Created attachment 215596 [details]
I'd like to use unixbench not only for performance measuring but also to create CPU load for multiprocessor in a convenient way.
The attached `unixbench-mp.patch' fixes to create CPU information structure.
`Run` scripts determines if multi processor benchmark should be run from it.
Of course, the structure highly depends on Linux systems.
We can't fill complete same data but at least number of CPUS, CPU model, and CPU architecture can be filled.
They are enough to execute multiprocessor benchmark.
So please apply the patch to Ports tree.
Build info is available at https://gitlab.com/swills/freebsd-ports/pipelines/161802943
Created attachment 216079 [details]
The maintainer doesn't respond to me.
Does anybody check and commit `unixbench-mp.patch`?
It will show parallel benchmarck result like `unixbench.log`.
Hi. Sorry for the delay, I was quite busy with personal stuff.
I can work on it and commit it.
However, to create CPU load, I can also suggest stress-ng (also available on FreeBSD)
It's a program that stress different subsystem.
This command will create load for all the available CPUs
stress-ng --cpu 0
A commit references this bug:
Date: Sat Jul 4 14:48:33 UTC 2020
New revision: 541217
benchmarks/unixbench: Add support to parallel execution
Add support to get the number of CPUs on FreeBSD.
Submitted by: Yuichiro NAITO <firstname.lastname@example.org>
I bumped the portrevision to force the re-build of the package and the update.
(In reply to Luca Pizzamiglio from comment #7)
I could confirm your commit.