Comparison of multiple-precision floating-point software