Skip to content

Add pinned-CPU native single-model timing - #35

Merged
PyYoshi merged 1 commit into
devfrom
v3/native-model-timing
Sep 20, 2026
Merged

PyYoshi merged 1 commit into
devfrom
v3/native-model-timing

Conversation

@PyYoshi

@PyYoshi PyYoshi commented Sep 20, 2026

Copy link
Copy Markdown
Owner

概要

  • native C++内部でreset/filter/feed/confidence/checksumを反復測定
  • I/O/Python/build/入力・scratch・prober確保は計測外。128 warm-up、単一CPU固定、モデル順を循環
  • 非計測時と全counter/state/confidence/reset一致を各試行で確認
  • 文書別median/min/maxと試行平均p95を分離。リクエストtail latencyとは呼ばない
  • 日本語仕様、純Python4件とnative/guard3件を追加

検証

  • モデル85 tests成功(native toolsあり)、なし74成功/11skip
  • Clangでprobe9 tests成功、Ruff/diff check成功
  • Paris16文書、CPU2、各20000反復×7試行×3モデル、独立3 run
  • 16文書の計時合計medianは約49〜50us。約1%差を高速化とは断定しない
  • run2で1文書のidentity試行平均p95が12.1%増、run1/3では再現せず。データは除外しない
  • engine/API/既存table/P01は不変

Related: PyYoshi/cChardet#123

@PyYoshi
PyYoshi marked this pull request as ready for review September 20, 2026 21:57
@PyYoshi
PyYoshi merged commit 61930cc into dev Sep 20, 2026
11 checks passed
@PyYoshi
PyYoshi deleted the v3/native-model-timing branch September 20, 2026 21:57
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