v1.0.0 — ETF Pattern Matching Acceleration · 形态匹配 ETF 策略加速

What is this?

Python+C++ hybrid programming practice: extracted pure-computation modules from a Chinese ETF pattern-matching strategy and accelerated with pybind11 + C++20. Not a trading system — no live trading code, no backtest claims.

Performance

Function	Python	C++	Speedup
DTW Distance (L=19)	98 µs	2.7 µs	37×
Pattern Match (single)	14.3 ms	0.23 ms	61×
Batch Match (100 calls)	50 ms	23 ms	2.2×

What is included

• 6 pure-computation Python modules (src/core/) — DTW, pattern matching, technical indicators, market features, risk controls, metrics
• 8-function C++ acceleration module (src/cpp/etf_core.cpp, ~1,100 lines) — pybind11 bindings with full GIL management
• 54 unit tests + 2 verification scripts — C++ vs Python parity verified
• Interactive Jupyter demo (31 cells) — walk through the full algorithm
• Performance analysis article (docs/performance-analysis.md, bilingual) — why 61× single-call → 2.2× batch
• Bilingual README (Chinese + English)

Review Chain

Round	Backend	Role	Findings
R1	Kimi-K2.7	Adversarial review	Code improvements (GIL, batch contract)
R2	GPT-5.5	Completeness review	Cross-file consistency
R3	GPT-5.5	Final review	Edge case coverage
R4	Kimi-K2.7	Code improvement	Boundary tests + batch convergence
—	GPT-5.6-Sol	Full code audit (2026-07-12)	32 findings (9 P0), all fixed
—	Kimi-K2.7	Regression audit (2026-07-12)	0 regressions

Build

cmake -B build -DPython_EXECUTABLE="<path-to-python.exe>"
cmake --build build --config Release
python verify_etf_core.py

Requirements: Python 3.12+, NumPy, pybind11 3.0.4, CMake 3.20+, MSVC 19.51 (or Clang 16+)

---

This project is a programming practice — not investment advice, not a trading system.
🤖 Generated with Claude Code