From: Bert Wesarg <bert.wesarg@googlemail.com>
Date: Sun, 06 Jul 2014 16:46:34 GMT
Subject: Re: [PATCH 5/5] Add a little script to compare two make perf runs
Message-ID: <CAKPyHN37FtG9=1zoMM3FxDzn1eSjbH7WGX8FOfkPLk8HJ5Scxg@mail.gmail.com>
In-Reply-To: <20140706161527.GU19781@tassilo.jf.intel.com>

On Sun, Jul 6, 2014 at 6:15 PM, Andi Kleen <ak@linux.intel.com> wrote:
>> a justification why the geometric mean is used here would increase my
>> confident significantly.
>
> It's just a standard way to summarize sets of benchmarks. For example SPEC
> uses the same approach.

No, SPEC would have calculated the geometric mean of the ratios
cmp[1][j] / cmp[0][j]. And this should also only be used under the
assumption that there is a multiplicative correlation.

Bert