Introduction
This document provides the installation guide, interface definitions, and sample code of Kunpeng Retrieval Library (KRL) to help you quickly get started with it.
KRL Overview
KRL, provided by Huawei, is an operator library optimized for the Kunpeng platform to accelerate vector retrieval. This library optimizes the instruction set architecture and memory access mechanism of the Kunpeng processor at the bottom layer. By combining low-precision quantization with high-precision re-ranking, the library significantly improves the computational efficiency and throughput of recall algorithms without compromising accuracy. These optimizations make it suitable for high-concurrency recall scenarios. KRL operators can be enabled to accelerate algorithms such as HNSW, PQFS, IVFPQ, and IVFPQFS—implemented in the open source Facebook AI Similarity Search (Faiss) library.
KRL applies to the Kunpeng 920 7282C processor, and supports NEON instructions (128-bit width) and SVE instructions (256-bit width).
Application Scenarios
KRL is suited for the following scenarios:
- Search: network search and multi-modal search
- Recommendation: recommendation systems
- Advertising: advertisement placements