KBest: Efficient Vector Search on Kunpeng CPU
Abstract
Vector search, which returns the vectors most similar to a given query vector from a large vector dataset, underlies many important applications such as search, recommendation, and LLMs. To be economic, vector search needs to be efficient to reduce the resources required by a given query workload. However, existing vector search libraries (e.g., Faiss and DiskANN) are optimized for x86 CPU architectures (i.e., Intel and AMD CPUs) while Huawei Kunpeng CPUs are based on the ARM architecture and competitive in compute power. In this paper, we present KBest as a vector search library tailored for the latest Kunpeng 920 CPUs. To be efficient, KBest incorporates extensive hardware-aware and algorithmic optimizations, which include single-instruction-multiple-data (SIMD) accelerated distance computation, data prefetch, index refinement, early termination, and vector quantization. Experiment results show that KBest outperforms SOTA vector search libraries running on x86 CPUs, and our optimizations can improve the query throughput by over 2x. Currently, KBest serves applications from both our internal business and external enterprise clients with tens of millions of queries on a daily basis.
Keywords
Cite
@article{arxiv.2508.03016,
title = {KBest: Efficient Vector Search on Kunpeng CPU},
author = {Kaihao Ma and Meiling Wang and Senkevich Oleg and Zijian Li and Daihao Xue and Dmitriy Malyshev and Yangming Lv and Shihai Xiao and Xiao Yan and Radionov Alexander and Weidi Zeng and Yuanzhan Gao and Zhiyu Zou and Xin Yao and Lin Liu and Junhao Wu and Yiding Liu and Yaoyao Fu and Gongyi Wang and Gong Zhang and Fei Yi and Yingfan Liu},
journal= {arXiv preprint arXiv:2508.03016},
year = {2025}
}