Performance study of a whole genome comparison tool on a hyper-threading multiprocessor

Abstract. We developed a multithreaded parallel implementation of a sequence alignment algorithm that is able to align whole genomes with reliable output and reasonable cost. This paper presents a performance evaluation of the whole genome comparison tool called ATGC — Another Tool for Genome Comparison, on a Hyper-Threading multiprocessor. We use our application to determine the system scalability for this partic-ular type of sequence comparison algorithm and the improvement due to Hyper-Threading technology. The experimental results show that de-spite of placing a great demand on the memory system, the multithreaded code generated by Intel compiler yields to a 3.3 absolute speedup on a quad-processor machine, with parallelization guided by OpenMP prag-mas. Additionally, a relatively high 1st level cache miss rate of 7-8% and a lack of memory bandwidth prevent logical processors with hyper-threading technology enabled from achieving further improvement.