Xiangpeng Hao’s bbtree reduces cache misses in a Rust ordered map by packing short key fragments into nodes and skipping shared prefixes. The project reports single-thread performance five times that of the standard BTreeMap and 30% lower memory usage, with no dependencies and no_std support. It provides separate single-thread replacement and concurrent-access APIs, and publishes experiments replacing in-memory tables in several databases. These are the project’s measurements on particular hardware and data types, not evidence that every database will receive the same speedup.
Back to Tech
bbtree introduces a cache-efficient ordered map for Rust
Its speed and memory gains are project benchmarks that should be checked against the intended key types and workload.