Don't stop early: Case-folding source code at memory speed
Summary
GitHub's post on case-folding optimization analyzes a memory-speed, branchless ASCII pass and a decode-free Unicode folding path, showing substantial performance gains for code search indexing. It also explains data structures (page bitmaps, runs) and compares single-pass, two-pass, and fused approaches, with benchmarks up to 45 GiB/s for ASCII cases. The article highlights an open-source crate (casefold) and practical lessons for high-performance text processing.