Neels Hofmeyr wrote on Wed, May 11, 2011 at 14:01:08 +0200: > On Wed, 2011-05-11 at 02:02 +0200, Bert Huijben wrote: > > > > > -----Original Message----- > > > From: Daniel Shahaf [mailto:d...@daniel.shahaf.name] > > > Are you sure? The script's output seems a little bit too consistent: > > > > > > ne...@apache.org wrote on Tue, May 10, 2011 at 11:57:35 +0000: > > > > ("1.23|+0.45" means factor=1.23, difference in seconds = 0.45 > > > > factor < 1 or difference < 0 means '5x5_trunk' is faster than '5x5_1.6') > > > .. > > This is just the explanation how you should read the numbers. The real > > numbers are in the tables (which you removed from this mail) > > heh lol exactly, that's just the legend. Didn't you notice? 1.23 .45? ;) > > It is probably healthy to be a little weary of those results, but still. > We had quite bad numbers, now the same thing gives pretty nice numbers. > That ought to say *some*thing. > > I did not believe it at first and double checked that the timings look > real. And they do. There are still very very few commands that are > slower by a very very small amount of seconds. The rest is mindblowing. >
Yep, I've re-read the result and noticed that most of them were improvement or very very slight deprovement (is that a word?). I've also noticed that in some cases that max/min time were quite far from the average; I wonder if we should be adding standard deviation (or some other statistic of 'spread') to the output. > ~Neels >
