基于龙芯的并行计算设计与实现

发布时间：2018-10-13 07:31

【摘要】：随着社会的发展进步,人们对科学计算的需求越来越高,单一计算机无法满足要求,大幅提高计算能力的并行计算系统应运而生。与此同时某些特殊大型科学计算,如武器试验模拟、激光制导等问题的仿真计算数据和结果又涉及国家机密和信息安全,设计开发基于国产芯片的并行计算平台显得尤为重要。首先,对龙芯、申威、飞腾等国产芯片发展现状、产品性能进行了广泛细致的资料收集和调研工作,按照芯片自主知识产权比例、芯片主频率、处理器浮点计算能力、芯片功耗等要求逐一筛选,认真对比分析,确定了使用龙芯系列芯片作为并行计算平台的核心计算芯片。其次,在并行计算系统体系结构选择上,通过对分布式共享存储处理机、对称多处理机、大规模并行计算机及工作站机群等常用并行计算系统的计算能力、可靠性,系统可扩展性与实现成本等多方面的对比分析,并结合现有实际硬件条件,明确了采用工作站机群方式构建并行计算平台。在前期充分准备的基础上完成了基于龙芯的并行计算的设计研究,实现了使用龙芯CPU、LINUX操作系统、GCC/GDB编译调试软件、MPICH2并行编程环境、高速互联网络构建工作站机群式并行计算环境。最后利用圆周率?值的并行计算程序对组建的并行计算平台进行了并行计算效率和可靠性的测试,得出的数据结果满足了研究预期的要求。
[Abstract]:With the development of society, the demand for scientific computing becomes higher and higher. The single computer can not meet the requirements, and the parallel computing system which greatly improves the computing ability emerges as the times require. At the same time, the data and results of some special large-scale scientific calculations, such as weapon test simulation, laser guidance and so on, are related to state secrets and information security, so it is very important to design and develop a parallel computing platform based on domestic chips. First of all, the status quo of the development of domestic chips, such as Longson, Shenway and Feteng, and the product performance have been extensively and meticulously collected and investigated, according to the proportion of independent intellectual property rights of the chip, the main frequency of the chip, and the floating point computing capability of the processor. The power consumption of the chip is selected one by one, and the core computing chip which uses the Godson series chips as the parallel computing platform is determined by careful comparison and analysis. Secondly, in the selection of parallel computing system architecture, through the distributed shared memory processor, symmetric multiprocessor, large-scale parallel computer and workstation cluster of common parallel computing system computing power, reliability, and so on, Comparing and analyzing the expansibility of the system and the cost of the system, and combining with the actual hardware conditions, the parallel computing platform is constructed by using the workstation cluster method. The design and research of parallel computing based on Godson is completed on the basis of sufficient preparation in the early stage. The workstation cluster parallel computing environment is constructed by using Godson CPU,LINUX operating system, GCC/GDB compiling and debugging software, MPICH2 parallel programming environment and high speed Internet. Finally using the circumference? The efficiency and reliability of the parallel computing platform are tested by the parallel computing program. The results meet the requirements of the research.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP338.6

【参考文献】