Correctly-rounded arbitrary precision decimal floating point arithmetic
