Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2508.03016 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911282101223424 |
|---|---|
| author | Ma, Kaihao Wang, Meiling Oleg, Senkevich Li, Zijian Xue, Daihao Malyshev, Dmitriy Lv, Yangming Xiao, Shihai Yan, Xiao Alexander, Radionov Zeng, Weidi Gao, Yuanzhan Zou, Zhiyu Yao, Xin Liu, Lin Wu, Junhao Liu, Yiding Fu, Yaoyao Wang, Gongyi Zhang, Gong Yi, Fei Liu, Yingfan |
| author_facet | Ma, Kaihao Wang, Meiling Oleg, Senkevich Li, Zijian Xue, Daihao Malyshev, Dmitriy Lv, Yangming Xiao, Shihai Yan, Xiao Alexander, Radionov Zeng, Weidi Gao, Yuanzhan Zou, Zhiyu Yao, Xin Liu, Lin Wu, Junhao Liu, Yiding Fu, Yaoyao Wang, Gongyi Zhang, Gong Yi, Fei Liu, Yingfan |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_03016 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | KBest: Efficient Vector Search on Kunpeng CPU Ma, Kaihao Wang, Meiling Oleg, Senkevich Li, Zijian Xue, Daihao Malyshev, Dmitriy Lv, Yangming Xiao, Shihai Yan, Xiao Alexander, Radionov Zeng, Weidi Gao, Yuanzhan Zou, Zhiyu Yao, Xin Liu, Lin Wu, Junhao Liu, Yiding Fu, Yaoyao Wang, Gongyi Zhang, Gong Yi, Fei Liu, Yingfan Information Retrieval 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. |
| title | KBest: Efficient Vector Search on Kunpeng CPU |
| topic | Information Retrieval |
| url | https://arxiv.org/abs/2508.03016 |