High-Performance Rust

Current stable at the time of writing: Rust 1.99.1 (released 2026-10-15; checked 2026-11-08); 1.100 ships 2026-11-12. Techniques from The Rust Performance Book, std docs, the Rust Reference, release notes and 2026 ecosystem surveys. Most are build-config or local rewrites; all are hints to a compiler that already does well, so measure each one.

Reference files (read one when a checklist row points at it):

• references/inlining-codegen.md — rustc's inlining rules, #[inline]/#[cold]/std::hint, asm verification.
• references/vectorization-simd.md — writing loops LLVM vectorizes, float reductions, #[target_feature] dispatch, SIMD crates.
• references/memory-layout.md — allocation avoidance, struct/enum layout (measured sizes), SoA, false sharing, interning.
• references/concurrency-locks.md — atomics, orderings, Mutex vs RwLock vs ArcSwap, sharding, channels, LazyLock.
• references/hot-path-idioms.md — entry API, hashers, sorting, memchr, number formatting, dispatch, const tables, new Vec/slice APIs.
• references/new-rust-features.md — version tables of every perf-relevant feature 1.75 -> 1.101 nightly.
• references/tokio-ecosystem.md — tokio runtime, bytes, serde/binary formats, rayon, caches, HTTP, tracing.

Measure first — never guess

Profile to find the hot spot, change one thing, re-measure. Effects (especially inlining and hints) are unpredictable and sometimes negative.