accurate 0.2.1

(more or less) accurate floating point algorithms