A Practical Approach to Tiling Zuker’s RNA Folding Using the Transitive Closure of Loop Dependence Graphs

Abstract

RNA secondary structure prediction is a computationally-intensive task that lies at the core of search applications in bioinformatics. In this paper, we consider Zuker’s RNA folding algorithm, which is challenging to optimize because it is resource intensive and has a large number of non-uniform dependences. We describe the application of a previously published approach, proposed by us, to automatic tiling Zuker’s RNA Folding loop nest using the exact polyhedral representation of dependences exposed for this nest. First, rectangular tiles are formed within the iteration space of Zuker’s loop nest. Then tiles are corrected to honor all dependences, exposed for the original loop nest, by means of applying the exact transitive closure of a dependence graph. We implemented our approach as a part of the source-to-source TRACO compiler. The experimental results present the significant speed-up factor of tiled code on a single core of a modern processor. Related work and future algorithm improvements are discussed.