grit://cooking/bench/fetch/noop@git-git
← cookingall benchmarksraw data

grit fetch · noop @git-git

No-op fetch when client already matches remote (git-git over smart HTTP)

repo: Upstream git/git pinned at v2.47.0, full history · 4,525 files · 75,031 commits · 954 refs

$ grit-bench daily --only fetch/noop@git-git2026-10-27 · grit 8dbc58e77 · git version 2.43.0
toolmedianminmaxrunspeak memory
git25 ms24 ms27 ms2054 MB
grit11 ms10 ms12 ms2018 MB
grit is 2.3× fastermemory: 3.0× less
every timed run (log scale)
git
grit
8.0 ms34 ms
How it's measured
setup
remote http://127.0.0.1:43713/remote.git served by git http-backend (git's own smart HTTP server)
before each run
grit-bench prepare-network-fetch-noop --git git --bare /tmp/grit-bench-scratch/daily-remote-git-git/remote.git
git runs
$ git fetch origin
grit runs
$ grit fetch origin
checked first
before timing, git and grit were each run once and their results compared; the result is verified after every timed run
Where the time goesone extra traced run per tool, outside the timed runs
git · median run 25 ms
run: git remote-http
23 ms
run: git-remote-http
22 ms
run: git rev-list
14 ms
fetch: remote_refs
7.7 ms
run: git maintenance
1.2 ms
index: do_read_index
0.3 ms
fetch: consume_refs
0.1 ms
cache_tree: read
0.0 ms
submodule: parallel/fetch
0.0 ms
grit · median run 11 ms
grit.repository.open ×2
0.3 ms
grit.config.load ×4
0.1 ms
grit.pack.index.open
0.0 ms
Every day8 days
dategrit commitgitgritgrit vs gitgrit peak
2026-10-278dbc58e7725 ms11 ms2.3× faster18 MB
2026-10-26c7cf2c44334 ms19 ms1.9× faster18 MB
2026-10-252824ad39124 ms11 ms2.2× faster17 MB
2026-10-24157a05dde27 ms13 ms2.1× faster17 MB
2026-10-23bddf92f8623 ms8.6 ms2.7× faster16 MB
2026-10-224c131b94223 ms8.7 ms2.7× faster15 MB
2026-10-21d4844e92a20 ms7.6 ms2.7× faster15 MB
2026-10-204754ad0b924 ms1.28 s53× slower101 MB