Valkey 9.0-rc1 Taps AVX-512 For String-To-Integer Conversion For ~19% Gain
- Reference: 0001569557
- News link: https://www.phoronix.com/news/Valkey-9.0-rc1-Released
- Source link:
One of the optimizations in Valkey 9.0 is [1]this merge optimizing the BITCOUNT command using AVX2. With this SIMD AVX2 optimization for a 1MB value size it's as much as a +514% change while for smaller 4KB value sizes is arorund a 10% improvement or +116% at 64KB.
Valkey also [2]added optimized string-to-integer performance and there it's relying on AVX-512 instructions. With that AVX-512 usage for string-to-integer conversions it's seeing a 17~19% performance boost.
Valkey 9.0-rc1 also brings better replication stability, support for using ARM NEON SIMD for optimized BITCOUNT, SSE2 SIMD for faster "findBucket" performance up to 6%, and a variety of other performance optimizations.
Some of the new features coming for Valkey 9.0 include multi-database support for the cluster mode, Multi-Patch TCP (MPTCP) support, atomic slot migration, and more.
Downloads and more details on the Valkey 9.0-rc1 release via [3]GitHub .
[1] https://github.com/valkey-io/valkey/pull/1741
[2] https://github.com/valkey-io/valkey/pull/1944
[3] https://github.com/valkey-io/valkey/releases/tag/9.0.0-rc1
phoronix