The express-lane transformation isolates and duplicates frequently executed program paths, aiming for better data-flow facts along the duplicated paths. An express-lane p is a copy of a frequently executed program path such that p has only one entry point at its beginning; p may have branches back to the original code, but the original code never branches into p. Classical data-flow analysis is likely to find sharper data-flow facts along an express-lane, because there are no join points.