Abstract
Pairwise sequence alignment is widely regarded as the most basic, and also the most computation-intensive, algorithm in computational biology. Sequence alignment identifies the similarity between the amino acid sequences of two proteins and is an important source of insight into many biological issues such as protein function and evolutionary pathways. In this paper, we present a high-performance hardware accelerator for sequence alignment problem. This accelerator relies on a set of simple processing cores connected by a reconfigurable network-on-chip. The reconfigurable network-on-chip provides very low inter-core latency and hence, allows integrating larger number of processing cores on a single chip. Evaluation results show that the proposed accelerator offers more than 47% performance gain over the baseline mesh topology and 18% over a state-of-the-art accelerator due to its faster network and finer-grained parallelism.