AgentSkillsCN

benchmark

为 crypto-rs-backtester 项目(Python 和 Rust)运行基准测试。

SKILL.md
--- frontmatter
name: benchmark
description: Run benchmarks for the crypto-rs-backtester project (Python and Rust).
metadata:
  owner: user
  version: "0.1.0"

Benchmark Skill

This skill allows you to run performance benchmarks for the project.

1. Python Benchmarks

Run the Python-side benchmarks using pytest.

bash
pytest -m bench -q

2. Rust Benchmarks (Standard)

Run the standard Rust core benchmarks.

bash
cargo bench -p backtester-core

3. Rust Benchmarks (Custom Configuration)

Run Rust benchmarks with custom parameters using the helper script. This allows you to tune the number of symbols, ticks per symbol, and batch window size.

bash
# Run with default 'heavy' settings (8 symbols, 500k ticks)
./.agent/skills/benchmark/scripts/run_custom_bench.sh

You can also set environment variables manually if you prefer:

bash
BACKTEST_BENCH_NSYMBOLS=8 \
BACKTEST_BENCH_TICKS_PER_SYMBOL=500000 \
cargo bench -p backtester-core --bench bench_core

4. Profile-Guided Optimization (PGO)

To build and benchmark with PGO (Linux/macOS). Note: This process is time-consuming as it runs a full build-profile-rebuild pipeline.

bash
make pgo