基于GPU实现允许k-差别近似串匹配并行算法

摘要

为了在GPU上实现允许k-差别近似串匹配并行算法,并进一步提高在GPU上的算法性能,在综合考虑GPU的存储层次前提下,采取了多种访存优化措施,并提出了避免bank冲突的解决方案.测试结果表明,所用优化措施及方案使算法性能明显提高.%To implement parallel algorithm for approximate string matching with k-differences on GPU and further improve performance with consideration of the GPU memory hierarchy, several optimal accessing schemes are discussed with solution for avoiding bank conflicting. Experimental results show that the proposed optimal accessing scheme can result in significant improvement of the algorithm.