, xz, or alike).
Would it be possible that a transparent uncompression step can be added. Similar to $file{...} feature in the config file?
See, e.g.:
https://github.com/daajoe/reprobench/blob/jkf_wip_pnode/lib/tcat.sh
for existing "transparent" uncompression.

Currently several metrics can be exported from the perf logs and included in the tabular results of the benchmark.
Similarly, the memory usage, which is tracked by runsolver and logged in the respective log file, should be extracted and included as well.