A.: Initial Experiences Porting a Bioinformatics Application to a Graphics

Abstract. Bioinformatics applications are one of the most relevant and compute-demanding applications today. While normally these applica-tions are executed on clusters or dedicated parallel systems, in this work we explore the use of an alternative architecture. We focus on exploiting the compute-intensive characteristics offered by the graphics processors (GPU) in order to accelerate a bioinformatics application. The GPU is a good match for these applications as it is an inexpensive, high-performance SIMD architecture. In our initial experiments we evaluate the use of a regular graphics card to improve the performance of RAxML, a bioinformatics program for phylogenetic tree inference. In this paper we focus on porting to the GPU the most time-consuming loop, which accounts for nearly 50 % of the total execution time. The preliminary results show that the loop code achieves a speedup of 3x while the whole application with a single loop optimization, achieves a speedup of 1.2x.