[PPL-devel] Use of ldexpl() in the floating_point branch