What problem does it solve? Rust performance work often fails because developers optimize without baselines, mix multiple changes into one measurement, or chase cold code. This Skill enforces a strict profiling discipline so every optimization is measured, isolated, and reversible. ## Core Features & Use Cases - Structured optimization loop: Enforces baseline → profile → hypothesis → single change → test → remeasure, with explicit stop conditions and revert rules. - Toolchain guidance: Covers hyperfine, criterion, samply, dhat, perf stat, and cargo asm, plus Cargo profile configuration (release, profiling, bench, dist) and PGO workflows. - Ranked optimization playbook: Orders techniques by expected impact, from algorithmic changes and allocation reduction (arena, SmallVec, Cow) through layout, hashing, micro-optimizations, and parallelism. - Use Case: A developer profiling the rsvelte Svelte compiler uses the Skill to identify hot functions with samply, replace serde_json::Value with typed AST nodes, and verify each change with perf_bench before committing. ## Quick Start Ask the assistant to profile the Rust project and run one measured optimization iteration on the hottest function.
