Performance of Sequence Alignment Bioinformatics Applications on General Purpose Processors: A Case Study

Aligning specific sequences against other known sequences in a database is a central aspect of bioinformatics. New experimental data being added continuously to these databases necessitates the use of efficient computational resources. Although processor clock frequencies speeds have been increasing steadily, it does not necessarily translate into better performance. In this paper, we analyze the performance of sequence alignment bioinformatics benchmarks from the BLAST and FASTA suites from a microarchitectural standpoint. Results indicate that these benchmarks are memory intensive with negligible floating point content. They exhibit parallelism at the instruction level and their performance can be improved significantly by using bigger caches.