Abstract: Focusing on the computing processing ability of critical application and localization, and based on the character analysis of Loongson 3A processor, a Loongson 3A high performance processing module based on NUMA parallel computing architecture is proposed. Furthermore, key points required for anti-harsh is analyzed and designed, and heat dissipation, power supervisor, power supply and boot up speed technology are solved. The performance can feed the demand of critical application through test. Consequently, the performance of memory accessing is improved. The relationship between the account of CPUs and the corresponding improvement of performance under SMP and NUMA is also be discussed.

Key words: Non Uniform Memory Access Architecture（NUMA）, Loongson 3A, VPX, COTS

摘要： 针对关键应用对信息处理能力提出的性能要求以及国产化需求，在分析龙芯3A处理器架构特点的基础上，设计了基于NUMA并行处理架构的龙芯3A高性能处理模块，并对抗恶劣环境的关键问题进行了分析和设计，解决了散热、电源监控及供电优化、启动速度等问题。通过测试验证，性能可以满足关键应用对信息处理能力的要求，从而有效解决了龙芯3A访存能力有限的问题。同时对SMP和NUMA架构下，龙芯3A处理器CPU数量的增加对访存性能的提升的关系进行了探讨。

关键词: 非统一内存访问架构（NUMA）, 龙芯3A, VPX, COTS