Re: [PATCH 2/2] pack-objects: optimize "recency order"

["Ævar Arnfjörð Bjarmason" <avarab@gmail.com>]

On Thu, Oct 27, 2011 at 23:01, Ævar Arnfjörð Bjarmason <avarab@gmail.com> wrote:

> I haven't dug into why this is but I'm pretty sure it's because this
> patch makes Git behave pathologically on repositories with a large
> amount of tags. git.git itself has ~27k revs / and ~450 tags, or a tag
> every ~60 revisions.
>
> Try it on e.g. a repository with a couple of hundred thousand revs and
> a tag every 10-20 revisions.

Actually it just seems slow in general, not just on repositories with
a lot of tags, on linux-2.6.git with this patch:

    $ time ~/g/git/git-pack-objects --keep-true-parents
--honor-pack-keep --non-empty --all --reflog
--delta-base-offs</dev/null
/home/avar/g/linux-2.6/.git/objects/pack/.tmp-pack
    Counting objects: 2184059, done.
    Delta compression using up to 8 threads.
    Compressing objects: 100% (338780/338780), done.
    130605d5b82492a38452b9bc7bddfb4a6ef0e942
    Writing objects: 100% (2184059/2184059), done.
    Total 2184059 (delta 1825129), reused 2184059 (delta 1825129)

    real    1m28.426s
    user    1m19.661s
    sys     0m3.008s

With it reverted:

    $ time ~/g/git/git-pack-objects --keep-true-parents
--honor-pack-keep --non-empty --all --reflog --delta-base-offset
</dev/null /home/avar/g/linux-2.6/.git/objects/pack/.tmp-pack
    Counting objects: 2184059, done.
    Delta compression using up to 8 threads.
    Compressing objects: 100% (338780/338780), done.
    130605d5b82492a38452b9bc7bddfb4a6ef0e942
    Writing objects: 100% (2184059/2184059), done.
    Total 2184059 (delta 1825129), reused 2184059 (delta 1825129)

    real    0m50.152s
    user    0m45.367s
    sys     0m2.920s

And if you create a lot of tags:

    git log --pretty=%h | perl -lne 'chomp(my $sha = <>); print $sha
if $i++ % 10 == 0' | parallel "git tag test-tag-{} {}"

Resulting in:

    $ git rev-list HEAD | wc -l
    269514
    $ git tag -l | wc -l
    13733

This'll be the time with this patch reverted, i.e. almost exactly the
same as with just ~250 tags:

    real    0m52.860s
    user    0m44.907s
    sys     0m3.172s

And with it applied, i.e. almost exactly the same as with just ~250
tags:

    real    1m29.080s
    user    1m21.825s
    sys     0m3.096s