一种基于国产平台的Hyperscan优化方法OA

As a crucial tool for text matching,regular expressions play a significant role in the processing of network data.However,the open-source regular expression library Hyperscan exhibits relatively low matching efficiency on the Weixin H8000 platform,which limits its application in network data processing.To solve this problem,we propose a method based on single instruction multiple data instruc-tion-level optimization to enhance the matching performance of Hyperscan on the Weixin H8000 platform.The optimization method first analyzed the parallel characteristics of Hyperscan and in combination with the performance analysis tools of the Sunway platform,the hotspot functions affecting matching performance were identified.Secondly,for these hotspot functions,SIMD assembly instructions were manually embedded to optimize the key functions.Experimental results demonstrate that the regular expression matching performance on the Weixin H8000 platform has been improved by an average of 77%,significantly enhancing the efficiency of Hyperscan on this plat-form.This optimization method not only provides an effective solution for improving Hyperscan's performance on Sunway processors but also offers important support for the application and development of domestic processors in the field of network data processing.