Skip to content

Optimize native exact-tree materialization - #28

Open
kvkenyon wants to merge 4 commits into
mainfrom
fm/rtc-native-101-ready-performance
Open

Optimize native exact-tree materialization#28
kvkenyon wants to merge 4 commits into
mainfrom
fm/rtc-native-101-ready-performance

Conversation

@kvkenyon

@kvkenyon kvkenyon commented Sep 4, 2026

Copy link
Copy Markdown
Owner

Cuts representative native exact-tree materialization from 8.93s p95 to 1.21s p95 on the measured M5/24 GB host. The measured cause was 611 serial Git launches per iteration; this batches committed blob metadata/content, reads numstats once, and limits patch concurrency to four while retaining exact revision identity, isolated object reads, shell-free Git, and existing caps.\n\nAdds a machine-readable 101-file/5,000-line benchmark with an M1/16 GiB reference budget and a stable 110-process cross-host regression gate, plus hostile-path/dirty-tree and process-ceiling coverage.\n\nValidated with the focused benchmark, hostile filename smoke, RTCIngestTests, scripts/native-check.sh, and npm run check. XCTest-only and full-Xcode UI/signing/notarization gates remain unclaimed on this Command Line Tools host.

@kvkenyon
kvkenyon force-pushed the fm/rtc-native-101-ready-performance branch from 9412fcf to a732f12 Compare September 4, 2026 03:13
@kvkenyon
kvkenyon force-pushed the fm/rtc-native-101-ready-performance branch 4 times, most recently from 3d142de to b7019e2 Compare September 4, 2026 04:56
@kvkenyon
kvkenyon force-pushed the fm/rtc-native-101-ready-performance branch from 567f7eb to 870afdd Compare September 4, 2026 13:43
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant