This system allows you to independently create and then combine the implementation of any format of real numbers and a set of benchmarks for testing the performance and correctness of calculations of the implementation in question. Posit (unum), float and bfloat tested. - View it on GitHub