Tutorial IV: The Promise of FPGA-Acceleration Processors to Bioinformatics Research

Anthony Johnson, Ph.D., University of Toledo
Location: Shriver Center
Monday 10:30 a.m. - 12:00 p.m.

Bioinformatics research inherently deals with huge data basis which contain genomes of ever increasing number of species. Inside those databases researchers would like to look for answers to many intriguing questions, but the available processing algorithms (well known examples of which, like Smith-Waterman, will be cited) running in sequential mode on classic general-purpose processors make their wishes computationally intractable in terms of the time needed to provide the sought for information. Since most of the bioinformatics processing tasks involve repetition of a huge number of the same independent processing steps on different sets of data, the massively parallel, adjustable-to-application computational resources of Field Programmable Gate Arrays (FPGAs) seem to be offering a viable, readily available solution, when added to the computer’s general purpose processor as an acceleration coprocessor. The idea of coprocessors has been already exploited in other applications; perhaps the most commonly known one of which is in real-time graphic data processing, where the so called "graphics cards" carry processors and memory dedicated to the computationally intensive image rendering. In processor-coprocessor cooperations, the classical general-purpose processors transfer massive processing tasks to the specialized coprocessor, and later further process the results collected from the coprocessor. Although FPGAs have been around for nearly two decades, and although their application in bioinformatics data processing has yet to become a common practice, FPGAs have been used to accelerate the bioinformatics data processing in the role of "acceleration coprocessors". Why that did not become a common practice, and what new developments can be expected in the near future, will be discussed. Classic general-purpose microprocessors, which are the working horse of regular computers, will be compared to FPGA-coprocessors in several respects, including their architecture, the way of programming, the way of processing data, and clock speed.

Back to tutorials