grit://cooking/bench/restore/paths@synth-10k

grit restore · paths @synth-10k

Restore ~1% paths from index: grit restore -- <paths> vs git restore -- <paths>

repo: 10,000-file synthetic repository; canonical cached shape, scenario variants documented by the runner · 10,000 files · 1,000 commits · 800 refs

$ grit-bench daily --only restore/paths@synth-10k2026-10-27 · grit 8dbc58e77 · git version 2.43.0

tool	median	min	max	runs	peak memory
git	12 ms	11 ms	35 ms	20	12 MB
grit	11 ms	11 ms	41 ms	20	22 MB

grit is parmemory: 1.9× more

every timed run (log scale)

git

grit

8.6 ms52 ms

How it's measured

before each run

grit-bench prepare-worktree-restore-paths --git git

git runs

$ set -eu
if [ -s /tmp/grit-bench-scratch/daily-synth-10k/.grit-bench-worktree-paths ]; then
  tr '\n' '\0' < /tmp/grit-bench-scratch/daily-synth-10k/.grit-bench-worktree-paths | xargs -0 git restore -- > /dev/null 2>&1
fi

grit runs

$ set -eu
if [ -s /tmp/grit-bench-scratch/daily-synth-10k/.grit-bench-worktree-paths ]; then
  tr '\n' '\0' < /tmp/grit-bench-scratch/daily-synth-10k/.grit-bench-worktree-paths | xargs -0 grit restore -- > /dev/null 2>&1
fi

Where the time goesone extra traced run per tool, outside the timed runs

git · median run 12 ms

index: preload

2.4 ms

index: do_write_index

2.2 ms

index: do_read_index

0.6 ms

cache_tree: read

0.0 ms

cache_tree: write

0.0 ms

grit · median run 11 ms

grit.index.write

3.1 ms

grit.index.read

0.8 ms

grit.repository.open

0.2 ms

grit.config.load ×2

0.1 ms

grit.pack.index.open

0.0 ms

Every day8 days

date	grit commit	git	grit	grit vs git	grit peak
2026-10-27	8dbc58e77	12 ms	11 ms	par	22 MB
2026-10-26	c7cf2c443	21 ms	23 ms	1.1× slower	24 MB
2026-10-25	2824ad391	12 ms	12 ms	par	21 MB
2026-10-24	157a05dde	13 ms	13 ms	par	22 MB
2026-10-23	bddf92f86	14 ms	17 ms	1.2× slower	22 MB
2026-10-22	4c131b942	13 ms	15 ms	1.2× slower	20 MB
2026-10-21	d4844e92a	12 ms	14 ms	1.1× slower	20 MB
2026-10-20	4754ad0b9	15 ms	1.34 s	89× slower	19 MB