期刊名称：International Journal of Distributed and Parallel Systems
印刷版ISSN：2229-3957
电子版ISSN：0976-9757
出版年度：2011
卷号：2
期号：1
DOI：10.5121/ijdps.2011.2102
出版社：Academy & Industry Research Collaboration Center (AIRCC)
摘要：In this paper, we describe the design of high-performance extended burrow wheeler transform based weighted sequence comparison algorithm for many core GPUs taking advantages of the full programmability offered by compute unified device architecture (CUDA) and its standard library thrust. Our extended burrow wheeler transform based weighted sequence comparison algorithm with thrust library implementation on CUDA is the fastest implementation of weighted sequence comparison algorithm than the our previous implementation of extended burrow wheeler transform based weighted sequence algorithm without using thrust library, and it is on average 56.3X times faster. Moreover, our present time implementation is also competitive with CPU implementations, being up to 2.9X times faster than comparable routine on 2.99 GHz Intel Pentium (R) 4 CPU with 3 GB RAM.
关键词：In this paper; we describe the design of high-performance extended burrow wheeler transform based;weighted sequence comparison algorithm for many core GPUs taking advantages of the full;programmability offered by compute unified device architecture (CUDA) and its standard library;thrust. Our extended burrow wheeler transform based weighted sequence comparison algorithm with;thrust library implementation on CUDA is the fastest implementation of weighted sequence comparison;algorithm than the our previous implementation of extended burrow wheeler transform based weighted;sequence algorithm without using thrust library; and it is on average 56.3X times faster. Moreover; our;present time implementation is also competitive with CPU implementations; being up to 2.9X times faster;than comparable routine on 2.99 GHz Intel Pentium (R) 4 CPU with 3 GB RAM.
