Re: [PATCH] fast-export: ensure we traverse commits in topological order

From: Elijah Newren <newren@gmail.com>
Date: Feb 11, 2009, 15:23 UTC
Message-ID: <51419b2c0902110723y75d0c4a4r34cf92cf371313ea@mail.gmail.com>
In-Reply-To: <20090211135640.GA19600@coredump.intra.peff.net>

On Wed, Feb 11, 2009 at 6:56 AM, Jeff King <peff@peff.net> wrote:

Show 27 quoted lines

> On Wed, Feb 11, 2009 at 10:48:18AM +0000, Mike Ralphson wrote:
>
>> > On Tue, 10 Feb 2009, newren@gmail.com wrote:
>> > fast-export will only list as parents those commits which have already
>> > been traversed (making it appear as if merges have been squashed if not
>> > all parents have been traversed).  To avoid this silent squashing of
>> > merge commits, we request commits in topological order.
>>
>> Any comparative timings? We don't need to rename this to 'git
>> reasonably-speedy-export'? 8-)
>
> Hmm.
>
> In git.git:
>
>  $ time git fast-export --all --signed-tags=strip >/dev/null
>  real    1m6.013s
>  user    1m3.840s
>  sys     0m2.140s
>
>  $ time git fast-export --all --signed-tags=strip --topo-order >/dev/null
>  real    0m49.018s
>  user    0m47.987s
>  sys     0m0.888s
>
> I certainly didn't expect it to be _faster_.  More efficient use of the
> delta cache, maybe?

Yeah, I also saw a speed-up, though the difference was less pronounced -- 19.7 seconds versus 19.5 (that was the average for 5 runs, after first doing a run that I ignored to make sure I was comparing hot-cache to hot-cache).

I'm a little surprised why my numbers were so much lower than yours, though. I don't think of my hard drive and CPU as being all that quick -- is this a difference in packing, by chance, with your repository not having recently been repacked?