Rust Parser Scans 37M LOC Linux Kernel in 43s, Author Seeks Call Graph Validation
blune-foo · reddit · 2026-10-04
- A developer shares a tree-sitter-based multilingual parser (Go, Rust, C/C++, TS/JS, Python, Java) building ASTs and approximate call graphs intended as retrieval input for RAG.
- Performance: a single algorithm handles both static (C) and dynamic (Python) languages; on the Linux kernel it parsed 64,460 files / 37.3M LOC / 648K functions in 43 seconds with a Rust multithreaded implementation; call graph test hit 774K nodes and 1.57M edges at 9GB maxRSS.
- The open question is validating call graph accuracy — the plan is to compare against clang and language-specific tools on small projects with sampled verification, and the author asks for better approaches.
More from coding & agent
- Blogger delegates visa applications entirely to AI agents, 3 approved so far — AlchainHust · 2026-10-04
- theo explains how he juggles 6 Claude subs and 3 Codex subs — 'sorry if it gets you banned' — 0xkarasy · 2026-10-04
- Super Mario 64 gets ported into Halo, the latest in AI coding-powered game mashups — mark_k · 2026-10-04
- Astra one-shots a zero-asset custom-engine game, developer impressed — Dimillian · 2026-10-04
- Solana AI Agent Project 'Poly' Outlines Q4 Roadmap: Own Models, Agent Toolkit, Tokenomics — DionysianAgent · 2026-10-04
- How to Build a Hiring-Worthy RAG Engineer Portfolio Project in One Month — ashishllm · 2026-10-04