Parallel Identifying(l,d)-Motifs in Biosequences Using CPU and GPU Computing
Parallel Identifying(l,d)-Motifs in Biosequences Using CPU and GPU Computing
复制标题
使用 CPU 和 GPU 计算并行识别生物序列中的 (l,d)-基序
DOI：
10.1007/978-3-319-39817-4_25
复制
发表时间：
2016
期刊：
影响因子：
--
通讯作者：
Zhengping Liu（刘正平）
中科院分区：
文献类型：
--
作者：
Cheng Zhong（钟诚）;Jing Zhang（张静）;Bei Hua （华蓓）;Feng Yang（杨锋）;Zhengping Liu（刘正平）
为了加速缓存访问并减少访问时间，在使用建模计算并行解决（l，d）基序识别问题时，将不同组合位置产生的大量数据和许多候选序列分配到GPU中的纹理存储器。根据组合位置数据的大小设置GPU中线程块的大小，找到线程块中运行线程的最佳数量，并通过CPU和GPU协同计算设计识别生物序列中(l，d)基序的高速缓存并行算法。实验结果表明，所提出的并行算法可以在较少的计算时间内解决一些大尺寸的（l，d）基序识别实例，并获得良好的加速性和可扩展性。
To accelerate cache access and reduce the access time, the large number of data produced with different combined positions and many candidate sequences are distributed to the texture memory in GPUs when the modeling computation is used to solve in parallel the (l,d)-motif identification problem. The size of thread blocks in GPUs is set according to the size of data in combined positions, the best number of running threads in a thread block is found, and a cache-efficient parallel algorithm for identifying (l,d)-motifs in biosequences is designed by CPU and GPUs cooperative computing. The experimental results show that the proposed parallel algorithm can solve some (l,d)-motif identification instances of large size in less computation time and obtain good speedup and scalability.