Re: [PATCH 2/2] for-each-ref: add --count-matches option

From: Phillip Wood <phillip.wood123@gmail.com>
Date: Jun 28, 2023, 13:12 UTC
Message-ID: <776c3682-d2eb-d2d7-3ea8-4a7db8cd7842@gmail.com>
In-Reply-To: <f6fd39bc-65d4-76e3-94b4-9163194c89dd@gmail.com>

On 27/06/2023 11:05, Phillip Wood wrote:

Show 29 quoted lines

> On 27/06/2023 08:30, Jeff King wrote:
>> I don't think this is a very realistic perf test, because for-each-ref
>> is doing a bunch of work to generate its default format, only to have
>> "wc" throw most of it away. Doing:
>>
>>    git for-each-ref --format='%(refname)' | wc -l
> 
> That's a good point. I wondered if using a short fixed format string was 
> even better so I tried
> 
> git init test
> cd test
> git commit --allow-empty -m initial
> seq 0 100000 | sed "s:\(.*\):create refs/heads/some-prefix/\1 $(git 
> rev-parse HEAD):" | git update-ref --stdin
> git pack-refs --all
> hyperfine -L fmt "","--format=%\(refname\)","--format=x" 'git 
> for-each-ref {fmt} refs/heads/ | wc -l'
> 
> Which gives
> [...] 
> Summary
>    git for-each-ref --format=x refs/heads/ | wc -l ran
>      1.05 ± 0.01 times faster than git for-each-ref 
> --format=%\(refname\) refs/heads/ | wc -l
>     18.25 ± 0.20 times faster than git for-each-ref  refs/heads/ | wc -l
> [...] 
> I'm a bit suspicious of the massive speed up I'm seeing by avoiding the 
> default format but it appears to be repeatable.

Having seen Peff's mail [1] I realized that my test repo above is looking up the commit from a loose object. If I repack the repository then the default format is still slower than using "--format=%(refname)" but is much more competitive.

$ git repack -a Enumerating objects: 2, done. Counting objects: 100% (2/2), done. Writing objects: 100% (2/2), done. Total 2 (delta 0), reused 0 (delta 0), pack-reused 0

$ hyperfine -L fmt "","--format=%\(refname\)","--format=x" 'git 
for-each-ref {fmt} refs/heads/ | wc'
Benchmark 1: git for-each-ref  refs/heads/ | wc -l
   Time (mean ± σ):     111.4 ms ±   1.4 ms    [User: 96.9 ms, System: 
19.6 ms]
   Range (min … max):   109.6 ms … 115.1 ms    25 runs

Benchmark 2: git for-each-ref --format=%\(refname\) refs/heads/ | wc -l
   Time (mean ± σ):      66.7 ms ±   0.7 ms    [User: 59.5 ms, System: 
9.5 ms]
   Range (min … max):    65.6 ms …  68.2 ms    42 runs

Benchmark 3: git for-each-ref --format=x refs/heads/ | wc -l
   Time (mean ± σ):      63.4 ms ±   0.7 ms    [User: 56.3 ms, System: 
8.0 ms]
   Range (min … max):    61.9 ms …  65.1 ms    44 runs

Summary
   git for-each-ref --format=x refs/heads/ | wc -l ran
     1.05 ± 0.02 times faster than git for-each-ref 
--format=%\(refname\) refs/heads/ | wc -l
     1.76 ± 0.03 times faster than git for-each-ref  refs/heads/ | wc -l

So it seems most of the slowdown I was seeing yesterday was due it looking up a loose object. I'm surprised repacking makes such a difference in a repository that only contains two objects.

Best Wishes

Phillip

[1] https://lore.kernel.org/git/20230627195900.GC1280909@coredump.intra.peff.net