GoKawiilRust's gearhash crate adds NEON backend, doubling ARM64 performance
The gearhash crate, used for content-defined chunking via the GEAR rolling hash, has been updated to version 0.1.4 with a new NEON-based backend for ARM64 processors. The change is automatically selected on aarch64 systems and remains backwards compatible, delivering roughly double the throughput at typical chunk sizes without requiring any code changes from users.
GoKawiil's interpretation of the reporting above, not reported fact.
When gearhash was first written, ARM64 servers and desktops were niche, so only x86 SIMD optimizations (SSE4.2, AVX2) made sense. Since then, Apple's M-series chips and AWS's Graviton line have made ARM64 mainstream in both consumer and cloud computing, so performance parity on that architecture now matters for real-world backup and deduplication workloads that rely on this crate.
- gearhash 0.1.4 introduces a NEON backend that roughly doubles speed on ARM64 hardware.
- The optimization is automatic on aarch64 targets and fully backwards compatible with existing code.
- The update reflects ARM64's rise from a niche server option to a mainstream computing platform via Apple Silicon and AWS Graviton.
Source: sam.dev, 2026-09-15
Published there as: “Speeding up gearhash on ARM64”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.