项目名称: CPU和GPU混合体系结构上生物网络比对并行算法研究
项目编号: No.61462005
项目类型: 地区科学基金项目
立项/批准年度: 2015
项目学科: 自动化技术、计算机技术
项目作者: 钟诚
作者单位: 广西大学
项目金额: 45万元
中文摘要: 新型的多核CPU和GPU混合体系结构对高性能并行算法研究与应用提出了新挑战、提供了新机遇。生物网络是一类复杂网络系统,由高通量实验建立的生物数据规模十分庞大。生物网络比对计算问题是生物数据分析领域重要的研究课题之一。生物网络比对问题是计算密集型、组合优化问题。本项目研究提出多核CPU和GPU协同计算、波前式并行和水平并行计算生物网络结点相似度和拓扑相似度得分矩阵的两层并行计算方法,提出适用于生物网络比对问题并行计算的多级缓存数据分配方法、CPU与GPU之间以及异构机群计算节点之间负载均衡的任务调度策略,设计实现与CPU和GPU混合体系结构相适应的高效、可扩展的生物网络局部比对和全局比对并行算法,以快速获得准确的比对结果,满足后基因组时代大规模生物网络分析的应用需求,促进新一代高性能计算机体系结构及并行算法在生物信息计算研究中的应用与发展。
中文关键词: 生物网络;比对;并行算法;混合体系结构
英文摘要: The hybrid multi-core CPU and GPU architectures propose new challenges and offer new opportunity for the research and applications of high-performance parallel algorithms. Biological networks are one of complex network systems. The scale of biological data established by high-throughput experiment is huge. Biological networks alignment is one of the important issues in the biological data analysis field and it is a computation-intensive and combinatorial optimization problem. This project will present a two-layer computation method for CPU and GPU cooperatively parallel computing the scoring matrix of nodes similarity and topology similarity for biological networks in the wavefront and horizontal ways, propose the data partitioning approach in multiple level caches and the loads-balance scheduling strategies between CPU and GPU and among computing nodes in the heterogeneous cluster, and design and implement the efficient and scalable parallel algorithms for local alignment and global alignment of biological networks on the hybrid CPU and GPU architectures to obtain quickly the accurate alignment results. The research results are helpful to satisfy the requirments for large scale biological network analysis in post-genome era and promote the applications and development of new generation high-performance computer architectures and parallel algorithms in biological information computation research.
英文关键词: Biological Network;Alignment;Parallel Algorithm;Hybrid Architectures