Performance: We're In This Together
summary
Argues Rust needs order-of-magnitude wins not 2% tweaks, citing Zed’s repo with 150 workspace members, nine deep dependency tree and Cargo/Rust‑Analyzer cache fights that forced full rebuilds; highlights a concrete case where an edition change cuts a workflow from 4m → 27s; lists actionable work: proc macro expansion caching, snapshot-on-publish, moving derives to declarative macros (could replace ~90% of proc macros), mirror-only arlibs and cross-workspace/remote caching plus an opaque-dependencies idea, and splitting the dev profile.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:06Rust build performance
- 3:14C++ parity limits
- 5:00User workflows expectations
- 7:32Talking to users
- 9:15Beyond Rust performance
- 11:45Indirect performance effects
- 13:07Config overhead
- 14:57Competing improvements
- 17:43Alternatives to proc macros
- 19:31Prioritizing performance
- 20:53Cross-workspace caching
- 23:18Caching trade-offs
- 26:28Project dependencies
- 27:33Coordinating improvements
- 28:15How to help
