Quantifying the common computational problems in contemporary applications

To quantify the potential opportunities for algorithmic choice, this work presents a quantitative characterization of the constituent computational problems (CPs) in contemporary applications. The feasibility of identifying well- defined CPs, occupying a significant portion of execution time, is studied. This is augmented by insight into the properties permitting separation of problem definitions from algorithms for their solution. It is conjectured that, since a large fraction (55 %) of the analyzed body of existing software was identified as CPs, algorithmic choice may be a viable approach.