What problem does it solve? Rust code that runs too slowly or compiles too slowly is hard to fix by intuition alone, because the optimizer, cache effects, and allocation patterns behave in non-obvious ways. This Skill enforces a measure-first methodology so you profile before optimizing and re-benchmark every change instead of guessing. ## Core Features & Use Cases - Benchmarking and Profiling: Set up criterion or divan microbenchmarks with correct black_box usage, and profile whole programs with flamegraph, samply, hyperfine, or the dhat heap profiler. - Concrete Optimizations: Apply targeted fixes for hot spots including build configuration (LTO, codegen-units, target-cpu), heap allocation reduction, faster hashing with FxHashMap, type size shrinking, iterator and bounds-check elision, buffered I/O, and SIMD vectorization. - Compile-Time Reduction: Speed up the edit-compile-run loop with faster linkers (mold/lld), dev profile tuning, crate splitting, monomorphization reduction, and sccache caching. - Use Case: A developer notices a Rust CLI processing files too slowly. They establish a criterion baseline, generate a flamegraph revealing hashing dominates, swap HashMap for FxHashMap, and confirm a measurable improvement before keeping the change. ## Quick Start Ask the assistant to profile your Rust project and optimize the hot path using the rust-performance methodology.
