kache-hash: A dynamic, concurrent, and cache-efficient hash table for streaming k-mer operations

kache-hash is an exact, dynamic, concurrent, resizable k-mer hash table that builds k-mer sets 3.5–7.0x faster and answers queries 1.5–5.4x faster at 56 threads.

• Why it matters: Streaming k-mer workloads repeatedly insert and look up overlapping k-mers, but general hash tables miss this locality and incur extra memory accesses and synchronization during concurrent builds.
• What they did: The method places k-mers by minimizers so runs sharing a minimizer reuse buckets with a single fetch and lock, using Iceberg hashing for skew, plus a compact spill filter for early-exit exact lookups.
• The result: Across five genomic datasets, it improves cache-line efficiency on the human genome (3.1–3.2x fewer per k-mer for positive and mixed queries), speeds PanGenie genotyping 1.96x, and accelerates K2Rmini read filtering 10.7–19.4x while keeping identical outputs and exact filtering.

AI summary of the abstract · may contain errors · not medical advice · read the paper.