[Monotone-devel] Re: cvs_import benchmark

[Lapo Luchini]

Lapo Luchini <lapo <at> lapo.it> writes:

> Lapo Luchini wrote:
> > So... 6% improvement between 0.26 and 0.30, and a 100% worsening from
> > 0.30 and yesterday's trunk! 0_0
> 
> Not having a working callgrind under FreeBSD nor Cygwin, I'll guess I
> will continue some kind of "binary search" in the remaining range to
> pinpoint the exact problematic commit.

But well... trying it again (this time on FreeBSD) I noticed that it was
actually solved in 0.31... good to know!

% for n in results3/*/stats.csv; do echo $n; cat $n; done
results3/cvsimp-mtn30-time/stats.csv
cvs_import-system-time,0.32,0.28,0.35
cvs_import-user-time,49.2,48.69,49.04
cvs_import-wall-time,94.42,51.55,51.22
results3/cvsimp-mtn30b-time/stats.csv
cvs_import-system-time,0.42,0.43,0.36
cvs_import-user-time,181.86,182.43,187.37
cvs_import-wall-time,186.34,187.47,191.86
results3/cvsimp-mtn31-time/stats.csv
cvs_import-system-time,0.37,0.34,0.37
cvs_import-user-time,48.92,47.98,47.79
cvs_import-wall-time,51.8,51.04,51.28

I still wonder why the problem was there in the first place, though (the
changelog in that range REALLY seemed "innocuous").

--
Lapo Luchini
address@hidden (OpenPGP & X.509)
www.lapo.it (Jabber, ICQ, MSN)