Abstract: The realization of longest prefix match algorithm traditionally uses double threads to lookup the route information, which not fully utilize the network processor’s no-delay thread switching and programmable characters. This article discusses how to fully utilize the hardware characteristics and asynchronous read/write of memory method in the micro engine on the IXP2350 platform to realize the whole route lookup with only one thread. The test result also shows that a performance improves 20% than double threads realization on the micro engine.

Key words: Network processor, Longest prefix match algorithm, Embedded system, Micro engine, Microcode

摘要： 传统的最长匹配路由算法都是基于双线程并行查找方法来实现的，没有充分利用新一代网络处理器无延时线程切换和可编程的特点。该文基于IXP2350平台对传统的最长匹配算法进行了优化改进，充分利用硬件特性和微引擎中异步内存读写的特点，用单线程来完成整个路由的查找。实验测试结果表明，这种优化改进使路由器的包转发效率提高了20%。

关键词: 网络处理器, 最长匹配算法, 嵌入式系统, 微引擎, 微码