On Tue, 31 May 2022 at 09:48, Peter Geoghegan <p...@bowt.ie> wrote: > Shouldn't this be using the geometric mean rather than the arithmetic > mean? That's pretty standard practice when summarizing a set of > benchmark results that are expressed as ratios to some baseline.
Maybe just comparing the SUM of the seconds of each version is the best way. That comes out to 86.6% David