Conversions from fixed-point numbers to floating-point numbers and the other direction aren't implemented in any of the solver backends