performance

Set of C11 libraries and CLI tools — parallel hashing, duplicate detection, file integrity manifests. Zero external runtime dependencies, GPLv3 (tools) / LGPLv3 (libraries). Lives in the bit-crafts monorepo on GitHub.

6 libraries, 3 tools, one C11 monorepo: bcduplicate finds duplicates at 758k files/s — 18× faster than rmlint on the same thread count.

Why find | xargs -P plateaus at scale, and how three phases separated by a strict barrier — parallel walk, sequential merge, parallel process — take a recursive hash from 12s to 7s on 650k files.

Find duplicates among a million files in 3.7 seconds: a size → fast-hash → full-hash pipeline, io_uring, and adaptive dispatch. Numbers, architecture, and what the hash choice actually changes.

Bump, pool, free-list, arena — four C allocators to replace malloc where it counts. Principle, use case, and minimal code for each.