Rust performance debugging with TUIs and LLMs
summary
A lightweight instrumentation crate called hotpath.rs crate shows live TUIs, exposes data to LLMs via TUI and MCP integration, and focuses on deterministic A/B benchmarks that compare Git branches automatically. It reports concrete wins — an example function becomes 92% faster and allocates 97% less after a fix — and the crate hits ~50k downloads, ~1k stars and 40+ project integrations. Instrumentation overhead measures ~40–80 ns (previously ~800 ns) and planned pprof-rs sampling aims to add CPU-sampling to wall-clock metrics.
AI-written summary · about
chapters (10)
- 0:00Rust performance debugging
- 1:50HotPuf profiler features
- 3:38Live TUI demo with Zed
- 7:30LLMs and agentic workflows
- 8:13Profiling HotPuf with benchmarks
- 13:38HotPuf implementation
- 16:47Performance overhead techniques
- 18:11Comparison with sampling profilers
- 20:15CPU sampling integration plans
- 21:39Summary and contributors
related talks
27mImproving Rust Performance Through Profiling and Benchmarking
28mRust on Morello
27mThe Wonderful World of Rust Tooling
57mDTrace for Rust... and everything else too!
31mBenchmarking and Optimization of Rust Libraries
23m"The Current State and Future of Rust Compiler Performance"
40mRust’s approach of getting things right
30mIntegrating Rust into Tor: Successes and Challenges