From: Marco Costalba <mcostalba@gmail.com>
Date: Tue, 17 Jul 2007 16:36:15 GMT
Subject: Re: [PATCH] Add --show-size to git log to print message size
Message-ID: <e5bfff550707170936kcd0c001nb4c4a90baacf4275@mail.gmail.com>
In-Reply-To: <200707170849.35842.andyparkins@gmail.com>

On 7/17/07, Andy Parkins <andyparkins@gmail.com> wrote:
>
> Does this really give a speedup?
>
> I'd be surprised, as long as the parse is being done during the output from
> git using the QProcess::readyRead() signal once and only once, then git is
> the bottle neck.  Parsing the stream is almost trivial in comparison to the
> work that git is doing.
>

Well, it's not exactly from readyRead() but you are right, the bottle
neck it's git.

A bare:
time git log --parents --boundary --pretty=raw -z --log-size
--topo-order HEAD > /dev/null

gives on linux and git repositories
git 960ms
linux 6920ms

Instead qgit gives, with the same command:

WITHOUT --log-size patch
git  1170ms  (+21%)
linux 7942ms (+15%)


WITH --log-size patch
git 1125ms (+17%)
linux 7820ms (+13%)

I don't know why overhead on Linux is less, probably on small repos
the GUI part of qgit that you can consider more or less a constant
time weights more regarding parsing.


Marco