Experiences with Parallelizing a Bio-informatics Program on the Cell BE

The Cell Broadband Engine Architecture is a new heterogeneous multi-core architecture targeted at compute-intensive workloads. The architecture of the Cell BE has several features that are unique in high-performance general-purpose processors, such as static instruction scheduling, extensive support for vectorization, scratch pad memories, explicit programming of DMAs, mailbox communication, multiple processor cores, etc. It is necessary to make explicit use of these features to obtain high performance. Yet, little work reports on how to apply them and how much each of them contributes to performance.This paper presents our experiences with programming the Cell BE architecture. Our test application is Clustal W, a bio-informatics program for multiple sequence alignment. We report on how we apply the unique features of the Cell BE to Clustal W and how important each is to obtain high performance. By making extensive use of vectorization and by parallelizing the applicationacross all cores, we speedup the pairwise alignment phase of Clustal W with a factor of 51.2 over PPU (superscalar) execution. The progressive alignment phase is sped up by a factor of 5.7 over PPU execution, resulting in an overall speedup by 9.1