Math behind unixbench tests

[jbalen]

I am writing an article on VPS benchmarking so I would like to know is there any document describing the math behind unixbench tests?

[meteorfox]

The release notes in the README[1] inside the UnixBench directory includes a lot more information about how the score is computed, specifically the v4.0 release note and up. I would start there.

Also, reading the code and the comments in the Run perl script[2], should also help. The results are normalized to a set of baseline results which you can see here.[3] Then to compute the aggregated score, I believe it's a geometric mean of the normalized scores.

As per what the tests are measuring the WRITING_TESTS [4] file explains their output format.

[1] https://github.com/kdlucas/byte-unixbench/blob/master/UnixBench/README
[2] https://github.com/kdlucas/byte-unixbench/blob/master/UnixBench/Run#L942
[3] https://github.com/kdlucas/byte-unixbench/blob/master/UnixBench/pgms/index.base
[4] https://github.com/kdlucas/byte-unixbench/blob/master/UnixBench/WRITING_TESTS

[gstrauss]

Thanks for those pointers, @meteorfox!

@jbalen, you might also look at #23 for a general discussion of some important limitations of using old benchmarks on newer technology, such as VPS.

[jbalen]

Sorry for my late reply :(
@meteorfox your comments are really helpful and I appreciate your help a lot!!